1 /*******************************************************************************
3 Intel 10 Gigabit PCI Express Linux driver
4 Copyright(c) 1999 - 2015 Intel Corporation.
6 This program is free software; you can redistribute it and/or modify it
7 under the terms and conditions of the GNU General Public License,
8 version 2, as published by the Free Software Foundation.
10 This program is distributed in the hope it will be useful, but WITHOUT
11 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
15 You should have received a copy of the GNU General Public License along with
16 this program; if not, write to the Free Software Foundation, Inc.,
17 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
19 The full GNU General Public License is included in this distribution in
20 the file called "COPYING".
23 Linux NICS <linux.nics@intel.com>
24 e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
25 Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
27 *******************************************************************************/
29 #include <linux/types.h>
30 #include <linux/module.h>
31 #include <linux/pci.h>
32 #include <linux/netdevice.h>
33 #include <linux/vmalloc.h>
34 #include <linux/string.h>
36 #include <linux/interrupt.h>
38 #include <linux/tcp.h>
39 #include <linux/sctp.h>
40 #include <linux/pkt_sched.h>
41 #include <linux/ipv6.h>
42 #include <linux/slab.h>
43 #include <net/checksum.h>
44 #include <net/ip6_checksum.h>
45 #include <linux/etherdevice.h>
46 #include <linux/ethtool.h>
48 #include <linux/if_vlan.h>
49 #include <linux/if_macvlan.h>
50 #include <linux/if_bridge.h>
51 #include <linux/prefetch.h>
52 #include <scsi/fc/fc_fcoe.h>
53 #include <net/vxlan.h>
56 #include <linux/of_net.h>
60 #include <asm/idprom.h>
65 #include "ixgbe_common.h"
66 #include "ixgbe_dcb_82599.h"
67 #include "ixgbe_sriov.h"
68 #ifdef CONFIG_IXGBE_VXLAN
69 #include <net/vxlan.h>
72 char ixgbe_driver_name
[] = "ixgbe";
73 static const char ixgbe_driver_string
[] =
74 "Intel(R) 10 Gigabit PCI Express Network Driver";
76 char ixgbe_default_device_descr
[] =
77 "Intel(R) 10 Gigabit Network Connection";
79 static char ixgbe_default_device_descr
[] =
80 "Intel(R) 10 Gigabit Network Connection";
82 #define DRV_VERSION "4.0.1-k"
83 const char ixgbe_driver_version
[] = DRV_VERSION
;
84 static const char ixgbe_copyright
[] =
85 "Copyright (c) 1999-2015 Intel Corporation.";
87 static const char ixgbe_overheat_msg
[] = "Network adapter has been stopped because it has over heated. Restart the computer. If the problem persists, power off the system and replace the adapter";
89 static const struct ixgbe_info
*ixgbe_info_tbl
[] = {
90 [board_82598
] = &ixgbe_82598_info
,
91 [board_82599
] = &ixgbe_82599_info
,
92 [board_X540
] = &ixgbe_X540_info
,
93 [board_X550
] = &ixgbe_X550_info
,
94 [board_X550EM_x
] = &ixgbe_X550EM_x_info
,
97 /* ixgbe_pci_tbl - PCI Device ID Table
99 * Wildcard entries (PCI_ANY_ID) should come last
100 * Last entry must be all 0s
102 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
103 * Class, Class Mask, private data (not used) }
105 static const struct pci_device_id ixgbe_pci_tbl
[] = {
106 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598
), board_82598
},
107 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AF_DUAL_PORT
), board_82598
},
108 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AF_SINGLE_PORT
), board_82598
},
109 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AT
), board_82598
},
110 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AT2
), board_82598
},
111 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598EB_CX4
), board_82598
},
112 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_CX4_DUAL_PORT
), board_82598
},
113 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_DA_DUAL_PORT
), board_82598
},
114 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM
), board_82598
},
115 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598EB_XF_LR
), board_82598
},
116 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598EB_SFP_LOM
), board_82598
},
117 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_BX
), board_82598
},
118 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_KX4
), board_82599
},
119 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_XAUI_LOM
), board_82599
},
120 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_KR
), board_82599
},
121 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP
), board_82599
},
122 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_EM
), board_82599
},
123 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_KX4_MEZZ
), board_82599
},
124 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_CX4
), board_82599
},
125 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_BACKPLANE_FCOE
), board_82599
},
126 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_FCOE
), board_82599
},
127 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_T3_LOM
), board_82599
},
128 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_COMBO_BACKPLANE
), board_82599
},
129 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X540T
), board_X540
},
130 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_SF2
), board_82599
},
131 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_LS
), board_82599
},
132 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_QSFP_SF_QP
), board_82599
},
133 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599EN_SFP
), board_82599
},
134 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_SF_QP
), board_82599
},
135 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X540T1
), board_X540
},
136 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X550T
), board_X550
},
137 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X550EM_X_KX4
), board_X550EM_x
},
138 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X550EM_X_KR
), board_X550EM_x
},
139 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X550EM_X_10G_T
), board_X550EM_x
},
140 /* required last entry */
143 MODULE_DEVICE_TABLE(pci
, ixgbe_pci_tbl
);
145 #ifdef CONFIG_IXGBE_DCA
146 static int ixgbe_notify_dca(struct notifier_block
*, unsigned long event
,
148 static struct notifier_block dca_notifier
= {
149 .notifier_call
= ixgbe_notify_dca
,
155 #ifdef CONFIG_PCI_IOV
156 static unsigned int max_vfs
;
157 module_param(max_vfs
, uint
, 0);
158 MODULE_PARM_DESC(max_vfs
,
159 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
160 #endif /* CONFIG_PCI_IOV */
162 static unsigned int allow_unsupported_sfp
;
163 module_param(allow_unsupported_sfp
, uint
, 0);
164 MODULE_PARM_DESC(allow_unsupported_sfp
,
165 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
167 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
168 static int debug
= -1;
169 module_param(debug
, int, 0);
170 MODULE_PARM_DESC(debug
, "Debug level (0=none,...,16=all)");
172 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
173 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
174 MODULE_LICENSE("GPL");
175 MODULE_VERSION(DRV_VERSION
);
177 static bool ixgbe_check_cfg_remove(struct ixgbe_hw
*hw
, struct pci_dev
*pdev
);
179 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter
*adapter
,
182 struct pci_dev
*parent_dev
;
183 struct pci_bus
*parent_bus
;
185 parent_bus
= adapter
->pdev
->bus
->parent
;
189 parent_dev
= parent_bus
->self
;
193 if (!pci_is_pcie(parent_dev
))
196 pcie_capability_read_word(parent_dev
, reg
, value
);
197 if (*value
== IXGBE_FAILED_READ_CFG_WORD
&&
198 ixgbe_check_cfg_remove(&adapter
->hw
, parent_dev
))
203 static s32
ixgbe_get_parent_bus_info(struct ixgbe_adapter
*adapter
)
205 struct ixgbe_hw
*hw
= &adapter
->hw
;
209 hw
->bus
.type
= ixgbe_bus_type_pci_express
;
211 /* Get the negotiated link width and speed from PCI config space of the
212 * parent, as this device is behind a switch
214 err
= ixgbe_read_pci_cfg_word_parent(adapter
, 18, &link_status
);
216 /* assume caller will handle error case */
220 hw
->bus
.width
= ixgbe_convert_bus_width(link_status
);
221 hw
->bus
.speed
= ixgbe_convert_bus_speed(link_status
);
227 * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
228 * @hw: hw specific details
230 * This function is used by probe to determine whether a device's PCI-Express
231 * bandwidth details should be gathered from the parent bus instead of from the
232 * device. Used to ensure that various locations all have the correct device ID
235 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw
*hw
)
237 switch (hw
->device_id
) {
238 case IXGBE_DEV_ID_82599_SFP_SF_QP
:
239 case IXGBE_DEV_ID_82599_QSFP_SF_QP
:
246 static void ixgbe_check_minimum_link(struct ixgbe_adapter
*adapter
,
249 struct ixgbe_hw
*hw
= &adapter
->hw
;
251 enum pci_bus_speed speed
= PCI_SPEED_UNKNOWN
;
252 enum pcie_link_width width
= PCIE_LNK_WIDTH_UNKNOWN
;
253 struct pci_dev
*pdev
;
255 /* Some devices are not connected over PCIe and thus do not negotiate
256 * speed. These devices do not have valid bus info, and thus any report
257 * we generate may not be correct.
259 if (hw
->bus
.type
== ixgbe_bus_type_internal
)
262 /* determine whether to use the parent device */
263 if (ixgbe_pcie_from_parent(&adapter
->hw
))
264 pdev
= adapter
->pdev
->bus
->parent
->self
;
266 pdev
= adapter
->pdev
;
268 if (pcie_get_minimum_link(pdev
, &speed
, &width
) ||
269 speed
== PCI_SPEED_UNKNOWN
|| width
== PCIE_LNK_WIDTH_UNKNOWN
) {
270 e_dev_warn("Unable to determine PCI Express bandwidth.\n");
275 case PCIE_SPEED_2_5GT
:
276 /* 8b/10b encoding reduces max throughput by 20% */
279 case PCIE_SPEED_5_0GT
:
280 /* 8b/10b encoding reduces max throughput by 20% */
283 case PCIE_SPEED_8_0GT
:
284 /* 128b/130b encoding reduces throughput by less than 2% */
288 e_dev_warn("Unable to determine PCI Express bandwidth.\n");
292 e_dev_info("PCI Express bandwidth of %dGT/s available\n",
294 e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
295 (speed
== PCIE_SPEED_8_0GT
? "8.0GT/s" :
296 speed
== PCIE_SPEED_5_0GT
? "5.0GT/s" :
297 speed
== PCIE_SPEED_2_5GT
? "2.5GT/s" :
300 (speed
== PCIE_SPEED_2_5GT
? "20%" :
301 speed
== PCIE_SPEED_5_0GT
? "20%" :
302 speed
== PCIE_SPEED_8_0GT
? "<2%" :
305 if (max_gts
< expected_gts
) {
306 e_dev_warn("This is not sufficient for optimal performance of this card.\n");
307 e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
309 e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
313 static void ixgbe_service_event_schedule(struct ixgbe_adapter
*adapter
)
315 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
) &&
316 !test_bit(__IXGBE_REMOVING
, &adapter
->state
) &&
317 !test_and_set_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
))
318 schedule_work(&adapter
->service_task
);
321 static void ixgbe_remove_adapter(struct ixgbe_hw
*hw
)
323 struct ixgbe_adapter
*adapter
= hw
->back
;
328 e_dev_err("Adapter removed\n");
329 if (test_bit(__IXGBE_SERVICE_INITED
, &adapter
->state
))
330 ixgbe_service_event_schedule(adapter
);
333 static void ixgbe_check_remove(struct ixgbe_hw
*hw
, u32 reg
)
337 /* The following check not only optimizes a bit by not
338 * performing a read on the status register when the
339 * register just read was a status register read that
340 * returned IXGBE_FAILED_READ_REG. It also blocks any
341 * potential recursion.
343 if (reg
== IXGBE_STATUS
) {
344 ixgbe_remove_adapter(hw
);
347 value
= ixgbe_read_reg(hw
, IXGBE_STATUS
);
348 if (value
== IXGBE_FAILED_READ_REG
)
349 ixgbe_remove_adapter(hw
);
353 * ixgbe_read_reg - Read from device register
354 * @hw: hw specific details
355 * @reg: offset of register to read
357 * Returns : value read or IXGBE_FAILED_READ_REG if removed
359 * This function is used to read device registers. It checks for device
360 * removal by confirming any read that returns all ones by checking the
361 * status register value for all ones. This function avoids reading from
362 * the hardware if a removal was previously detected in which case it
363 * returns IXGBE_FAILED_READ_REG (all ones).
365 u32
ixgbe_read_reg(struct ixgbe_hw
*hw
, u32 reg
)
367 u8 __iomem
*reg_addr
= ACCESS_ONCE(hw
->hw_addr
);
370 if (ixgbe_removed(reg_addr
))
371 return IXGBE_FAILED_READ_REG
;
372 value
= readl(reg_addr
+ reg
);
373 if (unlikely(value
== IXGBE_FAILED_READ_REG
))
374 ixgbe_check_remove(hw
, reg
);
378 static bool ixgbe_check_cfg_remove(struct ixgbe_hw
*hw
, struct pci_dev
*pdev
)
382 pci_read_config_word(pdev
, PCI_VENDOR_ID
, &value
);
383 if (value
== IXGBE_FAILED_READ_CFG_WORD
) {
384 ixgbe_remove_adapter(hw
);
390 u16
ixgbe_read_pci_cfg_word(struct ixgbe_hw
*hw
, u32 reg
)
392 struct ixgbe_adapter
*adapter
= hw
->back
;
395 if (ixgbe_removed(hw
->hw_addr
))
396 return IXGBE_FAILED_READ_CFG_WORD
;
397 pci_read_config_word(adapter
->pdev
, reg
, &value
);
398 if (value
== IXGBE_FAILED_READ_CFG_WORD
&&
399 ixgbe_check_cfg_remove(hw
, adapter
->pdev
))
400 return IXGBE_FAILED_READ_CFG_WORD
;
404 #ifdef CONFIG_PCI_IOV
405 static u32
ixgbe_read_pci_cfg_dword(struct ixgbe_hw
*hw
, u32 reg
)
407 struct ixgbe_adapter
*adapter
= hw
->back
;
410 if (ixgbe_removed(hw
->hw_addr
))
411 return IXGBE_FAILED_READ_CFG_DWORD
;
412 pci_read_config_dword(adapter
->pdev
, reg
, &value
);
413 if (value
== IXGBE_FAILED_READ_CFG_DWORD
&&
414 ixgbe_check_cfg_remove(hw
, adapter
->pdev
))
415 return IXGBE_FAILED_READ_CFG_DWORD
;
418 #endif /* CONFIG_PCI_IOV */
420 void ixgbe_write_pci_cfg_word(struct ixgbe_hw
*hw
, u32 reg
, u16 value
)
422 struct ixgbe_adapter
*adapter
= hw
->back
;
424 if (ixgbe_removed(hw
->hw_addr
))
426 pci_write_config_word(adapter
->pdev
, reg
, value
);
429 static void ixgbe_service_event_complete(struct ixgbe_adapter
*adapter
)
431 BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
));
433 /* flush memory to make sure state is correct before next watchdog */
434 smp_mb__before_atomic();
435 clear_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
);
438 struct ixgbe_reg_info
{
443 static const struct ixgbe_reg_info ixgbe_reg_info_tbl
[] = {
445 /* General Registers */
446 {IXGBE_CTRL
, "CTRL"},
447 {IXGBE_STATUS
, "STATUS"},
448 {IXGBE_CTRL_EXT
, "CTRL_EXT"},
450 /* Interrupt Registers */
451 {IXGBE_EICR
, "EICR"},
454 {IXGBE_SRRCTL(0), "SRRCTL"},
455 {IXGBE_DCA_RXCTRL(0), "DRXCTL"},
456 {IXGBE_RDLEN(0), "RDLEN"},
457 {IXGBE_RDH(0), "RDH"},
458 {IXGBE_RDT(0), "RDT"},
459 {IXGBE_RXDCTL(0), "RXDCTL"},
460 {IXGBE_RDBAL(0), "RDBAL"},
461 {IXGBE_RDBAH(0), "RDBAH"},
464 {IXGBE_TDBAL(0), "TDBAL"},
465 {IXGBE_TDBAH(0), "TDBAH"},
466 {IXGBE_TDLEN(0), "TDLEN"},
467 {IXGBE_TDH(0), "TDH"},
468 {IXGBE_TDT(0), "TDT"},
469 {IXGBE_TXDCTL(0), "TXDCTL"},
471 /* List Terminator */
477 * ixgbe_regdump - register printout routine
479 static void ixgbe_regdump(struct ixgbe_hw
*hw
, struct ixgbe_reg_info
*reginfo
)
485 switch (reginfo
->ofs
) {
486 case IXGBE_SRRCTL(0):
487 for (i
= 0; i
< 64; i
++)
488 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_SRRCTL(i
));
490 case IXGBE_DCA_RXCTRL(0):
491 for (i
= 0; i
< 64; i
++)
492 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_DCA_RXCTRL(i
));
495 for (i
= 0; i
< 64; i
++)
496 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDLEN(i
));
499 for (i
= 0; i
< 64; i
++)
500 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDH(i
));
503 for (i
= 0; i
< 64; i
++)
504 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDT(i
));
506 case IXGBE_RXDCTL(0):
507 for (i
= 0; i
< 64; i
++)
508 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RXDCTL(i
));
511 for (i
= 0; i
< 64; i
++)
512 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDBAL(i
));
515 for (i
= 0; i
< 64; i
++)
516 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDBAH(i
));
519 for (i
= 0; i
< 64; i
++)
520 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDBAL(i
));
523 for (i
= 0; i
< 64; i
++)
524 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDBAH(i
));
527 for (i
= 0; i
< 64; i
++)
528 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDLEN(i
));
531 for (i
= 0; i
< 64; i
++)
532 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDH(i
));
535 for (i
= 0; i
< 64; i
++)
536 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDT(i
));
538 case IXGBE_TXDCTL(0):
539 for (i
= 0; i
< 64; i
++)
540 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TXDCTL(i
));
543 pr_info("%-15s %08x\n", reginfo
->name
,
544 IXGBE_READ_REG(hw
, reginfo
->ofs
));
548 for (i
= 0; i
< 8; i
++) {
549 snprintf(rname
, 16, "%s[%d-%d]", reginfo
->name
, i
*8, i
*8+7);
550 pr_err("%-15s", rname
);
551 for (j
= 0; j
< 8; j
++)
552 pr_cont(" %08x", regs
[i
*8+j
]);
559 * ixgbe_dump - Print registers, tx-rings and rx-rings
561 static void ixgbe_dump(struct ixgbe_adapter
*adapter
)
563 struct net_device
*netdev
= adapter
->netdev
;
564 struct ixgbe_hw
*hw
= &adapter
->hw
;
565 struct ixgbe_reg_info
*reginfo
;
567 struct ixgbe_ring
*tx_ring
;
568 struct ixgbe_tx_buffer
*tx_buffer
;
569 union ixgbe_adv_tx_desc
*tx_desc
;
570 struct my_u0
{ u64 a
; u64 b
; } *u0
;
571 struct ixgbe_ring
*rx_ring
;
572 union ixgbe_adv_rx_desc
*rx_desc
;
573 struct ixgbe_rx_buffer
*rx_buffer_info
;
577 if (!netif_msg_hw(adapter
))
580 /* Print netdevice Info */
582 dev_info(&adapter
->pdev
->dev
, "Net device Info\n");
583 pr_info("Device Name state "
584 "trans_start last_rx\n");
585 pr_info("%-15s %016lX %016lX %016lX\n",
592 /* Print Registers */
593 dev_info(&adapter
->pdev
->dev
, "Register Dump\n");
594 pr_info(" Register Name Value\n");
595 for (reginfo
= (struct ixgbe_reg_info
*)ixgbe_reg_info_tbl
;
596 reginfo
->name
; reginfo
++) {
597 ixgbe_regdump(hw
, reginfo
);
600 /* Print TX Ring Summary */
601 if (!netdev
|| !netif_running(netdev
))
604 dev_info(&adapter
->pdev
->dev
, "TX Rings Summary\n");
605 pr_info(" %s %s %s %s\n",
606 "Queue [NTU] [NTC] [bi(ntc)->dma ]",
607 "leng", "ntw", "timestamp");
608 for (n
= 0; n
< adapter
->num_tx_queues
; n
++) {
609 tx_ring
= adapter
->tx_ring
[n
];
610 tx_buffer
= &tx_ring
->tx_buffer_info
[tx_ring
->next_to_clean
];
611 pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
612 n
, tx_ring
->next_to_use
, tx_ring
->next_to_clean
,
613 (u64
)dma_unmap_addr(tx_buffer
, dma
),
614 dma_unmap_len(tx_buffer
, len
),
615 tx_buffer
->next_to_watch
,
616 (u64
)tx_buffer
->time_stamp
);
620 if (!netif_msg_tx_done(adapter
))
621 goto rx_ring_summary
;
623 dev_info(&adapter
->pdev
->dev
, "TX Rings Dump\n");
625 /* Transmit Descriptor Formats
627 * 82598 Advanced Transmit Descriptor
628 * +--------------------------------------------------------------+
629 * 0 | Buffer Address [63:0] |
630 * +--------------------------------------------------------------+
631 * 8 | PAYLEN | POPTS | IDX | STA | DCMD |DTYP | RSV | DTALEN |
632 * +--------------------------------------------------------------+
633 * 63 46 45 40 39 36 35 32 31 24 23 20 19 0
635 * 82598 Advanced Transmit Descriptor (Write-Back Format)
636 * +--------------------------------------------------------------+
638 * +--------------------------------------------------------------+
639 * 8 | RSV | STA | NXTSEQ |
640 * +--------------------------------------------------------------+
643 * 82599+ Advanced Transmit Descriptor
644 * +--------------------------------------------------------------+
645 * 0 | Buffer Address [63:0] |
646 * +--------------------------------------------------------------+
647 * 8 |PAYLEN |POPTS|CC|IDX |STA |DCMD |DTYP |MAC |RSV |DTALEN |
648 * +--------------------------------------------------------------+
649 * 63 46 45 40 39 38 36 35 32 31 24 23 20 19 18 17 16 15 0
651 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
652 * +--------------------------------------------------------------+
654 * +--------------------------------------------------------------+
655 * 8 | RSV | STA | RSV |
656 * +--------------------------------------------------------------+
660 for (n
= 0; n
< adapter
->num_tx_queues
; n
++) {
661 tx_ring
= adapter
->tx_ring
[n
];
662 pr_info("------------------------------------\n");
663 pr_info("TX QUEUE INDEX = %d\n", tx_ring
->queue_index
);
664 pr_info("------------------------------------\n");
665 pr_info("%s%s %s %s %s %s\n",
666 "T [desc] [address 63:0 ] ",
667 "[PlPOIdStDDt Ln] [bi->dma ] ",
668 "leng", "ntw", "timestamp", "bi->skb");
670 for (i
= 0; tx_ring
->desc
&& (i
< tx_ring
->count
); i
++) {
671 tx_desc
= IXGBE_TX_DESC(tx_ring
, i
);
672 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
673 u0
= (struct my_u0
*)tx_desc
;
674 if (dma_unmap_len(tx_buffer
, len
) > 0) {
675 pr_info("T [0x%03X] %016llX %016llX %016llX %08X %p %016llX %p",
679 (u64
)dma_unmap_addr(tx_buffer
, dma
),
680 dma_unmap_len(tx_buffer
, len
),
681 tx_buffer
->next_to_watch
,
682 (u64
)tx_buffer
->time_stamp
,
684 if (i
== tx_ring
->next_to_use
&&
685 i
== tx_ring
->next_to_clean
)
687 else if (i
== tx_ring
->next_to_use
)
689 else if (i
== tx_ring
->next_to_clean
)
694 if (netif_msg_pktdata(adapter
) &&
696 print_hex_dump(KERN_INFO
, "",
697 DUMP_PREFIX_ADDRESS
, 16, 1,
698 tx_buffer
->skb
->data
,
699 dma_unmap_len(tx_buffer
, len
),
705 /* Print RX Rings Summary */
707 dev_info(&adapter
->pdev
->dev
, "RX Rings Summary\n");
708 pr_info("Queue [NTU] [NTC]\n");
709 for (n
= 0; n
< adapter
->num_rx_queues
; n
++) {
710 rx_ring
= adapter
->rx_ring
[n
];
711 pr_info("%5d %5X %5X\n",
712 n
, rx_ring
->next_to_use
, rx_ring
->next_to_clean
);
716 if (!netif_msg_rx_status(adapter
))
719 dev_info(&adapter
->pdev
->dev
, "RX Rings Dump\n");
721 /* Receive Descriptor Formats
723 * 82598 Advanced Receive Descriptor (Read) Format
725 * +-----------------------------------------------------+
726 * 0 | Packet Buffer Address [63:1] |A0/NSE|
727 * +----------------------------------------------+------+
728 * 8 | Header Buffer Address [63:1] | DD |
729 * +-----------------------------------------------------+
732 * 82598 Advanced Receive Descriptor (Write-Back) Format
734 * 63 48 47 32 31 30 21 20 16 15 4 3 0
735 * +------------------------------------------------------+
736 * 0 | RSS Hash / |SPH| HDR_LEN | RSV |Packet| RSS |
737 * | Packet | IP | | | | Type | Type |
738 * | Checksum | Ident | | | | | |
739 * +------------------------------------------------------+
740 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
741 * +------------------------------------------------------+
742 * 63 48 47 32 31 20 19 0
744 * 82599+ Advanced Receive Descriptor (Read) Format
746 * +-----------------------------------------------------+
747 * 0 | Packet Buffer Address [63:1] |A0/NSE|
748 * +----------------------------------------------+------+
749 * 8 | Header Buffer Address [63:1] | DD |
750 * +-----------------------------------------------------+
753 * 82599+ Advanced Receive Descriptor (Write-Back) Format
755 * 63 48 47 32 31 30 21 20 17 16 4 3 0
756 * +------------------------------------------------------+
757 * 0 |RSS / Frag Checksum|SPH| HDR_LEN |RSC- |Packet| RSS |
758 * |/ RTT / PCoE_PARAM | | | CNT | Type | Type |
759 * |/ Flow Dir Flt ID | | | | | |
760 * +------------------------------------------------------+
761 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
762 * +------------------------------------------------------+
763 * 63 48 47 32 31 20 19 0
766 for (n
= 0; n
< adapter
->num_rx_queues
; n
++) {
767 rx_ring
= adapter
->rx_ring
[n
];
768 pr_info("------------------------------------\n");
769 pr_info("RX QUEUE INDEX = %d\n", rx_ring
->queue_index
);
770 pr_info("------------------------------------\n");
772 "R [desc] [ PktBuf A0] ",
773 "[ HeadBuf DD] [bi->dma ] [bi->skb ] ",
774 "<-- Adv Rx Read format\n");
776 "RWB[desc] [PcsmIpSHl PtRs] ",
777 "[vl er S cks ln] ---------------- [bi->skb ] ",
778 "<-- Adv Rx Write-Back format\n");
780 for (i
= 0; i
< rx_ring
->count
; i
++) {
781 rx_buffer_info
= &rx_ring
->rx_buffer_info
[i
];
782 rx_desc
= IXGBE_RX_DESC(rx_ring
, i
);
783 u0
= (struct my_u0
*)rx_desc
;
784 staterr
= le32_to_cpu(rx_desc
->wb
.upper
.status_error
);
785 if (staterr
& IXGBE_RXD_STAT_DD
) {
786 /* Descriptor Done */
787 pr_info("RWB[0x%03X] %016llX "
788 "%016llX ---------------- %p", i
,
791 rx_buffer_info
->skb
);
793 pr_info("R [0x%03X] %016llX "
794 "%016llX %016llX %p", i
,
797 (u64
)rx_buffer_info
->dma
,
798 rx_buffer_info
->skb
);
800 if (netif_msg_pktdata(adapter
) &&
801 rx_buffer_info
->dma
) {
802 print_hex_dump(KERN_INFO
, "",
803 DUMP_PREFIX_ADDRESS
, 16, 1,
804 page_address(rx_buffer_info
->page
) +
805 rx_buffer_info
->page_offset
,
806 ixgbe_rx_bufsz(rx_ring
), true);
810 if (i
== rx_ring
->next_to_use
)
812 else if (i
== rx_ring
->next_to_clean
)
821 static void ixgbe_release_hw_control(struct ixgbe_adapter
*adapter
)
825 /* Let firmware take over control of h/w */
826 ctrl_ext
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_CTRL_EXT
);
827 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_CTRL_EXT
,
828 ctrl_ext
& ~IXGBE_CTRL_EXT_DRV_LOAD
);
831 static void ixgbe_get_hw_control(struct ixgbe_adapter
*adapter
)
835 /* Let firmware know the driver has taken over */
836 ctrl_ext
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_CTRL_EXT
);
837 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_CTRL_EXT
,
838 ctrl_ext
| IXGBE_CTRL_EXT_DRV_LOAD
);
842 * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
843 * @adapter: pointer to adapter struct
844 * @direction: 0 for Rx, 1 for Tx, -1 for other causes
845 * @queue: queue to map the corresponding interrupt to
846 * @msix_vector: the vector to map to the corresponding queue
849 static void ixgbe_set_ivar(struct ixgbe_adapter
*adapter
, s8 direction
,
850 u8 queue
, u8 msix_vector
)
853 struct ixgbe_hw
*hw
= &adapter
->hw
;
854 switch (hw
->mac
.type
) {
855 case ixgbe_mac_82598EB
:
856 msix_vector
|= IXGBE_IVAR_ALLOC_VAL
;
859 index
= (((direction
* 64) + queue
) >> 2) & 0x1F;
860 ivar
= IXGBE_READ_REG(hw
, IXGBE_IVAR(index
));
861 ivar
&= ~(0xFF << (8 * (queue
& 0x3)));
862 ivar
|= (msix_vector
<< (8 * (queue
& 0x3)));
863 IXGBE_WRITE_REG(hw
, IXGBE_IVAR(index
), ivar
);
865 case ixgbe_mac_82599EB
:
868 case ixgbe_mac_X550EM_x
:
869 if (direction
== -1) {
871 msix_vector
|= IXGBE_IVAR_ALLOC_VAL
;
872 index
= ((queue
& 1) * 8);
873 ivar
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_IVAR_MISC
);
874 ivar
&= ~(0xFF << index
);
875 ivar
|= (msix_vector
<< index
);
876 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_IVAR_MISC
, ivar
);
879 /* tx or rx causes */
880 msix_vector
|= IXGBE_IVAR_ALLOC_VAL
;
881 index
= ((16 * (queue
& 1)) + (8 * direction
));
882 ivar
= IXGBE_READ_REG(hw
, IXGBE_IVAR(queue
>> 1));
883 ivar
&= ~(0xFF << index
);
884 ivar
|= (msix_vector
<< index
);
885 IXGBE_WRITE_REG(hw
, IXGBE_IVAR(queue
>> 1), ivar
);
893 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter
*adapter
,
898 switch (adapter
->hw
.mac
.type
) {
899 case ixgbe_mac_82598EB
:
900 mask
= (IXGBE_EIMS_RTX_QUEUE
& qmask
);
901 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EICS
, mask
);
903 case ixgbe_mac_82599EB
:
906 case ixgbe_mac_X550EM_x
:
907 mask
= (qmask
& 0xFFFFFFFF);
908 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EICS_EX(0), mask
);
909 mask
= (qmask
>> 32);
910 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EICS_EX(1), mask
);
917 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring
*ring
,
918 struct ixgbe_tx_buffer
*tx_buffer
)
920 if (tx_buffer
->skb
) {
921 dev_kfree_skb_any(tx_buffer
->skb
);
922 if (dma_unmap_len(tx_buffer
, len
))
923 dma_unmap_single(ring
->dev
,
924 dma_unmap_addr(tx_buffer
, dma
),
925 dma_unmap_len(tx_buffer
, len
),
927 } else if (dma_unmap_len(tx_buffer
, len
)) {
928 dma_unmap_page(ring
->dev
,
929 dma_unmap_addr(tx_buffer
, dma
),
930 dma_unmap_len(tx_buffer
, len
),
933 tx_buffer
->next_to_watch
= NULL
;
934 tx_buffer
->skb
= NULL
;
935 dma_unmap_len_set(tx_buffer
, len
, 0);
936 /* tx_buffer must be completely set up in the transmit path */
939 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter
*adapter
)
941 struct ixgbe_hw
*hw
= &adapter
->hw
;
942 struct ixgbe_hw_stats
*hwstats
= &adapter
->stats
;
946 if ((hw
->fc
.current_mode
!= ixgbe_fc_full
) &&
947 (hw
->fc
.current_mode
!= ixgbe_fc_rx_pause
))
950 switch (hw
->mac
.type
) {
951 case ixgbe_mac_82598EB
:
952 data
= IXGBE_READ_REG(hw
, IXGBE_LXOFFRXC
);
955 data
= IXGBE_READ_REG(hw
, IXGBE_LXOFFRXCNT
);
957 hwstats
->lxoffrxc
+= data
;
959 /* refill credits (no tx hang) if we received xoff */
963 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
964 clear_bit(__IXGBE_HANG_CHECK_ARMED
,
965 &adapter
->tx_ring
[i
]->state
);
968 static void ixgbe_update_xoff_received(struct ixgbe_adapter
*adapter
)
970 struct ixgbe_hw
*hw
= &adapter
->hw
;
971 struct ixgbe_hw_stats
*hwstats
= &adapter
->stats
;
975 bool pfc_en
= adapter
->dcb_cfg
.pfc_mode_enable
;
977 if (adapter
->ixgbe_ieee_pfc
)
978 pfc_en
|= !!(adapter
->ixgbe_ieee_pfc
->pfc_en
);
980 if (!(adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
) || !pfc_en
) {
981 ixgbe_update_xoff_rx_lfc(adapter
);
985 /* update stats for each tc, only valid with PFC enabled */
986 for (i
= 0; i
< MAX_TX_PACKET_BUFFERS
; i
++) {
989 switch (hw
->mac
.type
) {
990 case ixgbe_mac_82598EB
:
991 pxoffrxc
= IXGBE_READ_REG(hw
, IXGBE_PXOFFRXC(i
));
994 pxoffrxc
= IXGBE_READ_REG(hw
, IXGBE_PXOFFRXCNT(i
));
996 hwstats
->pxoffrxc
[i
] += pxoffrxc
;
997 /* Get the TC for given UP */
998 tc
= netdev_get_prio_tc_map(adapter
->netdev
, i
);
999 xoff
[tc
] += pxoffrxc
;
1002 /* disarm tx queues that have received xoff frames */
1003 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
1004 struct ixgbe_ring
*tx_ring
= adapter
->tx_ring
[i
];
1006 tc
= tx_ring
->dcb_tc
;
1008 clear_bit(__IXGBE_HANG_CHECK_ARMED
, &tx_ring
->state
);
1012 static u64
ixgbe_get_tx_completed(struct ixgbe_ring
*ring
)
1014 return ring
->stats
.packets
;
1017 static u64
ixgbe_get_tx_pending(struct ixgbe_ring
*ring
)
1019 struct ixgbe_adapter
*adapter
;
1020 struct ixgbe_hw
*hw
;
1023 if (ring
->l2_accel_priv
)
1024 adapter
= ring
->l2_accel_priv
->real_adapter
;
1026 adapter
= netdev_priv(ring
->netdev
);
1029 head
= IXGBE_READ_REG(hw
, IXGBE_TDH(ring
->reg_idx
));
1030 tail
= IXGBE_READ_REG(hw
, IXGBE_TDT(ring
->reg_idx
));
1033 return (head
< tail
) ?
1034 tail
- head
: (tail
+ ring
->count
- head
);
1039 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring
*tx_ring
)
1041 u32 tx_done
= ixgbe_get_tx_completed(tx_ring
);
1042 u32 tx_done_old
= tx_ring
->tx_stats
.tx_done_old
;
1043 u32 tx_pending
= ixgbe_get_tx_pending(tx_ring
);
1045 clear_check_for_tx_hang(tx_ring
);
1048 * Check for a hung queue, but be thorough. This verifies
1049 * that a transmit has been completed since the previous
1050 * check AND there is at least one packet pending. The
1051 * ARMED bit is set to indicate a potential hang. The
1052 * bit is cleared if a pause frame is received to remove
1053 * false hang detection due to PFC or 802.3x frames. By
1054 * requiring this to fail twice we avoid races with
1055 * pfc clearing the ARMED bit and conditions where we
1056 * run the check_tx_hang logic with a transmit completion
1057 * pending but without time to complete it yet.
1059 if (tx_done_old
== tx_done
&& tx_pending
)
1060 /* make sure it is true for two checks in a row */
1061 return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED
,
1063 /* update completed stats and continue */
1064 tx_ring
->tx_stats
.tx_done_old
= tx_done
;
1065 /* reset the countdown */
1066 clear_bit(__IXGBE_HANG_CHECK_ARMED
, &tx_ring
->state
);
1072 * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1073 * @adapter: driver private struct
1075 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter
*adapter
)
1078 /* Do the reset outside of interrupt context */
1079 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
1080 adapter
->flags2
|= IXGBE_FLAG2_RESET_REQUESTED
;
1081 e_warn(drv
, "initiating reset due to tx timeout\n");
1082 ixgbe_service_event_schedule(adapter
);
1087 * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1088 * @q_vector: structure containing interrupt and ring information
1089 * @tx_ring: tx ring to clean
1091 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector
*q_vector
,
1092 struct ixgbe_ring
*tx_ring
)
1094 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
1095 struct ixgbe_tx_buffer
*tx_buffer
;
1096 union ixgbe_adv_tx_desc
*tx_desc
;
1097 unsigned int total_bytes
= 0, total_packets
= 0;
1098 unsigned int budget
= q_vector
->tx
.work_limit
;
1099 unsigned int i
= tx_ring
->next_to_clean
;
1101 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
1104 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
1105 tx_desc
= IXGBE_TX_DESC(tx_ring
, i
);
1106 i
-= tx_ring
->count
;
1109 union ixgbe_adv_tx_desc
*eop_desc
= tx_buffer
->next_to_watch
;
1111 /* if next_to_watch is not set then there is no work pending */
1115 /* prevent any other reads prior to eop_desc */
1116 read_barrier_depends();
1118 /* if DD is not set pending work has not been completed */
1119 if (!(eop_desc
->wb
.status
& cpu_to_le32(IXGBE_TXD_STAT_DD
)))
1122 /* clear next_to_watch to prevent false hangs */
1123 tx_buffer
->next_to_watch
= NULL
;
1125 /* update the statistics for this packet */
1126 total_bytes
+= tx_buffer
->bytecount
;
1127 total_packets
+= tx_buffer
->gso_segs
;
1130 dev_consume_skb_any(tx_buffer
->skb
);
1132 /* unmap skb header data */
1133 dma_unmap_single(tx_ring
->dev
,
1134 dma_unmap_addr(tx_buffer
, dma
),
1135 dma_unmap_len(tx_buffer
, len
),
1138 /* clear tx_buffer data */
1139 tx_buffer
->skb
= NULL
;
1140 dma_unmap_len_set(tx_buffer
, len
, 0);
1142 /* unmap remaining buffers */
1143 while (tx_desc
!= eop_desc
) {
1148 i
-= tx_ring
->count
;
1149 tx_buffer
= tx_ring
->tx_buffer_info
;
1150 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
1153 /* unmap any remaining paged data */
1154 if (dma_unmap_len(tx_buffer
, len
)) {
1155 dma_unmap_page(tx_ring
->dev
,
1156 dma_unmap_addr(tx_buffer
, dma
),
1157 dma_unmap_len(tx_buffer
, len
),
1159 dma_unmap_len_set(tx_buffer
, len
, 0);
1163 /* move us one more past the eop_desc for start of next pkt */
1168 i
-= tx_ring
->count
;
1169 tx_buffer
= tx_ring
->tx_buffer_info
;
1170 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
1173 /* issue prefetch for next Tx descriptor */
1176 /* update budget accounting */
1178 } while (likely(budget
));
1180 i
+= tx_ring
->count
;
1181 tx_ring
->next_to_clean
= i
;
1182 u64_stats_update_begin(&tx_ring
->syncp
);
1183 tx_ring
->stats
.bytes
+= total_bytes
;
1184 tx_ring
->stats
.packets
+= total_packets
;
1185 u64_stats_update_end(&tx_ring
->syncp
);
1186 q_vector
->tx
.total_bytes
+= total_bytes
;
1187 q_vector
->tx
.total_packets
+= total_packets
;
1189 if (check_for_tx_hang(tx_ring
) && ixgbe_check_tx_hang(tx_ring
)) {
1190 /* schedule immediate reset if we believe we hung */
1191 struct ixgbe_hw
*hw
= &adapter
->hw
;
1192 e_err(drv
, "Detected Tx Unit Hang\n"
1194 " TDH, TDT <%x>, <%x>\n"
1195 " next_to_use <%x>\n"
1196 " next_to_clean <%x>\n"
1197 "tx_buffer_info[next_to_clean]\n"
1198 " time_stamp <%lx>\n"
1200 tx_ring
->queue_index
,
1201 IXGBE_READ_REG(hw
, IXGBE_TDH(tx_ring
->reg_idx
)),
1202 IXGBE_READ_REG(hw
, IXGBE_TDT(tx_ring
->reg_idx
)),
1203 tx_ring
->next_to_use
, i
,
1204 tx_ring
->tx_buffer_info
[i
].time_stamp
, jiffies
);
1206 netif_stop_subqueue(tx_ring
->netdev
, tx_ring
->queue_index
);
1209 "tx hang %d detected on queue %d, resetting adapter\n",
1210 adapter
->tx_timeout_count
+ 1, tx_ring
->queue_index
);
1212 /* schedule immediate reset if we believe we hung */
1213 ixgbe_tx_timeout_reset(adapter
);
1215 /* the adapter is about to reset, no point in enabling stuff */
1219 netdev_tx_completed_queue(txring_txq(tx_ring
),
1220 total_packets
, total_bytes
);
1222 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1223 if (unlikely(total_packets
&& netif_carrier_ok(tx_ring
->netdev
) &&
1224 (ixgbe_desc_unused(tx_ring
) >= TX_WAKE_THRESHOLD
))) {
1225 /* Make sure that anybody stopping the queue after this
1226 * sees the new next_to_clean.
1229 if (__netif_subqueue_stopped(tx_ring
->netdev
,
1230 tx_ring
->queue_index
)
1231 && !test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
1232 netif_wake_subqueue(tx_ring
->netdev
,
1233 tx_ring
->queue_index
);
1234 ++tx_ring
->tx_stats
.restart_queue
;
1241 #ifdef CONFIG_IXGBE_DCA
1242 static void ixgbe_update_tx_dca(struct ixgbe_adapter
*adapter
,
1243 struct ixgbe_ring
*tx_ring
,
1246 struct ixgbe_hw
*hw
= &adapter
->hw
;
1247 u32 txctrl
= dca3_get_tag(tx_ring
->dev
, cpu
);
1250 switch (hw
->mac
.type
) {
1251 case ixgbe_mac_82598EB
:
1252 reg_offset
= IXGBE_DCA_TXCTRL(tx_ring
->reg_idx
);
1254 case ixgbe_mac_82599EB
:
1255 case ixgbe_mac_X540
:
1256 reg_offset
= IXGBE_DCA_TXCTRL_82599(tx_ring
->reg_idx
);
1257 txctrl
<<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599
;
1260 /* for unknown hardware do not write register */
1265 * We can enable relaxed ordering for reads, but not writes when
1266 * DCA is enabled. This is due to a known issue in some chipsets
1267 * which will cause the DCA tag to be cleared.
1269 txctrl
|= IXGBE_DCA_TXCTRL_DESC_RRO_EN
|
1270 IXGBE_DCA_TXCTRL_DATA_RRO_EN
|
1271 IXGBE_DCA_TXCTRL_DESC_DCA_EN
;
1273 IXGBE_WRITE_REG(hw
, reg_offset
, txctrl
);
1276 static void ixgbe_update_rx_dca(struct ixgbe_adapter
*adapter
,
1277 struct ixgbe_ring
*rx_ring
,
1280 struct ixgbe_hw
*hw
= &adapter
->hw
;
1281 u32 rxctrl
= dca3_get_tag(rx_ring
->dev
, cpu
);
1282 u8 reg_idx
= rx_ring
->reg_idx
;
1285 switch (hw
->mac
.type
) {
1286 case ixgbe_mac_82599EB
:
1287 case ixgbe_mac_X540
:
1288 rxctrl
<<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599
;
1295 * We can enable relaxed ordering for reads, but not writes when
1296 * DCA is enabled. This is due to a known issue in some chipsets
1297 * which will cause the DCA tag to be cleared.
1299 rxctrl
|= IXGBE_DCA_RXCTRL_DESC_RRO_EN
|
1300 IXGBE_DCA_RXCTRL_DESC_DCA_EN
;
1302 IXGBE_WRITE_REG(hw
, IXGBE_DCA_RXCTRL(reg_idx
), rxctrl
);
1305 static void ixgbe_update_dca(struct ixgbe_q_vector
*q_vector
)
1307 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
1308 struct ixgbe_ring
*ring
;
1309 int cpu
= get_cpu();
1311 if (q_vector
->cpu
== cpu
)
1314 ixgbe_for_each_ring(ring
, q_vector
->tx
)
1315 ixgbe_update_tx_dca(adapter
, ring
, cpu
);
1317 ixgbe_for_each_ring(ring
, q_vector
->rx
)
1318 ixgbe_update_rx_dca(adapter
, ring
, cpu
);
1320 q_vector
->cpu
= cpu
;
1325 static void ixgbe_setup_dca(struct ixgbe_adapter
*adapter
)
1329 if (!(adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
))
1332 /* always use CB2 mode, difference is masked in the CB driver */
1333 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_DCA_CTRL
, 2);
1335 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
1336 adapter
->q_vector
[i
]->cpu
= -1;
1337 ixgbe_update_dca(adapter
->q_vector
[i
]);
1341 static int __ixgbe_notify_dca(struct device
*dev
, void *data
)
1343 struct ixgbe_adapter
*adapter
= dev_get_drvdata(dev
);
1344 unsigned long event
= *(unsigned long *)data
;
1346 if (!(adapter
->flags
& IXGBE_FLAG_DCA_CAPABLE
))
1350 case DCA_PROVIDER_ADD
:
1351 /* if we're already enabled, don't do it again */
1352 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
)
1354 if (dca_add_requester(dev
) == 0) {
1355 adapter
->flags
|= IXGBE_FLAG_DCA_ENABLED
;
1356 ixgbe_setup_dca(adapter
);
1359 /* Fall Through since DCA is disabled. */
1360 case DCA_PROVIDER_REMOVE
:
1361 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
) {
1362 dca_remove_requester(dev
);
1363 adapter
->flags
&= ~IXGBE_FLAG_DCA_ENABLED
;
1364 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_DCA_CTRL
, 1);
1372 #endif /* CONFIG_IXGBE_DCA */
1374 #define IXGBE_RSS_L4_TYPES_MASK \
1375 ((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1376 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1377 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1378 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1380 static inline void ixgbe_rx_hash(struct ixgbe_ring
*ring
,
1381 union ixgbe_adv_rx_desc
*rx_desc
,
1382 struct sk_buff
*skb
)
1386 if (!(ring
->netdev
->features
& NETIF_F_RXHASH
))
1389 rss_type
= le16_to_cpu(rx_desc
->wb
.lower
.lo_dword
.hs_rss
.pkt_info
) &
1390 IXGBE_RXDADV_RSSTYPE_MASK
;
1395 skb_set_hash(skb
, le32_to_cpu(rx_desc
->wb
.lower
.hi_dword
.rss
),
1396 (IXGBE_RSS_L4_TYPES_MASK
& (1ul << rss_type
)) ?
1397 PKT_HASH_TYPE_L4
: PKT_HASH_TYPE_L3
);
1402 * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1403 * @ring: structure containing ring specific data
1404 * @rx_desc: advanced rx descriptor
1406 * Returns : true if it is FCoE pkt
1408 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring
*ring
,
1409 union ixgbe_adv_rx_desc
*rx_desc
)
1411 __le16 pkt_info
= rx_desc
->wb
.lower
.lo_dword
.hs_rss
.pkt_info
;
1413 return test_bit(__IXGBE_RX_FCOE
, &ring
->state
) &&
1414 ((pkt_info
& cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK
)) ==
1415 (cpu_to_le16(IXGBE_ETQF_FILTER_FCOE
<<
1416 IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT
)));
1419 #endif /* IXGBE_FCOE */
1421 * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1422 * @ring: structure containing ring specific data
1423 * @rx_desc: current Rx descriptor being processed
1424 * @skb: skb currently being received and modified
1426 static inline void ixgbe_rx_checksum(struct ixgbe_ring
*ring
,
1427 union ixgbe_adv_rx_desc
*rx_desc
,
1428 struct sk_buff
*skb
)
1430 __le16 pkt_info
= rx_desc
->wb
.lower
.lo_dword
.hs_rss
.pkt_info
;
1431 __le16 hdr_info
= rx_desc
->wb
.lower
.lo_dword
.hs_rss
.hdr_info
;
1432 bool encap_pkt
= false;
1434 skb_checksum_none_assert(skb
);
1436 /* Rx csum disabled */
1437 if (!(ring
->netdev
->features
& NETIF_F_RXCSUM
))
1440 if ((pkt_info
& cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN
)) &&
1441 (hdr_info
& cpu_to_le16(IXGBE_RXDADV_PKTTYPE_TUNNEL
>> 16))) {
1443 skb
->encapsulation
= 1;
1446 /* if IP and error */
1447 if (ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_IPCS
) &&
1448 ixgbe_test_staterr(rx_desc
, IXGBE_RXDADV_ERR_IPE
)) {
1449 ring
->rx_stats
.csum_err
++;
1453 if (!ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_L4CS
))
1456 if (ixgbe_test_staterr(rx_desc
, IXGBE_RXDADV_ERR_TCPE
)) {
1458 * 82599 errata, UDP frames with a 0 checksum can be marked as
1461 if ((pkt_info
& cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP
)) &&
1462 test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR
, &ring
->state
))
1465 ring
->rx_stats
.csum_err
++;
1469 /* It must be a TCP or UDP packet with a valid checksum */
1470 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1472 if (!ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_OUTERIPCS
))
1475 if (ixgbe_test_staterr(rx_desc
, IXGBE_RXDADV_ERR_OUTERIPER
)) {
1476 ring
->rx_stats
.csum_err
++;
1479 /* If we checked the outer header let the stack know */
1480 skb
->csum_level
= 1;
1484 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring
*rx_ring
,
1485 struct ixgbe_rx_buffer
*bi
)
1487 struct page
*page
= bi
->page
;
1490 /* since we are recycling buffers we should seldom need to alloc */
1494 /* alloc new page for storage */
1495 page
= dev_alloc_pages(ixgbe_rx_pg_order(rx_ring
));
1496 if (unlikely(!page
)) {
1497 rx_ring
->rx_stats
.alloc_rx_page_failed
++;
1501 /* map page for use */
1502 dma
= dma_map_page(rx_ring
->dev
, page
, 0,
1503 ixgbe_rx_pg_size(rx_ring
), DMA_FROM_DEVICE
);
1506 * if mapping failed free memory back to system since
1507 * there isn't much point in holding memory we can't use
1509 if (dma_mapping_error(rx_ring
->dev
, dma
)) {
1510 __free_pages(page
, ixgbe_rx_pg_order(rx_ring
));
1512 rx_ring
->rx_stats
.alloc_rx_page_failed
++;
1518 bi
->page_offset
= 0;
1524 * ixgbe_alloc_rx_buffers - Replace used receive buffers
1525 * @rx_ring: ring to place buffers on
1526 * @cleaned_count: number of buffers to replace
1528 void ixgbe_alloc_rx_buffers(struct ixgbe_ring
*rx_ring
, u16 cleaned_count
)
1530 union ixgbe_adv_rx_desc
*rx_desc
;
1531 struct ixgbe_rx_buffer
*bi
;
1532 u16 i
= rx_ring
->next_to_use
;
1538 rx_desc
= IXGBE_RX_DESC(rx_ring
, i
);
1539 bi
= &rx_ring
->rx_buffer_info
[i
];
1540 i
-= rx_ring
->count
;
1543 if (!ixgbe_alloc_mapped_page(rx_ring
, bi
))
1547 * Refresh the desc even if buffer_addrs didn't change
1548 * because each write-back erases this info.
1550 rx_desc
->read
.pkt_addr
= cpu_to_le64(bi
->dma
+ bi
->page_offset
);
1556 rx_desc
= IXGBE_RX_DESC(rx_ring
, 0);
1557 bi
= rx_ring
->rx_buffer_info
;
1558 i
-= rx_ring
->count
;
1561 /* clear the status bits for the next_to_use descriptor */
1562 rx_desc
->wb
.upper
.status_error
= 0;
1565 } while (cleaned_count
);
1567 i
+= rx_ring
->count
;
1569 if (rx_ring
->next_to_use
!= i
) {
1570 rx_ring
->next_to_use
= i
;
1572 /* update next to alloc since we have filled the ring */
1573 rx_ring
->next_to_alloc
= i
;
1575 /* Force memory writes to complete before letting h/w
1576 * know there are new descriptors to fetch. (Only
1577 * applicable for weak-ordered memory model archs,
1581 writel(i
, rx_ring
->tail
);
1585 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring
*ring
,
1586 struct sk_buff
*skb
)
1588 u16 hdr_len
= skb_headlen(skb
);
1590 /* set gso_size to avoid messing up TCP MSS */
1591 skb_shinfo(skb
)->gso_size
= DIV_ROUND_UP((skb
->len
- hdr_len
),
1592 IXGBE_CB(skb
)->append_cnt
);
1593 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV4
;
1596 static void ixgbe_update_rsc_stats(struct ixgbe_ring
*rx_ring
,
1597 struct sk_buff
*skb
)
1599 /* if append_cnt is 0 then frame is not RSC */
1600 if (!IXGBE_CB(skb
)->append_cnt
)
1603 rx_ring
->rx_stats
.rsc_count
+= IXGBE_CB(skb
)->append_cnt
;
1604 rx_ring
->rx_stats
.rsc_flush
++;
1606 ixgbe_set_rsc_gso_size(rx_ring
, skb
);
1608 /* gso_size is computed using append_cnt so always clear it last */
1609 IXGBE_CB(skb
)->append_cnt
= 0;
1613 * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1614 * @rx_ring: rx descriptor ring packet is being transacted on
1615 * @rx_desc: pointer to the EOP Rx descriptor
1616 * @skb: pointer to current skb being populated
1618 * This function checks the ring, descriptor, and packet information in
1619 * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1620 * other fields within the skb.
1622 static void ixgbe_process_skb_fields(struct ixgbe_ring
*rx_ring
,
1623 union ixgbe_adv_rx_desc
*rx_desc
,
1624 struct sk_buff
*skb
)
1626 struct net_device
*dev
= rx_ring
->netdev
;
1628 ixgbe_update_rsc_stats(rx_ring
, skb
);
1630 ixgbe_rx_hash(rx_ring
, rx_desc
, skb
);
1632 ixgbe_rx_checksum(rx_ring
, rx_desc
, skb
);
1634 if (unlikely(ixgbe_test_staterr(rx_desc
, IXGBE_RXDADV_STAT_TS
)))
1635 ixgbe_ptp_rx_hwtstamp(rx_ring
->q_vector
->adapter
, skb
);
1637 if ((dev
->features
& NETIF_F_HW_VLAN_CTAG_RX
) &&
1638 ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_VP
)) {
1639 u16 vid
= le16_to_cpu(rx_desc
->wb
.upper
.vlan
);
1640 __vlan_hwaccel_put_tag(skb
, htons(ETH_P_8021Q
), vid
);
1643 skb_record_rx_queue(skb
, rx_ring
->queue_index
);
1645 skb
->protocol
= eth_type_trans(skb
, dev
);
1648 static void ixgbe_rx_skb(struct ixgbe_q_vector
*q_vector
,
1649 struct sk_buff
*skb
)
1651 if (ixgbe_qv_busy_polling(q_vector
))
1652 netif_receive_skb(skb
);
1654 napi_gro_receive(&q_vector
->napi
, skb
);
1658 * ixgbe_is_non_eop - process handling of non-EOP buffers
1659 * @rx_ring: Rx ring being processed
1660 * @rx_desc: Rx descriptor for current buffer
1661 * @skb: Current socket buffer containing buffer in progress
1663 * This function updates next to clean. If the buffer is an EOP buffer
1664 * this function exits returning false, otherwise it will place the
1665 * sk_buff in the next buffer to be chained and return true indicating
1666 * that this is in fact a non-EOP buffer.
1668 static bool ixgbe_is_non_eop(struct ixgbe_ring
*rx_ring
,
1669 union ixgbe_adv_rx_desc
*rx_desc
,
1670 struct sk_buff
*skb
)
1672 u32 ntc
= rx_ring
->next_to_clean
+ 1;
1674 /* fetch, update, and store next to clean */
1675 ntc
= (ntc
< rx_ring
->count
) ? ntc
: 0;
1676 rx_ring
->next_to_clean
= ntc
;
1678 prefetch(IXGBE_RX_DESC(rx_ring
, ntc
));
1680 /* update RSC append count if present */
1681 if (ring_is_rsc_enabled(rx_ring
)) {
1682 __le32 rsc_enabled
= rx_desc
->wb
.lower
.lo_dword
.data
&
1683 cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK
);
1685 if (unlikely(rsc_enabled
)) {
1686 u32 rsc_cnt
= le32_to_cpu(rsc_enabled
);
1688 rsc_cnt
>>= IXGBE_RXDADV_RSCCNT_SHIFT
;
1689 IXGBE_CB(skb
)->append_cnt
+= rsc_cnt
- 1;
1691 /* update ntc based on RSC value */
1692 ntc
= le32_to_cpu(rx_desc
->wb
.upper
.status_error
);
1693 ntc
&= IXGBE_RXDADV_NEXTP_MASK
;
1694 ntc
>>= IXGBE_RXDADV_NEXTP_SHIFT
;
1698 /* if we are the last buffer then there is nothing else to do */
1699 if (likely(ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_EOP
)))
1702 /* place skb in next buffer to be received */
1703 rx_ring
->rx_buffer_info
[ntc
].skb
= skb
;
1704 rx_ring
->rx_stats
.non_eop_descs
++;
1710 * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1711 * @rx_ring: rx descriptor ring packet is being transacted on
1712 * @skb: pointer to current skb being adjusted
1714 * This function is an ixgbe specific version of __pskb_pull_tail. The
1715 * main difference between this version and the original function is that
1716 * this function can make several assumptions about the state of things
1717 * that allow for significant optimizations versus the standard function.
1718 * As a result we can do things like drop a frag and maintain an accurate
1719 * truesize for the skb.
1721 static void ixgbe_pull_tail(struct ixgbe_ring
*rx_ring
,
1722 struct sk_buff
*skb
)
1724 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[0];
1726 unsigned int pull_len
;
1729 * it is valid to use page_address instead of kmap since we are
1730 * working with pages allocated out of the lomem pool per
1731 * alloc_page(GFP_ATOMIC)
1733 va
= skb_frag_address(frag
);
1736 * we need the header to contain the greater of either ETH_HLEN or
1737 * 60 bytes if the skb->len is less than 60 for skb_pad.
1739 pull_len
= eth_get_headlen(va
, IXGBE_RX_HDR_SIZE
);
1741 /* align pull length to size of long to optimize memcpy performance */
1742 skb_copy_to_linear_data(skb
, va
, ALIGN(pull_len
, sizeof(long)));
1744 /* update all of the pointers */
1745 skb_frag_size_sub(frag
, pull_len
);
1746 frag
->page_offset
+= pull_len
;
1747 skb
->data_len
-= pull_len
;
1748 skb
->tail
+= pull_len
;
1752 * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1753 * @rx_ring: rx descriptor ring packet is being transacted on
1754 * @skb: pointer to current skb being updated
1756 * This function provides a basic DMA sync up for the first fragment of an
1757 * skb. The reason for doing this is that the first fragment cannot be
1758 * unmapped until we have reached the end of packet descriptor for a buffer
1761 static void ixgbe_dma_sync_frag(struct ixgbe_ring
*rx_ring
,
1762 struct sk_buff
*skb
)
1764 /* if the page was released unmap it, else just sync our portion */
1765 if (unlikely(IXGBE_CB(skb
)->page_released
)) {
1766 dma_unmap_page(rx_ring
->dev
, IXGBE_CB(skb
)->dma
,
1767 ixgbe_rx_pg_size(rx_ring
), DMA_FROM_DEVICE
);
1768 IXGBE_CB(skb
)->page_released
= false;
1770 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[0];
1772 dma_sync_single_range_for_cpu(rx_ring
->dev
,
1775 ixgbe_rx_bufsz(rx_ring
),
1778 IXGBE_CB(skb
)->dma
= 0;
1782 * ixgbe_cleanup_headers - Correct corrupted or empty headers
1783 * @rx_ring: rx descriptor ring packet is being transacted on
1784 * @rx_desc: pointer to the EOP Rx descriptor
1785 * @skb: pointer to current skb being fixed
1787 * Check for corrupted packet headers caused by senders on the local L2
1788 * embedded NIC switch not setting up their Tx Descriptors right. These
1789 * should be very rare.
1791 * Also address the case where we are pulling data in on pages only
1792 * and as such no data is present in the skb header.
1794 * In addition if skb is not at least 60 bytes we need to pad it so that
1795 * it is large enough to qualify as a valid Ethernet frame.
1797 * Returns true if an error was encountered and skb was freed.
1799 static bool ixgbe_cleanup_headers(struct ixgbe_ring
*rx_ring
,
1800 union ixgbe_adv_rx_desc
*rx_desc
,
1801 struct sk_buff
*skb
)
1803 struct net_device
*netdev
= rx_ring
->netdev
;
1805 /* verify that the packet does not have any known errors */
1806 if (unlikely(ixgbe_test_staterr(rx_desc
,
1807 IXGBE_RXDADV_ERR_FRAME_ERR_MASK
) &&
1808 !(netdev
->features
& NETIF_F_RXALL
))) {
1809 dev_kfree_skb_any(skb
);
1813 /* place header in linear portion of buffer */
1814 if (skb_is_nonlinear(skb
))
1815 ixgbe_pull_tail(rx_ring
, skb
);
1818 /* do not attempt to pad FCoE Frames as this will disrupt DDP */
1819 if (ixgbe_rx_is_fcoe(rx_ring
, rx_desc
))
1823 /* if eth_skb_pad returns an error the skb was freed */
1824 if (eth_skb_pad(skb
))
1831 * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1832 * @rx_ring: rx descriptor ring to store buffers on
1833 * @old_buff: donor buffer to have page reused
1835 * Synchronizes page for reuse by the adapter
1837 static void ixgbe_reuse_rx_page(struct ixgbe_ring
*rx_ring
,
1838 struct ixgbe_rx_buffer
*old_buff
)
1840 struct ixgbe_rx_buffer
*new_buff
;
1841 u16 nta
= rx_ring
->next_to_alloc
;
1843 new_buff
= &rx_ring
->rx_buffer_info
[nta
];
1845 /* update, and store next to alloc */
1847 rx_ring
->next_to_alloc
= (nta
< rx_ring
->count
) ? nta
: 0;
1849 /* transfer page from old buffer to new buffer */
1850 *new_buff
= *old_buff
;
1852 /* sync the buffer for use by the device */
1853 dma_sync_single_range_for_device(rx_ring
->dev
, new_buff
->dma
,
1854 new_buff
->page_offset
,
1855 ixgbe_rx_bufsz(rx_ring
),
1859 static inline bool ixgbe_page_is_reserved(struct page
*page
)
1861 return (page_to_nid(page
) != numa_mem_id()) || page_is_pfmemalloc(page
);
1865 * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1866 * @rx_ring: rx descriptor ring to transact packets on
1867 * @rx_buffer: buffer containing page to add
1868 * @rx_desc: descriptor containing length of buffer written by hardware
1869 * @skb: sk_buff to place the data into
1871 * This function will add the data contained in rx_buffer->page to the skb.
1872 * This is done either through a direct copy if the data in the buffer is
1873 * less than the skb header size, otherwise it will just attach the page as
1874 * a frag to the skb.
1876 * The function will then update the page offset if necessary and return
1877 * true if the buffer can be reused by the adapter.
1879 static bool ixgbe_add_rx_frag(struct ixgbe_ring
*rx_ring
,
1880 struct ixgbe_rx_buffer
*rx_buffer
,
1881 union ixgbe_adv_rx_desc
*rx_desc
,
1882 struct sk_buff
*skb
)
1884 struct page
*page
= rx_buffer
->page
;
1885 unsigned int size
= le16_to_cpu(rx_desc
->wb
.upper
.length
);
1886 #if (PAGE_SIZE < 8192)
1887 unsigned int truesize
= ixgbe_rx_bufsz(rx_ring
);
1889 unsigned int truesize
= ALIGN(size
, L1_CACHE_BYTES
);
1890 unsigned int last_offset
= ixgbe_rx_pg_size(rx_ring
) -
1891 ixgbe_rx_bufsz(rx_ring
);
1894 if ((size
<= IXGBE_RX_HDR_SIZE
) && !skb_is_nonlinear(skb
)) {
1895 unsigned char *va
= page_address(page
) + rx_buffer
->page_offset
;
1897 memcpy(__skb_put(skb
, size
), va
, ALIGN(size
, sizeof(long)));
1899 /* page is not reserved, we can reuse buffer as-is */
1900 if (likely(!ixgbe_page_is_reserved(page
)))
1903 /* this page cannot be reused so discard it */
1904 __free_pages(page
, ixgbe_rx_pg_order(rx_ring
));
1908 skb_add_rx_frag(skb
, skb_shinfo(skb
)->nr_frags
, page
,
1909 rx_buffer
->page_offset
, size
, truesize
);
1911 /* avoid re-using remote pages */
1912 if (unlikely(ixgbe_page_is_reserved(page
)))
1915 #if (PAGE_SIZE < 8192)
1916 /* if we are only owner of page we can reuse it */
1917 if (unlikely(page_count(page
) != 1))
1920 /* flip page offset to other buffer */
1921 rx_buffer
->page_offset
^= truesize
;
1923 /* move offset up to the next cache line */
1924 rx_buffer
->page_offset
+= truesize
;
1926 if (rx_buffer
->page_offset
> last_offset
)
1930 /* Even if we own the page, we are not allowed to use atomic_set()
1931 * This would break get_page_unless_zero() users.
1933 atomic_inc(&page
->_count
);
1938 static struct sk_buff
*ixgbe_fetch_rx_buffer(struct ixgbe_ring
*rx_ring
,
1939 union ixgbe_adv_rx_desc
*rx_desc
)
1941 struct ixgbe_rx_buffer
*rx_buffer
;
1942 struct sk_buff
*skb
;
1945 rx_buffer
= &rx_ring
->rx_buffer_info
[rx_ring
->next_to_clean
];
1946 page
= rx_buffer
->page
;
1949 skb
= rx_buffer
->skb
;
1952 void *page_addr
= page_address(page
) +
1953 rx_buffer
->page_offset
;
1955 /* prefetch first cache line of first page */
1956 prefetch(page_addr
);
1957 #if L1_CACHE_BYTES < 128
1958 prefetch(page_addr
+ L1_CACHE_BYTES
);
1961 /* allocate a skb to store the frags */
1962 skb
= napi_alloc_skb(&rx_ring
->q_vector
->napi
,
1964 if (unlikely(!skb
)) {
1965 rx_ring
->rx_stats
.alloc_rx_buff_failed
++;
1970 * we will be copying header into skb->data in
1971 * pskb_may_pull so it is in our interest to prefetch
1972 * it now to avoid a possible cache miss
1974 prefetchw(skb
->data
);
1977 * Delay unmapping of the first packet. It carries the
1978 * header information, HW may still access the header
1979 * after the writeback. Only unmap it when EOP is
1982 if (likely(ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_EOP
)))
1985 IXGBE_CB(skb
)->dma
= rx_buffer
->dma
;
1987 if (ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_EOP
))
1988 ixgbe_dma_sync_frag(rx_ring
, skb
);
1991 /* we are reusing so sync this buffer for CPU use */
1992 dma_sync_single_range_for_cpu(rx_ring
->dev
,
1994 rx_buffer
->page_offset
,
1995 ixgbe_rx_bufsz(rx_ring
),
1998 rx_buffer
->skb
= NULL
;
2001 /* pull page into skb */
2002 if (ixgbe_add_rx_frag(rx_ring
, rx_buffer
, rx_desc
, skb
)) {
2003 /* hand second half of page back to the ring */
2004 ixgbe_reuse_rx_page(rx_ring
, rx_buffer
);
2005 } else if (IXGBE_CB(skb
)->dma
== rx_buffer
->dma
) {
2006 /* the page has been released from the ring */
2007 IXGBE_CB(skb
)->page_released
= true;
2009 /* we are not reusing the buffer so unmap it */
2010 dma_unmap_page(rx_ring
->dev
, rx_buffer
->dma
,
2011 ixgbe_rx_pg_size(rx_ring
),
2015 /* clear contents of buffer_info */
2016 rx_buffer
->page
= NULL
;
2022 * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2023 * @q_vector: structure containing interrupt and ring information
2024 * @rx_ring: rx descriptor ring to transact packets on
2025 * @budget: Total limit on number of packets to process
2027 * This function provides a "bounce buffer" approach to Rx interrupt
2028 * processing. The advantage to this is that on systems that have
2029 * expensive overhead for IOMMU access this provides a means of avoiding
2030 * it by maintaining the mapping of the page to the syste.
2032 * Returns amount of work completed
2034 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector
*q_vector
,
2035 struct ixgbe_ring
*rx_ring
,
2038 unsigned int total_rx_bytes
= 0, total_rx_packets
= 0;
2040 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
2042 unsigned int mss
= 0;
2043 #endif /* IXGBE_FCOE */
2044 u16 cleaned_count
= ixgbe_desc_unused(rx_ring
);
2046 while (likely(total_rx_packets
< budget
)) {
2047 union ixgbe_adv_rx_desc
*rx_desc
;
2048 struct sk_buff
*skb
;
2050 /* return some buffers to hardware, one at a time is too slow */
2051 if (cleaned_count
>= IXGBE_RX_BUFFER_WRITE
) {
2052 ixgbe_alloc_rx_buffers(rx_ring
, cleaned_count
);
2056 rx_desc
= IXGBE_RX_DESC(rx_ring
, rx_ring
->next_to_clean
);
2058 if (!rx_desc
->wb
.upper
.status_error
)
2061 /* This memory barrier is needed to keep us from reading
2062 * any other fields out of the rx_desc until we know the
2063 * descriptor has been written back
2067 /* retrieve a buffer from the ring */
2068 skb
= ixgbe_fetch_rx_buffer(rx_ring
, rx_desc
);
2070 /* exit if we failed to retrieve a buffer */
2076 /* place incomplete frames back on ring for completion */
2077 if (ixgbe_is_non_eop(rx_ring
, rx_desc
, skb
))
2080 /* verify the packet layout is correct */
2081 if (ixgbe_cleanup_headers(rx_ring
, rx_desc
, skb
))
2084 /* probably a little skewed due to removing CRC */
2085 total_rx_bytes
+= skb
->len
;
2087 /* populate checksum, timestamp, VLAN, and protocol */
2088 ixgbe_process_skb_fields(rx_ring
, rx_desc
, skb
);
2091 /* if ddp, not passing to ULD unless for FCP_RSP or error */
2092 if (ixgbe_rx_is_fcoe(rx_ring
, rx_desc
)) {
2093 ddp_bytes
= ixgbe_fcoe_ddp(adapter
, rx_desc
, skb
);
2094 /* include DDPed FCoE data */
2095 if (ddp_bytes
> 0) {
2097 mss
= rx_ring
->netdev
->mtu
-
2098 sizeof(struct fcoe_hdr
) -
2099 sizeof(struct fc_frame_header
) -
2100 sizeof(struct fcoe_crc_eof
);
2104 total_rx_bytes
+= ddp_bytes
;
2105 total_rx_packets
+= DIV_ROUND_UP(ddp_bytes
,
2109 dev_kfree_skb_any(skb
);
2114 #endif /* IXGBE_FCOE */
2115 skb_mark_napi_id(skb
, &q_vector
->napi
);
2116 ixgbe_rx_skb(q_vector
, skb
);
2118 /* update budget accounting */
2122 u64_stats_update_begin(&rx_ring
->syncp
);
2123 rx_ring
->stats
.packets
+= total_rx_packets
;
2124 rx_ring
->stats
.bytes
+= total_rx_bytes
;
2125 u64_stats_update_end(&rx_ring
->syncp
);
2126 q_vector
->rx
.total_packets
+= total_rx_packets
;
2127 q_vector
->rx
.total_bytes
+= total_rx_bytes
;
2129 return total_rx_packets
;
2132 #ifdef CONFIG_NET_RX_BUSY_POLL
2133 /* must be called with local_bh_disable()d */
2134 static int ixgbe_low_latency_recv(struct napi_struct
*napi
)
2136 struct ixgbe_q_vector
*q_vector
=
2137 container_of(napi
, struct ixgbe_q_vector
, napi
);
2138 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
2139 struct ixgbe_ring
*ring
;
2142 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
2143 return LL_FLUSH_FAILED
;
2145 if (!ixgbe_qv_lock_poll(q_vector
))
2146 return LL_FLUSH_BUSY
;
2148 ixgbe_for_each_ring(ring
, q_vector
->rx
) {
2149 found
= ixgbe_clean_rx_irq(q_vector
, ring
, 4);
2150 #ifdef BP_EXTENDED_STATS
2152 ring
->stats
.cleaned
+= found
;
2154 ring
->stats
.misses
++;
2160 ixgbe_qv_unlock_poll(q_vector
);
2164 #endif /* CONFIG_NET_RX_BUSY_POLL */
2167 * ixgbe_configure_msix - Configure MSI-X hardware
2168 * @adapter: board private structure
2170 * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2173 static void ixgbe_configure_msix(struct ixgbe_adapter
*adapter
)
2175 struct ixgbe_q_vector
*q_vector
;
2179 /* Populate MSIX to EITR Select */
2180 if (adapter
->num_vfs
> 32) {
2181 u32 eitrsel
= (1 << (adapter
->num_vfs
- 32)) - 1;
2182 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EITRSEL
, eitrsel
);
2186 * Populate the IVAR table and set the ITR values to the
2187 * corresponding register.
2189 for (v_idx
= 0; v_idx
< adapter
->num_q_vectors
; v_idx
++) {
2190 struct ixgbe_ring
*ring
;
2191 q_vector
= adapter
->q_vector
[v_idx
];
2193 ixgbe_for_each_ring(ring
, q_vector
->rx
)
2194 ixgbe_set_ivar(adapter
, 0, ring
->reg_idx
, v_idx
);
2196 ixgbe_for_each_ring(ring
, q_vector
->tx
)
2197 ixgbe_set_ivar(adapter
, 1, ring
->reg_idx
, v_idx
);
2199 ixgbe_write_eitr(q_vector
);
2202 switch (adapter
->hw
.mac
.type
) {
2203 case ixgbe_mac_82598EB
:
2204 ixgbe_set_ivar(adapter
, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX
,
2207 case ixgbe_mac_82599EB
:
2208 case ixgbe_mac_X540
:
2209 case ixgbe_mac_X550
:
2210 case ixgbe_mac_X550EM_x
:
2211 ixgbe_set_ivar(adapter
, -1, 1, v_idx
);
2216 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EITR(v_idx
), 1950);
2218 /* set up to autoclear timer, and the vectors */
2219 mask
= IXGBE_EIMS_ENABLE_MASK
;
2220 mask
&= ~(IXGBE_EIMS_OTHER
|
2221 IXGBE_EIMS_MAILBOX
|
2224 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIAC
, mask
);
2227 enum latency_range
{
2231 latency_invalid
= 255
2235 * ixgbe_update_itr - update the dynamic ITR value based on statistics
2236 * @q_vector: structure containing interrupt and ring information
2237 * @ring_container: structure containing ring performance data
2239 * Stores a new ITR value based on packets and byte
2240 * counts during the last interrupt. The advantage of per interrupt
2241 * computation is faster updates and more accurate ITR for the current
2242 * traffic pattern. Constants in this function were computed
2243 * based on theoretical maximum wire speed and thresholds were set based
2244 * on testing data as well as attempting to minimize response time
2245 * while increasing bulk throughput.
2246 * this functionality is controlled by the InterruptThrottleRate module
2247 * parameter (see ixgbe_param.c)
2249 static void ixgbe_update_itr(struct ixgbe_q_vector
*q_vector
,
2250 struct ixgbe_ring_container
*ring_container
)
2252 int bytes
= ring_container
->total_bytes
;
2253 int packets
= ring_container
->total_packets
;
2256 u8 itr_setting
= ring_container
->itr
;
2261 /* simple throttlerate management
2262 * 0-10MB/s lowest (100000 ints/s)
2263 * 10-20MB/s low (20000 ints/s)
2264 * 20-1249MB/s bulk (12000 ints/s)
2266 /* what was last interrupt timeslice? */
2267 timepassed_us
= q_vector
->itr
>> 2;
2268 if (timepassed_us
== 0)
2271 bytes_perint
= bytes
/ timepassed_us
; /* bytes/usec */
2273 switch (itr_setting
) {
2274 case lowest_latency
:
2275 if (bytes_perint
> 10)
2276 itr_setting
= low_latency
;
2279 if (bytes_perint
> 20)
2280 itr_setting
= bulk_latency
;
2281 else if (bytes_perint
<= 10)
2282 itr_setting
= lowest_latency
;
2285 if (bytes_perint
<= 20)
2286 itr_setting
= low_latency
;
2290 /* clear work counters since we have the values we need */
2291 ring_container
->total_bytes
= 0;
2292 ring_container
->total_packets
= 0;
2294 /* write updated itr to ring container */
2295 ring_container
->itr
= itr_setting
;
2299 * ixgbe_write_eitr - write EITR register in hardware specific way
2300 * @q_vector: structure containing interrupt and ring information
2302 * This function is made to be called by ethtool and by the driver
2303 * when it needs to update EITR registers at runtime. Hardware
2304 * specific quirks/differences are taken care of here.
2306 void ixgbe_write_eitr(struct ixgbe_q_vector
*q_vector
)
2308 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
2309 struct ixgbe_hw
*hw
= &adapter
->hw
;
2310 int v_idx
= q_vector
->v_idx
;
2311 u32 itr_reg
= q_vector
->itr
& IXGBE_MAX_EITR
;
2313 switch (adapter
->hw
.mac
.type
) {
2314 case ixgbe_mac_82598EB
:
2315 /* must write high and low 16 bits to reset counter */
2316 itr_reg
|= (itr_reg
<< 16);
2318 case ixgbe_mac_82599EB
:
2319 case ixgbe_mac_X540
:
2320 case ixgbe_mac_X550
:
2321 case ixgbe_mac_X550EM_x
:
2323 * set the WDIS bit to not clear the timer bits and cause an
2324 * immediate assertion of the interrupt
2326 itr_reg
|= IXGBE_EITR_CNT_WDIS
;
2331 IXGBE_WRITE_REG(hw
, IXGBE_EITR(v_idx
), itr_reg
);
2334 static void ixgbe_set_itr(struct ixgbe_q_vector
*q_vector
)
2336 u32 new_itr
= q_vector
->itr
;
2339 ixgbe_update_itr(q_vector
, &q_vector
->tx
);
2340 ixgbe_update_itr(q_vector
, &q_vector
->rx
);
2342 current_itr
= max(q_vector
->rx
.itr
, q_vector
->tx
.itr
);
2344 switch (current_itr
) {
2345 /* counts and packets in update_itr are dependent on these numbers */
2346 case lowest_latency
:
2347 new_itr
= IXGBE_100K_ITR
;
2350 new_itr
= IXGBE_20K_ITR
;
2353 new_itr
= IXGBE_12K_ITR
;
2359 if (new_itr
!= q_vector
->itr
) {
2360 /* do an exponential smoothing */
2361 new_itr
= (10 * new_itr
* q_vector
->itr
) /
2362 ((9 * new_itr
) + q_vector
->itr
);
2364 /* save the algorithm value here */
2365 q_vector
->itr
= new_itr
;
2367 ixgbe_write_eitr(q_vector
);
2372 * ixgbe_check_overtemp_subtask - check for over temperature
2373 * @adapter: pointer to adapter
2375 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter
*adapter
)
2377 struct ixgbe_hw
*hw
= &adapter
->hw
;
2378 u32 eicr
= adapter
->interrupt_event
;
2380 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
2383 if (!(adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
) &&
2384 !(adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_EVENT
))
2387 adapter
->flags2
&= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT
;
2389 switch (hw
->device_id
) {
2390 case IXGBE_DEV_ID_82599_T3_LOM
:
2392 * Since the warning interrupt is for both ports
2393 * we don't have to check if:
2394 * - This interrupt wasn't for our port.
2395 * - We may have missed the interrupt so always have to
2396 * check if we got a LSC
2398 if (!(eicr
& IXGBE_EICR_GPI_SDP0_8259X
) &&
2399 !(eicr
& IXGBE_EICR_LSC
))
2402 if (!(eicr
& IXGBE_EICR_LSC
) && hw
->mac
.ops
.check_link
) {
2404 bool link_up
= false;
2406 hw
->mac
.ops
.check_link(hw
, &speed
, &link_up
, false);
2412 /* Check if this is not due to overtemp */
2413 if (hw
->phy
.ops
.check_overtemp(hw
) != IXGBE_ERR_OVERTEMP
)
2418 if (adapter
->hw
.mac
.type
>= ixgbe_mac_X540
)
2420 if (!(eicr
& IXGBE_EICR_GPI_SDP0(hw
)))
2424 e_crit(drv
, "%s\n", ixgbe_overheat_msg
);
2426 adapter
->interrupt_event
= 0;
2429 static void ixgbe_check_fan_failure(struct ixgbe_adapter
*adapter
, u32 eicr
)
2431 struct ixgbe_hw
*hw
= &adapter
->hw
;
2433 if ((adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
) &&
2434 (eicr
& IXGBE_EICR_GPI_SDP1(hw
))) {
2435 e_crit(probe
, "Fan has stopped, replace the adapter\n");
2436 /* write to clear the interrupt */
2437 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, IXGBE_EICR_GPI_SDP1(hw
));
2441 static void ixgbe_check_overtemp_event(struct ixgbe_adapter
*adapter
, u32 eicr
)
2443 struct ixgbe_hw
*hw
= &adapter
->hw
;
2445 if (!(adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
))
2448 switch (adapter
->hw
.mac
.type
) {
2449 case ixgbe_mac_82599EB
:
2451 * Need to check link state so complete overtemp check
2454 if (((eicr
& IXGBE_EICR_GPI_SDP0(hw
)) ||
2455 (eicr
& IXGBE_EICR_LSC
)) &&
2456 (!test_bit(__IXGBE_DOWN
, &adapter
->state
))) {
2457 adapter
->interrupt_event
= eicr
;
2458 adapter
->flags2
|= IXGBE_FLAG2_TEMP_SENSOR_EVENT
;
2459 ixgbe_service_event_schedule(adapter
);
2463 case ixgbe_mac_X540
:
2464 if (!(eicr
& IXGBE_EICR_TS
))
2471 e_crit(drv
, "%s\n", ixgbe_overheat_msg
);
2474 static inline bool ixgbe_is_sfp(struct ixgbe_hw
*hw
)
2476 switch (hw
->mac
.type
) {
2477 case ixgbe_mac_82598EB
:
2478 if (hw
->phy
.type
== ixgbe_phy_nl
)
2481 case ixgbe_mac_82599EB
:
2482 case ixgbe_mac_X550EM_x
:
2483 switch (hw
->mac
.ops
.get_media_type(hw
)) {
2484 case ixgbe_media_type_fiber
:
2485 case ixgbe_media_type_fiber_qsfp
:
2495 static void ixgbe_check_sfp_event(struct ixgbe_adapter
*adapter
, u32 eicr
)
2497 struct ixgbe_hw
*hw
= &adapter
->hw
;
2498 u32 eicr_mask
= IXGBE_EICR_GPI_SDP2(hw
);
2500 if (!ixgbe_is_sfp(hw
))
2503 /* Later MAC's use different SDP */
2504 if (hw
->mac
.type
>= ixgbe_mac_X540
)
2505 eicr_mask
= IXGBE_EICR_GPI_SDP0_X540
;
2507 if (eicr
& eicr_mask
) {
2508 /* Clear the interrupt */
2509 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, eicr_mask
);
2510 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
2511 adapter
->flags2
|= IXGBE_FLAG2_SFP_NEEDS_RESET
;
2512 adapter
->sfp_poll_time
= 0;
2513 ixgbe_service_event_schedule(adapter
);
2517 if (adapter
->hw
.mac
.type
== ixgbe_mac_82599EB
&&
2518 (eicr
& IXGBE_EICR_GPI_SDP1(hw
))) {
2519 /* Clear the interrupt */
2520 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, IXGBE_EICR_GPI_SDP1(hw
));
2521 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
2522 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_CONFIG
;
2523 ixgbe_service_event_schedule(adapter
);
2528 static void ixgbe_check_lsc(struct ixgbe_adapter
*adapter
)
2530 struct ixgbe_hw
*hw
= &adapter
->hw
;
2533 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_UPDATE
;
2534 adapter
->link_check_timeout
= jiffies
;
2535 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
2536 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, IXGBE_EIMC_LSC
);
2537 IXGBE_WRITE_FLUSH(hw
);
2538 ixgbe_service_event_schedule(adapter
);
2542 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter
*adapter
,
2546 struct ixgbe_hw
*hw
= &adapter
->hw
;
2548 switch (hw
->mac
.type
) {
2549 case ixgbe_mac_82598EB
:
2550 mask
= (IXGBE_EIMS_RTX_QUEUE
& qmask
);
2551 IXGBE_WRITE_REG(hw
, IXGBE_EIMS
, mask
);
2553 case ixgbe_mac_82599EB
:
2554 case ixgbe_mac_X540
:
2555 case ixgbe_mac_X550
:
2556 case ixgbe_mac_X550EM_x
:
2557 mask
= (qmask
& 0xFFFFFFFF);
2559 IXGBE_WRITE_REG(hw
, IXGBE_EIMS_EX(0), mask
);
2560 mask
= (qmask
>> 32);
2562 IXGBE_WRITE_REG(hw
, IXGBE_EIMS_EX(1), mask
);
2567 /* skip the flush */
2570 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter
*adapter
,
2574 struct ixgbe_hw
*hw
= &adapter
->hw
;
2576 switch (hw
->mac
.type
) {
2577 case ixgbe_mac_82598EB
:
2578 mask
= (IXGBE_EIMS_RTX_QUEUE
& qmask
);
2579 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, mask
);
2581 case ixgbe_mac_82599EB
:
2582 case ixgbe_mac_X540
:
2583 case ixgbe_mac_X550
:
2584 case ixgbe_mac_X550EM_x
:
2585 mask
= (qmask
& 0xFFFFFFFF);
2587 IXGBE_WRITE_REG(hw
, IXGBE_EIMC_EX(0), mask
);
2588 mask
= (qmask
>> 32);
2590 IXGBE_WRITE_REG(hw
, IXGBE_EIMC_EX(1), mask
);
2595 /* skip the flush */
2599 * ixgbe_irq_enable - Enable default interrupt generation settings
2600 * @adapter: board private structure
2602 static inline void ixgbe_irq_enable(struct ixgbe_adapter
*adapter
, bool queues
,
2605 struct ixgbe_hw
*hw
= &adapter
->hw
;
2606 u32 mask
= (IXGBE_EIMS_ENABLE_MASK
& ~IXGBE_EIMS_RTX_QUEUE
);
2608 /* don't reenable LSC while waiting for link */
2609 if (adapter
->flags
& IXGBE_FLAG_NEED_LINK_UPDATE
)
2610 mask
&= ~IXGBE_EIMS_LSC
;
2612 if (adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
)
2613 switch (adapter
->hw
.mac
.type
) {
2614 case ixgbe_mac_82599EB
:
2615 mask
|= IXGBE_EIMS_GPI_SDP0(hw
);
2617 case ixgbe_mac_X540
:
2618 case ixgbe_mac_X550
:
2619 case ixgbe_mac_X550EM_x
:
2620 mask
|= IXGBE_EIMS_TS
;
2625 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
)
2626 mask
|= IXGBE_EIMS_GPI_SDP1(hw
);
2627 switch (adapter
->hw
.mac
.type
) {
2628 case ixgbe_mac_82599EB
:
2629 mask
|= IXGBE_EIMS_GPI_SDP1(hw
);
2630 mask
|= IXGBE_EIMS_GPI_SDP2(hw
);
2632 case ixgbe_mac_X540
:
2633 case ixgbe_mac_X550
:
2634 case ixgbe_mac_X550EM_x
:
2635 if (adapter
->hw
.device_id
== IXGBE_DEV_ID_X550EM_X_SFP
)
2636 mask
|= IXGBE_EIMS_GPI_SDP0(&adapter
->hw
);
2637 if (adapter
->hw
.phy
.type
== ixgbe_phy_x550em_ext_t
)
2638 mask
|= IXGBE_EICR_GPI_SDP0_X540
;
2639 mask
|= IXGBE_EIMS_ECC
;
2640 mask
|= IXGBE_EIMS_MAILBOX
;
2646 if ((adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) &&
2647 !(adapter
->flags2
& IXGBE_FLAG2_FDIR_REQUIRES_REINIT
))
2648 mask
|= IXGBE_EIMS_FLOW_DIR
;
2650 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMS
, mask
);
2652 ixgbe_irq_enable_queues(adapter
, ~0);
2654 IXGBE_WRITE_FLUSH(&adapter
->hw
);
2657 static irqreturn_t
ixgbe_msix_other(int irq
, void *data
)
2659 struct ixgbe_adapter
*adapter
= data
;
2660 struct ixgbe_hw
*hw
= &adapter
->hw
;
2664 * Workaround for Silicon errata. Use clear-by-write instead
2665 * of clear-by-read. Reading with EICS will return the
2666 * interrupt causes without clearing, which later be done
2667 * with the write to EICR.
2669 eicr
= IXGBE_READ_REG(hw
, IXGBE_EICS
);
2671 /* The lower 16bits of the EICR register are for the queue interrupts
2672 * which should be masked here in order to not accidentally clear them if
2673 * the bits are high when ixgbe_msix_other is called. There is a race
2674 * condition otherwise which results in possible performance loss
2675 * especially if the ixgbe_msix_other interrupt is triggering
2676 * consistently (as it would when PPS is turned on for the X540 device)
2680 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, eicr
);
2682 if (eicr
& IXGBE_EICR_LSC
)
2683 ixgbe_check_lsc(adapter
);
2685 if (eicr
& IXGBE_EICR_MAILBOX
)
2686 ixgbe_msg_task(adapter
);
2688 switch (hw
->mac
.type
) {
2689 case ixgbe_mac_82599EB
:
2690 case ixgbe_mac_X540
:
2691 case ixgbe_mac_X550
:
2692 case ixgbe_mac_X550EM_x
:
2693 if (hw
->phy
.type
== ixgbe_phy_x550em_ext_t
&&
2694 (eicr
& IXGBE_EICR_GPI_SDP0_X540
)) {
2695 adapter
->flags2
|= IXGBE_FLAG2_PHY_INTERRUPT
;
2696 ixgbe_service_event_schedule(adapter
);
2697 IXGBE_WRITE_REG(hw
, IXGBE_EICR
,
2698 IXGBE_EICR_GPI_SDP0_X540
);
2700 if (eicr
& IXGBE_EICR_ECC
) {
2701 e_info(link
, "Received ECC Err, initiating reset\n");
2702 adapter
->flags2
|= IXGBE_FLAG2_RESET_REQUESTED
;
2703 ixgbe_service_event_schedule(adapter
);
2704 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, IXGBE_EICR_ECC
);
2706 /* Handle Flow Director Full threshold interrupt */
2707 if (eicr
& IXGBE_EICR_FLOW_DIR
) {
2708 int reinit_count
= 0;
2710 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
2711 struct ixgbe_ring
*ring
= adapter
->tx_ring
[i
];
2712 if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE
,
2717 /* no more flow director interrupts until after init */
2718 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, IXGBE_EIMC_FLOW_DIR
);
2719 adapter
->flags2
|= IXGBE_FLAG2_FDIR_REQUIRES_REINIT
;
2720 ixgbe_service_event_schedule(adapter
);
2723 ixgbe_check_sfp_event(adapter
, eicr
);
2724 ixgbe_check_overtemp_event(adapter
, eicr
);
2730 ixgbe_check_fan_failure(adapter
, eicr
);
2732 if (unlikely(eicr
& IXGBE_EICR_TIMESYNC
))
2733 ixgbe_ptp_check_pps_event(adapter
, eicr
);
2735 /* re-enable the original interrupt state, no lsc, no queues */
2736 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2737 ixgbe_irq_enable(adapter
, false, false);
2742 static irqreturn_t
ixgbe_msix_clean_rings(int irq
, void *data
)
2744 struct ixgbe_q_vector
*q_vector
= data
;
2746 /* EIAM disabled interrupts (on this vector) for us */
2748 if (q_vector
->rx
.ring
|| q_vector
->tx
.ring
)
2749 napi_schedule(&q_vector
->napi
);
2755 * ixgbe_poll - NAPI Rx polling callback
2756 * @napi: structure for representing this polling device
2757 * @budget: how many packets driver is allowed to clean
2759 * This function is used for legacy and MSI, NAPI mode
2761 int ixgbe_poll(struct napi_struct
*napi
, int budget
)
2763 struct ixgbe_q_vector
*q_vector
=
2764 container_of(napi
, struct ixgbe_q_vector
, napi
);
2765 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
2766 struct ixgbe_ring
*ring
;
2767 int per_ring_budget
;
2768 bool clean_complete
= true;
2770 #ifdef CONFIG_IXGBE_DCA
2771 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
)
2772 ixgbe_update_dca(q_vector
);
2775 ixgbe_for_each_ring(ring
, q_vector
->tx
)
2776 clean_complete
&= !!ixgbe_clean_tx_irq(q_vector
, ring
);
2778 if (!ixgbe_qv_lock_napi(q_vector
))
2781 /* attempt to distribute budget to each queue fairly, but don't allow
2782 * the budget to go below 1 because we'll exit polling */
2783 if (q_vector
->rx
.count
> 1)
2784 per_ring_budget
= max(budget
/q_vector
->rx
.count
, 1);
2786 per_ring_budget
= budget
;
2788 ixgbe_for_each_ring(ring
, q_vector
->rx
)
2789 clean_complete
&= (ixgbe_clean_rx_irq(q_vector
, ring
,
2790 per_ring_budget
) < per_ring_budget
);
2792 ixgbe_qv_unlock_napi(q_vector
);
2793 /* If all work not completed, return budget and keep polling */
2794 if (!clean_complete
)
2797 /* all work done, exit the polling mode */
2798 napi_complete(napi
);
2799 if (adapter
->rx_itr_setting
& 1)
2800 ixgbe_set_itr(q_vector
);
2801 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2802 ixgbe_irq_enable_queues(adapter
, ((u64
)1 << q_vector
->v_idx
));
2808 * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2809 * @adapter: board private structure
2811 * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2812 * interrupts from the kernel.
2814 static int ixgbe_request_msix_irqs(struct ixgbe_adapter
*adapter
)
2816 struct net_device
*netdev
= adapter
->netdev
;
2820 for (vector
= 0; vector
< adapter
->num_q_vectors
; vector
++) {
2821 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[vector
];
2822 struct msix_entry
*entry
= &adapter
->msix_entries
[vector
];
2824 if (q_vector
->tx
.ring
&& q_vector
->rx
.ring
) {
2825 snprintf(q_vector
->name
, sizeof(q_vector
->name
) - 1,
2826 "%s-%s-%d", netdev
->name
, "TxRx", ri
++);
2828 } else if (q_vector
->rx
.ring
) {
2829 snprintf(q_vector
->name
, sizeof(q_vector
->name
) - 1,
2830 "%s-%s-%d", netdev
->name
, "rx", ri
++);
2831 } else if (q_vector
->tx
.ring
) {
2832 snprintf(q_vector
->name
, sizeof(q_vector
->name
) - 1,
2833 "%s-%s-%d", netdev
->name
, "tx", ti
++);
2835 /* skip this unused q_vector */
2838 err
= request_irq(entry
->vector
, &ixgbe_msix_clean_rings
, 0,
2839 q_vector
->name
, q_vector
);
2841 e_err(probe
, "request_irq failed for MSIX interrupt "
2842 "Error: %d\n", err
);
2843 goto free_queue_irqs
;
2845 /* If Flow Director is enabled, set interrupt affinity */
2846 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) {
2847 /* assign the mask for this irq */
2848 irq_set_affinity_hint(entry
->vector
,
2849 &q_vector
->affinity_mask
);
2853 err
= request_irq(adapter
->msix_entries
[vector
].vector
,
2854 ixgbe_msix_other
, 0, netdev
->name
, adapter
);
2856 e_err(probe
, "request_irq for msix_other failed: %d\n", err
);
2857 goto free_queue_irqs
;
2865 irq_set_affinity_hint(adapter
->msix_entries
[vector
].vector
,
2867 free_irq(adapter
->msix_entries
[vector
].vector
,
2868 adapter
->q_vector
[vector
]);
2870 adapter
->flags
&= ~IXGBE_FLAG_MSIX_ENABLED
;
2871 pci_disable_msix(adapter
->pdev
);
2872 kfree(adapter
->msix_entries
);
2873 adapter
->msix_entries
= NULL
;
2878 * ixgbe_intr - legacy mode Interrupt Handler
2879 * @irq: interrupt number
2880 * @data: pointer to a network interface device structure
2882 static irqreturn_t
ixgbe_intr(int irq
, void *data
)
2884 struct ixgbe_adapter
*adapter
= data
;
2885 struct ixgbe_hw
*hw
= &adapter
->hw
;
2886 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[0];
2890 * Workaround for silicon errata #26 on 82598. Mask the interrupt
2891 * before the read of EICR.
2893 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, IXGBE_IRQ_CLEAR_MASK
);
2895 /* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2896 * therefore no explicit interrupt disable is necessary */
2897 eicr
= IXGBE_READ_REG(hw
, IXGBE_EICR
);
2900 * shared interrupt alert!
2901 * make sure interrupts are enabled because the read will
2902 * have disabled interrupts due to EIAM
2903 * finish the workaround of silicon errata on 82598. Unmask
2904 * the interrupt that we masked before the EICR read.
2906 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2907 ixgbe_irq_enable(adapter
, true, true);
2908 return IRQ_NONE
; /* Not our interrupt */
2911 if (eicr
& IXGBE_EICR_LSC
)
2912 ixgbe_check_lsc(adapter
);
2914 switch (hw
->mac
.type
) {
2915 case ixgbe_mac_82599EB
:
2916 ixgbe_check_sfp_event(adapter
, eicr
);
2918 case ixgbe_mac_X540
:
2919 case ixgbe_mac_X550
:
2920 case ixgbe_mac_X550EM_x
:
2921 if (eicr
& IXGBE_EICR_ECC
) {
2922 e_info(link
, "Received ECC Err, initiating reset\n");
2923 adapter
->flags2
|= IXGBE_FLAG2_RESET_REQUESTED
;
2924 ixgbe_service_event_schedule(adapter
);
2925 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, IXGBE_EICR_ECC
);
2927 ixgbe_check_overtemp_event(adapter
, eicr
);
2933 ixgbe_check_fan_failure(adapter
, eicr
);
2934 if (unlikely(eicr
& IXGBE_EICR_TIMESYNC
))
2935 ixgbe_ptp_check_pps_event(adapter
, eicr
);
2937 /* would disable interrupts here but EIAM disabled it */
2938 napi_schedule(&q_vector
->napi
);
2941 * re-enable link(maybe) and non-queue interrupts, no flush.
2942 * ixgbe_poll will re-enable the queue interrupts
2944 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2945 ixgbe_irq_enable(adapter
, false, false);
2951 * ixgbe_request_irq - initialize interrupts
2952 * @adapter: board private structure
2954 * Attempts to configure interrupts using the best available
2955 * capabilities of the hardware and kernel.
2957 static int ixgbe_request_irq(struct ixgbe_adapter
*adapter
)
2959 struct net_device
*netdev
= adapter
->netdev
;
2962 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)
2963 err
= ixgbe_request_msix_irqs(adapter
);
2964 else if (adapter
->flags
& IXGBE_FLAG_MSI_ENABLED
)
2965 err
= request_irq(adapter
->pdev
->irq
, ixgbe_intr
, 0,
2966 netdev
->name
, adapter
);
2968 err
= request_irq(adapter
->pdev
->irq
, ixgbe_intr
, IRQF_SHARED
,
2969 netdev
->name
, adapter
);
2972 e_err(probe
, "request_irq failed, Error %d\n", err
);
2977 static void ixgbe_free_irq(struct ixgbe_adapter
*adapter
)
2981 if (!(adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)) {
2982 free_irq(adapter
->pdev
->irq
, adapter
);
2986 for (vector
= 0; vector
< adapter
->num_q_vectors
; vector
++) {
2987 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[vector
];
2988 struct msix_entry
*entry
= &adapter
->msix_entries
[vector
];
2990 /* free only the irqs that were actually requested */
2991 if (!q_vector
->rx
.ring
&& !q_vector
->tx
.ring
)
2994 /* clear the affinity_mask in the IRQ descriptor */
2995 irq_set_affinity_hint(entry
->vector
, NULL
);
2997 free_irq(entry
->vector
, q_vector
);
3000 free_irq(adapter
->msix_entries
[vector
++].vector
, adapter
);
3004 * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3005 * @adapter: board private structure
3007 static inline void ixgbe_irq_disable(struct ixgbe_adapter
*adapter
)
3009 switch (adapter
->hw
.mac
.type
) {
3010 case ixgbe_mac_82598EB
:
3011 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC
, ~0);
3013 case ixgbe_mac_82599EB
:
3014 case ixgbe_mac_X540
:
3015 case ixgbe_mac_X550
:
3016 case ixgbe_mac_X550EM_x
:
3017 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC
, 0xFFFF0000);
3018 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC_EX(0), ~0);
3019 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC_EX(1), ~0);
3024 IXGBE_WRITE_FLUSH(&adapter
->hw
);
3025 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
) {
3028 for (vector
= 0; vector
< adapter
->num_q_vectors
; vector
++)
3029 synchronize_irq(adapter
->msix_entries
[vector
].vector
);
3031 synchronize_irq(adapter
->msix_entries
[vector
++].vector
);
3033 synchronize_irq(adapter
->pdev
->irq
);
3038 * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3041 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter
*adapter
)
3043 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[0];
3045 ixgbe_write_eitr(q_vector
);
3047 ixgbe_set_ivar(adapter
, 0, 0, 0);
3048 ixgbe_set_ivar(adapter
, 1, 0, 0);
3050 e_info(hw
, "Legacy interrupt IVAR setup done\n");
3054 * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3055 * @adapter: board private structure
3056 * @ring: structure containing ring specific data
3058 * Configure the Tx descriptor ring after a reset.
3060 void ixgbe_configure_tx_ring(struct ixgbe_adapter
*adapter
,
3061 struct ixgbe_ring
*ring
)
3063 struct ixgbe_hw
*hw
= &adapter
->hw
;
3064 u64 tdba
= ring
->dma
;
3066 u32 txdctl
= IXGBE_TXDCTL_ENABLE
;
3067 u8 reg_idx
= ring
->reg_idx
;
3069 /* disable queue to avoid issues while updating state */
3070 IXGBE_WRITE_REG(hw
, IXGBE_TXDCTL(reg_idx
), 0);
3071 IXGBE_WRITE_FLUSH(hw
);
3073 IXGBE_WRITE_REG(hw
, IXGBE_TDBAL(reg_idx
),
3074 (tdba
& DMA_BIT_MASK(32)));
3075 IXGBE_WRITE_REG(hw
, IXGBE_TDBAH(reg_idx
), (tdba
>> 32));
3076 IXGBE_WRITE_REG(hw
, IXGBE_TDLEN(reg_idx
),
3077 ring
->count
* sizeof(union ixgbe_adv_tx_desc
));
3078 IXGBE_WRITE_REG(hw
, IXGBE_TDH(reg_idx
), 0);
3079 IXGBE_WRITE_REG(hw
, IXGBE_TDT(reg_idx
), 0);
3080 ring
->tail
= adapter
->io_addr
+ IXGBE_TDT(reg_idx
);
3083 * set WTHRESH to encourage burst writeback, it should not be set
3084 * higher than 1 when:
3085 * - ITR is 0 as it could cause false TX hangs
3086 * - ITR is set to > 100k int/sec and BQL is enabled
3088 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3089 * to or less than the number of on chip descriptors, which is
3092 if (!ring
->q_vector
|| (ring
->q_vector
->itr
< IXGBE_100K_ITR
))
3093 txdctl
|= (1 << 16); /* WTHRESH = 1 */
3095 txdctl
|= (8 << 16); /* WTHRESH = 8 */
3098 * Setting PTHRESH to 32 both improves performance
3099 * and avoids a TX hang with DFP enabled
3101 txdctl
|= (1 << 8) | /* HTHRESH = 1 */
3102 32; /* PTHRESH = 32 */
3104 /* reinitialize flowdirector state */
3105 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) {
3106 ring
->atr_sample_rate
= adapter
->atr_sample_rate
;
3107 ring
->atr_count
= 0;
3108 set_bit(__IXGBE_TX_FDIR_INIT_DONE
, &ring
->state
);
3110 ring
->atr_sample_rate
= 0;
3113 /* initialize XPS */
3114 if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE
, &ring
->state
)) {
3115 struct ixgbe_q_vector
*q_vector
= ring
->q_vector
;
3118 netif_set_xps_queue(ring
->netdev
,
3119 &q_vector
->affinity_mask
,
3123 clear_bit(__IXGBE_HANG_CHECK_ARMED
, &ring
->state
);
3126 IXGBE_WRITE_REG(hw
, IXGBE_TXDCTL(reg_idx
), txdctl
);
3128 /* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3129 if (hw
->mac
.type
== ixgbe_mac_82598EB
&&
3130 !(IXGBE_READ_REG(hw
, IXGBE_LINKS
) & IXGBE_LINKS_UP
))
3133 /* poll to verify queue is enabled */
3135 usleep_range(1000, 2000);
3136 txdctl
= IXGBE_READ_REG(hw
, IXGBE_TXDCTL(reg_idx
));
3137 } while (--wait_loop
&& !(txdctl
& IXGBE_TXDCTL_ENABLE
));
3139 e_err(drv
, "Could not enable Tx Queue %d\n", reg_idx
);
3142 static void ixgbe_setup_mtqc(struct ixgbe_adapter
*adapter
)
3144 struct ixgbe_hw
*hw
= &adapter
->hw
;
3146 u8 tcs
= netdev_get_num_tc(adapter
->netdev
);
3148 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
3151 /* disable the arbiter while setting MTQC */
3152 rttdcs
= IXGBE_READ_REG(hw
, IXGBE_RTTDCS
);
3153 rttdcs
|= IXGBE_RTTDCS_ARBDIS
;
3154 IXGBE_WRITE_REG(hw
, IXGBE_RTTDCS
, rttdcs
);
3156 /* set transmit pool layout */
3157 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
3158 mtqc
= IXGBE_MTQC_VT_ENA
;
3160 mtqc
|= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_8TC_8TQ
;
3162 mtqc
|= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_4TC_4TQ
;
3163 else if (adapter
->ring_feature
[RING_F_RSS
].indices
== 4)
3164 mtqc
|= IXGBE_MTQC_32VF
;
3166 mtqc
|= IXGBE_MTQC_64VF
;
3169 mtqc
= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_8TC_8TQ
;
3171 mtqc
= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_4TC_4TQ
;
3173 mtqc
= IXGBE_MTQC_64Q_1PB
;
3176 IXGBE_WRITE_REG(hw
, IXGBE_MTQC
, mtqc
);
3178 /* Enable Security TX Buffer IFG for multiple pb */
3180 u32 sectx
= IXGBE_READ_REG(hw
, IXGBE_SECTXMINIFG
);
3181 sectx
|= IXGBE_SECTX_DCB
;
3182 IXGBE_WRITE_REG(hw
, IXGBE_SECTXMINIFG
, sectx
);
3185 /* re-enable the arbiter */
3186 rttdcs
&= ~IXGBE_RTTDCS_ARBDIS
;
3187 IXGBE_WRITE_REG(hw
, IXGBE_RTTDCS
, rttdcs
);
3191 * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3192 * @adapter: board private structure
3194 * Configure the Tx unit of the MAC after a reset.
3196 static void ixgbe_configure_tx(struct ixgbe_adapter
*adapter
)
3198 struct ixgbe_hw
*hw
= &adapter
->hw
;
3202 ixgbe_setup_mtqc(adapter
);
3204 if (hw
->mac
.type
!= ixgbe_mac_82598EB
) {
3205 /* DMATXCTL.EN must be before Tx queues are enabled */
3206 dmatxctl
= IXGBE_READ_REG(hw
, IXGBE_DMATXCTL
);
3207 dmatxctl
|= IXGBE_DMATXCTL_TE
;
3208 IXGBE_WRITE_REG(hw
, IXGBE_DMATXCTL
, dmatxctl
);
3211 /* Setup the HW Tx Head and Tail descriptor pointers */
3212 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
3213 ixgbe_configure_tx_ring(adapter
, adapter
->tx_ring
[i
]);
3216 static void ixgbe_enable_rx_drop(struct ixgbe_adapter
*adapter
,
3217 struct ixgbe_ring
*ring
)
3219 struct ixgbe_hw
*hw
= &adapter
->hw
;
3220 u8 reg_idx
= ring
->reg_idx
;
3221 u32 srrctl
= IXGBE_READ_REG(hw
, IXGBE_SRRCTL(reg_idx
));
3223 srrctl
|= IXGBE_SRRCTL_DROP_EN
;
3225 IXGBE_WRITE_REG(hw
, IXGBE_SRRCTL(reg_idx
), srrctl
);
3228 static void ixgbe_disable_rx_drop(struct ixgbe_adapter
*adapter
,
3229 struct ixgbe_ring
*ring
)
3231 struct ixgbe_hw
*hw
= &adapter
->hw
;
3232 u8 reg_idx
= ring
->reg_idx
;
3233 u32 srrctl
= IXGBE_READ_REG(hw
, IXGBE_SRRCTL(reg_idx
));
3235 srrctl
&= ~IXGBE_SRRCTL_DROP_EN
;
3237 IXGBE_WRITE_REG(hw
, IXGBE_SRRCTL(reg_idx
), srrctl
);
3240 #ifdef CONFIG_IXGBE_DCB
3241 void ixgbe_set_rx_drop_en(struct ixgbe_adapter
*adapter
)
3243 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter
*adapter
)
3247 bool pfc_en
= adapter
->dcb_cfg
.pfc_mode_enable
;
3249 if (adapter
->ixgbe_ieee_pfc
)
3250 pfc_en
|= !!(adapter
->ixgbe_ieee_pfc
->pfc_en
);
3253 * We should set the drop enable bit if:
3256 * Number of Rx queues > 1 and flow control is disabled
3258 * This allows us to avoid head of line blocking for security
3259 * and performance reasons.
3261 if (adapter
->num_vfs
|| (adapter
->num_rx_queues
> 1 &&
3262 !(adapter
->hw
.fc
.current_mode
& ixgbe_fc_tx_pause
) && !pfc_en
)) {
3263 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3264 ixgbe_enable_rx_drop(adapter
, adapter
->rx_ring
[i
]);
3266 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3267 ixgbe_disable_rx_drop(adapter
, adapter
->rx_ring
[i
]);
3271 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3273 static void ixgbe_configure_srrctl(struct ixgbe_adapter
*adapter
,
3274 struct ixgbe_ring
*rx_ring
)
3276 struct ixgbe_hw
*hw
= &adapter
->hw
;
3278 u8 reg_idx
= rx_ring
->reg_idx
;
3280 if (hw
->mac
.type
== ixgbe_mac_82598EB
) {
3281 u16 mask
= adapter
->ring_feature
[RING_F_RSS
].mask
;
3284 * if VMDq is not active we must program one srrctl register
3285 * per RSS queue since we have enabled RDRXCTL.MVMEN
3290 /* configure header buffer length, needed for RSC */
3291 srrctl
= IXGBE_RX_HDR_SIZE
<< IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT
;
3293 /* configure the packet buffer length */
3294 srrctl
|= ixgbe_rx_bufsz(rx_ring
) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT
;
3296 /* configure descriptor type */
3297 srrctl
|= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF
;
3299 IXGBE_WRITE_REG(hw
, IXGBE_SRRCTL(reg_idx
), srrctl
);
3303 * Return a number of entries in the RSS indirection table
3305 * @adapter: device handle
3307 * - 82598/82599/X540: 128
3308 * - X550(non-SRIOV mode): 512
3309 * - X550(SRIOV mode): 64
3311 u32
ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter
*adapter
)
3313 if (adapter
->hw
.mac
.type
< ixgbe_mac_X550
)
3315 else if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
3322 * Write the RETA table to HW
3324 * @adapter: device handle
3326 * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3328 void ixgbe_store_reta(struct ixgbe_adapter
*adapter
)
3330 u32 i
, reta_entries
= ixgbe_rss_indir_tbl_entries(adapter
);
3331 struct ixgbe_hw
*hw
= &adapter
->hw
;
3334 u8
*indir_tbl
= adapter
->rss_indir_tbl
;
3336 /* Fill out the redirection table as follows:
3337 * - 82598: 8 bit wide entries containing pair of 4 bit RSS
3339 * - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3340 * - X550: 8 bit wide entries containing 6 bit RSS index
3342 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
)
3343 indices_multi
= 0x11;
3345 indices_multi
= 0x1;
3347 /* Write redirection table to HW */
3348 for (i
= 0; i
< reta_entries
; i
++) {
3349 reta
|= indices_multi
* indir_tbl
[i
] << (i
& 0x3) * 8;
3352 IXGBE_WRITE_REG(hw
, IXGBE_RETA(i
>> 2), reta
);
3354 IXGBE_WRITE_REG(hw
, IXGBE_ERETA((i
>> 2) - 32),
3362 * Write the RETA table to HW (for x550 devices in SRIOV mode)
3364 * @adapter: device handle
3366 * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3368 static void ixgbe_store_vfreta(struct ixgbe_adapter
*adapter
)
3370 u32 i
, reta_entries
= ixgbe_rss_indir_tbl_entries(adapter
);
3371 struct ixgbe_hw
*hw
= &adapter
->hw
;
3373 unsigned int pf_pool
= adapter
->num_vfs
;
3375 /* Write redirection table to HW */
3376 for (i
= 0; i
< reta_entries
; i
++) {
3377 vfreta
|= (u32
)adapter
->rss_indir_tbl
[i
] << (i
& 0x3) * 8;
3379 IXGBE_WRITE_REG(hw
, IXGBE_PFVFRETA(i
>> 2, pf_pool
),
3386 static void ixgbe_setup_reta(struct ixgbe_adapter
*adapter
)
3388 struct ixgbe_hw
*hw
= &adapter
->hw
;
3390 u32 reta_entries
= ixgbe_rss_indir_tbl_entries(adapter
);
3391 u16 rss_i
= adapter
->ring_feature
[RING_F_RSS
].indices
;
3393 /* Program table for at least 2 queues w/ SR-IOV so that VFs can
3394 * make full use of any rings they may have. We will use the
3395 * PSRTYPE register to control how many rings we use within the PF.
3397 if ((adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) && (rss_i
< 2))
3400 /* Fill out hash function seeds */
3401 for (i
= 0; i
< 10; i
++)
3402 IXGBE_WRITE_REG(hw
, IXGBE_RSSRK(i
), adapter
->rss_key
[i
]);
3404 /* Fill out redirection table */
3405 memset(adapter
->rss_indir_tbl
, 0, sizeof(adapter
->rss_indir_tbl
));
3407 for (i
= 0, j
= 0; i
< reta_entries
; i
++, j
++) {
3411 adapter
->rss_indir_tbl
[i
] = j
;
3414 ixgbe_store_reta(adapter
);
3417 static void ixgbe_setup_vfreta(struct ixgbe_adapter
*adapter
)
3419 struct ixgbe_hw
*hw
= &adapter
->hw
;
3420 u16 rss_i
= adapter
->ring_feature
[RING_F_RSS
].indices
;
3421 unsigned int pf_pool
= adapter
->num_vfs
;
3424 /* Fill out hash function seeds */
3425 for (i
= 0; i
< 10; i
++)
3426 IXGBE_WRITE_REG(hw
, IXGBE_PFVFRSSRK(i
, pf_pool
),
3427 adapter
->rss_key
[i
]);
3429 /* Fill out the redirection table */
3430 for (i
= 0, j
= 0; i
< 64; i
++, j
++) {
3434 adapter
->rss_indir_tbl
[i
] = j
;
3437 ixgbe_store_vfreta(adapter
);
3440 static void ixgbe_setup_mrqc(struct ixgbe_adapter
*adapter
)
3442 struct ixgbe_hw
*hw
= &adapter
->hw
;
3443 u32 mrqc
= 0, rss_field
= 0, vfmrqc
= 0;
3446 /* Disable indicating checksum in descriptor, enables RSS hash */
3447 rxcsum
= IXGBE_READ_REG(hw
, IXGBE_RXCSUM
);
3448 rxcsum
|= IXGBE_RXCSUM_PCSD
;
3449 IXGBE_WRITE_REG(hw
, IXGBE_RXCSUM
, rxcsum
);
3451 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
) {
3452 if (adapter
->ring_feature
[RING_F_RSS
].mask
)
3453 mrqc
= IXGBE_MRQC_RSSEN
;
3455 u8 tcs
= netdev_get_num_tc(adapter
->netdev
);
3457 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
3459 mrqc
= IXGBE_MRQC_VMDQRT8TCEN
; /* 8 TCs */
3461 mrqc
= IXGBE_MRQC_VMDQRT4TCEN
; /* 4 TCs */
3462 else if (adapter
->ring_feature
[RING_F_RSS
].indices
== 4)
3463 mrqc
= IXGBE_MRQC_VMDQRSS32EN
;
3465 mrqc
= IXGBE_MRQC_VMDQRSS64EN
;
3468 mrqc
= IXGBE_MRQC_RTRSS8TCEN
;
3470 mrqc
= IXGBE_MRQC_RTRSS4TCEN
;
3472 mrqc
= IXGBE_MRQC_RSSEN
;
3476 /* Perform hash on these packet types */
3477 rss_field
|= IXGBE_MRQC_RSS_FIELD_IPV4
|
3478 IXGBE_MRQC_RSS_FIELD_IPV4_TCP
|
3479 IXGBE_MRQC_RSS_FIELD_IPV6
|
3480 IXGBE_MRQC_RSS_FIELD_IPV6_TCP
;
3482 if (adapter
->flags2
& IXGBE_FLAG2_RSS_FIELD_IPV4_UDP
)
3483 rss_field
|= IXGBE_MRQC_RSS_FIELD_IPV4_UDP
;
3484 if (adapter
->flags2
& IXGBE_FLAG2_RSS_FIELD_IPV6_UDP
)
3485 rss_field
|= IXGBE_MRQC_RSS_FIELD_IPV6_UDP
;
3487 netdev_rss_key_fill(adapter
->rss_key
, sizeof(adapter
->rss_key
));
3488 if ((hw
->mac
.type
>= ixgbe_mac_X550
) &&
3489 (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)) {
3490 unsigned int pf_pool
= adapter
->num_vfs
;
3492 /* Enable VF RSS mode */
3493 mrqc
|= IXGBE_MRQC_MULTIPLE_RSS
;
3494 IXGBE_WRITE_REG(hw
, IXGBE_MRQC
, mrqc
);
3496 /* Setup RSS through the VF registers */
3497 ixgbe_setup_vfreta(adapter
);
3498 vfmrqc
= IXGBE_MRQC_RSSEN
;
3499 vfmrqc
|= rss_field
;
3500 IXGBE_WRITE_REG(hw
, IXGBE_PFVFMRQC(pf_pool
), vfmrqc
);
3502 ixgbe_setup_reta(adapter
);
3504 IXGBE_WRITE_REG(hw
, IXGBE_MRQC
, mrqc
);
3509 * ixgbe_configure_rscctl - enable RSC for the indicated ring
3510 * @adapter: address of board private structure
3511 * @index: index of ring to set
3513 static void ixgbe_configure_rscctl(struct ixgbe_adapter
*adapter
,
3514 struct ixgbe_ring
*ring
)
3516 struct ixgbe_hw
*hw
= &adapter
->hw
;
3518 u8 reg_idx
= ring
->reg_idx
;
3520 if (!ring_is_rsc_enabled(ring
))
3523 rscctrl
= IXGBE_READ_REG(hw
, IXGBE_RSCCTL(reg_idx
));
3524 rscctrl
|= IXGBE_RSCCTL_RSCEN
;
3526 * we must limit the number of descriptors so that the
3527 * total size of max desc * buf_len is not greater
3530 rscctrl
|= IXGBE_RSCCTL_MAXDESC_16
;
3531 IXGBE_WRITE_REG(hw
, IXGBE_RSCCTL(reg_idx
), rscctrl
);
3534 #define IXGBE_MAX_RX_DESC_POLL 10
3535 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter
*adapter
,
3536 struct ixgbe_ring
*ring
)
3538 struct ixgbe_hw
*hw
= &adapter
->hw
;
3539 int wait_loop
= IXGBE_MAX_RX_DESC_POLL
;
3541 u8 reg_idx
= ring
->reg_idx
;
3543 if (ixgbe_removed(hw
->hw_addr
))
3545 /* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3546 if (hw
->mac
.type
== ixgbe_mac_82598EB
&&
3547 !(IXGBE_READ_REG(hw
, IXGBE_LINKS
) & IXGBE_LINKS_UP
))
3551 usleep_range(1000, 2000);
3552 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3553 } while (--wait_loop
&& !(rxdctl
& IXGBE_RXDCTL_ENABLE
));
3556 e_err(drv
, "RXDCTL.ENABLE on Rx queue %d not set within "
3557 "the polling period\n", reg_idx
);
3561 void ixgbe_disable_rx_queue(struct ixgbe_adapter
*adapter
,
3562 struct ixgbe_ring
*ring
)
3564 struct ixgbe_hw
*hw
= &adapter
->hw
;
3565 int wait_loop
= IXGBE_MAX_RX_DESC_POLL
;
3567 u8 reg_idx
= ring
->reg_idx
;
3569 if (ixgbe_removed(hw
->hw_addr
))
3571 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3572 rxdctl
&= ~IXGBE_RXDCTL_ENABLE
;
3574 /* write value back with RXDCTL.ENABLE bit cleared */
3575 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(reg_idx
), rxdctl
);
3577 if (hw
->mac
.type
== ixgbe_mac_82598EB
&&
3578 !(IXGBE_READ_REG(hw
, IXGBE_LINKS
) & IXGBE_LINKS_UP
))
3581 /* the hardware may take up to 100us to really disable the rx queue */
3584 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3585 } while (--wait_loop
&& (rxdctl
& IXGBE_RXDCTL_ENABLE
));
3588 e_err(drv
, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3589 "the polling period\n", reg_idx
);
3593 void ixgbe_configure_rx_ring(struct ixgbe_adapter
*adapter
,
3594 struct ixgbe_ring
*ring
)
3596 struct ixgbe_hw
*hw
= &adapter
->hw
;
3597 u64 rdba
= ring
->dma
;
3599 u8 reg_idx
= ring
->reg_idx
;
3601 /* disable queue to avoid issues while updating state */
3602 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3603 ixgbe_disable_rx_queue(adapter
, ring
);
3605 IXGBE_WRITE_REG(hw
, IXGBE_RDBAL(reg_idx
), (rdba
& DMA_BIT_MASK(32)));
3606 IXGBE_WRITE_REG(hw
, IXGBE_RDBAH(reg_idx
), (rdba
>> 32));
3607 IXGBE_WRITE_REG(hw
, IXGBE_RDLEN(reg_idx
),
3608 ring
->count
* sizeof(union ixgbe_adv_rx_desc
));
3609 IXGBE_WRITE_REG(hw
, IXGBE_RDH(reg_idx
), 0);
3610 IXGBE_WRITE_REG(hw
, IXGBE_RDT(reg_idx
), 0);
3611 ring
->tail
= adapter
->io_addr
+ IXGBE_RDT(reg_idx
);
3613 ixgbe_configure_srrctl(adapter
, ring
);
3614 ixgbe_configure_rscctl(adapter
, ring
);
3616 if (hw
->mac
.type
== ixgbe_mac_82598EB
) {
3618 * enable cache line friendly hardware writes:
3619 * PTHRESH=32 descriptors (half the internal cache),
3620 * this also removes ugly rx_no_buffer_count increment
3621 * HTHRESH=4 descriptors (to minimize latency on fetch)
3622 * WTHRESH=8 burst writeback up to two cache lines
3624 rxdctl
&= ~0x3FFFFF;
3628 /* enable receive descriptor ring */
3629 rxdctl
|= IXGBE_RXDCTL_ENABLE
;
3630 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(reg_idx
), rxdctl
);
3632 ixgbe_rx_desc_queue_enable(adapter
, ring
);
3633 ixgbe_alloc_rx_buffers(ring
, ixgbe_desc_unused(ring
));
3636 static void ixgbe_setup_psrtype(struct ixgbe_adapter
*adapter
)
3638 struct ixgbe_hw
*hw
= &adapter
->hw
;
3639 int rss_i
= adapter
->ring_feature
[RING_F_RSS
].indices
;
3642 /* PSRTYPE must be initialized in non 82598 adapters */
3643 u32 psrtype
= IXGBE_PSRTYPE_TCPHDR
|
3644 IXGBE_PSRTYPE_UDPHDR
|
3645 IXGBE_PSRTYPE_IPV4HDR
|
3646 IXGBE_PSRTYPE_L2HDR
|
3647 IXGBE_PSRTYPE_IPV6HDR
;
3649 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
3657 for_each_set_bit(pool
, &adapter
->fwd_bitmask
, 32)
3658 IXGBE_WRITE_REG(hw
, IXGBE_PSRTYPE(VMDQ_P(pool
)), psrtype
);
3661 static void ixgbe_configure_virtualization(struct ixgbe_adapter
*adapter
)
3663 struct ixgbe_hw
*hw
= &adapter
->hw
;
3664 u32 reg_offset
, vf_shift
;
3665 u32 gcr_ext
, vmdctl
;
3668 if (!(adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
))
3671 vmdctl
= IXGBE_READ_REG(hw
, IXGBE_VT_CTL
);
3672 vmdctl
|= IXGBE_VMD_CTL_VMDQ_EN
;
3673 vmdctl
&= ~IXGBE_VT_CTL_POOL_MASK
;
3674 vmdctl
|= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT
;
3675 vmdctl
|= IXGBE_VT_CTL_REPLEN
;
3676 IXGBE_WRITE_REG(hw
, IXGBE_VT_CTL
, vmdctl
);
3678 vf_shift
= VMDQ_P(0) % 32;
3679 reg_offset
= (VMDQ_P(0) >= 32) ? 1 : 0;
3681 /* Enable only the PF's pool for Tx/Rx */
3682 IXGBE_WRITE_REG(hw
, IXGBE_VFRE(reg_offset
), (~0) << vf_shift
);
3683 IXGBE_WRITE_REG(hw
, IXGBE_VFRE(reg_offset
^ 1), reg_offset
- 1);
3684 IXGBE_WRITE_REG(hw
, IXGBE_VFTE(reg_offset
), (~0) << vf_shift
);
3685 IXGBE_WRITE_REG(hw
, IXGBE_VFTE(reg_offset
^ 1), reg_offset
- 1);
3686 if (adapter
->bridge_mode
== BRIDGE_MODE_VEB
)
3687 IXGBE_WRITE_REG(hw
, IXGBE_PFDTXGSWC
, IXGBE_PFDTXGSWC_VT_LBEN
);
3689 /* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3690 hw
->mac
.ops
.set_vmdq(hw
, 0, VMDQ_P(0));
3693 * Set up VF register offsets for selected VT Mode,
3694 * i.e. 32 or 64 VFs for SR-IOV
3696 switch (adapter
->ring_feature
[RING_F_VMDQ
].mask
) {
3697 case IXGBE_82599_VMDQ_8Q_MASK
:
3698 gcr_ext
= IXGBE_GCR_EXT_VT_MODE_16
;
3700 case IXGBE_82599_VMDQ_4Q_MASK
:
3701 gcr_ext
= IXGBE_GCR_EXT_VT_MODE_32
;
3704 gcr_ext
= IXGBE_GCR_EXT_VT_MODE_64
;
3708 IXGBE_WRITE_REG(hw
, IXGBE_GCR_EXT
, gcr_ext
);
3711 /* Enable MAC Anti-Spoofing */
3712 hw
->mac
.ops
.set_mac_anti_spoofing(hw
, (adapter
->num_vfs
!= 0),
3715 /* Ensure LLDP is set for Ethertype Antispoofing if we will be
3716 * calling set_ethertype_anti_spoofing for each VF in loop below
3718 if (hw
->mac
.ops
.set_ethertype_anti_spoofing
)
3719 IXGBE_WRITE_REG(hw
, IXGBE_ETQF(IXGBE_ETQF_FILTER_LLDP
),
3720 (IXGBE_ETQF_FILTER_EN
| /* enable filter */
3721 IXGBE_ETQF_TX_ANTISPOOF
| /* tx antispoof */
3722 IXGBE_ETH_P_LLDP
)); /* LLDP eth type */
3724 /* For VFs that have spoof checking turned off */
3725 for (i
= 0; i
< adapter
->num_vfs
; i
++) {
3726 if (!adapter
->vfinfo
[i
].spoofchk_enabled
)
3727 ixgbe_ndo_set_vf_spoofchk(adapter
->netdev
, i
, false);
3729 /* enable ethertype anti spoofing if hw supports it */
3730 if (hw
->mac
.ops
.set_ethertype_anti_spoofing
)
3731 hw
->mac
.ops
.set_ethertype_anti_spoofing(hw
, true, i
);
3733 /* Enable/Disable RSS query feature */
3734 ixgbe_ndo_set_vf_rss_query_en(adapter
->netdev
, i
,
3735 adapter
->vfinfo
[i
].rss_query_enabled
);
3739 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter
*adapter
)
3741 struct ixgbe_hw
*hw
= &adapter
->hw
;
3742 struct net_device
*netdev
= adapter
->netdev
;
3743 int max_frame
= netdev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
3744 struct ixgbe_ring
*rx_ring
;
3749 /* adjust max frame to be able to do baby jumbo for FCoE */
3750 if ((adapter
->flags
& IXGBE_FLAG_FCOE_ENABLED
) &&
3751 (max_frame
< IXGBE_FCOE_JUMBO_FRAME_SIZE
))
3752 max_frame
= IXGBE_FCOE_JUMBO_FRAME_SIZE
;
3754 #endif /* IXGBE_FCOE */
3756 /* adjust max frame to be at least the size of a standard frame */
3757 if (max_frame
< (ETH_FRAME_LEN
+ ETH_FCS_LEN
))
3758 max_frame
= (ETH_FRAME_LEN
+ ETH_FCS_LEN
);
3760 mhadd
= IXGBE_READ_REG(hw
, IXGBE_MHADD
);
3761 if (max_frame
!= (mhadd
>> IXGBE_MHADD_MFS_SHIFT
)) {
3762 mhadd
&= ~IXGBE_MHADD_MFS_MASK
;
3763 mhadd
|= max_frame
<< IXGBE_MHADD_MFS_SHIFT
;
3765 IXGBE_WRITE_REG(hw
, IXGBE_MHADD
, mhadd
);
3768 hlreg0
= IXGBE_READ_REG(hw
, IXGBE_HLREG0
);
3769 /* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3770 hlreg0
|= IXGBE_HLREG0_JUMBOEN
;
3771 IXGBE_WRITE_REG(hw
, IXGBE_HLREG0
, hlreg0
);
3774 * Setup the HW Rx Head and Tail Descriptor Pointers and
3775 * the Base and Length of the Rx Descriptor Ring
3777 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3778 rx_ring
= adapter
->rx_ring
[i
];
3779 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)
3780 set_ring_rsc_enabled(rx_ring
);
3782 clear_ring_rsc_enabled(rx_ring
);
3786 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter
*adapter
)
3788 struct ixgbe_hw
*hw
= &adapter
->hw
;
3789 u32 rdrxctl
= IXGBE_READ_REG(hw
, IXGBE_RDRXCTL
);
3791 switch (hw
->mac
.type
) {
3792 case ixgbe_mac_82598EB
:
3794 * For VMDq support of different descriptor types or
3795 * buffer sizes through the use of multiple SRRCTL
3796 * registers, RDRXCTL.MVMEN must be set to 1
3798 * also, the manual doesn't mention it clearly but DCA hints
3799 * will only use queue 0's tags unless this bit is set. Side
3800 * effects of setting this bit are only that SRRCTL must be
3801 * fully programmed [0..15]
3803 rdrxctl
|= IXGBE_RDRXCTL_MVMEN
;
3805 case ixgbe_mac_X550
:
3806 case ixgbe_mac_X550EM_x
:
3807 case ixgbe_mac_82599EB
:
3808 case ixgbe_mac_X540
:
3809 /* Disable RSC for ACK packets */
3810 IXGBE_WRITE_REG(hw
, IXGBE_RSCDBU
,
3811 (IXGBE_RSCDBU_RSCACKDIS
| IXGBE_READ_REG(hw
, IXGBE_RSCDBU
)));
3812 rdrxctl
&= ~IXGBE_RDRXCTL_RSCFRSTSIZE
;
3813 /* hardware requires some bits to be set by default */
3814 rdrxctl
|= (IXGBE_RDRXCTL_RSCACKC
| IXGBE_RDRXCTL_FCOE_WRFIX
);
3815 rdrxctl
|= IXGBE_RDRXCTL_CRCSTRIP
;
3818 /* We should do nothing since we don't know this hardware */
3822 IXGBE_WRITE_REG(hw
, IXGBE_RDRXCTL
, rdrxctl
);
3826 * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3827 * @adapter: board private structure
3829 * Configure the Rx unit of the MAC after a reset.
3831 static void ixgbe_configure_rx(struct ixgbe_adapter
*adapter
)
3833 struct ixgbe_hw
*hw
= &adapter
->hw
;
3837 /* disable receives while setting up the descriptors */
3838 hw
->mac
.ops
.disable_rx(hw
);
3840 ixgbe_setup_psrtype(adapter
);
3841 ixgbe_setup_rdrxctl(adapter
);
3844 rfctl
= IXGBE_READ_REG(hw
, IXGBE_RFCTL
);
3845 rfctl
&= ~IXGBE_RFCTL_RSC_DIS
;
3846 if (!(adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
))
3847 rfctl
|= IXGBE_RFCTL_RSC_DIS
;
3848 IXGBE_WRITE_REG(hw
, IXGBE_RFCTL
, rfctl
);
3850 /* Program registers for the distribution of queues */
3851 ixgbe_setup_mrqc(adapter
);
3853 /* set_rx_buffer_len must be called before ring initialization */
3854 ixgbe_set_rx_buffer_len(adapter
);
3857 * Setup the HW Rx Head and Tail Descriptor Pointers and
3858 * the Base and Length of the Rx Descriptor Ring
3860 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3861 ixgbe_configure_rx_ring(adapter
, adapter
->rx_ring
[i
]);
3863 rxctrl
= IXGBE_READ_REG(hw
, IXGBE_RXCTRL
);
3864 /* disable drop enable for 82598 parts */
3865 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
3866 rxctrl
|= IXGBE_RXCTRL_DMBYPS
;
3868 /* enable all receives */
3869 rxctrl
|= IXGBE_RXCTRL_RXEN
;
3870 hw
->mac
.ops
.enable_rx_dma(hw
, rxctrl
);
3873 static int ixgbe_vlan_rx_add_vid(struct net_device
*netdev
,
3874 __be16 proto
, u16 vid
)
3876 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
3877 struct ixgbe_hw
*hw
= &adapter
->hw
;
3879 /* add VID to filter table */
3880 hw
->mac
.ops
.set_vfta(&adapter
->hw
, vid
, VMDQ_P(0), true);
3881 set_bit(vid
, adapter
->active_vlans
);
3886 static int ixgbe_vlan_rx_kill_vid(struct net_device
*netdev
,
3887 __be16 proto
, u16 vid
)
3889 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
3890 struct ixgbe_hw
*hw
= &adapter
->hw
;
3892 /* remove VID from filter table */
3893 hw
->mac
.ops
.set_vfta(&adapter
->hw
, vid
, VMDQ_P(0), false);
3894 clear_bit(vid
, adapter
->active_vlans
);
3900 * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3901 * @adapter: driver data
3903 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter
*adapter
)
3905 struct ixgbe_hw
*hw
= &adapter
->hw
;
3909 switch (hw
->mac
.type
) {
3910 case ixgbe_mac_82598EB
:
3911 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_VLNCTRL
);
3912 vlnctrl
&= ~IXGBE_VLNCTRL_VME
;
3913 IXGBE_WRITE_REG(hw
, IXGBE_VLNCTRL
, vlnctrl
);
3915 case ixgbe_mac_82599EB
:
3916 case ixgbe_mac_X540
:
3917 case ixgbe_mac_X550
:
3918 case ixgbe_mac_X550EM_x
:
3919 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3920 struct ixgbe_ring
*ring
= adapter
->rx_ring
[i
];
3922 if (ring
->l2_accel_priv
)
3925 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(j
));
3926 vlnctrl
&= ~IXGBE_RXDCTL_VME
;
3927 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(j
), vlnctrl
);
3936 * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3937 * @adapter: driver data
3939 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter
*adapter
)
3941 struct ixgbe_hw
*hw
= &adapter
->hw
;
3945 switch (hw
->mac
.type
) {
3946 case ixgbe_mac_82598EB
:
3947 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_VLNCTRL
);
3948 vlnctrl
|= IXGBE_VLNCTRL_VME
;
3949 IXGBE_WRITE_REG(hw
, IXGBE_VLNCTRL
, vlnctrl
);
3951 case ixgbe_mac_82599EB
:
3952 case ixgbe_mac_X540
:
3953 case ixgbe_mac_X550
:
3954 case ixgbe_mac_X550EM_x
:
3955 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3956 struct ixgbe_ring
*ring
= adapter
->rx_ring
[i
];
3958 if (ring
->l2_accel_priv
)
3961 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(j
));
3962 vlnctrl
|= IXGBE_RXDCTL_VME
;
3963 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(j
), vlnctrl
);
3971 static void ixgbe_restore_vlan(struct ixgbe_adapter
*adapter
)
3975 ixgbe_vlan_rx_add_vid(adapter
->netdev
, htons(ETH_P_8021Q
), 0);
3977 for_each_set_bit(vid
, adapter
->active_vlans
, VLAN_N_VID
)
3978 ixgbe_vlan_rx_add_vid(adapter
->netdev
, htons(ETH_P_8021Q
), vid
);
3982 * ixgbe_write_mc_addr_list - write multicast addresses to MTA
3983 * @netdev: network interface device structure
3985 * Writes multicast address list to the MTA hash table.
3986 * Returns: -ENOMEM on failure
3987 * 0 on no addresses written
3988 * X on writing X addresses to MTA
3990 static int ixgbe_write_mc_addr_list(struct net_device
*netdev
)
3992 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
3993 struct ixgbe_hw
*hw
= &adapter
->hw
;
3995 if (!netif_running(netdev
))
3998 if (hw
->mac
.ops
.update_mc_addr_list
)
3999 hw
->mac
.ops
.update_mc_addr_list(hw
, netdev
);
4003 #ifdef CONFIG_PCI_IOV
4004 ixgbe_restore_vf_multicasts(adapter
);
4007 return netdev_mc_count(netdev
);
4010 #ifdef CONFIG_PCI_IOV
4011 void ixgbe_full_sync_mac_table(struct ixgbe_adapter
*adapter
)
4013 struct ixgbe_hw
*hw
= &adapter
->hw
;
4015 for (i
= 0; i
< hw
->mac
.num_rar_entries
; i
++) {
4016 if (adapter
->mac_table
[i
].state
& IXGBE_MAC_STATE_IN_USE
)
4017 hw
->mac
.ops
.set_rar(hw
, i
, adapter
->mac_table
[i
].addr
,
4018 adapter
->mac_table
[i
].queue
,
4021 hw
->mac
.ops
.clear_rar(hw
, i
);
4023 adapter
->mac_table
[i
].state
&= ~(IXGBE_MAC_STATE_MODIFIED
);
4028 static void ixgbe_sync_mac_table(struct ixgbe_adapter
*adapter
)
4030 struct ixgbe_hw
*hw
= &adapter
->hw
;
4032 for (i
= 0; i
< hw
->mac
.num_rar_entries
; i
++) {
4033 if (adapter
->mac_table
[i
].state
& IXGBE_MAC_STATE_MODIFIED
) {
4034 if (adapter
->mac_table
[i
].state
&
4035 IXGBE_MAC_STATE_IN_USE
)
4036 hw
->mac
.ops
.set_rar(hw
, i
,
4037 adapter
->mac_table
[i
].addr
,
4038 adapter
->mac_table
[i
].queue
,
4041 hw
->mac
.ops
.clear_rar(hw
, i
);
4043 adapter
->mac_table
[i
].state
&=
4044 ~(IXGBE_MAC_STATE_MODIFIED
);
4049 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter
*adapter
)
4052 struct ixgbe_hw
*hw
= &adapter
->hw
;
4054 for (i
= 0; i
< hw
->mac
.num_rar_entries
; i
++) {
4055 adapter
->mac_table
[i
].state
|= IXGBE_MAC_STATE_MODIFIED
;
4056 adapter
->mac_table
[i
].state
&= ~IXGBE_MAC_STATE_IN_USE
;
4057 eth_zero_addr(adapter
->mac_table
[i
].addr
);
4058 adapter
->mac_table
[i
].queue
= 0;
4060 ixgbe_sync_mac_table(adapter
);
4063 static int ixgbe_available_rars(struct ixgbe_adapter
*adapter
)
4065 struct ixgbe_hw
*hw
= &adapter
->hw
;
4068 for (i
= 0; i
< hw
->mac
.num_rar_entries
; i
++) {
4069 if (adapter
->mac_table
[i
].state
== 0)
4075 /* this function destroys the first RAR entry */
4076 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter
*adapter
,
4079 struct ixgbe_hw
*hw
= &adapter
->hw
;
4081 memcpy(&adapter
->mac_table
[0].addr
, addr
, ETH_ALEN
);
4082 adapter
->mac_table
[0].queue
= VMDQ_P(0);
4083 adapter
->mac_table
[0].state
= (IXGBE_MAC_STATE_DEFAULT
|
4084 IXGBE_MAC_STATE_IN_USE
);
4085 hw
->mac
.ops
.set_rar(hw
, 0, adapter
->mac_table
[0].addr
,
4086 adapter
->mac_table
[0].queue
,
4090 int ixgbe_add_mac_filter(struct ixgbe_adapter
*adapter
, u8
*addr
, u16 queue
)
4092 struct ixgbe_hw
*hw
= &adapter
->hw
;
4095 if (is_zero_ether_addr(addr
))
4098 for (i
= 0; i
< hw
->mac
.num_rar_entries
; i
++) {
4099 if (adapter
->mac_table
[i
].state
& IXGBE_MAC_STATE_IN_USE
)
4101 adapter
->mac_table
[i
].state
|= (IXGBE_MAC_STATE_MODIFIED
|
4102 IXGBE_MAC_STATE_IN_USE
);
4103 ether_addr_copy(adapter
->mac_table
[i
].addr
, addr
);
4104 adapter
->mac_table
[i
].queue
= queue
;
4105 ixgbe_sync_mac_table(adapter
);
4111 int ixgbe_del_mac_filter(struct ixgbe_adapter
*adapter
, u8
*addr
, u16 queue
)
4113 /* search table for addr, if found, set to 0 and sync */
4115 struct ixgbe_hw
*hw
= &adapter
->hw
;
4117 if (is_zero_ether_addr(addr
))
4120 for (i
= 0; i
< hw
->mac
.num_rar_entries
; i
++) {
4121 if (ether_addr_equal(addr
, adapter
->mac_table
[i
].addr
) &&
4122 adapter
->mac_table
[i
].queue
== queue
) {
4123 adapter
->mac_table
[i
].state
|= IXGBE_MAC_STATE_MODIFIED
;
4124 adapter
->mac_table
[i
].state
&= ~IXGBE_MAC_STATE_IN_USE
;
4125 eth_zero_addr(adapter
->mac_table
[i
].addr
);
4126 adapter
->mac_table
[i
].queue
= 0;
4127 ixgbe_sync_mac_table(adapter
);
4134 * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
4135 * @netdev: network interface device structure
4137 * Writes unicast address list to the RAR table.
4138 * Returns: -ENOMEM on failure/insufficient address space
4139 * 0 on no addresses written
4140 * X on writing X addresses to the RAR table
4142 static int ixgbe_write_uc_addr_list(struct net_device
*netdev
, int vfn
)
4144 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
4147 /* return ENOMEM indicating insufficient memory for addresses */
4148 if (netdev_uc_count(netdev
) > ixgbe_available_rars(adapter
))
4151 if (!netdev_uc_empty(netdev
)) {
4152 struct netdev_hw_addr
*ha
;
4153 netdev_for_each_uc_addr(ha
, netdev
) {
4154 ixgbe_del_mac_filter(adapter
, ha
->addr
, vfn
);
4155 ixgbe_add_mac_filter(adapter
, ha
->addr
, vfn
);
4163 * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4164 * @netdev: network interface device structure
4166 * The set_rx_method entry point is called whenever the unicast/multicast
4167 * address list or the network interface flags are updated. This routine is
4168 * responsible for configuring the hardware for proper unicast, multicast and
4171 void ixgbe_set_rx_mode(struct net_device
*netdev
)
4173 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
4174 struct ixgbe_hw
*hw
= &adapter
->hw
;
4175 u32 fctrl
, vmolr
= IXGBE_VMOLR_BAM
| IXGBE_VMOLR_AUPE
;
4179 /* Check for Promiscuous and All Multicast modes */
4180 fctrl
= IXGBE_READ_REG(hw
, IXGBE_FCTRL
);
4181 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_VLNCTRL
);
4183 /* set all bits that we expect to always be set */
4184 fctrl
&= ~IXGBE_FCTRL_SBP
; /* disable store-bad-packets */
4185 fctrl
|= IXGBE_FCTRL_BAM
;
4186 fctrl
|= IXGBE_FCTRL_DPF
; /* discard pause frames when FC enabled */
4187 fctrl
|= IXGBE_FCTRL_PMCF
;
4189 /* clear the bits we are changing the status of */
4190 fctrl
&= ~(IXGBE_FCTRL_UPE
| IXGBE_FCTRL_MPE
);
4191 vlnctrl
&= ~(IXGBE_VLNCTRL_VFE
| IXGBE_VLNCTRL_CFIEN
);
4192 if (netdev
->flags
& IFF_PROMISC
) {
4193 hw
->addr_ctrl
.user_set_promisc
= true;
4194 fctrl
|= (IXGBE_FCTRL_UPE
| IXGBE_FCTRL_MPE
);
4195 vmolr
|= IXGBE_VMOLR_MPE
;
4196 /* Only disable hardware filter vlans in promiscuous mode
4197 * if SR-IOV and VMDQ are disabled - otherwise ensure
4198 * that hardware VLAN filters remain enabled.
4200 if (adapter
->flags
& (IXGBE_FLAG_VMDQ_ENABLED
|
4201 IXGBE_FLAG_SRIOV_ENABLED
))
4202 vlnctrl
|= (IXGBE_VLNCTRL_VFE
| IXGBE_VLNCTRL_CFIEN
);
4204 if (netdev
->flags
& IFF_ALLMULTI
) {
4205 fctrl
|= IXGBE_FCTRL_MPE
;
4206 vmolr
|= IXGBE_VMOLR_MPE
;
4208 vlnctrl
|= IXGBE_VLNCTRL_VFE
;
4209 hw
->addr_ctrl
.user_set_promisc
= false;
4213 * Write addresses to available RAR registers, if there is not
4214 * sufficient space to store all the addresses then enable
4215 * unicast promiscuous mode
4217 count
= ixgbe_write_uc_addr_list(netdev
, VMDQ_P(0));
4219 fctrl
|= IXGBE_FCTRL_UPE
;
4220 vmolr
|= IXGBE_VMOLR_ROPE
;
4223 /* Write addresses to the MTA, if the attempt fails
4224 * then we should just turn on promiscuous mode so
4225 * that we can at least receive multicast traffic
4227 count
= ixgbe_write_mc_addr_list(netdev
);
4229 fctrl
|= IXGBE_FCTRL_MPE
;
4230 vmolr
|= IXGBE_VMOLR_MPE
;
4232 vmolr
|= IXGBE_VMOLR_ROMPE
;
4235 if (hw
->mac
.type
!= ixgbe_mac_82598EB
) {
4236 vmolr
|= IXGBE_READ_REG(hw
, IXGBE_VMOLR(VMDQ_P(0))) &
4237 ~(IXGBE_VMOLR_MPE
| IXGBE_VMOLR_ROMPE
|
4239 IXGBE_WRITE_REG(hw
, IXGBE_VMOLR(VMDQ_P(0)), vmolr
);
4242 /* This is useful for sniffing bad packets. */
4243 if (adapter
->netdev
->features
& NETIF_F_RXALL
) {
4244 /* UPE and MPE will be handled by normal PROMISC logic
4245 * in e1000e_set_rx_mode */
4246 fctrl
|= (IXGBE_FCTRL_SBP
| /* Receive bad packets */
4247 IXGBE_FCTRL_BAM
| /* RX All Bcast Pkts */
4248 IXGBE_FCTRL_PMCF
); /* RX All MAC Ctrl Pkts */
4250 fctrl
&= ~(IXGBE_FCTRL_DPF
);
4251 /* NOTE: VLAN filtering is disabled by setting PROMISC */
4254 IXGBE_WRITE_REG(hw
, IXGBE_VLNCTRL
, vlnctrl
);
4255 IXGBE_WRITE_REG(hw
, IXGBE_FCTRL
, fctrl
);
4257 if (netdev
->features
& NETIF_F_HW_VLAN_CTAG_RX
)
4258 ixgbe_vlan_strip_enable(adapter
);
4260 ixgbe_vlan_strip_disable(adapter
);
4263 static void ixgbe_napi_enable_all(struct ixgbe_adapter
*adapter
)
4267 for (q_idx
= 0; q_idx
< adapter
->num_q_vectors
; q_idx
++) {
4268 ixgbe_qv_init_lock(adapter
->q_vector
[q_idx
]);
4269 napi_enable(&adapter
->q_vector
[q_idx
]->napi
);
4273 static void ixgbe_napi_disable_all(struct ixgbe_adapter
*adapter
)
4277 for (q_idx
= 0; q_idx
< adapter
->num_q_vectors
; q_idx
++) {
4278 napi_disable(&adapter
->q_vector
[q_idx
]->napi
);
4279 while (!ixgbe_qv_disable(adapter
->q_vector
[q_idx
])) {
4280 pr_info("QV %d locked\n", q_idx
);
4281 usleep_range(1000, 20000);
4286 static void ixgbe_clear_vxlan_port(struct ixgbe_adapter
*adapter
)
4288 switch (adapter
->hw
.mac
.type
) {
4289 case ixgbe_mac_X550
:
4290 case ixgbe_mac_X550EM_x
:
4291 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_VXLANCTRL
, 0);
4292 #ifdef CONFIG_IXGBE_VXLAN
4293 adapter
->vxlan_port
= 0;
4301 #ifdef CONFIG_IXGBE_DCB
4303 * ixgbe_configure_dcb - Configure DCB hardware
4304 * @adapter: ixgbe adapter struct
4306 * This is called by the driver on open to configure the DCB hardware.
4307 * This is also called by the gennetlink interface when reconfiguring
4310 static void ixgbe_configure_dcb(struct ixgbe_adapter
*adapter
)
4312 struct ixgbe_hw
*hw
= &adapter
->hw
;
4313 int max_frame
= adapter
->netdev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
4315 if (!(adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
)) {
4316 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
4317 netif_set_gso_max_size(adapter
->netdev
, 65536);
4321 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
4322 netif_set_gso_max_size(adapter
->netdev
, 32768);
4325 if (adapter
->netdev
->features
& NETIF_F_FCOE_MTU
)
4326 max_frame
= max(max_frame
, IXGBE_FCOE_JUMBO_FRAME_SIZE
);
4329 /* reconfigure the hardware */
4330 if (adapter
->dcbx_cap
& DCB_CAP_DCBX_VER_CEE
) {
4331 ixgbe_dcb_calculate_tc_credits(hw
, &adapter
->dcb_cfg
, max_frame
,
4333 ixgbe_dcb_calculate_tc_credits(hw
, &adapter
->dcb_cfg
, max_frame
,
4335 ixgbe_dcb_hw_config(hw
, &adapter
->dcb_cfg
);
4336 } else if (adapter
->ixgbe_ieee_ets
&& adapter
->ixgbe_ieee_pfc
) {
4337 ixgbe_dcb_hw_ets(&adapter
->hw
,
4338 adapter
->ixgbe_ieee_ets
,
4340 ixgbe_dcb_hw_pfc_config(&adapter
->hw
,
4341 adapter
->ixgbe_ieee_pfc
->pfc_en
,
4342 adapter
->ixgbe_ieee_ets
->prio_tc
);
4345 /* Enable RSS Hash per TC */
4346 if (hw
->mac
.type
!= ixgbe_mac_82598EB
) {
4348 u16 rss_i
= adapter
->ring_feature
[RING_F_RSS
].indices
- 1;
4355 /* write msb to all 8 TCs in one write */
4356 IXGBE_WRITE_REG(hw
, IXGBE_RQTC
, msb
* 0x11111111);
4361 /* Additional bittime to account for IXGBE framing */
4362 #define IXGBE_ETH_FRAMING 20
4365 * ixgbe_hpbthresh - calculate high water mark for flow control
4367 * @adapter: board private structure to calculate for
4368 * @pb: packet buffer to calculate
4370 static int ixgbe_hpbthresh(struct ixgbe_adapter
*adapter
, int pb
)
4372 struct ixgbe_hw
*hw
= &adapter
->hw
;
4373 struct net_device
*dev
= adapter
->netdev
;
4374 int link
, tc
, kb
, marker
;
4377 /* Calculate max LAN frame size */
4378 tc
= link
= dev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
+ IXGBE_ETH_FRAMING
;
4381 /* FCoE traffic class uses FCOE jumbo frames */
4382 if ((dev
->features
& NETIF_F_FCOE_MTU
) &&
4383 (tc
< IXGBE_FCOE_JUMBO_FRAME_SIZE
) &&
4384 (pb
== ixgbe_fcoe_get_tc(adapter
)))
4385 tc
= IXGBE_FCOE_JUMBO_FRAME_SIZE
;
4388 /* Calculate delay value for device */
4389 switch (hw
->mac
.type
) {
4390 case ixgbe_mac_X540
:
4391 case ixgbe_mac_X550
:
4392 case ixgbe_mac_X550EM_x
:
4393 dv_id
= IXGBE_DV_X540(link
, tc
);
4396 dv_id
= IXGBE_DV(link
, tc
);
4400 /* Loopback switch introduces additional latency */
4401 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
4402 dv_id
+= IXGBE_B2BT(tc
);
4404 /* Delay value is calculated in bit times convert to KB */
4405 kb
= IXGBE_BT2KB(dv_id
);
4406 rx_pba
= IXGBE_READ_REG(hw
, IXGBE_RXPBSIZE(pb
)) >> 10;
4408 marker
= rx_pba
- kb
;
4410 /* It is possible that the packet buffer is not large enough
4411 * to provide required headroom. In this case throw an error
4412 * to user and a do the best we can.
4415 e_warn(drv
, "Packet Buffer(%i) can not provide enough"
4416 "headroom to support flow control."
4417 "Decrease MTU or number of traffic classes\n", pb
);
4425 * ixgbe_lpbthresh - calculate low water mark for for flow control
4427 * @adapter: board private structure to calculate for
4428 * @pb: packet buffer to calculate
4430 static int ixgbe_lpbthresh(struct ixgbe_adapter
*adapter
, int pb
)
4432 struct ixgbe_hw
*hw
= &adapter
->hw
;
4433 struct net_device
*dev
= adapter
->netdev
;
4437 /* Calculate max LAN frame size */
4438 tc
= dev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
4441 /* FCoE traffic class uses FCOE jumbo frames */
4442 if ((dev
->features
& NETIF_F_FCOE_MTU
) &&
4443 (tc
< IXGBE_FCOE_JUMBO_FRAME_SIZE
) &&
4444 (pb
== netdev_get_prio_tc_map(dev
, adapter
->fcoe
.up
)))
4445 tc
= IXGBE_FCOE_JUMBO_FRAME_SIZE
;
4448 /* Calculate delay value for device */
4449 switch (hw
->mac
.type
) {
4450 case ixgbe_mac_X540
:
4451 case ixgbe_mac_X550
:
4452 case ixgbe_mac_X550EM_x
:
4453 dv_id
= IXGBE_LOW_DV_X540(tc
);
4456 dv_id
= IXGBE_LOW_DV(tc
);
4460 /* Delay value is calculated in bit times convert to KB */
4461 return IXGBE_BT2KB(dv_id
);
4465 * ixgbe_pbthresh_setup - calculate and setup high low water marks
4467 static void ixgbe_pbthresh_setup(struct ixgbe_adapter
*adapter
)
4469 struct ixgbe_hw
*hw
= &adapter
->hw
;
4470 int num_tc
= netdev_get_num_tc(adapter
->netdev
);
4476 for (i
= 0; i
< num_tc
; i
++) {
4477 hw
->fc
.high_water
[i
] = ixgbe_hpbthresh(adapter
, i
);
4478 hw
->fc
.low_water
[i
] = ixgbe_lpbthresh(adapter
, i
);
4480 /* Low water marks must not be larger than high water marks */
4481 if (hw
->fc
.low_water
[i
] > hw
->fc
.high_water
[i
])
4482 hw
->fc
.low_water
[i
] = 0;
4485 for (; i
< MAX_TRAFFIC_CLASS
; i
++)
4486 hw
->fc
.high_water
[i
] = 0;
4489 static void ixgbe_configure_pb(struct ixgbe_adapter
*adapter
)
4491 struct ixgbe_hw
*hw
= &adapter
->hw
;
4493 u8 tc
= netdev_get_num_tc(adapter
->netdev
);
4495 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
||
4496 adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
)
4497 hdrm
= 32 << adapter
->fdir_pballoc
;
4501 hw
->mac
.ops
.set_rxpba(hw
, tc
, hdrm
, PBA_STRATEGY_EQUAL
);
4502 ixgbe_pbthresh_setup(adapter
);
4505 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter
*adapter
)
4507 struct ixgbe_hw
*hw
= &adapter
->hw
;
4508 struct hlist_node
*node2
;
4509 struct ixgbe_fdir_filter
*filter
;
4511 spin_lock(&adapter
->fdir_perfect_lock
);
4513 if (!hlist_empty(&adapter
->fdir_filter_list
))
4514 ixgbe_fdir_set_input_mask_82599(hw
, &adapter
->fdir_mask
);
4516 hlist_for_each_entry_safe(filter
, node2
,
4517 &adapter
->fdir_filter_list
, fdir_node
) {
4518 ixgbe_fdir_write_perfect_filter_82599(hw
,
4521 (filter
->action
== IXGBE_FDIR_DROP_QUEUE
) ?
4522 IXGBE_FDIR_DROP_QUEUE
:
4523 adapter
->rx_ring
[filter
->action
]->reg_idx
);
4526 spin_unlock(&adapter
->fdir_perfect_lock
);
4529 static void ixgbe_macvlan_set_rx_mode(struct net_device
*dev
, unsigned int pool
,
4530 struct ixgbe_adapter
*adapter
)
4532 struct ixgbe_hw
*hw
= &adapter
->hw
;
4535 /* No unicast promiscuous support for VMDQ devices. */
4536 vmolr
= IXGBE_READ_REG(hw
, IXGBE_VMOLR(pool
));
4537 vmolr
|= (IXGBE_VMOLR_ROMPE
| IXGBE_VMOLR_BAM
| IXGBE_VMOLR_AUPE
);
4539 /* clear the affected bit */
4540 vmolr
&= ~IXGBE_VMOLR_MPE
;
4542 if (dev
->flags
& IFF_ALLMULTI
) {
4543 vmolr
|= IXGBE_VMOLR_MPE
;
4545 vmolr
|= IXGBE_VMOLR_ROMPE
;
4546 hw
->mac
.ops
.update_mc_addr_list(hw
, dev
);
4548 ixgbe_write_uc_addr_list(adapter
->netdev
, pool
);
4549 IXGBE_WRITE_REG(hw
, IXGBE_VMOLR(pool
), vmolr
);
4552 static void ixgbe_fwd_psrtype(struct ixgbe_fwd_adapter
*vadapter
)
4554 struct ixgbe_adapter
*adapter
= vadapter
->real_adapter
;
4555 int rss_i
= adapter
->num_rx_queues_per_pool
;
4556 struct ixgbe_hw
*hw
= &adapter
->hw
;
4557 u16 pool
= vadapter
->pool
;
4558 u32 psrtype
= IXGBE_PSRTYPE_TCPHDR
|
4559 IXGBE_PSRTYPE_UDPHDR
|
4560 IXGBE_PSRTYPE_IPV4HDR
|
4561 IXGBE_PSRTYPE_L2HDR
|
4562 IXGBE_PSRTYPE_IPV6HDR
;
4564 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
4572 IXGBE_WRITE_REG(hw
, IXGBE_PSRTYPE(VMDQ_P(pool
)), psrtype
);
4576 * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4577 * @rx_ring: ring to free buffers from
4579 static void ixgbe_clean_rx_ring(struct ixgbe_ring
*rx_ring
)
4581 struct device
*dev
= rx_ring
->dev
;
4585 /* ring already cleared, nothing to do */
4586 if (!rx_ring
->rx_buffer_info
)
4589 /* Free all the Rx ring sk_buffs */
4590 for (i
= 0; i
< rx_ring
->count
; i
++) {
4591 struct ixgbe_rx_buffer
*rx_buffer
= &rx_ring
->rx_buffer_info
[i
];
4593 if (rx_buffer
->skb
) {
4594 struct sk_buff
*skb
= rx_buffer
->skb
;
4595 if (IXGBE_CB(skb
)->page_released
)
4598 ixgbe_rx_bufsz(rx_ring
),
4601 rx_buffer
->skb
= NULL
;
4604 if (!rx_buffer
->page
)
4607 dma_unmap_page(dev
, rx_buffer
->dma
,
4608 ixgbe_rx_pg_size(rx_ring
), DMA_FROM_DEVICE
);
4609 __free_pages(rx_buffer
->page
, ixgbe_rx_pg_order(rx_ring
));
4611 rx_buffer
->page
= NULL
;
4614 size
= sizeof(struct ixgbe_rx_buffer
) * rx_ring
->count
;
4615 memset(rx_ring
->rx_buffer_info
, 0, size
);
4617 /* Zero out the descriptor ring */
4618 memset(rx_ring
->desc
, 0, rx_ring
->size
);
4620 rx_ring
->next_to_alloc
= 0;
4621 rx_ring
->next_to_clean
= 0;
4622 rx_ring
->next_to_use
= 0;
4625 static void ixgbe_disable_fwd_ring(struct ixgbe_fwd_adapter
*vadapter
,
4626 struct ixgbe_ring
*rx_ring
)
4628 struct ixgbe_adapter
*adapter
= vadapter
->real_adapter
;
4629 int index
= rx_ring
->queue_index
+ vadapter
->rx_base_queue
;
4631 /* shutdown specific queue receive and wait for dma to settle */
4632 ixgbe_disable_rx_queue(adapter
, rx_ring
);
4633 usleep_range(10000, 20000);
4634 ixgbe_irq_disable_queues(adapter
, ((u64
)1 << index
));
4635 ixgbe_clean_rx_ring(rx_ring
);
4636 rx_ring
->l2_accel_priv
= NULL
;
4639 static int ixgbe_fwd_ring_down(struct net_device
*vdev
,
4640 struct ixgbe_fwd_adapter
*accel
)
4642 struct ixgbe_adapter
*adapter
= accel
->real_adapter
;
4643 unsigned int rxbase
= accel
->rx_base_queue
;
4644 unsigned int txbase
= accel
->tx_base_queue
;
4647 netif_tx_stop_all_queues(vdev
);
4649 for (i
= 0; i
< adapter
->num_rx_queues_per_pool
; i
++) {
4650 ixgbe_disable_fwd_ring(accel
, adapter
->rx_ring
[rxbase
+ i
]);
4651 adapter
->rx_ring
[rxbase
+ i
]->netdev
= adapter
->netdev
;
4654 for (i
= 0; i
< adapter
->num_rx_queues_per_pool
; i
++) {
4655 adapter
->tx_ring
[txbase
+ i
]->l2_accel_priv
= NULL
;
4656 adapter
->tx_ring
[txbase
+ i
]->netdev
= adapter
->netdev
;
4663 static int ixgbe_fwd_ring_up(struct net_device
*vdev
,
4664 struct ixgbe_fwd_adapter
*accel
)
4666 struct ixgbe_adapter
*adapter
= accel
->real_adapter
;
4667 unsigned int rxbase
, txbase
, queues
;
4668 int i
, baseq
, err
= 0;
4670 if (!test_bit(accel
->pool
, &adapter
->fwd_bitmask
))
4673 baseq
= accel
->pool
* adapter
->num_rx_queues_per_pool
;
4674 netdev_dbg(vdev
, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
4675 accel
->pool
, adapter
->num_rx_pools
,
4676 baseq
, baseq
+ adapter
->num_rx_queues_per_pool
,
4677 adapter
->fwd_bitmask
);
4679 accel
->netdev
= vdev
;
4680 accel
->rx_base_queue
= rxbase
= baseq
;
4681 accel
->tx_base_queue
= txbase
= baseq
;
4683 for (i
= 0; i
< adapter
->num_rx_queues_per_pool
; i
++)
4684 ixgbe_disable_fwd_ring(accel
, adapter
->rx_ring
[rxbase
+ i
]);
4686 for (i
= 0; i
< adapter
->num_rx_queues_per_pool
; i
++) {
4687 adapter
->rx_ring
[rxbase
+ i
]->netdev
= vdev
;
4688 adapter
->rx_ring
[rxbase
+ i
]->l2_accel_priv
= accel
;
4689 ixgbe_configure_rx_ring(adapter
, adapter
->rx_ring
[rxbase
+ i
]);
4692 for (i
= 0; i
< adapter
->num_rx_queues_per_pool
; i
++) {
4693 adapter
->tx_ring
[txbase
+ i
]->netdev
= vdev
;
4694 adapter
->tx_ring
[txbase
+ i
]->l2_accel_priv
= accel
;
4697 queues
= min_t(unsigned int,
4698 adapter
->num_rx_queues_per_pool
, vdev
->num_tx_queues
);
4699 err
= netif_set_real_num_tx_queues(vdev
, queues
);
4703 err
= netif_set_real_num_rx_queues(vdev
, queues
);
4707 if (is_valid_ether_addr(vdev
->dev_addr
))
4708 ixgbe_add_mac_filter(adapter
, vdev
->dev_addr
, accel
->pool
);
4710 ixgbe_fwd_psrtype(accel
);
4711 ixgbe_macvlan_set_rx_mode(vdev
, accel
->pool
, adapter
);
4714 ixgbe_fwd_ring_down(vdev
, accel
);
4718 static void ixgbe_configure_dfwd(struct ixgbe_adapter
*adapter
)
4720 struct net_device
*upper
;
4721 struct list_head
*iter
;
4724 netdev_for_each_all_upper_dev_rcu(adapter
->netdev
, upper
, iter
) {
4725 if (netif_is_macvlan(upper
)) {
4726 struct macvlan_dev
*dfwd
= netdev_priv(upper
);
4727 struct ixgbe_fwd_adapter
*vadapter
= dfwd
->fwd_priv
;
4729 if (dfwd
->fwd_priv
) {
4730 err
= ixgbe_fwd_ring_up(upper
, vadapter
);
4738 static void ixgbe_configure(struct ixgbe_adapter
*adapter
)
4740 struct ixgbe_hw
*hw
= &adapter
->hw
;
4742 ixgbe_configure_pb(adapter
);
4743 #ifdef CONFIG_IXGBE_DCB
4744 ixgbe_configure_dcb(adapter
);
4747 * We must restore virtualization before VLANs or else
4748 * the VLVF registers will not be populated
4750 ixgbe_configure_virtualization(adapter
);
4752 ixgbe_set_rx_mode(adapter
->netdev
);
4753 ixgbe_restore_vlan(adapter
);
4755 switch (hw
->mac
.type
) {
4756 case ixgbe_mac_82599EB
:
4757 case ixgbe_mac_X540
:
4758 hw
->mac
.ops
.disable_rx_buff(hw
);
4764 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) {
4765 ixgbe_init_fdir_signature_82599(&adapter
->hw
,
4766 adapter
->fdir_pballoc
);
4767 } else if (adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
) {
4768 ixgbe_init_fdir_perfect_82599(&adapter
->hw
,
4769 adapter
->fdir_pballoc
);
4770 ixgbe_fdir_filter_restore(adapter
);
4773 switch (hw
->mac
.type
) {
4774 case ixgbe_mac_82599EB
:
4775 case ixgbe_mac_X540
:
4776 hw
->mac
.ops
.enable_rx_buff(hw
);
4783 /* configure FCoE L2 filters, redirection table, and Rx control */
4784 ixgbe_configure_fcoe(adapter
);
4786 #endif /* IXGBE_FCOE */
4787 ixgbe_configure_tx(adapter
);
4788 ixgbe_configure_rx(adapter
);
4789 ixgbe_configure_dfwd(adapter
);
4793 * ixgbe_sfp_link_config - set up SFP+ link
4794 * @adapter: pointer to private adapter struct
4796 static void ixgbe_sfp_link_config(struct ixgbe_adapter
*adapter
)
4799 * We are assuming the worst case scenario here, and that
4800 * is that an SFP was inserted/removed after the reset
4801 * but before SFP detection was enabled. As such the best
4802 * solution is to just start searching as soon as we start
4804 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
)
4805 adapter
->flags2
|= IXGBE_FLAG2_SEARCH_FOR_SFP
;
4807 adapter
->flags2
|= IXGBE_FLAG2_SFP_NEEDS_RESET
;
4808 adapter
->sfp_poll_time
= 0;
4812 * ixgbe_non_sfp_link_config - set up non-SFP+ link
4813 * @hw: pointer to private hardware struct
4815 * Returns 0 on success, negative on failure
4817 static int ixgbe_non_sfp_link_config(struct ixgbe_hw
*hw
)
4820 bool autoneg
, link_up
= false;
4821 int ret
= IXGBE_ERR_LINK_SETUP
;
4823 if (hw
->mac
.ops
.check_link
)
4824 ret
= hw
->mac
.ops
.check_link(hw
, &speed
, &link_up
, false);
4829 speed
= hw
->phy
.autoneg_advertised
;
4830 if ((!speed
) && (hw
->mac
.ops
.get_link_capabilities
))
4831 ret
= hw
->mac
.ops
.get_link_capabilities(hw
, &speed
,
4836 if (hw
->mac
.ops
.setup_link
)
4837 ret
= hw
->mac
.ops
.setup_link(hw
, speed
, link_up
);
4842 static void ixgbe_setup_gpie(struct ixgbe_adapter
*adapter
)
4844 struct ixgbe_hw
*hw
= &adapter
->hw
;
4847 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
) {
4848 gpie
= IXGBE_GPIE_MSIX_MODE
| IXGBE_GPIE_PBA_SUPPORT
|
4850 gpie
|= IXGBE_GPIE_EIAME
;
4852 * use EIAM to auto-mask when MSI-X interrupt is asserted
4853 * this saves a register write for every interrupt
4855 switch (hw
->mac
.type
) {
4856 case ixgbe_mac_82598EB
:
4857 IXGBE_WRITE_REG(hw
, IXGBE_EIAM
, IXGBE_EICS_RTX_QUEUE
);
4859 case ixgbe_mac_82599EB
:
4860 case ixgbe_mac_X540
:
4861 case ixgbe_mac_X550
:
4862 case ixgbe_mac_X550EM_x
:
4864 IXGBE_WRITE_REG(hw
, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4865 IXGBE_WRITE_REG(hw
, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4869 /* legacy interrupts, use EIAM to auto-mask when reading EICR,
4870 * specifically only auto mask tx and rx interrupts */
4871 IXGBE_WRITE_REG(hw
, IXGBE_EIAM
, IXGBE_EICS_RTX_QUEUE
);
4874 /* XXX: to interrupt immediately for EICS writes, enable this */
4875 /* gpie |= IXGBE_GPIE_EIMEN; */
4877 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
4878 gpie
&= ~IXGBE_GPIE_VTMODE_MASK
;
4880 switch (adapter
->ring_feature
[RING_F_VMDQ
].mask
) {
4881 case IXGBE_82599_VMDQ_8Q_MASK
:
4882 gpie
|= IXGBE_GPIE_VTMODE_16
;
4884 case IXGBE_82599_VMDQ_4Q_MASK
:
4885 gpie
|= IXGBE_GPIE_VTMODE_32
;
4888 gpie
|= IXGBE_GPIE_VTMODE_64
;
4893 /* Enable Thermal over heat sensor interrupt */
4894 if (adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
) {
4895 switch (adapter
->hw
.mac
.type
) {
4896 case ixgbe_mac_82599EB
:
4897 gpie
|= IXGBE_SDP0_GPIEN_8259X
;
4899 case ixgbe_mac_X540
:
4900 gpie
|= IXGBE_EIMS_TS
;
4907 /* Enable fan failure interrupt */
4908 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
)
4909 gpie
|= IXGBE_SDP1_GPIEN(hw
);
4911 switch (hw
->mac
.type
) {
4912 case ixgbe_mac_82599EB
:
4913 gpie
|= IXGBE_SDP1_GPIEN_8259X
| IXGBE_SDP2_GPIEN_8259X
;
4915 case ixgbe_mac_X550EM_x
:
4916 gpie
|= IXGBE_SDP0_GPIEN_X540
;
4922 IXGBE_WRITE_REG(hw
, IXGBE_GPIE
, gpie
);
4925 static void ixgbe_up_complete(struct ixgbe_adapter
*adapter
)
4927 struct ixgbe_hw
*hw
= &adapter
->hw
;
4931 ixgbe_get_hw_control(adapter
);
4932 ixgbe_setup_gpie(adapter
);
4934 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)
4935 ixgbe_configure_msix(adapter
);
4937 ixgbe_configure_msi_and_legacy(adapter
);
4939 /* enable the optics for 82599 SFP+ fiber */
4940 if (hw
->mac
.ops
.enable_tx_laser
)
4941 hw
->mac
.ops
.enable_tx_laser(hw
);
4943 if (hw
->phy
.ops
.set_phy_power
)
4944 hw
->phy
.ops
.set_phy_power(hw
, true);
4946 smp_mb__before_atomic();
4947 clear_bit(__IXGBE_DOWN
, &adapter
->state
);
4948 ixgbe_napi_enable_all(adapter
);
4950 if (ixgbe_is_sfp(hw
)) {
4951 ixgbe_sfp_link_config(adapter
);
4953 err
= ixgbe_non_sfp_link_config(hw
);
4955 e_err(probe
, "link_config FAILED %d\n", err
);
4958 /* clear any pending interrupts, may auto mask */
4959 IXGBE_READ_REG(hw
, IXGBE_EICR
);
4960 ixgbe_irq_enable(adapter
, true, true);
4963 * If this adapter has a fan, check to see if we had a failure
4964 * before we enabled the interrupt.
4966 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
) {
4967 u32 esdp
= IXGBE_READ_REG(hw
, IXGBE_ESDP
);
4968 if (esdp
& IXGBE_ESDP_SDP1
)
4969 e_crit(drv
, "Fan has stopped, replace the adapter\n");
4972 /* bring the link up in the watchdog, this could race with our first
4973 * link up interrupt but shouldn't be a problem */
4974 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_UPDATE
;
4975 adapter
->link_check_timeout
= jiffies
;
4976 mod_timer(&adapter
->service_timer
, jiffies
);
4978 /* Set PF Reset Done bit so PF/VF Mail Ops can work */
4979 ctrl_ext
= IXGBE_READ_REG(hw
, IXGBE_CTRL_EXT
);
4980 ctrl_ext
|= IXGBE_CTRL_EXT_PFRSTD
;
4981 IXGBE_WRITE_REG(hw
, IXGBE_CTRL_EXT
, ctrl_ext
);
4984 void ixgbe_reinit_locked(struct ixgbe_adapter
*adapter
)
4986 WARN_ON(in_interrupt());
4987 /* put off any impending NetWatchDogTimeout */
4988 adapter
->netdev
->trans_start
= jiffies
;
4990 while (test_and_set_bit(__IXGBE_RESETTING
, &adapter
->state
))
4991 usleep_range(1000, 2000);
4992 ixgbe_down(adapter
);
4994 * If SR-IOV enabled then wait a bit before bringing the adapter
4995 * back up to give the VFs time to respond to the reset. The
4996 * two second wait is based upon the watchdog timer cycle in
4999 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
5002 clear_bit(__IXGBE_RESETTING
, &adapter
->state
);
5005 void ixgbe_up(struct ixgbe_adapter
*adapter
)
5007 /* hardware has been reset, we need to reload some things */
5008 ixgbe_configure(adapter
);
5010 ixgbe_up_complete(adapter
);
5013 void ixgbe_reset(struct ixgbe_adapter
*adapter
)
5015 struct ixgbe_hw
*hw
= &adapter
->hw
;
5016 struct net_device
*netdev
= adapter
->netdev
;
5018 u8 old_addr
[ETH_ALEN
];
5020 if (ixgbe_removed(hw
->hw_addr
))
5022 /* lock SFP init bit to prevent race conditions with the watchdog */
5023 while (test_and_set_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
))
5024 usleep_range(1000, 2000);
5026 /* clear all SFP and link config related flags while holding SFP_INIT */
5027 adapter
->flags2
&= ~(IXGBE_FLAG2_SEARCH_FOR_SFP
|
5028 IXGBE_FLAG2_SFP_NEEDS_RESET
);
5029 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_CONFIG
;
5031 err
= hw
->mac
.ops
.init_hw(hw
);
5034 case IXGBE_ERR_SFP_NOT_PRESENT
:
5035 case IXGBE_ERR_SFP_NOT_SUPPORTED
:
5037 case IXGBE_ERR_MASTER_REQUESTS_PENDING
:
5038 e_dev_err("master disable timed out\n");
5040 case IXGBE_ERR_EEPROM_VERSION
:
5041 /* We are running on a pre-production device, log a warning */
5042 e_dev_warn("This device is a pre-production adapter/LOM. "
5043 "Please be aware there may be issues associated with "
5044 "your hardware. If you are experiencing problems "
5045 "please contact your Intel or hardware "
5046 "representative who provided you with this "
5050 e_dev_err("Hardware Error: %d\n", err
);
5053 clear_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
);
5054 /* do not flush user set addresses */
5055 memcpy(old_addr
, &adapter
->mac_table
[0].addr
, netdev
->addr_len
);
5056 ixgbe_flush_sw_mac_table(adapter
);
5057 ixgbe_mac_set_default_filter(adapter
, old_addr
);
5059 /* update SAN MAC vmdq pool selection */
5060 if (hw
->mac
.san_mac_rar_index
)
5061 hw
->mac
.ops
.set_vmdq_san_mac(hw
, VMDQ_P(0));
5063 if (test_bit(__IXGBE_PTP_RUNNING
, &adapter
->state
))
5064 ixgbe_ptp_reset(adapter
);
5066 if (hw
->phy
.ops
.set_phy_power
) {
5067 if (!netif_running(adapter
->netdev
) && !adapter
->wol
)
5068 hw
->phy
.ops
.set_phy_power(hw
, false);
5070 hw
->phy
.ops
.set_phy_power(hw
, true);
5075 * ixgbe_clean_tx_ring - Free Tx Buffers
5076 * @tx_ring: ring to be cleaned
5078 static void ixgbe_clean_tx_ring(struct ixgbe_ring
*tx_ring
)
5080 struct ixgbe_tx_buffer
*tx_buffer_info
;
5084 /* ring already cleared, nothing to do */
5085 if (!tx_ring
->tx_buffer_info
)
5088 /* Free all the Tx ring sk_buffs */
5089 for (i
= 0; i
< tx_ring
->count
; i
++) {
5090 tx_buffer_info
= &tx_ring
->tx_buffer_info
[i
];
5091 ixgbe_unmap_and_free_tx_resource(tx_ring
, tx_buffer_info
);
5094 netdev_tx_reset_queue(txring_txq(tx_ring
));
5096 size
= sizeof(struct ixgbe_tx_buffer
) * tx_ring
->count
;
5097 memset(tx_ring
->tx_buffer_info
, 0, size
);
5099 /* Zero out the descriptor ring */
5100 memset(tx_ring
->desc
, 0, tx_ring
->size
);
5102 tx_ring
->next_to_use
= 0;
5103 tx_ring
->next_to_clean
= 0;
5107 * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5108 * @adapter: board private structure
5110 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter
*adapter
)
5114 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
5115 ixgbe_clean_rx_ring(adapter
->rx_ring
[i
]);
5119 * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5120 * @adapter: board private structure
5122 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter
*adapter
)
5126 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
5127 ixgbe_clean_tx_ring(adapter
->tx_ring
[i
]);
5130 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter
*adapter
)
5132 struct hlist_node
*node2
;
5133 struct ixgbe_fdir_filter
*filter
;
5135 spin_lock(&adapter
->fdir_perfect_lock
);
5137 hlist_for_each_entry_safe(filter
, node2
,
5138 &adapter
->fdir_filter_list
, fdir_node
) {
5139 hlist_del(&filter
->fdir_node
);
5142 adapter
->fdir_filter_count
= 0;
5144 spin_unlock(&adapter
->fdir_perfect_lock
);
5147 void ixgbe_down(struct ixgbe_adapter
*adapter
)
5149 struct net_device
*netdev
= adapter
->netdev
;
5150 struct ixgbe_hw
*hw
= &adapter
->hw
;
5151 struct net_device
*upper
;
5152 struct list_head
*iter
;
5155 /* signal that we are down to the interrupt handler */
5156 if (test_and_set_bit(__IXGBE_DOWN
, &adapter
->state
))
5157 return; /* do nothing if already down */
5159 /* disable receives */
5160 hw
->mac
.ops
.disable_rx(hw
);
5162 /* disable all enabled rx queues */
5163 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
5164 /* this call also flushes the previous write */
5165 ixgbe_disable_rx_queue(adapter
, adapter
->rx_ring
[i
]);
5167 usleep_range(10000, 20000);
5169 netif_tx_stop_all_queues(netdev
);
5171 /* call carrier off first to avoid false dev_watchdog timeouts */
5172 netif_carrier_off(netdev
);
5173 netif_tx_disable(netdev
);
5175 /* disable any upper devices */
5176 netdev_for_each_all_upper_dev_rcu(adapter
->netdev
, upper
, iter
) {
5177 if (netif_is_macvlan(upper
)) {
5178 struct macvlan_dev
*vlan
= netdev_priv(upper
);
5180 if (vlan
->fwd_priv
) {
5181 netif_tx_stop_all_queues(upper
);
5182 netif_carrier_off(upper
);
5183 netif_tx_disable(upper
);
5188 ixgbe_irq_disable(adapter
);
5190 ixgbe_napi_disable_all(adapter
);
5192 adapter
->flags2
&= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT
|
5193 IXGBE_FLAG2_RESET_REQUESTED
);
5194 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_UPDATE
;
5196 del_timer_sync(&adapter
->service_timer
);
5198 if (adapter
->num_vfs
) {
5199 /* Clear EITR Select mapping */
5200 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EITRSEL
, 0);
5202 /* Mark all the VFs as inactive */
5203 for (i
= 0 ; i
< adapter
->num_vfs
; i
++)
5204 adapter
->vfinfo
[i
].clear_to_send
= false;
5206 /* ping all the active vfs to let them know we are going down */
5207 ixgbe_ping_all_vfs(adapter
);
5209 /* Disable all VFTE/VFRE TX/RX */
5210 ixgbe_disable_tx_rx(adapter
);
5213 /* disable transmits in the hardware now that interrupts are off */
5214 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
5215 u8 reg_idx
= adapter
->tx_ring
[i
]->reg_idx
;
5216 IXGBE_WRITE_REG(hw
, IXGBE_TXDCTL(reg_idx
), IXGBE_TXDCTL_SWFLSH
);
5219 /* Disable the Tx DMA engine on 82599 and later MAC */
5220 switch (hw
->mac
.type
) {
5221 case ixgbe_mac_82599EB
:
5222 case ixgbe_mac_X540
:
5223 case ixgbe_mac_X550
:
5224 case ixgbe_mac_X550EM_x
:
5225 IXGBE_WRITE_REG(hw
, IXGBE_DMATXCTL
,
5226 (IXGBE_READ_REG(hw
, IXGBE_DMATXCTL
) &
5227 ~IXGBE_DMATXCTL_TE
));
5233 if (!pci_channel_offline(adapter
->pdev
))
5234 ixgbe_reset(adapter
);
5236 /* power down the optics for 82599 SFP+ fiber */
5237 if (hw
->mac
.ops
.disable_tx_laser
)
5238 hw
->mac
.ops
.disable_tx_laser(hw
);
5240 ixgbe_clean_all_tx_rings(adapter
);
5241 ixgbe_clean_all_rx_rings(adapter
);
5243 #ifdef CONFIG_IXGBE_DCA
5244 /* since we reset the hardware DCA settings were cleared */
5245 ixgbe_setup_dca(adapter
);
5250 * ixgbe_tx_timeout - Respond to a Tx Hang
5251 * @netdev: network interface device structure
5253 static void ixgbe_tx_timeout(struct net_device
*netdev
)
5255 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
5257 /* Do the reset outside of interrupt context */
5258 ixgbe_tx_timeout_reset(adapter
);
5262 * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5263 * @adapter: board private structure to initialize
5265 * ixgbe_sw_init initializes the Adapter private data structure.
5266 * Fields are initialized based on PCI device information and
5267 * OS network device settings (MTU size).
5269 static int ixgbe_sw_init(struct ixgbe_adapter
*adapter
)
5271 struct ixgbe_hw
*hw
= &adapter
->hw
;
5272 struct pci_dev
*pdev
= adapter
->pdev
;
5273 unsigned int rss
, fdir
;
5275 #ifdef CONFIG_IXGBE_DCB
5277 struct tc_configuration
*tc
;
5280 /* PCI config space info */
5282 hw
->vendor_id
= pdev
->vendor
;
5283 hw
->device_id
= pdev
->device
;
5284 hw
->revision_id
= pdev
->revision
;
5285 hw
->subsystem_vendor_id
= pdev
->subsystem_vendor
;
5286 hw
->subsystem_device_id
= pdev
->subsystem_device
;
5288 /* Set common capability flags and settings */
5289 rss
= min_t(int, ixgbe_max_rss_indices(adapter
), num_online_cpus());
5290 adapter
->ring_feature
[RING_F_RSS
].limit
= rss
;
5291 adapter
->flags2
|= IXGBE_FLAG2_RSC_CAPABLE
;
5292 adapter
->flags2
|= IXGBE_FLAG2_RSC_ENABLED
;
5293 adapter
->max_q_vectors
= MAX_Q_VECTORS_82599
;
5294 adapter
->atr_sample_rate
= 20;
5295 fdir
= min_t(int, IXGBE_MAX_FDIR_INDICES
, num_online_cpus());
5296 adapter
->ring_feature
[RING_F_FDIR
].limit
= fdir
;
5297 adapter
->fdir_pballoc
= IXGBE_FDIR_PBALLOC_64K
;
5298 #ifdef CONFIG_IXGBE_DCA
5299 adapter
->flags
|= IXGBE_FLAG_DCA_CAPABLE
;
5302 adapter
->flags
|= IXGBE_FLAG_FCOE_CAPABLE
;
5303 adapter
->flags
&= ~IXGBE_FLAG_FCOE_ENABLED
;
5304 #ifdef CONFIG_IXGBE_DCB
5305 /* Default traffic class to use for FCoE */
5306 adapter
->fcoe
.up
= IXGBE_FCOE_DEFTC
;
5307 #endif /* CONFIG_IXGBE_DCB */
5308 #endif /* IXGBE_FCOE */
5310 adapter
->mac_table
= kzalloc(sizeof(struct ixgbe_mac_addr
) *
5311 hw
->mac
.num_rar_entries
,
5314 /* Set MAC specific capability flags and exceptions */
5315 switch (hw
->mac
.type
) {
5316 case ixgbe_mac_82598EB
:
5317 adapter
->flags2
&= ~IXGBE_FLAG2_RSC_CAPABLE
;
5318 adapter
->flags2
&= ~IXGBE_FLAG2_RSC_ENABLED
;
5320 if (hw
->device_id
== IXGBE_DEV_ID_82598AT
)
5321 adapter
->flags
|= IXGBE_FLAG_FAN_FAIL_CAPABLE
;
5323 adapter
->max_q_vectors
= MAX_Q_VECTORS_82598
;
5324 adapter
->ring_feature
[RING_F_FDIR
].limit
= 0;
5325 adapter
->atr_sample_rate
= 0;
5326 adapter
->fdir_pballoc
= 0;
5328 adapter
->flags
&= ~IXGBE_FLAG_FCOE_CAPABLE
;
5329 adapter
->flags
&= ~IXGBE_FLAG_FCOE_ENABLED
;
5330 #ifdef CONFIG_IXGBE_DCB
5331 adapter
->fcoe
.up
= 0;
5332 #endif /* IXGBE_DCB */
5333 #endif /* IXGBE_FCOE */
5335 case ixgbe_mac_82599EB
:
5336 if (hw
->device_id
== IXGBE_DEV_ID_82599_T3_LOM
)
5337 adapter
->flags2
|= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
;
5339 case ixgbe_mac_X540
:
5340 fwsm
= IXGBE_READ_REG(hw
, IXGBE_FWSM(hw
));
5341 if (fwsm
& IXGBE_FWSM_TS_ENABLED
)
5342 adapter
->flags2
|= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
;
5344 case ixgbe_mac_X550EM_x
:
5345 case ixgbe_mac_X550
:
5346 #ifdef CONFIG_IXGBE_DCA
5347 adapter
->flags
&= ~IXGBE_FLAG_DCA_CAPABLE
;
5349 #ifdef CONFIG_IXGBE_VXLAN
5350 adapter
->flags
|= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE
;
5358 /* FCoE support exists, always init the FCoE lock */
5359 spin_lock_init(&adapter
->fcoe
.lock
);
5362 /* n-tuple support exists, always init our spinlock */
5363 spin_lock_init(&adapter
->fdir_perfect_lock
);
5365 #ifdef CONFIG_IXGBE_DCB
5366 switch (hw
->mac
.type
) {
5367 case ixgbe_mac_X540
:
5368 case ixgbe_mac_X550
:
5369 case ixgbe_mac_X550EM_x
:
5370 adapter
->dcb_cfg
.num_tcs
.pg_tcs
= X540_TRAFFIC_CLASS
;
5371 adapter
->dcb_cfg
.num_tcs
.pfc_tcs
= X540_TRAFFIC_CLASS
;
5374 adapter
->dcb_cfg
.num_tcs
.pg_tcs
= MAX_TRAFFIC_CLASS
;
5375 adapter
->dcb_cfg
.num_tcs
.pfc_tcs
= MAX_TRAFFIC_CLASS
;
5379 /* Configure DCB traffic classes */
5380 for (j
= 0; j
< MAX_TRAFFIC_CLASS
; j
++) {
5381 tc
= &adapter
->dcb_cfg
.tc_config
[j
];
5382 tc
->path
[DCB_TX_CONFIG
].bwg_id
= 0;
5383 tc
->path
[DCB_TX_CONFIG
].bwg_percent
= 12 + (j
& 1);
5384 tc
->path
[DCB_RX_CONFIG
].bwg_id
= 0;
5385 tc
->path
[DCB_RX_CONFIG
].bwg_percent
= 12 + (j
& 1);
5386 tc
->dcb_pfc
= pfc_disabled
;
5389 /* Initialize default user to priority mapping, UPx->TC0 */
5390 tc
= &adapter
->dcb_cfg
.tc_config
[0];
5391 tc
->path
[DCB_TX_CONFIG
].up_to_tc_bitmap
= 0xFF;
5392 tc
->path
[DCB_RX_CONFIG
].up_to_tc_bitmap
= 0xFF;
5394 adapter
->dcb_cfg
.bw_percentage
[DCB_TX_CONFIG
][0] = 100;
5395 adapter
->dcb_cfg
.bw_percentage
[DCB_RX_CONFIG
][0] = 100;
5396 adapter
->dcb_cfg
.pfc_mode_enable
= false;
5397 adapter
->dcb_set_bitmap
= 0x00;
5398 adapter
->dcbx_cap
= DCB_CAP_DCBX_HOST
| DCB_CAP_DCBX_VER_CEE
;
5399 memcpy(&adapter
->temp_dcb_cfg
, &adapter
->dcb_cfg
,
5400 sizeof(adapter
->temp_dcb_cfg
));
5404 /* default flow control settings */
5405 hw
->fc
.requested_mode
= ixgbe_fc_full
;
5406 hw
->fc
.current_mode
= ixgbe_fc_full
; /* init for ethtool output */
5407 ixgbe_pbthresh_setup(adapter
);
5408 hw
->fc
.pause_time
= IXGBE_DEFAULT_FCPAUSE
;
5409 hw
->fc
.send_xon
= true;
5410 hw
->fc
.disable_fc_autoneg
= ixgbe_device_supports_autoneg_fc(hw
);
5412 #ifdef CONFIG_PCI_IOV
5414 e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
5416 /* assign number of SR-IOV VFs */
5417 if (hw
->mac
.type
!= ixgbe_mac_82598EB
) {
5418 if (max_vfs
> IXGBE_MAX_VFS_DRV_LIMIT
) {
5419 adapter
->num_vfs
= 0;
5420 e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
5422 adapter
->num_vfs
= max_vfs
;
5425 #endif /* CONFIG_PCI_IOV */
5427 /* enable itr by default in dynamic mode */
5428 adapter
->rx_itr_setting
= 1;
5429 adapter
->tx_itr_setting
= 1;
5431 /* set default ring sizes */
5432 adapter
->tx_ring_count
= IXGBE_DEFAULT_TXD
;
5433 adapter
->rx_ring_count
= IXGBE_DEFAULT_RXD
;
5435 /* set default work limits */
5436 adapter
->tx_work_limit
= IXGBE_DEFAULT_TX_WORK
;
5438 /* initialize eeprom parameters */
5439 if (ixgbe_init_eeprom_params_generic(hw
)) {
5440 e_dev_err("EEPROM initialization failed\n");
5444 /* PF holds first pool slot */
5445 set_bit(0, &adapter
->fwd_bitmask
);
5446 set_bit(__IXGBE_DOWN
, &adapter
->state
);
5452 * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
5453 * @tx_ring: tx descriptor ring (for a specific queue) to setup
5455 * Return 0 on success, negative on failure
5457 int ixgbe_setup_tx_resources(struct ixgbe_ring
*tx_ring
)
5459 struct device
*dev
= tx_ring
->dev
;
5460 int orig_node
= dev_to_node(dev
);
5464 size
= sizeof(struct ixgbe_tx_buffer
) * tx_ring
->count
;
5466 if (tx_ring
->q_vector
)
5467 ring_node
= tx_ring
->q_vector
->numa_node
;
5469 tx_ring
->tx_buffer_info
= vzalloc_node(size
, ring_node
);
5470 if (!tx_ring
->tx_buffer_info
)
5471 tx_ring
->tx_buffer_info
= vzalloc(size
);
5472 if (!tx_ring
->tx_buffer_info
)
5475 u64_stats_init(&tx_ring
->syncp
);
5477 /* round up to nearest 4K */
5478 tx_ring
->size
= tx_ring
->count
* sizeof(union ixgbe_adv_tx_desc
);
5479 tx_ring
->size
= ALIGN(tx_ring
->size
, 4096);
5481 set_dev_node(dev
, ring_node
);
5482 tx_ring
->desc
= dma_alloc_coherent(dev
,
5486 set_dev_node(dev
, orig_node
);
5488 tx_ring
->desc
= dma_alloc_coherent(dev
, tx_ring
->size
,
5489 &tx_ring
->dma
, GFP_KERNEL
);
5493 tx_ring
->next_to_use
= 0;
5494 tx_ring
->next_to_clean
= 0;
5498 vfree(tx_ring
->tx_buffer_info
);
5499 tx_ring
->tx_buffer_info
= NULL
;
5500 dev_err(dev
, "Unable to allocate memory for the Tx descriptor ring\n");
5505 * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
5506 * @adapter: board private structure
5508 * If this function returns with an error, then it's possible one or
5509 * more of the rings is populated (while the rest are not). It is the
5510 * callers duty to clean those orphaned rings.
5512 * Return 0 on success, negative on failure
5514 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter
*adapter
)
5518 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
5519 err
= ixgbe_setup_tx_resources(adapter
->tx_ring
[i
]);
5523 e_err(probe
, "Allocation for Tx Queue %u failed\n", i
);
5529 /* rewind the index freeing the rings as we go */
5531 ixgbe_free_tx_resources(adapter
->tx_ring
[i
]);
5536 * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
5537 * @rx_ring: rx descriptor ring (for a specific queue) to setup
5539 * Returns 0 on success, negative on failure
5541 int ixgbe_setup_rx_resources(struct ixgbe_ring
*rx_ring
)
5543 struct device
*dev
= rx_ring
->dev
;
5544 int orig_node
= dev_to_node(dev
);
5548 size
= sizeof(struct ixgbe_rx_buffer
) * rx_ring
->count
;
5550 if (rx_ring
->q_vector
)
5551 ring_node
= rx_ring
->q_vector
->numa_node
;
5553 rx_ring
->rx_buffer_info
= vzalloc_node(size
, ring_node
);
5554 if (!rx_ring
->rx_buffer_info
)
5555 rx_ring
->rx_buffer_info
= vzalloc(size
);
5556 if (!rx_ring
->rx_buffer_info
)
5559 u64_stats_init(&rx_ring
->syncp
);
5561 /* Round up to nearest 4K */
5562 rx_ring
->size
= rx_ring
->count
* sizeof(union ixgbe_adv_rx_desc
);
5563 rx_ring
->size
= ALIGN(rx_ring
->size
, 4096);
5565 set_dev_node(dev
, ring_node
);
5566 rx_ring
->desc
= dma_alloc_coherent(dev
,
5570 set_dev_node(dev
, orig_node
);
5572 rx_ring
->desc
= dma_alloc_coherent(dev
, rx_ring
->size
,
5573 &rx_ring
->dma
, GFP_KERNEL
);
5577 rx_ring
->next_to_clean
= 0;
5578 rx_ring
->next_to_use
= 0;
5582 vfree(rx_ring
->rx_buffer_info
);
5583 rx_ring
->rx_buffer_info
= NULL
;
5584 dev_err(dev
, "Unable to allocate memory for the Rx descriptor ring\n");
5589 * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
5590 * @adapter: board private structure
5592 * If this function returns with an error, then it's possible one or
5593 * more of the rings is populated (while the rest are not). It is the
5594 * callers duty to clean those orphaned rings.
5596 * Return 0 on success, negative on failure
5598 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter
*adapter
)
5602 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
5603 err
= ixgbe_setup_rx_resources(adapter
->rx_ring
[i
]);
5607 e_err(probe
, "Allocation for Rx Queue %u failed\n", i
);
5612 err
= ixgbe_setup_fcoe_ddp_resources(adapter
);
5617 /* rewind the index freeing the rings as we go */
5619 ixgbe_free_rx_resources(adapter
->rx_ring
[i
]);
5624 * ixgbe_free_tx_resources - Free Tx Resources per Queue
5625 * @tx_ring: Tx descriptor ring for a specific queue
5627 * Free all transmit software resources
5629 void ixgbe_free_tx_resources(struct ixgbe_ring
*tx_ring
)
5631 ixgbe_clean_tx_ring(tx_ring
);
5633 vfree(tx_ring
->tx_buffer_info
);
5634 tx_ring
->tx_buffer_info
= NULL
;
5636 /* if not set, then don't free */
5640 dma_free_coherent(tx_ring
->dev
, tx_ring
->size
,
5641 tx_ring
->desc
, tx_ring
->dma
);
5643 tx_ring
->desc
= NULL
;
5647 * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
5648 * @adapter: board private structure
5650 * Free all transmit software resources
5652 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter
*adapter
)
5656 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
5657 if (adapter
->tx_ring
[i
]->desc
)
5658 ixgbe_free_tx_resources(adapter
->tx_ring
[i
]);
5662 * ixgbe_free_rx_resources - Free Rx Resources
5663 * @rx_ring: ring to clean the resources from
5665 * Free all receive software resources
5667 void ixgbe_free_rx_resources(struct ixgbe_ring
*rx_ring
)
5669 ixgbe_clean_rx_ring(rx_ring
);
5671 vfree(rx_ring
->rx_buffer_info
);
5672 rx_ring
->rx_buffer_info
= NULL
;
5674 /* if not set, then don't free */
5678 dma_free_coherent(rx_ring
->dev
, rx_ring
->size
,
5679 rx_ring
->desc
, rx_ring
->dma
);
5681 rx_ring
->desc
= NULL
;
5685 * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
5686 * @adapter: board private structure
5688 * Free all receive software resources
5690 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter
*adapter
)
5695 ixgbe_free_fcoe_ddp_resources(adapter
);
5698 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
5699 if (adapter
->rx_ring
[i
]->desc
)
5700 ixgbe_free_rx_resources(adapter
->rx_ring
[i
]);
5704 * ixgbe_change_mtu - Change the Maximum Transfer Unit
5705 * @netdev: network interface device structure
5706 * @new_mtu: new value for maximum frame size
5708 * Returns 0 on success, negative on failure
5710 static int ixgbe_change_mtu(struct net_device
*netdev
, int new_mtu
)
5712 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
5713 int max_frame
= new_mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
5715 /* MTU < 68 is an error and causes problems on some kernels */
5716 if ((new_mtu
< 68) || (max_frame
> IXGBE_MAX_JUMBO_FRAME_SIZE
))
5720 * For 82599EB we cannot allow legacy VFs to enable their receive
5721 * paths when MTU greater than 1500 is configured. So display a
5722 * warning that legacy VFs will be disabled.
5724 if ((adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) &&
5725 (adapter
->hw
.mac
.type
== ixgbe_mac_82599EB
) &&
5726 (max_frame
> (ETH_FRAME_LEN
+ ETH_FCS_LEN
)))
5727 e_warn(probe
, "Setting MTU > 1500 will disable legacy VFs\n");
5729 e_info(probe
, "changing MTU from %d to %d\n", netdev
->mtu
, new_mtu
);
5731 /* must set new MTU before calling down or up */
5732 netdev
->mtu
= new_mtu
;
5734 if (netif_running(netdev
))
5735 ixgbe_reinit_locked(adapter
);
5741 * ixgbe_open - Called when a network interface is made active
5742 * @netdev: network interface device structure
5744 * Returns 0 on success, negative value on failure
5746 * The open entry point is called when a network interface is made
5747 * active by the system (IFF_UP). At this point all resources needed
5748 * for transmit and receive operations are allocated, the interrupt
5749 * handler is registered with the OS, the watchdog timer is started,
5750 * and the stack is notified that the interface is ready.
5752 static int ixgbe_open(struct net_device
*netdev
)
5754 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
5755 struct ixgbe_hw
*hw
= &adapter
->hw
;
5758 /* disallow open during test */
5759 if (test_bit(__IXGBE_TESTING
, &adapter
->state
))
5762 netif_carrier_off(netdev
);
5764 /* allocate transmit descriptors */
5765 err
= ixgbe_setup_all_tx_resources(adapter
);
5769 /* allocate receive descriptors */
5770 err
= ixgbe_setup_all_rx_resources(adapter
);
5774 ixgbe_configure(adapter
);
5776 err
= ixgbe_request_irq(adapter
);
5780 /* Notify the stack of the actual queue counts. */
5781 if (adapter
->num_rx_pools
> 1)
5782 queues
= adapter
->num_rx_queues_per_pool
;
5784 queues
= adapter
->num_tx_queues
;
5786 err
= netif_set_real_num_tx_queues(netdev
, queues
);
5788 goto err_set_queues
;
5790 if (adapter
->num_rx_pools
> 1 &&
5791 adapter
->num_rx_queues
> IXGBE_MAX_L2A_QUEUES
)
5792 queues
= IXGBE_MAX_L2A_QUEUES
;
5794 queues
= adapter
->num_rx_queues
;
5795 err
= netif_set_real_num_rx_queues(netdev
, queues
);
5797 goto err_set_queues
;
5799 ixgbe_ptp_init(adapter
);
5801 ixgbe_up_complete(adapter
);
5803 ixgbe_clear_vxlan_port(adapter
);
5804 #ifdef CONFIG_IXGBE_VXLAN
5805 vxlan_get_rx_port(netdev
);
5811 ixgbe_free_irq(adapter
);
5813 ixgbe_free_all_rx_resources(adapter
);
5814 if (hw
->phy
.ops
.set_phy_power
&& !adapter
->wol
)
5815 hw
->phy
.ops
.set_phy_power(&adapter
->hw
, false);
5817 ixgbe_free_all_tx_resources(adapter
);
5819 ixgbe_reset(adapter
);
5824 static void ixgbe_close_suspend(struct ixgbe_adapter
*adapter
)
5826 ixgbe_ptp_suspend(adapter
);
5828 if (adapter
->hw
.phy
.ops
.enter_lplu
) {
5829 adapter
->hw
.phy
.reset_disable
= true;
5830 ixgbe_down(adapter
);
5831 adapter
->hw
.phy
.ops
.enter_lplu(&adapter
->hw
);
5832 adapter
->hw
.phy
.reset_disable
= false;
5834 ixgbe_down(adapter
);
5837 ixgbe_free_irq(adapter
);
5839 ixgbe_free_all_tx_resources(adapter
);
5840 ixgbe_free_all_rx_resources(adapter
);
5844 * ixgbe_close - Disables a network interface
5845 * @netdev: network interface device structure
5847 * Returns 0, this is not allowed to fail
5849 * The close entry point is called when an interface is de-activated
5850 * by the OS. The hardware is still under the drivers control, but
5851 * needs to be disabled. A global MAC reset is issued to stop the
5852 * hardware, and all transmit and receive resources are freed.
5854 static int ixgbe_close(struct net_device
*netdev
)
5856 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
5858 ixgbe_ptp_stop(adapter
);
5860 ixgbe_close_suspend(adapter
);
5862 ixgbe_fdir_filter_exit(adapter
);
5864 ixgbe_release_hw_control(adapter
);
5870 static int ixgbe_resume(struct pci_dev
*pdev
)
5872 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
5873 struct net_device
*netdev
= adapter
->netdev
;
5876 adapter
->hw
.hw_addr
= adapter
->io_addr
;
5877 pci_set_power_state(pdev
, PCI_D0
);
5878 pci_restore_state(pdev
);
5880 * pci_restore_state clears dev->state_saved so call
5881 * pci_save_state to restore it.
5883 pci_save_state(pdev
);
5885 err
= pci_enable_device_mem(pdev
);
5887 e_dev_err("Cannot enable PCI device from suspend\n");
5890 smp_mb__before_atomic();
5891 clear_bit(__IXGBE_DISABLED
, &adapter
->state
);
5892 pci_set_master(pdev
);
5894 pci_wake_from_d3(pdev
, false);
5896 ixgbe_reset(adapter
);
5898 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_WUS
, ~0);
5901 err
= ixgbe_init_interrupt_scheme(adapter
);
5902 if (!err
&& netif_running(netdev
))
5903 err
= ixgbe_open(netdev
);
5910 netif_device_attach(netdev
);
5914 #endif /* CONFIG_PM */
5916 static int __ixgbe_shutdown(struct pci_dev
*pdev
, bool *enable_wake
)
5918 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
5919 struct net_device
*netdev
= adapter
->netdev
;
5920 struct ixgbe_hw
*hw
= &adapter
->hw
;
5922 u32 wufc
= adapter
->wol
;
5927 netif_device_detach(netdev
);
5930 if (netif_running(netdev
))
5931 ixgbe_close_suspend(adapter
);
5934 ixgbe_clear_interrupt_scheme(adapter
);
5937 retval
= pci_save_state(pdev
);
5942 if (hw
->mac
.ops
.stop_link_on_d3
)
5943 hw
->mac
.ops
.stop_link_on_d3(hw
);
5946 ixgbe_set_rx_mode(netdev
);
5948 /* enable the optics for 82599 SFP+ fiber as we can WoL */
5949 if (hw
->mac
.ops
.enable_tx_laser
)
5950 hw
->mac
.ops
.enable_tx_laser(hw
);
5952 /* turn on all-multi mode if wake on multicast is enabled */
5953 if (wufc
& IXGBE_WUFC_MC
) {
5954 fctrl
= IXGBE_READ_REG(hw
, IXGBE_FCTRL
);
5955 fctrl
|= IXGBE_FCTRL_MPE
;
5956 IXGBE_WRITE_REG(hw
, IXGBE_FCTRL
, fctrl
);
5959 ctrl
= IXGBE_READ_REG(hw
, IXGBE_CTRL
);
5960 ctrl
|= IXGBE_CTRL_GIO_DIS
;
5961 IXGBE_WRITE_REG(hw
, IXGBE_CTRL
, ctrl
);
5963 IXGBE_WRITE_REG(hw
, IXGBE_WUFC
, wufc
);
5965 IXGBE_WRITE_REG(hw
, IXGBE_WUC
, 0);
5966 IXGBE_WRITE_REG(hw
, IXGBE_WUFC
, 0);
5969 switch (hw
->mac
.type
) {
5970 case ixgbe_mac_82598EB
:
5971 pci_wake_from_d3(pdev
, false);
5973 case ixgbe_mac_82599EB
:
5974 case ixgbe_mac_X540
:
5975 case ixgbe_mac_X550
:
5976 case ixgbe_mac_X550EM_x
:
5977 pci_wake_from_d3(pdev
, !!wufc
);
5983 *enable_wake
= !!wufc
;
5984 if (hw
->phy
.ops
.set_phy_power
&& !*enable_wake
)
5985 hw
->phy
.ops
.set_phy_power(hw
, false);
5987 ixgbe_release_hw_control(adapter
);
5989 if (!test_and_set_bit(__IXGBE_DISABLED
, &adapter
->state
))
5990 pci_disable_device(pdev
);
5996 static int ixgbe_suspend(struct pci_dev
*pdev
, pm_message_t state
)
6001 retval
= __ixgbe_shutdown(pdev
, &wake
);
6006 pci_prepare_to_sleep(pdev
);
6008 pci_wake_from_d3(pdev
, false);
6009 pci_set_power_state(pdev
, PCI_D3hot
);
6014 #endif /* CONFIG_PM */
6016 static void ixgbe_shutdown(struct pci_dev
*pdev
)
6020 __ixgbe_shutdown(pdev
, &wake
);
6022 if (system_state
== SYSTEM_POWER_OFF
) {
6023 pci_wake_from_d3(pdev
, wake
);
6024 pci_set_power_state(pdev
, PCI_D3hot
);
6029 * ixgbe_update_stats - Update the board statistics counters.
6030 * @adapter: board private structure
6032 void ixgbe_update_stats(struct ixgbe_adapter
*adapter
)
6034 struct net_device
*netdev
= adapter
->netdev
;
6035 struct ixgbe_hw
*hw
= &adapter
->hw
;
6036 struct ixgbe_hw_stats
*hwstats
= &adapter
->stats
;
6038 u32 i
, missed_rx
= 0, mpc
, bprc
, lxon
, lxoff
, xon_off_tot
;
6039 u64 non_eop_descs
= 0, restart_queue
= 0, tx_busy
= 0;
6040 u64 alloc_rx_page_failed
= 0, alloc_rx_buff_failed
= 0;
6041 u64 bytes
= 0, packets
= 0, hw_csum_rx_error
= 0;
6043 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
6044 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
6047 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
) {
6050 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
6051 rsc_count
+= adapter
->rx_ring
[i
]->rx_stats
.rsc_count
;
6052 rsc_flush
+= adapter
->rx_ring
[i
]->rx_stats
.rsc_flush
;
6054 adapter
->rsc_total_count
= rsc_count
;
6055 adapter
->rsc_total_flush
= rsc_flush
;
6058 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
6059 struct ixgbe_ring
*rx_ring
= adapter
->rx_ring
[i
];
6060 non_eop_descs
+= rx_ring
->rx_stats
.non_eop_descs
;
6061 alloc_rx_page_failed
+= rx_ring
->rx_stats
.alloc_rx_page_failed
;
6062 alloc_rx_buff_failed
+= rx_ring
->rx_stats
.alloc_rx_buff_failed
;
6063 hw_csum_rx_error
+= rx_ring
->rx_stats
.csum_err
;
6064 bytes
+= rx_ring
->stats
.bytes
;
6065 packets
+= rx_ring
->stats
.packets
;
6067 adapter
->non_eop_descs
= non_eop_descs
;
6068 adapter
->alloc_rx_page_failed
= alloc_rx_page_failed
;
6069 adapter
->alloc_rx_buff_failed
= alloc_rx_buff_failed
;
6070 adapter
->hw_csum_rx_error
= hw_csum_rx_error
;
6071 netdev
->stats
.rx_bytes
= bytes
;
6072 netdev
->stats
.rx_packets
= packets
;
6076 /* gather some stats to the adapter struct that are per queue */
6077 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
6078 struct ixgbe_ring
*tx_ring
= adapter
->tx_ring
[i
];
6079 restart_queue
+= tx_ring
->tx_stats
.restart_queue
;
6080 tx_busy
+= tx_ring
->tx_stats
.tx_busy
;
6081 bytes
+= tx_ring
->stats
.bytes
;
6082 packets
+= tx_ring
->stats
.packets
;
6084 adapter
->restart_queue
= restart_queue
;
6085 adapter
->tx_busy
= tx_busy
;
6086 netdev
->stats
.tx_bytes
= bytes
;
6087 netdev
->stats
.tx_packets
= packets
;
6089 hwstats
->crcerrs
+= IXGBE_READ_REG(hw
, IXGBE_CRCERRS
);
6091 /* 8 register reads */
6092 for (i
= 0; i
< 8; i
++) {
6093 /* for packet buffers not used, the register should read 0 */
6094 mpc
= IXGBE_READ_REG(hw
, IXGBE_MPC(i
));
6096 hwstats
->mpc
[i
] += mpc
;
6097 total_mpc
+= hwstats
->mpc
[i
];
6098 hwstats
->pxontxc
[i
] += IXGBE_READ_REG(hw
, IXGBE_PXONTXC(i
));
6099 hwstats
->pxofftxc
[i
] += IXGBE_READ_REG(hw
, IXGBE_PXOFFTXC(i
));
6100 switch (hw
->mac
.type
) {
6101 case ixgbe_mac_82598EB
:
6102 hwstats
->rnbc
[i
] += IXGBE_READ_REG(hw
, IXGBE_RNBC(i
));
6103 hwstats
->qbtc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBTC(i
));
6104 hwstats
->qbrc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBRC(i
));
6105 hwstats
->pxonrxc
[i
] +=
6106 IXGBE_READ_REG(hw
, IXGBE_PXONRXC(i
));
6108 case ixgbe_mac_82599EB
:
6109 case ixgbe_mac_X540
:
6110 case ixgbe_mac_X550
:
6111 case ixgbe_mac_X550EM_x
:
6112 hwstats
->pxonrxc
[i
] +=
6113 IXGBE_READ_REG(hw
, IXGBE_PXONRXCNT(i
));
6120 /*16 register reads */
6121 for (i
= 0; i
< 16; i
++) {
6122 hwstats
->qptc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QPTC(i
));
6123 hwstats
->qprc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QPRC(i
));
6124 if ((hw
->mac
.type
== ixgbe_mac_82599EB
) ||
6125 (hw
->mac
.type
== ixgbe_mac_X540
) ||
6126 (hw
->mac
.type
== ixgbe_mac_X550
) ||
6127 (hw
->mac
.type
== ixgbe_mac_X550EM_x
)) {
6128 hwstats
->qbtc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBTC_L(i
));
6129 IXGBE_READ_REG(hw
, IXGBE_QBTC_H(i
)); /* to clear */
6130 hwstats
->qbrc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBRC_L(i
));
6131 IXGBE_READ_REG(hw
, IXGBE_QBRC_H(i
)); /* to clear */
6135 hwstats
->gprc
+= IXGBE_READ_REG(hw
, IXGBE_GPRC
);
6136 /* work around hardware counting issue */
6137 hwstats
->gprc
-= missed_rx
;
6139 ixgbe_update_xoff_received(adapter
);
6141 /* 82598 hardware only has a 32 bit counter in the high register */
6142 switch (hw
->mac
.type
) {
6143 case ixgbe_mac_82598EB
:
6144 hwstats
->lxonrxc
+= IXGBE_READ_REG(hw
, IXGBE_LXONRXC
);
6145 hwstats
->gorc
+= IXGBE_READ_REG(hw
, IXGBE_GORCH
);
6146 hwstats
->gotc
+= IXGBE_READ_REG(hw
, IXGBE_GOTCH
);
6147 hwstats
->tor
+= IXGBE_READ_REG(hw
, IXGBE_TORH
);
6149 case ixgbe_mac_X540
:
6150 case ixgbe_mac_X550
:
6151 case ixgbe_mac_X550EM_x
:
6152 /* OS2BMC stats are X540 and later */
6153 hwstats
->o2bgptc
+= IXGBE_READ_REG(hw
, IXGBE_O2BGPTC
);
6154 hwstats
->o2bspc
+= IXGBE_READ_REG(hw
, IXGBE_O2BSPC
);
6155 hwstats
->b2ospc
+= IXGBE_READ_REG(hw
, IXGBE_B2OSPC
);
6156 hwstats
->b2ogprc
+= IXGBE_READ_REG(hw
, IXGBE_B2OGPRC
);
6157 case ixgbe_mac_82599EB
:
6158 for (i
= 0; i
< 16; i
++)
6159 adapter
->hw_rx_no_dma_resources
+=
6160 IXGBE_READ_REG(hw
, IXGBE_QPRDC(i
));
6161 hwstats
->gorc
+= IXGBE_READ_REG(hw
, IXGBE_GORCL
);
6162 IXGBE_READ_REG(hw
, IXGBE_GORCH
); /* to clear */
6163 hwstats
->gotc
+= IXGBE_READ_REG(hw
, IXGBE_GOTCL
);
6164 IXGBE_READ_REG(hw
, IXGBE_GOTCH
); /* to clear */
6165 hwstats
->tor
+= IXGBE_READ_REG(hw
, IXGBE_TORL
);
6166 IXGBE_READ_REG(hw
, IXGBE_TORH
); /* to clear */
6167 hwstats
->lxonrxc
+= IXGBE_READ_REG(hw
, IXGBE_LXONRXCNT
);
6168 hwstats
->fdirmatch
+= IXGBE_READ_REG(hw
, IXGBE_FDIRMATCH
);
6169 hwstats
->fdirmiss
+= IXGBE_READ_REG(hw
, IXGBE_FDIRMISS
);
6171 hwstats
->fccrc
+= IXGBE_READ_REG(hw
, IXGBE_FCCRC
);
6172 hwstats
->fcoerpdc
+= IXGBE_READ_REG(hw
, IXGBE_FCOERPDC
);
6173 hwstats
->fcoeprc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEPRC
);
6174 hwstats
->fcoeptc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEPTC
);
6175 hwstats
->fcoedwrc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEDWRC
);
6176 hwstats
->fcoedwtc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEDWTC
);
6177 /* Add up per cpu counters for total ddp aloc fail */
6178 if (adapter
->fcoe
.ddp_pool
) {
6179 struct ixgbe_fcoe
*fcoe
= &adapter
->fcoe
;
6180 struct ixgbe_fcoe_ddp_pool
*ddp_pool
;
6182 u64 noddp
= 0, noddp_ext_buff
= 0;
6183 for_each_possible_cpu(cpu
) {
6184 ddp_pool
= per_cpu_ptr(fcoe
->ddp_pool
, cpu
);
6185 noddp
+= ddp_pool
->noddp
;
6186 noddp_ext_buff
+= ddp_pool
->noddp_ext_buff
;
6188 hwstats
->fcoe_noddp
= noddp
;
6189 hwstats
->fcoe_noddp_ext_buff
= noddp_ext_buff
;
6191 #endif /* IXGBE_FCOE */
6196 bprc
= IXGBE_READ_REG(hw
, IXGBE_BPRC
);
6197 hwstats
->bprc
+= bprc
;
6198 hwstats
->mprc
+= IXGBE_READ_REG(hw
, IXGBE_MPRC
);
6199 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
6200 hwstats
->mprc
-= bprc
;
6201 hwstats
->roc
+= IXGBE_READ_REG(hw
, IXGBE_ROC
);
6202 hwstats
->prc64
+= IXGBE_READ_REG(hw
, IXGBE_PRC64
);
6203 hwstats
->prc127
+= IXGBE_READ_REG(hw
, IXGBE_PRC127
);
6204 hwstats
->prc255
+= IXGBE_READ_REG(hw
, IXGBE_PRC255
);
6205 hwstats
->prc511
+= IXGBE_READ_REG(hw
, IXGBE_PRC511
);
6206 hwstats
->prc1023
+= IXGBE_READ_REG(hw
, IXGBE_PRC1023
);
6207 hwstats
->prc1522
+= IXGBE_READ_REG(hw
, IXGBE_PRC1522
);
6208 hwstats
->rlec
+= IXGBE_READ_REG(hw
, IXGBE_RLEC
);
6209 lxon
= IXGBE_READ_REG(hw
, IXGBE_LXONTXC
);
6210 hwstats
->lxontxc
+= lxon
;
6211 lxoff
= IXGBE_READ_REG(hw
, IXGBE_LXOFFTXC
);
6212 hwstats
->lxofftxc
+= lxoff
;
6213 hwstats
->gptc
+= IXGBE_READ_REG(hw
, IXGBE_GPTC
);
6214 hwstats
->mptc
+= IXGBE_READ_REG(hw
, IXGBE_MPTC
);
6216 * 82598 errata - tx of flow control packets is included in tx counters
6218 xon_off_tot
= lxon
+ lxoff
;
6219 hwstats
->gptc
-= xon_off_tot
;
6220 hwstats
->mptc
-= xon_off_tot
;
6221 hwstats
->gotc
-= (xon_off_tot
* (ETH_ZLEN
+ ETH_FCS_LEN
));
6222 hwstats
->ruc
+= IXGBE_READ_REG(hw
, IXGBE_RUC
);
6223 hwstats
->rfc
+= IXGBE_READ_REG(hw
, IXGBE_RFC
);
6224 hwstats
->rjc
+= IXGBE_READ_REG(hw
, IXGBE_RJC
);
6225 hwstats
->tpr
+= IXGBE_READ_REG(hw
, IXGBE_TPR
);
6226 hwstats
->ptc64
+= IXGBE_READ_REG(hw
, IXGBE_PTC64
);
6227 hwstats
->ptc64
-= xon_off_tot
;
6228 hwstats
->ptc127
+= IXGBE_READ_REG(hw
, IXGBE_PTC127
);
6229 hwstats
->ptc255
+= IXGBE_READ_REG(hw
, IXGBE_PTC255
);
6230 hwstats
->ptc511
+= IXGBE_READ_REG(hw
, IXGBE_PTC511
);
6231 hwstats
->ptc1023
+= IXGBE_READ_REG(hw
, IXGBE_PTC1023
);
6232 hwstats
->ptc1522
+= IXGBE_READ_REG(hw
, IXGBE_PTC1522
);
6233 hwstats
->bptc
+= IXGBE_READ_REG(hw
, IXGBE_BPTC
);
6235 /* Fill out the OS statistics structure */
6236 netdev
->stats
.multicast
= hwstats
->mprc
;
6239 netdev
->stats
.rx_errors
= hwstats
->crcerrs
+ hwstats
->rlec
;
6240 netdev
->stats
.rx_dropped
= 0;
6241 netdev
->stats
.rx_length_errors
= hwstats
->rlec
;
6242 netdev
->stats
.rx_crc_errors
= hwstats
->crcerrs
;
6243 netdev
->stats
.rx_missed_errors
= total_mpc
;
6247 * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
6248 * @adapter: pointer to the device adapter structure
6250 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter
*adapter
)
6252 struct ixgbe_hw
*hw
= &adapter
->hw
;
6255 if (!(adapter
->flags2
& IXGBE_FLAG2_FDIR_REQUIRES_REINIT
))
6258 adapter
->flags2
&= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT
;
6260 /* if interface is down do nothing */
6261 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
6264 /* do nothing if we are not using signature filters */
6265 if (!(adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
))
6268 adapter
->fdir_overflow
++;
6270 if (ixgbe_reinit_fdir_tables_82599(hw
) == 0) {
6271 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
6272 set_bit(__IXGBE_TX_FDIR_INIT_DONE
,
6273 &(adapter
->tx_ring
[i
]->state
));
6274 /* re-enable flow director interrupts */
6275 IXGBE_WRITE_REG(hw
, IXGBE_EIMS
, IXGBE_EIMS_FLOW_DIR
);
6277 e_err(probe
, "failed to finish FDIR re-initialization, "
6278 "ignored adding FDIR ATR filters\n");
6283 * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
6284 * @adapter: pointer to the device adapter structure
6286 * This function serves two purposes. First it strobes the interrupt lines
6287 * in order to make certain interrupts are occurring. Secondly it sets the
6288 * bits needed to check for TX hangs. As a result we should immediately
6289 * determine if a hang has occurred.
6291 static void ixgbe_check_hang_subtask(struct ixgbe_adapter
*adapter
)
6293 struct ixgbe_hw
*hw
= &adapter
->hw
;
6297 /* If we're down, removing or resetting, just bail */
6298 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
6299 test_bit(__IXGBE_REMOVING
, &adapter
->state
) ||
6300 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
6303 /* Force detection of hung controller */
6304 if (netif_carrier_ok(adapter
->netdev
)) {
6305 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
6306 set_check_for_tx_hang(adapter
->tx_ring
[i
]);
6309 if (!(adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)) {
6311 * for legacy and MSI interrupts don't set any bits
6312 * that are enabled for EIAM, because this operation
6313 * would set *both* EIMS and EICS for any bit in EIAM
6315 IXGBE_WRITE_REG(hw
, IXGBE_EICS
,
6316 (IXGBE_EICS_TCP_TIMER
| IXGBE_EICS_OTHER
));
6318 /* get one bit for every active tx/rx interrupt vector */
6319 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
6320 struct ixgbe_q_vector
*qv
= adapter
->q_vector
[i
];
6321 if (qv
->rx
.ring
|| qv
->tx
.ring
)
6322 eics
|= ((u64
)1 << i
);
6326 /* Cause software interrupt to ensure rings are cleaned */
6327 ixgbe_irq_rearm_queues(adapter
, eics
);
6331 * ixgbe_watchdog_update_link - update the link status
6332 * @adapter: pointer to the device adapter structure
6333 * @link_speed: pointer to a u32 to store the link_speed
6335 static void ixgbe_watchdog_update_link(struct ixgbe_adapter
*adapter
)
6337 struct ixgbe_hw
*hw
= &adapter
->hw
;
6338 u32 link_speed
= adapter
->link_speed
;
6339 bool link_up
= adapter
->link_up
;
6340 bool pfc_en
= adapter
->dcb_cfg
.pfc_mode_enable
;
6342 if (!(adapter
->flags
& IXGBE_FLAG_NEED_LINK_UPDATE
))
6345 if (hw
->mac
.ops
.check_link
) {
6346 hw
->mac
.ops
.check_link(hw
, &link_speed
, &link_up
, false);
6348 /* always assume link is up, if no check link function */
6349 link_speed
= IXGBE_LINK_SPEED_10GB_FULL
;
6353 if (adapter
->ixgbe_ieee_pfc
)
6354 pfc_en
|= !!(adapter
->ixgbe_ieee_pfc
->pfc_en
);
6356 if (link_up
&& !((adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
) && pfc_en
)) {
6357 hw
->mac
.ops
.fc_enable(hw
);
6358 ixgbe_set_rx_drop_en(adapter
);
6362 time_after(jiffies
, (adapter
->link_check_timeout
+
6363 IXGBE_TRY_LINK_TIMEOUT
))) {
6364 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_UPDATE
;
6365 IXGBE_WRITE_REG(hw
, IXGBE_EIMS
, IXGBE_EIMC_LSC
);
6366 IXGBE_WRITE_FLUSH(hw
);
6369 adapter
->link_up
= link_up
;
6370 adapter
->link_speed
= link_speed
;
6373 static void ixgbe_update_default_up(struct ixgbe_adapter
*adapter
)
6375 #ifdef CONFIG_IXGBE_DCB
6376 struct net_device
*netdev
= adapter
->netdev
;
6377 struct dcb_app app
= {
6378 .selector
= IEEE_8021QAZ_APP_SEL_ETHERTYPE
,
6383 if (adapter
->dcbx_cap
& DCB_CAP_DCBX_VER_IEEE
)
6384 up
= dcb_ieee_getapp_mask(netdev
, &app
);
6386 adapter
->default_up
= (up
> 1) ? (ffs(up
) - 1) : 0;
6391 * ixgbe_watchdog_link_is_up - update netif_carrier status and
6392 * print link up message
6393 * @adapter: pointer to the device adapter structure
6395 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter
*adapter
)
6397 struct net_device
*netdev
= adapter
->netdev
;
6398 struct ixgbe_hw
*hw
= &adapter
->hw
;
6399 struct net_device
*upper
;
6400 struct list_head
*iter
;
6401 u32 link_speed
= adapter
->link_speed
;
6402 const char *speed_str
;
6403 bool flow_rx
, flow_tx
;
6405 /* only continue if link was previously down */
6406 if (netif_carrier_ok(netdev
))
6409 adapter
->flags2
&= ~IXGBE_FLAG2_SEARCH_FOR_SFP
;
6411 switch (hw
->mac
.type
) {
6412 case ixgbe_mac_82598EB
: {
6413 u32 frctl
= IXGBE_READ_REG(hw
, IXGBE_FCTRL
);
6414 u32 rmcs
= IXGBE_READ_REG(hw
, IXGBE_RMCS
);
6415 flow_rx
= !!(frctl
& IXGBE_FCTRL_RFCE
);
6416 flow_tx
= !!(rmcs
& IXGBE_RMCS_TFCE_802_3X
);
6419 case ixgbe_mac_X540
:
6420 case ixgbe_mac_X550
:
6421 case ixgbe_mac_X550EM_x
:
6422 case ixgbe_mac_82599EB
: {
6423 u32 mflcn
= IXGBE_READ_REG(hw
, IXGBE_MFLCN
);
6424 u32 fccfg
= IXGBE_READ_REG(hw
, IXGBE_FCCFG
);
6425 flow_rx
= !!(mflcn
& IXGBE_MFLCN_RFCE
);
6426 flow_tx
= !!(fccfg
& IXGBE_FCCFG_TFCE_802_3X
);
6435 adapter
->last_rx_ptp_check
= jiffies
;
6437 if (test_bit(__IXGBE_PTP_RUNNING
, &adapter
->state
))
6438 ixgbe_ptp_start_cyclecounter(adapter
);
6440 switch (link_speed
) {
6441 case IXGBE_LINK_SPEED_10GB_FULL
:
6442 speed_str
= "10 Gbps";
6444 case IXGBE_LINK_SPEED_2_5GB_FULL
:
6445 speed_str
= "2.5 Gbps";
6447 case IXGBE_LINK_SPEED_1GB_FULL
:
6448 speed_str
= "1 Gbps";
6450 case IXGBE_LINK_SPEED_100_FULL
:
6451 speed_str
= "100 Mbps";
6454 speed_str
= "unknown speed";
6457 e_info(drv
, "NIC Link is Up %s, Flow Control: %s\n", speed_str
,
6458 ((flow_rx
&& flow_tx
) ? "RX/TX" :
6460 (flow_tx
? "TX" : "None"))));
6462 netif_carrier_on(netdev
);
6463 ixgbe_check_vf_rate_limit(adapter
);
6465 /* enable transmits */
6466 netif_tx_wake_all_queues(adapter
->netdev
);
6468 /* enable any upper devices */
6470 netdev_for_each_all_upper_dev_rcu(adapter
->netdev
, upper
, iter
) {
6471 if (netif_is_macvlan(upper
)) {
6472 struct macvlan_dev
*vlan
= netdev_priv(upper
);
6475 netif_tx_wake_all_queues(upper
);
6480 /* update the default user priority for VFs */
6481 ixgbe_update_default_up(adapter
);
6483 /* ping all the active vfs to let them know link has changed */
6484 ixgbe_ping_all_vfs(adapter
);
6488 * ixgbe_watchdog_link_is_down - update netif_carrier status and
6489 * print link down message
6490 * @adapter: pointer to the adapter structure
6492 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter
*adapter
)
6494 struct net_device
*netdev
= adapter
->netdev
;
6495 struct ixgbe_hw
*hw
= &adapter
->hw
;
6497 adapter
->link_up
= false;
6498 adapter
->link_speed
= 0;
6500 /* only continue if link was up previously */
6501 if (!netif_carrier_ok(netdev
))
6504 /* poll for SFP+ cable when link is down */
6505 if (ixgbe_is_sfp(hw
) && hw
->mac
.type
== ixgbe_mac_82598EB
)
6506 adapter
->flags2
|= IXGBE_FLAG2_SEARCH_FOR_SFP
;
6508 if (test_bit(__IXGBE_PTP_RUNNING
, &adapter
->state
))
6509 ixgbe_ptp_start_cyclecounter(adapter
);
6511 e_info(drv
, "NIC Link is Down\n");
6512 netif_carrier_off(netdev
);
6514 /* ping all the active vfs to let them know link has changed */
6515 ixgbe_ping_all_vfs(adapter
);
6518 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter
*adapter
)
6522 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
6523 struct ixgbe_ring
*tx_ring
= adapter
->tx_ring
[i
];
6525 if (tx_ring
->next_to_use
!= tx_ring
->next_to_clean
)
6532 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter
*adapter
)
6534 struct ixgbe_hw
*hw
= &adapter
->hw
;
6535 struct ixgbe_ring_feature
*vmdq
= &adapter
->ring_feature
[RING_F_VMDQ
];
6536 u32 q_per_pool
= __ALIGN_MASK(1, ~vmdq
->mask
);
6540 if (!adapter
->num_vfs
)
6543 /* resetting the PF is only needed for MAC before X550 */
6544 if (hw
->mac
.type
>= ixgbe_mac_X550
)
6547 for (i
= 0; i
< adapter
->num_vfs
; i
++) {
6548 for (j
= 0; j
< q_per_pool
; j
++) {
6551 h
= IXGBE_READ_REG(hw
, IXGBE_PVFTDHN(q_per_pool
, i
, j
));
6552 t
= IXGBE_READ_REG(hw
, IXGBE_PVFTDTN(q_per_pool
, i
, j
));
6563 * ixgbe_watchdog_flush_tx - flush queues on link down
6564 * @adapter: pointer to the device adapter structure
6566 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter
*adapter
)
6568 if (!netif_carrier_ok(adapter
->netdev
)) {
6569 if (ixgbe_ring_tx_pending(adapter
) ||
6570 ixgbe_vf_tx_pending(adapter
)) {
6571 /* We've lost link, so the controller stops DMA,
6572 * but we've got queued Tx work that's never going
6573 * to get done, so reset controller to flush Tx.
6574 * (Do the reset outside of interrupt context).
6576 e_warn(drv
, "initiating reset to clear Tx work after link loss\n");
6577 adapter
->flags2
|= IXGBE_FLAG2_RESET_REQUESTED
;
6582 #ifdef CONFIG_PCI_IOV
6583 static inline void ixgbe_issue_vf_flr(struct ixgbe_adapter
*adapter
,
6584 struct pci_dev
*vfdev
)
6586 if (!pci_wait_for_pending_transaction(vfdev
))
6587 e_dev_warn("Issuing VFLR with pending transactions\n");
6589 e_dev_err("Issuing VFLR for VF %s\n", pci_name(vfdev
));
6590 pcie_capability_set_word(vfdev
, PCI_EXP_DEVCTL
, PCI_EXP_DEVCTL_BCR_FLR
);
6595 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter
*adapter
)
6597 struct ixgbe_hw
*hw
= &adapter
->hw
;
6598 struct pci_dev
*pdev
= adapter
->pdev
;
6599 struct pci_dev
*vfdev
;
6602 unsigned short vf_id
;
6604 if (!(netif_carrier_ok(adapter
->netdev
)))
6607 gpc
= IXGBE_READ_REG(hw
, IXGBE_TXDGPC
);
6608 if (gpc
) /* If incrementing then no need for the check below */
6610 /* Check to see if a bad DMA write target from an errant or
6611 * malicious VF has caused a PCIe error. If so then we can
6612 * issue a VFLR to the offending VF(s) and then resume without
6613 * requesting a full slot reset.
6619 pos
= pci_find_ext_capability(pdev
, PCI_EXT_CAP_ID_SRIOV
);
6623 /* get the device ID for the VF */
6624 pci_read_config_word(pdev
, pos
+ PCI_SRIOV_VF_DID
, &vf_id
);
6626 /* check status reg for all VFs owned by this PF */
6627 vfdev
= pci_get_device(pdev
->vendor
, vf_id
, NULL
);
6629 if (vfdev
->is_virtfn
&& (vfdev
->physfn
== pdev
)) {
6632 pci_read_config_word(vfdev
, PCI_STATUS
, &status_reg
);
6633 if (status_reg
& PCI_STATUS_REC_MASTER_ABORT
)
6635 ixgbe_issue_vf_flr(adapter
, vfdev
);
6638 vfdev
= pci_get_device(pdev
->vendor
, vf_id
, vfdev
);
6642 static void ixgbe_spoof_check(struct ixgbe_adapter
*adapter
)
6646 /* Do not perform spoof check for 82598 or if not in IOV mode */
6647 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
||
6648 adapter
->num_vfs
== 0)
6651 ssvpc
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_SSVPC
);
6654 * ssvpc register is cleared on read, if zero then no
6655 * spoofed packets in the last interval.
6660 e_warn(drv
, "%u Spoofed packets detected\n", ssvpc
);
6663 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused
*adapter
)
6668 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused
*adapter
)
6671 #endif /* CONFIG_PCI_IOV */
6675 * ixgbe_watchdog_subtask - check and bring link up
6676 * @adapter: pointer to the device adapter structure
6678 static void ixgbe_watchdog_subtask(struct ixgbe_adapter
*adapter
)
6680 /* if interface is down, removing or resetting, do nothing */
6681 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
6682 test_bit(__IXGBE_REMOVING
, &adapter
->state
) ||
6683 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
6686 ixgbe_watchdog_update_link(adapter
);
6688 if (adapter
->link_up
)
6689 ixgbe_watchdog_link_is_up(adapter
);
6691 ixgbe_watchdog_link_is_down(adapter
);
6693 ixgbe_check_for_bad_vf(adapter
);
6694 ixgbe_spoof_check(adapter
);
6695 ixgbe_update_stats(adapter
);
6697 ixgbe_watchdog_flush_tx(adapter
);
6701 * ixgbe_sfp_detection_subtask - poll for SFP+ cable
6702 * @adapter: the ixgbe adapter structure
6704 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter
*adapter
)
6706 struct ixgbe_hw
*hw
= &adapter
->hw
;
6709 /* not searching for SFP so there is nothing to do here */
6710 if (!(adapter
->flags2
& IXGBE_FLAG2_SEARCH_FOR_SFP
) &&
6711 !(adapter
->flags2
& IXGBE_FLAG2_SFP_NEEDS_RESET
))
6714 if (adapter
->sfp_poll_time
&&
6715 time_after(adapter
->sfp_poll_time
, jiffies
))
6716 return; /* If not yet time to poll for SFP */
6718 /* someone else is in init, wait until next service event */
6719 if (test_and_set_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
))
6722 adapter
->sfp_poll_time
= jiffies
+ IXGBE_SFP_POLL_JIFFIES
- 1;
6724 err
= hw
->phy
.ops
.identify_sfp(hw
);
6725 if (err
== IXGBE_ERR_SFP_NOT_SUPPORTED
)
6728 if (err
== IXGBE_ERR_SFP_NOT_PRESENT
) {
6729 /* If no cable is present, then we need to reset
6730 * the next time we find a good cable. */
6731 adapter
->flags2
|= IXGBE_FLAG2_SFP_NEEDS_RESET
;
6738 /* exit if reset not needed */
6739 if (!(adapter
->flags2
& IXGBE_FLAG2_SFP_NEEDS_RESET
))
6742 adapter
->flags2
&= ~IXGBE_FLAG2_SFP_NEEDS_RESET
;
6745 * A module may be identified correctly, but the EEPROM may not have
6746 * support for that module. setup_sfp() will fail in that case, so
6747 * we should not allow that module to load.
6749 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
6750 err
= hw
->phy
.ops
.reset(hw
);
6752 err
= hw
->mac
.ops
.setup_sfp(hw
);
6754 if (err
== IXGBE_ERR_SFP_NOT_SUPPORTED
)
6757 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_CONFIG
;
6758 e_info(probe
, "detected SFP+: %d\n", hw
->phy
.sfp_type
);
6761 clear_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
);
6763 if ((err
== IXGBE_ERR_SFP_NOT_SUPPORTED
) &&
6764 (adapter
->netdev
->reg_state
== NETREG_REGISTERED
)) {
6765 e_dev_err("failed to initialize because an unsupported "
6766 "SFP+ module type was detected.\n");
6767 e_dev_err("Reload the driver after installing a "
6768 "supported module.\n");
6769 unregister_netdev(adapter
->netdev
);
6774 * ixgbe_sfp_link_config_subtask - set up link SFP after module install
6775 * @adapter: the ixgbe adapter structure
6777 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter
*adapter
)
6779 struct ixgbe_hw
*hw
= &adapter
->hw
;
6781 bool autoneg
= false;
6783 if (!(adapter
->flags
& IXGBE_FLAG_NEED_LINK_CONFIG
))
6786 /* someone else is in init, wait until next service event */
6787 if (test_and_set_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
))
6790 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_CONFIG
;
6792 speed
= hw
->phy
.autoneg_advertised
;
6793 if ((!speed
) && (hw
->mac
.ops
.get_link_capabilities
)) {
6794 hw
->mac
.ops
.get_link_capabilities(hw
, &speed
, &autoneg
);
6796 /* setup the highest link when no autoneg */
6798 if (speed
& IXGBE_LINK_SPEED_10GB_FULL
)
6799 speed
= IXGBE_LINK_SPEED_10GB_FULL
;
6803 if (hw
->mac
.ops
.setup_link
)
6804 hw
->mac
.ops
.setup_link(hw
, speed
, true);
6806 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_UPDATE
;
6807 adapter
->link_check_timeout
= jiffies
;
6808 clear_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
);
6812 * ixgbe_service_timer - Timer Call-back
6813 * @data: pointer to adapter cast into an unsigned long
6815 static void ixgbe_service_timer(unsigned long data
)
6817 struct ixgbe_adapter
*adapter
= (struct ixgbe_adapter
*)data
;
6818 unsigned long next_event_offset
;
6820 /* poll faster when waiting for link */
6821 if (adapter
->flags
& IXGBE_FLAG_NEED_LINK_UPDATE
)
6822 next_event_offset
= HZ
/ 10;
6824 next_event_offset
= HZ
* 2;
6826 /* Reset the timer */
6827 mod_timer(&adapter
->service_timer
, next_event_offset
+ jiffies
);
6829 ixgbe_service_event_schedule(adapter
);
6832 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter
*adapter
)
6834 struct ixgbe_hw
*hw
= &adapter
->hw
;
6837 if (!(adapter
->flags2
& IXGBE_FLAG2_PHY_INTERRUPT
))
6840 adapter
->flags2
&= ~IXGBE_FLAG2_PHY_INTERRUPT
;
6842 if (!hw
->phy
.ops
.handle_lasi
)
6845 status
= hw
->phy
.ops
.handle_lasi(&adapter
->hw
);
6846 if (status
!= IXGBE_ERR_OVERTEMP
)
6849 e_crit(drv
, "%s\n", ixgbe_overheat_msg
);
6852 static void ixgbe_reset_subtask(struct ixgbe_adapter
*adapter
)
6854 if (!(adapter
->flags2
& IXGBE_FLAG2_RESET_REQUESTED
))
6857 adapter
->flags2
&= ~IXGBE_FLAG2_RESET_REQUESTED
;
6859 /* If we're already down, removing or resetting, just bail */
6860 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
6861 test_bit(__IXGBE_REMOVING
, &adapter
->state
) ||
6862 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
6865 ixgbe_dump(adapter
);
6866 netdev_err(adapter
->netdev
, "Reset adapter\n");
6867 adapter
->tx_timeout_count
++;
6870 ixgbe_reinit_locked(adapter
);
6875 * ixgbe_service_task - manages and runs subtasks
6876 * @work: pointer to work_struct containing our data
6878 static void ixgbe_service_task(struct work_struct
*work
)
6880 struct ixgbe_adapter
*adapter
= container_of(work
,
6881 struct ixgbe_adapter
,
6883 if (ixgbe_removed(adapter
->hw
.hw_addr
)) {
6884 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
6886 ixgbe_down(adapter
);
6889 ixgbe_service_event_complete(adapter
);
6892 #ifdef CONFIG_IXGBE_VXLAN
6893 if (adapter
->flags2
& IXGBE_FLAG2_VXLAN_REREG_NEEDED
) {
6894 adapter
->flags2
&= ~IXGBE_FLAG2_VXLAN_REREG_NEEDED
;
6895 vxlan_get_rx_port(adapter
->netdev
);
6897 #endif /* CONFIG_IXGBE_VXLAN */
6898 ixgbe_reset_subtask(adapter
);
6899 ixgbe_phy_interrupt_subtask(adapter
);
6900 ixgbe_sfp_detection_subtask(adapter
);
6901 ixgbe_sfp_link_config_subtask(adapter
);
6902 ixgbe_check_overtemp_subtask(adapter
);
6903 ixgbe_watchdog_subtask(adapter
);
6904 ixgbe_fdir_reinit_subtask(adapter
);
6905 ixgbe_check_hang_subtask(adapter
);
6907 if (test_bit(__IXGBE_PTP_RUNNING
, &adapter
->state
)) {
6908 ixgbe_ptp_overflow_check(adapter
);
6909 ixgbe_ptp_rx_hang(adapter
);
6912 ixgbe_service_event_complete(adapter
);
6915 static int ixgbe_tso(struct ixgbe_ring
*tx_ring
,
6916 struct ixgbe_tx_buffer
*first
,
6919 struct sk_buff
*skb
= first
->skb
;
6920 u32 vlan_macip_lens
, type_tucmd
;
6921 u32 mss_l4len_idx
, l4len
;
6924 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
6927 if (!skb_is_gso(skb
))
6930 err
= skb_cow_head(skb
, 0);
6934 /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
6935 type_tucmd
= IXGBE_ADVTXD_TUCMD_L4T_TCP
;
6937 if (first
->protocol
== htons(ETH_P_IP
)) {
6938 struct iphdr
*iph
= ip_hdr(skb
);
6941 tcp_hdr(skb
)->check
= ~csum_tcpudp_magic(iph
->saddr
,
6945 type_tucmd
|= IXGBE_ADVTXD_TUCMD_IPV4
;
6946 first
->tx_flags
|= IXGBE_TX_FLAGS_TSO
|
6947 IXGBE_TX_FLAGS_CSUM
|
6948 IXGBE_TX_FLAGS_IPV4
;
6949 } else if (skb_is_gso_v6(skb
)) {
6950 ipv6_hdr(skb
)->payload_len
= 0;
6951 tcp_hdr(skb
)->check
=
6952 ~csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
6953 &ipv6_hdr(skb
)->daddr
,
6955 first
->tx_flags
|= IXGBE_TX_FLAGS_TSO
|
6956 IXGBE_TX_FLAGS_CSUM
;
6959 /* compute header lengths */
6960 l4len
= tcp_hdrlen(skb
);
6961 *hdr_len
= skb_transport_offset(skb
) + l4len
;
6963 /* update gso size and bytecount with header size */
6964 first
->gso_segs
= skb_shinfo(skb
)->gso_segs
;
6965 first
->bytecount
+= (first
->gso_segs
- 1) * *hdr_len
;
6967 /* mss_l4len_id: use 0 as index for TSO */
6968 mss_l4len_idx
= l4len
<< IXGBE_ADVTXD_L4LEN_SHIFT
;
6969 mss_l4len_idx
|= skb_shinfo(skb
)->gso_size
<< IXGBE_ADVTXD_MSS_SHIFT
;
6971 /* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
6972 vlan_macip_lens
= skb_network_header_len(skb
);
6973 vlan_macip_lens
|= skb_network_offset(skb
) << IXGBE_ADVTXD_MACLEN_SHIFT
;
6974 vlan_macip_lens
|= first
->tx_flags
& IXGBE_TX_FLAGS_VLAN_MASK
;
6976 ixgbe_tx_ctxtdesc(tx_ring
, vlan_macip_lens
, 0, type_tucmd
,
6982 static void ixgbe_tx_csum(struct ixgbe_ring
*tx_ring
,
6983 struct ixgbe_tx_buffer
*first
)
6985 struct sk_buff
*skb
= first
->skb
;
6986 u32 vlan_macip_lens
= 0;
6987 u32 mss_l4len_idx
= 0;
6990 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
6991 if (!(first
->tx_flags
& IXGBE_TX_FLAGS_HW_VLAN
) &&
6992 !(first
->tx_flags
& IXGBE_TX_FLAGS_CC
))
6994 vlan_macip_lens
= skb_network_offset(skb
) <<
6995 IXGBE_ADVTXD_MACLEN_SHIFT
;
7000 struct ipv6hdr
*ipv6
;
7004 struct tcphdr
*tcphdr
;
7008 if (skb
->encapsulation
) {
7009 network_hdr
.raw
= skb_inner_network_header(skb
);
7010 transport_hdr
.raw
= skb_inner_transport_header(skb
);
7011 vlan_macip_lens
= skb_inner_network_offset(skb
) <<
7012 IXGBE_ADVTXD_MACLEN_SHIFT
;
7014 network_hdr
.raw
= skb_network_header(skb
);
7015 transport_hdr
.raw
= skb_transport_header(skb
);
7016 vlan_macip_lens
= skb_network_offset(skb
) <<
7017 IXGBE_ADVTXD_MACLEN_SHIFT
;
7020 /* use first 4 bits to determine IP version */
7021 switch (network_hdr
.ipv4
->version
) {
7023 vlan_macip_lens
|= transport_hdr
.raw
- network_hdr
.raw
;
7024 type_tucmd
|= IXGBE_ADVTXD_TUCMD_IPV4
;
7025 l4_hdr
= network_hdr
.ipv4
->protocol
;
7028 vlan_macip_lens
|= transport_hdr
.raw
- network_hdr
.raw
;
7029 l4_hdr
= network_hdr
.ipv6
->nexthdr
;
7032 if (unlikely(net_ratelimit())) {
7033 dev_warn(tx_ring
->dev
,
7034 "partial checksum but version=%d\n",
7035 network_hdr
.ipv4
->version
);
7041 type_tucmd
|= IXGBE_ADVTXD_TUCMD_L4T_TCP
;
7042 mss_l4len_idx
= (transport_hdr
.tcphdr
->doff
* 4) <<
7043 IXGBE_ADVTXD_L4LEN_SHIFT
;
7046 type_tucmd
|= IXGBE_ADVTXD_TUCMD_L4T_SCTP
;
7047 mss_l4len_idx
= sizeof(struct sctphdr
) <<
7048 IXGBE_ADVTXD_L4LEN_SHIFT
;
7051 mss_l4len_idx
= sizeof(struct udphdr
) <<
7052 IXGBE_ADVTXD_L4LEN_SHIFT
;
7055 if (unlikely(net_ratelimit())) {
7056 dev_warn(tx_ring
->dev
,
7057 "partial checksum but l4 proto=%x!\n",
7063 /* update TX checksum flag */
7064 first
->tx_flags
|= IXGBE_TX_FLAGS_CSUM
;
7067 /* vlan_macip_lens: MACLEN, VLAN tag */
7068 vlan_macip_lens
|= first
->tx_flags
& IXGBE_TX_FLAGS_VLAN_MASK
;
7070 ixgbe_tx_ctxtdesc(tx_ring
, vlan_macip_lens
, 0,
7071 type_tucmd
, mss_l4len_idx
);
7074 #define IXGBE_SET_FLAG(_input, _flag, _result) \
7075 ((_flag <= _result) ? \
7076 ((u32)(_input & _flag) * (_result / _flag)) : \
7077 ((u32)(_input & _flag) / (_flag / _result)))
7079 static u32
ixgbe_tx_cmd_type(struct sk_buff
*skb
, u32 tx_flags
)
7081 /* set type for advanced descriptor with frame checksum insertion */
7082 u32 cmd_type
= IXGBE_ADVTXD_DTYP_DATA
|
7083 IXGBE_ADVTXD_DCMD_DEXT
|
7084 IXGBE_ADVTXD_DCMD_IFCS
;
7086 /* set HW vlan bit if vlan is present */
7087 cmd_type
|= IXGBE_SET_FLAG(tx_flags
, IXGBE_TX_FLAGS_HW_VLAN
,
7088 IXGBE_ADVTXD_DCMD_VLE
);
7090 /* set segmentation enable bits for TSO/FSO */
7091 cmd_type
|= IXGBE_SET_FLAG(tx_flags
, IXGBE_TX_FLAGS_TSO
,
7092 IXGBE_ADVTXD_DCMD_TSE
);
7094 /* set timestamp bit if present */
7095 cmd_type
|= IXGBE_SET_FLAG(tx_flags
, IXGBE_TX_FLAGS_TSTAMP
,
7096 IXGBE_ADVTXD_MAC_TSTAMP
);
7098 /* insert frame checksum */
7099 cmd_type
^= IXGBE_SET_FLAG(skb
->no_fcs
, 1, IXGBE_ADVTXD_DCMD_IFCS
);
7104 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc
*tx_desc
,
7105 u32 tx_flags
, unsigned int paylen
)
7107 u32 olinfo_status
= paylen
<< IXGBE_ADVTXD_PAYLEN_SHIFT
;
7109 /* enable L4 checksum for TSO and TX checksum offload */
7110 olinfo_status
|= IXGBE_SET_FLAG(tx_flags
,
7111 IXGBE_TX_FLAGS_CSUM
,
7112 IXGBE_ADVTXD_POPTS_TXSM
);
7114 /* enble IPv4 checksum for TSO */
7115 olinfo_status
|= IXGBE_SET_FLAG(tx_flags
,
7116 IXGBE_TX_FLAGS_IPV4
,
7117 IXGBE_ADVTXD_POPTS_IXSM
);
7120 * Check Context must be set if Tx switch is enabled, which it
7121 * always is for case where virtual functions are running
7123 olinfo_status
|= IXGBE_SET_FLAG(tx_flags
,
7127 tx_desc
->read
.olinfo_status
= cpu_to_le32(olinfo_status
);
7130 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring
*tx_ring
, u16 size
)
7132 netif_stop_subqueue(tx_ring
->netdev
, tx_ring
->queue_index
);
7134 /* Herbert's original patch had:
7135 * smp_mb__after_netif_stop_queue();
7136 * but since that doesn't exist yet, just open code it.
7140 /* We need to check again in a case another CPU has just
7141 * made room available.
7143 if (likely(ixgbe_desc_unused(tx_ring
) < size
))
7146 /* A reprieve! - use start_queue because it doesn't call schedule */
7147 netif_start_subqueue(tx_ring
->netdev
, tx_ring
->queue_index
);
7148 ++tx_ring
->tx_stats
.restart_queue
;
7152 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring
*tx_ring
, u16 size
)
7154 if (likely(ixgbe_desc_unused(tx_ring
) >= size
))
7157 return __ixgbe_maybe_stop_tx(tx_ring
, size
);
7160 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7163 static void ixgbe_tx_map(struct ixgbe_ring
*tx_ring
,
7164 struct ixgbe_tx_buffer
*first
,
7167 struct sk_buff
*skb
= first
->skb
;
7168 struct ixgbe_tx_buffer
*tx_buffer
;
7169 union ixgbe_adv_tx_desc
*tx_desc
;
7170 struct skb_frag_struct
*frag
;
7172 unsigned int data_len
, size
;
7173 u32 tx_flags
= first
->tx_flags
;
7174 u32 cmd_type
= ixgbe_tx_cmd_type(skb
, tx_flags
);
7175 u16 i
= tx_ring
->next_to_use
;
7177 tx_desc
= IXGBE_TX_DESC(tx_ring
, i
);
7179 ixgbe_tx_olinfo_status(tx_desc
, tx_flags
, skb
->len
- hdr_len
);
7181 size
= skb_headlen(skb
);
7182 data_len
= skb
->data_len
;
7185 if (tx_flags
& IXGBE_TX_FLAGS_FCOE
) {
7186 if (data_len
< sizeof(struct fcoe_crc_eof
)) {
7187 size
-= sizeof(struct fcoe_crc_eof
) - data_len
;
7190 data_len
-= sizeof(struct fcoe_crc_eof
);
7195 dma
= dma_map_single(tx_ring
->dev
, skb
->data
, size
, DMA_TO_DEVICE
);
7199 for (frag
= &skb_shinfo(skb
)->frags
[0];; frag
++) {
7200 if (dma_mapping_error(tx_ring
->dev
, dma
))
7203 /* record length, and DMA address */
7204 dma_unmap_len_set(tx_buffer
, len
, size
);
7205 dma_unmap_addr_set(tx_buffer
, dma
, dma
);
7207 tx_desc
->read
.buffer_addr
= cpu_to_le64(dma
);
7209 while (unlikely(size
> IXGBE_MAX_DATA_PER_TXD
)) {
7210 tx_desc
->read
.cmd_type_len
=
7211 cpu_to_le32(cmd_type
^ IXGBE_MAX_DATA_PER_TXD
);
7215 if (i
== tx_ring
->count
) {
7216 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
7219 tx_desc
->read
.olinfo_status
= 0;
7221 dma
+= IXGBE_MAX_DATA_PER_TXD
;
7222 size
-= IXGBE_MAX_DATA_PER_TXD
;
7224 tx_desc
->read
.buffer_addr
= cpu_to_le64(dma
);
7227 if (likely(!data_len
))
7230 tx_desc
->read
.cmd_type_len
= cpu_to_le32(cmd_type
^ size
);
7234 if (i
== tx_ring
->count
) {
7235 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
7238 tx_desc
->read
.olinfo_status
= 0;
7241 size
= min_t(unsigned int, data_len
, skb_frag_size(frag
));
7243 size
= skb_frag_size(frag
);
7247 dma
= skb_frag_dma_map(tx_ring
->dev
, frag
, 0, size
,
7250 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
7253 /* write last descriptor with RS and EOP bits */
7254 cmd_type
|= size
| IXGBE_TXD_CMD
;
7255 tx_desc
->read
.cmd_type_len
= cpu_to_le32(cmd_type
);
7257 netdev_tx_sent_queue(txring_txq(tx_ring
), first
->bytecount
);
7259 /* set the timestamp */
7260 first
->time_stamp
= jiffies
;
7263 * Force memory writes to complete before letting h/w know there
7264 * are new descriptors to fetch. (Only applicable for weak-ordered
7265 * memory model archs, such as IA-64).
7267 * We also need this memory barrier to make certain all of the
7268 * status bits have been updated before next_to_watch is written.
7272 /* set next_to_watch value indicating a packet is present */
7273 first
->next_to_watch
= tx_desc
;
7276 if (i
== tx_ring
->count
)
7279 tx_ring
->next_to_use
= i
;
7281 ixgbe_maybe_stop_tx(tx_ring
, DESC_NEEDED
);
7283 if (netif_xmit_stopped(txring_txq(tx_ring
)) || !skb
->xmit_more
) {
7284 writel(i
, tx_ring
->tail
);
7286 /* we need this if more than one processor can write to our tail
7287 * at a time, it synchronizes IO on IA64/Altix systems
7294 dev_err(tx_ring
->dev
, "TX DMA map failed\n");
7296 /* clear dma mappings for failed tx_buffer_info map */
7298 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
7299 ixgbe_unmap_and_free_tx_resource(tx_ring
, tx_buffer
);
7300 if (tx_buffer
== first
)
7307 tx_ring
->next_to_use
= i
;
7310 static void ixgbe_atr(struct ixgbe_ring
*ring
,
7311 struct ixgbe_tx_buffer
*first
)
7313 struct ixgbe_q_vector
*q_vector
= ring
->q_vector
;
7314 union ixgbe_atr_hash_dword input
= { .dword
= 0 };
7315 union ixgbe_atr_hash_dword common
= { .dword
= 0 };
7317 unsigned char *network
;
7319 struct ipv6hdr
*ipv6
;
7322 struct sk_buff
*skb
;
7323 #ifdef CONFIG_IXGBE_VXLAN
7325 #endif /* CONFIG_IXGBE_VXLAN */
7328 /* if ring doesn't have a interrupt vector, cannot perform ATR */
7332 /* do nothing if sampling is disabled */
7333 if (!ring
->atr_sample_rate
)
7338 /* snag network header to get L4 type and address */
7340 hdr
.network
= skb_network_header(skb
);
7341 if (skb
->encapsulation
) {
7342 #ifdef CONFIG_IXGBE_VXLAN
7343 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
7345 if (!adapter
->vxlan_port
)
7347 if (first
->protocol
!= htons(ETH_P_IP
) ||
7348 hdr
.ipv4
->version
!= IPVERSION
||
7349 hdr
.ipv4
->protocol
!= IPPROTO_UDP
) {
7352 if (ntohs(udp_hdr(skb
)->dest
) != adapter
->vxlan_port
)
7355 hdr
.network
= skb_inner_network_header(skb
);
7356 th
= inner_tcp_hdr(skb
);
7359 #endif /* CONFIG_IXGBE_VXLAN */
7361 /* Currently only IPv4/IPv6 with TCP is supported */
7362 if ((first
->protocol
!= htons(ETH_P_IPV6
) ||
7363 hdr
.ipv6
->nexthdr
!= IPPROTO_TCP
) &&
7364 (first
->protocol
!= htons(ETH_P_IP
) ||
7365 hdr
.ipv4
->protocol
!= IPPROTO_TCP
))
7370 /* skip this packet since it is invalid or the socket is closing */
7374 /* sample on all syn packets or once every atr sample count */
7375 if (!th
->syn
&& (ring
->atr_count
< ring
->atr_sample_rate
))
7378 /* reset sample count */
7379 ring
->atr_count
= 0;
7381 vlan_id
= htons(first
->tx_flags
>> IXGBE_TX_FLAGS_VLAN_SHIFT
);
7384 * src and dst are inverted, think how the receiver sees them
7386 * The input is broken into two sections, a non-compressed section
7387 * containing vm_pool, vlan_id, and flow_type. The rest of the data
7388 * is XORed together and stored in the compressed dword.
7390 input
.formatted
.vlan_id
= vlan_id
;
7393 * since src port and flex bytes occupy the same word XOR them together
7394 * and write the value to source port portion of compressed dword
7396 if (first
->tx_flags
& (IXGBE_TX_FLAGS_SW_VLAN
| IXGBE_TX_FLAGS_HW_VLAN
))
7397 common
.port
.src
^= th
->dest
^ htons(ETH_P_8021Q
);
7399 common
.port
.src
^= th
->dest
^ first
->protocol
;
7400 common
.port
.dst
^= th
->source
;
7402 if (first
->protocol
== htons(ETH_P_IP
)) {
7403 input
.formatted
.flow_type
= IXGBE_ATR_FLOW_TYPE_TCPV4
;
7404 common
.ip
^= hdr
.ipv4
->saddr
^ hdr
.ipv4
->daddr
;
7406 input
.formatted
.flow_type
= IXGBE_ATR_FLOW_TYPE_TCPV6
;
7407 common
.ip
^= hdr
.ipv6
->saddr
.s6_addr32
[0] ^
7408 hdr
.ipv6
->saddr
.s6_addr32
[1] ^
7409 hdr
.ipv6
->saddr
.s6_addr32
[2] ^
7410 hdr
.ipv6
->saddr
.s6_addr32
[3] ^
7411 hdr
.ipv6
->daddr
.s6_addr32
[0] ^
7412 hdr
.ipv6
->daddr
.s6_addr32
[1] ^
7413 hdr
.ipv6
->daddr
.s6_addr32
[2] ^
7414 hdr
.ipv6
->daddr
.s6_addr32
[3];
7417 #ifdef CONFIG_IXGBE_VXLAN
7419 input
.formatted
.flow_type
|= IXGBE_ATR_L4TYPE_TUNNEL_MASK
;
7420 #endif /* CONFIG_IXGBE_VXLAN */
7422 /* This assumes the Rx queue and Tx queue are bound to the same CPU */
7423 ixgbe_fdir_add_signature_filter_82599(&q_vector
->adapter
->hw
,
7424 input
, common
, ring
->queue_index
);
7427 static u16
ixgbe_select_queue(struct net_device
*dev
, struct sk_buff
*skb
,
7428 void *accel_priv
, select_queue_fallback_t fallback
)
7430 struct ixgbe_fwd_adapter
*fwd_adapter
= accel_priv
;
7432 struct ixgbe_adapter
*adapter
;
7433 struct ixgbe_ring_feature
*f
;
7438 return skb
->queue_mapping
+ fwd_adapter
->tx_base_queue
;
7443 * only execute the code below if protocol is FCoE
7444 * or FIP and we have FCoE enabled on the adapter
7446 switch (vlan_get_protocol(skb
)) {
7447 case htons(ETH_P_FCOE
):
7448 case htons(ETH_P_FIP
):
7449 adapter
= netdev_priv(dev
);
7451 if (adapter
->flags
& IXGBE_FLAG_FCOE_ENABLED
)
7454 return fallback(dev
, skb
);
7457 f
= &adapter
->ring_feature
[RING_F_FCOE
];
7459 txq
= skb_rx_queue_recorded(skb
) ? skb_get_rx_queue(skb
) :
7462 while (txq
>= f
->indices
)
7465 return txq
+ f
->offset
;
7467 return fallback(dev
, skb
);
7471 netdev_tx_t
ixgbe_xmit_frame_ring(struct sk_buff
*skb
,
7472 struct ixgbe_adapter
*adapter
,
7473 struct ixgbe_ring
*tx_ring
)
7475 struct ixgbe_tx_buffer
*first
;
7479 u16 count
= TXD_USE_COUNT(skb_headlen(skb
));
7480 __be16 protocol
= skb
->protocol
;
7484 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
7485 * + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
7486 * + 2 desc gap to keep tail from touching head,
7487 * + 1 desc for context descriptor,
7488 * otherwise try next time
7490 for (f
= 0; f
< skb_shinfo(skb
)->nr_frags
; f
++)
7491 count
+= TXD_USE_COUNT(skb_shinfo(skb
)->frags
[f
].size
);
7493 if (ixgbe_maybe_stop_tx(tx_ring
, count
+ 3)) {
7494 tx_ring
->tx_stats
.tx_busy
++;
7495 return NETDEV_TX_BUSY
;
7498 /* record the location of the first descriptor for this packet */
7499 first
= &tx_ring
->tx_buffer_info
[tx_ring
->next_to_use
];
7501 first
->bytecount
= skb
->len
;
7502 first
->gso_segs
= 1;
7504 /* if we have a HW VLAN tag being added default to the HW one */
7505 if (skb_vlan_tag_present(skb
)) {
7506 tx_flags
|= skb_vlan_tag_get(skb
) << IXGBE_TX_FLAGS_VLAN_SHIFT
;
7507 tx_flags
|= IXGBE_TX_FLAGS_HW_VLAN
;
7508 /* else if it is a SW VLAN check the next protocol and store the tag */
7509 } else if (protocol
== htons(ETH_P_8021Q
)) {
7510 struct vlan_hdr
*vhdr
, _vhdr
;
7511 vhdr
= skb_header_pointer(skb
, ETH_HLEN
, sizeof(_vhdr
), &_vhdr
);
7515 tx_flags
|= ntohs(vhdr
->h_vlan_TCI
) <<
7516 IXGBE_TX_FLAGS_VLAN_SHIFT
;
7517 tx_flags
|= IXGBE_TX_FLAGS_SW_VLAN
;
7519 protocol
= vlan_get_protocol(skb
);
7521 if (unlikely(skb_shinfo(skb
)->tx_flags
& SKBTX_HW_TSTAMP
) &&
7522 adapter
->ptp_clock
&&
7523 !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS
,
7525 skb_shinfo(skb
)->tx_flags
|= SKBTX_IN_PROGRESS
;
7526 tx_flags
|= IXGBE_TX_FLAGS_TSTAMP
;
7528 /* schedule check for Tx timestamp */
7529 adapter
->ptp_tx_skb
= skb_get(skb
);
7530 adapter
->ptp_tx_start
= jiffies
;
7531 schedule_work(&adapter
->ptp_tx_work
);
7534 skb_tx_timestamp(skb
);
7536 #ifdef CONFIG_PCI_IOV
7538 * Use the l2switch_enable flag - would be false if the DMA
7539 * Tx switch had been disabled.
7541 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
7542 tx_flags
|= IXGBE_TX_FLAGS_CC
;
7545 /* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
7546 if ((adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
) &&
7547 ((tx_flags
& (IXGBE_TX_FLAGS_HW_VLAN
| IXGBE_TX_FLAGS_SW_VLAN
)) ||
7548 (skb
->priority
!= TC_PRIO_CONTROL
))) {
7549 tx_flags
&= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK
;
7550 tx_flags
|= (skb
->priority
& 0x7) <<
7551 IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT
;
7552 if (tx_flags
& IXGBE_TX_FLAGS_SW_VLAN
) {
7553 struct vlan_ethhdr
*vhdr
;
7555 if (skb_cow_head(skb
, 0))
7557 vhdr
= (struct vlan_ethhdr
*)skb
->data
;
7558 vhdr
->h_vlan_TCI
= htons(tx_flags
>>
7559 IXGBE_TX_FLAGS_VLAN_SHIFT
);
7561 tx_flags
|= IXGBE_TX_FLAGS_HW_VLAN
;
7565 /* record initial flags and protocol */
7566 first
->tx_flags
= tx_flags
;
7567 first
->protocol
= protocol
;
7570 /* setup tx offload for FCoE */
7571 if ((protocol
== htons(ETH_P_FCOE
)) &&
7572 (tx_ring
->netdev
->features
& (NETIF_F_FSO
| NETIF_F_FCOE_CRC
))) {
7573 tso
= ixgbe_fso(tx_ring
, first
, &hdr_len
);
7580 #endif /* IXGBE_FCOE */
7581 tso
= ixgbe_tso(tx_ring
, first
, &hdr_len
);
7585 ixgbe_tx_csum(tx_ring
, first
);
7587 /* add the ATR filter if ATR is on */
7588 if (test_bit(__IXGBE_TX_FDIR_INIT_DONE
, &tx_ring
->state
))
7589 ixgbe_atr(tx_ring
, first
);
7593 #endif /* IXGBE_FCOE */
7594 ixgbe_tx_map(tx_ring
, first
, hdr_len
);
7596 return NETDEV_TX_OK
;
7599 dev_kfree_skb_any(first
->skb
);
7602 return NETDEV_TX_OK
;
7605 static netdev_tx_t
__ixgbe_xmit_frame(struct sk_buff
*skb
,
7606 struct net_device
*netdev
,
7607 struct ixgbe_ring
*ring
)
7609 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7610 struct ixgbe_ring
*tx_ring
;
7613 * The minimum packet size for olinfo paylen is 17 so pad the skb
7614 * in order to meet this minimum size requirement.
7616 if (skb_put_padto(skb
, 17))
7617 return NETDEV_TX_OK
;
7619 tx_ring
= ring
? ring
: adapter
->tx_ring
[skb
->queue_mapping
];
7621 return ixgbe_xmit_frame_ring(skb
, adapter
, tx_ring
);
7624 static netdev_tx_t
ixgbe_xmit_frame(struct sk_buff
*skb
,
7625 struct net_device
*netdev
)
7627 return __ixgbe_xmit_frame(skb
, netdev
, NULL
);
7631 * ixgbe_set_mac - Change the Ethernet Address of the NIC
7632 * @netdev: network interface device structure
7633 * @p: pointer to an address structure
7635 * Returns 0 on success, negative on failure
7637 static int ixgbe_set_mac(struct net_device
*netdev
, void *p
)
7639 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7640 struct ixgbe_hw
*hw
= &adapter
->hw
;
7641 struct sockaddr
*addr
= p
;
7644 if (!is_valid_ether_addr(addr
->sa_data
))
7645 return -EADDRNOTAVAIL
;
7647 ixgbe_del_mac_filter(adapter
, hw
->mac
.addr
, VMDQ_P(0));
7648 memcpy(netdev
->dev_addr
, addr
->sa_data
, netdev
->addr_len
);
7649 memcpy(hw
->mac
.addr
, addr
->sa_data
, netdev
->addr_len
);
7651 ret
= ixgbe_add_mac_filter(adapter
, hw
->mac
.addr
, VMDQ_P(0));
7652 return ret
> 0 ? 0 : ret
;
7656 ixgbe_mdio_read(struct net_device
*netdev
, int prtad
, int devad
, u16 addr
)
7658 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7659 struct ixgbe_hw
*hw
= &adapter
->hw
;
7663 if (prtad
!= hw
->phy
.mdio
.prtad
)
7665 rc
= hw
->phy
.ops
.read_reg(hw
, addr
, devad
, &value
);
7671 static int ixgbe_mdio_write(struct net_device
*netdev
, int prtad
, int devad
,
7672 u16 addr
, u16 value
)
7674 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7675 struct ixgbe_hw
*hw
= &adapter
->hw
;
7677 if (prtad
!= hw
->phy
.mdio
.prtad
)
7679 return hw
->phy
.ops
.write_reg(hw
, addr
, devad
, value
);
7682 static int ixgbe_ioctl(struct net_device
*netdev
, struct ifreq
*req
, int cmd
)
7684 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7688 return ixgbe_ptp_set_ts_config(adapter
, req
);
7690 return ixgbe_ptp_get_ts_config(adapter
, req
);
7692 return mdio_mii_ioctl(&adapter
->hw
.phy
.mdio
, if_mii(req
), cmd
);
7697 * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
7699 * @netdev: network interface device structure
7701 * Returns non-zero on failure
7703 static int ixgbe_add_sanmac_netdev(struct net_device
*dev
)
7706 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
7707 struct ixgbe_hw
*hw
= &adapter
->hw
;
7709 if (is_valid_ether_addr(hw
->mac
.san_addr
)) {
7711 err
= dev_addr_add(dev
, hw
->mac
.san_addr
, NETDEV_HW_ADDR_T_SAN
);
7714 /* update SAN MAC vmdq pool selection */
7715 hw
->mac
.ops
.set_vmdq_san_mac(hw
, VMDQ_P(0));
7721 * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
7723 * @netdev: network interface device structure
7725 * Returns non-zero on failure
7727 static int ixgbe_del_sanmac_netdev(struct net_device
*dev
)
7730 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
7731 struct ixgbe_mac_info
*mac
= &adapter
->hw
.mac
;
7733 if (is_valid_ether_addr(mac
->san_addr
)) {
7735 err
= dev_addr_del(dev
, mac
->san_addr
, NETDEV_HW_ADDR_T_SAN
);
7741 #ifdef CONFIG_NET_POLL_CONTROLLER
7743 * Polling 'interrupt' - used by things like netconsole to send skbs
7744 * without having to re-enable interrupts. It's not called while
7745 * the interrupt routine is executing.
7747 static void ixgbe_netpoll(struct net_device
*netdev
)
7749 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7752 /* if interface is down do nothing */
7753 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
7756 /* loop through and schedule all active queues */
7757 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
7758 ixgbe_msix_clean_rings(0, adapter
->q_vector
[i
]);
7762 static struct rtnl_link_stats64
*ixgbe_get_stats64(struct net_device
*netdev
,
7763 struct rtnl_link_stats64
*stats
)
7765 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7769 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
7770 struct ixgbe_ring
*ring
= ACCESS_ONCE(adapter
->rx_ring
[i
]);
7776 start
= u64_stats_fetch_begin_irq(&ring
->syncp
);
7777 packets
= ring
->stats
.packets
;
7778 bytes
= ring
->stats
.bytes
;
7779 } while (u64_stats_fetch_retry_irq(&ring
->syncp
, start
));
7780 stats
->rx_packets
+= packets
;
7781 stats
->rx_bytes
+= bytes
;
7785 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
7786 struct ixgbe_ring
*ring
= ACCESS_ONCE(adapter
->tx_ring
[i
]);
7792 start
= u64_stats_fetch_begin_irq(&ring
->syncp
);
7793 packets
= ring
->stats
.packets
;
7794 bytes
= ring
->stats
.bytes
;
7795 } while (u64_stats_fetch_retry_irq(&ring
->syncp
, start
));
7796 stats
->tx_packets
+= packets
;
7797 stats
->tx_bytes
+= bytes
;
7801 /* following stats updated by ixgbe_watchdog_task() */
7802 stats
->multicast
= netdev
->stats
.multicast
;
7803 stats
->rx_errors
= netdev
->stats
.rx_errors
;
7804 stats
->rx_length_errors
= netdev
->stats
.rx_length_errors
;
7805 stats
->rx_crc_errors
= netdev
->stats
.rx_crc_errors
;
7806 stats
->rx_missed_errors
= netdev
->stats
.rx_missed_errors
;
7810 #ifdef CONFIG_IXGBE_DCB
7812 * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
7813 * @adapter: pointer to ixgbe_adapter
7814 * @tc: number of traffic classes currently enabled
7816 * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
7817 * 802.1Q priority maps to a packet buffer that exists.
7819 static void ixgbe_validate_rtr(struct ixgbe_adapter
*adapter
, u8 tc
)
7821 struct ixgbe_hw
*hw
= &adapter
->hw
;
7825 /* 82598 have a static priority to TC mapping that can not
7826 * be changed so no validation is needed.
7828 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
7831 reg
= IXGBE_READ_REG(hw
, IXGBE_RTRUP2TC
);
7834 for (i
= 0; i
< MAX_TRAFFIC_CLASS
; i
++) {
7835 u8 up2tc
= reg
>> (i
* IXGBE_RTRUP2TC_UP_SHIFT
);
7837 /* If up2tc is out of bounds default to zero */
7839 reg
&= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT
);
7843 IXGBE_WRITE_REG(hw
, IXGBE_RTRUP2TC
, reg
);
7849 * ixgbe_set_prio_tc_map - Configure netdev prio tc map
7850 * @adapter: Pointer to adapter struct
7852 * Populate the netdev user priority to tc map
7854 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter
*adapter
)
7856 struct net_device
*dev
= adapter
->netdev
;
7857 struct ixgbe_dcb_config
*dcb_cfg
= &adapter
->dcb_cfg
;
7858 struct ieee_ets
*ets
= adapter
->ixgbe_ieee_ets
;
7861 for (prio
= 0; prio
< MAX_USER_PRIORITY
; prio
++) {
7864 if (adapter
->dcbx_cap
& DCB_CAP_DCBX_VER_CEE
)
7865 tc
= ixgbe_dcb_get_tc_from_up(dcb_cfg
, 0, prio
);
7867 tc
= ets
->prio_tc
[prio
];
7869 netdev_set_prio_tc_map(dev
, prio
, tc
);
7873 #endif /* CONFIG_IXGBE_DCB */
7875 * ixgbe_setup_tc - configure net_device for multiple traffic classes
7877 * @netdev: net device to configure
7878 * @tc: number of traffic classes to enable
7880 int ixgbe_setup_tc(struct net_device
*dev
, u8 tc
)
7882 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
7883 struct ixgbe_hw
*hw
= &adapter
->hw
;
7886 /* Hardware supports up to 8 traffic classes */
7887 if (tc
> adapter
->dcb_cfg
.num_tcs
.pg_tcs
)
7890 if (hw
->mac
.type
== ixgbe_mac_82598EB
&& tc
&& tc
< MAX_TRAFFIC_CLASS
)
7893 pools
= (find_first_zero_bit(&adapter
->fwd_bitmask
, 32) > 1);
7894 if (tc
&& pools
&& adapter
->num_rx_pools
> IXGBE_MAX_DCBMACVLANS
)
7897 /* Hardware has to reinitialize queues and interrupts to
7898 * match packet buffer alignment. Unfortunately, the
7899 * hardware is not flexible enough to do this dynamically.
7901 if (netif_running(dev
))
7903 ixgbe_clear_interrupt_scheme(adapter
);
7905 #ifdef CONFIG_IXGBE_DCB
7907 netdev_set_num_tc(dev
, tc
);
7908 ixgbe_set_prio_tc_map(adapter
);
7910 adapter
->flags
|= IXGBE_FLAG_DCB_ENABLED
;
7912 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
) {
7913 adapter
->last_lfc_mode
= adapter
->hw
.fc
.requested_mode
;
7914 adapter
->hw
.fc
.requested_mode
= ixgbe_fc_none
;
7917 netdev_reset_tc(dev
);
7919 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
)
7920 adapter
->hw
.fc
.requested_mode
= adapter
->last_lfc_mode
;
7922 adapter
->flags
&= ~IXGBE_FLAG_DCB_ENABLED
;
7924 adapter
->temp_dcb_cfg
.pfc_mode_enable
= false;
7925 adapter
->dcb_cfg
.pfc_mode_enable
= false;
7928 ixgbe_validate_rtr(adapter
, tc
);
7930 #endif /* CONFIG_IXGBE_DCB */
7931 ixgbe_init_interrupt_scheme(adapter
);
7933 if (netif_running(dev
))
7934 return ixgbe_open(dev
);
7939 #ifdef CONFIG_PCI_IOV
7940 void ixgbe_sriov_reinit(struct ixgbe_adapter
*adapter
)
7942 struct net_device
*netdev
= adapter
->netdev
;
7945 ixgbe_setup_tc(netdev
, netdev_get_num_tc(netdev
));
7950 void ixgbe_do_reset(struct net_device
*netdev
)
7952 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7954 if (netif_running(netdev
))
7955 ixgbe_reinit_locked(adapter
);
7957 ixgbe_reset(adapter
);
7960 static netdev_features_t
ixgbe_fix_features(struct net_device
*netdev
,
7961 netdev_features_t features
)
7963 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7965 /* If Rx checksum is disabled, then RSC/LRO should also be disabled */
7966 if (!(features
& NETIF_F_RXCSUM
))
7967 features
&= ~NETIF_F_LRO
;
7969 /* Turn off LRO if not RSC capable */
7970 if (!(adapter
->flags2
& IXGBE_FLAG2_RSC_CAPABLE
))
7971 features
&= ~NETIF_F_LRO
;
7976 static int ixgbe_set_features(struct net_device
*netdev
,
7977 netdev_features_t features
)
7979 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7980 netdev_features_t changed
= netdev
->features
^ features
;
7981 bool need_reset
= false;
7983 /* Make sure RSC matches LRO, reset if change */
7984 if (!(features
& NETIF_F_LRO
)) {
7985 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)
7987 adapter
->flags2
&= ~IXGBE_FLAG2_RSC_ENABLED
;
7988 } else if ((adapter
->flags2
& IXGBE_FLAG2_RSC_CAPABLE
) &&
7989 !(adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)) {
7990 if (adapter
->rx_itr_setting
== 1 ||
7991 adapter
->rx_itr_setting
> IXGBE_MIN_RSC_ITR
) {
7992 adapter
->flags2
|= IXGBE_FLAG2_RSC_ENABLED
;
7994 } else if ((changed
^ features
) & NETIF_F_LRO
) {
7995 e_info(probe
, "rx-usecs set too low, "
8001 * Check if Flow Director n-tuple support was enabled or disabled. If
8002 * the state changed, we need to reset.
8004 switch (features
& NETIF_F_NTUPLE
) {
8005 case NETIF_F_NTUPLE
:
8006 /* turn off ATR, enable perfect filters and reset */
8007 if (!(adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
))
8010 adapter
->flags
&= ~IXGBE_FLAG_FDIR_HASH_CAPABLE
;
8011 adapter
->flags
|= IXGBE_FLAG_FDIR_PERFECT_CAPABLE
;
8014 /* turn off perfect filters, enable ATR and reset */
8015 if (adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
)
8018 adapter
->flags
&= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE
;
8020 /* We cannot enable ATR if SR-IOV is enabled */
8021 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
8024 /* We cannot enable ATR if we have 2 or more traffic classes */
8025 if (netdev_get_num_tc(netdev
) > 1)
8028 /* We cannot enable ATR if RSS is disabled */
8029 if (adapter
->ring_feature
[RING_F_RSS
].limit
<= 1)
8032 /* A sample rate of 0 indicates ATR disabled */
8033 if (!adapter
->atr_sample_rate
)
8036 adapter
->flags
|= IXGBE_FLAG_FDIR_HASH_CAPABLE
;
8040 if (features
& NETIF_F_HW_VLAN_CTAG_RX
)
8041 ixgbe_vlan_strip_enable(adapter
);
8043 ixgbe_vlan_strip_disable(adapter
);
8045 if (changed
& NETIF_F_RXALL
)
8048 netdev
->features
= features
;
8050 #ifdef CONFIG_IXGBE_VXLAN
8051 if ((adapter
->flags
& IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE
)) {
8052 if (features
& NETIF_F_RXCSUM
)
8053 adapter
->flags2
|= IXGBE_FLAG2_VXLAN_REREG_NEEDED
;
8055 ixgbe_clear_vxlan_port(adapter
);
8057 #endif /* CONFIG_IXGBE_VXLAN */
8060 ixgbe_do_reset(netdev
);
8065 #ifdef CONFIG_IXGBE_VXLAN
8067 * ixgbe_add_vxlan_port - Get notifications about VXLAN ports that come up
8068 * @dev: The port's netdev
8069 * @sa_family: Socket Family that VXLAN is notifiying us about
8070 * @port: New UDP port number that VXLAN started listening to
8072 static void ixgbe_add_vxlan_port(struct net_device
*dev
, sa_family_t sa_family
,
8075 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
8076 struct ixgbe_hw
*hw
= &adapter
->hw
;
8077 u16 new_port
= ntohs(port
);
8079 if (!(adapter
->flags
& IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE
))
8082 if (sa_family
== AF_INET6
)
8085 if (adapter
->vxlan_port
== new_port
)
8088 if (adapter
->vxlan_port
) {
8090 "Hit Max num of VXLAN ports, not adding port %d\n",
8095 adapter
->vxlan_port
= new_port
;
8096 IXGBE_WRITE_REG(hw
, IXGBE_VXLANCTRL
, new_port
);
8100 * ixgbe_del_vxlan_port - Get notifications about VXLAN ports that go away
8101 * @dev: The port's netdev
8102 * @sa_family: Socket Family that VXLAN is notifying us about
8103 * @port: UDP port number that VXLAN stopped listening to
8105 static void ixgbe_del_vxlan_port(struct net_device
*dev
, sa_family_t sa_family
,
8108 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
8109 u16 new_port
= ntohs(port
);
8111 if (!(adapter
->flags
& IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE
))
8114 if (sa_family
== AF_INET6
)
8117 if (adapter
->vxlan_port
!= new_port
) {
8118 netdev_info(dev
, "Port %d was not found, not deleting\n",
8123 ixgbe_clear_vxlan_port(adapter
);
8124 adapter
->flags2
|= IXGBE_FLAG2_VXLAN_REREG_NEEDED
;
8126 #endif /* CONFIG_IXGBE_VXLAN */
8128 static int ixgbe_ndo_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
8129 struct net_device
*dev
,
8130 const unsigned char *addr
, u16 vid
,
8133 /* guarantee we can provide a unique filter for the unicast address */
8134 if (is_unicast_ether_addr(addr
) || is_link_local_ether_addr(addr
)) {
8135 if (IXGBE_MAX_PF_MACVLANS
<= netdev_uc_count(dev
))
8139 return ndo_dflt_fdb_add(ndm
, tb
, dev
, addr
, vid
, flags
);
8143 * ixgbe_configure_bridge_mode - set various bridge modes
8144 * @adapter - the private structure
8145 * @mode - requested bridge mode
8147 * Configure some settings require for various bridge modes.
8149 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter
*adapter
,
8152 struct ixgbe_hw
*hw
= &adapter
->hw
;
8153 unsigned int p
, num_pools
;
8157 case BRIDGE_MODE_VEPA
:
8158 /* disable Tx loopback, rely on switch hairpin mode */
8159 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_PFDTXGSWC
, 0);
8161 /* must enable Rx switching replication to allow multicast
8162 * packet reception on all VFs, and to enable source address
8165 vmdctl
= IXGBE_READ_REG(hw
, IXGBE_VMD_CTL
);
8166 vmdctl
|= IXGBE_VT_CTL_REPLEN
;
8167 IXGBE_WRITE_REG(hw
, IXGBE_VMD_CTL
, vmdctl
);
8169 /* enable Rx source address pruning. Note, this requires
8170 * replication to be enabled or else it does nothing.
8172 num_pools
= adapter
->num_vfs
+ adapter
->num_rx_pools
;
8173 for (p
= 0; p
< num_pools
; p
++) {
8174 if (hw
->mac
.ops
.set_source_address_pruning
)
8175 hw
->mac
.ops
.set_source_address_pruning(hw
,
8180 case BRIDGE_MODE_VEB
:
8181 /* enable Tx loopback for internal VF/PF communication */
8182 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_PFDTXGSWC
,
8183 IXGBE_PFDTXGSWC_VT_LBEN
);
8185 /* disable Rx switching replication unless we have SR-IOV
8188 vmdctl
= IXGBE_READ_REG(hw
, IXGBE_VMD_CTL
);
8189 if (!adapter
->num_vfs
)
8190 vmdctl
&= ~IXGBE_VT_CTL_REPLEN
;
8191 IXGBE_WRITE_REG(hw
, IXGBE_VMD_CTL
, vmdctl
);
8193 /* disable Rx source address pruning, since we don't expect to
8194 * be receiving external loopback of our transmitted frames.
8196 num_pools
= adapter
->num_vfs
+ adapter
->num_rx_pools
;
8197 for (p
= 0; p
< num_pools
; p
++) {
8198 if (hw
->mac
.ops
.set_source_address_pruning
)
8199 hw
->mac
.ops
.set_source_address_pruning(hw
,
8208 adapter
->bridge_mode
= mode
;
8210 e_info(drv
, "enabling bridge mode: %s\n",
8211 mode
== BRIDGE_MODE_VEPA
? "VEPA" : "VEB");
8216 static int ixgbe_ndo_bridge_setlink(struct net_device
*dev
,
8217 struct nlmsghdr
*nlh
, u16 flags
)
8219 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
8220 struct nlattr
*attr
, *br_spec
;
8223 if (!(adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
))
8226 br_spec
= nlmsg_find_attr(nlh
, sizeof(struct ifinfomsg
), IFLA_AF_SPEC
);
8230 nla_for_each_nested(attr
, br_spec
, rem
) {
8234 if (nla_type(attr
) != IFLA_BRIDGE_MODE
)
8237 if (nla_len(attr
) < sizeof(mode
))
8240 mode
= nla_get_u16(attr
);
8241 status
= ixgbe_configure_bridge_mode(adapter
, mode
);
8251 static int ixgbe_ndo_bridge_getlink(struct sk_buff
*skb
, u32 pid
, u32 seq
,
8252 struct net_device
*dev
,
8253 u32 filter_mask
, int nlflags
)
8255 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
8257 if (!(adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
))
8260 return ndo_dflt_bridge_getlink(skb
, pid
, seq
, dev
,
8261 adapter
->bridge_mode
, 0, 0, nlflags
,
8265 static void *ixgbe_fwd_add(struct net_device
*pdev
, struct net_device
*vdev
)
8267 struct ixgbe_fwd_adapter
*fwd_adapter
= NULL
;
8268 struct ixgbe_adapter
*adapter
= netdev_priv(pdev
);
8269 int used_pools
= adapter
->num_vfs
+ adapter
->num_rx_pools
;
8273 /* Hardware has a limited number of available pools. Each VF, and the
8274 * PF require a pool. Check to ensure we don't attempt to use more
8275 * then the available number of pools.
8277 if (used_pools
>= IXGBE_MAX_VF_FUNCTIONS
)
8278 return ERR_PTR(-EINVAL
);
8281 if (vdev
->num_rx_queues
!= vdev
->num_tx_queues
) {
8282 netdev_info(pdev
, "%s: Only supports a single queue count for TX and RX\n",
8284 return ERR_PTR(-EINVAL
);
8287 /* Check for hardware restriction on number of rx/tx queues */
8288 if (vdev
->num_tx_queues
> IXGBE_MAX_L2A_QUEUES
||
8289 vdev
->num_tx_queues
== IXGBE_BAD_L2A_QUEUE
) {
8291 "%s: Supports RX/TX Queue counts 1,2, and 4\n",
8293 return ERR_PTR(-EINVAL
);
8296 if (((adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
) &&
8297 adapter
->num_rx_pools
> IXGBE_MAX_DCBMACVLANS
- 1) ||
8298 (adapter
->num_rx_pools
> IXGBE_MAX_MACVLANS
))
8299 return ERR_PTR(-EBUSY
);
8301 fwd_adapter
= kzalloc(sizeof(*fwd_adapter
), GFP_KERNEL
);
8303 return ERR_PTR(-ENOMEM
);
8305 pool
= find_first_zero_bit(&adapter
->fwd_bitmask
, 32);
8306 adapter
->num_rx_pools
++;
8307 set_bit(pool
, &adapter
->fwd_bitmask
);
8308 limit
= find_last_bit(&adapter
->fwd_bitmask
, 32);
8310 /* Enable VMDq flag so device will be set in VM mode */
8311 adapter
->flags
|= IXGBE_FLAG_VMDQ_ENABLED
| IXGBE_FLAG_SRIOV_ENABLED
;
8312 adapter
->ring_feature
[RING_F_VMDQ
].limit
= limit
+ 1;
8313 adapter
->ring_feature
[RING_F_RSS
].limit
= vdev
->num_tx_queues
;
8315 /* Force reinit of ring allocation with VMDQ enabled */
8316 err
= ixgbe_setup_tc(pdev
, netdev_get_num_tc(pdev
));
8319 fwd_adapter
->pool
= pool
;
8320 fwd_adapter
->real_adapter
= adapter
;
8321 err
= ixgbe_fwd_ring_up(vdev
, fwd_adapter
);
8324 netif_tx_start_all_queues(vdev
);
8327 /* unwind counter and free adapter struct */
8329 "%s: dfwd hardware acceleration failed\n", vdev
->name
);
8330 clear_bit(pool
, &adapter
->fwd_bitmask
);
8331 adapter
->num_rx_pools
--;
8333 return ERR_PTR(err
);
8336 static void ixgbe_fwd_del(struct net_device
*pdev
, void *priv
)
8338 struct ixgbe_fwd_adapter
*fwd_adapter
= priv
;
8339 struct ixgbe_adapter
*adapter
= fwd_adapter
->real_adapter
;
8342 clear_bit(fwd_adapter
->pool
, &adapter
->fwd_bitmask
);
8343 adapter
->num_rx_pools
--;
8345 limit
= find_last_bit(&adapter
->fwd_bitmask
, 32);
8346 adapter
->ring_feature
[RING_F_VMDQ
].limit
= limit
+ 1;
8347 ixgbe_fwd_ring_down(fwd_adapter
->netdev
, fwd_adapter
);
8348 ixgbe_setup_tc(pdev
, netdev_get_num_tc(pdev
));
8349 netdev_dbg(pdev
, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
8350 fwd_adapter
->pool
, adapter
->num_rx_pools
,
8351 fwd_adapter
->rx_base_queue
,
8352 fwd_adapter
->rx_base_queue
+ adapter
->num_rx_queues_per_pool
,
8353 adapter
->fwd_bitmask
);
8357 #define IXGBE_MAX_TUNNEL_HDR_LEN 80
8358 static netdev_features_t
8359 ixgbe_features_check(struct sk_buff
*skb
, struct net_device
*dev
,
8360 netdev_features_t features
)
8362 if (!skb
->encapsulation
)
8365 if (unlikely(skb_inner_mac_header(skb
) - skb_transport_header(skb
) >
8366 IXGBE_MAX_TUNNEL_HDR_LEN
))
8367 return features
& ~NETIF_F_ALL_CSUM
;
8372 static const struct net_device_ops ixgbe_netdev_ops
= {
8373 .ndo_open
= ixgbe_open
,
8374 .ndo_stop
= ixgbe_close
,
8375 .ndo_start_xmit
= ixgbe_xmit_frame
,
8376 .ndo_select_queue
= ixgbe_select_queue
,
8377 .ndo_set_rx_mode
= ixgbe_set_rx_mode
,
8378 .ndo_validate_addr
= eth_validate_addr
,
8379 .ndo_set_mac_address
= ixgbe_set_mac
,
8380 .ndo_change_mtu
= ixgbe_change_mtu
,
8381 .ndo_tx_timeout
= ixgbe_tx_timeout
,
8382 .ndo_vlan_rx_add_vid
= ixgbe_vlan_rx_add_vid
,
8383 .ndo_vlan_rx_kill_vid
= ixgbe_vlan_rx_kill_vid
,
8384 .ndo_do_ioctl
= ixgbe_ioctl
,
8385 .ndo_set_vf_mac
= ixgbe_ndo_set_vf_mac
,
8386 .ndo_set_vf_vlan
= ixgbe_ndo_set_vf_vlan
,
8387 .ndo_set_vf_rate
= ixgbe_ndo_set_vf_bw
,
8388 .ndo_set_vf_spoofchk
= ixgbe_ndo_set_vf_spoofchk
,
8389 .ndo_set_vf_rss_query_en
= ixgbe_ndo_set_vf_rss_query_en
,
8390 .ndo_get_vf_config
= ixgbe_ndo_get_vf_config
,
8391 .ndo_get_stats64
= ixgbe_get_stats64
,
8392 #ifdef CONFIG_IXGBE_DCB
8393 .ndo_setup_tc
= ixgbe_setup_tc
,
8395 #ifdef CONFIG_NET_POLL_CONTROLLER
8396 .ndo_poll_controller
= ixgbe_netpoll
,
8398 #ifdef CONFIG_NET_RX_BUSY_POLL
8399 .ndo_busy_poll
= ixgbe_low_latency_recv
,
8402 .ndo_fcoe_ddp_setup
= ixgbe_fcoe_ddp_get
,
8403 .ndo_fcoe_ddp_target
= ixgbe_fcoe_ddp_target
,
8404 .ndo_fcoe_ddp_done
= ixgbe_fcoe_ddp_put
,
8405 .ndo_fcoe_enable
= ixgbe_fcoe_enable
,
8406 .ndo_fcoe_disable
= ixgbe_fcoe_disable
,
8407 .ndo_fcoe_get_wwn
= ixgbe_fcoe_get_wwn
,
8408 .ndo_fcoe_get_hbainfo
= ixgbe_fcoe_get_hbainfo
,
8409 #endif /* IXGBE_FCOE */
8410 .ndo_set_features
= ixgbe_set_features
,
8411 .ndo_fix_features
= ixgbe_fix_features
,
8412 .ndo_fdb_add
= ixgbe_ndo_fdb_add
,
8413 .ndo_bridge_setlink
= ixgbe_ndo_bridge_setlink
,
8414 .ndo_bridge_getlink
= ixgbe_ndo_bridge_getlink
,
8415 .ndo_dfwd_add_station
= ixgbe_fwd_add
,
8416 .ndo_dfwd_del_station
= ixgbe_fwd_del
,
8417 #ifdef CONFIG_IXGBE_VXLAN
8418 .ndo_add_vxlan_port
= ixgbe_add_vxlan_port
,
8419 .ndo_del_vxlan_port
= ixgbe_del_vxlan_port
,
8420 #endif /* CONFIG_IXGBE_VXLAN */
8421 .ndo_features_check
= ixgbe_features_check
,
8425 * ixgbe_enumerate_functions - Get the number of ports this device has
8426 * @adapter: adapter structure
8428 * This function enumerates the phsyical functions co-located on a single slot,
8429 * in order to determine how many ports a device has. This is most useful in
8430 * determining the required GT/s of PCIe bandwidth necessary for optimal
8433 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter
*adapter
)
8435 struct pci_dev
*entry
, *pdev
= adapter
->pdev
;
8438 /* Some cards can not use the generic count PCIe functions method,
8439 * because they are behind a parent switch, so we hardcode these with
8440 * the correct number of functions.
8442 if (ixgbe_pcie_from_parent(&adapter
->hw
))
8445 list_for_each_entry(entry
, &adapter
->pdev
->bus
->devices
, bus_list
) {
8446 /* don't count virtual functions */
8447 if (entry
->is_virtfn
)
8450 /* When the devices on the bus don't all match our device ID,
8451 * we can't reliably determine the correct number of
8452 * functions. This can occur if a function has been direct
8453 * attached to a virtual machine using VT-d, for example. In
8454 * this case, simply return -1 to indicate this.
8456 if ((entry
->vendor
!= pdev
->vendor
) ||
8457 (entry
->device
!= pdev
->device
))
8467 * ixgbe_wol_supported - Check whether device supports WoL
8468 * @hw: hw specific details
8469 * @device_id: the device ID
8470 * @subdev_id: the subsystem device ID
8472 * This function is used by probe and ethtool to determine
8473 * which devices have WoL support
8476 int ixgbe_wol_supported(struct ixgbe_adapter
*adapter
, u16 device_id
,
8479 struct ixgbe_hw
*hw
= &adapter
->hw
;
8480 u16 wol_cap
= adapter
->eeprom_cap
& IXGBE_DEVICE_CAPS_WOL_MASK
;
8481 int is_wol_supported
= 0;
8483 switch (device_id
) {
8484 case IXGBE_DEV_ID_82599_SFP
:
8485 /* Only these subdevices could supports WOL */
8486 switch (subdevice_id
) {
8487 case IXGBE_SUBDEV_ID_82599_SFP_WOL0
:
8488 case IXGBE_SUBDEV_ID_82599_560FLR
:
8489 /* only support first port */
8490 if (hw
->bus
.func
!= 0)
8492 case IXGBE_SUBDEV_ID_82599_SP_560FLR
:
8493 case IXGBE_SUBDEV_ID_82599_SFP
:
8494 case IXGBE_SUBDEV_ID_82599_RNDC
:
8495 case IXGBE_SUBDEV_ID_82599_ECNA_DP
:
8496 case IXGBE_SUBDEV_ID_82599_LOM_SFP
:
8497 is_wol_supported
= 1;
8501 case IXGBE_DEV_ID_82599EN_SFP
:
8502 /* Only this subdevice supports WOL */
8503 switch (subdevice_id
) {
8504 case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1
:
8505 is_wol_supported
= 1;
8509 case IXGBE_DEV_ID_82599_COMBO_BACKPLANE
:
8510 /* All except this subdevice support WOL */
8511 if (subdevice_id
!= IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ
)
8512 is_wol_supported
= 1;
8514 case IXGBE_DEV_ID_82599_KX4
:
8515 is_wol_supported
= 1;
8517 case IXGBE_DEV_ID_X540T
:
8518 case IXGBE_DEV_ID_X540T1
:
8519 case IXGBE_DEV_ID_X550T
:
8520 case IXGBE_DEV_ID_X550EM_X_KX4
:
8521 case IXGBE_DEV_ID_X550EM_X_KR
:
8522 case IXGBE_DEV_ID_X550EM_X_10G_T
:
8523 /* check eeprom to see if enabled wol */
8524 if ((wol_cap
== IXGBE_DEVICE_CAPS_WOL_PORT0_1
) ||
8525 ((wol_cap
== IXGBE_DEVICE_CAPS_WOL_PORT0
) &&
8526 (hw
->bus
.func
== 0))) {
8527 is_wol_supported
= 1;
8532 return is_wol_supported
;
8536 * ixgbe_get_platform_mac_addr - Look up MAC address in Open Firmware / IDPROM
8537 * @adapter: Pointer to adapter struct
8539 static void ixgbe_get_platform_mac_addr(struct ixgbe_adapter
*adapter
)
8542 struct device_node
*dp
= pci_device_to_OF_node(adapter
->pdev
);
8543 struct ixgbe_hw
*hw
= &adapter
->hw
;
8544 const unsigned char *addr
;
8546 addr
= of_get_mac_address(dp
);
8548 ether_addr_copy(hw
->mac
.perm_addr
, addr
);
8551 #endif /* CONFIG_OF */
8554 ether_addr_copy(hw
->mac
.perm_addr
, idprom
->id_ethaddr
);
8555 #endif /* CONFIG_SPARC */
8559 * ixgbe_probe - Device Initialization Routine
8560 * @pdev: PCI device information struct
8561 * @ent: entry in ixgbe_pci_tbl
8563 * Returns 0 on success, negative on failure
8565 * ixgbe_probe initializes an adapter identified by a pci_dev structure.
8566 * The OS initialization, configuring of the adapter private structure,
8567 * and a hardware reset occur.
8569 static int ixgbe_probe(struct pci_dev
*pdev
, const struct pci_device_id
*ent
)
8571 struct net_device
*netdev
;
8572 struct ixgbe_adapter
*adapter
= NULL
;
8573 struct ixgbe_hw
*hw
;
8574 const struct ixgbe_info
*ii
= ixgbe_info_tbl
[ent
->driver_data
];
8575 int i
, err
, pci_using_dac
, expected_gts
;
8576 unsigned int indices
= MAX_TX_QUEUES
;
8577 u8 part_str
[IXGBE_PBANUM_LENGTH
];
8578 bool disable_dev
= false;
8584 /* Catch broken hardware that put the wrong VF device ID in
8585 * the PCIe SR-IOV capability.
8587 if (pdev
->is_virtfn
) {
8588 WARN(1, KERN_ERR
"%s (%hx:%hx) should not be a VF!\n",
8589 pci_name(pdev
), pdev
->vendor
, pdev
->device
);
8593 err
= pci_enable_device_mem(pdev
);
8597 if (!dma_set_mask_and_coherent(&pdev
->dev
, DMA_BIT_MASK(64))) {
8600 err
= dma_set_mask_and_coherent(&pdev
->dev
, DMA_BIT_MASK(32));
8603 "No usable DMA configuration, aborting\n");
8609 err
= pci_request_selected_regions(pdev
, pci_select_bars(pdev
,
8610 IORESOURCE_MEM
), ixgbe_driver_name
);
8613 "pci_request_selected_regions failed 0x%x\n", err
);
8617 pci_enable_pcie_error_reporting(pdev
);
8619 pci_set_master(pdev
);
8620 pci_save_state(pdev
);
8622 if (ii
->mac
== ixgbe_mac_82598EB
) {
8623 #ifdef CONFIG_IXGBE_DCB
8624 /* 8 TC w/ 4 queues per TC */
8625 indices
= 4 * MAX_TRAFFIC_CLASS
;
8627 indices
= IXGBE_MAX_RSS_INDICES
;
8631 netdev
= alloc_etherdev_mq(sizeof(struct ixgbe_adapter
), indices
);
8634 goto err_alloc_etherdev
;
8637 SET_NETDEV_DEV(netdev
, &pdev
->dev
);
8639 adapter
= netdev_priv(netdev
);
8641 adapter
->netdev
= netdev
;
8642 adapter
->pdev
= pdev
;
8645 adapter
->msg_enable
= netif_msg_init(debug
, DEFAULT_MSG_ENABLE
);
8647 hw
->hw_addr
= ioremap(pci_resource_start(pdev
, 0),
8648 pci_resource_len(pdev
, 0));
8649 adapter
->io_addr
= hw
->hw_addr
;
8655 netdev
->netdev_ops
= &ixgbe_netdev_ops
;
8656 ixgbe_set_ethtool_ops(netdev
);
8657 netdev
->watchdog_timeo
= 5 * HZ
;
8658 strlcpy(netdev
->name
, pci_name(pdev
), sizeof(netdev
->name
));
8661 memcpy(&hw
->mac
.ops
, ii
->mac_ops
, sizeof(hw
->mac
.ops
));
8662 hw
->mac
.type
= ii
->mac
;
8663 hw
->mvals
= ii
->mvals
;
8666 memcpy(&hw
->eeprom
.ops
, ii
->eeprom_ops
, sizeof(hw
->eeprom
.ops
));
8667 eec
= IXGBE_READ_REG(hw
, IXGBE_EEC(hw
));
8668 if (ixgbe_removed(hw
->hw_addr
)) {
8672 /* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
8673 if (!(eec
& (1 << 8)))
8674 hw
->eeprom
.ops
.read
= &ixgbe_read_eeprom_bit_bang_generic
;
8677 memcpy(&hw
->phy
.ops
, ii
->phy_ops
, sizeof(hw
->phy
.ops
));
8678 hw
->phy
.sfp_type
= ixgbe_sfp_type_unknown
;
8679 /* ixgbe_identify_phy_generic will set prtad and mmds properly */
8680 hw
->phy
.mdio
.prtad
= MDIO_PRTAD_NONE
;
8681 hw
->phy
.mdio
.mmds
= 0;
8682 hw
->phy
.mdio
.mode_support
= MDIO_SUPPORTS_C45
| MDIO_EMULATE_C22
;
8683 hw
->phy
.mdio
.dev
= netdev
;
8684 hw
->phy
.mdio
.mdio_read
= ixgbe_mdio_read
;
8685 hw
->phy
.mdio
.mdio_write
= ixgbe_mdio_write
;
8687 ii
->get_invariants(hw
);
8689 /* setup the private structure */
8690 err
= ixgbe_sw_init(adapter
);
8694 /* Make it possible the adapter to be woken up via WOL */
8695 switch (adapter
->hw
.mac
.type
) {
8696 case ixgbe_mac_82599EB
:
8697 case ixgbe_mac_X540
:
8698 case ixgbe_mac_X550
:
8699 case ixgbe_mac_X550EM_x
:
8700 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_WUS
, ~0);
8707 * If there is a fan on this device and it has failed log the
8710 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
) {
8711 u32 esdp
= IXGBE_READ_REG(hw
, IXGBE_ESDP
);
8712 if (esdp
& IXGBE_ESDP_SDP1
)
8713 e_crit(probe
, "Fan has stopped, replace the adapter\n");
8716 if (allow_unsupported_sfp
)
8717 hw
->allow_unsupported_sfp
= allow_unsupported_sfp
;
8719 /* reset_hw fills in the perm_addr as well */
8720 hw
->phy
.reset_if_overtemp
= true;
8721 err
= hw
->mac
.ops
.reset_hw(hw
);
8722 hw
->phy
.reset_if_overtemp
= false;
8723 if (err
== IXGBE_ERR_SFP_NOT_PRESENT
) {
8725 } else if (err
== IXGBE_ERR_SFP_NOT_SUPPORTED
) {
8726 e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
8727 e_dev_err("Reload the driver after installing a supported module.\n");
8730 e_dev_err("HW Init failed: %d\n", err
);
8734 #ifdef CONFIG_PCI_IOV
8735 /* SR-IOV not supported on the 82598 */
8736 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
)
8739 ixgbe_init_mbx_params_pf(hw
);
8740 memcpy(&hw
->mbx
.ops
, ii
->mbx_ops
, sizeof(hw
->mbx
.ops
));
8741 pci_sriov_set_totalvfs(pdev
, IXGBE_MAX_VFS_DRV_LIMIT
);
8742 ixgbe_enable_sriov(adapter
);
8746 netdev
->features
= NETIF_F_SG
|
8749 NETIF_F_HW_VLAN_CTAG_TX
|
8750 NETIF_F_HW_VLAN_CTAG_RX
|
8756 netdev
->hw_features
= netdev
->features
| NETIF_F_HW_L2FW_DOFFLOAD
;
8758 switch (adapter
->hw
.mac
.type
) {
8759 case ixgbe_mac_82599EB
:
8760 case ixgbe_mac_X540
:
8761 case ixgbe_mac_X550
:
8762 case ixgbe_mac_X550EM_x
:
8763 netdev
->features
|= NETIF_F_SCTP_CSUM
;
8764 netdev
->hw_features
|= NETIF_F_SCTP_CSUM
|
8771 netdev
->hw_features
|= NETIF_F_RXALL
;
8772 netdev
->features
|= NETIF_F_HW_VLAN_CTAG_FILTER
;
8774 netdev
->vlan_features
|= NETIF_F_TSO
;
8775 netdev
->vlan_features
|= NETIF_F_TSO6
;
8776 netdev
->vlan_features
|= NETIF_F_IP_CSUM
;
8777 netdev
->vlan_features
|= NETIF_F_IPV6_CSUM
;
8778 netdev
->vlan_features
|= NETIF_F_SG
;
8780 netdev
->hw_enc_features
|= NETIF_F_SG
| NETIF_F_IP_CSUM
|
8783 netdev
->priv_flags
|= IFF_UNICAST_FLT
;
8784 netdev
->priv_flags
|= IFF_SUPP_NOFCS
;
8786 #ifdef CONFIG_IXGBE_VXLAN
8787 switch (adapter
->hw
.mac
.type
) {
8788 case ixgbe_mac_X550
:
8789 case ixgbe_mac_X550EM_x
:
8790 netdev
->hw_enc_features
|= NETIF_F_RXCSUM
|
8797 #endif /* CONFIG_IXGBE_VXLAN */
8799 #ifdef CONFIG_IXGBE_DCB
8800 netdev
->dcbnl_ops
= &dcbnl_ops
;
8804 if (adapter
->flags
& IXGBE_FLAG_FCOE_CAPABLE
) {
8805 unsigned int fcoe_l
;
8807 if (hw
->mac
.ops
.get_device_caps
) {
8808 hw
->mac
.ops
.get_device_caps(hw
, &device_caps
);
8809 if (device_caps
& IXGBE_DEVICE_CAPS_FCOE_OFFLOADS
)
8810 adapter
->flags
&= ~IXGBE_FLAG_FCOE_CAPABLE
;
8814 fcoe_l
= min_t(int, IXGBE_FCRETA_SIZE
, num_online_cpus());
8815 adapter
->ring_feature
[RING_F_FCOE
].limit
= fcoe_l
;
8817 netdev
->features
|= NETIF_F_FSO
|
8820 netdev
->vlan_features
|= NETIF_F_FSO
|
8824 #endif /* IXGBE_FCOE */
8825 if (pci_using_dac
) {
8826 netdev
->features
|= NETIF_F_HIGHDMA
;
8827 netdev
->vlan_features
|= NETIF_F_HIGHDMA
;
8830 if (adapter
->flags2
& IXGBE_FLAG2_RSC_CAPABLE
)
8831 netdev
->hw_features
|= NETIF_F_LRO
;
8832 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)
8833 netdev
->features
|= NETIF_F_LRO
;
8835 /* make sure the EEPROM is good */
8836 if (hw
->eeprom
.ops
.validate_checksum(hw
, NULL
) < 0) {
8837 e_dev_err("The EEPROM Checksum Is Not Valid\n");
8842 ixgbe_get_platform_mac_addr(adapter
);
8844 memcpy(netdev
->dev_addr
, hw
->mac
.perm_addr
, netdev
->addr_len
);
8846 if (!is_valid_ether_addr(netdev
->dev_addr
)) {
8847 e_dev_err("invalid MAC address\n");
8852 ixgbe_mac_set_default_filter(adapter
, hw
->mac
.perm_addr
);
8854 setup_timer(&adapter
->service_timer
, &ixgbe_service_timer
,
8855 (unsigned long) adapter
);
8857 if (ixgbe_removed(hw
->hw_addr
)) {
8861 INIT_WORK(&adapter
->service_task
, ixgbe_service_task
);
8862 set_bit(__IXGBE_SERVICE_INITED
, &adapter
->state
);
8863 clear_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
);
8865 err
= ixgbe_init_interrupt_scheme(adapter
);
8869 /* WOL not supported for all devices */
8871 hw
->eeprom
.ops
.read(hw
, 0x2c, &adapter
->eeprom_cap
);
8872 hw
->wol_enabled
= ixgbe_wol_supported(adapter
, pdev
->device
,
8873 pdev
->subsystem_device
);
8874 if (hw
->wol_enabled
)
8875 adapter
->wol
= IXGBE_WUFC_MAG
;
8877 device_set_wakeup_enable(&adapter
->pdev
->dev
, adapter
->wol
);
8879 /* save off EEPROM version number */
8880 hw
->eeprom
.ops
.read(hw
, 0x2e, &adapter
->eeprom_verh
);
8881 hw
->eeprom
.ops
.read(hw
, 0x2d, &adapter
->eeprom_verl
);
8883 /* pick up the PCI bus settings for reporting later */
8884 if (ixgbe_pcie_from_parent(hw
))
8885 ixgbe_get_parent_bus_info(adapter
);
8887 hw
->mac
.ops
.get_bus_info(hw
);
8889 /* calculate the expected PCIe bandwidth required for optimal
8890 * performance. Note that some older parts will never have enough
8891 * bandwidth due to being older generation PCIe parts. We clamp these
8892 * parts to ensure no warning is displayed if it can't be fixed.
8894 switch (hw
->mac
.type
) {
8895 case ixgbe_mac_82598EB
:
8896 expected_gts
= min(ixgbe_enumerate_functions(adapter
) * 10, 16);
8899 expected_gts
= ixgbe_enumerate_functions(adapter
) * 10;
8903 /* don't check link if we failed to enumerate functions */
8904 if (expected_gts
> 0)
8905 ixgbe_check_minimum_link(adapter
, expected_gts
);
8907 err
= ixgbe_read_pba_string_generic(hw
, part_str
, sizeof(part_str
));
8909 strlcpy(part_str
, "Unknown", sizeof(part_str
));
8910 if (ixgbe_is_sfp(hw
) && hw
->phy
.sfp_type
!= ixgbe_sfp_type_not_present
)
8911 e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
8912 hw
->mac
.type
, hw
->phy
.type
, hw
->phy
.sfp_type
,
8915 e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
8916 hw
->mac
.type
, hw
->phy
.type
, part_str
);
8918 e_dev_info("%pM\n", netdev
->dev_addr
);
8920 /* reset the hardware with the new settings */
8921 err
= hw
->mac
.ops
.start_hw(hw
);
8922 if (err
== IXGBE_ERR_EEPROM_VERSION
) {
8923 /* We are running on a pre-production device, log a warning */
8924 e_dev_warn("This device is a pre-production adapter/LOM. "
8925 "Please be aware there may be issues associated "
8926 "with your hardware. If you are experiencing "
8927 "problems please contact your Intel or hardware "
8928 "representative who provided you with this "
8931 strcpy(netdev
->name
, "eth%d");
8932 err
= register_netdev(netdev
);
8936 pci_set_drvdata(pdev
, adapter
);
8938 /* power down the optics for 82599 SFP+ fiber */
8939 if (hw
->mac
.ops
.disable_tx_laser
)
8940 hw
->mac
.ops
.disable_tx_laser(hw
);
8942 /* carrier off reporting is important to ethtool even BEFORE open */
8943 netif_carrier_off(netdev
);
8945 #ifdef CONFIG_IXGBE_DCA
8946 if (dca_add_requester(&pdev
->dev
) == 0) {
8947 adapter
->flags
|= IXGBE_FLAG_DCA_ENABLED
;
8948 ixgbe_setup_dca(adapter
);
8951 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
8952 e_info(probe
, "IOV is enabled with %d VFs\n", adapter
->num_vfs
);
8953 for (i
= 0; i
< adapter
->num_vfs
; i
++)
8954 ixgbe_vf_configuration(pdev
, (i
| 0x10000000));
8957 /* firmware requires driver version to be 0xFFFFFFFF
8958 * since os does not support feature
8960 if (hw
->mac
.ops
.set_fw_drv_ver
)
8961 hw
->mac
.ops
.set_fw_drv_ver(hw
, 0xFF, 0xFF, 0xFF,
8964 /* add san mac addr to netdev */
8965 ixgbe_add_sanmac_netdev(netdev
);
8967 e_dev_info("%s\n", ixgbe_default_device_descr
);
8969 #ifdef CONFIG_IXGBE_HWMON
8970 if (ixgbe_sysfs_init(adapter
))
8971 e_err(probe
, "failed to allocate sysfs resources\n");
8972 #endif /* CONFIG_IXGBE_HWMON */
8974 ixgbe_dbg_adapter_init(adapter
);
8976 /* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
8977 if (ixgbe_mng_enabled(hw
) && ixgbe_is_sfp(hw
) && hw
->mac
.ops
.setup_link
)
8978 hw
->mac
.ops
.setup_link(hw
,
8979 IXGBE_LINK_SPEED_10GB_FULL
| IXGBE_LINK_SPEED_1GB_FULL
,
8985 ixgbe_release_hw_control(adapter
);
8986 ixgbe_clear_interrupt_scheme(adapter
);
8988 ixgbe_disable_sriov(adapter
);
8989 adapter
->flags2
&= ~IXGBE_FLAG2_SEARCH_FOR_SFP
;
8990 iounmap(adapter
->io_addr
);
8991 kfree(adapter
->mac_table
);
8993 disable_dev
= !test_and_set_bit(__IXGBE_DISABLED
, &adapter
->state
);
8994 free_netdev(netdev
);
8996 pci_release_selected_regions(pdev
,
8997 pci_select_bars(pdev
, IORESOURCE_MEM
));
9000 if (!adapter
|| disable_dev
)
9001 pci_disable_device(pdev
);
9006 * ixgbe_remove - Device Removal Routine
9007 * @pdev: PCI device information struct
9009 * ixgbe_remove is called by the PCI subsystem to alert the driver
9010 * that it should release a PCI device. The could be caused by a
9011 * Hot-Plug event, or because the driver is going to be removed from
9014 static void ixgbe_remove(struct pci_dev
*pdev
)
9016 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
9017 struct net_device
*netdev
;
9020 /* if !adapter then we already cleaned up in probe */
9024 netdev
= adapter
->netdev
;
9025 ixgbe_dbg_adapter_exit(adapter
);
9027 set_bit(__IXGBE_REMOVING
, &adapter
->state
);
9028 cancel_work_sync(&adapter
->service_task
);
9031 #ifdef CONFIG_IXGBE_DCA
9032 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
) {
9033 adapter
->flags
&= ~IXGBE_FLAG_DCA_ENABLED
;
9034 dca_remove_requester(&pdev
->dev
);
9035 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_DCA_CTRL
, 1);
9039 #ifdef CONFIG_IXGBE_HWMON
9040 ixgbe_sysfs_exit(adapter
);
9041 #endif /* CONFIG_IXGBE_HWMON */
9043 /* remove the added san mac */
9044 ixgbe_del_sanmac_netdev(netdev
);
9046 #ifdef CONFIG_PCI_IOV
9047 ixgbe_disable_sriov(adapter
);
9049 if (netdev
->reg_state
== NETREG_REGISTERED
)
9050 unregister_netdev(netdev
);
9052 ixgbe_clear_interrupt_scheme(adapter
);
9054 ixgbe_release_hw_control(adapter
);
9057 kfree(adapter
->ixgbe_ieee_pfc
);
9058 kfree(adapter
->ixgbe_ieee_ets
);
9061 iounmap(adapter
->io_addr
);
9062 pci_release_selected_regions(pdev
, pci_select_bars(pdev
,
9065 e_dev_info("complete\n");
9067 kfree(adapter
->mac_table
);
9068 disable_dev
= !test_and_set_bit(__IXGBE_DISABLED
, &adapter
->state
);
9069 free_netdev(netdev
);
9071 pci_disable_pcie_error_reporting(pdev
);
9074 pci_disable_device(pdev
);
9078 * ixgbe_io_error_detected - called when PCI error is detected
9079 * @pdev: Pointer to PCI device
9080 * @state: The current pci connection state
9082 * This function is called after a PCI bus error affecting
9083 * this device has been detected.
9085 static pci_ers_result_t
ixgbe_io_error_detected(struct pci_dev
*pdev
,
9086 pci_channel_state_t state
)
9088 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
9089 struct net_device
*netdev
= adapter
->netdev
;
9091 #ifdef CONFIG_PCI_IOV
9092 struct ixgbe_hw
*hw
= &adapter
->hw
;
9093 struct pci_dev
*bdev
, *vfdev
;
9094 u32 dw0
, dw1
, dw2
, dw3
;
9096 u16 req_id
, pf_func
;
9098 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
||
9099 adapter
->num_vfs
== 0)
9100 goto skip_bad_vf_detection
;
9102 bdev
= pdev
->bus
->self
;
9103 while (bdev
&& (pci_pcie_type(bdev
) != PCI_EXP_TYPE_ROOT_PORT
))
9104 bdev
= bdev
->bus
->self
;
9107 goto skip_bad_vf_detection
;
9109 pos
= pci_find_ext_capability(bdev
, PCI_EXT_CAP_ID_ERR
);
9111 goto skip_bad_vf_detection
;
9113 dw0
= ixgbe_read_pci_cfg_dword(hw
, pos
+ PCI_ERR_HEADER_LOG
);
9114 dw1
= ixgbe_read_pci_cfg_dword(hw
, pos
+ PCI_ERR_HEADER_LOG
+ 4);
9115 dw2
= ixgbe_read_pci_cfg_dword(hw
, pos
+ PCI_ERR_HEADER_LOG
+ 8);
9116 dw3
= ixgbe_read_pci_cfg_dword(hw
, pos
+ PCI_ERR_HEADER_LOG
+ 12);
9117 if (ixgbe_removed(hw
->hw_addr
))
9118 goto skip_bad_vf_detection
;
9121 /* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
9122 if (!(req_id
& 0x0080))
9123 goto skip_bad_vf_detection
;
9125 pf_func
= req_id
& 0x01;
9126 if ((pf_func
& 1) == (pdev
->devfn
& 1)) {
9127 unsigned int device_id
;
9129 vf
= (req_id
& 0x7F) >> 1;
9130 e_dev_err("VF %d has caused a PCIe error\n", vf
);
9131 e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
9132 "%8.8x\tdw3: %8.8x\n",
9133 dw0
, dw1
, dw2
, dw3
);
9134 switch (adapter
->hw
.mac
.type
) {
9135 case ixgbe_mac_82599EB
:
9136 device_id
= IXGBE_82599_VF_DEVICE_ID
;
9138 case ixgbe_mac_X540
:
9139 device_id
= IXGBE_X540_VF_DEVICE_ID
;
9141 case ixgbe_mac_X550
:
9142 device_id
= IXGBE_DEV_ID_X550_VF
;
9144 case ixgbe_mac_X550EM_x
:
9145 device_id
= IXGBE_DEV_ID_X550EM_X_VF
;
9152 /* Find the pci device of the offending VF */
9153 vfdev
= pci_get_device(PCI_VENDOR_ID_INTEL
, device_id
, NULL
);
9155 if (vfdev
->devfn
== (req_id
& 0xFF))
9157 vfdev
= pci_get_device(PCI_VENDOR_ID_INTEL
,
9161 * There's a slim chance the VF could have been hot plugged,
9162 * so if it is no longer present we don't need to issue the
9163 * VFLR. Just clean up the AER in that case.
9166 ixgbe_issue_vf_flr(adapter
, vfdev
);
9167 /* Free device reference count */
9171 pci_cleanup_aer_uncorrect_error_status(pdev
);
9175 * Even though the error may have occurred on the other port
9176 * we still need to increment the vf error reference count for
9177 * both ports because the I/O resume function will be called
9180 adapter
->vferr_refcount
++;
9182 return PCI_ERS_RESULT_RECOVERED
;
9184 skip_bad_vf_detection
:
9185 #endif /* CONFIG_PCI_IOV */
9186 if (!test_bit(__IXGBE_SERVICE_INITED
, &adapter
->state
))
9187 return PCI_ERS_RESULT_DISCONNECT
;
9190 netif_device_detach(netdev
);
9192 if (state
== pci_channel_io_perm_failure
) {
9194 return PCI_ERS_RESULT_DISCONNECT
;
9197 if (netif_running(netdev
))
9198 ixgbe_down(adapter
);
9200 if (!test_and_set_bit(__IXGBE_DISABLED
, &adapter
->state
))
9201 pci_disable_device(pdev
);
9204 /* Request a slot reset. */
9205 return PCI_ERS_RESULT_NEED_RESET
;
9209 * ixgbe_io_slot_reset - called after the pci bus has been reset.
9210 * @pdev: Pointer to PCI device
9212 * Restart the card from scratch, as if from a cold-boot.
9214 static pci_ers_result_t
ixgbe_io_slot_reset(struct pci_dev
*pdev
)
9216 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
9217 pci_ers_result_t result
;
9220 if (pci_enable_device_mem(pdev
)) {
9221 e_err(probe
, "Cannot re-enable PCI device after reset.\n");
9222 result
= PCI_ERS_RESULT_DISCONNECT
;
9224 smp_mb__before_atomic();
9225 clear_bit(__IXGBE_DISABLED
, &adapter
->state
);
9226 adapter
->hw
.hw_addr
= adapter
->io_addr
;
9227 pci_set_master(pdev
);
9228 pci_restore_state(pdev
);
9229 pci_save_state(pdev
);
9231 pci_wake_from_d3(pdev
, false);
9233 ixgbe_reset(adapter
);
9234 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_WUS
, ~0);
9235 result
= PCI_ERS_RESULT_RECOVERED
;
9238 err
= pci_cleanup_aer_uncorrect_error_status(pdev
);
9240 e_dev_err("pci_cleanup_aer_uncorrect_error_status "
9241 "failed 0x%0x\n", err
);
9242 /* non-fatal, continue */
9249 * ixgbe_io_resume - called when traffic can start flowing again.
9250 * @pdev: Pointer to PCI device
9252 * This callback is called when the error recovery driver tells us that
9253 * its OK to resume normal operation.
9255 static void ixgbe_io_resume(struct pci_dev
*pdev
)
9257 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
9258 struct net_device
*netdev
= adapter
->netdev
;
9260 #ifdef CONFIG_PCI_IOV
9261 if (adapter
->vferr_refcount
) {
9262 e_info(drv
, "Resuming after VF err\n");
9263 adapter
->vferr_refcount
--;
9268 if (netif_running(netdev
))
9271 netif_device_attach(netdev
);
9274 static const struct pci_error_handlers ixgbe_err_handler
= {
9275 .error_detected
= ixgbe_io_error_detected
,
9276 .slot_reset
= ixgbe_io_slot_reset
,
9277 .resume
= ixgbe_io_resume
,
9280 static struct pci_driver ixgbe_driver
= {
9281 .name
= ixgbe_driver_name
,
9282 .id_table
= ixgbe_pci_tbl
,
9283 .probe
= ixgbe_probe
,
9284 .remove
= ixgbe_remove
,
9286 .suspend
= ixgbe_suspend
,
9287 .resume
= ixgbe_resume
,
9289 .shutdown
= ixgbe_shutdown
,
9290 .sriov_configure
= ixgbe_pci_sriov_configure
,
9291 .err_handler
= &ixgbe_err_handler
9295 * ixgbe_init_module - Driver Registration Routine
9297 * ixgbe_init_module is the first routine called when the driver is
9298 * loaded. All it does is register with the PCI subsystem.
9300 static int __init
ixgbe_init_module(void)
9303 pr_info("%s - version %s\n", ixgbe_driver_string
, ixgbe_driver_version
);
9304 pr_info("%s\n", ixgbe_copyright
);
9308 ret
= pci_register_driver(&ixgbe_driver
);
9314 #ifdef CONFIG_IXGBE_DCA
9315 dca_register_notify(&dca_notifier
);
9321 module_init(ixgbe_init_module
);
9324 * ixgbe_exit_module - Driver Exit Cleanup Routine
9326 * ixgbe_exit_module is called just before the driver is removed
9329 static void __exit
ixgbe_exit_module(void)
9331 #ifdef CONFIG_IXGBE_DCA
9332 dca_unregister_notify(&dca_notifier
);
9334 pci_unregister_driver(&ixgbe_driver
);
9339 #ifdef CONFIG_IXGBE_DCA
9340 static int ixgbe_notify_dca(struct notifier_block
*nb
, unsigned long event
,
9345 ret_val
= driver_for_each_device(&ixgbe_driver
.driver
, NULL
, &event
,
9346 __ixgbe_notify_dca
);
9348 return ret_val
? NOTIFY_BAD
: NOTIFY_DONE
;
9351 #endif /* CONFIG_IXGBE_DCA */
9353 module_exit(ixgbe_exit_module
);