1 /*******************************************************************************
3 Intel(R) Gigabit Ethernet Linux driver
4 Copyright(c) 2007-2014 Intel Corporation.
6 This program is free software; you can redistribute it and/or modify it
7 under the terms and conditions of the GNU General Public License,
8 version 2, as published by the Free Software Foundation.
10 This program is distributed in the hope it will be useful, but WITHOUT
11 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
15 You should have received a copy of the GNU General Public License along with
16 this program; if not, see <http://www.gnu.org/licenses/>.
18 The full GNU General Public License is included in this distribution in
19 the file called "COPYING".
22 e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
23 Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
25 *******************************************************************************/
27 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
29 #include <linux/module.h>
30 #include <linux/types.h>
31 #include <linux/init.h>
32 #include <linux/bitops.h>
33 #include <linux/vmalloc.h>
34 #include <linux/pagemap.h>
35 #include <linux/netdevice.h>
36 #include <linux/ipv6.h>
37 #include <linux/slab.h>
38 #include <net/checksum.h>
39 #include <net/ip6_checksum.h>
40 #include <linux/net_tstamp.h>
41 #include <linux/mii.h>
42 #include <linux/ethtool.h>
44 #include <linux/if_vlan.h>
45 #include <linux/pci.h>
46 #include <linux/pci-aspm.h>
47 #include <linux/delay.h>
48 #include <linux/interrupt.h>
50 #include <linux/tcp.h>
51 #include <linux/sctp.h>
52 #include <linux/if_ether.h>
53 #include <linux/aer.h>
54 #include <linux/prefetch.h>
55 #include <linux/pm_runtime.h>
57 #include <linux/dca.h>
59 #include <linux/i2c.h>
65 #define DRV_VERSION __stringify(MAJ) "." __stringify(MIN) "." \
66 __stringify(BUILD) "-k"
67 char igb_driver_name
[] = "igb";
68 char igb_driver_version
[] = DRV_VERSION
;
69 static const char igb_driver_string
[] =
70 "Intel(R) Gigabit Ethernet Network Driver";
71 static const char igb_copyright
[] =
72 "Copyright (c) 2007-2014 Intel Corporation.";
74 static const struct e1000_info
*igb_info_tbl
[] = {
75 [board_82575
] = &e1000_82575_info
,
78 static DEFINE_PCI_DEVICE_TABLE(igb_pci_tbl
) = {
79 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I354_BACKPLANE_1GBPS
) },
80 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I354_SGMII
) },
81 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS
) },
82 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I211_COPPER
), board_82575
},
83 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_COPPER
), board_82575
},
84 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_FIBER
), board_82575
},
85 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_SERDES
), board_82575
},
86 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_SGMII
), board_82575
},
87 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_COPPER_FLASHLESS
), board_82575
},
88 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_SERDES_FLASHLESS
), board_82575
},
89 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I350_COPPER
), board_82575
},
90 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I350_FIBER
), board_82575
},
91 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I350_SERDES
), board_82575
},
92 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I350_SGMII
), board_82575
},
93 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_COPPER
), board_82575
},
94 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_FIBER
), board_82575
},
95 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_QUAD_FIBER
), board_82575
},
96 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_SERDES
), board_82575
},
97 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_SGMII
), board_82575
},
98 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_COPPER_DUAL
), board_82575
},
99 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_DH89XXCC_SGMII
), board_82575
},
100 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_DH89XXCC_SERDES
), board_82575
},
101 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_DH89XXCC_BACKPLANE
), board_82575
},
102 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_DH89XXCC_SFP
), board_82575
},
103 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576
), board_82575
},
104 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_NS
), board_82575
},
105 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_NS_SERDES
), board_82575
},
106 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_FIBER
), board_82575
},
107 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_SERDES
), board_82575
},
108 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_SERDES_QUAD
), board_82575
},
109 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_QUAD_COPPER_ET2
), board_82575
},
110 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_QUAD_COPPER
), board_82575
},
111 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82575EB_COPPER
), board_82575
},
112 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82575EB_FIBER_SERDES
), board_82575
},
113 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82575GB_QUAD_COPPER
), board_82575
},
114 /* required last entry */
118 MODULE_DEVICE_TABLE(pci
, igb_pci_tbl
);
120 void igb_reset(struct igb_adapter
*);
121 static int igb_setup_all_tx_resources(struct igb_adapter
*);
122 static int igb_setup_all_rx_resources(struct igb_adapter
*);
123 static void igb_free_all_tx_resources(struct igb_adapter
*);
124 static void igb_free_all_rx_resources(struct igb_adapter
*);
125 static void igb_setup_mrqc(struct igb_adapter
*);
126 static int igb_probe(struct pci_dev
*, const struct pci_device_id
*);
127 static void igb_remove(struct pci_dev
*pdev
);
128 static int igb_sw_init(struct igb_adapter
*);
129 static int igb_open(struct net_device
*);
130 static int igb_close(struct net_device
*);
131 static void igb_configure(struct igb_adapter
*);
132 static void igb_configure_tx(struct igb_adapter
*);
133 static void igb_configure_rx(struct igb_adapter
*);
134 static void igb_clean_all_tx_rings(struct igb_adapter
*);
135 static void igb_clean_all_rx_rings(struct igb_adapter
*);
136 static void igb_clean_tx_ring(struct igb_ring
*);
137 static void igb_clean_rx_ring(struct igb_ring
*);
138 static void igb_set_rx_mode(struct net_device
*);
139 static void igb_update_phy_info(unsigned long);
140 static void igb_watchdog(unsigned long);
141 static void igb_watchdog_task(struct work_struct
*);
142 static netdev_tx_t
igb_xmit_frame(struct sk_buff
*skb
, struct net_device
*);
143 static struct rtnl_link_stats64
*igb_get_stats64(struct net_device
*dev
,
144 struct rtnl_link_stats64
*stats
);
145 static int igb_change_mtu(struct net_device
*, int);
146 static int igb_set_mac(struct net_device
*, void *);
147 static void igb_set_uta(struct igb_adapter
*adapter
);
148 static irqreturn_t
igb_intr(int irq
, void *);
149 static irqreturn_t
igb_intr_msi(int irq
, void *);
150 static irqreturn_t
igb_msix_other(int irq
, void *);
151 static irqreturn_t
igb_msix_ring(int irq
, void *);
152 #ifdef CONFIG_IGB_DCA
153 static void igb_update_dca(struct igb_q_vector
*);
154 static void igb_setup_dca(struct igb_adapter
*);
155 #endif /* CONFIG_IGB_DCA */
156 static int igb_poll(struct napi_struct
*, int);
157 static bool igb_clean_tx_irq(struct igb_q_vector
*);
158 static bool igb_clean_rx_irq(struct igb_q_vector
*, int);
159 static int igb_ioctl(struct net_device
*, struct ifreq
*, int cmd
);
160 static void igb_tx_timeout(struct net_device
*);
161 static void igb_reset_task(struct work_struct
*);
162 static void igb_vlan_mode(struct net_device
*netdev
, netdev_features_t features
);
163 static int igb_vlan_rx_add_vid(struct net_device
*, __be16
, u16
);
164 static int igb_vlan_rx_kill_vid(struct net_device
*, __be16
, u16
);
165 static void igb_restore_vlan(struct igb_adapter
*);
166 static void igb_rar_set_qsel(struct igb_adapter
*, u8
*, u32
, u8
);
167 static void igb_ping_all_vfs(struct igb_adapter
*);
168 static void igb_msg_task(struct igb_adapter
*);
169 static void igb_vmm_control(struct igb_adapter
*);
170 static int igb_set_vf_mac(struct igb_adapter
*, int, unsigned char *);
171 static void igb_restore_vf_multicasts(struct igb_adapter
*adapter
);
172 static int igb_ndo_set_vf_mac(struct net_device
*netdev
, int vf
, u8
*mac
);
173 static int igb_ndo_set_vf_vlan(struct net_device
*netdev
,
174 int vf
, u16 vlan
, u8 qos
);
175 static int igb_ndo_set_vf_bw(struct net_device
*netdev
, int vf
, int tx_rate
);
176 static int igb_ndo_set_vf_spoofchk(struct net_device
*netdev
, int vf
,
178 static int igb_ndo_get_vf_config(struct net_device
*netdev
, int vf
,
179 struct ifla_vf_info
*ivi
);
180 static void igb_check_vf_rate_limit(struct igb_adapter
*);
182 #ifdef CONFIG_PCI_IOV
183 static int igb_vf_configure(struct igb_adapter
*adapter
, int vf
);
184 static int igb_pci_enable_sriov(struct pci_dev
*dev
, int num_vfs
);
188 #ifdef CONFIG_PM_SLEEP
189 static int igb_suspend(struct device
*);
191 static int igb_resume(struct device
*);
192 #ifdef CONFIG_PM_RUNTIME
193 static int igb_runtime_suspend(struct device
*dev
);
194 static int igb_runtime_resume(struct device
*dev
);
195 static int igb_runtime_idle(struct device
*dev
);
197 static const struct dev_pm_ops igb_pm_ops
= {
198 SET_SYSTEM_SLEEP_PM_OPS(igb_suspend
, igb_resume
)
199 SET_RUNTIME_PM_OPS(igb_runtime_suspend
, igb_runtime_resume
,
203 static void igb_shutdown(struct pci_dev
*);
204 static int igb_pci_sriov_configure(struct pci_dev
*dev
, int num_vfs
);
205 #ifdef CONFIG_IGB_DCA
206 static int igb_notify_dca(struct notifier_block
*, unsigned long, void *);
207 static struct notifier_block dca_notifier
= {
208 .notifier_call
= igb_notify_dca
,
213 #ifdef CONFIG_NET_POLL_CONTROLLER
214 /* for netdump / net console */
215 static void igb_netpoll(struct net_device
*);
217 #ifdef CONFIG_PCI_IOV
218 static unsigned int max_vfs
= 0;
219 module_param(max_vfs
, uint
, 0);
220 MODULE_PARM_DESC(max_vfs
, "Maximum number of virtual functions to allocate "
221 "per physical function");
222 #endif /* CONFIG_PCI_IOV */
224 static pci_ers_result_t
igb_io_error_detected(struct pci_dev
*,
225 pci_channel_state_t
);
226 static pci_ers_result_t
igb_io_slot_reset(struct pci_dev
*);
227 static void igb_io_resume(struct pci_dev
*);
229 static const struct pci_error_handlers igb_err_handler
= {
230 .error_detected
= igb_io_error_detected
,
231 .slot_reset
= igb_io_slot_reset
,
232 .resume
= igb_io_resume
,
235 static void igb_init_dmac(struct igb_adapter
*adapter
, u32 pba
);
237 static struct pci_driver igb_driver
= {
238 .name
= igb_driver_name
,
239 .id_table
= igb_pci_tbl
,
241 .remove
= igb_remove
,
243 .driver
.pm
= &igb_pm_ops
,
245 .shutdown
= igb_shutdown
,
246 .sriov_configure
= igb_pci_sriov_configure
,
247 .err_handler
= &igb_err_handler
250 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
251 MODULE_DESCRIPTION("Intel(R) Gigabit Ethernet Network Driver");
252 MODULE_LICENSE("GPL");
253 MODULE_VERSION(DRV_VERSION
);
255 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
256 static int debug
= -1;
257 module_param(debug
, int, 0);
258 MODULE_PARM_DESC(debug
, "Debug level (0=none,...,16=all)");
260 struct igb_reg_info
{
265 static const struct igb_reg_info igb_reg_info_tbl
[] = {
267 /* General Registers */
268 {E1000_CTRL
, "CTRL"},
269 {E1000_STATUS
, "STATUS"},
270 {E1000_CTRL_EXT
, "CTRL_EXT"},
272 /* Interrupt Registers */
276 {E1000_RCTL
, "RCTL"},
277 {E1000_RDLEN(0), "RDLEN"},
278 {E1000_RDH(0), "RDH"},
279 {E1000_RDT(0), "RDT"},
280 {E1000_RXDCTL(0), "RXDCTL"},
281 {E1000_RDBAL(0), "RDBAL"},
282 {E1000_RDBAH(0), "RDBAH"},
285 {E1000_TCTL
, "TCTL"},
286 {E1000_TDBAL(0), "TDBAL"},
287 {E1000_TDBAH(0), "TDBAH"},
288 {E1000_TDLEN(0), "TDLEN"},
289 {E1000_TDH(0), "TDH"},
290 {E1000_TDT(0), "TDT"},
291 {E1000_TXDCTL(0), "TXDCTL"},
292 {E1000_TDFH
, "TDFH"},
293 {E1000_TDFT
, "TDFT"},
294 {E1000_TDFHS
, "TDFHS"},
295 {E1000_TDFPC
, "TDFPC"},
297 /* List Terminator */
301 /* igb_regdump - register printout routine */
302 static void igb_regdump(struct e1000_hw
*hw
, struct igb_reg_info
*reginfo
)
308 switch (reginfo
->ofs
) {
310 for (n
= 0; n
< 4; n
++)
311 regs
[n
] = rd32(E1000_RDLEN(n
));
314 for (n
= 0; n
< 4; n
++)
315 regs
[n
] = rd32(E1000_RDH(n
));
318 for (n
= 0; n
< 4; n
++)
319 regs
[n
] = rd32(E1000_RDT(n
));
321 case E1000_RXDCTL(0):
322 for (n
= 0; n
< 4; n
++)
323 regs
[n
] = rd32(E1000_RXDCTL(n
));
326 for (n
= 0; n
< 4; n
++)
327 regs
[n
] = rd32(E1000_RDBAL(n
));
330 for (n
= 0; n
< 4; n
++)
331 regs
[n
] = rd32(E1000_RDBAH(n
));
334 for (n
= 0; n
< 4; n
++)
335 regs
[n
] = rd32(E1000_RDBAL(n
));
338 for (n
= 0; n
< 4; n
++)
339 regs
[n
] = rd32(E1000_TDBAH(n
));
342 for (n
= 0; n
< 4; n
++)
343 regs
[n
] = rd32(E1000_TDLEN(n
));
346 for (n
= 0; n
< 4; n
++)
347 regs
[n
] = rd32(E1000_TDH(n
));
350 for (n
= 0; n
< 4; n
++)
351 regs
[n
] = rd32(E1000_TDT(n
));
353 case E1000_TXDCTL(0):
354 for (n
= 0; n
< 4; n
++)
355 regs
[n
] = rd32(E1000_TXDCTL(n
));
358 pr_info("%-15s %08x\n", reginfo
->name
, rd32(reginfo
->ofs
));
362 snprintf(rname
, 16, "%s%s", reginfo
->name
, "[0-3]");
363 pr_info("%-15s %08x %08x %08x %08x\n", rname
, regs
[0], regs
[1],
367 /* igb_dump - Print registers, Tx-rings and Rx-rings */
368 static void igb_dump(struct igb_adapter
*adapter
)
370 struct net_device
*netdev
= adapter
->netdev
;
371 struct e1000_hw
*hw
= &adapter
->hw
;
372 struct igb_reg_info
*reginfo
;
373 struct igb_ring
*tx_ring
;
374 union e1000_adv_tx_desc
*tx_desc
;
375 struct my_u0
{ u64 a
; u64 b
; } *u0
;
376 struct igb_ring
*rx_ring
;
377 union e1000_adv_rx_desc
*rx_desc
;
381 if (!netif_msg_hw(adapter
))
384 /* Print netdevice Info */
386 dev_info(&adapter
->pdev
->dev
, "Net device Info\n");
387 pr_info("Device Name state trans_start "
389 pr_info("%-15s %016lX %016lX %016lX\n", netdev
->name
,
390 netdev
->state
, netdev
->trans_start
, netdev
->last_rx
);
393 /* Print Registers */
394 dev_info(&adapter
->pdev
->dev
, "Register Dump\n");
395 pr_info(" Register Name Value\n");
396 for (reginfo
= (struct igb_reg_info
*)igb_reg_info_tbl
;
397 reginfo
->name
; reginfo
++) {
398 igb_regdump(hw
, reginfo
);
401 /* Print TX Ring Summary */
402 if (!netdev
|| !netif_running(netdev
))
405 dev_info(&adapter
->pdev
->dev
, "TX Rings Summary\n");
406 pr_info("Queue [NTU] [NTC] [bi(ntc)->dma ] leng ntw timestamp\n");
407 for (n
= 0; n
< adapter
->num_tx_queues
; n
++) {
408 struct igb_tx_buffer
*buffer_info
;
409 tx_ring
= adapter
->tx_ring
[n
];
410 buffer_info
= &tx_ring
->tx_buffer_info
[tx_ring
->next_to_clean
];
411 pr_info(" %5d %5X %5X %016llX %04X %p %016llX\n",
412 n
, tx_ring
->next_to_use
, tx_ring
->next_to_clean
,
413 (u64
)dma_unmap_addr(buffer_info
, dma
),
414 dma_unmap_len(buffer_info
, len
),
415 buffer_info
->next_to_watch
,
416 (u64
)buffer_info
->time_stamp
);
420 if (!netif_msg_tx_done(adapter
))
421 goto rx_ring_summary
;
423 dev_info(&adapter
->pdev
->dev
, "TX Rings Dump\n");
425 /* Transmit Descriptor Formats
427 * Advanced Transmit Descriptor
428 * +--------------------------------------------------------------+
429 * 0 | Buffer Address [63:0] |
430 * +--------------------------------------------------------------+
431 * 8 | PAYLEN | PORTS |CC|IDX | STA | DCMD |DTYP|MAC|RSV| DTALEN |
432 * +--------------------------------------------------------------+
433 * 63 46 45 40 39 38 36 35 32 31 24 15 0
436 for (n
= 0; n
< adapter
->num_tx_queues
; n
++) {
437 tx_ring
= adapter
->tx_ring
[n
];
438 pr_info("------------------------------------\n");
439 pr_info("TX QUEUE INDEX = %d\n", tx_ring
->queue_index
);
440 pr_info("------------------------------------\n");
441 pr_info("T [desc] [address 63:0 ] [PlPOCIStDDM Ln] "
442 "[bi->dma ] leng ntw timestamp "
445 for (i
= 0; tx_ring
->desc
&& (i
< tx_ring
->count
); i
++) {
446 const char *next_desc
;
447 struct igb_tx_buffer
*buffer_info
;
448 tx_desc
= IGB_TX_DESC(tx_ring
, i
);
449 buffer_info
= &tx_ring
->tx_buffer_info
[i
];
450 u0
= (struct my_u0
*)tx_desc
;
451 if (i
== tx_ring
->next_to_use
&&
452 i
== tx_ring
->next_to_clean
)
453 next_desc
= " NTC/U";
454 else if (i
== tx_ring
->next_to_use
)
456 else if (i
== tx_ring
->next_to_clean
)
461 pr_info("T [0x%03X] %016llX %016llX %016llX"
462 " %04X %p %016llX %p%s\n", i
,
465 (u64
)dma_unmap_addr(buffer_info
, dma
),
466 dma_unmap_len(buffer_info
, len
),
467 buffer_info
->next_to_watch
,
468 (u64
)buffer_info
->time_stamp
,
469 buffer_info
->skb
, next_desc
);
471 if (netif_msg_pktdata(adapter
) && buffer_info
->skb
)
472 print_hex_dump(KERN_INFO
, "",
474 16, 1, buffer_info
->skb
->data
,
475 dma_unmap_len(buffer_info
, len
),
480 /* Print RX Rings Summary */
482 dev_info(&adapter
->pdev
->dev
, "RX Rings Summary\n");
483 pr_info("Queue [NTU] [NTC]\n");
484 for (n
= 0; n
< adapter
->num_rx_queues
; n
++) {
485 rx_ring
= adapter
->rx_ring
[n
];
486 pr_info(" %5d %5X %5X\n",
487 n
, rx_ring
->next_to_use
, rx_ring
->next_to_clean
);
491 if (!netif_msg_rx_status(adapter
))
494 dev_info(&adapter
->pdev
->dev
, "RX Rings Dump\n");
496 /* Advanced Receive Descriptor (Read) Format
498 * +-----------------------------------------------------+
499 * 0 | Packet Buffer Address [63:1] |A0/NSE|
500 * +----------------------------------------------+------+
501 * 8 | Header Buffer Address [63:1] | DD |
502 * +-----------------------------------------------------+
505 * Advanced Receive Descriptor (Write-Back) Format
507 * 63 48 47 32 31 30 21 20 17 16 4 3 0
508 * +------------------------------------------------------+
509 * 0 | Packet IP |SPH| HDR_LEN | RSV|Packet| RSS |
510 * | Checksum Ident | | | | Type | Type |
511 * +------------------------------------------------------+
512 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
513 * +------------------------------------------------------+
514 * 63 48 47 32 31 20 19 0
517 for (n
= 0; n
< adapter
->num_rx_queues
; n
++) {
518 rx_ring
= adapter
->rx_ring
[n
];
519 pr_info("------------------------------------\n");
520 pr_info("RX QUEUE INDEX = %d\n", rx_ring
->queue_index
);
521 pr_info("------------------------------------\n");
522 pr_info("R [desc] [ PktBuf A0] [ HeadBuf DD] "
523 "[bi->dma ] [bi->skb] <-- Adv Rx Read format\n");
524 pr_info("RWB[desc] [PcsmIpSHl PtRs] [vl er S cks ln] -----"
525 "----------- [bi->skb] <-- Adv Rx Write-Back format\n");
527 for (i
= 0; i
< rx_ring
->count
; i
++) {
528 const char *next_desc
;
529 struct igb_rx_buffer
*buffer_info
;
530 buffer_info
= &rx_ring
->rx_buffer_info
[i
];
531 rx_desc
= IGB_RX_DESC(rx_ring
, i
);
532 u0
= (struct my_u0
*)rx_desc
;
533 staterr
= le32_to_cpu(rx_desc
->wb
.upper
.status_error
);
535 if (i
== rx_ring
->next_to_use
)
537 else if (i
== rx_ring
->next_to_clean
)
542 if (staterr
& E1000_RXD_STAT_DD
) {
543 /* Descriptor Done */
544 pr_info("%s[0x%03X] %016llX %016llX ---------------- %s\n",
550 pr_info("%s[0x%03X] %016llX %016llX %016llX %s\n",
554 (u64
)buffer_info
->dma
,
557 if (netif_msg_pktdata(adapter
) &&
558 buffer_info
->dma
&& buffer_info
->page
) {
559 print_hex_dump(KERN_INFO
, "",
562 page_address(buffer_info
->page
) +
563 buffer_info
->page_offset
,
575 * igb_get_i2c_data - Reads the I2C SDA data bit
576 * @hw: pointer to hardware structure
577 * @i2cctl: Current value of I2CCTL register
579 * Returns the I2C data bit value
581 static int igb_get_i2c_data(void *data
)
583 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
584 struct e1000_hw
*hw
= &adapter
->hw
;
585 s32 i2cctl
= rd32(E1000_I2CPARAMS
);
587 return ((i2cctl
& E1000_I2C_DATA_IN
) != 0);
591 * igb_set_i2c_data - Sets the I2C data bit
592 * @data: pointer to hardware structure
593 * @state: I2C data value (0 or 1) to set
595 * Sets the I2C data bit
597 static void igb_set_i2c_data(void *data
, int state
)
599 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
600 struct e1000_hw
*hw
= &adapter
->hw
;
601 s32 i2cctl
= rd32(E1000_I2CPARAMS
);
604 i2cctl
|= E1000_I2C_DATA_OUT
;
606 i2cctl
&= ~E1000_I2C_DATA_OUT
;
608 i2cctl
&= ~E1000_I2C_DATA_OE_N
;
609 i2cctl
|= E1000_I2C_CLK_OE_N
;
610 wr32(E1000_I2CPARAMS
, i2cctl
);
616 * igb_set_i2c_clk - Sets the I2C SCL clock
617 * @data: pointer to hardware structure
618 * @state: state to set clock
620 * Sets the I2C clock line to state
622 static void igb_set_i2c_clk(void *data
, int state
)
624 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
625 struct e1000_hw
*hw
= &adapter
->hw
;
626 s32 i2cctl
= rd32(E1000_I2CPARAMS
);
629 i2cctl
|= E1000_I2C_CLK_OUT
;
630 i2cctl
&= ~E1000_I2C_CLK_OE_N
;
632 i2cctl
&= ~E1000_I2C_CLK_OUT
;
633 i2cctl
&= ~E1000_I2C_CLK_OE_N
;
635 wr32(E1000_I2CPARAMS
, i2cctl
);
640 * igb_get_i2c_clk - Gets the I2C SCL clock state
641 * @data: pointer to hardware structure
643 * Gets the I2C clock state
645 static int igb_get_i2c_clk(void *data
)
647 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
648 struct e1000_hw
*hw
= &adapter
->hw
;
649 s32 i2cctl
= rd32(E1000_I2CPARAMS
);
651 return ((i2cctl
& E1000_I2C_CLK_IN
) != 0);
654 static const struct i2c_algo_bit_data igb_i2c_algo
= {
655 .setsda
= igb_set_i2c_data
,
656 .setscl
= igb_set_i2c_clk
,
657 .getsda
= igb_get_i2c_data
,
658 .getscl
= igb_get_i2c_clk
,
664 * igb_get_hw_dev - return device
665 * @hw: pointer to hardware structure
667 * used by hardware layer to print debugging information
669 struct net_device
*igb_get_hw_dev(struct e1000_hw
*hw
)
671 struct igb_adapter
*adapter
= hw
->back
;
672 return adapter
->netdev
;
676 * igb_init_module - Driver Registration Routine
678 * igb_init_module is the first routine called when the driver is
679 * loaded. All it does is register with the PCI subsystem.
681 static int __init
igb_init_module(void)
684 pr_info("%s - version %s\n",
685 igb_driver_string
, igb_driver_version
);
687 pr_info("%s\n", igb_copyright
);
689 #ifdef CONFIG_IGB_DCA
690 dca_register_notify(&dca_notifier
);
692 ret
= pci_register_driver(&igb_driver
);
696 module_init(igb_init_module
);
699 * igb_exit_module - Driver Exit Cleanup Routine
701 * igb_exit_module is called just before the driver is removed
704 static void __exit
igb_exit_module(void)
706 #ifdef CONFIG_IGB_DCA
707 dca_unregister_notify(&dca_notifier
);
709 pci_unregister_driver(&igb_driver
);
712 module_exit(igb_exit_module
);
714 #define Q_IDX_82576(i) (((i & 0x1) << 3) + (i >> 1))
716 * igb_cache_ring_register - Descriptor ring to register mapping
717 * @adapter: board private structure to initialize
719 * Once we know the feature-set enabled for the device, we'll cache
720 * the register offset the descriptor ring is assigned to.
722 static void igb_cache_ring_register(struct igb_adapter
*adapter
)
725 u32 rbase_offset
= adapter
->vfs_allocated_count
;
727 switch (adapter
->hw
.mac
.type
) {
729 /* The queues are allocated for virtualization such that VF 0
730 * is allocated queues 0 and 8, VF 1 queues 1 and 9, etc.
731 * In order to avoid collision we start at the first free queue
732 * and continue consuming queues in the same sequence
734 if (adapter
->vfs_allocated_count
) {
735 for (; i
< adapter
->rss_queues
; i
++)
736 adapter
->rx_ring
[i
]->reg_idx
= rbase_offset
+
746 for (; i
< adapter
->num_rx_queues
; i
++)
747 adapter
->rx_ring
[i
]->reg_idx
= rbase_offset
+ i
;
748 for (; j
< adapter
->num_tx_queues
; j
++)
749 adapter
->tx_ring
[j
]->reg_idx
= rbase_offset
+ j
;
755 * igb_write_ivar - configure ivar for given MSI-X vector
756 * @hw: pointer to the HW structure
757 * @msix_vector: vector number we are allocating to a given ring
758 * @index: row index of IVAR register to write within IVAR table
759 * @offset: column offset of in IVAR, should be multiple of 8
761 * This function is intended to handle the writing of the IVAR register
762 * for adapters 82576 and newer. The IVAR table consists of 2 columns,
763 * each containing an cause allocation for an Rx and Tx ring, and a
764 * variable number of rows depending on the number of queues supported.
766 static void igb_write_ivar(struct e1000_hw
*hw
, int msix_vector
,
767 int index
, int offset
)
769 u32 ivar
= array_rd32(E1000_IVAR0
, index
);
771 /* clear any bits that are currently set */
772 ivar
&= ~((u32
)0xFF << offset
);
774 /* write vector and valid bit */
775 ivar
|= (msix_vector
| E1000_IVAR_VALID
) << offset
;
777 array_wr32(E1000_IVAR0
, index
, ivar
);
780 #define IGB_N0_QUEUE -1
781 static void igb_assign_vector(struct igb_q_vector
*q_vector
, int msix_vector
)
783 struct igb_adapter
*adapter
= q_vector
->adapter
;
784 struct e1000_hw
*hw
= &adapter
->hw
;
785 int rx_queue
= IGB_N0_QUEUE
;
786 int tx_queue
= IGB_N0_QUEUE
;
789 if (q_vector
->rx
.ring
)
790 rx_queue
= q_vector
->rx
.ring
->reg_idx
;
791 if (q_vector
->tx
.ring
)
792 tx_queue
= q_vector
->tx
.ring
->reg_idx
;
794 switch (hw
->mac
.type
) {
796 /* The 82575 assigns vectors using a bitmask, which matches the
797 * bitmask for the EICR/EIMS/EIMC registers. To assign one
798 * or more queues to a vector, we write the appropriate bits
799 * into the MSIXBM register for that vector.
801 if (rx_queue
> IGB_N0_QUEUE
)
802 msixbm
= E1000_EICR_RX_QUEUE0
<< rx_queue
;
803 if (tx_queue
> IGB_N0_QUEUE
)
804 msixbm
|= E1000_EICR_TX_QUEUE0
<< tx_queue
;
805 if (!(adapter
->flags
& IGB_FLAG_HAS_MSIX
) && msix_vector
== 0)
806 msixbm
|= E1000_EIMS_OTHER
;
807 array_wr32(E1000_MSIXBM(0), msix_vector
, msixbm
);
808 q_vector
->eims_value
= msixbm
;
811 /* 82576 uses a table that essentially consists of 2 columns
812 * with 8 rows. The ordering is column-major so we use the
813 * lower 3 bits as the row index, and the 4th bit as the
816 if (rx_queue
> IGB_N0_QUEUE
)
817 igb_write_ivar(hw
, msix_vector
,
819 (rx_queue
& 0x8) << 1);
820 if (tx_queue
> IGB_N0_QUEUE
)
821 igb_write_ivar(hw
, msix_vector
,
823 ((tx_queue
& 0x8) << 1) + 8);
824 q_vector
->eims_value
= 1 << msix_vector
;
831 /* On 82580 and newer adapters the scheme is similar to 82576
832 * however instead of ordering column-major we have things
833 * ordered row-major. So we traverse the table by using
834 * bit 0 as the column offset, and the remaining bits as the
837 if (rx_queue
> IGB_N0_QUEUE
)
838 igb_write_ivar(hw
, msix_vector
,
840 (rx_queue
& 0x1) << 4);
841 if (tx_queue
> IGB_N0_QUEUE
)
842 igb_write_ivar(hw
, msix_vector
,
844 ((tx_queue
& 0x1) << 4) + 8);
845 q_vector
->eims_value
= 1 << msix_vector
;
852 /* add q_vector eims value to global eims_enable_mask */
853 adapter
->eims_enable_mask
|= q_vector
->eims_value
;
855 /* configure q_vector to set itr on first interrupt */
856 q_vector
->set_itr
= 1;
860 * igb_configure_msix - Configure MSI-X hardware
861 * @adapter: board private structure to initialize
863 * igb_configure_msix sets up the hardware to properly
864 * generate MSI-X interrupts.
866 static void igb_configure_msix(struct igb_adapter
*adapter
)
870 struct e1000_hw
*hw
= &adapter
->hw
;
872 adapter
->eims_enable_mask
= 0;
874 /* set vector for other causes, i.e. link changes */
875 switch (hw
->mac
.type
) {
877 tmp
= rd32(E1000_CTRL_EXT
);
878 /* enable MSI-X PBA support*/
879 tmp
|= E1000_CTRL_EXT_PBA_CLR
;
881 /* Auto-Mask interrupts upon ICR read. */
882 tmp
|= E1000_CTRL_EXT_EIAME
;
883 tmp
|= E1000_CTRL_EXT_IRCA
;
885 wr32(E1000_CTRL_EXT
, tmp
);
887 /* enable msix_other interrupt */
888 array_wr32(E1000_MSIXBM(0), vector
++, E1000_EIMS_OTHER
);
889 adapter
->eims_other
= E1000_EIMS_OTHER
;
899 /* Turn on MSI-X capability first, or our settings
900 * won't stick. And it will take days to debug.
902 wr32(E1000_GPIE
, E1000_GPIE_MSIX_MODE
|
903 E1000_GPIE_PBA
| E1000_GPIE_EIAME
|
906 /* enable msix_other interrupt */
907 adapter
->eims_other
= 1 << vector
;
908 tmp
= (vector
++ | E1000_IVAR_VALID
) << 8;
910 wr32(E1000_IVAR_MISC
, tmp
);
913 /* do nothing, since nothing else supports MSI-X */
915 } /* switch (hw->mac.type) */
917 adapter
->eims_enable_mask
|= adapter
->eims_other
;
919 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
920 igb_assign_vector(adapter
->q_vector
[i
], vector
++);
926 * igb_request_msix - Initialize MSI-X interrupts
927 * @adapter: board private structure to initialize
929 * igb_request_msix allocates MSI-X vectors and requests interrupts from the
932 static int igb_request_msix(struct igb_adapter
*adapter
)
934 struct net_device
*netdev
= adapter
->netdev
;
935 struct e1000_hw
*hw
= &adapter
->hw
;
936 int i
, err
= 0, vector
= 0, free_vector
= 0;
938 err
= request_irq(adapter
->msix_entries
[vector
].vector
,
939 igb_msix_other
, 0, netdev
->name
, adapter
);
943 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
944 struct igb_q_vector
*q_vector
= adapter
->q_vector
[i
];
948 q_vector
->itr_register
= hw
->hw_addr
+ E1000_EITR(vector
);
950 if (q_vector
->rx
.ring
&& q_vector
->tx
.ring
)
951 sprintf(q_vector
->name
, "%s-TxRx-%u", netdev
->name
,
952 q_vector
->rx
.ring
->queue_index
);
953 else if (q_vector
->tx
.ring
)
954 sprintf(q_vector
->name
, "%s-tx-%u", netdev
->name
,
955 q_vector
->tx
.ring
->queue_index
);
956 else if (q_vector
->rx
.ring
)
957 sprintf(q_vector
->name
, "%s-rx-%u", netdev
->name
,
958 q_vector
->rx
.ring
->queue_index
);
960 sprintf(q_vector
->name
, "%s-unused", netdev
->name
);
962 err
= request_irq(adapter
->msix_entries
[vector
].vector
,
963 igb_msix_ring
, 0, q_vector
->name
,
969 igb_configure_msix(adapter
);
973 /* free already assigned IRQs */
974 free_irq(adapter
->msix_entries
[free_vector
++].vector
, adapter
);
977 for (i
= 0; i
< vector
; i
++) {
978 free_irq(adapter
->msix_entries
[free_vector
++].vector
,
979 adapter
->q_vector
[i
]);
986 * igb_free_q_vector - Free memory allocated for specific interrupt vector
987 * @adapter: board private structure to initialize
988 * @v_idx: Index of vector to be freed
990 * This function frees the memory allocated to the q_vector.
992 static void igb_free_q_vector(struct igb_adapter
*adapter
, int v_idx
)
994 struct igb_q_vector
*q_vector
= adapter
->q_vector
[v_idx
];
996 adapter
->q_vector
[v_idx
] = NULL
;
998 /* igb_get_stats64() might access the rings on this vector,
999 * we must wait a grace period before freeing it.
1001 kfree_rcu(q_vector
, rcu
);
1005 * igb_reset_q_vector - Reset config for interrupt vector
1006 * @adapter: board private structure to initialize
1007 * @v_idx: Index of vector to be reset
1009 * If NAPI is enabled it will delete any references to the
1010 * NAPI struct. This is preparation for igb_free_q_vector.
1012 static void igb_reset_q_vector(struct igb_adapter
*adapter
, int v_idx
)
1014 struct igb_q_vector
*q_vector
= adapter
->q_vector
[v_idx
];
1016 if (q_vector
->tx
.ring
)
1017 adapter
->tx_ring
[q_vector
->tx
.ring
->queue_index
] = NULL
;
1019 if (q_vector
->rx
.ring
)
1020 adapter
->tx_ring
[q_vector
->rx
.ring
->queue_index
] = NULL
;
1022 netif_napi_del(&q_vector
->napi
);
1026 static void igb_reset_interrupt_capability(struct igb_adapter
*adapter
)
1028 int v_idx
= adapter
->num_q_vectors
;
1030 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
)
1031 pci_disable_msix(adapter
->pdev
);
1032 else if (adapter
->flags
& IGB_FLAG_HAS_MSI
)
1033 pci_disable_msi(adapter
->pdev
);
1036 igb_reset_q_vector(adapter
, v_idx
);
1040 * igb_free_q_vectors - Free memory allocated for interrupt vectors
1041 * @adapter: board private structure to initialize
1043 * This function frees the memory allocated to the q_vectors. In addition if
1044 * NAPI is enabled it will delete any references to the NAPI struct prior
1045 * to freeing the q_vector.
1047 static void igb_free_q_vectors(struct igb_adapter
*adapter
)
1049 int v_idx
= adapter
->num_q_vectors
;
1051 adapter
->num_tx_queues
= 0;
1052 adapter
->num_rx_queues
= 0;
1053 adapter
->num_q_vectors
= 0;
1056 igb_reset_q_vector(adapter
, v_idx
);
1057 igb_free_q_vector(adapter
, v_idx
);
1062 * igb_clear_interrupt_scheme - reset the device to a state of no interrupts
1063 * @adapter: board private structure to initialize
1065 * This function resets the device so that it has 0 Rx queues, Tx queues, and
1066 * MSI-X interrupts allocated.
1068 static void igb_clear_interrupt_scheme(struct igb_adapter
*adapter
)
1070 igb_free_q_vectors(adapter
);
1071 igb_reset_interrupt_capability(adapter
);
1075 * igb_set_interrupt_capability - set MSI or MSI-X if supported
1076 * @adapter: board private structure to initialize
1077 * @msix: boolean value of MSIX capability
1079 * Attempt to configure interrupts using the best available
1080 * capabilities of the hardware and kernel.
1082 static void igb_set_interrupt_capability(struct igb_adapter
*adapter
, bool msix
)
1089 adapter
->flags
|= IGB_FLAG_HAS_MSIX
;
1091 /* Number of supported queues. */
1092 adapter
->num_rx_queues
= adapter
->rss_queues
;
1093 if (adapter
->vfs_allocated_count
)
1094 adapter
->num_tx_queues
= 1;
1096 adapter
->num_tx_queues
= adapter
->rss_queues
;
1098 /* start with one vector for every Rx queue */
1099 numvecs
= adapter
->num_rx_queues
;
1101 /* if Tx handler is separate add 1 for every Tx queue */
1102 if (!(adapter
->flags
& IGB_FLAG_QUEUE_PAIRS
))
1103 numvecs
+= adapter
->num_tx_queues
;
1105 /* store the number of vectors reserved for queues */
1106 adapter
->num_q_vectors
= numvecs
;
1108 /* add 1 vector for link status interrupts */
1110 for (i
= 0; i
< numvecs
; i
++)
1111 adapter
->msix_entries
[i
].entry
= i
;
1113 err
= pci_enable_msix_range(adapter
->pdev
,
1114 adapter
->msix_entries
,
1120 igb_reset_interrupt_capability(adapter
);
1122 /* If we can't do MSI-X, try MSI */
1124 #ifdef CONFIG_PCI_IOV
1125 /* disable SR-IOV for non MSI-X configurations */
1126 if (adapter
->vf_data
) {
1127 struct e1000_hw
*hw
= &adapter
->hw
;
1128 /* disable iov and allow time for transactions to clear */
1129 pci_disable_sriov(adapter
->pdev
);
1132 kfree(adapter
->vf_data
);
1133 adapter
->vf_data
= NULL
;
1134 wr32(E1000_IOVCTL
, E1000_IOVCTL_REUSE_VFQ
);
1137 dev_info(&adapter
->pdev
->dev
, "IOV Disabled\n");
1140 adapter
->vfs_allocated_count
= 0;
1141 adapter
->rss_queues
= 1;
1142 adapter
->flags
|= IGB_FLAG_QUEUE_PAIRS
;
1143 adapter
->num_rx_queues
= 1;
1144 adapter
->num_tx_queues
= 1;
1145 adapter
->num_q_vectors
= 1;
1146 if (!pci_enable_msi(adapter
->pdev
))
1147 adapter
->flags
|= IGB_FLAG_HAS_MSI
;
1150 static void igb_add_ring(struct igb_ring
*ring
,
1151 struct igb_ring_container
*head
)
1158 * igb_alloc_q_vector - Allocate memory for a single interrupt vector
1159 * @adapter: board private structure to initialize
1160 * @v_count: q_vectors allocated on adapter, used for ring interleaving
1161 * @v_idx: index of vector in adapter struct
1162 * @txr_count: total number of Tx rings to allocate
1163 * @txr_idx: index of first Tx ring to allocate
1164 * @rxr_count: total number of Rx rings to allocate
1165 * @rxr_idx: index of first Rx ring to allocate
1167 * We allocate one q_vector. If allocation fails we return -ENOMEM.
1169 static int igb_alloc_q_vector(struct igb_adapter
*adapter
,
1170 int v_count
, int v_idx
,
1171 int txr_count
, int txr_idx
,
1172 int rxr_count
, int rxr_idx
)
1174 struct igb_q_vector
*q_vector
;
1175 struct igb_ring
*ring
;
1176 int ring_count
, size
;
1178 /* igb only supports 1 Tx and/or 1 Rx queue per vector */
1179 if (txr_count
> 1 || rxr_count
> 1)
1182 ring_count
= txr_count
+ rxr_count
;
1183 size
= sizeof(struct igb_q_vector
) +
1184 (sizeof(struct igb_ring
) * ring_count
);
1186 /* allocate q_vector and rings */
1187 q_vector
= adapter
->q_vector
[v_idx
];
1189 q_vector
= kzalloc(size
, GFP_KERNEL
);
1193 /* initialize NAPI */
1194 netif_napi_add(adapter
->netdev
, &q_vector
->napi
,
1197 /* tie q_vector and adapter together */
1198 adapter
->q_vector
[v_idx
] = q_vector
;
1199 q_vector
->adapter
= adapter
;
1201 /* initialize work limits */
1202 q_vector
->tx
.work_limit
= adapter
->tx_work_limit
;
1204 /* initialize ITR configuration */
1205 q_vector
->itr_register
= adapter
->hw
.hw_addr
+ E1000_EITR(0);
1206 q_vector
->itr_val
= IGB_START_ITR
;
1208 /* initialize pointer to rings */
1209 ring
= q_vector
->ring
;
1213 /* rx or rx/tx vector */
1214 if (!adapter
->rx_itr_setting
|| adapter
->rx_itr_setting
> 3)
1215 q_vector
->itr_val
= adapter
->rx_itr_setting
;
1217 /* tx only vector */
1218 if (!adapter
->tx_itr_setting
|| adapter
->tx_itr_setting
> 3)
1219 q_vector
->itr_val
= adapter
->tx_itr_setting
;
1223 /* assign generic ring traits */
1224 ring
->dev
= &adapter
->pdev
->dev
;
1225 ring
->netdev
= adapter
->netdev
;
1227 /* configure backlink on ring */
1228 ring
->q_vector
= q_vector
;
1230 /* update q_vector Tx values */
1231 igb_add_ring(ring
, &q_vector
->tx
);
1233 /* For 82575, context index must be unique per ring. */
1234 if (adapter
->hw
.mac
.type
== e1000_82575
)
1235 set_bit(IGB_RING_FLAG_TX_CTX_IDX
, &ring
->flags
);
1237 /* apply Tx specific ring traits */
1238 ring
->count
= adapter
->tx_ring_count
;
1239 ring
->queue_index
= txr_idx
;
1241 u64_stats_init(&ring
->tx_syncp
);
1242 u64_stats_init(&ring
->tx_syncp2
);
1244 /* assign ring to adapter */
1245 adapter
->tx_ring
[txr_idx
] = ring
;
1247 /* push pointer to next ring */
1252 /* assign generic ring traits */
1253 ring
->dev
= &adapter
->pdev
->dev
;
1254 ring
->netdev
= adapter
->netdev
;
1256 /* configure backlink on ring */
1257 ring
->q_vector
= q_vector
;
1259 /* update q_vector Rx values */
1260 igb_add_ring(ring
, &q_vector
->rx
);
1262 /* set flag indicating ring supports SCTP checksum offload */
1263 if (adapter
->hw
.mac
.type
>= e1000_82576
)
1264 set_bit(IGB_RING_FLAG_RX_SCTP_CSUM
, &ring
->flags
);
1267 * On i350, i354, i210, and i211, loopback VLAN packets
1268 * have the tag byte-swapped.
1270 if (adapter
->hw
.mac
.type
>= e1000_i350
)
1271 set_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP
, &ring
->flags
);
1273 /* apply Rx specific ring traits */
1274 ring
->count
= adapter
->rx_ring_count
;
1275 ring
->queue_index
= rxr_idx
;
1277 u64_stats_init(&ring
->rx_syncp
);
1279 /* assign ring to adapter */
1280 adapter
->rx_ring
[rxr_idx
] = ring
;
1288 * igb_alloc_q_vectors - Allocate memory for interrupt vectors
1289 * @adapter: board private structure to initialize
1291 * We allocate one q_vector per queue interrupt. If allocation fails we
1294 static int igb_alloc_q_vectors(struct igb_adapter
*adapter
)
1296 int q_vectors
= adapter
->num_q_vectors
;
1297 int rxr_remaining
= adapter
->num_rx_queues
;
1298 int txr_remaining
= adapter
->num_tx_queues
;
1299 int rxr_idx
= 0, txr_idx
= 0, v_idx
= 0;
1302 if (q_vectors
>= (rxr_remaining
+ txr_remaining
)) {
1303 for (; rxr_remaining
; v_idx
++) {
1304 err
= igb_alloc_q_vector(adapter
, q_vectors
, v_idx
,
1310 /* update counts and index */
1316 for (; v_idx
< q_vectors
; v_idx
++) {
1317 int rqpv
= DIV_ROUND_UP(rxr_remaining
, q_vectors
- v_idx
);
1318 int tqpv
= DIV_ROUND_UP(txr_remaining
, q_vectors
- v_idx
);
1319 err
= igb_alloc_q_vector(adapter
, q_vectors
, v_idx
,
1320 tqpv
, txr_idx
, rqpv
, rxr_idx
);
1325 /* update counts and index */
1326 rxr_remaining
-= rqpv
;
1327 txr_remaining
-= tqpv
;
1335 adapter
->num_tx_queues
= 0;
1336 adapter
->num_rx_queues
= 0;
1337 adapter
->num_q_vectors
= 0;
1340 igb_free_q_vector(adapter
, v_idx
);
1346 * igb_init_interrupt_scheme - initialize interrupts, allocate queues/vectors
1347 * @adapter: board private structure to initialize
1348 * @msix: boolean value of MSIX capability
1350 * This function initializes the interrupts and allocates all of the queues.
1352 static int igb_init_interrupt_scheme(struct igb_adapter
*adapter
, bool msix
)
1354 struct pci_dev
*pdev
= adapter
->pdev
;
1357 igb_set_interrupt_capability(adapter
, msix
);
1359 err
= igb_alloc_q_vectors(adapter
);
1361 dev_err(&pdev
->dev
, "Unable to allocate memory for vectors\n");
1362 goto err_alloc_q_vectors
;
1365 igb_cache_ring_register(adapter
);
1369 err_alloc_q_vectors
:
1370 igb_reset_interrupt_capability(adapter
);
1375 * igb_request_irq - initialize interrupts
1376 * @adapter: board private structure to initialize
1378 * Attempts to configure interrupts using the best available
1379 * capabilities of the hardware and kernel.
1381 static int igb_request_irq(struct igb_adapter
*adapter
)
1383 struct net_device
*netdev
= adapter
->netdev
;
1384 struct pci_dev
*pdev
= adapter
->pdev
;
1387 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1388 err
= igb_request_msix(adapter
);
1391 /* fall back to MSI */
1392 igb_free_all_tx_resources(adapter
);
1393 igb_free_all_rx_resources(adapter
);
1395 igb_clear_interrupt_scheme(adapter
);
1396 err
= igb_init_interrupt_scheme(adapter
, false);
1400 igb_setup_all_tx_resources(adapter
);
1401 igb_setup_all_rx_resources(adapter
);
1402 igb_configure(adapter
);
1405 igb_assign_vector(adapter
->q_vector
[0], 0);
1407 if (adapter
->flags
& IGB_FLAG_HAS_MSI
) {
1408 err
= request_irq(pdev
->irq
, igb_intr_msi
, 0,
1409 netdev
->name
, adapter
);
1413 /* fall back to legacy interrupts */
1414 igb_reset_interrupt_capability(adapter
);
1415 adapter
->flags
&= ~IGB_FLAG_HAS_MSI
;
1418 err
= request_irq(pdev
->irq
, igb_intr
, IRQF_SHARED
,
1419 netdev
->name
, adapter
);
1422 dev_err(&pdev
->dev
, "Error %d getting interrupt\n",
1429 static void igb_free_irq(struct igb_adapter
*adapter
)
1431 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1434 free_irq(adapter
->msix_entries
[vector
++].vector
, adapter
);
1436 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
1437 free_irq(adapter
->msix_entries
[vector
++].vector
,
1438 adapter
->q_vector
[i
]);
1440 free_irq(adapter
->pdev
->irq
, adapter
);
1445 * igb_irq_disable - Mask off interrupt generation on the NIC
1446 * @adapter: board private structure
1448 static void igb_irq_disable(struct igb_adapter
*adapter
)
1450 struct e1000_hw
*hw
= &adapter
->hw
;
1452 /* we need to be careful when disabling interrupts. The VFs are also
1453 * mapped into these registers and so clearing the bits can cause
1454 * issues on the VF drivers so we only need to clear what we set
1456 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1457 u32 regval
= rd32(E1000_EIAM
);
1458 wr32(E1000_EIAM
, regval
& ~adapter
->eims_enable_mask
);
1459 wr32(E1000_EIMC
, adapter
->eims_enable_mask
);
1460 regval
= rd32(E1000_EIAC
);
1461 wr32(E1000_EIAC
, regval
& ~adapter
->eims_enable_mask
);
1465 wr32(E1000_IMC
, ~0);
1467 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1469 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
1470 synchronize_irq(adapter
->msix_entries
[i
].vector
);
1472 synchronize_irq(adapter
->pdev
->irq
);
1477 * igb_irq_enable - Enable default interrupt generation settings
1478 * @adapter: board private structure
1480 static void igb_irq_enable(struct igb_adapter
*adapter
)
1482 struct e1000_hw
*hw
= &adapter
->hw
;
1484 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1485 u32 ims
= E1000_IMS_LSC
| E1000_IMS_DOUTSYNC
| E1000_IMS_DRSTA
;
1486 u32 regval
= rd32(E1000_EIAC
);
1487 wr32(E1000_EIAC
, regval
| adapter
->eims_enable_mask
);
1488 regval
= rd32(E1000_EIAM
);
1489 wr32(E1000_EIAM
, regval
| adapter
->eims_enable_mask
);
1490 wr32(E1000_EIMS
, adapter
->eims_enable_mask
);
1491 if (adapter
->vfs_allocated_count
) {
1492 wr32(E1000_MBVFIMR
, 0xFF);
1493 ims
|= E1000_IMS_VMMB
;
1495 wr32(E1000_IMS
, ims
);
1497 wr32(E1000_IMS
, IMS_ENABLE_MASK
|
1499 wr32(E1000_IAM
, IMS_ENABLE_MASK
|
1504 static void igb_update_mng_vlan(struct igb_adapter
*adapter
)
1506 struct e1000_hw
*hw
= &adapter
->hw
;
1507 u16 vid
= adapter
->hw
.mng_cookie
.vlan_id
;
1508 u16 old_vid
= adapter
->mng_vlan_id
;
1510 if (hw
->mng_cookie
.status
& E1000_MNG_DHCP_COOKIE_STATUS_VLAN
) {
1511 /* add VID to filter table */
1512 igb_vfta_set(hw
, vid
, true);
1513 adapter
->mng_vlan_id
= vid
;
1515 adapter
->mng_vlan_id
= IGB_MNG_VLAN_NONE
;
1518 if ((old_vid
!= (u16
)IGB_MNG_VLAN_NONE
) &&
1520 !test_bit(old_vid
, adapter
->active_vlans
)) {
1521 /* remove VID from filter table */
1522 igb_vfta_set(hw
, old_vid
, false);
1527 * igb_release_hw_control - release control of the h/w to f/w
1528 * @adapter: address of board private structure
1530 * igb_release_hw_control resets CTRL_EXT:DRV_LOAD bit.
1531 * For ASF and Pass Through versions of f/w this means that the
1532 * driver is no longer loaded.
1534 static void igb_release_hw_control(struct igb_adapter
*adapter
)
1536 struct e1000_hw
*hw
= &adapter
->hw
;
1539 /* Let firmware take over control of h/w */
1540 ctrl_ext
= rd32(E1000_CTRL_EXT
);
1541 wr32(E1000_CTRL_EXT
,
1542 ctrl_ext
& ~E1000_CTRL_EXT_DRV_LOAD
);
1546 * igb_get_hw_control - get control of the h/w from f/w
1547 * @adapter: address of board private structure
1549 * igb_get_hw_control sets CTRL_EXT:DRV_LOAD bit.
1550 * For ASF and Pass Through versions of f/w this means that
1551 * the driver is loaded.
1553 static void igb_get_hw_control(struct igb_adapter
*adapter
)
1555 struct e1000_hw
*hw
= &adapter
->hw
;
1558 /* Let firmware know the driver has taken over */
1559 ctrl_ext
= rd32(E1000_CTRL_EXT
);
1560 wr32(E1000_CTRL_EXT
,
1561 ctrl_ext
| E1000_CTRL_EXT_DRV_LOAD
);
1565 * igb_configure - configure the hardware for RX and TX
1566 * @adapter: private board structure
1568 static void igb_configure(struct igb_adapter
*adapter
)
1570 struct net_device
*netdev
= adapter
->netdev
;
1573 igb_get_hw_control(adapter
);
1574 igb_set_rx_mode(netdev
);
1576 igb_restore_vlan(adapter
);
1578 igb_setup_tctl(adapter
);
1579 igb_setup_mrqc(adapter
);
1580 igb_setup_rctl(adapter
);
1582 igb_configure_tx(adapter
);
1583 igb_configure_rx(adapter
);
1585 igb_rx_fifo_flush_82575(&adapter
->hw
);
1587 /* call igb_desc_unused which always leaves
1588 * at least 1 descriptor unused to make sure
1589 * next_to_use != next_to_clean
1591 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
1592 struct igb_ring
*ring
= adapter
->rx_ring
[i
];
1593 igb_alloc_rx_buffers(ring
, igb_desc_unused(ring
));
1598 * igb_power_up_link - Power up the phy/serdes link
1599 * @adapter: address of board private structure
1601 void igb_power_up_link(struct igb_adapter
*adapter
)
1603 igb_reset_phy(&adapter
->hw
);
1605 if (adapter
->hw
.phy
.media_type
== e1000_media_type_copper
)
1606 igb_power_up_phy_copper(&adapter
->hw
);
1608 igb_power_up_serdes_link_82575(&adapter
->hw
);
1612 * igb_power_down_link - Power down the phy/serdes link
1613 * @adapter: address of board private structure
1615 static void igb_power_down_link(struct igb_adapter
*adapter
)
1617 if (adapter
->hw
.phy
.media_type
== e1000_media_type_copper
)
1618 igb_power_down_phy_copper_82575(&adapter
->hw
);
1620 igb_shutdown_serdes_link_82575(&adapter
->hw
);
1624 * Detect and switch function for Media Auto Sense
1625 * @adapter: address of the board private structure
1627 static void igb_check_swap_media(struct igb_adapter
*adapter
)
1629 struct e1000_hw
*hw
= &adapter
->hw
;
1630 u32 ctrl_ext
, connsw
;
1631 bool swap_now
= false;
1633 ctrl_ext
= rd32(E1000_CTRL_EXT
);
1634 connsw
= rd32(E1000_CONNSW
);
1636 /* need to live swap if current media is copper and we have fiber/serdes
1640 if ((hw
->phy
.media_type
== e1000_media_type_copper
) &&
1641 (!(connsw
& E1000_CONNSW_AUTOSENSE_EN
))) {
1643 } else if (!(connsw
& E1000_CONNSW_SERDESD
)) {
1644 /* copper signal takes time to appear */
1645 if (adapter
->copper_tries
< 4) {
1646 adapter
->copper_tries
++;
1647 connsw
|= E1000_CONNSW_AUTOSENSE_CONF
;
1648 wr32(E1000_CONNSW
, connsw
);
1651 adapter
->copper_tries
= 0;
1652 if ((connsw
& E1000_CONNSW_PHYSD
) &&
1653 (!(connsw
& E1000_CONNSW_PHY_PDN
))) {
1655 connsw
&= ~E1000_CONNSW_AUTOSENSE_CONF
;
1656 wr32(E1000_CONNSW
, connsw
);
1664 switch (hw
->phy
.media_type
) {
1665 case e1000_media_type_copper
:
1666 netdev_info(adapter
->netdev
,
1667 "MAS: changing media to fiber/serdes\n");
1669 E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES
;
1670 adapter
->flags
|= IGB_FLAG_MEDIA_RESET
;
1671 adapter
->copper_tries
= 0;
1673 case e1000_media_type_internal_serdes
:
1674 case e1000_media_type_fiber
:
1675 netdev_info(adapter
->netdev
,
1676 "MAS: changing media to copper\n");
1678 ~E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES
;
1679 adapter
->flags
|= IGB_FLAG_MEDIA_RESET
;
1682 /* shouldn't get here during regular operation */
1683 netdev_err(adapter
->netdev
,
1684 "AMS: Invalid media type found, returning\n");
1687 wr32(E1000_CTRL_EXT
, ctrl_ext
);
1691 * igb_up - Open the interface and prepare it to handle traffic
1692 * @adapter: board private structure
1694 int igb_up(struct igb_adapter
*adapter
)
1696 struct e1000_hw
*hw
= &adapter
->hw
;
1699 /* hardware has been reset, we need to reload some things */
1700 igb_configure(adapter
);
1702 clear_bit(__IGB_DOWN
, &adapter
->state
);
1704 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
1705 napi_enable(&(adapter
->q_vector
[i
]->napi
));
1707 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
)
1708 igb_configure_msix(adapter
);
1710 igb_assign_vector(adapter
->q_vector
[0], 0);
1712 /* Clear any pending interrupts. */
1714 igb_irq_enable(adapter
);
1716 /* notify VFs that reset has been completed */
1717 if (adapter
->vfs_allocated_count
) {
1718 u32 reg_data
= rd32(E1000_CTRL_EXT
);
1719 reg_data
|= E1000_CTRL_EXT_PFRSTD
;
1720 wr32(E1000_CTRL_EXT
, reg_data
);
1723 netif_tx_start_all_queues(adapter
->netdev
);
1725 /* start the watchdog. */
1726 hw
->mac
.get_link_status
= 1;
1727 schedule_work(&adapter
->watchdog_task
);
1729 if ((adapter
->flags
& IGB_FLAG_EEE
) &&
1730 (!hw
->dev_spec
._82575
.eee_disable
))
1731 adapter
->eee_advert
= MDIO_EEE_100TX
| MDIO_EEE_1000T
;
1736 void igb_down(struct igb_adapter
*adapter
)
1738 struct net_device
*netdev
= adapter
->netdev
;
1739 struct e1000_hw
*hw
= &adapter
->hw
;
1743 /* signal that we're down so the interrupt handler does not
1744 * reschedule our watchdog timer
1746 set_bit(__IGB_DOWN
, &adapter
->state
);
1748 /* disable receives in the hardware */
1749 rctl
= rd32(E1000_RCTL
);
1750 wr32(E1000_RCTL
, rctl
& ~E1000_RCTL_EN
);
1751 /* flush and sleep below */
1753 netif_tx_stop_all_queues(netdev
);
1755 /* disable transmits in the hardware */
1756 tctl
= rd32(E1000_TCTL
);
1757 tctl
&= ~E1000_TCTL_EN
;
1758 wr32(E1000_TCTL
, tctl
);
1759 /* flush both disables and wait for them to finish */
1763 igb_irq_disable(adapter
);
1765 adapter
->flags
&= ~IGB_FLAG_NEED_LINK_UPDATE
;
1767 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
1768 napi_synchronize(&(adapter
->q_vector
[i
]->napi
));
1769 napi_disable(&(adapter
->q_vector
[i
]->napi
));
1773 del_timer_sync(&adapter
->watchdog_timer
);
1774 del_timer_sync(&adapter
->phy_info_timer
);
1776 netif_carrier_off(netdev
);
1778 /* record the stats before reset*/
1779 spin_lock(&adapter
->stats64_lock
);
1780 igb_update_stats(adapter
, &adapter
->stats64
);
1781 spin_unlock(&adapter
->stats64_lock
);
1783 adapter
->link_speed
= 0;
1784 adapter
->link_duplex
= 0;
1786 if (!pci_channel_offline(adapter
->pdev
))
1788 igb_clean_all_tx_rings(adapter
);
1789 igb_clean_all_rx_rings(adapter
);
1790 #ifdef CONFIG_IGB_DCA
1792 /* since we reset the hardware DCA settings were cleared */
1793 igb_setup_dca(adapter
);
1797 void igb_reinit_locked(struct igb_adapter
*adapter
)
1799 WARN_ON(in_interrupt());
1800 while (test_and_set_bit(__IGB_RESETTING
, &adapter
->state
))
1804 clear_bit(__IGB_RESETTING
, &adapter
->state
);
1807 /** igb_enable_mas - Media Autosense re-enable after swap
1809 * @adapter: adapter struct
1811 static s32
igb_enable_mas(struct igb_adapter
*adapter
)
1813 struct e1000_hw
*hw
= &adapter
->hw
;
1817 connsw
= rd32(E1000_CONNSW
);
1818 if (!(hw
->phy
.media_type
== e1000_media_type_copper
))
1821 /* configure for SerDes media detect */
1822 if (!(connsw
& E1000_CONNSW_SERDESD
)) {
1823 connsw
|= E1000_CONNSW_ENRGSRC
;
1824 connsw
|= E1000_CONNSW_AUTOSENSE_EN
;
1825 wr32(E1000_CONNSW
, connsw
);
1827 } else if (connsw
& E1000_CONNSW_SERDESD
) {
1828 /* already SerDes, no need to enable anything */
1831 netdev_info(adapter
->netdev
,
1832 "MAS: Unable to configure feature, disabling..\n");
1833 adapter
->flags
&= ~IGB_FLAG_MAS_ENABLE
;
1838 void igb_reset(struct igb_adapter
*adapter
)
1840 struct pci_dev
*pdev
= adapter
->pdev
;
1841 struct e1000_hw
*hw
= &adapter
->hw
;
1842 struct e1000_mac_info
*mac
= &hw
->mac
;
1843 struct e1000_fc_info
*fc
= &hw
->fc
;
1844 u32 pba
= 0, tx_space
, min_tx_space
, min_rx_space
, hwm
;
1846 /* Repartition Pba for greater than 9k mtu
1847 * To take effect CTRL.RST is required.
1849 switch (mac
->type
) {
1853 pba
= rd32(E1000_RXPBS
);
1854 pba
= igb_rxpbs_adjust_82580(pba
);
1857 pba
= rd32(E1000_RXPBS
);
1858 pba
&= E1000_RXPBS_SIZE_MASK_82576
;
1864 pba
= E1000_PBA_34K
;
1868 if ((adapter
->max_frame_size
> ETH_FRAME_LEN
+ ETH_FCS_LEN
) &&
1869 (mac
->type
< e1000_82576
)) {
1870 /* adjust PBA for jumbo frames */
1871 wr32(E1000_PBA
, pba
);
1873 /* To maintain wire speed transmits, the Tx FIFO should be
1874 * large enough to accommodate two full transmit packets,
1875 * rounded up to the next 1KB and expressed in KB. Likewise,
1876 * the Rx FIFO should be large enough to accommodate at least
1877 * one full receive packet and is similarly rounded up and
1880 pba
= rd32(E1000_PBA
);
1881 /* upper 16 bits has Tx packet buffer allocation size in KB */
1882 tx_space
= pba
>> 16;
1883 /* lower 16 bits has Rx packet buffer allocation size in KB */
1885 /* the Tx fifo also stores 16 bytes of information about the Tx
1886 * but don't include ethernet FCS because hardware appends it
1888 min_tx_space
= (adapter
->max_frame_size
+
1889 sizeof(union e1000_adv_tx_desc
) -
1891 min_tx_space
= ALIGN(min_tx_space
, 1024);
1892 min_tx_space
>>= 10;
1893 /* software strips receive CRC, so leave room for it */
1894 min_rx_space
= adapter
->max_frame_size
;
1895 min_rx_space
= ALIGN(min_rx_space
, 1024);
1896 min_rx_space
>>= 10;
1898 /* If current Tx allocation is less than the min Tx FIFO size,
1899 * and the min Tx FIFO size is less than the current Rx FIFO
1900 * allocation, take space away from current Rx allocation
1902 if (tx_space
< min_tx_space
&&
1903 ((min_tx_space
- tx_space
) < pba
)) {
1904 pba
= pba
- (min_tx_space
- tx_space
);
1906 /* if short on Rx space, Rx wins and must trump Tx
1909 if (pba
< min_rx_space
)
1912 wr32(E1000_PBA
, pba
);
1915 /* flow control settings */
1916 /* The high water mark must be low enough to fit one full frame
1917 * (or the size used for early receive) above it in the Rx FIFO.
1918 * Set it to the lower of:
1919 * - 90% of the Rx FIFO size, or
1920 * - the full Rx FIFO size minus one full frame
1922 hwm
= min(((pba
<< 10) * 9 / 10),
1923 ((pba
<< 10) - 2 * adapter
->max_frame_size
));
1925 fc
->high_water
= hwm
& 0xFFFFFFF0; /* 16-byte granularity */
1926 fc
->low_water
= fc
->high_water
- 16;
1927 fc
->pause_time
= 0xFFFF;
1929 fc
->current_mode
= fc
->requested_mode
;
1931 /* disable receive for all VFs and wait one second */
1932 if (adapter
->vfs_allocated_count
) {
1934 for (i
= 0 ; i
< adapter
->vfs_allocated_count
; i
++)
1935 adapter
->vf_data
[i
].flags
&= IGB_VF_FLAG_PF_SET_MAC
;
1937 /* ping all the active vfs to let them know we are going down */
1938 igb_ping_all_vfs(adapter
);
1940 /* disable transmits and receives */
1941 wr32(E1000_VFRE
, 0);
1942 wr32(E1000_VFTE
, 0);
1945 /* Allow time for pending master requests to run */
1946 hw
->mac
.ops
.reset_hw(hw
);
1949 if (adapter
->flags
& IGB_FLAG_MEDIA_RESET
) {
1950 /* need to resetup here after media swap */
1951 adapter
->ei
.get_invariants(hw
);
1952 adapter
->flags
&= ~IGB_FLAG_MEDIA_RESET
;
1954 if (adapter
->flags
& IGB_FLAG_MAS_ENABLE
) {
1955 if (igb_enable_mas(adapter
))
1957 "Error enabling Media Auto Sense\n");
1959 if (hw
->mac
.ops
.init_hw(hw
))
1960 dev_err(&pdev
->dev
, "Hardware Error\n");
1962 /* Flow control settings reset on hardware reset, so guarantee flow
1963 * control is off when forcing speed.
1965 if (!hw
->mac
.autoneg
)
1966 igb_force_mac_fc(hw
);
1968 igb_init_dmac(adapter
, pba
);
1969 #ifdef CONFIG_IGB_HWMON
1970 /* Re-initialize the thermal sensor on i350 devices. */
1971 if (!test_bit(__IGB_DOWN
, &adapter
->state
)) {
1972 if (mac
->type
== e1000_i350
&& hw
->bus
.func
== 0) {
1973 /* If present, re-initialize the external thermal sensor
1977 mac
->ops
.init_thermal_sensor_thresh(hw
);
1981 /* Re-establish EEE setting */
1982 if (hw
->phy
.media_type
== e1000_media_type_copper
) {
1983 switch (mac
->type
) {
1987 igb_set_eee_i350(hw
);
1990 igb_set_eee_i354(hw
);
1996 if (!netif_running(adapter
->netdev
))
1997 igb_power_down_link(adapter
);
1999 igb_update_mng_vlan(adapter
);
2001 /* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
2002 wr32(E1000_VET
, ETHERNET_IEEE_VLAN_TYPE
);
2004 /* Re-enable PTP, where applicable. */
2005 igb_ptp_reset(adapter
);
2007 igb_get_phy_info(hw
);
2010 static netdev_features_t
igb_fix_features(struct net_device
*netdev
,
2011 netdev_features_t features
)
2013 /* Since there is no support for separate Rx/Tx vlan accel
2014 * enable/disable make sure Tx flag is always in same state as Rx.
2016 if (features
& NETIF_F_HW_VLAN_CTAG_RX
)
2017 features
|= NETIF_F_HW_VLAN_CTAG_TX
;
2019 features
&= ~NETIF_F_HW_VLAN_CTAG_TX
;
2024 static int igb_set_features(struct net_device
*netdev
,
2025 netdev_features_t features
)
2027 netdev_features_t changed
= netdev
->features
^ features
;
2028 struct igb_adapter
*adapter
= netdev_priv(netdev
);
2030 if (changed
& NETIF_F_HW_VLAN_CTAG_RX
)
2031 igb_vlan_mode(netdev
, features
);
2033 if (!(changed
& NETIF_F_RXALL
))
2036 netdev
->features
= features
;
2038 if (netif_running(netdev
))
2039 igb_reinit_locked(adapter
);
2046 static const struct net_device_ops igb_netdev_ops
= {
2047 .ndo_open
= igb_open
,
2048 .ndo_stop
= igb_close
,
2049 .ndo_start_xmit
= igb_xmit_frame
,
2050 .ndo_get_stats64
= igb_get_stats64
,
2051 .ndo_set_rx_mode
= igb_set_rx_mode
,
2052 .ndo_set_mac_address
= igb_set_mac
,
2053 .ndo_change_mtu
= igb_change_mtu
,
2054 .ndo_do_ioctl
= igb_ioctl
,
2055 .ndo_tx_timeout
= igb_tx_timeout
,
2056 .ndo_validate_addr
= eth_validate_addr
,
2057 .ndo_vlan_rx_add_vid
= igb_vlan_rx_add_vid
,
2058 .ndo_vlan_rx_kill_vid
= igb_vlan_rx_kill_vid
,
2059 .ndo_set_vf_mac
= igb_ndo_set_vf_mac
,
2060 .ndo_set_vf_vlan
= igb_ndo_set_vf_vlan
,
2061 .ndo_set_vf_tx_rate
= igb_ndo_set_vf_bw
,
2062 .ndo_set_vf_spoofchk
= igb_ndo_set_vf_spoofchk
,
2063 .ndo_get_vf_config
= igb_ndo_get_vf_config
,
2064 #ifdef CONFIG_NET_POLL_CONTROLLER
2065 .ndo_poll_controller
= igb_netpoll
,
2067 .ndo_fix_features
= igb_fix_features
,
2068 .ndo_set_features
= igb_set_features
,
2072 * igb_set_fw_version - Configure version string for ethtool
2073 * @adapter: adapter struct
2075 void igb_set_fw_version(struct igb_adapter
*adapter
)
2077 struct e1000_hw
*hw
= &adapter
->hw
;
2078 struct e1000_fw_version fw
;
2080 igb_get_fw_version(hw
, &fw
);
2082 switch (hw
->mac
.type
) {
2085 if (!(igb_get_flash_presence_i210(hw
))) {
2086 snprintf(adapter
->fw_version
,
2087 sizeof(adapter
->fw_version
),
2089 fw
.invm_major
, fw
.invm_minor
,
2095 /* if option is rom valid, display its version too */
2097 snprintf(adapter
->fw_version
,
2098 sizeof(adapter
->fw_version
),
2099 "%d.%d, 0x%08x, %d.%d.%d",
2100 fw
.eep_major
, fw
.eep_minor
, fw
.etrack_id
,
2101 fw
.or_major
, fw
.or_build
, fw
.or_patch
);
2103 } else if (fw
.etrack_id
!= 0X0000) {
2104 snprintf(adapter
->fw_version
,
2105 sizeof(adapter
->fw_version
),
2107 fw
.eep_major
, fw
.eep_minor
, fw
.etrack_id
);
2109 snprintf(adapter
->fw_version
,
2110 sizeof(adapter
->fw_version
),
2112 fw
.eep_major
, fw
.eep_minor
, fw
.eep_build
);
2120 * igb_init_mas - init Media Autosense feature if enabled in the NVM
2122 * @adapter: adapter struct
2124 static void igb_init_mas(struct igb_adapter
*adapter
)
2126 struct e1000_hw
*hw
= &adapter
->hw
;
2129 hw
->nvm
.ops
.read(hw
, NVM_COMPAT
, 1, &eeprom_data
);
2130 switch (hw
->bus
.func
) {
2132 if (eeprom_data
& IGB_MAS_ENABLE_0
) {
2133 adapter
->flags
|= IGB_FLAG_MAS_ENABLE
;
2134 netdev_info(adapter
->netdev
,
2135 "MAS: Enabling Media Autosense for port %d\n",
2140 if (eeprom_data
& IGB_MAS_ENABLE_1
) {
2141 adapter
->flags
|= IGB_FLAG_MAS_ENABLE
;
2142 netdev_info(adapter
->netdev
,
2143 "MAS: Enabling Media Autosense for port %d\n",
2148 if (eeprom_data
& IGB_MAS_ENABLE_2
) {
2149 adapter
->flags
|= IGB_FLAG_MAS_ENABLE
;
2150 netdev_info(adapter
->netdev
,
2151 "MAS: Enabling Media Autosense for port %d\n",
2156 if (eeprom_data
& IGB_MAS_ENABLE_3
) {
2157 adapter
->flags
|= IGB_FLAG_MAS_ENABLE
;
2158 netdev_info(adapter
->netdev
,
2159 "MAS: Enabling Media Autosense for port %d\n",
2164 /* Shouldn't get here */
2165 netdev_err(adapter
->netdev
,
2166 "MAS: Invalid port configuration, returning\n");
2172 * igb_init_i2c - Init I2C interface
2173 * @adapter: pointer to adapter structure
2175 static s32
igb_init_i2c(struct igb_adapter
*adapter
)
2177 s32 status
= E1000_SUCCESS
;
2179 /* I2C interface supported on i350 devices */
2180 if (adapter
->hw
.mac
.type
!= e1000_i350
)
2181 return E1000_SUCCESS
;
2183 /* Initialize the i2c bus which is controlled by the registers.
2184 * This bus will use the i2c_algo_bit structue that implements
2185 * the protocol through toggling of the 4 bits in the register.
2187 adapter
->i2c_adap
.owner
= THIS_MODULE
;
2188 adapter
->i2c_algo
= igb_i2c_algo
;
2189 adapter
->i2c_algo
.data
= adapter
;
2190 adapter
->i2c_adap
.algo_data
= &adapter
->i2c_algo
;
2191 adapter
->i2c_adap
.dev
.parent
= &adapter
->pdev
->dev
;
2192 strlcpy(adapter
->i2c_adap
.name
, "igb BB",
2193 sizeof(adapter
->i2c_adap
.name
));
2194 status
= i2c_bit_add_bus(&adapter
->i2c_adap
);
2199 * igb_probe - Device Initialization Routine
2200 * @pdev: PCI device information struct
2201 * @ent: entry in igb_pci_tbl
2203 * Returns 0 on success, negative on failure
2205 * igb_probe initializes an adapter identified by a pci_dev structure.
2206 * The OS initialization, configuring of the adapter private structure,
2207 * and a hardware reset occur.
2209 static int igb_probe(struct pci_dev
*pdev
, const struct pci_device_id
*ent
)
2211 struct net_device
*netdev
;
2212 struct igb_adapter
*adapter
;
2213 struct e1000_hw
*hw
;
2214 u16 eeprom_data
= 0;
2216 static int global_quad_port_a
; /* global quad port a indication */
2217 const struct e1000_info
*ei
= igb_info_tbl
[ent
->driver_data
];
2218 int err
, pci_using_dac
;
2219 u8 part_str
[E1000_PBANUM_LENGTH
];
2221 /* Catch broken hardware that put the wrong VF device ID in
2222 * the PCIe SR-IOV capability.
2224 if (pdev
->is_virtfn
) {
2225 WARN(1, KERN_ERR
"%s (%hx:%hx) should not be a VF!\n",
2226 pci_name(pdev
), pdev
->vendor
, pdev
->device
);
2230 err
= pci_enable_device_mem(pdev
);
2235 err
= dma_set_mask_and_coherent(&pdev
->dev
, DMA_BIT_MASK(64));
2239 err
= dma_set_mask_and_coherent(&pdev
->dev
, DMA_BIT_MASK(32));
2242 "No usable DMA configuration, aborting\n");
2247 err
= pci_request_selected_regions(pdev
, pci_select_bars(pdev
,
2253 pci_enable_pcie_error_reporting(pdev
);
2255 pci_set_master(pdev
);
2256 pci_save_state(pdev
);
2259 netdev
= alloc_etherdev_mq(sizeof(struct igb_adapter
),
2262 goto err_alloc_etherdev
;
2264 SET_NETDEV_DEV(netdev
, &pdev
->dev
);
2266 pci_set_drvdata(pdev
, netdev
);
2267 adapter
= netdev_priv(netdev
);
2268 adapter
->netdev
= netdev
;
2269 adapter
->pdev
= pdev
;
2272 adapter
->msg_enable
= netif_msg_init(debug
, DEFAULT_MSG_ENABLE
);
2275 hw
->hw_addr
= pci_iomap(pdev
, 0, 0);
2279 netdev
->netdev_ops
= &igb_netdev_ops
;
2280 igb_set_ethtool_ops(netdev
);
2281 netdev
->watchdog_timeo
= 5 * HZ
;
2283 strncpy(netdev
->name
, pci_name(pdev
), sizeof(netdev
->name
) - 1);
2285 netdev
->mem_start
= pci_resource_start(pdev
, 0);
2286 netdev
->mem_end
= pci_resource_end(pdev
, 0);
2288 /* PCI config space info */
2289 hw
->vendor_id
= pdev
->vendor
;
2290 hw
->device_id
= pdev
->device
;
2291 hw
->revision_id
= pdev
->revision
;
2292 hw
->subsystem_vendor_id
= pdev
->subsystem_vendor
;
2293 hw
->subsystem_device_id
= pdev
->subsystem_device
;
2295 /* Copy the default MAC, PHY and NVM function pointers */
2296 memcpy(&hw
->mac
.ops
, ei
->mac_ops
, sizeof(hw
->mac
.ops
));
2297 memcpy(&hw
->phy
.ops
, ei
->phy_ops
, sizeof(hw
->phy
.ops
));
2298 memcpy(&hw
->nvm
.ops
, ei
->nvm_ops
, sizeof(hw
->nvm
.ops
));
2299 /* Initialize skew-specific constants */
2300 err
= ei
->get_invariants(hw
);
2304 /* setup the private structure */
2305 err
= igb_sw_init(adapter
);
2309 igb_get_bus_info_pcie(hw
);
2311 hw
->phy
.autoneg_wait_to_complete
= false;
2313 /* Copper options */
2314 if (hw
->phy
.media_type
== e1000_media_type_copper
) {
2315 hw
->phy
.mdix
= AUTO_ALL_MODES
;
2316 hw
->phy
.disable_polarity_correction
= false;
2317 hw
->phy
.ms_type
= e1000_ms_hw_default
;
2320 if (igb_check_reset_block(hw
))
2321 dev_info(&pdev
->dev
,
2322 "PHY reset is blocked due to SOL/IDER session.\n");
2324 /* features is initialized to 0 in allocation, it might have bits
2325 * set by igb_sw_init so we should use an or instead of an
2328 netdev
->features
|= NETIF_F_SG
|
2335 NETIF_F_HW_VLAN_CTAG_RX
|
2336 NETIF_F_HW_VLAN_CTAG_TX
;
2338 /* copy netdev features into list of user selectable features */
2339 netdev
->hw_features
|= netdev
->features
;
2340 netdev
->hw_features
|= NETIF_F_RXALL
;
2342 /* set this bit last since it cannot be part of hw_features */
2343 netdev
->features
|= NETIF_F_HW_VLAN_CTAG_FILTER
;
2345 netdev
->vlan_features
|= NETIF_F_TSO
|
2351 netdev
->priv_flags
|= IFF_SUPP_NOFCS
;
2353 if (pci_using_dac
) {
2354 netdev
->features
|= NETIF_F_HIGHDMA
;
2355 netdev
->vlan_features
|= NETIF_F_HIGHDMA
;
2358 if (hw
->mac
.type
>= e1000_82576
) {
2359 netdev
->hw_features
|= NETIF_F_SCTP_CSUM
;
2360 netdev
->features
|= NETIF_F_SCTP_CSUM
;
2363 netdev
->priv_flags
|= IFF_UNICAST_FLT
;
2365 adapter
->en_mng_pt
= igb_enable_mng_pass_thru(hw
);
2367 /* before reading the NVM, reset the controller to put the device in a
2368 * known good starting state
2370 hw
->mac
.ops
.reset_hw(hw
);
2372 /* make sure the NVM is good , i211/i210 parts can have special NVM
2373 * that doesn't contain a checksum
2375 switch (hw
->mac
.type
) {
2378 if (igb_get_flash_presence_i210(hw
)) {
2379 if (hw
->nvm
.ops
.validate(hw
) < 0) {
2381 "The NVM Checksum Is Not Valid\n");
2388 if (hw
->nvm
.ops
.validate(hw
) < 0) {
2389 dev_err(&pdev
->dev
, "The NVM Checksum Is Not Valid\n");
2396 /* copy the MAC address out of the NVM */
2397 if (hw
->mac
.ops
.read_mac_addr(hw
))
2398 dev_err(&pdev
->dev
, "NVM Read Error\n");
2400 memcpy(netdev
->dev_addr
, hw
->mac
.addr
, netdev
->addr_len
);
2402 if (!is_valid_ether_addr(netdev
->dev_addr
)) {
2403 dev_err(&pdev
->dev
, "Invalid MAC Address\n");
2408 /* get firmware version for ethtool -i */
2409 igb_set_fw_version(adapter
);
2411 setup_timer(&adapter
->watchdog_timer
, igb_watchdog
,
2412 (unsigned long) adapter
);
2413 setup_timer(&adapter
->phy_info_timer
, igb_update_phy_info
,
2414 (unsigned long) adapter
);
2416 INIT_WORK(&adapter
->reset_task
, igb_reset_task
);
2417 INIT_WORK(&adapter
->watchdog_task
, igb_watchdog_task
);
2419 /* Initialize link properties that are user-changeable */
2420 adapter
->fc_autoneg
= true;
2421 hw
->mac
.autoneg
= true;
2422 hw
->phy
.autoneg_advertised
= 0x2f;
2424 hw
->fc
.requested_mode
= e1000_fc_default
;
2425 hw
->fc
.current_mode
= e1000_fc_default
;
2427 igb_validate_mdi_setting(hw
);
2429 /* By default, support wake on port A */
2430 if (hw
->bus
.func
== 0)
2431 adapter
->flags
|= IGB_FLAG_WOL_SUPPORTED
;
2433 /* Check the NVM for wake support on non-port A ports */
2434 if (hw
->mac
.type
>= e1000_82580
)
2435 hw
->nvm
.ops
.read(hw
, NVM_INIT_CONTROL3_PORT_A
+
2436 NVM_82580_LAN_FUNC_OFFSET(hw
->bus
.func
), 1,
2438 else if (hw
->bus
.func
== 1)
2439 hw
->nvm
.ops
.read(hw
, NVM_INIT_CONTROL3_PORT_B
, 1, &eeprom_data
);
2441 if (eeprom_data
& IGB_EEPROM_APME
)
2442 adapter
->flags
|= IGB_FLAG_WOL_SUPPORTED
;
2444 /* now that we have the eeprom settings, apply the special cases where
2445 * the eeprom may be wrong or the board simply won't support wake on
2446 * lan on a particular port
2448 switch (pdev
->device
) {
2449 case E1000_DEV_ID_82575GB_QUAD_COPPER
:
2450 adapter
->flags
&= ~IGB_FLAG_WOL_SUPPORTED
;
2452 case E1000_DEV_ID_82575EB_FIBER_SERDES
:
2453 case E1000_DEV_ID_82576_FIBER
:
2454 case E1000_DEV_ID_82576_SERDES
:
2455 /* Wake events only supported on port A for dual fiber
2456 * regardless of eeprom setting
2458 if (rd32(E1000_STATUS
) & E1000_STATUS_FUNC_1
)
2459 adapter
->flags
&= ~IGB_FLAG_WOL_SUPPORTED
;
2461 case E1000_DEV_ID_82576_QUAD_COPPER
:
2462 case E1000_DEV_ID_82576_QUAD_COPPER_ET2
:
2463 /* if quad port adapter, disable WoL on all but port A */
2464 if (global_quad_port_a
!= 0)
2465 adapter
->flags
&= ~IGB_FLAG_WOL_SUPPORTED
;
2467 adapter
->flags
|= IGB_FLAG_QUAD_PORT_A
;
2468 /* Reset for multiple quad port adapters */
2469 if (++global_quad_port_a
== 4)
2470 global_quad_port_a
= 0;
2473 /* If the device can't wake, don't set software support */
2474 if (!device_can_wakeup(&adapter
->pdev
->dev
))
2475 adapter
->flags
&= ~IGB_FLAG_WOL_SUPPORTED
;
2478 /* initialize the wol settings based on the eeprom settings */
2479 if (adapter
->flags
& IGB_FLAG_WOL_SUPPORTED
)
2480 adapter
->wol
|= E1000_WUFC_MAG
;
2482 /* Some vendors want WoL disabled by default, but still supported */
2483 if ((hw
->mac
.type
== e1000_i350
) &&
2484 (pdev
->subsystem_vendor
== PCI_VENDOR_ID_HP
)) {
2485 adapter
->flags
|= IGB_FLAG_WOL_SUPPORTED
;
2489 device_set_wakeup_enable(&adapter
->pdev
->dev
,
2490 adapter
->flags
& IGB_FLAG_WOL_SUPPORTED
);
2492 /* reset the hardware with the new settings */
2495 /* Init the I2C interface */
2496 err
= igb_init_i2c(adapter
);
2498 dev_err(&pdev
->dev
, "failed to init i2c interface\n");
2502 /* let the f/w know that the h/w is now under the control of the
2504 igb_get_hw_control(adapter
);
2506 strcpy(netdev
->name
, "eth%d");
2507 err
= register_netdev(netdev
);
2511 /* carrier off reporting is important to ethtool even BEFORE open */
2512 netif_carrier_off(netdev
);
2514 #ifdef CONFIG_IGB_DCA
2515 if (dca_add_requester(&pdev
->dev
) == 0) {
2516 adapter
->flags
|= IGB_FLAG_DCA_ENABLED
;
2517 dev_info(&pdev
->dev
, "DCA enabled\n");
2518 igb_setup_dca(adapter
);
2522 #ifdef CONFIG_IGB_HWMON
2523 /* Initialize the thermal sensor on i350 devices. */
2524 if (hw
->mac
.type
== e1000_i350
&& hw
->bus
.func
== 0) {
2527 /* Read the NVM to determine if this i350 device supports an
2528 * external thermal sensor.
2530 hw
->nvm
.ops
.read(hw
, NVM_ETS_CFG
, 1, &ets_word
);
2531 if (ets_word
!= 0x0000 && ets_word
!= 0xFFFF)
2532 adapter
->ets
= true;
2534 adapter
->ets
= false;
2535 if (igb_sysfs_init(adapter
))
2537 "failed to allocate sysfs resources\n");
2539 adapter
->ets
= false;
2542 /* Check if Media Autosense is enabled */
2544 if (hw
->dev_spec
._82575
.mas_capable
)
2545 igb_init_mas(adapter
);
2547 /* do hw tstamp init after resetting */
2548 igb_ptp_init(adapter
);
2550 dev_info(&pdev
->dev
, "Intel(R) Gigabit Ethernet Network Connection\n");
2551 /* print bus type/speed/width info, not applicable to i354 */
2552 if (hw
->mac
.type
!= e1000_i354
) {
2553 dev_info(&pdev
->dev
, "%s: (PCIe:%s:%s) %pM\n",
2555 ((hw
->bus
.speed
== e1000_bus_speed_2500
) ? "2.5Gb/s" :
2556 (hw
->bus
.speed
== e1000_bus_speed_5000
) ? "5.0Gb/s" :
2558 ((hw
->bus
.width
== e1000_bus_width_pcie_x4
) ?
2560 (hw
->bus
.width
== e1000_bus_width_pcie_x2
) ?
2562 (hw
->bus
.width
== e1000_bus_width_pcie_x1
) ?
2563 "Width x1" : "unknown"), netdev
->dev_addr
);
2566 if ((hw
->mac
.type
>= e1000_i210
||
2567 igb_get_flash_presence_i210(hw
))) {
2568 ret_val
= igb_read_part_string(hw
, part_str
,
2569 E1000_PBANUM_LENGTH
);
2571 ret_val
= -E1000_ERR_INVM_VALUE_NOT_FOUND
;
2575 strcpy(part_str
, "Unknown");
2576 dev_info(&pdev
->dev
, "%s: PBA No: %s\n", netdev
->name
, part_str
);
2577 dev_info(&pdev
->dev
,
2578 "Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
2579 (adapter
->flags
& IGB_FLAG_HAS_MSIX
) ? "MSI-X" :
2580 (adapter
->flags
& IGB_FLAG_HAS_MSI
) ? "MSI" : "legacy",
2581 adapter
->num_rx_queues
, adapter
->num_tx_queues
);
2582 if (hw
->phy
.media_type
== e1000_media_type_copper
) {
2583 switch (hw
->mac
.type
) {
2587 /* Enable EEE for internal copper PHY devices */
2588 err
= igb_set_eee_i350(hw
);
2590 (!hw
->dev_spec
._82575
.eee_disable
)) {
2591 adapter
->eee_advert
=
2592 MDIO_EEE_100TX
| MDIO_EEE_1000T
;
2593 adapter
->flags
|= IGB_FLAG_EEE
;
2597 if ((rd32(E1000_CTRL_EXT
) &
2598 E1000_CTRL_EXT_LINK_MODE_SGMII
)) {
2599 err
= igb_set_eee_i354(hw
);
2601 (!hw
->dev_spec
._82575
.eee_disable
)) {
2602 adapter
->eee_advert
=
2603 MDIO_EEE_100TX
| MDIO_EEE_1000T
;
2604 adapter
->flags
|= IGB_FLAG_EEE
;
2612 pm_runtime_put_noidle(&pdev
->dev
);
2616 igb_release_hw_control(adapter
);
2617 memset(&adapter
->i2c_adap
, 0, sizeof(adapter
->i2c_adap
));
2619 if (!igb_check_reset_block(hw
))
2622 if (hw
->flash_address
)
2623 iounmap(hw
->flash_address
);
2625 igb_clear_interrupt_scheme(adapter
);
2626 iounmap(hw
->hw_addr
);
2628 free_netdev(netdev
);
2630 pci_release_selected_regions(pdev
,
2631 pci_select_bars(pdev
, IORESOURCE_MEM
));
2634 pci_disable_device(pdev
);
2638 #ifdef CONFIG_PCI_IOV
2639 static int igb_disable_sriov(struct pci_dev
*pdev
)
2641 struct net_device
*netdev
= pci_get_drvdata(pdev
);
2642 struct igb_adapter
*adapter
= netdev_priv(netdev
);
2643 struct e1000_hw
*hw
= &adapter
->hw
;
2645 /* reclaim resources allocated to VFs */
2646 if (adapter
->vf_data
) {
2647 /* disable iov and allow time for transactions to clear */
2648 if (pci_vfs_assigned(pdev
)) {
2649 dev_warn(&pdev
->dev
,
2650 "Cannot deallocate SR-IOV virtual functions while they are assigned - VFs will not be deallocated\n");
2653 pci_disable_sriov(pdev
);
2657 kfree(adapter
->vf_data
);
2658 adapter
->vf_data
= NULL
;
2659 adapter
->vfs_allocated_count
= 0;
2660 wr32(E1000_IOVCTL
, E1000_IOVCTL_REUSE_VFQ
);
2663 dev_info(&pdev
->dev
, "IOV Disabled\n");
2665 /* Re-enable DMA Coalescing flag since IOV is turned off */
2666 adapter
->flags
|= IGB_FLAG_DMAC
;
2672 static int igb_enable_sriov(struct pci_dev
*pdev
, int num_vfs
)
2674 struct net_device
*netdev
= pci_get_drvdata(pdev
);
2675 struct igb_adapter
*adapter
= netdev_priv(netdev
);
2676 int old_vfs
= pci_num_vf(pdev
);
2680 if (!(adapter
->flags
& IGB_FLAG_HAS_MSIX
) || num_vfs
> 7) {
2688 dev_info(&pdev
->dev
, "%d pre-allocated VFs found - override max_vfs setting of %d\n",
2690 adapter
->vfs_allocated_count
= old_vfs
;
2692 adapter
->vfs_allocated_count
= num_vfs
;
2694 adapter
->vf_data
= kcalloc(adapter
->vfs_allocated_count
,
2695 sizeof(struct vf_data_storage
), GFP_KERNEL
);
2697 /* if allocation failed then we do not support SR-IOV */
2698 if (!adapter
->vf_data
) {
2699 adapter
->vfs_allocated_count
= 0;
2701 "Unable to allocate memory for VF Data Storage\n");
2706 /* only call pci_enable_sriov() if no VFs are allocated already */
2708 err
= pci_enable_sriov(pdev
, adapter
->vfs_allocated_count
);
2712 dev_info(&pdev
->dev
, "%d VFs allocated\n",
2713 adapter
->vfs_allocated_count
);
2714 for (i
= 0; i
< adapter
->vfs_allocated_count
; i
++)
2715 igb_vf_configure(adapter
, i
);
2717 /* DMA Coalescing is not supported in IOV mode. */
2718 adapter
->flags
&= ~IGB_FLAG_DMAC
;
2722 kfree(adapter
->vf_data
);
2723 adapter
->vf_data
= NULL
;
2724 adapter
->vfs_allocated_count
= 0;
2731 * igb_remove_i2c - Cleanup I2C interface
2732 * @adapter: pointer to adapter structure
2734 static void igb_remove_i2c(struct igb_adapter
*adapter
)
2736 /* free the adapter bus structure */
2737 i2c_del_adapter(&adapter
->i2c_adap
);
2741 * igb_remove - Device Removal Routine
2742 * @pdev: PCI device information struct
2744 * igb_remove is called by the PCI subsystem to alert the driver
2745 * that it should release a PCI device. The could be caused by a
2746 * Hot-Plug event, or because the driver is going to be removed from
2749 static void igb_remove(struct pci_dev
*pdev
)
2751 struct net_device
*netdev
= pci_get_drvdata(pdev
);
2752 struct igb_adapter
*adapter
= netdev_priv(netdev
);
2753 struct e1000_hw
*hw
= &adapter
->hw
;
2755 pm_runtime_get_noresume(&pdev
->dev
);
2756 #ifdef CONFIG_IGB_HWMON
2757 igb_sysfs_exit(adapter
);
2759 igb_remove_i2c(adapter
);
2760 igb_ptp_stop(adapter
);
2761 /* The watchdog timer may be rescheduled, so explicitly
2762 * disable watchdog from being rescheduled.
2764 set_bit(__IGB_DOWN
, &adapter
->state
);
2765 del_timer_sync(&adapter
->watchdog_timer
);
2766 del_timer_sync(&adapter
->phy_info_timer
);
2768 cancel_work_sync(&adapter
->reset_task
);
2769 cancel_work_sync(&adapter
->watchdog_task
);
2771 #ifdef CONFIG_IGB_DCA
2772 if (adapter
->flags
& IGB_FLAG_DCA_ENABLED
) {
2773 dev_info(&pdev
->dev
, "DCA disabled\n");
2774 dca_remove_requester(&pdev
->dev
);
2775 adapter
->flags
&= ~IGB_FLAG_DCA_ENABLED
;
2776 wr32(E1000_DCA_CTRL
, E1000_DCA_CTRL_DCA_MODE_DISABLE
);
2780 /* Release control of h/w to f/w. If f/w is AMT enabled, this
2781 * would have already happened in close and is redundant.
2783 igb_release_hw_control(adapter
);
2785 unregister_netdev(netdev
);
2787 igb_clear_interrupt_scheme(adapter
);
2789 #ifdef CONFIG_PCI_IOV
2790 igb_disable_sriov(pdev
);
2793 iounmap(hw
->hw_addr
);
2794 if (hw
->flash_address
)
2795 iounmap(hw
->flash_address
);
2796 pci_release_selected_regions(pdev
,
2797 pci_select_bars(pdev
, IORESOURCE_MEM
));
2799 kfree(adapter
->shadow_vfta
);
2800 free_netdev(netdev
);
2802 pci_disable_pcie_error_reporting(pdev
);
2804 pci_disable_device(pdev
);
2808 * igb_probe_vfs - Initialize vf data storage and add VFs to pci config space
2809 * @adapter: board private structure to initialize
2811 * This function initializes the vf specific data storage and then attempts to
2812 * allocate the VFs. The reason for ordering it this way is because it is much
2813 * mor expensive time wise to disable SR-IOV than it is to allocate and free
2814 * the memory for the VFs.
2816 static void igb_probe_vfs(struct igb_adapter
*adapter
)
2818 #ifdef CONFIG_PCI_IOV
2819 struct pci_dev
*pdev
= adapter
->pdev
;
2820 struct e1000_hw
*hw
= &adapter
->hw
;
2822 /* Virtualization features not supported on i210 family. */
2823 if ((hw
->mac
.type
== e1000_i210
) || (hw
->mac
.type
== e1000_i211
))
2826 pci_sriov_set_totalvfs(pdev
, 7);
2827 igb_pci_enable_sriov(pdev
, max_vfs
);
2829 #endif /* CONFIG_PCI_IOV */
2832 static void igb_init_queue_configuration(struct igb_adapter
*adapter
)
2834 struct e1000_hw
*hw
= &adapter
->hw
;
2837 /* Determine the maximum number of RSS queues supported. */
2838 switch (hw
->mac
.type
) {
2840 max_rss_queues
= IGB_MAX_RX_QUEUES_I211
;
2844 max_rss_queues
= IGB_MAX_RX_QUEUES_82575
;
2847 /* I350 cannot do RSS and SR-IOV at the same time */
2848 if (!!adapter
->vfs_allocated_count
) {
2854 if (!!adapter
->vfs_allocated_count
) {
2862 max_rss_queues
= IGB_MAX_RX_QUEUES
;
2866 adapter
->rss_queues
= min_t(u32
, max_rss_queues
, num_online_cpus());
2868 /* Determine if we need to pair queues. */
2869 switch (hw
->mac
.type
) {
2872 /* Device supports enough interrupts without queue pairing. */
2875 /* If VFs are going to be allocated with RSS queues then we
2876 * should pair the queues in order to conserve interrupts due
2877 * to limited supply.
2879 if ((adapter
->rss_queues
> 1) &&
2880 (adapter
->vfs_allocated_count
> 6))
2881 adapter
->flags
|= IGB_FLAG_QUEUE_PAIRS
;
2888 /* If rss_queues > half of max_rss_queues, pair the queues in
2889 * order to conserve interrupts due to limited supply.
2891 if (adapter
->rss_queues
> (max_rss_queues
/ 2))
2892 adapter
->flags
|= IGB_FLAG_QUEUE_PAIRS
;
2898 * igb_sw_init - Initialize general software structures (struct igb_adapter)
2899 * @adapter: board private structure to initialize
2901 * igb_sw_init initializes the Adapter private data structure.
2902 * Fields are initialized based on PCI device information and
2903 * OS network device settings (MTU size).
2905 static int igb_sw_init(struct igb_adapter
*adapter
)
2907 struct e1000_hw
*hw
= &adapter
->hw
;
2908 struct net_device
*netdev
= adapter
->netdev
;
2909 struct pci_dev
*pdev
= adapter
->pdev
;
2911 pci_read_config_word(pdev
, PCI_COMMAND
, &hw
->bus
.pci_cmd_word
);
2913 /* set default ring sizes */
2914 adapter
->tx_ring_count
= IGB_DEFAULT_TXD
;
2915 adapter
->rx_ring_count
= IGB_DEFAULT_RXD
;
2917 /* set default ITR values */
2918 adapter
->rx_itr_setting
= IGB_DEFAULT_ITR
;
2919 adapter
->tx_itr_setting
= IGB_DEFAULT_ITR
;
2921 /* set default work limits */
2922 adapter
->tx_work_limit
= IGB_DEFAULT_TX_WORK
;
2924 adapter
->max_frame_size
= netdev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
+
2926 adapter
->min_frame_size
= ETH_ZLEN
+ ETH_FCS_LEN
;
2928 spin_lock_init(&adapter
->stats64_lock
);
2929 #ifdef CONFIG_PCI_IOV
2930 switch (hw
->mac
.type
) {
2934 dev_warn(&pdev
->dev
,
2935 "Maximum of 7 VFs per PF, using max\n");
2936 max_vfs
= adapter
->vfs_allocated_count
= 7;
2938 adapter
->vfs_allocated_count
= max_vfs
;
2939 if (adapter
->vfs_allocated_count
)
2940 dev_warn(&pdev
->dev
,
2941 "Enabling SR-IOV VFs using the module parameter is deprecated - please use the pci sysfs interface.\n");
2946 #endif /* CONFIG_PCI_IOV */
2948 igb_init_queue_configuration(adapter
);
2950 /* Setup and initialize a copy of the hw vlan table array */
2951 adapter
->shadow_vfta
= kcalloc(E1000_VLAN_FILTER_TBL_SIZE
, sizeof(u32
),
2954 /* This call may decrease the number of queues */
2955 if (igb_init_interrupt_scheme(adapter
, true)) {
2956 dev_err(&pdev
->dev
, "Unable to allocate memory for queues\n");
2960 igb_probe_vfs(adapter
);
2962 /* Explicitly disable IRQ since the NIC can be in any state. */
2963 igb_irq_disable(adapter
);
2965 if (hw
->mac
.type
>= e1000_i350
)
2966 adapter
->flags
&= ~IGB_FLAG_DMAC
;
2968 set_bit(__IGB_DOWN
, &adapter
->state
);
2973 * igb_open - Called when a network interface is made active
2974 * @netdev: network interface device structure
2976 * Returns 0 on success, negative value on failure
2978 * The open entry point is called when a network interface is made
2979 * active by the system (IFF_UP). At this point all resources needed
2980 * for transmit and receive operations are allocated, the interrupt
2981 * handler is registered with the OS, the watchdog timer is started,
2982 * and the stack is notified that the interface is ready.
2984 static int __igb_open(struct net_device
*netdev
, bool resuming
)
2986 struct igb_adapter
*adapter
= netdev_priv(netdev
);
2987 struct e1000_hw
*hw
= &adapter
->hw
;
2988 struct pci_dev
*pdev
= adapter
->pdev
;
2992 /* disallow open during test */
2993 if (test_bit(__IGB_TESTING
, &adapter
->state
)) {
2999 pm_runtime_get_sync(&pdev
->dev
);
3001 netif_carrier_off(netdev
);
3003 /* allocate transmit descriptors */
3004 err
= igb_setup_all_tx_resources(adapter
);
3008 /* allocate receive descriptors */
3009 err
= igb_setup_all_rx_resources(adapter
);
3013 igb_power_up_link(adapter
);
3015 /* before we allocate an interrupt, we must be ready to handle it.
3016 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
3017 * as soon as we call pci_request_irq, so we have to setup our
3018 * clean_rx handler before we do so.
3020 igb_configure(adapter
);
3022 err
= igb_request_irq(adapter
);
3026 /* Notify the stack of the actual queue counts. */
3027 err
= netif_set_real_num_tx_queues(adapter
->netdev
,
3028 adapter
->num_tx_queues
);
3030 goto err_set_queues
;
3032 err
= netif_set_real_num_rx_queues(adapter
->netdev
,
3033 adapter
->num_rx_queues
);
3035 goto err_set_queues
;
3037 /* From here on the code is the same as igb_up() */
3038 clear_bit(__IGB_DOWN
, &adapter
->state
);
3040 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
3041 napi_enable(&(adapter
->q_vector
[i
]->napi
));
3043 /* Clear any pending interrupts. */
3046 igb_irq_enable(adapter
);
3048 /* notify VFs that reset has been completed */
3049 if (adapter
->vfs_allocated_count
) {
3050 u32 reg_data
= rd32(E1000_CTRL_EXT
);
3051 reg_data
|= E1000_CTRL_EXT_PFRSTD
;
3052 wr32(E1000_CTRL_EXT
, reg_data
);
3055 netif_tx_start_all_queues(netdev
);
3058 pm_runtime_put(&pdev
->dev
);
3060 /* start the watchdog. */
3061 hw
->mac
.get_link_status
= 1;
3062 schedule_work(&adapter
->watchdog_task
);
3067 igb_free_irq(adapter
);
3069 igb_release_hw_control(adapter
);
3070 igb_power_down_link(adapter
);
3071 igb_free_all_rx_resources(adapter
);
3073 igb_free_all_tx_resources(adapter
);
3077 pm_runtime_put(&pdev
->dev
);
3082 static int igb_open(struct net_device
*netdev
)
3084 return __igb_open(netdev
, false);
3088 * igb_close - Disables a network interface
3089 * @netdev: network interface device structure
3091 * Returns 0, this is not allowed to fail
3093 * The close entry point is called when an interface is de-activated
3094 * by the OS. The hardware is still under the driver's control, but
3095 * needs to be disabled. A global MAC reset is issued to stop the
3096 * hardware, and all transmit and receive resources are freed.
3098 static int __igb_close(struct net_device
*netdev
, bool suspending
)
3100 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3101 struct pci_dev
*pdev
= adapter
->pdev
;
3103 WARN_ON(test_bit(__IGB_RESETTING
, &adapter
->state
));
3106 pm_runtime_get_sync(&pdev
->dev
);
3109 igb_free_irq(adapter
);
3111 igb_free_all_tx_resources(adapter
);
3112 igb_free_all_rx_resources(adapter
);
3115 pm_runtime_put_sync(&pdev
->dev
);
3119 static int igb_close(struct net_device
*netdev
)
3121 return __igb_close(netdev
, false);
3125 * igb_setup_tx_resources - allocate Tx resources (Descriptors)
3126 * @tx_ring: tx descriptor ring (for a specific queue) to setup
3128 * Return 0 on success, negative on failure
3130 int igb_setup_tx_resources(struct igb_ring
*tx_ring
)
3132 struct device
*dev
= tx_ring
->dev
;
3135 size
= sizeof(struct igb_tx_buffer
) * tx_ring
->count
;
3137 tx_ring
->tx_buffer_info
= vzalloc(size
);
3138 if (!tx_ring
->tx_buffer_info
)
3141 /* round up to nearest 4K */
3142 tx_ring
->size
= tx_ring
->count
* sizeof(union e1000_adv_tx_desc
);
3143 tx_ring
->size
= ALIGN(tx_ring
->size
, 4096);
3145 tx_ring
->desc
= dma_alloc_coherent(dev
, tx_ring
->size
,
3146 &tx_ring
->dma
, GFP_KERNEL
);
3150 tx_ring
->next_to_use
= 0;
3151 tx_ring
->next_to_clean
= 0;
3156 vfree(tx_ring
->tx_buffer_info
);
3157 tx_ring
->tx_buffer_info
= NULL
;
3158 dev_err(dev
, "Unable to allocate memory for the Tx descriptor ring\n");
3163 * igb_setup_all_tx_resources - wrapper to allocate Tx resources
3164 * (Descriptors) for all queues
3165 * @adapter: board private structure
3167 * Return 0 on success, negative on failure
3169 static int igb_setup_all_tx_resources(struct igb_adapter
*adapter
)
3171 struct pci_dev
*pdev
= adapter
->pdev
;
3174 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
3175 err
= igb_setup_tx_resources(adapter
->tx_ring
[i
]);
3178 "Allocation for Tx Queue %u failed\n", i
);
3179 for (i
--; i
>= 0; i
--)
3180 igb_free_tx_resources(adapter
->tx_ring
[i
]);
3189 * igb_setup_tctl - configure the transmit control registers
3190 * @adapter: Board private structure
3192 void igb_setup_tctl(struct igb_adapter
*adapter
)
3194 struct e1000_hw
*hw
= &adapter
->hw
;
3197 /* disable queue 0 which is enabled by default on 82575 and 82576 */
3198 wr32(E1000_TXDCTL(0), 0);
3200 /* Program the Transmit Control Register */
3201 tctl
= rd32(E1000_TCTL
);
3202 tctl
&= ~E1000_TCTL_CT
;
3203 tctl
|= E1000_TCTL_PSP
| E1000_TCTL_RTLC
|
3204 (E1000_COLLISION_THRESHOLD
<< E1000_CT_SHIFT
);
3206 igb_config_collision_dist(hw
);
3208 /* Enable transmits */
3209 tctl
|= E1000_TCTL_EN
;
3211 wr32(E1000_TCTL
, tctl
);
3215 * igb_configure_tx_ring - Configure transmit ring after Reset
3216 * @adapter: board private structure
3217 * @ring: tx ring to configure
3219 * Configure a transmit ring after a reset.
3221 void igb_configure_tx_ring(struct igb_adapter
*adapter
,
3222 struct igb_ring
*ring
)
3224 struct e1000_hw
*hw
= &adapter
->hw
;
3226 u64 tdba
= ring
->dma
;
3227 int reg_idx
= ring
->reg_idx
;
3229 /* disable the queue */
3230 wr32(E1000_TXDCTL(reg_idx
), 0);
3234 wr32(E1000_TDLEN(reg_idx
),
3235 ring
->count
* sizeof(union e1000_adv_tx_desc
));
3236 wr32(E1000_TDBAL(reg_idx
),
3237 tdba
& 0x00000000ffffffffULL
);
3238 wr32(E1000_TDBAH(reg_idx
), tdba
>> 32);
3240 ring
->tail
= hw
->hw_addr
+ E1000_TDT(reg_idx
);
3241 wr32(E1000_TDH(reg_idx
), 0);
3242 writel(0, ring
->tail
);
3244 txdctl
|= IGB_TX_PTHRESH
;
3245 txdctl
|= IGB_TX_HTHRESH
<< 8;
3246 txdctl
|= IGB_TX_WTHRESH
<< 16;
3248 txdctl
|= E1000_TXDCTL_QUEUE_ENABLE
;
3249 wr32(E1000_TXDCTL(reg_idx
), txdctl
);
3253 * igb_configure_tx - Configure transmit Unit after Reset
3254 * @adapter: board private structure
3256 * Configure the Tx unit of the MAC after a reset.
3258 static void igb_configure_tx(struct igb_adapter
*adapter
)
3262 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
3263 igb_configure_tx_ring(adapter
, adapter
->tx_ring
[i
]);
3267 * igb_setup_rx_resources - allocate Rx resources (Descriptors)
3268 * @rx_ring: Rx descriptor ring (for a specific queue) to setup
3270 * Returns 0 on success, negative on failure
3272 int igb_setup_rx_resources(struct igb_ring
*rx_ring
)
3274 struct device
*dev
= rx_ring
->dev
;
3277 size
= sizeof(struct igb_rx_buffer
) * rx_ring
->count
;
3279 rx_ring
->rx_buffer_info
= vzalloc(size
);
3280 if (!rx_ring
->rx_buffer_info
)
3283 /* Round up to nearest 4K */
3284 rx_ring
->size
= rx_ring
->count
* sizeof(union e1000_adv_rx_desc
);
3285 rx_ring
->size
= ALIGN(rx_ring
->size
, 4096);
3287 rx_ring
->desc
= dma_alloc_coherent(dev
, rx_ring
->size
,
3288 &rx_ring
->dma
, GFP_KERNEL
);
3292 rx_ring
->next_to_alloc
= 0;
3293 rx_ring
->next_to_clean
= 0;
3294 rx_ring
->next_to_use
= 0;
3299 vfree(rx_ring
->rx_buffer_info
);
3300 rx_ring
->rx_buffer_info
= NULL
;
3301 dev_err(dev
, "Unable to allocate memory for the Rx descriptor ring\n");
3306 * igb_setup_all_rx_resources - wrapper to allocate Rx resources
3307 * (Descriptors) for all queues
3308 * @adapter: board private structure
3310 * Return 0 on success, negative on failure
3312 static int igb_setup_all_rx_resources(struct igb_adapter
*adapter
)
3314 struct pci_dev
*pdev
= adapter
->pdev
;
3317 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3318 err
= igb_setup_rx_resources(adapter
->rx_ring
[i
]);
3321 "Allocation for Rx Queue %u failed\n", i
);
3322 for (i
--; i
>= 0; i
--)
3323 igb_free_rx_resources(adapter
->rx_ring
[i
]);
3332 * igb_setup_mrqc - configure the multiple receive queue control registers
3333 * @adapter: Board private structure
3335 static void igb_setup_mrqc(struct igb_adapter
*adapter
)
3337 struct e1000_hw
*hw
= &adapter
->hw
;
3339 u32 j
, num_rx_queues
;
3340 static const u32 rsskey
[10] = { 0xDA565A6D, 0xC20E5B25, 0x3D256741,
3341 0xB08FA343, 0xCB2BCAD0, 0xB4307BAE,
3342 0xA32DCB77, 0x0CF23080, 0x3BB7426A,
3345 /* Fill out hash function seeds */
3346 for (j
= 0; j
< 10; j
++)
3347 wr32(E1000_RSSRK(j
), rsskey
[j
]);
3349 num_rx_queues
= adapter
->rss_queues
;
3351 switch (hw
->mac
.type
) {
3353 /* 82576 supports 2 RSS queues for SR-IOV */
3354 if (adapter
->vfs_allocated_count
)
3361 if (adapter
->rss_indir_tbl_init
!= num_rx_queues
) {
3362 for (j
= 0; j
< IGB_RETA_SIZE
; j
++)
3363 adapter
->rss_indir_tbl
[j
] = (j
* num_rx_queues
) / IGB_RETA_SIZE
;
3364 adapter
->rss_indir_tbl_init
= num_rx_queues
;
3366 igb_write_rss_indir_tbl(adapter
);
3368 /* Disable raw packet checksumming so that RSS hash is placed in
3369 * descriptor on writeback. No need to enable TCP/UDP/IP checksum
3370 * offloads as they are enabled by default
3372 rxcsum
= rd32(E1000_RXCSUM
);
3373 rxcsum
|= E1000_RXCSUM_PCSD
;
3375 if (adapter
->hw
.mac
.type
>= e1000_82576
)
3376 /* Enable Receive Checksum Offload for SCTP */
3377 rxcsum
|= E1000_RXCSUM_CRCOFL
;
3379 /* Don't need to set TUOFL or IPOFL, they default to 1 */
3380 wr32(E1000_RXCSUM
, rxcsum
);
3382 /* Generate RSS hash based on packet types, TCP/UDP
3383 * port numbers and/or IPv4/v6 src and dst addresses
3385 mrqc
= E1000_MRQC_RSS_FIELD_IPV4
|
3386 E1000_MRQC_RSS_FIELD_IPV4_TCP
|
3387 E1000_MRQC_RSS_FIELD_IPV6
|
3388 E1000_MRQC_RSS_FIELD_IPV6_TCP
|
3389 E1000_MRQC_RSS_FIELD_IPV6_TCP_EX
;
3391 if (adapter
->flags
& IGB_FLAG_RSS_FIELD_IPV4_UDP
)
3392 mrqc
|= E1000_MRQC_RSS_FIELD_IPV4_UDP
;
3393 if (adapter
->flags
& IGB_FLAG_RSS_FIELD_IPV6_UDP
)
3394 mrqc
|= E1000_MRQC_RSS_FIELD_IPV6_UDP
;
3396 /* If VMDq is enabled then we set the appropriate mode for that, else
3397 * we default to RSS so that an RSS hash is calculated per packet even
3398 * if we are only using one queue
3400 if (adapter
->vfs_allocated_count
) {
3401 if (hw
->mac
.type
> e1000_82575
) {
3402 /* Set the default pool for the PF's first queue */
3403 u32 vtctl
= rd32(E1000_VT_CTL
);
3404 vtctl
&= ~(E1000_VT_CTL_DEFAULT_POOL_MASK
|
3405 E1000_VT_CTL_DISABLE_DEF_POOL
);
3406 vtctl
|= adapter
->vfs_allocated_count
<<
3407 E1000_VT_CTL_DEFAULT_POOL_SHIFT
;
3408 wr32(E1000_VT_CTL
, vtctl
);
3410 if (adapter
->rss_queues
> 1)
3411 mrqc
|= E1000_MRQC_ENABLE_VMDQ_RSS_2Q
;
3413 mrqc
|= E1000_MRQC_ENABLE_VMDQ
;
3415 if (hw
->mac
.type
!= e1000_i211
)
3416 mrqc
|= E1000_MRQC_ENABLE_RSS_4Q
;
3418 igb_vmm_control(adapter
);
3420 wr32(E1000_MRQC
, mrqc
);
3424 * igb_setup_rctl - configure the receive control registers
3425 * @adapter: Board private structure
3427 void igb_setup_rctl(struct igb_adapter
*adapter
)
3429 struct e1000_hw
*hw
= &adapter
->hw
;
3432 rctl
= rd32(E1000_RCTL
);
3434 rctl
&= ~(3 << E1000_RCTL_MO_SHIFT
);
3435 rctl
&= ~(E1000_RCTL_LBM_TCVR
| E1000_RCTL_LBM_MAC
);
3437 rctl
|= E1000_RCTL_EN
| E1000_RCTL_BAM
| E1000_RCTL_RDMTS_HALF
|
3438 (hw
->mac
.mc_filter_type
<< E1000_RCTL_MO_SHIFT
);
3440 /* enable stripping of CRC. It's unlikely this will break BMC
3441 * redirection as it did with e1000. Newer features require
3442 * that the HW strips the CRC.
3444 rctl
|= E1000_RCTL_SECRC
;
3446 /* disable store bad packets and clear size bits. */
3447 rctl
&= ~(E1000_RCTL_SBP
| E1000_RCTL_SZ_256
);
3449 /* enable LPE to prevent packets larger than max_frame_size */
3450 rctl
|= E1000_RCTL_LPE
;
3452 /* disable queue 0 to prevent tail write w/o re-config */
3453 wr32(E1000_RXDCTL(0), 0);
3455 /* Attention!!! For SR-IOV PF driver operations you must enable
3456 * queue drop for all VF and PF queues to prevent head of line blocking
3457 * if an un-trusted VF does not provide descriptors to hardware.
3459 if (adapter
->vfs_allocated_count
) {
3460 /* set all queue drop enable bits */
3461 wr32(E1000_QDE
, ALL_QUEUES
);
3464 /* This is useful for sniffing bad packets. */
3465 if (adapter
->netdev
->features
& NETIF_F_RXALL
) {
3466 /* UPE and MPE will be handled by normal PROMISC logic
3467 * in e1000e_set_rx_mode
3469 rctl
|= (E1000_RCTL_SBP
| /* Receive bad packets */
3470 E1000_RCTL_BAM
| /* RX All Bcast Pkts */
3471 E1000_RCTL_PMCF
); /* RX All MAC Ctrl Pkts */
3473 rctl
&= ~(E1000_RCTL_VFE
| /* Disable VLAN filter */
3474 E1000_RCTL_DPF
| /* Allow filtered pause */
3475 E1000_RCTL_CFIEN
); /* Dis VLAN CFIEN Filter */
3476 /* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3477 * and that breaks VLANs.
3481 wr32(E1000_RCTL
, rctl
);
3484 static inline int igb_set_vf_rlpml(struct igb_adapter
*adapter
, int size
,
3487 struct e1000_hw
*hw
= &adapter
->hw
;
3490 /* if it isn't the PF check to see if VFs are enabled and
3491 * increase the size to support vlan tags
3493 if (vfn
< adapter
->vfs_allocated_count
&&
3494 adapter
->vf_data
[vfn
].vlans_enabled
)
3495 size
+= VLAN_TAG_SIZE
;
3497 vmolr
= rd32(E1000_VMOLR(vfn
));
3498 vmolr
&= ~E1000_VMOLR_RLPML_MASK
;
3499 vmolr
|= size
| E1000_VMOLR_LPE
;
3500 wr32(E1000_VMOLR(vfn
), vmolr
);
3506 * igb_rlpml_set - set maximum receive packet size
3507 * @adapter: board private structure
3509 * Configure maximum receivable packet size.
3511 static void igb_rlpml_set(struct igb_adapter
*adapter
)
3513 u32 max_frame_size
= adapter
->max_frame_size
;
3514 struct e1000_hw
*hw
= &adapter
->hw
;
3515 u16 pf_id
= adapter
->vfs_allocated_count
;
3518 igb_set_vf_rlpml(adapter
, max_frame_size
, pf_id
);
3519 /* If we're in VMDQ or SR-IOV mode, then set global RLPML
3520 * to our max jumbo frame size, in case we need to enable
3521 * jumbo frames on one of the rings later.
3522 * This will not pass over-length frames into the default
3523 * queue because it's gated by the VMOLR.RLPML.
3525 max_frame_size
= MAX_JUMBO_FRAME_SIZE
;
3528 wr32(E1000_RLPML
, max_frame_size
);
3531 static inline void igb_set_vmolr(struct igb_adapter
*adapter
,
3534 struct e1000_hw
*hw
= &adapter
->hw
;
3537 /* This register exists only on 82576 and newer so if we are older then
3538 * we should exit and do nothing
3540 if (hw
->mac
.type
< e1000_82576
)
3543 vmolr
= rd32(E1000_VMOLR(vfn
));
3544 vmolr
|= E1000_VMOLR_STRVLAN
; /* Strip vlan tags */
3546 vmolr
|= E1000_VMOLR_AUPE
; /* Accept untagged packets */
3548 vmolr
&= ~(E1000_VMOLR_AUPE
); /* Tagged packets ONLY */
3550 /* clear all bits that might not be set */
3551 vmolr
&= ~(E1000_VMOLR_BAM
| E1000_VMOLR_RSSE
);
3553 if (adapter
->rss_queues
> 1 && vfn
== adapter
->vfs_allocated_count
)
3554 vmolr
|= E1000_VMOLR_RSSE
; /* enable RSS */
3555 /* for VMDq only allow the VFs and pool 0 to accept broadcast and
3558 if (vfn
<= adapter
->vfs_allocated_count
)
3559 vmolr
|= E1000_VMOLR_BAM
; /* Accept broadcast */
3561 wr32(E1000_VMOLR(vfn
), vmolr
);
3565 * igb_configure_rx_ring - Configure a receive ring after Reset
3566 * @adapter: board private structure
3567 * @ring: receive ring to be configured
3569 * Configure the Rx unit of the MAC after a reset.
3571 void igb_configure_rx_ring(struct igb_adapter
*adapter
,
3572 struct igb_ring
*ring
)
3574 struct e1000_hw
*hw
= &adapter
->hw
;
3575 u64 rdba
= ring
->dma
;
3576 int reg_idx
= ring
->reg_idx
;
3577 u32 srrctl
= 0, rxdctl
= 0;
3579 /* disable the queue */
3580 wr32(E1000_RXDCTL(reg_idx
), 0);
3582 /* Set DMA base address registers */
3583 wr32(E1000_RDBAL(reg_idx
),
3584 rdba
& 0x00000000ffffffffULL
);
3585 wr32(E1000_RDBAH(reg_idx
), rdba
>> 32);
3586 wr32(E1000_RDLEN(reg_idx
),
3587 ring
->count
* sizeof(union e1000_adv_rx_desc
));
3589 /* initialize head and tail */
3590 ring
->tail
= hw
->hw_addr
+ E1000_RDT(reg_idx
);
3591 wr32(E1000_RDH(reg_idx
), 0);
3592 writel(0, ring
->tail
);
3594 /* set descriptor configuration */
3595 srrctl
= IGB_RX_HDR_LEN
<< E1000_SRRCTL_BSIZEHDRSIZE_SHIFT
;
3596 srrctl
|= IGB_RX_BUFSZ
>> E1000_SRRCTL_BSIZEPKT_SHIFT
;
3597 srrctl
|= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF
;
3598 if (hw
->mac
.type
>= e1000_82580
)
3599 srrctl
|= E1000_SRRCTL_TIMESTAMP
;
3600 /* Only set Drop Enable if we are supporting multiple queues */
3601 if (adapter
->vfs_allocated_count
|| adapter
->num_rx_queues
> 1)
3602 srrctl
|= E1000_SRRCTL_DROP_EN
;
3604 wr32(E1000_SRRCTL(reg_idx
), srrctl
);
3606 /* set filtering for VMDQ pools */
3607 igb_set_vmolr(adapter
, reg_idx
& 0x7, true);
3609 rxdctl
|= IGB_RX_PTHRESH
;
3610 rxdctl
|= IGB_RX_HTHRESH
<< 8;
3611 rxdctl
|= IGB_RX_WTHRESH
<< 16;
3613 /* enable receive descriptor fetching */
3614 rxdctl
|= E1000_RXDCTL_QUEUE_ENABLE
;
3615 wr32(E1000_RXDCTL(reg_idx
), rxdctl
);
3619 * igb_configure_rx - Configure receive Unit after Reset
3620 * @adapter: board private structure
3622 * Configure the Rx unit of the MAC after a reset.
3624 static void igb_configure_rx(struct igb_adapter
*adapter
)
3628 /* set UTA to appropriate mode */
3629 igb_set_uta(adapter
);
3631 /* set the correct pool for the PF default MAC address in entry 0 */
3632 igb_rar_set_qsel(adapter
, adapter
->hw
.mac
.addr
, 0,
3633 adapter
->vfs_allocated_count
);
3635 /* Setup the HW Rx Head and Tail Descriptor Pointers and
3636 * the Base and Length of the Rx Descriptor Ring
3638 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3639 igb_configure_rx_ring(adapter
, adapter
->rx_ring
[i
]);
3643 * igb_free_tx_resources - Free Tx Resources per Queue
3644 * @tx_ring: Tx descriptor ring for a specific queue
3646 * Free all transmit software resources
3648 void igb_free_tx_resources(struct igb_ring
*tx_ring
)
3650 igb_clean_tx_ring(tx_ring
);
3652 vfree(tx_ring
->tx_buffer_info
);
3653 tx_ring
->tx_buffer_info
= NULL
;
3655 /* if not set, then don't free */
3659 dma_free_coherent(tx_ring
->dev
, tx_ring
->size
,
3660 tx_ring
->desc
, tx_ring
->dma
);
3662 tx_ring
->desc
= NULL
;
3666 * igb_free_all_tx_resources - Free Tx Resources for All Queues
3667 * @adapter: board private structure
3669 * Free all transmit software resources
3671 static void igb_free_all_tx_resources(struct igb_adapter
*adapter
)
3675 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
3676 igb_free_tx_resources(adapter
->tx_ring
[i
]);
3679 void igb_unmap_and_free_tx_resource(struct igb_ring
*ring
,
3680 struct igb_tx_buffer
*tx_buffer
)
3682 if (tx_buffer
->skb
) {
3683 dev_kfree_skb_any(tx_buffer
->skb
);
3684 if (dma_unmap_len(tx_buffer
, len
))
3685 dma_unmap_single(ring
->dev
,
3686 dma_unmap_addr(tx_buffer
, dma
),
3687 dma_unmap_len(tx_buffer
, len
),
3689 } else if (dma_unmap_len(tx_buffer
, len
)) {
3690 dma_unmap_page(ring
->dev
,
3691 dma_unmap_addr(tx_buffer
, dma
),
3692 dma_unmap_len(tx_buffer
, len
),
3695 tx_buffer
->next_to_watch
= NULL
;
3696 tx_buffer
->skb
= NULL
;
3697 dma_unmap_len_set(tx_buffer
, len
, 0);
3698 /* buffer_info must be completely set up in the transmit path */
3702 * igb_clean_tx_ring - Free Tx Buffers
3703 * @tx_ring: ring to be cleaned
3705 static void igb_clean_tx_ring(struct igb_ring
*tx_ring
)
3707 struct igb_tx_buffer
*buffer_info
;
3711 if (!tx_ring
->tx_buffer_info
)
3713 /* Free all the Tx ring sk_buffs */
3715 for (i
= 0; i
< tx_ring
->count
; i
++) {
3716 buffer_info
= &tx_ring
->tx_buffer_info
[i
];
3717 igb_unmap_and_free_tx_resource(tx_ring
, buffer_info
);
3720 netdev_tx_reset_queue(txring_txq(tx_ring
));
3722 size
= sizeof(struct igb_tx_buffer
) * tx_ring
->count
;
3723 memset(tx_ring
->tx_buffer_info
, 0, size
);
3725 /* Zero out the descriptor ring */
3726 memset(tx_ring
->desc
, 0, tx_ring
->size
);
3728 tx_ring
->next_to_use
= 0;
3729 tx_ring
->next_to_clean
= 0;
3733 * igb_clean_all_tx_rings - Free Tx Buffers for all queues
3734 * @adapter: board private structure
3736 static void igb_clean_all_tx_rings(struct igb_adapter
*adapter
)
3740 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
3741 igb_clean_tx_ring(adapter
->tx_ring
[i
]);
3745 * igb_free_rx_resources - Free Rx Resources
3746 * @rx_ring: ring to clean the resources from
3748 * Free all receive software resources
3750 void igb_free_rx_resources(struct igb_ring
*rx_ring
)
3752 igb_clean_rx_ring(rx_ring
);
3754 vfree(rx_ring
->rx_buffer_info
);
3755 rx_ring
->rx_buffer_info
= NULL
;
3757 /* if not set, then don't free */
3761 dma_free_coherent(rx_ring
->dev
, rx_ring
->size
,
3762 rx_ring
->desc
, rx_ring
->dma
);
3764 rx_ring
->desc
= NULL
;
3768 * igb_free_all_rx_resources - Free Rx Resources for All Queues
3769 * @adapter: board private structure
3771 * Free all receive software resources
3773 static void igb_free_all_rx_resources(struct igb_adapter
*adapter
)
3777 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3778 igb_free_rx_resources(adapter
->rx_ring
[i
]);
3782 * igb_clean_rx_ring - Free Rx Buffers per Queue
3783 * @rx_ring: ring to free buffers from
3785 static void igb_clean_rx_ring(struct igb_ring
*rx_ring
)
3791 dev_kfree_skb(rx_ring
->skb
);
3792 rx_ring
->skb
= NULL
;
3794 if (!rx_ring
->rx_buffer_info
)
3797 /* Free all the Rx ring sk_buffs */
3798 for (i
= 0; i
< rx_ring
->count
; i
++) {
3799 struct igb_rx_buffer
*buffer_info
= &rx_ring
->rx_buffer_info
[i
];
3801 if (!buffer_info
->page
)
3804 dma_unmap_page(rx_ring
->dev
,
3808 __free_page(buffer_info
->page
);
3810 buffer_info
->page
= NULL
;
3813 size
= sizeof(struct igb_rx_buffer
) * rx_ring
->count
;
3814 memset(rx_ring
->rx_buffer_info
, 0, size
);
3816 /* Zero out the descriptor ring */
3817 memset(rx_ring
->desc
, 0, rx_ring
->size
);
3819 rx_ring
->next_to_alloc
= 0;
3820 rx_ring
->next_to_clean
= 0;
3821 rx_ring
->next_to_use
= 0;
3825 * igb_clean_all_rx_rings - Free Rx Buffers for all queues
3826 * @adapter: board private structure
3828 static void igb_clean_all_rx_rings(struct igb_adapter
*adapter
)
3832 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3833 igb_clean_rx_ring(adapter
->rx_ring
[i
]);
3837 * igb_set_mac - Change the Ethernet Address of the NIC
3838 * @netdev: network interface device structure
3839 * @p: pointer to an address structure
3841 * Returns 0 on success, negative on failure
3843 static int igb_set_mac(struct net_device
*netdev
, void *p
)
3845 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3846 struct e1000_hw
*hw
= &adapter
->hw
;
3847 struct sockaddr
*addr
= p
;
3849 if (!is_valid_ether_addr(addr
->sa_data
))
3850 return -EADDRNOTAVAIL
;
3852 memcpy(netdev
->dev_addr
, addr
->sa_data
, netdev
->addr_len
);
3853 memcpy(hw
->mac
.addr
, addr
->sa_data
, netdev
->addr_len
);
3855 /* set the correct pool for the new PF MAC address in entry 0 */
3856 igb_rar_set_qsel(adapter
, hw
->mac
.addr
, 0,
3857 adapter
->vfs_allocated_count
);
3863 * igb_write_mc_addr_list - write multicast addresses to MTA
3864 * @netdev: network interface device structure
3866 * Writes multicast address list to the MTA hash table.
3867 * Returns: -ENOMEM on failure
3868 * 0 on no addresses written
3869 * X on writing X addresses to MTA
3871 static int igb_write_mc_addr_list(struct net_device
*netdev
)
3873 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3874 struct e1000_hw
*hw
= &adapter
->hw
;
3875 struct netdev_hw_addr
*ha
;
3879 if (netdev_mc_empty(netdev
)) {
3880 /* nothing to program, so clear mc list */
3881 igb_update_mc_addr_list(hw
, NULL
, 0);
3882 igb_restore_vf_multicasts(adapter
);
3886 mta_list
= kzalloc(netdev_mc_count(netdev
) * 6, GFP_ATOMIC
);
3890 /* The shared function expects a packed array of only addresses. */
3892 netdev_for_each_mc_addr(ha
, netdev
)
3893 memcpy(mta_list
+ (i
++ * ETH_ALEN
), ha
->addr
, ETH_ALEN
);
3895 igb_update_mc_addr_list(hw
, mta_list
, i
);
3898 return netdev_mc_count(netdev
);
3902 * igb_write_uc_addr_list - write unicast addresses to RAR table
3903 * @netdev: network interface device structure
3905 * Writes unicast address list to the RAR table.
3906 * Returns: -ENOMEM on failure/insufficient address space
3907 * 0 on no addresses written
3908 * X on writing X addresses to the RAR table
3910 static int igb_write_uc_addr_list(struct net_device
*netdev
)
3912 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3913 struct e1000_hw
*hw
= &adapter
->hw
;
3914 unsigned int vfn
= adapter
->vfs_allocated_count
;
3915 unsigned int rar_entries
= hw
->mac
.rar_entry_count
- (vfn
+ 1);
3918 /* return ENOMEM indicating insufficient memory for addresses */
3919 if (netdev_uc_count(netdev
) > rar_entries
)
3922 if (!netdev_uc_empty(netdev
) && rar_entries
) {
3923 struct netdev_hw_addr
*ha
;
3925 netdev_for_each_uc_addr(ha
, netdev
) {
3928 igb_rar_set_qsel(adapter
, ha
->addr
,
3934 /* write the addresses in reverse order to avoid write combining */
3935 for (; rar_entries
> 0 ; rar_entries
--) {
3936 wr32(E1000_RAH(rar_entries
), 0);
3937 wr32(E1000_RAL(rar_entries
), 0);
3945 * igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
3946 * @netdev: network interface device structure
3948 * The set_rx_mode entry point is called whenever the unicast or multicast
3949 * address lists or the network interface flags are updated. This routine is
3950 * responsible for configuring the hardware for proper unicast, multicast,
3951 * promiscuous mode, and all-multi behavior.
3953 static void igb_set_rx_mode(struct net_device
*netdev
)
3955 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3956 struct e1000_hw
*hw
= &adapter
->hw
;
3957 unsigned int vfn
= adapter
->vfs_allocated_count
;
3958 u32 rctl
, vmolr
= 0;
3961 /* Check for Promiscuous and All Multicast modes */
3962 rctl
= rd32(E1000_RCTL
);
3964 /* clear the effected bits */
3965 rctl
&= ~(E1000_RCTL_UPE
| E1000_RCTL_MPE
| E1000_RCTL_VFE
);
3967 if (netdev
->flags
& IFF_PROMISC
) {
3968 /* retain VLAN HW filtering if in VT mode */
3969 if (adapter
->vfs_allocated_count
)
3970 rctl
|= E1000_RCTL_VFE
;
3971 rctl
|= (E1000_RCTL_UPE
| E1000_RCTL_MPE
);
3972 vmolr
|= (E1000_VMOLR_ROPE
| E1000_VMOLR_MPME
);
3974 if (netdev
->flags
& IFF_ALLMULTI
) {
3975 rctl
|= E1000_RCTL_MPE
;
3976 vmolr
|= E1000_VMOLR_MPME
;
3978 /* Write addresses to the MTA, if the attempt fails
3979 * then we should just turn on promiscuous mode so
3980 * that we can at least receive multicast traffic
3982 count
= igb_write_mc_addr_list(netdev
);
3984 rctl
|= E1000_RCTL_MPE
;
3985 vmolr
|= E1000_VMOLR_MPME
;
3987 vmolr
|= E1000_VMOLR_ROMPE
;
3990 /* Write addresses to available RAR registers, if there is not
3991 * sufficient space to store all the addresses then enable
3992 * unicast promiscuous mode
3994 count
= igb_write_uc_addr_list(netdev
);
3996 rctl
|= E1000_RCTL_UPE
;
3997 vmolr
|= E1000_VMOLR_ROPE
;
3999 rctl
|= E1000_RCTL_VFE
;
4001 wr32(E1000_RCTL
, rctl
);
4003 /* In order to support SR-IOV and eventually VMDq it is necessary to set
4004 * the VMOLR to enable the appropriate modes. Without this workaround
4005 * we will have issues with VLAN tag stripping not being done for frames
4006 * that are only arriving because we are the default pool
4008 if ((hw
->mac
.type
< e1000_82576
) || (hw
->mac
.type
> e1000_i350
))
4011 vmolr
|= rd32(E1000_VMOLR(vfn
)) &
4012 ~(E1000_VMOLR_ROPE
| E1000_VMOLR_MPME
| E1000_VMOLR_ROMPE
);
4013 wr32(E1000_VMOLR(vfn
), vmolr
);
4014 igb_restore_vf_multicasts(adapter
);
4017 static void igb_check_wvbr(struct igb_adapter
*adapter
)
4019 struct e1000_hw
*hw
= &adapter
->hw
;
4022 switch (hw
->mac
.type
) {
4025 if (!(wvbr
= rd32(E1000_WVBR
)))
4032 adapter
->wvbr
|= wvbr
;
4035 #define IGB_STAGGERED_QUEUE_OFFSET 8
4037 static void igb_spoof_check(struct igb_adapter
*adapter
)
4044 for(j
= 0; j
< adapter
->vfs_allocated_count
; j
++) {
4045 if (adapter
->wvbr
& (1 << j
) ||
4046 adapter
->wvbr
& (1 << (j
+ IGB_STAGGERED_QUEUE_OFFSET
))) {
4047 dev_warn(&adapter
->pdev
->dev
,
4048 "Spoof event(s) detected on VF %d\n", j
);
4051 (1 << (j
+ IGB_STAGGERED_QUEUE_OFFSET
)));
4056 /* Need to wait a few seconds after link up to get diagnostic information from
4059 static void igb_update_phy_info(unsigned long data
)
4061 struct igb_adapter
*adapter
= (struct igb_adapter
*) data
;
4062 igb_get_phy_info(&adapter
->hw
);
4066 * igb_has_link - check shared code for link and determine up/down
4067 * @adapter: pointer to driver private info
4069 bool igb_has_link(struct igb_adapter
*adapter
)
4071 struct e1000_hw
*hw
= &adapter
->hw
;
4072 bool link_active
= false;
4074 /* get_link_status is set on LSC (link status) interrupt or
4075 * rx sequence error interrupt. get_link_status will stay
4076 * false until the e1000_check_for_link establishes link
4077 * for copper adapters ONLY
4079 switch (hw
->phy
.media_type
) {
4080 case e1000_media_type_copper
:
4081 if (!hw
->mac
.get_link_status
)
4083 case e1000_media_type_internal_serdes
:
4084 hw
->mac
.ops
.check_for_link(hw
);
4085 link_active
= !hw
->mac
.get_link_status
;
4088 case e1000_media_type_unknown
:
4092 if (((hw
->mac
.type
== e1000_i210
) ||
4093 (hw
->mac
.type
== e1000_i211
)) &&
4094 (hw
->phy
.id
== I210_I_PHY_ID
)) {
4095 if (!netif_carrier_ok(adapter
->netdev
)) {
4096 adapter
->flags
&= ~IGB_FLAG_NEED_LINK_UPDATE
;
4097 } else if (!(adapter
->flags
& IGB_FLAG_NEED_LINK_UPDATE
)) {
4098 adapter
->flags
|= IGB_FLAG_NEED_LINK_UPDATE
;
4099 adapter
->link_check_timeout
= jiffies
;
4106 static bool igb_thermal_sensor_event(struct e1000_hw
*hw
, u32 event
)
4109 u32 ctrl_ext
, thstat
;
4111 /* check for thermal sensor event on i350 copper only */
4112 if (hw
->mac
.type
== e1000_i350
) {
4113 thstat
= rd32(E1000_THSTAT
);
4114 ctrl_ext
= rd32(E1000_CTRL_EXT
);
4116 if ((hw
->phy
.media_type
== e1000_media_type_copper
) &&
4117 !(ctrl_ext
& E1000_CTRL_EXT_LINK_MODE_SGMII
))
4118 ret
= !!(thstat
& event
);
4125 * igb_watchdog - Timer Call-back
4126 * @data: pointer to adapter cast into an unsigned long
4128 static void igb_watchdog(unsigned long data
)
4130 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
4131 /* Do the rest outside of interrupt context */
4132 schedule_work(&adapter
->watchdog_task
);
4135 static void igb_watchdog_task(struct work_struct
*work
)
4137 struct igb_adapter
*adapter
= container_of(work
,
4140 struct e1000_hw
*hw
= &adapter
->hw
;
4141 struct e1000_phy_info
*phy
= &hw
->phy
;
4142 struct net_device
*netdev
= adapter
->netdev
;
4147 link
= igb_has_link(adapter
);
4149 if (adapter
->flags
& IGB_FLAG_NEED_LINK_UPDATE
) {
4150 if (time_after(jiffies
, (adapter
->link_check_timeout
+ HZ
)))
4151 adapter
->flags
&= ~IGB_FLAG_NEED_LINK_UPDATE
;
4156 /* Force link down if we have fiber to swap to */
4157 if (adapter
->flags
& IGB_FLAG_MAS_ENABLE
) {
4158 if (hw
->phy
.media_type
== e1000_media_type_copper
) {
4159 connsw
= rd32(E1000_CONNSW
);
4160 if (!(connsw
& E1000_CONNSW_AUTOSENSE_EN
))
4165 /* Perform a reset if the media type changed. */
4166 if (hw
->dev_spec
._82575
.media_changed
) {
4167 hw
->dev_spec
._82575
.media_changed
= false;
4168 adapter
->flags
|= IGB_FLAG_MEDIA_RESET
;
4171 /* Cancel scheduled suspend requests. */
4172 pm_runtime_resume(netdev
->dev
.parent
);
4174 if (!netif_carrier_ok(netdev
)) {
4176 hw
->mac
.ops
.get_speed_and_duplex(hw
,
4177 &adapter
->link_speed
,
4178 &adapter
->link_duplex
);
4180 ctrl
= rd32(E1000_CTRL
);
4181 /* Links status message must follow this format */
4182 printk(KERN_INFO
"igb: %s NIC Link is Up %d Mbps %s "
4183 "Duplex, Flow Control: %s\n",
4185 adapter
->link_speed
,
4186 adapter
->link_duplex
== FULL_DUPLEX
?
4188 (ctrl
& E1000_CTRL_TFCE
) &&
4189 (ctrl
& E1000_CTRL_RFCE
) ? "RX/TX" :
4190 (ctrl
& E1000_CTRL_RFCE
) ? "RX" :
4191 (ctrl
& E1000_CTRL_TFCE
) ? "TX" : "None");
4193 /* disable EEE if enabled */
4194 if ((adapter
->flags
& IGB_FLAG_EEE
) &&
4195 (adapter
->link_duplex
== HALF_DUPLEX
)) {
4196 dev_info(&adapter
->pdev
->dev
,
4197 "EEE Disabled: unsupported at half duplex. Re-enable using ethtool when at full duplex.\n");
4198 adapter
->hw
.dev_spec
._82575
.eee_disable
= true;
4199 adapter
->flags
&= ~IGB_FLAG_EEE
;
4202 /* check if SmartSpeed worked */
4203 igb_check_downshift(hw
);
4204 if (phy
->speed_downgraded
)
4205 netdev_warn(netdev
, "Link Speed was downgraded by SmartSpeed\n");
4207 /* check for thermal sensor event */
4208 if (igb_thermal_sensor_event(hw
,
4209 E1000_THSTAT_LINK_THROTTLE
)) {
4210 netdev_info(netdev
, "The network adapter link "
4211 "speed was downshifted because it "
4215 /* adjust timeout factor according to speed/duplex */
4216 adapter
->tx_timeout_factor
= 1;
4217 switch (adapter
->link_speed
) {
4219 adapter
->tx_timeout_factor
= 14;
4222 /* maybe add some timeout factor ? */
4226 netif_carrier_on(netdev
);
4228 igb_ping_all_vfs(adapter
);
4229 igb_check_vf_rate_limit(adapter
);
4231 /* link state has changed, schedule phy info update */
4232 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
4233 mod_timer(&adapter
->phy_info_timer
,
4234 round_jiffies(jiffies
+ 2 * HZ
));
4237 if (netif_carrier_ok(netdev
)) {
4238 adapter
->link_speed
= 0;
4239 adapter
->link_duplex
= 0;
4241 /* check for thermal sensor event */
4242 if (igb_thermal_sensor_event(hw
,
4243 E1000_THSTAT_PWR_DOWN
)) {
4244 netdev_err(netdev
, "The network adapter was "
4245 "stopped because it overheated\n");
4248 /* Links status message must follow this format */
4249 printk(KERN_INFO
"igb: %s NIC Link is Down\n",
4251 netif_carrier_off(netdev
);
4253 igb_ping_all_vfs(adapter
);
4255 /* link state has changed, schedule phy info update */
4256 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
4257 mod_timer(&adapter
->phy_info_timer
,
4258 round_jiffies(jiffies
+ 2 * HZ
));
4260 /* link is down, time to check for alternate media */
4261 if (adapter
->flags
& IGB_FLAG_MAS_ENABLE
) {
4262 igb_check_swap_media(adapter
);
4263 if (adapter
->flags
& IGB_FLAG_MEDIA_RESET
) {
4264 schedule_work(&adapter
->reset_task
);
4265 /* return immediately */
4269 pm_schedule_suspend(netdev
->dev
.parent
,
4272 /* also check for alternate media here */
4273 } else if (!netif_carrier_ok(netdev
) &&
4274 (adapter
->flags
& IGB_FLAG_MAS_ENABLE
)) {
4275 igb_check_swap_media(adapter
);
4276 if (adapter
->flags
& IGB_FLAG_MEDIA_RESET
) {
4277 schedule_work(&adapter
->reset_task
);
4278 /* return immediately */
4284 spin_lock(&adapter
->stats64_lock
);
4285 igb_update_stats(adapter
, &adapter
->stats64
);
4286 spin_unlock(&adapter
->stats64_lock
);
4288 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
4289 struct igb_ring
*tx_ring
= adapter
->tx_ring
[i
];
4290 if (!netif_carrier_ok(netdev
)) {
4291 /* We've lost link, so the controller stops DMA,
4292 * but we've got queued Tx work that's never going
4293 * to get done, so reset controller to flush Tx.
4294 * (Do the reset outside of interrupt context).
4296 if (igb_desc_unused(tx_ring
) + 1 < tx_ring
->count
) {
4297 adapter
->tx_timeout_count
++;
4298 schedule_work(&adapter
->reset_task
);
4299 /* return immediately since reset is imminent */
4304 /* Force detection of hung controller every watchdog period */
4305 set_bit(IGB_RING_FLAG_TX_DETECT_HANG
, &tx_ring
->flags
);
4308 /* Cause software interrupt to ensure Rx ring is cleaned */
4309 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
4311 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
4312 eics
|= adapter
->q_vector
[i
]->eims_value
;
4313 wr32(E1000_EICS
, eics
);
4315 wr32(E1000_ICS
, E1000_ICS_RXDMT0
);
4318 igb_spoof_check(adapter
);
4319 igb_ptp_rx_hang(adapter
);
4321 /* Reset the timer */
4322 if (!test_bit(__IGB_DOWN
, &adapter
->state
)) {
4323 if (adapter
->flags
& IGB_FLAG_NEED_LINK_UPDATE
)
4324 mod_timer(&adapter
->watchdog_timer
,
4325 round_jiffies(jiffies
+ HZ
));
4327 mod_timer(&adapter
->watchdog_timer
,
4328 round_jiffies(jiffies
+ 2 * HZ
));
4332 enum latency_range
{
4336 latency_invalid
= 255
4340 * igb_update_ring_itr - update the dynamic ITR value based on packet size
4341 * @q_vector: pointer to q_vector
4343 * Stores a new ITR value based on strictly on packet size. This
4344 * algorithm is less sophisticated than that used in igb_update_itr,
4345 * due to the difficulty of synchronizing statistics across multiple
4346 * receive rings. The divisors and thresholds used by this function
4347 * were determined based on theoretical maximum wire speed and testing
4348 * data, in order to minimize response time while increasing bulk
4350 * This functionality is controlled by the InterruptThrottleRate module
4351 * parameter (see igb_param.c)
4352 * NOTE: This function is called only when operating in a multiqueue
4353 * receive environment.
4355 static void igb_update_ring_itr(struct igb_q_vector
*q_vector
)
4357 int new_val
= q_vector
->itr_val
;
4358 int avg_wire_size
= 0;
4359 struct igb_adapter
*adapter
= q_vector
->adapter
;
4360 unsigned int packets
;
4362 /* For non-gigabit speeds, just fix the interrupt rate at 4000
4363 * ints/sec - ITR timer value of 120 ticks.
4365 if (adapter
->link_speed
!= SPEED_1000
) {
4366 new_val
= IGB_4K_ITR
;
4370 packets
= q_vector
->rx
.total_packets
;
4372 avg_wire_size
= q_vector
->rx
.total_bytes
/ packets
;
4374 packets
= q_vector
->tx
.total_packets
;
4376 avg_wire_size
= max_t(u32
, avg_wire_size
,
4377 q_vector
->tx
.total_bytes
/ packets
);
4379 /* if avg_wire_size isn't set no work was done */
4383 /* Add 24 bytes to size to account for CRC, preamble, and gap */
4384 avg_wire_size
+= 24;
4386 /* Don't starve jumbo frames */
4387 avg_wire_size
= min(avg_wire_size
, 3000);
4389 /* Give a little boost to mid-size frames */
4390 if ((avg_wire_size
> 300) && (avg_wire_size
< 1200))
4391 new_val
= avg_wire_size
/ 3;
4393 new_val
= avg_wire_size
/ 2;
4395 /* conservative mode (itr 3) eliminates the lowest_latency setting */
4396 if (new_val
< IGB_20K_ITR
&&
4397 ((q_vector
->rx
.ring
&& adapter
->rx_itr_setting
== 3) ||
4398 (!q_vector
->rx
.ring
&& adapter
->tx_itr_setting
== 3)))
4399 new_val
= IGB_20K_ITR
;
4402 if (new_val
!= q_vector
->itr_val
) {
4403 q_vector
->itr_val
= new_val
;
4404 q_vector
->set_itr
= 1;
4407 q_vector
->rx
.total_bytes
= 0;
4408 q_vector
->rx
.total_packets
= 0;
4409 q_vector
->tx
.total_bytes
= 0;
4410 q_vector
->tx
.total_packets
= 0;
4414 * igb_update_itr - update the dynamic ITR value based on statistics
4415 * @q_vector: pointer to q_vector
4416 * @ring_container: ring info to update the itr for
4418 * Stores a new ITR value based on packets and byte
4419 * counts during the last interrupt. The advantage of per interrupt
4420 * computation is faster updates and more accurate ITR for the current
4421 * traffic pattern. Constants in this function were computed
4422 * based on theoretical maximum wire speed and thresholds were set based
4423 * on testing data as well as attempting to minimize response time
4424 * while increasing bulk throughput.
4425 * this functionality is controlled by the InterruptThrottleRate module
4426 * parameter (see igb_param.c)
4427 * NOTE: These calculations are only valid when operating in a single-
4428 * queue environment.
4430 static void igb_update_itr(struct igb_q_vector
*q_vector
,
4431 struct igb_ring_container
*ring_container
)
4433 unsigned int packets
= ring_container
->total_packets
;
4434 unsigned int bytes
= ring_container
->total_bytes
;
4435 u8 itrval
= ring_container
->itr
;
4437 /* no packets, exit with status unchanged */
4442 case lowest_latency
:
4443 /* handle TSO and jumbo frames */
4444 if (bytes
/packets
> 8000)
4445 itrval
= bulk_latency
;
4446 else if ((packets
< 5) && (bytes
> 512))
4447 itrval
= low_latency
;
4449 case low_latency
: /* 50 usec aka 20000 ints/s */
4450 if (bytes
> 10000) {
4451 /* this if handles the TSO accounting */
4452 if (bytes
/packets
> 8000) {
4453 itrval
= bulk_latency
;
4454 } else if ((packets
< 10) || ((bytes
/packets
) > 1200)) {
4455 itrval
= bulk_latency
;
4456 } else if ((packets
> 35)) {
4457 itrval
= lowest_latency
;
4459 } else if (bytes
/packets
> 2000) {
4460 itrval
= bulk_latency
;
4461 } else if (packets
<= 2 && bytes
< 512) {
4462 itrval
= lowest_latency
;
4465 case bulk_latency
: /* 250 usec aka 4000 ints/s */
4466 if (bytes
> 25000) {
4468 itrval
= low_latency
;
4469 } else if (bytes
< 1500) {
4470 itrval
= low_latency
;
4475 /* clear work counters since we have the values we need */
4476 ring_container
->total_bytes
= 0;
4477 ring_container
->total_packets
= 0;
4479 /* write updated itr to ring container */
4480 ring_container
->itr
= itrval
;
4483 static void igb_set_itr(struct igb_q_vector
*q_vector
)
4485 struct igb_adapter
*adapter
= q_vector
->adapter
;
4486 u32 new_itr
= q_vector
->itr_val
;
4489 /* for non-gigabit speeds, just fix the interrupt rate at 4000 */
4490 if (adapter
->link_speed
!= SPEED_1000
) {
4492 new_itr
= IGB_4K_ITR
;
4496 igb_update_itr(q_vector
, &q_vector
->tx
);
4497 igb_update_itr(q_vector
, &q_vector
->rx
);
4499 current_itr
= max(q_vector
->rx
.itr
, q_vector
->tx
.itr
);
4501 /* conservative mode (itr 3) eliminates the lowest_latency setting */
4502 if (current_itr
== lowest_latency
&&
4503 ((q_vector
->rx
.ring
&& adapter
->rx_itr_setting
== 3) ||
4504 (!q_vector
->rx
.ring
&& adapter
->tx_itr_setting
== 3)))
4505 current_itr
= low_latency
;
4507 switch (current_itr
) {
4508 /* counts and packets in update_itr are dependent on these numbers */
4509 case lowest_latency
:
4510 new_itr
= IGB_70K_ITR
; /* 70,000 ints/sec */
4513 new_itr
= IGB_20K_ITR
; /* 20,000 ints/sec */
4516 new_itr
= IGB_4K_ITR
; /* 4,000 ints/sec */
4523 if (new_itr
!= q_vector
->itr_val
) {
4524 /* this attempts to bias the interrupt rate towards Bulk
4525 * by adding intermediate steps when interrupt rate is
4528 new_itr
= new_itr
> q_vector
->itr_val
?
4529 max((new_itr
* q_vector
->itr_val
) /
4530 (new_itr
+ (q_vector
->itr_val
>> 2)),
4532 /* Don't write the value here; it resets the adapter's
4533 * internal timer, and causes us to delay far longer than
4534 * we should between interrupts. Instead, we write the ITR
4535 * value at the beginning of the next interrupt so the timing
4536 * ends up being correct.
4538 q_vector
->itr_val
= new_itr
;
4539 q_vector
->set_itr
= 1;
4543 static void igb_tx_ctxtdesc(struct igb_ring
*tx_ring
, u32 vlan_macip_lens
,
4544 u32 type_tucmd
, u32 mss_l4len_idx
)
4546 struct e1000_adv_tx_context_desc
*context_desc
;
4547 u16 i
= tx_ring
->next_to_use
;
4549 context_desc
= IGB_TX_CTXTDESC(tx_ring
, i
);
4552 tx_ring
->next_to_use
= (i
< tx_ring
->count
) ? i
: 0;
4554 /* set bits to identify this as an advanced context descriptor */
4555 type_tucmd
|= E1000_TXD_CMD_DEXT
| E1000_ADVTXD_DTYP_CTXT
;
4557 /* For 82575, context index must be unique per ring. */
4558 if (test_bit(IGB_RING_FLAG_TX_CTX_IDX
, &tx_ring
->flags
))
4559 mss_l4len_idx
|= tx_ring
->reg_idx
<< 4;
4561 context_desc
->vlan_macip_lens
= cpu_to_le32(vlan_macip_lens
);
4562 context_desc
->seqnum_seed
= 0;
4563 context_desc
->type_tucmd_mlhl
= cpu_to_le32(type_tucmd
);
4564 context_desc
->mss_l4len_idx
= cpu_to_le32(mss_l4len_idx
);
4567 static int igb_tso(struct igb_ring
*tx_ring
,
4568 struct igb_tx_buffer
*first
,
4571 struct sk_buff
*skb
= first
->skb
;
4572 u32 vlan_macip_lens
, type_tucmd
;
4573 u32 mss_l4len_idx
, l4len
;
4575 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
4578 if (!skb_is_gso(skb
))
4581 if (skb_header_cloned(skb
)) {
4582 int err
= pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
);
4587 /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
4588 type_tucmd
= E1000_ADVTXD_TUCMD_L4T_TCP
;
4590 if (first
->protocol
== __constant_htons(ETH_P_IP
)) {
4591 struct iphdr
*iph
= ip_hdr(skb
);
4594 tcp_hdr(skb
)->check
= ~csum_tcpudp_magic(iph
->saddr
,
4598 type_tucmd
|= E1000_ADVTXD_TUCMD_IPV4
;
4599 first
->tx_flags
|= IGB_TX_FLAGS_TSO
|
4602 } else if (skb_is_gso_v6(skb
)) {
4603 ipv6_hdr(skb
)->payload_len
= 0;
4604 tcp_hdr(skb
)->check
= ~csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
4605 &ipv6_hdr(skb
)->daddr
,
4607 first
->tx_flags
|= IGB_TX_FLAGS_TSO
|
4611 /* compute header lengths */
4612 l4len
= tcp_hdrlen(skb
);
4613 *hdr_len
= skb_transport_offset(skb
) + l4len
;
4615 /* update gso size and bytecount with header size */
4616 first
->gso_segs
= skb_shinfo(skb
)->gso_segs
;
4617 first
->bytecount
+= (first
->gso_segs
- 1) * *hdr_len
;
4620 mss_l4len_idx
= l4len
<< E1000_ADVTXD_L4LEN_SHIFT
;
4621 mss_l4len_idx
|= skb_shinfo(skb
)->gso_size
<< E1000_ADVTXD_MSS_SHIFT
;
4623 /* VLAN MACLEN IPLEN */
4624 vlan_macip_lens
= skb_network_header_len(skb
);
4625 vlan_macip_lens
|= skb_network_offset(skb
) << E1000_ADVTXD_MACLEN_SHIFT
;
4626 vlan_macip_lens
|= first
->tx_flags
& IGB_TX_FLAGS_VLAN_MASK
;
4628 igb_tx_ctxtdesc(tx_ring
, vlan_macip_lens
, type_tucmd
, mss_l4len_idx
);
4633 static void igb_tx_csum(struct igb_ring
*tx_ring
, struct igb_tx_buffer
*first
)
4635 struct sk_buff
*skb
= first
->skb
;
4636 u32 vlan_macip_lens
= 0;
4637 u32 mss_l4len_idx
= 0;
4640 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
4641 if (!(first
->tx_flags
& IGB_TX_FLAGS_VLAN
))
4645 switch (first
->protocol
) {
4646 case __constant_htons(ETH_P_IP
):
4647 vlan_macip_lens
|= skb_network_header_len(skb
);
4648 type_tucmd
|= E1000_ADVTXD_TUCMD_IPV4
;
4649 l4_hdr
= ip_hdr(skb
)->protocol
;
4651 case __constant_htons(ETH_P_IPV6
):
4652 vlan_macip_lens
|= skb_network_header_len(skb
);
4653 l4_hdr
= ipv6_hdr(skb
)->nexthdr
;
4656 if (unlikely(net_ratelimit())) {
4657 dev_warn(tx_ring
->dev
,
4658 "partial checksum but proto=%x!\n",
4666 type_tucmd
|= E1000_ADVTXD_TUCMD_L4T_TCP
;
4667 mss_l4len_idx
= tcp_hdrlen(skb
) <<
4668 E1000_ADVTXD_L4LEN_SHIFT
;
4671 type_tucmd
|= E1000_ADVTXD_TUCMD_L4T_SCTP
;
4672 mss_l4len_idx
= sizeof(struct sctphdr
) <<
4673 E1000_ADVTXD_L4LEN_SHIFT
;
4676 mss_l4len_idx
= sizeof(struct udphdr
) <<
4677 E1000_ADVTXD_L4LEN_SHIFT
;
4680 if (unlikely(net_ratelimit())) {
4681 dev_warn(tx_ring
->dev
,
4682 "partial checksum but l4 proto=%x!\n",
4688 /* update TX checksum flag */
4689 first
->tx_flags
|= IGB_TX_FLAGS_CSUM
;
4692 vlan_macip_lens
|= skb_network_offset(skb
) << E1000_ADVTXD_MACLEN_SHIFT
;
4693 vlan_macip_lens
|= first
->tx_flags
& IGB_TX_FLAGS_VLAN_MASK
;
4695 igb_tx_ctxtdesc(tx_ring
, vlan_macip_lens
, type_tucmd
, mss_l4len_idx
);
4698 #define IGB_SET_FLAG(_input, _flag, _result) \
4699 ((_flag <= _result) ? \
4700 ((u32)(_input & _flag) * (_result / _flag)) : \
4701 ((u32)(_input & _flag) / (_flag / _result)))
4703 static u32
igb_tx_cmd_type(struct sk_buff
*skb
, u32 tx_flags
)
4705 /* set type for advanced descriptor with frame checksum insertion */
4706 u32 cmd_type
= E1000_ADVTXD_DTYP_DATA
|
4707 E1000_ADVTXD_DCMD_DEXT
|
4708 E1000_ADVTXD_DCMD_IFCS
;
4710 /* set HW vlan bit if vlan is present */
4711 cmd_type
|= IGB_SET_FLAG(tx_flags
, IGB_TX_FLAGS_VLAN
,
4712 (E1000_ADVTXD_DCMD_VLE
));
4714 /* set segmentation bits for TSO */
4715 cmd_type
|= IGB_SET_FLAG(tx_flags
, IGB_TX_FLAGS_TSO
,
4716 (E1000_ADVTXD_DCMD_TSE
));
4718 /* set timestamp bit if present */
4719 cmd_type
|= IGB_SET_FLAG(tx_flags
, IGB_TX_FLAGS_TSTAMP
,
4720 (E1000_ADVTXD_MAC_TSTAMP
));
4722 /* insert frame checksum */
4723 cmd_type
^= IGB_SET_FLAG(skb
->no_fcs
, 1, E1000_ADVTXD_DCMD_IFCS
);
4728 static void igb_tx_olinfo_status(struct igb_ring
*tx_ring
,
4729 union e1000_adv_tx_desc
*tx_desc
,
4730 u32 tx_flags
, unsigned int paylen
)
4732 u32 olinfo_status
= paylen
<< E1000_ADVTXD_PAYLEN_SHIFT
;
4734 /* 82575 requires a unique index per ring */
4735 if (test_bit(IGB_RING_FLAG_TX_CTX_IDX
, &tx_ring
->flags
))
4736 olinfo_status
|= tx_ring
->reg_idx
<< 4;
4738 /* insert L4 checksum */
4739 olinfo_status
|= IGB_SET_FLAG(tx_flags
,
4741 (E1000_TXD_POPTS_TXSM
<< 8));
4743 /* insert IPv4 checksum */
4744 olinfo_status
|= IGB_SET_FLAG(tx_flags
,
4746 (E1000_TXD_POPTS_IXSM
<< 8));
4748 tx_desc
->read
.olinfo_status
= cpu_to_le32(olinfo_status
);
4751 static void igb_tx_map(struct igb_ring
*tx_ring
,
4752 struct igb_tx_buffer
*first
,
4755 struct sk_buff
*skb
= first
->skb
;
4756 struct igb_tx_buffer
*tx_buffer
;
4757 union e1000_adv_tx_desc
*tx_desc
;
4758 struct skb_frag_struct
*frag
;
4760 unsigned int data_len
, size
;
4761 u32 tx_flags
= first
->tx_flags
;
4762 u32 cmd_type
= igb_tx_cmd_type(skb
, tx_flags
);
4763 u16 i
= tx_ring
->next_to_use
;
4765 tx_desc
= IGB_TX_DESC(tx_ring
, i
);
4767 igb_tx_olinfo_status(tx_ring
, tx_desc
, tx_flags
, skb
->len
- hdr_len
);
4769 size
= skb_headlen(skb
);
4770 data_len
= skb
->data_len
;
4772 dma
= dma_map_single(tx_ring
->dev
, skb
->data
, size
, DMA_TO_DEVICE
);
4776 for (frag
= &skb_shinfo(skb
)->frags
[0];; frag
++) {
4777 if (dma_mapping_error(tx_ring
->dev
, dma
))
4780 /* record length, and DMA address */
4781 dma_unmap_len_set(tx_buffer
, len
, size
);
4782 dma_unmap_addr_set(tx_buffer
, dma
, dma
);
4784 tx_desc
->read
.buffer_addr
= cpu_to_le64(dma
);
4786 while (unlikely(size
> IGB_MAX_DATA_PER_TXD
)) {
4787 tx_desc
->read
.cmd_type_len
=
4788 cpu_to_le32(cmd_type
^ IGB_MAX_DATA_PER_TXD
);
4792 if (i
== tx_ring
->count
) {
4793 tx_desc
= IGB_TX_DESC(tx_ring
, 0);
4796 tx_desc
->read
.olinfo_status
= 0;
4798 dma
+= IGB_MAX_DATA_PER_TXD
;
4799 size
-= IGB_MAX_DATA_PER_TXD
;
4801 tx_desc
->read
.buffer_addr
= cpu_to_le64(dma
);
4804 if (likely(!data_len
))
4807 tx_desc
->read
.cmd_type_len
= cpu_to_le32(cmd_type
^ size
);
4811 if (i
== tx_ring
->count
) {
4812 tx_desc
= IGB_TX_DESC(tx_ring
, 0);
4815 tx_desc
->read
.olinfo_status
= 0;
4817 size
= skb_frag_size(frag
);
4820 dma
= skb_frag_dma_map(tx_ring
->dev
, frag
, 0,
4821 size
, DMA_TO_DEVICE
);
4823 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
4826 /* write last descriptor with RS and EOP bits */
4827 cmd_type
|= size
| IGB_TXD_DCMD
;
4828 tx_desc
->read
.cmd_type_len
= cpu_to_le32(cmd_type
);
4830 netdev_tx_sent_queue(txring_txq(tx_ring
), first
->bytecount
);
4832 /* set the timestamp */
4833 first
->time_stamp
= jiffies
;
4835 /* Force memory writes to complete before letting h/w know there
4836 * are new descriptors to fetch. (Only applicable for weak-ordered
4837 * memory model archs, such as IA-64).
4839 * We also need this memory barrier to make certain all of the
4840 * status bits have been updated before next_to_watch is written.
4844 /* set next_to_watch value indicating a packet is present */
4845 first
->next_to_watch
= tx_desc
;
4848 if (i
== tx_ring
->count
)
4851 tx_ring
->next_to_use
= i
;
4853 writel(i
, tx_ring
->tail
);
4855 /* we need this if more than one processor can write to our tail
4856 * at a time, it synchronizes IO on IA64/Altix systems
4863 dev_err(tx_ring
->dev
, "TX DMA map failed\n");
4865 /* clear dma mappings for failed tx_buffer_info map */
4867 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
4868 igb_unmap_and_free_tx_resource(tx_ring
, tx_buffer
);
4869 if (tx_buffer
== first
)
4876 tx_ring
->next_to_use
= i
;
4879 static int __igb_maybe_stop_tx(struct igb_ring
*tx_ring
, const u16 size
)
4881 struct net_device
*netdev
= tx_ring
->netdev
;
4883 netif_stop_subqueue(netdev
, tx_ring
->queue_index
);
4885 /* Herbert's original patch had:
4886 * smp_mb__after_netif_stop_queue();
4887 * but since that doesn't exist yet, just open code it.
4891 /* We need to check again in a case another CPU has just
4892 * made room available.
4894 if (igb_desc_unused(tx_ring
) < size
)
4898 netif_wake_subqueue(netdev
, tx_ring
->queue_index
);
4900 u64_stats_update_begin(&tx_ring
->tx_syncp2
);
4901 tx_ring
->tx_stats
.restart_queue2
++;
4902 u64_stats_update_end(&tx_ring
->tx_syncp2
);
4907 static inline int igb_maybe_stop_tx(struct igb_ring
*tx_ring
, const u16 size
)
4909 if (igb_desc_unused(tx_ring
) >= size
)
4911 return __igb_maybe_stop_tx(tx_ring
, size
);
4914 netdev_tx_t
igb_xmit_frame_ring(struct sk_buff
*skb
,
4915 struct igb_ring
*tx_ring
)
4917 struct igb_tx_buffer
*first
;
4920 u16 count
= TXD_USE_COUNT(skb_headlen(skb
));
4921 __be16 protocol
= vlan_get_protocol(skb
);
4924 /* need: 1 descriptor per page * PAGE_SIZE/IGB_MAX_DATA_PER_TXD,
4925 * + 1 desc for skb_headlen/IGB_MAX_DATA_PER_TXD,
4926 * + 2 desc gap to keep tail from touching head,
4927 * + 1 desc for context descriptor,
4928 * otherwise try next time
4930 if (NETDEV_FRAG_PAGE_MAX_SIZE
> IGB_MAX_DATA_PER_TXD
) {
4932 for (f
= 0; f
< skb_shinfo(skb
)->nr_frags
; f
++)
4933 count
+= TXD_USE_COUNT(skb_shinfo(skb
)->frags
[f
].size
);
4935 count
+= skb_shinfo(skb
)->nr_frags
;
4938 if (igb_maybe_stop_tx(tx_ring
, count
+ 3)) {
4939 /* this is a hard error */
4940 return NETDEV_TX_BUSY
;
4943 /* record the location of the first descriptor for this packet */
4944 first
= &tx_ring
->tx_buffer_info
[tx_ring
->next_to_use
];
4946 first
->bytecount
= skb
->len
;
4947 first
->gso_segs
= 1;
4949 skb_tx_timestamp(skb
);
4951 if (unlikely(skb_shinfo(skb
)->tx_flags
& SKBTX_HW_TSTAMP
)) {
4952 struct igb_adapter
*adapter
= netdev_priv(tx_ring
->netdev
);
4954 if (!(adapter
->ptp_tx_skb
)) {
4955 skb_shinfo(skb
)->tx_flags
|= SKBTX_IN_PROGRESS
;
4956 tx_flags
|= IGB_TX_FLAGS_TSTAMP
;
4958 adapter
->ptp_tx_skb
= skb_get(skb
);
4959 adapter
->ptp_tx_start
= jiffies
;
4960 if (adapter
->hw
.mac
.type
== e1000_82576
)
4961 schedule_work(&adapter
->ptp_tx_work
);
4965 if (vlan_tx_tag_present(skb
)) {
4966 tx_flags
|= IGB_TX_FLAGS_VLAN
;
4967 tx_flags
|= (vlan_tx_tag_get(skb
) << IGB_TX_FLAGS_VLAN_SHIFT
);
4970 /* record initial flags and protocol */
4971 first
->tx_flags
= tx_flags
;
4972 first
->protocol
= protocol
;
4974 tso
= igb_tso(tx_ring
, first
, &hdr_len
);
4978 igb_tx_csum(tx_ring
, first
);
4980 igb_tx_map(tx_ring
, first
, hdr_len
);
4982 /* Make sure there is space in the ring for the next send. */
4983 igb_maybe_stop_tx(tx_ring
, DESC_NEEDED
);
4985 return NETDEV_TX_OK
;
4988 igb_unmap_and_free_tx_resource(tx_ring
, first
);
4990 return NETDEV_TX_OK
;
4993 static inline struct igb_ring
*igb_tx_queue_mapping(struct igb_adapter
*adapter
,
4994 struct sk_buff
*skb
)
4996 unsigned int r_idx
= skb
->queue_mapping
;
4998 if (r_idx
>= adapter
->num_tx_queues
)
4999 r_idx
= r_idx
% adapter
->num_tx_queues
;
5001 return adapter
->tx_ring
[r_idx
];
5004 static netdev_tx_t
igb_xmit_frame(struct sk_buff
*skb
,
5005 struct net_device
*netdev
)
5007 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5009 if (test_bit(__IGB_DOWN
, &adapter
->state
)) {
5010 dev_kfree_skb_any(skb
);
5011 return NETDEV_TX_OK
;
5014 if (skb
->len
<= 0) {
5015 dev_kfree_skb_any(skb
);
5016 return NETDEV_TX_OK
;
5019 /* The minimum packet size with TCTL.PSP set is 17 so pad the skb
5020 * in order to meet this minimum size requirement.
5022 if (unlikely(skb
->len
< 17)) {
5023 if (skb_pad(skb
, 17 - skb
->len
))
5024 return NETDEV_TX_OK
;
5026 skb_set_tail_pointer(skb
, 17);
5029 return igb_xmit_frame_ring(skb
, igb_tx_queue_mapping(adapter
, skb
));
5033 * igb_tx_timeout - Respond to a Tx Hang
5034 * @netdev: network interface device structure
5036 static void igb_tx_timeout(struct net_device
*netdev
)
5038 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5039 struct e1000_hw
*hw
= &adapter
->hw
;
5041 /* Do the reset outside of interrupt context */
5042 adapter
->tx_timeout_count
++;
5044 if (hw
->mac
.type
>= e1000_82580
)
5045 hw
->dev_spec
._82575
.global_device_reset
= true;
5047 schedule_work(&adapter
->reset_task
);
5049 (adapter
->eims_enable_mask
& ~adapter
->eims_other
));
5052 static void igb_reset_task(struct work_struct
*work
)
5054 struct igb_adapter
*adapter
;
5055 adapter
= container_of(work
, struct igb_adapter
, reset_task
);
5058 netdev_err(adapter
->netdev
, "Reset adapter\n");
5059 igb_reinit_locked(adapter
);
5063 * igb_get_stats64 - Get System Network Statistics
5064 * @netdev: network interface device structure
5065 * @stats: rtnl_link_stats64 pointer
5067 static struct rtnl_link_stats64
*igb_get_stats64(struct net_device
*netdev
,
5068 struct rtnl_link_stats64
*stats
)
5070 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5072 spin_lock(&adapter
->stats64_lock
);
5073 igb_update_stats(adapter
, &adapter
->stats64
);
5074 memcpy(stats
, &adapter
->stats64
, sizeof(*stats
));
5075 spin_unlock(&adapter
->stats64_lock
);
5081 * igb_change_mtu - Change the Maximum Transfer Unit
5082 * @netdev: network interface device structure
5083 * @new_mtu: new value for maximum frame size
5085 * Returns 0 on success, negative on failure
5087 static int igb_change_mtu(struct net_device
*netdev
, int new_mtu
)
5089 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5090 struct pci_dev
*pdev
= adapter
->pdev
;
5091 int max_frame
= new_mtu
+ ETH_HLEN
+ ETH_FCS_LEN
+ VLAN_HLEN
;
5093 if ((new_mtu
< 68) || (max_frame
> MAX_JUMBO_FRAME_SIZE
)) {
5094 dev_err(&pdev
->dev
, "Invalid MTU setting\n");
5098 #define MAX_STD_JUMBO_FRAME_SIZE 9238
5099 if (max_frame
> MAX_STD_JUMBO_FRAME_SIZE
) {
5100 dev_err(&pdev
->dev
, "MTU > 9216 not supported.\n");
5104 /* adjust max frame to be at least the size of a standard frame */
5105 if (max_frame
< (ETH_FRAME_LEN
+ ETH_FCS_LEN
))
5106 max_frame
= ETH_FRAME_LEN
+ ETH_FCS_LEN
;
5108 while (test_and_set_bit(__IGB_RESETTING
, &adapter
->state
))
5111 /* igb_down has a dependency on max_frame_size */
5112 adapter
->max_frame_size
= max_frame
;
5114 if (netif_running(netdev
))
5117 dev_info(&pdev
->dev
, "changing MTU from %d to %d\n",
5118 netdev
->mtu
, new_mtu
);
5119 netdev
->mtu
= new_mtu
;
5121 if (netif_running(netdev
))
5126 clear_bit(__IGB_RESETTING
, &adapter
->state
);
5132 * igb_update_stats - Update the board statistics counters
5133 * @adapter: board private structure
5135 void igb_update_stats(struct igb_adapter
*adapter
,
5136 struct rtnl_link_stats64
*net_stats
)
5138 struct e1000_hw
*hw
= &adapter
->hw
;
5139 struct pci_dev
*pdev
= adapter
->pdev
;
5145 u64 _bytes
, _packets
;
5147 #define PHY_IDLE_ERROR_COUNT_MASK 0x00FF
5149 /* Prevent stats update while adapter is being reset, or if the pci
5150 * connection is down.
5152 if (adapter
->link_speed
== 0)
5154 if (pci_channel_offline(pdev
))
5161 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
5162 u32 rqdpc
= rd32(E1000_RQDPC(i
));
5163 struct igb_ring
*ring
= adapter
->rx_ring
[i
];
5166 ring
->rx_stats
.drops
+= rqdpc
;
5167 net_stats
->rx_fifo_errors
+= rqdpc
;
5171 start
= u64_stats_fetch_begin_bh(&ring
->rx_syncp
);
5172 _bytes
= ring
->rx_stats
.bytes
;
5173 _packets
= ring
->rx_stats
.packets
;
5174 } while (u64_stats_fetch_retry_bh(&ring
->rx_syncp
, start
));
5176 packets
+= _packets
;
5179 net_stats
->rx_bytes
= bytes
;
5180 net_stats
->rx_packets
= packets
;
5184 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
5185 struct igb_ring
*ring
= adapter
->tx_ring
[i
];
5187 start
= u64_stats_fetch_begin_bh(&ring
->tx_syncp
);
5188 _bytes
= ring
->tx_stats
.bytes
;
5189 _packets
= ring
->tx_stats
.packets
;
5190 } while (u64_stats_fetch_retry_bh(&ring
->tx_syncp
, start
));
5192 packets
+= _packets
;
5194 net_stats
->tx_bytes
= bytes
;
5195 net_stats
->tx_packets
= packets
;
5198 /* read stats registers */
5199 adapter
->stats
.crcerrs
+= rd32(E1000_CRCERRS
);
5200 adapter
->stats
.gprc
+= rd32(E1000_GPRC
);
5201 adapter
->stats
.gorc
+= rd32(E1000_GORCL
);
5202 rd32(E1000_GORCH
); /* clear GORCL */
5203 adapter
->stats
.bprc
+= rd32(E1000_BPRC
);
5204 adapter
->stats
.mprc
+= rd32(E1000_MPRC
);
5205 adapter
->stats
.roc
+= rd32(E1000_ROC
);
5207 adapter
->stats
.prc64
+= rd32(E1000_PRC64
);
5208 adapter
->stats
.prc127
+= rd32(E1000_PRC127
);
5209 adapter
->stats
.prc255
+= rd32(E1000_PRC255
);
5210 adapter
->stats
.prc511
+= rd32(E1000_PRC511
);
5211 adapter
->stats
.prc1023
+= rd32(E1000_PRC1023
);
5212 adapter
->stats
.prc1522
+= rd32(E1000_PRC1522
);
5213 adapter
->stats
.symerrs
+= rd32(E1000_SYMERRS
);
5214 adapter
->stats
.sec
+= rd32(E1000_SEC
);
5216 mpc
= rd32(E1000_MPC
);
5217 adapter
->stats
.mpc
+= mpc
;
5218 net_stats
->rx_fifo_errors
+= mpc
;
5219 adapter
->stats
.scc
+= rd32(E1000_SCC
);
5220 adapter
->stats
.ecol
+= rd32(E1000_ECOL
);
5221 adapter
->stats
.mcc
+= rd32(E1000_MCC
);
5222 adapter
->stats
.latecol
+= rd32(E1000_LATECOL
);
5223 adapter
->stats
.dc
+= rd32(E1000_DC
);
5224 adapter
->stats
.rlec
+= rd32(E1000_RLEC
);
5225 adapter
->stats
.xonrxc
+= rd32(E1000_XONRXC
);
5226 adapter
->stats
.xontxc
+= rd32(E1000_XONTXC
);
5227 adapter
->stats
.xoffrxc
+= rd32(E1000_XOFFRXC
);
5228 adapter
->stats
.xofftxc
+= rd32(E1000_XOFFTXC
);
5229 adapter
->stats
.fcruc
+= rd32(E1000_FCRUC
);
5230 adapter
->stats
.gptc
+= rd32(E1000_GPTC
);
5231 adapter
->stats
.gotc
+= rd32(E1000_GOTCL
);
5232 rd32(E1000_GOTCH
); /* clear GOTCL */
5233 adapter
->stats
.rnbc
+= rd32(E1000_RNBC
);
5234 adapter
->stats
.ruc
+= rd32(E1000_RUC
);
5235 adapter
->stats
.rfc
+= rd32(E1000_RFC
);
5236 adapter
->stats
.rjc
+= rd32(E1000_RJC
);
5237 adapter
->stats
.tor
+= rd32(E1000_TORH
);
5238 adapter
->stats
.tot
+= rd32(E1000_TOTH
);
5239 adapter
->stats
.tpr
+= rd32(E1000_TPR
);
5241 adapter
->stats
.ptc64
+= rd32(E1000_PTC64
);
5242 adapter
->stats
.ptc127
+= rd32(E1000_PTC127
);
5243 adapter
->stats
.ptc255
+= rd32(E1000_PTC255
);
5244 adapter
->stats
.ptc511
+= rd32(E1000_PTC511
);
5245 adapter
->stats
.ptc1023
+= rd32(E1000_PTC1023
);
5246 adapter
->stats
.ptc1522
+= rd32(E1000_PTC1522
);
5248 adapter
->stats
.mptc
+= rd32(E1000_MPTC
);
5249 adapter
->stats
.bptc
+= rd32(E1000_BPTC
);
5251 adapter
->stats
.tpt
+= rd32(E1000_TPT
);
5252 adapter
->stats
.colc
+= rd32(E1000_COLC
);
5254 adapter
->stats
.algnerrc
+= rd32(E1000_ALGNERRC
);
5255 /* read internal phy specific stats */
5256 reg
= rd32(E1000_CTRL_EXT
);
5257 if (!(reg
& E1000_CTRL_EXT_LINK_MODE_MASK
)) {
5258 adapter
->stats
.rxerrc
+= rd32(E1000_RXERRC
);
5260 /* this stat has invalid values on i210/i211 */
5261 if ((hw
->mac
.type
!= e1000_i210
) &&
5262 (hw
->mac
.type
!= e1000_i211
))
5263 adapter
->stats
.tncrs
+= rd32(E1000_TNCRS
);
5266 adapter
->stats
.tsctc
+= rd32(E1000_TSCTC
);
5267 adapter
->stats
.tsctfc
+= rd32(E1000_TSCTFC
);
5269 adapter
->stats
.iac
+= rd32(E1000_IAC
);
5270 adapter
->stats
.icrxoc
+= rd32(E1000_ICRXOC
);
5271 adapter
->stats
.icrxptc
+= rd32(E1000_ICRXPTC
);
5272 adapter
->stats
.icrxatc
+= rd32(E1000_ICRXATC
);
5273 adapter
->stats
.ictxptc
+= rd32(E1000_ICTXPTC
);
5274 adapter
->stats
.ictxatc
+= rd32(E1000_ICTXATC
);
5275 adapter
->stats
.ictxqec
+= rd32(E1000_ICTXQEC
);
5276 adapter
->stats
.ictxqmtc
+= rd32(E1000_ICTXQMTC
);
5277 adapter
->stats
.icrxdmtc
+= rd32(E1000_ICRXDMTC
);
5279 /* Fill out the OS statistics structure */
5280 net_stats
->multicast
= adapter
->stats
.mprc
;
5281 net_stats
->collisions
= adapter
->stats
.colc
;
5285 /* RLEC on some newer hardware can be incorrect so build
5286 * our own version based on RUC and ROC
5288 net_stats
->rx_errors
= adapter
->stats
.rxerrc
+
5289 adapter
->stats
.crcerrs
+ adapter
->stats
.algnerrc
+
5290 adapter
->stats
.ruc
+ adapter
->stats
.roc
+
5291 adapter
->stats
.cexterr
;
5292 net_stats
->rx_length_errors
= adapter
->stats
.ruc
+
5294 net_stats
->rx_crc_errors
= adapter
->stats
.crcerrs
;
5295 net_stats
->rx_frame_errors
= adapter
->stats
.algnerrc
;
5296 net_stats
->rx_missed_errors
= adapter
->stats
.mpc
;
5299 net_stats
->tx_errors
= adapter
->stats
.ecol
+
5300 adapter
->stats
.latecol
;
5301 net_stats
->tx_aborted_errors
= adapter
->stats
.ecol
;
5302 net_stats
->tx_window_errors
= adapter
->stats
.latecol
;
5303 net_stats
->tx_carrier_errors
= adapter
->stats
.tncrs
;
5305 /* Tx Dropped needs to be maintained elsewhere */
5308 if (hw
->phy
.media_type
== e1000_media_type_copper
) {
5309 if ((adapter
->link_speed
== SPEED_1000
) &&
5310 (!igb_read_phy_reg(hw
, PHY_1000T_STATUS
, &phy_tmp
))) {
5311 phy_tmp
&= PHY_IDLE_ERROR_COUNT_MASK
;
5312 adapter
->phy_stats
.idle_errors
+= phy_tmp
;
5316 /* Management Stats */
5317 adapter
->stats
.mgptc
+= rd32(E1000_MGTPTC
);
5318 adapter
->stats
.mgprc
+= rd32(E1000_MGTPRC
);
5319 adapter
->stats
.mgpdc
+= rd32(E1000_MGTPDC
);
5322 reg
= rd32(E1000_MANC
);
5323 if (reg
& E1000_MANC_EN_BMC2OS
) {
5324 adapter
->stats
.o2bgptc
+= rd32(E1000_O2BGPTC
);
5325 adapter
->stats
.o2bspc
+= rd32(E1000_O2BSPC
);
5326 adapter
->stats
.b2ospc
+= rd32(E1000_B2OSPC
);
5327 adapter
->stats
.b2ogprc
+= rd32(E1000_B2OGPRC
);
5331 static irqreturn_t
igb_msix_other(int irq
, void *data
)
5333 struct igb_adapter
*adapter
= data
;
5334 struct e1000_hw
*hw
= &adapter
->hw
;
5335 u32 icr
= rd32(E1000_ICR
);
5336 /* reading ICR causes bit 31 of EICR to be cleared */
5338 if (icr
& E1000_ICR_DRSTA
)
5339 schedule_work(&adapter
->reset_task
);
5341 if (icr
& E1000_ICR_DOUTSYNC
) {
5342 /* HW is reporting DMA is out of sync */
5343 adapter
->stats
.doosync
++;
5344 /* The DMA Out of Sync is also indication of a spoof event
5345 * in IOV mode. Check the Wrong VM Behavior register to
5346 * see if it is really a spoof event.
5348 igb_check_wvbr(adapter
);
5351 /* Check for a mailbox event */
5352 if (icr
& E1000_ICR_VMMB
)
5353 igb_msg_task(adapter
);
5355 if (icr
& E1000_ICR_LSC
) {
5356 hw
->mac
.get_link_status
= 1;
5357 /* guard against interrupt when we're going down */
5358 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
5359 mod_timer(&adapter
->watchdog_timer
, jiffies
+ 1);
5362 if (icr
& E1000_ICR_TS
) {
5363 u32 tsicr
= rd32(E1000_TSICR
);
5365 if (tsicr
& E1000_TSICR_TXTS
) {
5366 /* acknowledge the interrupt */
5367 wr32(E1000_TSICR
, E1000_TSICR_TXTS
);
5368 /* retrieve hardware timestamp */
5369 schedule_work(&adapter
->ptp_tx_work
);
5373 wr32(E1000_EIMS
, adapter
->eims_other
);
5378 static void igb_write_itr(struct igb_q_vector
*q_vector
)
5380 struct igb_adapter
*adapter
= q_vector
->adapter
;
5381 u32 itr_val
= q_vector
->itr_val
& 0x7FFC;
5383 if (!q_vector
->set_itr
)
5389 if (adapter
->hw
.mac
.type
== e1000_82575
)
5390 itr_val
|= itr_val
<< 16;
5392 itr_val
|= E1000_EITR_CNT_IGNR
;
5394 writel(itr_val
, q_vector
->itr_register
);
5395 q_vector
->set_itr
= 0;
5398 static irqreturn_t
igb_msix_ring(int irq
, void *data
)
5400 struct igb_q_vector
*q_vector
= data
;
5402 /* Write the ITR value calculated from the previous interrupt. */
5403 igb_write_itr(q_vector
);
5405 napi_schedule(&q_vector
->napi
);
5410 #ifdef CONFIG_IGB_DCA
5411 static void igb_update_tx_dca(struct igb_adapter
*adapter
,
5412 struct igb_ring
*tx_ring
,
5415 struct e1000_hw
*hw
= &adapter
->hw
;
5416 u32 txctrl
= dca3_get_tag(tx_ring
->dev
, cpu
);
5418 if (hw
->mac
.type
!= e1000_82575
)
5419 txctrl
<<= E1000_DCA_TXCTRL_CPUID_SHIFT
;
5421 /* We can enable relaxed ordering for reads, but not writes when
5422 * DCA is enabled. This is due to a known issue in some chipsets
5423 * which will cause the DCA tag to be cleared.
5425 txctrl
|= E1000_DCA_TXCTRL_DESC_RRO_EN
|
5426 E1000_DCA_TXCTRL_DATA_RRO_EN
|
5427 E1000_DCA_TXCTRL_DESC_DCA_EN
;
5429 wr32(E1000_DCA_TXCTRL(tx_ring
->reg_idx
), txctrl
);
5432 static void igb_update_rx_dca(struct igb_adapter
*adapter
,
5433 struct igb_ring
*rx_ring
,
5436 struct e1000_hw
*hw
= &adapter
->hw
;
5437 u32 rxctrl
= dca3_get_tag(&adapter
->pdev
->dev
, cpu
);
5439 if (hw
->mac
.type
!= e1000_82575
)
5440 rxctrl
<<= E1000_DCA_RXCTRL_CPUID_SHIFT
;
5442 /* We can enable relaxed ordering for reads, but not writes when
5443 * DCA is enabled. This is due to a known issue in some chipsets
5444 * which will cause the DCA tag to be cleared.
5446 rxctrl
|= E1000_DCA_RXCTRL_DESC_RRO_EN
|
5447 E1000_DCA_RXCTRL_DESC_DCA_EN
;
5449 wr32(E1000_DCA_RXCTRL(rx_ring
->reg_idx
), rxctrl
);
5452 static void igb_update_dca(struct igb_q_vector
*q_vector
)
5454 struct igb_adapter
*adapter
= q_vector
->adapter
;
5455 int cpu
= get_cpu();
5457 if (q_vector
->cpu
== cpu
)
5460 if (q_vector
->tx
.ring
)
5461 igb_update_tx_dca(adapter
, q_vector
->tx
.ring
, cpu
);
5463 if (q_vector
->rx
.ring
)
5464 igb_update_rx_dca(adapter
, q_vector
->rx
.ring
, cpu
);
5466 q_vector
->cpu
= cpu
;
5471 static void igb_setup_dca(struct igb_adapter
*adapter
)
5473 struct e1000_hw
*hw
= &adapter
->hw
;
5476 if (!(adapter
->flags
& IGB_FLAG_DCA_ENABLED
))
5479 /* Always use CB2 mode, difference is masked in the CB driver. */
5480 wr32(E1000_DCA_CTRL
, E1000_DCA_CTRL_DCA_MODE_CB2
);
5482 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
5483 adapter
->q_vector
[i
]->cpu
= -1;
5484 igb_update_dca(adapter
->q_vector
[i
]);
5488 static int __igb_notify_dca(struct device
*dev
, void *data
)
5490 struct net_device
*netdev
= dev_get_drvdata(dev
);
5491 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5492 struct pci_dev
*pdev
= adapter
->pdev
;
5493 struct e1000_hw
*hw
= &adapter
->hw
;
5494 unsigned long event
= *(unsigned long *)data
;
5497 case DCA_PROVIDER_ADD
:
5498 /* if already enabled, don't do it again */
5499 if (adapter
->flags
& IGB_FLAG_DCA_ENABLED
)
5501 if (dca_add_requester(dev
) == 0) {
5502 adapter
->flags
|= IGB_FLAG_DCA_ENABLED
;
5503 dev_info(&pdev
->dev
, "DCA enabled\n");
5504 igb_setup_dca(adapter
);
5507 /* Fall Through since DCA is disabled. */
5508 case DCA_PROVIDER_REMOVE
:
5509 if (adapter
->flags
& IGB_FLAG_DCA_ENABLED
) {
5510 /* without this a class_device is left
5511 * hanging around in the sysfs model
5513 dca_remove_requester(dev
);
5514 dev_info(&pdev
->dev
, "DCA disabled\n");
5515 adapter
->flags
&= ~IGB_FLAG_DCA_ENABLED
;
5516 wr32(E1000_DCA_CTRL
, E1000_DCA_CTRL_DCA_MODE_DISABLE
);
5524 static int igb_notify_dca(struct notifier_block
*nb
, unsigned long event
,
5529 ret_val
= driver_for_each_device(&igb_driver
.driver
, NULL
, &event
,
5532 return ret_val
? NOTIFY_BAD
: NOTIFY_DONE
;
5534 #endif /* CONFIG_IGB_DCA */
5536 #ifdef CONFIG_PCI_IOV
5537 static int igb_vf_configure(struct igb_adapter
*adapter
, int vf
)
5539 unsigned char mac_addr
[ETH_ALEN
];
5541 eth_zero_addr(mac_addr
);
5542 igb_set_vf_mac(adapter
, vf
, mac_addr
);
5544 /* By default spoof check is enabled for all VFs */
5545 adapter
->vf_data
[vf
].spoofchk_enabled
= true;
5551 static void igb_ping_all_vfs(struct igb_adapter
*adapter
)
5553 struct e1000_hw
*hw
= &adapter
->hw
;
5557 for (i
= 0 ; i
< adapter
->vfs_allocated_count
; i
++) {
5558 ping
= E1000_PF_CONTROL_MSG
;
5559 if (adapter
->vf_data
[i
].flags
& IGB_VF_FLAG_CTS
)
5560 ping
|= E1000_VT_MSGTYPE_CTS
;
5561 igb_write_mbx(hw
, &ping
, 1, i
);
5565 static int igb_set_vf_promisc(struct igb_adapter
*adapter
, u32
*msgbuf
, u32 vf
)
5567 struct e1000_hw
*hw
= &adapter
->hw
;
5568 u32 vmolr
= rd32(E1000_VMOLR(vf
));
5569 struct vf_data_storage
*vf_data
= &adapter
->vf_data
[vf
];
5571 vf_data
->flags
&= ~(IGB_VF_FLAG_UNI_PROMISC
|
5572 IGB_VF_FLAG_MULTI_PROMISC
);
5573 vmolr
&= ~(E1000_VMOLR_ROPE
| E1000_VMOLR_ROMPE
| E1000_VMOLR_MPME
);
5575 if (*msgbuf
& E1000_VF_SET_PROMISC_MULTICAST
) {
5576 vmolr
|= E1000_VMOLR_MPME
;
5577 vf_data
->flags
|= IGB_VF_FLAG_MULTI_PROMISC
;
5578 *msgbuf
&= ~E1000_VF_SET_PROMISC_MULTICAST
;
5580 /* if we have hashes and we are clearing a multicast promisc
5581 * flag we need to write the hashes to the MTA as this step
5582 * was previously skipped
5584 if (vf_data
->num_vf_mc_hashes
> 30) {
5585 vmolr
|= E1000_VMOLR_MPME
;
5586 } else if (vf_data
->num_vf_mc_hashes
) {
5588 vmolr
|= E1000_VMOLR_ROMPE
;
5589 for (j
= 0; j
< vf_data
->num_vf_mc_hashes
; j
++)
5590 igb_mta_set(hw
, vf_data
->vf_mc_hashes
[j
]);
5594 wr32(E1000_VMOLR(vf
), vmolr
);
5596 /* there are flags left unprocessed, likely not supported */
5597 if (*msgbuf
& E1000_VT_MSGINFO_MASK
)
5603 static int igb_set_vf_multicasts(struct igb_adapter
*adapter
,
5604 u32
*msgbuf
, u32 vf
)
5606 int n
= (msgbuf
[0] & E1000_VT_MSGINFO_MASK
) >> E1000_VT_MSGINFO_SHIFT
;
5607 u16
*hash_list
= (u16
*)&msgbuf
[1];
5608 struct vf_data_storage
*vf_data
= &adapter
->vf_data
[vf
];
5611 /* salt away the number of multicast addresses assigned
5612 * to this VF for later use to restore when the PF multi cast
5615 vf_data
->num_vf_mc_hashes
= n
;
5617 /* only up to 30 hash values supported */
5621 /* store the hashes for later use */
5622 for (i
= 0; i
< n
; i
++)
5623 vf_data
->vf_mc_hashes
[i
] = hash_list
[i
];
5625 /* Flush and reset the mta with the new values */
5626 igb_set_rx_mode(adapter
->netdev
);
5631 static void igb_restore_vf_multicasts(struct igb_adapter
*adapter
)
5633 struct e1000_hw
*hw
= &adapter
->hw
;
5634 struct vf_data_storage
*vf_data
;
5637 for (i
= 0; i
< adapter
->vfs_allocated_count
; i
++) {
5638 u32 vmolr
= rd32(E1000_VMOLR(i
));
5639 vmolr
&= ~(E1000_VMOLR_ROMPE
| E1000_VMOLR_MPME
);
5641 vf_data
= &adapter
->vf_data
[i
];
5643 if ((vf_data
->num_vf_mc_hashes
> 30) ||
5644 (vf_data
->flags
& IGB_VF_FLAG_MULTI_PROMISC
)) {
5645 vmolr
|= E1000_VMOLR_MPME
;
5646 } else if (vf_data
->num_vf_mc_hashes
) {
5647 vmolr
|= E1000_VMOLR_ROMPE
;
5648 for (j
= 0; j
< vf_data
->num_vf_mc_hashes
; j
++)
5649 igb_mta_set(hw
, vf_data
->vf_mc_hashes
[j
]);
5651 wr32(E1000_VMOLR(i
), vmolr
);
5655 static void igb_clear_vf_vfta(struct igb_adapter
*adapter
, u32 vf
)
5657 struct e1000_hw
*hw
= &adapter
->hw
;
5658 u32 pool_mask
, reg
, vid
;
5661 pool_mask
= 1 << (E1000_VLVF_POOLSEL_SHIFT
+ vf
);
5663 /* Find the vlan filter for this id */
5664 for (i
= 0; i
< E1000_VLVF_ARRAY_SIZE
; i
++) {
5665 reg
= rd32(E1000_VLVF(i
));
5667 /* remove the vf from the pool */
5670 /* if pool is empty then remove entry from vfta */
5671 if (!(reg
& E1000_VLVF_POOLSEL_MASK
) &&
5672 (reg
& E1000_VLVF_VLANID_ENABLE
)) {
5674 vid
= reg
& E1000_VLVF_VLANID_MASK
;
5675 igb_vfta_set(hw
, vid
, false);
5678 wr32(E1000_VLVF(i
), reg
);
5681 adapter
->vf_data
[vf
].vlans_enabled
= 0;
5684 static s32
igb_vlvf_set(struct igb_adapter
*adapter
, u32 vid
, bool add
, u32 vf
)
5686 struct e1000_hw
*hw
= &adapter
->hw
;
5689 /* The vlvf table only exists on 82576 hardware and newer */
5690 if (hw
->mac
.type
< e1000_82576
)
5693 /* we only need to do this if VMDq is enabled */
5694 if (!adapter
->vfs_allocated_count
)
5697 /* Find the vlan filter for this id */
5698 for (i
= 0; i
< E1000_VLVF_ARRAY_SIZE
; i
++) {
5699 reg
= rd32(E1000_VLVF(i
));
5700 if ((reg
& E1000_VLVF_VLANID_ENABLE
) &&
5701 vid
== (reg
& E1000_VLVF_VLANID_MASK
))
5706 if (i
== E1000_VLVF_ARRAY_SIZE
) {
5707 /* Did not find a matching VLAN ID entry that was
5708 * enabled. Search for a free filter entry, i.e.
5709 * one without the enable bit set
5711 for (i
= 0; i
< E1000_VLVF_ARRAY_SIZE
; i
++) {
5712 reg
= rd32(E1000_VLVF(i
));
5713 if (!(reg
& E1000_VLVF_VLANID_ENABLE
))
5717 if (i
< E1000_VLVF_ARRAY_SIZE
) {
5718 /* Found an enabled/available entry */
5719 reg
|= 1 << (E1000_VLVF_POOLSEL_SHIFT
+ vf
);
5721 /* if !enabled we need to set this up in vfta */
5722 if (!(reg
& E1000_VLVF_VLANID_ENABLE
)) {
5723 /* add VID to filter table */
5724 igb_vfta_set(hw
, vid
, true);
5725 reg
|= E1000_VLVF_VLANID_ENABLE
;
5727 reg
&= ~E1000_VLVF_VLANID_MASK
;
5729 wr32(E1000_VLVF(i
), reg
);
5731 /* do not modify RLPML for PF devices */
5732 if (vf
>= adapter
->vfs_allocated_count
)
5735 if (!adapter
->vf_data
[vf
].vlans_enabled
) {
5737 reg
= rd32(E1000_VMOLR(vf
));
5738 size
= reg
& E1000_VMOLR_RLPML_MASK
;
5740 reg
&= ~E1000_VMOLR_RLPML_MASK
;
5742 wr32(E1000_VMOLR(vf
), reg
);
5745 adapter
->vf_data
[vf
].vlans_enabled
++;
5748 if (i
< E1000_VLVF_ARRAY_SIZE
) {
5749 /* remove vf from the pool */
5750 reg
&= ~(1 << (E1000_VLVF_POOLSEL_SHIFT
+ vf
));
5751 /* if pool is empty then remove entry from vfta */
5752 if (!(reg
& E1000_VLVF_POOLSEL_MASK
)) {
5754 igb_vfta_set(hw
, vid
, false);
5756 wr32(E1000_VLVF(i
), reg
);
5758 /* do not modify RLPML for PF devices */
5759 if (vf
>= adapter
->vfs_allocated_count
)
5762 adapter
->vf_data
[vf
].vlans_enabled
--;
5763 if (!adapter
->vf_data
[vf
].vlans_enabled
) {
5765 reg
= rd32(E1000_VMOLR(vf
));
5766 size
= reg
& E1000_VMOLR_RLPML_MASK
;
5768 reg
&= ~E1000_VMOLR_RLPML_MASK
;
5770 wr32(E1000_VMOLR(vf
), reg
);
5777 static void igb_set_vmvir(struct igb_adapter
*adapter
, u32 vid
, u32 vf
)
5779 struct e1000_hw
*hw
= &adapter
->hw
;
5782 wr32(E1000_VMVIR(vf
), (vid
| E1000_VMVIR_VLANA_DEFAULT
));
5784 wr32(E1000_VMVIR(vf
), 0);
5787 static int igb_ndo_set_vf_vlan(struct net_device
*netdev
,
5788 int vf
, u16 vlan
, u8 qos
)
5791 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5793 if ((vf
>= adapter
->vfs_allocated_count
) || (vlan
> 4095) || (qos
> 7))
5796 err
= igb_vlvf_set(adapter
, vlan
, !!vlan
, vf
);
5799 igb_set_vmvir(adapter
, vlan
| (qos
<< VLAN_PRIO_SHIFT
), vf
);
5800 igb_set_vmolr(adapter
, vf
, !vlan
);
5801 adapter
->vf_data
[vf
].pf_vlan
= vlan
;
5802 adapter
->vf_data
[vf
].pf_qos
= qos
;
5803 dev_info(&adapter
->pdev
->dev
,
5804 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan
, qos
, vf
);
5805 if (test_bit(__IGB_DOWN
, &adapter
->state
)) {
5806 dev_warn(&adapter
->pdev
->dev
,
5807 "The VF VLAN has been set, but the PF device is not up.\n");
5808 dev_warn(&adapter
->pdev
->dev
,
5809 "Bring the PF device up before attempting to use the VF device.\n");
5812 igb_vlvf_set(adapter
, adapter
->vf_data
[vf
].pf_vlan
,
5814 igb_set_vmvir(adapter
, vlan
, vf
);
5815 igb_set_vmolr(adapter
, vf
, true);
5816 adapter
->vf_data
[vf
].pf_vlan
= 0;
5817 adapter
->vf_data
[vf
].pf_qos
= 0;
5823 static int igb_find_vlvf_entry(struct igb_adapter
*adapter
, int vid
)
5825 struct e1000_hw
*hw
= &adapter
->hw
;
5829 /* Find the vlan filter for this id */
5830 for (i
= 0; i
< E1000_VLVF_ARRAY_SIZE
; i
++) {
5831 reg
= rd32(E1000_VLVF(i
));
5832 if ((reg
& E1000_VLVF_VLANID_ENABLE
) &&
5833 vid
== (reg
& E1000_VLVF_VLANID_MASK
))
5837 if (i
>= E1000_VLVF_ARRAY_SIZE
)
5843 static int igb_set_vf_vlan(struct igb_adapter
*adapter
, u32
*msgbuf
, u32 vf
)
5845 struct e1000_hw
*hw
= &adapter
->hw
;
5846 int add
= (msgbuf
[0] & E1000_VT_MSGINFO_MASK
) >> E1000_VT_MSGINFO_SHIFT
;
5847 int vid
= (msgbuf
[1] & E1000_VLVF_VLANID_MASK
);
5850 /* If in promiscuous mode we need to make sure the PF also has
5851 * the VLAN filter set.
5853 if (add
&& (adapter
->netdev
->flags
& IFF_PROMISC
))
5854 err
= igb_vlvf_set(adapter
, vid
, add
,
5855 adapter
->vfs_allocated_count
);
5859 err
= igb_vlvf_set(adapter
, vid
, add
, vf
);
5864 /* Go through all the checks to see if the VLAN filter should
5865 * be wiped completely.
5867 if (!add
&& (adapter
->netdev
->flags
& IFF_PROMISC
)) {
5870 int regndx
= igb_find_vlvf_entry(adapter
, vid
);
5873 /* See if any other pools are set for this VLAN filter
5874 * entry other than the PF.
5876 vlvf
= bits
= rd32(E1000_VLVF(regndx
));
5877 bits
&= 1 << (E1000_VLVF_POOLSEL_SHIFT
+
5878 adapter
->vfs_allocated_count
);
5879 /* If the filter was removed then ensure PF pool bit
5880 * is cleared if the PF only added itself to the pool
5881 * because the PF is in promiscuous mode.
5883 if ((vlvf
& VLAN_VID_MASK
) == vid
&&
5884 !test_bit(vid
, adapter
->active_vlans
) &&
5886 igb_vlvf_set(adapter
, vid
, add
,
5887 adapter
->vfs_allocated_count
);
5894 static inline void igb_vf_reset(struct igb_adapter
*adapter
, u32 vf
)
5896 /* clear flags - except flag that indicates PF has set the MAC */
5897 adapter
->vf_data
[vf
].flags
&= IGB_VF_FLAG_PF_SET_MAC
;
5898 adapter
->vf_data
[vf
].last_nack
= jiffies
;
5900 /* reset offloads to defaults */
5901 igb_set_vmolr(adapter
, vf
, true);
5903 /* reset vlans for device */
5904 igb_clear_vf_vfta(adapter
, vf
);
5905 if (adapter
->vf_data
[vf
].pf_vlan
)
5906 igb_ndo_set_vf_vlan(adapter
->netdev
, vf
,
5907 adapter
->vf_data
[vf
].pf_vlan
,
5908 adapter
->vf_data
[vf
].pf_qos
);
5910 igb_clear_vf_vfta(adapter
, vf
);
5912 /* reset multicast table array for vf */
5913 adapter
->vf_data
[vf
].num_vf_mc_hashes
= 0;
5915 /* Flush and reset the mta with the new values */
5916 igb_set_rx_mode(adapter
->netdev
);
5919 static void igb_vf_reset_event(struct igb_adapter
*adapter
, u32 vf
)
5921 unsigned char *vf_mac
= adapter
->vf_data
[vf
].vf_mac_addresses
;
5923 /* clear mac address as we were hotplug removed/added */
5924 if (!(adapter
->vf_data
[vf
].flags
& IGB_VF_FLAG_PF_SET_MAC
))
5925 eth_zero_addr(vf_mac
);
5927 /* process remaining reset events */
5928 igb_vf_reset(adapter
, vf
);
5931 static void igb_vf_reset_msg(struct igb_adapter
*adapter
, u32 vf
)
5933 struct e1000_hw
*hw
= &adapter
->hw
;
5934 unsigned char *vf_mac
= adapter
->vf_data
[vf
].vf_mac_addresses
;
5935 int rar_entry
= hw
->mac
.rar_entry_count
- (vf
+ 1);
5937 u8
*addr
= (u8
*)(&msgbuf
[1]);
5939 /* process all the same items cleared in a function level reset */
5940 igb_vf_reset(adapter
, vf
);
5942 /* set vf mac address */
5943 igb_rar_set_qsel(adapter
, vf_mac
, rar_entry
, vf
);
5945 /* enable transmit and receive for vf */
5946 reg
= rd32(E1000_VFTE
);
5947 wr32(E1000_VFTE
, reg
| (1 << vf
));
5948 reg
= rd32(E1000_VFRE
);
5949 wr32(E1000_VFRE
, reg
| (1 << vf
));
5951 adapter
->vf_data
[vf
].flags
|= IGB_VF_FLAG_CTS
;
5953 /* reply to reset with ack and vf mac address */
5954 msgbuf
[0] = E1000_VF_RESET
| E1000_VT_MSGTYPE_ACK
;
5955 memcpy(addr
, vf_mac
, ETH_ALEN
);
5956 igb_write_mbx(hw
, msgbuf
, 3, vf
);
5959 static int igb_set_vf_mac_addr(struct igb_adapter
*adapter
, u32
*msg
, int vf
)
5961 /* The VF MAC Address is stored in a packed array of bytes
5962 * starting at the second 32 bit word of the msg array
5964 unsigned char *addr
= (char *)&msg
[1];
5967 if (is_valid_ether_addr(addr
))
5968 err
= igb_set_vf_mac(adapter
, vf
, addr
);
5973 static void igb_rcv_ack_from_vf(struct igb_adapter
*adapter
, u32 vf
)
5975 struct e1000_hw
*hw
= &adapter
->hw
;
5976 struct vf_data_storage
*vf_data
= &adapter
->vf_data
[vf
];
5977 u32 msg
= E1000_VT_MSGTYPE_NACK
;
5979 /* if device isn't clear to send it shouldn't be reading either */
5980 if (!(vf_data
->flags
& IGB_VF_FLAG_CTS
) &&
5981 time_after(jiffies
, vf_data
->last_nack
+ (2 * HZ
))) {
5982 igb_write_mbx(hw
, &msg
, 1, vf
);
5983 vf_data
->last_nack
= jiffies
;
5987 static void igb_rcv_msg_from_vf(struct igb_adapter
*adapter
, u32 vf
)
5989 struct pci_dev
*pdev
= adapter
->pdev
;
5990 u32 msgbuf
[E1000_VFMAILBOX_SIZE
];
5991 struct e1000_hw
*hw
= &adapter
->hw
;
5992 struct vf_data_storage
*vf_data
= &adapter
->vf_data
[vf
];
5995 retval
= igb_read_mbx(hw
, msgbuf
, E1000_VFMAILBOX_SIZE
, vf
);
5998 /* if receive failed revoke VF CTS stats and restart init */
5999 dev_err(&pdev
->dev
, "Error receiving message from VF\n");
6000 vf_data
->flags
&= ~IGB_VF_FLAG_CTS
;
6001 if (!time_after(jiffies
, vf_data
->last_nack
+ (2 * HZ
)))
6006 /* this is a message we already processed, do nothing */
6007 if (msgbuf
[0] & (E1000_VT_MSGTYPE_ACK
| E1000_VT_MSGTYPE_NACK
))
6010 /* until the vf completes a reset it should not be
6011 * allowed to start any configuration.
6013 if (msgbuf
[0] == E1000_VF_RESET
) {
6014 igb_vf_reset_msg(adapter
, vf
);
6018 if (!(vf_data
->flags
& IGB_VF_FLAG_CTS
)) {
6019 if (!time_after(jiffies
, vf_data
->last_nack
+ (2 * HZ
)))
6025 switch ((msgbuf
[0] & 0xFFFF)) {
6026 case E1000_VF_SET_MAC_ADDR
:
6028 if (!(vf_data
->flags
& IGB_VF_FLAG_PF_SET_MAC
))
6029 retval
= igb_set_vf_mac_addr(adapter
, msgbuf
, vf
);
6031 dev_warn(&pdev
->dev
,
6032 "VF %d attempted to override administratively set MAC address\nReload the VF driver to resume operations\n",
6035 case E1000_VF_SET_PROMISC
:
6036 retval
= igb_set_vf_promisc(adapter
, msgbuf
, vf
);
6038 case E1000_VF_SET_MULTICAST
:
6039 retval
= igb_set_vf_multicasts(adapter
, msgbuf
, vf
);
6041 case E1000_VF_SET_LPE
:
6042 retval
= igb_set_vf_rlpml(adapter
, msgbuf
[1], vf
);
6044 case E1000_VF_SET_VLAN
:
6046 if (vf_data
->pf_vlan
)
6047 dev_warn(&pdev
->dev
,
6048 "VF %d attempted to override administratively set VLAN tag\nReload the VF driver to resume operations\n",
6051 retval
= igb_set_vf_vlan(adapter
, msgbuf
, vf
);
6054 dev_err(&pdev
->dev
, "Unhandled Msg %08x\n", msgbuf
[0]);
6059 msgbuf
[0] |= E1000_VT_MSGTYPE_CTS
;
6061 /* notify the VF of the results of what it sent us */
6063 msgbuf
[0] |= E1000_VT_MSGTYPE_NACK
;
6065 msgbuf
[0] |= E1000_VT_MSGTYPE_ACK
;
6067 igb_write_mbx(hw
, msgbuf
, 1, vf
);
6070 static void igb_msg_task(struct igb_adapter
*adapter
)
6072 struct e1000_hw
*hw
= &adapter
->hw
;
6075 for (vf
= 0; vf
< adapter
->vfs_allocated_count
; vf
++) {
6076 /* process any reset requests */
6077 if (!igb_check_for_rst(hw
, vf
))
6078 igb_vf_reset_event(adapter
, vf
);
6080 /* process any messages pending */
6081 if (!igb_check_for_msg(hw
, vf
))
6082 igb_rcv_msg_from_vf(adapter
, vf
);
6084 /* process any acks */
6085 if (!igb_check_for_ack(hw
, vf
))
6086 igb_rcv_ack_from_vf(adapter
, vf
);
6091 * igb_set_uta - Set unicast filter table address
6092 * @adapter: board private structure
6094 * The unicast table address is a register array of 32-bit registers.
6095 * The table is meant to be used in a way similar to how the MTA is used
6096 * however due to certain limitations in the hardware it is necessary to
6097 * set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
6098 * enable bit to allow vlan tag stripping when promiscuous mode is enabled
6100 static void igb_set_uta(struct igb_adapter
*adapter
)
6102 struct e1000_hw
*hw
= &adapter
->hw
;
6105 /* The UTA table only exists on 82576 hardware and newer */
6106 if (hw
->mac
.type
< e1000_82576
)
6109 /* we only need to do this if VMDq is enabled */
6110 if (!adapter
->vfs_allocated_count
)
6113 for (i
= 0; i
< hw
->mac
.uta_reg_count
; i
++)
6114 array_wr32(E1000_UTA
, i
, ~0);
6118 * igb_intr_msi - Interrupt Handler
6119 * @irq: interrupt number
6120 * @data: pointer to a network interface device structure
6122 static irqreturn_t
igb_intr_msi(int irq
, void *data
)
6124 struct igb_adapter
*adapter
= data
;
6125 struct igb_q_vector
*q_vector
= adapter
->q_vector
[0];
6126 struct e1000_hw
*hw
= &adapter
->hw
;
6127 /* read ICR disables interrupts using IAM */
6128 u32 icr
= rd32(E1000_ICR
);
6130 igb_write_itr(q_vector
);
6132 if (icr
& E1000_ICR_DRSTA
)
6133 schedule_work(&adapter
->reset_task
);
6135 if (icr
& E1000_ICR_DOUTSYNC
) {
6136 /* HW is reporting DMA is out of sync */
6137 adapter
->stats
.doosync
++;
6140 if (icr
& (E1000_ICR_RXSEQ
| E1000_ICR_LSC
)) {
6141 hw
->mac
.get_link_status
= 1;
6142 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
6143 mod_timer(&adapter
->watchdog_timer
, jiffies
+ 1);
6146 if (icr
& E1000_ICR_TS
) {
6147 u32 tsicr
= rd32(E1000_TSICR
);
6149 if (tsicr
& E1000_TSICR_TXTS
) {
6150 /* acknowledge the interrupt */
6151 wr32(E1000_TSICR
, E1000_TSICR_TXTS
);
6152 /* retrieve hardware timestamp */
6153 schedule_work(&adapter
->ptp_tx_work
);
6157 napi_schedule(&q_vector
->napi
);
6163 * igb_intr - Legacy Interrupt Handler
6164 * @irq: interrupt number
6165 * @data: pointer to a network interface device structure
6167 static irqreturn_t
igb_intr(int irq
, void *data
)
6169 struct igb_adapter
*adapter
= data
;
6170 struct igb_q_vector
*q_vector
= adapter
->q_vector
[0];
6171 struct e1000_hw
*hw
= &adapter
->hw
;
6172 /* Interrupt Auto-Mask...upon reading ICR, interrupts are masked. No
6173 * need for the IMC write
6175 u32 icr
= rd32(E1000_ICR
);
6177 /* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
6178 * not set, then the adapter didn't send an interrupt
6180 if (!(icr
& E1000_ICR_INT_ASSERTED
))
6183 igb_write_itr(q_vector
);
6185 if (icr
& E1000_ICR_DRSTA
)
6186 schedule_work(&adapter
->reset_task
);
6188 if (icr
& E1000_ICR_DOUTSYNC
) {
6189 /* HW is reporting DMA is out of sync */
6190 adapter
->stats
.doosync
++;
6193 if (icr
& (E1000_ICR_RXSEQ
| E1000_ICR_LSC
)) {
6194 hw
->mac
.get_link_status
= 1;
6195 /* guard against interrupt when we're going down */
6196 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
6197 mod_timer(&adapter
->watchdog_timer
, jiffies
+ 1);
6200 if (icr
& E1000_ICR_TS
) {
6201 u32 tsicr
= rd32(E1000_TSICR
);
6203 if (tsicr
& E1000_TSICR_TXTS
) {
6204 /* acknowledge the interrupt */
6205 wr32(E1000_TSICR
, E1000_TSICR_TXTS
);
6206 /* retrieve hardware timestamp */
6207 schedule_work(&adapter
->ptp_tx_work
);
6211 napi_schedule(&q_vector
->napi
);
6216 static void igb_ring_irq_enable(struct igb_q_vector
*q_vector
)
6218 struct igb_adapter
*adapter
= q_vector
->adapter
;
6219 struct e1000_hw
*hw
= &adapter
->hw
;
6221 if ((q_vector
->rx
.ring
&& (adapter
->rx_itr_setting
& 3)) ||
6222 (!q_vector
->rx
.ring
&& (adapter
->tx_itr_setting
& 3))) {
6223 if ((adapter
->num_q_vectors
== 1) && !adapter
->vf_data
)
6224 igb_set_itr(q_vector
);
6226 igb_update_ring_itr(q_vector
);
6229 if (!test_bit(__IGB_DOWN
, &adapter
->state
)) {
6230 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
)
6231 wr32(E1000_EIMS
, q_vector
->eims_value
);
6233 igb_irq_enable(adapter
);
6238 * igb_poll - NAPI Rx polling callback
6239 * @napi: napi polling structure
6240 * @budget: count of how many packets we should handle
6242 static int igb_poll(struct napi_struct
*napi
, int budget
)
6244 struct igb_q_vector
*q_vector
= container_of(napi
,
6245 struct igb_q_vector
,
6247 bool clean_complete
= true;
6249 #ifdef CONFIG_IGB_DCA
6250 if (q_vector
->adapter
->flags
& IGB_FLAG_DCA_ENABLED
)
6251 igb_update_dca(q_vector
);
6253 if (q_vector
->tx
.ring
)
6254 clean_complete
= igb_clean_tx_irq(q_vector
);
6256 if (q_vector
->rx
.ring
)
6257 clean_complete
&= igb_clean_rx_irq(q_vector
, budget
);
6259 /* If all work not completed, return budget and keep polling */
6260 if (!clean_complete
)
6263 /* If not enough Rx work done, exit the polling mode */
6264 napi_complete(napi
);
6265 igb_ring_irq_enable(q_vector
);
6271 * igb_clean_tx_irq - Reclaim resources after transmit completes
6272 * @q_vector: pointer to q_vector containing needed info
6274 * returns true if ring is completely cleaned
6276 static bool igb_clean_tx_irq(struct igb_q_vector
*q_vector
)
6278 struct igb_adapter
*adapter
= q_vector
->adapter
;
6279 struct igb_ring
*tx_ring
= q_vector
->tx
.ring
;
6280 struct igb_tx_buffer
*tx_buffer
;
6281 union e1000_adv_tx_desc
*tx_desc
;
6282 unsigned int total_bytes
= 0, total_packets
= 0;
6283 unsigned int budget
= q_vector
->tx
.work_limit
;
6284 unsigned int i
= tx_ring
->next_to_clean
;
6286 if (test_bit(__IGB_DOWN
, &adapter
->state
))
6289 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
6290 tx_desc
= IGB_TX_DESC(tx_ring
, i
);
6291 i
-= tx_ring
->count
;
6294 union e1000_adv_tx_desc
*eop_desc
= tx_buffer
->next_to_watch
;
6296 /* if next_to_watch is not set then there is no work pending */
6300 /* prevent any other reads prior to eop_desc */
6301 read_barrier_depends();
6303 /* if DD is not set pending work has not been completed */
6304 if (!(eop_desc
->wb
.status
& cpu_to_le32(E1000_TXD_STAT_DD
)))
6307 /* clear next_to_watch to prevent false hangs */
6308 tx_buffer
->next_to_watch
= NULL
;
6310 /* update the statistics for this packet */
6311 total_bytes
+= tx_buffer
->bytecount
;
6312 total_packets
+= tx_buffer
->gso_segs
;
6315 dev_kfree_skb_any(tx_buffer
->skb
);
6317 /* unmap skb header data */
6318 dma_unmap_single(tx_ring
->dev
,
6319 dma_unmap_addr(tx_buffer
, dma
),
6320 dma_unmap_len(tx_buffer
, len
),
6323 /* clear tx_buffer data */
6324 tx_buffer
->skb
= NULL
;
6325 dma_unmap_len_set(tx_buffer
, len
, 0);
6327 /* clear last DMA location and unmap remaining buffers */
6328 while (tx_desc
!= eop_desc
) {
6333 i
-= tx_ring
->count
;
6334 tx_buffer
= tx_ring
->tx_buffer_info
;
6335 tx_desc
= IGB_TX_DESC(tx_ring
, 0);
6338 /* unmap any remaining paged data */
6339 if (dma_unmap_len(tx_buffer
, len
)) {
6340 dma_unmap_page(tx_ring
->dev
,
6341 dma_unmap_addr(tx_buffer
, dma
),
6342 dma_unmap_len(tx_buffer
, len
),
6344 dma_unmap_len_set(tx_buffer
, len
, 0);
6348 /* move us one more past the eop_desc for start of next pkt */
6353 i
-= tx_ring
->count
;
6354 tx_buffer
= tx_ring
->tx_buffer_info
;
6355 tx_desc
= IGB_TX_DESC(tx_ring
, 0);
6358 /* issue prefetch for next Tx descriptor */
6361 /* update budget accounting */
6363 } while (likely(budget
));
6365 netdev_tx_completed_queue(txring_txq(tx_ring
),
6366 total_packets
, total_bytes
);
6367 i
+= tx_ring
->count
;
6368 tx_ring
->next_to_clean
= i
;
6369 u64_stats_update_begin(&tx_ring
->tx_syncp
);
6370 tx_ring
->tx_stats
.bytes
+= total_bytes
;
6371 tx_ring
->tx_stats
.packets
+= total_packets
;
6372 u64_stats_update_end(&tx_ring
->tx_syncp
);
6373 q_vector
->tx
.total_bytes
+= total_bytes
;
6374 q_vector
->tx
.total_packets
+= total_packets
;
6376 if (test_bit(IGB_RING_FLAG_TX_DETECT_HANG
, &tx_ring
->flags
)) {
6377 struct e1000_hw
*hw
= &adapter
->hw
;
6379 /* Detect a transmit hang in hardware, this serializes the
6380 * check with the clearing of time_stamp and movement of i
6382 clear_bit(IGB_RING_FLAG_TX_DETECT_HANG
, &tx_ring
->flags
);
6383 if (tx_buffer
->next_to_watch
&&
6384 time_after(jiffies
, tx_buffer
->time_stamp
+
6385 (adapter
->tx_timeout_factor
* HZ
)) &&
6386 !(rd32(E1000_STATUS
) & E1000_STATUS_TXOFF
)) {
6388 /* detected Tx unit hang */
6389 dev_err(tx_ring
->dev
,
6390 "Detected Tx Unit Hang\n"
6394 " next_to_use <%x>\n"
6395 " next_to_clean <%x>\n"
6396 "buffer_info[next_to_clean]\n"
6397 " time_stamp <%lx>\n"
6398 " next_to_watch <%p>\n"
6400 " desc.status <%x>\n",
6401 tx_ring
->queue_index
,
6402 rd32(E1000_TDH(tx_ring
->reg_idx
)),
6403 readl(tx_ring
->tail
),
6404 tx_ring
->next_to_use
,
6405 tx_ring
->next_to_clean
,
6406 tx_buffer
->time_stamp
,
6407 tx_buffer
->next_to_watch
,
6409 tx_buffer
->next_to_watch
->wb
.status
);
6410 netif_stop_subqueue(tx_ring
->netdev
,
6411 tx_ring
->queue_index
);
6413 /* we are about to reset, no point in enabling stuff */
6418 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
6419 if (unlikely(total_packets
&&
6420 netif_carrier_ok(tx_ring
->netdev
) &&
6421 igb_desc_unused(tx_ring
) >= TX_WAKE_THRESHOLD
)) {
6422 /* Make sure that anybody stopping the queue after this
6423 * sees the new next_to_clean.
6426 if (__netif_subqueue_stopped(tx_ring
->netdev
,
6427 tx_ring
->queue_index
) &&
6428 !(test_bit(__IGB_DOWN
, &adapter
->state
))) {
6429 netif_wake_subqueue(tx_ring
->netdev
,
6430 tx_ring
->queue_index
);
6432 u64_stats_update_begin(&tx_ring
->tx_syncp
);
6433 tx_ring
->tx_stats
.restart_queue
++;
6434 u64_stats_update_end(&tx_ring
->tx_syncp
);
6442 * igb_reuse_rx_page - page flip buffer and store it back on the ring
6443 * @rx_ring: rx descriptor ring to store buffers on
6444 * @old_buff: donor buffer to have page reused
6446 * Synchronizes page for reuse by the adapter
6448 static void igb_reuse_rx_page(struct igb_ring
*rx_ring
,
6449 struct igb_rx_buffer
*old_buff
)
6451 struct igb_rx_buffer
*new_buff
;
6452 u16 nta
= rx_ring
->next_to_alloc
;
6454 new_buff
= &rx_ring
->rx_buffer_info
[nta
];
6456 /* update, and store next to alloc */
6458 rx_ring
->next_to_alloc
= (nta
< rx_ring
->count
) ? nta
: 0;
6460 /* transfer page from old buffer to new buffer */
6461 memcpy(new_buff
, old_buff
, sizeof(struct igb_rx_buffer
));
6463 /* sync the buffer for use by the device */
6464 dma_sync_single_range_for_device(rx_ring
->dev
, old_buff
->dma
,
6465 old_buff
->page_offset
,
6470 static bool igb_can_reuse_rx_page(struct igb_rx_buffer
*rx_buffer
,
6472 unsigned int truesize
)
6474 /* avoid re-using remote pages */
6475 if (unlikely(page_to_nid(page
) != numa_node_id()))
6478 #if (PAGE_SIZE < 8192)
6479 /* if we are only owner of page we can reuse it */
6480 if (unlikely(page_count(page
) != 1))
6483 /* flip page offset to other buffer */
6484 rx_buffer
->page_offset
^= IGB_RX_BUFSZ
;
6486 /* since we are the only owner of the page and we need to
6487 * increment it, just set the value to 2 in order to avoid
6488 * an unnecessary locked operation
6490 atomic_set(&page
->_count
, 2);
6492 /* move offset up to the next cache line */
6493 rx_buffer
->page_offset
+= truesize
;
6495 if (rx_buffer
->page_offset
> (PAGE_SIZE
- IGB_RX_BUFSZ
))
6498 /* bump ref count on page before it is given to the stack */
6506 * igb_add_rx_frag - Add contents of Rx buffer to sk_buff
6507 * @rx_ring: rx descriptor ring to transact packets on
6508 * @rx_buffer: buffer containing page to add
6509 * @rx_desc: descriptor containing length of buffer written by hardware
6510 * @skb: sk_buff to place the data into
6512 * This function will add the data contained in rx_buffer->page to the skb.
6513 * This is done either through a direct copy if the data in the buffer is
6514 * less than the skb header size, otherwise it will just attach the page as
6515 * a frag to the skb.
6517 * The function will then update the page offset if necessary and return
6518 * true if the buffer can be reused by the adapter.
6520 static bool igb_add_rx_frag(struct igb_ring
*rx_ring
,
6521 struct igb_rx_buffer
*rx_buffer
,
6522 union e1000_adv_rx_desc
*rx_desc
,
6523 struct sk_buff
*skb
)
6525 struct page
*page
= rx_buffer
->page
;
6526 unsigned int size
= le16_to_cpu(rx_desc
->wb
.upper
.length
);
6527 #if (PAGE_SIZE < 8192)
6528 unsigned int truesize
= IGB_RX_BUFSZ
;
6530 unsigned int truesize
= ALIGN(size
, L1_CACHE_BYTES
);
6533 if ((size
<= IGB_RX_HDR_LEN
) && !skb_is_nonlinear(skb
)) {
6534 unsigned char *va
= page_address(page
) + rx_buffer
->page_offset
;
6536 if (igb_test_staterr(rx_desc
, E1000_RXDADV_STAT_TSIP
)) {
6537 igb_ptp_rx_pktstamp(rx_ring
->q_vector
, va
, skb
);
6538 va
+= IGB_TS_HDR_LEN
;
6539 size
-= IGB_TS_HDR_LEN
;
6542 memcpy(__skb_put(skb
, size
), va
, ALIGN(size
, sizeof(long)));
6544 /* we can reuse buffer as-is, just make sure it is local */
6545 if (likely(page_to_nid(page
) == numa_node_id()))
6548 /* this page cannot be reused so discard it */
6553 skb_add_rx_frag(skb
, skb_shinfo(skb
)->nr_frags
, page
,
6554 rx_buffer
->page_offset
, size
, truesize
);
6556 return igb_can_reuse_rx_page(rx_buffer
, page
, truesize
);
6559 static struct sk_buff
*igb_fetch_rx_buffer(struct igb_ring
*rx_ring
,
6560 union e1000_adv_rx_desc
*rx_desc
,
6561 struct sk_buff
*skb
)
6563 struct igb_rx_buffer
*rx_buffer
;
6566 rx_buffer
= &rx_ring
->rx_buffer_info
[rx_ring
->next_to_clean
];
6568 page
= rx_buffer
->page
;
6572 void *page_addr
= page_address(page
) +
6573 rx_buffer
->page_offset
;
6575 /* prefetch first cache line of first page */
6576 prefetch(page_addr
);
6577 #if L1_CACHE_BYTES < 128
6578 prefetch(page_addr
+ L1_CACHE_BYTES
);
6581 /* allocate a skb to store the frags */
6582 skb
= netdev_alloc_skb_ip_align(rx_ring
->netdev
,
6584 if (unlikely(!skb
)) {
6585 rx_ring
->rx_stats
.alloc_failed
++;
6589 /* we will be copying header into skb->data in
6590 * pskb_may_pull so it is in our interest to prefetch
6591 * it now to avoid a possible cache miss
6593 prefetchw(skb
->data
);
6596 /* we are reusing so sync this buffer for CPU use */
6597 dma_sync_single_range_for_cpu(rx_ring
->dev
,
6599 rx_buffer
->page_offset
,
6603 /* pull page into skb */
6604 if (igb_add_rx_frag(rx_ring
, rx_buffer
, rx_desc
, skb
)) {
6605 /* hand second half of page back to the ring */
6606 igb_reuse_rx_page(rx_ring
, rx_buffer
);
6608 /* we are not reusing the buffer so unmap it */
6609 dma_unmap_page(rx_ring
->dev
, rx_buffer
->dma
,
6610 PAGE_SIZE
, DMA_FROM_DEVICE
);
6613 /* clear contents of rx_buffer */
6614 rx_buffer
->page
= NULL
;
6619 static inline void igb_rx_checksum(struct igb_ring
*ring
,
6620 union e1000_adv_rx_desc
*rx_desc
,
6621 struct sk_buff
*skb
)
6623 skb_checksum_none_assert(skb
);
6625 /* Ignore Checksum bit is set */
6626 if (igb_test_staterr(rx_desc
, E1000_RXD_STAT_IXSM
))
6629 /* Rx checksum disabled via ethtool */
6630 if (!(ring
->netdev
->features
& NETIF_F_RXCSUM
))
6633 /* TCP/UDP checksum error bit is set */
6634 if (igb_test_staterr(rx_desc
,
6635 E1000_RXDEXT_STATERR_TCPE
|
6636 E1000_RXDEXT_STATERR_IPE
)) {
6637 /* work around errata with sctp packets where the TCPE aka
6638 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
6639 * packets, (aka let the stack check the crc32c)
6641 if (!((skb
->len
== 60) &&
6642 test_bit(IGB_RING_FLAG_RX_SCTP_CSUM
, &ring
->flags
))) {
6643 u64_stats_update_begin(&ring
->rx_syncp
);
6644 ring
->rx_stats
.csum_err
++;
6645 u64_stats_update_end(&ring
->rx_syncp
);
6647 /* let the stack verify checksum errors */
6650 /* It must be a TCP or UDP packet with a valid checksum */
6651 if (igb_test_staterr(rx_desc
, E1000_RXD_STAT_TCPCS
|
6652 E1000_RXD_STAT_UDPCS
))
6653 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
6655 dev_dbg(ring
->dev
, "cksum success: bits %08X\n",
6656 le32_to_cpu(rx_desc
->wb
.upper
.status_error
));
6659 static inline void igb_rx_hash(struct igb_ring
*ring
,
6660 union e1000_adv_rx_desc
*rx_desc
,
6661 struct sk_buff
*skb
)
6663 if (ring
->netdev
->features
& NETIF_F_RXHASH
)
6665 le32_to_cpu(rx_desc
->wb
.lower
.hi_dword
.rss
),
6670 * igb_is_non_eop - process handling of non-EOP buffers
6671 * @rx_ring: Rx ring being processed
6672 * @rx_desc: Rx descriptor for current buffer
6673 * @skb: current socket buffer containing buffer in progress
6675 * This function updates next to clean. If the buffer is an EOP buffer
6676 * this function exits returning false, otherwise it will place the
6677 * sk_buff in the next buffer to be chained and return true indicating
6678 * that this is in fact a non-EOP buffer.
6680 static bool igb_is_non_eop(struct igb_ring
*rx_ring
,
6681 union e1000_adv_rx_desc
*rx_desc
)
6683 u32 ntc
= rx_ring
->next_to_clean
+ 1;
6685 /* fetch, update, and store next to clean */
6686 ntc
= (ntc
< rx_ring
->count
) ? ntc
: 0;
6687 rx_ring
->next_to_clean
= ntc
;
6689 prefetch(IGB_RX_DESC(rx_ring
, ntc
));
6691 if (likely(igb_test_staterr(rx_desc
, E1000_RXD_STAT_EOP
)))
6698 * igb_get_headlen - determine size of header for LRO/GRO
6699 * @data: pointer to the start of the headers
6700 * @max_len: total length of section to find headers in
6702 * This function is meant to determine the length of headers that will
6703 * be recognized by hardware for LRO, and GRO offloads. The main
6704 * motivation of doing this is to only perform one pull for IPv4 TCP
6705 * packets so that we can do basic things like calculating the gso_size
6706 * based on the average data per packet.
6708 static unsigned int igb_get_headlen(unsigned char *data
,
6709 unsigned int max_len
)
6712 unsigned char *network
;
6715 struct vlan_hdr
*vlan
;
6718 struct ipv6hdr
*ipv6
;
6721 u8 nexthdr
= 0; /* default to not TCP */
6724 /* this should never happen, but better safe than sorry */
6725 if (max_len
< ETH_HLEN
)
6728 /* initialize network frame pointer */
6731 /* set first protocol and move network header forward */
6732 protocol
= hdr
.eth
->h_proto
;
6733 hdr
.network
+= ETH_HLEN
;
6735 /* handle any vlan tag if present */
6736 if (protocol
== __constant_htons(ETH_P_8021Q
)) {
6737 if ((hdr
.network
- data
) > (max_len
- VLAN_HLEN
))
6740 protocol
= hdr
.vlan
->h_vlan_encapsulated_proto
;
6741 hdr
.network
+= VLAN_HLEN
;
6744 /* handle L3 protocols */
6745 if (protocol
== __constant_htons(ETH_P_IP
)) {
6746 if ((hdr
.network
- data
) > (max_len
- sizeof(struct iphdr
)))
6749 /* access ihl as a u8 to avoid unaligned access on ia64 */
6750 hlen
= (hdr
.network
[0] & 0x0F) << 2;
6752 /* verify hlen meets minimum size requirements */
6753 if (hlen
< sizeof(struct iphdr
))
6754 return hdr
.network
- data
;
6756 /* record next protocol if header is present */
6757 if (!(hdr
.ipv4
->frag_off
& htons(IP_OFFSET
)))
6758 nexthdr
= hdr
.ipv4
->protocol
;
6759 } else if (protocol
== __constant_htons(ETH_P_IPV6
)) {
6760 if ((hdr
.network
- data
) > (max_len
- sizeof(struct ipv6hdr
)))
6763 /* record next protocol */
6764 nexthdr
= hdr
.ipv6
->nexthdr
;
6765 hlen
= sizeof(struct ipv6hdr
);
6767 return hdr
.network
- data
;
6770 /* relocate pointer to start of L4 header */
6771 hdr
.network
+= hlen
;
6773 /* finally sort out TCP */
6774 if (nexthdr
== IPPROTO_TCP
) {
6775 if ((hdr
.network
- data
) > (max_len
- sizeof(struct tcphdr
)))
6778 /* access doff as a u8 to avoid unaligned access on ia64 */
6779 hlen
= (hdr
.network
[12] & 0xF0) >> 2;
6781 /* verify hlen meets minimum size requirements */
6782 if (hlen
< sizeof(struct tcphdr
))
6783 return hdr
.network
- data
;
6785 hdr
.network
+= hlen
;
6786 } else if (nexthdr
== IPPROTO_UDP
) {
6787 if ((hdr
.network
- data
) > (max_len
- sizeof(struct udphdr
)))
6790 hdr
.network
+= sizeof(struct udphdr
);
6793 /* If everything has gone correctly hdr.network should be the
6794 * data section of the packet and will be the end of the header.
6795 * If not then it probably represents the end of the last recognized
6798 if ((hdr
.network
- data
) < max_len
)
6799 return hdr
.network
- data
;
6805 * igb_pull_tail - igb specific version of skb_pull_tail
6806 * @rx_ring: rx descriptor ring packet is being transacted on
6807 * @rx_desc: pointer to the EOP Rx descriptor
6808 * @skb: pointer to current skb being adjusted
6810 * This function is an igb specific version of __pskb_pull_tail. The
6811 * main difference between this version and the original function is that
6812 * this function can make several assumptions about the state of things
6813 * that allow for significant optimizations versus the standard function.
6814 * As a result we can do things like drop a frag and maintain an accurate
6815 * truesize for the skb.
6817 static void igb_pull_tail(struct igb_ring
*rx_ring
,
6818 union e1000_adv_rx_desc
*rx_desc
,
6819 struct sk_buff
*skb
)
6821 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[0];
6823 unsigned int pull_len
;
6825 /* it is valid to use page_address instead of kmap since we are
6826 * working with pages allocated out of the lomem pool per
6827 * alloc_page(GFP_ATOMIC)
6829 va
= skb_frag_address(frag
);
6831 if (igb_test_staterr(rx_desc
, E1000_RXDADV_STAT_TSIP
)) {
6832 /* retrieve timestamp from buffer */
6833 igb_ptp_rx_pktstamp(rx_ring
->q_vector
, va
, skb
);
6835 /* update pointers to remove timestamp header */
6836 skb_frag_size_sub(frag
, IGB_TS_HDR_LEN
);
6837 frag
->page_offset
+= IGB_TS_HDR_LEN
;
6838 skb
->data_len
-= IGB_TS_HDR_LEN
;
6839 skb
->len
-= IGB_TS_HDR_LEN
;
6841 /* move va to start of packet data */
6842 va
+= IGB_TS_HDR_LEN
;
6845 /* we need the header to contain the greater of either ETH_HLEN or
6846 * 60 bytes if the skb->len is less than 60 for skb_pad.
6848 pull_len
= igb_get_headlen(va
, IGB_RX_HDR_LEN
);
6850 /* align pull length to size of long to optimize memcpy performance */
6851 skb_copy_to_linear_data(skb
, va
, ALIGN(pull_len
, sizeof(long)));
6853 /* update all of the pointers */
6854 skb_frag_size_sub(frag
, pull_len
);
6855 frag
->page_offset
+= pull_len
;
6856 skb
->data_len
-= pull_len
;
6857 skb
->tail
+= pull_len
;
6861 * igb_cleanup_headers - Correct corrupted or empty headers
6862 * @rx_ring: rx descriptor ring packet is being transacted on
6863 * @rx_desc: pointer to the EOP Rx descriptor
6864 * @skb: pointer to current skb being fixed
6866 * Address the case where we are pulling data in on pages only
6867 * and as such no data is present in the skb header.
6869 * In addition if skb is not at least 60 bytes we need to pad it so that
6870 * it is large enough to qualify as a valid Ethernet frame.
6872 * Returns true if an error was encountered and skb was freed.
6874 static bool igb_cleanup_headers(struct igb_ring
*rx_ring
,
6875 union e1000_adv_rx_desc
*rx_desc
,
6876 struct sk_buff
*skb
)
6878 if (unlikely((igb_test_staterr(rx_desc
,
6879 E1000_RXDEXT_ERR_FRAME_ERR_MASK
)))) {
6880 struct net_device
*netdev
= rx_ring
->netdev
;
6881 if (!(netdev
->features
& NETIF_F_RXALL
)) {
6882 dev_kfree_skb_any(skb
);
6887 /* place header in linear portion of buffer */
6888 if (skb_is_nonlinear(skb
))
6889 igb_pull_tail(rx_ring
, rx_desc
, skb
);
6891 /* if skb_pad returns an error the skb was freed */
6892 if (unlikely(skb
->len
< 60)) {
6893 int pad_len
= 60 - skb
->len
;
6895 if (skb_pad(skb
, pad_len
))
6897 __skb_put(skb
, pad_len
);
6904 * igb_process_skb_fields - Populate skb header fields from Rx descriptor
6905 * @rx_ring: rx descriptor ring packet is being transacted on
6906 * @rx_desc: pointer to the EOP Rx descriptor
6907 * @skb: pointer to current skb being populated
6909 * This function checks the ring, descriptor, and packet information in
6910 * order to populate the hash, checksum, VLAN, timestamp, protocol, and
6911 * other fields within the skb.
6913 static void igb_process_skb_fields(struct igb_ring
*rx_ring
,
6914 union e1000_adv_rx_desc
*rx_desc
,
6915 struct sk_buff
*skb
)
6917 struct net_device
*dev
= rx_ring
->netdev
;
6919 igb_rx_hash(rx_ring
, rx_desc
, skb
);
6921 igb_rx_checksum(rx_ring
, rx_desc
, skb
);
6923 igb_ptp_rx_hwtstamp(rx_ring
, rx_desc
, skb
);
6925 if ((dev
->features
& NETIF_F_HW_VLAN_CTAG_RX
) &&
6926 igb_test_staterr(rx_desc
, E1000_RXD_STAT_VP
)) {
6928 if (igb_test_staterr(rx_desc
, E1000_RXDEXT_STATERR_LB
) &&
6929 test_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP
, &rx_ring
->flags
))
6930 vid
= be16_to_cpu(rx_desc
->wb
.upper
.vlan
);
6932 vid
= le16_to_cpu(rx_desc
->wb
.upper
.vlan
);
6934 __vlan_hwaccel_put_tag(skb
, htons(ETH_P_8021Q
), vid
);
6937 skb_record_rx_queue(skb
, rx_ring
->queue_index
);
6939 skb
->protocol
= eth_type_trans(skb
, rx_ring
->netdev
);
6942 static bool igb_clean_rx_irq(struct igb_q_vector
*q_vector
, const int budget
)
6944 struct igb_ring
*rx_ring
= q_vector
->rx
.ring
;
6945 struct sk_buff
*skb
= rx_ring
->skb
;
6946 unsigned int total_bytes
= 0, total_packets
= 0;
6947 u16 cleaned_count
= igb_desc_unused(rx_ring
);
6950 union e1000_adv_rx_desc
*rx_desc
;
6952 /* return some buffers to hardware, one at a time is too slow */
6953 if (cleaned_count
>= IGB_RX_BUFFER_WRITE
) {
6954 igb_alloc_rx_buffers(rx_ring
, cleaned_count
);
6958 rx_desc
= IGB_RX_DESC(rx_ring
, rx_ring
->next_to_clean
);
6960 if (!igb_test_staterr(rx_desc
, E1000_RXD_STAT_DD
))
6963 /* This memory barrier is needed to keep us from reading
6964 * any other fields out of the rx_desc until we know the
6965 * RXD_STAT_DD bit is set
6969 /* retrieve a buffer from the ring */
6970 skb
= igb_fetch_rx_buffer(rx_ring
, rx_desc
, skb
);
6972 /* exit if we failed to retrieve a buffer */
6978 /* fetch next buffer in frame if non-eop */
6979 if (igb_is_non_eop(rx_ring
, rx_desc
))
6982 /* verify the packet layout is correct */
6983 if (igb_cleanup_headers(rx_ring
, rx_desc
, skb
)) {
6988 /* probably a little skewed due to removing CRC */
6989 total_bytes
+= skb
->len
;
6991 /* populate checksum, timestamp, VLAN, and protocol */
6992 igb_process_skb_fields(rx_ring
, rx_desc
, skb
);
6994 napi_gro_receive(&q_vector
->napi
, skb
);
6996 /* reset skb pointer */
6999 /* update budget accounting */
7001 } while (likely(total_packets
< budget
));
7003 /* place incomplete frames back on ring for completion */
7006 u64_stats_update_begin(&rx_ring
->rx_syncp
);
7007 rx_ring
->rx_stats
.packets
+= total_packets
;
7008 rx_ring
->rx_stats
.bytes
+= total_bytes
;
7009 u64_stats_update_end(&rx_ring
->rx_syncp
);
7010 q_vector
->rx
.total_packets
+= total_packets
;
7011 q_vector
->rx
.total_bytes
+= total_bytes
;
7014 igb_alloc_rx_buffers(rx_ring
, cleaned_count
);
7016 return (total_packets
< budget
);
7019 static bool igb_alloc_mapped_page(struct igb_ring
*rx_ring
,
7020 struct igb_rx_buffer
*bi
)
7022 struct page
*page
= bi
->page
;
7025 /* since we are recycling buffers we should seldom need to alloc */
7029 /* alloc new page for storage */
7030 page
= __skb_alloc_page(GFP_ATOMIC
| __GFP_COLD
, NULL
);
7031 if (unlikely(!page
)) {
7032 rx_ring
->rx_stats
.alloc_failed
++;
7036 /* map page for use */
7037 dma
= dma_map_page(rx_ring
->dev
, page
, 0, PAGE_SIZE
, DMA_FROM_DEVICE
);
7039 /* if mapping failed free memory back to system since
7040 * there isn't much point in holding memory we can't use
7042 if (dma_mapping_error(rx_ring
->dev
, dma
)) {
7045 rx_ring
->rx_stats
.alloc_failed
++;
7051 bi
->page_offset
= 0;
7057 * igb_alloc_rx_buffers - Replace used receive buffers; packet split
7058 * @adapter: address of board private structure
7060 void igb_alloc_rx_buffers(struct igb_ring
*rx_ring
, u16 cleaned_count
)
7062 union e1000_adv_rx_desc
*rx_desc
;
7063 struct igb_rx_buffer
*bi
;
7064 u16 i
= rx_ring
->next_to_use
;
7070 rx_desc
= IGB_RX_DESC(rx_ring
, i
);
7071 bi
= &rx_ring
->rx_buffer_info
[i
];
7072 i
-= rx_ring
->count
;
7075 if (!igb_alloc_mapped_page(rx_ring
, bi
))
7078 /* Refresh the desc even if buffer_addrs didn't change
7079 * because each write-back erases this info.
7081 rx_desc
->read
.pkt_addr
= cpu_to_le64(bi
->dma
+ bi
->page_offset
);
7087 rx_desc
= IGB_RX_DESC(rx_ring
, 0);
7088 bi
= rx_ring
->rx_buffer_info
;
7089 i
-= rx_ring
->count
;
7092 /* clear the hdr_addr for the next_to_use descriptor */
7093 rx_desc
->read
.hdr_addr
= 0;
7096 } while (cleaned_count
);
7098 i
+= rx_ring
->count
;
7100 if (rx_ring
->next_to_use
!= i
) {
7101 /* record the next descriptor to use */
7102 rx_ring
->next_to_use
= i
;
7104 /* update next to alloc since we have filled the ring */
7105 rx_ring
->next_to_alloc
= i
;
7107 /* Force memory writes to complete before letting h/w
7108 * know there are new descriptors to fetch. (Only
7109 * applicable for weak-ordered memory model archs,
7113 writel(i
, rx_ring
->tail
);
7123 static int igb_mii_ioctl(struct net_device
*netdev
, struct ifreq
*ifr
, int cmd
)
7125 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7126 struct mii_ioctl_data
*data
= if_mii(ifr
);
7128 if (adapter
->hw
.phy
.media_type
!= e1000_media_type_copper
)
7133 data
->phy_id
= adapter
->hw
.phy
.addr
;
7136 if (igb_read_phy_reg(&adapter
->hw
, data
->reg_num
& 0x1F,
7153 static int igb_ioctl(struct net_device
*netdev
, struct ifreq
*ifr
, int cmd
)
7159 return igb_mii_ioctl(netdev
, ifr
, cmd
);
7161 return igb_ptp_hwtstamp_ioctl(netdev
, ifr
, cmd
);
7167 s32
igb_read_pcie_cap_reg(struct e1000_hw
*hw
, u32 reg
, u16
*value
)
7169 struct igb_adapter
*adapter
= hw
->back
;
7171 if (pcie_capability_read_word(adapter
->pdev
, reg
, value
))
7172 return -E1000_ERR_CONFIG
;
7177 s32
igb_write_pcie_cap_reg(struct e1000_hw
*hw
, u32 reg
, u16
*value
)
7179 struct igb_adapter
*adapter
= hw
->back
;
7181 if (pcie_capability_write_word(adapter
->pdev
, reg
, *value
))
7182 return -E1000_ERR_CONFIG
;
7187 static void igb_vlan_mode(struct net_device
*netdev
, netdev_features_t features
)
7189 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7190 struct e1000_hw
*hw
= &adapter
->hw
;
7192 bool enable
= !!(features
& NETIF_F_HW_VLAN_CTAG_RX
);
7195 /* enable VLAN tag insert/strip */
7196 ctrl
= rd32(E1000_CTRL
);
7197 ctrl
|= E1000_CTRL_VME
;
7198 wr32(E1000_CTRL
, ctrl
);
7200 /* Disable CFI check */
7201 rctl
= rd32(E1000_RCTL
);
7202 rctl
&= ~E1000_RCTL_CFIEN
;
7203 wr32(E1000_RCTL
, rctl
);
7205 /* disable VLAN tag insert/strip */
7206 ctrl
= rd32(E1000_CTRL
);
7207 ctrl
&= ~E1000_CTRL_VME
;
7208 wr32(E1000_CTRL
, ctrl
);
7211 igb_rlpml_set(adapter
);
7214 static int igb_vlan_rx_add_vid(struct net_device
*netdev
,
7215 __be16 proto
, u16 vid
)
7217 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7218 struct e1000_hw
*hw
= &adapter
->hw
;
7219 int pf_id
= adapter
->vfs_allocated_count
;
7221 /* attempt to add filter to vlvf array */
7222 igb_vlvf_set(adapter
, vid
, true, pf_id
);
7224 /* add the filter since PF can receive vlans w/o entry in vlvf */
7225 igb_vfta_set(hw
, vid
, true);
7227 set_bit(vid
, adapter
->active_vlans
);
7232 static int igb_vlan_rx_kill_vid(struct net_device
*netdev
,
7233 __be16 proto
, u16 vid
)
7235 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7236 struct e1000_hw
*hw
= &adapter
->hw
;
7237 int pf_id
= adapter
->vfs_allocated_count
;
7240 /* remove vlan from VLVF table array */
7241 err
= igb_vlvf_set(adapter
, vid
, false, pf_id
);
7243 /* if vid was not present in VLVF just remove it from table */
7245 igb_vfta_set(hw
, vid
, false);
7247 clear_bit(vid
, adapter
->active_vlans
);
7252 static void igb_restore_vlan(struct igb_adapter
*adapter
)
7256 igb_vlan_mode(adapter
->netdev
, adapter
->netdev
->features
);
7258 for_each_set_bit(vid
, adapter
->active_vlans
, VLAN_N_VID
)
7259 igb_vlan_rx_add_vid(adapter
->netdev
, htons(ETH_P_8021Q
), vid
);
7262 int igb_set_spd_dplx(struct igb_adapter
*adapter
, u32 spd
, u8 dplx
)
7264 struct pci_dev
*pdev
= adapter
->pdev
;
7265 struct e1000_mac_info
*mac
= &adapter
->hw
.mac
;
7269 /* Make sure dplx is at most 1 bit and lsb of speed is not set
7270 * for the switch() below to work
7272 if ((spd
& 1) || (dplx
& ~1))
7275 /* Fiber NIC's only allow 1000 gbps Full duplex
7276 * and 100Mbps Full duplex for 100baseFx sfp
7278 if (adapter
->hw
.phy
.media_type
== e1000_media_type_internal_serdes
) {
7279 switch (spd
+ dplx
) {
7280 case SPEED_10
+ DUPLEX_HALF
:
7281 case SPEED_10
+ DUPLEX_FULL
:
7282 case SPEED_100
+ DUPLEX_HALF
:
7289 switch (spd
+ dplx
) {
7290 case SPEED_10
+ DUPLEX_HALF
:
7291 mac
->forced_speed_duplex
= ADVERTISE_10_HALF
;
7293 case SPEED_10
+ DUPLEX_FULL
:
7294 mac
->forced_speed_duplex
= ADVERTISE_10_FULL
;
7296 case SPEED_100
+ DUPLEX_HALF
:
7297 mac
->forced_speed_duplex
= ADVERTISE_100_HALF
;
7299 case SPEED_100
+ DUPLEX_FULL
:
7300 mac
->forced_speed_duplex
= ADVERTISE_100_FULL
;
7302 case SPEED_1000
+ DUPLEX_FULL
:
7304 adapter
->hw
.phy
.autoneg_advertised
= ADVERTISE_1000_FULL
;
7306 case SPEED_1000
+ DUPLEX_HALF
: /* not supported */
7311 /* clear MDI, MDI(-X) override is only allowed when autoneg enabled */
7312 adapter
->hw
.phy
.mdix
= AUTO_ALL_MODES
;
7317 dev_err(&pdev
->dev
, "Unsupported Speed/Duplex configuration\n");
7321 static int __igb_shutdown(struct pci_dev
*pdev
, bool *enable_wake
,
7324 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7325 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7326 struct e1000_hw
*hw
= &adapter
->hw
;
7327 u32 ctrl
, rctl
, status
;
7328 u32 wufc
= runtime
? E1000_WUFC_LNKC
: adapter
->wol
;
7333 netif_device_detach(netdev
);
7335 if (netif_running(netdev
))
7336 __igb_close(netdev
, true);
7338 igb_clear_interrupt_scheme(adapter
);
7341 retval
= pci_save_state(pdev
);
7346 status
= rd32(E1000_STATUS
);
7347 if (status
& E1000_STATUS_LU
)
7348 wufc
&= ~E1000_WUFC_LNKC
;
7351 igb_setup_rctl(adapter
);
7352 igb_set_rx_mode(netdev
);
7354 /* turn on all-multi mode if wake on multicast is enabled */
7355 if (wufc
& E1000_WUFC_MC
) {
7356 rctl
= rd32(E1000_RCTL
);
7357 rctl
|= E1000_RCTL_MPE
;
7358 wr32(E1000_RCTL
, rctl
);
7361 ctrl
= rd32(E1000_CTRL
);
7362 /* advertise wake from D3Cold */
7363 #define E1000_CTRL_ADVD3WUC 0x00100000
7364 /* phy power management enable */
7365 #define E1000_CTRL_EN_PHY_PWR_MGMT 0x00200000
7366 ctrl
|= E1000_CTRL_ADVD3WUC
;
7367 wr32(E1000_CTRL
, ctrl
);
7369 /* Allow time for pending master requests to run */
7370 igb_disable_pcie_master(hw
);
7372 wr32(E1000_WUC
, E1000_WUC_PME_EN
);
7373 wr32(E1000_WUFC
, wufc
);
7376 wr32(E1000_WUFC
, 0);
7379 *enable_wake
= wufc
|| adapter
->en_mng_pt
;
7381 igb_power_down_link(adapter
);
7383 igb_power_up_link(adapter
);
7385 /* Release control of h/w to f/w. If f/w is AMT enabled, this
7386 * would have already happened in close and is redundant.
7388 igb_release_hw_control(adapter
);
7390 pci_disable_device(pdev
);
7396 #ifdef CONFIG_PM_SLEEP
7397 static int igb_suspend(struct device
*dev
)
7401 struct pci_dev
*pdev
= to_pci_dev(dev
);
7403 retval
= __igb_shutdown(pdev
, &wake
, 0);
7408 pci_prepare_to_sleep(pdev
);
7410 pci_wake_from_d3(pdev
, false);
7411 pci_set_power_state(pdev
, PCI_D3hot
);
7416 #endif /* CONFIG_PM_SLEEP */
7418 static int igb_resume(struct device
*dev
)
7420 struct pci_dev
*pdev
= to_pci_dev(dev
);
7421 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7422 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7423 struct e1000_hw
*hw
= &adapter
->hw
;
7426 pci_set_power_state(pdev
, PCI_D0
);
7427 pci_restore_state(pdev
);
7428 pci_save_state(pdev
);
7430 err
= pci_enable_device_mem(pdev
);
7433 "igb: Cannot enable PCI device from suspend\n");
7436 pci_set_master(pdev
);
7438 pci_enable_wake(pdev
, PCI_D3hot
, 0);
7439 pci_enable_wake(pdev
, PCI_D3cold
, 0);
7441 if (igb_init_interrupt_scheme(adapter
, true)) {
7442 dev_err(&pdev
->dev
, "Unable to allocate memory for queues\n");
7448 /* let the f/w know that the h/w is now under the control of the
7451 igb_get_hw_control(adapter
);
7453 wr32(E1000_WUS
, ~0);
7455 if (netdev
->flags
& IFF_UP
) {
7457 err
= __igb_open(netdev
, true);
7463 netif_device_attach(netdev
);
7467 #ifdef CONFIG_PM_RUNTIME
7468 static int igb_runtime_idle(struct device
*dev
)
7470 struct pci_dev
*pdev
= to_pci_dev(dev
);
7471 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7472 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7474 if (!igb_has_link(adapter
))
7475 pm_schedule_suspend(dev
, MSEC_PER_SEC
* 5);
7480 static int igb_runtime_suspend(struct device
*dev
)
7482 struct pci_dev
*pdev
= to_pci_dev(dev
);
7486 retval
= __igb_shutdown(pdev
, &wake
, 1);
7491 pci_prepare_to_sleep(pdev
);
7493 pci_wake_from_d3(pdev
, false);
7494 pci_set_power_state(pdev
, PCI_D3hot
);
7500 static int igb_runtime_resume(struct device
*dev
)
7502 return igb_resume(dev
);
7504 #endif /* CONFIG_PM_RUNTIME */
7507 static void igb_shutdown(struct pci_dev
*pdev
)
7511 __igb_shutdown(pdev
, &wake
, 0);
7513 if (system_state
== SYSTEM_POWER_OFF
) {
7514 pci_wake_from_d3(pdev
, wake
);
7515 pci_set_power_state(pdev
, PCI_D3hot
);
7519 #ifdef CONFIG_PCI_IOV
7520 static int igb_sriov_reinit(struct pci_dev
*dev
)
7522 struct net_device
*netdev
= pci_get_drvdata(dev
);
7523 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7524 struct pci_dev
*pdev
= adapter
->pdev
;
7528 if (netif_running(netdev
))
7531 igb_clear_interrupt_scheme(adapter
);
7533 igb_init_queue_configuration(adapter
);
7535 if (igb_init_interrupt_scheme(adapter
, true)) {
7536 dev_err(&pdev
->dev
, "Unable to allocate memory for queues\n");
7540 if (netif_running(netdev
))
7548 static int igb_pci_disable_sriov(struct pci_dev
*dev
)
7550 int err
= igb_disable_sriov(dev
);
7553 err
= igb_sriov_reinit(dev
);
7558 static int igb_pci_enable_sriov(struct pci_dev
*dev
, int num_vfs
)
7560 int err
= igb_enable_sriov(dev
, num_vfs
);
7565 err
= igb_sriov_reinit(dev
);
7574 static int igb_pci_sriov_configure(struct pci_dev
*dev
, int num_vfs
)
7576 #ifdef CONFIG_PCI_IOV
7578 return igb_pci_disable_sriov(dev
);
7580 return igb_pci_enable_sriov(dev
, num_vfs
);
7585 #ifdef CONFIG_NET_POLL_CONTROLLER
7586 /* Polling 'interrupt' - used by things like netconsole to send skbs
7587 * without having to re-enable interrupts. It's not called while
7588 * the interrupt routine is executing.
7590 static void igb_netpoll(struct net_device
*netdev
)
7592 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7593 struct e1000_hw
*hw
= &adapter
->hw
;
7594 struct igb_q_vector
*q_vector
;
7597 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
7598 q_vector
= adapter
->q_vector
[i
];
7599 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
)
7600 wr32(E1000_EIMC
, q_vector
->eims_value
);
7602 igb_irq_disable(adapter
);
7603 napi_schedule(&q_vector
->napi
);
7606 #endif /* CONFIG_NET_POLL_CONTROLLER */
7609 * igb_io_error_detected - called when PCI error is detected
7610 * @pdev: Pointer to PCI device
7611 * @state: The current pci connection state
7613 * This function is called after a PCI bus error affecting
7614 * this device has been detected.
7616 static pci_ers_result_t
igb_io_error_detected(struct pci_dev
*pdev
,
7617 pci_channel_state_t state
)
7619 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7620 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7622 netif_device_detach(netdev
);
7624 if (state
== pci_channel_io_perm_failure
)
7625 return PCI_ERS_RESULT_DISCONNECT
;
7627 if (netif_running(netdev
))
7629 pci_disable_device(pdev
);
7631 /* Request a slot slot reset. */
7632 return PCI_ERS_RESULT_NEED_RESET
;
7636 * igb_io_slot_reset - called after the pci bus has been reset.
7637 * @pdev: Pointer to PCI device
7639 * Restart the card from scratch, as if from a cold-boot. Implementation
7640 * resembles the first-half of the igb_resume routine.
7642 static pci_ers_result_t
igb_io_slot_reset(struct pci_dev
*pdev
)
7644 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7645 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7646 struct e1000_hw
*hw
= &adapter
->hw
;
7647 pci_ers_result_t result
;
7650 if (pci_enable_device_mem(pdev
)) {
7652 "Cannot re-enable PCI device after reset.\n");
7653 result
= PCI_ERS_RESULT_DISCONNECT
;
7655 pci_set_master(pdev
);
7656 pci_restore_state(pdev
);
7657 pci_save_state(pdev
);
7659 pci_enable_wake(pdev
, PCI_D3hot
, 0);
7660 pci_enable_wake(pdev
, PCI_D3cold
, 0);
7663 wr32(E1000_WUS
, ~0);
7664 result
= PCI_ERS_RESULT_RECOVERED
;
7667 err
= pci_cleanup_aer_uncorrect_error_status(pdev
);
7670 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
7672 /* non-fatal, continue */
7679 * igb_io_resume - called when traffic can start flowing again.
7680 * @pdev: Pointer to PCI device
7682 * This callback is called when the error recovery driver tells us that
7683 * its OK to resume normal operation. Implementation resembles the
7684 * second-half of the igb_resume routine.
7686 static void igb_io_resume(struct pci_dev
*pdev
)
7688 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7689 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7691 if (netif_running(netdev
)) {
7692 if (igb_up(adapter
)) {
7693 dev_err(&pdev
->dev
, "igb_up failed after reset\n");
7698 netif_device_attach(netdev
);
7700 /* let the f/w know that the h/w is now under the control of the
7703 igb_get_hw_control(adapter
);
7706 static void igb_rar_set_qsel(struct igb_adapter
*adapter
, u8
*addr
, u32 index
,
7709 u32 rar_low
, rar_high
;
7710 struct e1000_hw
*hw
= &adapter
->hw
;
7712 /* HW expects these in little endian so we reverse the byte order
7713 * from network order (big endian) to little endian
7715 rar_low
= ((u32
) addr
[0] | ((u32
) addr
[1] << 8) |
7716 ((u32
) addr
[2] << 16) | ((u32
) addr
[3] << 24));
7717 rar_high
= ((u32
) addr
[4] | ((u32
) addr
[5] << 8));
7719 /* Indicate to hardware the Address is Valid. */
7720 rar_high
|= E1000_RAH_AV
;
7722 if (hw
->mac
.type
== e1000_82575
)
7723 rar_high
|= E1000_RAH_POOL_1
* qsel
;
7725 rar_high
|= E1000_RAH_POOL_1
<< qsel
;
7727 wr32(E1000_RAL(index
), rar_low
);
7729 wr32(E1000_RAH(index
), rar_high
);
7733 static int igb_set_vf_mac(struct igb_adapter
*adapter
,
7734 int vf
, unsigned char *mac_addr
)
7736 struct e1000_hw
*hw
= &adapter
->hw
;
7737 /* VF MAC addresses start at end of receive addresses and moves
7738 * towards the first, as a result a collision should not be possible
7740 int rar_entry
= hw
->mac
.rar_entry_count
- (vf
+ 1);
7742 memcpy(adapter
->vf_data
[vf
].vf_mac_addresses
, mac_addr
, ETH_ALEN
);
7744 igb_rar_set_qsel(adapter
, mac_addr
, rar_entry
, vf
);
7749 static int igb_ndo_set_vf_mac(struct net_device
*netdev
, int vf
, u8
*mac
)
7751 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7752 if (!is_valid_ether_addr(mac
) || (vf
>= adapter
->vfs_allocated_count
))
7754 adapter
->vf_data
[vf
].flags
|= IGB_VF_FLAG_PF_SET_MAC
;
7755 dev_info(&adapter
->pdev
->dev
, "setting MAC %pM on VF %d\n", mac
, vf
);
7756 dev_info(&adapter
->pdev
->dev
,
7757 "Reload the VF driver to make this change effective.");
7758 if (test_bit(__IGB_DOWN
, &adapter
->state
)) {
7759 dev_warn(&adapter
->pdev
->dev
,
7760 "The VF MAC address has been set, but the PF device is not up.\n");
7761 dev_warn(&adapter
->pdev
->dev
,
7762 "Bring the PF device up before attempting to use the VF device.\n");
7764 return igb_set_vf_mac(adapter
, vf
, mac
);
7767 static int igb_link_mbps(int internal_link_speed
)
7769 switch (internal_link_speed
) {
7779 static void igb_set_vf_rate_limit(struct e1000_hw
*hw
, int vf
, int tx_rate
,
7786 /* Calculate the rate factor values to set */
7787 rf_int
= link_speed
/ tx_rate
;
7788 rf_dec
= (link_speed
- (rf_int
* tx_rate
));
7789 rf_dec
= (rf_dec
* (1 << E1000_RTTBCNRC_RF_INT_SHIFT
)) /
7792 bcnrc_val
= E1000_RTTBCNRC_RS_ENA
;
7793 bcnrc_val
|= ((rf_int
<< E1000_RTTBCNRC_RF_INT_SHIFT
) &
7794 E1000_RTTBCNRC_RF_INT_MASK
);
7795 bcnrc_val
|= (rf_dec
& E1000_RTTBCNRC_RF_DEC_MASK
);
7800 wr32(E1000_RTTDQSEL
, vf
); /* vf X uses queue X */
7801 /* Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
7802 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported.
7804 wr32(E1000_RTTBCNRM
, 0x14);
7805 wr32(E1000_RTTBCNRC
, bcnrc_val
);
7808 static void igb_check_vf_rate_limit(struct igb_adapter
*adapter
)
7810 int actual_link_speed
, i
;
7811 bool reset_rate
= false;
7813 /* VF TX rate limit was not set or not supported */
7814 if ((adapter
->vf_rate_link_speed
== 0) ||
7815 (adapter
->hw
.mac
.type
!= e1000_82576
))
7818 actual_link_speed
= igb_link_mbps(adapter
->link_speed
);
7819 if (actual_link_speed
!= adapter
->vf_rate_link_speed
) {
7821 adapter
->vf_rate_link_speed
= 0;
7822 dev_info(&adapter
->pdev
->dev
,
7823 "Link speed has been changed. VF Transmit rate is disabled\n");
7826 for (i
= 0; i
< adapter
->vfs_allocated_count
; i
++) {
7828 adapter
->vf_data
[i
].tx_rate
= 0;
7830 igb_set_vf_rate_limit(&adapter
->hw
, i
,
7831 adapter
->vf_data
[i
].tx_rate
,
7836 static int igb_ndo_set_vf_bw(struct net_device
*netdev
, int vf
, int tx_rate
)
7838 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7839 struct e1000_hw
*hw
= &adapter
->hw
;
7840 int actual_link_speed
;
7842 if (hw
->mac
.type
!= e1000_82576
)
7845 actual_link_speed
= igb_link_mbps(adapter
->link_speed
);
7846 if ((vf
>= adapter
->vfs_allocated_count
) ||
7847 (!(rd32(E1000_STATUS
) & E1000_STATUS_LU
)) ||
7848 (tx_rate
< 0) || (tx_rate
> actual_link_speed
))
7851 adapter
->vf_rate_link_speed
= actual_link_speed
;
7852 adapter
->vf_data
[vf
].tx_rate
= (u16
)tx_rate
;
7853 igb_set_vf_rate_limit(hw
, vf
, tx_rate
, actual_link_speed
);
7858 static int igb_ndo_set_vf_spoofchk(struct net_device
*netdev
, int vf
,
7861 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7862 struct e1000_hw
*hw
= &adapter
->hw
;
7863 u32 reg_val
, reg_offset
;
7865 if (!adapter
->vfs_allocated_count
)
7868 if (vf
>= adapter
->vfs_allocated_count
)
7871 reg_offset
= (hw
->mac
.type
== e1000_82576
) ? E1000_DTXSWC
: E1000_TXSWC
;
7872 reg_val
= rd32(reg_offset
);
7874 reg_val
|= ((1 << vf
) |
7875 (1 << (vf
+ E1000_DTXSWC_VLAN_SPOOF_SHIFT
)));
7877 reg_val
&= ~((1 << vf
) |
7878 (1 << (vf
+ E1000_DTXSWC_VLAN_SPOOF_SHIFT
)));
7879 wr32(reg_offset
, reg_val
);
7881 adapter
->vf_data
[vf
].spoofchk_enabled
= setting
;
7882 return E1000_SUCCESS
;
7885 static int igb_ndo_get_vf_config(struct net_device
*netdev
,
7886 int vf
, struct ifla_vf_info
*ivi
)
7888 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7889 if (vf
>= adapter
->vfs_allocated_count
)
7892 memcpy(&ivi
->mac
, adapter
->vf_data
[vf
].vf_mac_addresses
, ETH_ALEN
);
7893 ivi
->tx_rate
= adapter
->vf_data
[vf
].tx_rate
;
7894 ivi
->vlan
= adapter
->vf_data
[vf
].pf_vlan
;
7895 ivi
->qos
= adapter
->vf_data
[vf
].pf_qos
;
7896 ivi
->spoofchk
= adapter
->vf_data
[vf
].spoofchk_enabled
;
7900 static void igb_vmm_control(struct igb_adapter
*adapter
)
7902 struct e1000_hw
*hw
= &adapter
->hw
;
7905 switch (hw
->mac
.type
) {
7911 /* replication is not supported for 82575 */
7914 /* notify HW that the MAC is adding vlan tags */
7915 reg
= rd32(E1000_DTXCTL
);
7916 reg
|= E1000_DTXCTL_VLAN_ADDED
;
7917 wr32(E1000_DTXCTL
, reg
);
7919 /* enable replication vlan tag stripping */
7920 reg
= rd32(E1000_RPLOLR
);
7921 reg
|= E1000_RPLOLR_STRVLAN
;
7922 wr32(E1000_RPLOLR
, reg
);
7924 /* none of the above registers are supported by i350 */
7928 if (adapter
->vfs_allocated_count
) {
7929 igb_vmdq_set_loopback_pf(hw
, true);
7930 igb_vmdq_set_replication_pf(hw
, true);
7931 igb_vmdq_set_anti_spoofing_pf(hw
, true,
7932 adapter
->vfs_allocated_count
);
7934 igb_vmdq_set_loopback_pf(hw
, false);
7935 igb_vmdq_set_replication_pf(hw
, false);
7939 static void igb_init_dmac(struct igb_adapter
*adapter
, u32 pba
)
7941 struct e1000_hw
*hw
= &adapter
->hw
;
7945 if (hw
->mac
.type
> e1000_82580
) {
7946 if (adapter
->flags
& IGB_FLAG_DMAC
) {
7949 /* force threshold to 0. */
7950 wr32(E1000_DMCTXTH
, 0);
7952 /* DMA Coalescing high water mark needs to be greater
7953 * than the Rx threshold. Set hwm to PBA - max frame
7954 * size in 16B units, capping it at PBA - 6KB.
7956 hwm
= 64 * pba
- adapter
->max_frame_size
/ 16;
7957 if (hwm
< 64 * (pba
- 6))
7958 hwm
= 64 * (pba
- 6);
7959 reg
= rd32(E1000_FCRTC
);
7960 reg
&= ~E1000_FCRTC_RTH_COAL_MASK
;
7961 reg
|= ((hwm
<< E1000_FCRTC_RTH_COAL_SHIFT
)
7962 & E1000_FCRTC_RTH_COAL_MASK
);
7963 wr32(E1000_FCRTC
, reg
);
7965 /* Set the DMA Coalescing Rx threshold to PBA - 2 * max
7966 * frame size, capping it at PBA - 10KB.
7968 dmac_thr
= pba
- adapter
->max_frame_size
/ 512;
7969 if (dmac_thr
< pba
- 10)
7970 dmac_thr
= pba
- 10;
7971 reg
= rd32(E1000_DMACR
);
7972 reg
&= ~E1000_DMACR_DMACTHR_MASK
;
7973 reg
|= ((dmac_thr
<< E1000_DMACR_DMACTHR_SHIFT
)
7974 & E1000_DMACR_DMACTHR_MASK
);
7976 /* transition to L0x or L1 if available..*/
7977 reg
|= (E1000_DMACR_DMAC_EN
| E1000_DMACR_DMAC_LX_MASK
);
7979 /* watchdog timer= +-1000 usec in 32usec intervals */
7982 /* Disable BMC-to-OS Watchdog Enable */
7983 if (hw
->mac
.type
!= e1000_i354
)
7984 reg
&= ~E1000_DMACR_DC_BMC2OSW_EN
;
7986 wr32(E1000_DMACR
, reg
);
7988 /* no lower threshold to disable
7989 * coalescing(smart fifb)-UTRESH=0
7991 wr32(E1000_DMCRTRH
, 0);
7993 reg
= (IGB_DMCTLX_DCFLUSH_DIS
| 0x4);
7995 wr32(E1000_DMCTLX
, reg
);
7997 /* free space in tx packet buffer to wake from
8000 wr32(E1000_DMCTXTH
, (IGB_MIN_TXPBSIZE
-
8001 (IGB_TX_BUF_4096
+ adapter
->max_frame_size
)) >> 6);
8003 /* make low power state decision controlled
8006 reg
= rd32(E1000_PCIEMISC
);
8007 reg
&= ~E1000_PCIEMISC_LX_DECISION
;
8008 wr32(E1000_PCIEMISC
, reg
);
8009 } /* endif adapter->dmac is not disabled */
8010 } else if (hw
->mac
.type
== e1000_82580
) {
8011 u32 reg
= rd32(E1000_PCIEMISC
);
8012 wr32(E1000_PCIEMISC
, reg
& ~E1000_PCIEMISC_LX_DECISION
);
8013 wr32(E1000_DMACR
, 0);
8018 * igb_read_i2c_byte - Reads 8 bit word over I2C
8019 * @hw: pointer to hardware structure
8020 * @byte_offset: byte offset to read
8021 * @dev_addr: device address
8024 * Performs byte read operation over I2C interface at
8025 * a specified device address.
8027 s32
igb_read_i2c_byte(struct e1000_hw
*hw
, u8 byte_offset
,
8028 u8 dev_addr
, u8
*data
)
8030 struct igb_adapter
*adapter
= container_of(hw
, struct igb_adapter
, hw
);
8031 struct i2c_client
*this_client
= adapter
->i2c_client
;
8036 return E1000_ERR_I2C
;
8038 swfw_mask
= E1000_SWFW_PHY0_SM
;
8040 if (hw
->mac
.ops
.acquire_swfw_sync(hw
, swfw_mask
)
8042 return E1000_ERR_SWFW_SYNC
;
8044 status
= i2c_smbus_read_byte_data(this_client
, byte_offset
);
8045 hw
->mac
.ops
.release_swfw_sync(hw
, swfw_mask
);
8048 return E1000_ERR_I2C
;
8051 return E1000_SUCCESS
;
8056 * igb_write_i2c_byte - Writes 8 bit word over I2C
8057 * @hw: pointer to hardware structure
8058 * @byte_offset: byte offset to write
8059 * @dev_addr: device address
8060 * @data: value to write
8062 * Performs byte write operation over I2C interface at
8063 * a specified device address.
8065 s32
igb_write_i2c_byte(struct e1000_hw
*hw
, u8 byte_offset
,
8066 u8 dev_addr
, u8 data
)
8068 struct igb_adapter
*adapter
= container_of(hw
, struct igb_adapter
, hw
);
8069 struct i2c_client
*this_client
= adapter
->i2c_client
;
8071 u16 swfw_mask
= E1000_SWFW_PHY0_SM
;
8074 return E1000_ERR_I2C
;
8076 if (hw
->mac
.ops
.acquire_swfw_sync(hw
, swfw_mask
) != E1000_SUCCESS
)
8077 return E1000_ERR_SWFW_SYNC
;
8078 status
= i2c_smbus_write_byte_data(this_client
, byte_offset
, data
);
8079 hw
->mac
.ops
.release_swfw_sync(hw
, swfw_mask
);
8082 return E1000_ERR_I2C
;
8084 return E1000_SUCCESS
;
8088 int igb_reinit_queues(struct igb_adapter
*adapter
)
8090 struct net_device
*netdev
= adapter
->netdev
;
8091 struct pci_dev
*pdev
= adapter
->pdev
;
8094 if (netif_running(netdev
))
8097 igb_reset_interrupt_capability(adapter
);
8099 if (igb_init_interrupt_scheme(adapter
, true)) {
8100 dev_err(&pdev
->dev
, "Unable to allocate memory for queues\n");
8104 if (netif_running(netdev
))
8105 err
= igb_open(netdev
);