1 /*******************************************************************************
3 Intel 10 Gigabit PCI Express Linux driver
4 Copyright(c) 1999 - 2013 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 e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
24 Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
26 *******************************************************************************/
28 #include <linux/types.h>
29 #include <linux/module.h>
30 #include <linux/pci.h>
31 #include <linux/netdevice.h>
32 #include <linux/vmalloc.h>
33 #include <linux/string.h>
35 #include <linux/interrupt.h>
37 #include <linux/tcp.h>
38 #include <linux/sctp.h>
39 #include <linux/pkt_sched.h>
40 #include <linux/ipv6.h>
41 #include <linux/slab.h>
42 #include <net/checksum.h>
43 #include <net/ip6_checksum.h>
44 #include <linux/ethtool.h>
46 #include <linux/if_vlan.h>
47 #include <linux/if_bridge.h>
48 #include <linux/prefetch.h>
49 #include <scsi/fc/fc_fcoe.h>
52 #include "ixgbe_common.h"
53 #include "ixgbe_dcb_82599.h"
54 #include "ixgbe_sriov.h"
56 char ixgbe_driver_name
[] = "ixgbe";
57 static const char ixgbe_driver_string
[] =
58 "Intel(R) 10 Gigabit PCI Express Network Driver";
60 char ixgbe_default_device_descr
[] =
61 "Intel(R) 10 Gigabit Network Connection";
63 static char ixgbe_default_device_descr
[] =
64 "Intel(R) 10 Gigabit Network Connection";
66 #define DRV_VERSION "3.13.10-k"
67 const char ixgbe_driver_version
[] = DRV_VERSION
;
68 static const char ixgbe_copyright
[] =
69 "Copyright (c) 1999-2013 Intel Corporation.";
71 static const struct ixgbe_info
*ixgbe_info_tbl
[] = {
72 [board_82598
] = &ixgbe_82598_info
,
73 [board_82599
] = &ixgbe_82599_info
,
74 [board_X540
] = &ixgbe_X540_info
,
77 /* ixgbe_pci_tbl - PCI Device ID Table
79 * Wildcard entries (PCI_ANY_ID) should come last
80 * Last entry must be all 0s
82 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
83 * Class, Class Mask, private data (not used) }
85 static DEFINE_PCI_DEVICE_TABLE(ixgbe_pci_tbl
) = {
86 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598
), board_82598
},
87 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AF_DUAL_PORT
), board_82598
},
88 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AF_SINGLE_PORT
), board_82598
},
89 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AT
), board_82598
},
90 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AT2
), board_82598
},
91 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598EB_CX4
), board_82598
},
92 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_CX4_DUAL_PORT
), board_82598
},
93 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_DA_DUAL_PORT
), board_82598
},
94 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM
), board_82598
},
95 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598EB_XF_LR
), board_82598
},
96 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598EB_SFP_LOM
), board_82598
},
97 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_BX
), board_82598
},
98 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_KX4
), board_82599
},
99 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_XAUI_LOM
), board_82599
},
100 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_KR
), board_82599
},
101 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP
), board_82599
},
102 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_EM
), board_82599
},
103 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_KX4_MEZZ
), board_82599
},
104 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_CX4
), board_82599
},
105 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_BACKPLANE_FCOE
), board_82599
},
106 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_FCOE
), board_82599
},
107 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_T3_LOM
), board_82599
},
108 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_COMBO_BACKPLANE
), board_82599
},
109 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X540T
), board_X540
},
110 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_SF2
), board_82599
},
111 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_LS
), board_82599
},
112 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599EN_SFP
), board_82599
},
113 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_SF_QP
), board_82599
},
114 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X540T1
), board_X540
},
115 /* required last entry */
118 MODULE_DEVICE_TABLE(pci
, ixgbe_pci_tbl
);
120 #ifdef CONFIG_IXGBE_DCA
121 static int ixgbe_notify_dca(struct notifier_block
*, unsigned long event
,
123 static struct notifier_block dca_notifier
= {
124 .notifier_call
= ixgbe_notify_dca
,
130 #ifdef CONFIG_PCI_IOV
131 static unsigned int max_vfs
;
132 module_param(max_vfs
, uint
, 0);
133 MODULE_PARM_DESC(max_vfs
,
134 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63");
135 #endif /* CONFIG_PCI_IOV */
137 static unsigned int allow_unsupported_sfp
;
138 module_param(allow_unsupported_sfp
, uint
, 0);
139 MODULE_PARM_DESC(allow_unsupported_sfp
,
140 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
142 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
143 static int debug
= -1;
144 module_param(debug
, int, 0);
145 MODULE_PARM_DESC(debug
, "Debug level (0=none,...,16=all)");
147 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
148 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
149 MODULE_LICENSE("GPL");
150 MODULE_VERSION(DRV_VERSION
);
152 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter
*adapter
,
156 struct pci_dev
*parent_dev
;
157 struct pci_bus
*parent_bus
;
159 parent_bus
= adapter
->pdev
->bus
->parent
;
163 parent_dev
= parent_bus
->self
;
167 pos
= pci_find_capability(parent_dev
, PCI_CAP_ID_EXP
);
171 pci_read_config_word(parent_dev
, pos
+ reg
, value
);
175 static s32
ixgbe_get_parent_bus_info(struct ixgbe_adapter
*adapter
)
177 struct ixgbe_hw
*hw
= &adapter
->hw
;
181 hw
->bus
.type
= ixgbe_bus_type_pci_express
;
183 /* Get the negotiated link width and speed from PCI config space of the
184 * parent, as this device is behind a switch
186 err
= ixgbe_read_pci_cfg_word_parent(adapter
, 18, &link_status
);
188 /* assume caller will handle error case */
192 hw
->bus
.width
= ixgbe_convert_bus_width(link_status
);
193 hw
->bus
.speed
= ixgbe_convert_bus_speed(link_status
);
198 static void ixgbe_service_event_schedule(struct ixgbe_adapter
*adapter
)
200 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
) &&
201 !test_and_set_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
))
202 schedule_work(&adapter
->service_task
);
205 static void ixgbe_service_event_complete(struct ixgbe_adapter
*adapter
)
207 BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
));
209 /* flush memory to make sure state is correct before next watchdog */
210 smp_mb__before_clear_bit();
211 clear_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
);
214 struct ixgbe_reg_info
{
219 static const struct ixgbe_reg_info ixgbe_reg_info_tbl
[] = {
221 /* General Registers */
222 {IXGBE_CTRL
, "CTRL"},
223 {IXGBE_STATUS
, "STATUS"},
224 {IXGBE_CTRL_EXT
, "CTRL_EXT"},
226 /* Interrupt Registers */
227 {IXGBE_EICR
, "EICR"},
230 {IXGBE_SRRCTL(0), "SRRCTL"},
231 {IXGBE_DCA_RXCTRL(0), "DRXCTL"},
232 {IXGBE_RDLEN(0), "RDLEN"},
233 {IXGBE_RDH(0), "RDH"},
234 {IXGBE_RDT(0), "RDT"},
235 {IXGBE_RXDCTL(0), "RXDCTL"},
236 {IXGBE_RDBAL(0), "RDBAL"},
237 {IXGBE_RDBAH(0), "RDBAH"},
240 {IXGBE_TDBAL(0), "TDBAL"},
241 {IXGBE_TDBAH(0), "TDBAH"},
242 {IXGBE_TDLEN(0), "TDLEN"},
243 {IXGBE_TDH(0), "TDH"},
244 {IXGBE_TDT(0), "TDT"},
245 {IXGBE_TXDCTL(0), "TXDCTL"},
247 /* List Terminator */
253 * ixgbe_regdump - register printout routine
255 static void ixgbe_regdump(struct ixgbe_hw
*hw
, struct ixgbe_reg_info
*reginfo
)
261 switch (reginfo
->ofs
) {
262 case IXGBE_SRRCTL(0):
263 for (i
= 0; i
< 64; i
++)
264 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_SRRCTL(i
));
266 case IXGBE_DCA_RXCTRL(0):
267 for (i
= 0; i
< 64; i
++)
268 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_DCA_RXCTRL(i
));
271 for (i
= 0; i
< 64; i
++)
272 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDLEN(i
));
275 for (i
= 0; i
< 64; i
++)
276 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDH(i
));
279 for (i
= 0; i
< 64; i
++)
280 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDT(i
));
282 case IXGBE_RXDCTL(0):
283 for (i
= 0; i
< 64; i
++)
284 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RXDCTL(i
));
287 for (i
= 0; i
< 64; i
++)
288 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDBAL(i
));
291 for (i
= 0; i
< 64; i
++)
292 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDBAH(i
));
295 for (i
= 0; i
< 64; i
++)
296 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDBAL(i
));
299 for (i
= 0; i
< 64; i
++)
300 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDBAH(i
));
303 for (i
= 0; i
< 64; i
++)
304 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDLEN(i
));
307 for (i
= 0; i
< 64; i
++)
308 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDH(i
));
311 for (i
= 0; i
< 64; i
++)
312 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDT(i
));
314 case IXGBE_TXDCTL(0):
315 for (i
= 0; i
< 64; i
++)
316 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TXDCTL(i
));
319 pr_info("%-15s %08x\n", reginfo
->name
,
320 IXGBE_READ_REG(hw
, reginfo
->ofs
));
324 for (i
= 0; i
< 8; i
++) {
325 snprintf(rname
, 16, "%s[%d-%d]", reginfo
->name
, i
*8, i
*8+7);
326 pr_err("%-15s", rname
);
327 for (j
= 0; j
< 8; j
++)
328 pr_cont(" %08x", regs
[i
*8+j
]);
335 * ixgbe_dump - Print registers, tx-rings and rx-rings
337 static void ixgbe_dump(struct ixgbe_adapter
*adapter
)
339 struct net_device
*netdev
= adapter
->netdev
;
340 struct ixgbe_hw
*hw
= &adapter
->hw
;
341 struct ixgbe_reg_info
*reginfo
;
343 struct ixgbe_ring
*tx_ring
;
344 struct ixgbe_tx_buffer
*tx_buffer
;
345 union ixgbe_adv_tx_desc
*tx_desc
;
346 struct my_u0
{ u64 a
; u64 b
; } *u0
;
347 struct ixgbe_ring
*rx_ring
;
348 union ixgbe_adv_rx_desc
*rx_desc
;
349 struct ixgbe_rx_buffer
*rx_buffer_info
;
353 if (!netif_msg_hw(adapter
))
356 /* Print netdevice Info */
358 dev_info(&adapter
->pdev
->dev
, "Net device Info\n");
359 pr_info("Device Name state "
360 "trans_start last_rx\n");
361 pr_info("%-15s %016lX %016lX %016lX\n",
368 /* Print Registers */
369 dev_info(&adapter
->pdev
->dev
, "Register Dump\n");
370 pr_info(" Register Name Value\n");
371 for (reginfo
= (struct ixgbe_reg_info
*)ixgbe_reg_info_tbl
;
372 reginfo
->name
; reginfo
++) {
373 ixgbe_regdump(hw
, reginfo
);
376 /* Print TX Ring Summary */
377 if (!netdev
|| !netif_running(netdev
))
380 dev_info(&adapter
->pdev
->dev
, "TX Rings Summary\n");
381 pr_info(" %s %s %s %s\n",
382 "Queue [NTU] [NTC] [bi(ntc)->dma ]",
383 "leng", "ntw", "timestamp");
384 for (n
= 0; n
< adapter
->num_tx_queues
; n
++) {
385 tx_ring
= adapter
->tx_ring
[n
];
386 tx_buffer
= &tx_ring
->tx_buffer_info
[tx_ring
->next_to_clean
];
387 pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
388 n
, tx_ring
->next_to_use
, tx_ring
->next_to_clean
,
389 (u64
)dma_unmap_addr(tx_buffer
, dma
),
390 dma_unmap_len(tx_buffer
, len
),
391 tx_buffer
->next_to_watch
,
392 (u64
)tx_buffer
->time_stamp
);
396 if (!netif_msg_tx_done(adapter
))
397 goto rx_ring_summary
;
399 dev_info(&adapter
->pdev
->dev
, "TX Rings Dump\n");
401 /* Transmit Descriptor Formats
403 * 82598 Advanced Transmit Descriptor
404 * +--------------------------------------------------------------+
405 * 0 | Buffer Address [63:0] |
406 * +--------------------------------------------------------------+
407 * 8 | PAYLEN | POPTS | IDX | STA | DCMD |DTYP | RSV | DTALEN |
408 * +--------------------------------------------------------------+
409 * 63 46 45 40 39 36 35 32 31 24 23 20 19 0
411 * 82598 Advanced Transmit Descriptor (Write-Back Format)
412 * +--------------------------------------------------------------+
414 * +--------------------------------------------------------------+
415 * 8 | RSV | STA | NXTSEQ |
416 * +--------------------------------------------------------------+
419 * 82599+ Advanced Transmit Descriptor
420 * +--------------------------------------------------------------+
421 * 0 | Buffer Address [63:0] |
422 * +--------------------------------------------------------------+
423 * 8 |PAYLEN |POPTS|CC|IDX |STA |DCMD |DTYP |MAC |RSV |DTALEN |
424 * +--------------------------------------------------------------+
425 * 63 46 45 40 39 38 36 35 32 31 24 23 20 19 18 17 16 15 0
427 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
428 * +--------------------------------------------------------------+
430 * +--------------------------------------------------------------+
431 * 8 | RSV | STA | RSV |
432 * +--------------------------------------------------------------+
436 for (n
= 0; n
< adapter
->num_tx_queues
; n
++) {
437 tx_ring
= adapter
->tx_ring
[n
];
438 pr_info("------------------------------------\n");
439 pr_info("TX QUEUE INDEX = %d\n", tx_ring
->queue_index
);
440 pr_info("------------------------------------\n");
441 pr_info("%s%s %s %s %s %s\n",
442 "T [desc] [address 63:0 ] ",
443 "[PlPOIdStDDt Ln] [bi->dma ] ",
444 "leng", "ntw", "timestamp", "bi->skb");
446 for (i
= 0; tx_ring
->desc
&& (i
< tx_ring
->count
); i
++) {
447 tx_desc
= IXGBE_TX_DESC(tx_ring
, i
);
448 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
449 u0
= (struct my_u0
*)tx_desc
;
450 if (dma_unmap_len(tx_buffer
, len
) > 0) {
451 pr_info("T [0x%03X] %016llX %016llX %016llX %08X %p %016llX %p",
455 (u64
)dma_unmap_addr(tx_buffer
, dma
),
456 dma_unmap_len(tx_buffer
, len
),
457 tx_buffer
->next_to_watch
,
458 (u64
)tx_buffer
->time_stamp
,
460 if (i
== tx_ring
->next_to_use
&&
461 i
== tx_ring
->next_to_clean
)
463 else if (i
== tx_ring
->next_to_use
)
465 else if (i
== tx_ring
->next_to_clean
)
470 if (netif_msg_pktdata(adapter
) &&
472 print_hex_dump(KERN_INFO
, "",
473 DUMP_PREFIX_ADDRESS
, 16, 1,
474 tx_buffer
->skb
->data
,
475 dma_unmap_len(tx_buffer
, len
),
481 /* Print RX Rings Summary */
483 dev_info(&adapter
->pdev
->dev
, "RX Rings Summary\n");
484 pr_info("Queue [NTU] [NTC]\n");
485 for (n
= 0; n
< adapter
->num_rx_queues
; n
++) {
486 rx_ring
= adapter
->rx_ring
[n
];
487 pr_info("%5d %5X %5X\n",
488 n
, rx_ring
->next_to_use
, rx_ring
->next_to_clean
);
492 if (!netif_msg_rx_status(adapter
))
495 dev_info(&adapter
->pdev
->dev
, "RX Rings Dump\n");
497 /* Receive Descriptor Formats
499 * 82598 Advanced Receive Descriptor (Read) Format
501 * +-----------------------------------------------------+
502 * 0 | Packet Buffer Address [63:1] |A0/NSE|
503 * +----------------------------------------------+------+
504 * 8 | Header Buffer Address [63:1] | DD |
505 * +-----------------------------------------------------+
508 * 82598 Advanced Receive Descriptor (Write-Back) Format
510 * 63 48 47 32 31 30 21 20 16 15 4 3 0
511 * +------------------------------------------------------+
512 * 0 | RSS Hash / |SPH| HDR_LEN | RSV |Packet| RSS |
513 * | Packet | IP | | | | Type | Type |
514 * | Checksum | Ident | | | | | |
515 * +------------------------------------------------------+
516 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
517 * +------------------------------------------------------+
518 * 63 48 47 32 31 20 19 0
520 * 82599+ Advanced Receive Descriptor (Read) Format
522 * +-----------------------------------------------------+
523 * 0 | Packet Buffer Address [63:1] |A0/NSE|
524 * +----------------------------------------------+------+
525 * 8 | Header Buffer Address [63:1] | DD |
526 * +-----------------------------------------------------+
529 * 82599+ Advanced Receive Descriptor (Write-Back) Format
531 * 63 48 47 32 31 30 21 20 17 16 4 3 0
532 * +------------------------------------------------------+
533 * 0 |RSS / Frag Checksum|SPH| HDR_LEN |RSC- |Packet| RSS |
534 * |/ RTT / PCoE_PARAM | | | CNT | Type | Type |
535 * |/ Flow Dir Flt ID | | | | | |
536 * +------------------------------------------------------+
537 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
538 * +------------------------------------------------------+
539 * 63 48 47 32 31 20 19 0
542 for (n
= 0; n
< adapter
->num_rx_queues
; n
++) {
543 rx_ring
= adapter
->rx_ring
[n
];
544 pr_info("------------------------------------\n");
545 pr_info("RX QUEUE INDEX = %d\n", rx_ring
->queue_index
);
546 pr_info("------------------------------------\n");
548 "R [desc] [ PktBuf A0] ",
549 "[ HeadBuf DD] [bi->dma ] [bi->skb ] ",
550 "<-- Adv Rx Read format\n");
552 "RWB[desc] [PcsmIpSHl PtRs] ",
553 "[vl er S cks ln] ---------------- [bi->skb ] ",
554 "<-- Adv Rx Write-Back format\n");
556 for (i
= 0; i
< rx_ring
->count
; i
++) {
557 rx_buffer_info
= &rx_ring
->rx_buffer_info
[i
];
558 rx_desc
= IXGBE_RX_DESC(rx_ring
, i
);
559 u0
= (struct my_u0
*)rx_desc
;
560 staterr
= le32_to_cpu(rx_desc
->wb
.upper
.status_error
);
561 if (staterr
& IXGBE_RXD_STAT_DD
) {
562 /* Descriptor Done */
563 pr_info("RWB[0x%03X] %016llX "
564 "%016llX ---------------- %p", i
,
567 rx_buffer_info
->skb
);
569 pr_info("R [0x%03X] %016llX "
570 "%016llX %016llX %p", i
,
573 (u64
)rx_buffer_info
->dma
,
574 rx_buffer_info
->skb
);
576 if (netif_msg_pktdata(adapter
) &&
577 rx_buffer_info
->dma
) {
578 print_hex_dump(KERN_INFO
, "",
579 DUMP_PREFIX_ADDRESS
, 16, 1,
580 page_address(rx_buffer_info
->page
) +
581 rx_buffer_info
->page_offset
,
582 ixgbe_rx_bufsz(rx_ring
), true);
586 if (i
== rx_ring
->next_to_use
)
588 else if (i
== rx_ring
->next_to_clean
)
600 static void ixgbe_release_hw_control(struct ixgbe_adapter
*adapter
)
604 /* Let firmware take over control of h/w */
605 ctrl_ext
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_CTRL_EXT
);
606 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_CTRL_EXT
,
607 ctrl_ext
& ~IXGBE_CTRL_EXT_DRV_LOAD
);
610 static void ixgbe_get_hw_control(struct ixgbe_adapter
*adapter
)
614 /* Let firmware know the driver has taken over */
615 ctrl_ext
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_CTRL_EXT
);
616 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_CTRL_EXT
,
617 ctrl_ext
| IXGBE_CTRL_EXT_DRV_LOAD
);
621 * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
622 * @adapter: pointer to adapter struct
623 * @direction: 0 for Rx, 1 for Tx, -1 for other causes
624 * @queue: queue to map the corresponding interrupt to
625 * @msix_vector: the vector to map to the corresponding queue
628 static void ixgbe_set_ivar(struct ixgbe_adapter
*adapter
, s8 direction
,
629 u8 queue
, u8 msix_vector
)
632 struct ixgbe_hw
*hw
= &adapter
->hw
;
633 switch (hw
->mac
.type
) {
634 case ixgbe_mac_82598EB
:
635 msix_vector
|= IXGBE_IVAR_ALLOC_VAL
;
638 index
= (((direction
* 64) + queue
) >> 2) & 0x1F;
639 ivar
= IXGBE_READ_REG(hw
, IXGBE_IVAR(index
));
640 ivar
&= ~(0xFF << (8 * (queue
& 0x3)));
641 ivar
|= (msix_vector
<< (8 * (queue
& 0x3)));
642 IXGBE_WRITE_REG(hw
, IXGBE_IVAR(index
), ivar
);
644 case ixgbe_mac_82599EB
:
646 if (direction
== -1) {
648 msix_vector
|= IXGBE_IVAR_ALLOC_VAL
;
649 index
= ((queue
& 1) * 8);
650 ivar
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_IVAR_MISC
);
651 ivar
&= ~(0xFF << index
);
652 ivar
|= (msix_vector
<< index
);
653 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_IVAR_MISC
, ivar
);
656 /* tx or rx causes */
657 msix_vector
|= IXGBE_IVAR_ALLOC_VAL
;
658 index
= ((16 * (queue
& 1)) + (8 * direction
));
659 ivar
= IXGBE_READ_REG(hw
, IXGBE_IVAR(queue
>> 1));
660 ivar
&= ~(0xFF << index
);
661 ivar
|= (msix_vector
<< index
);
662 IXGBE_WRITE_REG(hw
, IXGBE_IVAR(queue
>> 1), ivar
);
670 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter
*adapter
,
675 switch (adapter
->hw
.mac
.type
) {
676 case ixgbe_mac_82598EB
:
677 mask
= (IXGBE_EIMS_RTX_QUEUE
& qmask
);
678 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EICS
, mask
);
680 case ixgbe_mac_82599EB
:
682 mask
= (qmask
& 0xFFFFFFFF);
683 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EICS_EX(0), mask
);
684 mask
= (qmask
>> 32);
685 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EICS_EX(1), mask
);
692 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring
*ring
,
693 struct ixgbe_tx_buffer
*tx_buffer
)
695 if (tx_buffer
->skb
) {
696 dev_kfree_skb_any(tx_buffer
->skb
);
697 if (dma_unmap_len(tx_buffer
, len
))
698 dma_unmap_single(ring
->dev
,
699 dma_unmap_addr(tx_buffer
, dma
),
700 dma_unmap_len(tx_buffer
, len
),
702 } else if (dma_unmap_len(tx_buffer
, len
)) {
703 dma_unmap_page(ring
->dev
,
704 dma_unmap_addr(tx_buffer
, dma
),
705 dma_unmap_len(tx_buffer
, len
),
708 tx_buffer
->next_to_watch
= NULL
;
709 tx_buffer
->skb
= NULL
;
710 dma_unmap_len_set(tx_buffer
, len
, 0);
711 /* tx_buffer must be completely set up in the transmit path */
714 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter
*adapter
)
716 struct ixgbe_hw
*hw
= &adapter
->hw
;
717 struct ixgbe_hw_stats
*hwstats
= &adapter
->stats
;
721 if ((hw
->fc
.current_mode
!= ixgbe_fc_full
) &&
722 (hw
->fc
.current_mode
!= ixgbe_fc_rx_pause
))
725 switch (hw
->mac
.type
) {
726 case ixgbe_mac_82598EB
:
727 data
= IXGBE_READ_REG(hw
, IXGBE_LXOFFRXC
);
730 data
= IXGBE_READ_REG(hw
, IXGBE_LXOFFRXCNT
);
732 hwstats
->lxoffrxc
+= data
;
734 /* refill credits (no tx hang) if we received xoff */
738 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
739 clear_bit(__IXGBE_HANG_CHECK_ARMED
,
740 &adapter
->tx_ring
[i
]->state
);
743 static void ixgbe_update_xoff_received(struct ixgbe_adapter
*adapter
)
745 struct ixgbe_hw
*hw
= &adapter
->hw
;
746 struct ixgbe_hw_stats
*hwstats
= &adapter
->stats
;
750 bool pfc_en
= adapter
->dcb_cfg
.pfc_mode_enable
;
752 if (adapter
->ixgbe_ieee_pfc
)
753 pfc_en
|= !!(adapter
->ixgbe_ieee_pfc
->pfc_en
);
755 if (!(adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
) || !pfc_en
) {
756 ixgbe_update_xoff_rx_lfc(adapter
);
760 /* update stats for each tc, only valid with PFC enabled */
761 for (i
= 0; i
< MAX_TX_PACKET_BUFFERS
; i
++) {
764 switch (hw
->mac
.type
) {
765 case ixgbe_mac_82598EB
:
766 pxoffrxc
= IXGBE_READ_REG(hw
, IXGBE_PXOFFRXC(i
));
769 pxoffrxc
= IXGBE_READ_REG(hw
, IXGBE_PXOFFRXCNT(i
));
771 hwstats
->pxoffrxc
[i
] += pxoffrxc
;
772 /* Get the TC for given UP */
773 tc
= netdev_get_prio_tc_map(adapter
->netdev
, i
);
774 xoff
[tc
] += pxoffrxc
;
777 /* disarm tx queues that have received xoff frames */
778 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
779 struct ixgbe_ring
*tx_ring
= adapter
->tx_ring
[i
];
781 tc
= tx_ring
->dcb_tc
;
783 clear_bit(__IXGBE_HANG_CHECK_ARMED
, &tx_ring
->state
);
787 static u64
ixgbe_get_tx_completed(struct ixgbe_ring
*ring
)
789 return ring
->stats
.packets
;
792 static u64
ixgbe_get_tx_pending(struct ixgbe_ring
*ring
)
794 struct ixgbe_adapter
*adapter
= netdev_priv(ring
->netdev
);
795 struct ixgbe_hw
*hw
= &adapter
->hw
;
797 u32 head
= IXGBE_READ_REG(hw
, IXGBE_TDH(ring
->reg_idx
));
798 u32 tail
= IXGBE_READ_REG(hw
, IXGBE_TDT(ring
->reg_idx
));
801 return (head
< tail
) ?
802 tail
- head
: (tail
+ ring
->count
- head
);
807 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring
*tx_ring
)
809 u32 tx_done
= ixgbe_get_tx_completed(tx_ring
);
810 u32 tx_done_old
= tx_ring
->tx_stats
.tx_done_old
;
811 u32 tx_pending
= ixgbe_get_tx_pending(tx_ring
);
814 clear_check_for_tx_hang(tx_ring
);
817 * Check for a hung queue, but be thorough. This verifies
818 * that a transmit has been completed since the previous
819 * check AND there is at least one packet pending. The
820 * ARMED bit is set to indicate a potential hang. The
821 * bit is cleared if a pause frame is received to remove
822 * false hang detection due to PFC or 802.3x frames. By
823 * requiring this to fail twice we avoid races with
824 * pfc clearing the ARMED bit and conditions where we
825 * run the check_tx_hang logic with a transmit completion
826 * pending but without time to complete it yet.
828 if ((tx_done_old
== tx_done
) && tx_pending
) {
829 /* make sure it is true for two checks in a row */
830 ret
= test_and_set_bit(__IXGBE_HANG_CHECK_ARMED
,
833 /* update completed stats and continue */
834 tx_ring
->tx_stats
.tx_done_old
= tx_done
;
835 /* reset the countdown */
836 clear_bit(__IXGBE_HANG_CHECK_ARMED
, &tx_ring
->state
);
843 * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
844 * @adapter: driver private struct
846 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter
*adapter
)
849 /* Do the reset outside of interrupt context */
850 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
851 adapter
->flags2
|= IXGBE_FLAG2_RESET_REQUESTED
;
852 e_warn(drv
, "initiating reset due to tx timeout\n");
853 ixgbe_service_event_schedule(adapter
);
858 * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
859 * @q_vector: structure containing interrupt and ring information
860 * @tx_ring: tx ring to clean
862 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector
*q_vector
,
863 struct ixgbe_ring
*tx_ring
)
865 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
866 struct ixgbe_tx_buffer
*tx_buffer
;
867 union ixgbe_adv_tx_desc
*tx_desc
;
868 unsigned int total_bytes
= 0, total_packets
= 0;
869 unsigned int budget
= q_vector
->tx
.work_limit
;
870 unsigned int i
= tx_ring
->next_to_clean
;
872 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
875 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
876 tx_desc
= IXGBE_TX_DESC(tx_ring
, i
);
880 union ixgbe_adv_tx_desc
*eop_desc
= tx_buffer
->next_to_watch
;
882 /* if next_to_watch is not set then there is no work pending */
886 /* prevent any other reads prior to eop_desc */
887 read_barrier_depends();
889 /* if DD is not set pending work has not been completed */
890 if (!(eop_desc
->wb
.status
& cpu_to_le32(IXGBE_TXD_STAT_DD
)))
893 /* clear next_to_watch to prevent false hangs */
894 tx_buffer
->next_to_watch
= NULL
;
896 /* update the statistics for this packet */
897 total_bytes
+= tx_buffer
->bytecount
;
898 total_packets
+= tx_buffer
->gso_segs
;
901 dev_kfree_skb_any(tx_buffer
->skb
);
903 /* unmap skb header data */
904 dma_unmap_single(tx_ring
->dev
,
905 dma_unmap_addr(tx_buffer
, dma
),
906 dma_unmap_len(tx_buffer
, len
),
909 /* clear tx_buffer data */
910 tx_buffer
->skb
= NULL
;
911 dma_unmap_len_set(tx_buffer
, len
, 0);
913 /* unmap remaining buffers */
914 while (tx_desc
!= eop_desc
) {
920 tx_buffer
= tx_ring
->tx_buffer_info
;
921 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
924 /* unmap any remaining paged data */
925 if (dma_unmap_len(tx_buffer
, len
)) {
926 dma_unmap_page(tx_ring
->dev
,
927 dma_unmap_addr(tx_buffer
, dma
),
928 dma_unmap_len(tx_buffer
, len
),
930 dma_unmap_len_set(tx_buffer
, len
, 0);
934 /* move us one more past the eop_desc for start of next pkt */
940 tx_buffer
= tx_ring
->tx_buffer_info
;
941 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
944 /* issue prefetch for next Tx descriptor */
947 /* update budget accounting */
949 } while (likely(budget
));
952 tx_ring
->next_to_clean
= i
;
953 u64_stats_update_begin(&tx_ring
->syncp
);
954 tx_ring
->stats
.bytes
+= total_bytes
;
955 tx_ring
->stats
.packets
+= total_packets
;
956 u64_stats_update_end(&tx_ring
->syncp
);
957 q_vector
->tx
.total_bytes
+= total_bytes
;
958 q_vector
->tx
.total_packets
+= total_packets
;
960 if (check_for_tx_hang(tx_ring
) && ixgbe_check_tx_hang(tx_ring
)) {
961 /* schedule immediate reset if we believe we hung */
962 struct ixgbe_hw
*hw
= &adapter
->hw
;
963 e_err(drv
, "Detected Tx Unit Hang\n"
965 " TDH, TDT <%x>, <%x>\n"
966 " next_to_use <%x>\n"
967 " next_to_clean <%x>\n"
968 "tx_buffer_info[next_to_clean]\n"
969 " time_stamp <%lx>\n"
971 tx_ring
->queue_index
,
972 IXGBE_READ_REG(hw
, IXGBE_TDH(tx_ring
->reg_idx
)),
973 IXGBE_READ_REG(hw
, IXGBE_TDT(tx_ring
->reg_idx
)),
974 tx_ring
->next_to_use
, i
,
975 tx_ring
->tx_buffer_info
[i
].time_stamp
, jiffies
);
977 netif_stop_subqueue(tx_ring
->netdev
, tx_ring
->queue_index
);
980 "tx hang %d detected on queue %d, resetting adapter\n",
981 adapter
->tx_timeout_count
+ 1, tx_ring
->queue_index
);
983 /* schedule immediate reset if we believe we hung */
984 ixgbe_tx_timeout_reset(adapter
);
986 /* the adapter is about to reset, no point in enabling stuff */
990 netdev_tx_completed_queue(txring_txq(tx_ring
),
991 total_packets
, total_bytes
);
993 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
994 if (unlikely(total_packets
&& netif_carrier_ok(tx_ring
->netdev
) &&
995 (ixgbe_desc_unused(tx_ring
) >= TX_WAKE_THRESHOLD
))) {
996 /* Make sure that anybody stopping the queue after this
997 * sees the new next_to_clean.
1000 if (__netif_subqueue_stopped(tx_ring
->netdev
,
1001 tx_ring
->queue_index
)
1002 && !test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
1003 netif_wake_subqueue(tx_ring
->netdev
,
1004 tx_ring
->queue_index
);
1005 ++tx_ring
->tx_stats
.restart_queue
;
1012 #ifdef CONFIG_IXGBE_DCA
1013 static void ixgbe_update_tx_dca(struct ixgbe_adapter
*adapter
,
1014 struct ixgbe_ring
*tx_ring
,
1017 struct ixgbe_hw
*hw
= &adapter
->hw
;
1018 u32 txctrl
= dca3_get_tag(tx_ring
->dev
, cpu
);
1021 switch (hw
->mac
.type
) {
1022 case ixgbe_mac_82598EB
:
1023 reg_offset
= IXGBE_DCA_TXCTRL(tx_ring
->reg_idx
);
1025 case ixgbe_mac_82599EB
:
1026 case ixgbe_mac_X540
:
1027 reg_offset
= IXGBE_DCA_TXCTRL_82599(tx_ring
->reg_idx
);
1028 txctrl
<<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599
;
1031 /* for unknown hardware do not write register */
1036 * We can enable relaxed ordering for reads, but not writes when
1037 * DCA is enabled. This is due to a known issue in some chipsets
1038 * which will cause the DCA tag to be cleared.
1040 txctrl
|= IXGBE_DCA_TXCTRL_DESC_RRO_EN
|
1041 IXGBE_DCA_TXCTRL_DATA_RRO_EN
|
1042 IXGBE_DCA_TXCTRL_DESC_DCA_EN
;
1044 IXGBE_WRITE_REG(hw
, reg_offset
, txctrl
);
1047 static void ixgbe_update_rx_dca(struct ixgbe_adapter
*adapter
,
1048 struct ixgbe_ring
*rx_ring
,
1051 struct ixgbe_hw
*hw
= &adapter
->hw
;
1052 u32 rxctrl
= dca3_get_tag(rx_ring
->dev
, cpu
);
1053 u8 reg_idx
= rx_ring
->reg_idx
;
1056 switch (hw
->mac
.type
) {
1057 case ixgbe_mac_82599EB
:
1058 case ixgbe_mac_X540
:
1059 rxctrl
<<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599
;
1066 * We can enable relaxed ordering for reads, but not writes when
1067 * DCA is enabled. This is due to a known issue in some chipsets
1068 * which will cause the DCA tag to be cleared.
1070 rxctrl
|= IXGBE_DCA_RXCTRL_DESC_RRO_EN
|
1071 IXGBE_DCA_RXCTRL_DESC_DCA_EN
;
1073 IXGBE_WRITE_REG(hw
, IXGBE_DCA_RXCTRL(reg_idx
), rxctrl
);
1076 static void ixgbe_update_dca(struct ixgbe_q_vector
*q_vector
)
1078 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
1079 struct ixgbe_ring
*ring
;
1080 int cpu
= get_cpu();
1082 if (q_vector
->cpu
== cpu
)
1085 ixgbe_for_each_ring(ring
, q_vector
->tx
)
1086 ixgbe_update_tx_dca(adapter
, ring
, cpu
);
1088 ixgbe_for_each_ring(ring
, q_vector
->rx
)
1089 ixgbe_update_rx_dca(adapter
, ring
, cpu
);
1091 q_vector
->cpu
= cpu
;
1096 static void ixgbe_setup_dca(struct ixgbe_adapter
*adapter
)
1100 if (!(adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
))
1103 /* always use CB2 mode, difference is masked in the CB driver */
1104 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_DCA_CTRL
, 2);
1106 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
1107 adapter
->q_vector
[i
]->cpu
= -1;
1108 ixgbe_update_dca(adapter
->q_vector
[i
]);
1112 static int __ixgbe_notify_dca(struct device
*dev
, void *data
)
1114 struct ixgbe_adapter
*adapter
= dev_get_drvdata(dev
);
1115 unsigned long event
= *(unsigned long *)data
;
1117 if (!(adapter
->flags
& IXGBE_FLAG_DCA_CAPABLE
))
1121 case DCA_PROVIDER_ADD
:
1122 /* if we're already enabled, don't do it again */
1123 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
)
1125 if (dca_add_requester(dev
) == 0) {
1126 adapter
->flags
|= IXGBE_FLAG_DCA_ENABLED
;
1127 ixgbe_setup_dca(adapter
);
1130 /* Fall Through since DCA is disabled. */
1131 case DCA_PROVIDER_REMOVE
:
1132 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
) {
1133 dca_remove_requester(dev
);
1134 adapter
->flags
&= ~IXGBE_FLAG_DCA_ENABLED
;
1135 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_DCA_CTRL
, 1);
1143 #endif /* CONFIG_IXGBE_DCA */
1144 static inline void ixgbe_rx_hash(struct ixgbe_ring
*ring
,
1145 union ixgbe_adv_rx_desc
*rx_desc
,
1146 struct sk_buff
*skb
)
1148 if (ring
->netdev
->features
& NETIF_F_RXHASH
)
1149 skb
->rxhash
= le32_to_cpu(rx_desc
->wb
.lower
.hi_dword
.rss
);
1154 * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1155 * @ring: structure containing ring specific data
1156 * @rx_desc: advanced rx descriptor
1158 * Returns : true if it is FCoE pkt
1160 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring
*ring
,
1161 union ixgbe_adv_rx_desc
*rx_desc
)
1163 __le16 pkt_info
= rx_desc
->wb
.lower
.lo_dword
.hs_rss
.pkt_info
;
1165 return test_bit(__IXGBE_RX_FCOE
, &ring
->state
) &&
1166 ((pkt_info
& cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK
)) ==
1167 (cpu_to_le16(IXGBE_ETQF_FILTER_FCOE
<<
1168 IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT
)));
1171 #endif /* IXGBE_FCOE */
1173 * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1174 * @ring: structure containing ring specific data
1175 * @rx_desc: current Rx descriptor being processed
1176 * @skb: skb currently being received and modified
1178 static inline void ixgbe_rx_checksum(struct ixgbe_ring
*ring
,
1179 union ixgbe_adv_rx_desc
*rx_desc
,
1180 struct sk_buff
*skb
)
1182 skb_checksum_none_assert(skb
);
1184 /* Rx csum disabled */
1185 if (!(ring
->netdev
->features
& NETIF_F_RXCSUM
))
1188 /* if IP and error */
1189 if (ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_IPCS
) &&
1190 ixgbe_test_staterr(rx_desc
, IXGBE_RXDADV_ERR_IPE
)) {
1191 ring
->rx_stats
.csum_err
++;
1195 if (!ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_L4CS
))
1198 if (ixgbe_test_staterr(rx_desc
, IXGBE_RXDADV_ERR_TCPE
)) {
1199 __le16 pkt_info
= rx_desc
->wb
.lower
.lo_dword
.hs_rss
.pkt_info
;
1202 * 82599 errata, UDP frames with a 0 checksum can be marked as
1205 if ((pkt_info
& cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP
)) &&
1206 test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR
, &ring
->state
))
1209 ring
->rx_stats
.csum_err
++;
1213 /* It must be a TCP or UDP packet with a valid checksum */
1214 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1217 static inline void ixgbe_release_rx_desc(struct ixgbe_ring
*rx_ring
, u32 val
)
1219 rx_ring
->next_to_use
= val
;
1221 /* update next to alloc since we have filled the ring */
1222 rx_ring
->next_to_alloc
= val
;
1224 * Force memory writes to complete before letting h/w
1225 * know there are new descriptors to fetch. (Only
1226 * applicable for weak-ordered memory model archs,
1230 writel(val
, rx_ring
->tail
);
1233 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring
*rx_ring
,
1234 struct ixgbe_rx_buffer
*bi
)
1236 struct page
*page
= bi
->page
;
1237 dma_addr_t dma
= bi
->dma
;
1239 /* since we are recycling buffers we should seldom need to alloc */
1243 /* alloc new page for storage */
1244 if (likely(!page
)) {
1245 page
= __skb_alloc_pages(GFP_ATOMIC
| __GFP_COLD
| __GFP_COMP
,
1246 bi
->skb
, ixgbe_rx_pg_order(rx_ring
));
1247 if (unlikely(!page
)) {
1248 rx_ring
->rx_stats
.alloc_rx_page_failed
++;
1254 /* map page for use */
1255 dma
= dma_map_page(rx_ring
->dev
, page
, 0,
1256 ixgbe_rx_pg_size(rx_ring
), DMA_FROM_DEVICE
);
1259 * if mapping failed free memory back to system since
1260 * there isn't much point in holding memory we can't use
1262 if (dma_mapping_error(rx_ring
->dev
, dma
)) {
1263 __free_pages(page
, ixgbe_rx_pg_order(rx_ring
));
1266 rx_ring
->rx_stats
.alloc_rx_page_failed
++;
1271 bi
->page_offset
= 0;
1277 * ixgbe_alloc_rx_buffers - Replace used receive buffers
1278 * @rx_ring: ring to place buffers on
1279 * @cleaned_count: number of buffers to replace
1281 void ixgbe_alloc_rx_buffers(struct ixgbe_ring
*rx_ring
, u16 cleaned_count
)
1283 union ixgbe_adv_rx_desc
*rx_desc
;
1284 struct ixgbe_rx_buffer
*bi
;
1285 u16 i
= rx_ring
->next_to_use
;
1291 rx_desc
= IXGBE_RX_DESC(rx_ring
, i
);
1292 bi
= &rx_ring
->rx_buffer_info
[i
];
1293 i
-= rx_ring
->count
;
1296 if (!ixgbe_alloc_mapped_page(rx_ring
, bi
))
1300 * Refresh the desc even if buffer_addrs didn't change
1301 * because each write-back erases this info.
1303 rx_desc
->read
.pkt_addr
= cpu_to_le64(bi
->dma
+ bi
->page_offset
);
1309 rx_desc
= IXGBE_RX_DESC(rx_ring
, 0);
1310 bi
= rx_ring
->rx_buffer_info
;
1311 i
-= rx_ring
->count
;
1314 /* clear the hdr_addr for the next_to_use descriptor */
1315 rx_desc
->read
.hdr_addr
= 0;
1318 } while (cleaned_count
);
1320 i
+= rx_ring
->count
;
1322 if (rx_ring
->next_to_use
!= i
)
1323 ixgbe_release_rx_desc(rx_ring
, i
);
1327 * ixgbe_get_headlen - determine size of header for RSC/LRO/GRO/FCOE
1328 * @data: pointer to the start of the headers
1329 * @max_len: total length of section to find headers in
1331 * This function is meant to determine the length of headers that will
1332 * be recognized by hardware for LRO, GRO, and RSC offloads. The main
1333 * motivation of doing this is to only perform one pull for IPv4 TCP
1334 * packets so that we can do basic things like calculating the gso_size
1335 * based on the average data per packet.
1337 static unsigned int ixgbe_get_headlen(unsigned char *data
,
1338 unsigned int max_len
)
1341 unsigned char *network
;
1344 struct vlan_hdr
*vlan
;
1347 struct ipv6hdr
*ipv6
;
1350 u8 nexthdr
= 0; /* default to not TCP */
1353 /* this should never happen, but better safe than sorry */
1354 if (max_len
< ETH_HLEN
)
1357 /* initialize network frame pointer */
1360 /* set first protocol and move network header forward */
1361 protocol
= hdr
.eth
->h_proto
;
1362 hdr
.network
+= ETH_HLEN
;
1364 /* handle any vlan tag if present */
1365 if (protocol
== __constant_htons(ETH_P_8021Q
)) {
1366 if ((hdr
.network
- data
) > (max_len
- VLAN_HLEN
))
1369 protocol
= hdr
.vlan
->h_vlan_encapsulated_proto
;
1370 hdr
.network
+= VLAN_HLEN
;
1373 /* handle L3 protocols */
1374 if (protocol
== __constant_htons(ETH_P_IP
)) {
1375 if ((hdr
.network
- data
) > (max_len
- sizeof(struct iphdr
)))
1378 /* access ihl as a u8 to avoid unaligned access on ia64 */
1379 hlen
= (hdr
.network
[0] & 0x0F) << 2;
1381 /* verify hlen meets minimum size requirements */
1382 if (hlen
< sizeof(struct iphdr
))
1383 return hdr
.network
- data
;
1385 /* record next protocol if header is present */
1386 if (!(hdr
.ipv4
->frag_off
& htons(IP_OFFSET
)))
1387 nexthdr
= hdr
.ipv4
->protocol
;
1388 } else if (protocol
== __constant_htons(ETH_P_IPV6
)) {
1389 if ((hdr
.network
- data
) > (max_len
- sizeof(struct ipv6hdr
)))
1392 /* record next protocol */
1393 nexthdr
= hdr
.ipv6
->nexthdr
;
1394 hlen
= sizeof(struct ipv6hdr
);
1396 } else if (protocol
== __constant_htons(ETH_P_FCOE
)) {
1397 if ((hdr
.network
- data
) > (max_len
- FCOE_HEADER_LEN
))
1399 hlen
= FCOE_HEADER_LEN
;
1402 return hdr
.network
- data
;
1405 /* relocate pointer to start of L4 header */
1406 hdr
.network
+= hlen
;
1408 /* finally sort out TCP/UDP */
1409 if (nexthdr
== IPPROTO_TCP
) {
1410 if ((hdr
.network
- data
) > (max_len
- sizeof(struct tcphdr
)))
1413 /* access doff as a u8 to avoid unaligned access on ia64 */
1414 hlen
= (hdr
.network
[12] & 0xF0) >> 2;
1416 /* verify hlen meets minimum size requirements */
1417 if (hlen
< sizeof(struct tcphdr
))
1418 return hdr
.network
- data
;
1420 hdr
.network
+= hlen
;
1421 } else if (nexthdr
== IPPROTO_UDP
) {
1422 if ((hdr
.network
- data
) > (max_len
- sizeof(struct udphdr
)))
1425 hdr
.network
+= sizeof(struct udphdr
);
1429 * If everything has gone correctly hdr.network should be the
1430 * data section of the packet and will be the end of the header.
1431 * If not then it probably represents the end of the last recognized
1434 if ((hdr
.network
- data
) < max_len
)
1435 return hdr
.network
- data
;
1440 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring
*ring
,
1441 struct sk_buff
*skb
)
1443 u16 hdr_len
= skb_headlen(skb
);
1445 /* set gso_size to avoid messing up TCP MSS */
1446 skb_shinfo(skb
)->gso_size
= DIV_ROUND_UP((skb
->len
- hdr_len
),
1447 IXGBE_CB(skb
)->append_cnt
);
1448 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV4
;
1451 static void ixgbe_update_rsc_stats(struct ixgbe_ring
*rx_ring
,
1452 struct sk_buff
*skb
)
1454 /* if append_cnt is 0 then frame is not RSC */
1455 if (!IXGBE_CB(skb
)->append_cnt
)
1458 rx_ring
->rx_stats
.rsc_count
+= IXGBE_CB(skb
)->append_cnt
;
1459 rx_ring
->rx_stats
.rsc_flush
++;
1461 ixgbe_set_rsc_gso_size(rx_ring
, skb
);
1463 /* gso_size is computed using append_cnt so always clear it last */
1464 IXGBE_CB(skb
)->append_cnt
= 0;
1468 * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1469 * @rx_ring: rx descriptor ring packet is being transacted on
1470 * @rx_desc: pointer to the EOP Rx descriptor
1471 * @skb: pointer to current skb being populated
1473 * This function checks the ring, descriptor, and packet information in
1474 * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1475 * other fields within the skb.
1477 static void ixgbe_process_skb_fields(struct ixgbe_ring
*rx_ring
,
1478 union ixgbe_adv_rx_desc
*rx_desc
,
1479 struct sk_buff
*skb
)
1481 struct net_device
*dev
= rx_ring
->netdev
;
1483 ixgbe_update_rsc_stats(rx_ring
, skb
);
1485 ixgbe_rx_hash(rx_ring
, rx_desc
, skb
);
1487 ixgbe_rx_checksum(rx_ring
, rx_desc
, skb
);
1489 ixgbe_ptp_rx_hwtstamp(rx_ring
, rx_desc
, skb
);
1491 if ((dev
->features
& NETIF_F_HW_VLAN_CTAG_RX
) &&
1492 ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_VP
)) {
1493 u16 vid
= le16_to_cpu(rx_desc
->wb
.upper
.vlan
);
1494 __vlan_hwaccel_put_tag(skb
, vid
);
1497 skb_record_rx_queue(skb
, rx_ring
->queue_index
);
1499 skb
->protocol
= eth_type_trans(skb
, dev
);
1502 static void ixgbe_rx_skb(struct ixgbe_q_vector
*q_vector
,
1503 struct sk_buff
*skb
)
1505 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
1507 if (!(adapter
->flags
& IXGBE_FLAG_IN_NETPOLL
))
1508 napi_gro_receive(&q_vector
->napi
, skb
);
1514 * ixgbe_is_non_eop - process handling of non-EOP buffers
1515 * @rx_ring: Rx ring being processed
1516 * @rx_desc: Rx descriptor for current buffer
1517 * @skb: Current socket buffer containing buffer in progress
1519 * This function updates next to clean. If the buffer is an EOP buffer
1520 * this function exits returning false, otherwise it will place the
1521 * sk_buff in the next buffer to be chained and return true indicating
1522 * that this is in fact a non-EOP buffer.
1524 static bool ixgbe_is_non_eop(struct ixgbe_ring
*rx_ring
,
1525 union ixgbe_adv_rx_desc
*rx_desc
,
1526 struct sk_buff
*skb
)
1528 u32 ntc
= rx_ring
->next_to_clean
+ 1;
1530 /* fetch, update, and store next to clean */
1531 ntc
= (ntc
< rx_ring
->count
) ? ntc
: 0;
1532 rx_ring
->next_to_clean
= ntc
;
1534 prefetch(IXGBE_RX_DESC(rx_ring
, ntc
));
1536 /* update RSC append count if present */
1537 if (ring_is_rsc_enabled(rx_ring
)) {
1538 __le32 rsc_enabled
= rx_desc
->wb
.lower
.lo_dword
.data
&
1539 cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK
);
1541 if (unlikely(rsc_enabled
)) {
1542 u32 rsc_cnt
= le32_to_cpu(rsc_enabled
);
1544 rsc_cnt
>>= IXGBE_RXDADV_RSCCNT_SHIFT
;
1545 IXGBE_CB(skb
)->append_cnt
+= rsc_cnt
- 1;
1547 /* update ntc based on RSC value */
1548 ntc
= le32_to_cpu(rx_desc
->wb
.upper
.status_error
);
1549 ntc
&= IXGBE_RXDADV_NEXTP_MASK
;
1550 ntc
>>= IXGBE_RXDADV_NEXTP_SHIFT
;
1554 /* if we are the last buffer then there is nothing else to do */
1555 if (likely(ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_EOP
)))
1558 /* place skb in next buffer to be received */
1559 rx_ring
->rx_buffer_info
[ntc
].skb
= skb
;
1560 rx_ring
->rx_stats
.non_eop_descs
++;
1566 * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1567 * @rx_ring: rx descriptor ring packet is being transacted on
1568 * @skb: pointer to current skb being adjusted
1570 * This function is an ixgbe specific version of __pskb_pull_tail. The
1571 * main difference between this version and the original function is that
1572 * this function can make several assumptions about the state of things
1573 * that allow for significant optimizations versus the standard function.
1574 * As a result we can do things like drop a frag and maintain an accurate
1575 * truesize for the skb.
1577 static void ixgbe_pull_tail(struct ixgbe_ring
*rx_ring
,
1578 struct sk_buff
*skb
)
1580 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[0];
1582 unsigned int pull_len
;
1585 * it is valid to use page_address instead of kmap since we are
1586 * working with pages allocated out of the lomem pool per
1587 * alloc_page(GFP_ATOMIC)
1589 va
= skb_frag_address(frag
);
1592 * we need the header to contain the greater of either ETH_HLEN or
1593 * 60 bytes if the skb->len is less than 60 for skb_pad.
1595 pull_len
= ixgbe_get_headlen(va
, IXGBE_RX_HDR_SIZE
);
1597 /* align pull length to size of long to optimize memcpy performance */
1598 skb_copy_to_linear_data(skb
, va
, ALIGN(pull_len
, sizeof(long)));
1600 /* update all of the pointers */
1601 skb_frag_size_sub(frag
, pull_len
);
1602 frag
->page_offset
+= pull_len
;
1603 skb
->data_len
-= pull_len
;
1604 skb
->tail
+= pull_len
;
1608 * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1609 * @rx_ring: rx descriptor ring packet is being transacted on
1610 * @skb: pointer to current skb being updated
1612 * This function provides a basic DMA sync up for the first fragment of an
1613 * skb. The reason for doing this is that the first fragment cannot be
1614 * unmapped until we have reached the end of packet descriptor for a buffer
1617 static void ixgbe_dma_sync_frag(struct ixgbe_ring
*rx_ring
,
1618 struct sk_buff
*skb
)
1620 /* if the page was released unmap it, else just sync our portion */
1621 if (unlikely(IXGBE_CB(skb
)->page_released
)) {
1622 dma_unmap_page(rx_ring
->dev
, IXGBE_CB(skb
)->dma
,
1623 ixgbe_rx_pg_size(rx_ring
), DMA_FROM_DEVICE
);
1624 IXGBE_CB(skb
)->page_released
= false;
1626 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[0];
1628 dma_sync_single_range_for_cpu(rx_ring
->dev
,
1631 ixgbe_rx_bufsz(rx_ring
),
1634 IXGBE_CB(skb
)->dma
= 0;
1638 * ixgbe_cleanup_headers - Correct corrupted or empty headers
1639 * @rx_ring: rx descriptor ring packet is being transacted on
1640 * @rx_desc: pointer to the EOP Rx descriptor
1641 * @skb: pointer to current skb being fixed
1643 * Check for corrupted packet headers caused by senders on the local L2
1644 * embedded NIC switch not setting up their Tx Descriptors right. These
1645 * should be very rare.
1647 * Also address the case where we are pulling data in on pages only
1648 * and as such no data is present in the skb header.
1650 * In addition if skb is not at least 60 bytes we need to pad it so that
1651 * it is large enough to qualify as a valid Ethernet frame.
1653 * Returns true if an error was encountered and skb was freed.
1655 static bool ixgbe_cleanup_headers(struct ixgbe_ring
*rx_ring
,
1656 union ixgbe_adv_rx_desc
*rx_desc
,
1657 struct sk_buff
*skb
)
1659 struct net_device
*netdev
= rx_ring
->netdev
;
1661 /* verify that the packet does not have any known errors */
1662 if (unlikely(ixgbe_test_staterr(rx_desc
,
1663 IXGBE_RXDADV_ERR_FRAME_ERR_MASK
) &&
1664 !(netdev
->features
& NETIF_F_RXALL
))) {
1665 dev_kfree_skb_any(skb
);
1669 /* place header in linear portion of buffer */
1670 if (skb_is_nonlinear(skb
))
1671 ixgbe_pull_tail(rx_ring
, skb
);
1674 /* do not attempt to pad FCoE Frames as this will disrupt DDP */
1675 if (ixgbe_rx_is_fcoe(rx_ring
, rx_desc
))
1679 /* if skb_pad returns an error the skb was freed */
1680 if (unlikely(skb
->len
< 60)) {
1681 int pad_len
= 60 - skb
->len
;
1683 if (skb_pad(skb
, pad_len
))
1685 __skb_put(skb
, pad_len
);
1692 * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1693 * @rx_ring: rx descriptor ring to store buffers on
1694 * @old_buff: donor buffer to have page reused
1696 * Synchronizes page for reuse by the adapter
1698 static void ixgbe_reuse_rx_page(struct ixgbe_ring
*rx_ring
,
1699 struct ixgbe_rx_buffer
*old_buff
)
1701 struct ixgbe_rx_buffer
*new_buff
;
1702 u16 nta
= rx_ring
->next_to_alloc
;
1704 new_buff
= &rx_ring
->rx_buffer_info
[nta
];
1706 /* update, and store next to alloc */
1708 rx_ring
->next_to_alloc
= (nta
< rx_ring
->count
) ? nta
: 0;
1710 /* transfer page from old buffer to new buffer */
1711 new_buff
->page
= old_buff
->page
;
1712 new_buff
->dma
= old_buff
->dma
;
1713 new_buff
->page_offset
= old_buff
->page_offset
;
1715 /* sync the buffer for use by the device */
1716 dma_sync_single_range_for_device(rx_ring
->dev
, new_buff
->dma
,
1717 new_buff
->page_offset
,
1718 ixgbe_rx_bufsz(rx_ring
),
1723 * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1724 * @rx_ring: rx descriptor ring to transact packets on
1725 * @rx_buffer: buffer containing page to add
1726 * @rx_desc: descriptor containing length of buffer written by hardware
1727 * @skb: sk_buff to place the data into
1729 * This function will add the data contained in rx_buffer->page to the skb.
1730 * This is done either through a direct copy if the data in the buffer is
1731 * less than the skb header size, otherwise it will just attach the page as
1732 * a frag to the skb.
1734 * The function will then update the page offset if necessary and return
1735 * true if the buffer can be reused by the adapter.
1737 static bool ixgbe_add_rx_frag(struct ixgbe_ring
*rx_ring
,
1738 struct ixgbe_rx_buffer
*rx_buffer
,
1739 union ixgbe_adv_rx_desc
*rx_desc
,
1740 struct sk_buff
*skb
)
1742 struct page
*page
= rx_buffer
->page
;
1743 unsigned int size
= le16_to_cpu(rx_desc
->wb
.upper
.length
);
1744 #if (PAGE_SIZE < 8192)
1745 unsigned int truesize
= ixgbe_rx_bufsz(rx_ring
);
1747 unsigned int truesize
= ALIGN(size
, L1_CACHE_BYTES
);
1748 unsigned int last_offset
= ixgbe_rx_pg_size(rx_ring
) -
1749 ixgbe_rx_bufsz(rx_ring
);
1752 if ((size
<= IXGBE_RX_HDR_SIZE
) && !skb_is_nonlinear(skb
)) {
1753 unsigned char *va
= page_address(page
) + rx_buffer
->page_offset
;
1755 memcpy(__skb_put(skb
, size
), va
, ALIGN(size
, sizeof(long)));
1757 /* we can reuse buffer as-is, just make sure it is local */
1758 if (likely(page_to_nid(page
) == numa_node_id()))
1761 /* this page cannot be reused so discard it */
1766 skb_add_rx_frag(skb
, skb_shinfo(skb
)->nr_frags
, page
,
1767 rx_buffer
->page_offset
, size
, truesize
);
1769 /* avoid re-using remote pages */
1770 if (unlikely(page_to_nid(page
) != numa_node_id()))
1773 #if (PAGE_SIZE < 8192)
1774 /* if we are only owner of page we can reuse it */
1775 if (unlikely(page_count(page
) != 1))
1778 /* flip page offset to other buffer */
1779 rx_buffer
->page_offset
^= truesize
;
1782 * since we are the only owner of the page and we need to
1783 * increment it, just set the value to 2 in order to avoid
1784 * an unecessary locked operation
1786 atomic_set(&page
->_count
, 2);
1788 /* move offset up to the next cache line */
1789 rx_buffer
->page_offset
+= truesize
;
1791 if (rx_buffer
->page_offset
> last_offset
)
1794 /* bump ref count on page before it is given to the stack */
1801 static struct sk_buff
*ixgbe_fetch_rx_buffer(struct ixgbe_ring
*rx_ring
,
1802 union ixgbe_adv_rx_desc
*rx_desc
)
1804 struct ixgbe_rx_buffer
*rx_buffer
;
1805 struct sk_buff
*skb
;
1808 rx_buffer
= &rx_ring
->rx_buffer_info
[rx_ring
->next_to_clean
];
1809 page
= rx_buffer
->page
;
1812 skb
= rx_buffer
->skb
;
1815 void *page_addr
= page_address(page
) +
1816 rx_buffer
->page_offset
;
1818 /* prefetch first cache line of first page */
1819 prefetch(page_addr
);
1820 #if L1_CACHE_BYTES < 128
1821 prefetch(page_addr
+ L1_CACHE_BYTES
);
1824 /* allocate a skb to store the frags */
1825 skb
= netdev_alloc_skb_ip_align(rx_ring
->netdev
,
1827 if (unlikely(!skb
)) {
1828 rx_ring
->rx_stats
.alloc_rx_buff_failed
++;
1833 * we will be copying header into skb->data in
1834 * pskb_may_pull so it is in our interest to prefetch
1835 * it now to avoid a possible cache miss
1837 prefetchw(skb
->data
);
1840 * Delay unmapping of the first packet. It carries the
1841 * header information, HW may still access the header
1842 * after the writeback. Only unmap it when EOP is
1845 if (likely(ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_EOP
)))
1848 IXGBE_CB(skb
)->dma
= rx_buffer
->dma
;
1850 if (ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_EOP
))
1851 ixgbe_dma_sync_frag(rx_ring
, skb
);
1854 /* we are reusing so sync this buffer for CPU use */
1855 dma_sync_single_range_for_cpu(rx_ring
->dev
,
1857 rx_buffer
->page_offset
,
1858 ixgbe_rx_bufsz(rx_ring
),
1862 /* pull page into skb */
1863 if (ixgbe_add_rx_frag(rx_ring
, rx_buffer
, rx_desc
, skb
)) {
1864 /* hand second half of page back to the ring */
1865 ixgbe_reuse_rx_page(rx_ring
, rx_buffer
);
1866 } else if (IXGBE_CB(skb
)->dma
== rx_buffer
->dma
) {
1867 /* the page has been released from the ring */
1868 IXGBE_CB(skb
)->page_released
= true;
1870 /* we are not reusing the buffer so unmap it */
1871 dma_unmap_page(rx_ring
->dev
, rx_buffer
->dma
,
1872 ixgbe_rx_pg_size(rx_ring
),
1876 /* clear contents of buffer_info */
1877 rx_buffer
->skb
= NULL
;
1879 rx_buffer
->page
= NULL
;
1885 * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
1886 * @q_vector: structure containing interrupt and ring information
1887 * @rx_ring: rx descriptor ring to transact packets on
1888 * @budget: Total limit on number of packets to process
1890 * This function provides a "bounce buffer" approach to Rx interrupt
1891 * processing. The advantage to this is that on systems that have
1892 * expensive overhead for IOMMU access this provides a means of avoiding
1893 * it by maintaining the mapping of the page to the syste.
1895 * Returns true if all work is completed without reaching budget
1897 static bool ixgbe_clean_rx_irq(struct ixgbe_q_vector
*q_vector
,
1898 struct ixgbe_ring
*rx_ring
,
1901 unsigned int total_rx_bytes
= 0, total_rx_packets
= 0;
1903 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
1905 unsigned int mss
= 0;
1906 #endif /* IXGBE_FCOE */
1907 u16 cleaned_count
= ixgbe_desc_unused(rx_ring
);
1910 union ixgbe_adv_rx_desc
*rx_desc
;
1911 struct sk_buff
*skb
;
1913 /* return some buffers to hardware, one at a time is too slow */
1914 if (cleaned_count
>= IXGBE_RX_BUFFER_WRITE
) {
1915 ixgbe_alloc_rx_buffers(rx_ring
, cleaned_count
);
1919 rx_desc
= IXGBE_RX_DESC(rx_ring
, rx_ring
->next_to_clean
);
1921 if (!ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_DD
))
1925 * This memory barrier is needed to keep us from reading
1926 * any other fields out of the rx_desc until we know the
1927 * RXD_STAT_DD bit is set
1931 /* retrieve a buffer from the ring */
1932 skb
= ixgbe_fetch_rx_buffer(rx_ring
, rx_desc
);
1934 /* exit if we failed to retrieve a buffer */
1940 /* place incomplete frames back on ring for completion */
1941 if (ixgbe_is_non_eop(rx_ring
, rx_desc
, skb
))
1944 /* verify the packet layout is correct */
1945 if (ixgbe_cleanup_headers(rx_ring
, rx_desc
, skb
))
1948 /* probably a little skewed due to removing CRC */
1949 total_rx_bytes
+= skb
->len
;
1951 /* populate checksum, timestamp, VLAN, and protocol */
1952 ixgbe_process_skb_fields(rx_ring
, rx_desc
, skb
);
1955 /* if ddp, not passing to ULD unless for FCP_RSP or error */
1956 if (ixgbe_rx_is_fcoe(rx_ring
, rx_desc
)) {
1957 ddp_bytes
= ixgbe_fcoe_ddp(adapter
, rx_desc
, skb
);
1958 /* include DDPed FCoE data */
1959 if (ddp_bytes
> 0) {
1961 mss
= rx_ring
->netdev
->mtu
-
1962 sizeof(struct fcoe_hdr
) -
1963 sizeof(struct fc_frame_header
) -
1964 sizeof(struct fcoe_crc_eof
);
1968 total_rx_bytes
+= ddp_bytes
;
1969 total_rx_packets
+= DIV_ROUND_UP(ddp_bytes
,
1973 dev_kfree_skb_any(skb
);
1978 #endif /* IXGBE_FCOE */
1979 ixgbe_rx_skb(q_vector
, skb
);
1981 /* update budget accounting */
1983 } while (likely(total_rx_packets
< budget
));
1985 u64_stats_update_begin(&rx_ring
->syncp
);
1986 rx_ring
->stats
.packets
+= total_rx_packets
;
1987 rx_ring
->stats
.bytes
+= total_rx_bytes
;
1988 u64_stats_update_end(&rx_ring
->syncp
);
1989 q_vector
->rx
.total_packets
+= total_rx_packets
;
1990 q_vector
->rx
.total_bytes
+= total_rx_bytes
;
1993 ixgbe_alloc_rx_buffers(rx_ring
, cleaned_count
);
1995 return (total_rx_packets
< budget
);
1999 * ixgbe_configure_msix - Configure MSI-X hardware
2000 * @adapter: board private structure
2002 * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2005 static void ixgbe_configure_msix(struct ixgbe_adapter
*adapter
)
2007 struct ixgbe_q_vector
*q_vector
;
2011 /* Populate MSIX to EITR Select */
2012 if (adapter
->num_vfs
> 32) {
2013 u32 eitrsel
= (1 << (adapter
->num_vfs
- 32)) - 1;
2014 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EITRSEL
, eitrsel
);
2018 * Populate the IVAR table and set the ITR values to the
2019 * corresponding register.
2021 for (v_idx
= 0; v_idx
< adapter
->num_q_vectors
; v_idx
++) {
2022 struct ixgbe_ring
*ring
;
2023 q_vector
= adapter
->q_vector
[v_idx
];
2025 ixgbe_for_each_ring(ring
, q_vector
->rx
)
2026 ixgbe_set_ivar(adapter
, 0, ring
->reg_idx
, v_idx
);
2028 ixgbe_for_each_ring(ring
, q_vector
->tx
)
2029 ixgbe_set_ivar(adapter
, 1, ring
->reg_idx
, v_idx
);
2031 ixgbe_write_eitr(q_vector
);
2034 switch (adapter
->hw
.mac
.type
) {
2035 case ixgbe_mac_82598EB
:
2036 ixgbe_set_ivar(adapter
, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX
,
2039 case ixgbe_mac_82599EB
:
2040 case ixgbe_mac_X540
:
2041 ixgbe_set_ivar(adapter
, -1, 1, v_idx
);
2046 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EITR(v_idx
), 1950);
2048 /* set up to autoclear timer, and the vectors */
2049 mask
= IXGBE_EIMS_ENABLE_MASK
;
2050 mask
&= ~(IXGBE_EIMS_OTHER
|
2051 IXGBE_EIMS_MAILBOX
|
2054 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIAC
, mask
);
2057 enum latency_range
{
2061 latency_invalid
= 255
2065 * ixgbe_update_itr - update the dynamic ITR value based on statistics
2066 * @q_vector: structure containing interrupt and ring information
2067 * @ring_container: structure containing ring performance data
2069 * Stores a new ITR value based on packets and byte
2070 * counts during the last interrupt. The advantage of per interrupt
2071 * computation is faster updates and more accurate ITR for the current
2072 * traffic pattern. Constants in this function were computed
2073 * based on theoretical maximum wire speed and thresholds were set based
2074 * on testing data as well as attempting to minimize response time
2075 * while increasing bulk throughput.
2076 * this functionality is controlled by the InterruptThrottleRate module
2077 * parameter (see ixgbe_param.c)
2079 static void ixgbe_update_itr(struct ixgbe_q_vector
*q_vector
,
2080 struct ixgbe_ring_container
*ring_container
)
2082 int bytes
= ring_container
->total_bytes
;
2083 int packets
= ring_container
->total_packets
;
2086 u8 itr_setting
= ring_container
->itr
;
2091 /* simple throttlerate management
2092 * 0-10MB/s lowest (100000 ints/s)
2093 * 10-20MB/s low (20000 ints/s)
2094 * 20-1249MB/s bulk (8000 ints/s)
2096 /* what was last interrupt timeslice? */
2097 timepassed_us
= q_vector
->itr
>> 2;
2098 bytes_perint
= bytes
/ timepassed_us
; /* bytes/usec */
2100 switch (itr_setting
) {
2101 case lowest_latency
:
2102 if (bytes_perint
> 10)
2103 itr_setting
= low_latency
;
2106 if (bytes_perint
> 20)
2107 itr_setting
= bulk_latency
;
2108 else if (bytes_perint
<= 10)
2109 itr_setting
= lowest_latency
;
2112 if (bytes_perint
<= 20)
2113 itr_setting
= low_latency
;
2117 /* clear work counters since we have the values we need */
2118 ring_container
->total_bytes
= 0;
2119 ring_container
->total_packets
= 0;
2121 /* write updated itr to ring container */
2122 ring_container
->itr
= itr_setting
;
2126 * ixgbe_write_eitr - write EITR register in hardware specific way
2127 * @q_vector: structure containing interrupt and ring information
2129 * This function is made to be called by ethtool and by the driver
2130 * when it needs to update EITR registers at runtime. Hardware
2131 * specific quirks/differences are taken care of here.
2133 void ixgbe_write_eitr(struct ixgbe_q_vector
*q_vector
)
2135 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
2136 struct ixgbe_hw
*hw
= &adapter
->hw
;
2137 int v_idx
= q_vector
->v_idx
;
2138 u32 itr_reg
= q_vector
->itr
& IXGBE_MAX_EITR
;
2140 switch (adapter
->hw
.mac
.type
) {
2141 case ixgbe_mac_82598EB
:
2142 /* must write high and low 16 bits to reset counter */
2143 itr_reg
|= (itr_reg
<< 16);
2145 case ixgbe_mac_82599EB
:
2146 case ixgbe_mac_X540
:
2148 * set the WDIS bit to not clear the timer bits and cause an
2149 * immediate assertion of the interrupt
2151 itr_reg
|= IXGBE_EITR_CNT_WDIS
;
2156 IXGBE_WRITE_REG(hw
, IXGBE_EITR(v_idx
), itr_reg
);
2159 static void ixgbe_set_itr(struct ixgbe_q_vector
*q_vector
)
2161 u32 new_itr
= q_vector
->itr
;
2164 ixgbe_update_itr(q_vector
, &q_vector
->tx
);
2165 ixgbe_update_itr(q_vector
, &q_vector
->rx
);
2167 current_itr
= max(q_vector
->rx
.itr
, q_vector
->tx
.itr
);
2169 switch (current_itr
) {
2170 /* counts and packets in update_itr are dependent on these numbers */
2171 case lowest_latency
:
2172 new_itr
= IXGBE_100K_ITR
;
2175 new_itr
= IXGBE_20K_ITR
;
2178 new_itr
= IXGBE_8K_ITR
;
2184 if (new_itr
!= q_vector
->itr
) {
2185 /* do an exponential smoothing */
2186 new_itr
= (10 * new_itr
* q_vector
->itr
) /
2187 ((9 * new_itr
) + q_vector
->itr
);
2189 /* save the algorithm value here */
2190 q_vector
->itr
= new_itr
;
2192 ixgbe_write_eitr(q_vector
);
2197 * ixgbe_check_overtemp_subtask - check for over temperature
2198 * @adapter: pointer to adapter
2200 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter
*adapter
)
2202 struct ixgbe_hw
*hw
= &adapter
->hw
;
2203 u32 eicr
= adapter
->interrupt_event
;
2205 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
2208 if (!(adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
) &&
2209 !(adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_EVENT
))
2212 adapter
->flags2
&= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT
;
2214 switch (hw
->device_id
) {
2215 case IXGBE_DEV_ID_82599_T3_LOM
:
2217 * Since the warning interrupt is for both ports
2218 * we don't have to check if:
2219 * - This interrupt wasn't for our port.
2220 * - We may have missed the interrupt so always have to
2221 * check if we got a LSC
2223 if (!(eicr
& IXGBE_EICR_GPI_SDP0
) &&
2224 !(eicr
& IXGBE_EICR_LSC
))
2227 if (!(eicr
& IXGBE_EICR_LSC
) && hw
->mac
.ops
.check_link
) {
2229 bool link_up
= false;
2231 hw
->mac
.ops
.check_link(hw
, &speed
, &link_up
, false);
2237 /* Check if this is not due to overtemp */
2238 if (hw
->phy
.ops
.check_overtemp(hw
) != IXGBE_ERR_OVERTEMP
)
2243 if (!(eicr
& IXGBE_EICR_GPI_SDP0
))
2248 "Network adapter has been stopped because it has over heated. "
2249 "Restart the computer. If the problem persists, "
2250 "power off the system and replace the adapter\n");
2252 adapter
->interrupt_event
= 0;
2255 static void ixgbe_check_fan_failure(struct ixgbe_adapter
*adapter
, u32 eicr
)
2257 struct ixgbe_hw
*hw
= &adapter
->hw
;
2259 if ((adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
) &&
2260 (eicr
& IXGBE_EICR_GPI_SDP1
)) {
2261 e_crit(probe
, "Fan has stopped, replace the adapter\n");
2262 /* write to clear the interrupt */
2263 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, IXGBE_EICR_GPI_SDP1
);
2267 static void ixgbe_check_overtemp_event(struct ixgbe_adapter
*adapter
, u32 eicr
)
2269 if (!(adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
))
2272 switch (adapter
->hw
.mac
.type
) {
2273 case ixgbe_mac_82599EB
:
2275 * Need to check link state so complete overtemp check
2278 if (((eicr
& IXGBE_EICR_GPI_SDP0
) || (eicr
& IXGBE_EICR_LSC
)) &&
2279 (!test_bit(__IXGBE_DOWN
, &adapter
->state
))) {
2280 adapter
->interrupt_event
= eicr
;
2281 adapter
->flags2
|= IXGBE_FLAG2_TEMP_SENSOR_EVENT
;
2282 ixgbe_service_event_schedule(adapter
);
2286 case ixgbe_mac_X540
:
2287 if (!(eicr
& IXGBE_EICR_TS
))
2295 "Network adapter has been stopped because it has over heated. "
2296 "Restart the computer. If the problem persists, "
2297 "power off the system and replace the adapter\n");
2300 static void ixgbe_check_sfp_event(struct ixgbe_adapter
*adapter
, u32 eicr
)
2302 struct ixgbe_hw
*hw
= &adapter
->hw
;
2304 if (eicr
& IXGBE_EICR_GPI_SDP2
) {
2305 /* Clear the interrupt */
2306 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, IXGBE_EICR_GPI_SDP2
);
2307 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
2308 adapter
->flags2
|= IXGBE_FLAG2_SFP_NEEDS_RESET
;
2309 ixgbe_service_event_schedule(adapter
);
2313 if (eicr
& IXGBE_EICR_GPI_SDP1
) {
2314 /* Clear the interrupt */
2315 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, IXGBE_EICR_GPI_SDP1
);
2316 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
2317 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_CONFIG
;
2318 ixgbe_service_event_schedule(adapter
);
2323 static void ixgbe_check_lsc(struct ixgbe_adapter
*adapter
)
2325 struct ixgbe_hw
*hw
= &adapter
->hw
;
2328 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_UPDATE
;
2329 adapter
->link_check_timeout
= jiffies
;
2330 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
2331 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, IXGBE_EIMC_LSC
);
2332 IXGBE_WRITE_FLUSH(hw
);
2333 ixgbe_service_event_schedule(adapter
);
2337 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter
*adapter
,
2341 struct ixgbe_hw
*hw
= &adapter
->hw
;
2343 switch (hw
->mac
.type
) {
2344 case ixgbe_mac_82598EB
:
2345 mask
= (IXGBE_EIMS_RTX_QUEUE
& qmask
);
2346 IXGBE_WRITE_REG(hw
, IXGBE_EIMS
, mask
);
2348 case ixgbe_mac_82599EB
:
2349 case ixgbe_mac_X540
:
2350 mask
= (qmask
& 0xFFFFFFFF);
2352 IXGBE_WRITE_REG(hw
, IXGBE_EIMS_EX(0), mask
);
2353 mask
= (qmask
>> 32);
2355 IXGBE_WRITE_REG(hw
, IXGBE_EIMS_EX(1), mask
);
2360 /* skip the flush */
2363 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter
*adapter
,
2367 struct ixgbe_hw
*hw
= &adapter
->hw
;
2369 switch (hw
->mac
.type
) {
2370 case ixgbe_mac_82598EB
:
2371 mask
= (IXGBE_EIMS_RTX_QUEUE
& qmask
);
2372 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, mask
);
2374 case ixgbe_mac_82599EB
:
2375 case ixgbe_mac_X540
:
2376 mask
= (qmask
& 0xFFFFFFFF);
2378 IXGBE_WRITE_REG(hw
, IXGBE_EIMC_EX(0), mask
);
2379 mask
= (qmask
>> 32);
2381 IXGBE_WRITE_REG(hw
, IXGBE_EIMC_EX(1), mask
);
2386 /* skip the flush */
2390 * ixgbe_irq_enable - Enable default interrupt generation settings
2391 * @adapter: board private structure
2393 static inline void ixgbe_irq_enable(struct ixgbe_adapter
*adapter
, bool queues
,
2396 u32 mask
= (IXGBE_EIMS_ENABLE_MASK
& ~IXGBE_EIMS_RTX_QUEUE
);
2398 /* don't reenable LSC while waiting for link */
2399 if (adapter
->flags
& IXGBE_FLAG_NEED_LINK_UPDATE
)
2400 mask
&= ~IXGBE_EIMS_LSC
;
2402 if (adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
)
2403 switch (adapter
->hw
.mac
.type
) {
2404 case ixgbe_mac_82599EB
:
2405 mask
|= IXGBE_EIMS_GPI_SDP0
;
2407 case ixgbe_mac_X540
:
2408 mask
|= IXGBE_EIMS_TS
;
2413 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
)
2414 mask
|= IXGBE_EIMS_GPI_SDP1
;
2415 switch (adapter
->hw
.mac
.type
) {
2416 case ixgbe_mac_82599EB
:
2417 mask
|= IXGBE_EIMS_GPI_SDP1
;
2418 mask
|= IXGBE_EIMS_GPI_SDP2
;
2419 case ixgbe_mac_X540
:
2420 mask
|= IXGBE_EIMS_ECC
;
2421 mask
|= IXGBE_EIMS_MAILBOX
;
2427 if (adapter
->hw
.mac
.type
== ixgbe_mac_X540
)
2428 mask
|= IXGBE_EIMS_TIMESYNC
;
2430 if ((adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) &&
2431 !(adapter
->flags2
& IXGBE_FLAG2_FDIR_REQUIRES_REINIT
))
2432 mask
|= IXGBE_EIMS_FLOW_DIR
;
2434 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMS
, mask
);
2436 ixgbe_irq_enable_queues(adapter
, ~0);
2438 IXGBE_WRITE_FLUSH(&adapter
->hw
);
2441 static irqreturn_t
ixgbe_msix_other(int irq
, void *data
)
2443 struct ixgbe_adapter
*adapter
= data
;
2444 struct ixgbe_hw
*hw
= &adapter
->hw
;
2448 * Workaround for Silicon errata. Use clear-by-write instead
2449 * of clear-by-read. Reading with EICS will return the
2450 * interrupt causes without clearing, which later be done
2451 * with the write to EICR.
2453 eicr
= IXGBE_READ_REG(hw
, IXGBE_EICS
);
2454 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, eicr
);
2456 if (eicr
& IXGBE_EICR_LSC
)
2457 ixgbe_check_lsc(adapter
);
2459 if (eicr
& IXGBE_EICR_MAILBOX
)
2460 ixgbe_msg_task(adapter
);
2462 switch (hw
->mac
.type
) {
2463 case ixgbe_mac_82599EB
:
2464 case ixgbe_mac_X540
:
2465 if (eicr
& IXGBE_EICR_ECC
)
2466 e_info(link
, "Received unrecoverable ECC Err, please "
2468 /* Handle Flow Director Full threshold interrupt */
2469 if (eicr
& IXGBE_EICR_FLOW_DIR
) {
2470 int reinit_count
= 0;
2472 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
2473 struct ixgbe_ring
*ring
= adapter
->tx_ring
[i
];
2474 if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE
,
2479 /* no more flow director interrupts until after init */
2480 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, IXGBE_EIMC_FLOW_DIR
);
2481 adapter
->flags2
|= IXGBE_FLAG2_FDIR_REQUIRES_REINIT
;
2482 ixgbe_service_event_schedule(adapter
);
2485 ixgbe_check_sfp_event(adapter
, eicr
);
2486 ixgbe_check_overtemp_event(adapter
, eicr
);
2492 ixgbe_check_fan_failure(adapter
, eicr
);
2494 if (unlikely(eicr
& IXGBE_EICR_TIMESYNC
))
2495 ixgbe_ptp_check_pps_event(adapter
, eicr
);
2497 /* re-enable the original interrupt state, no lsc, no queues */
2498 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2499 ixgbe_irq_enable(adapter
, false, false);
2504 static irqreturn_t
ixgbe_msix_clean_rings(int irq
, void *data
)
2506 struct ixgbe_q_vector
*q_vector
= data
;
2508 /* EIAM disabled interrupts (on this vector) for us */
2510 if (q_vector
->rx
.ring
|| q_vector
->tx
.ring
)
2511 napi_schedule(&q_vector
->napi
);
2517 * ixgbe_poll - NAPI Rx polling callback
2518 * @napi: structure for representing this polling device
2519 * @budget: how many packets driver is allowed to clean
2521 * This function is used for legacy and MSI, NAPI mode
2523 int ixgbe_poll(struct napi_struct
*napi
, int budget
)
2525 struct ixgbe_q_vector
*q_vector
=
2526 container_of(napi
, struct ixgbe_q_vector
, napi
);
2527 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
2528 struct ixgbe_ring
*ring
;
2529 int per_ring_budget
;
2530 bool clean_complete
= true;
2532 #ifdef CONFIG_IXGBE_DCA
2533 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
)
2534 ixgbe_update_dca(q_vector
);
2537 ixgbe_for_each_ring(ring
, q_vector
->tx
)
2538 clean_complete
&= !!ixgbe_clean_tx_irq(q_vector
, ring
);
2540 /* attempt to distribute budget to each queue fairly, but don't allow
2541 * the budget to go below 1 because we'll exit polling */
2542 if (q_vector
->rx
.count
> 1)
2543 per_ring_budget
= max(budget
/q_vector
->rx
.count
, 1);
2545 per_ring_budget
= budget
;
2547 ixgbe_for_each_ring(ring
, q_vector
->rx
)
2548 clean_complete
&= ixgbe_clean_rx_irq(q_vector
, ring
,
2551 /* If all work not completed, return budget and keep polling */
2552 if (!clean_complete
)
2555 /* all work done, exit the polling mode */
2556 napi_complete(napi
);
2557 if (adapter
->rx_itr_setting
& 1)
2558 ixgbe_set_itr(q_vector
);
2559 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2560 ixgbe_irq_enable_queues(adapter
, ((u64
)1 << q_vector
->v_idx
));
2566 * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2567 * @adapter: board private structure
2569 * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2570 * interrupts from the kernel.
2572 static int ixgbe_request_msix_irqs(struct ixgbe_adapter
*adapter
)
2574 struct net_device
*netdev
= adapter
->netdev
;
2578 for (vector
= 0; vector
< adapter
->num_q_vectors
; vector
++) {
2579 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[vector
];
2580 struct msix_entry
*entry
= &adapter
->msix_entries
[vector
];
2582 if (q_vector
->tx
.ring
&& q_vector
->rx
.ring
) {
2583 snprintf(q_vector
->name
, sizeof(q_vector
->name
) - 1,
2584 "%s-%s-%d", netdev
->name
, "TxRx", ri
++);
2586 } else if (q_vector
->rx
.ring
) {
2587 snprintf(q_vector
->name
, sizeof(q_vector
->name
) - 1,
2588 "%s-%s-%d", netdev
->name
, "rx", ri
++);
2589 } else if (q_vector
->tx
.ring
) {
2590 snprintf(q_vector
->name
, sizeof(q_vector
->name
) - 1,
2591 "%s-%s-%d", netdev
->name
, "tx", ti
++);
2593 /* skip this unused q_vector */
2596 err
= request_irq(entry
->vector
, &ixgbe_msix_clean_rings
, 0,
2597 q_vector
->name
, q_vector
);
2599 e_err(probe
, "request_irq failed for MSIX interrupt "
2600 "Error: %d\n", err
);
2601 goto free_queue_irqs
;
2603 /* If Flow Director is enabled, set interrupt affinity */
2604 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) {
2605 /* assign the mask for this irq */
2606 irq_set_affinity_hint(entry
->vector
,
2607 &q_vector
->affinity_mask
);
2611 err
= request_irq(adapter
->msix_entries
[vector
].vector
,
2612 ixgbe_msix_other
, 0, netdev
->name
, adapter
);
2614 e_err(probe
, "request_irq for msix_other failed: %d\n", err
);
2615 goto free_queue_irqs
;
2623 irq_set_affinity_hint(adapter
->msix_entries
[vector
].vector
,
2625 free_irq(adapter
->msix_entries
[vector
].vector
,
2626 adapter
->q_vector
[vector
]);
2628 adapter
->flags
&= ~IXGBE_FLAG_MSIX_ENABLED
;
2629 pci_disable_msix(adapter
->pdev
);
2630 kfree(adapter
->msix_entries
);
2631 adapter
->msix_entries
= NULL
;
2636 * ixgbe_intr - legacy mode Interrupt Handler
2637 * @irq: interrupt number
2638 * @data: pointer to a network interface device structure
2640 static irqreturn_t
ixgbe_intr(int irq
, void *data
)
2642 struct ixgbe_adapter
*adapter
= data
;
2643 struct ixgbe_hw
*hw
= &adapter
->hw
;
2644 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[0];
2648 * Workaround for silicon errata #26 on 82598. Mask the interrupt
2649 * before the read of EICR.
2651 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, IXGBE_IRQ_CLEAR_MASK
);
2653 /* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2654 * therefore no explicit interrupt disable is necessary */
2655 eicr
= IXGBE_READ_REG(hw
, IXGBE_EICR
);
2658 * shared interrupt alert!
2659 * make sure interrupts are enabled because the read will
2660 * have disabled interrupts due to EIAM
2661 * finish the workaround of silicon errata on 82598. Unmask
2662 * the interrupt that we masked before the EICR read.
2664 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2665 ixgbe_irq_enable(adapter
, true, true);
2666 return IRQ_NONE
; /* Not our interrupt */
2669 if (eicr
& IXGBE_EICR_LSC
)
2670 ixgbe_check_lsc(adapter
);
2672 switch (hw
->mac
.type
) {
2673 case ixgbe_mac_82599EB
:
2674 ixgbe_check_sfp_event(adapter
, eicr
);
2676 case ixgbe_mac_X540
:
2677 if (eicr
& IXGBE_EICR_ECC
)
2678 e_info(link
, "Received unrecoverable ECC err, please "
2680 ixgbe_check_overtemp_event(adapter
, eicr
);
2686 ixgbe_check_fan_failure(adapter
, eicr
);
2687 if (unlikely(eicr
& IXGBE_EICR_TIMESYNC
))
2688 ixgbe_ptp_check_pps_event(adapter
, eicr
);
2690 /* would disable interrupts here but EIAM disabled it */
2691 napi_schedule(&q_vector
->napi
);
2694 * re-enable link(maybe) and non-queue interrupts, no flush.
2695 * ixgbe_poll will re-enable the queue interrupts
2697 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2698 ixgbe_irq_enable(adapter
, false, false);
2704 * ixgbe_request_irq - initialize interrupts
2705 * @adapter: board private structure
2707 * Attempts to configure interrupts using the best available
2708 * capabilities of the hardware and kernel.
2710 static int ixgbe_request_irq(struct ixgbe_adapter
*adapter
)
2712 struct net_device
*netdev
= adapter
->netdev
;
2715 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)
2716 err
= ixgbe_request_msix_irqs(adapter
);
2717 else if (adapter
->flags
& IXGBE_FLAG_MSI_ENABLED
)
2718 err
= request_irq(adapter
->pdev
->irq
, ixgbe_intr
, 0,
2719 netdev
->name
, adapter
);
2721 err
= request_irq(adapter
->pdev
->irq
, ixgbe_intr
, IRQF_SHARED
,
2722 netdev
->name
, adapter
);
2725 e_err(probe
, "request_irq failed, Error %d\n", err
);
2730 static void ixgbe_free_irq(struct ixgbe_adapter
*adapter
)
2734 if (!(adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)) {
2735 free_irq(adapter
->pdev
->irq
, adapter
);
2739 for (vector
= 0; vector
< adapter
->num_q_vectors
; vector
++) {
2740 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[vector
];
2741 struct msix_entry
*entry
= &adapter
->msix_entries
[vector
];
2743 /* free only the irqs that were actually requested */
2744 if (!q_vector
->rx
.ring
&& !q_vector
->tx
.ring
)
2747 /* clear the affinity_mask in the IRQ descriptor */
2748 irq_set_affinity_hint(entry
->vector
, NULL
);
2750 free_irq(entry
->vector
, q_vector
);
2753 free_irq(adapter
->msix_entries
[vector
++].vector
, adapter
);
2757 * ixgbe_irq_disable - Mask off interrupt generation on the NIC
2758 * @adapter: board private structure
2760 static inline void ixgbe_irq_disable(struct ixgbe_adapter
*adapter
)
2762 switch (adapter
->hw
.mac
.type
) {
2763 case ixgbe_mac_82598EB
:
2764 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC
, ~0);
2766 case ixgbe_mac_82599EB
:
2767 case ixgbe_mac_X540
:
2768 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC
, 0xFFFF0000);
2769 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC_EX(0), ~0);
2770 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC_EX(1), ~0);
2775 IXGBE_WRITE_FLUSH(&adapter
->hw
);
2776 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
) {
2779 for (vector
= 0; vector
< adapter
->num_q_vectors
; vector
++)
2780 synchronize_irq(adapter
->msix_entries
[vector
].vector
);
2782 synchronize_irq(adapter
->msix_entries
[vector
++].vector
);
2784 synchronize_irq(adapter
->pdev
->irq
);
2789 * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
2792 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter
*adapter
)
2794 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[0];
2796 ixgbe_write_eitr(q_vector
);
2798 ixgbe_set_ivar(adapter
, 0, 0, 0);
2799 ixgbe_set_ivar(adapter
, 1, 0, 0);
2801 e_info(hw
, "Legacy interrupt IVAR setup done\n");
2805 * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
2806 * @adapter: board private structure
2807 * @ring: structure containing ring specific data
2809 * Configure the Tx descriptor ring after a reset.
2811 void ixgbe_configure_tx_ring(struct ixgbe_adapter
*adapter
,
2812 struct ixgbe_ring
*ring
)
2814 struct ixgbe_hw
*hw
= &adapter
->hw
;
2815 u64 tdba
= ring
->dma
;
2817 u32 txdctl
= IXGBE_TXDCTL_ENABLE
;
2818 u8 reg_idx
= ring
->reg_idx
;
2820 /* disable queue to avoid issues while updating state */
2821 IXGBE_WRITE_REG(hw
, IXGBE_TXDCTL(reg_idx
), 0);
2822 IXGBE_WRITE_FLUSH(hw
);
2824 IXGBE_WRITE_REG(hw
, IXGBE_TDBAL(reg_idx
),
2825 (tdba
& DMA_BIT_MASK(32)));
2826 IXGBE_WRITE_REG(hw
, IXGBE_TDBAH(reg_idx
), (tdba
>> 32));
2827 IXGBE_WRITE_REG(hw
, IXGBE_TDLEN(reg_idx
),
2828 ring
->count
* sizeof(union ixgbe_adv_tx_desc
));
2829 IXGBE_WRITE_REG(hw
, IXGBE_TDH(reg_idx
), 0);
2830 IXGBE_WRITE_REG(hw
, IXGBE_TDT(reg_idx
), 0);
2831 ring
->tail
= hw
->hw_addr
+ IXGBE_TDT(reg_idx
);
2834 * set WTHRESH to encourage burst writeback, it should not be set
2835 * higher than 1 when:
2836 * - ITR is 0 as it could cause false TX hangs
2837 * - ITR is set to > 100k int/sec and BQL is enabled
2839 * In order to avoid issues WTHRESH + PTHRESH should always be equal
2840 * to or less than the number of on chip descriptors, which is
2843 #if IS_ENABLED(CONFIG_BQL)
2844 if (!ring
->q_vector
|| (ring
->q_vector
->itr
< IXGBE_100K_ITR
))
2846 if (!ring
->q_vector
|| (ring
->q_vector
->itr
< 8))
2848 txdctl
|= (1 << 16); /* WTHRESH = 1 */
2850 txdctl
|= (8 << 16); /* WTHRESH = 8 */
2853 * Setting PTHRESH to 32 both improves performance
2854 * and avoids a TX hang with DFP enabled
2856 txdctl
|= (1 << 8) | /* HTHRESH = 1 */
2857 32; /* PTHRESH = 32 */
2859 /* reinitialize flowdirector state */
2860 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) {
2861 ring
->atr_sample_rate
= adapter
->atr_sample_rate
;
2862 ring
->atr_count
= 0;
2863 set_bit(__IXGBE_TX_FDIR_INIT_DONE
, &ring
->state
);
2865 ring
->atr_sample_rate
= 0;
2868 /* initialize XPS */
2869 if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE
, &ring
->state
)) {
2870 struct ixgbe_q_vector
*q_vector
= ring
->q_vector
;
2873 netif_set_xps_queue(adapter
->netdev
,
2874 &q_vector
->affinity_mask
,
2878 clear_bit(__IXGBE_HANG_CHECK_ARMED
, &ring
->state
);
2881 IXGBE_WRITE_REG(hw
, IXGBE_TXDCTL(reg_idx
), txdctl
);
2883 /* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
2884 if (hw
->mac
.type
== ixgbe_mac_82598EB
&&
2885 !(IXGBE_READ_REG(hw
, IXGBE_LINKS
) & IXGBE_LINKS_UP
))
2888 /* poll to verify queue is enabled */
2890 usleep_range(1000, 2000);
2891 txdctl
= IXGBE_READ_REG(hw
, IXGBE_TXDCTL(reg_idx
));
2892 } while (--wait_loop
&& !(txdctl
& IXGBE_TXDCTL_ENABLE
));
2894 e_err(drv
, "Could not enable Tx Queue %d\n", reg_idx
);
2897 static void ixgbe_setup_mtqc(struct ixgbe_adapter
*adapter
)
2899 struct ixgbe_hw
*hw
= &adapter
->hw
;
2901 u8 tcs
= netdev_get_num_tc(adapter
->netdev
);
2903 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
2906 /* disable the arbiter while setting MTQC */
2907 rttdcs
= IXGBE_READ_REG(hw
, IXGBE_RTTDCS
);
2908 rttdcs
|= IXGBE_RTTDCS_ARBDIS
;
2909 IXGBE_WRITE_REG(hw
, IXGBE_RTTDCS
, rttdcs
);
2911 /* set transmit pool layout */
2912 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
2913 mtqc
= IXGBE_MTQC_VT_ENA
;
2915 mtqc
|= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_8TC_8TQ
;
2917 mtqc
|= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_4TC_4TQ
;
2918 else if (adapter
->ring_feature
[RING_F_RSS
].indices
== 4)
2919 mtqc
|= IXGBE_MTQC_32VF
;
2921 mtqc
|= IXGBE_MTQC_64VF
;
2924 mtqc
= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_8TC_8TQ
;
2926 mtqc
= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_4TC_4TQ
;
2928 mtqc
= IXGBE_MTQC_64Q_1PB
;
2931 IXGBE_WRITE_REG(hw
, IXGBE_MTQC
, mtqc
);
2933 /* Enable Security TX Buffer IFG for multiple pb */
2935 u32 sectx
= IXGBE_READ_REG(hw
, IXGBE_SECTXMINIFG
);
2936 sectx
|= IXGBE_SECTX_DCB
;
2937 IXGBE_WRITE_REG(hw
, IXGBE_SECTXMINIFG
, sectx
);
2940 /* re-enable the arbiter */
2941 rttdcs
&= ~IXGBE_RTTDCS_ARBDIS
;
2942 IXGBE_WRITE_REG(hw
, IXGBE_RTTDCS
, rttdcs
);
2946 * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
2947 * @adapter: board private structure
2949 * Configure the Tx unit of the MAC after a reset.
2951 static void ixgbe_configure_tx(struct ixgbe_adapter
*adapter
)
2953 struct ixgbe_hw
*hw
= &adapter
->hw
;
2957 ixgbe_setup_mtqc(adapter
);
2959 if (hw
->mac
.type
!= ixgbe_mac_82598EB
) {
2960 /* DMATXCTL.EN must be before Tx queues are enabled */
2961 dmatxctl
= IXGBE_READ_REG(hw
, IXGBE_DMATXCTL
);
2962 dmatxctl
|= IXGBE_DMATXCTL_TE
;
2963 IXGBE_WRITE_REG(hw
, IXGBE_DMATXCTL
, dmatxctl
);
2966 /* Setup the HW Tx Head and Tail descriptor pointers */
2967 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
2968 ixgbe_configure_tx_ring(adapter
, adapter
->tx_ring
[i
]);
2971 static void ixgbe_enable_rx_drop(struct ixgbe_adapter
*adapter
,
2972 struct ixgbe_ring
*ring
)
2974 struct ixgbe_hw
*hw
= &adapter
->hw
;
2975 u8 reg_idx
= ring
->reg_idx
;
2976 u32 srrctl
= IXGBE_READ_REG(hw
, IXGBE_SRRCTL(reg_idx
));
2978 srrctl
|= IXGBE_SRRCTL_DROP_EN
;
2980 IXGBE_WRITE_REG(hw
, IXGBE_SRRCTL(reg_idx
), srrctl
);
2983 static void ixgbe_disable_rx_drop(struct ixgbe_adapter
*adapter
,
2984 struct ixgbe_ring
*ring
)
2986 struct ixgbe_hw
*hw
= &adapter
->hw
;
2987 u8 reg_idx
= ring
->reg_idx
;
2988 u32 srrctl
= IXGBE_READ_REG(hw
, IXGBE_SRRCTL(reg_idx
));
2990 srrctl
&= ~IXGBE_SRRCTL_DROP_EN
;
2992 IXGBE_WRITE_REG(hw
, IXGBE_SRRCTL(reg_idx
), srrctl
);
2995 #ifdef CONFIG_IXGBE_DCB
2996 void ixgbe_set_rx_drop_en(struct ixgbe_adapter
*adapter
)
2998 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter
*adapter
)
3002 bool pfc_en
= adapter
->dcb_cfg
.pfc_mode_enable
;
3004 if (adapter
->ixgbe_ieee_pfc
)
3005 pfc_en
|= !!(adapter
->ixgbe_ieee_pfc
->pfc_en
);
3008 * We should set the drop enable bit if:
3011 * Number of Rx queues > 1 and flow control is disabled
3013 * This allows us to avoid head of line blocking for security
3014 * and performance reasons.
3016 if (adapter
->num_vfs
|| (adapter
->num_rx_queues
> 1 &&
3017 !(adapter
->hw
.fc
.current_mode
& ixgbe_fc_tx_pause
) && !pfc_en
)) {
3018 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3019 ixgbe_enable_rx_drop(adapter
, adapter
->rx_ring
[i
]);
3021 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3022 ixgbe_disable_rx_drop(adapter
, adapter
->rx_ring
[i
]);
3026 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3028 static void ixgbe_configure_srrctl(struct ixgbe_adapter
*adapter
,
3029 struct ixgbe_ring
*rx_ring
)
3031 struct ixgbe_hw
*hw
= &adapter
->hw
;
3033 u8 reg_idx
= rx_ring
->reg_idx
;
3035 if (hw
->mac
.type
== ixgbe_mac_82598EB
) {
3036 u16 mask
= adapter
->ring_feature
[RING_F_RSS
].mask
;
3039 * if VMDq is not active we must program one srrctl register
3040 * per RSS queue since we have enabled RDRXCTL.MVMEN
3045 /* configure header buffer length, needed for RSC */
3046 srrctl
= IXGBE_RX_HDR_SIZE
<< IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT
;
3048 /* configure the packet buffer length */
3049 srrctl
|= ixgbe_rx_bufsz(rx_ring
) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT
;
3051 /* configure descriptor type */
3052 srrctl
|= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF
;
3054 IXGBE_WRITE_REG(hw
, IXGBE_SRRCTL(reg_idx
), srrctl
);
3057 static void ixgbe_setup_mrqc(struct ixgbe_adapter
*adapter
)
3059 struct ixgbe_hw
*hw
= &adapter
->hw
;
3060 static const u32 seed
[10] = { 0xE291D73D, 0x1805EC6C, 0x2A94B30D,
3061 0xA54F2BEC, 0xEA49AF7C, 0xE214AD3D, 0xB855AABE,
3062 0x6A3E67EA, 0x14364D17, 0x3BED200D};
3063 u32 mrqc
= 0, reta
= 0;
3066 u16 rss_i
= adapter
->ring_feature
[RING_F_RSS
].indices
;
3069 * Program table for at least 2 queues w/ SR-IOV so that VFs can
3070 * make full use of any rings they may have. We will use the
3071 * PSRTYPE register to control how many rings we use within the PF.
3073 if ((adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) && (rss_i
< 2))
3076 /* Fill out hash function seeds */
3077 for (i
= 0; i
< 10; i
++)
3078 IXGBE_WRITE_REG(hw
, IXGBE_RSSRK(i
), seed
[i
]);
3080 /* Fill out redirection table */
3081 for (i
= 0, j
= 0; i
< 128; i
++, j
++) {
3084 /* reta = 4-byte sliding window of
3085 * 0x00..(indices-1)(indices-1)00..etc. */
3086 reta
= (reta
<< 8) | (j
* 0x11);
3088 IXGBE_WRITE_REG(hw
, IXGBE_RETA(i
>> 2), reta
);
3091 /* Disable indicating checksum in descriptor, enables RSS hash */
3092 rxcsum
= IXGBE_READ_REG(hw
, IXGBE_RXCSUM
);
3093 rxcsum
|= IXGBE_RXCSUM_PCSD
;
3094 IXGBE_WRITE_REG(hw
, IXGBE_RXCSUM
, rxcsum
);
3096 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
) {
3097 if (adapter
->ring_feature
[RING_F_RSS
].mask
)
3098 mrqc
= IXGBE_MRQC_RSSEN
;
3100 u8 tcs
= netdev_get_num_tc(adapter
->netdev
);
3102 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
3104 mrqc
= IXGBE_MRQC_VMDQRT8TCEN
; /* 8 TCs */
3106 mrqc
= IXGBE_MRQC_VMDQRT4TCEN
; /* 4 TCs */
3107 else if (adapter
->ring_feature
[RING_F_RSS
].indices
== 4)
3108 mrqc
= IXGBE_MRQC_VMDQRSS32EN
;
3110 mrqc
= IXGBE_MRQC_VMDQRSS64EN
;
3113 mrqc
= IXGBE_MRQC_RTRSS8TCEN
;
3115 mrqc
= IXGBE_MRQC_RTRSS4TCEN
;
3117 mrqc
= IXGBE_MRQC_RSSEN
;
3121 /* Perform hash on these packet types */
3122 mrqc
|= IXGBE_MRQC_RSS_FIELD_IPV4
|
3123 IXGBE_MRQC_RSS_FIELD_IPV4_TCP
|
3124 IXGBE_MRQC_RSS_FIELD_IPV6
|
3125 IXGBE_MRQC_RSS_FIELD_IPV6_TCP
;
3127 if (adapter
->flags2
& IXGBE_FLAG2_RSS_FIELD_IPV4_UDP
)
3128 mrqc
|= IXGBE_MRQC_RSS_FIELD_IPV4_UDP
;
3129 if (adapter
->flags2
& IXGBE_FLAG2_RSS_FIELD_IPV6_UDP
)
3130 mrqc
|= IXGBE_MRQC_RSS_FIELD_IPV6_UDP
;
3132 IXGBE_WRITE_REG(hw
, IXGBE_MRQC
, mrqc
);
3136 * ixgbe_configure_rscctl - enable RSC for the indicated ring
3137 * @adapter: address of board private structure
3138 * @index: index of ring to set
3140 static void ixgbe_configure_rscctl(struct ixgbe_adapter
*adapter
,
3141 struct ixgbe_ring
*ring
)
3143 struct ixgbe_hw
*hw
= &adapter
->hw
;
3145 u8 reg_idx
= ring
->reg_idx
;
3147 if (!ring_is_rsc_enabled(ring
))
3150 rscctrl
= IXGBE_READ_REG(hw
, IXGBE_RSCCTL(reg_idx
));
3151 rscctrl
|= IXGBE_RSCCTL_RSCEN
;
3153 * we must limit the number of descriptors so that the
3154 * total size of max desc * buf_len is not greater
3157 rscctrl
|= IXGBE_RSCCTL_MAXDESC_16
;
3158 IXGBE_WRITE_REG(hw
, IXGBE_RSCCTL(reg_idx
), rscctrl
);
3161 #define IXGBE_MAX_RX_DESC_POLL 10
3162 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter
*adapter
,
3163 struct ixgbe_ring
*ring
)
3165 struct ixgbe_hw
*hw
= &adapter
->hw
;
3166 int wait_loop
= IXGBE_MAX_RX_DESC_POLL
;
3168 u8 reg_idx
= ring
->reg_idx
;
3170 /* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3171 if (hw
->mac
.type
== ixgbe_mac_82598EB
&&
3172 !(IXGBE_READ_REG(hw
, IXGBE_LINKS
) & IXGBE_LINKS_UP
))
3176 usleep_range(1000, 2000);
3177 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3178 } while (--wait_loop
&& !(rxdctl
& IXGBE_RXDCTL_ENABLE
));
3181 e_err(drv
, "RXDCTL.ENABLE on Rx queue %d not set within "
3182 "the polling period\n", reg_idx
);
3186 void ixgbe_disable_rx_queue(struct ixgbe_adapter
*adapter
,
3187 struct ixgbe_ring
*ring
)
3189 struct ixgbe_hw
*hw
= &adapter
->hw
;
3190 int wait_loop
= IXGBE_MAX_RX_DESC_POLL
;
3192 u8 reg_idx
= ring
->reg_idx
;
3194 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3195 rxdctl
&= ~IXGBE_RXDCTL_ENABLE
;
3197 /* write value back with RXDCTL.ENABLE bit cleared */
3198 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(reg_idx
), rxdctl
);
3200 if (hw
->mac
.type
== ixgbe_mac_82598EB
&&
3201 !(IXGBE_READ_REG(hw
, IXGBE_LINKS
) & IXGBE_LINKS_UP
))
3204 /* the hardware may take up to 100us to really disable the rx queue */
3207 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3208 } while (--wait_loop
&& (rxdctl
& IXGBE_RXDCTL_ENABLE
));
3211 e_err(drv
, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3212 "the polling period\n", reg_idx
);
3216 void ixgbe_configure_rx_ring(struct ixgbe_adapter
*adapter
,
3217 struct ixgbe_ring
*ring
)
3219 struct ixgbe_hw
*hw
= &adapter
->hw
;
3220 u64 rdba
= ring
->dma
;
3222 u8 reg_idx
= ring
->reg_idx
;
3224 /* disable queue to avoid issues while updating state */
3225 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3226 ixgbe_disable_rx_queue(adapter
, ring
);
3228 IXGBE_WRITE_REG(hw
, IXGBE_RDBAL(reg_idx
), (rdba
& DMA_BIT_MASK(32)));
3229 IXGBE_WRITE_REG(hw
, IXGBE_RDBAH(reg_idx
), (rdba
>> 32));
3230 IXGBE_WRITE_REG(hw
, IXGBE_RDLEN(reg_idx
),
3231 ring
->count
* sizeof(union ixgbe_adv_rx_desc
));
3232 IXGBE_WRITE_REG(hw
, IXGBE_RDH(reg_idx
), 0);
3233 IXGBE_WRITE_REG(hw
, IXGBE_RDT(reg_idx
), 0);
3234 ring
->tail
= hw
->hw_addr
+ IXGBE_RDT(reg_idx
);
3236 ixgbe_configure_srrctl(adapter
, ring
);
3237 ixgbe_configure_rscctl(adapter
, ring
);
3239 if (hw
->mac
.type
== ixgbe_mac_82598EB
) {
3241 * enable cache line friendly hardware writes:
3242 * PTHRESH=32 descriptors (half the internal cache),
3243 * this also removes ugly rx_no_buffer_count increment
3244 * HTHRESH=4 descriptors (to minimize latency on fetch)
3245 * WTHRESH=8 burst writeback up to two cache lines
3247 rxdctl
&= ~0x3FFFFF;
3251 /* enable receive descriptor ring */
3252 rxdctl
|= IXGBE_RXDCTL_ENABLE
;
3253 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(reg_idx
), rxdctl
);
3255 ixgbe_rx_desc_queue_enable(adapter
, ring
);
3256 ixgbe_alloc_rx_buffers(ring
, ixgbe_desc_unused(ring
));
3259 static void ixgbe_setup_psrtype(struct ixgbe_adapter
*adapter
)
3261 struct ixgbe_hw
*hw
= &adapter
->hw
;
3262 int rss_i
= adapter
->ring_feature
[RING_F_RSS
].indices
;
3265 /* PSRTYPE must be initialized in non 82598 adapters */
3266 u32 psrtype
= IXGBE_PSRTYPE_TCPHDR
|
3267 IXGBE_PSRTYPE_UDPHDR
|
3268 IXGBE_PSRTYPE_IPV4HDR
|
3269 IXGBE_PSRTYPE_L2HDR
|
3270 IXGBE_PSRTYPE_IPV6HDR
;
3272 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
3280 for (p
= 0; p
< adapter
->num_rx_pools
; p
++)
3281 IXGBE_WRITE_REG(hw
, IXGBE_PSRTYPE(VMDQ_P(p
)),
3285 static void ixgbe_configure_virtualization(struct ixgbe_adapter
*adapter
)
3287 struct ixgbe_hw
*hw
= &adapter
->hw
;
3288 u32 reg_offset
, vf_shift
;
3289 u32 gcr_ext
, vmdctl
;
3292 if (!(adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
))
3295 vmdctl
= IXGBE_READ_REG(hw
, IXGBE_VT_CTL
);
3296 vmdctl
|= IXGBE_VMD_CTL_VMDQ_EN
;
3297 vmdctl
&= ~IXGBE_VT_CTL_POOL_MASK
;
3298 vmdctl
|= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT
;
3299 vmdctl
|= IXGBE_VT_CTL_REPLEN
;
3300 IXGBE_WRITE_REG(hw
, IXGBE_VT_CTL
, vmdctl
);
3302 vf_shift
= VMDQ_P(0) % 32;
3303 reg_offset
= (VMDQ_P(0) >= 32) ? 1 : 0;
3305 /* Enable only the PF's pool for Tx/Rx */
3306 IXGBE_WRITE_REG(hw
, IXGBE_VFRE(reg_offset
), (~0) << vf_shift
);
3307 IXGBE_WRITE_REG(hw
, IXGBE_VFRE(reg_offset
^ 1), reg_offset
- 1);
3308 IXGBE_WRITE_REG(hw
, IXGBE_VFTE(reg_offset
), (~0) << vf_shift
);
3309 IXGBE_WRITE_REG(hw
, IXGBE_VFTE(reg_offset
^ 1), reg_offset
- 1);
3310 if (adapter
->flags2
& IXGBE_FLAG2_BRIDGE_MODE_VEB
)
3311 IXGBE_WRITE_REG(hw
, IXGBE_PFDTXGSWC
, IXGBE_PFDTXGSWC_VT_LBEN
);
3313 /* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3314 hw
->mac
.ops
.set_vmdq(hw
, 0, VMDQ_P(0));
3317 * Set up VF register offsets for selected VT Mode,
3318 * i.e. 32 or 64 VFs for SR-IOV
3320 switch (adapter
->ring_feature
[RING_F_VMDQ
].mask
) {
3321 case IXGBE_82599_VMDQ_8Q_MASK
:
3322 gcr_ext
= IXGBE_GCR_EXT_VT_MODE_16
;
3324 case IXGBE_82599_VMDQ_4Q_MASK
:
3325 gcr_ext
= IXGBE_GCR_EXT_VT_MODE_32
;
3328 gcr_ext
= IXGBE_GCR_EXT_VT_MODE_64
;
3332 IXGBE_WRITE_REG(hw
, IXGBE_GCR_EXT
, gcr_ext
);
3335 /* Enable MAC Anti-Spoofing */
3336 hw
->mac
.ops
.set_mac_anti_spoofing(hw
, (adapter
->num_vfs
!= 0),
3338 /* For VFs that have spoof checking turned off */
3339 for (i
= 0; i
< adapter
->num_vfs
; i
++) {
3340 if (!adapter
->vfinfo
[i
].spoofchk_enabled
)
3341 ixgbe_ndo_set_vf_spoofchk(adapter
->netdev
, i
, false);
3345 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter
*adapter
)
3347 struct ixgbe_hw
*hw
= &adapter
->hw
;
3348 struct net_device
*netdev
= adapter
->netdev
;
3349 int max_frame
= netdev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
3350 struct ixgbe_ring
*rx_ring
;
3355 /* adjust max frame to be able to do baby jumbo for FCoE */
3356 if ((adapter
->flags
& IXGBE_FLAG_FCOE_ENABLED
) &&
3357 (max_frame
< IXGBE_FCOE_JUMBO_FRAME_SIZE
))
3358 max_frame
= IXGBE_FCOE_JUMBO_FRAME_SIZE
;
3360 #endif /* IXGBE_FCOE */
3362 /* adjust max frame to be at least the size of a standard frame */
3363 if (max_frame
< (ETH_FRAME_LEN
+ ETH_FCS_LEN
))
3364 max_frame
= (ETH_FRAME_LEN
+ ETH_FCS_LEN
);
3366 mhadd
= IXGBE_READ_REG(hw
, IXGBE_MHADD
);
3367 if (max_frame
!= (mhadd
>> IXGBE_MHADD_MFS_SHIFT
)) {
3368 mhadd
&= ~IXGBE_MHADD_MFS_MASK
;
3369 mhadd
|= max_frame
<< IXGBE_MHADD_MFS_SHIFT
;
3371 IXGBE_WRITE_REG(hw
, IXGBE_MHADD
, mhadd
);
3374 hlreg0
= IXGBE_READ_REG(hw
, IXGBE_HLREG0
);
3375 /* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3376 hlreg0
|= IXGBE_HLREG0_JUMBOEN
;
3377 IXGBE_WRITE_REG(hw
, IXGBE_HLREG0
, hlreg0
);
3380 * Setup the HW Rx Head and Tail Descriptor Pointers and
3381 * the Base and Length of the Rx Descriptor Ring
3383 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3384 rx_ring
= adapter
->rx_ring
[i
];
3385 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)
3386 set_ring_rsc_enabled(rx_ring
);
3388 clear_ring_rsc_enabled(rx_ring
);
3392 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter
*adapter
)
3394 struct ixgbe_hw
*hw
= &adapter
->hw
;
3395 u32 rdrxctl
= IXGBE_READ_REG(hw
, IXGBE_RDRXCTL
);
3397 switch (hw
->mac
.type
) {
3398 case ixgbe_mac_82598EB
:
3400 * For VMDq support of different descriptor types or
3401 * buffer sizes through the use of multiple SRRCTL
3402 * registers, RDRXCTL.MVMEN must be set to 1
3404 * also, the manual doesn't mention it clearly but DCA hints
3405 * will only use queue 0's tags unless this bit is set. Side
3406 * effects of setting this bit are only that SRRCTL must be
3407 * fully programmed [0..15]
3409 rdrxctl
|= IXGBE_RDRXCTL_MVMEN
;
3411 case ixgbe_mac_82599EB
:
3412 case ixgbe_mac_X540
:
3413 /* Disable RSC for ACK packets */
3414 IXGBE_WRITE_REG(hw
, IXGBE_RSCDBU
,
3415 (IXGBE_RSCDBU_RSCACKDIS
| IXGBE_READ_REG(hw
, IXGBE_RSCDBU
)));
3416 rdrxctl
&= ~IXGBE_RDRXCTL_RSCFRSTSIZE
;
3417 /* hardware requires some bits to be set by default */
3418 rdrxctl
|= (IXGBE_RDRXCTL_RSCACKC
| IXGBE_RDRXCTL_FCOE_WRFIX
);
3419 rdrxctl
|= IXGBE_RDRXCTL_CRCSTRIP
;
3422 /* We should do nothing since we don't know this hardware */
3426 IXGBE_WRITE_REG(hw
, IXGBE_RDRXCTL
, rdrxctl
);
3430 * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3431 * @adapter: board private structure
3433 * Configure the Rx unit of the MAC after a reset.
3435 static void ixgbe_configure_rx(struct ixgbe_adapter
*adapter
)
3437 struct ixgbe_hw
*hw
= &adapter
->hw
;
3441 /* disable receives while setting up the descriptors */
3442 rxctrl
= IXGBE_READ_REG(hw
, IXGBE_RXCTRL
);
3443 IXGBE_WRITE_REG(hw
, IXGBE_RXCTRL
, rxctrl
& ~IXGBE_RXCTRL_RXEN
);
3445 ixgbe_setup_psrtype(adapter
);
3446 ixgbe_setup_rdrxctl(adapter
);
3448 /* Program registers for the distribution of queues */
3449 ixgbe_setup_mrqc(adapter
);
3451 /* set_rx_buffer_len must be called before ring initialization */
3452 ixgbe_set_rx_buffer_len(adapter
);
3455 * Setup the HW Rx Head and Tail Descriptor Pointers and
3456 * the Base and Length of the Rx Descriptor Ring
3458 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3459 ixgbe_configure_rx_ring(adapter
, adapter
->rx_ring
[i
]);
3461 /* disable drop enable for 82598 parts */
3462 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
3463 rxctrl
|= IXGBE_RXCTRL_DMBYPS
;
3465 /* enable all receives */
3466 rxctrl
|= IXGBE_RXCTRL_RXEN
;
3467 hw
->mac
.ops
.enable_rx_dma(hw
, rxctrl
);
3470 static int ixgbe_vlan_rx_add_vid(struct net_device
*netdev
, u16 vid
)
3472 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
3473 struct ixgbe_hw
*hw
= &adapter
->hw
;
3475 /* add VID to filter table */
3476 hw
->mac
.ops
.set_vfta(&adapter
->hw
, vid
, VMDQ_P(0), true);
3477 set_bit(vid
, adapter
->active_vlans
);
3482 static int ixgbe_vlan_rx_kill_vid(struct net_device
*netdev
, u16 vid
)
3484 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
3485 struct ixgbe_hw
*hw
= &adapter
->hw
;
3487 /* remove VID from filter table */
3488 hw
->mac
.ops
.set_vfta(&adapter
->hw
, vid
, VMDQ_P(0), false);
3489 clear_bit(vid
, adapter
->active_vlans
);
3495 * ixgbe_vlan_filter_disable - helper to disable hw vlan filtering
3496 * @adapter: driver data
3498 static void ixgbe_vlan_filter_disable(struct ixgbe_adapter
*adapter
)
3500 struct ixgbe_hw
*hw
= &adapter
->hw
;
3503 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_VLNCTRL
);
3504 vlnctrl
&= ~(IXGBE_VLNCTRL_VFE
| IXGBE_VLNCTRL_CFIEN
);
3505 IXGBE_WRITE_REG(hw
, IXGBE_VLNCTRL
, vlnctrl
);
3509 * ixgbe_vlan_filter_enable - helper to enable hw vlan filtering
3510 * @adapter: driver data
3512 static void ixgbe_vlan_filter_enable(struct ixgbe_adapter
*adapter
)
3514 struct ixgbe_hw
*hw
= &adapter
->hw
;
3517 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_VLNCTRL
);
3518 vlnctrl
|= IXGBE_VLNCTRL_VFE
;
3519 vlnctrl
&= ~IXGBE_VLNCTRL_CFIEN
;
3520 IXGBE_WRITE_REG(hw
, IXGBE_VLNCTRL
, vlnctrl
);
3524 * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3525 * @adapter: driver data
3527 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter
*adapter
)
3529 struct ixgbe_hw
*hw
= &adapter
->hw
;
3533 switch (hw
->mac
.type
) {
3534 case ixgbe_mac_82598EB
:
3535 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_VLNCTRL
);
3536 vlnctrl
&= ~IXGBE_VLNCTRL_VME
;
3537 IXGBE_WRITE_REG(hw
, IXGBE_VLNCTRL
, vlnctrl
);
3539 case ixgbe_mac_82599EB
:
3540 case ixgbe_mac_X540
:
3541 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3542 j
= adapter
->rx_ring
[i
]->reg_idx
;
3543 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(j
));
3544 vlnctrl
&= ~IXGBE_RXDCTL_VME
;
3545 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(j
), vlnctrl
);
3554 * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3555 * @adapter: driver data
3557 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter
*adapter
)
3559 struct ixgbe_hw
*hw
= &adapter
->hw
;
3563 switch (hw
->mac
.type
) {
3564 case ixgbe_mac_82598EB
:
3565 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_VLNCTRL
);
3566 vlnctrl
|= IXGBE_VLNCTRL_VME
;
3567 IXGBE_WRITE_REG(hw
, IXGBE_VLNCTRL
, vlnctrl
);
3569 case ixgbe_mac_82599EB
:
3570 case ixgbe_mac_X540
:
3571 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3572 j
= adapter
->rx_ring
[i
]->reg_idx
;
3573 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(j
));
3574 vlnctrl
|= IXGBE_RXDCTL_VME
;
3575 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(j
), vlnctrl
);
3583 static void ixgbe_restore_vlan(struct ixgbe_adapter
*adapter
)
3587 ixgbe_vlan_rx_add_vid(adapter
->netdev
, 0);
3589 for_each_set_bit(vid
, adapter
->active_vlans
, VLAN_N_VID
)
3590 ixgbe_vlan_rx_add_vid(adapter
->netdev
, vid
);
3594 * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
3595 * @netdev: network interface device structure
3597 * Writes unicast address list to the RAR table.
3598 * Returns: -ENOMEM on failure/insufficient address space
3599 * 0 on no addresses written
3600 * X on writing X addresses to the RAR table
3602 static int ixgbe_write_uc_addr_list(struct net_device
*netdev
)
3604 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
3605 struct ixgbe_hw
*hw
= &adapter
->hw
;
3606 unsigned int rar_entries
= hw
->mac
.num_rar_entries
- 1;
3609 /* In SR-IOV mode significantly less RAR entries are available */
3610 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
3611 rar_entries
= IXGBE_MAX_PF_MACVLANS
- 1;
3613 /* return ENOMEM indicating insufficient memory for addresses */
3614 if (netdev_uc_count(netdev
) > rar_entries
)
3617 if (!netdev_uc_empty(netdev
)) {
3618 struct netdev_hw_addr
*ha
;
3619 /* return error if we do not support writing to RAR table */
3620 if (!hw
->mac
.ops
.set_rar
)
3623 netdev_for_each_uc_addr(ha
, netdev
) {
3626 hw
->mac
.ops
.set_rar(hw
, rar_entries
--, ha
->addr
,
3627 VMDQ_P(0), IXGBE_RAH_AV
);
3631 /* write the addresses in reverse order to avoid write combining */
3632 for (; rar_entries
> 0 ; rar_entries
--)
3633 hw
->mac
.ops
.clear_rar(hw
, rar_entries
);
3639 * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
3640 * @netdev: network interface device structure
3642 * The set_rx_method entry point is called whenever the unicast/multicast
3643 * address list or the network interface flags are updated. This routine is
3644 * responsible for configuring the hardware for proper unicast, multicast and
3647 void ixgbe_set_rx_mode(struct net_device
*netdev
)
3649 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
3650 struct ixgbe_hw
*hw
= &adapter
->hw
;
3651 u32 fctrl
, vmolr
= IXGBE_VMOLR_BAM
| IXGBE_VMOLR_AUPE
;
3654 /* Check for Promiscuous and All Multicast modes */
3656 fctrl
= IXGBE_READ_REG(hw
, IXGBE_FCTRL
);
3658 /* set all bits that we expect to always be set */
3659 fctrl
&= ~IXGBE_FCTRL_SBP
; /* disable store-bad-packets */
3660 fctrl
|= IXGBE_FCTRL_BAM
;
3661 fctrl
|= IXGBE_FCTRL_DPF
; /* discard pause frames when FC enabled */
3662 fctrl
|= IXGBE_FCTRL_PMCF
;
3664 /* clear the bits we are changing the status of */
3665 fctrl
&= ~(IXGBE_FCTRL_UPE
| IXGBE_FCTRL_MPE
);
3667 if (netdev
->flags
& IFF_PROMISC
) {
3668 hw
->addr_ctrl
.user_set_promisc
= true;
3669 fctrl
|= (IXGBE_FCTRL_UPE
| IXGBE_FCTRL_MPE
);
3670 vmolr
|= (IXGBE_VMOLR_ROPE
| IXGBE_VMOLR_MPE
);
3671 /* don't hardware filter vlans in promisc mode */
3672 ixgbe_vlan_filter_disable(adapter
);
3674 if (netdev
->flags
& IFF_ALLMULTI
) {
3675 fctrl
|= IXGBE_FCTRL_MPE
;
3676 vmolr
|= IXGBE_VMOLR_MPE
;
3679 * Write addresses to the MTA, if the attempt fails
3680 * then we should just turn on promiscuous mode so
3681 * that we can at least receive multicast traffic
3683 hw
->mac
.ops
.update_mc_addr_list(hw
, netdev
);
3684 vmolr
|= IXGBE_VMOLR_ROMPE
;
3686 ixgbe_vlan_filter_enable(adapter
);
3687 hw
->addr_ctrl
.user_set_promisc
= false;
3691 * Write addresses to available RAR registers, if there is not
3692 * sufficient space to store all the addresses then enable
3693 * unicast promiscuous mode
3695 count
= ixgbe_write_uc_addr_list(netdev
);
3697 fctrl
|= IXGBE_FCTRL_UPE
;
3698 vmolr
|= IXGBE_VMOLR_ROPE
;
3701 if (adapter
->num_vfs
)
3702 ixgbe_restore_vf_multicasts(adapter
);
3704 if (hw
->mac
.type
!= ixgbe_mac_82598EB
) {
3705 vmolr
|= IXGBE_READ_REG(hw
, IXGBE_VMOLR(VMDQ_P(0))) &
3706 ~(IXGBE_VMOLR_MPE
| IXGBE_VMOLR_ROMPE
|
3708 IXGBE_WRITE_REG(hw
, IXGBE_VMOLR(VMDQ_P(0)), vmolr
);
3711 /* This is useful for sniffing bad packets. */
3712 if (adapter
->netdev
->features
& NETIF_F_RXALL
) {
3713 /* UPE and MPE will be handled by normal PROMISC logic
3714 * in e1000e_set_rx_mode */
3715 fctrl
|= (IXGBE_FCTRL_SBP
| /* Receive bad packets */
3716 IXGBE_FCTRL_BAM
| /* RX All Bcast Pkts */
3717 IXGBE_FCTRL_PMCF
); /* RX All MAC Ctrl Pkts */
3719 fctrl
&= ~(IXGBE_FCTRL_DPF
);
3720 /* NOTE: VLAN filtering is disabled by setting PROMISC */
3723 IXGBE_WRITE_REG(hw
, IXGBE_FCTRL
, fctrl
);
3725 if (netdev
->features
& NETIF_F_HW_VLAN_CTAG_RX
)
3726 ixgbe_vlan_strip_enable(adapter
);
3728 ixgbe_vlan_strip_disable(adapter
);
3731 static void ixgbe_napi_enable_all(struct ixgbe_adapter
*adapter
)
3735 for (q_idx
= 0; q_idx
< adapter
->num_q_vectors
; q_idx
++)
3736 napi_enable(&adapter
->q_vector
[q_idx
]->napi
);
3739 static void ixgbe_napi_disable_all(struct ixgbe_adapter
*adapter
)
3743 for (q_idx
= 0; q_idx
< adapter
->num_q_vectors
; q_idx
++)
3744 napi_disable(&adapter
->q_vector
[q_idx
]->napi
);
3747 #ifdef CONFIG_IXGBE_DCB
3749 * ixgbe_configure_dcb - Configure DCB hardware
3750 * @adapter: ixgbe adapter struct
3752 * This is called by the driver on open to configure the DCB hardware.
3753 * This is also called by the gennetlink interface when reconfiguring
3756 static void ixgbe_configure_dcb(struct ixgbe_adapter
*adapter
)
3758 struct ixgbe_hw
*hw
= &adapter
->hw
;
3759 int max_frame
= adapter
->netdev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
3761 if (!(adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
)) {
3762 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
3763 netif_set_gso_max_size(adapter
->netdev
, 65536);
3767 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
3768 netif_set_gso_max_size(adapter
->netdev
, 32768);
3771 if (adapter
->netdev
->features
& NETIF_F_FCOE_MTU
)
3772 max_frame
= max(max_frame
, IXGBE_FCOE_JUMBO_FRAME_SIZE
);
3775 /* reconfigure the hardware */
3776 if (adapter
->dcbx_cap
& DCB_CAP_DCBX_VER_CEE
) {
3777 ixgbe_dcb_calculate_tc_credits(hw
, &adapter
->dcb_cfg
, max_frame
,
3779 ixgbe_dcb_calculate_tc_credits(hw
, &adapter
->dcb_cfg
, max_frame
,
3781 ixgbe_dcb_hw_config(hw
, &adapter
->dcb_cfg
);
3782 } else if (adapter
->ixgbe_ieee_ets
&& adapter
->ixgbe_ieee_pfc
) {
3783 ixgbe_dcb_hw_ets(&adapter
->hw
,
3784 adapter
->ixgbe_ieee_ets
,
3786 ixgbe_dcb_hw_pfc_config(&adapter
->hw
,
3787 adapter
->ixgbe_ieee_pfc
->pfc_en
,
3788 adapter
->ixgbe_ieee_ets
->prio_tc
);
3791 /* Enable RSS Hash per TC */
3792 if (hw
->mac
.type
!= ixgbe_mac_82598EB
) {
3794 u16 rss_i
= adapter
->ring_feature
[RING_F_RSS
].indices
- 1;
3801 /* write msb to all 8 TCs in one write */
3802 IXGBE_WRITE_REG(hw
, IXGBE_RQTC
, msb
* 0x11111111);
3807 /* Additional bittime to account for IXGBE framing */
3808 #define IXGBE_ETH_FRAMING 20
3811 * ixgbe_hpbthresh - calculate high water mark for flow control
3813 * @adapter: board private structure to calculate for
3814 * @pb: packet buffer to calculate
3816 static int ixgbe_hpbthresh(struct ixgbe_adapter
*adapter
, int pb
)
3818 struct ixgbe_hw
*hw
= &adapter
->hw
;
3819 struct net_device
*dev
= adapter
->netdev
;
3820 int link
, tc
, kb
, marker
;
3823 /* Calculate max LAN frame size */
3824 tc
= link
= dev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
+ IXGBE_ETH_FRAMING
;
3827 /* FCoE traffic class uses FCOE jumbo frames */
3828 if ((dev
->features
& NETIF_F_FCOE_MTU
) &&
3829 (tc
< IXGBE_FCOE_JUMBO_FRAME_SIZE
) &&
3830 (pb
== ixgbe_fcoe_get_tc(adapter
)))
3831 tc
= IXGBE_FCOE_JUMBO_FRAME_SIZE
;
3834 /* Calculate delay value for device */
3835 switch (hw
->mac
.type
) {
3836 case ixgbe_mac_X540
:
3837 dv_id
= IXGBE_DV_X540(link
, tc
);
3840 dv_id
= IXGBE_DV(link
, tc
);
3844 /* Loopback switch introduces additional latency */
3845 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
3846 dv_id
+= IXGBE_B2BT(tc
);
3848 /* Delay value is calculated in bit times convert to KB */
3849 kb
= IXGBE_BT2KB(dv_id
);
3850 rx_pba
= IXGBE_READ_REG(hw
, IXGBE_RXPBSIZE(pb
)) >> 10;
3852 marker
= rx_pba
- kb
;
3854 /* It is possible that the packet buffer is not large enough
3855 * to provide required headroom. In this case throw an error
3856 * to user and a do the best we can.
3859 e_warn(drv
, "Packet Buffer(%i) can not provide enough"
3860 "headroom to support flow control."
3861 "Decrease MTU or number of traffic classes\n", pb
);
3869 * ixgbe_lpbthresh - calculate low water mark for for flow control
3871 * @adapter: board private structure to calculate for
3872 * @pb: packet buffer to calculate
3874 static int ixgbe_lpbthresh(struct ixgbe_adapter
*adapter
)
3876 struct ixgbe_hw
*hw
= &adapter
->hw
;
3877 struct net_device
*dev
= adapter
->netdev
;
3881 /* Calculate max LAN frame size */
3882 tc
= dev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
3884 /* Calculate delay value for device */
3885 switch (hw
->mac
.type
) {
3886 case ixgbe_mac_X540
:
3887 dv_id
= IXGBE_LOW_DV_X540(tc
);
3890 dv_id
= IXGBE_LOW_DV(tc
);
3894 /* Delay value is calculated in bit times convert to KB */
3895 return IXGBE_BT2KB(dv_id
);
3899 * ixgbe_pbthresh_setup - calculate and setup high low water marks
3901 static void ixgbe_pbthresh_setup(struct ixgbe_adapter
*adapter
)
3903 struct ixgbe_hw
*hw
= &adapter
->hw
;
3904 int num_tc
= netdev_get_num_tc(adapter
->netdev
);
3910 hw
->fc
.low_water
= ixgbe_lpbthresh(adapter
);
3912 for (i
= 0; i
< num_tc
; i
++) {
3913 hw
->fc
.high_water
[i
] = ixgbe_hpbthresh(adapter
, i
);
3915 /* Low water marks must not be larger than high water marks */
3916 if (hw
->fc
.low_water
> hw
->fc
.high_water
[i
])
3917 hw
->fc
.low_water
= 0;
3921 static void ixgbe_configure_pb(struct ixgbe_adapter
*adapter
)
3923 struct ixgbe_hw
*hw
= &adapter
->hw
;
3925 u8 tc
= netdev_get_num_tc(adapter
->netdev
);
3927 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
||
3928 adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
)
3929 hdrm
= 32 << adapter
->fdir_pballoc
;
3933 hw
->mac
.ops
.set_rxpba(hw
, tc
, hdrm
, PBA_STRATEGY_EQUAL
);
3934 ixgbe_pbthresh_setup(adapter
);
3937 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter
*adapter
)
3939 struct ixgbe_hw
*hw
= &adapter
->hw
;
3940 struct hlist_node
*node2
;
3941 struct ixgbe_fdir_filter
*filter
;
3943 spin_lock(&adapter
->fdir_perfect_lock
);
3945 if (!hlist_empty(&adapter
->fdir_filter_list
))
3946 ixgbe_fdir_set_input_mask_82599(hw
, &adapter
->fdir_mask
);
3948 hlist_for_each_entry_safe(filter
, node2
,
3949 &adapter
->fdir_filter_list
, fdir_node
) {
3950 ixgbe_fdir_write_perfect_filter_82599(hw
,
3953 (filter
->action
== IXGBE_FDIR_DROP_QUEUE
) ?
3954 IXGBE_FDIR_DROP_QUEUE
:
3955 adapter
->rx_ring
[filter
->action
]->reg_idx
);
3958 spin_unlock(&adapter
->fdir_perfect_lock
);
3961 static void ixgbe_configure(struct ixgbe_adapter
*adapter
)
3963 struct ixgbe_hw
*hw
= &adapter
->hw
;
3965 ixgbe_configure_pb(adapter
);
3966 #ifdef CONFIG_IXGBE_DCB
3967 ixgbe_configure_dcb(adapter
);
3970 * We must restore virtualization before VLANs or else
3971 * the VLVF registers will not be populated
3973 ixgbe_configure_virtualization(adapter
);
3975 ixgbe_set_rx_mode(adapter
->netdev
);
3976 ixgbe_restore_vlan(adapter
);
3978 switch (hw
->mac
.type
) {
3979 case ixgbe_mac_82599EB
:
3980 case ixgbe_mac_X540
:
3981 hw
->mac
.ops
.disable_rx_buff(hw
);
3987 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) {
3988 ixgbe_init_fdir_signature_82599(&adapter
->hw
,
3989 adapter
->fdir_pballoc
);
3990 } else if (adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
) {
3991 ixgbe_init_fdir_perfect_82599(&adapter
->hw
,
3992 adapter
->fdir_pballoc
);
3993 ixgbe_fdir_filter_restore(adapter
);
3996 switch (hw
->mac
.type
) {
3997 case ixgbe_mac_82599EB
:
3998 case ixgbe_mac_X540
:
3999 hw
->mac
.ops
.enable_rx_buff(hw
);
4006 /* configure FCoE L2 filters, redirection table, and Rx control */
4007 ixgbe_configure_fcoe(adapter
);
4009 #endif /* IXGBE_FCOE */
4010 ixgbe_configure_tx(adapter
);
4011 ixgbe_configure_rx(adapter
);
4014 static inline bool ixgbe_is_sfp(struct ixgbe_hw
*hw
)
4016 switch (hw
->phy
.type
) {
4017 case ixgbe_phy_sfp_avago
:
4018 case ixgbe_phy_sfp_ftl
:
4019 case ixgbe_phy_sfp_intel
:
4020 case ixgbe_phy_sfp_unknown
:
4021 case ixgbe_phy_sfp_passive_tyco
:
4022 case ixgbe_phy_sfp_passive_unknown
:
4023 case ixgbe_phy_sfp_active_unknown
:
4024 case ixgbe_phy_sfp_ftl_active
:
4027 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
4035 * ixgbe_sfp_link_config - set up SFP+ link
4036 * @adapter: pointer to private adapter struct
4038 static void ixgbe_sfp_link_config(struct ixgbe_adapter
*adapter
)
4041 * We are assuming the worst case scenario here, and that
4042 * is that an SFP was inserted/removed after the reset
4043 * but before SFP detection was enabled. As such the best
4044 * solution is to just start searching as soon as we start
4046 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
)
4047 adapter
->flags2
|= IXGBE_FLAG2_SEARCH_FOR_SFP
;
4049 adapter
->flags2
|= IXGBE_FLAG2_SFP_NEEDS_RESET
;
4053 * ixgbe_non_sfp_link_config - set up non-SFP+ link
4054 * @hw: pointer to private hardware struct
4056 * Returns 0 on success, negative on failure
4058 static int ixgbe_non_sfp_link_config(struct ixgbe_hw
*hw
)
4061 bool autoneg
, link_up
= false;
4062 u32 ret
= IXGBE_ERR_LINK_SETUP
;
4064 if (hw
->mac
.ops
.check_link
)
4065 ret
= hw
->mac
.ops
.check_link(hw
, &speed
, &link_up
, false);
4070 speed
= hw
->phy
.autoneg_advertised
;
4071 if ((!speed
) && (hw
->mac
.ops
.get_link_capabilities
))
4072 ret
= hw
->mac
.ops
.get_link_capabilities(hw
, &speed
,
4077 if (hw
->mac
.ops
.setup_link
)
4078 ret
= hw
->mac
.ops
.setup_link(hw
, speed
, link_up
);
4083 static void ixgbe_setup_gpie(struct ixgbe_adapter
*adapter
)
4085 struct ixgbe_hw
*hw
= &adapter
->hw
;
4088 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
) {
4089 gpie
= IXGBE_GPIE_MSIX_MODE
| IXGBE_GPIE_PBA_SUPPORT
|
4091 gpie
|= IXGBE_GPIE_EIAME
;
4093 * use EIAM to auto-mask when MSI-X interrupt is asserted
4094 * this saves a register write for every interrupt
4096 switch (hw
->mac
.type
) {
4097 case ixgbe_mac_82598EB
:
4098 IXGBE_WRITE_REG(hw
, IXGBE_EIAM
, IXGBE_EICS_RTX_QUEUE
);
4100 case ixgbe_mac_82599EB
:
4101 case ixgbe_mac_X540
:
4103 IXGBE_WRITE_REG(hw
, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4104 IXGBE_WRITE_REG(hw
, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4108 /* legacy interrupts, use EIAM to auto-mask when reading EICR,
4109 * specifically only auto mask tx and rx interrupts */
4110 IXGBE_WRITE_REG(hw
, IXGBE_EIAM
, IXGBE_EICS_RTX_QUEUE
);
4113 /* XXX: to interrupt immediately for EICS writes, enable this */
4114 /* gpie |= IXGBE_GPIE_EIMEN; */
4116 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
4117 gpie
&= ~IXGBE_GPIE_VTMODE_MASK
;
4119 switch (adapter
->ring_feature
[RING_F_VMDQ
].mask
) {
4120 case IXGBE_82599_VMDQ_8Q_MASK
:
4121 gpie
|= IXGBE_GPIE_VTMODE_16
;
4123 case IXGBE_82599_VMDQ_4Q_MASK
:
4124 gpie
|= IXGBE_GPIE_VTMODE_32
;
4127 gpie
|= IXGBE_GPIE_VTMODE_64
;
4132 /* Enable Thermal over heat sensor interrupt */
4133 if (adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
) {
4134 switch (adapter
->hw
.mac
.type
) {
4135 case ixgbe_mac_82599EB
:
4136 gpie
|= IXGBE_SDP0_GPIEN
;
4138 case ixgbe_mac_X540
:
4139 gpie
|= IXGBE_EIMS_TS
;
4146 /* Enable fan failure interrupt */
4147 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
)
4148 gpie
|= IXGBE_SDP1_GPIEN
;
4150 if (hw
->mac
.type
== ixgbe_mac_82599EB
) {
4151 gpie
|= IXGBE_SDP1_GPIEN
;
4152 gpie
|= IXGBE_SDP2_GPIEN
;
4155 IXGBE_WRITE_REG(hw
, IXGBE_GPIE
, gpie
);
4158 static void ixgbe_up_complete(struct ixgbe_adapter
*adapter
)
4160 struct ixgbe_hw
*hw
= &adapter
->hw
;
4164 ixgbe_get_hw_control(adapter
);
4165 ixgbe_setup_gpie(adapter
);
4167 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)
4168 ixgbe_configure_msix(adapter
);
4170 ixgbe_configure_msi_and_legacy(adapter
);
4172 /* enable the optics for 82599 SFP+ fiber */
4173 if (hw
->mac
.ops
.enable_tx_laser
)
4174 hw
->mac
.ops
.enable_tx_laser(hw
);
4176 clear_bit(__IXGBE_DOWN
, &adapter
->state
);
4177 ixgbe_napi_enable_all(adapter
);
4179 if (ixgbe_is_sfp(hw
)) {
4180 ixgbe_sfp_link_config(adapter
);
4182 err
= ixgbe_non_sfp_link_config(hw
);
4184 e_err(probe
, "link_config FAILED %d\n", err
);
4187 /* clear any pending interrupts, may auto mask */
4188 IXGBE_READ_REG(hw
, IXGBE_EICR
);
4189 ixgbe_irq_enable(adapter
, true, true);
4192 * If this adapter has a fan, check to see if we had a failure
4193 * before we enabled the interrupt.
4195 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
) {
4196 u32 esdp
= IXGBE_READ_REG(hw
, IXGBE_ESDP
);
4197 if (esdp
& IXGBE_ESDP_SDP1
)
4198 e_crit(drv
, "Fan has stopped, replace the adapter\n");
4201 /* enable transmits */
4202 netif_tx_start_all_queues(adapter
->netdev
);
4204 /* bring the link up in the watchdog, this could race with our first
4205 * link up interrupt but shouldn't be a problem */
4206 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_UPDATE
;
4207 adapter
->link_check_timeout
= jiffies
;
4208 mod_timer(&adapter
->service_timer
, jiffies
);
4210 /* Set PF Reset Done bit so PF/VF Mail Ops can work */
4211 ctrl_ext
= IXGBE_READ_REG(hw
, IXGBE_CTRL_EXT
);
4212 ctrl_ext
|= IXGBE_CTRL_EXT_PFRSTD
;
4213 IXGBE_WRITE_REG(hw
, IXGBE_CTRL_EXT
, ctrl_ext
);
4216 void ixgbe_reinit_locked(struct ixgbe_adapter
*adapter
)
4218 WARN_ON(in_interrupt());
4219 /* put off any impending NetWatchDogTimeout */
4220 adapter
->netdev
->trans_start
= jiffies
;
4222 while (test_and_set_bit(__IXGBE_RESETTING
, &adapter
->state
))
4223 usleep_range(1000, 2000);
4224 ixgbe_down(adapter
);
4226 * If SR-IOV enabled then wait a bit before bringing the adapter
4227 * back up to give the VFs time to respond to the reset. The
4228 * two second wait is based upon the watchdog timer cycle in
4231 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
4234 clear_bit(__IXGBE_RESETTING
, &adapter
->state
);
4237 void ixgbe_up(struct ixgbe_adapter
*adapter
)
4239 /* hardware has been reset, we need to reload some things */
4240 ixgbe_configure(adapter
);
4242 ixgbe_up_complete(adapter
);
4245 void ixgbe_reset(struct ixgbe_adapter
*adapter
)
4247 struct ixgbe_hw
*hw
= &adapter
->hw
;
4250 /* lock SFP init bit to prevent race conditions with the watchdog */
4251 while (test_and_set_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
))
4252 usleep_range(1000, 2000);
4254 /* clear all SFP and link config related flags while holding SFP_INIT */
4255 adapter
->flags2
&= ~(IXGBE_FLAG2_SEARCH_FOR_SFP
|
4256 IXGBE_FLAG2_SFP_NEEDS_RESET
);
4257 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_CONFIG
;
4259 err
= hw
->mac
.ops
.init_hw(hw
);
4262 case IXGBE_ERR_SFP_NOT_PRESENT
:
4263 case IXGBE_ERR_SFP_NOT_SUPPORTED
:
4265 case IXGBE_ERR_MASTER_REQUESTS_PENDING
:
4266 e_dev_err("master disable timed out\n");
4268 case IXGBE_ERR_EEPROM_VERSION
:
4269 /* We are running on a pre-production device, log a warning */
4270 e_dev_warn("This device is a pre-production adapter/LOM. "
4271 "Please be aware there may be issues associated with "
4272 "your hardware. If you are experiencing problems "
4273 "please contact your Intel or hardware "
4274 "representative who provided you with this "
4278 e_dev_err("Hardware Error: %d\n", err
);
4281 clear_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
);
4283 /* reprogram the RAR[0] in case user changed it. */
4284 hw
->mac
.ops
.set_rar(hw
, 0, hw
->mac
.addr
, VMDQ_P(0), IXGBE_RAH_AV
);
4286 /* update SAN MAC vmdq pool selection */
4287 if (hw
->mac
.san_mac_rar_index
)
4288 hw
->mac
.ops
.set_vmdq_san_mac(hw
, VMDQ_P(0));
4290 if (adapter
->flags2
& IXGBE_FLAG2_PTP_ENABLED
)
4291 ixgbe_ptp_reset(adapter
);
4295 * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4296 * @rx_ring: ring to free buffers from
4298 static void ixgbe_clean_rx_ring(struct ixgbe_ring
*rx_ring
)
4300 struct device
*dev
= rx_ring
->dev
;
4304 /* ring already cleared, nothing to do */
4305 if (!rx_ring
->rx_buffer_info
)
4308 /* Free all the Rx ring sk_buffs */
4309 for (i
= 0; i
< rx_ring
->count
; i
++) {
4310 struct ixgbe_rx_buffer
*rx_buffer
;
4312 rx_buffer
= &rx_ring
->rx_buffer_info
[i
];
4313 if (rx_buffer
->skb
) {
4314 struct sk_buff
*skb
= rx_buffer
->skb
;
4315 if (IXGBE_CB(skb
)->page_released
) {
4318 ixgbe_rx_bufsz(rx_ring
),
4320 IXGBE_CB(skb
)->page_released
= false;
4324 rx_buffer
->skb
= NULL
;
4326 dma_unmap_page(dev
, rx_buffer
->dma
,
4327 ixgbe_rx_pg_size(rx_ring
),
4330 if (rx_buffer
->page
)
4331 __free_pages(rx_buffer
->page
,
4332 ixgbe_rx_pg_order(rx_ring
));
4333 rx_buffer
->page
= NULL
;
4336 size
= sizeof(struct ixgbe_rx_buffer
) * rx_ring
->count
;
4337 memset(rx_ring
->rx_buffer_info
, 0, size
);
4339 /* Zero out the descriptor ring */
4340 memset(rx_ring
->desc
, 0, rx_ring
->size
);
4342 rx_ring
->next_to_alloc
= 0;
4343 rx_ring
->next_to_clean
= 0;
4344 rx_ring
->next_to_use
= 0;
4348 * ixgbe_clean_tx_ring - Free Tx Buffers
4349 * @tx_ring: ring to be cleaned
4351 static void ixgbe_clean_tx_ring(struct ixgbe_ring
*tx_ring
)
4353 struct ixgbe_tx_buffer
*tx_buffer_info
;
4357 /* ring already cleared, nothing to do */
4358 if (!tx_ring
->tx_buffer_info
)
4361 /* Free all the Tx ring sk_buffs */
4362 for (i
= 0; i
< tx_ring
->count
; i
++) {
4363 tx_buffer_info
= &tx_ring
->tx_buffer_info
[i
];
4364 ixgbe_unmap_and_free_tx_resource(tx_ring
, tx_buffer_info
);
4367 netdev_tx_reset_queue(txring_txq(tx_ring
));
4369 size
= sizeof(struct ixgbe_tx_buffer
) * tx_ring
->count
;
4370 memset(tx_ring
->tx_buffer_info
, 0, size
);
4372 /* Zero out the descriptor ring */
4373 memset(tx_ring
->desc
, 0, tx_ring
->size
);
4375 tx_ring
->next_to_use
= 0;
4376 tx_ring
->next_to_clean
= 0;
4380 * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
4381 * @adapter: board private structure
4383 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter
*adapter
)
4387 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
4388 ixgbe_clean_rx_ring(adapter
->rx_ring
[i
]);
4392 * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
4393 * @adapter: board private structure
4395 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter
*adapter
)
4399 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
4400 ixgbe_clean_tx_ring(adapter
->tx_ring
[i
]);
4403 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter
*adapter
)
4405 struct hlist_node
*node2
;
4406 struct ixgbe_fdir_filter
*filter
;
4408 spin_lock(&adapter
->fdir_perfect_lock
);
4410 hlist_for_each_entry_safe(filter
, node2
,
4411 &adapter
->fdir_filter_list
, fdir_node
) {
4412 hlist_del(&filter
->fdir_node
);
4415 adapter
->fdir_filter_count
= 0;
4417 spin_unlock(&adapter
->fdir_perfect_lock
);
4420 void ixgbe_down(struct ixgbe_adapter
*adapter
)
4422 struct net_device
*netdev
= adapter
->netdev
;
4423 struct ixgbe_hw
*hw
= &adapter
->hw
;
4427 /* signal that we are down to the interrupt handler */
4428 set_bit(__IXGBE_DOWN
, &adapter
->state
);
4430 /* disable receives */
4431 rxctrl
= IXGBE_READ_REG(hw
, IXGBE_RXCTRL
);
4432 IXGBE_WRITE_REG(hw
, IXGBE_RXCTRL
, rxctrl
& ~IXGBE_RXCTRL_RXEN
);
4434 /* disable all enabled rx queues */
4435 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
4436 /* this call also flushes the previous write */
4437 ixgbe_disable_rx_queue(adapter
, adapter
->rx_ring
[i
]);
4439 usleep_range(10000, 20000);
4441 netif_tx_stop_all_queues(netdev
);
4443 /* call carrier off first to avoid false dev_watchdog timeouts */
4444 netif_carrier_off(netdev
);
4445 netif_tx_disable(netdev
);
4447 ixgbe_irq_disable(adapter
);
4449 ixgbe_napi_disable_all(adapter
);
4451 adapter
->flags2
&= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT
|
4452 IXGBE_FLAG2_RESET_REQUESTED
);
4453 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_UPDATE
;
4455 del_timer_sync(&adapter
->service_timer
);
4457 if (adapter
->num_vfs
) {
4458 /* Clear EITR Select mapping */
4459 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EITRSEL
, 0);
4461 /* Mark all the VFs as inactive */
4462 for (i
= 0 ; i
< adapter
->num_vfs
; i
++)
4463 adapter
->vfinfo
[i
].clear_to_send
= false;
4465 /* ping all the active vfs to let them know we are going down */
4466 ixgbe_ping_all_vfs(adapter
);
4468 /* Disable all VFTE/VFRE TX/RX */
4469 ixgbe_disable_tx_rx(adapter
);
4472 /* disable transmits in the hardware now that interrupts are off */
4473 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
4474 u8 reg_idx
= adapter
->tx_ring
[i
]->reg_idx
;
4475 IXGBE_WRITE_REG(hw
, IXGBE_TXDCTL(reg_idx
), IXGBE_TXDCTL_SWFLSH
);
4478 /* Disable the Tx DMA engine on 82599 and X540 */
4479 switch (hw
->mac
.type
) {
4480 case ixgbe_mac_82599EB
:
4481 case ixgbe_mac_X540
:
4482 IXGBE_WRITE_REG(hw
, IXGBE_DMATXCTL
,
4483 (IXGBE_READ_REG(hw
, IXGBE_DMATXCTL
) &
4484 ~IXGBE_DMATXCTL_TE
));
4490 if (!pci_channel_offline(adapter
->pdev
))
4491 ixgbe_reset(adapter
);
4493 /* power down the optics for 82599 SFP+ fiber */
4494 if (hw
->mac
.ops
.disable_tx_laser
)
4495 hw
->mac
.ops
.disable_tx_laser(hw
);
4497 ixgbe_clean_all_tx_rings(adapter
);
4498 ixgbe_clean_all_rx_rings(adapter
);
4500 #ifdef CONFIG_IXGBE_DCA
4501 /* since we reset the hardware DCA settings were cleared */
4502 ixgbe_setup_dca(adapter
);
4507 * ixgbe_tx_timeout - Respond to a Tx Hang
4508 * @netdev: network interface device structure
4510 static void ixgbe_tx_timeout(struct net_device
*netdev
)
4512 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
4514 /* Do the reset outside of interrupt context */
4515 ixgbe_tx_timeout_reset(adapter
);
4519 * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
4520 * @adapter: board private structure to initialize
4522 * ixgbe_sw_init initializes the Adapter private data structure.
4523 * Fields are initialized based on PCI device information and
4524 * OS network device settings (MTU size).
4526 static int ixgbe_sw_init(struct ixgbe_adapter
*adapter
)
4528 struct ixgbe_hw
*hw
= &adapter
->hw
;
4529 struct pci_dev
*pdev
= adapter
->pdev
;
4530 unsigned int rss
, fdir
;
4532 #ifdef CONFIG_IXGBE_DCB
4534 struct tc_configuration
*tc
;
4537 /* PCI config space info */
4539 hw
->vendor_id
= pdev
->vendor
;
4540 hw
->device_id
= pdev
->device
;
4541 hw
->revision_id
= pdev
->revision
;
4542 hw
->subsystem_vendor_id
= pdev
->subsystem_vendor
;
4543 hw
->subsystem_device_id
= pdev
->subsystem_device
;
4545 /* Set common capability flags and settings */
4546 rss
= min_t(int, IXGBE_MAX_RSS_INDICES
, num_online_cpus());
4547 adapter
->ring_feature
[RING_F_RSS
].limit
= rss
;
4548 adapter
->flags2
|= IXGBE_FLAG2_RSC_CAPABLE
;
4549 adapter
->flags2
|= IXGBE_FLAG2_RSC_ENABLED
;
4550 adapter
->max_q_vectors
= MAX_Q_VECTORS_82599
;
4551 adapter
->atr_sample_rate
= 20;
4552 fdir
= min_t(int, IXGBE_MAX_FDIR_INDICES
, num_online_cpus());
4553 adapter
->ring_feature
[RING_F_FDIR
].limit
= fdir
;
4554 adapter
->fdir_pballoc
= IXGBE_FDIR_PBALLOC_64K
;
4555 #ifdef CONFIG_IXGBE_DCA
4556 adapter
->flags
|= IXGBE_FLAG_DCA_CAPABLE
;
4559 adapter
->flags
|= IXGBE_FLAG_FCOE_CAPABLE
;
4560 adapter
->flags
&= ~IXGBE_FLAG_FCOE_ENABLED
;
4561 #ifdef CONFIG_IXGBE_DCB
4562 /* Default traffic class to use for FCoE */
4563 adapter
->fcoe
.up
= IXGBE_FCOE_DEFTC
;
4564 #endif /* CONFIG_IXGBE_DCB */
4565 #endif /* IXGBE_FCOE */
4567 /* Set MAC specific capability flags and exceptions */
4568 switch (hw
->mac
.type
) {
4569 case ixgbe_mac_82598EB
:
4570 adapter
->flags2
&= ~IXGBE_FLAG2_RSC_CAPABLE
;
4571 adapter
->flags2
&= ~IXGBE_FLAG2_RSC_ENABLED
;
4573 if (hw
->device_id
== IXGBE_DEV_ID_82598AT
)
4574 adapter
->flags
|= IXGBE_FLAG_FAN_FAIL_CAPABLE
;
4576 adapter
->max_q_vectors
= MAX_Q_VECTORS_82598
;
4577 adapter
->ring_feature
[RING_F_FDIR
].limit
= 0;
4578 adapter
->atr_sample_rate
= 0;
4579 adapter
->fdir_pballoc
= 0;
4581 adapter
->flags
&= ~IXGBE_FLAG_FCOE_CAPABLE
;
4582 adapter
->flags
&= ~IXGBE_FLAG_FCOE_ENABLED
;
4583 #ifdef CONFIG_IXGBE_DCB
4584 adapter
->fcoe
.up
= 0;
4585 #endif /* IXGBE_DCB */
4586 #endif /* IXGBE_FCOE */
4588 case ixgbe_mac_82599EB
:
4589 if (hw
->device_id
== IXGBE_DEV_ID_82599_T3_LOM
)
4590 adapter
->flags2
|= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
;
4592 case ixgbe_mac_X540
:
4593 fwsm
= IXGBE_READ_REG(hw
, IXGBE_FWSM
);
4594 if (fwsm
& IXGBE_FWSM_TS_ENABLED
)
4595 adapter
->flags2
|= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
;
4602 /* FCoE support exists, always init the FCoE lock */
4603 spin_lock_init(&adapter
->fcoe
.lock
);
4606 /* n-tuple support exists, always init our spinlock */
4607 spin_lock_init(&adapter
->fdir_perfect_lock
);
4609 #ifdef CONFIG_IXGBE_DCB
4610 switch (hw
->mac
.type
) {
4611 case ixgbe_mac_X540
:
4612 adapter
->dcb_cfg
.num_tcs
.pg_tcs
= X540_TRAFFIC_CLASS
;
4613 adapter
->dcb_cfg
.num_tcs
.pfc_tcs
= X540_TRAFFIC_CLASS
;
4616 adapter
->dcb_cfg
.num_tcs
.pg_tcs
= MAX_TRAFFIC_CLASS
;
4617 adapter
->dcb_cfg
.num_tcs
.pfc_tcs
= MAX_TRAFFIC_CLASS
;
4621 /* Configure DCB traffic classes */
4622 for (j
= 0; j
< MAX_TRAFFIC_CLASS
; j
++) {
4623 tc
= &adapter
->dcb_cfg
.tc_config
[j
];
4624 tc
->path
[DCB_TX_CONFIG
].bwg_id
= 0;
4625 tc
->path
[DCB_TX_CONFIG
].bwg_percent
= 12 + (j
& 1);
4626 tc
->path
[DCB_RX_CONFIG
].bwg_id
= 0;
4627 tc
->path
[DCB_RX_CONFIG
].bwg_percent
= 12 + (j
& 1);
4628 tc
->dcb_pfc
= pfc_disabled
;
4631 /* Initialize default user to priority mapping, UPx->TC0 */
4632 tc
= &adapter
->dcb_cfg
.tc_config
[0];
4633 tc
->path
[DCB_TX_CONFIG
].up_to_tc_bitmap
= 0xFF;
4634 tc
->path
[DCB_RX_CONFIG
].up_to_tc_bitmap
= 0xFF;
4636 adapter
->dcb_cfg
.bw_percentage
[DCB_TX_CONFIG
][0] = 100;
4637 adapter
->dcb_cfg
.bw_percentage
[DCB_RX_CONFIG
][0] = 100;
4638 adapter
->dcb_cfg
.pfc_mode_enable
= false;
4639 adapter
->dcb_set_bitmap
= 0x00;
4640 adapter
->dcbx_cap
= DCB_CAP_DCBX_HOST
| DCB_CAP_DCBX_VER_CEE
;
4641 memcpy(&adapter
->temp_dcb_cfg
, &adapter
->dcb_cfg
,
4642 sizeof(adapter
->temp_dcb_cfg
));
4646 /* default flow control settings */
4647 hw
->fc
.requested_mode
= ixgbe_fc_full
;
4648 hw
->fc
.current_mode
= ixgbe_fc_full
; /* init for ethtool output */
4649 ixgbe_pbthresh_setup(adapter
);
4650 hw
->fc
.pause_time
= IXGBE_DEFAULT_FCPAUSE
;
4651 hw
->fc
.send_xon
= true;
4652 hw
->fc
.disable_fc_autoneg
=
4653 (ixgbe_device_supports_autoneg_fc(hw
) == 0) ? false : true;
4655 #ifdef CONFIG_PCI_IOV
4656 /* assign number of SR-IOV VFs */
4657 if (hw
->mac
.type
!= ixgbe_mac_82598EB
)
4658 adapter
->num_vfs
= (max_vfs
> 63) ? 0 : max_vfs
;
4661 /* enable itr by default in dynamic mode */
4662 adapter
->rx_itr_setting
= 1;
4663 adapter
->tx_itr_setting
= 1;
4665 /* set default ring sizes */
4666 adapter
->tx_ring_count
= IXGBE_DEFAULT_TXD
;
4667 adapter
->rx_ring_count
= IXGBE_DEFAULT_RXD
;
4669 /* set default work limits */
4670 adapter
->tx_work_limit
= IXGBE_DEFAULT_TX_WORK
;
4672 /* initialize eeprom parameters */
4673 if (ixgbe_init_eeprom_params_generic(hw
)) {
4674 e_dev_err("EEPROM initialization failed\n");
4678 set_bit(__IXGBE_DOWN
, &adapter
->state
);
4684 * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
4685 * @tx_ring: tx descriptor ring (for a specific queue) to setup
4687 * Return 0 on success, negative on failure
4689 int ixgbe_setup_tx_resources(struct ixgbe_ring
*tx_ring
)
4691 struct device
*dev
= tx_ring
->dev
;
4692 int orig_node
= dev_to_node(dev
);
4696 size
= sizeof(struct ixgbe_tx_buffer
) * tx_ring
->count
;
4698 if (tx_ring
->q_vector
)
4699 numa_node
= tx_ring
->q_vector
->numa_node
;
4701 tx_ring
->tx_buffer_info
= vzalloc_node(size
, numa_node
);
4702 if (!tx_ring
->tx_buffer_info
)
4703 tx_ring
->tx_buffer_info
= vzalloc(size
);
4704 if (!tx_ring
->tx_buffer_info
)
4707 /* round up to nearest 4K */
4708 tx_ring
->size
= tx_ring
->count
* sizeof(union ixgbe_adv_tx_desc
);
4709 tx_ring
->size
= ALIGN(tx_ring
->size
, 4096);
4711 set_dev_node(dev
, numa_node
);
4712 tx_ring
->desc
= dma_alloc_coherent(dev
,
4716 set_dev_node(dev
, orig_node
);
4718 tx_ring
->desc
= dma_alloc_coherent(dev
, tx_ring
->size
,
4719 &tx_ring
->dma
, GFP_KERNEL
);
4723 tx_ring
->next_to_use
= 0;
4724 tx_ring
->next_to_clean
= 0;
4728 vfree(tx_ring
->tx_buffer_info
);
4729 tx_ring
->tx_buffer_info
= NULL
;
4730 dev_err(dev
, "Unable to allocate memory for the Tx descriptor ring\n");
4735 * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
4736 * @adapter: board private structure
4738 * If this function returns with an error, then it's possible one or
4739 * more of the rings is populated (while the rest are not). It is the
4740 * callers duty to clean those orphaned rings.
4742 * Return 0 on success, negative on failure
4744 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter
*adapter
)
4748 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
4749 err
= ixgbe_setup_tx_resources(adapter
->tx_ring
[i
]);
4753 e_err(probe
, "Allocation for Tx Queue %u failed\n", i
);
4759 /* rewind the index freeing the rings as we go */
4761 ixgbe_free_tx_resources(adapter
->tx_ring
[i
]);
4766 * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
4767 * @rx_ring: rx descriptor ring (for a specific queue) to setup
4769 * Returns 0 on success, negative on failure
4771 int ixgbe_setup_rx_resources(struct ixgbe_ring
*rx_ring
)
4773 struct device
*dev
= rx_ring
->dev
;
4774 int orig_node
= dev_to_node(dev
);
4778 size
= sizeof(struct ixgbe_rx_buffer
) * rx_ring
->count
;
4780 if (rx_ring
->q_vector
)
4781 numa_node
= rx_ring
->q_vector
->numa_node
;
4783 rx_ring
->rx_buffer_info
= vzalloc_node(size
, numa_node
);
4784 if (!rx_ring
->rx_buffer_info
)
4785 rx_ring
->rx_buffer_info
= vzalloc(size
);
4786 if (!rx_ring
->rx_buffer_info
)
4789 /* Round up to nearest 4K */
4790 rx_ring
->size
= rx_ring
->count
* sizeof(union ixgbe_adv_rx_desc
);
4791 rx_ring
->size
= ALIGN(rx_ring
->size
, 4096);
4793 set_dev_node(dev
, numa_node
);
4794 rx_ring
->desc
= dma_alloc_coherent(dev
,
4798 set_dev_node(dev
, orig_node
);
4800 rx_ring
->desc
= dma_alloc_coherent(dev
, rx_ring
->size
,
4801 &rx_ring
->dma
, GFP_KERNEL
);
4805 rx_ring
->next_to_clean
= 0;
4806 rx_ring
->next_to_use
= 0;
4810 vfree(rx_ring
->rx_buffer_info
);
4811 rx_ring
->rx_buffer_info
= NULL
;
4812 dev_err(dev
, "Unable to allocate memory for the Rx descriptor ring\n");
4817 * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
4818 * @adapter: board private structure
4820 * If this function returns with an error, then it's possible one or
4821 * more of the rings is populated (while the rest are not). It is the
4822 * callers duty to clean those orphaned rings.
4824 * Return 0 on success, negative on failure
4826 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter
*adapter
)
4830 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
4831 err
= ixgbe_setup_rx_resources(adapter
->rx_ring
[i
]);
4835 e_err(probe
, "Allocation for Rx Queue %u failed\n", i
);
4840 err
= ixgbe_setup_fcoe_ddp_resources(adapter
);
4845 /* rewind the index freeing the rings as we go */
4847 ixgbe_free_rx_resources(adapter
->rx_ring
[i
]);
4852 * ixgbe_free_tx_resources - Free Tx Resources per Queue
4853 * @tx_ring: Tx descriptor ring for a specific queue
4855 * Free all transmit software resources
4857 void ixgbe_free_tx_resources(struct ixgbe_ring
*tx_ring
)
4859 ixgbe_clean_tx_ring(tx_ring
);
4861 vfree(tx_ring
->tx_buffer_info
);
4862 tx_ring
->tx_buffer_info
= NULL
;
4864 /* if not set, then don't free */
4868 dma_free_coherent(tx_ring
->dev
, tx_ring
->size
,
4869 tx_ring
->desc
, tx_ring
->dma
);
4871 tx_ring
->desc
= NULL
;
4875 * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
4876 * @adapter: board private structure
4878 * Free all transmit software resources
4880 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter
*adapter
)
4884 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
4885 if (adapter
->tx_ring
[i
]->desc
)
4886 ixgbe_free_tx_resources(adapter
->tx_ring
[i
]);
4890 * ixgbe_free_rx_resources - Free Rx Resources
4891 * @rx_ring: ring to clean the resources from
4893 * Free all receive software resources
4895 void ixgbe_free_rx_resources(struct ixgbe_ring
*rx_ring
)
4897 ixgbe_clean_rx_ring(rx_ring
);
4899 vfree(rx_ring
->rx_buffer_info
);
4900 rx_ring
->rx_buffer_info
= NULL
;
4902 /* if not set, then don't free */
4906 dma_free_coherent(rx_ring
->dev
, rx_ring
->size
,
4907 rx_ring
->desc
, rx_ring
->dma
);
4909 rx_ring
->desc
= NULL
;
4913 * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
4914 * @adapter: board private structure
4916 * Free all receive software resources
4918 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter
*adapter
)
4923 ixgbe_free_fcoe_ddp_resources(adapter
);
4926 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
4927 if (adapter
->rx_ring
[i
]->desc
)
4928 ixgbe_free_rx_resources(adapter
->rx_ring
[i
]);
4932 * ixgbe_change_mtu - Change the Maximum Transfer Unit
4933 * @netdev: network interface device structure
4934 * @new_mtu: new value for maximum frame size
4936 * Returns 0 on success, negative on failure
4938 static int ixgbe_change_mtu(struct net_device
*netdev
, int new_mtu
)
4940 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
4941 int max_frame
= new_mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
4943 /* MTU < 68 is an error and causes problems on some kernels */
4944 if ((new_mtu
< 68) || (max_frame
> IXGBE_MAX_JUMBO_FRAME_SIZE
))
4948 * For 82599EB we cannot allow legacy VFs to enable their receive
4949 * paths when MTU greater than 1500 is configured. So display a
4950 * warning that legacy VFs will be disabled.
4952 if ((adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) &&
4953 (adapter
->hw
.mac
.type
== ixgbe_mac_82599EB
) &&
4954 (max_frame
> (ETH_FRAME_LEN
+ ETH_FCS_LEN
)))
4955 e_warn(probe
, "Setting MTU > 1500 will disable legacy VFs\n");
4957 e_info(probe
, "changing MTU from %d to %d\n", netdev
->mtu
, new_mtu
);
4959 /* must set new MTU before calling down or up */
4960 netdev
->mtu
= new_mtu
;
4962 if (netif_running(netdev
))
4963 ixgbe_reinit_locked(adapter
);
4969 * ixgbe_open - Called when a network interface is made active
4970 * @netdev: network interface device structure
4972 * Returns 0 on success, negative value on failure
4974 * The open entry point is called when a network interface is made
4975 * active by the system (IFF_UP). At this point all resources needed
4976 * for transmit and receive operations are allocated, the interrupt
4977 * handler is registered with the OS, the watchdog timer is started,
4978 * and the stack is notified that the interface is ready.
4980 static int ixgbe_open(struct net_device
*netdev
)
4982 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
4985 /* disallow open during test */
4986 if (test_bit(__IXGBE_TESTING
, &adapter
->state
))
4989 netif_carrier_off(netdev
);
4991 /* allocate transmit descriptors */
4992 err
= ixgbe_setup_all_tx_resources(adapter
);
4996 /* allocate receive descriptors */
4997 err
= ixgbe_setup_all_rx_resources(adapter
);
5001 ixgbe_configure(adapter
);
5003 err
= ixgbe_request_irq(adapter
);
5007 /* Notify the stack of the actual queue counts. */
5008 err
= netif_set_real_num_tx_queues(netdev
,
5009 adapter
->num_rx_pools
> 1 ? 1 :
5010 adapter
->num_tx_queues
);
5012 goto err_set_queues
;
5015 err
= netif_set_real_num_rx_queues(netdev
,
5016 adapter
->num_rx_pools
> 1 ? 1 :
5017 adapter
->num_rx_queues
);
5019 goto err_set_queues
;
5021 ixgbe_ptp_init(adapter
);
5023 ixgbe_up_complete(adapter
);
5028 ixgbe_free_irq(adapter
);
5030 ixgbe_free_all_rx_resources(adapter
);
5032 ixgbe_free_all_tx_resources(adapter
);
5034 ixgbe_reset(adapter
);
5040 * ixgbe_close - Disables a network interface
5041 * @netdev: network interface device structure
5043 * Returns 0, this is not allowed to fail
5045 * The close entry point is called when an interface is de-activated
5046 * by the OS. The hardware is still under the drivers control, but
5047 * needs to be disabled. A global MAC reset is issued to stop the
5048 * hardware, and all transmit and receive resources are freed.
5050 static int ixgbe_close(struct net_device
*netdev
)
5052 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
5054 ixgbe_ptp_stop(adapter
);
5056 ixgbe_down(adapter
);
5057 ixgbe_free_irq(adapter
);
5059 ixgbe_fdir_filter_exit(adapter
);
5061 ixgbe_free_all_tx_resources(adapter
);
5062 ixgbe_free_all_rx_resources(adapter
);
5064 ixgbe_release_hw_control(adapter
);
5070 static int ixgbe_resume(struct pci_dev
*pdev
)
5072 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
5073 struct net_device
*netdev
= adapter
->netdev
;
5076 pci_set_power_state(pdev
, PCI_D0
);
5077 pci_restore_state(pdev
);
5079 * pci_restore_state clears dev->state_saved so call
5080 * pci_save_state to restore it.
5082 pci_save_state(pdev
);
5084 err
= pci_enable_device_mem(pdev
);
5086 e_dev_err("Cannot enable PCI device from suspend\n");
5089 pci_set_master(pdev
);
5091 pci_wake_from_d3(pdev
, false);
5093 ixgbe_reset(adapter
);
5095 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_WUS
, ~0);
5098 err
= ixgbe_init_interrupt_scheme(adapter
);
5099 if (!err
&& netif_running(netdev
))
5100 err
= ixgbe_open(netdev
);
5107 netif_device_attach(netdev
);
5111 #endif /* CONFIG_PM */
5113 static int __ixgbe_shutdown(struct pci_dev
*pdev
, bool *enable_wake
)
5115 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
5116 struct net_device
*netdev
= adapter
->netdev
;
5117 struct ixgbe_hw
*hw
= &adapter
->hw
;
5119 u32 wufc
= adapter
->wol
;
5124 netif_device_detach(netdev
);
5127 if (netif_running(netdev
)) {
5128 ixgbe_down(adapter
);
5129 ixgbe_free_irq(adapter
);
5130 ixgbe_free_all_tx_resources(adapter
);
5131 ixgbe_free_all_rx_resources(adapter
);
5135 ixgbe_clear_interrupt_scheme(adapter
);
5138 retval
= pci_save_state(pdev
);
5144 ixgbe_set_rx_mode(netdev
);
5146 /* enable the optics for 82599 SFP+ fiber as we can WoL */
5147 if (hw
->mac
.ops
.enable_tx_laser
)
5148 hw
->mac
.ops
.enable_tx_laser(hw
);
5150 /* turn on all-multi mode if wake on multicast is enabled */
5151 if (wufc
& IXGBE_WUFC_MC
) {
5152 fctrl
= IXGBE_READ_REG(hw
, IXGBE_FCTRL
);
5153 fctrl
|= IXGBE_FCTRL_MPE
;
5154 IXGBE_WRITE_REG(hw
, IXGBE_FCTRL
, fctrl
);
5157 ctrl
= IXGBE_READ_REG(hw
, IXGBE_CTRL
);
5158 ctrl
|= IXGBE_CTRL_GIO_DIS
;
5159 IXGBE_WRITE_REG(hw
, IXGBE_CTRL
, ctrl
);
5161 IXGBE_WRITE_REG(hw
, IXGBE_WUFC
, wufc
);
5163 IXGBE_WRITE_REG(hw
, IXGBE_WUC
, 0);
5164 IXGBE_WRITE_REG(hw
, IXGBE_WUFC
, 0);
5167 switch (hw
->mac
.type
) {
5168 case ixgbe_mac_82598EB
:
5169 pci_wake_from_d3(pdev
, false);
5171 case ixgbe_mac_82599EB
:
5172 case ixgbe_mac_X540
:
5173 pci_wake_from_d3(pdev
, !!wufc
);
5179 *enable_wake
= !!wufc
;
5181 ixgbe_release_hw_control(adapter
);
5183 pci_disable_device(pdev
);
5189 static int ixgbe_suspend(struct pci_dev
*pdev
, pm_message_t state
)
5194 retval
= __ixgbe_shutdown(pdev
, &wake
);
5199 pci_prepare_to_sleep(pdev
);
5201 pci_wake_from_d3(pdev
, false);
5202 pci_set_power_state(pdev
, PCI_D3hot
);
5207 #endif /* CONFIG_PM */
5209 static void ixgbe_shutdown(struct pci_dev
*pdev
)
5213 __ixgbe_shutdown(pdev
, &wake
);
5215 if (system_state
== SYSTEM_POWER_OFF
) {
5216 pci_wake_from_d3(pdev
, wake
);
5217 pci_set_power_state(pdev
, PCI_D3hot
);
5222 * ixgbe_update_stats - Update the board statistics counters.
5223 * @adapter: board private structure
5225 void ixgbe_update_stats(struct ixgbe_adapter
*adapter
)
5227 struct net_device
*netdev
= adapter
->netdev
;
5228 struct ixgbe_hw
*hw
= &adapter
->hw
;
5229 struct ixgbe_hw_stats
*hwstats
= &adapter
->stats
;
5231 u32 i
, missed_rx
= 0, mpc
, bprc
, lxon
, lxoff
, xon_off_tot
;
5232 u64 non_eop_descs
= 0, restart_queue
= 0, tx_busy
= 0;
5233 u64 alloc_rx_page_failed
= 0, alloc_rx_buff_failed
= 0;
5234 u64 bytes
= 0, packets
= 0, hw_csum_rx_error
= 0;
5236 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
5237 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
5240 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
) {
5243 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
5244 rsc_count
+= adapter
->rx_ring
[i
]->rx_stats
.rsc_count
;
5245 rsc_flush
+= adapter
->rx_ring
[i
]->rx_stats
.rsc_flush
;
5247 adapter
->rsc_total_count
= rsc_count
;
5248 adapter
->rsc_total_flush
= rsc_flush
;
5251 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
5252 struct ixgbe_ring
*rx_ring
= adapter
->rx_ring
[i
];
5253 non_eop_descs
+= rx_ring
->rx_stats
.non_eop_descs
;
5254 alloc_rx_page_failed
+= rx_ring
->rx_stats
.alloc_rx_page_failed
;
5255 alloc_rx_buff_failed
+= rx_ring
->rx_stats
.alloc_rx_buff_failed
;
5256 hw_csum_rx_error
+= rx_ring
->rx_stats
.csum_err
;
5257 bytes
+= rx_ring
->stats
.bytes
;
5258 packets
+= rx_ring
->stats
.packets
;
5260 adapter
->non_eop_descs
= non_eop_descs
;
5261 adapter
->alloc_rx_page_failed
= alloc_rx_page_failed
;
5262 adapter
->alloc_rx_buff_failed
= alloc_rx_buff_failed
;
5263 adapter
->hw_csum_rx_error
= hw_csum_rx_error
;
5264 netdev
->stats
.rx_bytes
= bytes
;
5265 netdev
->stats
.rx_packets
= packets
;
5269 /* gather some stats to the adapter struct that are per queue */
5270 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
5271 struct ixgbe_ring
*tx_ring
= adapter
->tx_ring
[i
];
5272 restart_queue
+= tx_ring
->tx_stats
.restart_queue
;
5273 tx_busy
+= tx_ring
->tx_stats
.tx_busy
;
5274 bytes
+= tx_ring
->stats
.bytes
;
5275 packets
+= tx_ring
->stats
.packets
;
5277 adapter
->restart_queue
= restart_queue
;
5278 adapter
->tx_busy
= tx_busy
;
5279 netdev
->stats
.tx_bytes
= bytes
;
5280 netdev
->stats
.tx_packets
= packets
;
5282 hwstats
->crcerrs
+= IXGBE_READ_REG(hw
, IXGBE_CRCERRS
);
5284 /* 8 register reads */
5285 for (i
= 0; i
< 8; i
++) {
5286 /* for packet buffers not used, the register should read 0 */
5287 mpc
= IXGBE_READ_REG(hw
, IXGBE_MPC(i
));
5289 hwstats
->mpc
[i
] += mpc
;
5290 total_mpc
+= hwstats
->mpc
[i
];
5291 hwstats
->pxontxc
[i
] += IXGBE_READ_REG(hw
, IXGBE_PXONTXC(i
));
5292 hwstats
->pxofftxc
[i
] += IXGBE_READ_REG(hw
, IXGBE_PXOFFTXC(i
));
5293 switch (hw
->mac
.type
) {
5294 case ixgbe_mac_82598EB
:
5295 hwstats
->rnbc
[i
] += IXGBE_READ_REG(hw
, IXGBE_RNBC(i
));
5296 hwstats
->qbtc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBTC(i
));
5297 hwstats
->qbrc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBRC(i
));
5298 hwstats
->pxonrxc
[i
] +=
5299 IXGBE_READ_REG(hw
, IXGBE_PXONRXC(i
));
5301 case ixgbe_mac_82599EB
:
5302 case ixgbe_mac_X540
:
5303 hwstats
->pxonrxc
[i
] +=
5304 IXGBE_READ_REG(hw
, IXGBE_PXONRXCNT(i
));
5311 /*16 register reads */
5312 for (i
= 0; i
< 16; i
++) {
5313 hwstats
->qptc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QPTC(i
));
5314 hwstats
->qprc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QPRC(i
));
5315 if ((hw
->mac
.type
== ixgbe_mac_82599EB
) ||
5316 (hw
->mac
.type
== ixgbe_mac_X540
)) {
5317 hwstats
->qbtc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBTC_L(i
));
5318 IXGBE_READ_REG(hw
, IXGBE_QBTC_H(i
)); /* to clear */
5319 hwstats
->qbrc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBRC_L(i
));
5320 IXGBE_READ_REG(hw
, IXGBE_QBRC_H(i
)); /* to clear */
5324 hwstats
->gprc
+= IXGBE_READ_REG(hw
, IXGBE_GPRC
);
5325 /* work around hardware counting issue */
5326 hwstats
->gprc
-= missed_rx
;
5328 ixgbe_update_xoff_received(adapter
);
5330 /* 82598 hardware only has a 32 bit counter in the high register */
5331 switch (hw
->mac
.type
) {
5332 case ixgbe_mac_82598EB
:
5333 hwstats
->lxonrxc
+= IXGBE_READ_REG(hw
, IXGBE_LXONRXC
);
5334 hwstats
->gorc
+= IXGBE_READ_REG(hw
, IXGBE_GORCH
);
5335 hwstats
->gotc
+= IXGBE_READ_REG(hw
, IXGBE_GOTCH
);
5336 hwstats
->tor
+= IXGBE_READ_REG(hw
, IXGBE_TORH
);
5338 case ixgbe_mac_X540
:
5339 /* OS2BMC stats are X540 only*/
5340 hwstats
->o2bgptc
+= IXGBE_READ_REG(hw
, IXGBE_O2BGPTC
);
5341 hwstats
->o2bspc
+= IXGBE_READ_REG(hw
, IXGBE_O2BSPC
);
5342 hwstats
->b2ospc
+= IXGBE_READ_REG(hw
, IXGBE_B2OSPC
);
5343 hwstats
->b2ogprc
+= IXGBE_READ_REG(hw
, IXGBE_B2OGPRC
);
5344 case ixgbe_mac_82599EB
:
5345 for (i
= 0; i
< 16; i
++)
5346 adapter
->hw_rx_no_dma_resources
+=
5347 IXGBE_READ_REG(hw
, IXGBE_QPRDC(i
));
5348 hwstats
->gorc
+= IXGBE_READ_REG(hw
, IXGBE_GORCL
);
5349 IXGBE_READ_REG(hw
, IXGBE_GORCH
); /* to clear */
5350 hwstats
->gotc
+= IXGBE_READ_REG(hw
, IXGBE_GOTCL
);
5351 IXGBE_READ_REG(hw
, IXGBE_GOTCH
); /* to clear */
5352 hwstats
->tor
+= IXGBE_READ_REG(hw
, IXGBE_TORL
);
5353 IXGBE_READ_REG(hw
, IXGBE_TORH
); /* to clear */
5354 hwstats
->lxonrxc
+= IXGBE_READ_REG(hw
, IXGBE_LXONRXCNT
);
5355 hwstats
->fdirmatch
+= IXGBE_READ_REG(hw
, IXGBE_FDIRMATCH
);
5356 hwstats
->fdirmiss
+= IXGBE_READ_REG(hw
, IXGBE_FDIRMISS
);
5358 hwstats
->fccrc
+= IXGBE_READ_REG(hw
, IXGBE_FCCRC
);
5359 hwstats
->fcoerpdc
+= IXGBE_READ_REG(hw
, IXGBE_FCOERPDC
);
5360 hwstats
->fcoeprc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEPRC
);
5361 hwstats
->fcoeptc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEPTC
);
5362 hwstats
->fcoedwrc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEDWRC
);
5363 hwstats
->fcoedwtc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEDWTC
);
5364 /* Add up per cpu counters for total ddp aloc fail */
5365 if (adapter
->fcoe
.ddp_pool
) {
5366 struct ixgbe_fcoe
*fcoe
= &adapter
->fcoe
;
5367 struct ixgbe_fcoe_ddp_pool
*ddp_pool
;
5369 u64 noddp
= 0, noddp_ext_buff
= 0;
5370 for_each_possible_cpu(cpu
) {
5371 ddp_pool
= per_cpu_ptr(fcoe
->ddp_pool
, cpu
);
5372 noddp
+= ddp_pool
->noddp
;
5373 noddp_ext_buff
+= ddp_pool
->noddp_ext_buff
;
5375 hwstats
->fcoe_noddp
= noddp
;
5376 hwstats
->fcoe_noddp_ext_buff
= noddp_ext_buff
;
5378 #endif /* IXGBE_FCOE */
5383 bprc
= IXGBE_READ_REG(hw
, IXGBE_BPRC
);
5384 hwstats
->bprc
+= bprc
;
5385 hwstats
->mprc
+= IXGBE_READ_REG(hw
, IXGBE_MPRC
);
5386 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
5387 hwstats
->mprc
-= bprc
;
5388 hwstats
->roc
+= IXGBE_READ_REG(hw
, IXGBE_ROC
);
5389 hwstats
->prc64
+= IXGBE_READ_REG(hw
, IXGBE_PRC64
);
5390 hwstats
->prc127
+= IXGBE_READ_REG(hw
, IXGBE_PRC127
);
5391 hwstats
->prc255
+= IXGBE_READ_REG(hw
, IXGBE_PRC255
);
5392 hwstats
->prc511
+= IXGBE_READ_REG(hw
, IXGBE_PRC511
);
5393 hwstats
->prc1023
+= IXGBE_READ_REG(hw
, IXGBE_PRC1023
);
5394 hwstats
->prc1522
+= IXGBE_READ_REG(hw
, IXGBE_PRC1522
);
5395 hwstats
->rlec
+= IXGBE_READ_REG(hw
, IXGBE_RLEC
);
5396 lxon
= IXGBE_READ_REG(hw
, IXGBE_LXONTXC
);
5397 hwstats
->lxontxc
+= lxon
;
5398 lxoff
= IXGBE_READ_REG(hw
, IXGBE_LXOFFTXC
);
5399 hwstats
->lxofftxc
+= lxoff
;
5400 hwstats
->gptc
+= IXGBE_READ_REG(hw
, IXGBE_GPTC
);
5401 hwstats
->mptc
+= IXGBE_READ_REG(hw
, IXGBE_MPTC
);
5403 * 82598 errata - tx of flow control packets is included in tx counters
5405 xon_off_tot
= lxon
+ lxoff
;
5406 hwstats
->gptc
-= xon_off_tot
;
5407 hwstats
->mptc
-= xon_off_tot
;
5408 hwstats
->gotc
-= (xon_off_tot
* (ETH_ZLEN
+ ETH_FCS_LEN
));
5409 hwstats
->ruc
+= IXGBE_READ_REG(hw
, IXGBE_RUC
);
5410 hwstats
->rfc
+= IXGBE_READ_REG(hw
, IXGBE_RFC
);
5411 hwstats
->rjc
+= IXGBE_READ_REG(hw
, IXGBE_RJC
);
5412 hwstats
->tpr
+= IXGBE_READ_REG(hw
, IXGBE_TPR
);
5413 hwstats
->ptc64
+= IXGBE_READ_REG(hw
, IXGBE_PTC64
);
5414 hwstats
->ptc64
-= xon_off_tot
;
5415 hwstats
->ptc127
+= IXGBE_READ_REG(hw
, IXGBE_PTC127
);
5416 hwstats
->ptc255
+= IXGBE_READ_REG(hw
, IXGBE_PTC255
);
5417 hwstats
->ptc511
+= IXGBE_READ_REG(hw
, IXGBE_PTC511
);
5418 hwstats
->ptc1023
+= IXGBE_READ_REG(hw
, IXGBE_PTC1023
);
5419 hwstats
->ptc1522
+= IXGBE_READ_REG(hw
, IXGBE_PTC1522
);
5420 hwstats
->bptc
+= IXGBE_READ_REG(hw
, IXGBE_BPTC
);
5422 /* Fill out the OS statistics structure */
5423 netdev
->stats
.multicast
= hwstats
->mprc
;
5426 netdev
->stats
.rx_errors
= hwstats
->crcerrs
+ hwstats
->rlec
;
5427 netdev
->stats
.rx_dropped
= 0;
5428 netdev
->stats
.rx_length_errors
= hwstats
->rlec
;
5429 netdev
->stats
.rx_crc_errors
= hwstats
->crcerrs
;
5430 netdev
->stats
.rx_missed_errors
= total_mpc
;
5434 * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
5435 * @adapter: pointer to the device adapter structure
5437 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter
*adapter
)
5439 struct ixgbe_hw
*hw
= &adapter
->hw
;
5442 if (!(adapter
->flags2
& IXGBE_FLAG2_FDIR_REQUIRES_REINIT
))
5445 adapter
->flags2
&= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT
;
5447 /* if interface is down do nothing */
5448 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
5451 /* do nothing if we are not using signature filters */
5452 if (!(adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
))
5455 adapter
->fdir_overflow
++;
5457 if (ixgbe_reinit_fdir_tables_82599(hw
) == 0) {
5458 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
5459 set_bit(__IXGBE_TX_FDIR_INIT_DONE
,
5460 &(adapter
->tx_ring
[i
]->state
));
5461 /* re-enable flow director interrupts */
5462 IXGBE_WRITE_REG(hw
, IXGBE_EIMS
, IXGBE_EIMS_FLOW_DIR
);
5464 e_err(probe
, "failed to finish FDIR re-initialization, "
5465 "ignored adding FDIR ATR filters\n");
5470 * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
5471 * @adapter: pointer to the device adapter structure
5473 * This function serves two purposes. First it strobes the interrupt lines
5474 * in order to make certain interrupts are occurring. Secondly it sets the
5475 * bits needed to check for TX hangs. As a result we should immediately
5476 * determine if a hang has occurred.
5478 static void ixgbe_check_hang_subtask(struct ixgbe_adapter
*adapter
)
5480 struct ixgbe_hw
*hw
= &adapter
->hw
;
5484 /* If we're down or resetting, just bail */
5485 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
5486 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
5489 /* Force detection of hung controller */
5490 if (netif_carrier_ok(adapter
->netdev
)) {
5491 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
5492 set_check_for_tx_hang(adapter
->tx_ring
[i
]);
5495 if (!(adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)) {
5497 * for legacy and MSI interrupts don't set any bits
5498 * that are enabled for EIAM, because this operation
5499 * would set *both* EIMS and EICS for any bit in EIAM
5501 IXGBE_WRITE_REG(hw
, IXGBE_EICS
,
5502 (IXGBE_EICS_TCP_TIMER
| IXGBE_EICS_OTHER
));
5504 /* get one bit for every active tx/rx interrupt vector */
5505 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
5506 struct ixgbe_q_vector
*qv
= adapter
->q_vector
[i
];
5507 if (qv
->rx
.ring
|| qv
->tx
.ring
)
5508 eics
|= ((u64
)1 << i
);
5512 /* Cause software interrupt to ensure rings are cleaned */
5513 ixgbe_irq_rearm_queues(adapter
, eics
);
5518 * ixgbe_watchdog_update_link - update the link status
5519 * @adapter: pointer to the device adapter structure
5520 * @link_speed: pointer to a u32 to store the link_speed
5522 static void ixgbe_watchdog_update_link(struct ixgbe_adapter
*adapter
)
5524 struct ixgbe_hw
*hw
= &adapter
->hw
;
5525 u32 link_speed
= adapter
->link_speed
;
5526 bool link_up
= adapter
->link_up
;
5527 bool pfc_en
= adapter
->dcb_cfg
.pfc_mode_enable
;
5529 if (!(adapter
->flags
& IXGBE_FLAG_NEED_LINK_UPDATE
))
5532 if (hw
->mac
.ops
.check_link
) {
5533 hw
->mac
.ops
.check_link(hw
, &link_speed
, &link_up
, false);
5535 /* always assume link is up, if no check link function */
5536 link_speed
= IXGBE_LINK_SPEED_10GB_FULL
;
5540 if (adapter
->ixgbe_ieee_pfc
)
5541 pfc_en
|= !!(adapter
->ixgbe_ieee_pfc
->pfc_en
);
5543 if (link_up
&& !((adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
) && pfc_en
)) {
5544 hw
->mac
.ops
.fc_enable(hw
);
5545 ixgbe_set_rx_drop_en(adapter
);
5549 time_after(jiffies
, (adapter
->link_check_timeout
+
5550 IXGBE_TRY_LINK_TIMEOUT
))) {
5551 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_UPDATE
;
5552 IXGBE_WRITE_REG(hw
, IXGBE_EIMS
, IXGBE_EIMC_LSC
);
5553 IXGBE_WRITE_FLUSH(hw
);
5556 adapter
->link_up
= link_up
;
5557 adapter
->link_speed
= link_speed
;
5560 static void ixgbe_update_default_up(struct ixgbe_adapter
*adapter
)
5562 #ifdef CONFIG_IXGBE_DCB
5563 struct net_device
*netdev
= adapter
->netdev
;
5564 struct dcb_app app
= {
5565 .selector
= IEEE_8021QAZ_APP_SEL_ETHERTYPE
,
5570 if (adapter
->dcbx_cap
& DCB_CAP_DCBX_VER_IEEE
)
5571 up
= dcb_ieee_getapp_mask(netdev
, &app
);
5573 adapter
->default_up
= (up
> 1) ? (ffs(up
) - 1) : 0;
5578 * ixgbe_watchdog_link_is_up - update netif_carrier status and
5579 * print link up message
5580 * @adapter: pointer to the device adapter structure
5582 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter
*adapter
)
5584 struct net_device
*netdev
= adapter
->netdev
;
5585 struct ixgbe_hw
*hw
= &adapter
->hw
;
5586 u32 link_speed
= adapter
->link_speed
;
5587 bool flow_rx
, flow_tx
;
5589 /* only continue if link was previously down */
5590 if (netif_carrier_ok(netdev
))
5593 adapter
->flags2
&= ~IXGBE_FLAG2_SEARCH_FOR_SFP
;
5595 switch (hw
->mac
.type
) {
5596 case ixgbe_mac_82598EB
: {
5597 u32 frctl
= IXGBE_READ_REG(hw
, IXGBE_FCTRL
);
5598 u32 rmcs
= IXGBE_READ_REG(hw
, IXGBE_RMCS
);
5599 flow_rx
= !!(frctl
& IXGBE_FCTRL_RFCE
);
5600 flow_tx
= !!(rmcs
& IXGBE_RMCS_TFCE_802_3X
);
5603 case ixgbe_mac_X540
:
5604 case ixgbe_mac_82599EB
: {
5605 u32 mflcn
= IXGBE_READ_REG(hw
, IXGBE_MFLCN
);
5606 u32 fccfg
= IXGBE_READ_REG(hw
, IXGBE_FCCFG
);
5607 flow_rx
= !!(mflcn
& IXGBE_MFLCN_RFCE
);
5608 flow_tx
= !!(fccfg
& IXGBE_FCCFG_TFCE_802_3X
);
5617 adapter
->last_rx_ptp_check
= jiffies
;
5619 if (adapter
->flags2
& IXGBE_FLAG2_PTP_ENABLED
)
5620 ixgbe_ptp_start_cyclecounter(adapter
);
5622 e_info(drv
, "NIC Link is Up %s, Flow Control: %s\n",
5623 (link_speed
== IXGBE_LINK_SPEED_10GB_FULL
?
5625 (link_speed
== IXGBE_LINK_SPEED_1GB_FULL
?
5627 (link_speed
== IXGBE_LINK_SPEED_100_FULL
?
5630 ((flow_rx
&& flow_tx
) ? "RX/TX" :
5632 (flow_tx
? "TX" : "None"))));
5634 netif_carrier_on(netdev
);
5635 ixgbe_check_vf_rate_limit(adapter
);
5637 /* update the default user priority for VFs */
5638 ixgbe_update_default_up(adapter
);
5640 /* ping all the active vfs to let them know link has changed */
5641 ixgbe_ping_all_vfs(adapter
);
5645 * ixgbe_watchdog_link_is_down - update netif_carrier status and
5646 * print link down message
5647 * @adapter: pointer to the adapter structure
5649 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter
*adapter
)
5651 struct net_device
*netdev
= adapter
->netdev
;
5652 struct ixgbe_hw
*hw
= &adapter
->hw
;
5654 adapter
->link_up
= false;
5655 adapter
->link_speed
= 0;
5657 /* only continue if link was up previously */
5658 if (!netif_carrier_ok(netdev
))
5661 /* poll for SFP+ cable when link is down */
5662 if (ixgbe_is_sfp(hw
) && hw
->mac
.type
== ixgbe_mac_82598EB
)
5663 adapter
->flags2
|= IXGBE_FLAG2_SEARCH_FOR_SFP
;
5665 if (adapter
->flags2
& IXGBE_FLAG2_PTP_ENABLED
)
5666 ixgbe_ptp_start_cyclecounter(adapter
);
5668 e_info(drv
, "NIC Link is Down\n");
5669 netif_carrier_off(netdev
);
5671 /* ping all the active vfs to let them know link has changed */
5672 ixgbe_ping_all_vfs(adapter
);
5676 * ixgbe_watchdog_flush_tx - flush queues on link down
5677 * @adapter: pointer to the device adapter structure
5679 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter
*adapter
)
5682 int some_tx_pending
= 0;
5684 if (!netif_carrier_ok(adapter
->netdev
)) {
5685 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
5686 struct ixgbe_ring
*tx_ring
= adapter
->tx_ring
[i
];
5687 if (tx_ring
->next_to_use
!= tx_ring
->next_to_clean
) {
5688 some_tx_pending
= 1;
5693 if (some_tx_pending
) {
5694 /* We've lost link, so the controller stops DMA,
5695 * but we've got queued Tx work that's never going
5696 * to get done, so reset controller to flush Tx.
5697 * (Do the reset outside of interrupt context).
5699 e_warn(drv
, "initiating reset to clear Tx work after link loss\n");
5700 adapter
->flags2
|= IXGBE_FLAG2_RESET_REQUESTED
;
5705 static void ixgbe_spoof_check(struct ixgbe_adapter
*adapter
)
5709 /* Do not perform spoof check for 82598 or if not in IOV mode */
5710 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
||
5711 adapter
->num_vfs
== 0)
5714 ssvpc
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_SSVPC
);
5717 * ssvpc register is cleared on read, if zero then no
5718 * spoofed packets in the last interval.
5723 e_warn(drv
, "%u Spoofed packets detected\n", ssvpc
);
5727 * ixgbe_watchdog_subtask - check and bring link up
5728 * @adapter: pointer to the device adapter structure
5730 static void ixgbe_watchdog_subtask(struct ixgbe_adapter
*adapter
)
5732 /* if interface is down do nothing */
5733 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
5734 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
5737 ixgbe_watchdog_update_link(adapter
);
5739 if (adapter
->link_up
)
5740 ixgbe_watchdog_link_is_up(adapter
);
5742 ixgbe_watchdog_link_is_down(adapter
);
5744 ixgbe_spoof_check(adapter
);
5745 ixgbe_update_stats(adapter
);
5747 ixgbe_watchdog_flush_tx(adapter
);
5751 * ixgbe_sfp_detection_subtask - poll for SFP+ cable
5752 * @adapter: the ixgbe adapter structure
5754 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter
*adapter
)
5756 struct ixgbe_hw
*hw
= &adapter
->hw
;
5759 /* not searching for SFP so there is nothing to do here */
5760 if (!(adapter
->flags2
& IXGBE_FLAG2_SEARCH_FOR_SFP
) &&
5761 !(adapter
->flags2
& IXGBE_FLAG2_SFP_NEEDS_RESET
))
5764 /* concurent i2c reads are not supported */
5765 if (test_bit(__IXGBE_READ_I2C
, &adapter
->state
))
5768 /* someone else is in init, wait until next service event */
5769 if (test_and_set_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
))
5772 err
= hw
->phy
.ops
.identify_sfp(hw
);
5773 if (err
== IXGBE_ERR_SFP_NOT_SUPPORTED
)
5776 if (err
== IXGBE_ERR_SFP_NOT_PRESENT
) {
5777 /* If no cable is present, then we need to reset
5778 * the next time we find a good cable. */
5779 adapter
->flags2
|= IXGBE_FLAG2_SFP_NEEDS_RESET
;
5786 /* exit if reset not needed */
5787 if (!(adapter
->flags2
& IXGBE_FLAG2_SFP_NEEDS_RESET
))
5790 adapter
->flags2
&= ~IXGBE_FLAG2_SFP_NEEDS_RESET
;
5793 * A module may be identified correctly, but the EEPROM may not have
5794 * support for that module. setup_sfp() will fail in that case, so
5795 * we should not allow that module to load.
5797 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
5798 err
= hw
->phy
.ops
.reset(hw
);
5800 err
= hw
->mac
.ops
.setup_sfp(hw
);
5802 if (err
== IXGBE_ERR_SFP_NOT_SUPPORTED
)
5805 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_CONFIG
;
5806 e_info(probe
, "detected SFP+: %d\n", hw
->phy
.sfp_type
);
5809 clear_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
);
5811 if ((err
== IXGBE_ERR_SFP_NOT_SUPPORTED
) &&
5812 (adapter
->netdev
->reg_state
== NETREG_REGISTERED
)) {
5813 e_dev_err("failed to initialize because an unsupported "
5814 "SFP+ module type was detected.\n");
5815 e_dev_err("Reload the driver after installing a "
5816 "supported module.\n");
5817 unregister_netdev(adapter
->netdev
);
5822 * ixgbe_sfp_link_config_subtask - set up link SFP after module install
5823 * @adapter: the ixgbe adapter structure
5825 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter
*adapter
)
5827 struct ixgbe_hw
*hw
= &adapter
->hw
;
5829 bool autoneg
= false;
5831 if (!(adapter
->flags
& IXGBE_FLAG_NEED_LINK_CONFIG
))
5834 /* someone else is in init, wait until next service event */
5835 if (test_and_set_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
))
5838 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_CONFIG
;
5840 speed
= hw
->phy
.autoneg_advertised
;
5841 if ((!speed
) && (hw
->mac
.ops
.get_link_capabilities
))
5842 hw
->mac
.ops
.get_link_capabilities(hw
, &speed
, &autoneg
);
5843 if (hw
->mac
.ops
.setup_link
)
5844 hw
->mac
.ops
.setup_link(hw
, speed
, true);
5846 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_UPDATE
;
5847 adapter
->link_check_timeout
= jiffies
;
5848 clear_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
);
5851 #ifdef CONFIG_PCI_IOV
5852 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter
*adapter
)
5855 struct ixgbe_hw
*hw
= &adapter
->hw
;
5856 struct net_device
*netdev
= adapter
->netdev
;
5860 gpc
= IXGBE_READ_REG(hw
, IXGBE_TXDGPC
);
5861 if (gpc
) /* If incrementing then no need for the check below */
5864 * Check to see if a bad DMA write target from an errant or
5865 * malicious VF has caused a PCIe error. If so then we can
5866 * issue a VFLR to the offending VF(s) and then resume without
5867 * requesting a full slot reset.
5870 for (vf
= 0; vf
< adapter
->num_vfs
; vf
++) {
5871 ciaa
= (vf
<< 16) | 0x80000000;
5872 /* 32 bit read so align, we really want status at offset 6 */
5873 ciaa
|= PCI_COMMAND
;
5874 IXGBE_WRITE_REG(hw
, IXGBE_CIAA_82599
, ciaa
);
5875 ciad
= IXGBE_READ_REG(hw
, IXGBE_CIAD_82599
);
5877 /* disable debug mode asap after reading data */
5878 IXGBE_WRITE_REG(hw
, IXGBE_CIAA_82599
, ciaa
);
5879 /* Get the upper 16 bits which will be the PCI status reg */
5881 if (ciad
& PCI_STATUS_REC_MASTER_ABORT
) {
5882 netdev_err(netdev
, "VF %d Hung DMA\n", vf
);
5884 ciaa
= (vf
<< 16) | 0x80000000;
5886 IXGBE_WRITE_REG(hw
, IXGBE_CIAA_82599
, ciaa
);
5887 ciad
= 0x00008000; /* VFLR */
5888 IXGBE_WRITE_REG(hw
, IXGBE_CIAD_82599
, ciad
);
5890 IXGBE_WRITE_REG(hw
, IXGBE_CIAA_82599
, ciaa
);
5897 * ixgbe_service_timer - Timer Call-back
5898 * @data: pointer to adapter cast into an unsigned long
5900 static void ixgbe_service_timer(unsigned long data
)
5902 struct ixgbe_adapter
*adapter
= (struct ixgbe_adapter
*)data
;
5903 unsigned long next_event_offset
;
5906 /* poll faster when waiting for link */
5907 if (adapter
->flags
& IXGBE_FLAG_NEED_LINK_UPDATE
)
5908 next_event_offset
= HZ
/ 10;
5910 next_event_offset
= HZ
* 2;
5912 #ifdef CONFIG_PCI_IOV
5914 * don't bother with SR-IOV VF DMA hang check if there are
5915 * no VFs or the link is down
5917 if (!adapter
->num_vfs
||
5918 (adapter
->flags
& IXGBE_FLAG_NEED_LINK_UPDATE
))
5919 goto normal_timer_service
;
5921 /* If we have VFs allocated then we must check for DMA hangs */
5922 ixgbe_check_for_bad_vf(adapter
);
5923 next_event_offset
= HZ
/ 50;
5924 adapter
->timer_event_accumulator
++;
5926 if (adapter
->timer_event_accumulator
>= 100)
5927 adapter
->timer_event_accumulator
= 0;
5931 normal_timer_service
:
5933 /* Reset the timer */
5934 mod_timer(&adapter
->service_timer
, next_event_offset
+ jiffies
);
5937 ixgbe_service_event_schedule(adapter
);
5940 static void ixgbe_reset_subtask(struct ixgbe_adapter
*adapter
)
5942 if (!(adapter
->flags2
& IXGBE_FLAG2_RESET_REQUESTED
))
5945 adapter
->flags2
&= ~IXGBE_FLAG2_RESET_REQUESTED
;
5947 /* If we're already down or resetting, just bail */
5948 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
5949 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
5952 ixgbe_dump(adapter
);
5953 netdev_err(adapter
->netdev
, "Reset adapter\n");
5954 adapter
->tx_timeout_count
++;
5956 ixgbe_reinit_locked(adapter
);
5960 * ixgbe_service_task - manages and runs subtasks
5961 * @work: pointer to work_struct containing our data
5963 static void ixgbe_service_task(struct work_struct
*work
)
5965 struct ixgbe_adapter
*adapter
= container_of(work
,
5966 struct ixgbe_adapter
,
5968 ixgbe_reset_subtask(adapter
);
5969 ixgbe_sfp_detection_subtask(adapter
);
5970 ixgbe_sfp_link_config_subtask(adapter
);
5971 ixgbe_check_overtemp_subtask(adapter
);
5972 ixgbe_watchdog_subtask(adapter
);
5973 ixgbe_fdir_reinit_subtask(adapter
);
5974 ixgbe_check_hang_subtask(adapter
);
5976 if (adapter
->flags2
& IXGBE_FLAG2_PTP_ENABLED
) {
5977 ixgbe_ptp_overflow_check(adapter
);
5978 ixgbe_ptp_rx_hang(adapter
);
5981 ixgbe_service_event_complete(adapter
);
5984 static int ixgbe_tso(struct ixgbe_ring
*tx_ring
,
5985 struct ixgbe_tx_buffer
*first
,
5988 struct sk_buff
*skb
= first
->skb
;
5989 u32 vlan_macip_lens
, type_tucmd
;
5990 u32 mss_l4len_idx
, l4len
;
5992 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
5995 if (!skb_is_gso(skb
))
5998 if (skb_header_cloned(skb
)) {
5999 int err
= pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
);
6004 /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
6005 type_tucmd
= IXGBE_ADVTXD_TUCMD_L4T_TCP
;
6007 if (first
->protocol
== __constant_htons(ETH_P_IP
)) {
6008 struct iphdr
*iph
= ip_hdr(skb
);
6011 tcp_hdr(skb
)->check
= ~csum_tcpudp_magic(iph
->saddr
,
6015 type_tucmd
|= IXGBE_ADVTXD_TUCMD_IPV4
;
6016 first
->tx_flags
|= IXGBE_TX_FLAGS_TSO
|
6017 IXGBE_TX_FLAGS_CSUM
|
6018 IXGBE_TX_FLAGS_IPV4
;
6019 } else if (skb_is_gso_v6(skb
)) {
6020 ipv6_hdr(skb
)->payload_len
= 0;
6021 tcp_hdr(skb
)->check
=
6022 ~csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
6023 &ipv6_hdr(skb
)->daddr
,
6025 first
->tx_flags
|= IXGBE_TX_FLAGS_TSO
|
6026 IXGBE_TX_FLAGS_CSUM
;
6029 /* compute header lengths */
6030 l4len
= tcp_hdrlen(skb
);
6031 *hdr_len
= skb_transport_offset(skb
) + l4len
;
6033 /* update gso size and bytecount with header size */
6034 first
->gso_segs
= skb_shinfo(skb
)->gso_segs
;
6035 first
->bytecount
+= (first
->gso_segs
- 1) * *hdr_len
;
6037 /* mss_l4len_id: use 0 as index for TSO */
6038 mss_l4len_idx
= l4len
<< IXGBE_ADVTXD_L4LEN_SHIFT
;
6039 mss_l4len_idx
|= skb_shinfo(skb
)->gso_size
<< IXGBE_ADVTXD_MSS_SHIFT
;
6041 /* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
6042 vlan_macip_lens
= skb_network_header_len(skb
);
6043 vlan_macip_lens
|= skb_network_offset(skb
) << IXGBE_ADVTXD_MACLEN_SHIFT
;
6044 vlan_macip_lens
|= first
->tx_flags
& IXGBE_TX_FLAGS_VLAN_MASK
;
6046 ixgbe_tx_ctxtdesc(tx_ring
, vlan_macip_lens
, 0, type_tucmd
,
6052 static void ixgbe_tx_csum(struct ixgbe_ring
*tx_ring
,
6053 struct ixgbe_tx_buffer
*first
)
6055 struct sk_buff
*skb
= first
->skb
;
6056 u32 vlan_macip_lens
= 0;
6057 u32 mss_l4len_idx
= 0;
6060 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
6061 if (!(first
->tx_flags
& IXGBE_TX_FLAGS_HW_VLAN
) &&
6062 !(first
->tx_flags
& IXGBE_TX_FLAGS_CC
))
6066 switch (first
->protocol
) {
6067 case __constant_htons(ETH_P_IP
):
6068 vlan_macip_lens
|= skb_network_header_len(skb
);
6069 type_tucmd
|= IXGBE_ADVTXD_TUCMD_IPV4
;
6070 l4_hdr
= ip_hdr(skb
)->protocol
;
6072 case __constant_htons(ETH_P_IPV6
):
6073 vlan_macip_lens
|= skb_network_header_len(skb
);
6074 l4_hdr
= ipv6_hdr(skb
)->nexthdr
;
6077 if (unlikely(net_ratelimit())) {
6078 dev_warn(tx_ring
->dev
,
6079 "partial checksum but proto=%x!\n",
6087 type_tucmd
|= IXGBE_ADVTXD_TUCMD_L4T_TCP
;
6088 mss_l4len_idx
= tcp_hdrlen(skb
) <<
6089 IXGBE_ADVTXD_L4LEN_SHIFT
;
6092 type_tucmd
|= IXGBE_ADVTXD_TUCMD_L4T_SCTP
;
6093 mss_l4len_idx
= sizeof(struct sctphdr
) <<
6094 IXGBE_ADVTXD_L4LEN_SHIFT
;
6097 mss_l4len_idx
= sizeof(struct udphdr
) <<
6098 IXGBE_ADVTXD_L4LEN_SHIFT
;
6101 if (unlikely(net_ratelimit())) {
6102 dev_warn(tx_ring
->dev
,
6103 "partial checksum but l4 proto=%x!\n",
6109 /* update TX checksum flag */
6110 first
->tx_flags
|= IXGBE_TX_FLAGS_CSUM
;
6113 /* vlan_macip_lens: MACLEN, VLAN tag */
6114 vlan_macip_lens
|= skb_network_offset(skb
) << IXGBE_ADVTXD_MACLEN_SHIFT
;
6115 vlan_macip_lens
|= first
->tx_flags
& IXGBE_TX_FLAGS_VLAN_MASK
;
6117 ixgbe_tx_ctxtdesc(tx_ring
, vlan_macip_lens
, 0,
6118 type_tucmd
, mss_l4len_idx
);
6121 #define IXGBE_SET_FLAG(_input, _flag, _result) \
6122 ((_flag <= _result) ? \
6123 ((u32)(_input & _flag) * (_result / _flag)) : \
6124 ((u32)(_input & _flag) / (_flag / _result)))
6126 static u32
ixgbe_tx_cmd_type(struct sk_buff
*skb
, u32 tx_flags
)
6128 /* set type for advanced descriptor with frame checksum insertion */
6129 u32 cmd_type
= IXGBE_ADVTXD_DTYP_DATA
|
6130 IXGBE_ADVTXD_DCMD_DEXT
|
6131 IXGBE_ADVTXD_DCMD_IFCS
;
6133 /* set HW vlan bit if vlan is present */
6134 cmd_type
|= IXGBE_SET_FLAG(tx_flags
, IXGBE_TX_FLAGS_HW_VLAN
,
6135 IXGBE_ADVTXD_DCMD_VLE
);
6137 /* set segmentation enable bits for TSO/FSO */
6138 cmd_type
|= IXGBE_SET_FLAG(tx_flags
, IXGBE_TX_FLAGS_TSO
,
6139 IXGBE_ADVTXD_DCMD_TSE
);
6141 /* set timestamp bit if present */
6142 cmd_type
|= IXGBE_SET_FLAG(tx_flags
, IXGBE_TX_FLAGS_TSTAMP
,
6143 IXGBE_ADVTXD_MAC_TSTAMP
);
6145 /* insert frame checksum */
6146 cmd_type
^= IXGBE_SET_FLAG(skb
->no_fcs
, 1, IXGBE_ADVTXD_DCMD_IFCS
);
6151 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc
*tx_desc
,
6152 u32 tx_flags
, unsigned int paylen
)
6154 u32 olinfo_status
= paylen
<< IXGBE_ADVTXD_PAYLEN_SHIFT
;
6156 /* enable L4 checksum for TSO and TX checksum offload */
6157 olinfo_status
|= IXGBE_SET_FLAG(tx_flags
,
6158 IXGBE_TX_FLAGS_CSUM
,
6159 IXGBE_ADVTXD_POPTS_TXSM
);
6161 /* enble IPv4 checksum for TSO */
6162 olinfo_status
|= IXGBE_SET_FLAG(tx_flags
,
6163 IXGBE_TX_FLAGS_IPV4
,
6164 IXGBE_ADVTXD_POPTS_IXSM
);
6167 * Check Context must be set if Tx switch is enabled, which it
6168 * always is for case where virtual functions are running
6170 olinfo_status
|= IXGBE_SET_FLAG(tx_flags
,
6174 tx_desc
->read
.olinfo_status
= cpu_to_le32(olinfo_status
);
6177 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
6180 static void ixgbe_tx_map(struct ixgbe_ring
*tx_ring
,
6181 struct ixgbe_tx_buffer
*first
,
6184 struct sk_buff
*skb
= first
->skb
;
6185 struct ixgbe_tx_buffer
*tx_buffer
;
6186 union ixgbe_adv_tx_desc
*tx_desc
;
6187 struct skb_frag_struct
*frag
;
6189 unsigned int data_len
, size
;
6190 u32 tx_flags
= first
->tx_flags
;
6191 u32 cmd_type
= ixgbe_tx_cmd_type(skb
, tx_flags
);
6192 u16 i
= tx_ring
->next_to_use
;
6194 tx_desc
= IXGBE_TX_DESC(tx_ring
, i
);
6196 ixgbe_tx_olinfo_status(tx_desc
, tx_flags
, skb
->len
- hdr_len
);
6198 size
= skb_headlen(skb
);
6199 data_len
= skb
->data_len
;
6202 if (tx_flags
& IXGBE_TX_FLAGS_FCOE
) {
6203 if (data_len
< sizeof(struct fcoe_crc_eof
)) {
6204 size
-= sizeof(struct fcoe_crc_eof
) - data_len
;
6207 data_len
-= sizeof(struct fcoe_crc_eof
);
6212 dma
= dma_map_single(tx_ring
->dev
, skb
->data
, size
, DMA_TO_DEVICE
);
6216 for (frag
= &skb_shinfo(skb
)->frags
[0];; frag
++) {
6217 if (dma_mapping_error(tx_ring
->dev
, dma
))
6220 /* record length, and DMA address */
6221 dma_unmap_len_set(tx_buffer
, len
, size
);
6222 dma_unmap_addr_set(tx_buffer
, dma
, dma
);
6224 tx_desc
->read
.buffer_addr
= cpu_to_le64(dma
);
6226 while (unlikely(size
> IXGBE_MAX_DATA_PER_TXD
)) {
6227 tx_desc
->read
.cmd_type_len
=
6228 cpu_to_le32(cmd_type
^ IXGBE_MAX_DATA_PER_TXD
);
6232 if (i
== tx_ring
->count
) {
6233 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
6236 tx_desc
->read
.olinfo_status
= 0;
6238 dma
+= IXGBE_MAX_DATA_PER_TXD
;
6239 size
-= IXGBE_MAX_DATA_PER_TXD
;
6241 tx_desc
->read
.buffer_addr
= cpu_to_le64(dma
);
6244 if (likely(!data_len
))
6247 tx_desc
->read
.cmd_type_len
= cpu_to_le32(cmd_type
^ size
);
6251 if (i
== tx_ring
->count
) {
6252 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
6255 tx_desc
->read
.olinfo_status
= 0;
6258 size
= min_t(unsigned int, data_len
, skb_frag_size(frag
));
6260 size
= skb_frag_size(frag
);
6264 dma
= skb_frag_dma_map(tx_ring
->dev
, frag
, 0, size
,
6267 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
6270 /* write last descriptor with RS and EOP bits */
6271 cmd_type
|= size
| IXGBE_TXD_CMD
;
6272 tx_desc
->read
.cmd_type_len
= cpu_to_le32(cmd_type
);
6274 netdev_tx_sent_queue(txring_txq(tx_ring
), first
->bytecount
);
6276 /* set the timestamp */
6277 first
->time_stamp
= jiffies
;
6280 * Force memory writes to complete before letting h/w know there
6281 * are new descriptors to fetch. (Only applicable for weak-ordered
6282 * memory model archs, such as IA-64).
6284 * We also need this memory barrier to make certain all of the
6285 * status bits have been updated before next_to_watch is written.
6289 /* set next_to_watch value indicating a packet is present */
6290 first
->next_to_watch
= tx_desc
;
6293 if (i
== tx_ring
->count
)
6296 tx_ring
->next_to_use
= i
;
6298 /* notify HW of packet */
6299 writel(i
, tx_ring
->tail
);
6303 dev_err(tx_ring
->dev
, "TX DMA map failed\n");
6305 /* clear dma mappings for failed tx_buffer_info map */
6307 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
6308 ixgbe_unmap_and_free_tx_resource(tx_ring
, tx_buffer
);
6309 if (tx_buffer
== first
)
6316 tx_ring
->next_to_use
= i
;
6319 static void ixgbe_atr(struct ixgbe_ring
*ring
,
6320 struct ixgbe_tx_buffer
*first
)
6322 struct ixgbe_q_vector
*q_vector
= ring
->q_vector
;
6323 union ixgbe_atr_hash_dword input
= { .dword
= 0 };
6324 union ixgbe_atr_hash_dword common
= { .dword
= 0 };
6326 unsigned char *network
;
6328 struct ipv6hdr
*ipv6
;
6333 /* if ring doesn't have a interrupt vector, cannot perform ATR */
6337 /* do nothing if sampling is disabled */
6338 if (!ring
->atr_sample_rate
)
6343 /* snag network header to get L4 type and address */
6344 hdr
.network
= skb_network_header(first
->skb
);
6346 /* Currently only IPv4/IPv6 with TCP is supported */
6347 if ((first
->protocol
!= __constant_htons(ETH_P_IPV6
) ||
6348 hdr
.ipv6
->nexthdr
!= IPPROTO_TCP
) &&
6349 (first
->protocol
!= __constant_htons(ETH_P_IP
) ||
6350 hdr
.ipv4
->protocol
!= IPPROTO_TCP
))
6353 th
= tcp_hdr(first
->skb
);
6355 /* skip this packet since it is invalid or the socket is closing */
6359 /* sample on all syn packets or once every atr sample count */
6360 if (!th
->syn
&& (ring
->atr_count
< ring
->atr_sample_rate
))
6363 /* reset sample count */
6364 ring
->atr_count
= 0;
6366 vlan_id
= htons(first
->tx_flags
>> IXGBE_TX_FLAGS_VLAN_SHIFT
);
6369 * src and dst are inverted, think how the receiver sees them
6371 * The input is broken into two sections, a non-compressed section
6372 * containing vm_pool, vlan_id, and flow_type. The rest of the data
6373 * is XORed together and stored in the compressed dword.
6375 input
.formatted
.vlan_id
= vlan_id
;
6378 * since src port and flex bytes occupy the same word XOR them together
6379 * and write the value to source port portion of compressed dword
6381 if (first
->tx_flags
& (IXGBE_TX_FLAGS_SW_VLAN
| IXGBE_TX_FLAGS_HW_VLAN
))
6382 common
.port
.src
^= th
->dest
^ __constant_htons(ETH_P_8021Q
);
6384 common
.port
.src
^= th
->dest
^ first
->protocol
;
6385 common
.port
.dst
^= th
->source
;
6387 if (first
->protocol
== __constant_htons(ETH_P_IP
)) {
6388 input
.formatted
.flow_type
= IXGBE_ATR_FLOW_TYPE_TCPV4
;
6389 common
.ip
^= hdr
.ipv4
->saddr
^ hdr
.ipv4
->daddr
;
6391 input
.formatted
.flow_type
= IXGBE_ATR_FLOW_TYPE_TCPV6
;
6392 common
.ip
^= hdr
.ipv6
->saddr
.s6_addr32
[0] ^
6393 hdr
.ipv6
->saddr
.s6_addr32
[1] ^
6394 hdr
.ipv6
->saddr
.s6_addr32
[2] ^
6395 hdr
.ipv6
->saddr
.s6_addr32
[3] ^
6396 hdr
.ipv6
->daddr
.s6_addr32
[0] ^
6397 hdr
.ipv6
->daddr
.s6_addr32
[1] ^
6398 hdr
.ipv6
->daddr
.s6_addr32
[2] ^
6399 hdr
.ipv6
->daddr
.s6_addr32
[3];
6402 /* This assumes the Rx queue and Tx queue are bound to the same CPU */
6403 ixgbe_fdir_add_signature_filter_82599(&q_vector
->adapter
->hw
,
6404 input
, common
, ring
->queue_index
);
6407 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring
*tx_ring
, u16 size
)
6409 netif_stop_subqueue(tx_ring
->netdev
, tx_ring
->queue_index
);
6410 /* Herbert's original patch had:
6411 * smp_mb__after_netif_stop_queue();
6412 * but since that doesn't exist yet, just open code it. */
6415 /* We need to check again in a case another CPU has just
6416 * made room available. */
6417 if (likely(ixgbe_desc_unused(tx_ring
) < size
))
6420 /* A reprieve! - use start_queue because it doesn't call schedule */
6421 netif_start_subqueue(tx_ring
->netdev
, tx_ring
->queue_index
);
6422 ++tx_ring
->tx_stats
.restart_queue
;
6426 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring
*tx_ring
, u16 size
)
6428 if (likely(ixgbe_desc_unused(tx_ring
) >= size
))
6430 return __ixgbe_maybe_stop_tx(tx_ring
, size
);
6434 static u16
ixgbe_select_queue(struct net_device
*dev
, struct sk_buff
*skb
)
6436 struct ixgbe_adapter
*adapter
;
6437 struct ixgbe_ring_feature
*f
;
6441 * only execute the code below if protocol is FCoE
6442 * or FIP and we have FCoE enabled on the adapter
6444 switch (vlan_get_protocol(skb
)) {
6445 case __constant_htons(ETH_P_FCOE
):
6446 case __constant_htons(ETH_P_FIP
):
6447 adapter
= netdev_priv(dev
);
6449 if (adapter
->flags
& IXGBE_FLAG_FCOE_ENABLED
)
6452 return __netdev_pick_tx(dev
, skb
);
6455 f
= &adapter
->ring_feature
[RING_F_FCOE
];
6457 txq
= skb_rx_queue_recorded(skb
) ? skb_get_rx_queue(skb
) :
6460 while (txq
>= f
->indices
)
6463 return txq
+ f
->offset
;
6467 netdev_tx_t
ixgbe_xmit_frame_ring(struct sk_buff
*skb
,
6468 struct ixgbe_adapter
*adapter
,
6469 struct ixgbe_ring
*tx_ring
)
6471 struct ixgbe_tx_buffer
*first
;
6475 u16 count
= TXD_USE_COUNT(skb_headlen(skb
));
6476 __be16 protocol
= skb
->protocol
;
6480 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
6481 * + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
6482 * + 2 desc gap to keep tail from touching head,
6483 * + 1 desc for context descriptor,
6484 * otherwise try next time
6486 for (f
= 0; f
< skb_shinfo(skb
)->nr_frags
; f
++)
6487 count
+= TXD_USE_COUNT(skb_shinfo(skb
)->frags
[f
].size
);
6489 if (ixgbe_maybe_stop_tx(tx_ring
, count
+ 3)) {
6490 tx_ring
->tx_stats
.tx_busy
++;
6491 return NETDEV_TX_BUSY
;
6494 /* record the location of the first descriptor for this packet */
6495 first
= &tx_ring
->tx_buffer_info
[tx_ring
->next_to_use
];
6497 first
->bytecount
= skb
->len
;
6498 first
->gso_segs
= 1;
6500 /* if we have a HW VLAN tag being added default to the HW one */
6501 if (vlan_tx_tag_present(skb
)) {
6502 tx_flags
|= vlan_tx_tag_get(skb
) << IXGBE_TX_FLAGS_VLAN_SHIFT
;
6503 tx_flags
|= IXGBE_TX_FLAGS_HW_VLAN
;
6504 /* else if it is a SW VLAN check the next protocol and store the tag */
6505 } else if (protocol
== __constant_htons(ETH_P_8021Q
)) {
6506 struct vlan_hdr
*vhdr
, _vhdr
;
6507 vhdr
= skb_header_pointer(skb
, ETH_HLEN
, sizeof(_vhdr
), &_vhdr
);
6511 protocol
= vhdr
->h_vlan_encapsulated_proto
;
6512 tx_flags
|= ntohs(vhdr
->h_vlan_TCI
) <<
6513 IXGBE_TX_FLAGS_VLAN_SHIFT
;
6514 tx_flags
|= IXGBE_TX_FLAGS_SW_VLAN
;
6517 skb_tx_timestamp(skb
);
6519 if (unlikely(skb_shinfo(skb
)->tx_flags
& SKBTX_HW_TSTAMP
)) {
6520 skb_shinfo(skb
)->tx_flags
|= SKBTX_IN_PROGRESS
;
6521 tx_flags
|= IXGBE_TX_FLAGS_TSTAMP
;
6523 /* schedule check for Tx timestamp */
6524 adapter
->ptp_tx_skb
= skb_get(skb
);
6525 adapter
->ptp_tx_start
= jiffies
;
6526 schedule_work(&adapter
->ptp_tx_work
);
6529 #ifdef CONFIG_PCI_IOV
6531 * Use the l2switch_enable flag - would be false if the DMA
6532 * Tx switch had been disabled.
6534 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
6535 tx_flags
|= IXGBE_TX_FLAGS_CC
;
6538 /* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
6539 if ((adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
) &&
6540 ((tx_flags
& (IXGBE_TX_FLAGS_HW_VLAN
| IXGBE_TX_FLAGS_SW_VLAN
)) ||
6541 (skb
->priority
!= TC_PRIO_CONTROL
))) {
6542 tx_flags
&= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK
;
6543 tx_flags
|= (skb
->priority
& 0x7) <<
6544 IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT
;
6545 if (tx_flags
& IXGBE_TX_FLAGS_SW_VLAN
) {
6546 struct vlan_ethhdr
*vhdr
;
6547 if (skb_header_cloned(skb
) &&
6548 pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
))
6550 vhdr
= (struct vlan_ethhdr
*)skb
->data
;
6551 vhdr
->h_vlan_TCI
= htons(tx_flags
>>
6552 IXGBE_TX_FLAGS_VLAN_SHIFT
);
6554 tx_flags
|= IXGBE_TX_FLAGS_HW_VLAN
;
6558 /* record initial flags and protocol */
6559 first
->tx_flags
= tx_flags
;
6560 first
->protocol
= protocol
;
6563 /* setup tx offload for FCoE */
6564 if ((protocol
== __constant_htons(ETH_P_FCOE
)) &&
6565 (tx_ring
->netdev
->features
& (NETIF_F_FSO
| NETIF_F_FCOE_CRC
))) {
6566 tso
= ixgbe_fso(tx_ring
, first
, &hdr_len
);
6573 #endif /* IXGBE_FCOE */
6574 tso
= ixgbe_tso(tx_ring
, first
, &hdr_len
);
6578 ixgbe_tx_csum(tx_ring
, first
);
6580 /* add the ATR filter if ATR is on */
6581 if (test_bit(__IXGBE_TX_FDIR_INIT_DONE
, &tx_ring
->state
))
6582 ixgbe_atr(tx_ring
, first
);
6586 #endif /* IXGBE_FCOE */
6587 ixgbe_tx_map(tx_ring
, first
, hdr_len
);
6589 ixgbe_maybe_stop_tx(tx_ring
, DESC_NEEDED
);
6591 return NETDEV_TX_OK
;
6594 dev_kfree_skb_any(first
->skb
);
6597 return NETDEV_TX_OK
;
6600 static netdev_tx_t
ixgbe_xmit_frame(struct sk_buff
*skb
,
6601 struct net_device
*netdev
)
6603 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6604 struct ixgbe_ring
*tx_ring
;
6607 * The minimum packet size for olinfo paylen is 17 so pad the skb
6608 * in order to meet this minimum size requirement.
6610 if (unlikely(skb
->len
< 17)) {
6611 if (skb_pad(skb
, 17 - skb
->len
))
6612 return NETDEV_TX_OK
;
6614 skb_set_tail_pointer(skb
, 17);
6617 tx_ring
= adapter
->tx_ring
[skb
->queue_mapping
];
6618 return ixgbe_xmit_frame_ring(skb
, adapter
, tx_ring
);
6622 * ixgbe_set_mac - Change the Ethernet Address of the NIC
6623 * @netdev: network interface device structure
6624 * @p: pointer to an address structure
6626 * Returns 0 on success, negative on failure
6628 static int ixgbe_set_mac(struct net_device
*netdev
, void *p
)
6630 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6631 struct ixgbe_hw
*hw
= &adapter
->hw
;
6632 struct sockaddr
*addr
= p
;
6634 if (!is_valid_ether_addr(addr
->sa_data
))
6635 return -EADDRNOTAVAIL
;
6637 memcpy(netdev
->dev_addr
, addr
->sa_data
, netdev
->addr_len
);
6638 memcpy(hw
->mac
.addr
, addr
->sa_data
, netdev
->addr_len
);
6640 hw
->mac
.ops
.set_rar(hw
, 0, hw
->mac
.addr
, VMDQ_P(0), IXGBE_RAH_AV
);
6646 ixgbe_mdio_read(struct net_device
*netdev
, int prtad
, int devad
, u16 addr
)
6648 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6649 struct ixgbe_hw
*hw
= &adapter
->hw
;
6653 if (prtad
!= hw
->phy
.mdio
.prtad
)
6655 rc
= hw
->phy
.ops
.read_reg(hw
, addr
, devad
, &value
);
6661 static int ixgbe_mdio_write(struct net_device
*netdev
, int prtad
, int devad
,
6662 u16 addr
, u16 value
)
6664 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6665 struct ixgbe_hw
*hw
= &adapter
->hw
;
6667 if (prtad
!= hw
->phy
.mdio
.prtad
)
6669 return hw
->phy
.ops
.write_reg(hw
, addr
, devad
, value
);
6672 static int ixgbe_ioctl(struct net_device
*netdev
, struct ifreq
*req
, int cmd
)
6674 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6678 return ixgbe_ptp_hwtstamp_ioctl(adapter
, req
, cmd
);
6680 return mdio_mii_ioctl(&adapter
->hw
.phy
.mdio
, if_mii(req
), cmd
);
6685 * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
6687 * @netdev: network interface device structure
6689 * Returns non-zero on failure
6691 static int ixgbe_add_sanmac_netdev(struct net_device
*dev
)
6694 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
6695 struct ixgbe_hw
*hw
= &adapter
->hw
;
6697 if (is_valid_ether_addr(hw
->mac
.san_addr
)) {
6699 err
= dev_addr_add(dev
, hw
->mac
.san_addr
, NETDEV_HW_ADDR_T_SAN
);
6702 /* update SAN MAC vmdq pool selection */
6703 hw
->mac
.ops
.set_vmdq_san_mac(hw
, VMDQ_P(0));
6709 * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
6711 * @netdev: network interface device structure
6713 * Returns non-zero on failure
6715 static int ixgbe_del_sanmac_netdev(struct net_device
*dev
)
6718 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
6719 struct ixgbe_mac_info
*mac
= &adapter
->hw
.mac
;
6721 if (is_valid_ether_addr(mac
->san_addr
)) {
6723 err
= dev_addr_del(dev
, mac
->san_addr
, NETDEV_HW_ADDR_T_SAN
);
6729 #ifdef CONFIG_NET_POLL_CONTROLLER
6731 * Polling 'interrupt' - used by things like netconsole to send skbs
6732 * without having to re-enable interrupts. It's not called while
6733 * the interrupt routine is executing.
6735 static void ixgbe_netpoll(struct net_device
*netdev
)
6737 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6740 /* if interface is down do nothing */
6741 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
6744 adapter
->flags
|= IXGBE_FLAG_IN_NETPOLL
;
6745 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
) {
6746 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
6747 ixgbe_msix_clean_rings(0, adapter
->q_vector
[i
]);
6749 ixgbe_intr(adapter
->pdev
->irq
, netdev
);
6751 adapter
->flags
&= ~IXGBE_FLAG_IN_NETPOLL
;
6755 static struct rtnl_link_stats64
*ixgbe_get_stats64(struct net_device
*netdev
,
6756 struct rtnl_link_stats64
*stats
)
6758 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6762 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
6763 struct ixgbe_ring
*ring
= ACCESS_ONCE(adapter
->rx_ring
[i
]);
6769 start
= u64_stats_fetch_begin_bh(&ring
->syncp
);
6770 packets
= ring
->stats
.packets
;
6771 bytes
= ring
->stats
.bytes
;
6772 } while (u64_stats_fetch_retry_bh(&ring
->syncp
, start
));
6773 stats
->rx_packets
+= packets
;
6774 stats
->rx_bytes
+= bytes
;
6778 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
6779 struct ixgbe_ring
*ring
= ACCESS_ONCE(adapter
->tx_ring
[i
]);
6785 start
= u64_stats_fetch_begin_bh(&ring
->syncp
);
6786 packets
= ring
->stats
.packets
;
6787 bytes
= ring
->stats
.bytes
;
6788 } while (u64_stats_fetch_retry_bh(&ring
->syncp
, start
));
6789 stats
->tx_packets
+= packets
;
6790 stats
->tx_bytes
+= bytes
;
6794 /* following stats updated by ixgbe_watchdog_task() */
6795 stats
->multicast
= netdev
->stats
.multicast
;
6796 stats
->rx_errors
= netdev
->stats
.rx_errors
;
6797 stats
->rx_length_errors
= netdev
->stats
.rx_length_errors
;
6798 stats
->rx_crc_errors
= netdev
->stats
.rx_crc_errors
;
6799 stats
->rx_missed_errors
= netdev
->stats
.rx_missed_errors
;
6803 #ifdef CONFIG_IXGBE_DCB
6805 * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
6806 * @adapter: pointer to ixgbe_adapter
6807 * @tc: number of traffic classes currently enabled
6809 * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
6810 * 802.1Q priority maps to a packet buffer that exists.
6812 static void ixgbe_validate_rtr(struct ixgbe_adapter
*adapter
, u8 tc
)
6814 struct ixgbe_hw
*hw
= &adapter
->hw
;
6818 /* 82598 have a static priority to TC mapping that can not
6819 * be changed so no validation is needed.
6821 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
6824 reg
= IXGBE_READ_REG(hw
, IXGBE_RTRUP2TC
);
6827 for (i
= 0; i
< MAX_TRAFFIC_CLASS
; i
++) {
6828 u8 up2tc
= reg
>> (i
* IXGBE_RTRUP2TC_UP_SHIFT
);
6830 /* If up2tc is out of bounds default to zero */
6832 reg
&= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT
);
6836 IXGBE_WRITE_REG(hw
, IXGBE_RTRUP2TC
, reg
);
6842 * ixgbe_set_prio_tc_map - Configure netdev prio tc map
6843 * @adapter: Pointer to adapter struct
6845 * Populate the netdev user priority to tc map
6847 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter
*adapter
)
6849 struct net_device
*dev
= adapter
->netdev
;
6850 struct ixgbe_dcb_config
*dcb_cfg
= &adapter
->dcb_cfg
;
6851 struct ieee_ets
*ets
= adapter
->ixgbe_ieee_ets
;
6854 for (prio
= 0; prio
< MAX_USER_PRIORITY
; prio
++) {
6857 if (adapter
->dcbx_cap
& DCB_CAP_DCBX_VER_CEE
)
6858 tc
= ixgbe_dcb_get_tc_from_up(dcb_cfg
, 0, prio
);
6860 tc
= ets
->prio_tc
[prio
];
6862 netdev_set_prio_tc_map(dev
, prio
, tc
);
6866 #endif /* CONFIG_IXGBE_DCB */
6868 * ixgbe_setup_tc - configure net_device for multiple traffic classes
6870 * @netdev: net device to configure
6871 * @tc: number of traffic classes to enable
6873 int ixgbe_setup_tc(struct net_device
*dev
, u8 tc
)
6875 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
6876 struct ixgbe_hw
*hw
= &adapter
->hw
;
6878 /* Hardware supports up to 8 traffic classes */
6879 if (tc
> adapter
->dcb_cfg
.num_tcs
.pg_tcs
||
6880 (hw
->mac
.type
== ixgbe_mac_82598EB
&&
6881 tc
< MAX_TRAFFIC_CLASS
))
6884 /* Hardware has to reinitialize queues and interrupts to
6885 * match packet buffer alignment. Unfortunately, the
6886 * hardware is not flexible enough to do this dynamically.
6888 if (netif_running(dev
))
6890 ixgbe_clear_interrupt_scheme(adapter
);
6892 #ifdef CONFIG_IXGBE_DCB
6894 netdev_set_num_tc(dev
, tc
);
6895 ixgbe_set_prio_tc_map(adapter
);
6897 adapter
->flags
|= IXGBE_FLAG_DCB_ENABLED
;
6899 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
) {
6900 adapter
->last_lfc_mode
= adapter
->hw
.fc
.requested_mode
;
6901 adapter
->hw
.fc
.requested_mode
= ixgbe_fc_none
;
6904 netdev_reset_tc(dev
);
6906 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
)
6907 adapter
->hw
.fc
.requested_mode
= adapter
->last_lfc_mode
;
6909 adapter
->flags
&= ~IXGBE_FLAG_DCB_ENABLED
;
6911 adapter
->temp_dcb_cfg
.pfc_mode_enable
= false;
6912 adapter
->dcb_cfg
.pfc_mode_enable
= false;
6915 ixgbe_validate_rtr(adapter
, tc
);
6917 #endif /* CONFIG_IXGBE_DCB */
6918 ixgbe_init_interrupt_scheme(adapter
);
6920 if (netif_running(dev
))
6921 return ixgbe_open(dev
);
6926 #ifdef CONFIG_PCI_IOV
6927 void ixgbe_sriov_reinit(struct ixgbe_adapter
*adapter
)
6929 struct net_device
*netdev
= adapter
->netdev
;
6932 ixgbe_setup_tc(netdev
, netdev_get_num_tc(netdev
));
6937 void ixgbe_do_reset(struct net_device
*netdev
)
6939 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6941 if (netif_running(netdev
))
6942 ixgbe_reinit_locked(adapter
);
6944 ixgbe_reset(adapter
);
6947 static netdev_features_t
ixgbe_fix_features(struct net_device
*netdev
,
6948 netdev_features_t features
)
6950 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6952 /* If Rx checksum is disabled, then RSC/LRO should also be disabled */
6953 if (!(features
& NETIF_F_RXCSUM
))
6954 features
&= ~NETIF_F_LRO
;
6956 /* Turn off LRO if not RSC capable */
6957 if (!(adapter
->flags2
& IXGBE_FLAG2_RSC_CAPABLE
))
6958 features
&= ~NETIF_F_LRO
;
6963 static int ixgbe_set_features(struct net_device
*netdev
,
6964 netdev_features_t features
)
6966 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6967 netdev_features_t changed
= netdev
->features
^ features
;
6968 bool need_reset
= false;
6970 /* Make sure RSC matches LRO, reset if change */
6971 if (!(features
& NETIF_F_LRO
)) {
6972 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)
6974 adapter
->flags2
&= ~IXGBE_FLAG2_RSC_ENABLED
;
6975 } else if ((adapter
->flags2
& IXGBE_FLAG2_RSC_CAPABLE
) &&
6976 !(adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)) {
6977 if (adapter
->rx_itr_setting
== 1 ||
6978 adapter
->rx_itr_setting
> IXGBE_MIN_RSC_ITR
) {
6979 adapter
->flags2
|= IXGBE_FLAG2_RSC_ENABLED
;
6981 } else if ((changed
^ features
) & NETIF_F_LRO
) {
6982 e_info(probe
, "rx-usecs set too low, "
6988 * Check if Flow Director n-tuple support was enabled or disabled. If
6989 * the state changed, we need to reset.
6991 switch (features
& NETIF_F_NTUPLE
) {
6992 case NETIF_F_NTUPLE
:
6993 /* turn off ATR, enable perfect filters and reset */
6994 if (!(adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
))
6997 adapter
->flags
&= ~IXGBE_FLAG_FDIR_HASH_CAPABLE
;
6998 adapter
->flags
|= IXGBE_FLAG_FDIR_PERFECT_CAPABLE
;
7001 /* turn off perfect filters, enable ATR and reset */
7002 if (adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
)
7005 adapter
->flags
&= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE
;
7007 /* We cannot enable ATR if SR-IOV is enabled */
7008 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
7011 /* We cannot enable ATR if we have 2 or more traffic classes */
7012 if (netdev_get_num_tc(netdev
) > 1)
7015 /* We cannot enable ATR if RSS is disabled */
7016 if (adapter
->ring_feature
[RING_F_RSS
].limit
<= 1)
7019 /* A sample rate of 0 indicates ATR disabled */
7020 if (!adapter
->atr_sample_rate
)
7023 adapter
->flags
|= IXGBE_FLAG_FDIR_HASH_CAPABLE
;
7027 if (features
& NETIF_F_HW_VLAN_CTAG_RX
)
7028 ixgbe_vlan_strip_enable(adapter
);
7030 ixgbe_vlan_strip_disable(adapter
);
7032 if (changed
& NETIF_F_RXALL
)
7035 netdev
->features
= features
;
7037 ixgbe_do_reset(netdev
);
7042 static int ixgbe_ndo_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
7043 struct net_device
*dev
,
7044 const unsigned char *addr
,
7047 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
7050 if (!(adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
))
7051 return ndo_dflt_fdb_add(ndm
, tb
, dev
, addr
, flags
);
7053 /* Hardware does not support aging addresses so if a
7054 * ndm_state is given only allow permanent addresses
7056 if (ndm
->ndm_state
&& !(ndm
->ndm_state
& NUD_PERMANENT
)) {
7057 pr_info("%s: FDB only supports static addresses\n",
7062 if (is_unicast_ether_addr(addr
) || is_link_local_ether_addr(addr
)) {
7063 u32 rar_uc_entries
= IXGBE_MAX_PF_MACVLANS
;
7065 if (netdev_uc_count(dev
) < rar_uc_entries
)
7066 err
= dev_uc_add_excl(dev
, addr
);
7069 } else if (is_multicast_ether_addr(addr
)) {
7070 err
= dev_mc_add_excl(dev
, addr
);
7075 /* Only return duplicate errors if NLM_F_EXCL is set */
7076 if (err
== -EEXIST
&& !(flags
& NLM_F_EXCL
))
7082 static int ixgbe_ndo_bridge_setlink(struct net_device
*dev
,
7083 struct nlmsghdr
*nlh
)
7085 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
7086 struct nlattr
*attr
, *br_spec
;
7089 if (!(adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
))
7092 br_spec
= nlmsg_find_attr(nlh
, sizeof(struct ifinfomsg
), IFLA_AF_SPEC
);
7094 nla_for_each_nested(attr
, br_spec
, rem
) {
7098 if (nla_type(attr
) != IFLA_BRIDGE_MODE
)
7101 mode
= nla_get_u16(attr
);
7102 if (mode
== BRIDGE_MODE_VEPA
) {
7104 adapter
->flags2
&= ~IXGBE_FLAG2_BRIDGE_MODE_VEB
;
7105 } else if (mode
== BRIDGE_MODE_VEB
) {
7106 reg
= IXGBE_PFDTXGSWC_VT_LBEN
;
7107 adapter
->flags2
|= IXGBE_FLAG2_BRIDGE_MODE_VEB
;
7111 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_PFDTXGSWC
, reg
);
7113 e_info(drv
, "enabling bridge mode: %s\n",
7114 mode
== BRIDGE_MODE_VEPA
? "VEPA" : "VEB");
7120 static int ixgbe_ndo_bridge_getlink(struct sk_buff
*skb
, u32 pid
, u32 seq
,
7121 struct net_device
*dev
,
7124 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
7127 if (!(adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
))
7130 if (adapter
->flags2
& IXGBE_FLAG2_BRIDGE_MODE_VEB
)
7131 mode
= BRIDGE_MODE_VEB
;
7133 mode
= BRIDGE_MODE_VEPA
;
7135 return ndo_dflt_bridge_getlink(skb
, pid
, seq
, dev
, mode
);
7138 static const struct net_device_ops ixgbe_netdev_ops
= {
7139 .ndo_open
= ixgbe_open
,
7140 .ndo_stop
= ixgbe_close
,
7141 .ndo_start_xmit
= ixgbe_xmit_frame
,
7143 .ndo_select_queue
= ixgbe_select_queue
,
7145 .ndo_set_rx_mode
= ixgbe_set_rx_mode
,
7146 .ndo_validate_addr
= eth_validate_addr
,
7147 .ndo_set_mac_address
= ixgbe_set_mac
,
7148 .ndo_change_mtu
= ixgbe_change_mtu
,
7149 .ndo_tx_timeout
= ixgbe_tx_timeout
,
7150 .ndo_vlan_rx_add_vid
= ixgbe_vlan_rx_add_vid
,
7151 .ndo_vlan_rx_kill_vid
= ixgbe_vlan_rx_kill_vid
,
7152 .ndo_do_ioctl
= ixgbe_ioctl
,
7153 .ndo_set_vf_mac
= ixgbe_ndo_set_vf_mac
,
7154 .ndo_set_vf_vlan
= ixgbe_ndo_set_vf_vlan
,
7155 .ndo_set_vf_tx_rate
= ixgbe_ndo_set_vf_bw
,
7156 .ndo_set_vf_spoofchk
= ixgbe_ndo_set_vf_spoofchk
,
7157 .ndo_get_vf_config
= ixgbe_ndo_get_vf_config
,
7158 .ndo_get_stats64
= ixgbe_get_stats64
,
7159 #ifdef CONFIG_IXGBE_DCB
7160 .ndo_setup_tc
= ixgbe_setup_tc
,
7162 #ifdef CONFIG_NET_POLL_CONTROLLER
7163 .ndo_poll_controller
= ixgbe_netpoll
,
7166 .ndo_fcoe_ddp_setup
= ixgbe_fcoe_ddp_get
,
7167 .ndo_fcoe_ddp_target
= ixgbe_fcoe_ddp_target
,
7168 .ndo_fcoe_ddp_done
= ixgbe_fcoe_ddp_put
,
7169 .ndo_fcoe_enable
= ixgbe_fcoe_enable
,
7170 .ndo_fcoe_disable
= ixgbe_fcoe_disable
,
7171 .ndo_fcoe_get_wwn
= ixgbe_fcoe_get_wwn
,
7172 .ndo_fcoe_get_hbainfo
= ixgbe_fcoe_get_hbainfo
,
7173 #endif /* IXGBE_FCOE */
7174 .ndo_set_features
= ixgbe_set_features
,
7175 .ndo_fix_features
= ixgbe_fix_features
,
7176 .ndo_fdb_add
= ixgbe_ndo_fdb_add
,
7177 .ndo_bridge_setlink
= ixgbe_ndo_bridge_setlink
,
7178 .ndo_bridge_getlink
= ixgbe_ndo_bridge_getlink
,
7182 * ixgbe_wol_supported - Check whether device supports WoL
7183 * @hw: hw specific details
7184 * @device_id: the device ID
7185 * @subdev_id: the subsystem device ID
7187 * This function is used by probe and ethtool to determine
7188 * which devices have WoL support
7191 int ixgbe_wol_supported(struct ixgbe_adapter
*adapter
, u16 device_id
,
7194 struct ixgbe_hw
*hw
= &adapter
->hw
;
7195 u16 wol_cap
= adapter
->eeprom_cap
& IXGBE_DEVICE_CAPS_WOL_MASK
;
7196 int is_wol_supported
= 0;
7198 switch (device_id
) {
7199 case IXGBE_DEV_ID_82599_SFP
:
7200 /* Only these subdevices could supports WOL */
7201 switch (subdevice_id
) {
7202 case IXGBE_SUBDEV_ID_82599_560FLR
:
7203 /* only support first port */
7204 if (hw
->bus
.func
!= 0)
7206 case IXGBE_SUBDEV_ID_82599_SFP
:
7207 case IXGBE_SUBDEV_ID_82599_RNDC
:
7208 case IXGBE_SUBDEV_ID_82599_ECNA_DP
:
7209 case IXGBE_SUBDEV_ID_82599_LOM_SFP
:
7210 is_wol_supported
= 1;
7214 case IXGBE_DEV_ID_82599_COMBO_BACKPLANE
:
7215 /* All except this subdevice support WOL */
7216 if (subdevice_id
!= IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ
)
7217 is_wol_supported
= 1;
7219 case IXGBE_DEV_ID_82599_KX4
:
7220 is_wol_supported
= 1;
7222 case IXGBE_DEV_ID_X540T
:
7223 case IXGBE_DEV_ID_X540T1
:
7224 /* check eeprom to see if enabled wol */
7225 if ((wol_cap
== IXGBE_DEVICE_CAPS_WOL_PORT0_1
) ||
7226 ((wol_cap
== IXGBE_DEVICE_CAPS_WOL_PORT0
) &&
7227 (hw
->bus
.func
== 0))) {
7228 is_wol_supported
= 1;
7233 return is_wol_supported
;
7237 * ixgbe_probe - Device Initialization Routine
7238 * @pdev: PCI device information struct
7239 * @ent: entry in ixgbe_pci_tbl
7241 * Returns 0 on success, negative on failure
7243 * ixgbe_probe initializes an adapter identified by a pci_dev structure.
7244 * The OS initialization, configuring of the adapter private structure,
7245 * and a hardware reset occur.
7247 static int ixgbe_probe(struct pci_dev
*pdev
, const struct pci_device_id
*ent
)
7249 struct net_device
*netdev
;
7250 struct ixgbe_adapter
*adapter
= NULL
;
7251 struct ixgbe_hw
*hw
;
7252 const struct ixgbe_info
*ii
= ixgbe_info_tbl
[ent
->driver_data
];
7253 static int cards_found
;
7254 int i
, err
, pci_using_dac
;
7255 unsigned int indices
= MAX_TX_QUEUES
;
7256 u8 part_str
[IXGBE_PBANUM_LENGTH
];
7262 /* Catch broken hardware that put the wrong VF device ID in
7263 * the PCIe SR-IOV capability.
7265 if (pdev
->is_virtfn
) {
7266 WARN(1, KERN_ERR
"%s (%hx:%hx) should not be a VF!\n",
7267 pci_name(pdev
), pdev
->vendor
, pdev
->device
);
7271 err
= pci_enable_device_mem(pdev
);
7275 if (!dma_set_mask(&pdev
->dev
, DMA_BIT_MASK(64)) &&
7276 !dma_set_coherent_mask(&pdev
->dev
, DMA_BIT_MASK(64))) {
7279 err
= dma_set_mask(&pdev
->dev
, DMA_BIT_MASK(32));
7281 err
= dma_set_coherent_mask(&pdev
->dev
,
7285 "No usable DMA configuration, aborting\n");
7292 err
= pci_request_selected_regions(pdev
, pci_select_bars(pdev
,
7293 IORESOURCE_MEM
), ixgbe_driver_name
);
7296 "pci_request_selected_regions failed 0x%x\n", err
);
7300 pci_enable_pcie_error_reporting(pdev
);
7302 pci_set_master(pdev
);
7303 pci_save_state(pdev
);
7305 if (ii
->mac
== ixgbe_mac_82598EB
) {
7306 #ifdef CONFIG_IXGBE_DCB
7307 /* 8 TC w/ 4 queues per TC */
7308 indices
= 4 * MAX_TRAFFIC_CLASS
;
7310 indices
= IXGBE_MAX_RSS_INDICES
;
7314 netdev
= alloc_etherdev_mq(sizeof(struct ixgbe_adapter
), indices
);
7317 goto err_alloc_etherdev
;
7320 SET_NETDEV_DEV(netdev
, &pdev
->dev
);
7322 adapter
= netdev_priv(netdev
);
7323 pci_set_drvdata(pdev
, adapter
);
7325 adapter
->netdev
= netdev
;
7326 adapter
->pdev
= pdev
;
7329 adapter
->msg_enable
= netif_msg_init(debug
, DEFAULT_MSG_ENABLE
);
7331 hw
->hw_addr
= ioremap(pci_resource_start(pdev
, 0),
7332 pci_resource_len(pdev
, 0));
7338 netdev
->netdev_ops
= &ixgbe_netdev_ops
;
7339 ixgbe_set_ethtool_ops(netdev
);
7340 netdev
->watchdog_timeo
= 5 * HZ
;
7341 strncpy(netdev
->name
, pci_name(pdev
), sizeof(netdev
->name
) - 1);
7343 adapter
->bd_number
= cards_found
;
7346 memcpy(&hw
->mac
.ops
, ii
->mac_ops
, sizeof(hw
->mac
.ops
));
7347 hw
->mac
.type
= ii
->mac
;
7350 memcpy(&hw
->eeprom
.ops
, ii
->eeprom_ops
, sizeof(hw
->eeprom
.ops
));
7351 eec
= IXGBE_READ_REG(hw
, IXGBE_EEC
);
7352 /* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
7353 if (!(eec
& (1 << 8)))
7354 hw
->eeprom
.ops
.read
= &ixgbe_read_eeprom_bit_bang_generic
;
7357 memcpy(&hw
->phy
.ops
, ii
->phy_ops
, sizeof(hw
->phy
.ops
));
7358 hw
->phy
.sfp_type
= ixgbe_sfp_type_unknown
;
7359 /* ixgbe_identify_phy_generic will set prtad and mmds properly */
7360 hw
->phy
.mdio
.prtad
= MDIO_PRTAD_NONE
;
7361 hw
->phy
.mdio
.mmds
= 0;
7362 hw
->phy
.mdio
.mode_support
= MDIO_SUPPORTS_C45
| MDIO_EMULATE_C22
;
7363 hw
->phy
.mdio
.dev
= netdev
;
7364 hw
->phy
.mdio
.mdio_read
= ixgbe_mdio_read
;
7365 hw
->phy
.mdio
.mdio_write
= ixgbe_mdio_write
;
7367 ii
->get_invariants(hw
);
7369 /* setup the private structure */
7370 err
= ixgbe_sw_init(adapter
);
7374 /* Cache if MNG FW is up so we don't have to read the REG later */
7375 if (hw
->mac
.ops
.mng_fw_enabled
)
7376 hw
->mng_fw_enabled
= hw
->mac
.ops
.mng_fw_enabled(hw
);
7378 /* Make it possible the adapter to be woken up via WOL */
7379 switch (adapter
->hw
.mac
.type
) {
7380 case ixgbe_mac_82599EB
:
7381 case ixgbe_mac_X540
:
7382 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_WUS
, ~0);
7389 * If there is a fan on this device and it has failed log the
7392 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
) {
7393 u32 esdp
= IXGBE_READ_REG(hw
, IXGBE_ESDP
);
7394 if (esdp
& IXGBE_ESDP_SDP1
)
7395 e_crit(probe
, "Fan has stopped, replace the adapter\n");
7398 if (allow_unsupported_sfp
)
7399 hw
->allow_unsupported_sfp
= allow_unsupported_sfp
;
7401 /* reset_hw fills in the perm_addr as well */
7402 hw
->phy
.reset_if_overtemp
= true;
7403 err
= hw
->mac
.ops
.reset_hw(hw
);
7404 hw
->phy
.reset_if_overtemp
= false;
7405 if (err
== IXGBE_ERR_SFP_NOT_PRESENT
&&
7406 hw
->mac
.type
== ixgbe_mac_82598EB
) {
7408 } else if (err
== IXGBE_ERR_SFP_NOT_SUPPORTED
) {
7409 e_dev_err("failed to load because an unsupported SFP+ "
7410 "module type was detected.\n");
7411 e_dev_err("Reload the driver after installing a supported "
7415 e_dev_err("HW Init failed: %d\n", err
);
7419 #ifdef CONFIG_PCI_IOV
7420 /* SR-IOV not supported on the 82598 */
7421 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
)
7424 ixgbe_init_mbx_params_pf(hw
);
7425 memcpy(&hw
->mbx
.ops
, ii
->mbx_ops
, sizeof(hw
->mbx
.ops
));
7426 ixgbe_enable_sriov(adapter
);
7427 pci_sriov_set_totalvfs(pdev
, 63);
7431 netdev
->features
= NETIF_F_SG
|
7434 NETIF_F_HW_VLAN_CTAG_TX
|
7435 NETIF_F_HW_VLAN_CTAG_RX
|
7436 NETIF_F_HW_VLAN_CTAG_FILTER
|
7442 netdev
->hw_features
= netdev
->features
;
7444 switch (adapter
->hw
.mac
.type
) {
7445 case ixgbe_mac_82599EB
:
7446 case ixgbe_mac_X540
:
7447 netdev
->features
|= NETIF_F_SCTP_CSUM
;
7448 netdev
->hw_features
|= NETIF_F_SCTP_CSUM
|
7455 netdev
->hw_features
|= NETIF_F_RXALL
;
7457 netdev
->vlan_features
|= NETIF_F_TSO
;
7458 netdev
->vlan_features
|= NETIF_F_TSO6
;
7459 netdev
->vlan_features
|= NETIF_F_IP_CSUM
;
7460 netdev
->vlan_features
|= NETIF_F_IPV6_CSUM
;
7461 netdev
->vlan_features
|= NETIF_F_SG
;
7463 netdev
->priv_flags
|= IFF_UNICAST_FLT
;
7464 netdev
->priv_flags
|= IFF_SUPP_NOFCS
;
7466 #ifdef CONFIG_IXGBE_DCB
7467 netdev
->dcbnl_ops
= &dcbnl_ops
;
7471 if (adapter
->flags
& IXGBE_FLAG_FCOE_CAPABLE
) {
7472 unsigned int fcoe_l
;
7474 if (hw
->mac
.ops
.get_device_caps
) {
7475 hw
->mac
.ops
.get_device_caps(hw
, &device_caps
);
7476 if (device_caps
& IXGBE_DEVICE_CAPS_FCOE_OFFLOADS
)
7477 adapter
->flags
&= ~IXGBE_FLAG_FCOE_CAPABLE
;
7481 fcoe_l
= min_t(int, IXGBE_FCRETA_SIZE
, num_online_cpus());
7482 adapter
->ring_feature
[RING_F_FCOE
].limit
= fcoe_l
;
7484 netdev
->features
|= NETIF_F_FSO
|
7487 netdev
->vlan_features
|= NETIF_F_FSO
|
7491 #endif /* IXGBE_FCOE */
7492 if (pci_using_dac
) {
7493 netdev
->features
|= NETIF_F_HIGHDMA
;
7494 netdev
->vlan_features
|= NETIF_F_HIGHDMA
;
7497 if (adapter
->flags2
& IXGBE_FLAG2_RSC_CAPABLE
)
7498 netdev
->hw_features
|= NETIF_F_LRO
;
7499 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)
7500 netdev
->features
|= NETIF_F_LRO
;
7502 /* make sure the EEPROM is good */
7503 if (hw
->eeprom
.ops
.validate_checksum(hw
, NULL
) < 0) {
7504 e_dev_err("The EEPROM Checksum Is Not Valid\n");
7509 memcpy(netdev
->dev_addr
, hw
->mac
.perm_addr
, netdev
->addr_len
);
7511 if (!is_valid_ether_addr(netdev
->dev_addr
)) {
7512 e_dev_err("invalid MAC address\n");
7517 setup_timer(&adapter
->service_timer
, &ixgbe_service_timer
,
7518 (unsigned long) adapter
);
7520 INIT_WORK(&adapter
->service_task
, ixgbe_service_task
);
7521 clear_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
);
7523 err
= ixgbe_init_interrupt_scheme(adapter
);
7527 /* WOL not supported for all devices */
7529 hw
->eeprom
.ops
.read(hw
, 0x2c, &adapter
->eeprom_cap
);
7530 hw
->wol_supported
= ixgbe_wol_supported(adapter
, pdev
->device
,
7531 pdev
->subsystem_device
);
7532 if (hw
->wol_supported
)
7533 adapter
->wol
= IXGBE_WUFC_MAG
;
7535 device_set_wakeup_enable(&adapter
->pdev
->dev
, adapter
->wol
);
7537 /* save off EEPROM version number */
7538 hw
->eeprom
.ops
.read(hw
, 0x2e, &adapter
->eeprom_verh
);
7539 hw
->eeprom
.ops
.read(hw
, 0x2d, &adapter
->eeprom_verl
);
7541 /* pick up the PCI bus settings for reporting later */
7542 hw
->mac
.ops
.get_bus_info(hw
);
7543 if (hw
->device_id
== IXGBE_DEV_ID_82599_SFP_SF_QP
)
7544 ixgbe_get_parent_bus_info(adapter
);
7546 /* print bus type/speed/width info */
7547 e_dev_info("(PCI Express:%s:%s) %pM\n",
7548 (hw
->bus
.speed
== ixgbe_bus_speed_8000
? "8.0GT/s" :
7549 hw
->bus
.speed
== ixgbe_bus_speed_5000
? "5.0GT/s" :
7550 hw
->bus
.speed
== ixgbe_bus_speed_2500
? "2.5GT/s" :
7552 (hw
->bus
.width
== ixgbe_bus_width_pcie_x8
? "Width x8" :
7553 hw
->bus
.width
== ixgbe_bus_width_pcie_x4
? "Width x4" :
7554 hw
->bus
.width
== ixgbe_bus_width_pcie_x1
? "Width x1" :
7558 err
= ixgbe_read_pba_string_generic(hw
, part_str
, IXGBE_PBANUM_LENGTH
);
7560 strncpy(part_str
, "Unknown", IXGBE_PBANUM_LENGTH
);
7561 if (ixgbe_is_sfp(hw
) && hw
->phy
.sfp_type
!= ixgbe_sfp_type_not_present
)
7562 e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
7563 hw
->mac
.type
, hw
->phy
.type
, hw
->phy
.sfp_type
,
7566 e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
7567 hw
->mac
.type
, hw
->phy
.type
, part_str
);
7569 if (hw
->bus
.width
<= ixgbe_bus_width_pcie_x4
) {
7570 e_dev_warn("PCI-Express bandwidth available for this card is "
7571 "not sufficient for optimal performance.\n");
7572 e_dev_warn("For optimal performance a x8 PCI-Express slot "
7576 /* reset the hardware with the new settings */
7577 err
= hw
->mac
.ops
.start_hw(hw
);
7578 if (err
== IXGBE_ERR_EEPROM_VERSION
) {
7579 /* We are running on a pre-production device, log a warning */
7580 e_dev_warn("This device is a pre-production adapter/LOM. "
7581 "Please be aware there may be issues associated "
7582 "with your hardware. If you are experiencing "
7583 "problems please contact your Intel or hardware "
7584 "representative who provided you with this "
7587 strcpy(netdev
->name
, "eth%d");
7588 err
= register_netdev(netdev
);
7592 /* power down the optics for 82599 SFP+ fiber */
7593 if (hw
->mac
.ops
.disable_tx_laser
)
7594 hw
->mac
.ops
.disable_tx_laser(hw
);
7596 /* carrier off reporting is important to ethtool even BEFORE open */
7597 netif_carrier_off(netdev
);
7599 #ifdef CONFIG_IXGBE_DCA
7600 if (dca_add_requester(&pdev
->dev
) == 0) {
7601 adapter
->flags
|= IXGBE_FLAG_DCA_ENABLED
;
7602 ixgbe_setup_dca(adapter
);
7605 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
7606 e_info(probe
, "IOV is enabled with %d VFs\n", adapter
->num_vfs
);
7607 for (i
= 0; i
< adapter
->num_vfs
; i
++)
7608 ixgbe_vf_configuration(pdev
, (i
| 0x10000000));
7611 /* firmware requires driver version to be 0xFFFFFFFF
7612 * since os does not support feature
7614 if (hw
->mac
.ops
.set_fw_drv_ver
)
7615 hw
->mac
.ops
.set_fw_drv_ver(hw
, 0xFF, 0xFF, 0xFF,
7618 /* add san mac addr to netdev */
7619 ixgbe_add_sanmac_netdev(netdev
);
7621 e_dev_info("%s\n", ixgbe_default_device_descr
);
7624 #ifdef CONFIG_IXGBE_HWMON
7625 if (ixgbe_sysfs_init(adapter
))
7626 e_err(probe
, "failed to allocate sysfs resources\n");
7627 #endif /* CONFIG_IXGBE_HWMON */
7629 ixgbe_dbg_adapter_init(adapter
);
7631 /* Need link setup for MNG FW, else wait for IXGBE_UP */
7632 if (hw
->mng_fw_enabled
&& hw
->mac
.ops
.setup_link
)
7633 hw
->mac
.ops
.setup_link(hw
,
7634 IXGBE_LINK_SPEED_10GB_FULL
| IXGBE_LINK_SPEED_1GB_FULL
,
7640 ixgbe_release_hw_control(adapter
);
7641 ixgbe_clear_interrupt_scheme(adapter
);
7643 ixgbe_disable_sriov(adapter
);
7644 adapter
->flags2
&= ~IXGBE_FLAG2_SEARCH_FOR_SFP
;
7645 iounmap(hw
->hw_addr
);
7647 free_netdev(netdev
);
7649 pci_release_selected_regions(pdev
,
7650 pci_select_bars(pdev
, IORESOURCE_MEM
));
7653 pci_disable_device(pdev
);
7658 * ixgbe_remove - Device Removal Routine
7659 * @pdev: PCI device information struct
7661 * ixgbe_remove is called by the PCI subsystem to alert the driver
7662 * that it should release a PCI device. The could be caused by a
7663 * Hot-Plug event, or because the driver is going to be removed from
7666 static void ixgbe_remove(struct pci_dev
*pdev
)
7668 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
7669 struct net_device
*netdev
= adapter
->netdev
;
7671 ixgbe_dbg_adapter_exit(adapter
);
7673 set_bit(__IXGBE_DOWN
, &adapter
->state
);
7674 cancel_work_sync(&adapter
->service_task
);
7677 #ifdef CONFIG_IXGBE_DCA
7678 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
) {
7679 adapter
->flags
&= ~IXGBE_FLAG_DCA_ENABLED
;
7680 dca_remove_requester(&pdev
->dev
);
7681 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_DCA_CTRL
, 1);
7685 #ifdef CONFIG_IXGBE_HWMON
7686 ixgbe_sysfs_exit(adapter
);
7687 #endif /* CONFIG_IXGBE_HWMON */
7689 /* remove the added san mac */
7690 ixgbe_del_sanmac_netdev(netdev
);
7692 if (netdev
->reg_state
== NETREG_REGISTERED
)
7693 unregister_netdev(netdev
);
7695 #ifdef CONFIG_PCI_IOV
7697 * Only disable SR-IOV on unload if the user specified the now
7698 * deprecated max_vfs module parameter.
7701 ixgbe_disable_sriov(adapter
);
7703 ixgbe_clear_interrupt_scheme(adapter
);
7705 ixgbe_release_hw_control(adapter
);
7708 kfree(adapter
->ixgbe_ieee_pfc
);
7709 kfree(adapter
->ixgbe_ieee_ets
);
7712 iounmap(adapter
->hw
.hw_addr
);
7713 pci_release_selected_regions(pdev
, pci_select_bars(pdev
,
7716 e_dev_info("complete\n");
7718 free_netdev(netdev
);
7720 pci_disable_pcie_error_reporting(pdev
);
7722 pci_disable_device(pdev
);
7726 * ixgbe_io_error_detected - called when PCI error is detected
7727 * @pdev: Pointer to PCI device
7728 * @state: The current pci connection state
7730 * This function is called after a PCI bus error affecting
7731 * this device has been detected.
7733 static pci_ers_result_t
ixgbe_io_error_detected(struct pci_dev
*pdev
,
7734 pci_channel_state_t state
)
7736 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
7737 struct net_device
*netdev
= adapter
->netdev
;
7739 #ifdef CONFIG_PCI_IOV
7740 struct pci_dev
*bdev
, *vfdev
;
7741 u32 dw0
, dw1
, dw2
, dw3
;
7743 u16 req_id
, pf_func
;
7745 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
||
7746 adapter
->num_vfs
== 0)
7747 goto skip_bad_vf_detection
;
7749 bdev
= pdev
->bus
->self
;
7750 while (bdev
&& (pci_pcie_type(bdev
) != PCI_EXP_TYPE_ROOT_PORT
))
7751 bdev
= bdev
->bus
->self
;
7754 goto skip_bad_vf_detection
;
7756 pos
= pci_find_ext_capability(bdev
, PCI_EXT_CAP_ID_ERR
);
7758 goto skip_bad_vf_detection
;
7760 pci_read_config_dword(bdev
, pos
+ PCI_ERR_HEADER_LOG
, &dw0
);
7761 pci_read_config_dword(bdev
, pos
+ PCI_ERR_HEADER_LOG
+ 4, &dw1
);
7762 pci_read_config_dword(bdev
, pos
+ PCI_ERR_HEADER_LOG
+ 8, &dw2
);
7763 pci_read_config_dword(bdev
, pos
+ PCI_ERR_HEADER_LOG
+ 12, &dw3
);
7766 /* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
7767 if (!(req_id
& 0x0080))
7768 goto skip_bad_vf_detection
;
7770 pf_func
= req_id
& 0x01;
7771 if ((pf_func
& 1) == (pdev
->devfn
& 1)) {
7772 unsigned int device_id
;
7774 vf
= (req_id
& 0x7F) >> 1;
7775 e_dev_err("VF %d has caused a PCIe error\n", vf
);
7776 e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
7777 "%8.8x\tdw3: %8.8x\n",
7778 dw0
, dw1
, dw2
, dw3
);
7779 switch (adapter
->hw
.mac
.type
) {
7780 case ixgbe_mac_82599EB
:
7781 device_id
= IXGBE_82599_VF_DEVICE_ID
;
7783 case ixgbe_mac_X540
:
7784 device_id
= IXGBE_X540_VF_DEVICE_ID
;
7791 /* Find the pci device of the offending VF */
7792 vfdev
= pci_get_device(PCI_VENDOR_ID_INTEL
, device_id
, NULL
);
7794 if (vfdev
->devfn
== (req_id
& 0xFF))
7796 vfdev
= pci_get_device(PCI_VENDOR_ID_INTEL
,
7800 * There's a slim chance the VF could have been hot plugged,
7801 * so if it is no longer present we don't need to issue the
7802 * VFLR. Just clean up the AER in that case.
7805 e_dev_err("Issuing VFLR to VF %d\n", vf
);
7806 pci_write_config_dword(vfdev
, 0xA8, 0x00008000);
7807 /* Free device reference count */
7811 pci_cleanup_aer_uncorrect_error_status(pdev
);
7815 * Even though the error may have occurred on the other port
7816 * we still need to increment the vf error reference count for
7817 * both ports because the I/O resume function will be called
7820 adapter
->vferr_refcount
++;
7822 return PCI_ERS_RESULT_RECOVERED
;
7824 skip_bad_vf_detection
:
7825 #endif /* CONFIG_PCI_IOV */
7826 netif_device_detach(netdev
);
7828 if (state
== pci_channel_io_perm_failure
)
7829 return PCI_ERS_RESULT_DISCONNECT
;
7831 if (netif_running(netdev
))
7832 ixgbe_down(adapter
);
7833 pci_disable_device(pdev
);
7835 /* Request a slot reset. */
7836 return PCI_ERS_RESULT_NEED_RESET
;
7840 * ixgbe_io_slot_reset - called after the pci bus has been reset.
7841 * @pdev: Pointer to PCI device
7843 * Restart the card from scratch, as if from a cold-boot.
7845 static pci_ers_result_t
ixgbe_io_slot_reset(struct pci_dev
*pdev
)
7847 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
7848 pci_ers_result_t result
;
7851 if (pci_enable_device_mem(pdev
)) {
7852 e_err(probe
, "Cannot re-enable PCI device after reset.\n");
7853 result
= PCI_ERS_RESULT_DISCONNECT
;
7855 pci_set_master(pdev
);
7856 pci_restore_state(pdev
);
7857 pci_save_state(pdev
);
7859 pci_wake_from_d3(pdev
, false);
7861 ixgbe_reset(adapter
);
7862 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_WUS
, ~0);
7863 result
= PCI_ERS_RESULT_RECOVERED
;
7866 err
= pci_cleanup_aer_uncorrect_error_status(pdev
);
7868 e_dev_err("pci_cleanup_aer_uncorrect_error_status "
7869 "failed 0x%0x\n", err
);
7870 /* non-fatal, continue */
7877 * ixgbe_io_resume - called when traffic can start flowing again.
7878 * @pdev: Pointer to PCI device
7880 * This callback is called when the error recovery driver tells us that
7881 * its OK to resume normal operation.
7883 static void ixgbe_io_resume(struct pci_dev
*pdev
)
7885 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
7886 struct net_device
*netdev
= adapter
->netdev
;
7888 #ifdef CONFIG_PCI_IOV
7889 if (adapter
->vferr_refcount
) {
7890 e_info(drv
, "Resuming after VF err\n");
7891 adapter
->vferr_refcount
--;
7896 if (netif_running(netdev
))
7899 netif_device_attach(netdev
);
7902 static const struct pci_error_handlers ixgbe_err_handler
= {
7903 .error_detected
= ixgbe_io_error_detected
,
7904 .slot_reset
= ixgbe_io_slot_reset
,
7905 .resume
= ixgbe_io_resume
,
7908 static struct pci_driver ixgbe_driver
= {
7909 .name
= ixgbe_driver_name
,
7910 .id_table
= ixgbe_pci_tbl
,
7911 .probe
= ixgbe_probe
,
7912 .remove
= ixgbe_remove
,
7914 .suspend
= ixgbe_suspend
,
7915 .resume
= ixgbe_resume
,
7917 .shutdown
= ixgbe_shutdown
,
7918 .sriov_configure
= ixgbe_pci_sriov_configure
,
7919 .err_handler
= &ixgbe_err_handler
7923 * ixgbe_init_module - Driver Registration Routine
7925 * ixgbe_init_module is the first routine called when the driver is
7926 * loaded. All it does is register with the PCI subsystem.
7928 static int __init
ixgbe_init_module(void)
7931 pr_info("%s - version %s\n", ixgbe_driver_string
, ixgbe_driver_version
);
7932 pr_info("%s\n", ixgbe_copyright
);
7936 ret
= pci_register_driver(&ixgbe_driver
);
7942 #ifdef CONFIG_IXGBE_DCA
7943 dca_register_notify(&dca_notifier
);
7949 module_init(ixgbe_init_module
);
7952 * ixgbe_exit_module - Driver Exit Cleanup Routine
7954 * ixgbe_exit_module is called just before the driver is removed
7957 static void __exit
ixgbe_exit_module(void)
7959 #ifdef CONFIG_IXGBE_DCA
7960 dca_unregister_notify(&dca_notifier
);
7962 pci_unregister_driver(&ixgbe_driver
);
7966 rcu_barrier(); /* Wait for completion of call_rcu()'s */
7969 #ifdef CONFIG_IXGBE_DCA
7970 static int ixgbe_notify_dca(struct notifier_block
*nb
, unsigned long event
,
7975 ret_val
= driver_for_each_device(&ixgbe_driver
.driver
, NULL
, &event
,
7976 __ixgbe_notify_dca
);
7978 return ret_val
? NOTIFY_BAD
: NOTIFY_DONE
;
7981 #endif /* CONFIG_IXGBE_DCA */
7983 module_exit(ixgbe_exit_module
);