1 /* Intel(R) Gigabit Ethernet Linux driver
2 * Copyright(c) 2007-2014 Intel Corporation.
4 * This program is free software; you can redistribute it and/or modify it
5 * under the terms and conditions of the GNU General Public License,
6 * version 2, as published by the Free Software Foundation.
8 * This program is distributed in the hope it will be useful, but WITHOUT
9 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
13 * You should have received a copy of the GNU General Public License along with
14 * this program; if not, see <http://www.gnu.org/licenses/>.
16 * The full GNU General Public License is included in this distribution in
17 * the file called "COPYING".
19 * Contact Information:
20 * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
21 * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
24 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
26 #include <linux/module.h>
27 #include <linux/types.h>
28 #include <linux/init.h>
29 #include <linux/bitops.h>
30 #include <linux/vmalloc.h>
31 #include <linux/pagemap.h>
32 #include <linux/netdevice.h>
33 #include <linux/ipv6.h>
34 #include <linux/slab.h>
35 #include <net/checksum.h>
36 #include <net/ip6_checksum.h>
37 #include <linux/net_tstamp.h>
38 #include <linux/mii.h>
39 #include <linux/ethtool.h>
41 #include <linux/if_vlan.h>
42 #include <linux/pci.h>
43 #include <linux/pci-aspm.h>
44 #include <linux/delay.h>
45 #include <linux/interrupt.h>
47 #include <linux/tcp.h>
48 #include <linux/sctp.h>
49 #include <linux/if_ether.h>
50 #include <linux/aer.h>
51 #include <linux/prefetch.h>
52 #include <linux/pm_runtime.h>
54 #include <linux/dca.h>
56 #include <linux/i2c.h>
62 #define DRV_VERSION __stringify(MAJ) "." __stringify(MIN) "." \
63 __stringify(BUILD) "-k"
64 char igb_driver_name
[] = "igb";
65 char igb_driver_version
[] = DRV_VERSION
;
66 static const char igb_driver_string
[] =
67 "Intel(R) Gigabit Ethernet Network Driver";
68 static const char igb_copyright
[] =
69 "Copyright (c) 2007-2014 Intel Corporation.";
71 static const struct e1000_info
*igb_info_tbl
[] = {
72 [board_82575
] = &e1000_82575_info
,
75 static const struct pci_device_id igb_pci_tbl
[] = {
76 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I354_BACKPLANE_1GBPS
) },
77 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I354_SGMII
) },
78 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS
) },
79 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I211_COPPER
), board_82575
},
80 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_COPPER
), board_82575
},
81 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_FIBER
), board_82575
},
82 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_SERDES
), board_82575
},
83 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_SGMII
), board_82575
},
84 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_COPPER_FLASHLESS
), board_82575
},
85 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_SERDES_FLASHLESS
), board_82575
},
86 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I350_COPPER
), board_82575
},
87 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I350_FIBER
), board_82575
},
88 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I350_SERDES
), board_82575
},
89 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I350_SGMII
), board_82575
},
90 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_COPPER
), board_82575
},
91 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_FIBER
), board_82575
},
92 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_QUAD_FIBER
), board_82575
},
93 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_SERDES
), board_82575
},
94 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_SGMII
), board_82575
},
95 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_COPPER_DUAL
), board_82575
},
96 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_DH89XXCC_SGMII
), board_82575
},
97 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_DH89XXCC_SERDES
), board_82575
},
98 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_DH89XXCC_BACKPLANE
), board_82575
},
99 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_DH89XXCC_SFP
), board_82575
},
100 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576
), board_82575
},
101 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_NS
), board_82575
},
102 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_NS_SERDES
), board_82575
},
103 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_FIBER
), board_82575
},
104 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_SERDES
), board_82575
},
105 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_SERDES_QUAD
), board_82575
},
106 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_QUAD_COPPER_ET2
), board_82575
},
107 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_QUAD_COPPER
), board_82575
},
108 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82575EB_COPPER
), board_82575
},
109 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82575EB_FIBER_SERDES
), board_82575
},
110 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82575GB_QUAD_COPPER
), board_82575
},
111 /* required last entry */
115 MODULE_DEVICE_TABLE(pci
, igb_pci_tbl
);
117 static int igb_setup_all_tx_resources(struct igb_adapter
*);
118 static int igb_setup_all_rx_resources(struct igb_adapter
*);
119 static void igb_free_all_tx_resources(struct igb_adapter
*);
120 static void igb_free_all_rx_resources(struct igb_adapter
*);
121 static void igb_setup_mrqc(struct igb_adapter
*);
122 static int igb_probe(struct pci_dev
*, const struct pci_device_id
*);
123 static void igb_remove(struct pci_dev
*pdev
);
124 static int igb_sw_init(struct igb_adapter
*);
125 static int igb_open(struct net_device
*);
126 static int igb_close(struct net_device
*);
127 static void igb_configure(struct igb_adapter
*);
128 static void igb_configure_tx(struct igb_adapter
*);
129 static void igb_configure_rx(struct igb_adapter
*);
130 static void igb_clean_all_tx_rings(struct igb_adapter
*);
131 static void igb_clean_all_rx_rings(struct igb_adapter
*);
132 static void igb_clean_tx_ring(struct igb_ring
*);
133 static void igb_clean_rx_ring(struct igb_ring
*);
134 static void igb_set_rx_mode(struct net_device
*);
135 static void igb_update_phy_info(unsigned long);
136 static void igb_watchdog(unsigned long);
137 static void igb_watchdog_task(struct work_struct
*);
138 static netdev_tx_t
igb_xmit_frame(struct sk_buff
*skb
, struct net_device
*);
139 static struct rtnl_link_stats64
*igb_get_stats64(struct net_device
*dev
,
140 struct rtnl_link_stats64
*stats
);
141 static int igb_change_mtu(struct net_device
*, int);
142 static int igb_set_mac(struct net_device
*, void *);
143 static void igb_set_uta(struct igb_adapter
*adapter
);
144 static irqreturn_t
igb_intr(int irq
, void *);
145 static irqreturn_t
igb_intr_msi(int irq
, void *);
146 static irqreturn_t
igb_msix_other(int irq
, void *);
147 static irqreturn_t
igb_msix_ring(int irq
, void *);
148 #ifdef CONFIG_IGB_DCA
149 static void igb_update_dca(struct igb_q_vector
*);
150 static void igb_setup_dca(struct igb_adapter
*);
151 #endif /* CONFIG_IGB_DCA */
152 static int igb_poll(struct napi_struct
*, int);
153 static bool igb_clean_tx_irq(struct igb_q_vector
*);
154 static bool igb_clean_rx_irq(struct igb_q_vector
*, int);
155 static int igb_ioctl(struct net_device
*, struct ifreq
*, int cmd
);
156 static void igb_tx_timeout(struct net_device
*);
157 static void igb_reset_task(struct work_struct
*);
158 static void igb_vlan_mode(struct net_device
*netdev
,
159 netdev_features_t features
);
160 static int igb_vlan_rx_add_vid(struct net_device
*, __be16
, u16
);
161 static int igb_vlan_rx_kill_vid(struct net_device
*, __be16
, u16
);
162 static void igb_restore_vlan(struct igb_adapter
*);
163 static void igb_rar_set_qsel(struct igb_adapter
*, u8
*, u32
, u8
);
164 static void igb_ping_all_vfs(struct igb_adapter
*);
165 static void igb_msg_task(struct igb_adapter
*);
166 static void igb_vmm_control(struct igb_adapter
*);
167 static int igb_set_vf_mac(struct igb_adapter
*, int, unsigned char *);
168 static void igb_restore_vf_multicasts(struct igb_adapter
*adapter
);
169 static int igb_ndo_set_vf_mac(struct net_device
*netdev
, int vf
, u8
*mac
);
170 static int igb_ndo_set_vf_vlan(struct net_device
*netdev
,
171 int vf
, u16 vlan
, u8 qos
);
172 static int igb_ndo_set_vf_bw(struct net_device
*, int, int, int);
173 static int igb_ndo_set_vf_spoofchk(struct net_device
*netdev
, int vf
,
175 static int igb_ndo_get_vf_config(struct net_device
*netdev
, int vf
,
176 struct ifla_vf_info
*ivi
);
177 static void igb_check_vf_rate_limit(struct igb_adapter
*);
179 #ifdef CONFIG_PCI_IOV
180 static int igb_vf_configure(struct igb_adapter
*adapter
, int vf
);
181 static int igb_pci_enable_sriov(struct pci_dev
*dev
, int num_vfs
);
185 #ifdef CONFIG_PM_SLEEP
186 static int igb_suspend(struct device
*);
188 static int igb_resume(struct device
*);
189 static int igb_runtime_suspend(struct device
*dev
);
190 static int igb_runtime_resume(struct device
*dev
);
191 static int igb_runtime_idle(struct device
*dev
);
192 static const struct dev_pm_ops igb_pm_ops
= {
193 SET_SYSTEM_SLEEP_PM_OPS(igb_suspend
, igb_resume
)
194 SET_RUNTIME_PM_OPS(igb_runtime_suspend
, igb_runtime_resume
,
198 static void igb_shutdown(struct pci_dev
*);
199 static int igb_pci_sriov_configure(struct pci_dev
*dev
, int num_vfs
);
200 #ifdef CONFIG_IGB_DCA
201 static int igb_notify_dca(struct notifier_block
*, unsigned long, void *);
202 static struct notifier_block dca_notifier
= {
203 .notifier_call
= igb_notify_dca
,
208 #ifdef CONFIG_NET_POLL_CONTROLLER
209 /* for netdump / net console */
210 static void igb_netpoll(struct net_device
*);
212 #ifdef CONFIG_PCI_IOV
213 static unsigned int max_vfs
;
214 module_param(max_vfs
, uint
, 0);
215 MODULE_PARM_DESC(max_vfs
, "Maximum number of virtual functions to allocate per physical function");
216 #endif /* CONFIG_PCI_IOV */
218 static pci_ers_result_t
igb_io_error_detected(struct pci_dev
*,
219 pci_channel_state_t
);
220 static pci_ers_result_t
igb_io_slot_reset(struct pci_dev
*);
221 static void igb_io_resume(struct pci_dev
*);
223 static const struct pci_error_handlers igb_err_handler
= {
224 .error_detected
= igb_io_error_detected
,
225 .slot_reset
= igb_io_slot_reset
,
226 .resume
= igb_io_resume
,
229 static void igb_init_dmac(struct igb_adapter
*adapter
, u32 pba
);
231 static struct pci_driver igb_driver
= {
232 .name
= igb_driver_name
,
233 .id_table
= igb_pci_tbl
,
235 .remove
= igb_remove
,
237 .driver
.pm
= &igb_pm_ops
,
239 .shutdown
= igb_shutdown
,
240 .sriov_configure
= igb_pci_sriov_configure
,
241 .err_handler
= &igb_err_handler
244 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
245 MODULE_DESCRIPTION("Intel(R) Gigabit Ethernet Network Driver");
246 MODULE_LICENSE("GPL");
247 MODULE_VERSION(DRV_VERSION
);
249 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
250 static int debug
= -1;
251 module_param(debug
, int, 0);
252 MODULE_PARM_DESC(debug
, "Debug level (0=none,...,16=all)");
254 struct igb_reg_info
{
259 static const struct igb_reg_info igb_reg_info_tbl
[] = {
261 /* General Registers */
262 {E1000_CTRL
, "CTRL"},
263 {E1000_STATUS
, "STATUS"},
264 {E1000_CTRL_EXT
, "CTRL_EXT"},
266 /* Interrupt Registers */
270 {E1000_RCTL
, "RCTL"},
271 {E1000_RDLEN(0), "RDLEN"},
272 {E1000_RDH(0), "RDH"},
273 {E1000_RDT(0), "RDT"},
274 {E1000_RXDCTL(0), "RXDCTL"},
275 {E1000_RDBAL(0), "RDBAL"},
276 {E1000_RDBAH(0), "RDBAH"},
279 {E1000_TCTL
, "TCTL"},
280 {E1000_TDBAL(0), "TDBAL"},
281 {E1000_TDBAH(0), "TDBAH"},
282 {E1000_TDLEN(0), "TDLEN"},
283 {E1000_TDH(0), "TDH"},
284 {E1000_TDT(0), "TDT"},
285 {E1000_TXDCTL(0), "TXDCTL"},
286 {E1000_TDFH
, "TDFH"},
287 {E1000_TDFT
, "TDFT"},
288 {E1000_TDFHS
, "TDFHS"},
289 {E1000_TDFPC
, "TDFPC"},
291 /* List Terminator */
295 /* igb_regdump - register printout routine */
296 static void igb_regdump(struct e1000_hw
*hw
, struct igb_reg_info
*reginfo
)
302 switch (reginfo
->ofs
) {
304 for (n
= 0; n
< 4; n
++)
305 regs
[n
] = rd32(E1000_RDLEN(n
));
308 for (n
= 0; n
< 4; n
++)
309 regs
[n
] = rd32(E1000_RDH(n
));
312 for (n
= 0; n
< 4; n
++)
313 regs
[n
] = rd32(E1000_RDT(n
));
315 case E1000_RXDCTL(0):
316 for (n
= 0; n
< 4; n
++)
317 regs
[n
] = rd32(E1000_RXDCTL(n
));
320 for (n
= 0; n
< 4; n
++)
321 regs
[n
] = rd32(E1000_RDBAL(n
));
324 for (n
= 0; n
< 4; n
++)
325 regs
[n
] = rd32(E1000_RDBAH(n
));
328 for (n
= 0; n
< 4; n
++)
329 regs
[n
] = rd32(E1000_RDBAL(n
));
332 for (n
= 0; n
< 4; n
++)
333 regs
[n
] = rd32(E1000_TDBAH(n
));
336 for (n
= 0; n
< 4; n
++)
337 regs
[n
] = rd32(E1000_TDLEN(n
));
340 for (n
= 0; n
< 4; n
++)
341 regs
[n
] = rd32(E1000_TDH(n
));
344 for (n
= 0; n
< 4; n
++)
345 regs
[n
] = rd32(E1000_TDT(n
));
347 case E1000_TXDCTL(0):
348 for (n
= 0; n
< 4; n
++)
349 regs
[n
] = rd32(E1000_TXDCTL(n
));
352 pr_info("%-15s %08x\n", reginfo
->name
, rd32(reginfo
->ofs
));
356 snprintf(rname
, 16, "%s%s", reginfo
->name
, "[0-3]");
357 pr_info("%-15s %08x %08x %08x %08x\n", rname
, regs
[0], regs
[1],
361 /* igb_dump - Print registers, Tx-rings and Rx-rings */
362 static void igb_dump(struct igb_adapter
*adapter
)
364 struct net_device
*netdev
= adapter
->netdev
;
365 struct e1000_hw
*hw
= &adapter
->hw
;
366 struct igb_reg_info
*reginfo
;
367 struct igb_ring
*tx_ring
;
368 union e1000_adv_tx_desc
*tx_desc
;
369 struct my_u0
{ u64 a
; u64 b
; } *u0
;
370 struct igb_ring
*rx_ring
;
371 union e1000_adv_rx_desc
*rx_desc
;
375 if (!netif_msg_hw(adapter
))
378 /* Print netdevice Info */
380 dev_info(&adapter
->pdev
->dev
, "Net device Info\n");
381 pr_info("Device Name state trans_start last_rx\n");
382 pr_info("%-15s %016lX %016lX %016lX\n", netdev
->name
,
383 netdev
->state
, netdev
->trans_start
, netdev
->last_rx
);
386 /* Print Registers */
387 dev_info(&adapter
->pdev
->dev
, "Register Dump\n");
388 pr_info(" Register Name Value\n");
389 for (reginfo
= (struct igb_reg_info
*)igb_reg_info_tbl
;
390 reginfo
->name
; reginfo
++) {
391 igb_regdump(hw
, reginfo
);
394 /* Print TX Ring Summary */
395 if (!netdev
|| !netif_running(netdev
))
398 dev_info(&adapter
->pdev
->dev
, "TX Rings Summary\n");
399 pr_info("Queue [NTU] [NTC] [bi(ntc)->dma ] leng ntw timestamp\n");
400 for (n
= 0; n
< adapter
->num_tx_queues
; n
++) {
401 struct igb_tx_buffer
*buffer_info
;
402 tx_ring
= adapter
->tx_ring
[n
];
403 buffer_info
= &tx_ring
->tx_buffer_info
[tx_ring
->next_to_clean
];
404 pr_info(" %5d %5X %5X %016llX %04X %p %016llX\n",
405 n
, tx_ring
->next_to_use
, tx_ring
->next_to_clean
,
406 (u64
)dma_unmap_addr(buffer_info
, dma
),
407 dma_unmap_len(buffer_info
, len
),
408 buffer_info
->next_to_watch
,
409 (u64
)buffer_info
->time_stamp
);
413 if (!netif_msg_tx_done(adapter
))
414 goto rx_ring_summary
;
416 dev_info(&adapter
->pdev
->dev
, "TX Rings Dump\n");
418 /* Transmit Descriptor Formats
420 * Advanced Transmit Descriptor
421 * +--------------------------------------------------------------+
422 * 0 | Buffer Address [63:0] |
423 * +--------------------------------------------------------------+
424 * 8 | PAYLEN | PORTS |CC|IDX | STA | DCMD |DTYP|MAC|RSV| DTALEN |
425 * +--------------------------------------------------------------+
426 * 63 46 45 40 39 38 36 35 32 31 24 15 0
429 for (n
= 0; n
< adapter
->num_tx_queues
; n
++) {
430 tx_ring
= adapter
->tx_ring
[n
];
431 pr_info("------------------------------------\n");
432 pr_info("TX QUEUE INDEX = %d\n", tx_ring
->queue_index
);
433 pr_info("------------------------------------\n");
434 pr_info("T [desc] [address 63:0 ] [PlPOCIStDDM Ln] [bi->dma ] leng ntw timestamp bi->skb\n");
436 for (i
= 0; tx_ring
->desc
&& (i
< tx_ring
->count
); i
++) {
437 const char *next_desc
;
438 struct igb_tx_buffer
*buffer_info
;
439 tx_desc
= IGB_TX_DESC(tx_ring
, i
);
440 buffer_info
= &tx_ring
->tx_buffer_info
[i
];
441 u0
= (struct my_u0
*)tx_desc
;
442 if (i
== tx_ring
->next_to_use
&&
443 i
== tx_ring
->next_to_clean
)
444 next_desc
= " NTC/U";
445 else if (i
== tx_ring
->next_to_use
)
447 else if (i
== tx_ring
->next_to_clean
)
452 pr_info("T [0x%03X] %016llX %016llX %016llX %04X %p %016llX %p%s\n",
453 i
, le64_to_cpu(u0
->a
),
455 (u64
)dma_unmap_addr(buffer_info
, dma
),
456 dma_unmap_len(buffer_info
, len
),
457 buffer_info
->next_to_watch
,
458 (u64
)buffer_info
->time_stamp
,
459 buffer_info
->skb
, next_desc
);
461 if (netif_msg_pktdata(adapter
) && buffer_info
->skb
)
462 print_hex_dump(KERN_INFO
, "",
464 16, 1, buffer_info
->skb
->data
,
465 dma_unmap_len(buffer_info
, len
),
470 /* Print RX Rings Summary */
472 dev_info(&adapter
->pdev
->dev
, "RX Rings Summary\n");
473 pr_info("Queue [NTU] [NTC]\n");
474 for (n
= 0; n
< adapter
->num_rx_queues
; n
++) {
475 rx_ring
= adapter
->rx_ring
[n
];
476 pr_info(" %5d %5X %5X\n",
477 n
, rx_ring
->next_to_use
, rx_ring
->next_to_clean
);
481 if (!netif_msg_rx_status(adapter
))
484 dev_info(&adapter
->pdev
->dev
, "RX Rings Dump\n");
486 /* Advanced Receive Descriptor (Read) Format
488 * +-----------------------------------------------------+
489 * 0 | Packet Buffer Address [63:1] |A0/NSE|
490 * +----------------------------------------------+------+
491 * 8 | Header Buffer Address [63:1] | DD |
492 * +-----------------------------------------------------+
495 * Advanced Receive Descriptor (Write-Back) Format
497 * 63 48 47 32 31 30 21 20 17 16 4 3 0
498 * +------------------------------------------------------+
499 * 0 | Packet IP |SPH| HDR_LEN | RSV|Packet| RSS |
500 * | Checksum Ident | | | | Type | Type |
501 * +------------------------------------------------------+
502 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
503 * +------------------------------------------------------+
504 * 63 48 47 32 31 20 19 0
507 for (n
= 0; n
< adapter
->num_rx_queues
; n
++) {
508 rx_ring
= adapter
->rx_ring
[n
];
509 pr_info("------------------------------------\n");
510 pr_info("RX QUEUE INDEX = %d\n", rx_ring
->queue_index
);
511 pr_info("------------------------------------\n");
512 pr_info("R [desc] [ PktBuf A0] [ HeadBuf DD] [bi->dma ] [bi->skb] <-- Adv Rx Read format\n");
513 pr_info("RWB[desc] [PcsmIpSHl PtRs] [vl er S cks ln] ---------------- [bi->skb] <-- Adv Rx Write-Back format\n");
515 for (i
= 0; i
< rx_ring
->count
; i
++) {
516 const char *next_desc
;
517 struct igb_rx_buffer
*buffer_info
;
518 buffer_info
= &rx_ring
->rx_buffer_info
[i
];
519 rx_desc
= IGB_RX_DESC(rx_ring
, i
);
520 u0
= (struct my_u0
*)rx_desc
;
521 staterr
= le32_to_cpu(rx_desc
->wb
.upper
.status_error
);
523 if (i
== rx_ring
->next_to_use
)
525 else if (i
== rx_ring
->next_to_clean
)
530 if (staterr
& E1000_RXD_STAT_DD
) {
531 /* Descriptor Done */
532 pr_info("%s[0x%03X] %016llX %016llX ---------------- %s\n",
538 pr_info("%s[0x%03X] %016llX %016llX %016llX %s\n",
542 (u64
)buffer_info
->dma
,
545 if (netif_msg_pktdata(adapter
) &&
546 buffer_info
->dma
&& buffer_info
->page
) {
547 print_hex_dump(KERN_INFO
, "",
550 page_address(buffer_info
->page
) +
551 buffer_info
->page_offset
,
563 * igb_get_i2c_data - Reads the I2C SDA data bit
564 * @hw: pointer to hardware structure
565 * @i2cctl: Current value of I2CCTL register
567 * Returns the I2C data bit value
569 static int igb_get_i2c_data(void *data
)
571 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
572 struct e1000_hw
*hw
= &adapter
->hw
;
573 s32 i2cctl
= rd32(E1000_I2CPARAMS
);
575 return !!(i2cctl
& E1000_I2C_DATA_IN
);
579 * igb_set_i2c_data - Sets the I2C data bit
580 * @data: pointer to hardware structure
581 * @state: I2C data value (0 or 1) to set
583 * Sets the I2C data bit
585 static void igb_set_i2c_data(void *data
, int state
)
587 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
588 struct e1000_hw
*hw
= &adapter
->hw
;
589 s32 i2cctl
= rd32(E1000_I2CPARAMS
);
592 i2cctl
|= E1000_I2C_DATA_OUT
;
594 i2cctl
&= ~E1000_I2C_DATA_OUT
;
596 i2cctl
&= ~E1000_I2C_DATA_OE_N
;
597 i2cctl
|= E1000_I2C_CLK_OE_N
;
598 wr32(E1000_I2CPARAMS
, i2cctl
);
604 * igb_set_i2c_clk - Sets the I2C SCL clock
605 * @data: pointer to hardware structure
606 * @state: state to set clock
608 * Sets the I2C clock line to state
610 static void igb_set_i2c_clk(void *data
, int state
)
612 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
613 struct e1000_hw
*hw
= &adapter
->hw
;
614 s32 i2cctl
= rd32(E1000_I2CPARAMS
);
617 i2cctl
|= E1000_I2C_CLK_OUT
;
618 i2cctl
&= ~E1000_I2C_CLK_OE_N
;
620 i2cctl
&= ~E1000_I2C_CLK_OUT
;
621 i2cctl
&= ~E1000_I2C_CLK_OE_N
;
623 wr32(E1000_I2CPARAMS
, i2cctl
);
628 * igb_get_i2c_clk - Gets the I2C SCL clock state
629 * @data: pointer to hardware structure
631 * Gets the I2C clock state
633 static int igb_get_i2c_clk(void *data
)
635 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
636 struct e1000_hw
*hw
= &adapter
->hw
;
637 s32 i2cctl
= rd32(E1000_I2CPARAMS
);
639 return !!(i2cctl
& E1000_I2C_CLK_IN
);
642 static const struct i2c_algo_bit_data igb_i2c_algo
= {
643 .setsda
= igb_set_i2c_data
,
644 .setscl
= igb_set_i2c_clk
,
645 .getsda
= igb_get_i2c_data
,
646 .getscl
= igb_get_i2c_clk
,
652 * igb_get_hw_dev - return device
653 * @hw: pointer to hardware structure
655 * used by hardware layer to print debugging information
657 struct net_device
*igb_get_hw_dev(struct e1000_hw
*hw
)
659 struct igb_adapter
*adapter
= hw
->back
;
660 return adapter
->netdev
;
664 * igb_init_module - Driver Registration Routine
666 * igb_init_module is the first routine called when the driver is
667 * loaded. All it does is register with the PCI subsystem.
669 static int __init
igb_init_module(void)
673 pr_info("%s - version %s\n",
674 igb_driver_string
, igb_driver_version
);
675 pr_info("%s\n", igb_copyright
);
677 #ifdef CONFIG_IGB_DCA
678 dca_register_notify(&dca_notifier
);
680 ret
= pci_register_driver(&igb_driver
);
684 module_init(igb_init_module
);
687 * igb_exit_module - Driver Exit Cleanup Routine
689 * igb_exit_module is called just before the driver is removed
692 static void __exit
igb_exit_module(void)
694 #ifdef CONFIG_IGB_DCA
695 dca_unregister_notify(&dca_notifier
);
697 pci_unregister_driver(&igb_driver
);
700 module_exit(igb_exit_module
);
702 #define Q_IDX_82576(i) (((i & 0x1) << 3) + (i >> 1))
704 * igb_cache_ring_register - Descriptor ring to register mapping
705 * @adapter: board private structure to initialize
707 * Once we know the feature-set enabled for the device, we'll cache
708 * the register offset the descriptor ring is assigned to.
710 static void igb_cache_ring_register(struct igb_adapter
*adapter
)
713 u32 rbase_offset
= adapter
->vfs_allocated_count
;
715 switch (adapter
->hw
.mac
.type
) {
717 /* The queues are allocated for virtualization such that VF 0
718 * is allocated queues 0 and 8, VF 1 queues 1 and 9, etc.
719 * In order to avoid collision we start at the first free queue
720 * and continue consuming queues in the same sequence
722 if (adapter
->vfs_allocated_count
) {
723 for (; i
< adapter
->rss_queues
; i
++)
724 adapter
->rx_ring
[i
]->reg_idx
= rbase_offset
+
736 for (; i
< adapter
->num_rx_queues
; i
++)
737 adapter
->rx_ring
[i
]->reg_idx
= rbase_offset
+ i
;
738 for (; j
< adapter
->num_tx_queues
; j
++)
739 adapter
->tx_ring
[j
]->reg_idx
= rbase_offset
+ j
;
744 u32
igb_rd32(struct e1000_hw
*hw
, u32 reg
)
746 struct igb_adapter
*igb
= container_of(hw
, struct igb_adapter
, hw
);
747 u8 __iomem
*hw_addr
= ACCESS_ONCE(hw
->hw_addr
);
750 if (E1000_REMOVED(hw_addr
))
753 value
= readl(&hw_addr
[reg
]);
755 /* reads should not return all F's */
756 if (!(~value
) && (!reg
|| !(~readl(hw_addr
)))) {
757 struct net_device
*netdev
= igb
->netdev
;
759 netif_device_detach(netdev
);
760 netdev_err(netdev
, "PCIe link lost, device now detached\n");
767 * igb_write_ivar - configure ivar for given MSI-X vector
768 * @hw: pointer to the HW structure
769 * @msix_vector: vector number we are allocating to a given ring
770 * @index: row index of IVAR register to write within IVAR table
771 * @offset: column offset of in IVAR, should be multiple of 8
773 * This function is intended to handle the writing of the IVAR register
774 * for adapters 82576 and newer. The IVAR table consists of 2 columns,
775 * each containing an cause allocation for an Rx and Tx ring, and a
776 * variable number of rows depending on the number of queues supported.
778 static void igb_write_ivar(struct e1000_hw
*hw
, int msix_vector
,
779 int index
, int offset
)
781 u32 ivar
= array_rd32(E1000_IVAR0
, index
);
783 /* clear any bits that are currently set */
784 ivar
&= ~((u32
)0xFF << offset
);
786 /* write vector and valid bit */
787 ivar
|= (msix_vector
| E1000_IVAR_VALID
) << offset
;
789 array_wr32(E1000_IVAR0
, index
, ivar
);
792 #define IGB_N0_QUEUE -1
793 static void igb_assign_vector(struct igb_q_vector
*q_vector
, int msix_vector
)
795 struct igb_adapter
*adapter
= q_vector
->adapter
;
796 struct e1000_hw
*hw
= &adapter
->hw
;
797 int rx_queue
= IGB_N0_QUEUE
;
798 int tx_queue
= IGB_N0_QUEUE
;
801 if (q_vector
->rx
.ring
)
802 rx_queue
= q_vector
->rx
.ring
->reg_idx
;
803 if (q_vector
->tx
.ring
)
804 tx_queue
= q_vector
->tx
.ring
->reg_idx
;
806 switch (hw
->mac
.type
) {
808 /* The 82575 assigns vectors using a bitmask, which matches the
809 * bitmask for the EICR/EIMS/EIMC registers. To assign one
810 * or more queues to a vector, we write the appropriate bits
811 * into the MSIXBM register for that vector.
813 if (rx_queue
> IGB_N0_QUEUE
)
814 msixbm
= E1000_EICR_RX_QUEUE0
<< rx_queue
;
815 if (tx_queue
> IGB_N0_QUEUE
)
816 msixbm
|= E1000_EICR_TX_QUEUE0
<< tx_queue
;
817 if (!(adapter
->flags
& IGB_FLAG_HAS_MSIX
) && msix_vector
== 0)
818 msixbm
|= E1000_EIMS_OTHER
;
819 array_wr32(E1000_MSIXBM(0), msix_vector
, msixbm
);
820 q_vector
->eims_value
= msixbm
;
823 /* 82576 uses a table that essentially consists of 2 columns
824 * with 8 rows. The ordering is column-major so we use the
825 * lower 3 bits as the row index, and the 4th bit as the
828 if (rx_queue
> IGB_N0_QUEUE
)
829 igb_write_ivar(hw
, msix_vector
,
831 (rx_queue
& 0x8) << 1);
832 if (tx_queue
> IGB_N0_QUEUE
)
833 igb_write_ivar(hw
, msix_vector
,
835 ((tx_queue
& 0x8) << 1) + 8);
836 q_vector
->eims_value
= 1 << msix_vector
;
843 /* On 82580 and newer adapters the scheme is similar to 82576
844 * however instead of ordering column-major we have things
845 * ordered row-major. So we traverse the table by using
846 * bit 0 as the column offset, and the remaining bits as the
849 if (rx_queue
> IGB_N0_QUEUE
)
850 igb_write_ivar(hw
, msix_vector
,
852 (rx_queue
& 0x1) << 4);
853 if (tx_queue
> IGB_N0_QUEUE
)
854 igb_write_ivar(hw
, msix_vector
,
856 ((tx_queue
& 0x1) << 4) + 8);
857 q_vector
->eims_value
= 1 << msix_vector
;
864 /* add q_vector eims value to global eims_enable_mask */
865 adapter
->eims_enable_mask
|= q_vector
->eims_value
;
867 /* configure q_vector to set itr on first interrupt */
868 q_vector
->set_itr
= 1;
872 * igb_configure_msix - Configure MSI-X hardware
873 * @adapter: board private structure to initialize
875 * igb_configure_msix sets up the hardware to properly
876 * generate MSI-X interrupts.
878 static void igb_configure_msix(struct igb_adapter
*adapter
)
882 struct e1000_hw
*hw
= &adapter
->hw
;
884 adapter
->eims_enable_mask
= 0;
886 /* set vector for other causes, i.e. link changes */
887 switch (hw
->mac
.type
) {
889 tmp
= rd32(E1000_CTRL_EXT
);
890 /* enable MSI-X PBA support*/
891 tmp
|= E1000_CTRL_EXT_PBA_CLR
;
893 /* Auto-Mask interrupts upon ICR read. */
894 tmp
|= E1000_CTRL_EXT_EIAME
;
895 tmp
|= E1000_CTRL_EXT_IRCA
;
897 wr32(E1000_CTRL_EXT
, tmp
);
899 /* enable msix_other interrupt */
900 array_wr32(E1000_MSIXBM(0), vector
++, E1000_EIMS_OTHER
);
901 adapter
->eims_other
= E1000_EIMS_OTHER
;
911 /* Turn on MSI-X capability first, or our settings
912 * won't stick. And it will take days to debug.
914 wr32(E1000_GPIE
, E1000_GPIE_MSIX_MODE
|
915 E1000_GPIE_PBA
| E1000_GPIE_EIAME
|
918 /* enable msix_other interrupt */
919 adapter
->eims_other
= 1 << vector
;
920 tmp
= (vector
++ | E1000_IVAR_VALID
) << 8;
922 wr32(E1000_IVAR_MISC
, tmp
);
925 /* do nothing, since nothing else supports MSI-X */
927 } /* switch (hw->mac.type) */
929 adapter
->eims_enable_mask
|= adapter
->eims_other
;
931 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
932 igb_assign_vector(adapter
->q_vector
[i
], vector
++);
938 * igb_request_msix - Initialize MSI-X interrupts
939 * @adapter: board private structure to initialize
941 * igb_request_msix allocates MSI-X vectors and requests interrupts from the
944 static int igb_request_msix(struct igb_adapter
*adapter
)
946 struct net_device
*netdev
= adapter
->netdev
;
947 struct e1000_hw
*hw
= &adapter
->hw
;
948 int i
, err
= 0, vector
= 0, free_vector
= 0;
950 err
= request_irq(adapter
->msix_entries
[vector
].vector
,
951 igb_msix_other
, 0, netdev
->name
, adapter
);
955 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
956 struct igb_q_vector
*q_vector
= adapter
->q_vector
[i
];
960 q_vector
->itr_register
= hw
->hw_addr
+ E1000_EITR(vector
);
962 if (q_vector
->rx
.ring
&& q_vector
->tx
.ring
)
963 sprintf(q_vector
->name
, "%s-TxRx-%u", netdev
->name
,
964 q_vector
->rx
.ring
->queue_index
);
965 else if (q_vector
->tx
.ring
)
966 sprintf(q_vector
->name
, "%s-tx-%u", netdev
->name
,
967 q_vector
->tx
.ring
->queue_index
);
968 else if (q_vector
->rx
.ring
)
969 sprintf(q_vector
->name
, "%s-rx-%u", netdev
->name
,
970 q_vector
->rx
.ring
->queue_index
);
972 sprintf(q_vector
->name
, "%s-unused", netdev
->name
);
974 err
= request_irq(adapter
->msix_entries
[vector
].vector
,
975 igb_msix_ring
, 0, q_vector
->name
,
981 igb_configure_msix(adapter
);
985 /* free already assigned IRQs */
986 free_irq(adapter
->msix_entries
[free_vector
++].vector
, adapter
);
989 for (i
= 0; i
< vector
; i
++) {
990 free_irq(adapter
->msix_entries
[free_vector
++].vector
,
991 adapter
->q_vector
[i
]);
998 * igb_free_q_vector - Free memory allocated for specific interrupt vector
999 * @adapter: board private structure to initialize
1000 * @v_idx: Index of vector to be freed
1002 * This function frees the memory allocated to the q_vector.
1004 static void igb_free_q_vector(struct igb_adapter
*adapter
, int v_idx
)
1006 struct igb_q_vector
*q_vector
= adapter
->q_vector
[v_idx
];
1008 adapter
->q_vector
[v_idx
] = NULL
;
1010 /* igb_get_stats64() might access the rings on this vector,
1011 * we must wait a grace period before freeing it.
1014 kfree_rcu(q_vector
, rcu
);
1018 * igb_reset_q_vector - Reset config for interrupt vector
1019 * @adapter: board private structure to initialize
1020 * @v_idx: Index of vector to be reset
1022 * If NAPI is enabled it will delete any references to the
1023 * NAPI struct. This is preparation for igb_free_q_vector.
1025 static void igb_reset_q_vector(struct igb_adapter
*adapter
, int v_idx
)
1027 struct igb_q_vector
*q_vector
= adapter
->q_vector
[v_idx
];
1029 /* Coming from igb_set_interrupt_capability, the vectors are not yet
1030 * allocated. So, q_vector is NULL so we should stop here.
1035 if (q_vector
->tx
.ring
)
1036 adapter
->tx_ring
[q_vector
->tx
.ring
->queue_index
] = NULL
;
1038 if (q_vector
->rx
.ring
)
1039 adapter
->tx_ring
[q_vector
->rx
.ring
->queue_index
] = NULL
;
1041 netif_napi_del(&q_vector
->napi
);
1045 static void igb_reset_interrupt_capability(struct igb_adapter
*adapter
)
1047 int v_idx
= adapter
->num_q_vectors
;
1049 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
)
1050 pci_disable_msix(adapter
->pdev
);
1051 else if (adapter
->flags
& IGB_FLAG_HAS_MSI
)
1052 pci_disable_msi(adapter
->pdev
);
1055 igb_reset_q_vector(adapter
, v_idx
);
1059 * igb_free_q_vectors - Free memory allocated for interrupt vectors
1060 * @adapter: board private structure to initialize
1062 * This function frees the memory allocated to the q_vectors. In addition if
1063 * NAPI is enabled it will delete any references to the NAPI struct prior
1064 * to freeing the q_vector.
1066 static void igb_free_q_vectors(struct igb_adapter
*adapter
)
1068 int v_idx
= adapter
->num_q_vectors
;
1070 adapter
->num_tx_queues
= 0;
1071 adapter
->num_rx_queues
= 0;
1072 adapter
->num_q_vectors
= 0;
1075 igb_reset_q_vector(adapter
, v_idx
);
1076 igb_free_q_vector(adapter
, v_idx
);
1081 * igb_clear_interrupt_scheme - reset the device to a state of no interrupts
1082 * @adapter: board private structure to initialize
1084 * This function resets the device so that it has 0 Rx queues, Tx queues, and
1085 * MSI-X interrupts allocated.
1087 static void igb_clear_interrupt_scheme(struct igb_adapter
*adapter
)
1089 igb_free_q_vectors(adapter
);
1090 igb_reset_interrupt_capability(adapter
);
1094 * igb_set_interrupt_capability - set MSI or MSI-X if supported
1095 * @adapter: board private structure to initialize
1096 * @msix: boolean value of MSIX capability
1098 * Attempt to configure interrupts using the best available
1099 * capabilities of the hardware and kernel.
1101 static void igb_set_interrupt_capability(struct igb_adapter
*adapter
, bool msix
)
1108 adapter
->flags
|= IGB_FLAG_HAS_MSIX
;
1110 /* Number of supported queues. */
1111 adapter
->num_rx_queues
= adapter
->rss_queues
;
1112 if (adapter
->vfs_allocated_count
)
1113 adapter
->num_tx_queues
= 1;
1115 adapter
->num_tx_queues
= adapter
->rss_queues
;
1117 /* start with one vector for every Rx queue */
1118 numvecs
= adapter
->num_rx_queues
;
1120 /* if Tx handler is separate add 1 for every Tx queue */
1121 if (!(adapter
->flags
& IGB_FLAG_QUEUE_PAIRS
))
1122 numvecs
+= adapter
->num_tx_queues
;
1124 /* store the number of vectors reserved for queues */
1125 adapter
->num_q_vectors
= numvecs
;
1127 /* add 1 vector for link status interrupts */
1129 for (i
= 0; i
< numvecs
; i
++)
1130 adapter
->msix_entries
[i
].entry
= i
;
1132 err
= pci_enable_msix_range(adapter
->pdev
,
1133 adapter
->msix_entries
,
1139 igb_reset_interrupt_capability(adapter
);
1141 /* If we can't do MSI-X, try MSI */
1143 adapter
->flags
&= ~IGB_FLAG_HAS_MSIX
;
1144 #ifdef CONFIG_PCI_IOV
1145 /* disable SR-IOV for non MSI-X configurations */
1146 if (adapter
->vf_data
) {
1147 struct e1000_hw
*hw
= &adapter
->hw
;
1148 /* disable iov and allow time for transactions to clear */
1149 pci_disable_sriov(adapter
->pdev
);
1152 kfree(adapter
->vf_data
);
1153 adapter
->vf_data
= NULL
;
1154 wr32(E1000_IOVCTL
, E1000_IOVCTL_REUSE_VFQ
);
1157 dev_info(&adapter
->pdev
->dev
, "IOV Disabled\n");
1160 adapter
->vfs_allocated_count
= 0;
1161 adapter
->rss_queues
= 1;
1162 adapter
->flags
|= IGB_FLAG_QUEUE_PAIRS
;
1163 adapter
->num_rx_queues
= 1;
1164 adapter
->num_tx_queues
= 1;
1165 adapter
->num_q_vectors
= 1;
1166 if (!pci_enable_msi(adapter
->pdev
))
1167 adapter
->flags
|= IGB_FLAG_HAS_MSI
;
1170 static void igb_add_ring(struct igb_ring
*ring
,
1171 struct igb_ring_container
*head
)
1178 * igb_alloc_q_vector - Allocate memory for a single interrupt vector
1179 * @adapter: board private structure to initialize
1180 * @v_count: q_vectors allocated on adapter, used for ring interleaving
1181 * @v_idx: index of vector in adapter struct
1182 * @txr_count: total number of Tx rings to allocate
1183 * @txr_idx: index of first Tx ring to allocate
1184 * @rxr_count: total number of Rx rings to allocate
1185 * @rxr_idx: index of first Rx ring to allocate
1187 * We allocate one q_vector. If allocation fails we return -ENOMEM.
1189 static int igb_alloc_q_vector(struct igb_adapter
*adapter
,
1190 int v_count
, int v_idx
,
1191 int txr_count
, int txr_idx
,
1192 int rxr_count
, int rxr_idx
)
1194 struct igb_q_vector
*q_vector
;
1195 struct igb_ring
*ring
;
1196 int ring_count
, size
;
1198 /* igb only supports 1 Tx and/or 1 Rx queue per vector */
1199 if (txr_count
> 1 || rxr_count
> 1)
1202 ring_count
= txr_count
+ rxr_count
;
1203 size
= sizeof(struct igb_q_vector
) +
1204 (sizeof(struct igb_ring
) * ring_count
);
1206 /* allocate q_vector and rings */
1207 q_vector
= adapter
->q_vector
[v_idx
];
1209 q_vector
= kzalloc(size
, GFP_KERNEL
);
1213 /* initialize NAPI */
1214 netif_napi_add(adapter
->netdev
, &q_vector
->napi
,
1217 /* tie q_vector and adapter together */
1218 adapter
->q_vector
[v_idx
] = q_vector
;
1219 q_vector
->adapter
= adapter
;
1221 /* initialize work limits */
1222 q_vector
->tx
.work_limit
= adapter
->tx_work_limit
;
1224 /* initialize ITR configuration */
1225 q_vector
->itr_register
= adapter
->hw
.hw_addr
+ E1000_EITR(0);
1226 q_vector
->itr_val
= IGB_START_ITR
;
1228 /* initialize pointer to rings */
1229 ring
= q_vector
->ring
;
1233 /* rx or rx/tx vector */
1234 if (!adapter
->rx_itr_setting
|| adapter
->rx_itr_setting
> 3)
1235 q_vector
->itr_val
= adapter
->rx_itr_setting
;
1237 /* tx only vector */
1238 if (!adapter
->tx_itr_setting
|| adapter
->tx_itr_setting
> 3)
1239 q_vector
->itr_val
= adapter
->tx_itr_setting
;
1243 /* assign generic ring traits */
1244 ring
->dev
= &adapter
->pdev
->dev
;
1245 ring
->netdev
= adapter
->netdev
;
1247 /* configure backlink on ring */
1248 ring
->q_vector
= q_vector
;
1250 /* update q_vector Tx values */
1251 igb_add_ring(ring
, &q_vector
->tx
);
1253 /* For 82575, context index must be unique per ring. */
1254 if (adapter
->hw
.mac
.type
== e1000_82575
)
1255 set_bit(IGB_RING_FLAG_TX_CTX_IDX
, &ring
->flags
);
1257 /* apply Tx specific ring traits */
1258 ring
->count
= adapter
->tx_ring_count
;
1259 ring
->queue_index
= txr_idx
;
1261 u64_stats_init(&ring
->tx_syncp
);
1262 u64_stats_init(&ring
->tx_syncp2
);
1264 /* assign ring to adapter */
1265 adapter
->tx_ring
[txr_idx
] = ring
;
1267 /* push pointer to next ring */
1272 /* assign generic ring traits */
1273 ring
->dev
= &adapter
->pdev
->dev
;
1274 ring
->netdev
= adapter
->netdev
;
1276 /* configure backlink on ring */
1277 ring
->q_vector
= q_vector
;
1279 /* update q_vector Rx values */
1280 igb_add_ring(ring
, &q_vector
->rx
);
1282 /* set flag indicating ring supports SCTP checksum offload */
1283 if (adapter
->hw
.mac
.type
>= e1000_82576
)
1284 set_bit(IGB_RING_FLAG_RX_SCTP_CSUM
, &ring
->flags
);
1286 /* On i350, i354, i210, and i211, loopback VLAN packets
1287 * have the tag byte-swapped.
1289 if (adapter
->hw
.mac
.type
>= e1000_i350
)
1290 set_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP
, &ring
->flags
);
1292 /* apply Rx specific ring traits */
1293 ring
->count
= adapter
->rx_ring_count
;
1294 ring
->queue_index
= rxr_idx
;
1296 u64_stats_init(&ring
->rx_syncp
);
1298 /* assign ring to adapter */
1299 adapter
->rx_ring
[rxr_idx
] = ring
;
1307 * igb_alloc_q_vectors - Allocate memory for interrupt vectors
1308 * @adapter: board private structure to initialize
1310 * We allocate one q_vector per queue interrupt. If allocation fails we
1313 static int igb_alloc_q_vectors(struct igb_adapter
*adapter
)
1315 int q_vectors
= adapter
->num_q_vectors
;
1316 int rxr_remaining
= adapter
->num_rx_queues
;
1317 int txr_remaining
= adapter
->num_tx_queues
;
1318 int rxr_idx
= 0, txr_idx
= 0, v_idx
= 0;
1321 if (q_vectors
>= (rxr_remaining
+ txr_remaining
)) {
1322 for (; rxr_remaining
; v_idx
++) {
1323 err
= igb_alloc_q_vector(adapter
, q_vectors
, v_idx
,
1329 /* update counts and index */
1335 for (; v_idx
< q_vectors
; v_idx
++) {
1336 int rqpv
= DIV_ROUND_UP(rxr_remaining
, q_vectors
- v_idx
);
1337 int tqpv
= DIV_ROUND_UP(txr_remaining
, q_vectors
- v_idx
);
1339 err
= igb_alloc_q_vector(adapter
, q_vectors
, v_idx
,
1340 tqpv
, txr_idx
, rqpv
, rxr_idx
);
1345 /* update counts and index */
1346 rxr_remaining
-= rqpv
;
1347 txr_remaining
-= tqpv
;
1355 adapter
->num_tx_queues
= 0;
1356 adapter
->num_rx_queues
= 0;
1357 adapter
->num_q_vectors
= 0;
1360 igb_free_q_vector(adapter
, v_idx
);
1366 * igb_init_interrupt_scheme - initialize interrupts, allocate queues/vectors
1367 * @adapter: board private structure to initialize
1368 * @msix: boolean value of MSIX capability
1370 * This function initializes the interrupts and allocates all of the queues.
1372 static int igb_init_interrupt_scheme(struct igb_adapter
*adapter
, bool msix
)
1374 struct pci_dev
*pdev
= adapter
->pdev
;
1377 igb_set_interrupt_capability(adapter
, msix
);
1379 err
= igb_alloc_q_vectors(adapter
);
1381 dev_err(&pdev
->dev
, "Unable to allocate memory for vectors\n");
1382 goto err_alloc_q_vectors
;
1385 igb_cache_ring_register(adapter
);
1389 err_alloc_q_vectors
:
1390 igb_reset_interrupt_capability(adapter
);
1395 * igb_request_irq - initialize interrupts
1396 * @adapter: board private structure to initialize
1398 * Attempts to configure interrupts using the best available
1399 * capabilities of the hardware and kernel.
1401 static int igb_request_irq(struct igb_adapter
*adapter
)
1403 struct net_device
*netdev
= adapter
->netdev
;
1404 struct pci_dev
*pdev
= adapter
->pdev
;
1407 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1408 err
= igb_request_msix(adapter
);
1411 /* fall back to MSI */
1412 igb_free_all_tx_resources(adapter
);
1413 igb_free_all_rx_resources(adapter
);
1415 igb_clear_interrupt_scheme(adapter
);
1416 err
= igb_init_interrupt_scheme(adapter
, false);
1420 igb_setup_all_tx_resources(adapter
);
1421 igb_setup_all_rx_resources(adapter
);
1422 igb_configure(adapter
);
1425 igb_assign_vector(adapter
->q_vector
[0], 0);
1427 if (adapter
->flags
& IGB_FLAG_HAS_MSI
) {
1428 err
= request_irq(pdev
->irq
, igb_intr_msi
, 0,
1429 netdev
->name
, adapter
);
1433 /* fall back to legacy interrupts */
1434 igb_reset_interrupt_capability(adapter
);
1435 adapter
->flags
&= ~IGB_FLAG_HAS_MSI
;
1438 err
= request_irq(pdev
->irq
, igb_intr
, IRQF_SHARED
,
1439 netdev
->name
, adapter
);
1442 dev_err(&pdev
->dev
, "Error %d getting interrupt\n",
1449 static void igb_free_irq(struct igb_adapter
*adapter
)
1451 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1454 free_irq(adapter
->msix_entries
[vector
++].vector
, adapter
);
1456 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
1457 free_irq(adapter
->msix_entries
[vector
++].vector
,
1458 adapter
->q_vector
[i
]);
1460 free_irq(adapter
->pdev
->irq
, adapter
);
1465 * igb_irq_disable - Mask off interrupt generation on the NIC
1466 * @adapter: board private structure
1468 static void igb_irq_disable(struct igb_adapter
*adapter
)
1470 struct e1000_hw
*hw
= &adapter
->hw
;
1472 /* we need to be careful when disabling interrupts. The VFs are also
1473 * mapped into these registers and so clearing the bits can cause
1474 * issues on the VF drivers so we only need to clear what we set
1476 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1477 u32 regval
= rd32(E1000_EIAM
);
1479 wr32(E1000_EIAM
, regval
& ~adapter
->eims_enable_mask
);
1480 wr32(E1000_EIMC
, adapter
->eims_enable_mask
);
1481 regval
= rd32(E1000_EIAC
);
1482 wr32(E1000_EIAC
, regval
& ~adapter
->eims_enable_mask
);
1486 wr32(E1000_IMC
, ~0);
1488 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1491 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
1492 synchronize_irq(adapter
->msix_entries
[i
].vector
);
1494 synchronize_irq(adapter
->pdev
->irq
);
1499 * igb_irq_enable - Enable default interrupt generation settings
1500 * @adapter: board private structure
1502 static void igb_irq_enable(struct igb_adapter
*adapter
)
1504 struct e1000_hw
*hw
= &adapter
->hw
;
1506 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1507 u32 ims
= E1000_IMS_LSC
| E1000_IMS_DOUTSYNC
| E1000_IMS_DRSTA
;
1508 u32 regval
= rd32(E1000_EIAC
);
1510 wr32(E1000_EIAC
, regval
| adapter
->eims_enable_mask
);
1511 regval
= rd32(E1000_EIAM
);
1512 wr32(E1000_EIAM
, regval
| adapter
->eims_enable_mask
);
1513 wr32(E1000_EIMS
, adapter
->eims_enable_mask
);
1514 if (adapter
->vfs_allocated_count
) {
1515 wr32(E1000_MBVFIMR
, 0xFF);
1516 ims
|= E1000_IMS_VMMB
;
1518 wr32(E1000_IMS
, ims
);
1520 wr32(E1000_IMS
, IMS_ENABLE_MASK
|
1522 wr32(E1000_IAM
, IMS_ENABLE_MASK
|
1527 static void igb_update_mng_vlan(struct igb_adapter
*adapter
)
1529 struct e1000_hw
*hw
= &adapter
->hw
;
1530 u16 vid
= adapter
->hw
.mng_cookie
.vlan_id
;
1531 u16 old_vid
= adapter
->mng_vlan_id
;
1533 if (hw
->mng_cookie
.status
& E1000_MNG_DHCP_COOKIE_STATUS_VLAN
) {
1534 /* add VID to filter table */
1535 igb_vfta_set(hw
, vid
, true);
1536 adapter
->mng_vlan_id
= vid
;
1538 adapter
->mng_vlan_id
= IGB_MNG_VLAN_NONE
;
1541 if ((old_vid
!= (u16
)IGB_MNG_VLAN_NONE
) &&
1543 !test_bit(old_vid
, adapter
->active_vlans
)) {
1544 /* remove VID from filter table */
1545 igb_vfta_set(hw
, old_vid
, false);
1550 * igb_release_hw_control - release control of the h/w to f/w
1551 * @adapter: address of board private structure
1553 * igb_release_hw_control resets CTRL_EXT:DRV_LOAD bit.
1554 * For ASF and Pass Through versions of f/w this means that the
1555 * driver is no longer loaded.
1557 static void igb_release_hw_control(struct igb_adapter
*adapter
)
1559 struct e1000_hw
*hw
= &adapter
->hw
;
1562 /* Let firmware take over control of h/w */
1563 ctrl_ext
= rd32(E1000_CTRL_EXT
);
1564 wr32(E1000_CTRL_EXT
,
1565 ctrl_ext
& ~E1000_CTRL_EXT_DRV_LOAD
);
1569 * igb_get_hw_control - get control of the h/w from f/w
1570 * @adapter: address of board private structure
1572 * igb_get_hw_control sets CTRL_EXT:DRV_LOAD bit.
1573 * For ASF and Pass Through versions of f/w this means that
1574 * the driver is loaded.
1576 static void igb_get_hw_control(struct igb_adapter
*adapter
)
1578 struct e1000_hw
*hw
= &adapter
->hw
;
1581 /* Let firmware know the driver has taken over */
1582 ctrl_ext
= rd32(E1000_CTRL_EXT
);
1583 wr32(E1000_CTRL_EXT
,
1584 ctrl_ext
| E1000_CTRL_EXT_DRV_LOAD
);
1588 * igb_configure - configure the hardware for RX and TX
1589 * @adapter: private board structure
1591 static void igb_configure(struct igb_adapter
*adapter
)
1593 struct net_device
*netdev
= adapter
->netdev
;
1596 igb_get_hw_control(adapter
);
1597 igb_set_rx_mode(netdev
);
1599 igb_restore_vlan(adapter
);
1601 igb_setup_tctl(adapter
);
1602 igb_setup_mrqc(adapter
);
1603 igb_setup_rctl(adapter
);
1605 igb_configure_tx(adapter
);
1606 igb_configure_rx(adapter
);
1608 igb_rx_fifo_flush_82575(&adapter
->hw
);
1610 /* call igb_desc_unused which always leaves
1611 * at least 1 descriptor unused to make sure
1612 * next_to_use != next_to_clean
1614 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
1615 struct igb_ring
*ring
= adapter
->rx_ring
[i
];
1616 igb_alloc_rx_buffers(ring
, igb_desc_unused(ring
));
1621 * igb_power_up_link - Power up the phy/serdes link
1622 * @adapter: address of board private structure
1624 void igb_power_up_link(struct igb_adapter
*adapter
)
1626 igb_reset_phy(&adapter
->hw
);
1628 if (adapter
->hw
.phy
.media_type
== e1000_media_type_copper
)
1629 igb_power_up_phy_copper(&adapter
->hw
);
1631 igb_power_up_serdes_link_82575(&adapter
->hw
);
1633 igb_setup_link(&adapter
->hw
);
1637 * igb_power_down_link - Power down the phy/serdes link
1638 * @adapter: address of board private structure
1640 static void igb_power_down_link(struct igb_adapter
*adapter
)
1642 if (adapter
->hw
.phy
.media_type
== e1000_media_type_copper
)
1643 igb_power_down_phy_copper_82575(&adapter
->hw
);
1645 igb_shutdown_serdes_link_82575(&adapter
->hw
);
1649 * Detect and switch function for Media Auto Sense
1650 * @adapter: address of the board private structure
1652 static void igb_check_swap_media(struct igb_adapter
*adapter
)
1654 struct e1000_hw
*hw
= &adapter
->hw
;
1655 u32 ctrl_ext
, connsw
;
1656 bool swap_now
= false;
1658 ctrl_ext
= rd32(E1000_CTRL_EXT
);
1659 connsw
= rd32(E1000_CONNSW
);
1661 /* need to live swap if current media is copper and we have fiber/serdes
1665 if ((hw
->phy
.media_type
== e1000_media_type_copper
) &&
1666 (!(connsw
& E1000_CONNSW_AUTOSENSE_EN
))) {
1668 } else if (!(connsw
& E1000_CONNSW_SERDESD
)) {
1669 /* copper signal takes time to appear */
1670 if (adapter
->copper_tries
< 4) {
1671 adapter
->copper_tries
++;
1672 connsw
|= E1000_CONNSW_AUTOSENSE_CONF
;
1673 wr32(E1000_CONNSW
, connsw
);
1676 adapter
->copper_tries
= 0;
1677 if ((connsw
& E1000_CONNSW_PHYSD
) &&
1678 (!(connsw
& E1000_CONNSW_PHY_PDN
))) {
1680 connsw
&= ~E1000_CONNSW_AUTOSENSE_CONF
;
1681 wr32(E1000_CONNSW
, connsw
);
1689 switch (hw
->phy
.media_type
) {
1690 case e1000_media_type_copper
:
1691 netdev_info(adapter
->netdev
,
1692 "MAS: changing media to fiber/serdes\n");
1694 E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES
;
1695 adapter
->flags
|= IGB_FLAG_MEDIA_RESET
;
1696 adapter
->copper_tries
= 0;
1698 case e1000_media_type_internal_serdes
:
1699 case e1000_media_type_fiber
:
1700 netdev_info(adapter
->netdev
,
1701 "MAS: changing media to copper\n");
1703 ~E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES
;
1704 adapter
->flags
|= IGB_FLAG_MEDIA_RESET
;
1707 /* shouldn't get here during regular operation */
1708 netdev_err(adapter
->netdev
,
1709 "AMS: Invalid media type found, returning\n");
1712 wr32(E1000_CTRL_EXT
, ctrl_ext
);
1716 * igb_up - Open the interface and prepare it to handle traffic
1717 * @adapter: board private structure
1719 int igb_up(struct igb_adapter
*adapter
)
1721 struct e1000_hw
*hw
= &adapter
->hw
;
1724 /* hardware has been reset, we need to reload some things */
1725 igb_configure(adapter
);
1727 clear_bit(__IGB_DOWN
, &adapter
->state
);
1729 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
1730 napi_enable(&(adapter
->q_vector
[i
]->napi
));
1732 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
)
1733 igb_configure_msix(adapter
);
1735 igb_assign_vector(adapter
->q_vector
[0], 0);
1737 /* Clear any pending interrupts. */
1739 igb_irq_enable(adapter
);
1741 /* notify VFs that reset has been completed */
1742 if (adapter
->vfs_allocated_count
) {
1743 u32 reg_data
= rd32(E1000_CTRL_EXT
);
1745 reg_data
|= E1000_CTRL_EXT_PFRSTD
;
1746 wr32(E1000_CTRL_EXT
, reg_data
);
1749 netif_tx_start_all_queues(adapter
->netdev
);
1751 /* start the watchdog. */
1752 hw
->mac
.get_link_status
= 1;
1753 schedule_work(&adapter
->watchdog_task
);
1755 if ((adapter
->flags
& IGB_FLAG_EEE
) &&
1756 (!hw
->dev_spec
._82575
.eee_disable
))
1757 adapter
->eee_advert
= MDIO_EEE_100TX
| MDIO_EEE_1000T
;
1762 void igb_down(struct igb_adapter
*adapter
)
1764 struct net_device
*netdev
= adapter
->netdev
;
1765 struct e1000_hw
*hw
= &adapter
->hw
;
1769 /* signal that we're down so the interrupt handler does not
1770 * reschedule our watchdog timer
1772 set_bit(__IGB_DOWN
, &adapter
->state
);
1774 /* disable receives in the hardware */
1775 rctl
= rd32(E1000_RCTL
);
1776 wr32(E1000_RCTL
, rctl
& ~E1000_RCTL_EN
);
1777 /* flush and sleep below */
1779 netif_carrier_off(netdev
);
1780 netif_tx_stop_all_queues(netdev
);
1782 /* disable transmits in the hardware */
1783 tctl
= rd32(E1000_TCTL
);
1784 tctl
&= ~E1000_TCTL_EN
;
1785 wr32(E1000_TCTL
, tctl
);
1786 /* flush both disables and wait for them to finish */
1788 usleep_range(10000, 11000);
1790 igb_irq_disable(adapter
);
1792 adapter
->flags
&= ~IGB_FLAG_NEED_LINK_UPDATE
;
1794 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
1795 if (adapter
->q_vector
[i
]) {
1796 napi_synchronize(&adapter
->q_vector
[i
]->napi
);
1797 napi_disable(&adapter
->q_vector
[i
]->napi
);
1801 del_timer_sync(&adapter
->watchdog_timer
);
1802 del_timer_sync(&adapter
->phy_info_timer
);
1804 /* record the stats before reset*/
1805 spin_lock(&adapter
->stats64_lock
);
1806 igb_update_stats(adapter
, &adapter
->stats64
);
1807 spin_unlock(&adapter
->stats64_lock
);
1809 adapter
->link_speed
= 0;
1810 adapter
->link_duplex
= 0;
1812 if (!pci_channel_offline(adapter
->pdev
))
1814 igb_clean_all_tx_rings(adapter
);
1815 igb_clean_all_rx_rings(adapter
);
1816 #ifdef CONFIG_IGB_DCA
1818 /* since we reset the hardware DCA settings were cleared */
1819 igb_setup_dca(adapter
);
1823 void igb_reinit_locked(struct igb_adapter
*adapter
)
1825 WARN_ON(in_interrupt());
1826 while (test_and_set_bit(__IGB_RESETTING
, &adapter
->state
))
1827 usleep_range(1000, 2000);
1830 clear_bit(__IGB_RESETTING
, &adapter
->state
);
1833 /** igb_enable_mas - Media Autosense re-enable after swap
1835 * @adapter: adapter struct
1837 static void igb_enable_mas(struct igb_adapter
*adapter
)
1839 struct e1000_hw
*hw
= &adapter
->hw
;
1840 u32 connsw
= rd32(E1000_CONNSW
);
1842 /* configure for SerDes media detect */
1843 if ((hw
->phy
.media_type
== e1000_media_type_copper
) &&
1844 (!(connsw
& E1000_CONNSW_SERDESD
))) {
1845 connsw
|= E1000_CONNSW_ENRGSRC
;
1846 connsw
|= E1000_CONNSW_AUTOSENSE_EN
;
1847 wr32(E1000_CONNSW
, connsw
);
1852 void igb_reset(struct igb_adapter
*adapter
)
1854 struct pci_dev
*pdev
= adapter
->pdev
;
1855 struct e1000_hw
*hw
= &adapter
->hw
;
1856 struct e1000_mac_info
*mac
= &hw
->mac
;
1857 struct e1000_fc_info
*fc
= &hw
->fc
;
1858 u32 pba
= 0, tx_space
, min_tx_space
, min_rx_space
, hwm
;
1860 /* Repartition Pba for greater than 9k mtu
1861 * To take effect CTRL.RST is required.
1863 switch (mac
->type
) {
1867 pba
= rd32(E1000_RXPBS
);
1868 pba
= igb_rxpbs_adjust_82580(pba
);
1871 pba
= rd32(E1000_RXPBS
);
1872 pba
&= E1000_RXPBS_SIZE_MASK_82576
;
1878 pba
= E1000_PBA_34K
;
1882 if ((adapter
->max_frame_size
> ETH_FRAME_LEN
+ ETH_FCS_LEN
) &&
1883 (mac
->type
< e1000_82576
)) {
1884 /* adjust PBA for jumbo frames */
1885 wr32(E1000_PBA
, pba
);
1887 /* To maintain wire speed transmits, the Tx FIFO should be
1888 * large enough to accommodate two full transmit packets,
1889 * rounded up to the next 1KB and expressed in KB. Likewise,
1890 * the Rx FIFO should be large enough to accommodate at least
1891 * one full receive packet and is similarly rounded up and
1894 pba
= rd32(E1000_PBA
);
1895 /* upper 16 bits has Tx packet buffer allocation size in KB */
1896 tx_space
= pba
>> 16;
1897 /* lower 16 bits has Rx packet buffer allocation size in KB */
1899 /* the Tx fifo also stores 16 bytes of information about the Tx
1900 * but don't include ethernet FCS because hardware appends it
1902 min_tx_space
= (adapter
->max_frame_size
+
1903 sizeof(union e1000_adv_tx_desc
) -
1905 min_tx_space
= ALIGN(min_tx_space
, 1024);
1906 min_tx_space
>>= 10;
1907 /* software strips receive CRC, so leave room for it */
1908 min_rx_space
= adapter
->max_frame_size
;
1909 min_rx_space
= ALIGN(min_rx_space
, 1024);
1910 min_rx_space
>>= 10;
1912 /* If current Tx allocation is less than the min Tx FIFO size,
1913 * and the min Tx FIFO size is less than the current Rx FIFO
1914 * allocation, take space away from current Rx allocation
1916 if (tx_space
< min_tx_space
&&
1917 ((min_tx_space
- tx_space
) < pba
)) {
1918 pba
= pba
- (min_tx_space
- tx_space
);
1920 /* if short on Rx space, Rx wins and must trump Tx
1923 if (pba
< min_rx_space
)
1926 wr32(E1000_PBA
, pba
);
1929 /* flow control settings */
1930 /* The high water mark must be low enough to fit one full frame
1931 * (or the size used for early receive) above it in the Rx FIFO.
1932 * Set it to the lower of:
1933 * - 90% of the Rx FIFO size, or
1934 * - the full Rx FIFO size minus one full frame
1936 hwm
= min(((pba
<< 10) * 9 / 10),
1937 ((pba
<< 10) - 2 * adapter
->max_frame_size
));
1939 fc
->high_water
= hwm
& 0xFFFFFFF0; /* 16-byte granularity */
1940 fc
->low_water
= fc
->high_water
- 16;
1941 fc
->pause_time
= 0xFFFF;
1943 fc
->current_mode
= fc
->requested_mode
;
1945 /* disable receive for all VFs and wait one second */
1946 if (adapter
->vfs_allocated_count
) {
1949 for (i
= 0 ; i
< adapter
->vfs_allocated_count
; i
++)
1950 adapter
->vf_data
[i
].flags
&= IGB_VF_FLAG_PF_SET_MAC
;
1952 /* ping all the active vfs to let them know we are going down */
1953 igb_ping_all_vfs(adapter
);
1955 /* disable transmits and receives */
1956 wr32(E1000_VFRE
, 0);
1957 wr32(E1000_VFTE
, 0);
1960 /* Allow time for pending master requests to run */
1961 hw
->mac
.ops
.reset_hw(hw
);
1964 if (adapter
->flags
& IGB_FLAG_MEDIA_RESET
) {
1965 /* need to resetup here after media swap */
1966 adapter
->ei
.get_invariants(hw
);
1967 adapter
->flags
&= ~IGB_FLAG_MEDIA_RESET
;
1969 if ((mac
->type
== e1000_82575
) &&
1970 (adapter
->flags
& IGB_FLAG_MAS_ENABLE
)) {
1971 igb_enable_mas(adapter
);
1973 if (hw
->mac
.ops
.init_hw(hw
))
1974 dev_err(&pdev
->dev
, "Hardware Error\n");
1976 /* Flow control settings reset on hardware reset, so guarantee flow
1977 * control is off when forcing speed.
1979 if (!hw
->mac
.autoneg
)
1980 igb_force_mac_fc(hw
);
1982 igb_init_dmac(adapter
, pba
);
1983 #ifdef CONFIG_IGB_HWMON
1984 /* Re-initialize the thermal sensor on i350 devices. */
1985 if (!test_bit(__IGB_DOWN
, &adapter
->state
)) {
1986 if (mac
->type
== e1000_i350
&& hw
->bus
.func
== 0) {
1987 /* If present, re-initialize the external thermal sensor
1991 mac
->ops
.init_thermal_sensor_thresh(hw
);
1995 /* Re-establish EEE setting */
1996 if (hw
->phy
.media_type
== e1000_media_type_copper
) {
1997 switch (mac
->type
) {
2001 igb_set_eee_i350(hw
, true, true);
2004 igb_set_eee_i354(hw
, true, true);
2010 if (!netif_running(adapter
->netdev
))
2011 igb_power_down_link(adapter
);
2013 igb_update_mng_vlan(adapter
);
2015 /* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
2016 wr32(E1000_VET
, ETHERNET_IEEE_VLAN_TYPE
);
2018 /* Re-enable PTP, where applicable. */
2019 igb_ptp_reset(adapter
);
2021 igb_get_phy_info(hw
);
2024 static netdev_features_t
igb_fix_features(struct net_device
*netdev
,
2025 netdev_features_t features
)
2027 /* Since there is no support for separate Rx/Tx vlan accel
2028 * enable/disable make sure Tx flag is always in same state as Rx.
2030 if (features
& NETIF_F_HW_VLAN_CTAG_RX
)
2031 features
|= NETIF_F_HW_VLAN_CTAG_TX
;
2033 features
&= ~NETIF_F_HW_VLAN_CTAG_TX
;
2038 static int igb_set_features(struct net_device
*netdev
,
2039 netdev_features_t features
)
2041 netdev_features_t changed
= netdev
->features
^ features
;
2042 struct igb_adapter
*adapter
= netdev_priv(netdev
);
2044 if (changed
& NETIF_F_HW_VLAN_CTAG_RX
)
2045 igb_vlan_mode(netdev
, features
);
2047 if (!(changed
& NETIF_F_RXALL
))
2050 netdev
->features
= features
;
2052 if (netif_running(netdev
))
2053 igb_reinit_locked(adapter
);
2060 static const struct net_device_ops igb_netdev_ops
= {
2061 .ndo_open
= igb_open
,
2062 .ndo_stop
= igb_close
,
2063 .ndo_start_xmit
= igb_xmit_frame
,
2064 .ndo_get_stats64
= igb_get_stats64
,
2065 .ndo_set_rx_mode
= igb_set_rx_mode
,
2066 .ndo_set_mac_address
= igb_set_mac
,
2067 .ndo_change_mtu
= igb_change_mtu
,
2068 .ndo_do_ioctl
= igb_ioctl
,
2069 .ndo_tx_timeout
= igb_tx_timeout
,
2070 .ndo_validate_addr
= eth_validate_addr
,
2071 .ndo_vlan_rx_add_vid
= igb_vlan_rx_add_vid
,
2072 .ndo_vlan_rx_kill_vid
= igb_vlan_rx_kill_vid
,
2073 .ndo_set_vf_mac
= igb_ndo_set_vf_mac
,
2074 .ndo_set_vf_vlan
= igb_ndo_set_vf_vlan
,
2075 .ndo_set_vf_rate
= igb_ndo_set_vf_bw
,
2076 .ndo_set_vf_spoofchk
= igb_ndo_set_vf_spoofchk
,
2077 .ndo_get_vf_config
= igb_ndo_get_vf_config
,
2078 #ifdef CONFIG_NET_POLL_CONTROLLER
2079 .ndo_poll_controller
= igb_netpoll
,
2081 .ndo_fix_features
= igb_fix_features
,
2082 .ndo_set_features
= igb_set_features
,
2083 .ndo_features_check
= passthru_features_check
,
2087 * igb_set_fw_version - Configure version string for ethtool
2088 * @adapter: adapter struct
2090 void igb_set_fw_version(struct igb_adapter
*adapter
)
2092 struct e1000_hw
*hw
= &adapter
->hw
;
2093 struct e1000_fw_version fw
;
2095 igb_get_fw_version(hw
, &fw
);
2097 switch (hw
->mac
.type
) {
2100 if (!(igb_get_flash_presence_i210(hw
))) {
2101 snprintf(adapter
->fw_version
,
2102 sizeof(adapter
->fw_version
),
2104 fw
.invm_major
, fw
.invm_minor
,
2110 /* if option is rom valid, display its version too */
2112 snprintf(adapter
->fw_version
,
2113 sizeof(adapter
->fw_version
),
2114 "%d.%d, 0x%08x, %d.%d.%d",
2115 fw
.eep_major
, fw
.eep_minor
, fw
.etrack_id
,
2116 fw
.or_major
, fw
.or_build
, fw
.or_patch
);
2118 } else if (fw
.etrack_id
!= 0X0000) {
2119 snprintf(adapter
->fw_version
,
2120 sizeof(adapter
->fw_version
),
2122 fw
.eep_major
, fw
.eep_minor
, fw
.etrack_id
);
2124 snprintf(adapter
->fw_version
,
2125 sizeof(adapter
->fw_version
),
2127 fw
.eep_major
, fw
.eep_minor
, fw
.eep_build
);
2134 * igb_init_mas - init Media Autosense feature if enabled in the NVM
2136 * @adapter: adapter struct
2138 static void igb_init_mas(struct igb_adapter
*adapter
)
2140 struct e1000_hw
*hw
= &adapter
->hw
;
2143 hw
->nvm
.ops
.read(hw
, NVM_COMPAT
, 1, &eeprom_data
);
2144 switch (hw
->bus
.func
) {
2146 if (eeprom_data
& IGB_MAS_ENABLE_0
) {
2147 adapter
->flags
|= IGB_FLAG_MAS_ENABLE
;
2148 netdev_info(adapter
->netdev
,
2149 "MAS: Enabling Media Autosense for port %d\n",
2154 if (eeprom_data
& IGB_MAS_ENABLE_1
) {
2155 adapter
->flags
|= IGB_FLAG_MAS_ENABLE
;
2156 netdev_info(adapter
->netdev
,
2157 "MAS: Enabling Media Autosense for port %d\n",
2162 if (eeprom_data
& IGB_MAS_ENABLE_2
) {
2163 adapter
->flags
|= IGB_FLAG_MAS_ENABLE
;
2164 netdev_info(adapter
->netdev
,
2165 "MAS: Enabling Media Autosense for port %d\n",
2170 if (eeprom_data
& IGB_MAS_ENABLE_3
) {
2171 adapter
->flags
|= IGB_FLAG_MAS_ENABLE
;
2172 netdev_info(adapter
->netdev
,
2173 "MAS: Enabling Media Autosense for port %d\n",
2178 /* Shouldn't get here */
2179 netdev_err(adapter
->netdev
,
2180 "MAS: Invalid port configuration, returning\n");
2186 * igb_init_i2c - Init I2C interface
2187 * @adapter: pointer to adapter structure
2189 static s32
igb_init_i2c(struct igb_adapter
*adapter
)
2193 /* I2C interface supported on i350 devices */
2194 if (adapter
->hw
.mac
.type
!= e1000_i350
)
2197 /* Initialize the i2c bus which is controlled by the registers.
2198 * This bus will use the i2c_algo_bit structue that implements
2199 * the protocol through toggling of the 4 bits in the register.
2201 adapter
->i2c_adap
.owner
= THIS_MODULE
;
2202 adapter
->i2c_algo
= igb_i2c_algo
;
2203 adapter
->i2c_algo
.data
= adapter
;
2204 adapter
->i2c_adap
.algo_data
= &adapter
->i2c_algo
;
2205 adapter
->i2c_adap
.dev
.parent
= &adapter
->pdev
->dev
;
2206 strlcpy(adapter
->i2c_adap
.name
, "igb BB",
2207 sizeof(adapter
->i2c_adap
.name
));
2208 status
= i2c_bit_add_bus(&adapter
->i2c_adap
);
2213 * igb_probe - Device Initialization Routine
2214 * @pdev: PCI device information struct
2215 * @ent: entry in igb_pci_tbl
2217 * Returns 0 on success, negative on failure
2219 * igb_probe initializes an adapter identified by a pci_dev structure.
2220 * The OS initialization, configuring of the adapter private structure,
2221 * and a hardware reset occur.
2223 static int igb_probe(struct pci_dev
*pdev
, const struct pci_device_id
*ent
)
2225 struct net_device
*netdev
;
2226 struct igb_adapter
*adapter
;
2227 struct e1000_hw
*hw
;
2228 u16 eeprom_data
= 0;
2230 static int global_quad_port_a
; /* global quad port a indication */
2231 const struct e1000_info
*ei
= igb_info_tbl
[ent
->driver_data
];
2232 int err
, pci_using_dac
;
2233 u8 part_str
[E1000_PBANUM_LENGTH
];
2235 /* Catch broken hardware that put the wrong VF device ID in
2236 * the PCIe SR-IOV capability.
2238 if (pdev
->is_virtfn
) {
2239 WARN(1, KERN_ERR
"%s (%hx:%hx) should not be a VF!\n",
2240 pci_name(pdev
), pdev
->vendor
, pdev
->device
);
2244 err
= pci_enable_device_mem(pdev
);
2249 err
= dma_set_mask_and_coherent(&pdev
->dev
, DMA_BIT_MASK(64));
2253 err
= dma_set_mask_and_coherent(&pdev
->dev
, DMA_BIT_MASK(32));
2256 "No usable DMA configuration, aborting\n");
2261 err
= pci_request_selected_regions(pdev
, pci_select_bars(pdev
,
2267 pci_enable_pcie_error_reporting(pdev
);
2269 pci_set_master(pdev
);
2270 pci_save_state(pdev
);
2273 netdev
= alloc_etherdev_mq(sizeof(struct igb_adapter
),
2276 goto err_alloc_etherdev
;
2278 SET_NETDEV_DEV(netdev
, &pdev
->dev
);
2280 pci_set_drvdata(pdev
, netdev
);
2281 adapter
= netdev_priv(netdev
);
2282 adapter
->netdev
= netdev
;
2283 adapter
->pdev
= pdev
;
2286 adapter
->msg_enable
= netif_msg_init(debug
, DEFAULT_MSG_ENABLE
);
2289 hw
->hw_addr
= pci_iomap(pdev
, 0, 0);
2293 netdev
->netdev_ops
= &igb_netdev_ops
;
2294 igb_set_ethtool_ops(netdev
);
2295 netdev
->watchdog_timeo
= 5 * HZ
;
2297 strncpy(netdev
->name
, pci_name(pdev
), sizeof(netdev
->name
) - 1);
2299 netdev
->mem_start
= pci_resource_start(pdev
, 0);
2300 netdev
->mem_end
= pci_resource_end(pdev
, 0);
2302 /* PCI config space info */
2303 hw
->vendor_id
= pdev
->vendor
;
2304 hw
->device_id
= pdev
->device
;
2305 hw
->revision_id
= pdev
->revision
;
2306 hw
->subsystem_vendor_id
= pdev
->subsystem_vendor
;
2307 hw
->subsystem_device_id
= pdev
->subsystem_device
;
2309 /* Copy the default MAC, PHY and NVM function pointers */
2310 memcpy(&hw
->mac
.ops
, ei
->mac_ops
, sizeof(hw
->mac
.ops
));
2311 memcpy(&hw
->phy
.ops
, ei
->phy_ops
, sizeof(hw
->phy
.ops
));
2312 memcpy(&hw
->nvm
.ops
, ei
->nvm_ops
, sizeof(hw
->nvm
.ops
));
2313 /* Initialize skew-specific constants */
2314 err
= ei
->get_invariants(hw
);
2318 /* setup the private structure */
2319 err
= igb_sw_init(adapter
);
2323 igb_get_bus_info_pcie(hw
);
2325 hw
->phy
.autoneg_wait_to_complete
= false;
2327 /* Copper options */
2328 if (hw
->phy
.media_type
== e1000_media_type_copper
) {
2329 hw
->phy
.mdix
= AUTO_ALL_MODES
;
2330 hw
->phy
.disable_polarity_correction
= false;
2331 hw
->phy
.ms_type
= e1000_ms_hw_default
;
2334 if (igb_check_reset_block(hw
))
2335 dev_info(&pdev
->dev
,
2336 "PHY reset is blocked due to SOL/IDER session.\n");
2338 /* features is initialized to 0 in allocation, it might have bits
2339 * set by igb_sw_init so we should use an or instead of an
2342 netdev
->features
|= NETIF_F_SG
|
2349 NETIF_F_HW_VLAN_CTAG_RX
|
2350 NETIF_F_HW_VLAN_CTAG_TX
;
2352 /* copy netdev features into list of user selectable features */
2353 netdev
->hw_features
|= netdev
->features
;
2354 netdev
->hw_features
|= NETIF_F_RXALL
;
2356 /* set this bit last since it cannot be part of hw_features */
2357 netdev
->features
|= NETIF_F_HW_VLAN_CTAG_FILTER
;
2359 netdev
->vlan_features
|= NETIF_F_TSO
|
2365 netdev
->priv_flags
|= IFF_SUPP_NOFCS
;
2367 if (pci_using_dac
) {
2368 netdev
->features
|= NETIF_F_HIGHDMA
;
2369 netdev
->vlan_features
|= NETIF_F_HIGHDMA
;
2372 if (hw
->mac
.type
>= e1000_82576
) {
2373 netdev
->hw_features
|= NETIF_F_SCTP_CSUM
;
2374 netdev
->features
|= NETIF_F_SCTP_CSUM
;
2377 netdev
->priv_flags
|= IFF_UNICAST_FLT
;
2379 adapter
->en_mng_pt
= igb_enable_mng_pass_thru(hw
);
2381 /* before reading the NVM, reset the controller to put the device in a
2382 * known good starting state
2384 hw
->mac
.ops
.reset_hw(hw
);
2386 /* make sure the NVM is good , i211/i210 parts can have special NVM
2387 * that doesn't contain a checksum
2389 switch (hw
->mac
.type
) {
2392 if (igb_get_flash_presence_i210(hw
)) {
2393 if (hw
->nvm
.ops
.validate(hw
) < 0) {
2395 "The NVM Checksum Is Not Valid\n");
2402 if (hw
->nvm
.ops
.validate(hw
) < 0) {
2403 dev_err(&pdev
->dev
, "The NVM Checksum Is Not Valid\n");
2410 /* copy the MAC address out of the NVM */
2411 if (hw
->mac
.ops
.read_mac_addr(hw
))
2412 dev_err(&pdev
->dev
, "NVM Read Error\n");
2414 memcpy(netdev
->dev_addr
, hw
->mac
.addr
, netdev
->addr_len
);
2416 if (!is_valid_ether_addr(netdev
->dev_addr
)) {
2417 dev_err(&pdev
->dev
, "Invalid MAC Address\n");
2422 /* get firmware version for ethtool -i */
2423 igb_set_fw_version(adapter
);
2425 /* configure RXPBSIZE and TXPBSIZE */
2426 if (hw
->mac
.type
== e1000_i210
) {
2427 wr32(E1000_RXPBS
, I210_RXPBSIZE_DEFAULT
);
2428 wr32(E1000_TXPBS
, I210_TXPBSIZE_DEFAULT
);
2431 setup_timer(&adapter
->watchdog_timer
, igb_watchdog
,
2432 (unsigned long) adapter
);
2433 setup_timer(&adapter
->phy_info_timer
, igb_update_phy_info
,
2434 (unsigned long) adapter
);
2436 INIT_WORK(&adapter
->reset_task
, igb_reset_task
);
2437 INIT_WORK(&adapter
->watchdog_task
, igb_watchdog_task
);
2439 /* Initialize link properties that are user-changeable */
2440 adapter
->fc_autoneg
= true;
2441 hw
->mac
.autoneg
= true;
2442 hw
->phy
.autoneg_advertised
= 0x2f;
2444 hw
->fc
.requested_mode
= e1000_fc_default
;
2445 hw
->fc
.current_mode
= e1000_fc_default
;
2447 igb_validate_mdi_setting(hw
);
2449 /* By default, support wake on port A */
2450 if (hw
->bus
.func
== 0)
2451 adapter
->flags
|= IGB_FLAG_WOL_SUPPORTED
;
2453 /* Check the NVM for wake support on non-port A ports */
2454 if (hw
->mac
.type
>= e1000_82580
)
2455 hw
->nvm
.ops
.read(hw
, NVM_INIT_CONTROL3_PORT_A
+
2456 NVM_82580_LAN_FUNC_OFFSET(hw
->bus
.func
), 1,
2458 else if (hw
->bus
.func
== 1)
2459 hw
->nvm
.ops
.read(hw
, NVM_INIT_CONTROL3_PORT_B
, 1, &eeprom_data
);
2461 if (eeprom_data
& IGB_EEPROM_APME
)
2462 adapter
->flags
|= IGB_FLAG_WOL_SUPPORTED
;
2464 /* now that we have the eeprom settings, apply the special cases where
2465 * the eeprom may be wrong or the board simply won't support wake on
2466 * lan on a particular port
2468 switch (pdev
->device
) {
2469 case E1000_DEV_ID_82575GB_QUAD_COPPER
:
2470 adapter
->flags
&= ~IGB_FLAG_WOL_SUPPORTED
;
2472 case E1000_DEV_ID_82575EB_FIBER_SERDES
:
2473 case E1000_DEV_ID_82576_FIBER
:
2474 case E1000_DEV_ID_82576_SERDES
:
2475 /* Wake events only supported on port A for dual fiber
2476 * regardless of eeprom setting
2478 if (rd32(E1000_STATUS
) & E1000_STATUS_FUNC_1
)
2479 adapter
->flags
&= ~IGB_FLAG_WOL_SUPPORTED
;
2481 case E1000_DEV_ID_82576_QUAD_COPPER
:
2482 case E1000_DEV_ID_82576_QUAD_COPPER_ET2
:
2483 /* if quad port adapter, disable WoL on all but port A */
2484 if (global_quad_port_a
!= 0)
2485 adapter
->flags
&= ~IGB_FLAG_WOL_SUPPORTED
;
2487 adapter
->flags
|= IGB_FLAG_QUAD_PORT_A
;
2488 /* Reset for multiple quad port adapters */
2489 if (++global_quad_port_a
== 4)
2490 global_quad_port_a
= 0;
2493 /* If the device can't wake, don't set software support */
2494 if (!device_can_wakeup(&adapter
->pdev
->dev
))
2495 adapter
->flags
&= ~IGB_FLAG_WOL_SUPPORTED
;
2498 /* initialize the wol settings based on the eeprom settings */
2499 if (adapter
->flags
& IGB_FLAG_WOL_SUPPORTED
)
2500 adapter
->wol
|= E1000_WUFC_MAG
;
2502 /* Some vendors want WoL disabled by default, but still supported */
2503 if ((hw
->mac
.type
== e1000_i350
) &&
2504 (pdev
->subsystem_vendor
== PCI_VENDOR_ID_HP
)) {
2505 adapter
->flags
|= IGB_FLAG_WOL_SUPPORTED
;
2509 device_set_wakeup_enable(&adapter
->pdev
->dev
,
2510 adapter
->flags
& IGB_FLAG_WOL_SUPPORTED
);
2512 /* reset the hardware with the new settings */
2515 /* Init the I2C interface */
2516 err
= igb_init_i2c(adapter
);
2518 dev_err(&pdev
->dev
, "failed to init i2c interface\n");
2522 /* let the f/w know that the h/w is now under the control of the
2525 igb_get_hw_control(adapter
);
2527 strcpy(netdev
->name
, "eth%d");
2528 err
= register_netdev(netdev
);
2532 /* carrier off reporting is important to ethtool even BEFORE open */
2533 netif_carrier_off(netdev
);
2535 #ifdef CONFIG_IGB_DCA
2536 if (dca_add_requester(&pdev
->dev
) == 0) {
2537 adapter
->flags
|= IGB_FLAG_DCA_ENABLED
;
2538 dev_info(&pdev
->dev
, "DCA enabled\n");
2539 igb_setup_dca(adapter
);
2543 #ifdef CONFIG_IGB_HWMON
2544 /* Initialize the thermal sensor on i350 devices. */
2545 if (hw
->mac
.type
== e1000_i350
&& hw
->bus
.func
== 0) {
2548 /* Read the NVM to determine if this i350 device supports an
2549 * external thermal sensor.
2551 hw
->nvm
.ops
.read(hw
, NVM_ETS_CFG
, 1, &ets_word
);
2552 if (ets_word
!= 0x0000 && ets_word
!= 0xFFFF)
2553 adapter
->ets
= true;
2555 adapter
->ets
= false;
2556 if (igb_sysfs_init(adapter
))
2558 "failed to allocate sysfs resources\n");
2560 adapter
->ets
= false;
2563 /* Check if Media Autosense is enabled */
2565 if (hw
->dev_spec
._82575
.mas_capable
)
2566 igb_init_mas(adapter
);
2568 /* do hw tstamp init after resetting */
2569 igb_ptp_init(adapter
);
2571 dev_info(&pdev
->dev
, "Intel(R) Gigabit Ethernet Network Connection\n");
2572 /* print bus type/speed/width info, not applicable to i354 */
2573 if (hw
->mac
.type
!= e1000_i354
) {
2574 dev_info(&pdev
->dev
, "%s: (PCIe:%s:%s) %pM\n",
2576 ((hw
->bus
.speed
== e1000_bus_speed_2500
) ? "2.5Gb/s" :
2577 (hw
->bus
.speed
== e1000_bus_speed_5000
) ? "5.0Gb/s" :
2579 ((hw
->bus
.width
== e1000_bus_width_pcie_x4
) ?
2581 (hw
->bus
.width
== e1000_bus_width_pcie_x2
) ?
2583 (hw
->bus
.width
== e1000_bus_width_pcie_x1
) ?
2584 "Width x1" : "unknown"), netdev
->dev_addr
);
2587 if ((hw
->mac
.type
>= e1000_i210
||
2588 igb_get_flash_presence_i210(hw
))) {
2589 ret_val
= igb_read_part_string(hw
, part_str
,
2590 E1000_PBANUM_LENGTH
);
2592 ret_val
= -E1000_ERR_INVM_VALUE_NOT_FOUND
;
2596 strcpy(part_str
, "Unknown");
2597 dev_info(&pdev
->dev
, "%s: PBA No: %s\n", netdev
->name
, part_str
);
2598 dev_info(&pdev
->dev
,
2599 "Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
2600 (adapter
->flags
& IGB_FLAG_HAS_MSIX
) ? "MSI-X" :
2601 (adapter
->flags
& IGB_FLAG_HAS_MSI
) ? "MSI" : "legacy",
2602 adapter
->num_rx_queues
, adapter
->num_tx_queues
);
2603 if (hw
->phy
.media_type
== e1000_media_type_copper
) {
2604 switch (hw
->mac
.type
) {
2608 /* Enable EEE for internal copper PHY devices */
2609 err
= igb_set_eee_i350(hw
, true, true);
2611 (!hw
->dev_spec
._82575
.eee_disable
)) {
2612 adapter
->eee_advert
=
2613 MDIO_EEE_100TX
| MDIO_EEE_1000T
;
2614 adapter
->flags
|= IGB_FLAG_EEE
;
2618 if ((rd32(E1000_CTRL_EXT
) &
2619 E1000_CTRL_EXT_LINK_MODE_SGMII
)) {
2620 err
= igb_set_eee_i354(hw
, true, true);
2622 (!hw
->dev_spec
._82575
.eee_disable
)) {
2623 adapter
->eee_advert
=
2624 MDIO_EEE_100TX
| MDIO_EEE_1000T
;
2625 adapter
->flags
|= IGB_FLAG_EEE
;
2633 pm_runtime_put_noidle(&pdev
->dev
);
2637 igb_release_hw_control(adapter
);
2638 memset(&adapter
->i2c_adap
, 0, sizeof(adapter
->i2c_adap
));
2640 if (!igb_check_reset_block(hw
))
2643 if (hw
->flash_address
)
2644 iounmap(hw
->flash_address
);
2646 igb_clear_interrupt_scheme(adapter
);
2647 pci_iounmap(pdev
, hw
->hw_addr
);
2649 free_netdev(netdev
);
2651 pci_release_selected_regions(pdev
,
2652 pci_select_bars(pdev
, IORESOURCE_MEM
));
2655 pci_disable_device(pdev
);
2659 #ifdef CONFIG_PCI_IOV
2660 static int igb_disable_sriov(struct pci_dev
*pdev
)
2662 struct net_device
*netdev
= pci_get_drvdata(pdev
);
2663 struct igb_adapter
*adapter
= netdev_priv(netdev
);
2664 struct e1000_hw
*hw
= &adapter
->hw
;
2666 /* reclaim resources allocated to VFs */
2667 if (adapter
->vf_data
) {
2668 /* disable iov and allow time for transactions to clear */
2669 if (pci_vfs_assigned(pdev
)) {
2670 dev_warn(&pdev
->dev
,
2671 "Cannot deallocate SR-IOV virtual functions while they are assigned - VFs will not be deallocated\n");
2674 pci_disable_sriov(pdev
);
2678 kfree(adapter
->vf_data
);
2679 adapter
->vf_data
= NULL
;
2680 adapter
->vfs_allocated_count
= 0;
2681 wr32(E1000_IOVCTL
, E1000_IOVCTL_REUSE_VFQ
);
2684 dev_info(&pdev
->dev
, "IOV Disabled\n");
2686 /* Re-enable DMA Coalescing flag since IOV is turned off */
2687 adapter
->flags
|= IGB_FLAG_DMAC
;
2693 static int igb_enable_sriov(struct pci_dev
*pdev
, int num_vfs
)
2695 struct net_device
*netdev
= pci_get_drvdata(pdev
);
2696 struct igb_adapter
*adapter
= netdev_priv(netdev
);
2697 int old_vfs
= pci_num_vf(pdev
);
2701 if (!(adapter
->flags
& IGB_FLAG_HAS_MSIX
) || num_vfs
> 7) {
2709 dev_info(&pdev
->dev
, "%d pre-allocated VFs found - override max_vfs setting of %d\n",
2711 adapter
->vfs_allocated_count
= old_vfs
;
2713 adapter
->vfs_allocated_count
= num_vfs
;
2715 adapter
->vf_data
= kcalloc(adapter
->vfs_allocated_count
,
2716 sizeof(struct vf_data_storage
), GFP_KERNEL
);
2718 /* if allocation failed then we do not support SR-IOV */
2719 if (!adapter
->vf_data
) {
2720 adapter
->vfs_allocated_count
= 0;
2722 "Unable to allocate memory for VF Data Storage\n");
2727 /* only call pci_enable_sriov() if no VFs are allocated already */
2729 err
= pci_enable_sriov(pdev
, adapter
->vfs_allocated_count
);
2733 dev_info(&pdev
->dev
, "%d VFs allocated\n",
2734 adapter
->vfs_allocated_count
);
2735 for (i
= 0; i
< adapter
->vfs_allocated_count
; i
++)
2736 igb_vf_configure(adapter
, i
);
2738 /* DMA Coalescing is not supported in IOV mode. */
2739 adapter
->flags
&= ~IGB_FLAG_DMAC
;
2743 kfree(adapter
->vf_data
);
2744 adapter
->vf_data
= NULL
;
2745 adapter
->vfs_allocated_count
= 0;
2752 * igb_remove_i2c - Cleanup I2C interface
2753 * @adapter: pointer to adapter structure
2755 static void igb_remove_i2c(struct igb_adapter
*adapter
)
2757 /* free the adapter bus structure */
2758 i2c_del_adapter(&adapter
->i2c_adap
);
2762 * igb_remove - Device Removal Routine
2763 * @pdev: PCI device information struct
2765 * igb_remove is called by the PCI subsystem to alert the driver
2766 * that it should release a PCI device. The could be caused by a
2767 * Hot-Plug event, or because the driver is going to be removed from
2770 static void igb_remove(struct pci_dev
*pdev
)
2772 struct net_device
*netdev
= pci_get_drvdata(pdev
);
2773 struct igb_adapter
*adapter
= netdev_priv(netdev
);
2774 struct e1000_hw
*hw
= &adapter
->hw
;
2776 pm_runtime_get_noresume(&pdev
->dev
);
2777 #ifdef CONFIG_IGB_HWMON
2778 igb_sysfs_exit(adapter
);
2780 igb_remove_i2c(adapter
);
2781 igb_ptp_stop(adapter
);
2782 /* The watchdog timer may be rescheduled, so explicitly
2783 * disable watchdog from being rescheduled.
2785 set_bit(__IGB_DOWN
, &adapter
->state
);
2786 del_timer_sync(&adapter
->watchdog_timer
);
2787 del_timer_sync(&adapter
->phy_info_timer
);
2789 cancel_work_sync(&adapter
->reset_task
);
2790 cancel_work_sync(&adapter
->watchdog_task
);
2792 #ifdef CONFIG_IGB_DCA
2793 if (adapter
->flags
& IGB_FLAG_DCA_ENABLED
) {
2794 dev_info(&pdev
->dev
, "DCA disabled\n");
2795 dca_remove_requester(&pdev
->dev
);
2796 adapter
->flags
&= ~IGB_FLAG_DCA_ENABLED
;
2797 wr32(E1000_DCA_CTRL
, E1000_DCA_CTRL_DCA_MODE_DISABLE
);
2801 /* Release control of h/w to f/w. If f/w is AMT enabled, this
2802 * would have already happened in close and is redundant.
2804 igb_release_hw_control(adapter
);
2806 unregister_netdev(netdev
);
2808 igb_clear_interrupt_scheme(adapter
);
2810 #ifdef CONFIG_PCI_IOV
2811 igb_disable_sriov(pdev
);
2814 pci_iounmap(pdev
, hw
->hw_addr
);
2815 if (hw
->flash_address
)
2816 iounmap(hw
->flash_address
);
2817 pci_release_selected_regions(pdev
,
2818 pci_select_bars(pdev
, IORESOURCE_MEM
));
2820 kfree(adapter
->shadow_vfta
);
2821 free_netdev(netdev
);
2823 pci_disable_pcie_error_reporting(pdev
);
2825 pci_disable_device(pdev
);
2829 * igb_probe_vfs - Initialize vf data storage and add VFs to pci config space
2830 * @adapter: board private structure to initialize
2832 * This function initializes the vf specific data storage and then attempts to
2833 * allocate the VFs. The reason for ordering it this way is because it is much
2834 * mor expensive time wise to disable SR-IOV than it is to allocate and free
2835 * the memory for the VFs.
2837 static void igb_probe_vfs(struct igb_adapter
*adapter
)
2839 #ifdef CONFIG_PCI_IOV
2840 struct pci_dev
*pdev
= adapter
->pdev
;
2841 struct e1000_hw
*hw
= &adapter
->hw
;
2843 /* Virtualization features not supported on i210 family. */
2844 if ((hw
->mac
.type
== e1000_i210
) || (hw
->mac
.type
== e1000_i211
))
2847 pci_sriov_set_totalvfs(pdev
, 7);
2848 igb_pci_enable_sriov(pdev
, max_vfs
);
2850 #endif /* CONFIG_PCI_IOV */
2853 static void igb_init_queue_configuration(struct igb_adapter
*adapter
)
2855 struct e1000_hw
*hw
= &adapter
->hw
;
2858 /* Determine the maximum number of RSS queues supported. */
2859 switch (hw
->mac
.type
) {
2861 max_rss_queues
= IGB_MAX_RX_QUEUES_I211
;
2865 max_rss_queues
= IGB_MAX_RX_QUEUES_82575
;
2868 /* I350 cannot do RSS and SR-IOV at the same time */
2869 if (!!adapter
->vfs_allocated_count
) {
2875 if (!!adapter
->vfs_allocated_count
) {
2883 max_rss_queues
= IGB_MAX_RX_QUEUES
;
2887 adapter
->rss_queues
= min_t(u32
, max_rss_queues
, num_online_cpus());
2889 /* Determine if we need to pair queues. */
2890 switch (hw
->mac
.type
) {
2893 /* Device supports enough interrupts without queue pairing. */
2896 /* If VFs are going to be allocated with RSS queues then we
2897 * should pair the queues in order to conserve interrupts due
2898 * to limited supply.
2900 if ((adapter
->rss_queues
> 1) &&
2901 (adapter
->vfs_allocated_count
> 6))
2902 adapter
->flags
|= IGB_FLAG_QUEUE_PAIRS
;
2909 /* If rss_queues > half of max_rss_queues, pair the queues in
2910 * order to conserve interrupts due to limited supply.
2912 if (adapter
->rss_queues
> (max_rss_queues
/ 2))
2913 adapter
->flags
|= IGB_FLAG_QUEUE_PAIRS
;
2919 * igb_sw_init - Initialize general software structures (struct igb_adapter)
2920 * @adapter: board private structure to initialize
2922 * igb_sw_init initializes the Adapter private data structure.
2923 * Fields are initialized based on PCI device information and
2924 * OS network device settings (MTU size).
2926 static int igb_sw_init(struct igb_adapter
*adapter
)
2928 struct e1000_hw
*hw
= &adapter
->hw
;
2929 struct net_device
*netdev
= adapter
->netdev
;
2930 struct pci_dev
*pdev
= adapter
->pdev
;
2932 pci_read_config_word(pdev
, PCI_COMMAND
, &hw
->bus
.pci_cmd_word
);
2934 /* set default ring sizes */
2935 adapter
->tx_ring_count
= IGB_DEFAULT_TXD
;
2936 adapter
->rx_ring_count
= IGB_DEFAULT_RXD
;
2938 /* set default ITR values */
2939 adapter
->rx_itr_setting
= IGB_DEFAULT_ITR
;
2940 adapter
->tx_itr_setting
= IGB_DEFAULT_ITR
;
2942 /* set default work limits */
2943 adapter
->tx_work_limit
= IGB_DEFAULT_TX_WORK
;
2945 adapter
->max_frame_size
= netdev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
+
2947 adapter
->min_frame_size
= ETH_ZLEN
+ ETH_FCS_LEN
;
2949 spin_lock_init(&adapter
->stats64_lock
);
2950 #ifdef CONFIG_PCI_IOV
2951 switch (hw
->mac
.type
) {
2955 dev_warn(&pdev
->dev
,
2956 "Maximum of 7 VFs per PF, using max\n");
2957 max_vfs
= adapter
->vfs_allocated_count
= 7;
2959 adapter
->vfs_allocated_count
= max_vfs
;
2960 if (adapter
->vfs_allocated_count
)
2961 dev_warn(&pdev
->dev
,
2962 "Enabling SR-IOV VFs using the module parameter is deprecated - please use the pci sysfs interface.\n");
2967 #endif /* CONFIG_PCI_IOV */
2969 igb_init_queue_configuration(adapter
);
2971 /* Setup and initialize a copy of the hw vlan table array */
2972 adapter
->shadow_vfta
= kcalloc(E1000_VLAN_FILTER_TBL_SIZE
, sizeof(u32
),
2975 /* This call may decrease the number of queues */
2976 if (igb_init_interrupt_scheme(adapter
, true)) {
2977 dev_err(&pdev
->dev
, "Unable to allocate memory for queues\n");
2981 igb_probe_vfs(adapter
);
2983 /* Explicitly disable IRQ since the NIC can be in any state. */
2984 igb_irq_disable(adapter
);
2986 if (hw
->mac
.type
>= e1000_i350
)
2987 adapter
->flags
&= ~IGB_FLAG_DMAC
;
2989 set_bit(__IGB_DOWN
, &adapter
->state
);
2994 * igb_open - Called when a network interface is made active
2995 * @netdev: network interface device structure
2997 * Returns 0 on success, negative value on failure
2999 * The open entry point is called when a network interface is made
3000 * active by the system (IFF_UP). At this point all resources needed
3001 * for transmit and receive operations are allocated, the interrupt
3002 * handler is registered with the OS, the watchdog timer is started,
3003 * and the stack is notified that the interface is ready.
3005 static int __igb_open(struct net_device
*netdev
, bool resuming
)
3007 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3008 struct e1000_hw
*hw
= &adapter
->hw
;
3009 struct pci_dev
*pdev
= adapter
->pdev
;
3013 /* disallow open during test */
3014 if (test_bit(__IGB_TESTING
, &adapter
->state
)) {
3020 pm_runtime_get_sync(&pdev
->dev
);
3022 netif_carrier_off(netdev
);
3024 /* allocate transmit descriptors */
3025 err
= igb_setup_all_tx_resources(adapter
);
3029 /* allocate receive descriptors */
3030 err
= igb_setup_all_rx_resources(adapter
);
3034 igb_power_up_link(adapter
);
3036 /* before we allocate an interrupt, we must be ready to handle it.
3037 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
3038 * as soon as we call pci_request_irq, so we have to setup our
3039 * clean_rx handler before we do so.
3041 igb_configure(adapter
);
3043 err
= igb_request_irq(adapter
);
3047 /* Notify the stack of the actual queue counts. */
3048 err
= netif_set_real_num_tx_queues(adapter
->netdev
,
3049 adapter
->num_tx_queues
);
3051 goto err_set_queues
;
3053 err
= netif_set_real_num_rx_queues(adapter
->netdev
,
3054 adapter
->num_rx_queues
);
3056 goto err_set_queues
;
3058 /* From here on the code is the same as igb_up() */
3059 clear_bit(__IGB_DOWN
, &adapter
->state
);
3061 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
3062 napi_enable(&(adapter
->q_vector
[i
]->napi
));
3064 /* Clear any pending interrupts. */
3067 igb_irq_enable(adapter
);
3069 /* notify VFs that reset has been completed */
3070 if (adapter
->vfs_allocated_count
) {
3071 u32 reg_data
= rd32(E1000_CTRL_EXT
);
3073 reg_data
|= E1000_CTRL_EXT_PFRSTD
;
3074 wr32(E1000_CTRL_EXT
, reg_data
);
3077 netif_tx_start_all_queues(netdev
);
3080 pm_runtime_put(&pdev
->dev
);
3082 /* start the watchdog. */
3083 hw
->mac
.get_link_status
= 1;
3084 schedule_work(&adapter
->watchdog_task
);
3089 igb_free_irq(adapter
);
3091 igb_release_hw_control(adapter
);
3092 igb_power_down_link(adapter
);
3093 igb_free_all_rx_resources(adapter
);
3095 igb_free_all_tx_resources(adapter
);
3099 pm_runtime_put(&pdev
->dev
);
3104 static int igb_open(struct net_device
*netdev
)
3106 return __igb_open(netdev
, false);
3110 * igb_close - Disables a network interface
3111 * @netdev: network interface device structure
3113 * Returns 0, this is not allowed to fail
3115 * The close entry point is called when an interface is de-activated
3116 * by the OS. The hardware is still under the driver's control, but
3117 * needs to be disabled. A global MAC reset is issued to stop the
3118 * hardware, and all transmit and receive resources are freed.
3120 static int __igb_close(struct net_device
*netdev
, bool suspending
)
3122 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3123 struct pci_dev
*pdev
= adapter
->pdev
;
3125 WARN_ON(test_bit(__IGB_RESETTING
, &adapter
->state
));
3128 pm_runtime_get_sync(&pdev
->dev
);
3131 igb_free_irq(adapter
);
3133 igb_free_all_tx_resources(adapter
);
3134 igb_free_all_rx_resources(adapter
);
3137 pm_runtime_put_sync(&pdev
->dev
);
3141 static int igb_close(struct net_device
*netdev
)
3143 return __igb_close(netdev
, false);
3147 * igb_setup_tx_resources - allocate Tx resources (Descriptors)
3148 * @tx_ring: tx descriptor ring (for a specific queue) to setup
3150 * Return 0 on success, negative on failure
3152 int igb_setup_tx_resources(struct igb_ring
*tx_ring
)
3154 struct device
*dev
= tx_ring
->dev
;
3157 size
= sizeof(struct igb_tx_buffer
) * tx_ring
->count
;
3159 tx_ring
->tx_buffer_info
= vzalloc(size
);
3160 if (!tx_ring
->tx_buffer_info
)
3163 /* round up to nearest 4K */
3164 tx_ring
->size
= tx_ring
->count
* sizeof(union e1000_adv_tx_desc
);
3165 tx_ring
->size
= ALIGN(tx_ring
->size
, 4096);
3167 tx_ring
->desc
= dma_alloc_coherent(dev
, tx_ring
->size
,
3168 &tx_ring
->dma
, GFP_KERNEL
);
3172 tx_ring
->next_to_use
= 0;
3173 tx_ring
->next_to_clean
= 0;
3178 vfree(tx_ring
->tx_buffer_info
);
3179 tx_ring
->tx_buffer_info
= NULL
;
3180 dev_err(dev
, "Unable to allocate memory for the Tx descriptor ring\n");
3185 * igb_setup_all_tx_resources - wrapper to allocate Tx resources
3186 * (Descriptors) for all queues
3187 * @adapter: board private structure
3189 * Return 0 on success, negative on failure
3191 static int igb_setup_all_tx_resources(struct igb_adapter
*adapter
)
3193 struct pci_dev
*pdev
= adapter
->pdev
;
3196 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
3197 err
= igb_setup_tx_resources(adapter
->tx_ring
[i
]);
3200 "Allocation for Tx Queue %u failed\n", i
);
3201 for (i
--; i
>= 0; i
--)
3202 igb_free_tx_resources(adapter
->tx_ring
[i
]);
3211 * igb_setup_tctl - configure the transmit control registers
3212 * @adapter: Board private structure
3214 void igb_setup_tctl(struct igb_adapter
*adapter
)
3216 struct e1000_hw
*hw
= &adapter
->hw
;
3219 /* disable queue 0 which is enabled by default on 82575 and 82576 */
3220 wr32(E1000_TXDCTL(0), 0);
3222 /* Program the Transmit Control Register */
3223 tctl
= rd32(E1000_TCTL
);
3224 tctl
&= ~E1000_TCTL_CT
;
3225 tctl
|= E1000_TCTL_PSP
| E1000_TCTL_RTLC
|
3226 (E1000_COLLISION_THRESHOLD
<< E1000_CT_SHIFT
);
3228 igb_config_collision_dist(hw
);
3230 /* Enable transmits */
3231 tctl
|= E1000_TCTL_EN
;
3233 wr32(E1000_TCTL
, tctl
);
3237 * igb_configure_tx_ring - Configure transmit ring after Reset
3238 * @adapter: board private structure
3239 * @ring: tx ring to configure
3241 * Configure a transmit ring after a reset.
3243 void igb_configure_tx_ring(struct igb_adapter
*adapter
,
3244 struct igb_ring
*ring
)
3246 struct e1000_hw
*hw
= &adapter
->hw
;
3248 u64 tdba
= ring
->dma
;
3249 int reg_idx
= ring
->reg_idx
;
3251 /* disable the queue */
3252 wr32(E1000_TXDCTL(reg_idx
), 0);
3256 wr32(E1000_TDLEN(reg_idx
),
3257 ring
->count
* sizeof(union e1000_adv_tx_desc
));
3258 wr32(E1000_TDBAL(reg_idx
),
3259 tdba
& 0x00000000ffffffffULL
);
3260 wr32(E1000_TDBAH(reg_idx
), tdba
>> 32);
3262 ring
->tail
= hw
->hw_addr
+ E1000_TDT(reg_idx
);
3263 wr32(E1000_TDH(reg_idx
), 0);
3264 writel(0, ring
->tail
);
3266 txdctl
|= IGB_TX_PTHRESH
;
3267 txdctl
|= IGB_TX_HTHRESH
<< 8;
3268 txdctl
|= IGB_TX_WTHRESH
<< 16;
3270 txdctl
|= E1000_TXDCTL_QUEUE_ENABLE
;
3271 wr32(E1000_TXDCTL(reg_idx
), txdctl
);
3275 * igb_configure_tx - Configure transmit Unit after Reset
3276 * @adapter: board private structure
3278 * Configure the Tx unit of the MAC after a reset.
3280 static void igb_configure_tx(struct igb_adapter
*adapter
)
3284 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
3285 igb_configure_tx_ring(adapter
, adapter
->tx_ring
[i
]);
3289 * igb_setup_rx_resources - allocate Rx resources (Descriptors)
3290 * @rx_ring: Rx descriptor ring (for a specific queue) to setup
3292 * Returns 0 on success, negative on failure
3294 int igb_setup_rx_resources(struct igb_ring
*rx_ring
)
3296 struct device
*dev
= rx_ring
->dev
;
3299 size
= sizeof(struct igb_rx_buffer
) * rx_ring
->count
;
3301 rx_ring
->rx_buffer_info
= vzalloc(size
);
3302 if (!rx_ring
->rx_buffer_info
)
3305 /* Round up to nearest 4K */
3306 rx_ring
->size
= rx_ring
->count
* sizeof(union e1000_adv_rx_desc
);
3307 rx_ring
->size
= ALIGN(rx_ring
->size
, 4096);
3309 rx_ring
->desc
= dma_alloc_coherent(dev
, rx_ring
->size
,
3310 &rx_ring
->dma
, GFP_KERNEL
);
3314 rx_ring
->next_to_alloc
= 0;
3315 rx_ring
->next_to_clean
= 0;
3316 rx_ring
->next_to_use
= 0;
3321 vfree(rx_ring
->rx_buffer_info
);
3322 rx_ring
->rx_buffer_info
= NULL
;
3323 dev_err(dev
, "Unable to allocate memory for the Rx descriptor ring\n");
3328 * igb_setup_all_rx_resources - wrapper to allocate Rx resources
3329 * (Descriptors) for all queues
3330 * @adapter: board private structure
3332 * Return 0 on success, negative on failure
3334 static int igb_setup_all_rx_resources(struct igb_adapter
*adapter
)
3336 struct pci_dev
*pdev
= adapter
->pdev
;
3339 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3340 err
= igb_setup_rx_resources(adapter
->rx_ring
[i
]);
3343 "Allocation for Rx Queue %u failed\n", i
);
3344 for (i
--; i
>= 0; i
--)
3345 igb_free_rx_resources(adapter
->rx_ring
[i
]);
3354 * igb_setup_mrqc - configure the multiple receive queue control registers
3355 * @adapter: Board private structure
3357 static void igb_setup_mrqc(struct igb_adapter
*adapter
)
3359 struct e1000_hw
*hw
= &adapter
->hw
;
3361 u32 j
, num_rx_queues
;
3364 netdev_rss_key_fill(rss_key
, sizeof(rss_key
));
3365 for (j
= 0; j
< 10; j
++)
3366 wr32(E1000_RSSRK(j
), rss_key
[j
]);
3368 num_rx_queues
= adapter
->rss_queues
;
3370 switch (hw
->mac
.type
) {
3372 /* 82576 supports 2 RSS queues for SR-IOV */
3373 if (adapter
->vfs_allocated_count
)
3380 if (adapter
->rss_indir_tbl_init
!= num_rx_queues
) {
3381 for (j
= 0; j
< IGB_RETA_SIZE
; j
++)
3382 adapter
->rss_indir_tbl
[j
] =
3383 (j
* num_rx_queues
) / IGB_RETA_SIZE
;
3384 adapter
->rss_indir_tbl_init
= num_rx_queues
;
3386 igb_write_rss_indir_tbl(adapter
);
3388 /* Disable raw packet checksumming so that RSS hash is placed in
3389 * descriptor on writeback. No need to enable TCP/UDP/IP checksum
3390 * offloads as they are enabled by default
3392 rxcsum
= rd32(E1000_RXCSUM
);
3393 rxcsum
|= E1000_RXCSUM_PCSD
;
3395 if (adapter
->hw
.mac
.type
>= e1000_82576
)
3396 /* Enable Receive Checksum Offload for SCTP */
3397 rxcsum
|= E1000_RXCSUM_CRCOFL
;
3399 /* Don't need to set TUOFL or IPOFL, they default to 1 */
3400 wr32(E1000_RXCSUM
, rxcsum
);
3402 /* Generate RSS hash based on packet types, TCP/UDP
3403 * port numbers and/or IPv4/v6 src and dst addresses
3405 mrqc
= E1000_MRQC_RSS_FIELD_IPV4
|
3406 E1000_MRQC_RSS_FIELD_IPV4_TCP
|
3407 E1000_MRQC_RSS_FIELD_IPV6
|
3408 E1000_MRQC_RSS_FIELD_IPV6_TCP
|
3409 E1000_MRQC_RSS_FIELD_IPV6_TCP_EX
;
3411 if (adapter
->flags
& IGB_FLAG_RSS_FIELD_IPV4_UDP
)
3412 mrqc
|= E1000_MRQC_RSS_FIELD_IPV4_UDP
;
3413 if (adapter
->flags
& IGB_FLAG_RSS_FIELD_IPV6_UDP
)
3414 mrqc
|= E1000_MRQC_RSS_FIELD_IPV6_UDP
;
3416 /* If VMDq is enabled then we set the appropriate mode for that, else
3417 * we default to RSS so that an RSS hash is calculated per packet even
3418 * if we are only using one queue
3420 if (adapter
->vfs_allocated_count
) {
3421 if (hw
->mac
.type
> e1000_82575
) {
3422 /* Set the default pool for the PF's first queue */
3423 u32 vtctl
= rd32(E1000_VT_CTL
);
3425 vtctl
&= ~(E1000_VT_CTL_DEFAULT_POOL_MASK
|
3426 E1000_VT_CTL_DISABLE_DEF_POOL
);
3427 vtctl
|= adapter
->vfs_allocated_count
<<
3428 E1000_VT_CTL_DEFAULT_POOL_SHIFT
;
3429 wr32(E1000_VT_CTL
, vtctl
);
3431 if (adapter
->rss_queues
> 1)
3432 mrqc
|= E1000_MRQC_ENABLE_VMDQ_RSS_2Q
;
3434 mrqc
|= E1000_MRQC_ENABLE_VMDQ
;
3436 if (hw
->mac
.type
!= e1000_i211
)
3437 mrqc
|= E1000_MRQC_ENABLE_RSS_4Q
;
3439 igb_vmm_control(adapter
);
3441 wr32(E1000_MRQC
, mrqc
);
3445 * igb_setup_rctl - configure the receive control registers
3446 * @adapter: Board private structure
3448 void igb_setup_rctl(struct igb_adapter
*adapter
)
3450 struct e1000_hw
*hw
= &adapter
->hw
;
3453 rctl
= rd32(E1000_RCTL
);
3455 rctl
&= ~(3 << E1000_RCTL_MO_SHIFT
);
3456 rctl
&= ~(E1000_RCTL_LBM_TCVR
| E1000_RCTL_LBM_MAC
);
3458 rctl
|= E1000_RCTL_EN
| E1000_RCTL_BAM
| E1000_RCTL_RDMTS_HALF
|
3459 (hw
->mac
.mc_filter_type
<< E1000_RCTL_MO_SHIFT
);
3461 /* enable stripping of CRC. It's unlikely this will break BMC
3462 * redirection as it did with e1000. Newer features require
3463 * that the HW strips the CRC.
3465 rctl
|= E1000_RCTL_SECRC
;
3467 /* disable store bad packets and clear size bits. */
3468 rctl
&= ~(E1000_RCTL_SBP
| E1000_RCTL_SZ_256
);
3470 /* enable LPE to prevent packets larger than max_frame_size */
3471 rctl
|= E1000_RCTL_LPE
;
3473 /* disable queue 0 to prevent tail write w/o re-config */
3474 wr32(E1000_RXDCTL(0), 0);
3476 /* Attention!!! For SR-IOV PF driver operations you must enable
3477 * queue drop for all VF and PF queues to prevent head of line blocking
3478 * if an un-trusted VF does not provide descriptors to hardware.
3480 if (adapter
->vfs_allocated_count
) {
3481 /* set all queue drop enable bits */
3482 wr32(E1000_QDE
, ALL_QUEUES
);
3485 /* This is useful for sniffing bad packets. */
3486 if (adapter
->netdev
->features
& NETIF_F_RXALL
) {
3487 /* UPE and MPE will be handled by normal PROMISC logic
3488 * in e1000e_set_rx_mode
3490 rctl
|= (E1000_RCTL_SBP
| /* Receive bad packets */
3491 E1000_RCTL_BAM
| /* RX All Bcast Pkts */
3492 E1000_RCTL_PMCF
); /* RX All MAC Ctrl Pkts */
3494 rctl
&= ~(E1000_RCTL_VFE
| /* Disable VLAN filter */
3495 E1000_RCTL_DPF
| /* Allow filtered pause */
3496 E1000_RCTL_CFIEN
); /* Dis VLAN CFIEN Filter */
3497 /* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3498 * and that breaks VLANs.
3502 wr32(E1000_RCTL
, rctl
);
3505 static inline int igb_set_vf_rlpml(struct igb_adapter
*adapter
, int size
,
3508 struct e1000_hw
*hw
= &adapter
->hw
;
3511 /* if it isn't the PF check to see if VFs are enabled and
3512 * increase the size to support vlan tags
3514 if (vfn
< adapter
->vfs_allocated_count
&&
3515 adapter
->vf_data
[vfn
].vlans_enabled
)
3516 size
+= VLAN_TAG_SIZE
;
3518 vmolr
= rd32(E1000_VMOLR(vfn
));
3519 vmolr
&= ~E1000_VMOLR_RLPML_MASK
;
3520 vmolr
|= size
| E1000_VMOLR_LPE
;
3521 wr32(E1000_VMOLR(vfn
), vmolr
);
3527 * igb_rlpml_set - set maximum receive packet size
3528 * @adapter: board private structure
3530 * Configure maximum receivable packet size.
3532 static void igb_rlpml_set(struct igb_adapter
*adapter
)
3534 u32 max_frame_size
= adapter
->max_frame_size
;
3535 struct e1000_hw
*hw
= &adapter
->hw
;
3536 u16 pf_id
= adapter
->vfs_allocated_count
;
3539 igb_set_vf_rlpml(adapter
, max_frame_size
, pf_id
);
3540 /* If we're in VMDQ or SR-IOV mode, then set global RLPML
3541 * to our max jumbo frame size, in case we need to enable
3542 * jumbo frames on one of the rings later.
3543 * This will not pass over-length frames into the default
3544 * queue because it's gated by the VMOLR.RLPML.
3546 max_frame_size
= MAX_JUMBO_FRAME_SIZE
;
3549 wr32(E1000_RLPML
, max_frame_size
);
3552 static inline void igb_set_vmolr(struct igb_adapter
*adapter
,
3555 struct e1000_hw
*hw
= &adapter
->hw
;
3558 /* This register exists only on 82576 and newer so if we are older then
3559 * we should exit and do nothing
3561 if (hw
->mac
.type
< e1000_82576
)
3564 vmolr
= rd32(E1000_VMOLR(vfn
));
3565 vmolr
|= E1000_VMOLR_STRVLAN
; /* Strip vlan tags */
3566 if (hw
->mac
.type
== e1000_i350
) {
3569 dvmolr
= rd32(E1000_DVMOLR(vfn
));
3570 dvmolr
|= E1000_DVMOLR_STRVLAN
;
3571 wr32(E1000_DVMOLR(vfn
), dvmolr
);
3574 vmolr
|= E1000_VMOLR_AUPE
; /* Accept untagged packets */
3576 vmolr
&= ~(E1000_VMOLR_AUPE
); /* Tagged packets ONLY */
3578 /* clear all bits that might not be set */
3579 vmolr
&= ~(E1000_VMOLR_BAM
| E1000_VMOLR_RSSE
);
3581 if (adapter
->rss_queues
> 1 && vfn
== adapter
->vfs_allocated_count
)
3582 vmolr
|= E1000_VMOLR_RSSE
; /* enable RSS */
3583 /* for VMDq only allow the VFs and pool 0 to accept broadcast and
3586 if (vfn
<= adapter
->vfs_allocated_count
)
3587 vmolr
|= E1000_VMOLR_BAM
; /* Accept broadcast */
3589 wr32(E1000_VMOLR(vfn
), vmolr
);
3593 * igb_configure_rx_ring - Configure a receive ring after Reset
3594 * @adapter: board private structure
3595 * @ring: receive ring to be configured
3597 * Configure the Rx unit of the MAC after a reset.
3599 void igb_configure_rx_ring(struct igb_adapter
*adapter
,
3600 struct igb_ring
*ring
)
3602 struct e1000_hw
*hw
= &adapter
->hw
;
3603 u64 rdba
= ring
->dma
;
3604 int reg_idx
= ring
->reg_idx
;
3605 u32 srrctl
= 0, rxdctl
= 0;
3607 /* disable the queue */
3608 wr32(E1000_RXDCTL(reg_idx
), 0);
3610 /* Set DMA base address registers */
3611 wr32(E1000_RDBAL(reg_idx
),
3612 rdba
& 0x00000000ffffffffULL
);
3613 wr32(E1000_RDBAH(reg_idx
), rdba
>> 32);
3614 wr32(E1000_RDLEN(reg_idx
),
3615 ring
->count
* sizeof(union e1000_adv_rx_desc
));
3617 /* initialize head and tail */
3618 ring
->tail
= hw
->hw_addr
+ E1000_RDT(reg_idx
);
3619 wr32(E1000_RDH(reg_idx
), 0);
3620 writel(0, ring
->tail
);
3622 /* set descriptor configuration */
3623 srrctl
= IGB_RX_HDR_LEN
<< E1000_SRRCTL_BSIZEHDRSIZE_SHIFT
;
3624 srrctl
|= IGB_RX_BUFSZ
>> E1000_SRRCTL_BSIZEPKT_SHIFT
;
3625 srrctl
|= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF
;
3626 if (hw
->mac
.type
>= e1000_82580
)
3627 srrctl
|= E1000_SRRCTL_TIMESTAMP
;
3628 /* Only set Drop Enable if we are supporting multiple queues */
3629 if (adapter
->vfs_allocated_count
|| adapter
->num_rx_queues
> 1)
3630 srrctl
|= E1000_SRRCTL_DROP_EN
;
3632 wr32(E1000_SRRCTL(reg_idx
), srrctl
);
3634 /* set filtering for VMDQ pools */
3635 igb_set_vmolr(adapter
, reg_idx
& 0x7, true);
3637 rxdctl
|= IGB_RX_PTHRESH
;
3638 rxdctl
|= IGB_RX_HTHRESH
<< 8;
3639 rxdctl
|= IGB_RX_WTHRESH
<< 16;
3641 /* enable receive descriptor fetching */
3642 rxdctl
|= E1000_RXDCTL_QUEUE_ENABLE
;
3643 wr32(E1000_RXDCTL(reg_idx
), rxdctl
);
3647 * igb_configure_rx - Configure receive Unit after Reset
3648 * @adapter: board private structure
3650 * Configure the Rx unit of the MAC after a reset.
3652 static void igb_configure_rx(struct igb_adapter
*adapter
)
3656 /* set UTA to appropriate mode */
3657 igb_set_uta(adapter
);
3659 /* set the correct pool for the PF default MAC address in entry 0 */
3660 igb_rar_set_qsel(adapter
, adapter
->hw
.mac
.addr
, 0,
3661 adapter
->vfs_allocated_count
);
3663 /* Setup the HW Rx Head and Tail Descriptor Pointers and
3664 * the Base and Length of the Rx Descriptor Ring
3666 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3667 igb_configure_rx_ring(adapter
, adapter
->rx_ring
[i
]);
3671 * igb_free_tx_resources - Free Tx Resources per Queue
3672 * @tx_ring: Tx descriptor ring for a specific queue
3674 * Free all transmit software resources
3676 void igb_free_tx_resources(struct igb_ring
*tx_ring
)
3678 igb_clean_tx_ring(tx_ring
);
3680 vfree(tx_ring
->tx_buffer_info
);
3681 tx_ring
->tx_buffer_info
= NULL
;
3683 /* if not set, then don't free */
3687 dma_free_coherent(tx_ring
->dev
, tx_ring
->size
,
3688 tx_ring
->desc
, tx_ring
->dma
);
3690 tx_ring
->desc
= NULL
;
3694 * igb_free_all_tx_resources - Free Tx Resources for All Queues
3695 * @adapter: board private structure
3697 * Free all transmit software resources
3699 static void igb_free_all_tx_resources(struct igb_adapter
*adapter
)
3703 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
3704 if (adapter
->tx_ring
[i
])
3705 igb_free_tx_resources(adapter
->tx_ring
[i
]);
3708 void igb_unmap_and_free_tx_resource(struct igb_ring
*ring
,
3709 struct igb_tx_buffer
*tx_buffer
)
3711 if (tx_buffer
->skb
) {
3712 dev_kfree_skb_any(tx_buffer
->skb
);
3713 if (dma_unmap_len(tx_buffer
, len
))
3714 dma_unmap_single(ring
->dev
,
3715 dma_unmap_addr(tx_buffer
, dma
),
3716 dma_unmap_len(tx_buffer
, len
),
3718 } else if (dma_unmap_len(tx_buffer
, len
)) {
3719 dma_unmap_page(ring
->dev
,
3720 dma_unmap_addr(tx_buffer
, dma
),
3721 dma_unmap_len(tx_buffer
, len
),
3724 tx_buffer
->next_to_watch
= NULL
;
3725 tx_buffer
->skb
= NULL
;
3726 dma_unmap_len_set(tx_buffer
, len
, 0);
3727 /* buffer_info must be completely set up in the transmit path */
3731 * igb_clean_tx_ring - Free Tx Buffers
3732 * @tx_ring: ring to be cleaned
3734 static void igb_clean_tx_ring(struct igb_ring
*tx_ring
)
3736 struct igb_tx_buffer
*buffer_info
;
3740 if (!tx_ring
->tx_buffer_info
)
3742 /* Free all the Tx ring sk_buffs */
3744 for (i
= 0; i
< tx_ring
->count
; i
++) {
3745 buffer_info
= &tx_ring
->tx_buffer_info
[i
];
3746 igb_unmap_and_free_tx_resource(tx_ring
, buffer_info
);
3749 netdev_tx_reset_queue(txring_txq(tx_ring
));
3751 size
= sizeof(struct igb_tx_buffer
) * tx_ring
->count
;
3752 memset(tx_ring
->tx_buffer_info
, 0, size
);
3754 /* Zero out the descriptor ring */
3755 memset(tx_ring
->desc
, 0, tx_ring
->size
);
3757 tx_ring
->next_to_use
= 0;
3758 tx_ring
->next_to_clean
= 0;
3762 * igb_clean_all_tx_rings - Free Tx Buffers for all queues
3763 * @adapter: board private structure
3765 static void igb_clean_all_tx_rings(struct igb_adapter
*adapter
)
3769 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
3770 if (adapter
->tx_ring
[i
])
3771 igb_clean_tx_ring(adapter
->tx_ring
[i
]);
3775 * igb_free_rx_resources - Free Rx Resources
3776 * @rx_ring: ring to clean the resources from
3778 * Free all receive software resources
3780 void igb_free_rx_resources(struct igb_ring
*rx_ring
)
3782 igb_clean_rx_ring(rx_ring
);
3784 vfree(rx_ring
->rx_buffer_info
);
3785 rx_ring
->rx_buffer_info
= NULL
;
3787 /* if not set, then don't free */
3791 dma_free_coherent(rx_ring
->dev
, rx_ring
->size
,
3792 rx_ring
->desc
, rx_ring
->dma
);
3794 rx_ring
->desc
= NULL
;
3798 * igb_free_all_rx_resources - Free Rx Resources for All Queues
3799 * @adapter: board private structure
3801 * Free all receive software resources
3803 static void igb_free_all_rx_resources(struct igb_adapter
*adapter
)
3807 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3808 if (adapter
->rx_ring
[i
])
3809 igb_free_rx_resources(adapter
->rx_ring
[i
]);
3813 * igb_clean_rx_ring - Free Rx Buffers per Queue
3814 * @rx_ring: ring to free buffers from
3816 static void igb_clean_rx_ring(struct igb_ring
*rx_ring
)
3822 dev_kfree_skb(rx_ring
->skb
);
3823 rx_ring
->skb
= NULL
;
3825 if (!rx_ring
->rx_buffer_info
)
3828 /* Free all the Rx ring sk_buffs */
3829 for (i
= 0; i
< rx_ring
->count
; i
++) {
3830 struct igb_rx_buffer
*buffer_info
= &rx_ring
->rx_buffer_info
[i
];
3832 if (!buffer_info
->page
)
3835 dma_unmap_page(rx_ring
->dev
,
3839 __free_page(buffer_info
->page
);
3841 buffer_info
->page
= NULL
;
3844 size
= sizeof(struct igb_rx_buffer
) * rx_ring
->count
;
3845 memset(rx_ring
->rx_buffer_info
, 0, size
);
3847 /* Zero out the descriptor ring */
3848 memset(rx_ring
->desc
, 0, rx_ring
->size
);
3850 rx_ring
->next_to_alloc
= 0;
3851 rx_ring
->next_to_clean
= 0;
3852 rx_ring
->next_to_use
= 0;
3856 * igb_clean_all_rx_rings - Free Rx Buffers for all queues
3857 * @adapter: board private structure
3859 static void igb_clean_all_rx_rings(struct igb_adapter
*adapter
)
3863 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3864 if (adapter
->rx_ring
[i
])
3865 igb_clean_rx_ring(adapter
->rx_ring
[i
]);
3869 * igb_set_mac - Change the Ethernet Address of the NIC
3870 * @netdev: network interface device structure
3871 * @p: pointer to an address structure
3873 * Returns 0 on success, negative on failure
3875 static int igb_set_mac(struct net_device
*netdev
, void *p
)
3877 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3878 struct e1000_hw
*hw
= &adapter
->hw
;
3879 struct sockaddr
*addr
= p
;
3881 if (!is_valid_ether_addr(addr
->sa_data
))
3882 return -EADDRNOTAVAIL
;
3884 memcpy(netdev
->dev_addr
, addr
->sa_data
, netdev
->addr_len
);
3885 memcpy(hw
->mac
.addr
, addr
->sa_data
, netdev
->addr_len
);
3887 /* set the correct pool for the new PF MAC address in entry 0 */
3888 igb_rar_set_qsel(adapter
, hw
->mac
.addr
, 0,
3889 adapter
->vfs_allocated_count
);
3895 * igb_write_mc_addr_list - write multicast addresses to MTA
3896 * @netdev: network interface device structure
3898 * Writes multicast address list to the MTA hash table.
3899 * Returns: -ENOMEM on failure
3900 * 0 on no addresses written
3901 * X on writing X addresses to MTA
3903 static int igb_write_mc_addr_list(struct net_device
*netdev
)
3905 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3906 struct e1000_hw
*hw
= &adapter
->hw
;
3907 struct netdev_hw_addr
*ha
;
3911 if (netdev_mc_empty(netdev
)) {
3912 /* nothing to program, so clear mc list */
3913 igb_update_mc_addr_list(hw
, NULL
, 0);
3914 igb_restore_vf_multicasts(adapter
);
3918 mta_list
= kzalloc(netdev_mc_count(netdev
) * 6, GFP_ATOMIC
);
3922 /* The shared function expects a packed array of only addresses. */
3924 netdev_for_each_mc_addr(ha
, netdev
)
3925 memcpy(mta_list
+ (i
++ * ETH_ALEN
), ha
->addr
, ETH_ALEN
);
3927 igb_update_mc_addr_list(hw
, mta_list
, i
);
3930 return netdev_mc_count(netdev
);
3934 * igb_write_uc_addr_list - write unicast addresses to RAR table
3935 * @netdev: network interface device structure
3937 * Writes unicast address list to the RAR table.
3938 * Returns: -ENOMEM on failure/insufficient address space
3939 * 0 on no addresses written
3940 * X on writing X addresses to the RAR table
3942 static int igb_write_uc_addr_list(struct net_device
*netdev
)
3944 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3945 struct e1000_hw
*hw
= &adapter
->hw
;
3946 unsigned int vfn
= adapter
->vfs_allocated_count
;
3947 unsigned int rar_entries
= hw
->mac
.rar_entry_count
- (vfn
+ 1);
3950 /* return ENOMEM indicating insufficient memory for addresses */
3951 if (netdev_uc_count(netdev
) > rar_entries
)
3954 if (!netdev_uc_empty(netdev
) && rar_entries
) {
3955 struct netdev_hw_addr
*ha
;
3957 netdev_for_each_uc_addr(ha
, netdev
) {
3960 igb_rar_set_qsel(adapter
, ha
->addr
,
3966 /* write the addresses in reverse order to avoid write combining */
3967 for (; rar_entries
> 0 ; rar_entries
--) {
3968 wr32(E1000_RAH(rar_entries
), 0);
3969 wr32(E1000_RAL(rar_entries
), 0);
3977 * igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
3978 * @netdev: network interface device structure
3980 * The set_rx_mode entry point is called whenever the unicast or multicast
3981 * address lists or the network interface flags are updated. This routine is
3982 * responsible for configuring the hardware for proper unicast, multicast,
3983 * promiscuous mode, and all-multi behavior.
3985 static void igb_set_rx_mode(struct net_device
*netdev
)
3987 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3988 struct e1000_hw
*hw
= &adapter
->hw
;
3989 unsigned int vfn
= adapter
->vfs_allocated_count
;
3990 u32 rctl
, vmolr
= 0;
3993 /* Check for Promiscuous and All Multicast modes */
3994 rctl
= rd32(E1000_RCTL
);
3996 /* clear the effected bits */
3997 rctl
&= ~(E1000_RCTL_UPE
| E1000_RCTL_MPE
| E1000_RCTL_VFE
);
3999 if (netdev
->flags
& IFF_PROMISC
) {
4000 /* retain VLAN HW filtering if in VT mode */
4001 if (adapter
->vfs_allocated_count
)
4002 rctl
|= E1000_RCTL_VFE
;
4003 rctl
|= (E1000_RCTL_UPE
| E1000_RCTL_MPE
);
4004 vmolr
|= (E1000_VMOLR_ROPE
| E1000_VMOLR_MPME
);
4006 if (netdev
->flags
& IFF_ALLMULTI
) {
4007 rctl
|= E1000_RCTL_MPE
;
4008 vmolr
|= E1000_VMOLR_MPME
;
4010 /* Write addresses to the MTA, if the attempt fails
4011 * then we should just turn on promiscuous mode so
4012 * that we can at least receive multicast traffic
4014 count
= igb_write_mc_addr_list(netdev
);
4016 rctl
|= E1000_RCTL_MPE
;
4017 vmolr
|= E1000_VMOLR_MPME
;
4019 vmolr
|= E1000_VMOLR_ROMPE
;
4022 /* Write addresses to available RAR registers, if there is not
4023 * sufficient space to store all the addresses then enable
4024 * unicast promiscuous mode
4026 count
= igb_write_uc_addr_list(netdev
);
4028 rctl
|= E1000_RCTL_UPE
;
4029 vmolr
|= E1000_VMOLR_ROPE
;
4031 rctl
|= E1000_RCTL_VFE
;
4033 wr32(E1000_RCTL
, rctl
);
4035 /* In order to support SR-IOV and eventually VMDq it is necessary to set
4036 * the VMOLR to enable the appropriate modes. Without this workaround
4037 * we will have issues with VLAN tag stripping not being done for frames
4038 * that are only arriving because we are the default pool
4040 if ((hw
->mac
.type
< e1000_82576
) || (hw
->mac
.type
> e1000_i350
))
4043 vmolr
|= rd32(E1000_VMOLR(vfn
)) &
4044 ~(E1000_VMOLR_ROPE
| E1000_VMOLR_MPME
| E1000_VMOLR_ROMPE
);
4045 wr32(E1000_VMOLR(vfn
), vmolr
);
4046 igb_restore_vf_multicasts(adapter
);
4049 static void igb_check_wvbr(struct igb_adapter
*adapter
)
4051 struct e1000_hw
*hw
= &adapter
->hw
;
4054 switch (hw
->mac
.type
) {
4057 wvbr
= rd32(E1000_WVBR
);
4065 adapter
->wvbr
|= wvbr
;
4068 #define IGB_STAGGERED_QUEUE_OFFSET 8
4070 static void igb_spoof_check(struct igb_adapter
*adapter
)
4077 for (j
= 0; j
< adapter
->vfs_allocated_count
; j
++) {
4078 if (adapter
->wvbr
& (1 << j
) ||
4079 adapter
->wvbr
& (1 << (j
+ IGB_STAGGERED_QUEUE_OFFSET
))) {
4080 dev_warn(&adapter
->pdev
->dev
,
4081 "Spoof event(s) detected on VF %d\n", j
);
4084 (1 << (j
+ IGB_STAGGERED_QUEUE_OFFSET
)));
4089 /* Need to wait a few seconds after link up to get diagnostic information from
4092 static void igb_update_phy_info(unsigned long data
)
4094 struct igb_adapter
*adapter
= (struct igb_adapter
*) data
;
4095 igb_get_phy_info(&adapter
->hw
);
4099 * igb_has_link - check shared code for link and determine up/down
4100 * @adapter: pointer to driver private info
4102 bool igb_has_link(struct igb_adapter
*adapter
)
4104 struct e1000_hw
*hw
= &adapter
->hw
;
4105 bool link_active
= false;
4107 /* get_link_status is set on LSC (link status) interrupt or
4108 * rx sequence error interrupt. get_link_status will stay
4109 * false until the e1000_check_for_link establishes link
4110 * for copper adapters ONLY
4112 switch (hw
->phy
.media_type
) {
4113 case e1000_media_type_copper
:
4114 if (!hw
->mac
.get_link_status
)
4116 case e1000_media_type_internal_serdes
:
4117 hw
->mac
.ops
.check_for_link(hw
);
4118 link_active
= !hw
->mac
.get_link_status
;
4121 case e1000_media_type_unknown
:
4125 if (((hw
->mac
.type
== e1000_i210
) ||
4126 (hw
->mac
.type
== e1000_i211
)) &&
4127 (hw
->phy
.id
== I210_I_PHY_ID
)) {
4128 if (!netif_carrier_ok(adapter
->netdev
)) {
4129 adapter
->flags
&= ~IGB_FLAG_NEED_LINK_UPDATE
;
4130 } else if (!(adapter
->flags
& IGB_FLAG_NEED_LINK_UPDATE
)) {
4131 adapter
->flags
|= IGB_FLAG_NEED_LINK_UPDATE
;
4132 adapter
->link_check_timeout
= jiffies
;
4139 static bool igb_thermal_sensor_event(struct e1000_hw
*hw
, u32 event
)
4142 u32 ctrl_ext
, thstat
;
4144 /* check for thermal sensor event on i350 copper only */
4145 if (hw
->mac
.type
== e1000_i350
) {
4146 thstat
= rd32(E1000_THSTAT
);
4147 ctrl_ext
= rd32(E1000_CTRL_EXT
);
4149 if ((hw
->phy
.media_type
== e1000_media_type_copper
) &&
4150 !(ctrl_ext
& E1000_CTRL_EXT_LINK_MODE_SGMII
))
4151 ret
= !!(thstat
& event
);
4158 * igb_check_lvmmc - check for malformed packets received
4159 * and indicated in LVMMC register
4160 * @adapter: pointer to adapter
4162 static void igb_check_lvmmc(struct igb_adapter
*adapter
)
4164 struct e1000_hw
*hw
= &adapter
->hw
;
4167 lvmmc
= rd32(E1000_LVMMC
);
4169 if (unlikely(net_ratelimit())) {
4170 netdev_warn(adapter
->netdev
,
4171 "malformed Tx packet detected and dropped, LVMMC:0x%08x\n",
4178 * igb_watchdog - Timer Call-back
4179 * @data: pointer to adapter cast into an unsigned long
4181 static void igb_watchdog(unsigned long data
)
4183 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
4184 /* Do the rest outside of interrupt context */
4185 schedule_work(&adapter
->watchdog_task
);
4188 static void igb_watchdog_task(struct work_struct
*work
)
4190 struct igb_adapter
*adapter
= container_of(work
,
4193 struct e1000_hw
*hw
= &adapter
->hw
;
4194 struct e1000_phy_info
*phy
= &hw
->phy
;
4195 struct net_device
*netdev
= adapter
->netdev
;
4200 link
= igb_has_link(adapter
);
4202 if (adapter
->flags
& IGB_FLAG_NEED_LINK_UPDATE
) {
4203 if (time_after(jiffies
, (adapter
->link_check_timeout
+ HZ
)))
4204 adapter
->flags
&= ~IGB_FLAG_NEED_LINK_UPDATE
;
4209 /* Force link down if we have fiber to swap to */
4210 if (adapter
->flags
& IGB_FLAG_MAS_ENABLE
) {
4211 if (hw
->phy
.media_type
== e1000_media_type_copper
) {
4212 connsw
= rd32(E1000_CONNSW
);
4213 if (!(connsw
& E1000_CONNSW_AUTOSENSE_EN
))
4218 /* Perform a reset if the media type changed. */
4219 if (hw
->dev_spec
._82575
.media_changed
) {
4220 hw
->dev_spec
._82575
.media_changed
= false;
4221 adapter
->flags
|= IGB_FLAG_MEDIA_RESET
;
4224 /* Cancel scheduled suspend requests. */
4225 pm_runtime_resume(netdev
->dev
.parent
);
4227 if (!netif_carrier_ok(netdev
)) {
4230 hw
->mac
.ops
.get_speed_and_duplex(hw
,
4231 &adapter
->link_speed
,
4232 &adapter
->link_duplex
);
4234 ctrl
= rd32(E1000_CTRL
);
4235 /* Links status message must follow this format */
4237 "igb: %s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
4239 adapter
->link_speed
,
4240 adapter
->link_duplex
== FULL_DUPLEX
?
4242 (ctrl
& E1000_CTRL_TFCE
) &&
4243 (ctrl
& E1000_CTRL_RFCE
) ? "RX/TX" :
4244 (ctrl
& E1000_CTRL_RFCE
) ? "RX" :
4245 (ctrl
& E1000_CTRL_TFCE
) ? "TX" : "None");
4247 /* disable EEE if enabled */
4248 if ((adapter
->flags
& IGB_FLAG_EEE
) &&
4249 (adapter
->link_duplex
== HALF_DUPLEX
)) {
4250 dev_info(&adapter
->pdev
->dev
,
4251 "EEE Disabled: unsupported at half duplex. Re-enable using ethtool when at full duplex.\n");
4252 adapter
->hw
.dev_spec
._82575
.eee_disable
= true;
4253 adapter
->flags
&= ~IGB_FLAG_EEE
;
4256 /* check if SmartSpeed worked */
4257 igb_check_downshift(hw
);
4258 if (phy
->speed_downgraded
)
4259 netdev_warn(netdev
, "Link Speed was downgraded by SmartSpeed\n");
4261 /* check for thermal sensor event */
4262 if (igb_thermal_sensor_event(hw
,
4263 E1000_THSTAT_LINK_THROTTLE
))
4264 netdev_info(netdev
, "The network adapter link speed was downshifted because it overheated\n");
4266 /* adjust timeout factor according to speed/duplex */
4267 adapter
->tx_timeout_factor
= 1;
4268 switch (adapter
->link_speed
) {
4270 adapter
->tx_timeout_factor
= 14;
4273 /* maybe add some timeout factor ? */
4277 netif_carrier_on(netdev
);
4279 igb_ping_all_vfs(adapter
);
4280 igb_check_vf_rate_limit(adapter
);
4282 /* link state has changed, schedule phy info update */
4283 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
4284 mod_timer(&adapter
->phy_info_timer
,
4285 round_jiffies(jiffies
+ 2 * HZ
));
4288 if (netif_carrier_ok(netdev
)) {
4289 adapter
->link_speed
= 0;
4290 adapter
->link_duplex
= 0;
4292 /* check for thermal sensor event */
4293 if (igb_thermal_sensor_event(hw
,
4294 E1000_THSTAT_PWR_DOWN
)) {
4295 netdev_err(netdev
, "The network adapter was stopped because it overheated\n");
4298 /* Links status message must follow this format */
4299 netdev_info(netdev
, "igb: %s NIC Link is Down\n",
4301 netif_carrier_off(netdev
);
4303 igb_ping_all_vfs(adapter
);
4305 /* link state has changed, schedule phy info update */
4306 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
4307 mod_timer(&adapter
->phy_info_timer
,
4308 round_jiffies(jiffies
+ 2 * HZ
));
4310 /* link is down, time to check for alternate media */
4311 if (adapter
->flags
& IGB_FLAG_MAS_ENABLE
) {
4312 igb_check_swap_media(adapter
);
4313 if (adapter
->flags
& IGB_FLAG_MEDIA_RESET
) {
4314 schedule_work(&adapter
->reset_task
);
4315 /* return immediately */
4319 pm_schedule_suspend(netdev
->dev
.parent
,
4322 /* also check for alternate media here */
4323 } else if (!netif_carrier_ok(netdev
) &&
4324 (adapter
->flags
& IGB_FLAG_MAS_ENABLE
)) {
4325 igb_check_swap_media(adapter
);
4326 if (adapter
->flags
& IGB_FLAG_MEDIA_RESET
) {
4327 schedule_work(&adapter
->reset_task
);
4328 /* return immediately */
4334 spin_lock(&adapter
->stats64_lock
);
4335 igb_update_stats(adapter
, &adapter
->stats64
);
4336 spin_unlock(&adapter
->stats64_lock
);
4338 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
4339 struct igb_ring
*tx_ring
= adapter
->tx_ring
[i
];
4340 if (!netif_carrier_ok(netdev
)) {
4341 /* We've lost link, so the controller stops DMA,
4342 * but we've got queued Tx work that's never going
4343 * to get done, so reset controller to flush Tx.
4344 * (Do the reset outside of interrupt context).
4346 if (igb_desc_unused(tx_ring
) + 1 < tx_ring
->count
) {
4347 adapter
->tx_timeout_count
++;
4348 schedule_work(&adapter
->reset_task
);
4349 /* return immediately since reset is imminent */
4354 /* Force detection of hung controller every watchdog period */
4355 set_bit(IGB_RING_FLAG_TX_DETECT_HANG
, &tx_ring
->flags
);
4358 /* Cause software interrupt to ensure Rx ring is cleaned */
4359 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
4362 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
4363 eics
|= adapter
->q_vector
[i
]->eims_value
;
4364 wr32(E1000_EICS
, eics
);
4366 wr32(E1000_ICS
, E1000_ICS_RXDMT0
);
4369 igb_spoof_check(adapter
);
4370 igb_ptp_rx_hang(adapter
);
4372 /* Check LVMMC register on i350/i354 only */
4373 if ((adapter
->hw
.mac
.type
== e1000_i350
) ||
4374 (adapter
->hw
.mac
.type
== e1000_i354
))
4375 igb_check_lvmmc(adapter
);
4377 /* Reset the timer */
4378 if (!test_bit(__IGB_DOWN
, &adapter
->state
)) {
4379 if (adapter
->flags
& IGB_FLAG_NEED_LINK_UPDATE
)
4380 mod_timer(&adapter
->watchdog_timer
,
4381 round_jiffies(jiffies
+ HZ
));
4383 mod_timer(&adapter
->watchdog_timer
,
4384 round_jiffies(jiffies
+ 2 * HZ
));
4388 enum latency_range
{
4392 latency_invalid
= 255
4396 * igb_update_ring_itr - update the dynamic ITR value based on packet size
4397 * @q_vector: pointer to q_vector
4399 * Stores a new ITR value based on strictly on packet size. This
4400 * algorithm is less sophisticated than that used in igb_update_itr,
4401 * due to the difficulty of synchronizing statistics across multiple
4402 * receive rings. The divisors and thresholds used by this function
4403 * were determined based on theoretical maximum wire speed and testing
4404 * data, in order to minimize response time while increasing bulk
4406 * This functionality is controlled by ethtool's coalescing settings.
4407 * NOTE: This function is called only when operating in a multiqueue
4408 * receive environment.
4410 static void igb_update_ring_itr(struct igb_q_vector
*q_vector
)
4412 int new_val
= q_vector
->itr_val
;
4413 int avg_wire_size
= 0;
4414 struct igb_adapter
*adapter
= q_vector
->adapter
;
4415 unsigned int packets
;
4417 /* For non-gigabit speeds, just fix the interrupt rate at 4000
4418 * ints/sec - ITR timer value of 120 ticks.
4420 if (adapter
->link_speed
!= SPEED_1000
) {
4421 new_val
= IGB_4K_ITR
;
4425 packets
= q_vector
->rx
.total_packets
;
4427 avg_wire_size
= q_vector
->rx
.total_bytes
/ packets
;
4429 packets
= q_vector
->tx
.total_packets
;
4431 avg_wire_size
= max_t(u32
, avg_wire_size
,
4432 q_vector
->tx
.total_bytes
/ packets
);
4434 /* if avg_wire_size isn't set no work was done */
4438 /* Add 24 bytes to size to account for CRC, preamble, and gap */
4439 avg_wire_size
+= 24;
4441 /* Don't starve jumbo frames */
4442 avg_wire_size
= min(avg_wire_size
, 3000);
4444 /* Give a little boost to mid-size frames */
4445 if ((avg_wire_size
> 300) && (avg_wire_size
< 1200))
4446 new_val
= avg_wire_size
/ 3;
4448 new_val
= avg_wire_size
/ 2;
4450 /* conservative mode (itr 3) eliminates the lowest_latency setting */
4451 if (new_val
< IGB_20K_ITR
&&
4452 ((q_vector
->rx
.ring
&& adapter
->rx_itr_setting
== 3) ||
4453 (!q_vector
->rx
.ring
&& adapter
->tx_itr_setting
== 3)))
4454 new_val
= IGB_20K_ITR
;
4457 if (new_val
!= q_vector
->itr_val
) {
4458 q_vector
->itr_val
= new_val
;
4459 q_vector
->set_itr
= 1;
4462 q_vector
->rx
.total_bytes
= 0;
4463 q_vector
->rx
.total_packets
= 0;
4464 q_vector
->tx
.total_bytes
= 0;
4465 q_vector
->tx
.total_packets
= 0;
4469 * igb_update_itr - update the dynamic ITR value based on statistics
4470 * @q_vector: pointer to q_vector
4471 * @ring_container: ring info to update the itr for
4473 * Stores a new ITR value based on packets and byte
4474 * counts during the last interrupt. The advantage of per interrupt
4475 * computation is faster updates and more accurate ITR for the current
4476 * traffic pattern. Constants in this function were computed
4477 * based on theoretical maximum wire speed and thresholds were set based
4478 * on testing data as well as attempting to minimize response time
4479 * while increasing bulk throughput.
4480 * This functionality is controlled by ethtool's coalescing settings.
4481 * NOTE: These calculations are only valid when operating in a single-
4482 * queue environment.
4484 static void igb_update_itr(struct igb_q_vector
*q_vector
,
4485 struct igb_ring_container
*ring_container
)
4487 unsigned int packets
= ring_container
->total_packets
;
4488 unsigned int bytes
= ring_container
->total_bytes
;
4489 u8 itrval
= ring_container
->itr
;
4491 /* no packets, exit with status unchanged */
4496 case lowest_latency
:
4497 /* handle TSO and jumbo frames */
4498 if (bytes
/packets
> 8000)
4499 itrval
= bulk_latency
;
4500 else if ((packets
< 5) && (bytes
> 512))
4501 itrval
= low_latency
;
4503 case low_latency
: /* 50 usec aka 20000 ints/s */
4504 if (bytes
> 10000) {
4505 /* this if handles the TSO accounting */
4506 if (bytes
/packets
> 8000)
4507 itrval
= bulk_latency
;
4508 else if ((packets
< 10) || ((bytes
/packets
) > 1200))
4509 itrval
= bulk_latency
;
4510 else if ((packets
> 35))
4511 itrval
= lowest_latency
;
4512 } else if (bytes
/packets
> 2000) {
4513 itrval
= bulk_latency
;
4514 } else if (packets
<= 2 && bytes
< 512) {
4515 itrval
= lowest_latency
;
4518 case bulk_latency
: /* 250 usec aka 4000 ints/s */
4519 if (bytes
> 25000) {
4521 itrval
= low_latency
;
4522 } else if (bytes
< 1500) {
4523 itrval
= low_latency
;
4528 /* clear work counters since we have the values we need */
4529 ring_container
->total_bytes
= 0;
4530 ring_container
->total_packets
= 0;
4532 /* write updated itr to ring container */
4533 ring_container
->itr
= itrval
;
4536 static void igb_set_itr(struct igb_q_vector
*q_vector
)
4538 struct igb_adapter
*adapter
= q_vector
->adapter
;
4539 u32 new_itr
= q_vector
->itr_val
;
4542 /* for non-gigabit speeds, just fix the interrupt rate at 4000 */
4543 if (adapter
->link_speed
!= SPEED_1000
) {
4545 new_itr
= IGB_4K_ITR
;
4549 igb_update_itr(q_vector
, &q_vector
->tx
);
4550 igb_update_itr(q_vector
, &q_vector
->rx
);
4552 current_itr
= max(q_vector
->rx
.itr
, q_vector
->tx
.itr
);
4554 /* conservative mode (itr 3) eliminates the lowest_latency setting */
4555 if (current_itr
== lowest_latency
&&
4556 ((q_vector
->rx
.ring
&& adapter
->rx_itr_setting
== 3) ||
4557 (!q_vector
->rx
.ring
&& adapter
->tx_itr_setting
== 3)))
4558 current_itr
= low_latency
;
4560 switch (current_itr
) {
4561 /* counts and packets in update_itr are dependent on these numbers */
4562 case lowest_latency
:
4563 new_itr
= IGB_70K_ITR
; /* 70,000 ints/sec */
4566 new_itr
= IGB_20K_ITR
; /* 20,000 ints/sec */
4569 new_itr
= IGB_4K_ITR
; /* 4,000 ints/sec */
4576 if (new_itr
!= q_vector
->itr_val
) {
4577 /* this attempts to bias the interrupt rate towards Bulk
4578 * by adding intermediate steps when interrupt rate is
4581 new_itr
= new_itr
> q_vector
->itr_val
?
4582 max((new_itr
* q_vector
->itr_val
) /
4583 (new_itr
+ (q_vector
->itr_val
>> 2)),
4585 /* Don't write the value here; it resets the adapter's
4586 * internal timer, and causes us to delay far longer than
4587 * we should between interrupts. Instead, we write the ITR
4588 * value at the beginning of the next interrupt so the timing
4589 * ends up being correct.
4591 q_vector
->itr_val
= new_itr
;
4592 q_vector
->set_itr
= 1;
4596 static void igb_tx_ctxtdesc(struct igb_ring
*tx_ring
, u32 vlan_macip_lens
,
4597 u32 type_tucmd
, u32 mss_l4len_idx
)
4599 struct e1000_adv_tx_context_desc
*context_desc
;
4600 u16 i
= tx_ring
->next_to_use
;
4602 context_desc
= IGB_TX_CTXTDESC(tx_ring
, i
);
4605 tx_ring
->next_to_use
= (i
< tx_ring
->count
) ? i
: 0;
4607 /* set bits to identify this as an advanced context descriptor */
4608 type_tucmd
|= E1000_TXD_CMD_DEXT
| E1000_ADVTXD_DTYP_CTXT
;
4610 /* For 82575, context index must be unique per ring. */
4611 if (test_bit(IGB_RING_FLAG_TX_CTX_IDX
, &tx_ring
->flags
))
4612 mss_l4len_idx
|= tx_ring
->reg_idx
<< 4;
4614 context_desc
->vlan_macip_lens
= cpu_to_le32(vlan_macip_lens
);
4615 context_desc
->seqnum_seed
= 0;
4616 context_desc
->type_tucmd_mlhl
= cpu_to_le32(type_tucmd
);
4617 context_desc
->mss_l4len_idx
= cpu_to_le32(mss_l4len_idx
);
4620 static int igb_tso(struct igb_ring
*tx_ring
,
4621 struct igb_tx_buffer
*first
,
4624 struct sk_buff
*skb
= first
->skb
;
4625 u32 vlan_macip_lens
, type_tucmd
;
4626 u32 mss_l4len_idx
, l4len
;
4629 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
4632 if (!skb_is_gso(skb
))
4635 err
= skb_cow_head(skb
, 0);
4639 /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
4640 type_tucmd
= E1000_ADVTXD_TUCMD_L4T_TCP
;
4642 if (first
->protocol
== htons(ETH_P_IP
)) {
4643 struct iphdr
*iph
= ip_hdr(skb
);
4646 tcp_hdr(skb
)->check
= ~csum_tcpudp_magic(iph
->saddr
,
4650 type_tucmd
|= E1000_ADVTXD_TUCMD_IPV4
;
4651 first
->tx_flags
|= IGB_TX_FLAGS_TSO
|
4654 } else if (skb_is_gso_v6(skb
)) {
4655 ipv6_hdr(skb
)->payload_len
= 0;
4656 tcp_hdr(skb
)->check
= ~csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
4657 &ipv6_hdr(skb
)->daddr
,
4659 first
->tx_flags
|= IGB_TX_FLAGS_TSO
|
4663 /* compute header lengths */
4664 l4len
= tcp_hdrlen(skb
);
4665 *hdr_len
= skb_transport_offset(skb
) + l4len
;
4667 /* update gso size and bytecount with header size */
4668 first
->gso_segs
= skb_shinfo(skb
)->gso_segs
;
4669 first
->bytecount
+= (first
->gso_segs
- 1) * *hdr_len
;
4672 mss_l4len_idx
= l4len
<< E1000_ADVTXD_L4LEN_SHIFT
;
4673 mss_l4len_idx
|= skb_shinfo(skb
)->gso_size
<< E1000_ADVTXD_MSS_SHIFT
;
4675 /* VLAN MACLEN IPLEN */
4676 vlan_macip_lens
= skb_network_header_len(skb
);
4677 vlan_macip_lens
|= skb_network_offset(skb
) << E1000_ADVTXD_MACLEN_SHIFT
;
4678 vlan_macip_lens
|= first
->tx_flags
& IGB_TX_FLAGS_VLAN_MASK
;
4680 igb_tx_ctxtdesc(tx_ring
, vlan_macip_lens
, type_tucmd
, mss_l4len_idx
);
4685 static void igb_tx_csum(struct igb_ring
*tx_ring
, struct igb_tx_buffer
*first
)
4687 struct sk_buff
*skb
= first
->skb
;
4688 u32 vlan_macip_lens
= 0;
4689 u32 mss_l4len_idx
= 0;
4692 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
4693 if (!(first
->tx_flags
& IGB_TX_FLAGS_VLAN
))
4698 switch (first
->protocol
) {
4699 case htons(ETH_P_IP
):
4700 vlan_macip_lens
|= skb_network_header_len(skb
);
4701 type_tucmd
|= E1000_ADVTXD_TUCMD_IPV4
;
4702 l4_hdr
= ip_hdr(skb
)->protocol
;
4704 case htons(ETH_P_IPV6
):
4705 vlan_macip_lens
|= skb_network_header_len(skb
);
4706 l4_hdr
= ipv6_hdr(skb
)->nexthdr
;
4709 if (unlikely(net_ratelimit())) {
4710 dev_warn(tx_ring
->dev
,
4711 "partial checksum but proto=%x!\n",
4719 type_tucmd
|= E1000_ADVTXD_TUCMD_L4T_TCP
;
4720 mss_l4len_idx
= tcp_hdrlen(skb
) <<
4721 E1000_ADVTXD_L4LEN_SHIFT
;
4724 type_tucmd
|= E1000_ADVTXD_TUCMD_L4T_SCTP
;
4725 mss_l4len_idx
= sizeof(struct sctphdr
) <<
4726 E1000_ADVTXD_L4LEN_SHIFT
;
4729 mss_l4len_idx
= sizeof(struct udphdr
) <<
4730 E1000_ADVTXD_L4LEN_SHIFT
;
4733 if (unlikely(net_ratelimit())) {
4734 dev_warn(tx_ring
->dev
,
4735 "partial checksum but l4 proto=%x!\n",
4741 /* update TX checksum flag */
4742 first
->tx_flags
|= IGB_TX_FLAGS_CSUM
;
4745 vlan_macip_lens
|= skb_network_offset(skb
) << E1000_ADVTXD_MACLEN_SHIFT
;
4746 vlan_macip_lens
|= first
->tx_flags
& IGB_TX_FLAGS_VLAN_MASK
;
4748 igb_tx_ctxtdesc(tx_ring
, vlan_macip_lens
, type_tucmd
, mss_l4len_idx
);
4751 #define IGB_SET_FLAG(_input, _flag, _result) \
4752 ((_flag <= _result) ? \
4753 ((u32)(_input & _flag) * (_result / _flag)) : \
4754 ((u32)(_input & _flag) / (_flag / _result)))
4756 static u32
igb_tx_cmd_type(struct sk_buff
*skb
, u32 tx_flags
)
4758 /* set type for advanced descriptor with frame checksum insertion */
4759 u32 cmd_type
= E1000_ADVTXD_DTYP_DATA
|
4760 E1000_ADVTXD_DCMD_DEXT
|
4761 E1000_ADVTXD_DCMD_IFCS
;
4763 /* set HW vlan bit if vlan is present */
4764 cmd_type
|= IGB_SET_FLAG(tx_flags
, IGB_TX_FLAGS_VLAN
,
4765 (E1000_ADVTXD_DCMD_VLE
));
4767 /* set segmentation bits for TSO */
4768 cmd_type
|= IGB_SET_FLAG(tx_flags
, IGB_TX_FLAGS_TSO
,
4769 (E1000_ADVTXD_DCMD_TSE
));
4771 /* set timestamp bit if present */
4772 cmd_type
|= IGB_SET_FLAG(tx_flags
, IGB_TX_FLAGS_TSTAMP
,
4773 (E1000_ADVTXD_MAC_TSTAMP
));
4775 /* insert frame checksum */
4776 cmd_type
^= IGB_SET_FLAG(skb
->no_fcs
, 1, E1000_ADVTXD_DCMD_IFCS
);
4781 static void igb_tx_olinfo_status(struct igb_ring
*tx_ring
,
4782 union e1000_adv_tx_desc
*tx_desc
,
4783 u32 tx_flags
, unsigned int paylen
)
4785 u32 olinfo_status
= paylen
<< E1000_ADVTXD_PAYLEN_SHIFT
;
4787 /* 82575 requires a unique index per ring */
4788 if (test_bit(IGB_RING_FLAG_TX_CTX_IDX
, &tx_ring
->flags
))
4789 olinfo_status
|= tx_ring
->reg_idx
<< 4;
4791 /* insert L4 checksum */
4792 olinfo_status
|= IGB_SET_FLAG(tx_flags
,
4794 (E1000_TXD_POPTS_TXSM
<< 8));
4796 /* insert IPv4 checksum */
4797 olinfo_status
|= IGB_SET_FLAG(tx_flags
,
4799 (E1000_TXD_POPTS_IXSM
<< 8));
4801 tx_desc
->read
.olinfo_status
= cpu_to_le32(olinfo_status
);
4804 static int __igb_maybe_stop_tx(struct igb_ring
*tx_ring
, const u16 size
)
4806 struct net_device
*netdev
= tx_ring
->netdev
;
4808 netif_stop_subqueue(netdev
, tx_ring
->queue_index
);
4810 /* Herbert's original patch had:
4811 * smp_mb__after_netif_stop_queue();
4812 * but since that doesn't exist yet, just open code it.
4816 /* We need to check again in a case another CPU has just
4817 * made room available.
4819 if (igb_desc_unused(tx_ring
) < size
)
4823 netif_wake_subqueue(netdev
, tx_ring
->queue_index
);
4825 u64_stats_update_begin(&tx_ring
->tx_syncp2
);
4826 tx_ring
->tx_stats
.restart_queue2
++;
4827 u64_stats_update_end(&tx_ring
->tx_syncp2
);
4832 static inline int igb_maybe_stop_tx(struct igb_ring
*tx_ring
, const u16 size
)
4834 if (igb_desc_unused(tx_ring
) >= size
)
4836 return __igb_maybe_stop_tx(tx_ring
, size
);
4839 static void igb_tx_map(struct igb_ring
*tx_ring
,
4840 struct igb_tx_buffer
*first
,
4843 struct sk_buff
*skb
= first
->skb
;
4844 struct igb_tx_buffer
*tx_buffer
;
4845 union e1000_adv_tx_desc
*tx_desc
;
4846 struct skb_frag_struct
*frag
;
4848 unsigned int data_len
, size
;
4849 u32 tx_flags
= first
->tx_flags
;
4850 u32 cmd_type
= igb_tx_cmd_type(skb
, tx_flags
);
4851 u16 i
= tx_ring
->next_to_use
;
4853 tx_desc
= IGB_TX_DESC(tx_ring
, i
);
4855 igb_tx_olinfo_status(tx_ring
, tx_desc
, tx_flags
, skb
->len
- hdr_len
);
4857 size
= skb_headlen(skb
);
4858 data_len
= skb
->data_len
;
4860 dma
= dma_map_single(tx_ring
->dev
, skb
->data
, size
, DMA_TO_DEVICE
);
4864 for (frag
= &skb_shinfo(skb
)->frags
[0];; frag
++) {
4865 if (dma_mapping_error(tx_ring
->dev
, dma
))
4868 /* record length, and DMA address */
4869 dma_unmap_len_set(tx_buffer
, len
, size
);
4870 dma_unmap_addr_set(tx_buffer
, dma
, dma
);
4872 tx_desc
->read
.buffer_addr
= cpu_to_le64(dma
);
4874 while (unlikely(size
> IGB_MAX_DATA_PER_TXD
)) {
4875 tx_desc
->read
.cmd_type_len
=
4876 cpu_to_le32(cmd_type
^ IGB_MAX_DATA_PER_TXD
);
4880 if (i
== tx_ring
->count
) {
4881 tx_desc
= IGB_TX_DESC(tx_ring
, 0);
4884 tx_desc
->read
.olinfo_status
= 0;
4886 dma
+= IGB_MAX_DATA_PER_TXD
;
4887 size
-= IGB_MAX_DATA_PER_TXD
;
4889 tx_desc
->read
.buffer_addr
= cpu_to_le64(dma
);
4892 if (likely(!data_len
))
4895 tx_desc
->read
.cmd_type_len
= cpu_to_le32(cmd_type
^ size
);
4899 if (i
== tx_ring
->count
) {
4900 tx_desc
= IGB_TX_DESC(tx_ring
, 0);
4903 tx_desc
->read
.olinfo_status
= 0;
4905 size
= skb_frag_size(frag
);
4908 dma
= skb_frag_dma_map(tx_ring
->dev
, frag
, 0,
4909 size
, DMA_TO_DEVICE
);
4911 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
4914 /* write last descriptor with RS and EOP bits */
4915 cmd_type
|= size
| IGB_TXD_DCMD
;
4916 tx_desc
->read
.cmd_type_len
= cpu_to_le32(cmd_type
);
4918 netdev_tx_sent_queue(txring_txq(tx_ring
), first
->bytecount
);
4920 /* set the timestamp */
4921 first
->time_stamp
= jiffies
;
4923 /* Force memory writes to complete before letting h/w know there
4924 * are new descriptors to fetch. (Only applicable for weak-ordered
4925 * memory model archs, such as IA-64).
4927 * We also need this memory barrier to make certain all of the
4928 * status bits have been updated before next_to_watch is written.
4932 /* set next_to_watch value indicating a packet is present */
4933 first
->next_to_watch
= tx_desc
;
4936 if (i
== tx_ring
->count
)
4939 tx_ring
->next_to_use
= i
;
4941 /* Make sure there is space in the ring for the next send. */
4942 igb_maybe_stop_tx(tx_ring
, DESC_NEEDED
);
4944 if (netif_xmit_stopped(txring_txq(tx_ring
)) || !skb
->xmit_more
) {
4945 writel(i
, tx_ring
->tail
);
4947 /* we need this if more than one processor can write to our tail
4948 * at a time, it synchronizes IO on IA64/Altix systems
4955 dev_err(tx_ring
->dev
, "TX DMA map failed\n");
4957 /* clear dma mappings for failed tx_buffer_info map */
4959 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
4960 igb_unmap_and_free_tx_resource(tx_ring
, tx_buffer
);
4961 if (tx_buffer
== first
)
4968 tx_ring
->next_to_use
= i
;
4971 netdev_tx_t
igb_xmit_frame_ring(struct sk_buff
*skb
,
4972 struct igb_ring
*tx_ring
)
4974 struct igb_tx_buffer
*first
;
4977 u16 count
= TXD_USE_COUNT(skb_headlen(skb
));
4978 __be16 protocol
= vlan_get_protocol(skb
);
4981 /* need: 1 descriptor per page * PAGE_SIZE/IGB_MAX_DATA_PER_TXD,
4982 * + 1 desc for skb_headlen/IGB_MAX_DATA_PER_TXD,
4983 * + 2 desc gap to keep tail from touching head,
4984 * + 1 desc for context descriptor,
4985 * otherwise try next time
4987 if (NETDEV_FRAG_PAGE_MAX_SIZE
> IGB_MAX_DATA_PER_TXD
) {
4990 for (f
= 0; f
< skb_shinfo(skb
)->nr_frags
; f
++)
4991 count
+= TXD_USE_COUNT(skb_shinfo(skb
)->frags
[f
].size
);
4993 count
+= skb_shinfo(skb
)->nr_frags
;
4996 if (igb_maybe_stop_tx(tx_ring
, count
+ 3)) {
4997 /* this is a hard error */
4998 return NETDEV_TX_BUSY
;
5001 /* record the location of the first descriptor for this packet */
5002 first
= &tx_ring
->tx_buffer_info
[tx_ring
->next_to_use
];
5004 first
->bytecount
= skb
->len
;
5005 first
->gso_segs
= 1;
5007 if (unlikely(skb_shinfo(skb
)->tx_flags
& SKBTX_HW_TSTAMP
)) {
5008 struct igb_adapter
*adapter
= netdev_priv(tx_ring
->netdev
);
5010 if (!test_and_set_bit_lock(__IGB_PTP_TX_IN_PROGRESS
,
5012 skb_shinfo(skb
)->tx_flags
|= SKBTX_IN_PROGRESS
;
5013 tx_flags
|= IGB_TX_FLAGS_TSTAMP
;
5015 adapter
->ptp_tx_skb
= skb_get(skb
);
5016 adapter
->ptp_tx_start
= jiffies
;
5017 if (adapter
->hw
.mac
.type
== e1000_82576
)
5018 schedule_work(&adapter
->ptp_tx_work
);
5022 skb_tx_timestamp(skb
);
5024 if (skb_vlan_tag_present(skb
)) {
5025 tx_flags
|= IGB_TX_FLAGS_VLAN
;
5026 tx_flags
|= (skb_vlan_tag_get(skb
) << IGB_TX_FLAGS_VLAN_SHIFT
);
5029 /* record initial flags and protocol */
5030 first
->tx_flags
= tx_flags
;
5031 first
->protocol
= protocol
;
5033 tso
= igb_tso(tx_ring
, first
, &hdr_len
);
5037 igb_tx_csum(tx_ring
, first
);
5039 igb_tx_map(tx_ring
, first
, hdr_len
);
5041 return NETDEV_TX_OK
;
5044 igb_unmap_and_free_tx_resource(tx_ring
, first
);
5046 return NETDEV_TX_OK
;
5049 static inline struct igb_ring
*igb_tx_queue_mapping(struct igb_adapter
*adapter
,
5050 struct sk_buff
*skb
)
5052 unsigned int r_idx
= skb
->queue_mapping
;
5054 if (r_idx
>= adapter
->num_tx_queues
)
5055 r_idx
= r_idx
% adapter
->num_tx_queues
;
5057 return adapter
->tx_ring
[r_idx
];
5060 static netdev_tx_t
igb_xmit_frame(struct sk_buff
*skb
,
5061 struct net_device
*netdev
)
5063 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5065 if (test_bit(__IGB_DOWN
, &adapter
->state
)) {
5066 dev_kfree_skb_any(skb
);
5067 return NETDEV_TX_OK
;
5070 if (skb
->len
<= 0) {
5071 dev_kfree_skb_any(skb
);
5072 return NETDEV_TX_OK
;
5075 /* The minimum packet size with TCTL.PSP set is 17 so pad the skb
5076 * in order to meet this minimum size requirement.
5078 if (skb_put_padto(skb
, 17))
5079 return NETDEV_TX_OK
;
5081 return igb_xmit_frame_ring(skb
, igb_tx_queue_mapping(adapter
, skb
));
5085 * igb_tx_timeout - Respond to a Tx Hang
5086 * @netdev: network interface device structure
5088 static void igb_tx_timeout(struct net_device
*netdev
)
5090 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5091 struct e1000_hw
*hw
= &adapter
->hw
;
5093 /* Do the reset outside of interrupt context */
5094 adapter
->tx_timeout_count
++;
5096 if (hw
->mac
.type
>= e1000_82580
)
5097 hw
->dev_spec
._82575
.global_device_reset
= true;
5099 schedule_work(&adapter
->reset_task
);
5101 (adapter
->eims_enable_mask
& ~adapter
->eims_other
));
5104 static void igb_reset_task(struct work_struct
*work
)
5106 struct igb_adapter
*adapter
;
5107 adapter
= container_of(work
, struct igb_adapter
, reset_task
);
5110 netdev_err(adapter
->netdev
, "Reset adapter\n");
5111 igb_reinit_locked(adapter
);
5115 * igb_get_stats64 - Get System Network Statistics
5116 * @netdev: network interface device structure
5117 * @stats: rtnl_link_stats64 pointer
5119 static struct rtnl_link_stats64
*igb_get_stats64(struct net_device
*netdev
,
5120 struct rtnl_link_stats64
*stats
)
5122 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5124 spin_lock(&adapter
->stats64_lock
);
5125 igb_update_stats(adapter
, &adapter
->stats64
);
5126 memcpy(stats
, &adapter
->stats64
, sizeof(*stats
));
5127 spin_unlock(&adapter
->stats64_lock
);
5133 * igb_change_mtu - Change the Maximum Transfer Unit
5134 * @netdev: network interface device structure
5135 * @new_mtu: new value for maximum frame size
5137 * Returns 0 on success, negative on failure
5139 static int igb_change_mtu(struct net_device
*netdev
, int new_mtu
)
5141 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5142 struct pci_dev
*pdev
= adapter
->pdev
;
5143 int max_frame
= new_mtu
+ ETH_HLEN
+ ETH_FCS_LEN
+ VLAN_HLEN
;
5145 if ((new_mtu
< 68) || (max_frame
> MAX_JUMBO_FRAME_SIZE
)) {
5146 dev_err(&pdev
->dev
, "Invalid MTU setting\n");
5150 #define MAX_STD_JUMBO_FRAME_SIZE 9238
5151 if (max_frame
> MAX_STD_JUMBO_FRAME_SIZE
) {
5152 dev_err(&pdev
->dev
, "MTU > 9216 not supported.\n");
5156 /* adjust max frame to be at least the size of a standard frame */
5157 if (max_frame
< (ETH_FRAME_LEN
+ ETH_FCS_LEN
))
5158 max_frame
= ETH_FRAME_LEN
+ ETH_FCS_LEN
;
5160 while (test_and_set_bit(__IGB_RESETTING
, &adapter
->state
))
5161 usleep_range(1000, 2000);
5163 /* igb_down has a dependency on max_frame_size */
5164 adapter
->max_frame_size
= max_frame
;
5166 if (netif_running(netdev
))
5169 dev_info(&pdev
->dev
, "changing MTU from %d to %d\n",
5170 netdev
->mtu
, new_mtu
);
5171 netdev
->mtu
= new_mtu
;
5173 if (netif_running(netdev
))
5178 clear_bit(__IGB_RESETTING
, &adapter
->state
);
5184 * igb_update_stats - Update the board statistics counters
5185 * @adapter: board private structure
5187 void igb_update_stats(struct igb_adapter
*adapter
,
5188 struct rtnl_link_stats64
*net_stats
)
5190 struct e1000_hw
*hw
= &adapter
->hw
;
5191 struct pci_dev
*pdev
= adapter
->pdev
;
5196 u64 _bytes
, _packets
;
5198 /* Prevent stats update while adapter is being reset, or if the pci
5199 * connection is down.
5201 if (adapter
->link_speed
== 0)
5203 if (pci_channel_offline(pdev
))
5210 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
5211 struct igb_ring
*ring
= adapter
->rx_ring
[i
];
5212 u32 rqdpc
= rd32(E1000_RQDPC(i
));
5213 if (hw
->mac
.type
>= e1000_i210
)
5214 wr32(E1000_RQDPC(i
), 0);
5217 ring
->rx_stats
.drops
+= rqdpc
;
5218 net_stats
->rx_fifo_errors
+= rqdpc
;
5222 start
= u64_stats_fetch_begin_irq(&ring
->rx_syncp
);
5223 _bytes
= ring
->rx_stats
.bytes
;
5224 _packets
= ring
->rx_stats
.packets
;
5225 } while (u64_stats_fetch_retry_irq(&ring
->rx_syncp
, start
));
5227 packets
+= _packets
;
5230 net_stats
->rx_bytes
= bytes
;
5231 net_stats
->rx_packets
= packets
;
5235 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
5236 struct igb_ring
*ring
= adapter
->tx_ring
[i
];
5238 start
= u64_stats_fetch_begin_irq(&ring
->tx_syncp
);
5239 _bytes
= ring
->tx_stats
.bytes
;
5240 _packets
= ring
->tx_stats
.packets
;
5241 } while (u64_stats_fetch_retry_irq(&ring
->tx_syncp
, start
));
5243 packets
+= _packets
;
5245 net_stats
->tx_bytes
= bytes
;
5246 net_stats
->tx_packets
= packets
;
5249 /* read stats registers */
5250 adapter
->stats
.crcerrs
+= rd32(E1000_CRCERRS
);
5251 adapter
->stats
.gprc
+= rd32(E1000_GPRC
);
5252 adapter
->stats
.gorc
+= rd32(E1000_GORCL
);
5253 rd32(E1000_GORCH
); /* clear GORCL */
5254 adapter
->stats
.bprc
+= rd32(E1000_BPRC
);
5255 adapter
->stats
.mprc
+= rd32(E1000_MPRC
);
5256 adapter
->stats
.roc
+= rd32(E1000_ROC
);
5258 adapter
->stats
.prc64
+= rd32(E1000_PRC64
);
5259 adapter
->stats
.prc127
+= rd32(E1000_PRC127
);
5260 adapter
->stats
.prc255
+= rd32(E1000_PRC255
);
5261 adapter
->stats
.prc511
+= rd32(E1000_PRC511
);
5262 adapter
->stats
.prc1023
+= rd32(E1000_PRC1023
);
5263 adapter
->stats
.prc1522
+= rd32(E1000_PRC1522
);
5264 adapter
->stats
.symerrs
+= rd32(E1000_SYMERRS
);
5265 adapter
->stats
.sec
+= rd32(E1000_SEC
);
5267 mpc
= rd32(E1000_MPC
);
5268 adapter
->stats
.mpc
+= mpc
;
5269 net_stats
->rx_fifo_errors
+= mpc
;
5270 adapter
->stats
.scc
+= rd32(E1000_SCC
);
5271 adapter
->stats
.ecol
+= rd32(E1000_ECOL
);
5272 adapter
->stats
.mcc
+= rd32(E1000_MCC
);
5273 adapter
->stats
.latecol
+= rd32(E1000_LATECOL
);
5274 adapter
->stats
.dc
+= rd32(E1000_DC
);
5275 adapter
->stats
.rlec
+= rd32(E1000_RLEC
);
5276 adapter
->stats
.xonrxc
+= rd32(E1000_XONRXC
);
5277 adapter
->stats
.xontxc
+= rd32(E1000_XONTXC
);
5278 adapter
->stats
.xoffrxc
+= rd32(E1000_XOFFRXC
);
5279 adapter
->stats
.xofftxc
+= rd32(E1000_XOFFTXC
);
5280 adapter
->stats
.fcruc
+= rd32(E1000_FCRUC
);
5281 adapter
->stats
.gptc
+= rd32(E1000_GPTC
);
5282 adapter
->stats
.gotc
+= rd32(E1000_GOTCL
);
5283 rd32(E1000_GOTCH
); /* clear GOTCL */
5284 adapter
->stats
.rnbc
+= rd32(E1000_RNBC
);
5285 adapter
->stats
.ruc
+= rd32(E1000_RUC
);
5286 adapter
->stats
.rfc
+= rd32(E1000_RFC
);
5287 adapter
->stats
.rjc
+= rd32(E1000_RJC
);
5288 adapter
->stats
.tor
+= rd32(E1000_TORH
);
5289 adapter
->stats
.tot
+= rd32(E1000_TOTH
);
5290 adapter
->stats
.tpr
+= rd32(E1000_TPR
);
5292 adapter
->stats
.ptc64
+= rd32(E1000_PTC64
);
5293 adapter
->stats
.ptc127
+= rd32(E1000_PTC127
);
5294 adapter
->stats
.ptc255
+= rd32(E1000_PTC255
);
5295 adapter
->stats
.ptc511
+= rd32(E1000_PTC511
);
5296 adapter
->stats
.ptc1023
+= rd32(E1000_PTC1023
);
5297 adapter
->stats
.ptc1522
+= rd32(E1000_PTC1522
);
5299 adapter
->stats
.mptc
+= rd32(E1000_MPTC
);
5300 adapter
->stats
.bptc
+= rd32(E1000_BPTC
);
5302 adapter
->stats
.tpt
+= rd32(E1000_TPT
);
5303 adapter
->stats
.colc
+= rd32(E1000_COLC
);
5305 adapter
->stats
.algnerrc
+= rd32(E1000_ALGNERRC
);
5306 /* read internal phy specific stats */
5307 reg
= rd32(E1000_CTRL_EXT
);
5308 if (!(reg
& E1000_CTRL_EXT_LINK_MODE_MASK
)) {
5309 adapter
->stats
.rxerrc
+= rd32(E1000_RXERRC
);
5311 /* this stat has invalid values on i210/i211 */
5312 if ((hw
->mac
.type
!= e1000_i210
) &&
5313 (hw
->mac
.type
!= e1000_i211
))
5314 adapter
->stats
.tncrs
+= rd32(E1000_TNCRS
);
5317 adapter
->stats
.tsctc
+= rd32(E1000_TSCTC
);
5318 adapter
->stats
.tsctfc
+= rd32(E1000_TSCTFC
);
5320 adapter
->stats
.iac
+= rd32(E1000_IAC
);
5321 adapter
->stats
.icrxoc
+= rd32(E1000_ICRXOC
);
5322 adapter
->stats
.icrxptc
+= rd32(E1000_ICRXPTC
);
5323 adapter
->stats
.icrxatc
+= rd32(E1000_ICRXATC
);
5324 adapter
->stats
.ictxptc
+= rd32(E1000_ICTXPTC
);
5325 adapter
->stats
.ictxatc
+= rd32(E1000_ICTXATC
);
5326 adapter
->stats
.ictxqec
+= rd32(E1000_ICTXQEC
);
5327 adapter
->stats
.ictxqmtc
+= rd32(E1000_ICTXQMTC
);
5328 adapter
->stats
.icrxdmtc
+= rd32(E1000_ICRXDMTC
);
5330 /* Fill out the OS statistics structure */
5331 net_stats
->multicast
= adapter
->stats
.mprc
;
5332 net_stats
->collisions
= adapter
->stats
.colc
;
5336 /* RLEC on some newer hardware can be incorrect so build
5337 * our own version based on RUC and ROC
5339 net_stats
->rx_errors
= adapter
->stats
.rxerrc
+
5340 adapter
->stats
.crcerrs
+ adapter
->stats
.algnerrc
+
5341 adapter
->stats
.ruc
+ adapter
->stats
.roc
+
5342 adapter
->stats
.cexterr
;
5343 net_stats
->rx_length_errors
= adapter
->stats
.ruc
+
5345 net_stats
->rx_crc_errors
= adapter
->stats
.crcerrs
;
5346 net_stats
->rx_frame_errors
= adapter
->stats
.algnerrc
;
5347 net_stats
->rx_missed_errors
= adapter
->stats
.mpc
;
5350 net_stats
->tx_errors
= adapter
->stats
.ecol
+
5351 adapter
->stats
.latecol
;
5352 net_stats
->tx_aborted_errors
= adapter
->stats
.ecol
;
5353 net_stats
->tx_window_errors
= adapter
->stats
.latecol
;
5354 net_stats
->tx_carrier_errors
= adapter
->stats
.tncrs
;
5356 /* Tx Dropped needs to be maintained elsewhere */
5358 /* Management Stats */
5359 adapter
->stats
.mgptc
+= rd32(E1000_MGTPTC
);
5360 adapter
->stats
.mgprc
+= rd32(E1000_MGTPRC
);
5361 adapter
->stats
.mgpdc
+= rd32(E1000_MGTPDC
);
5364 reg
= rd32(E1000_MANC
);
5365 if (reg
& E1000_MANC_EN_BMC2OS
) {
5366 adapter
->stats
.o2bgptc
+= rd32(E1000_O2BGPTC
);
5367 adapter
->stats
.o2bspc
+= rd32(E1000_O2BSPC
);
5368 adapter
->stats
.b2ospc
+= rd32(E1000_B2OSPC
);
5369 adapter
->stats
.b2ogprc
+= rd32(E1000_B2OGPRC
);
5373 static void igb_tsync_interrupt(struct igb_adapter
*adapter
)
5375 struct e1000_hw
*hw
= &adapter
->hw
;
5376 struct ptp_clock_event event
;
5378 u32 ack
= 0, tsauxc
, sec
, nsec
, tsicr
= rd32(E1000_TSICR
);
5380 if (tsicr
& TSINTR_SYS_WRAP
) {
5381 event
.type
= PTP_CLOCK_PPS
;
5382 if (adapter
->ptp_caps
.pps
)
5383 ptp_clock_event(adapter
->ptp_clock
, &event
);
5385 dev_err(&adapter
->pdev
->dev
, "unexpected SYS WRAP");
5386 ack
|= TSINTR_SYS_WRAP
;
5389 if (tsicr
& E1000_TSICR_TXTS
) {
5390 /* retrieve hardware timestamp */
5391 schedule_work(&adapter
->ptp_tx_work
);
5392 ack
|= E1000_TSICR_TXTS
;
5395 if (tsicr
& TSINTR_TT0
) {
5396 spin_lock(&adapter
->tmreg_lock
);
5397 ts
= timespec_add(adapter
->perout
[0].start
,
5398 adapter
->perout
[0].period
);
5399 wr32(E1000_TRGTTIML0
, ts
.tv_nsec
);
5400 wr32(E1000_TRGTTIMH0
, ts
.tv_sec
);
5401 tsauxc
= rd32(E1000_TSAUXC
);
5402 tsauxc
|= TSAUXC_EN_TT0
;
5403 wr32(E1000_TSAUXC
, tsauxc
);
5404 adapter
->perout
[0].start
= ts
;
5405 spin_unlock(&adapter
->tmreg_lock
);
5409 if (tsicr
& TSINTR_TT1
) {
5410 spin_lock(&adapter
->tmreg_lock
);
5411 ts
= timespec_add(adapter
->perout
[1].start
,
5412 adapter
->perout
[1].period
);
5413 wr32(E1000_TRGTTIML1
, ts
.tv_nsec
);
5414 wr32(E1000_TRGTTIMH1
, ts
.tv_sec
);
5415 tsauxc
= rd32(E1000_TSAUXC
);
5416 tsauxc
|= TSAUXC_EN_TT1
;
5417 wr32(E1000_TSAUXC
, tsauxc
);
5418 adapter
->perout
[1].start
= ts
;
5419 spin_unlock(&adapter
->tmreg_lock
);
5423 if (tsicr
& TSINTR_AUTT0
) {
5424 nsec
= rd32(E1000_AUXSTMPL0
);
5425 sec
= rd32(E1000_AUXSTMPH0
);
5426 event
.type
= PTP_CLOCK_EXTTS
;
5428 event
.timestamp
= sec
* 1000000000ULL + nsec
;
5429 ptp_clock_event(adapter
->ptp_clock
, &event
);
5430 ack
|= TSINTR_AUTT0
;
5433 if (tsicr
& TSINTR_AUTT1
) {
5434 nsec
= rd32(E1000_AUXSTMPL1
);
5435 sec
= rd32(E1000_AUXSTMPH1
);
5436 event
.type
= PTP_CLOCK_EXTTS
;
5438 event
.timestamp
= sec
* 1000000000ULL + nsec
;
5439 ptp_clock_event(adapter
->ptp_clock
, &event
);
5440 ack
|= TSINTR_AUTT1
;
5443 /* acknowledge the interrupts */
5444 wr32(E1000_TSICR
, ack
);
5447 static irqreturn_t
igb_msix_other(int irq
, void *data
)
5449 struct igb_adapter
*adapter
= data
;
5450 struct e1000_hw
*hw
= &adapter
->hw
;
5451 u32 icr
= rd32(E1000_ICR
);
5452 /* reading ICR causes bit 31 of EICR to be cleared */
5454 if (icr
& E1000_ICR_DRSTA
)
5455 schedule_work(&adapter
->reset_task
);
5457 if (icr
& E1000_ICR_DOUTSYNC
) {
5458 /* HW is reporting DMA is out of sync */
5459 adapter
->stats
.doosync
++;
5460 /* The DMA Out of Sync is also indication of a spoof event
5461 * in IOV mode. Check the Wrong VM Behavior register to
5462 * see if it is really a spoof event.
5464 igb_check_wvbr(adapter
);
5467 /* Check for a mailbox event */
5468 if (icr
& E1000_ICR_VMMB
)
5469 igb_msg_task(adapter
);
5471 if (icr
& E1000_ICR_LSC
) {
5472 hw
->mac
.get_link_status
= 1;
5473 /* guard against interrupt when we're going down */
5474 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
5475 mod_timer(&adapter
->watchdog_timer
, jiffies
+ 1);
5478 if (icr
& E1000_ICR_TS
)
5479 igb_tsync_interrupt(adapter
);
5481 wr32(E1000_EIMS
, adapter
->eims_other
);
5486 static void igb_write_itr(struct igb_q_vector
*q_vector
)
5488 struct igb_adapter
*adapter
= q_vector
->adapter
;
5489 u32 itr_val
= q_vector
->itr_val
& 0x7FFC;
5491 if (!q_vector
->set_itr
)
5497 if (adapter
->hw
.mac
.type
== e1000_82575
)
5498 itr_val
|= itr_val
<< 16;
5500 itr_val
|= E1000_EITR_CNT_IGNR
;
5502 writel(itr_val
, q_vector
->itr_register
);
5503 q_vector
->set_itr
= 0;
5506 static irqreturn_t
igb_msix_ring(int irq
, void *data
)
5508 struct igb_q_vector
*q_vector
= data
;
5510 /* Write the ITR value calculated from the previous interrupt. */
5511 igb_write_itr(q_vector
);
5513 napi_schedule(&q_vector
->napi
);
5518 #ifdef CONFIG_IGB_DCA
5519 static void igb_update_tx_dca(struct igb_adapter
*adapter
,
5520 struct igb_ring
*tx_ring
,
5523 struct e1000_hw
*hw
= &adapter
->hw
;
5524 u32 txctrl
= dca3_get_tag(tx_ring
->dev
, cpu
);
5526 if (hw
->mac
.type
!= e1000_82575
)
5527 txctrl
<<= E1000_DCA_TXCTRL_CPUID_SHIFT
;
5529 /* We can enable relaxed ordering for reads, but not writes when
5530 * DCA is enabled. This is due to a known issue in some chipsets
5531 * which will cause the DCA tag to be cleared.
5533 txctrl
|= E1000_DCA_TXCTRL_DESC_RRO_EN
|
5534 E1000_DCA_TXCTRL_DATA_RRO_EN
|
5535 E1000_DCA_TXCTRL_DESC_DCA_EN
;
5537 wr32(E1000_DCA_TXCTRL(tx_ring
->reg_idx
), txctrl
);
5540 static void igb_update_rx_dca(struct igb_adapter
*adapter
,
5541 struct igb_ring
*rx_ring
,
5544 struct e1000_hw
*hw
= &adapter
->hw
;
5545 u32 rxctrl
= dca3_get_tag(&adapter
->pdev
->dev
, cpu
);
5547 if (hw
->mac
.type
!= e1000_82575
)
5548 rxctrl
<<= E1000_DCA_RXCTRL_CPUID_SHIFT
;
5550 /* We can enable relaxed ordering for reads, but not writes when
5551 * DCA is enabled. This is due to a known issue in some chipsets
5552 * which will cause the DCA tag to be cleared.
5554 rxctrl
|= E1000_DCA_RXCTRL_DESC_RRO_EN
|
5555 E1000_DCA_RXCTRL_DESC_DCA_EN
;
5557 wr32(E1000_DCA_RXCTRL(rx_ring
->reg_idx
), rxctrl
);
5560 static void igb_update_dca(struct igb_q_vector
*q_vector
)
5562 struct igb_adapter
*adapter
= q_vector
->adapter
;
5563 int cpu
= get_cpu();
5565 if (q_vector
->cpu
== cpu
)
5568 if (q_vector
->tx
.ring
)
5569 igb_update_tx_dca(adapter
, q_vector
->tx
.ring
, cpu
);
5571 if (q_vector
->rx
.ring
)
5572 igb_update_rx_dca(adapter
, q_vector
->rx
.ring
, cpu
);
5574 q_vector
->cpu
= cpu
;
5579 static void igb_setup_dca(struct igb_adapter
*adapter
)
5581 struct e1000_hw
*hw
= &adapter
->hw
;
5584 if (!(adapter
->flags
& IGB_FLAG_DCA_ENABLED
))
5587 /* Always use CB2 mode, difference is masked in the CB driver. */
5588 wr32(E1000_DCA_CTRL
, E1000_DCA_CTRL_DCA_MODE_CB2
);
5590 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
5591 adapter
->q_vector
[i
]->cpu
= -1;
5592 igb_update_dca(adapter
->q_vector
[i
]);
5596 static int __igb_notify_dca(struct device
*dev
, void *data
)
5598 struct net_device
*netdev
= dev_get_drvdata(dev
);
5599 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5600 struct pci_dev
*pdev
= adapter
->pdev
;
5601 struct e1000_hw
*hw
= &adapter
->hw
;
5602 unsigned long event
= *(unsigned long *)data
;
5605 case DCA_PROVIDER_ADD
:
5606 /* if already enabled, don't do it again */
5607 if (adapter
->flags
& IGB_FLAG_DCA_ENABLED
)
5609 if (dca_add_requester(dev
) == 0) {
5610 adapter
->flags
|= IGB_FLAG_DCA_ENABLED
;
5611 dev_info(&pdev
->dev
, "DCA enabled\n");
5612 igb_setup_dca(adapter
);
5615 /* Fall Through since DCA is disabled. */
5616 case DCA_PROVIDER_REMOVE
:
5617 if (adapter
->flags
& IGB_FLAG_DCA_ENABLED
) {
5618 /* without this a class_device is left
5619 * hanging around in the sysfs model
5621 dca_remove_requester(dev
);
5622 dev_info(&pdev
->dev
, "DCA disabled\n");
5623 adapter
->flags
&= ~IGB_FLAG_DCA_ENABLED
;
5624 wr32(E1000_DCA_CTRL
, E1000_DCA_CTRL_DCA_MODE_DISABLE
);
5632 static int igb_notify_dca(struct notifier_block
*nb
, unsigned long event
,
5637 ret_val
= driver_for_each_device(&igb_driver
.driver
, NULL
, &event
,
5640 return ret_val
? NOTIFY_BAD
: NOTIFY_DONE
;
5642 #endif /* CONFIG_IGB_DCA */
5644 #ifdef CONFIG_PCI_IOV
5645 static int igb_vf_configure(struct igb_adapter
*adapter
, int vf
)
5647 unsigned char mac_addr
[ETH_ALEN
];
5649 eth_zero_addr(mac_addr
);
5650 igb_set_vf_mac(adapter
, vf
, mac_addr
);
5652 /* By default spoof check is enabled for all VFs */
5653 adapter
->vf_data
[vf
].spoofchk_enabled
= true;
5659 static void igb_ping_all_vfs(struct igb_adapter
*adapter
)
5661 struct e1000_hw
*hw
= &adapter
->hw
;
5665 for (i
= 0 ; i
< adapter
->vfs_allocated_count
; i
++) {
5666 ping
= E1000_PF_CONTROL_MSG
;
5667 if (adapter
->vf_data
[i
].flags
& IGB_VF_FLAG_CTS
)
5668 ping
|= E1000_VT_MSGTYPE_CTS
;
5669 igb_write_mbx(hw
, &ping
, 1, i
);
5673 static int igb_set_vf_promisc(struct igb_adapter
*adapter
, u32
*msgbuf
, u32 vf
)
5675 struct e1000_hw
*hw
= &adapter
->hw
;
5676 u32 vmolr
= rd32(E1000_VMOLR(vf
));
5677 struct vf_data_storage
*vf_data
= &adapter
->vf_data
[vf
];
5679 vf_data
->flags
&= ~(IGB_VF_FLAG_UNI_PROMISC
|
5680 IGB_VF_FLAG_MULTI_PROMISC
);
5681 vmolr
&= ~(E1000_VMOLR_ROPE
| E1000_VMOLR_ROMPE
| E1000_VMOLR_MPME
);
5683 if (*msgbuf
& E1000_VF_SET_PROMISC_MULTICAST
) {
5684 vmolr
|= E1000_VMOLR_MPME
;
5685 vf_data
->flags
|= IGB_VF_FLAG_MULTI_PROMISC
;
5686 *msgbuf
&= ~E1000_VF_SET_PROMISC_MULTICAST
;
5688 /* if we have hashes and we are clearing a multicast promisc
5689 * flag we need to write the hashes to the MTA as this step
5690 * was previously skipped
5692 if (vf_data
->num_vf_mc_hashes
> 30) {
5693 vmolr
|= E1000_VMOLR_MPME
;
5694 } else if (vf_data
->num_vf_mc_hashes
) {
5697 vmolr
|= E1000_VMOLR_ROMPE
;
5698 for (j
= 0; j
< vf_data
->num_vf_mc_hashes
; j
++)
5699 igb_mta_set(hw
, vf_data
->vf_mc_hashes
[j
]);
5703 wr32(E1000_VMOLR(vf
), vmolr
);
5705 /* there are flags left unprocessed, likely not supported */
5706 if (*msgbuf
& E1000_VT_MSGINFO_MASK
)
5712 static int igb_set_vf_multicasts(struct igb_adapter
*adapter
,
5713 u32
*msgbuf
, u32 vf
)
5715 int n
= (msgbuf
[0] & E1000_VT_MSGINFO_MASK
) >> E1000_VT_MSGINFO_SHIFT
;
5716 u16
*hash_list
= (u16
*)&msgbuf
[1];
5717 struct vf_data_storage
*vf_data
= &adapter
->vf_data
[vf
];
5720 /* salt away the number of multicast addresses assigned
5721 * to this VF for later use to restore when the PF multi cast
5724 vf_data
->num_vf_mc_hashes
= n
;
5726 /* only up to 30 hash values supported */
5730 /* store the hashes for later use */
5731 for (i
= 0; i
< n
; i
++)
5732 vf_data
->vf_mc_hashes
[i
] = hash_list
[i
];
5734 /* Flush and reset the mta with the new values */
5735 igb_set_rx_mode(adapter
->netdev
);
5740 static void igb_restore_vf_multicasts(struct igb_adapter
*adapter
)
5742 struct e1000_hw
*hw
= &adapter
->hw
;
5743 struct vf_data_storage
*vf_data
;
5746 for (i
= 0; i
< adapter
->vfs_allocated_count
; i
++) {
5747 u32 vmolr
= rd32(E1000_VMOLR(i
));
5749 vmolr
&= ~(E1000_VMOLR_ROMPE
| E1000_VMOLR_MPME
);
5751 vf_data
= &adapter
->vf_data
[i
];
5753 if ((vf_data
->num_vf_mc_hashes
> 30) ||
5754 (vf_data
->flags
& IGB_VF_FLAG_MULTI_PROMISC
)) {
5755 vmolr
|= E1000_VMOLR_MPME
;
5756 } else if (vf_data
->num_vf_mc_hashes
) {
5757 vmolr
|= E1000_VMOLR_ROMPE
;
5758 for (j
= 0; j
< vf_data
->num_vf_mc_hashes
; j
++)
5759 igb_mta_set(hw
, vf_data
->vf_mc_hashes
[j
]);
5761 wr32(E1000_VMOLR(i
), vmolr
);
5765 static void igb_clear_vf_vfta(struct igb_adapter
*adapter
, u32 vf
)
5767 struct e1000_hw
*hw
= &adapter
->hw
;
5768 u32 pool_mask
, reg
, vid
;
5771 pool_mask
= 1 << (E1000_VLVF_POOLSEL_SHIFT
+ vf
);
5773 /* Find the vlan filter for this id */
5774 for (i
= 0; i
< E1000_VLVF_ARRAY_SIZE
; i
++) {
5775 reg
= rd32(E1000_VLVF(i
));
5777 /* remove the vf from the pool */
5780 /* if pool is empty then remove entry from vfta */
5781 if (!(reg
& E1000_VLVF_POOLSEL_MASK
) &&
5782 (reg
& E1000_VLVF_VLANID_ENABLE
)) {
5784 vid
= reg
& E1000_VLVF_VLANID_MASK
;
5785 igb_vfta_set(hw
, vid
, false);
5788 wr32(E1000_VLVF(i
), reg
);
5791 adapter
->vf_data
[vf
].vlans_enabled
= 0;
5794 static s32
igb_vlvf_set(struct igb_adapter
*adapter
, u32 vid
, bool add
, u32 vf
)
5796 struct e1000_hw
*hw
= &adapter
->hw
;
5799 /* The vlvf table only exists on 82576 hardware and newer */
5800 if (hw
->mac
.type
< e1000_82576
)
5803 /* we only need to do this if VMDq is enabled */
5804 if (!adapter
->vfs_allocated_count
)
5807 /* Find the vlan filter for this id */
5808 for (i
= 0; i
< E1000_VLVF_ARRAY_SIZE
; i
++) {
5809 reg
= rd32(E1000_VLVF(i
));
5810 if ((reg
& E1000_VLVF_VLANID_ENABLE
) &&
5811 vid
== (reg
& E1000_VLVF_VLANID_MASK
))
5816 if (i
== E1000_VLVF_ARRAY_SIZE
) {
5817 /* Did not find a matching VLAN ID entry that was
5818 * enabled. Search for a free filter entry, i.e.
5819 * one without the enable bit set
5821 for (i
= 0; i
< E1000_VLVF_ARRAY_SIZE
; i
++) {
5822 reg
= rd32(E1000_VLVF(i
));
5823 if (!(reg
& E1000_VLVF_VLANID_ENABLE
))
5827 if (i
< E1000_VLVF_ARRAY_SIZE
) {
5828 /* Found an enabled/available entry */
5829 reg
|= 1 << (E1000_VLVF_POOLSEL_SHIFT
+ vf
);
5831 /* if !enabled we need to set this up in vfta */
5832 if (!(reg
& E1000_VLVF_VLANID_ENABLE
)) {
5833 /* add VID to filter table */
5834 igb_vfta_set(hw
, vid
, true);
5835 reg
|= E1000_VLVF_VLANID_ENABLE
;
5837 reg
&= ~E1000_VLVF_VLANID_MASK
;
5839 wr32(E1000_VLVF(i
), reg
);
5841 /* do not modify RLPML for PF devices */
5842 if (vf
>= adapter
->vfs_allocated_count
)
5845 if (!adapter
->vf_data
[vf
].vlans_enabled
) {
5848 reg
= rd32(E1000_VMOLR(vf
));
5849 size
= reg
& E1000_VMOLR_RLPML_MASK
;
5851 reg
&= ~E1000_VMOLR_RLPML_MASK
;
5853 wr32(E1000_VMOLR(vf
), reg
);
5856 adapter
->vf_data
[vf
].vlans_enabled
++;
5859 if (i
< E1000_VLVF_ARRAY_SIZE
) {
5860 /* remove vf from the pool */
5861 reg
&= ~(1 << (E1000_VLVF_POOLSEL_SHIFT
+ vf
));
5862 /* if pool is empty then remove entry from vfta */
5863 if (!(reg
& E1000_VLVF_POOLSEL_MASK
)) {
5865 igb_vfta_set(hw
, vid
, false);
5867 wr32(E1000_VLVF(i
), reg
);
5869 /* do not modify RLPML for PF devices */
5870 if (vf
>= adapter
->vfs_allocated_count
)
5873 adapter
->vf_data
[vf
].vlans_enabled
--;
5874 if (!adapter
->vf_data
[vf
].vlans_enabled
) {
5877 reg
= rd32(E1000_VMOLR(vf
));
5878 size
= reg
& E1000_VMOLR_RLPML_MASK
;
5880 reg
&= ~E1000_VMOLR_RLPML_MASK
;
5882 wr32(E1000_VMOLR(vf
), reg
);
5889 static void igb_set_vmvir(struct igb_adapter
*adapter
, u32 vid
, u32 vf
)
5891 struct e1000_hw
*hw
= &adapter
->hw
;
5894 wr32(E1000_VMVIR(vf
), (vid
| E1000_VMVIR_VLANA_DEFAULT
));
5896 wr32(E1000_VMVIR(vf
), 0);
5899 static int igb_ndo_set_vf_vlan(struct net_device
*netdev
,
5900 int vf
, u16 vlan
, u8 qos
)
5903 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5905 if ((vf
>= adapter
->vfs_allocated_count
) || (vlan
> 4095) || (qos
> 7))
5908 err
= igb_vlvf_set(adapter
, vlan
, !!vlan
, vf
);
5911 igb_set_vmvir(adapter
, vlan
| (qos
<< VLAN_PRIO_SHIFT
), vf
);
5912 igb_set_vmolr(adapter
, vf
, !vlan
);
5913 adapter
->vf_data
[vf
].pf_vlan
= vlan
;
5914 adapter
->vf_data
[vf
].pf_qos
= qos
;
5915 dev_info(&adapter
->pdev
->dev
,
5916 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan
, qos
, vf
);
5917 if (test_bit(__IGB_DOWN
, &adapter
->state
)) {
5918 dev_warn(&adapter
->pdev
->dev
,
5919 "The VF VLAN has been set, but the PF device is not up.\n");
5920 dev_warn(&adapter
->pdev
->dev
,
5921 "Bring the PF device up before attempting to use the VF device.\n");
5924 igb_vlvf_set(adapter
, adapter
->vf_data
[vf
].pf_vlan
,
5926 igb_set_vmvir(adapter
, vlan
, vf
);
5927 igb_set_vmolr(adapter
, vf
, true);
5928 adapter
->vf_data
[vf
].pf_vlan
= 0;
5929 adapter
->vf_data
[vf
].pf_qos
= 0;
5935 static int igb_find_vlvf_entry(struct igb_adapter
*adapter
, int vid
)
5937 struct e1000_hw
*hw
= &adapter
->hw
;
5941 /* Find the vlan filter for this id */
5942 for (i
= 0; i
< E1000_VLVF_ARRAY_SIZE
; i
++) {
5943 reg
= rd32(E1000_VLVF(i
));
5944 if ((reg
& E1000_VLVF_VLANID_ENABLE
) &&
5945 vid
== (reg
& E1000_VLVF_VLANID_MASK
))
5949 if (i
>= E1000_VLVF_ARRAY_SIZE
)
5955 static int igb_set_vf_vlan(struct igb_adapter
*adapter
, u32
*msgbuf
, u32 vf
)
5957 struct e1000_hw
*hw
= &adapter
->hw
;
5958 int add
= (msgbuf
[0] & E1000_VT_MSGINFO_MASK
) >> E1000_VT_MSGINFO_SHIFT
;
5959 int vid
= (msgbuf
[1] & E1000_VLVF_VLANID_MASK
);
5962 /* If in promiscuous mode we need to make sure the PF also has
5963 * the VLAN filter set.
5965 if (add
&& (adapter
->netdev
->flags
& IFF_PROMISC
))
5966 err
= igb_vlvf_set(adapter
, vid
, add
,
5967 adapter
->vfs_allocated_count
);
5971 err
= igb_vlvf_set(adapter
, vid
, add
, vf
);
5976 /* Go through all the checks to see if the VLAN filter should
5977 * be wiped completely.
5979 if (!add
&& (adapter
->netdev
->flags
& IFF_PROMISC
)) {
5981 int regndx
= igb_find_vlvf_entry(adapter
, vid
);
5985 /* See if any other pools are set for this VLAN filter
5986 * entry other than the PF.
5988 vlvf
= bits
= rd32(E1000_VLVF(regndx
));
5989 bits
&= 1 << (E1000_VLVF_POOLSEL_SHIFT
+
5990 adapter
->vfs_allocated_count
);
5991 /* If the filter was removed then ensure PF pool bit
5992 * is cleared if the PF only added itself to the pool
5993 * because the PF is in promiscuous mode.
5995 if ((vlvf
& VLAN_VID_MASK
) == vid
&&
5996 !test_bit(vid
, adapter
->active_vlans
) &&
5998 igb_vlvf_set(adapter
, vid
, add
,
5999 adapter
->vfs_allocated_count
);
6006 static inline void igb_vf_reset(struct igb_adapter
*adapter
, u32 vf
)
6008 /* clear flags - except flag that indicates PF has set the MAC */
6009 adapter
->vf_data
[vf
].flags
&= IGB_VF_FLAG_PF_SET_MAC
;
6010 adapter
->vf_data
[vf
].last_nack
= jiffies
;
6012 /* reset offloads to defaults */
6013 igb_set_vmolr(adapter
, vf
, true);
6015 /* reset vlans for device */
6016 igb_clear_vf_vfta(adapter
, vf
);
6017 if (adapter
->vf_data
[vf
].pf_vlan
)
6018 igb_ndo_set_vf_vlan(adapter
->netdev
, vf
,
6019 adapter
->vf_data
[vf
].pf_vlan
,
6020 adapter
->vf_data
[vf
].pf_qos
);
6022 igb_clear_vf_vfta(adapter
, vf
);
6024 /* reset multicast table array for vf */
6025 adapter
->vf_data
[vf
].num_vf_mc_hashes
= 0;
6027 /* Flush and reset the mta with the new values */
6028 igb_set_rx_mode(adapter
->netdev
);
6031 static void igb_vf_reset_event(struct igb_adapter
*adapter
, u32 vf
)
6033 unsigned char *vf_mac
= adapter
->vf_data
[vf
].vf_mac_addresses
;
6035 /* clear mac address as we were hotplug removed/added */
6036 if (!(adapter
->vf_data
[vf
].flags
& IGB_VF_FLAG_PF_SET_MAC
))
6037 eth_zero_addr(vf_mac
);
6039 /* process remaining reset events */
6040 igb_vf_reset(adapter
, vf
);
6043 static void igb_vf_reset_msg(struct igb_adapter
*adapter
, u32 vf
)
6045 struct e1000_hw
*hw
= &adapter
->hw
;
6046 unsigned char *vf_mac
= adapter
->vf_data
[vf
].vf_mac_addresses
;
6047 int rar_entry
= hw
->mac
.rar_entry_count
- (vf
+ 1);
6049 u8
*addr
= (u8
*)(&msgbuf
[1]);
6051 /* process all the same items cleared in a function level reset */
6052 igb_vf_reset(adapter
, vf
);
6054 /* set vf mac address */
6055 igb_rar_set_qsel(adapter
, vf_mac
, rar_entry
, vf
);
6057 /* enable transmit and receive for vf */
6058 reg
= rd32(E1000_VFTE
);
6059 wr32(E1000_VFTE
, reg
| (1 << vf
));
6060 reg
= rd32(E1000_VFRE
);
6061 wr32(E1000_VFRE
, reg
| (1 << vf
));
6063 adapter
->vf_data
[vf
].flags
|= IGB_VF_FLAG_CTS
;
6065 /* reply to reset with ack and vf mac address */
6066 if (!is_zero_ether_addr(vf_mac
)) {
6067 msgbuf
[0] = E1000_VF_RESET
| E1000_VT_MSGTYPE_ACK
;
6068 memcpy(addr
, vf_mac
, ETH_ALEN
);
6070 msgbuf
[0] = E1000_VF_RESET
| E1000_VT_MSGTYPE_NACK
;
6072 igb_write_mbx(hw
, msgbuf
, 3, vf
);
6075 static int igb_set_vf_mac_addr(struct igb_adapter
*adapter
, u32
*msg
, int vf
)
6077 /* The VF MAC Address is stored in a packed array of bytes
6078 * starting at the second 32 bit word of the msg array
6080 unsigned char *addr
= (char *)&msg
[1];
6083 if (is_valid_ether_addr(addr
))
6084 err
= igb_set_vf_mac(adapter
, vf
, addr
);
6089 static void igb_rcv_ack_from_vf(struct igb_adapter
*adapter
, u32 vf
)
6091 struct e1000_hw
*hw
= &adapter
->hw
;
6092 struct vf_data_storage
*vf_data
= &adapter
->vf_data
[vf
];
6093 u32 msg
= E1000_VT_MSGTYPE_NACK
;
6095 /* if device isn't clear to send it shouldn't be reading either */
6096 if (!(vf_data
->flags
& IGB_VF_FLAG_CTS
) &&
6097 time_after(jiffies
, vf_data
->last_nack
+ (2 * HZ
))) {
6098 igb_write_mbx(hw
, &msg
, 1, vf
);
6099 vf_data
->last_nack
= jiffies
;
6103 static void igb_rcv_msg_from_vf(struct igb_adapter
*adapter
, u32 vf
)
6105 struct pci_dev
*pdev
= adapter
->pdev
;
6106 u32 msgbuf
[E1000_VFMAILBOX_SIZE
];
6107 struct e1000_hw
*hw
= &adapter
->hw
;
6108 struct vf_data_storage
*vf_data
= &adapter
->vf_data
[vf
];
6111 retval
= igb_read_mbx(hw
, msgbuf
, E1000_VFMAILBOX_SIZE
, vf
);
6114 /* if receive failed revoke VF CTS stats and restart init */
6115 dev_err(&pdev
->dev
, "Error receiving message from VF\n");
6116 vf_data
->flags
&= ~IGB_VF_FLAG_CTS
;
6117 if (!time_after(jiffies
, vf_data
->last_nack
+ (2 * HZ
)))
6122 /* this is a message we already processed, do nothing */
6123 if (msgbuf
[0] & (E1000_VT_MSGTYPE_ACK
| E1000_VT_MSGTYPE_NACK
))
6126 /* until the vf completes a reset it should not be
6127 * allowed to start any configuration.
6129 if (msgbuf
[0] == E1000_VF_RESET
) {
6130 igb_vf_reset_msg(adapter
, vf
);
6134 if (!(vf_data
->flags
& IGB_VF_FLAG_CTS
)) {
6135 if (!time_after(jiffies
, vf_data
->last_nack
+ (2 * HZ
)))
6141 switch ((msgbuf
[0] & 0xFFFF)) {
6142 case E1000_VF_SET_MAC_ADDR
:
6144 if (!(vf_data
->flags
& IGB_VF_FLAG_PF_SET_MAC
))
6145 retval
= igb_set_vf_mac_addr(adapter
, msgbuf
, vf
);
6147 dev_warn(&pdev
->dev
,
6148 "VF %d attempted to override administratively set MAC address\nReload the VF driver to resume operations\n",
6151 case E1000_VF_SET_PROMISC
:
6152 retval
= igb_set_vf_promisc(adapter
, msgbuf
, vf
);
6154 case E1000_VF_SET_MULTICAST
:
6155 retval
= igb_set_vf_multicasts(adapter
, msgbuf
, vf
);
6157 case E1000_VF_SET_LPE
:
6158 retval
= igb_set_vf_rlpml(adapter
, msgbuf
[1], vf
);
6160 case E1000_VF_SET_VLAN
:
6162 if (vf_data
->pf_vlan
)
6163 dev_warn(&pdev
->dev
,
6164 "VF %d attempted to override administratively set VLAN tag\nReload the VF driver to resume operations\n",
6167 retval
= igb_set_vf_vlan(adapter
, msgbuf
, vf
);
6170 dev_err(&pdev
->dev
, "Unhandled Msg %08x\n", msgbuf
[0]);
6175 msgbuf
[0] |= E1000_VT_MSGTYPE_CTS
;
6177 /* notify the VF of the results of what it sent us */
6179 msgbuf
[0] |= E1000_VT_MSGTYPE_NACK
;
6181 msgbuf
[0] |= E1000_VT_MSGTYPE_ACK
;
6183 igb_write_mbx(hw
, msgbuf
, 1, vf
);
6186 static void igb_msg_task(struct igb_adapter
*adapter
)
6188 struct e1000_hw
*hw
= &adapter
->hw
;
6191 for (vf
= 0; vf
< adapter
->vfs_allocated_count
; vf
++) {
6192 /* process any reset requests */
6193 if (!igb_check_for_rst(hw
, vf
))
6194 igb_vf_reset_event(adapter
, vf
);
6196 /* process any messages pending */
6197 if (!igb_check_for_msg(hw
, vf
))
6198 igb_rcv_msg_from_vf(adapter
, vf
);
6200 /* process any acks */
6201 if (!igb_check_for_ack(hw
, vf
))
6202 igb_rcv_ack_from_vf(adapter
, vf
);
6207 * igb_set_uta - Set unicast filter table address
6208 * @adapter: board private structure
6210 * The unicast table address is a register array of 32-bit registers.
6211 * The table is meant to be used in a way similar to how the MTA is used
6212 * however due to certain limitations in the hardware it is necessary to
6213 * set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
6214 * enable bit to allow vlan tag stripping when promiscuous mode is enabled
6216 static void igb_set_uta(struct igb_adapter
*adapter
)
6218 struct e1000_hw
*hw
= &adapter
->hw
;
6221 /* The UTA table only exists on 82576 hardware and newer */
6222 if (hw
->mac
.type
< e1000_82576
)
6225 /* we only need to do this if VMDq is enabled */
6226 if (!adapter
->vfs_allocated_count
)
6229 for (i
= 0; i
< hw
->mac
.uta_reg_count
; i
++)
6230 array_wr32(E1000_UTA
, i
, ~0);
6234 * igb_intr_msi - Interrupt Handler
6235 * @irq: interrupt number
6236 * @data: pointer to a network interface device structure
6238 static irqreturn_t
igb_intr_msi(int irq
, void *data
)
6240 struct igb_adapter
*adapter
= data
;
6241 struct igb_q_vector
*q_vector
= adapter
->q_vector
[0];
6242 struct e1000_hw
*hw
= &adapter
->hw
;
6243 /* read ICR disables interrupts using IAM */
6244 u32 icr
= rd32(E1000_ICR
);
6246 igb_write_itr(q_vector
);
6248 if (icr
& E1000_ICR_DRSTA
)
6249 schedule_work(&adapter
->reset_task
);
6251 if (icr
& E1000_ICR_DOUTSYNC
) {
6252 /* HW is reporting DMA is out of sync */
6253 adapter
->stats
.doosync
++;
6256 if (icr
& (E1000_ICR_RXSEQ
| E1000_ICR_LSC
)) {
6257 hw
->mac
.get_link_status
= 1;
6258 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
6259 mod_timer(&adapter
->watchdog_timer
, jiffies
+ 1);
6262 if (icr
& E1000_ICR_TS
)
6263 igb_tsync_interrupt(adapter
);
6265 napi_schedule(&q_vector
->napi
);
6271 * igb_intr - Legacy Interrupt Handler
6272 * @irq: interrupt number
6273 * @data: pointer to a network interface device structure
6275 static irqreturn_t
igb_intr(int irq
, void *data
)
6277 struct igb_adapter
*adapter
= data
;
6278 struct igb_q_vector
*q_vector
= adapter
->q_vector
[0];
6279 struct e1000_hw
*hw
= &adapter
->hw
;
6280 /* Interrupt Auto-Mask...upon reading ICR, interrupts are masked. No
6281 * need for the IMC write
6283 u32 icr
= rd32(E1000_ICR
);
6285 /* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
6286 * not set, then the adapter didn't send an interrupt
6288 if (!(icr
& E1000_ICR_INT_ASSERTED
))
6291 igb_write_itr(q_vector
);
6293 if (icr
& E1000_ICR_DRSTA
)
6294 schedule_work(&adapter
->reset_task
);
6296 if (icr
& E1000_ICR_DOUTSYNC
) {
6297 /* HW is reporting DMA is out of sync */
6298 adapter
->stats
.doosync
++;
6301 if (icr
& (E1000_ICR_RXSEQ
| E1000_ICR_LSC
)) {
6302 hw
->mac
.get_link_status
= 1;
6303 /* guard against interrupt when we're going down */
6304 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
6305 mod_timer(&adapter
->watchdog_timer
, jiffies
+ 1);
6308 if (icr
& E1000_ICR_TS
)
6309 igb_tsync_interrupt(adapter
);
6311 napi_schedule(&q_vector
->napi
);
6316 static void igb_ring_irq_enable(struct igb_q_vector
*q_vector
)
6318 struct igb_adapter
*adapter
= q_vector
->adapter
;
6319 struct e1000_hw
*hw
= &adapter
->hw
;
6321 if ((q_vector
->rx
.ring
&& (adapter
->rx_itr_setting
& 3)) ||
6322 (!q_vector
->rx
.ring
&& (adapter
->tx_itr_setting
& 3))) {
6323 if ((adapter
->num_q_vectors
== 1) && !adapter
->vf_data
)
6324 igb_set_itr(q_vector
);
6326 igb_update_ring_itr(q_vector
);
6329 if (!test_bit(__IGB_DOWN
, &adapter
->state
)) {
6330 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
)
6331 wr32(E1000_EIMS
, q_vector
->eims_value
);
6333 igb_irq_enable(adapter
);
6338 * igb_poll - NAPI Rx polling callback
6339 * @napi: napi polling structure
6340 * @budget: count of how many packets we should handle
6342 static int igb_poll(struct napi_struct
*napi
, int budget
)
6344 struct igb_q_vector
*q_vector
= container_of(napi
,
6345 struct igb_q_vector
,
6347 bool clean_complete
= true;
6349 #ifdef CONFIG_IGB_DCA
6350 if (q_vector
->adapter
->flags
& IGB_FLAG_DCA_ENABLED
)
6351 igb_update_dca(q_vector
);
6353 if (q_vector
->tx
.ring
)
6354 clean_complete
= igb_clean_tx_irq(q_vector
);
6356 if (q_vector
->rx
.ring
)
6357 clean_complete
&= igb_clean_rx_irq(q_vector
, budget
);
6359 /* If all work not completed, return budget and keep polling */
6360 if (!clean_complete
)
6363 /* If not enough Rx work done, exit the polling mode */
6364 napi_complete(napi
);
6365 igb_ring_irq_enable(q_vector
);
6371 * igb_clean_tx_irq - Reclaim resources after transmit completes
6372 * @q_vector: pointer to q_vector containing needed info
6374 * returns true if ring is completely cleaned
6376 static bool igb_clean_tx_irq(struct igb_q_vector
*q_vector
)
6378 struct igb_adapter
*adapter
= q_vector
->adapter
;
6379 struct igb_ring
*tx_ring
= q_vector
->tx
.ring
;
6380 struct igb_tx_buffer
*tx_buffer
;
6381 union e1000_adv_tx_desc
*tx_desc
;
6382 unsigned int total_bytes
= 0, total_packets
= 0;
6383 unsigned int budget
= q_vector
->tx
.work_limit
;
6384 unsigned int i
= tx_ring
->next_to_clean
;
6386 if (test_bit(__IGB_DOWN
, &adapter
->state
))
6389 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
6390 tx_desc
= IGB_TX_DESC(tx_ring
, i
);
6391 i
-= tx_ring
->count
;
6394 union e1000_adv_tx_desc
*eop_desc
= tx_buffer
->next_to_watch
;
6396 /* if next_to_watch is not set then there is no work pending */
6400 /* prevent any other reads prior to eop_desc */
6401 read_barrier_depends();
6403 /* if DD is not set pending work has not been completed */
6404 if (!(eop_desc
->wb
.status
& cpu_to_le32(E1000_TXD_STAT_DD
)))
6407 /* clear next_to_watch to prevent false hangs */
6408 tx_buffer
->next_to_watch
= NULL
;
6410 /* update the statistics for this packet */
6411 total_bytes
+= tx_buffer
->bytecount
;
6412 total_packets
+= tx_buffer
->gso_segs
;
6415 dev_consume_skb_any(tx_buffer
->skb
);
6417 /* unmap skb header data */
6418 dma_unmap_single(tx_ring
->dev
,
6419 dma_unmap_addr(tx_buffer
, dma
),
6420 dma_unmap_len(tx_buffer
, len
),
6423 /* clear tx_buffer data */
6424 tx_buffer
->skb
= NULL
;
6425 dma_unmap_len_set(tx_buffer
, len
, 0);
6427 /* clear last DMA location and unmap remaining buffers */
6428 while (tx_desc
!= eop_desc
) {
6433 i
-= tx_ring
->count
;
6434 tx_buffer
= tx_ring
->tx_buffer_info
;
6435 tx_desc
= IGB_TX_DESC(tx_ring
, 0);
6438 /* unmap any remaining paged data */
6439 if (dma_unmap_len(tx_buffer
, len
)) {
6440 dma_unmap_page(tx_ring
->dev
,
6441 dma_unmap_addr(tx_buffer
, dma
),
6442 dma_unmap_len(tx_buffer
, len
),
6444 dma_unmap_len_set(tx_buffer
, len
, 0);
6448 /* move us one more past the eop_desc for start of next pkt */
6453 i
-= tx_ring
->count
;
6454 tx_buffer
= tx_ring
->tx_buffer_info
;
6455 tx_desc
= IGB_TX_DESC(tx_ring
, 0);
6458 /* issue prefetch for next Tx descriptor */
6461 /* update budget accounting */
6463 } while (likely(budget
));
6465 netdev_tx_completed_queue(txring_txq(tx_ring
),
6466 total_packets
, total_bytes
);
6467 i
+= tx_ring
->count
;
6468 tx_ring
->next_to_clean
= i
;
6469 u64_stats_update_begin(&tx_ring
->tx_syncp
);
6470 tx_ring
->tx_stats
.bytes
+= total_bytes
;
6471 tx_ring
->tx_stats
.packets
+= total_packets
;
6472 u64_stats_update_end(&tx_ring
->tx_syncp
);
6473 q_vector
->tx
.total_bytes
+= total_bytes
;
6474 q_vector
->tx
.total_packets
+= total_packets
;
6476 if (test_bit(IGB_RING_FLAG_TX_DETECT_HANG
, &tx_ring
->flags
)) {
6477 struct e1000_hw
*hw
= &adapter
->hw
;
6479 /* Detect a transmit hang in hardware, this serializes the
6480 * check with the clearing of time_stamp and movement of i
6482 clear_bit(IGB_RING_FLAG_TX_DETECT_HANG
, &tx_ring
->flags
);
6483 if (tx_buffer
->next_to_watch
&&
6484 time_after(jiffies
, tx_buffer
->time_stamp
+
6485 (adapter
->tx_timeout_factor
* HZ
)) &&
6486 !(rd32(E1000_STATUS
) & E1000_STATUS_TXOFF
)) {
6488 /* detected Tx unit hang */
6489 dev_err(tx_ring
->dev
,
6490 "Detected Tx Unit Hang\n"
6494 " next_to_use <%x>\n"
6495 " next_to_clean <%x>\n"
6496 "buffer_info[next_to_clean]\n"
6497 " time_stamp <%lx>\n"
6498 " next_to_watch <%p>\n"
6500 " desc.status <%x>\n",
6501 tx_ring
->queue_index
,
6502 rd32(E1000_TDH(tx_ring
->reg_idx
)),
6503 readl(tx_ring
->tail
),
6504 tx_ring
->next_to_use
,
6505 tx_ring
->next_to_clean
,
6506 tx_buffer
->time_stamp
,
6507 tx_buffer
->next_to_watch
,
6509 tx_buffer
->next_to_watch
->wb
.status
);
6510 netif_stop_subqueue(tx_ring
->netdev
,
6511 tx_ring
->queue_index
);
6513 /* we are about to reset, no point in enabling stuff */
6518 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
6519 if (unlikely(total_packets
&&
6520 netif_carrier_ok(tx_ring
->netdev
) &&
6521 igb_desc_unused(tx_ring
) >= TX_WAKE_THRESHOLD
)) {
6522 /* Make sure that anybody stopping the queue after this
6523 * sees the new next_to_clean.
6526 if (__netif_subqueue_stopped(tx_ring
->netdev
,
6527 tx_ring
->queue_index
) &&
6528 !(test_bit(__IGB_DOWN
, &adapter
->state
))) {
6529 netif_wake_subqueue(tx_ring
->netdev
,
6530 tx_ring
->queue_index
);
6532 u64_stats_update_begin(&tx_ring
->tx_syncp
);
6533 tx_ring
->tx_stats
.restart_queue
++;
6534 u64_stats_update_end(&tx_ring
->tx_syncp
);
6542 * igb_reuse_rx_page - page flip buffer and store it back on the ring
6543 * @rx_ring: rx descriptor ring to store buffers on
6544 * @old_buff: donor buffer to have page reused
6546 * Synchronizes page for reuse by the adapter
6548 static void igb_reuse_rx_page(struct igb_ring
*rx_ring
,
6549 struct igb_rx_buffer
*old_buff
)
6551 struct igb_rx_buffer
*new_buff
;
6552 u16 nta
= rx_ring
->next_to_alloc
;
6554 new_buff
= &rx_ring
->rx_buffer_info
[nta
];
6556 /* update, and store next to alloc */
6558 rx_ring
->next_to_alloc
= (nta
< rx_ring
->count
) ? nta
: 0;
6560 /* transfer page from old buffer to new buffer */
6561 *new_buff
= *old_buff
;
6563 /* sync the buffer for use by the device */
6564 dma_sync_single_range_for_device(rx_ring
->dev
, old_buff
->dma
,
6565 old_buff
->page_offset
,
6570 static inline bool igb_page_is_reserved(struct page
*page
)
6572 return (page_to_nid(page
) != numa_mem_id()) || page
->pfmemalloc
;
6575 static bool igb_can_reuse_rx_page(struct igb_rx_buffer
*rx_buffer
,
6577 unsigned int truesize
)
6579 /* avoid re-using remote pages */
6580 if (unlikely(igb_page_is_reserved(page
)))
6583 #if (PAGE_SIZE < 8192)
6584 /* if we are only owner of page we can reuse it */
6585 if (unlikely(page_count(page
) != 1))
6588 /* flip page offset to other buffer */
6589 rx_buffer
->page_offset
^= IGB_RX_BUFSZ
;
6591 /* move offset up to the next cache line */
6592 rx_buffer
->page_offset
+= truesize
;
6594 if (rx_buffer
->page_offset
> (PAGE_SIZE
- IGB_RX_BUFSZ
))
6598 /* Even if we own the page, we are not allowed to use atomic_set()
6599 * This would break get_page_unless_zero() users.
6601 atomic_inc(&page
->_count
);
6607 * igb_add_rx_frag - Add contents of Rx buffer to sk_buff
6608 * @rx_ring: rx descriptor ring to transact packets on
6609 * @rx_buffer: buffer containing page to add
6610 * @rx_desc: descriptor containing length of buffer written by hardware
6611 * @skb: sk_buff to place the data into
6613 * This function will add the data contained in rx_buffer->page to the skb.
6614 * This is done either through a direct copy if the data in the buffer is
6615 * less than the skb header size, otherwise it will just attach the page as
6616 * a frag to the skb.
6618 * The function will then update the page offset if necessary and return
6619 * true if the buffer can be reused by the adapter.
6621 static bool igb_add_rx_frag(struct igb_ring
*rx_ring
,
6622 struct igb_rx_buffer
*rx_buffer
,
6623 union e1000_adv_rx_desc
*rx_desc
,
6624 struct sk_buff
*skb
)
6626 struct page
*page
= rx_buffer
->page
;
6627 unsigned int size
= le16_to_cpu(rx_desc
->wb
.upper
.length
);
6628 #if (PAGE_SIZE < 8192)
6629 unsigned int truesize
= IGB_RX_BUFSZ
;
6631 unsigned int truesize
= ALIGN(size
, L1_CACHE_BYTES
);
6634 if ((size
<= IGB_RX_HDR_LEN
) && !skb_is_nonlinear(skb
)) {
6635 unsigned char *va
= page_address(page
) + rx_buffer
->page_offset
;
6637 if (igb_test_staterr(rx_desc
, E1000_RXDADV_STAT_TSIP
)) {
6638 igb_ptp_rx_pktstamp(rx_ring
->q_vector
, va
, skb
);
6639 va
+= IGB_TS_HDR_LEN
;
6640 size
-= IGB_TS_HDR_LEN
;
6643 memcpy(__skb_put(skb
, size
), va
, ALIGN(size
, sizeof(long)));
6645 /* page is not reserved, we can reuse buffer as-is */
6646 if (likely(!igb_page_is_reserved(page
)))
6649 /* this page cannot be reused so discard it */
6654 skb_add_rx_frag(skb
, skb_shinfo(skb
)->nr_frags
, page
,
6655 rx_buffer
->page_offset
, size
, truesize
);
6657 return igb_can_reuse_rx_page(rx_buffer
, page
, truesize
);
6660 static struct sk_buff
*igb_fetch_rx_buffer(struct igb_ring
*rx_ring
,
6661 union e1000_adv_rx_desc
*rx_desc
,
6662 struct sk_buff
*skb
)
6664 struct igb_rx_buffer
*rx_buffer
;
6667 rx_buffer
= &rx_ring
->rx_buffer_info
[rx_ring
->next_to_clean
];
6668 page
= rx_buffer
->page
;
6672 void *page_addr
= page_address(page
) +
6673 rx_buffer
->page_offset
;
6675 /* prefetch first cache line of first page */
6676 prefetch(page_addr
);
6677 #if L1_CACHE_BYTES < 128
6678 prefetch(page_addr
+ L1_CACHE_BYTES
);
6681 /* allocate a skb to store the frags */
6682 skb
= napi_alloc_skb(&rx_ring
->q_vector
->napi
, IGB_RX_HDR_LEN
);
6683 if (unlikely(!skb
)) {
6684 rx_ring
->rx_stats
.alloc_failed
++;
6688 /* we will be copying header into skb->data in
6689 * pskb_may_pull so it is in our interest to prefetch
6690 * it now to avoid a possible cache miss
6692 prefetchw(skb
->data
);
6695 /* we are reusing so sync this buffer for CPU use */
6696 dma_sync_single_range_for_cpu(rx_ring
->dev
,
6698 rx_buffer
->page_offset
,
6702 /* pull page into skb */
6703 if (igb_add_rx_frag(rx_ring
, rx_buffer
, rx_desc
, skb
)) {
6704 /* hand second half of page back to the ring */
6705 igb_reuse_rx_page(rx_ring
, rx_buffer
);
6707 /* we are not reusing the buffer so unmap it */
6708 dma_unmap_page(rx_ring
->dev
, rx_buffer
->dma
,
6709 PAGE_SIZE
, DMA_FROM_DEVICE
);
6712 /* clear contents of rx_buffer */
6713 rx_buffer
->page
= NULL
;
6718 static inline void igb_rx_checksum(struct igb_ring
*ring
,
6719 union e1000_adv_rx_desc
*rx_desc
,
6720 struct sk_buff
*skb
)
6722 skb_checksum_none_assert(skb
);
6724 /* Ignore Checksum bit is set */
6725 if (igb_test_staterr(rx_desc
, E1000_RXD_STAT_IXSM
))
6728 /* Rx checksum disabled via ethtool */
6729 if (!(ring
->netdev
->features
& NETIF_F_RXCSUM
))
6732 /* TCP/UDP checksum error bit is set */
6733 if (igb_test_staterr(rx_desc
,
6734 E1000_RXDEXT_STATERR_TCPE
|
6735 E1000_RXDEXT_STATERR_IPE
)) {
6736 /* work around errata with sctp packets where the TCPE aka
6737 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
6738 * packets, (aka let the stack check the crc32c)
6740 if (!((skb
->len
== 60) &&
6741 test_bit(IGB_RING_FLAG_RX_SCTP_CSUM
, &ring
->flags
))) {
6742 u64_stats_update_begin(&ring
->rx_syncp
);
6743 ring
->rx_stats
.csum_err
++;
6744 u64_stats_update_end(&ring
->rx_syncp
);
6746 /* let the stack verify checksum errors */
6749 /* It must be a TCP or UDP packet with a valid checksum */
6750 if (igb_test_staterr(rx_desc
, E1000_RXD_STAT_TCPCS
|
6751 E1000_RXD_STAT_UDPCS
))
6752 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
6754 dev_dbg(ring
->dev
, "cksum success: bits %08X\n",
6755 le32_to_cpu(rx_desc
->wb
.upper
.status_error
));
6758 static inline void igb_rx_hash(struct igb_ring
*ring
,
6759 union e1000_adv_rx_desc
*rx_desc
,
6760 struct sk_buff
*skb
)
6762 if (ring
->netdev
->features
& NETIF_F_RXHASH
)
6764 le32_to_cpu(rx_desc
->wb
.lower
.hi_dword
.rss
),
6769 * igb_is_non_eop - process handling of non-EOP buffers
6770 * @rx_ring: Rx ring being processed
6771 * @rx_desc: Rx descriptor for current buffer
6772 * @skb: current socket buffer containing buffer in progress
6774 * This function updates next to clean. If the buffer is an EOP buffer
6775 * this function exits returning false, otherwise it will place the
6776 * sk_buff in the next buffer to be chained and return true indicating
6777 * that this is in fact a non-EOP buffer.
6779 static bool igb_is_non_eop(struct igb_ring
*rx_ring
,
6780 union e1000_adv_rx_desc
*rx_desc
)
6782 u32 ntc
= rx_ring
->next_to_clean
+ 1;
6784 /* fetch, update, and store next to clean */
6785 ntc
= (ntc
< rx_ring
->count
) ? ntc
: 0;
6786 rx_ring
->next_to_clean
= ntc
;
6788 prefetch(IGB_RX_DESC(rx_ring
, ntc
));
6790 if (likely(igb_test_staterr(rx_desc
, E1000_RXD_STAT_EOP
)))
6797 * igb_pull_tail - igb specific version of skb_pull_tail
6798 * @rx_ring: rx descriptor ring packet is being transacted on
6799 * @rx_desc: pointer to the EOP Rx descriptor
6800 * @skb: pointer to current skb being adjusted
6802 * This function is an igb specific version of __pskb_pull_tail. The
6803 * main difference between this version and the original function is that
6804 * this function can make several assumptions about the state of things
6805 * that allow for significant optimizations versus the standard function.
6806 * As a result we can do things like drop a frag and maintain an accurate
6807 * truesize for the skb.
6809 static void igb_pull_tail(struct igb_ring
*rx_ring
,
6810 union e1000_adv_rx_desc
*rx_desc
,
6811 struct sk_buff
*skb
)
6813 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[0];
6815 unsigned int pull_len
;
6817 /* it is valid to use page_address instead of kmap since we are
6818 * working with pages allocated out of the lomem pool per
6819 * alloc_page(GFP_ATOMIC)
6821 va
= skb_frag_address(frag
);
6823 if (igb_test_staterr(rx_desc
, E1000_RXDADV_STAT_TSIP
)) {
6824 /* retrieve timestamp from buffer */
6825 igb_ptp_rx_pktstamp(rx_ring
->q_vector
, va
, skb
);
6827 /* update pointers to remove timestamp header */
6828 skb_frag_size_sub(frag
, IGB_TS_HDR_LEN
);
6829 frag
->page_offset
+= IGB_TS_HDR_LEN
;
6830 skb
->data_len
-= IGB_TS_HDR_LEN
;
6831 skb
->len
-= IGB_TS_HDR_LEN
;
6833 /* move va to start of packet data */
6834 va
+= IGB_TS_HDR_LEN
;
6837 /* we need the header to contain the greater of either ETH_HLEN or
6838 * 60 bytes if the skb->len is less than 60 for skb_pad.
6840 pull_len
= eth_get_headlen(va
, IGB_RX_HDR_LEN
);
6842 /* align pull length to size of long to optimize memcpy performance */
6843 skb_copy_to_linear_data(skb
, va
, ALIGN(pull_len
, sizeof(long)));
6845 /* update all of the pointers */
6846 skb_frag_size_sub(frag
, pull_len
);
6847 frag
->page_offset
+= pull_len
;
6848 skb
->data_len
-= pull_len
;
6849 skb
->tail
+= pull_len
;
6853 * igb_cleanup_headers - Correct corrupted or empty headers
6854 * @rx_ring: rx descriptor ring packet is being transacted on
6855 * @rx_desc: pointer to the EOP Rx descriptor
6856 * @skb: pointer to current skb being fixed
6858 * Address the case where we are pulling data in on pages only
6859 * and as such no data is present in the skb header.
6861 * In addition if skb is not at least 60 bytes we need to pad it so that
6862 * it is large enough to qualify as a valid Ethernet frame.
6864 * Returns true if an error was encountered and skb was freed.
6866 static bool igb_cleanup_headers(struct igb_ring
*rx_ring
,
6867 union e1000_adv_rx_desc
*rx_desc
,
6868 struct sk_buff
*skb
)
6870 if (unlikely((igb_test_staterr(rx_desc
,
6871 E1000_RXDEXT_ERR_FRAME_ERR_MASK
)))) {
6872 struct net_device
*netdev
= rx_ring
->netdev
;
6873 if (!(netdev
->features
& NETIF_F_RXALL
)) {
6874 dev_kfree_skb_any(skb
);
6879 /* place header in linear portion of buffer */
6880 if (skb_is_nonlinear(skb
))
6881 igb_pull_tail(rx_ring
, rx_desc
, skb
);
6883 /* if eth_skb_pad returns an error the skb was freed */
6884 if (eth_skb_pad(skb
))
6891 * igb_process_skb_fields - Populate skb header fields from Rx descriptor
6892 * @rx_ring: rx descriptor ring packet is being transacted on
6893 * @rx_desc: pointer to the EOP Rx descriptor
6894 * @skb: pointer to current skb being populated
6896 * This function checks the ring, descriptor, and packet information in
6897 * order to populate the hash, checksum, VLAN, timestamp, protocol, and
6898 * other fields within the skb.
6900 static void igb_process_skb_fields(struct igb_ring
*rx_ring
,
6901 union e1000_adv_rx_desc
*rx_desc
,
6902 struct sk_buff
*skb
)
6904 struct net_device
*dev
= rx_ring
->netdev
;
6906 igb_rx_hash(rx_ring
, rx_desc
, skb
);
6908 igb_rx_checksum(rx_ring
, rx_desc
, skb
);
6910 if (igb_test_staterr(rx_desc
, E1000_RXDADV_STAT_TS
) &&
6911 !igb_test_staterr(rx_desc
, E1000_RXDADV_STAT_TSIP
))
6912 igb_ptp_rx_rgtstamp(rx_ring
->q_vector
, skb
);
6914 if ((dev
->features
& NETIF_F_HW_VLAN_CTAG_RX
) &&
6915 igb_test_staterr(rx_desc
, E1000_RXD_STAT_VP
)) {
6918 if (igb_test_staterr(rx_desc
, E1000_RXDEXT_STATERR_LB
) &&
6919 test_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP
, &rx_ring
->flags
))
6920 vid
= be16_to_cpu(rx_desc
->wb
.upper
.vlan
);
6922 vid
= le16_to_cpu(rx_desc
->wb
.upper
.vlan
);
6924 __vlan_hwaccel_put_tag(skb
, htons(ETH_P_8021Q
), vid
);
6927 skb_record_rx_queue(skb
, rx_ring
->queue_index
);
6929 skb
->protocol
= eth_type_trans(skb
, rx_ring
->netdev
);
6932 static bool igb_clean_rx_irq(struct igb_q_vector
*q_vector
, const int budget
)
6934 struct igb_ring
*rx_ring
= q_vector
->rx
.ring
;
6935 struct sk_buff
*skb
= rx_ring
->skb
;
6936 unsigned int total_bytes
= 0, total_packets
= 0;
6937 u16 cleaned_count
= igb_desc_unused(rx_ring
);
6939 while (likely(total_packets
< budget
)) {
6940 union e1000_adv_rx_desc
*rx_desc
;
6942 /* return some buffers to hardware, one at a time is too slow */
6943 if (cleaned_count
>= IGB_RX_BUFFER_WRITE
) {
6944 igb_alloc_rx_buffers(rx_ring
, cleaned_count
);
6948 rx_desc
= IGB_RX_DESC(rx_ring
, rx_ring
->next_to_clean
);
6950 if (!rx_desc
->wb
.upper
.status_error
)
6953 /* This memory barrier is needed to keep us from reading
6954 * any other fields out of the rx_desc until we know the
6955 * descriptor has been written back
6959 /* retrieve a buffer from the ring */
6960 skb
= igb_fetch_rx_buffer(rx_ring
, rx_desc
, skb
);
6962 /* exit if we failed to retrieve a buffer */
6968 /* fetch next buffer in frame if non-eop */
6969 if (igb_is_non_eop(rx_ring
, rx_desc
))
6972 /* verify the packet layout is correct */
6973 if (igb_cleanup_headers(rx_ring
, rx_desc
, skb
)) {
6978 /* probably a little skewed due to removing CRC */
6979 total_bytes
+= skb
->len
;
6981 /* populate checksum, timestamp, VLAN, and protocol */
6982 igb_process_skb_fields(rx_ring
, rx_desc
, skb
);
6984 napi_gro_receive(&q_vector
->napi
, skb
);
6986 /* reset skb pointer */
6989 /* update budget accounting */
6993 /* place incomplete frames back on ring for completion */
6996 u64_stats_update_begin(&rx_ring
->rx_syncp
);
6997 rx_ring
->rx_stats
.packets
+= total_packets
;
6998 rx_ring
->rx_stats
.bytes
+= total_bytes
;
6999 u64_stats_update_end(&rx_ring
->rx_syncp
);
7000 q_vector
->rx
.total_packets
+= total_packets
;
7001 q_vector
->rx
.total_bytes
+= total_bytes
;
7004 igb_alloc_rx_buffers(rx_ring
, cleaned_count
);
7006 return total_packets
< budget
;
7009 static bool igb_alloc_mapped_page(struct igb_ring
*rx_ring
,
7010 struct igb_rx_buffer
*bi
)
7012 struct page
*page
= bi
->page
;
7015 /* since we are recycling buffers we should seldom need to alloc */
7019 /* alloc new page for storage */
7020 page
= dev_alloc_page();
7021 if (unlikely(!page
)) {
7022 rx_ring
->rx_stats
.alloc_failed
++;
7026 /* map page for use */
7027 dma
= dma_map_page(rx_ring
->dev
, page
, 0, PAGE_SIZE
, DMA_FROM_DEVICE
);
7029 /* if mapping failed free memory back to system since
7030 * there isn't much point in holding memory we can't use
7032 if (dma_mapping_error(rx_ring
->dev
, dma
)) {
7035 rx_ring
->rx_stats
.alloc_failed
++;
7041 bi
->page_offset
= 0;
7047 * igb_alloc_rx_buffers - Replace used receive buffers; packet split
7048 * @adapter: address of board private structure
7050 void igb_alloc_rx_buffers(struct igb_ring
*rx_ring
, u16 cleaned_count
)
7052 union e1000_adv_rx_desc
*rx_desc
;
7053 struct igb_rx_buffer
*bi
;
7054 u16 i
= rx_ring
->next_to_use
;
7060 rx_desc
= IGB_RX_DESC(rx_ring
, i
);
7061 bi
= &rx_ring
->rx_buffer_info
[i
];
7062 i
-= rx_ring
->count
;
7065 if (!igb_alloc_mapped_page(rx_ring
, bi
))
7068 /* Refresh the desc even if buffer_addrs didn't change
7069 * because each write-back erases this info.
7071 rx_desc
->read
.pkt_addr
= cpu_to_le64(bi
->dma
+ bi
->page_offset
);
7077 rx_desc
= IGB_RX_DESC(rx_ring
, 0);
7078 bi
= rx_ring
->rx_buffer_info
;
7079 i
-= rx_ring
->count
;
7082 /* clear the status bits for the next_to_use descriptor */
7083 rx_desc
->wb
.upper
.status_error
= 0;
7086 } while (cleaned_count
);
7088 i
+= rx_ring
->count
;
7090 if (rx_ring
->next_to_use
!= i
) {
7091 /* record the next descriptor to use */
7092 rx_ring
->next_to_use
= i
;
7094 /* update next to alloc since we have filled the ring */
7095 rx_ring
->next_to_alloc
= i
;
7097 /* Force memory writes to complete before letting h/w
7098 * know there are new descriptors to fetch. (Only
7099 * applicable for weak-ordered memory model archs,
7103 writel(i
, rx_ring
->tail
);
7113 static int igb_mii_ioctl(struct net_device
*netdev
, struct ifreq
*ifr
, int cmd
)
7115 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7116 struct mii_ioctl_data
*data
= if_mii(ifr
);
7118 if (adapter
->hw
.phy
.media_type
!= e1000_media_type_copper
)
7123 data
->phy_id
= adapter
->hw
.phy
.addr
;
7126 if (igb_read_phy_reg(&adapter
->hw
, data
->reg_num
& 0x1F,
7143 static int igb_ioctl(struct net_device
*netdev
, struct ifreq
*ifr
, int cmd
)
7149 return igb_mii_ioctl(netdev
, ifr
, cmd
);
7151 return igb_ptp_get_ts_config(netdev
, ifr
);
7153 return igb_ptp_set_ts_config(netdev
, ifr
);
7159 void igb_read_pci_cfg(struct e1000_hw
*hw
, u32 reg
, u16
*value
)
7161 struct igb_adapter
*adapter
= hw
->back
;
7163 pci_read_config_word(adapter
->pdev
, reg
, value
);
7166 void igb_write_pci_cfg(struct e1000_hw
*hw
, u32 reg
, u16
*value
)
7168 struct igb_adapter
*adapter
= hw
->back
;
7170 pci_write_config_word(adapter
->pdev
, reg
, *value
);
7173 s32
igb_read_pcie_cap_reg(struct e1000_hw
*hw
, u32 reg
, u16
*value
)
7175 struct igb_adapter
*adapter
= hw
->back
;
7177 if (pcie_capability_read_word(adapter
->pdev
, reg
, value
))
7178 return -E1000_ERR_CONFIG
;
7183 s32
igb_write_pcie_cap_reg(struct e1000_hw
*hw
, u32 reg
, u16
*value
)
7185 struct igb_adapter
*adapter
= hw
->back
;
7187 if (pcie_capability_write_word(adapter
->pdev
, reg
, *value
))
7188 return -E1000_ERR_CONFIG
;
7193 static void igb_vlan_mode(struct net_device
*netdev
, netdev_features_t features
)
7195 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7196 struct e1000_hw
*hw
= &adapter
->hw
;
7198 bool enable
= !!(features
& NETIF_F_HW_VLAN_CTAG_RX
);
7201 /* enable VLAN tag insert/strip */
7202 ctrl
= rd32(E1000_CTRL
);
7203 ctrl
|= E1000_CTRL_VME
;
7204 wr32(E1000_CTRL
, ctrl
);
7206 /* Disable CFI check */
7207 rctl
= rd32(E1000_RCTL
);
7208 rctl
&= ~E1000_RCTL_CFIEN
;
7209 wr32(E1000_RCTL
, rctl
);
7211 /* disable VLAN tag insert/strip */
7212 ctrl
= rd32(E1000_CTRL
);
7213 ctrl
&= ~E1000_CTRL_VME
;
7214 wr32(E1000_CTRL
, ctrl
);
7217 igb_rlpml_set(adapter
);
7220 static int igb_vlan_rx_add_vid(struct net_device
*netdev
,
7221 __be16 proto
, u16 vid
)
7223 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7224 struct e1000_hw
*hw
= &adapter
->hw
;
7225 int pf_id
= adapter
->vfs_allocated_count
;
7227 /* attempt to add filter to vlvf array */
7228 igb_vlvf_set(adapter
, vid
, true, pf_id
);
7230 /* add the filter since PF can receive vlans w/o entry in vlvf */
7231 igb_vfta_set(hw
, vid
, true);
7233 set_bit(vid
, adapter
->active_vlans
);
7238 static int igb_vlan_rx_kill_vid(struct net_device
*netdev
,
7239 __be16 proto
, u16 vid
)
7241 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7242 struct e1000_hw
*hw
= &adapter
->hw
;
7243 int pf_id
= adapter
->vfs_allocated_count
;
7246 /* remove vlan from VLVF table array */
7247 err
= igb_vlvf_set(adapter
, vid
, false, pf_id
);
7249 /* if vid was not present in VLVF just remove it from table */
7251 igb_vfta_set(hw
, vid
, false);
7253 clear_bit(vid
, adapter
->active_vlans
);
7258 static void igb_restore_vlan(struct igb_adapter
*adapter
)
7262 igb_vlan_mode(adapter
->netdev
, adapter
->netdev
->features
);
7264 for_each_set_bit(vid
, adapter
->active_vlans
, VLAN_N_VID
)
7265 igb_vlan_rx_add_vid(adapter
->netdev
, htons(ETH_P_8021Q
), vid
);
7268 int igb_set_spd_dplx(struct igb_adapter
*adapter
, u32 spd
, u8 dplx
)
7270 struct pci_dev
*pdev
= adapter
->pdev
;
7271 struct e1000_mac_info
*mac
= &adapter
->hw
.mac
;
7275 /* Make sure dplx is at most 1 bit and lsb of speed is not set
7276 * for the switch() below to work
7278 if ((spd
& 1) || (dplx
& ~1))
7281 /* Fiber NIC's only allow 1000 gbps Full duplex
7282 * and 100Mbps Full duplex for 100baseFx sfp
7284 if (adapter
->hw
.phy
.media_type
== e1000_media_type_internal_serdes
) {
7285 switch (spd
+ dplx
) {
7286 case SPEED_10
+ DUPLEX_HALF
:
7287 case SPEED_10
+ DUPLEX_FULL
:
7288 case SPEED_100
+ DUPLEX_HALF
:
7295 switch (spd
+ dplx
) {
7296 case SPEED_10
+ DUPLEX_HALF
:
7297 mac
->forced_speed_duplex
= ADVERTISE_10_HALF
;
7299 case SPEED_10
+ DUPLEX_FULL
:
7300 mac
->forced_speed_duplex
= ADVERTISE_10_FULL
;
7302 case SPEED_100
+ DUPLEX_HALF
:
7303 mac
->forced_speed_duplex
= ADVERTISE_100_HALF
;
7305 case SPEED_100
+ DUPLEX_FULL
:
7306 mac
->forced_speed_duplex
= ADVERTISE_100_FULL
;
7308 case SPEED_1000
+ DUPLEX_FULL
:
7310 adapter
->hw
.phy
.autoneg_advertised
= ADVERTISE_1000_FULL
;
7312 case SPEED_1000
+ DUPLEX_HALF
: /* not supported */
7317 /* clear MDI, MDI(-X) override is only allowed when autoneg enabled */
7318 adapter
->hw
.phy
.mdix
= AUTO_ALL_MODES
;
7323 dev_err(&pdev
->dev
, "Unsupported Speed/Duplex configuration\n");
7327 static int __igb_shutdown(struct pci_dev
*pdev
, bool *enable_wake
,
7330 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7331 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7332 struct e1000_hw
*hw
= &adapter
->hw
;
7333 u32 ctrl
, rctl
, status
;
7334 u32 wufc
= runtime
? E1000_WUFC_LNKC
: adapter
->wol
;
7339 netif_device_detach(netdev
);
7341 if (netif_running(netdev
))
7342 __igb_close(netdev
, true);
7344 igb_clear_interrupt_scheme(adapter
);
7347 retval
= pci_save_state(pdev
);
7352 status
= rd32(E1000_STATUS
);
7353 if (status
& E1000_STATUS_LU
)
7354 wufc
&= ~E1000_WUFC_LNKC
;
7357 igb_setup_rctl(adapter
);
7358 igb_set_rx_mode(netdev
);
7360 /* turn on all-multi mode if wake on multicast is enabled */
7361 if (wufc
& E1000_WUFC_MC
) {
7362 rctl
= rd32(E1000_RCTL
);
7363 rctl
|= E1000_RCTL_MPE
;
7364 wr32(E1000_RCTL
, rctl
);
7367 ctrl
= rd32(E1000_CTRL
);
7368 /* advertise wake from D3Cold */
7369 #define E1000_CTRL_ADVD3WUC 0x00100000
7370 /* phy power management enable */
7371 #define E1000_CTRL_EN_PHY_PWR_MGMT 0x00200000
7372 ctrl
|= E1000_CTRL_ADVD3WUC
;
7373 wr32(E1000_CTRL
, ctrl
);
7375 /* Allow time for pending master requests to run */
7376 igb_disable_pcie_master(hw
);
7378 wr32(E1000_WUC
, E1000_WUC_PME_EN
);
7379 wr32(E1000_WUFC
, wufc
);
7382 wr32(E1000_WUFC
, 0);
7385 *enable_wake
= wufc
|| adapter
->en_mng_pt
;
7387 igb_power_down_link(adapter
);
7389 igb_power_up_link(adapter
);
7391 /* Release control of h/w to f/w. If f/w is AMT enabled, this
7392 * would have already happened in close and is redundant.
7394 igb_release_hw_control(adapter
);
7396 pci_disable_device(pdev
);
7402 #ifdef CONFIG_PM_SLEEP
7403 static int igb_suspend(struct device
*dev
)
7407 struct pci_dev
*pdev
= to_pci_dev(dev
);
7409 retval
= __igb_shutdown(pdev
, &wake
, 0);
7414 pci_prepare_to_sleep(pdev
);
7416 pci_wake_from_d3(pdev
, false);
7417 pci_set_power_state(pdev
, PCI_D3hot
);
7422 #endif /* CONFIG_PM_SLEEP */
7424 static int igb_resume(struct device
*dev
)
7426 struct pci_dev
*pdev
= to_pci_dev(dev
);
7427 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7428 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7429 struct e1000_hw
*hw
= &adapter
->hw
;
7432 pci_set_power_state(pdev
, PCI_D0
);
7433 pci_restore_state(pdev
);
7434 pci_save_state(pdev
);
7436 if (!pci_device_is_present(pdev
))
7438 err
= pci_enable_device_mem(pdev
);
7441 "igb: Cannot enable PCI device from suspend\n");
7444 pci_set_master(pdev
);
7446 pci_enable_wake(pdev
, PCI_D3hot
, 0);
7447 pci_enable_wake(pdev
, PCI_D3cold
, 0);
7449 if (igb_init_interrupt_scheme(adapter
, true)) {
7450 dev_err(&pdev
->dev
, "Unable to allocate memory for queues\n");
7456 /* let the f/w know that the h/w is now under the control of the
7459 igb_get_hw_control(adapter
);
7461 wr32(E1000_WUS
, ~0);
7463 if (netdev
->flags
& IFF_UP
) {
7465 err
= __igb_open(netdev
, true);
7471 netif_device_attach(netdev
);
7475 static int igb_runtime_idle(struct device
*dev
)
7477 struct pci_dev
*pdev
= to_pci_dev(dev
);
7478 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7479 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7481 if (!igb_has_link(adapter
))
7482 pm_schedule_suspend(dev
, MSEC_PER_SEC
* 5);
7487 static int igb_runtime_suspend(struct device
*dev
)
7489 struct pci_dev
*pdev
= to_pci_dev(dev
);
7493 retval
= __igb_shutdown(pdev
, &wake
, 1);
7498 pci_prepare_to_sleep(pdev
);
7500 pci_wake_from_d3(pdev
, false);
7501 pci_set_power_state(pdev
, PCI_D3hot
);
7507 static int igb_runtime_resume(struct device
*dev
)
7509 return igb_resume(dev
);
7511 #endif /* CONFIG_PM */
7513 static void igb_shutdown(struct pci_dev
*pdev
)
7517 __igb_shutdown(pdev
, &wake
, 0);
7519 if (system_state
== SYSTEM_POWER_OFF
) {
7520 pci_wake_from_d3(pdev
, wake
);
7521 pci_set_power_state(pdev
, PCI_D3hot
);
7525 #ifdef CONFIG_PCI_IOV
7526 static int igb_sriov_reinit(struct pci_dev
*dev
)
7528 struct net_device
*netdev
= pci_get_drvdata(dev
);
7529 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7530 struct pci_dev
*pdev
= adapter
->pdev
;
7534 if (netif_running(netdev
))
7539 igb_clear_interrupt_scheme(adapter
);
7541 igb_init_queue_configuration(adapter
);
7543 if (igb_init_interrupt_scheme(adapter
, true)) {
7544 dev_err(&pdev
->dev
, "Unable to allocate memory for queues\n");
7548 if (netif_running(netdev
))
7556 static int igb_pci_disable_sriov(struct pci_dev
*dev
)
7558 int err
= igb_disable_sriov(dev
);
7561 err
= igb_sriov_reinit(dev
);
7566 static int igb_pci_enable_sriov(struct pci_dev
*dev
, int num_vfs
)
7568 int err
= igb_enable_sriov(dev
, num_vfs
);
7573 err
= igb_sriov_reinit(dev
);
7582 static int igb_pci_sriov_configure(struct pci_dev
*dev
, int num_vfs
)
7584 #ifdef CONFIG_PCI_IOV
7586 return igb_pci_disable_sriov(dev
);
7588 return igb_pci_enable_sriov(dev
, num_vfs
);
7593 #ifdef CONFIG_NET_POLL_CONTROLLER
7594 /* Polling 'interrupt' - used by things like netconsole to send skbs
7595 * without having to re-enable interrupts. It's not called while
7596 * the interrupt routine is executing.
7598 static void igb_netpoll(struct net_device
*netdev
)
7600 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7601 struct e1000_hw
*hw
= &adapter
->hw
;
7602 struct igb_q_vector
*q_vector
;
7605 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
7606 q_vector
= adapter
->q_vector
[i
];
7607 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
)
7608 wr32(E1000_EIMC
, q_vector
->eims_value
);
7610 igb_irq_disable(adapter
);
7611 napi_schedule(&q_vector
->napi
);
7614 #endif /* CONFIG_NET_POLL_CONTROLLER */
7617 * igb_io_error_detected - called when PCI error is detected
7618 * @pdev: Pointer to PCI device
7619 * @state: The current pci connection state
7621 * This function is called after a PCI bus error affecting
7622 * this device has been detected.
7624 static pci_ers_result_t
igb_io_error_detected(struct pci_dev
*pdev
,
7625 pci_channel_state_t state
)
7627 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7628 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7630 netif_device_detach(netdev
);
7632 if (state
== pci_channel_io_perm_failure
)
7633 return PCI_ERS_RESULT_DISCONNECT
;
7635 if (netif_running(netdev
))
7637 pci_disable_device(pdev
);
7639 /* Request a slot slot reset. */
7640 return PCI_ERS_RESULT_NEED_RESET
;
7644 * igb_io_slot_reset - called after the pci bus has been reset.
7645 * @pdev: Pointer to PCI device
7647 * Restart the card from scratch, as if from a cold-boot. Implementation
7648 * resembles the first-half of the igb_resume routine.
7650 static pci_ers_result_t
igb_io_slot_reset(struct pci_dev
*pdev
)
7652 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7653 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7654 struct e1000_hw
*hw
= &adapter
->hw
;
7655 pci_ers_result_t result
;
7658 if (pci_enable_device_mem(pdev
)) {
7660 "Cannot re-enable PCI device after reset.\n");
7661 result
= PCI_ERS_RESULT_DISCONNECT
;
7663 pci_set_master(pdev
);
7664 pci_restore_state(pdev
);
7665 pci_save_state(pdev
);
7667 pci_enable_wake(pdev
, PCI_D3hot
, 0);
7668 pci_enable_wake(pdev
, PCI_D3cold
, 0);
7671 wr32(E1000_WUS
, ~0);
7672 result
= PCI_ERS_RESULT_RECOVERED
;
7675 err
= pci_cleanup_aer_uncorrect_error_status(pdev
);
7678 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
7680 /* non-fatal, continue */
7687 * igb_io_resume - called when traffic can start flowing again.
7688 * @pdev: Pointer to PCI device
7690 * This callback is called when the error recovery driver tells us that
7691 * its OK to resume normal operation. Implementation resembles the
7692 * second-half of the igb_resume routine.
7694 static void igb_io_resume(struct pci_dev
*pdev
)
7696 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7697 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7699 if (netif_running(netdev
)) {
7700 if (igb_up(adapter
)) {
7701 dev_err(&pdev
->dev
, "igb_up failed after reset\n");
7706 netif_device_attach(netdev
);
7708 /* let the f/w know that the h/w is now under the control of the
7711 igb_get_hw_control(adapter
);
7714 static void igb_rar_set_qsel(struct igb_adapter
*adapter
, u8
*addr
, u32 index
,
7717 u32 rar_low
, rar_high
;
7718 struct e1000_hw
*hw
= &adapter
->hw
;
7720 /* HW expects these in little endian so we reverse the byte order
7721 * from network order (big endian) to little endian
7723 rar_low
= ((u32
) addr
[0] | ((u32
) addr
[1] << 8) |
7724 ((u32
) addr
[2] << 16) | ((u32
) addr
[3] << 24));
7725 rar_high
= ((u32
) addr
[4] | ((u32
) addr
[5] << 8));
7727 /* Indicate to hardware the Address is Valid. */
7728 rar_high
|= E1000_RAH_AV
;
7730 if (hw
->mac
.type
== e1000_82575
)
7731 rar_high
|= E1000_RAH_POOL_1
* qsel
;
7733 rar_high
|= E1000_RAH_POOL_1
<< qsel
;
7735 wr32(E1000_RAL(index
), rar_low
);
7737 wr32(E1000_RAH(index
), rar_high
);
7741 static int igb_set_vf_mac(struct igb_adapter
*adapter
,
7742 int vf
, unsigned char *mac_addr
)
7744 struct e1000_hw
*hw
= &adapter
->hw
;
7745 /* VF MAC addresses start at end of receive addresses and moves
7746 * towards the first, as a result a collision should not be possible
7748 int rar_entry
= hw
->mac
.rar_entry_count
- (vf
+ 1);
7750 memcpy(adapter
->vf_data
[vf
].vf_mac_addresses
, mac_addr
, ETH_ALEN
);
7752 igb_rar_set_qsel(adapter
, mac_addr
, rar_entry
, vf
);
7757 static int igb_ndo_set_vf_mac(struct net_device
*netdev
, int vf
, u8
*mac
)
7759 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7760 if (!is_valid_ether_addr(mac
) || (vf
>= adapter
->vfs_allocated_count
))
7762 adapter
->vf_data
[vf
].flags
|= IGB_VF_FLAG_PF_SET_MAC
;
7763 dev_info(&adapter
->pdev
->dev
, "setting MAC %pM on VF %d\n", mac
, vf
);
7764 dev_info(&adapter
->pdev
->dev
,
7765 "Reload the VF driver to make this change effective.");
7766 if (test_bit(__IGB_DOWN
, &adapter
->state
)) {
7767 dev_warn(&adapter
->pdev
->dev
,
7768 "The VF MAC address has been set, but the PF device is not up.\n");
7769 dev_warn(&adapter
->pdev
->dev
,
7770 "Bring the PF device up before attempting to use the VF device.\n");
7772 return igb_set_vf_mac(adapter
, vf
, mac
);
7775 static int igb_link_mbps(int internal_link_speed
)
7777 switch (internal_link_speed
) {
7787 static void igb_set_vf_rate_limit(struct e1000_hw
*hw
, int vf
, int tx_rate
,
7794 /* Calculate the rate factor values to set */
7795 rf_int
= link_speed
/ tx_rate
;
7796 rf_dec
= (link_speed
- (rf_int
* tx_rate
));
7797 rf_dec
= (rf_dec
* (1 << E1000_RTTBCNRC_RF_INT_SHIFT
)) /
7800 bcnrc_val
= E1000_RTTBCNRC_RS_ENA
;
7801 bcnrc_val
|= ((rf_int
<< E1000_RTTBCNRC_RF_INT_SHIFT
) &
7802 E1000_RTTBCNRC_RF_INT_MASK
);
7803 bcnrc_val
|= (rf_dec
& E1000_RTTBCNRC_RF_DEC_MASK
);
7808 wr32(E1000_RTTDQSEL
, vf
); /* vf X uses queue X */
7809 /* Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
7810 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported.
7812 wr32(E1000_RTTBCNRM
, 0x14);
7813 wr32(E1000_RTTBCNRC
, bcnrc_val
);
7816 static void igb_check_vf_rate_limit(struct igb_adapter
*adapter
)
7818 int actual_link_speed
, i
;
7819 bool reset_rate
= false;
7821 /* VF TX rate limit was not set or not supported */
7822 if ((adapter
->vf_rate_link_speed
== 0) ||
7823 (adapter
->hw
.mac
.type
!= e1000_82576
))
7826 actual_link_speed
= igb_link_mbps(adapter
->link_speed
);
7827 if (actual_link_speed
!= adapter
->vf_rate_link_speed
) {
7829 adapter
->vf_rate_link_speed
= 0;
7830 dev_info(&adapter
->pdev
->dev
,
7831 "Link speed has been changed. VF Transmit rate is disabled\n");
7834 for (i
= 0; i
< adapter
->vfs_allocated_count
; i
++) {
7836 adapter
->vf_data
[i
].tx_rate
= 0;
7838 igb_set_vf_rate_limit(&adapter
->hw
, i
,
7839 adapter
->vf_data
[i
].tx_rate
,
7844 static int igb_ndo_set_vf_bw(struct net_device
*netdev
, int vf
,
7845 int min_tx_rate
, int max_tx_rate
)
7847 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7848 struct e1000_hw
*hw
= &adapter
->hw
;
7849 int actual_link_speed
;
7851 if (hw
->mac
.type
!= e1000_82576
)
7857 actual_link_speed
= igb_link_mbps(adapter
->link_speed
);
7858 if ((vf
>= adapter
->vfs_allocated_count
) ||
7859 (!(rd32(E1000_STATUS
) & E1000_STATUS_LU
)) ||
7860 (max_tx_rate
< 0) ||
7861 (max_tx_rate
> actual_link_speed
))
7864 adapter
->vf_rate_link_speed
= actual_link_speed
;
7865 adapter
->vf_data
[vf
].tx_rate
= (u16
)max_tx_rate
;
7866 igb_set_vf_rate_limit(hw
, vf
, max_tx_rate
, actual_link_speed
);
7871 static int igb_ndo_set_vf_spoofchk(struct net_device
*netdev
, int vf
,
7874 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7875 struct e1000_hw
*hw
= &adapter
->hw
;
7876 u32 reg_val
, reg_offset
;
7878 if (!adapter
->vfs_allocated_count
)
7881 if (vf
>= adapter
->vfs_allocated_count
)
7884 reg_offset
= (hw
->mac
.type
== e1000_82576
) ? E1000_DTXSWC
: E1000_TXSWC
;
7885 reg_val
= rd32(reg_offset
);
7887 reg_val
|= ((1 << vf
) |
7888 (1 << (vf
+ E1000_DTXSWC_VLAN_SPOOF_SHIFT
)));
7890 reg_val
&= ~((1 << vf
) |
7891 (1 << (vf
+ E1000_DTXSWC_VLAN_SPOOF_SHIFT
)));
7892 wr32(reg_offset
, reg_val
);
7894 adapter
->vf_data
[vf
].spoofchk_enabled
= setting
;
7898 static int igb_ndo_get_vf_config(struct net_device
*netdev
,
7899 int vf
, struct ifla_vf_info
*ivi
)
7901 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7902 if (vf
>= adapter
->vfs_allocated_count
)
7905 memcpy(&ivi
->mac
, adapter
->vf_data
[vf
].vf_mac_addresses
, ETH_ALEN
);
7906 ivi
->max_tx_rate
= adapter
->vf_data
[vf
].tx_rate
;
7907 ivi
->min_tx_rate
= 0;
7908 ivi
->vlan
= adapter
->vf_data
[vf
].pf_vlan
;
7909 ivi
->qos
= adapter
->vf_data
[vf
].pf_qos
;
7910 ivi
->spoofchk
= adapter
->vf_data
[vf
].spoofchk_enabled
;
7914 static void igb_vmm_control(struct igb_adapter
*adapter
)
7916 struct e1000_hw
*hw
= &adapter
->hw
;
7919 switch (hw
->mac
.type
) {
7925 /* replication is not supported for 82575 */
7928 /* notify HW that the MAC is adding vlan tags */
7929 reg
= rd32(E1000_DTXCTL
);
7930 reg
|= E1000_DTXCTL_VLAN_ADDED
;
7931 wr32(E1000_DTXCTL
, reg
);
7934 /* enable replication vlan tag stripping */
7935 reg
= rd32(E1000_RPLOLR
);
7936 reg
|= E1000_RPLOLR_STRVLAN
;
7937 wr32(E1000_RPLOLR
, reg
);
7940 /* none of the above registers are supported by i350 */
7944 if (adapter
->vfs_allocated_count
) {
7945 igb_vmdq_set_loopback_pf(hw
, true);
7946 igb_vmdq_set_replication_pf(hw
, true);
7947 igb_vmdq_set_anti_spoofing_pf(hw
, true,
7948 adapter
->vfs_allocated_count
);
7950 igb_vmdq_set_loopback_pf(hw
, false);
7951 igb_vmdq_set_replication_pf(hw
, false);
7955 static void igb_init_dmac(struct igb_adapter
*adapter
, u32 pba
)
7957 struct e1000_hw
*hw
= &adapter
->hw
;
7961 if (hw
->mac
.type
> e1000_82580
) {
7962 if (adapter
->flags
& IGB_FLAG_DMAC
) {
7965 /* force threshold to 0. */
7966 wr32(E1000_DMCTXTH
, 0);
7968 /* DMA Coalescing high water mark needs to be greater
7969 * than the Rx threshold. Set hwm to PBA - max frame
7970 * size in 16B units, capping it at PBA - 6KB.
7972 hwm
= 64 * pba
- adapter
->max_frame_size
/ 16;
7973 if (hwm
< 64 * (pba
- 6))
7974 hwm
= 64 * (pba
- 6);
7975 reg
= rd32(E1000_FCRTC
);
7976 reg
&= ~E1000_FCRTC_RTH_COAL_MASK
;
7977 reg
|= ((hwm
<< E1000_FCRTC_RTH_COAL_SHIFT
)
7978 & E1000_FCRTC_RTH_COAL_MASK
);
7979 wr32(E1000_FCRTC
, reg
);
7981 /* Set the DMA Coalescing Rx threshold to PBA - 2 * max
7982 * frame size, capping it at PBA - 10KB.
7984 dmac_thr
= pba
- adapter
->max_frame_size
/ 512;
7985 if (dmac_thr
< pba
- 10)
7986 dmac_thr
= pba
- 10;
7987 reg
= rd32(E1000_DMACR
);
7988 reg
&= ~E1000_DMACR_DMACTHR_MASK
;
7989 reg
|= ((dmac_thr
<< E1000_DMACR_DMACTHR_SHIFT
)
7990 & E1000_DMACR_DMACTHR_MASK
);
7992 /* transition to L0x or L1 if available..*/
7993 reg
|= (E1000_DMACR_DMAC_EN
| E1000_DMACR_DMAC_LX_MASK
);
7995 /* watchdog timer= +-1000 usec in 32usec intervals */
7998 /* Disable BMC-to-OS Watchdog Enable */
7999 if (hw
->mac
.type
!= e1000_i354
)
8000 reg
&= ~E1000_DMACR_DC_BMC2OSW_EN
;
8002 wr32(E1000_DMACR
, reg
);
8004 /* no lower threshold to disable
8005 * coalescing(smart fifb)-UTRESH=0
8007 wr32(E1000_DMCRTRH
, 0);
8009 reg
= (IGB_DMCTLX_DCFLUSH_DIS
| 0x4);
8011 wr32(E1000_DMCTLX
, reg
);
8013 /* free space in tx packet buffer to wake from
8016 wr32(E1000_DMCTXTH
, (IGB_MIN_TXPBSIZE
-
8017 (IGB_TX_BUF_4096
+ adapter
->max_frame_size
)) >> 6);
8019 /* make low power state decision controlled
8022 reg
= rd32(E1000_PCIEMISC
);
8023 reg
&= ~E1000_PCIEMISC_LX_DECISION
;
8024 wr32(E1000_PCIEMISC
, reg
);
8025 } /* endif adapter->dmac is not disabled */
8026 } else if (hw
->mac
.type
== e1000_82580
) {
8027 u32 reg
= rd32(E1000_PCIEMISC
);
8029 wr32(E1000_PCIEMISC
, reg
& ~E1000_PCIEMISC_LX_DECISION
);
8030 wr32(E1000_DMACR
, 0);
8035 * igb_read_i2c_byte - Reads 8 bit word over I2C
8036 * @hw: pointer to hardware structure
8037 * @byte_offset: byte offset to read
8038 * @dev_addr: device address
8041 * Performs byte read operation over I2C interface at
8042 * a specified device address.
8044 s32
igb_read_i2c_byte(struct e1000_hw
*hw
, u8 byte_offset
,
8045 u8 dev_addr
, u8
*data
)
8047 struct igb_adapter
*adapter
= container_of(hw
, struct igb_adapter
, hw
);
8048 struct i2c_client
*this_client
= adapter
->i2c_client
;
8053 return E1000_ERR_I2C
;
8055 swfw_mask
= E1000_SWFW_PHY0_SM
;
8057 if (hw
->mac
.ops
.acquire_swfw_sync(hw
, swfw_mask
))
8058 return E1000_ERR_SWFW_SYNC
;
8060 status
= i2c_smbus_read_byte_data(this_client
, byte_offset
);
8061 hw
->mac
.ops
.release_swfw_sync(hw
, swfw_mask
);
8064 return E1000_ERR_I2C
;
8072 * igb_write_i2c_byte - Writes 8 bit word over I2C
8073 * @hw: pointer to hardware structure
8074 * @byte_offset: byte offset to write
8075 * @dev_addr: device address
8076 * @data: value to write
8078 * Performs byte write operation over I2C interface at
8079 * a specified device address.
8081 s32
igb_write_i2c_byte(struct e1000_hw
*hw
, u8 byte_offset
,
8082 u8 dev_addr
, u8 data
)
8084 struct igb_adapter
*adapter
= container_of(hw
, struct igb_adapter
, hw
);
8085 struct i2c_client
*this_client
= adapter
->i2c_client
;
8087 u16 swfw_mask
= E1000_SWFW_PHY0_SM
;
8090 return E1000_ERR_I2C
;
8092 if (hw
->mac
.ops
.acquire_swfw_sync(hw
, swfw_mask
))
8093 return E1000_ERR_SWFW_SYNC
;
8094 status
= i2c_smbus_write_byte_data(this_client
, byte_offset
, data
);
8095 hw
->mac
.ops
.release_swfw_sync(hw
, swfw_mask
);
8098 return E1000_ERR_I2C
;
8104 int igb_reinit_queues(struct igb_adapter
*adapter
)
8106 struct net_device
*netdev
= adapter
->netdev
;
8107 struct pci_dev
*pdev
= adapter
->pdev
;
8110 if (netif_running(netdev
))
8113 igb_reset_interrupt_capability(adapter
);
8115 if (igb_init_interrupt_scheme(adapter
, true)) {
8116 dev_err(&pdev
->dev
, "Unable to allocate memory for queues\n");
8120 if (netif_running(netdev
))
8121 err
= igb_open(netdev
);