1 /* Intel(R) Gigabit Ethernet Linux driver
2 * Copyright(c) 2007-2014 Intel Corporation.
4 * This program is free software; you can redistribute it and/or modify it
5 * under the terms and conditions of the GNU General Public License,
6 * version 2, as published by the Free Software Foundation.
8 * This program is distributed in the hope it will be useful, but WITHOUT
9 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
13 * You should have received a copy of the GNU General Public License along with
14 * this program; if not, see <http://www.gnu.org/licenses/>.
16 * The full GNU General Public License is included in this distribution in
17 * the file called "COPYING".
19 * Contact Information:
20 * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
21 * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
24 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
26 #include <linux/module.h>
27 #include <linux/types.h>
28 #include <linux/init.h>
29 #include <linux/bitops.h>
30 #include <linux/vmalloc.h>
31 #include <linux/pagemap.h>
32 #include <linux/netdevice.h>
33 #include <linux/ipv6.h>
34 #include <linux/slab.h>
35 #include <net/checksum.h>
36 #include <net/ip6_checksum.h>
37 #include <linux/net_tstamp.h>
38 #include <linux/mii.h>
39 #include <linux/ethtool.h>
41 #include <linux/if_vlan.h>
42 #include <linux/pci.h>
43 #include <linux/pci-aspm.h>
44 #include <linux/delay.h>
45 #include <linux/interrupt.h>
47 #include <linux/tcp.h>
48 #include <linux/sctp.h>
49 #include <linux/if_ether.h>
50 #include <linux/aer.h>
51 #include <linux/prefetch.h>
52 #include <linux/pm_runtime.h>
54 #include <linux/dca.h>
56 #include <linux/i2c.h>
62 #define DRV_VERSION __stringify(MAJ) "." __stringify(MIN) "." \
63 __stringify(BUILD) "-k"
64 char igb_driver_name
[] = "igb";
65 char igb_driver_version
[] = DRV_VERSION
;
66 static const char igb_driver_string
[] =
67 "Intel(R) Gigabit Ethernet Network Driver";
68 static const char igb_copyright
[] =
69 "Copyright (c) 2007-2014 Intel Corporation.";
71 static const struct e1000_info
*igb_info_tbl
[] = {
72 [board_82575
] = &e1000_82575_info
,
75 static const struct pci_device_id igb_pci_tbl
[] = {
76 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I354_BACKPLANE_1GBPS
) },
77 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I354_SGMII
) },
78 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS
) },
79 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I211_COPPER
), board_82575
},
80 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_COPPER
), board_82575
},
81 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_FIBER
), board_82575
},
82 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_SERDES
), board_82575
},
83 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_SGMII
), board_82575
},
84 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_COPPER_FLASHLESS
), board_82575
},
85 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I210_SERDES_FLASHLESS
), board_82575
},
86 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I350_COPPER
), board_82575
},
87 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I350_FIBER
), board_82575
},
88 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I350_SERDES
), board_82575
},
89 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_I350_SGMII
), board_82575
},
90 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_COPPER
), board_82575
},
91 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_FIBER
), board_82575
},
92 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_QUAD_FIBER
), board_82575
},
93 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_SERDES
), board_82575
},
94 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_SGMII
), board_82575
},
95 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82580_COPPER_DUAL
), board_82575
},
96 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_DH89XXCC_SGMII
), board_82575
},
97 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_DH89XXCC_SERDES
), board_82575
},
98 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_DH89XXCC_BACKPLANE
), board_82575
},
99 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_DH89XXCC_SFP
), board_82575
},
100 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576
), board_82575
},
101 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_NS
), board_82575
},
102 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_NS_SERDES
), board_82575
},
103 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_FIBER
), board_82575
},
104 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_SERDES
), board_82575
},
105 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_SERDES_QUAD
), board_82575
},
106 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_QUAD_COPPER_ET2
), board_82575
},
107 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82576_QUAD_COPPER
), board_82575
},
108 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82575EB_COPPER
), board_82575
},
109 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82575EB_FIBER_SERDES
), board_82575
},
110 { PCI_VDEVICE(INTEL
, E1000_DEV_ID_82575GB_QUAD_COPPER
), board_82575
},
111 /* required last entry */
115 MODULE_DEVICE_TABLE(pci
, igb_pci_tbl
);
117 static int igb_setup_all_tx_resources(struct igb_adapter
*);
118 static int igb_setup_all_rx_resources(struct igb_adapter
*);
119 static void igb_free_all_tx_resources(struct igb_adapter
*);
120 static void igb_free_all_rx_resources(struct igb_adapter
*);
121 static void igb_setup_mrqc(struct igb_adapter
*);
122 static int igb_probe(struct pci_dev
*, const struct pci_device_id
*);
123 static void igb_remove(struct pci_dev
*pdev
);
124 static int igb_sw_init(struct igb_adapter
*);
125 static int igb_open(struct net_device
*);
126 static int igb_close(struct net_device
*);
127 static void igb_configure(struct igb_adapter
*);
128 static void igb_configure_tx(struct igb_adapter
*);
129 static void igb_configure_rx(struct igb_adapter
*);
130 static void igb_clean_all_tx_rings(struct igb_adapter
*);
131 static void igb_clean_all_rx_rings(struct igb_adapter
*);
132 static void igb_clean_tx_ring(struct igb_ring
*);
133 static void igb_clean_rx_ring(struct igb_ring
*);
134 static void igb_set_rx_mode(struct net_device
*);
135 static void igb_update_phy_info(unsigned long);
136 static void igb_watchdog(unsigned long);
137 static void igb_watchdog_task(struct work_struct
*);
138 static netdev_tx_t
igb_xmit_frame(struct sk_buff
*skb
, struct net_device
*);
139 static struct rtnl_link_stats64
*igb_get_stats64(struct net_device
*dev
,
140 struct rtnl_link_stats64
*stats
);
141 static int igb_change_mtu(struct net_device
*, int);
142 static int igb_set_mac(struct net_device
*, void *);
143 static void igb_set_uta(struct igb_adapter
*adapter
);
144 static irqreturn_t
igb_intr(int irq
, void *);
145 static irqreturn_t
igb_intr_msi(int irq
, void *);
146 static irqreturn_t
igb_msix_other(int irq
, void *);
147 static irqreturn_t
igb_msix_ring(int irq
, void *);
148 #ifdef CONFIG_IGB_DCA
149 static void igb_update_dca(struct igb_q_vector
*);
150 static void igb_setup_dca(struct igb_adapter
*);
151 #endif /* CONFIG_IGB_DCA */
152 static int igb_poll(struct napi_struct
*, int);
153 static bool igb_clean_tx_irq(struct igb_q_vector
*);
154 static int igb_clean_rx_irq(struct igb_q_vector
*, int);
155 static int igb_ioctl(struct net_device
*, struct ifreq
*, int cmd
);
156 static void igb_tx_timeout(struct net_device
*);
157 static void igb_reset_task(struct work_struct
*);
158 static void igb_vlan_mode(struct net_device
*netdev
,
159 netdev_features_t features
);
160 static int igb_vlan_rx_add_vid(struct net_device
*, __be16
, u16
);
161 static int igb_vlan_rx_kill_vid(struct net_device
*, __be16
, u16
);
162 static void igb_restore_vlan(struct igb_adapter
*);
163 static void igb_rar_set_qsel(struct igb_adapter
*, u8
*, u32
, u8
);
164 static void igb_ping_all_vfs(struct igb_adapter
*);
165 static void igb_msg_task(struct igb_adapter
*);
166 static void igb_vmm_control(struct igb_adapter
*);
167 static int igb_set_vf_mac(struct igb_adapter
*, int, unsigned char *);
168 static void igb_restore_vf_multicasts(struct igb_adapter
*adapter
);
169 static int igb_ndo_set_vf_mac(struct net_device
*netdev
, int vf
, u8
*mac
);
170 static int igb_ndo_set_vf_vlan(struct net_device
*netdev
,
171 int vf
, u16 vlan
, u8 qos
);
172 static int igb_ndo_set_vf_bw(struct net_device
*, int, int, int);
173 static int igb_ndo_set_vf_spoofchk(struct net_device
*netdev
, int vf
,
175 static int igb_ndo_get_vf_config(struct net_device
*netdev
, int vf
,
176 struct ifla_vf_info
*ivi
);
177 static void igb_check_vf_rate_limit(struct igb_adapter
*);
179 #ifdef CONFIG_PCI_IOV
180 static int igb_vf_configure(struct igb_adapter
*adapter
, int vf
);
181 static int igb_pci_enable_sriov(struct pci_dev
*dev
, int num_vfs
);
182 static int igb_disable_sriov(struct pci_dev
*dev
);
183 static int igb_pci_disable_sriov(struct pci_dev
*dev
);
187 #ifdef CONFIG_PM_SLEEP
188 static int igb_suspend(struct device
*);
190 static int igb_resume(struct device
*);
191 static int igb_runtime_suspend(struct device
*dev
);
192 static int igb_runtime_resume(struct device
*dev
);
193 static int igb_runtime_idle(struct device
*dev
);
194 static const struct dev_pm_ops igb_pm_ops
= {
195 SET_SYSTEM_SLEEP_PM_OPS(igb_suspend
, igb_resume
)
196 SET_RUNTIME_PM_OPS(igb_runtime_suspend
, igb_runtime_resume
,
200 static void igb_shutdown(struct pci_dev
*);
201 static int igb_pci_sriov_configure(struct pci_dev
*dev
, int num_vfs
);
202 #ifdef CONFIG_IGB_DCA
203 static int igb_notify_dca(struct notifier_block
*, unsigned long, void *);
204 static struct notifier_block dca_notifier
= {
205 .notifier_call
= igb_notify_dca
,
210 #ifdef CONFIG_NET_POLL_CONTROLLER
211 /* for netdump / net console */
212 static void igb_netpoll(struct net_device
*);
214 #ifdef CONFIG_PCI_IOV
215 static unsigned int max_vfs
;
216 module_param(max_vfs
, uint
, 0);
217 MODULE_PARM_DESC(max_vfs
, "Maximum number of virtual functions to allocate per physical function");
218 #endif /* CONFIG_PCI_IOV */
220 static pci_ers_result_t
igb_io_error_detected(struct pci_dev
*,
221 pci_channel_state_t
);
222 static pci_ers_result_t
igb_io_slot_reset(struct pci_dev
*);
223 static void igb_io_resume(struct pci_dev
*);
225 static const struct pci_error_handlers igb_err_handler
= {
226 .error_detected
= igb_io_error_detected
,
227 .slot_reset
= igb_io_slot_reset
,
228 .resume
= igb_io_resume
,
231 static void igb_init_dmac(struct igb_adapter
*adapter
, u32 pba
);
233 static struct pci_driver igb_driver
= {
234 .name
= igb_driver_name
,
235 .id_table
= igb_pci_tbl
,
237 .remove
= igb_remove
,
239 .driver
.pm
= &igb_pm_ops
,
241 .shutdown
= igb_shutdown
,
242 .sriov_configure
= igb_pci_sriov_configure
,
243 .err_handler
= &igb_err_handler
246 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
247 MODULE_DESCRIPTION("Intel(R) Gigabit Ethernet Network Driver");
248 MODULE_LICENSE("GPL");
249 MODULE_VERSION(DRV_VERSION
);
251 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
252 static int debug
= -1;
253 module_param(debug
, int, 0);
254 MODULE_PARM_DESC(debug
, "Debug level (0=none,...,16=all)");
256 struct igb_reg_info
{
261 static const struct igb_reg_info igb_reg_info_tbl
[] = {
263 /* General Registers */
264 {E1000_CTRL
, "CTRL"},
265 {E1000_STATUS
, "STATUS"},
266 {E1000_CTRL_EXT
, "CTRL_EXT"},
268 /* Interrupt Registers */
272 {E1000_RCTL
, "RCTL"},
273 {E1000_RDLEN(0), "RDLEN"},
274 {E1000_RDH(0), "RDH"},
275 {E1000_RDT(0), "RDT"},
276 {E1000_RXDCTL(0), "RXDCTL"},
277 {E1000_RDBAL(0), "RDBAL"},
278 {E1000_RDBAH(0), "RDBAH"},
281 {E1000_TCTL
, "TCTL"},
282 {E1000_TDBAL(0), "TDBAL"},
283 {E1000_TDBAH(0), "TDBAH"},
284 {E1000_TDLEN(0), "TDLEN"},
285 {E1000_TDH(0), "TDH"},
286 {E1000_TDT(0), "TDT"},
287 {E1000_TXDCTL(0), "TXDCTL"},
288 {E1000_TDFH
, "TDFH"},
289 {E1000_TDFT
, "TDFT"},
290 {E1000_TDFHS
, "TDFHS"},
291 {E1000_TDFPC
, "TDFPC"},
293 /* List Terminator */
297 /* igb_regdump - register printout routine */
298 static void igb_regdump(struct e1000_hw
*hw
, struct igb_reg_info
*reginfo
)
304 switch (reginfo
->ofs
) {
306 for (n
= 0; n
< 4; n
++)
307 regs
[n
] = rd32(E1000_RDLEN(n
));
310 for (n
= 0; n
< 4; n
++)
311 regs
[n
] = rd32(E1000_RDH(n
));
314 for (n
= 0; n
< 4; n
++)
315 regs
[n
] = rd32(E1000_RDT(n
));
317 case E1000_RXDCTL(0):
318 for (n
= 0; n
< 4; n
++)
319 regs
[n
] = rd32(E1000_RXDCTL(n
));
322 for (n
= 0; n
< 4; n
++)
323 regs
[n
] = rd32(E1000_RDBAL(n
));
326 for (n
= 0; n
< 4; n
++)
327 regs
[n
] = rd32(E1000_RDBAH(n
));
330 for (n
= 0; n
< 4; n
++)
331 regs
[n
] = rd32(E1000_RDBAL(n
));
334 for (n
= 0; n
< 4; n
++)
335 regs
[n
] = rd32(E1000_TDBAH(n
));
338 for (n
= 0; n
< 4; n
++)
339 regs
[n
] = rd32(E1000_TDLEN(n
));
342 for (n
= 0; n
< 4; n
++)
343 regs
[n
] = rd32(E1000_TDH(n
));
346 for (n
= 0; n
< 4; n
++)
347 regs
[n
] = rd32(E1000_TDT(n
));
349 case E1000_TXDCTL(0):
350 for (n
= 0; n
< 4; n
++)
351 regs
[n
] = rd32(E1000_TXDCTL(n
));
354 pr_info("%-15s %08x\n", reginfo
->name
, rd32(reginfo
->ofs
));
358 snprintf(rname
, 16, "%s%s", reginfo
->name
, "[0-3]");
359 pr_info("%-15s %08x %08x %08x %08x\n", rname
, regs
[0], regs
[1],
363 /* igb_dump - Print registers, Tx-rings and Rx-rings */
364 static void igb_dump(struct igb_adapter
*adapter
)
366 struct net_device
*netdev
= adapter
->netdev
;
367 struct e1000_hw
*hw
= &adapter
->hw
;
368 struct igb_reg_info
*reginfo
;
369 struct igb_ring
*tx_ring
;
370 union e1000_adv_tx_desc
*tx_desc
;
371 struct my_u0
{ u64 a
; u64 b
; } *u0
;
372 struct igb_ring
*rx_ring
;
373 union e1000_adv_rx_desc
*rx_desc
;
377 if (!netif_msg_hw(adapter
))
380 /* Print netdevice Info */
382 dev_info(&adapter
->pdev
->dev
, "Net device Info\n");
383 pr_info("Device Name state trans_start last_rx\n");
384 pr_info("%-15s %016lX %016lX %016lX\n", netdev
->name
,
385 netdev
->state
, netdev
->trans_start
, netdev
->last_rx
);
388 /* Print Registers */
389 dev_info(&adapter
->pdev
->dev
, "Register Dump\n");
390 pr_info(" Register Name Value\n");
391 for (reginfo
= (struct igb_reg_info
*)igb_reg_info_tbl
;
392 reginfo
->name
; reginfo
++) {
393 igb_regdump(hw
, reginfo
);
396 /* Print TX Ring Summary */
397 if (!netdev
|| !netif_running(netdev
))
400 dev_info(&adapter
->pdev
->dev
, "TX Rings Summary\n");
401 pr_info("Queue [NTU] [NTC] [bi(ntc)->dma ] leng ntw timestamp\n");
402 for (n
= 0; n
< adapter
->num_tx_queues
; n
++) {
403 struct igb_tx_buffer
*buffer_info
;
404 tx_ring
= adapter
->tx_ring
[n
];
405 buffer_info
= &tx_ring
->tx_buffer_info
[tx_ring
->next_to_clean
];
406 pr_info(" %5d %5X %5X %016llX %04X %p %016llX\n",
407 n
, tx_ring
->next_to_use
, tx_ring
->next_to_clean
,
408 (u64
)dma_unmap_addr(buffer_info
, dma
),
409 dma_unmap_len(buffer_info
, len
),
410 buffer_info
->next_to_watch
,
411 (u64
)buffer_info
->time_stamp
);
415 if (!netif_msg_tx_done(adapter
))
416 goto rx_ring_summary
;
418 dev_info(&adapter
->pdev
->dev
, "TX Rings Dump\n");
420 /* Transmit Descriptor Formats
422 * Advanced Transmit Descriptor
423 * +--------------------------------------------------------------+
424 * 0 | Buffer Address [63:0] |
425 * +--------------------------------------------------------------+
426 * 8 | PAYLEN | PORTS |CC|IDX | STA | DCMD |DTYP|MAC|RSV| DTALEN |
427 * +--------------------------------------------------------------+
428 * 63 46 45 40 39 38 36 35 32 31 24 15 0
431 for (n
= 0; n
< adapter
->num_tx_queues
; n
++) {
432 tx_ring
= adapter
->tx_ring
[n
];
433 pr_info("------------------------------------\n");
434 pr_info("TX QUEUE INDEX = %d\n", tx_ring
->queue_index
);
435 pr_info("------------------------------------\n");
436 pr_info("T [desc] [address 63:0 ] [PlPOCIStDDM Ln] [bi->dma ] leng ntw timestamp bi->skb\n");
438 for (i
= 0; tx_ring
->desc
&& (i
< tx_ring
->count
); i
++) {
439 const char *next_desc
;
440 struct igb_tx_buffer
*buffer_info
;
441 tx_desc
= IGB_TX_DESC(tx_ring
, i
);
442 buffer_info
= &tx_ring
->tx_buffer_info
[i
];
443 u0
= (struct my_u0
*)tx_desc
;
444 if (i
== tx_ring
->next_to_use
&&
445 i
== tx_ring
->next_to_clean
)
446 next_desc
= " NTC/U";
447 else if (i
== tx_ring
->next_to_use
)
449 else if (i
== tx_ring
->next_to_clean
)
454 pr_info("T [0x%03X] %016llX %016llX %016llX %04X %p %016llX %p%s\n",
455 i
, le64_to_cpu(u0
->a
),
457 (u64
)dma_unmap_addr(buffer_info
, dma
),
458 dma_unmap_len(buffer_info
, len
),
459 buffer_info
->next_to_watch
,
460 (u64
)buffer_info
->time_stamp
,
461 buffer_info
->skb
, next_desc
);
463 if (netif_msg_pktdata(adapter
) && buffer_info
->skb
)
464 print_hex_dump(KERN_INFO
, "",
466 16, 1, buffer_info
->skb
->data
,
467 dma_unmap_len(buffer_info
, len
),
472 /* Print RX Rings Summary */
474 dev_info(&adapter
->pdev
->dev
, "RX Rings Summary\n");
475 pr_info("Queue [NTU] [NTC]\n");
476 for (n
= 0; n
< adapter
->num_rx_queues
; n
++) {
477 rx_ring
= adapter
->rx_ring
[n
];
478 pr_info(" %5d %5X %5X\n",
479 n
, rx_ring
->next_to_use
, rx_ring
->next_to_clean
);
483 if (!netif_msg_rx_status(adapter
))
486 dev_info(&adapter
->pdev
->dev
, "RX Rings Dump\n");
488 /* Advanced Receive Descriptor (Read) Format
490 * +-----------------------------------------------------+
491 * 0 | Packet Buffer Address [63:1] |A0/NSE|
492 * +----------------------------------------------+------+
493 * 8 | Header Buffer Address [63:1] | DD |
494 * +-----------------------------------------------------+
497 * Advanced Receive Descriptor (Write-Back) Format
499 * 63 48 47 32 31 30 21 20 17 16 4 3 0
500 * +------------------------------------------------------+
501 * 0 | Packet IP |SPH| HDR_LEN | RSV|Packet| RSS |
502 * | Checksum Ident | | | | Type | Type |
503 * +------------------------------------------------------+
504 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
505 * +------------------------------------------------------+
506 * 63 48 47 32 31 20 19 0
509 for (n
= 0; n
< adapter
->num_rx_queues
; n
++) {
510 rx_ring
= adapter
->rx_ring
[n
];
511 pr_info("------------------------------------\n");
512 pr_info("RX QUEUE INDEX = %d\n", rx_ring
->queue_index
);
513 pr_info("------------------------------------\n");
514 pr_info("R [desc] [ PktBuf A0] [ HeadBuf DD] [bi->dma ] [bi->skb] <-- Adv Rx Read format\n");
515 pr_info("RWB[desc] [PcsmIpSHl PtRs] [vl er S cks ln] ---------------- [bi->skb] <-- Adv Rx Write-Back format\n");
517 for (i
= 0; i
< rx_ring
->count
; i
++) {
518 const char *next_desc
;
519 struct igb_rx_buffer
*buffer_info
;
520 buffer_info
= &rx_ring
->rx_buffer_info
[i
];
521 rx_desc
= IGB_RX_DESC(rx_ring
, i
);
522 u0
= (struct my_u0
*)rx_desc
;
523 staterr
= le32_to_cpu(rx_desc
->wb
.upper
.status_error
);
525 if (i
== rx_ring
->next_to_use
)
527 else if (i
== rx_ring
->next_to_clean
)
532 if (staterr
& E1000_RXD_STAT_DD
) {
533 /* Descriptor Done */
534 pr_info("%s[0x%03X] %016llX %016llX ---------------- %s\n",
540 pr_info("%s[0x%03X] %016llX %016llX %016llX %s\n",
544 (u64
)buffer_info
->dma
,
547 if (netif_msg_pktdata(adapter
) &&
548 buffer_info
->dma
&& buffer_info
->page
) {
549 print_hex_dump(KERN_INFO
, "",
552 page_address(buffer_info
->page
) +
553 buffer_info
->page_offset
,
565 * igb_get_i2c_data - Reads the I2C SDA data bit
566 * @hw: pointer to hardware structure
567 * @i2cctl: Current value of I2CCTL register
569 * Returns the I2C data bit value
571 static int igb_get_i2c_data(void *data
)
573 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
574 struct e1000_hw
*hw
= &adapter
->hw
;
575 s32 i2cctl
= rd32(E1000_I2CPARAMS
);
577 return !!(i2cctl
& E1000_I2C_DATA_IN
);
581 * igb_set_i2c_data - Sets the I2C data bit
582 * @data: pointer to hardware structure
583 * @state: I2C data value (0 or 1) to set
585 * Sets the I2C data bit
587 static void igb_set_i2c_data(void *data
, int state
)
589 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
590 struct e1000_hw
*hw
= &adapter
->hw
;
591 s32 i2cctl
= rd32(E1000_I2CPARAMS
);
594 i2cctl
|= E1000_I2C_DATA_OUT
;
596 i2cctl
&= ~E1000_I2C_DATA_OUT
;
598 i2cctl
&= ~E1000_I2C_DATA_OE_N
;
599 i2cctl
|= E1000_I2C_CLK_OE_N
;
600 wr32(E1000_I2CPARAMS
, i2cctl
);
606 * igb_set_i2c_clk - Sets the I2C SCL clock
607 * @data: pointer to hardware structure
608 * @state: state to set clock
610 * Sets the I2C clock line to state
612 static void igb_set_i2c_clk(void *data
, int state
)
614 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
615 struct e1000_hw
*hw
= &adapter
->hw
;
616 s32 i2cctl
= rd32(E1000_I2CPARAMS
);
619 i2cctl
|= E1000_I2C_CLK_OUT
;
620 i2cctl
&= ~E1000_I2C_CLK_OE_N
;
622 i2cctl
&= ~E1000_I2C_CLK_OUT
;
623 i2cctl
&= ~E1000_I2C_CLK_OE_N
;
625 wr32(E1000_I2CPARAMS
, i2cctl
);
630 * igb_get_i2c_clk - Gets the I2C SCL clock state
631 * @data: pointer to hardware structure
633 * Gets the I2C clock state
635 static int igb_get_i2c_clk(void *data
)
637 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
638 struct e1000_hw
*hw
= &adapter
->hw
;
639 s32 i2cctl
= rd32(E1000_I2CPARAMS
);
641 return !!(i2cctl
& E1000_I2C_CLK_IN
);
644 static const struct i2c_algo_bit_data igb_i2c_algo
= {
645 .setsda
= igb_set_i2c_data
,
646 .setscl
= igb_set_i2c_clk
,
647 .getsda
= igb_get_i2c_data
,
648 .getscl
= igb_get_i2c_clk
,
654 * igb_get_hw_dev - return device
655 * @hw: pointer to hardware structure
657 * used by hardware layer to print debugging information
659 struct net_device
*igb_get_hw_dev(struct e1000_hw
*hw
)
661 struct igb_adapter
*adapter
= hw
->back
;
662 return adapter
->netdev
;
666 * igb_init_module - Driver Registration Routine
668 * igb_init_module is the first routine called when the driver is
669 * loaded. All it does is register with the PCI subsystem.
671 static int __init
igb_init_module(void)
675 pr_info("%s - version %s\n",
676 igb_driver_string
, igb_driver_version
);
677 pr_info("%s\n", igb_copyright
);
679 #ifdef CONFIG_IGB_DCA
680 dca_register_notify(&dca_notifier
);
682 ret
= pci_register_driver(&igb_driver
);
686 module_init(igb_init_module
);
689 * igb_exit_module - Driver Exit Cleanup Routine
691 * igb_exit_module is called just before the driver is removed
694 static void __exit
igb_exit_module(void)
696 #ifdef CONFIG_IGB_DCA
697 dca_unregister_notify(&dca_notifier
);
699 pci_unregister_driver(&igb_driver
);
702 module_exit(igb_exit_module
);
704 #define Q_IDX_82576(i) (((i & 0x1) << 3) + (i >> 1))
706 * igb_cache_ring_register - Descriptor ring to register mapping
707 * @adapter: board private structure to initialize
709 * Once we know the feature-set enabled for the device, we'll cache
710 * the register offset the descriptor ring is assigned to.
712 static void igb_cache_ring_register(struct igb_adapter
*adapter
)
715 u32 rbase_offset
= adapter
->vfs_allocated_count
;
717 switch (adapter
->hw
.mac
.type
) {
719 /* The queues are allocated for virtualization such that VF 0
720 * is allocated queues 0 and 8, VF 1 queues 1 and 9, etc.
721 * In order to avoid collision we start at the first free queue
722 * and continue consuming queues in the same sequence
724 if (adapter
->vfs_allocated_count
) {
725 for (; i
< adapter
->rss_queues
; i
++)
726 adapter
->rx_ring
[i
]->reg_idx
= rbase_offset
+
738 for (; i
< adapter
->num_rx_queues
; i
++)
739 adapter
->rx_ring
[i
]->reg_idx
= rbase_offset
+ i
;
740 for (; j
< adapter
->num_tx_queues
; j
++)
741 adapter
->tx_ring
[j
]->reg_idx
= rbase_offset
+ j
;
746 u32
igb_rd32(struct e1000_hw
*hw
, u32 reg
)
748 struct igb_adapter
*igb
= container_of(hw
, struct igb_adapter
, hw
);
749 u8 __iomem
*hw_addr
= ACCESS_ONCE(hw
->hw_addr
);
752 if (E1000_REMOVED(hw_addr
))
755 value
= readl(&hw_addr
[reg
]);
757 /* reads should not return all F's */
758 if (!(~value
) && (!reg
|| !(~readl(hw_addr
)))) {
759 struct net_device
*netdev
= igb
->netdev
;
761 netif_device_detach(netdev
);
762 netdev_err(netdev
, "PCIe link lost, device now detached\n");
769 * igb_write_ivar - configure ivar for given MSI-X vector
770 * @hw: pointer to the HW structure
771 * @msix_vector: vector number we are allocating to a given ring
772 * @index: row index of IVAR register to write within IVAR table
773 * @offset: column offset of in IVAR, should be multiple of 8
775 * This function is intended to handle the writing of the IVAR register
776 * for adapters 82576 and newer. The IVAR table consists of 2 columns,
777 * each containing an cause allocation for an Rx and Tx ring, and a
778 * variable number of rows depending on the number of queues supported.
780 static void igb_write_ivar(struct e1000_hw
*hw
, int msix_vector
,
781 int index
, int offset
)
783 u32 ivar
= array_rd32(E1000_IVAR0
, index
);
785 /* clear any bits that are currently set */
786 ivar
&= ~((u32
)0xFF << offset
);
788 /* write vector and valid bit */
789 ivar
|= (msix_vector
| E1000_IVAR_VALID
) << offset
;
791 array_wr32(E1000_IVAR0
, index
, ivar
);
794 #define IGB_N0_QUEUE -1
795 static void igb_assign_vector(struct igb_q_vector
*q_vector
, int msix_vector
)
797 struct igb_adapter
*adapter
= q_vector
->adapter
;
798 struct e1000_hw
*hw
= &adapter
->hw
;
799 int rx_queue
= IGB_N0_QUEUE
;
800 int tx_queue
= IGB_N0_QUEUE
;
803 if (q_vector
->rx
.ring
)
804 rx_queue
= q_vector
->rx
.ring
->reg_idx
;
805 if (q_vector
->tx
.ring
)
806 tx_queue
= q_vector
->tx
.ring
->reg_idx
;
808 switch (hw
->mac
.type
) {
810 /* The 82575 assigns vectors using a bitmask, which matches the
811 * bitmask for the EICR/EIMS/EIMC registers. To assign one
812 * or more queues to a vector, we write the appropriate bits
813 * into the MSIXBM register for that vector.
815 if (rx_queue
> IGB_N0_QUEUE
)
816 msixbm
= E1000_EICR_RX_QUEUE0
<< rx_queue
;
817 if (tx_queue
> IGB_N0_QUEUE
)
818 msixbm
|= E1000_EICR_TX_QUEUE0
<< tx_queue
;
819 if (!(adapter
->flags
& IGB_FLAG_HAS_MSIX
) && msix_vector
== 0)
820 msixbm
|= E1000_EIMS_OTHER
;
821 array_wr32(E1000_MSIXBM(0), msix_vector
, msixbm
);
822 q_vector
->eims_value
= msixbm
;
825 /* 82576 uses a table that essentially consists of 2 columns
826 * with 8 rows. The ordering is column-major so we use the
827 * lower 3 bits as the row index, and the 4th bit as the
830 if (rx_queue
> IGB_N0_QUEUE
)
831 igb_write_ivar(hw
, msix_vector
,
833 (rx_queue
& 0x8) << 1);
834 if (tx_queue
> IGB_N0_QUEUE
)
835 igb_write_ivar(hw
, msix_vector
,
837 ((tx_queue
& 0x8) << 1) + 8);
838 q_vector
->eims_value
= 1 << msix_vector
;
845 /* On 82580 and newer adapters the scheme is similar to 82576
846 * however instead of ordering column-major we have things
847 * ordered row-major. So we traverse the table by using
848 * bit 0 as the column offset, and the remaining bits as the
851 if (rx_queue
> IGB_N0_QUEUE
)
852 igb_write_ivar(hw
, msix_vector
,
854 (rx_queue
& 0x1) << 4);
855 if (tx_queue
> IGB_N0_QUEUE
)
856 igb_write_ivar(hw
, msix_vector
,
858 ((tx_queue
& 0x1) << 4) + 8);
859 q_vector
->eims_value
= 1 << msix_vector
;
866 /* add q_vector eims value to global eims_enable_mask */
867 adapter
->eims_enable_mask
|= q_vector
->eims_value
;
869 /* configure q_vector to set itr on first interrupt */
870 q_vector
->set_itr
= 1;
874 * igb_configure_msix - Configure MSI-X hardware
875 * @adapter: board private structure to initialize
877 * igb_configure_msix sets up the hardware to properly
878 * generate MSI-X interrupts.
880 static void igb_configure_msix(struct igb_adapter
*adapter
)
884 struct e1000_hw
*hw
= &adapter
->hw
;
886 adapter
->eims_enable_mask
= 0;
888 /* set vector for other causes, i.e. link changes */
889 switch (hw
->mac
.type
) {
891 tmp
= rd32(E1000_CTRL_EXT
);
892 /* enable MSI-X PBA support*/
893 tmp
|= E1000_CTRL_EXT_PBA_CLR
;
895 /* Auto-Mask interrupts upon ICR read. */
896 tmp
|= E1000_CTRL_EXT_EIAME
;
897 tmp
|= E1000_CTRL_EXT_IRCA
;
899 wr32(E1000_CTRL_EXT
, tmp
);
901 /* enable msix_other interrupt */
902 array_wr32(E1000_MSIXBM(0), vector
++, E1000_EIMS_OTHER
);
903 adapter
->eims_other
= E1000_EIMS_OTHER
;
913 /* Turn on MSI-X capability first, or our settings
914 * won't stick. And it will take days to debug.
916 wr32(E1000_GPIE
, E1000_GPIE_MSIX_MODE
|
917 E1000_GPIE_PBA
| E1000_GPIE_EIAME
|
920 /* enable msix_other interrupt */
921 adapter
->eims_other
= 1 << vector
;
922 tmp
= (vector
++ | E1000_IVAR_VALID
) << 8;
924 wr32(E1000_IVAR_MISC
, tmp
);
927 /* do nothing, since nothing else supports MSI-X */
929 } /* switch (hw->mac.type) */
931 adapter
->eims_enable_mask
|= adapter
->eims_other
;
933 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
934 igb_assign_vector(adapter
->q_vector
[i
], vector
++);
940 * igb_request_msix - Initialize MSI-X interrupts
941 * @adapter: board private structure to initialize
943 * igb_request_msix allocates MSI-X vectors and requests interrupts from the
946 static int igb_request_msix(struct igb_adapter
*adapter
)
948 struct net_device
*netdev
= adapter
->netdev
;
949 int i
, err
= 0, vector
= 0, free_vector
= 0;
951 err
= request_irq(adapter
->msix_entries
[vector
].vector
,
952 igb_msix_other
, 0, netdev
->name
, adapter
);
956 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
957 struct igb_q_vector
*q_vector
= adapter
->q_vector
[i
];
961 q_vector
->itr_register
= adapter
->io_addr
+ E1000_EITR(vector
);
963 if (q_vector
->rx
.ring
&& q_vector
->tx
.ring
)
964 sprintf(q_vector
->name
, "%s-TxRx-%u", netdev
->name
,
965 q_vector
->rx
.ring
->queue_index
);
966 else if (q_vector
->tx
.ring
)
967 sprintf(q_vector
->name
, "%s-tx-%u", netdev
->name
,
968 q_vector
->tx
.ring
->queue_index
);
969 else if (q_vector
->rx
.ring
)
970 sprintf(q_vector
->name
, "%s-rx-%u", netdev
->name
,
971 q_vector
->rx
.ring
->queue_index
);
973 sprintf(q_vector
->name
, "%s-unused", netdev
->name
);
975 err
= request_irq(adapter
->msix_entries
[vector
].vector
,
976 igb_msix_ring
, 0, q_vector
->name
,
982 igb_configure_msix(adapter
);
986 /* free already assigned IRQs */
987 free_irq(adapter
->msix_entries
[free_vector
++].vector
, adapter
);
990 for (i
= 0; i
< vector
; i
++) {
991 free_irq(adapter
->msix_entries
[free_vector
++].vector
,
992 adapter
->q_vector
[i
]);
999 * igb_free_q_vector - Free memory allocated for specific interrupt vector
1000 * @adapter: board private structure to initialize
1001 * @v_idx: Index of vector to be freed
1003 * This function frees the memory allocated to the q_vector.
1005 static void igb_free_q_vector(struct igb_adapter
*adapter
, int v_idx
)
1007 struct igb_q_vector
*q_vector
= adapter
->q_vector
[v_idx
];
1009 adapter
->q_vector
[v_idx
] = NULL
;
1011 /* igb_get_stats64() might access the rings on this vector,
1012 * we must wait a grace period before freeing it.
1015 kfree_rcu(q_vector
, rcu
);
1019 * igb_reset_q_vector - Reset config for interrupt vector
1020 * @adapter: board private structure to initialize
1021 * @v_idx: Index of vector to be reset
1023 * If NAPI is enabled it will delete any references to the
1024 * NAPI struct. This is preparation for igb_free_q_vector.
1026 static void igb_reset_q_vector(struct igb_adapter
*adapter
, int v_idx
)
1028 struct igb_q_vector
*q_vector
= adapter
->q_vector
[v_idx
];
1030 /* Coming from igb_set_interrupt_capability, the vectors are not yet
1031 * allocated. So, q_vector is NULL so we should stop here.
1036 if (q_vector
->tx
.ring
)
1037 adapter
->tx_ring
[q_vector
->tx
.ring
->queue_index
] = NULL
;
1039 if (q_vector
->rx
.ring
)
1040 adapter
->rx_ring
[q_vector
->rx
.ring
->queue_index
] = NULL
;
1042 netif_napi_del(&q_vector
->napi
);
1046 static void igb_reset_interrupt_capability(struct igb_adapter
*adapter
)
1048 int v_idx
= adapter
->num_q_vectors
;
1050 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
)
1051 pci_disable_msix(adapter
->pdev
);
1052 else if (adapter
->flags
& IGB_FLAG_HAS_MSI
)
1053 pci_disable_msi(adapter
->pdev
);
1056 igb_reset_q_vector(adapter
, v_idx
);
1060 * igb_free_q_vectors - Free memory allocated for interrupt vectors
1061 * @adapter: board private structure to initialize
1063 * This function frees the memory allocated to the q_vectors. In addition if
1064 * NAPI is enabled it will delete any references to the NAPI struct prior
1065 * to freeing the q_vector.
1067 static void igb_free_q_vectors(struct igb_adapter
*adapter
)
1069 int v_idx
= adapter
->num_q_vectors
;
1071 adapter
->num_tx_queues
= 0;
1072 adapter
->num_rx_queues
= 0;
1073 adapter
->num_q_vectors
= 0;
1076 igb_reset_q_vector(adapter
, v_idx
);
1077 igb_free_q_vector(adapter
, v_idx
);
1082 * igb_clear_interrupt_scheme - reset the device to a state of no interrupts
1083 * @adapter: board private structure to initialize
1085 * This function resets the device so that it has 0 Rx queues, Tx queues, and
1086 * MSI-X interrupts allocated.
1088 static void igb_clear_interrupt_scheme(struct igb_adapter
*adapter
)
1090 igb_free_q_vectors(adapter
);
1091 igb_reset_interrupt_capability(adapter
);
1095 * igb_set_interrupt_capability - set MSI or MSI-X if supported
1096 * @adapter: board private structure to initialize
1097 * @msix: boolean value of MSIX capability
1099 * Attempt to configure interrupts using the best available
1100 * capabilities of the hardware and kernel.
1102 static void igb_set_interrupt_capability(struct igb_adapter
*adapter
, bool msix
)
1109 adapter
->flags
|= IGB_FLAG_HAS_MSIX
;
1111 /* Number of supported queues. */
1112 adapter
->num_rx_queues
= adapter
->rss_queues
;
1113 if (adapter
->vfs_allocated_count
)
1114 adapter
->num_tx_queues
= 1;
1116 adapter
->num_tx_queues
= adapter
->rss_queues
;
1118 /* start with one vector for every Rx queue */
1119 numvecs
= adapter
->num_rx_queues
;
1121 /* if Tx handler is separate add 1 for every Tx queue */
1122 if (!(adapter
->flags
& IGB_FLAG_QUEUE_PAIRS
))
1123 numvecs
+= adapter
->num_tx_queues
;
1125 /* store the number of vectors reserved for queues */
1126 adapter
->num_q_vectors
= numvecs
;
1128 /* add 1 vector for link status interrupts */
1130 for (i
= 0; i
< numvecs
; i
++)
1131 adapter
->msix_entries
[i
].entry
= i
;
1133 err
= pci_enable_msix_range(adapter
->pdev
,
1134 adapter
->msix_entries
,
1140 igb_reset_interrupt_capability(adapter
);
1142 /* If we can't do MSI-X, try MSI */
1144 adapter
->flags
&= ~IGB_FLAG_HAS_MSIX
;
1145 #ifdef CONFIG_PCI_IOV
1146 /* disable SR-IOV for non MSI-X configurations */
1147 if (adapter
->vf_data
) {
1148 struct e1000_hw
*hw
= &adapter
->hw
;
1149 /* disable iov and allow time for transactions to clear */
1150 pci_disable_sriov(adapter
->pdev
);
1153 kfree(adapter
->vf_data
);
1154 adapter
->vf_data
= NULL
;
1155 wr32(E1000_IOVCTL
, E1000_IOVCTL_REUSE_VFQ
);
1158 dev_info(&adapter
->pdev
->dev
, "IOV Disabled\n");
1161 adapter
->vfs_allocated_count
= 0;
1162 adapter
->rss_queues
= 1;
1163 adapter
->flags
|= IGB_FLAG_QUEUE_PAIRS
;
1164 adapter
->num_rx_queues
= 1;
1165 adapter
->num_tx_queues
= 1;
1166 adapter
->num_q_vectors
= 1;
1167 if (!pci_enable_msi(adapter
->pdev
))
1168 adapter
->flags
|= IGB_FLAG_HAS_MSI
;
1171 static void igb_add_ring(struct igb_ring
*ring
,
1172 struct igb_ring_container
*head
)
1179 * igb_alloc_q_vector - Allocate memory for a single interrupt vector
1180 * @adapter: board private structure to initialize
1181 * @v_count: q_vectors allocated on adapter, used for ring interleaving
1182 * @v_idx: index of vector in adapter struct
1183 * @txr_count: total number of Tx rings to allocate
1184 * @txr_idx: index of first Tx ring to allocate
1185 * @rxr_count: total number of Rx rings to allocate
1186 * @rxr_idx: index of first Rx ring to allocate
1188 * We allocate one q_vector. If allocation fails we return -ENOMEM.
1190 static int igb_alloc_q_vector(struct igb_adapter
*adapter
,
1191 int v_count
, int v_idx
,
1192 int txr_count
, int txr_idx
,
1193 int rxr_count
, int rxr_idx
)
1195 struct igb_q_vector
*q_vector
;
1196 struct igb_ring
*ring
;
1197 int ring_count
, size
;
1199 /* igb only supports 1 Tx and/or 1 Rx queue per vector */
1200 if (txr_count
> 1 || rxr_count
> 1)
1203 ring_count
= txr_count
+ rxr_count
;
1204 size
= sizeof(struct igb_q_vector
) +
1205 (sizeof(struct igb_ring
) * ring_count
);
1207 /* allocate q_vector and rings */
1208 q_vector
= adapter
->q_vector
[v_idx
];
1210 q_vector
= kzalloc(size
, GFP_KERNEL
);
1211 } else if (size
> ksize(q_vector
)) {
1212 kfree_rcu(q_vector
, rcu
);
1213 q_vector
= kzalloc(size
, GFP_KERNEL
);
1215 memset(q_vector
, 0, size
);
1220 /* initialize NAPI */
1221 netif_napi_add(adapter
->netdev
, &q_vector
->napi
,
1224 /* tie q_vector and adapter together */
1225 adapter
->q_vector
[v_idx
] = q_vector
;
1226 q_vector
->adapter
= adapter
;
1228 /* initialize work limits */
1229 q_vector
->tx
.work_limit
= adapter
->tx_work_limit
;
1231 /* initialize ITR configuration */
1232 q_vector
->itr_register
= adapter
->io_addr
+ E1000_EITR(0);
1233 q_vector
->itr_val
= IGB_START_ITR
;
1235 /* initialize pointer to rings */
1236 ring
= q_vector
->ring
;
1240 /* rx or rx/tx vector */
1241 if (!adapter
->rx_itr_setting
|| adapter
->rx_itr_setting
> 3)
1242 q_vector
->itr_val
= adapter
->rx_itr_setting
;
1244 /* tx only vector */
1245 if (!adapter
->tx_itr_setting
|| adapter
->tx_itr_setting
> 3)
1246 q_vector
->itr_val
= adapter
->tx_itr_setting
;
1250 /* assign generic ring traits */
1251 ring
->dev
= &adapter
->pdev
->dev
;
1252 ring
->netdev
= adapter
->netdev
;
1254 /* configure backlink on ring */
1255 ring
->q_vector
= q_vector
;
1257 /* update q_vector Tx values */
1258 igb_add_ring(ring
, &q_vector
->tx
);
1260 /* For 82575, context index must be unique per ring. */
1261 if (adapter
->hw
.mac
.type
== e1000_82575
)
1262 set_bit(IGB_RING_FLAG_TX_CTX_IDX
, &ring
->flags
);
1264 /* apply Tx specific ring traits */
1265 ring
->count
= adapter
->tx_ring_count
;
1266 ring
->queue_index
= txr_idx
;
1268 u64_stats_init(&ring
->tx_syncp
);
1269 u64_stats_init(&ring
->tx_syncp2
);
1271 /* assign ring to adapter */
1272 adapter
->tx_ring
[txr_idx
] = ring
;
1274 /* push pointer to next ring */
1279 /* assign generic ring traits */
1280 ring
->dev
= &adapter
->pdev
->dev
;
1281 ring
->netdev
= adapter
->netdev
;
1283 /* configure backlink on ring */
1284 ring
->q_vector
= q_vector
;
1286 /* update q_vector Rx values */
1287 igb_add_ring(ring
, &q_vector
->rx
);
1289 /* set flag indicating ring supports SCTP checksum offload */
1290 if (adapter
->hw
.mac
.type
>= e1000_82576
)
1291 set_bit(IGB_RING_FLAG_RX_SCTP_CSUM
, &ring
->flags
);
1293 /* On i350, i354, i210, and i211, loopback VLAN packets
1294 * have the tag byte-swapped.
1296 if (adapter
->hw
.mac
.type
>= e1000_i350
)
1297 set_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP
, &ring
->flags
);
1299 /* apply Rx specific ring traits */
1300 ring
->count
= adapter
->rx_ring_count
;
1301 ring
->queue_index
= rxr_idx
;
1303 u64_stats_init(&ring
->rx_syncp
);
1305 /* assign ring to adapter */
1306 adapter
->rx_ring
[rxr_idx
] = ring
;
1314 * igb_alloc_q_vectors - Allocate memory for interrupt vectors
1315 * @adapter: board private structure to initialize
1317 * We allocate one q_vector per queue interrupt. If allocation fails we
1320 static int igb_alloc_q_vectors(struct igb_adapter
*adapter
)
1322 int q_vectors
= adapter
->num_q_vectors
;
1323 int rxr_remaining
= adapter
->num_rx_queues
;
1324 int txr_remaining
= adapter
->num_tx_queues
;
1325 int rxr_idx
= 0, txr_idx
= 0, v_idx
= 0;
1328 if (q_vectors
>= (rxr_remaining
+ txr_remaining
)) {
1329 for (; rxr_remaining
; v_idx
++) {
1330 err
= igb_alloc_q_vector(adapter
, q_vectors
, v_idx
,
1336 /* update counts and index */
1342 for (; v_idx
< q_vectors
; v_idx
++) {
1343 int rqpv
= DIV_ROUND_UP(rxr_remaining
, q_vectors
- v_idx
);
1344 int tqpv
= DIV_ROUND_UP(txr_remaining
, q_vectors
- v_idx
);
1346 err
= igb_alloc_q_vector(adapter
, q_vectors
, v_idx
,
1347 tqpv
, txr_idx
, rqpv
, rxr_idx
);
1352 /* update counts and index */
1353 rxr_remaining
-= rqpv
;
1354 txr_remaining
-= tqpv
;
1362 adapter
->num_tx_queues
= 0;
1363 adapter
->num_rx_queues
= 0;
1364 adapter
->num_q_vectors
= 0;
1367 igb_free_q_vector(adapter
, v_idx
);
1373 * igb_init_interrupt_scheme - initialize interrupts, allocate queues/vectors
1374 * @adapter: board private structure to initialize
1375 * @msix: boolean value of MSIX capability
1377 * This function initializes the interrupts and allocates all of the queues.
1379 static int igb_init_interrupt_scheme(struct igb_adapter
*adapter
, bool msix
)
1381 struct pci_dev
*pdev
= adapter
->pdev
;
1384 igb_set_interrupt_capability(adapter
, msix
);
1386 err
= igb_alloc_q_vectors(adapter
);
1388 dev_err(&pdev
->dev
, "Unable to allocate memory for vectors\n");
1389 goto err_alloc_q_vectors
;
1392 igb_cache_ring_register(adapter
);
1396 err_alloc_q_vectors
:
1397 igb_reset_interrupt_capability(adapter
);
1402 * igb_request_irq - initialize interrupts
1403 * @adapter: board private structure to initialize
1405 * Attempts to configure interrupts using the best available
1406 * capabilities of the hardware and kernel.
1408 static int igb_request_irq(struct igb_adapter
*adapter
)
1410 struct net_device
*netdev
= adapter
->netdev
;
1411 struct pci_dev
*pdev
= adapter
->pdev
;
1414 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1415 err
= igb_request_msix(adapter
);
1418 /* fall back to MSI */
1419 igb_free_all_tx_resources(adapter
);
1420 igb_free_all_rx_resources(adapter
);
1422 igb_clear_interrupt_scheme(adapter
);
1423 err
= igb_init_interrupt_scheme(adapter
, false);
1427 igb_setup_all_tx_resources(adapter
);
1428 igb_setup_all_rx_resources(adapter
);
1429 igb_configure(adapter
);
1432 igb_assign_vector(adapter
->q_vector
[0], 0);
1434 if (adapter
->flags
& IGB_FLAG_HAS_MSI
) {
1435 err
= request_irq(pdev
->irq
, igb_intr_msi
, 0,
1436 netdev
->name
, adapter
);
1440 /* fall back to legacy interrupts */
1441 igb_reset_interrupt_capability(adapter
);
1442 adapter
->flags
&= ~IGB_FLAG_HAS_MSI
;
1445 err
= request_irq(pdev
->irq
, igb_intr
, IRQF_SHARED
,
1446 netdev
->name
, adapter
);
1449 dev_err(&pdev
->dev
, "Error %d getting interrupt\n",
1456 static void igb_free_irq(struct igb_adapter
*adapter
)
1458 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1461 free_irq(adapter
->msix_entries
[vector
++].vector
, adapter
);
1463 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
1464 free_irq(adapter
->msix_entries
[vector
++].vector
,
1465 adapter
->q_vector
[i
]);
1467 free_irq(adapter
->pdev
->irq
, adapter
);
1472 * igb_irq_disable - Mask off interrupt generation on the NIC
1473 * @adapter: board private structure
1475 static void igb_irq_disable(struct igb_adapter
*adapter
)
1477 struct e1000_hw
*hw
= &adapter
->hw
;
1479 /* we need to be careful when disabling interrupts. The VFs are also
1480 * mapped into these registers and so clearing the bits can cause
1481 * issues on the VF drivers so we only need to clear what we set
1483 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1484 u32 regval
= rd32(E1000_EIAM
);
1486 wr32(E1000_EIAM
, regval
& ~adapter
->eims_enable_mask
);
1487 wr32(E1000_EIMC
, adapter
->eims_enable_mask
);
1488 regval
= rd32(E1000_EIAC
);
1489 wr32(E1000_EIAC
, regval
& ~adapter
->eims_enable_mask
);
1493 wr32(E1000_IMC
, ~0);
1495 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1498 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
1499 synchronize_irq(adapter
->msix_entries
[i
].vector
);
1501 synchronize_irq(adapter
->pdev
->irq
);
1506 * igb_irq_enable - Enable default interrupt generation settings
1507 * @adapter: board private structure
1509 static void igb_irq_enable(struct igb_adapter
*adapter
)
1511 struct e1000_hw
*hw
= &adapter
->hw
;
1513 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
1514 u32 ims
= E1000_IMS_LSC
| E1000_IMS_DOUTSYNC
| E1000_IMS_DRSTA
;
1515 u32 regval
= rd32(E1000_EIAC
);
1517 wr32(E1000_EIAC
, regval
| adapter
->eims_enable_mask
);
1518 regval
= rd32(E1000_EIAM
);
1519 wr32(E1000_EIAM
, regval
| adapter
->eims_enable_mask
);
1520 wr32(E1000_EIMS
, adapter
->eims_enable_mask
);
1521 if (adapter
->vfs_allocated_count
) {
1522 wr32(E1000_MBVFIMR
, 0xFF);
1523 ims
|= E1000_IMS_VMMB
;
1525 wr32(E1000_IMS
, ims
);
1527 wr32(E1000_IMS
, IMS_ENABLE_MASK
|
1529 wr32(E1000_IAM
, IMS_ENABLE_MASK
|
1534 static void igb_update_mng_vlan(struct igb_adapter
*adapter
)
1536 struct e1000_hw
*hw
= &adapter
->hw
;
1537 u16 pf_id
= adapter
->vfs_allocated_count
;
1538 u16 vid
= adapter
->hw
.mng_cookie
.vlan_id
;
1539 u16 old_vid
= adapter
->mng_vlan_id
;
1541 if (hw
->mng_cookie
.status
& E1000_MNG_DHCP_COOKIE_STATUS_VLAN
) {
1542 /* add VID to filter table */
1543 igb_vfta_set(hw
, vid
, pf_id
, true, true);
1544 adapter
->mng_vlan_id
= vid
;
1546 adapter
->mng_vlan_id
= IGB_MNG_VLAN_NONE
;
1549 if ((old_vid
!= (u16
)IGB_MNG_VLAN_NONE
) &&
1551 !test_bit(old_vid
, adapter
->active_vlans
)) {
1552 /* remove VID from filter table */
1553 igb_vfta_set(hw
, vid
, pf_id
, false, true);
1558 * igb_release_hw_control - release control of the h/w to f/w
1559 * @adapter: address of board private structure
1561 * igb_release_hw_control resets CTRL_EXT:DRV_LOAD bit.
1562 * For ASF and Pass Through versions of f/w this means that the
1563 * driver is no longer loaded.
1565 static void igb_release_hw_control(struct igb_adapter
*adapter
)
1567 struct e1000_hw
*hw
= &adapter
->hw
;
1570 /* Let firmware take over control of h/w */
1571 ctrl_ext
= rd32(E1000_CTRL_EXT
);
1572 wr32(E1000_CTRL_EXT
,
1573 ctrl_ext
& ~E1000_CTRL_EXT_DRV_LOAD
);
1577 * igb_get_hw_control - get control of the h/w from f/w
1578 * @adapter: address of board private structure
1580 * igb_get_hw_control sets CTRL_EXT:DRV_LOAD bit.
1581 * For ASF and Pass Through versions of f/w this means that
1582 * the driver is loaded.
1584 static void igb_get_hw_control(struct igb_adapter
*adapter
)
1586 struct e1000_hw
*hw
= &adapter
->hw
;
1589 /* Let firmware know the driver has taken over */
1590 ctrl_ext
= rd32(E1000_CTRL_EXT
);
1591 wr32(E1000_CTRL_EXT
,
1592 ctrl_ext
| E1000_CTRL_EXT_DRV_LOAD
);
1596 * igb_configure - configure the hardware for RX and TX
1597 * @adapter: private board structure
1599 static void igb_configure(struct igb_adapter
*adapter
)
1601 struct net_device
*netdev
= adapter
->netdev
;
1604 igb_get_hw_control(adapter
);
1605 igb_set_rx_mode(netdev
);
1607 igb_restore_vlan(adapter
);
1609 igb_setup_tctl(adapter
);
1610 igb_setup_mrqc(adapter
);
1611 igb_setup_rctl(adapter
);
1613 igb_configure_tx(adapter
);
1614 igb_configure_rx(adapter
);
1616 igb_rx_fifo_flush_82575(&adapter
->hw
);
1618 /* call igb_desc_unused which always leaves
1619 * at least 1 descriptor unused to make sure
1620 * next_to_use != next_to_clean
1622 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
1623 struct igb_ring
*ring
= adapter
->rx_ring
[i
];
1624 igb_alloc_rx_buffers(ring
, igb_desc_unused(ring
));
1629 * igb_power_up_link - Power up the phy/serdes link
1630 * @adapter: address of board private structure
1632 void igb_power_up_link(struct igb_adapter
*adapter
)
1634 igb_reset_phy(&adapter
->hw
);
1636 if (adapter
->hw
.phy
.media_type
== e1000_media_type_copper
)
1637 igb_power_up_phy_copper(&adapter
->hw
);
1639 igb_power_up_serdes_link_82575(&adapter
->hw
);
1641 igb_setup_link(&adapter
->hw
);
1645 * igb_power_down_link - Power down the phy/serdes link
1646 * @adapter: address of board private structure
1648 static void igb_power_down_link(struct igb_adapter
*adapter
)
1650 if (adapter
->hw
.phy
.media_type
== e1000_media_type_copper
)
1651 igb_power_down_phy_copper_82575(&adapter
->hw
);
1653 igb_shutdown_serdes_link_82575(&adapter
->hw
);
1657 * Detect and switch function for Media Auto Sense
1658 * @adapter: address of the board private structure
1660 static void igb_check_swap_media(struct igb_adapter
*adapter
)
1662 struct e1000_hw
*hw
= &adapter
->hw
;
1663 u32 ctrl_ext
, connsw
;
1664 bool swap_now
= false;
1666 ctrl_ext
= rd32(E1000_CTRL_EXT
);
1667 connsw
= rd32(E1000_CONNSW
);
1669 /* need to live swap if current media is copper and we have fiber/serdes
1673 if ((hw
->phy
.media_type
== e1000_media_type_copper
) &&
1674 (!(connsw
& E1000_CONNSW_AUTOSENSE_EN
))) {
1676 } else if (!(connsw
& E1000_CONNSW_SERDESD
)) {
1677 /* copper signal takes time to appear */
1678 if (adapter
->copper_tries
< 4) {
1679 adapter
->copper_tries
++;
1680 connsw
|= E1000_CONNSW_AUTOSENSE_CONF
;
1681 wr32(E1000_CONNSW
, connsw
);
1684 adapter
->copper_tries
= 0;
1685 if ((connsw
& E1000_CONNSW_PHYSD
) &&
1686 (!(connsw
& E1000_CONNSW_PHY_PDN
))) {
1688 connsw
&= ~E1000_CONNSW_AUTOSENSE_CONF
;
1689 wr32(E1000_CONNSW
, connsw
);
1697 switch (hw
->phy
.media_type
) {
1698 case e1000_media_type_copper
:
1699 netdev_info(adapter
->netdev
,
1700 "MAS: changing media to fiber/serdes\n");
1702 E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES
;
1703 adapter
->flags
|= IGB_FLAG_MEDIA_RESET
;
1704 adapter
->copper_tries
= 0;
1706 case e1000_media_type_internal_serdes
:
1707 case e1000_media_type_fiber
:
1708 netdev_info(adapter
->netdev
,
1709 "MAS: changing media to copper\n");
1711 ~E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES
;
1712 adapter
->flags
|= IGB_FLAG_MEDIA_RESET
;
1715 /* shouldn't get here during regular operation */
1716 netdev_err(adapter
->netdev
,
1717 "AMS: Invalid media type found, returning\n");
1720 wr32(E1000_CTRL_EXT
, ctrl_ext
);
1724 * igb_up - Open the interface and prepare it to handle traffic
1725 * @adapter: board private structure
1727 int igb_up(struct igb_adapter
*adapter
)
1729 struct e1000_hw
*hw
= &adapter
->hw
;
1732 /* hardware has been reset, we need to reload some things */
1733 igb_configure(adapter
);
1735 clear_bit(__IGB_DOWN
, &adapter
->state
);
1737 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
1738 napi_enable(&(adapter
->q_vector
[i
]->napi
));
1740 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
)
1741 igb_configure_msix(adapter
);
1743 igb_assign_vector(adapter
->q_vector
[0], 0);
1745 /* Clear any pending interrupts. */
1747 igb_irq_enable(adapter
);
1749 /* notify VFs that reset has been completed */
1750 if (adapter
->vfs_allocated_count
) {
1751 u32 reg_data
= rd32(E1000_CTRL_EXT
);
1753 reg_data
|= E1000_CTRL_EXT_PFRSTD
;
1754 wr32(E1000_CTRL_EXT
, reg_data
);
1757 netif_tx_start_all_queues(adapter
->netdev
);
1759 /* start the watchdog. */
1760 hw
->mac
.get_link_status
= 1;
1761 schedule_work(&adapter
->watchdog_task
);
1763 if ((adapter
->flags
& IGB_FLAG_EEE
) &&
1764 (!hw
->dev_spec
._82575
.eee_disable
))
1765 adapter
->eee_advert
= MDIO_EEE_100TX
| MDIO_EEE_1000T
;
1770 void igb_down(struct igb_adapter
*adapter
)
1772 struct net_device
*netdev
= adapter
->netdev
;
1773 struct e1000_hw
*hw
= &adapter
->hw
;
1777 /* signal that we're down so the interrupt handler does not
1778 * reschedule our watchdog timer
1780 set_bit(__IGB_DOWN
, &adapter
->state
);
1782 /* disable receives in the hardware */
1783 rctl
= rd32(E1000_RCTL
);
1784 wr32(E1000_RCTL
, rctl
& ~E1000_RCTL_EN
);
1785 /* flush and sleep below */
1787 netif_carrier_off(netdev
);
1788 netif_tx_stop_all_queues(netdev
);
1790 /* disable transmits in the hardware */
1791 tctl
= rd32(E1000_TCTL
);
1792 tctl
&= ~E1000_TCTL_EN
;
1793 wr32(E1000_TCTL
, tctl
);
1794 /* flush both disables and wait for them to finish */
1796 usleep_range(10000, 11000);
1798 igb_irq_disable(adapter
);
1800 adapter
->flags
&= ~IGB_FLAG_NEED_LINK_UPDATE
;
1802 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
1803 if (adapter
->q_vector
[i
]) {
1804 napi_synchronize(&adapter
->q_vector
[i
]->napi
);
1805 napi_disable(&adapter
->q_vector
[i
]->napi
);
1809 del_timer_sync(&adapter
->watchdog_timer
);
1810 del_timer_sync(&adapter
->phy_info_timer
);
1812 /* record the stats before reset*/
1813 spin_lock(&adapter
->stats64_lock
);
1814 igb_update_stats(adapter
, &adapter
->stats64
);
1815 spin_unlock(&adapter
->stats64_lock
);
1817 adapter
->link_speed
= 0;
1818 adapter
->link_duplex
= 0;
1820 if (!pci_channel_offline(adapter
->pdev
))
1822 igb_clean_all_tx_rings(adapter
);
1823 igb_clean_all_rx_rings(adapter
);
1824 #ifdef CONFIG_IGB_DCA
1826 /* since we reset the hardware DCA settings were cleared */
1827 igb_setup_dca(adapter
);
1831 void igb_reinit_locked(struct igb_adapter
*adapter
)
1833 WARN_ON(in_interrupt());
1834 while (test_and_set_bit(__IGB_RESETTING
, &adapter
->state
))
1835 usleep_range(1000, 2000);
1838 clear_bit(__IGB_RESETTING
, &adapter
->state
);
1841 /** igb_enable_mas - Media Autosense re-enable after swap
1843 * @adapter: adapter struct
1845 static void igb_enable_mas(struct igb_adapter
*adapter
)
1847 struct e1000_hw
*hw
= &adapter
->hw
;
1848 u32 connsw
= rd32(E1000_CONNSW
);
1850 /* configure for SerDes media detect */
1851 if ((hw
->phy
.media_type
== e1000_media_type_copper
) &&
1852 (!(connsw
& E1000_CONNSW_SERDESD
))) {
1853 connsw
|= E1000_CONNSW_ENRGSRC
;
1854 connsw
|= E1000_CONNSW_AUTOSENSE_EN
;
1855 wr32(E1000_CONNSW
, connsw
);
1860 void igb_reset(struct igb_adapter
*adapter
)
1862 struct pci_dev
*pdev
= adapter
->pdev
;
1863 struct e1000_hw
*hw
= &adapter
->hw
;
1864 struct e1000_mac_info
*mac
= &hw
->mac
;
1865 struct e1000_fc_info
*fc
= &hw
->fc
;
1868 /* Repartition Pba for greater than 9k mtu
1869 * To take effect CTRL.RST is required.
1871 switch (mac
->type
) {
1875 pba
= rd32(E1000_RXPBS
);
1876 pba
= igb_rxpbs_adjust_82580(pba
);
1879 pba
= rd32(E1000_RXPBS
);
1880 pba
&= E1000_RXPBS_SIZE_MASK_82576
;
1886 pba
= E1000_PBA_34K
;
1890 if (mac
->type
== e1000_82575
) {
1891 u32 min_rx_space
, min_tx_space
, needed_tx_space
;
1893 /* write Rx PBA so that hardware can report correct Tx PBA */
1894 wr32(E1000_PBA
, pba
);
1896 /* To maintain wire speed transmits, the Tx FIFO should be
1897 * large enough to accommodate two full transmit packets,
1898 * rounded up to the next 1KB and expressed in KB. Likewise,
1899 * the Rx FIFO should be large enough to accommodate at least
1900 * one full receive packet and is similarly rounded up and
1903 min_rx_space
= DIV_ROUND_UP(MAX_JUMBO_FRAME_SIZE
, 1024);
1905 /* The Tx FIFO also stores 16 bytes of information about the Tx
1906 * but don't include Ethernet FCS because hardware appends it.
1907 * We only need to round down to the nearest 512 byte block
1908 * count since the value we care about is 2 frames, not 1.
1910 min_tx_space
= adapter
->max_frame_size
;
1911 min_tx_space
+= sizeof(union e1000_adv_tx_desc
) - ETH_FCS_LEN
;
1912 min_tx_space
= DIV_ROUND_UP(min_tx_space
, 512);
1914 /* upper 16 bits has Tx packet buffer allocation size in KB */
1915 needed_tx_space
= min_tx_space
- (rd32(E1000_PBA
) >> 16);
1917 /* If current Tx allocation is less than the min Tx FIFO size,
1918 * and the min Tx FIFO size is less than the current Rx FIFO
1919 * allocation, take space away from current Rx allocation.
1921 if (needed_tx_space
< pba
) {
1922 pba
-= needed_tx_space
;
1924 /* if short on Rx space, Rx wins and must trump Tx
1927 if (pba
< min_rx_space
)
1931 /* adjust PBA for jumbo frames */
1932 wr32(E1000_PBA
, pba
);
1935 /* flow control settings
1936 * The high water mark must be low enough to fit one full frame
1937 * after transmitting the pause frame. As such we must have enough
1938 * space to allow for us to complete our current transmit and then
1939 * receive the frame that is in progress from the link partner.
1941 * - the full Rx FIFO size minus one full Tx plus one full Rx frame
1943 hwm
= (pba
<< 10) - (adapter
->max_frame_size
+ MAX_JUMBO_FRAME_SIZE
);
1945 fc
->high_water
= hwm
& 0xFFFFFFF0; /* 16-byte granularity */
1946 fc
->low_water
= fc
->high_water
- 16;
1947 fc
->pause_time
= 0xFFFF;
1949 fc
->current_mode
= fc
->requested_mode
;
1951 /* disable receive for all VFs and wait one second */
1952 if (adapter
->vfs_allocated_count
) {
1955 for (i
= 0 ; i
< adapter
->vfs_allocated_count
; i
++)
1956 adapter
->vf_data
[i
].flags
&= IGB_VF_FLAG_PF_SET_MAC
;
1958 /* ping all the active vfs to let them know we are going down */
1959 igb_ping_all_vfs(adapter
);
1961 /* disable transmits and receives */
1962 wr32(E1000_VFRE
, 0);
1963 wr32(E1000_VFTE
, 0);
1966 /* Allow time for pending master requests to run */
1967 hw
->mac
.ops
.reset_hw(hw
);
1970 if (adapter
->flags
& IGB_FLAG_MEDIA_RESET
) {
1971 /* need to resetup here after media swap */
1972 adapter
->ei
.get_invariants(hw
);
1973 adapter
->flags
&= ~IGB_FLAG_MEDIA_RESET
;
1975 if ((mac
->type
== e1000_82575
) &&
1976 (adapter
->flags
& IGB_FLAG_MAS_ENABLE
)) {
1977 igb_enable_mas(adapter
);
1979 if (hw
->mac
.ops
.init_hw(hw
))
1980 dev_err(&pdev
->dev
, "Hardware Error\n");
1982 /* Flow control settings reset on hardware reset, so guarantee flow
1983 * control is off when forcing speed.
1985 if (!hw
->mac
.autoneg
)
1986 igb_force_mac_fc(hw
);
1988 igb_init_dmac(adapter
, pba
);
1989 #ifdef CONFIG_IGB_HWMON
1990 /* Re-initialize the thermal sensor on i350 devices. */
1991 if (!test_bit(__IGB_DOWN
, &adapter
->state
)) {
1992 if (mac
->type
== e1000_i350
&& hw
->bus
.func
== 0) {
1993 /* If present, re-initialize the external thermal sensor
1997 mac
->ops
.init_thermal_sensor_thresh(hw
);
2001 /* Re-establish EEE setting */
2002 if (hw
->phy
.media_type
== e1000_media_type_copper
) {
2003 switch (mac
->type
) {
2007 igb_set_eee_i350(hw
, true, true);
2010 igb_set_eee_i354(hw
, true, true);
2016 if (!netif_running(adapter
->netdev
))
2017 igb_power_down_link(adapter
);
2019 igb_update_mng_vlan(adapter
);
2021 /* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
2022 wr32(E1000_VET
, ETHERNET_IEEE_VLAN_TYPE
);
2024 /* Re-enable PTP, where applicable. */
2025 igb_ptp_reset(adapter
);
2027 igb_get_phy_info(hw
);
2030 static netdev_features_t
igb_fix_features(struct net_device
*netdev
,
2031 netdev_features_t features
)
2033 /* Since there is no support for separate Rx/Tx vlan accel
2034 * enable/disable make sure Tx flag is always in same state as Rx.
2036 if (features
& NETIF_F_HW_VLAN_CTAG_RX
)
2037 features
|= NETIF_F_HW_VLAN_CTAG_TX
;
2039 features
&= ~NETIF_F_HW_VLAN_CTAG_TX
;
2044 static int igb_set_features(struct net_device
*netdev
,
2045 netdev_features_t features
)
2047 netdev_features_t changed
= netdev
->features
^ features
;
2048 struct igb_adapter
*adapter
= netdev_priv(netdev
);
2050 if (changed
& NETIF_F_HW_VLAN_CTAG_RX
)
2051 igb_vlan_mode(netdev
, features
);
2053 if (!(changed
& NETIF_F_RXALL
))
2056 netdev
->features
= features
;
2058 if (netif_running(netdev
))
2059 igb_reinit_locked(adapter
);
2066 static const struct net_device_ops igb_netdev_ops
= {
2067 .ndo_open
= igb_open
,
2068 .ndo_stop
= igb_close
,
2069 .ndo_start_xmit
= igb_xmit_frame
,
2070 .ndo_get_stats64
= igb_get_stats64
,
2071 .ndo_set_rx_mode
= igb_set_rx_mode
,
2072 .ndo_set_mac_address
= igb_set_mac
,
2073 .ndo_change_mtu
= igb_change_mtu
,
2074 .ndo_do_ioctl
= igb_ioctl
,
2075 .ndo_tx_timeout
= igb_tx_timeout
,
2076 .ndo_validate_addr
= eth_validate_addr
,
2077 .ndo_vlan_rx_add_vid
= igb_vlan_rx_add_vid
,
2078 .ndo_vlan_rx_kill_vid
= igb_vlan_rx_kill_vid
,
2079 .ndo_set_vf_mac
= igb_ndo_set_vf_mac
,
2080 .ndo_set_vf_vlan
= igb_ndo_set_vf_vlan
,
2081 .ndo_set_vf_rate
= igb_ndo_set_vf_bw
,
2082 .ndo_set_vf_spoofchk
= igb_ndo_set_vf_spoofchk
,
2083 .ndo_get_vf_config
= igb_ndo_get_vf_config
,
2084 #ifdef CONFIG_NET_POLL_CONTROLLER
2085 .ndo_poll_controller
= igb_netpoll
,
2087 .ndo_fix_features
= igb_fix_features
,
2088 .ndo_set_features
= igb_set_features
,
2089 .ndo_features_check
= passthru_features_check
,
2093 * igb_set_fw_version - Configure version string for ethtool
2094 * @adapter: adapter struct
2096 void igb_set_fw_version(struct igb_adapter
*adapter
)
2098 struct e1000_hw
*hw
= &adapter
->hw
;
2099 struct e1000_fw_version fw
;
2101 igb_get_fw_version(hw
, &fw
);
2103 switch (hw
->mac
.type
) {
2106 if (!(igb_get_flash_presence_i210(hw
))) {
2107 snprintf(adapter
->fw_version
,
2108 sizeof(adapter
->fw_version
),
2110 fw
.invm_major
, fw
.invm_minor
,
2116 /* if option is rom valid, display its version too */
2118 snprintf(adapter
->fw_version
,
2119 sizeof(adapter
->fw_version
),
2120 "%d.%d, 0x%08x, %d.%d.%d",
2121 fw
.eep_major
, fw
.eep_minor
, fw
.etrack_id
,
2122 fw
.or_major
, fw
.or_build
, fw
.or_patch
);
2124 } else if (fw
.etrack_id
!= 0X0000) {
2125 snprintf(adapter
->fw_version
,
2126 sizeof(adapter
->fw_version
),
2128 fw
.eep_major
, fw
.eep_minor
, fw
.etrack_id
);
2130 snprintf(adapter
->fw_version
,
2131 sizeof(adapter
->fw_version
),
2133 fw
.eep_major
, fw
.eep_minor
, fw
.eep_build
);
2140 * igb_init_mas - init Media Autosense feature if enabled in the NVM
2142 * @adapter: adapter struct
2144 static void igb_init_mas(struct igb_adapter
*adapter
)
2146 struct e1000_hw
*hw
= &adapter
->hw
;
2149 hw
->nvm
.ops
.read(hw
, NVM_COMPAT
, 1, &eeprom_data
);
2150 switch (hw
->bus
.func
) {
2152 if (eeprom_data
& IGB_MAS_ENABLE_0
) {
2153 adapter
->flags
|= IGB_FLAG_MAS_ENABLE
;
2154 netdev_info(adapter
->netdev
,
2155 "MAS: Enabling Media Autosense for port %d\n",
2160 if (eeprom_data
& IGB_MAS_ENABLE_1
) {
2161 adapter
->flags
|= IGB_FLAG_MAS_ENABLE
;
2162 netdev_info(adapter
->netdev
,
2163 "MAS: Enabling Media Autosense for port %d\n",
2168 if (eeprom_data
& IGB_MAS_ENABLE_2
) {
2169 adapter
->flags
|= IGB_FLAG_MAS_ENABLE
;
2170 netdev_info(adapter
->netdev
,
2171 "MAS: Enabling Media Autosense for port %d\n",
2176 if (eeprom_data
& IGB_MAS_ENABLE_3
) {
2177 adapter
->flags
|= IGB_FLAG_MAS_ENABLE
;
2178 netdev_info(adapter
->netdev
,
2179 "MAS: Enabling Media Autosense for port %d\n",
2184 /* Shouldn't get here */
2185 netdev_err(adapter
->netdev
,
2186 "MAS: Invalid port configuration, returning\n");
2192 * igb_init_i2c - Init I2C interface
2193 * @adapter: pointer to adapter structure
2195 static s32
igb_init_i2c(struct igb_adapter
*adapter
)
2199 /* I2C interface supported on i350 devices */
2200 if (adapter
->hw
.mac
.type
!= e1000_i350
)
2203 /* Initialize the i2c bus which is controlled by the registers.
2204 * This bus will use the i2c_algo_bit structue that implements
2205 * the protocol through toggling of the 4 bits in the register.
2207 adapter
->i2c_adap
.owner
= THIS_MODULE
;
2208 adapter
->i2c_algo
= igb_i2c_algo
;
2209 adapter
->i2c_algo
.data
= adapter
;
2210 adapter
->i2c_adap
.algo_data
= &adapter
->i2c_algo
;
2211 adapter
->i2c_adap
.dev
.parent
= &adapter
->pdev
->dev
;
2212 strlcpy(adapter
->i2c_adap
.name
, "igb BB",
2213 sizeof(adapter
->i2c_adap
.name
));
2214 status
= i2c_bit_add_bus(&adapter
->i2c_adap
);
2219 * igb_probe - Device Initialization Routine
2220 * @pdev: PCI device information struct
2221 * @ent: entry in igb_pci_tbl
2223 * Returns 0 on success, negative on failure
2225 * igb_probe initializes an adapter identified by a pci_dev structure.
2226 * The OS initialization, configuring of the adapter private structure,
2227 * and a hardware reset occur.
2229 static int igb_probe(struct pci_dev
*pdev
, const struct pci_device_id
*ent
)
2231 struct net_device
*netdev
;
2232 struct igb_adapter
*adapter
;
2233 struct e1000_hw
*hw
;
2234 u16 eeprom_data
= 0;
2236 static int global_quad_port_a
; /* global quad port a indication */
2237 const struct e1000_info
*ei
= igb_info_tbl
[ent
->driver_data
];
2238 int err
, pci_using_dac
;
2239 u8 part_str
[E1000_PBANUM_LENGTH
];
2241 /* Catch broken hardware that put the wrong VF device ID in
2242 * the PCIe SR-IOV capability.
2244 if (pdev
->is_virtfn
) {
2245 WARN(1, KERN_ERR
"%s (%hx:%hx) should not be a VF!\n",
2246 pci_name(pdev
), pdev
->vendor
, pdev
->device
);
2250 err
= pci_enable_device_mem(pdev
);
2255 err
= dma_set_mask_and_coherent(&pdev
->dev
, DMA_BIT_MASK(64));
2259 err
= dma_set_mask_and_coherent(&pdev
->dev
, DMA_BIT_MASK(32));
2262 "No usable DMA configuration, aborting\n");
2267 err
= pci_request_selected_regions(pdev
, pci_select_bars(pdev
,
2273 pci_enable_pcie_error_reporting(pdev
);
2275 pci_set_master(pdev
);
2276 pci_save_state(pdev
);
2279 netdev
= alloc_etherdev_mq(sizeof(struct igb_adapter
),
2282 goto err_alloc_etherdev
;
2284 SET_NETDEV_DEV(netdev
, &pdev
->dev
);
2286 pci_set_drvdata(pdev
, netdev
);
2287 adapter
= netdev_priv(netdev
);
2288 adapter
->netdev
= netdev
;
2289 adapter
->pdev
= pdev
;
2292 adapter
->msg_enable
= netif_msg_init(debug
, DEFAULT_MSG_ENABLE
);
2295 adapter
->io_addr
= pci_iomap(pdev
, 0, 0);
2296 if (!adapter
->io_addr
)
2298 /* hw->hw_addr can be altered, we'll use adapter->io_addr for unmap */
2299 hw
->hw_addr
= adapter
->io_addr
;
2301 netdev
->netdev_ops
= &igb_netdev_ops
;
2302 igb_set_ethtool_ops(netdev
);
2303 netdev
->watchdog_timeo
= 5 * HZ
;
2305 strncpy(netdev
->name
, pci_name(pdev
), sizeof(netdev
->name
) - 1);
2307 netdev
->mem_start
= pci_resource_start(pdev
, 0);
2308 netdev
->mem_end
= pci_resource_end(pdev
, 0);
2310 /* PCI config space info */
2311 hw
->vendor_id
= pdev
->vendor
;
2312 hw
->device_id
= pdev
->device
;
2313 hw
->revision_id
= pdev
->revision
;
2314 hw
->subsystem_vendor_id
= pdev
->subsystem_vendor
;
2315 hw
->subsystem_device_id
= pdev
->subsystem_device
;
2317 /* Copy the default MAC, PHY and NVM function pointers */
2318 memcpy(&hw
->mac
.ops
, ei
->mac_ops
, sizeof(hw
->mac
.ops
));
2319 memcpy(&hw
->phy
.ops
, ei
->phy_ops
, sizeof(hw
->phy
.ops
));
2320 memcpy(&hw
->nvm
.ops
, ei
->nvm_ops
, sizeof(hw
->nvm
.ops
));
2321 /* Initialize skew-specific constants */
2322 err
= ei
->get_invariants(hw
);
2326 /* setup the private structure */
2327 err
= igb_sw_init(adapter
);
2331 igb_get_bus_info_pcie(hw
);
2333 hw
->phy
.autoneg_wait_to_complete
= false;
2335 /* Copper options */
2336 if (hw
->phy
.media_type
== e1000_media_type_copper
) {
2337 hw
->phy
.mdix
= AUTO_ALL_MODES
;
2338 hw
->phy
.disable_polarity_correction
= false;
2339 hw
->phy
.ms_type
= e1000_ms_hw_default
;
2342 if (igb_check_reset_block(hw
))
2343 dev_info(&pdev
->dev
,
2344 "PHY reset is blocked due to SOL/IDER session.\n");
2346 /* features is initialized to 0 in allocation, it might have bits
2347 * set by igb_sw_init so we should use an or instead of an
2350 netdev
->features
|= NETIF_F_SG
|
2357 NETIF_F_HW_VLAN_CTAG_RX
|
2358 NETIF_F_HW_VLAN_CTAG_TX
;
2360 /* copy netdev features into list of user selectable features */
2361 netdev
->hw_features
|= netdev
->features
;
2362 netdev
->hw_features
|= NETIF_F_RXALL
;
2364 /* set this bit last since it cannot be part of hw_features */
2365 netdev
->features
|= NETIF_F_HW_VLAN_CTAG_FILTER
;
2367 netdev
->vlan_features
|= NETIF_F_TSO
|
2373 netdev
->priv_flags
|= IFF_SUPP_NOFCS
;
2375 if (pci_using_dac
) {
2376 netdev
->features
|= NETIF_F_HIGHDMA
;
2377 netdev
->vlan_features
|= NETIF_F_HIGHDMA
;
2380 if (hw
->mac
.type
>= e1000_82576
) {
2381 netdev
->hw_features
|= NETIF_F_SCTP_CRC
;
2382 netdev
->features
|= NETIF_F_SCTP_CRC
;
2385 netdev
->priv_flags
|= IFF_UNICAST_FLT
;
2387 adapter
->en_mng_pt
= igb_enable_mng_pass_thru(hw
);
2389 /* before reading the NVM, reset the controller to put the device in a
2390 * known good starting state
2392 hw
->mac
.ops
.reset_hw(hw
);
2394 /* make sure the NVM is good , i211/i210 parts can have special NVM
2395 * that doesn't contain a checksum
2397 switch (hw
->mac
.type
) {
2400 if (igb_get_flash_presence_i210(hw
)) {
2401 if (hw
->nvm
.ops
.validate(hw
) < 0) {
2403 "The NVM Checksum Is Not Valid\n");
2410 if (hw
->nvm
.ops
.validate(hw
) < 0) {
2411 dev_err(&pdev
->dev
, "The NVM Checksum Is Not Valid\n");
2418 /* copy the MAC address out of the NVM */
2419 if (hw
->mac
.ops
.read_mac_addr(hw
))
2420 dev_err(&pdev
->dev
, "NVM Read Error\n");
2422 memcpy(netdev
->dev_addr
, hw
->mac
.addr
, netdev
->addr_len
);
2424 if (!is_valid_ether_addr(netdev
->dev_addr
)) {
2425 dev_err(&pdev
->dev
, "Invalid MAC Address\n");
2430 /* get firmware version for ethtool -i */
2431 igb_set_fw_version(adapter
);
2433 /* configure RXPBSIZE and TXPBSIZE */
2434 if (hw
->mac
.type
== e1000_i210
) {
2435 wr32(E1000_RXPBS
, I210_RXPBSIZE_DEFAULT
);
2436 wr32(E1000_TXPBS
, I210_TXPBSIZE_DEFAULT
);
2439 setup_timer(&adapter
->watchdog_timer
, igb_watchdog
,
2440 (unsigned long) adapter
);
2441 setup_timer(&adapter
->phy_info_timer
, igb_update_phy_info
,
2442 (unsigned long) adapter
);
2444 INIT_WORK(&adapter
->reset_task
, igb_reset_task
);
2445 INIT_WORK(&adapter
->watchdog_task
, igb_watchdog_task
);
2447 /* Initialize link properties that are user-changeable */
2448 adapter
->fc_autoneg
= true;
2449 hw
->mac
.autoneg
= true;
2450 hw
->phy
.autoneg_advertised
= 0x2f;
2452 hw
->fc
.requested_mode
= e1000_fc_default
;
2453 hw
->fc
.current_mode
= e1000_fc_default
;
2455 igb_validate_mdi_setting(hw
);
2457 /* By default, support wake on port A */
2458 if (hw
->bus
.func
== 0)
2459 adapter
->flags
|= IGB_FLAG_WOL_SUPPORTED
;
2461 /* Check the NVM for wake support on non-port A ports */
2462 if (hw
->mac
.type
>= e1000_82580
)
2463 hw
->nvm
.ops
.read(hw
, NVM_INIT_CONTROL3_PORT_A
+
2464 NVM_82580_LAN_FUNC_OFFSET(hw
->bus
.func
), 1,
2466 else if (hw
->bus
.func
== 1)
2467 hw
->nvm
.ops
.read(hw
, NVM_INIT_CONTROL3_PORT_B
, 1, &eeprom_data
);
2469 if (eeprom_data
& IGB_EEPROM_APME
)
2470 adapter
->flags
|= IGB_FLAG_WOL_SUPPORTED
;
2472 /* now that we have the eeprom settings, apply the special cases where
2473 * the eeprom may be wrong or the board simply won't support wake on
2474 * lan on a particular port
2476 switch (pdev
->device
) {
2477 case E1000_DEV_ID_82575GB_QUAD_COPPER
:
2478 adapter
->flags
&= ~IGB_FLAG_WOL_SUPPORTED
;
2480 case E1000_DEV_ID_82575EB_FIBER_SERDES
:
2481 case E1000_DEV_ID_82576_FIBER
:
2482 case E1000_DEV_ID_82576_SERDES
:
2483 /* Wake events only supported on port A for dual fiber
2484 * regardless of eeprom setting
2486 if (rd32(E1000_STATUS
) & E1000_STATUS_FUNC_1
)
2487 adapter
->flags
&= ~IGB_FLAG_WOL_SUPPORTED
;
2489 case E1000_DEV_ID_82576_QUAD_COPPER
:
2490 case E1000_DEV_ID_82576_QUAD_COPPER_ET2
:
2491 /* if quad port adapter, disable WoL on all but port A */
2492 if (global_quad_port_a
!= 0)
2493 adapter
->flags
&= ~IGB_FLAG_WOL_SUPPORTED
;
2495 adapter
->flags
|= IGB_FLAG_QUAD_PORT_A
;
2496 /* Reset for multiple quad port adapters */
2497 if (++global_quad_port_a
== 4)
2498 global_quad_port_a
= 0;
2501 /* If the device can't wake, don't set software support */
2502 if (!device_can_wakeup(&adapter
->pdev
->dev
))
2503 adapter
->flags
&= ~IGB_FLAG_WOL_SUPPORTED
;
2506 /* initialize the wol settings based on the eeprom settings */
2507 if (adapter
->flags
& IGB_FLAG_WOL_SUPPORTED
)
2508 adapter
->wol
|= E1000_WUFC_MAG
;
2510 /* Some vendors want WoL disabled by default, but still supported */
2511 if ((hw
->mac
.type
== e1000_i350
) &&
2512 (pdev
->subsystem_vendor
== PCI_VENDOR_ID_HP
)) {
2513 adapter
->flags
|= IGB_FLAG_WOL_SUPPORTED
;
2517 device_set_wakeup_enable(&adapter
->pdev
->dev
,
2518 adapter
->flags
& IGB_FLAG_WOL_SUPPORTED
);
2520 /* reset the hardware with the new settings */
2523 /* Init the I2C interface */
2524 err
= igb_init_i2c(adapter
);
2526 dev_err(&pdev
->dev
, "failed to init i2c interface\n");
2530 /* let the f/w know that the h/w is now under the control of the
2533 igb_get_hw_control(adapter
);
2535 strcpy(netdev
->name
, "eth%d");
2536 err
= register_netdev(netdev
);
2540 /* carrier off reporting is important to ethtool even BEFORE open */
2541 netif_carrier_off(netdev
);
2543 #ifdef CONFIG_IGB_DCA
2544 if (dca_add_requester(&pdev
->dev
) == 0) {
2545 adapter
->flags
|= IGB_FLAG_DCA_ENABLED
;
2546 dev_info(&pdev
->dev
, "DCA enabled\n");
2547 igb_setup_dca(adapter
);
2551 #ifdef CONFIG_IGB_HWMON
2552 /* Initialize the thermal sensor on i350 devices. */
2553 if (hw
->mac
.type
== e1000_i350
&& hw
->bus
.func
== 0) {
2556 /* Read the NVM to determine if this i350 device supports an
2557 * external thermal sensor.
2559 hw
->nvm
.ops
.read(hw
, NVM_ETS_CFG
, 1, &ets_word
);
2560 if (ets_word
!= 0x0000 && ets_word
!= 0xFFFF)
2561 adapter
->ets
= true;
2563 adapter
->ets
= false;
2564 if (igb_sysfs_init(adapter
))
2566 "failed to allocate sysfs resources\n");
2568 adapter
->ets
= false;
2571 /* Check if Media Autosense is enabled */
2573 if (hw
->dev_spec
._82575
.mas_capable
)
2574 igb_init_mas(adapter
);
2576 /* do hw tstamp init after resetting */
2577 igb_ptp_init(adapter
);
2579 dev_info(&pdev
->dev
, "Intel(R) Gigabit Ethernet Network Connection\n");
2580 /* print bus type/speed/width info, not applicable to i354 */
2581 if (hw
->mac
.type
!= e1000_i354
) {
2582 dev_info(&pdev
->dev
, "%s: (PCIe:%s:%s) %pM\n",
2584 ((hw
->bus
.speed
== e1000_bus_speed_2500
) ? "2.5Gb/s" :
2585 (hw
->bus
.speed
== e1000_bus_speed_5000
) ? "5.0Gb/s" :
2587 ((hw
->bus
.width
== e1000_bus_width_pcie_x4
) ?
2589 (hw
->bus
.width
== e1000_bus_width_pcie_x2
) ?
2591 (hw
->bus
.width
== e1000_bus_width_pcie_x1
) ?
2592 "Width x1" : "unknown"), netdev
->dev_addr
);
2595 if ((hw
->mac
.type
>= e1000_i210
||
2596 igb_get_flash_presence_i210(hw
))) {
2597 ret_val
= igb_read_part_string(hw
, part_str
,
2598 E1000_PBANUM_LENGTH
);
2600 ret_val
= -E1000_ERR_INVM_VALUE_NOT_FOUND
;
2604 strcpy(part_str
, "Unknown");
2605 dev_info(&pdev
->dev
, "%s: PBA No: %s\n", netdev
->name
, part_str
);
2606 dev_info(&pdev
->dev
,
2607 "Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
2608 (adapter
->flags
& IGB_FLAG_HAS_MSIX
) ? "MSI-X" :
2609 (adapter
->flags
& IGB_FLAG_HAS_MSI
) ? "MSI" : "legacy",
2610 adapter
->num_rx_queues
, adapter
->num_tx_queues
);
2611 if (hw
->phy
.media_type
== e1000_media_type_copper
) {
2612 switch (hw
->mac
.type
) {
2616 /* Enable EEE for internal copper PHY devices */
2617 err
= igb_set_eee_i350(hw
, true, true);
2619 (!hw
->dev_spec
._82575
.eee_disable
)) {
2620 adapter
->eee_advert
=
2621 MDIO_EEE_100TX
| MDIO_EEE_1000T
;
2622 adapter
->flags
|= IGB_FLAG_EEE
;
2626 if ((rd32(E1000_CTRL_EXT
) &
2627 E1000_CTRL_EXT_LINK_MODE_SGMII
)) {
2628 err
= igb_set_eee_i354(hw
, true, true);
2630 (!hw
->dev_spec
._82575
.eee_disable
)) {
2631 adapter
->eee_advert
=
2632 MDIO_EEE_100TX
| MDIO_EEE_1000T
;
2633 adapter
->flags
|= IGB_FLAG_EEE
;
2641 pm_runtime_put_noidle(&pdev
->dev
);
2645 igb_release_hw_control(adapter
);
2646 memset(&adapter
->i2c_adap
, 0, sizeof(adapter
->i2c_adap
));
2648 if (!igb_check_reset_block(hw
))
2651 if (hw
->flash_address
)
2652 iounmap(hw
->flash_address
);
2654 kfree(adapter
->shadow_vfta
);
2655 igb_clear_interrupt_scheme(adapter
);
2656 #ifdef CONFIG_PCI_IOV
2657 igb_disable_sriov(pdev
);
2659 pci_iounmap(pdev
, adapter
->io_addr
);
2661 free_netdev(netdev
);
2663 pci_release_selected_regions(pdev
,
2664 pci_select_bars(pdev
, IORESOURCE_MEM
));
2667 pci_disable_device(pdev
);
2671 #ifdef CONFIG_PCI_IOV
2672 static int igb_disable_sriov(struct pci_dev
*pdev
)
2674 struct net_device
*netdev
= pci_get_drvdata(pdev
);
2675 struct igb_adapter
*adapter
= netdev_priv(netdev
);
2676 struct e1000_hw
*hw
= &adapter
->hw
;
2678 /* reclaim resources allocated to VFs */
2679 if (adapter
->vf_data
) {
2680 /* disable iov and allow time for transactions to clear */
2681 if (pci_vfs_assigned(pdev
)) {
2682 dev_warn(&pdev
->dev
,
2683 "Cannot deallocate SR-IOV virtual functions while they are assigned - VFs will not be deallocated\n");
2686 pci_disable_sriov(pdev
);
2690 kfree(adapter
->vf_data
);
2691 adapter
->vf_data
= NULL
;
2692 adapter
->vfs_allocated_count
= 0;
2693 wr32(E1000_IOVCTL
, E1000_IOVCTL_REUSE_VFQ
);
2696 dev_info(&pdev
->dev
, "IOV Disabled\n");
2698 /* Re-enable DMA Coalescing flag since IOV is turned off */
2699 adapter
->flags
|= IGB_FLAG_DMAC
;
2705 static int igb_enable_sriov(struct pci_dev
*pdev
, int num_vfs
)
2707 struct net_device
*netdev
= pci_get_drvdata(pdev
);
2708 struct igb_adapter
*adapter
= netdev_priv(netdev
);
2709 int old_vfs
= pci_num_vf(pdev
);
2713 if (!(adapter
->flags
& IGB_FLAG_HAS_MSIX
) || num_vfs
> 7) {
2721 dev_info(&pdev
->dev
, "%d pre-allocated VFs found - override max_vfs setting of %d\n",
2723 adapter
->vfs_allocated_count
= old_vfs
;
2725 adapter
->vfs_allocated_count
= num_vfs
;
2727 adapter
->vf_data
= kcalloc(adapter
->vfs_allocated_count
,
2728 sizeof(struct vf_data_storage
), GFP_KERNEL
);
2730 /* if allocation failed then we do not support SR-IOV */
2731 if (!adapter
->vf_data
) {
2732 adapter
->vfs_allocated_count
= 0;
2734 "Unable to allocate memory for VF Data Storage\n");
2739 /* only call pci_enable_sriov() if no VFs are allocated already */
2741 err
= pci_enable_sriov(pdev
, adapter
->vfs_allocated_count
);
2745 dev_info(&pdev
->dev
, "%d VFs allocated\n",
2746 adapter
->vfs_allocated_count
);
2747 for (i
= 0; i
< adapter
->vfs_allocated_count
; i
++)
2748 igb_vf_configure(adapter
, i
);
2750 /* DMA Coalescing is not supported in IOV mode. */
2751 adapter
->flags
&= ~IGB_FLAG_DMAC
;
2755 kfree(adapter
->vf_data
);
2756 adapter
->vf_data
= NULL
;
2757 adapter
->vfs_allocated_count
= 0;
2764 * igb_remove_i2c - Cleanup I2C interface
2765 * @adapter: pointer to adapter structure
2767 static void igb_remove_i2c(struct igb_adapter
*adapter
)
2769 /* free the adapter bus structure */
2770 i2c_del_adapter(&adapter
->i2c_adap
);
2774 * igb_remove - Device Removal Routine
2775 * @pdev: PCI device information struct
2777 * igb_remove is called by the PCI subsystem to alert the driver
2778 * that it should release a PCI device. The could be caused by a
2779 * Hot-Plug event, or because the driver is going to be removed from
2782 static void igb_remove(struct pci_dev
*pdev
)
2784 struct net_device
*netdev
= pci_get_drvdata(pdev
);
2785 struct igb_adapter
*adapter
= netdev_priv(netdev
);
2786 struct e1000_hw
*hw
= &adapter
->hw
;
2788 pm_runtime_get_noresume(&pdev
->dev
);
2789 #ifdef CONFIG_IGB_HWMON
2790 igb_sysfs_exit(adapter
);
2792 igb_remove_i2c(adapter
);
2793 igb_ptp_stop(adapter
);
2794 /* The watchdog timer may be rescheduled, so explicitly
2795 * disable watchdog from being rescheduled.
2797 set_bit(__IGB_DOWN
, &adapter
->state
);
2798 del_timer_sync(&adapter
->watchdog_timer
);
2799 del_timer_sync(&adapter
->phy_info_timer
);
2801 cancel_work_sync(&adapter
->reset_task
);
2802 cancel_work_sync(&adapter
->watchdog_task
);
2804 #ifdef CONFIG_IGB_DCA
2805 if (adapter
->flags
& IGB_FLAG_DCA_ENABLED
) {
2806 dev_info(&pdev
->dev
, "DCA disabled\n");
2807 dca_remove_requester(&pdev
->dev
);
2808 adapter
->flags
&= ~IGB_FLAG_DCA_ENABLED
;
2809 wr32(E1000_DCA_CTRL
, E1000_DCA_CTRL_DCA_MODE_DISABLE
);
2813 /* Release control of h/w to f/w. If f/w is AMT enabled, this
2814 * would have already happened in close and is redundant.
2816 igb_release_hw_control(adapter
);
2818 #ifdef CONFIG_PCI_IOV
2819 igb_disable_sriov(pdev
);
2822 unregister_netdev(netdev
);
2824 igb_clear_interrupt_scheme(adapter
);
2826 pci_iounmap(pdev
, adapter
->io_addr
);
2827 if (hw
->flash_address
)
2828 iounmap(hw
->flash_address
);
2829 pci_release_selected_regions(pdev
,
2830 pci_select_bars(pdev
, IORESOURCE_MEM
));
2832 kfree(adapter
->shadow_vfta
);
2833 free_netdev(netdev
);
2835 pci_disable_pcie_error_reporting(pdev
);
2837 pci_disable_device(pdev
);
2841 * igb_probe_vfs - Initialize vf data storage and add VFs to pci config space
2842 * @adapter: board private structure to initialize
2844 * This function initializes the vf specific data storage and then attempts to
2845 * allocate the VFs. The reason for ordering it this way is because it is much
2846 * mor expensive time wise to disable SR-IOV than it is to allocate and free
2847 * the memory for the VFs.
2849 static void igb_probe_vfs(struct igb_adapter
*adapter
)
2851 #ifdef CONFIG_PCI_IOV
2852 struct pci_dev
*pdev
= adapter
->pdev
;
2853 struct e1000_hw
*hw
= &adapter
->hw
;
2855 /* Virtualization features not supported on i210 family. */
2856 if ((hw
->mac
.type
== e1000_i210
) || (hw
->mac
.type
== e1000_i211
))
2859 /* Of the below we really only want the effect of getting
2860 * IGB_FLAG_HAS_MSIX set (if available), without which
2861 * igb_enable_sriov() has no effect.
2863 igb_set_interrupt_capability(adapter
, true);
2864 igb_reset_interrupt_capability(adapter
);
2866 pci_sriov_set_totalvfs(pdev
, 7);
2867 igb_enable_sriov(pdev
, max_vfs
);
2869 #endif /* CONFIG_PCI_IOV */
2872 static void igb_init_queue_configuration(struct igb_adapter
*adapter
)
2874 struct e1000_hw
*hw
= &adapter
->hw
;
2877 /* Determine the maximum number of RSS queues supported. */
2878 switch (hw
->mac
.type
) {
2880 max_rss_queues
= IGB_MAX_RX_QUEUES_I211
;
2884 max_rss_queues
= IGB_MAX_RX_QUEUES_82575
;
2887 /* I350 cannot do RSS and SR-IOV at the same time */
2888 if (!!adapter
->vfs_allocated_count
) {
2894 if (!!adapter
->vfs_allocated_count
) {
2902 max_rss_queues
= IGB_MAX_RX_QUEUES
;
2906 adapter
->rss_queues
= min_t(u32
, max_rss_queues
, num_online_cpus());
2908 igb_set_flag_queue_pairs(adapter
, max_rss_queues
);
2911 void igb_set_flag_queue_pairs(struct igb_adapter
*adapter
,
2912 const u32 max_rss_queues
)
2914 struct e1000_hw
*hw
= &adapter
->hw
;
2916 /* Determine if we need to pair queues. */
2917 switch (hw
->mac
.type
) {
2920 /* Device supports enough interrupts without queue pairing. */
2928 /* If rss_queues > half of max_rss_queues, pair the queues in
2929 * order to conserve interrupts due to limited supply.
2931 if (adapter
->rss_queues
> (max_rss_queues
/ 2))
2932 adapter
->flags
|= IGB_FLAG_QUEUE_PAIRS
;
2934 adapter
->flags
&= ~IGB_FLAG_QUEUE_PAIRS
;
2940 * igb_sw_init - Initialize general software structures (struct igb_adapter)
2941 * @adapter: board private structure to initialize
2943 * igb_sw_init initializes the Adapter private data structure.
2944 * Fields are initialized based on PCI device information and
2945 * OS network device settings (MTU size).
2947 static int igb_sw_init(struct igb_adapter
*adapter
)
2949 struct e1000_hw
*hw
= &adapter
->hw
;
2950 struct net_device
*netdev
= adapter
->netdev
;
2951 struct pci_dev
*pdev
= adapter
->pdev
;
2953 pci_read_config_word(pdev
, PCI_COMMAND
, &hw
->bus
.pci_cmd_word
);
2955 /* set default ring sizes */
2956 adapter
->tx_ring_count
= IGB_DEFAULT_TXD
;
2957 adapter
->rx_ring_count
= IGB_DEFAULT_RXD
;
2959 /* set default ITR values */
2960 adapter
->rx_itr_setting
= IGB_DEFAULT_ITR
;
2961 adapter
->tx_itr_setting
= IGB_DEFAULT_ITR
;
2963 /* set default work limits */
2964 adapter
->tx_work_limit
= IGB_DEFAULT_TX_WORK
;
2966 adapter
->max_frame_size
= netdev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
+
2968 adapter
->min_frame_size
= ETH_ZLEN
+ ETH_FCS_LEN
;
2970 spin_lock_init(&adapter
->stats64_lock
);
2971 #ifdef CONFIG_PCI_IOV
2972 switch (hw
->mac
.type
) {
2976 dev_warn(&pdev
->dev
,
2977 "Maximum of 7 VFs per PF, using max\n");
2978 max_vfs
= adapter
->vfs_allocated_count
= 7;
2980 adapter
->vfs_allocated_count
= max_vfs
;
2981 if (adapter
->vfs_allocated_count
)
2982 dev_warn(&pdev
->dev
,
2983 "Enabling SR-IOV VFs using the module parameter is deprecated - please use the pci sysfs interface.\n");
2988 #endif /* CONFIG_PCI_IOV */
2990 /* Assume MSI-X interrupts, will be checked during IRQ allocation */
2991 adapter
->flags
|= IGB_FLAG_HAS_MSIX
;
2993 igb_probe_vfs(adapter
);
2995 igb_init_queue_configuration(adapter
);
2997 /* Setup and initialize a copy of the hw vlan table array */
2998 adapter
->shadow_vfta
= kcalloc(E1000_VLAN_FILTER_TBL_SIZE
, sizeof(u32
),
3001 /* This call may decrease the number of queues */
3002 if (igb_init_interrupt_scheme(adapter
, true)) {
3003 dev_err(&pdev
->dev
, "Unable to allocate memory for queues\n");
3007 /* Explicitly disable IRQ since the NIC can be in any state. */
3008 igb_irq_disable(adapter
);
3010 if (hw
->mac
.type
>= e1000_i350
)
3011 adapter
->flags
&= ~IGB_FLAG_DMAC
;
3013 set_bit(__IGB_DOWN
, &adapter
->state
);
3018 * igb_open - Called when a network interface is made active
3019 * @netdev: network interface device structure
3021 * Returns 0 on success, negative value on failure
3023 * The open entry point is called when a network interface is made
3024 * active by the system (IFF_UP). At this point all resources needed
3025 * for transmit and receive operations are allocated, the interrupt
3026 * handler is registered with the OS, the watchdog timer is started,
3027 * and the stack is notified that the interface is ready.
3029 static int __igb_open(struct net_device
*netdev
, bool resuming
)
3031 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3032 struct e1000_hw
*hw
= &adapter
->hw
;
3033 struct pci_dev
*pdev
= adapter
->pdev
;
3037 /* disallow open during test */
3038 if (test_bit(__IGB_TESTING
, &adapter
->state
)) {
3044 pm_runtime_get_sync(&pdev
->dev
);
3046 netif_carrier_off(netdev
);
3048 /* allocate transmit descriptors */
3049 err
= igb_setup_all_tx_resources(adapter
);
3053 /* allocate receive descriptors */
3054 err
= igb_setup_all_rx_resources(adapter
);
3058 igb_power_up_link(adapter
);
3060 /* before we allocate an interrupt, we must be ready to handle it.
3061 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
3062 * as soon as we call pci_request_irq, so we have to setup our
3063 * clean_rx handler before we do so.
3065 igb_configure(adapter
);
3067 err
= igb_request_irq(adapter
);
3071 /* Notify the stack of the actual queue counts. */
3072 err
= netif_set_real_num_tx_queues(adapter
->netdev
,
3073 adapter
->num_tx_queues
);
3075 goto err_set_queues
;
3077 err
= netif_set_real_num_rx_queues(adapter
->netdev
,
3078 adapter
->num_rx_queues
);
3080 goto err_set_queues
;
3082 /* From here on the code is the same as igb_up() */
3083 clear_bit(__IGB_DOWN
, &adapter
->state
);
3085 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
3086 napi_enable(&(adapter
->q_vector
[i
]->napi
));
3088 /* Clear any pending interrupts. */
3091 igb_irq_enable(adapter
);
3093 /* notify VFs that reset has been completed */
3094 if (adapter
->vfs_allocated_count
) {
3095 u32 reg_data
= rd32(E1000_CTRL_EXT
);
3097 reg_data
|= E1000_CTRL_EXT_PFRSTD
;
3098 wr32(E1000_CTRL_EXT
, reg_data
);
3101 netif_tx_start_all_queues(netdev
);
3104 pm_runtime_put(&pdev
->dev
);
3106 /* start the watchdog. */
3107 hw
->mac
.get_link_status
= 1;
3108 schedule_work(&adapter
->watchdog_task
);
3113 igb_free_irq(adapter
);
3115 igb_release_hw_control(adapter
);
3116 igb_power_down_link(adapter
);
3117 igb_free_all_rx_resources(adapter
);
3119 igb_free_all_tx_resources(adapter
);
3123 pm_runtime_put(&pdev
->dev
);
3128 static int igb_open(struct net_device
*netdev
)
3130 return __igb_open(netdev
, false);
3134 * igb_close - Disables a network interface
3135 * @netdev: network interface device structure
3137 * Returns 0, this is not allowed to fail
3139 * The close entry point is called when an interface is de-activated
3140 * by the OS. The hardware is still under the driver's control, but
3141 * needs to be disabled. A global MAC reset is issued to stop the
3142 * hardware, and all transmit and receive resources are freed.
3144 static int __igb_close(struct net_device
*netdev
, bool suspending
)
3146 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3147 struct pci_dev
*pdev
= adapter
->pdev
;
3149 WARN_ON(test_bit(__IGB_RESETTING
, &adapter
->state
));
3152 pm_runtime_get_sync(&pdev
->dev
);
3155 igb_free_irq(adapter
);
3157 igb_free_all_tx_resources(adapter
);
3158 igb_free_all_rx_resources(adapter
);
3161 pm_runtime_put_sync(&pdev
->dev
);
3165 static int igb_close(struct net_device
*netdev
)
3167 return __igb_close(netdev
, false);
3171 * igb_setup_tx_resources - allocate Tx resources (Descriptors)
3172 * @tx_ring: tx descriptor ring (for a specific queue) to setup
3174 * Return 0 on success, negative on failure
3176 int igb_setup_tx_resources(struct igb_ring
*tx_ring
)
3178 struct device
*dev
= tx_ring
->dev
;
3181 size
= sizeof(struct igb_tx_buffer
) * tx_ring
->count
;
3183 tx_ring
->tx_buffer_info
= vzalloc(size
);
3184 if (!tx_ring
->tx_buffer_info
)
3187 /* round up to nearest 4K */
3188 tx_ring
->size
= tx_ring
->count
* sizeof(union e1000_adv_tx_desc
);
3189 tx_ring
->size
= ALIGN(tx_ring
->size
, 4096);
3191 tx_ring
->desc
= dma_alloc_coherent(dev
, tx_ring
->size
,
3192 &tx_ring
->dma
, GFP_KERNEL
);
3196 tx_ring
->next_to_use
= 0;
3197 tx_ring
->next_to_clean
= 0;
3202 vfree(tx_ring
->tx_buffer_info
);
3203 tx_ring
->tx_buffer_info
= NULL
;
3204 dev_err(dev
, "Unable to allocate memory for the Tx descriptor ring\n");
3209 * igb_setup_all_tx_resources - wrapper to allocate Tx resources
3210 * (Descriptors) for all queues
3211 * @adapter: board private structure
3213 * Return 0 on success, negative on failure
3215 static int igb_setup_all_tx_resources(struct igb_adapter
*adapter
)
3217 struct pci_dev
*pdev
= adapter
->pdev
;
3220 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
3221 err
= igb_setup_tx_resources(adapter
->tx_ring
[i
]);
3224 "Allocation for Tx Queue %u failed\n", i
);
3225 for (i
--; i
>= 0; i
--)
3226 igb_free_tx_resources(adapter
->tx_ring
[i
]);
3235 * igb_setup_tctl - configure the transmit control registers
3236 * @adapter: Board private structure
3238 void igb_setup_tctl(struct igb_adapter
*adapter
)
3240 struct e1000_hw
*hw
= &adapter
->hw
;
3243 /* disable queue 0 which is enabled by default on 82575 and 82576 */
3244 wr32(E1000_TXDCTL(0), 0);
3246 /* Program the Transmit Control Register */
3247 tctl
= rd32(E1000_TCTL
);
3248 tctl
&= ~E1000_TCTL_CT
;
3249 tctl
|= E1000_TCTL_PSP
| E1000_TCTL_RTLC
|
3250 (E1000_COLLISION_THRESHOLD
<< E1000_CT_SHIFT
);
3252 igb_config_collision_dist(hw
);
3254 /* Enable transmits */
3255 tctl
|= E1000_TCTL_EN
;
3257 wr32(E1000_TCTL
, tctl
);
3261 * igb_configure_tx_ring - Configure transmit ring after Reset
3262 * @adapter: board private structure
3263 * @ring: tx ring to configure
3265 * Configure a transmit ring after a reset.
3267 void igb_configure_tx_ring(struct igb_adapter
*adapter
,
3268 struct igb_ring
*ring
)
3270 struct e1000_hw
*hw
= &adapter
->hw
;
3272 u64 tdba
= ring
->dma
;
3273 int reg_idx
= ring
->reg_idx
;
3275 /* disable the queue */
3276 wr32(E1000_TXDCTL(reg_idx
), 0);
3280 wr32(E1000_TDLEN(reg_idx
),
3281 ring
->count
* sizeof(union e1000_adv_tx_desc
));
3282 wr32(E1000_TDBAL(reg_idx
),
3283 tdba
& 0x00000000ffffffffULL
);
3284 wr32(E1000_TDBAH(reg_idx
), tdba
>> 32);
3286 ring
->tail
= hw
->hw_addr
+ E1000_TDT(reg_idx
);
3287 wr32(E1000_TDH(reg_idx
), 0);
3288 writel(0, ring
->tail
);
3290 txdctl
|= IGB_TX_PTHRESH
;
3291 txdctl
|= IGB_TX_HTHRESH
<< 8;
3292 txdctl
|= IGB_TX_WTHRESH
<< 16;
3294 txdctl
|= E1000_TXDCTL_QUEUE_ENABLE
;
3295 wr32(E1000_TXDCTL(reg_idx
), txdctl
);
3299 * igb_configure_tx - Configure transmit Unit after Reset
3300 * @adapter: board private structure
3302 * Configure the Tx unit of the MAC after a reset.
3304 static void igb_configure_tx(struct igb_adapter
*adapter
)
3308 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
3309 igb_configure_tx_ring(adapter
, adapter
->tx_ring
[i
]);
3313 * igb_setup_rx_resources - allocate Rx resources (Descriptors)
3314 * @rx_ring: Rx descriptor ring (for a specific queue) to setup
3316 * Returns 0 on success, negative on failure
3318 int igb_setup_rx_resources(struct igb_ring
*rx_ring
)
3320 struct device
*dev
= rx_ring
->dev
;
3323 size
= sizeof(struct igb_rx_buffer
) * rx_ring
->count
;
3325 rx_ring
->rx_buffer_info
= vzalloc(size
);
3326 if (!rx_ring
->rx_buffer_info
)
3329 /* Round up to nearest 4K */
3330 rx_ring
->size
= rx_ring
->count
* sizeof(union e1000_adv_rx_desc
);
3331 rx_ring
->size
= ALIGN(rx_ring
->size
, 4096);
3333 rx_ring
->desc
= dma_alloc_coherent(dev
, rx_ring
->size
,
3334 &rx_ring
->dma
, GFP_KERNEL
);
3338 rx_ring
->next_to_alloc
= 0;
3339 rx_ring
->next_to_clean
= 0;
3340 rx_ring
->next_to_use
= 0;
3345 vfree(rx_ring
->rx_buffer_info
);
3346 rx_ring
->rx_buffer_info
= NULL
;
3347 dev_err(dev
, "Unable to allocate memory for the Rx descriptor ring\n");
3352 * igb_setup_all_rx_resources - wrapper to allocate Rx resources
3353 * (Descriptors) for all queues
3354 * @adapter: board private structure
3356 * Return 0 on success, negative on failure
3358 static int igb_setup_all_rx_resources(struct igb_adapter
*adapter
)
3360 struct pci_dev
*pdev
= adapter
->pdev
;
3363 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3364 err
= igb_setup_rx_resources(adapter
->rx_ring
[i
]);
3367 "Allocation for Rx Queue %u failed\n", i
);
3368 for (i
--; i
>= 0; i
--)
3369 igb_free_rx_resources(adapter
->rx_ring
[i
]);
3378 * igb_setup_mrqc - configure the multiple receive queue control registers
3379 * @adapter: Board private structure
3381 static void igb_setup_mrqc(struct igb_adapter
*adapter
)
3383 struct e1000_hw
*hw
= &adapter
->hw
;
3385 u32 j
, num_rx_queues
;
3388 netdev_rss_key_fill(rss_key
, sizeof(rss_key
));
3389 for (j
= 0; j
< 10; j
++)
3390 wr32(E1000_RSSRK(j
), rss_key
[j
]);
3392 num_rx_queues
= adapter
->rss_queues
;
3394 switch (hw
->mac
.type
) {
3396 /* 82576 supports 2 RSS queues for SR-IOV */
3397 if (adapter
->vfs_allocated_count
)
3404 if (adapter
->rss_indir_tbl_init
!= num_rx_queues
) {
3405 for (j
= 0; j
< IGB_RETA_SIZE
; j
++)
3406 adapter
->rss_indir_tbl
[j
] =
3407 (j
* num_rx_queues
) / IGB_RETA_SIZE
;
3408 adapter
->rss_indir_tbl_init
= num_rx_queues
;
3410 igb_write_rss_indir_tbl(adapter
);
3412 /* Disable raw packet checksumming so that RSS hash is placed in
3413 * descriptor on writeback. No need to enable TCP/UDP/IP checksum
3414 * offloads as they are enabled by default
3416 rxcsum
= rd32(E1000_RXCSUM
);
3417 rxcsum
|= E1000_RXCSUM_PCSD
;
3419 if (adapter
->hw
.mac
.type
>= e1000_82576
)
3420 /* Enable Receive Checksum Offload for SCTP */
3421 rxcsum
|= E1000_RXCSUM_CRCOFL
;
3423 /* Don't need to set TUOFL or IPOFL, they default to 1 */
3424 wr32(E1000_RXCSUM
, rxcsum
);
3426 /* Generate RSS hash based on packet types, TCP/UDP
3427 * port numbers and/or IPv4/v6 src and dst addresses
3429 mrqc
= E1000_MRQC_RSS_FIELD_IPV4
|
3430 E1000_MRQC_RSS_FIELD_IPV4_TCP
|
3431 E1000_MRQC_RSS_FIELD_IPV6
|
3432 E1000_MRQC_RSS_FIELD_IPV6_TCP
|
3433 E1000_MRQC_RSS_FIELD_IPV6_TCP_EX
;
3435 if (adapter
->flags
& IGB_FLAG_RSS_FIELD_IPV4_UDP
)
3436 mrqc
|= E1000_MRQC_RSS_FIELD_IPV4_UDP
;
3437 if (adapter
->flags
& IGB_FLAG_RSS_FIELD_IPV6_UDP
)
3438 mrqc
|= E1000_MRQC_RSS_FIELD_IPV6_UDP
;
3440 /* If VMDq is enabled then we set the appropriate mode for that, else
3441 * we default to RSS so that an RSS hash is calculated per packet even
3442 * if we are only using one queue
3444 if (adapter
->vfs_allocated_count
) {
3445 if (hw
->mac
.type
> e1000_82575
) {
3446 /* Set the default pool for the PF's first queue */
3447 u32 vtctl
= rd32(E1000_VT_CTL
);
3449 vtctl
&= ~(E1000_VT_CTL_DEFAULT_POOL_MASK
|
3450 E1000_VT_CTL_DISABLE_DEF_POOL
);
3451 vtctl
|= adapter
->vfs_allocated_count
<<
3452 E1000_VT_CTL_DEFAULT_POOL_SHIFT
;
3453 wr32(E1000_VT_CTL
, vtctl
);
3455 if (adapter
->rss_queues
> 1)
3456 mrqc
|= E1000_MRQC_ENABLE_VMDQ_RSS_2Q
;
3458 mrqc
|= E1000_MRQC_ENABLE_VMDQ
;
3460 if (hw
->mac
.type
!= e1000_i211
)
3461 mrqc
|= E1000_MRQC_ENABLE_RSS_4Q
;
3463 igb_vmm_control(adapter
);
3465 wr32(E1000_MRQC
, mrqc
);
3469 * igb_setup_rctl - configure the receive control registers
3470 * @adapter: Board private structure
3472 void igb_setup_rctl(struct igb_adapter
*adapter
)
3474 struct e1000_hw
*hw
= &adapter
->hw
;
3477 rctl
= rd32(E1000_RCTL
);
3479 rctl
&= ~(3 << E1000_RCTL_MO_SHIFT
);
3480 rctl
&= ~(E1000_RCTL_LBM_TCVR
| E1000_RCTL_LBM_MAC
);
3482 rctl
|= E1000_RCTL_EN
| E1000_RCTL_BAM
| E1000_RCTL_RDMTS_HALF
|
3483 (hw
->mac
.mc_filter_type
<< E1000_RCTL_MO_SHIFT
);
3485 /* enable stripping of CRC. It's unlikely this will break BMC
3486 * redirection as it did with e1000. Newer features require
3487 * that the HW strips the CRC.
3489 rctl
|= E1000_RCTL_SECRC
;
3491 /* disable store bad packets and clear size bits. */
3492 rctl
&= ~(E1000_RCTL_SBP
| E1000_RCTL_SZ_256
);
3494 /* enable LPE to allow for reception of jumbo frames */
3495 rctl
|= E1000_RCTL_LPE
;
3497 /* disable queue 0 to prevent tail write w/o re-config */
3498 wr32(E1000_RXDCTL(0), 0);
3500 /* Attention!!! For SR-IOV PF driver operations you must enable
3501 * queue drop for all VF and PF queues to prevent head of line blocking
3502 * if an un-trusted VF does not provide descriptors to hardware.
3504 if (adapter
->vfs_allocated_count
) {
3505 /* set all queue drop enable bits */
3506 wr32(E1000_QDE
, ALL_QUEUES
);
3509 /* This is useful for sniffing bad packets. */
3510 if (adapter
->netdev
->features
& NETIF_F_RXALL
) {
3511 /* UPE and MPE will be handled by normal PROMISC logic
3512 * in e1000e_set_rx_mode
3514 rctl
|= (E1000_RCTL_SBP
| /* Receive bad packets */
3515 E1000_RCTL_BAM
| /* RX All Bcast Pkts */
3516 E1000_RCTL_PMCF
); /* RX All MAC Ctrl Pkts */
3518 rctl
&= ~(E1000_RCTL_VFE
| /* Disable VLAN filter */
3519 E1000_RCTL_DPF
| /* Allow filtered pause */
3520 E1000_RCTL_CFIEN
); /* Dis VLAN CFIEN Filter */
3521 /* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3522 * and that breaks VLANs.
3526 wr32(E1000_RCTL
, rctl
);
3529 static inline int igb_set_vf_rlpml(struct igb_adapter
*adapter
, int size
,
3532 struct e1000_hw
*hw
= &adapter
->hw
;
3535 if (size
> MAX_JUMBO_FRAME_SIZE
)
3536 size
= MAX_JUMBO_FRAME_SIZE
;
3538 vmolr
= rd32(E1000_VMOLR(vfn
));
3539 vmolr
&= ~E1000_VMOLR_RLPML_MASK
;
3540 vmolr
|= size
| E1000_VMOLR_LPE
;
3541 wr32(E1000_VMOLR(vfn
), vmolr
);
3546 static inline void igb_set_vmolr(struct igb_adapter
*adapter
,
3549 struct e1000_hw
*hw
= &adapter
->hw
;
3552 /* This register exists only on 82576 and newer so if we are older then
3553 * we should exit and do nothing
3555 if (hw
->mac
.type
< e1000_82576
)
3558 vmolr
= rd32(E1000_VMOLR(vfn
));
3559 vmolr
|= E1000_VMOLR_STRVLAN
; /* Strip vlan tags */
3560 if (hw
->mac
.type
== e1000_i350
) {
3563 dvmolr
= rd32(E1000_DVMOLR(vfn
));
3564 dvmolr
|= E1000_DVMOLR_STRVLAN
;
3565 wr32(E1000_DVMOLR(vfn
), dvmolr
);
3568 vmolr
|= E1000_VMOLR_AUPE
; /* Accept untagged packets */
3570 vmolr
&= ~(E1000_VMOLR_AUPE
); /* Tagged packets ONLY */
3572 /* clear all bits that might not be set */
3573 vmolr
&= ~(E1000_VMOLR_BAM
| E1000_VMOLR_RSSE
);
3575 if (adapter
->rss_queues
> 1 && vfn
== adapter
->vfs_allocated_count
)
3576 vmolr
|= E1000_VMOLR_RSSE
; /* enable RSS */
3577 /* for VMDq only allow the VFs and pool 0 to accept broadcast and
3580 if (vfn
<= adapter
->vfs_allocated_count
)
3581 vmolr
|= E1000_VMOLR_BAM
; /* Accept broadcast */
3583 wr32(E1000_VMOLR(vfn
), vmolr
);
3587 * igb_configure_rx_ring - Configure a receive ring after Reset
3588 * @adapter: board private structure
3589 * @ring: receive ring to be configured
3591 * Configure the Rx unit of the MAC after a reset.
3593 void igb_configure_rx_ring(struct igb_adapter
*adapter
,
3594 struct igb_ring
*ring
)
3596 struct e1000_hw
*hw
= &adapter
->hw
;
3597 u64 rdba
= ring
->dma
;
3598 int reg_idx
= ring
->reg_idx
;
3599 u32 srrctl
= 0, rxdctl
= 0;
3601 /* disable the queue */
3602 wr32(E1000_RXDCTL(reg_idx
), 0);
3604 /* Set DMA base address registers */
3605 wr32(E1000_RDBAL(reg_idx
),
3606 rdba
& 0x00000000ffffffffULL
);
3607 wr32(E1000_RDBAH(reg_idx
), rdba
>> 32);
3608 wr32(E1000_RDLEN(reg_idx
),
3609 ring
->count
* sizeof(union e1000_adv_rx_desc
));
3611 /* initialize head and tail */
3612 ring
->tail
= hw
->hw_addr
+ E1000_RDT(reg_idx
);
3613 wr32(E1000_RDH(reg_idx
), 0);
3614 writel(0, ring
->tail
);
3616 /* set descriptor configuration */
3617 srrctl
= IGB_RX_HDR_LEN
<< E1000_SRRCTL_BSIZEHDRSIZE_SHIFT
;
3618 srrctl
|= IGB_RX_BUFSZ
>> E1000_SRRCTL_BSIZEPKT_SHIFT
;
3619 srrctl
|= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF
;
3620 if (hw
->mac
.type
>= e1000_82580
)
3621 srrctl
|= E1000_SRRCTL_TIMESTAMP
;
3622 /* Only set Drop Enable if we are supporting multiple queues */
3623 if (adapter
->vfs_allocated_count
|| adapter
->num_rx_queues
> 1)
3624 srrctl
|= E1000_SRRCTL_DROP_EN
;
3626 wr32(E1000_SRRCTL(reg_idx
), srrctl
);
3628 /* set filtering for VMDQ pools */
3629 igb_set_vmolr(adapter
, reg_idx
& 0x7, true);
3631 rxdctl
|= IGB_RX_PTHRESH
;
3632 rxdctl
|= IGB_RX_HTHRESH
<< 8;
3633 rxdctl
|= IGB_RX_WTHRESH
<< 16;
3635 /* enable receive descriptor fetching */
3636 rxdctl
|= E1000_RXDCTL_QUEUE_ENABLE
;
3637 wr32(E1000_RXDCTL(reg_idx
), rxdctl
);
3641 * igb_configure_rx - Configure receive Unit after Reset
3642 * @adapter: board private structure
3644 * Configure the Rx unit of the MAC after a reset.
3646 static void igb_configure_rx(struct igb_adapter
*adapter
)
3650 /* set UTA to appropriate mode */
3651 igb_set_uta(adapter
);
3653 /* set the correct pool for the PF default MAC address in entry 0 */
3654 igb_rar_set_qsel(adapter
, adapter
->hw
.mac
.addr
, 0,
3655 adapter
->vfs_allocated_count
);
3657 /* Setup the HW Rx Head and Tail Descriptor Pointers and
3658 * the Base and Length of the Rx Descriptor Ring
3660 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3661 igb_configure_rx_ring(adapter
, adapter
->rx_ring
[i
]);
3665 * igb_free_tx_resources - Free Tx Resources per Queue
3666 * @tx_ring: Tx descriptor ring for a specific queue
3668 * Free all transmit software resources
3670 void igb_free_tx_resources(struct igb_ring
*tx_ring
)
3672 igb_clean_tx_ring(tx_ring
);
3674 vfree(tx_ring
->tx_buffer_info
);
3675 tx_ring
->tx_buffer_info
= NULL
;
3677 /* if not set, then don't free */
3681 dma_free_coherent(tx_ring
->dev
, tx_ring
->size
,
3682 tx_ring
->desc
, tx_ring
->dma
);
3684 tx_ring
->desc
= NULL
;
3688 * igb_free_all_tx_resources - Free Tx Resources for All Queues
3689 * @adapter: board private structure
3691 * Free all transmit software resources
3693 static void igb_free_all_tx_resources(struct igb_adapter
*adapter
)
3697 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
3698 if (adapter
->tx_ring
[i
])
3699 igb_free_tx_resources(adapter
->tx_ring
[i
]);
3702 void igb_unmap_and_free_tx_resource(struct igb_ring
*ring
,
3703 struct igb_tx_buffer
*tx_buffer
)
3705 if (tx_buffer
->skb
) {
3706 dev_kfree_skb_any(tx_buffer
->skb
);
3707 if (dma_unmap_len(tx_buffer
, len
))
3708 dma_unmap_single(ring
->dev
,
3709 dma_unmap_addr(tx_buffer
, dma
),
3710 dma_unmap_len(tx_buffer
, len
),
3712 } else if (dma_unmap_len(tx_buffer
, len
)) {
3713 dma_unmap_page(ring
->dev
,
3714 dma_unmap_addr(tx_buffer
, dma
),
3715 dma_unmap_len(tx_buffer
, len
),
3718 tx_buffer
->next_to_watch
= NULL
;
3719 tx_buffer
->skb
= NULL
;
3720 dma_unmap_len_set(tx_buffer
, len
, 0);
3721 /* buffer_info must be completely set up in the transmit path */
3725 * igb_clean_tx_ring - Free Tx Buffers
3726 * @tx_ring: ring to be cleaned
3728 static void igb_clean_tx_ring(struct igb_ring
*tx_ring
)
3730 struct igb_tx_buffer
*buffer_info
;
3734 if (!tx_ring
->tx_buffer_info
)
3736 /* Free all the Tx ring sk_buffs */
3738 for (i
= 0; i
< tx_ring
->count
; i
++) {
3739 buffer_info
= &tx_ring
->tx_buffer_info
[i
];
3740 igb_unmap_and_free_tx_resource(tx_ring
, buffer_info
);
3743 netdev_tx_reset_queue(txring_txq(tx_ring
));
3745 size
= sizeof(struct igb_tx_buffer
) * tx_ring
->count
;
3746 memset(tx_ring
->tx_buffer_info
, 0, size
);
3748 /* Zero out the descriptor ring */
3749 memset(tx_ring
->desc
, 0, tx_ring
->size
);
3751 tx_ring
->next_to_use
= 0;
3752 tx_ring
->next_to_clean
= 0;
3756 * igb_clean_all_tx_rings - Free Tx Buffers for all queues
3757 * @adapter: board private structure
3759 static void igb_clean_all_tx_rings(struct igb_adapter
*adapter
)
3763 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
3764 if (adapter
->tx_ring
[i
])
3765 igb_clean_tx_ring(adapter
->tx_ring
[i
]);
3769 * igb_free_rx_resources - Free Rx Resources
3770 * @rx_ring: ring to clean the resources from
3772 * Free all receive software resources
3774 void igb_free_rx_resources(struct igb_ring
*rx_ring
)
3776 igb_clean_rx_ring(rx_ring
);
3778 vfree(rx_ring
->rx_buffer_info
);
3779 rx_ring
->rx_buffer_info
= NULL
;
3781 /* if not set, then don't free */
3785 dma_free_coherent(rx_ring
->dev
, rx_ring
->size
,
3786 rx_ring
->desc
, rx_ring
->dma
);
3788 rx_ring
->desc
= NULL
;
3792 * igb_free_all_rx_resources - Free Rx Resources for All Queues
3793 * @adapter: board private structure
3795 * Free all receive software resources
3797 static void igb_free_all_rx_resources(struct igb_adapter
*adapter
)
3801 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3802 if (adapter
->rx_ring
[i
])
3803 igb_free_rx_resources(adapter
->rx_ring
[i
]);
3807 * igb_clean_rx_ring - Free Rx Buffers per Queue
3808 * @rx_ring: ring to free buffers from
3810 static void igb_clean_rx_ring(struct igb_ring
*rx_ring
)
3816 dev_kfree_skb(rx_ring
->skb
);
3817 rx_ring
->skb
= NULL
;
3819 if (!rx_ring
->rx_buffer_info
)
3822 /* Free all the Rx ring sk_buffs */
3823 for (i
= 0; i
< rx_ring
->count
; i
++) {
3824 struct igb_rx_buffer
*buffer_info
= &rx_ring
->rx_buffer_info
[i
];
3826 if (!buffer_info
->page
)
3829 dma_unmap_page(rx_ring
->dev
,
3833 __free_page(buffer_info
->page
);
3835 buffer_info
->page
= NULL
;
3838 size
= sizeof(struct igb_rx_buffer
) * rx_ring
->count
;
3839 memset(rx_ring
->rx_buffer_info
, 0, size
);
3841 /* Zero out the descriptor ring */
3842 memset(rx_ring
->desc
, 0, rx_ring
->size
);
3844 rx_ring
->next_to_alloc
= 0;
3845 rx_ring
->next_to_clean
= 0;
3846 rx_ring
->next_to_use
= 0;
3850 * igb_clean_all_rx_rings - Free Rx Buffers for all queues
3851 * @adapter: board private structure
3853 static void igb_clean_all_rx_rings(struct igb_adapter
*adapter
)
3857 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3858 if (adapter
->rx_ring
[i
])
3859 igb_clean_rx_ring(adapter
->rx_ring
[i
]);
3863 * igb_set_mac - Change the Ethernet Address of the NIC
3864 * @netdev: network interface device structure
3865 * @p: pointer to an address structure
3867 * Returns 0 on success, negative on failure
3869 static int igb_set_mac(struct net_device
*netdev
, void *p
)
3871 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3872 struct e1000_hw
*hw
= &adapter
->hw
;
3873 struct sockaddr
*addr
= p
;
3875 if (!is_valid_ether_addr(addr
->sa_data
))
3876 return -EADDRNOTAVAIL
;
3878 memcpy(netdev
->dev_addr
, addr
->sa_data
, netdev
->addr_len
);
3879 memcpy(hw
->mac
.addr
, addr
->sa_data
, netdev
->addr_len
);
3881 /* set the correct pool for the new PF MAC address in entry 0 */
3882 igb_rar_set_qsel(adapter
, hw
->mac
.addr
, 0,
3883 adapter
->vfs_allocated_count
);
3889 * igb_write_mc_addr_list - write multicast addresses to MTA
3890 * @netdev: network interface device structure
3892 * Writes multicast address list to the MTA hash table.
3893 * Returns: -ENOMEM on failure
3894 * 0 on no addresses written
3895 * X on writing X addresses to MTA
3897 static int igb_write_mc_addr_list(struct net_device
*netdev
)
3899 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3900 struct e1000_hw
*hw
= &adapter
->hw
;
3901 struct netdev_hw_addr
*ha
;
3905 if (netdev_mc_empty(netdev
)) {
3906 /* nothing to program, so clear mc list */
3907 igb_update_mc_addr_list(hw
, NULL
, 0);
3908 igb_restore_vf_multicasts(adapter
);
3912 mta_list
= kzalloc(netdev_mc_count(netdev
) * 6, GFP_ATOMIC
);
3916 /* The shared function expects a packed array of only addresses. */
3918 netdev_for_each_mc_addr(ha
, netdev
)
3919 memcpy(mta_list
+ (i
++ * ETH_ALEN
), ha
->addr
, ETH_ALEN
);
3921 igb_update_mc_addr_list(hw
, mta_list
, i
);
3924 return netdev_mc_count(netdev
);
3928 * igb_write_uc_addr_list - write unicast addresses to RAR table
3929 * @netdev: network interface device structure
3931 * Writes unicast address list to the RAR table.
3932 * Returns: -ENOMEM on failure/insufficient address space
3933 * 0 on no addresses written
3934 * X on writing X addresses to the RAR table
3936 static int igb_write_uc_addr_list(struct net_device
*netdev
)
3938 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3939 struct e1000_hw
*hw
= &adapter
->hw
;
3940 unsigned int vfn
= adapter
->vfs_allocated_count
;
3941 unsigned int rar_entries
= hw
->mac
.rar_entry_count
- (vfn
+ 1);
3944 /* return ENOMEM indicating insufficient memory for addresses */
3945 if (netdev_uc_count(netdev
) > rar_entries
)
3948 if (!netdev_uc_empty(netdev
) && rar_entries
) {
3949 struct netdev_hw_addr
*ha
;
3951 netdev_for_each_uc_addr(ha
, netdev
) {
3954 igb_rar_set_qsel(adapter
, ha
->addr
,
3960 /* write the addresses in reverse order to avoid write combining */
3961 for (; rar_entries
> 0 ; rar_entries
--) {
3962 wr32(E1000_RAH(rar_entries
), 0);
3963 wr32(E1000_RAL(rar_entries
), 0);
3971 * igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
3972 * @netdev: network interface device structure
3974 * The set_rx_mode entry point is called whenever the unicast or multicast
3975 * address lists or the network interface flags are updated. This routine is
3976 * responsible for configuring the hardware for proper unicast, multicast,
3977 * promiscuous mode, and all-multi behavior.
3979 static void igb_set_rx_mode(struct net_device
*netdev
)
3981 struct igb_adapter
*adapter
= netdev_priv(netdev
);
3982 struct e1000_hw
*hw
= &adapter
->hw
;
3983 unsigned int vfn
= adapter
->vfs_allocated_count
;
3984 u32 rctl
, vmolr
= 0;
3987 /* Check for Promiscuous and All Multicast modes */
3988 rctl
= rd32(E1000_RCTL
);
3990 /* clear the effected bits */
3991 rctl
&= ~(E1000_RCTL_UPE
| E1000_RCTL_MPE
| E1000_RCTL_VFE
);
3993 if (netdev
->flags
& IFF_PROMISC
) {
3994 /* retain VLAN HW filtering if in VT mode */
3995 if (adapter
->vfs_allocated_count
)
3996 rctl
|= E1000_RCTL_VFE
;
3997 rctl
|= (E1000_RCTL_UPE
| E1000_RCTL_MPE
);
3998 vmolr
|= (E1000_VMOLR_ROPE
| E1000_VMOLR_MPME
);
4000 if (netdev
->flags
& IFF_ALLMULTI
) {
4001 rctl
|= E1000_RCTL_MPE
;
4002 vmolr
|= E1000_VMOLR_MPME
;
4004 /* Write addresses to the MTA, if the attempt fails
4005 * then we should just turn on promiscuous mode so
4006 * that we can at least receive multicast traffic
4008 count
= igb_write_mc_addr_list(netdev
);
4010 rctl
|= E1000_RCTL_MPE
;
4011 vmolr
|= E1000_VMOLR_MPME
;
4013 vmolr
|= E1000_VMOLR_ROMPE
;
4016 /* Write addresses to available RAR registers, if there is not
4017 * sufficient space to store all the addresses then enable
4018 * unicast promiscuous mode
4020 count
= igb_write_uc_addr_list(netdev
);
4022 rctl
|= E1000_RCTL_UPE
;
4023 vmolr
|= E1000_VMOLR_ROPE
;
4025 rctl
|= E1000_RCTL_VFE
;
4027 wr32(E1000_RCTL
, rctl
);
4029 /* In order to support SR-IOV and eventually VMDq it is necessary to set
4030 * the VMOLR to enable the appropriate modes. Without this workaround
4031 * we will have issues with VLAN tag stripping not being done for frames
4032 * that are only arriving because we are the default pool
4034 if ((hw
->mac
.type
< e1000_82576
) || (hw
->mac
.type
> e1000_i350
))
4037 vmolr
|= rd32(E1000_VMOLR(vfn
)) &
4038 ~(E1000_VMOLR_ROPE
| E1000_VMOLR_MPME
| E1000_VMOLR_ROMPE
);
4040 /* enable Rx jumbo frames, no need for restriction */
4041 vmolr
&= ~E1000_VMOLR_RLPML_MASK
;
4042 vmolr
|= MAX_JUMBO_FRAME_SIZE
| E1000_VMOLR_LPE
;
4044 wr32(E1000_VMOLR(vfn
), vmolr
);
4045 wr32(E1000_RLPML
, MAX_JUMBO_FRAME_SIZE
);
4047 igb_restore_vf_multicasts(adapter
);
4050 static void igb_check_wvbr(struct igb_adapter
*adapter
)
4052 struct e1000_hw
*hw
= &adapter
->hw
;
4055 switch (hw
->mac
.type
) {
4058 wvbr
= rd32(E1000_WVBR
);
4066 adapter
->wvbr
|= wvbr
;
4069 #define IGB_STAGGERED_QUEUE_OFFSET 8
4071 static void igb_spoof_check(struct igb_adapter
*adapter
)
4078 for (j
= 0; j
< adapter
->vfs_allocated_count
; j
++) {
4079 if (adapter
->wvbr
& (1 << j
) ||
4080 adapter
->wvbr
& (1 << (j
+ IGB_STAGGERED_QUEUE_OFFSET
))) {
4081 dev_warn(&adapter
->pdev
->dev
,
4082 "Spoof event(s) detected on VF %d\n", j
);
4085 (1 << (j
+ IGB_STAGGERED_QUEUE_OFFSET
)));
4090 /* Need to wait a few seconds after link up to get diagnostic information from
4093 static void igb_update_phy_info(unsigned long data
)
4095 struct igb_adapter
*adapter
= (struct igb_adapter
*) data
;
4096 igb_get_phy_info(&adapter
->hw
);
4100 * igb_has_link - check shared code for link and determine up/down
4101 * @adapter: pointer to driver private info
4103 bool igb_has_link(struct igb_adapter
*adapter
)
4105 struct e1000_hw
*hw
= &adapter
->hw
;
4106 bool link_active
= false;
4108 /* get_link_status is set on LSC (link status) interrupt or
4109 * rx sequence error interrupt. get_link_status will stay
4110 * false until the e1000_check_for_link establishes link
4111 * for copper adapters ONLY
4113 switch (hw
->phy
.media_type
) {
4114 case e1000_media_type_copper
:
4115 if (!hw
->mac
.get_link_status
)
4117 case e1000_media_type_internal_serdes
:
4118 hw
->mac
.ops
.check_for_link(hw
);
4119 link_active
= !hw
->mac
.get_link_status
;
4122 case e1000_media_type_unknown
:
4126 if (((hw
->mac
.type
== e1000_i210
) ||
4127 (hw
->mac
.type
== e1000_i211
)) &&
4128 (hw
->phy
.id
== I210_I_PHY_ID
)) {
4129 if (!netif_carrier_ok(adapter
->netdev
)) {
4130 adapter
->flags
&= ~IGB_FLAG_NEED_LINK_UPDATE
;
4131 } else if (!(adapter
->flags
& IGB_FLAG_NEED_LINK_UPDATE
)) {
4132 adapter
->flags
|= IGB_FLAG_NEED_LINK_UPDATE
;
4133 adapter
->link_check_timeout
= jiffies
;
4140 static bool igb_thermal_sensor_event(struct e1000_hw
*hw
, u32 event
)
4143 u32 ctrl_ext
, thstat
;
4145 /* check for thermal sensor event on i350 copper only */
4146 if (hw
->mac
.type
== e1000_i350
) {
4147 thstat
= rd32(E1000_THSTAT
);
4148 ctrl_ext
= rd32(E1000_CTRL_EXT
);
4150 if ((hw
->phy
.media_type
== e1000_media_type_copper
) &&
4151 !(ctrl_ext
& E1000_CTRL_EXT_LINK_MODE_SGMII
))
4152 ret
= !!(thstat
& event
);
4159 * igb_check_lvmmc - check for malformed packets received
4160 * and indicated in LVMMC register
4161 * @adapter: pointer to adapter
4163 static void igb_check_lvmmc(struct igb_adapter
*adapter
)
4165 struct e1000_hw
*hw
= &adapter
->hw
;
4168 lvmmc
= rd32(E1000_LVMMC
);
4170 if (unlikely(net_ratelimit())) {
4171 netdev_warn(adapter
->netdev
,
4172 "malformed Tx packet detected and dropped, LVMMC:0x%08x\n",
4179 * igb_watchdog - Timer Call-back
4180 * @data: pointer to adapter cast into an unsigned long
4182 static void igb_watchdog(unsigned long data
)
4184 struct igb_adapter
*adapter
= (struct igb_adapter
*)data
;
4185 /* Do the rest outside of interrupt context */
4186 schedule_work(&adapter
->watchdog_task
);
4189 static void igb_watchdog_task(struct work_struct
*work
)
4191 struct igb_adapter
*adapter
= container_of(work
,
4194 struct e1000_hw
*hw
= &adapter
->hw
;
4195 struct e1000_phy_info
*phy
= &hw
->phy
;
4196 struct net_device
*netdev
= adapter
->netdev
;
4201 link
= igb_has_link(adapter
);
4203 if (adapter
->flags
& IGB_FLAG_NEED_LINK_UPDATE
) {
4204 if (time_after(jiffies
, (adapter
->link_check_timeout
+ HZ
)))
4205 adapter
->flags
&= ~IGB_FLAG_NEED_LINK_UPDATE
;
4210 /* Force link down if we have fiber to swap to */
4211 if (adapter
->flags
& IGB_FLAG_MAS_ENABLE
) {
4212 if (hw
->phy
.media_type
== e1000_media_type_copper
) {
4213 connsw
= rd32(E1000_CONNSW
);
4214 if (!(connsw
& E1000_CONNSW_AUTOSENSE_EN
))
4219 /* Perform a reset if the media type changed. */
4220 if (hw
->dev_spec
._82575
.media_changed
) {
4221 hw
->dev_spec
._82575
.media_changed
= false;
4222 adapter
->flags
|= IGB_FLAG_MEDIA_RESET
;
4225 /* Cancel scheduled suspend requests. */
4226 pm_runtime_resume(netdev
->dev
.parent
);
4228 if (!netif_carrier_ok(netdev
)) {
4231 hw
->mac
.ops
.get_speed_and_duplex(hw
,
4232 &adapter
->link_speed
,
4233 &adapter
->link_duplex
);
4235 ctrl
= rd32(E1000_CTRL
);
4236 /* Links status message must follow this format */
4238 "igb: %s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
4240 adapter
->link_speed
,
4241 adapter
->link_duplex
== FULL_DUPLEX
?
4243 (ctrl
& E1000_CTRL_TFCE
) &&
4244 (ctrl
& E1000_CTRL_RFCE
) ? "RX/TX" :
4245 (ctrl
& E1000_CTRL_RFCE
) ? "RX" :
4246 (ctrl
& E1000_CTRL_TFCE
) ? "TX" : "None");
4248 /* disable EEE if enabled */
4249 if ((adapter
->flags
& IGB_FLAG_EEE
) &&
4250 (adapter
->link_duplex
== HALF_DUPLEX
)) {
4251 dev_info(&adapter
->pdev
->dev
,
4252 "EEE Disabled: unsupported at half duplex. Re-enable using ethtool when at full duplex.\n");
4253 adapter
->hw
.dev_spec
._82575
.eee_disable
= true;
4254 adapter
->flags
&= ~IGB_FLAG_EEE
;
4257 /* check if SmartSpeed worked */
4258 igb_check_downshift(hw
);
4259 if (phy
->speed_downgraded
)
4260 netdev_warn(netdev
, "Link Speed was downgraded by SmartSpeed\n");
4262 /* check for thermal sensor event */
4263 if (igb_thermal_sensor_event(hw
,
4264 E1000_THSTAT_LINK_THROTTLE
))
4265 netdev_info(netdev
, "The network adapter link speed was downshifted because it overheated\n");
4267 /* adjust timeout factor according to speed/duplex */
4268 adapter
->tx_timeout_factor
= 1;
4269 switch (adapter
->link_speed
) {
4271 adapter
->tx_timeout_factor
= 14;
4274 /* maybe add some timeout factor ? */
4278 netif_carrier_on(netdev
);
4280 igb_ping_all_vfs(adapter
);
4281 igb_check_vf_rate_limit(adapter
);
4283 /* link state has changed, schedule phy info update */
4284 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
4285 mod_timer(&adapter
->phy_info_timer
,
4286 round_jiffies(jiffies
+ 2 * HZ
));
4289 if (netif_carrier_ok(netdev
)) {
4290 adapter
->link_speed
= 0;
4291 adapter
->link_duplex
= 0;
4293 /* check for thermal sensor event */
4294 if (igb_thermal_sensor_event(hw
,
4295 E1000_THSTAT_PWR_DOWN
)) {
4296 netdev_err(netdev
, "The network adapter was stopped because it overheated\n");
4299 /* Links status message must follow this format */
4300 netdev_info(netdev
, "igb: %s NIC Link is Down\n",
4302 netif_carrier_off(netdev
);
4304 igb_ping_all_vfs(adapter
);
4306 /* link state has changed, schedule phy info update */
4307 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
4308 mod_timer(&adapter
->phy_info_timer
,
4309 round_jiffies(jiffies
+ 2 * HZ
));
4311 /* link is down, time to check for alternate media */
4312 if (adapter
->flags
& IGB_FLAG_MAS_ENABLE
) {
4313 igb_check_swap_media(adapter
);
4314 if (adapter
->flags
& IGB_FLAG_MEDIA_RESET
) {
4315 schedule_work(&adapter
->reset_task
);
4316 /* return immediately */
4320 pm_schedule_suspend(netdev
->dev
.parent
,
4323 /* also check for alternate media here */
4324 } else if (!netif_carrier_ok(netdev
) &&
4325 (adapter
->flags
& IGB_FLAG_MAS_ENABLE
)) {
4326 igb_check_swap_media(adapter
);
4327 if (adapter
->flags
& IGB_FLAG_MEDIA_RESET
) {
4328 schedule_work(&adapter
->reset_task
);
4329 /* return immediately */
4335 spin_lock(&adapter
->stats64_lock
);
4336 igb_update_stats(adapter
, &adapter
->stats64
);
4337 spin_unlock(&adapter
->stats64_lock
);
4339 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
4340 struct igb_ring
*tx_ring
= adapter
->tx_ring
[i
];
4341 if (!netif_carrier_ok(netdev
)) {
4342 /* We've lost link, so the controller stops DMA,
4343 * but we've got queued Tx work that's never going
4344 * to get done, so reset controller to flush Tx.
4345 * (Do the reset outside of interrupt context).
4347 if (igb_desc_unused(tx_ring
) + 1 < tx_ring
->count
) {
4348 adapter
->tx_timeout_count
++;
4349 schedule_work(&adapter
->reset_task
);
4350 /* return immediately since reset is imminent */
4355 /* Force detection of hung controller every watchdog period */
4356 set_bit(IGB_RING_FLAG_TX_DETECT_HANG
, &tx_ring
->flags
);
4359 /* Cause software interrupt to ensure Rx ring is cleaned */
4360 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
) {
4363 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
4364 eics
|= adapter
->q_vector
[i
]->eims_value
;
4365 wr32(E1000_EICS
, eics
);
4367 wr32(E1000_ICS
, E1000_ICS_RXDMT0
);
4370 igb_spoof_check(adapter
);
4371 igb_ptp_rx_hang(adapter
);
4373 /* Check LVMMC register on i350/i354 only */
4374 if ((adapter
->hw
.mac
.type
== e1000_i350
) ||
4375 (adapter
->hw
.mac
.type
== e1000_i354
))
4376 igb_check_lvmmc(adapter
);
4378 /* Reset the timer */
4379 if (!test_bit(__IGB_DOWN
, &adapter
->state
)) {
4380 if (adapter
->flags
& IGB_FLAG_NEED_LINK_UPDATE
)
4381 mod_timer(&adapter
->watchdog_timer
,
4382 round_jiffies(jiffies
+ HZ
));
4384 mod_timer(&adapter
->watchdog_timer
,
4385 round_jiffies(jiffies
+ 2 * HZ
));
4389 enum latency_range
{
4393 latency_invalid
= 255
4397 * igb_update_ring_itr - update the dynamic ITR value based on packet size
4398 * @q_vector: pointer to q_vector
4400 * Stores a new ITR value based on strictly on packet size. This
4401 * algorithm is less sophisticated than that used in igb_update_itr,
4402 * due to the difficulty of synchronizing statistics across multiple
4403 * receive rings. The divisors and thresholds used by this function
4404 * were determined based on theoretical maximum wire speed and testing
4405 * data, in order to minimize response time while increasing bulk
4407 * This functionality is controlled by ethtool's coalescing settings.
4408 * NOTE: This function is called only when operating in a multiqueue
4409 * receive environment.
4411 static void igb_update_ring_itr(struct igb_q_vector
*q_vector
)
4413 int new_val
= q_vector
->itr_val
;
4414 int avg_wire_size
= 0;
4415 struct igb_adapter
*adapter
= q_vector
->adapter
;
4416 unsigned int packets
;
4418 /* For non-gigabit speeds, just fix the interrupt rate at 4000
4419 * ints/sec - ITR timer value of 120 ticks.
4421 if (adapter
->link_speed
!= SPEED_1000
) {
4422 new_val
= IGB_4K_ITR
;
4426 packets
= q_vector
->rx
.total_packets
;
4428 avg_wire_size
= q_vector
->rx
.total_bytes
/ packets
;
4430 packets
= q_vector
->tx
.total_packets
;
4432 avg_wire_size
= max_t(u32
, avg_wire_size
,
4433 q_vector
->tx
.total_bytes
/ packets
);
4435 /* if avg_wire_size isn't set no work was done */
4439 /* Add 24 bytes to size to account for CRC, preamble, and gap */
4440 avg_wire_size
+= 24;
4442 /* Don't starve jumbo frames */
4443 avg_wire_size
= min(avg_wire_size
, 3000);
4445 /* Give a little boost to mid-size frames */
4446 if ((avg_wire_size
> 300) && (avg_wire_size
< 1200))
4447 new_val
= avg_wire_size
/ 3;
4449 new_val
= avg_wire_size
/ 2;
4451 /* conservative mode (itr 3) eliminates the lowest_latency setting */
4452 if (new_val
< IGB_20K_ITR
&&
4453 ((q_vector
->rx
.ring
&& adapter
->rx_itr_setting
== 3) ||
4454 (!q_vector
->rx
.ring
&& adapter
->tx_itr_setting
== 3)))
4455 new_val
= IGB_20K_ITR
;
4458 if (new_val
!= q_vector
->itr_val
) {
4459 q_vector
->itr_val
= new_val
;
4460 q_vector
->set_itr
= 1;
4463 q_vector
->rx
.total_bytes
= 0;
4464 q_vector
->rx
.total_packets
= 0;
4465 q_vector
->tx
.total_bytes
= 0;
4466 q_vector
->tx
.total_packets
= 0;
4470 * igb_update_itr - update the dynamic ITR value based on statistics
4471 * @q_vector: pointer to q_vector
4472 * @ring_container: ring info to update the itr for
4474 * Stores a new ITR value based on packets and byte
4475 * counts during the last interrupt. The advantage of per interrupt
4476 * computation is faster updates and more accurate ITR for the current
4477 * traffic pattern. Constants in this function were computed
4478 * based on theoretical maximum wire speed and thresholds were set based
4479 * on testing data as well as attempting to minimize response time
4480 * while increasing bulk throughput.
4481 * This functionality is controlled by ethtool's coalescing settings.
4482 * NOTE: These calculations are only valid when operating in a single-
4483 * queue environment.
4485 static void igb_update_itr(struct igb_q_vector
*q_vector
,
4486 struct igb_ring_container
*ring_container
)
4488 unsigned int packets
= ring_container
->total_packets
;
4489 unsigned int bytes
= ring_container
->total_bytes
;
4490 u8 itrval
= ring_container
->itr
;
4492 /* no packets, exit with status unchanged */
4497 case lowest_latency
:
4498 /* handle TSO and jumbo frames */
4499 if (bytes
/packets
> 8000)
4500 itrval
= bulk_latency
;
4501 else if ((packets
< 5) && (bytes
> 512))
4502 itrval
= low_latency
;
4504 case low_latency
: /* 50 usec aka 20000 ints/s */
4505 if (bytes
> 10000) {
4506 /* this if handles the TSO accounting */
4507 if (bytes
/packets
> 8000)
4508 itrval
= bulk_latency
;
4509 else if ((packets
< 10) || ((bytes
/packets
) > 1200))
4510 itrval
= bulk_latency
;
4511 else if ((packets
> 35))
4512 itrval
= lowest_latency
;
4513 } else if (bytes
/packets
> 2000) {
4514 itrval
= bulk_latency
;
4515 } else if (packets
<= 2 && bytes
< 512) {
4516 itrval
= lowest_latency
;
4519 case bulk_latency
: /* 250 usec aka 4000 ints/s */
4520 if (bytes
> 25000) {
4522 itrval
= low_latency
;
4523 } else if (bytes
< 1500) {
4524 itrval
= low_latency
;
4529 /* clear work counters since we have the values we need */
4530 ring_container
->total_bytes
= 0;
4531 ring_container
->total_packets
= 0;
4533 /* write updated itr to ring container */
4534 ring_container
->itr
= itrval
;
4537 static void igb_set_itr(struct igb_q_vector
*q_vector
)
4539 struct igb_adapter
*adapter
= q_vector
->adapter
;
4540 u32 new_itr
= q_vector
->itr_val
;
4543 /* for non-gigabit speeds, just fix the interrupt rate at 4000 */
4544 if (adapter
->link_speed
!= SPEED_1000
) {
4546 new_itr
= IGB_4K_ITR
;
4550 igb_update_itr(q_vector
, &q_vector
->tx
);
4551 igb_update_itr(q_vector
, &q_vector
->rx
);
4553 current_itr
= max(q_vector
->rx
.itr
, q_vector
->tx
.itr
);
4555 /* conservative mode (itr 3) eliminates the lowest_latency setting */
4556 if (current_itr
== lowest_latency
&&
4557 ((q_vector
->rx
.ring
&& adapter
->rx_itr_setting
== 3) ||
4558 (!q_vector
->rx
.ring
&& adapter
->tx_itr_setting
== 3)))
4559 current_itr
= low_latency
;
4561 switch (current_itr
) {
4562 /* counts and packets in update_itr are dependent on these numbers */
4563 case lowest_latency
:
4564 new_itr
= IGB_70K_ITR
; /* 70,000 ints/sec */
4567 new_itr
= IGB_20K_ITR
; /* 20,000 ints/sec */
4570 new_itr
= IGB_4K_ITR
; /* 4,000 ints/sec */
4577 if (new_itr
!= q_vector
->itr_val
) {
4578 /* this attempts to bias the interrupt rate towards Bulk
4579 * by adding intermediate steps when interrupt rate is
4582 new_itr
= new_itr
> q_vector
->itr_val
?
4583 max((new_itr
* q_vector
->itr_val
) /
4584 (new_itr
+ (q_vector
->itr_val
>> 2)),
4586 /* Don't write the value here; it resets the adapter's
4587 * internal timer, and causes us to delay far longer than
4588 * we should between interrupts. Instead, we write the ITR
4589 * value at the beginning of the next interrupt so the timing
4590 * ends up being correct.
4592 q_vector
->itr_val
= new_itr
;
4593 q_vector
->set_itr
= 1;
4597 static void igb_tx_ctxtdesc(struct igb_ring
*tx_ring
, u32 vlan_macip_lens
,
4598 u32 type_tucmd
, u32 mss_l4len_idx
)
4600 struct e1000_adv_tx_context_desc
*context_desc
;
4601 u16 i
= tx_ring
->next_to_use
;
4603 context_desc
= IGB_TX_CTXTDESC(tx_ring
, i
);
4606 tx_ring
->next_to_use
= (i
< tx_ring
->count
) ? i
: 0;
4608 /* set bits to identify this as an advanced context descriptor */
4609 type_tucmd
|= E1000_TXD_CMD_DEXT
| E1000_ADVTXD_DTYP_CTXT
;
4611 /* For 82575, context index must be unique per ring. */
4612 if (test_bit(IGB_RING_FLAG_TX_CTX_IDX
, &tx_ring
->flags
))
4613 mss_l4len_idx
|= tx_ring
->reg_idx
<< 4;
4615 context_desc
->vlan_macip_lens
= cpu_to_le32(vlan_macip_lens
);
4616 context_desc
->seqnum_seed
= 0;
4617 context_desc
->type_tucmd_mlhl
= cpu_to_le32(type_tucmd
);
4618 context_desc
->mss_l4len_idx
= cpu_to_le32(mss_l4len_idx
);
4621 static int igb_tso(struct igb_ring
*tx_ring
,
4622 struct igb_tx_buffer
*first
,
4625 struct sk_buff
*skb
= first
->skb
;
4626 u32 vlan_macip_lens
, type_tucmd
;
4627 u32 mss_l4len_idx
, l4len
;
4630 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
4633 if (!skb_is_gso(skb
))
4636 err
= skb_cow_head(skb
, 0);
4640 /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
4641 type_tucmd
= E1000_ADVTXD_TUCMD_L4T_TCP
;
4643 if (first
->protocol
== htons(ETH_P_IP
)) {
4644 struct iphdr
*iph
= ip_hdr(skb
);
4647 tcp_hdr(skb
)->check
= ~csum_tcpudp_magic(iph
->saddr
,
4651 type_tucmd
|= E1000_ADVTXD_TUCMD_IPV4
;
4652 first
->tx_flags
|= IGB_TX_FLAGS_TSO
|
4655 } else if (skb_is_gso_v6(skb
)) {
4656 ipv6_hdr(skb
)->payload_len
= 0;
4657 tcp_hdr(skb
)->check
= ~csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
4658 &ipv6_hdr(skb
)->daddr
,
4660 first
->tx_flags
|= IGB_TX_FLAGS_TSO
|
4664 /* compute header lengths */
4665 l4len
= tcp_hdrlen(skb
);
4666 *hdr_len
= skb_transport_offset(skb
) + l4len
;
4668 /* update gso size and bytecount with header size */
4669 first
->gso_segs
= skb_shinfo(skb
)->gso_segs
;
4670 first
->bytecount
+= (first
->gso_segs
- 1) * *hdr_len
;
4673 mss_l4len_idx
= l4len
<< E1000_ADVTXD_L4LEN_SHIFT
;
4674 mss_l4len_idx
|= skb_shinfo(skb
)->gso_size
<< E1000_ADVTXD_MSS_SHIFT
;
4676 /* VLAN MACLEN IPLEN */
4677 vlan_macip_lens
= skb_network_header_len(skb
);
4678 vlan_macip_lens
|= skb_network_offset(skb
) << E1000_ADVTXD_MACLEN_SHIFT
;
4679 vlan_macip_lens
|= first
->tx_flags
& IGB_TX_FLAGS_VLAN_MASK
;
4681 igb_tx_ctxtdesc(tx_ring
, vlan_macip_lens
, type_tucmd
, mss_l4len_idx
);
4686 static void igb_tx_csum(struct igb_ring
*tx_ring
, struct igb_tx_buffer
*first
)
4688 struct sk_buff
*skb
= first
->skb
;
4689 u32 vlan_macip_lens
= 0;
4690 u32 mss_l4len_idx
= 0;
4693 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
4694 if (!(first
->tx_flags
& IGB_TX_FLAGS_VLAN
))
4699 switch (first
->protocol
) {
4700 case htons(ETH_P_IP
):
4701 vlan_macip_lens
|= skb_network_header_len(skb
);
4702 type_tucmd
|= E1000_ADVTXD_TUCMD_IPV4
;
4703 l4_hdr
= ip_hdr(skb
)->protocol
;
4705 case htons(ETH_P_IPV6
):
4706 vlan_macip_lens
|= skb_network_header_len(skb
);
4707 l4_hdr
= ipv6_hdr(skb
)->nexthdr
;
4710 if (unlikely(net_ratelimit())) {
4711 dev_warn(tx_ring
->dev
,
4712 "partial checksum but proto=%x!\n",
4720 type_tucmd
|= E1000_ADVTXD_TUCMD_L4T_TCP
;
4721 mss_l4len_idx
= tcp_hdrlen(skb
) <<
4722 E1000_ADVTXD_L4LEN_SHIFT
;
4725 type_tucmd
|= E1000_ADVTXD_TUCMD_L4T_SCTP
;
4726 mss_l4len_idx
= sizeof(struct sctphdr
) <<
4727 E1000_ADVTXD_L4LEN_SHIFT
;
4730 mss_l4len_idx
= sizeof(struct udphdr
) <<
4731 E1000_ADVTXD_L4LEN_SHIFT
;
4734 if (unlikely(net_ratelimit())) {
4735 dev_warn(tx_ring
->dev
,
4736 "partial checksum but l4 proto=%x!\n",
4742 /* update TX checksum flag */
4743 first
->tx_flags
|= IGB_TX_FLAGS_CSUM
;
4746 vlan_macip_lens
|= skb_network_offset(skb
) << E1000_ADVTXD_MACLEN_SHIFT
;
4747 vlan_macip_lens
|= first
->tx_flags
& IGB_TX_FLAGS_VLAN_MASK
;
4749 igb_tx_ctxtdesc(tx_ring
, vlan_macip_lens
, type_tucmd
, mss_l4len_idx
);
4752 #define IGB_SET_FLAG(_input, _flag, _result) \
4753 ((_flag <= _result) ? \
4754 ((u32)(_input & _flag) * (_result / _flag)) : \
4755 ((u32)(_input & _flag) / (_flag / _result)))
4757 static u32
igb_tx_cmd_type(struct sk_buff
*skb
, u32 tx_flags
)
4759 /* set type for advanced descriptor with frame checksum insertion */
4760 u32 cmd_type
= E1000_ADVTXD_DTYP_DATA
|
4761 E1000_ADVTXD_DCMD_DEXT
|
4762 E1000_ADVTXD_DCMD_IFCS
;
4764 /* set HW vlan bit if vlan is present */
4765 cmd_type
|= IGB_SET_FLAG(tx_flags
, IGB_TX_FLAGS_VLAN
,
4766 (E1000_ADVTXD_DCMD_VLE
));
4768 /* set segmentation bits for TSO */
4769 cmd_type
|= IGB_SET_FLAG(tx_flags
, IGB_TX_FLAGS_TSO
,
4770 (E1000_ADVTXD_DCMD_TSE
));
4772 /* set timestamp bit if present */
4773 cmd_type
|= IGB_SET_FLAG(tx_flags
, IGB_TX_FLAGS_TSTAMP
,
4774 (E1000_ADVTXD_MAC_TSTAMP
));
4776 /* insert frame checksum */
4777 cmd_type
^= IGB_SET_FLAG(skb
->no_fcs
, 1, E1000_ADVTXD_DCMD_IFCS
);
4782 static void igb_tx_olinfo_status(struct igb_ring
*tx_ring
,
4783 union e1000_adv_tx_desc
*tx_desc
,
4784 u32 tx_flags
, unsigned int paylen
)
4786 u32 olinfo_status
= paylen
<< E1000_ADVTXD_PAYLEN_SHIFT
;
4788 /* 82575 requires a unique index per ring */
4789 if (test_bit(IGB_RING_FLAG_TX_CTX_IDX
, &tx_ring
->flags
))
4790 olinfo_status
|= tx_ring
->reg_idx
<< 4;
4792 /* insert L4 checksum */
4793 olinfo_status
|= IGB_SET_FLAG(tx_flags
,
4795 (E1000_TXD_POPTS_TXSM
<< 8));
4797 /* insert IPv4 checksum */
4798 olinfo_status
|= IGB_SET_FLAG(tx_flags
,
4800 (E1000_TXD_POPTS_IXSM
<< 8));
4802 tx_desc
->read
.olinfo_status
= cpu_to_le32(olinfo_status
);
4805 static int __igb_maybe_stop_tx(struct igb_ring
*tx_ring
, const u16 size
)
4807 struct net_device
*netdev
= tx_ring
->netdev
;
4809 netif_stop_subqueue(netdev
, tx_ring
->queue_index
);
4811 /* Herbert's original patch had:
4812 * smp_mb__after_netif_stop_queue();
4813 * but since that doesn't exist yet, just open code it.
4817 /* We need to check again in a case another CPU has just
4818 * made room available.
4820 if (igb_desc_unused(tx_ring
) < size
)
4824 netif_wake_subqueue(netdev
, tx_ring
->queue_index
);
4826 u64_stats_update_begin(&tx_ring
->tx_syncp2
);
4827 tx_ring
->tx_stats
.restart_queue2
++;
4828 u64_stats_update_end(&tx_ring
->tx_syncp2
);
4833 static inline int igb_maybe_stop_tx(struct igb_ring
*tx_ring
, const u16 size
)
4835 if (igb_desc_unused(tx_ring
) >= size
)
4837 return __igb_maybe_stop_tx(tx_ring
, size
);
4840 static void igb_tx_map(struct igb_ring
*tx_ring
,
4841 struct igb_tx_buffer
*first
,
4844 struct sk_buff
*skb
= first
->skb
;
4845 struct igb_tx_buffer
*tx_buffer
;
4846 union e1000_adv_tx_desc
*tx_desc
;
4847 struct skb_frag_struct
*frag
;
4849 unsigned int data_len
, size
;
4850 u32 tx_flags
= first
->tx_flags
;
4851 u32 cmd_type
= igb_tx_cmd_type(skb
, tx_flags
);
4852 u16 i
= tx_ring
->next_to_use
;
4854 tx_desc
= IGB_TX_DESC(tx_ring
, i
);
4856 igb_tx_olinfo_status(tx_ring
, tx_desc
, tx_flags
, skb
->len
- hdr_len
);
4858 size
= skb_headlen(skb
);
4859 data_len
= skb
->data_len
;
4861 dma
= dma_map_single(tx_ring
->dev
, skb
->data
, size
, DMA_TO_DEVICE
);
4865 for (frag
= &skb_shinfo(skb
)->frags
[0];; frag
++) {
4866 if (dma_mapping_error(tx_ring
->dev
, dma
))
4869 /* record length, and DMA address */
4870 dma_unmap_len_set(tx_buffer
, len
, size
);
4871 dma_unmap_addr_set(tx_buffer
, dma
, dma
);
4873 tx_desc
->read
.buffer_addr
= cpu_to_le64(dma
);
4875 while (unlikely(size
> IGB_MAX_DATA_PER_TXD
)) {
4876 tx_desc
->read
.cmd_type_len
=
4877 cpu_to_le32(cmd_type
^ IGB_MAX_DATA_PER_TXD
);
4881 if (i
== tx_ring
->count
) {
4882 tx_desc
= IGB_TX_DESC(tx_ring
, 0);
4885 tx_desc
->read
.olinfo_status
= 0;
4887 dma
+= IGB_MAX_DATA_PER_TXD
;
4888 size
-= IGB_MAX_DATA_PER_TXD
;
4890 tx_desc
->read
.buffer_addr
= cpu_to_le64(dma
);
4893 if (likely(!data_len
))
4896 tx_desc
->read
.cmd_type_len
= cpu_to_le32(cmd_type
^ size
);
4900 if (i
== tx_ring
->count
) {
4901 tx_desc
= IGB_TX_DESC(tx_ring
, 0);
4904 tx_desc
->read
.olinfo_status
= 0;
4906 size
= skb_frag_size(frag
);
4909 dma
= skb_frag_dma_map(tx_ring
->dev
, frag
, 0,
4910 size
, DMA_TO_DEVICE
);
4912 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
4915 /* write last descriptor with RS and EOP bits */
4916 cmd_type
|= size
| IGB_TXD_DCMD
;
4917 tx_desc
->read
.cmd_type_len
= cpu_to_le32(cmd_type
);
4919 netdev_tx_sent_queue(txring_txq(tx_ring
), first
->bytecount
);
4921 /* set the timestamp */
4922 first
->time_stamp
= jiffies
;
4924 /* Force memory writes to complete before letting h/w know there
4925 * are new descriptors to fetch. (Only applicable for weak-ordered
4926 * memory model archs, such as IA-64).
4928 * We also need this memory barrier to make certain all of the
4929 * status bits have been updated before next_to_watch is written.
4933 /* set next_to_watch value indicating a packet is present */
4934 first
->next_to_watch
= tx_desc
;
4937 if (i
== tx_ring
->count
)
4940 tx_ring
->next_to_use
= i
;
4942 /* Make sure there is space in the ring for the next send. */
4943 igb_maybe_stop_tx(tx_ring
, DESC_NEEDED
);
4945 if (netif_xmit_stopped(txring_txq(tx_ring
)) || !skb
->xmit_more
) {
4946 writel(i
, tx_ring
->tail
);
4948 /* we need this if more than one processor can write to our tail
4949 * at a time, it synchronizes IO on IA64/Altix systems
4956 dev_err(tx_ring
->dev
, "TX DMA map failed\n");
4958 /* clear dma mappings for failed tx_buffer_info map */
4960 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
4961 igb_unmap_and_free_tx_resource(tx_ring
, tx_buffer
);
4962 if (tx_buffer
== first
)
4969 tx_ring
->next_to_use
= i
;
4972 netdev_tx_t
igb_xmit_frame_ring(struct sk_buff
*skb
,
4973 struct igb_ring
*tx_ring
)
4975 struct igb_tx_buffer
*first
;
4979 u16 count
= TXD_USE_COUNT(skb_headlen(skb
));
4980 __be16 protocol
= vlan_get_protocol(skb
);
4983 /* need: 1 descriptor per page * PAGE_SIZE/IGB_MAX_DATA_PER_TXD,
4984 * + 1 desc for skb_headlen/IGB_MAX_DATA_PER_TXD,
4985 * + 2 desc gap to keep tail from touching head,
4986 * + 1 desc for context descriptor,
4987 * otherwise try next time
4989 for (f
= 0; f
< skb_shinfo(skb
)->nr_frags
; f
++)
4990 count
+= TXD_USE_COUNT(skb_shinfo(skb
)->frags
[f
].size
);
4992 if (igb_maybe_stop_tx(tx_ring
, count
+ 3)) {
4993 /* this is a hard error */
4994 return NETDEV_TX_BUSY
;
4997 /* record the location of the first descriptor for this packet */
4998 first
= &tx_ring
->tx_buffer_info
[tx_ring
->next_to_use
];
5000 first
->bytecount
= skb
->len
;
5001 first
->gso_segs
= 1;
5003 if (unlikely(skb_shinfo(skb
)->tx_flags
& SKBTX_HW_TSTAMP
)) {
5004 struct igb_adapter
*adapter
= netdev_priv(tx_ring
->netdev
);
5006 if (!test_and_set_bit_lock(__IGB_PTP_TX_IN_PROGRESS
,
5008 skb_shinfo(skb
)->tx_flags
|= SKBTX_IN_PROGRESS
;
5009 tx_flags
|= IGB_TX_FLAGS_TSTAMP
;
5011 adapter
->ptp_tx_skb
= skb_get(skb
);
5012 adapter
->ptp_tx_start
= jiffies
;
5013 if (adapter
->hw
.mac
.type
== e1000_82576
)
5014 schedule_work(&adapter
->ptp_tx_work
);
5018 skb_tx_timestamp(skb
);
5020 if (skb_vlan_tag_present(skb
)) {
5021 tx_flags
|= IGB_TX_FLAGS_VLAN
;
5022 tx_flags
|= (skb_vlan_tag_get(skb
) << IGB_TX_FLAGS_VLAN_SHIFT
);
5025 /* record initial flags and protocol */
5026 first
->tx_flags
= tx_flags
;
5027 first
->protocol
= protocol
;
5029 tso
= igb_tso(tx_ring
, first
, &hdr_len
);
5033 igb_tx_csum(tx_ring
, first
);
5035 igb_tx_map(tx_ring
, first
, hdr_len
);
5037 return NETDEV_TX_OK
;
5040 igb_unmap_and_free_tx_resource(tx_ring
, first
);
5042 return NETDEV_TX_OK
;
5045 static inline struct igb_ring
*igb_tx_queue_mapping(struct igb_adapter
*adapter
,
5046 struct sk_buff
*skb
)
5048 unsigned int r_idx
= skb
->queue_mapping
;
5050 if (r_idx
>= adapter
->num_tx_queues
)
5051 r_idx
= r_idx
% adapter
->num_tx_queues
;
5053 return adapter
->tx_ring
[r_idx
];
5056 static netdev_tx_t
igb_xmit_frame(struct sk_buff
*skb
,
5057 struct net_device
*netdev
)
5059 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5061 if (test_bit(__IGB_DOWN
, &adapter
->state
)) {
5062 dev_kfree_skb_any(skb
);
5063 return NETDEV_TX_OK
;
5066 if (skb
->len
<= 0) {
5067 dev_kfree_skb_any(skb
);
5068 return NETDEV_TX_OK
;
5071 /* The minimum packet size with TCTL.PSP set is 17 so pad the skb
5072 * in order to meet this minimum size requirement.
5074 if (skb_put_padto(skb
, 17))
5075 return NETDEV_TX_OK
;
5077 return igb_xmit_frame_ring(skb
, igb_tx_queue_mapping(adapter
, skb
));
5081 * igb_tx_timeout - Respond to a Tx Hang
5082 * @netdev: network interface device structure
5084 static void igb_tx_timeout(struct net_device
*netdev
)
5086 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5087 struct e1000_hw
*hw
= &adapter
->hw
;
5089 /* Do the reset outside of interrupt context */
5090 adapter
->tx_timeout_count
++;
5092 if (hw
->mac
.type
>= e1000_82580
)
5093 hw
->dev_spec
._82575
.global_device_reset
= true;
5095 schedule_work(&adapter
->reset_task
);
5097 (adapter
->eims_enable_mask
& ~adapter
->eims_other
));
5100 static void igb_reset_task(struct work_struct
*work
)
5102 struct igb_adapter
*adapter
;
5103 adapter
= container_of(work
, struct igb_adapter
, reset_task
);
5106 netdev_err(adapter
->netdev
, "Reset adapter\n");
5107 igb_reinit_locked(adapter
);
5111 * igb_get_stats64 - Get System Network Statistics
5112 * @netdev: network interface device structure
5113 * @stats: rtnl_link_stats64 pointer
5115 static struct rtnl_link_stats64
*igb_get_stats64(struct net_device
*netdev
,
5116 struct rtnl_link_stats64
*stats
)
5118 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5120 spin_lock(&adapter
->stats64_lock
);
5121 igb_update_stats(adapter
, &adapter
->stats64
);
5122 memcpy(stats
, &adapter
->stats64
, sizeof(*stats
));
5123 spin_unlock(&adapter
->stats64_lock
);
5129 * igb_change_mtu - Change the Maximum Transfer Unit
5130 * @netdev: network interface device structure
5131 * @new_mtu: new value for maximum frame size
5133 * Returns 0 on success, negative on failure
5135 static int igb_change_mtu(struct net_device
*netdev
, int new_mtu
)
5137 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5138 struct pci_dev
*pdev
= adapter
->pdev
;
5139 int max_frame
= new_mtu
+ ETH_HLEN
+ ETH_FCS_LEN
+ VLAN_HLEN
;
5141 if ((new_mtu
< 68) || (max_frame
> MAX_JUMBO_FRAME_SIZE
)) {
5142 dev_err(&pdev
->dev
, "Invalid MTU setting\n");
5146 #define MAX_STD_JUMBO_FRAME_SIZE 9238
5147 if (max_frame
> MAX_STD_JUMBO_FRAME_SIZE
) {
5148 dev_err(&pdev
->dev
, "MTU > 9216 not supported.\n");
5152 /* adjust max frame to be at least the size of a standard frame */
5153 if (max_frame
< (ETH_FRAME_LEN
+ ETH_FCS_LEN
))
5154 max_frame
= ETH_FRAME_LEN
+ ETH_FCS_LEN
;
5156 while (test_and_set_bit(__IGB_RESETTING
, &adapter
->state
))
5157 usleep_range(1000, 2000);
5159 /* igb_down has a dependency on max_frame_size */
5160 adapter
->max_frame_size
= max_frame
;
5162 if (netif_running(netdev
))
5165 dev_info(&pdev
->dev
, "changing MTU from %d to %d\n",
5166 netdev
->mtu
, new_mtu
);
5167 netdev
->mtu
= new_mtu
;
5169 if (netif_running(netdev
))
5174 clear_bit(__IGB_RESETTING
, &adapter
->state
);
5180 * igb_update_stats - Update the board statistics counters
5181 * @adapter: board private structure
5183 void igb_update_stats(struct igb_adapter
*adapter
,
5184 struct rtnl_link_stats64
*net_stats
)
5186 struct e1000_hw
*hw
= &adapter
->hw
;
5187 struct pci_dev
*pdev
= adapter
->pdev
;
5192 u64 _bytes
, _packets
;
5194 /* Prevent stats update while adapter is being reset, or if the pci
5195 * connection is down.
5197 if (adapter
->link_speed
== 0)
5199 if (pci_channel_offline(pdev
))
5206 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
5207 struct igb_ring
*ring
= adapter
->rx_ring
[i
];
5208 u32 rqdpc
= rd32(E1000_RQDPC(i
));
5209 if (hw
->mac
.type
>= e1000_i210
)
5210 wr32(E1000_RQDPC(i
), 0);
5213 ring
->rx_stats
.drops
+= rqdpc
;
5214 net_stats
->rx_fifo_errors
+= rqdpc
;
5218 start
= u64_stats_fetch_begin_irq(&ring
->rx_syncp
);
5219 _bytes
= ring
->rx_stats
.bytes
;
5220 _packets
= ring
->rx_stats
.packets
;
5221 } while (u64_stats_fetch_retry_irq(&ring
->rx_syncp
, start
));
5223 packets
+= _packets
;
5226 net_stats
->rx_bytes
= bytes
;
5227 net_stats
->rx_packets
= packets
;
5231 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
5232 struct igb_ring
*ring
= adapter
->tx_ring
[i
];
5234 start
= u64_stats_fetch_begin_irq(&ring
->tx_syncp
);
5235 _bytes
= ring
->tx_stats
.bytes
;
5236 _packets
= ring
->tx_stats
.packets
;
5237 } while (u64_stats_fetch_retry_irq(&ring
->tx_syncp
, start
));
5239 packets
+= _packets
;
5241 net_stats
->tx_bytes
= bytes
;
5242 net_stats
->tx_packets
= packets
;
5245 /* read stats registers */
5246 adapter
->stats
.crcerrs
+= rd32(E1000_CRCERRS
);
5247 adapter
->stats
.gprc
+= rd32(E1000_GPRC
);
5248 adapter
->stats
.gorc
+= rd32(E1000_GORCL
);
5249 rd32(E1000_GORCH
); /* clear GORCL */
5250 adapter
->stats
.bprc
+= rd32(E1000_BPRC
);
5251 adapter
->stats
.mprc
+= rd32(E1000_MPRC
);
5252 adapter
->stats
.roc
+= rd32(E1000_ROC
);
5254 adapter
->stats
.prc64
+= rd32(E1000_PRC64
);
5255 adapter
->stats
.prc127
+= rd32(E1000_PRC127
);
5256 adapter
->stats
.prc255
+= rd32(E1000_PRC255
);
5257 adapter
->stats
.prc511
+= rd32(E1000_PRC511
);
5258 adapter
->stats
.prc1023
+= rd32(E1000_PRC1023
);
5259 adapter
->stats
.prc1522
+= rd32(E1000_PRC1522
);
5260 adapter
->stats
.symerrs
+= rd32(E1000_SYMERRS
);
5261 adapter
->stats
.sec
+= rd32(E1000_SEC
);
5263 mpc
= rd32(E1000_MPC
);
5264 adapter
->stats
.mpc
+= mpc
;
5265 net_stats
->rx_fifo_errors
+= mpc
;
5266 adapter
->stats
.scc
+= rd32(E1000_SCC
);
5267 adapter
->stats
.ecol
+= rd32(E1000_ECOL
);
5268 adapter
->stats
.mcc
+= rd32(E1000_MCC
);
5269 adapter
->stats
.latecol
+= rd32(E1000_LATECOL
);
5270 adapter
->stats
.dc
+= rd32(E1000_DC
);
5271 adapter
->stats
.rlec
+= rd32(E1000_RLEC
);
5272 adapter
->stats
.xonrxc
+= rd32(E1000_XONRXC
);
5273 adapter
->stats
.xontxc
+= rd32(E1000_XONTXC
);
5274 adapter
->stats
.xoffrxc
+= rd32(E1000_XOFFRXC
);
5275 adapter
->stats
.xofftxc
+= rd32(E1000_XOFFTXC
);
5276 adapter
->stats
.fcruc
+= rd32(E1000_FCRUC
);
5277 adapter
->stats
.gptc
+= rd32(E1000_GPTC
);
5278 adapter
->stats
.gotc
+= rd32(E1000_GOTCL
);
5279 rd32(E1000_GOTCH
); /* clear GOTCL */
5280 adapter
->stats
.rnbc
+= rd32(E1000_RNBC
);
5281 adapter
->stats
.ruc
+= rd32(E1000_RUC
);
5282 adapter
->stats
.rfc
+= rd32(E1000_RFC
);
5283 adapter
->stats
.rjc
+= rd32(E1000_RJC
);
5284 adapter
->stats
.tor
+= rd32(E1000_TORH
);
5285 adapter
->stats
.tot
+= rd32(E1000_TOTH
);
5286 adapter
->stats
.tpr
+= rd32(E1000_TPR
);
5288 adapter
->stats
.ptc64
+= rd32(E1000_PTC64
);
5289 adapter
->stats
.ptc127
+= rd32(E1000_PTC127
);
5290 adapter
->stats
.ptc255
+= rd32(E1000_PTC255
);
5291 adapter
->stats
.ptc511
+= rd32(E1000_PTC511
);
5292 adapter
->stats
.ptc1023
+= rd32(E1000_PTC1023
);
5293 adapter
->stats
.ptc1522
+= rd32(E1000_PTC1522
);
5295 adapter
->stats
.mptc
+= rd32(E1000_MPTC
);
5296 adapter
->stats
.bptc
+= rd32(E1000_BPTC
);
5298 adapter
->stats
.tpt
+= rd32(E1000_TPT
);
5299 adapter
->stats
.colc
+= rd32(E1000_COLC
);
5301 adapter
->stats
.algnerrc
+= rd32(E1000_ALGNERRC
);
5302 /* read internal phy specific stats */
5303 reg
= rd32(E1000_CTRL_EXT
);
5304 if (!(reg
& E1000_CTRL_EXT_LINK_MODE_MASK
)) {
5305 adapter
->stats
.rxerrc
+= rd32(E1000_RXERRC
);
5307 /* this stat has invalid values on i210/i211 */
5308 if ((hw
->mac
.type
!= e1000_i210
) &&
5309 (hw
->mac
.type
!= e1000_i211
))
5310 adapter
->stats
.tncrs
+= rd32(E1000_TNCRS
);
5313 adapter
->stats
.tsctc
+= rd32(E1000_TSCTC
);
5314 adapter
->stats
.tsctfc
+= rd32(E1000_TSCTFC
);
5316 adapter
->stats
.iac
+= rd32(E1000_IAC
);
5317 adapter
->stats
.icrxoc
+= rd32(E1000_ICRXOC
);
5318 adapter
->stats
.icrxptc
+= rd32(E1000_ICRXPTC
);
5319 adapter
->stats
.icrxatc
+= rd32(E1000_ICRXATC
);
5320 adapter
->stats
.ictxptc
+= rd32(E1000_ICTXPTC
);
5321 adapter
->stats
.ictxatc
+= rd32(E1000_ICTXATC
);
5322 adapter
->stats
.ictxqec
+= rd32(E1000_ICTXQEC
);
5323 adapter
->stats
.ictxqmtc
+= rd32(E1000_ICTXQMTC
);
5324 adapter
->stats
.icrxdmtc
+= rd32(E1000_ICRXDMTC
);
5326 /* Fill out the OS statistics structure */
5327 net_stats
->multicast
= adapter
->stats
.mprc
;
5328 net_stats
->collisions
= adapter
->stats
.colc
;
5332 /* RLEC on some newer hardware can be incorrect so build
5333 * our own version based on RUC and ROC
5335 net_stats
->rx_errors
= adapter
->stats
.rxerrc
+
5336 adapter
->stats
.crcerrs
+ adapter
->stats
.algnerrc
+
5337 adapter
->stats
.ruc
+ adapter
->stats
.roc
+
5338 adapter
->stats
.cexterr
;
5339 net_stats
->rx_length_errors
= adapter
->stats
.ruc
+
5341 net_stats
->rx_crc_errors
= adapter
->stats
.crcerrs
;
5342 net_stats
->rx_frame_errors
= adapter
->stats
.algnerrc
;
5343 net_stats
->rx_missed_errors
= adapter
->stats
.mpc
;
5346 net_stats
->tx_errors
= adapter
->stats
.ecol
+
5347 adapter
->stats
.latecol
;
5348 net_stats
->tx_aborted_errors
= adapter
->stats
.ecol
;
5349 net_stats
->tx_window_errors
= adapter
->stats
.latecol
;
5350 net_stats
->tx_carrier_errors
= adapter
->stats
.tncrs
;
5352 /* Tx Dropped needs to be maintained elsewhere */
5354 /* Management Stats */
5355 adapter
->stats
.mgptc
+= rd32(E1000_MGTPTC
);
5356 adapter
->stats
.mgprc
+= rd32(E1000_MGTPRC
);
5357 adapter
->stats
.mgpdc
+= rd32(E1000_MGTPDC
);
5360 reg
= rd32(E1000_MANC
);
5361 if (reg
& E1000_MANC_EN_BMC2OS
) {
5362 adapter
->stats
.o2bgptc
+= rd32(E1000_O2BGPTC
);
5363 adapter
->stats
.o2bspc
+= rd32(E1000_O2BSPC
);
5364 adapter
->stats
.b2ospc
+= rd32(E1000_B2OSPC
);
5365 adapter
->stats
.b2ogprc
+= rd32(E1000_B2OGPRC
);
5369 static void igb_tsync_interrupt(struct igb_adapter
*adapter
)
5371 struct e1000_hw
*hw
= &adapter
->hw
;
5372 struct ptp_clock_event event
;
5373 struct timespec64 ts
;
5374 u32 ack
= 0, tsauxc
, sec
, nsec
, tsicr
= rd32(E1000_TSICR
);
5376 if (tsicr
& TSINTR_SYS_WRAP
) {
5377 event
.type
= PTP_CLOCK_PPS
;
5378 if (adapter
->ptp_caps
.pps
)
5379 ptp_clock_event(adapter
->ptp_clock
, &event
);
5381 dev_err(&adapter
->pdev
->dev
, "unexpected SYS WRAP");
5382 ack
|= TSINTR_SYS_WRAP
;
5385 if (tsicr
& E1000_TSICR_TXTS
) {
5386 /* retrieve hardware timestamp */
5387 schedule_work(&adapter
->ptp_tx_work
);
5388 ack
|= E1000_TSICR_TXTS
;
5391 if (tsicr
& TSINTR_TT0
) {
5392 spin_lock(&adapter
->tmreg_lock
);
5393 ts
= timespec64_add(adapter
->perout
[0].start
,
5394 adapter
->perout
[0].period
);
5395 /* u32 conversion of tv_sec is safe until y2106 */
5396 wr32(E1000_TRGTTIML0
, ts
.tv_nsec
);
5397 wr32(E1000_TRGTTIMH0
, (u32
)ts
.tv_sec
);
5398 tsauxc
= rd32(E1000_TSAUXC
);
5399 tsauxc
|= TSAUXC_EN_TT0
;
5400 wr32(E1000_TSAUXC
, tsauxc
);
5401 adapter
->perout
[0].start
= ts
;
5402 spin_unlock(&adapter
->tmreg_lock
);
5406 if (tsicr
& TSINTR_TT1
) {
5407 spin_lock(&adapter
->tmreg_lock
);
5408 ts
= timespec64_add(adapter
->perout
[1].start
,
5409 adapter
->perout
[1].period
);
5410 wr32(E1000_TRGTTIML1
, ts
.tv_nsec
);
5411 wr32(E1000_TRGTTIMH1
, (u32
)ts
.tv_sec
);
5412 tsauxc
= rd32(E1000_TSAUXC
);
5413 tsauxc
|= TSAUXC_EN_TT1
;
5414 wr32(E1000_TSAUXC
, tsauxc
);
5415 adapter
->perout
[1].start
= ts
;
5416 spin_unlock(&adapter
->tmreg_lock
);
5420 if (tsicr
& TSINTR_AUTT0
) {
5421 nsec
= rd32(E1000_AUXSTMPL0
);
5422 sec
= rd32(E1000_AUXSTMPH0
);
5423 event
.type
= PTP_CLOCK_EXTTS
;
5425 event
.timestamp
= sec
* 1000000000ULL + nsec
;
5426 ptp_clock_event(adapter
->ptp_clock
, &event
);
5427 ack
|= TSINTR_AUTT0
;
5430 if (tsicr
& TSINTR_AUTT1
) {
5431 nsec
= rd32(E1000_AUXSTMPL1
);
5432 sec
= rd32(E1000_AUXSTMPH1
);
5433 event
.type
= PTP_CLOCK_EXTTS
;
5435 event
.timestamp
= sec
* 1000000000ULL + nsec
;
5436 ptp_clock_event(adapter
->ptp_clock
, &event
);
5437 ack
|= TSINTR_AUTT1
;
5440 /* acknowledge the interrupts */
5441 wr32(E1000_TSICR
, ack
);
5444 static irqreturn_t
igb_msix_other(int irq
, void *data
)
5446 struct igb_adapter
*adapter
= data
;
5447 struct e1000_hw
*hw
= &adapter
->hw
;
5448 u32 icr
= rd32(E1000_ICR
);
5449 /* reading ICR causes bit 31 of EICR to be cleared */
5451 if (icr
& E1000_ICR_DRSTA
)
5452 schedule_work(&adapter
->reset_task
);
5454 if (icr
& E1000_ICR_DOUTSYNC
) {
5455 /* HW is reporting DMA is out of sync */
5456 adapter
->stats
.doosync
++;
5457 /* The DMA Out of Sync is also indication of a spoof event
5458 * in IOV mode. Check the Wrong VM Behavior register to
5459 * see if it is really a spoof event.
5461 igb_check_wvbr(adapter
);
5464 /* Check for a mailbox event */
5465 if (icr
& E1000_ICR_VMMB
)
5466 igb_msg_task(adapter
);
5468 if (icr
& E1000_ICR_LSC
) {
5469 hw
->mac
.get_link_status
= 1;
5470 /* guard against interrupt when we're going down */
5471 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
5472 mod_timer(&adapter
->watchdog_timer
, jiffies
+ 1);
5475 if (icr
& E1000_ICR_TS
)
5476 igb_tsync_interrupt(adapter
);
5478 wr32(E1000_EIMS
, adapter
->eims_other
);
5483 static void igb_write_itr(struct igb_q_vector
*q_vector
)
5485 struct igb_adapter
*adapter
= q_vector
->adapter
;
5486 u32 itr_val
= q_vector
->itr_val
& 0x7FFC;
5488 if (!q_vector
->set_itr
)
5494 if (adapter
->hw
.mac
.type
== e1000_82575
)
5495 itr_val
|= itr_val
<< 16;
5497 itr_val
|= E1000_EITR_CNT_IGNR
;
5499 writel(itr_val
, q_vector
->itr_register
);
5500 q_vector
->set_itr
= 0;
5503 static irqreturn_t
igb_msix_ring(int irq
, void *data
)
5505 struct igb_q_vector
*q_vector
= data
;
5507 /* Write the ITR value calculated from the previous interrupt. */
5508 igb_write_itr(q_vector
);
5510 napi_schedule(&q_vector
->napi
);
5515 #ifdef CONFIG_IGB_DCA
5516 static void igb_update_tx_dca(struct igb_adapter
*adapter
,
5517 struct igb_ring
*tx_ring
,
5520 struct e1000_hw
*hw
= &adapter
->hw
;
5521 u32 txctrl
= dca3_get_tag(tx_ring
->dev
, cpu
);
5523 if (hw
->mac
.type
!= e1000_82575
)
5524 txctrl
<<= E1000_DCA_TXCTRL_CPUID_SHIFT
;
5526 /* We can enable relaxed ordering for reads, but not writes when
5527 * DCA is enabled. This is due to a known issue in some chipsets
5528 * which will cause the DCA tag to be cleared.
5530 txctrl
|= E1000_DCA_TXCTRL_DESC_RRO_EN
|
5531 E1000_DCA_TXCTRL_DATA_RRO_EN
|
5532 E1000_DCA_TXCTRL_DESC_DCA_EN
;
5534 wr32(E1000_DCA_TXCTRL(tx_ring
->reg_idx
), txctrl
);
5537 static void igb_update_rx_dca(struct igb_adapter
*adapter
,
5538 struct igb_ring
*rx_ring
,
5541 struct e1000_hw
*hw
= &adapter
->hw
;
5542 u32 rxctrl
= dca3_get_tag(&adapter
->pdev
->dev
, cpu
);
5544 if (hw
->mac
.type
!= e1000_82575
)
5545 rxctrl
<<= E1000_DCA_RXCTRL_CPUID_SHIFT
;
5547 /* We can enable relaxed ordering for reads, but not writes when
5548 * DCA is enabled. This is due to a known issue in some chipsets
5549 * which will cause the DCA tag to be cleared.
5551 rxctrl
|= E1000_DCA_RXCTRL_DESC_RRO_EN
|
5552 E1000_DCA_RXCTRL_DESC_DCA_EN
;
5554 wr32(E1000_DCA_RXCTRL(rx_ring
->reg_idx
), rxctrl
);
5557 static void igb_update_dca(struct igb_q_vector
*q_vector
)
5559 struct igb_adapter
*adapter
= q_vector
->adapter
;
5560 int cpu
= get_cpu();
5562 if (q_vector
->cpu
== cpu
)
5565 if (q_vector
->tx
.ring
)
5566 igb_update_tx_dca(adapter
, q_vector
->tx
.ring
, cpu
);
5568 if (q_vector
->rx
.ring
)
5569 igb_update_rx_dca(adapter
, q_vector
->rx
.ring
, cpu
);
5571 q_vector
->cpu
= cpu
;
5576 static void igb_setup_dca(struct igb_adapter
*adapter
)
5578 struct e1000_hw
*hw
= &adapter
->hw
;
5581 if (!(adapter
->flags
& IGB_FLAG_DCA_ENABLED
))
5584 /* Always use CB2 mode, difference is masked in the CB driver. */
5585 wr32(E1000_DCA_CTRL
, E1000_DCA_CTRL_DCA_MODE_CB2
);
5587 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
5588 adapter
->q_vector
[i
]->cpu
= -1;
5589 igb_update_dca(adapter
->q_vector
[i
]);
5593 static int __igb_notify_dca(struct device
*dev
, void *data
)
5595 struct net_device
*netdev
= dev_get_drvdata(dev
);
5596 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5597 struct pci_dev
*pdev
= adapter
->pdev
;
5598 struct e1000_hw
*hw
= &adapter
->hw
;
5599 unsigned long event
= *(unsigned long *)data
;
5602 case DCA_PROVIDER_ADD
:
5603 /* if already enabled, don't do it again */
5604 if (adapter
->flags
& IGB_FLAG_DCA_ENABLED
)
5606 if (dca_add_requester(dev
) == 0) {
5607 adapter
->flags
|= IGB_FLAG_DCA_ENABLED
;
5608 dev_info(&pdev
->dev
, "DCA enabled\n");
5609 igb_setup_dca(adapter
);
5612 /* Fall Through since DCA is disabled. */
5613 case DCA_PROVIDER_REMOVE
:
5614 if (adapter
->flags
& IGB_FLAG_DCA_ENABLED
) {
5615 /* without this a class_device is left
5616 * hanging around in the sysfs model
5618 dca_remove_requester(dev
);
5619 dev_info(&pdev
->dev
, "DCA disabled\n");
5620 adapter
->flags
&= ~IGB_FLAG_DCA_ENABLED
;
5621 wr32(E1000_DCA_CTRL
, E1000_DCA_CTRL_DCA_MODE_DISABLE
);
5629 static int igb_notify_dca(struct notifier_block
*nb
, unsigned long event
,
5634 ret_val
= driver_for_each_device(&igb_driver
.driver
, NULL
, &event
,
5637 return ret_val
? NOTIFY_BAD
: NOTIFY_DONE
;
5639 #endif /* CONFIG_IGB_DCA */
5641 #ifdef CONFIG_PCI_IOV
5642 static int igb_vf_configure(struct igb_adapter
*adapter
, int vf
)
5644 unsigned char mac_addr
[ETH_ALEN
];
5646 eth_zero_addr(mac_addr
);
5647 igb_set_vf_mac(adapter
, vf
, mac_addr
);
5649 /* By default spoof check is enabled for all VFs */
5650 adapter
->vf_data
[vf
].spoofchk_enabled
= true;
5656 static void igb_ping_all_vfs(struct igb_adapter
*adapter
)
5658 struct e1000_hw
*hw
= &adapter
->hw
;
5662 for (i
= 0 ; i
< adapter
->vfs_allocated_count
; i
++) {
5663 ping
= E1000_PF_CONTROL_MSG
;
5664 if (adapter
->vf_data
[i
].flags
& IGB_VF_FLAG_CTS
)
5665 ping
|= E1000_VT_MSGTYPE_CTS
;
5666 igb_write_mbx(hw
, &ping
, 1, i
);
5670 static int igb_set_vf_promisc(struct igb_adapter
*adapter
, u32
*msgbuf
, u32 vf
)
5672 struct e1000_hw
*hw
= &adapter
->hw
;
5673 u32 vmolr
= rd32(E1000_VMOLR(vf
));
5674 struct vf_data_storage
*vf_data
= &adapter
->vf_data
[vf
];
5676 vf_data
->flags
&= ~(IGB_VF_FLAG_UNI_PROMISC
|
5677 IGB_VF_FLAG_MULTI_PROMISC
);
5678 vmolr
&= ~(E1000_VMOLR_ROPE
| E1000_VMOLR_ROMPE
| E1000_VMOLR_MPME
);
5680 if (*msgbuf
& E1000_VF_SET_PROMISC_MULTICAST
) {
5681 vmolr
|= E1000_VMOLR_MPME
;
5682 vf_data
->flags
|= IGB_VF_FLAG_MULTI_PROMISC
;
5683 *msgbuf
&= ~E1000_VF_SET_PROMISC_MULTICAST
;
5685 /* if we have hashes and we are clearing a multicast promisc
5686 * flag we need to write the hashes to the MTA as this step
5687 * was previously skipped
5689 if (vf_data
->num_vf_mc_hashes
> 30) {
5690 vmolr
|= E1000_VMOLR_MPME
;
5691 } else if (vf_data
->num_vf_mc_hashes
) {
5694 vmolr
|= E1000_VMOLR_ROMPE
;
5695 for (j
= 0; j
< vf_data
->num_vf_mc_hashes
; j
++)
5696 igb_mta_set(hw
, vf_data
->vf_mc_hashes
[j
]);
5700 wr32(E1000_VMOLR(vf
), vmolr
);
5702 /* there are flags left unprocessed, likely not supported */
5703 if (*msgbuf
& E1000_VT_MSGINFO_MASK
)
5709 static int igb_set_vf_multicasts(struct igb_adapter
*adapter
,
5710 u32
*msgbuf
, u32 vf
)
5712 int n
= (msgbuf
[0] & E1000_VT_MSGINFO_MASK
) >> E1000_VT_MSGINFO_SHIFT
;
5713 u16
*hash_list
= (u16
*)&msgbuf
[1];
5714 struct vf_data_storage
*vf_data
= &adapter
->vf_data
[vf
];
5717 /* salt away the number of multicast addresses assigned
5718 * to this VF for later use to restore when the PF multi cast
5721 vf_data
->num_vf_mc_hashes
= n
;
5723 /* only up to 30 hash values supported */
5727 /* store the hashes for later use */
5728 for (i
= 0; i
< n
; i
++)
5729 vf_data
->vf_mc_hashes
[i
] = hash_list
[i
];
5731 /* Flush and reset the mta with the new values */
5732 igb_set_rx_mode(adapter
->netdev
);
5737 static void igb_restore_vf_multicasts(struct igb_adapter
*adapter
)
5739 struct e1000_hw
*hw
= &adapter
->hw
;
5740 struct vf_data_storage
*vf_data
;
5743 for (i
= 0; i
< adapter
->vfs_allocated_count
; i
++) {
5744 u32 vmolr
= rd32(E1000_VMOLR(i
));
5746 vmolr
&= ~(E1000_VMOLR_ROMPE
| E1000_VMOLR_MPME
);
5748 vf_data
= &adapter
->vf_data
[i
];
5750 if ((vf_data
->num_vf_mc_hashes
> 30) ||
5751 (vf_data
->flags
& IGB_VF_FLAG_MULTI_PROMISC
)) {
5752 vmolr
|= E1000_VMOLR_MPME
;
5753 } else if (vf_data
->num_vf_mc_hashes
) {
5754 vmolr
|= E1000_VMOLR_ROMPE
;
5755 for (j
= 0; j
< vf_data
->num_vf_mc_hashes
; j
++)
5756 igb_mta_set(hw
, vf_data
->vf_mc_hashes
[j
]);
5758 wr32(E1000_VMOLR(i
), vmolr
);
5762 static void igb_clear_vf_vfta(struct igb_adapter
*adapter
, u32 vf
)
5764 struct e1000_hw
*hw
= &adapter
->hw
;
5765 u32 pool_mask
, reg
, vid
;
5768 pool_mask
= 1 << (E1000_VLVF_POOLSEL_SHIFT
+ vf
);
5770 /* Find the vlan filter for this id */
5771 for (i
= 0; i
< E1000_VLVF_ARRAY_SIZE
; i
++) {
5772 reg
= rd32(E1000_VLVF(i
));
5774 /* remove the vf from the pool */
5777 /* if pool is empty then remove entry from vfta */
5778 if (!(reg
& E1000_VLVF_POOLSEL_MASK
) &&
5779 (reg
& E1000_VLVF_VLANID_ENABLE
)) {
5781 vid
= reg
& E1000_VLVF_VLANID_MASK
;
5782 igb_vfta_set(hw
, vid
, vf
, false, true);
5785 wr32(E1000_VLVF(i
), reg
);
5789 static int igb_find_vlvf_entry(struct igb_adapter
*adapter
, int vid
)
5791 struct e1000_hw
*hw
= &adapter
->hw
;
5795 /* Find the vlan filter for this id */
5796 for (i
= 0; i
< E1000_VLVF_ARRAY_SIZE
; i
++) {
5797 reg
= rd32(E1000_VLVF(i
));
5798 if ((reg
& E1000_VLVF_VLANID_ENABLE
) &&
5799 vid
== (reg
& E1000_VLVF_VLANID_MASK
))
5803 if (i
>= E1000_VLVF_ARRAY_SIZE
)
5809 static s32
igb_set_vf_vlan(struct igb_adapter
*adapter
, u32 vid
,
5812 int pf_id
= adapter
->vfs_allocated_count
;
5813 struct e1000_hw
*hw
= &adapter
->hw
;
5816 /* If VLAN overlaps with one the PF is currently monitoring make
5817 * sure that we are able to allocate a VLVF entry. This may be
5818 * redundant but it guarantees PF will maintain visibility to
5821 if (add
&& (adapter
->netdev
->flags
& IFF_PROMISC
)) {
5822 err
= igb_vfta_set(hw
, vid
, pf_id
, true, false);
5827 err
= igb_vfta_set(hw
, vid
, vf
, add
, false);
5832 /* Go through all the checks to see if the VLAN filter should
5833 * be wiped completely.
5835 if (!add
&& (adapter
->netdev
->flags
& IFF_PROMISC
)) {
5837 int regndx
= igb_find_vlvf_entry(adapter
, vid
);
5841 /* See if any other pools are set for this VLAN filter
5842 * entry other than the PF.
5844 vlvf
= bits
= rd32(E1000_VLVF(regndx
));
5845 bits
&= 1 << (E1000_VLVF_POOLSEL_SHIFT
+
5846 adapter
->vfs_allocated_count
);
5847 /* If the filter was removed then ensure PF pool bit
5848 * is cleared if the PF only added itself to the pool
5849 * because the PF is in promiscuous mode.
5851 if ((vlvf
& VLAN_VID_MASK
) == vid
&&
5852 !test_bit(vid
, adapter
->active_vlans
) &&
5854 igb_vfta_set(hw
, vid
, adapter
->vfs_allocated_count
,
5862 static void igb_set_vmvir(struct igb_adapter
*adapter
, u32 vid
, u32 vf
)
5864 struct e1000_hw
*hw
= &adapter
->hw
;
5867 wr32(E1000_VMVIR(vf
), (vid
| E1000_VMVIR_VLANA_DEFAULT
));
5869 wr32(E1000_VMVIR(vf
), 0);
5872 static int igb_enable_port_vlan(struct igb_adapter
*adapter
, int vf
,
5877 err
= igb_set_vf_vlan(adapter
, vlan
, true, vf
);
5881 igb_set_vmvir(adapter
, vlan
| (qos
<< VLAN_PRIO_SHIFT
), vf
);
5882 igb_set_vmolr(adapter
, vf
, !vlan
);
5884 /* revoke access to previous VLAN */
5885 if (vlan
!= adapter
->vf_data
[vf
].pf_vlan
)
5886 igb_set_vf_vlan(adapter
, adapter
->vf_data
[vf
].pf_vlan
,
5889 adapter
->vf_data
[vf
].pf_vlan
= vlan
;
5890 adapter
->vf_data
[vf
].pf_qos
= qos
;
5891 dev_info(&adapter
->pdev
->dev
,
5892 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan
, qos
, vf
);
5893 if (test_bit(__IGB_DOWN
, &adapter
->state
)) {
5894 dev_warn(&adapter
->pdev
->dev
,
5895 "The VF VLAN has been set, but the PF device is not up.\n");
5896 dev_warn(&adapter
->pdev
->dev
,
5897 "Bring the PF device up before attempting to use the VF device.\n");
5903 static int igb_disable_port_vlan(struct igb_adapter
*adapter
, int vf
)
5905 /* Restore tagless access via VLAN 0 */
5906 igb_set_vf_vlan(adapter
, 0, true, vf
);
5908 igb_set_vmvir(adapter
, 0, vf
);
5909 igb_set_vmolr(adapter
, vf
, true);
5911 /* Remove any PF assigned VLAN */
5912 if (adapter
->vf_data
[vf
].pf_vlan
)
5913 igb_set_vf_vlan(adapter
, adapter
->vf_data
[vf
].pf_vlan
,
5916 adapter
->vf_data
[vf
].pf_vlan
= 0;
5917 adapter
->vf_data
[vf
].pf_qos
= 0;
5922 static int igb_ndo_set_vf_vlan(struct net_device
*netdev
,
5923 int vf
, u16 vlan
, u8 qos
)
5925 struct igb_adapter
*adapter
= netdev_priv(netdev
);
5927 if ((vf
>= adapter
->vfs_allocated_count
) || (vlan
> 4095) || (qos
> 7))
5930 return (vlan
|| qos
) ? igb_enable_port_vlan(adapter
, vf
, vlan
, qos
) :
5931 igb_disable_port_vlan(adapter
, vf
);
5934 static int igb_set_vf_vlan_msg(struct igb_adapter
*adapter
, u32
*msgbuf
, u32 vf
)
5936 int add
= (msgbuf
[0] & E1000_VT_MSGINFO_MASK
) >> E1000_VT_MSGINFO_SHIFT
;
5937 int vid
= (msgbuf
[1] & E1000_VLVF_VLANID_MASK
);
5939 if (adapter
->vf_data
[vf
].pf_vlan
)
5942 /* VLAN 0 is a special case, don't allow it to be removed */
5946 return igb_set_vf_vlan(adapter
, vid
, !!add
, vf
);
5949 static inline void igb_vf_reset(struct igb_adapter
*adapter
, u32 vf
)
5951 struct vf_data_storage
*vf_data
= &adapter
->vf_data
[vf
];
5953 /* clear flags - except flag that indicates PF has set the MAC */
5954 vf_data
->flags
&= IGB_VF_FLAG_PF_SET_MAC
;
5955 vf_data
->last_nack
= jiffies
;
5957 /* reset vlans for device */
5958 igb_clear_vf_vfta(adapter
, vf
);
5959 igb_set_vf_vlan(adapter
, vf_data
->pf_vlan
, true, vf
);
5960 igb_set_vmvir(adapter
, vf_data
->pf_vlan
|
5961 (vf_data
->pf_qos
<< VLAN_PRIO_SHIFT
), vf
);
5962 igb_set_vmolr(adapter
, vf
, !vf_data
->pf_vlan
);
5964 /* reset multicast table array for vf */
5965 adapter
->vf_data
[vf
].num_vf_mc_hashes
= 0;
5967 /* Flush and reset the mta with the new values */
5968 igb_set_rx_mode(adapter
->netdev
);
5971 static void igb_vf_reset_event(struct igb_adapter
*adapter
, u32 vf
)
5973 unsigned char *vf_mac
= adapter
->vf_data
[vf
].vf_mac_addresses
;
5975 /* clear mac address as we were hotplug removed/added */
5976 if (!(adapter
->vf_data
[vf
].flags
& IGB_VF_FLAG_PF_SET_MAC
))
5977 eth_zero_addr(vf_mac
);
5979 /* process remaining reset events */
5980 igb_vf_reset(adapter
, vf
);
5983 static void igb_vf_reset_msg(struct igb_adapter
*adapter
, u32 vf
)
5985 struct e1000_hw
*hw
= &adapter
->hw
;
5986 unsigned char *vf_mac
= adapter
->vf_data
[vf
].vf_mac_addresses
;
5987 int rar_entry
= hw
->mac
.rar_entry_count
- (vf
+ 1);
5989 u8
*addr
= (u8
*)(&msgbuf
[1]);
5991 /* process all the same items cleared in a function level reset */
5992 igb_vf_reset(adapter
, vf
);
5994 /* set vf mac address */
5995 igb_rar_set_qsel(adapter
, vf_mac
, rar_entry
, vf
);
5997 /* enable transmit and receive for vf */
5998 reg
= rd32(E1000_VFTE
);
5999 wr32(E1000_VFTE
, reg
| (1 << vf
));
6000 reg
= rd32(E1000_VFRE
);
6001 wr32(E1000_VFRE
, reg
| (1 << vf
));
6003 adapter
->vf_data
[vf
].flags
|= IGB_VF_FLAG_CTS
;
6005 /* reply to reset with ack and vf mac address */
6006 if (!is_zero_ether_addr(vf_mac
)) {
6007 msgbuf
[0] = E1000_VF_RESET
| E1000_VT_MSGTYPE_ACK
;
6008 memcpy(addr
, vf_mac
, ETH_ALEN
);
6010 msgbuf
[0] = E1000_VF_RESET
| E1000_VT_MSGTYPE_NACK
;
6012 igb_write_mbx(hw
, msgbuf
, 3, vf
);
6015 static int igb_set_vf_mac_addr(struct igb_adapter
*adapter
, u32
*msg
, int vf
)
6017 /* The VF MAC Address is stored in a packed array of bytes
6018 * starting at the second 32 bit word of the msg array
6020 unsigned char *addr
= (char *)&msg
[1];
6023 if (is_valid_ether_addr(addr
))
6024 err
= igb_set_vf_mac(adapter
, vf
, addr
);
6029 static void igb_rcv_ack_from_vf(struct igb_adapter
*adapter
, u32 vf
)
6031 struct e1000_hw
*hw
= &adapter
->hw
;
6032 struct vf_data_storage
*vf_data
= &adapter
->vf_data
[vf
];
6033 u32 msg
= E1000_VT_MSGTYPE_NACK
;
6035 /* if device isn't clear to send it shouldn't be reading either */
6036 if (!(vf_data
->flags
& IGB_VF_FLAG_CTS
) &&
6037 time_after(jiffies
, vf_data
->last_nack
+ (2 * HZ
))) {
6038 igb_write_mbx(hw
, &msg
, 1, vf
);
6039 vf_data
->last_nack
= jiffies
;
6043 static void igb_rcv_msg_from_vf(struct igb_adapter
*adapter
, u32 vf
)
6045 struct pci_dev
*pdev
= adapter
->pdev
;
6046 u32 msgbuf
[E1000_VFMAILBOX_SIZE
];
6047 struct e1000_hw
*hw
= &adapter
->hw
;
6048 struct vf_data_storage
*vf_data
= &adapter
->vf_data
[vf
];
6051 retval
= igb_read_mbx(hw
, msgbuf
, E1000_VFMAILBOX_SIZE
, vf
);
6054 /* if receive failed revoke VF CTS stats and restart init */
6055 dev_err(&pdev
->dev
, "Error receiving message from VF\n");
6056 vf_data
->flags
&= ~IGB_VF_FLAG_CTS
;
6057 if (!time_after(jiffies
, vf_data
->last_nack
+ (2 * HZ
)))
6062 /* this is a message we already processed, do nothing */
6063 if (msgbuf
[0] & (E1000_VT_MSGTYPE_ACK
| E1000_VT_MSGTYPE_NACK
))
6066 /* until the vf completes a reset it should not be
6067 * allowed to start any configuration.
6069 if (msgbuf
[0] == E1000_VF_RESET
) {
6070 igb_vf_reset_msg(adapter
, vf
);
6074 if (!(vf_data
->flags
& IGB_VF_FLAG_CTS
)) {
6075 if (!time_after(jiffies
, vf_data
->last_nack
+ (2 * HZ
)))
6081 switch ((msgbuf
[0] & 0xFFFF)) {
6082 case E1000_VF_SET_MAC_ADDR
:
6084 if (!(vf_data
->flags
& IGB_VF_FLAG_PF_SET_MAC
))
6085 retval
= igb_set_vf_mac_addr(adapter
, msgbuf
, vf
);
6087 dev_warn(&pdev
->dev
,
6088 "VF %d attempted to override administratively set MAC address\nReload the VF driver to resume operations\n",
6091 case E1000_VF_SET_PROMISC
:
6092 retval
= igb_set_vf_promisc(adapter
, msgbuf
, vf
);
6094 case E1000_VF_SET_MULTICAST
:
6095 retval
= igb_set_vf_multicasts(adapter
, msgbuf
, vf
);
6097 case E1000_VF_SET_LPE
:
6098 retval
= igb_set_vf_rlpml(adapter
, msgbuf
[1], vf
);
6100 case E1000_VF_SET_VLAN
:
6102 if (vf_data
->pf_vlan
)
6103 dev_warn(&pdev
->dev
,
6104 "VF %d attempted to override administratively set VLAN tag\nReload the VF driver to resume operations\n",
6107 retval
= igb_set_vf_vlan_msg(adapter
, msgbuf
, vf
);
6110 dev_err(&pdev
->dev
, "Unhandled Msg %08x\n", msgbuf
[0]);
6115 msgbuf
[0] |= E1000_VT_MSGTYPE_CTS
;
6117 /* notify the VF of the results of what it sent us */
6119 msgbuf
[0] |= E1000_VT_MSGTYPE_NACK
;
6121 msgbuf
[0] |= E1000_VT_MSGTYPE_ACK
;
6123 igb_write_mbx(hw
, msgbuf
, 1, vf
);
6126 static void igb_msg_task(struct igb_adapter
*adapter
)
6128 struct e1000_hw
*hw
= &adapter
->hw
;
6131 for (vf
= 0; vf
< adapter
->vfs_allocated_count
; vf
++) {
6132 /* process any reset requests */
6133 if (!igb_check_for_rst(hw
, vf
))
6134 igb_vf_reset_event(adapter
, vf
);
6136 /* process any messages pending */
6137 if (!igb_check_for_msg(hw
, vf
))
6138 igb_rcv_msg_from_vf(adapter
, vf
);
6140 /* process any acks */
6141 if (!igb_check_for_ack(hw
, vf
))
6142 igb_rcv_ack_from_vf(adapter
, vf
);
6147 * igb_set_uta - Set unicast filter table address
6148 * @adapter: board private structure
6150 * The unicast table address is a register array of 32-bit registers.
6151 * The table is meant to be used in a way similar to how the MTA is used
6152 * however due to certain limitations in the hardware it is necessary to
6153 * set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
6154 * enable bit to allow vlan tag stripping when promiscuous mode is enabled
6156 static void igb_set_uta(struct igb_adapter
*adapter
)
6158 struct e1000_hw
*hw
= &adapter
->hw
;
6161 /* The UTA table only exists on 82576 hardware and newer */
6162 if (hw
->mac
.type
< e1000_82576
)
6165 /* we only need to do this if VMDq is enabled */
6166 if (!adapter
->vfs_allocated_count
)
6169 for (i
= 0; i
< hw
->mac
.uta_reg_count
; i
++)
6170 array_wr32(E1000_UTA
, i
, ~0);
6174 * igb_intr_msi - Interrupt Handler
6175 * @irq: interrupt number
6176 * @data: pointer to a network interface device structure
6178 static irqreturn_t
igb_intr_msi(int irq
, void *data
)
6180 struct igb_adapter
*adapter
= data
;
6181 struct igb_q_vector
*q_vector
= adapter
->q_vector
[0];
6182 struct e1000_hw
*hw
= &adapter
->hw
;
6183 /* read ICR disables interrupts using IAM */
6184 u32 icr
= rd32(E1000_ICR
);
6186 igb_write_itr(q_vector
);
6188 if (icr
& E1000_ICR_DRSTA
)
6189 schedule_work(&adapter
->reset_task
);
6191 if (icr
& E1000_ICR_DOUTSYNC
) {
6192 /* HW is reporting DMA is out of sync */
6193 adapter
->stats
.doosync
++;
6196 if (icr
& (E1000_ICR_RXSEQ
| E1000_ICR_LSC
)) {
6197 hw
->mac
.get_link_status
= 1;
6198 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
6199 mod_timer(&adapter
->watchdog_timer
, jiffies
+ 1);
6202 if (icr
& E1000_ICR_TS
)
6203 igb_tsync_interrupt(adapter
);
6205 napi_schedule(&q_vector
->napi
);
6211 * igb_intr - Legacy Interrupt Handler
6212 * @irq: interrupt number
6213 * @data: pointer to a network interface device structure
6215 static irqreturn_t
igb_intr(int irq
, void *data
)
6217 struct igb_adapter
*adapter
= data
;
6218 struct igb_q_vector
*q_vector
= adapter
->q_vector
[0];
6219 struct e1000_hw
*hw
= &adapter
->hw
;
6220 /* Interrupt Auto-Mask...upon reading ICR, interrupts are masked. No
6221 * need for the IMC write
6223 u32 icr
= rd32(E1000_ICR
);
6225 /* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
6226 * not set, then the adapter didn't send an interrupt
6228 if (!(icr
& E1000_ICR_INT_ASSERTED
))
6231 igb_write_itr(q_vector
);
6233 if (icr
& E1000_ICR_DRSTA
)
6234 schedule_work(&adapter
->reset_task
);
6236 if (icr
& E1000_ICR_DOUTSYNC
) {
6237 /* HW is reporting DMA is out of sync */
6238 adapter
->stats
.doosync
++;
6241 if (icr
& (E1000_ICR_RXSEQ
| E1000_ICR_LSC
)) {
6242 hw
->mac
.get_link_status
= 1;
6243 /* guard against interrupt when we're going down */
6244 if (!test_bit(__IGB_DOWN
, &adapter
->state
))
6245 mod_timer(&adapter
->watchdog_timer
, jiffies
+ 1);
6248 if (icr
& E1000_ICR_TS
)
6249 igb_tsync_interrupt(adapter
);
6251 napi_schedule(&q_vector
->napi
);
6256 static void igb_ring_irq_enable(struct igb_q_vector
*q_vector
)
6258 struct igb_adapter
*adapter
= q_vector
->adapter
;
6259 struct e1000_hw
*hw
= &adapter
->hw
;
6261 if ((q_vector
->rx
.ring
&& (adapter
->rx_itr_setting
& 3)) ||
6262 (!q_vector
->rx
.ring
&& (adapter
->tx_itr_setting
& 3))) {
6263 if ((adapter
->num_q_vectors
== 1) && !adapter
->vf_data
)
6264 igb_set_itr(q_vector
);
6266 igb_update_ring_itr(q_vector
);
6269 if (!test_bit(__IGB_DOWN
, &adapter
->state
)) {
6270 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
)
6271 wr32(E1000_EIMS
, q_vector
->eims_value
);
6273 igb_irq_enable(adapter
);
6278 * igb_poll - NAPI Rx polling callback
6279 * @napi: napi polling structure
6280 * @budget: count of how many packets we should handle
6282 static int igb_poll(struct napi_struct
*napi
, int budget
)
6284 struct igb_q_vector
*q_vector
= container_of(napi
,
6285 struct igb_q_vector
,
6287 bool clean_complete
= true;
6290 #ifdef CONFIG_IGB_DCA
6291 if (q_vector
->adapter
->flags
& IGB_FLAG_DCA_ENABLED
)
6292 igb_update_dca(q_vector
);
6294 if (q_vector
->tx
.ring
)
6295 clean_complete
= igb_clean_tx_irq(q_vector
);
6297 if (q_vector
->rx
.ring
) {
6298 int cleaned
= igb_clean_rx_irq(q_vector
, budget
);
6300 work_done
+= cleaned
;
6301 clean_complete
&= (cleaned
< budget
);
6304 /* If all work not completed, return budget and keep polling */
6305 if (!clean_complete
)
6308 /* If not enough Rx work done, exit the polling mode */
6309 napi_complete_done(napi
, work_done
);
6310 igb_ring_irq_enable(q_vector
);
6316 * igb_clean_tx_irq - Reclaim resources after transmit completes
6317 * @q_vector: pointer to q_vector containing needed info
6319 * returns true if ring is completely cleaned
6321 static bool igb_clean_tx_irq(struct igb_q_vector
*q_vector
)
6323 struct igb_adapter
*adapter
= q_vector
->adapter
;
6324 struct igb_ring
*tx_ring
= q_vector
->tx
.ring
;
6325 struct igb_tx_buffer
*tx_buffer
;
6326 union e1000_adv_tx_desc
*tx_desc
;
6327 unsigned int total_bytes
= 0, total_packets
= 0;
6328 unsigned int budget
= q_vector
->tx
.work_limit
;
6329 unsigned int i
= tx_ring
->next_to_clean
;
6331 if (test_bit(__IGB_DOWN
, &adapter
->state
))
6334 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
6335 tx_desc
= IGB_TX_DESC(tx_ring
, i
);
6336 i
-= tx_ring
->count
;
6339 union e1000_adv_tx_desc
*eop_desc
= tx_buffer
->next_to_watch
;
6341 /* if next_to_watch is not set then there is no work pending */
6345 /* prevent any other reads prior to eop_desc */
6346 read_barrier_depends();
6348 /* if DD is not set pending work has not been completed */
6349 if (!(eop_desc
->wb
.status
& cpu_to_le32(E1000_TXD_STAT_DD
)))
6352 /* clear next_to_watch to prevent false hangs */
6353 tx_buffer
->next_to_watch
= NULL
;
6355 /* update the statistics for this packet */
6356 total_bytes
+= tx_buffer
->bytecount
;
6357 total_packets
+= tx_buffer
->gso_segs
;
6360 dev_consume_skb_any(tx_buffer
->skb
);
6362 /* unmap skb header data */
6363 dma_unmap_single(tx_ring
->dev
,
6364 dma_unmap_addr(tx_buffer
, dma
),
6365 dma_unmap_len(tx_buffer
, len
),
6368 /* clear tx_buffer data */
6369 tx_buffer
->skb
= NULL
;
6370 dma_unmap_len_set(tx_buffer
, len
, 0);
6372 /* clear last DMA location and unmap remaining buffers */
6373 while (tx_desc
!= eop_desc
) {
6378 i
-= tx_ring
->count
;
6379 tx_buffer
= tx_ring
->tx_buffer_info
;
6380 tx_desc
= IGB_TX_DESC(tx_ring
, 0);
6383 /* unmap any remaining paged data */
6384 if (dma_unmap_len(tx_buffer
, len
)) {
6385 dma_unmap_page(tx_ring
->dev
,
6386 dma_unmap_addr(tx_buffer
, dma
),
6387 dma_unmap_len(tx_buffer
, len
),
6389 dma_unmap_len_set(tx_buffer
, len
, 0);
6393 /* move us one more past the eop_desc for start of next pkt */
6398 i
-= tx_ring
->count
;
6399 tx_buffer
= tx_ring
->tx_buffer_info
;
6400 tx_desc
= IGB_TX_DESC(tx_ring
, 0);
6403 /* issue prefetch for next Tx descriptor */
6406 /* update budget accounting */
6408 } while (likely(budget
));
6410 netdev_tx_completed_queue(txring_txq(tx_ring
),
6411 total_packets
, total_bytes
);
6412 i
+= tx_ring
->count
;
6413 tx_ring
->next_to_clean
= i
;
6414 u64_stats_update_begin(&tx_ring
->tx_syncp
);
6415 tx_ring
->tx_stats
.bytes
+= total_bytes
;
6416 tx_ring
->tx_stats
.packets
+= total_packets
;
6417 u64_stats_update_end(&tx_ring
->tx_syncp
);
6418 q_vector
->tx
.total_bytes
+= total_bytes
;
6419 q_vector
->tx
.total_packets
+= total_packets
;
6421 if (test_bit(IGB_RING_FLAG_TX_DETECT_HANG
, &tx_ring
->flags
)) {
6422 struct e1000_hw
*hw
= &adapter
->hw
;
6424 /* Detect a transmit hang in hardware, this serializes the
6425 * check with the clearing of time_stamp and movement of i
6427 clear_bit(IGB_RING_FLAG_TX_DETECT_HANG
, &tx_ring
->flags
);
6428 if (tx_buffer
->next_to_watch
&&
6429 time_after(jiffies
, tx_buffer
->time_stamp
+
6430 (adapter
->tx_timeout_factor
* HZ
)) &&
6431 !(rd32(E1000_STATUS
) & E1000_STATUS_TXOFF
)) {
6433 /* detected Tx unit hang */
6434 dev_err(tx_ring
->dev
,
6435 "Detected Tx Unit Hang\n"
6439 " next_to_use <%x>\n"
6440 " next_to_clean <%x>\n"
6441 "buffer_info[next_to_clean]\n"
6442 " time_stamp <%lx>\n"
6443 " next_to_watch <%p>\n"
6445 " desc.status <%x>\n",
6446 tx_ring
->queue_index
,
6447 rd32(E1000_TDH(tx_ring
->reg_idx
)),
6448 readl(tx_ring
->tail
),
6449 tx_ring
->next_to_use
,
6450 tx_ring
->next_to_clean
,
6451 tx_buffer
->time_stamp
,
6452 tx_buffer
->next_to_watch
,
6454 tx_buffer
->next_to_watch
->wb
.status
);
6455 netif_stop_subqueue(tx_ring
->netdev
,
6456 tx_ring
->queue_index
);
6458 /* we are about to reset, no point in enabling stuff */
6463 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
6464 if (unlikely(total_packets
&&
6465 netif_carrier_ok(tx_ring
->netdev
) &&
6466 igb_desc_unused(tx_ring
) >= TX_WAKE_THRESHOLD
)) {
6467 /* Make sure that anybody stopping the queue after this
6468 * sees the new next_to_clean.
6471 if (__netif_subqueue_stopped(tx_ring
->netdev
,
6472 tx_ring
->queue_index
) &&
6473 !(test_bit(__IGB_DOWN
, &adapter
->state
))) {
6474 netif_wake_subqueue(tx_ring
->netdev
,
6475 tx_ring
->queue_index
);
6477 u64_stats_update_begin(&tx_ring
->tx_syncp
);
6478 tx_ring
->tx_stats
.restart_queue
++;
6479 u64_stats_update_end(&tx_ring
->tx_syncp
);
6487 * igb_reuse_rx_page - page flip buffer and store it back on the ring
6488 * @rx_ring: rx descriptor ring to store buffers on
6489 * @old_buff: donor buffer to have page reused
6491 * Synchronizes page for reuse by the adapter
6493 static void igb_reuse_rx_page(struct igb_ring
*rx_ring
,
6494 struct igb_rx_buffer
*old_buff
)
6496 struct igb_rx_buffer
*new_buff
;
6497 u16 nta
= rx_ring
->next_to_alloc
;
6499 new_buff
= &rx_ring
->rx_buffer_info
[nta
];
6501 /* update, and store next to alloc */
6503 rx_ring
->next_to_alloc
= (nta
< rx_ring
->count
) ? nta
: 0;
6505 /* transfer page from old buffer to new buffer */
6506 *new_buff
= *old_buff
;
6508 /* sync the buffer for use by the device */
6509 dma_sync_single_range_for_device(rx_ring
->dev
, old_buff
->dma
,
6510 old_buff
->page_offset
,
6515 static inline bool igb_page_is_reserved(struct page
*page
)
6517 return (page_to_nid(page
) != numa_mem_id()) || page_is_pfmemalloc(page
);
6520 static bool igb_can_reuse_rx_page(struct igb_rx_buffer
*rx_buffer
,
6522 unsigned int truesize
)
6524 /* avoid re-using remote pages */
6525 if (unlikely(igb_page_is_reserved(page
)))
6528 #if (PAGE_SIZE < 8192)
6529 /* if we are only owner of page we can reuse it */
6530 if (unlikely(page_count(page
) != 1))
6533 /* flip page offset to other buffer */
6534 rx_buffer
->page_offset
^= IGB_RX_BUFSZ
;
6536 /* move offset up to the next cache line */
6537 rx_buffer
->page_offset
+= truesize
;
6539 if (rx_buffer
->page_offset
> (PAGE_SIZE
- IGB_RX_BUFSZ
))
6543 /* Even if we own the page, we are not allowed to use atomic_set()
6544 * This would break get_page_unless_zero() users.
6546 atomic_inc(&page
->_count
);
6552 * igb_add_rx_frag - Add contents of Rx buffer to sk_buff
6553 * @rx_ring: rx descriptor ring to transact packets on
6554 * @rx_buffer: buffer containing page to add
6555 * @rx_desc: descriptor containing length of buffer written by hardware
6556 * @skb: sk_buff to place the data into
6558 * This function will add the data contained in rx_buffer->page to the skb.
6559 * This is done either through a direct copy if the data in the buffer is
6560 * less than the skb header size, otherwise it will just attach the page as
6561 * a frag to the skb.
6563 * The function will then update the page offset if necessary and return
6564 * true if the buffer can be reused by the adapter.
6566 static bool igb_add_rx_frag(struct igb_ring
*rx_ring
,
6567 struct igb_rx_buffer
*rx_buffer
,
6568 union e1000_adv_rx_desc
*rx_desc
,
6569 struct sk_buff
*skb
)
6571 struct page
*page
= rx_buffer
->page
;
6572 unsigned char *va
= page_address(page
) + rx_buffer
->page_offset
;
6573 unsigned int size
= le16_to_cpu(rx_desc
->wb
.upper
.length
);
6574 #if (PAGE_SIZE < 8192)
6575 unsigned int truesize
= IGB_RX_BUFSZ
;
6577 unsigned int truesize
= SKB_DATA_ALIGN(size
);
6579 unsigned int pull_len
;
6581 if (unlikely(skb_is_nonlinear(skb
)))
6584 if (unlikely(igb_test_staterr(rx_desc
, E1000_RXDADV_STAT_TSIP
))) {
6585 igb_ptp_rx_pktstamp(rx_ring
->q_vector
, va
, skb
);
6586 va
+= IGB_TS_HDR_LEN
;
6587 size
-= IGB_TS_HDR_LEN
;
6590 if (likely(size
<= IGB_RX_HDR_LEN
)) {
6591 memcpy(__skb_put(skb
, size
), va
, ALIGN(size
, sizeof(long)));
6593 /* page is not reserved, we can reuse buffer as-is */
6594 if (likely(!igb_page_is_reserved(page
)))
6597 /* this page cannot be reused so discard it */
6602 /* we need the header to contain the greater of either ETH_HLEN or
6603 * 60 bytes if the skb->len is less than 60 for skb_pad.
6605 pull_len
= eth_get_headlen(va
, IGB_RX_HDR_LEN
);
6607 /* align pull length to size of long to optimize memcpy performance */
6608 memcpy(__skb_put(skb
, pull_len
), va
, ALIGN(pull_len
, sizeof(long)));
6610 /* update all of the pointers */
6615 skb_add_rx_frag(skb
, skb_shinfo(skb
)->nr_frags
, page
,
6616 (unsigned long)va
& ~PAGE_MASK
, size
, truesize
);
6618 return igb_can_reuse_rx_page(rx_buffer
, page
, truesize
);
6621 static struct sk_buff
*igb_fetch_rx_buffer(struct igb_ring
*rx_ring
,
6622 union e1000_adv_rx_desc
*rx_desc
,
6623 struct sk_buff
*skb
)
6625 struct igb_rx_buffer
*rx_buffer
;
6628 rx_buffer
= &rx_ring
->rx_buffer_info
[rx_ring
->next_to_clean
];
6629 page
= rx_buffer
->page
;
6633 void *page_addr
= page_address(page
) +
6634 rx_buffer
->page_offset
;
6636 /* prefetch first cache line of first page */
6637 prefetch(page_addr
);
6638 #if L1_CACHE_BYTES < 128
6639 prefetch(page_addr
+ L1_CACHE_BYTES
);
6642 /* allocate a skb to store the frags */
6643 skb
= napi_alloc_skb(&rx_ring
->q_vector
->napi
, IGB_RX_HDR_LEN
);
6644 if (unlikely(!skb
)) {
6645 rx_ring
->rx_stats
.alloc_failed
++;
6649 /* we will be copying header into skb->data in
6650 * pskb_may_pull so it is in our interest to prefetch
6651 * it now to avoid a possible cache miss
6653 prefetchw(skb
->data
);
6656 /* we are reusing so sync this buffer for CPU use */
6657 dma_sync_single_range_for_cpu(rx_ring
->dev
,
6659 rx_buffer
->page_offset
,
6663 /* pull page into skb */
6664 if (igb_add_rx_frag(rx_ring
, rx_buffer
, rx_desc
, skb
)) {
6665 /* hand second half of page back to the ring */
6666 igb_reuse_rx_page(rx_ring
, rx_buffer
);
6668 /* we are not reusing the buffer so unmap it */
6669 dma_unmap_page(rx_ring
->dev
, rx_buffer
->dma
,
6670 PAGE_SIZE
, DMA_FROM_DEVICE
);
6673 /* clear contents of rx_buffer */
6674 rx_buffer
->page
= NULL
;
6679 static inline void igb_rx_checksum(struct igb_ring
*ring
,
6680 union e1000_adv_rx_desc
*rx_desc
,
6681 struct sk_buff
*skb
)
6683 skb_checksum_none_assert(skb
);
6685 /* Ignore Checksum bit is set */
6686 if (igb_test_staterr(rx_desc
, E1000_RXD_STAT_IXSM
))
6689 /* Rx checksum disabled via ethtool */
6690 if (!(ring
->netdev
->features
& NETIF_F_RXCSUM
))
6693 /* TCP/UDP checksum error bit is set */
6694 if (igb_test_staterr(rx_desc
,
6695 E1000_RXDEXT_STATERR_TCPE
|
6696 E1000_RXDEXT_STATERR_IPE
)) {
6697 /* work around errata with sctp packets where the TCPE aka
6698 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
6699 * packets, (aka let the stack check the crc32c)
6701 if (!((skb
->len
== 60) &&
6702 test_bit(IGB_RING_FLAG_RX_SCTP_CSUM
, &ring
->flags
))) {
6703 u64_stats_update_begin(&ring
->rx_syncp
);
6704 ring
->rx_stats
.csum_err
++;
6705 u64_stats_update_end(&ring
->rx_syncp
);
6707 /* let the stack verify checksum errors */
6710 /* It must be a TCP or UDP packet with a valid checksum */
6711 if (igb_test_staterr(rx_desc
, E1000_RXD_STAT_TCPCS
|
6712 E1000_RXD_STAT_UDPCS
))
6713 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
6715 dev_dbg(ring
->dev
, "cksum success: bits %08X\n",
6716 le32_to_cpu(rx_desc
->wb
.upper
.status_error
));
6719 static inline void igb_rx_hash(struct igb_ring
*ring
,
6720 union e1000_adv_rx_desc
*rx_desc
,
6721 struct sk_buff
*skb
)
6723 if (ring
->netdev
->features
& NETIF_F_RXHASH
)
6725 le32_to_cpu(rx_desc
->wb
.lower
.hi_dword
.rss
),
6730 * igb_is_non_eop - process handling of non-EOP buffers
6731 * @rx_ring: Rx ring being processed
6732 * @rx_desc: Rx descriptor for current buffer
6733 * @skb: current socket buffer containing buffer in progress
6735 * This function updates next to clean. If the buffer is an EOP buffer
6736 * this function exits returning false, otherwise it will place the
6737 * sk_buff in the next buffer to be chained and return true indicating
6738 * that this is in fact a non-EOP buffer.
6740 static bool igb_is_non_eop(struct igb_ring
*rx_ring
,
6741 union e1000_adv_rx_desc
*rx_desc
)
6743 u32 ntc
= rx_ring
->next_to_clean
+ 1;
6745 /* fetch, update, and store next to clean */
6746 ntc
= (ntc
< rx_ring
->count
) ? ntc
: 0;
6747 rx_ring
->next_to_clean
= ntc
;
6749 prefetch(IGB_RX_DESC(rx_ring
, ntc
));
6751 if (likely(igb_test_staterr(rx_desc
, E1000_RXD_STAT_EOP
)))
6758 * igb_cleanup_headers - Correct corrupted or empty headers
6759 * @rx_ring: rx descriptor ring packet is being transacted on
6760 * @rx_desc: pointer to the EOP Rx descriptor
6761 * @skb: pointer to current skb being fixed
6763 * Address the case where we are pulling data in on pages only
6764 * and as such no data is present in the skb header.
6766 * In addition if skb is not at least 60 bytes we need to pad it so that
6767 * it is large enough to qualify as a valid Ethernet frame.
6769 * Returns true if an error was encountered and skb was freed.
6771 static bool igb_cleanup_headers(struct igb_ring
*rx_ring
,
6772 union e1000_adv_rx_desc
*rx_desc
,
6773 struct sk_buff
*skb
)
6775 if (unlikely((igb_test_staterr(rx_desc
,
6776 E1000_RXDEXT_ERR_FRAME_ERR_MASK
)))) {
6777 struct net_device
*netdev
= rx_ring
->netdev
;
6778 if (!(netdev
->features
& NETIF_F_RXALL
)) {
6779 dev_kfree_skb_any(skb
);
6784 /* if eth_skb_pad returns an error the skb was freed */
6785 if (eth_skb_pad(skb
))
6792 * igb_process_skb_fields - Populate skb header fields from Rx descriptor
6793 * @rx_ring: rx descriptor ring packet is being transacted on
6794 * @rx_desc: pointer to the EOP Rx descriptor
6795 * @skb: pointer to current skb being populated
6797 * This function checks the ring, descriptor, and packet information in
6798 * order to populate the hash, checksum, VLAN, timestamp, protocol, and
6799 * other fields within the skb.
6801 static void igb_process_skb_fields(struct igb_ring
*rx_ring
,
6802 union e1000_adv_rx_desc
*rx_desc
,
6803 struct sk_buff
*skb
)
6805 struct net_device
*dev
= rx_ring
->netdev
;
6807 igb_rx_hash(rx_ring
, rx_desc
, skb
);
6809 igb_rx_checksum(rx_ring
, rx_desc
, skb
);
6811 if (igb_test_staterr(rx_desc
, E1000_RXDADV_STAT_TS
) &&
6812 !igb_test_staterr(rx_desc
, E1000_RXDADV_STAT_TSIP
))
6813 igb_ptp_rx_rgtstamp(rx_ring
->q_vector
, skb
);
6815 if ((dev
->features
& NETIF_F_HW_VLAN_CTAG_RX
) &&
6816 igb_test_staterr(rx_desc
, E1000_RXD_STAT_VP
)) {
6819 if (igb_test_staterr(rx_desc
, E1000_RXDEXT_STATERR_LB
) &&
6820 test_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP
, &rx_ring
->flags
))
6821 vid
= be16_to_cpu(rx_desc
->wb
.upper
.vlan
);
6823 vid
= le16_to_cpu(rx_desc
->wb
.upper
.vlan
);
6825 __vlan_hwaccel_put_tag(skb
, htons(ETH_P_8021Q
), vid
);
6828 skb_record_rx_queue(skb
, rx_ring
->queue_index
);
6830 skb
->protocol
= eth_type_trans(skb
, rx_ring
->netdev
);
6833 static int igb_clean_rx_irq(struct igb_q_vector
*q_vector
, const int budget
)
6835 struct igb_ring
*rx_ring
= q_vector
->rx
.ring
;
6836 struct sk_buff
*skb
= rx_ring
->skb
;
6837 unsigned int total_bytes
= 0, total_packets
= 0;
6838 u16 cleaned_count
= igb_desc_unused(rx_ring
);
6840 while (likely(total_packets
< budget
)) {
6841 union e1000_adv_rx_desc
*rx_desc
;
6843 /* return some buffers to hardware, one at a time is too slow */
6844 if (cleaned_count
>= IGB_RX_BUFFER_WRITE
) {
6845 igb_alloc_rx_buffers(rx_ring
, cleaned_count
);
6849 rx_desc
= IGB_RX_DESC(rx_ring
, rx_ring
->next_to_clean
);
6851 if (!rx_desc
->wb
.upper
.status_error
)
6854 /* This memory barrier is needed to keep us from reading
6855 * any other fields out of the rx_desc until we know the
6856 * descriptor has been written back
6860 /* retrieve a buffer from the ring */
6861 skb
= igb_fetch_rx_buffer(rx_ring
, rx_desc
, skb
);
6863 /* exit if we failed to retrieve a buffer */
6869 /* fetch next buffer in frame if non-eop */
6870 if (igb_is_non_eop(rx_ring
, rx_desc
))
6873 /* verify the packet layout is correct */
6874 if (igb_cleanup_headers(rx_ring
, rx_desc
, skb
)) {
6879 /* probably a little skewed due to removing CRC */
6880 total_bytes
+= skb
->len
;
6882 /* populate checksum, timestamp, VLAN, and protocol */
6883 igb_process_skb_fields(rx_ring
, rx_desc
, skb
);
6885 napi_gro_receive(&q_vector
->napi
, skb
);
6887 /* reset skb pointer */
6890 /* update budget accounting */
6894 /* place incomplete frames back on ring for completion */
6897 u64_stats_update_begin(&rx_ring
->rx_syncp
);
6898 rx_ring
->rx_stats
.packets
+= total_packets
;
6899 rx_ring
->rx_stats
.bytes
+= total_bytes
;
6900 u64_stats_update_end(&rx_ring
->rx_syncp
);
6901 q_vector
->rx
.total_packets
+= total_packets
;
6902 q_vector
->rx
.total_bytes
+= total_bytes
;
6905 igb_alloc_rx_buffers(rx_ring
, cleaned_count
);
6907 return total_packets
;
6910 static bool igb_alloc_mapped_page(struct igb_ring
*rx_ring
,
6911 struct igb_rx_buffer
*bi
)
6913 struct page
*page
= bi
->page
;
6916 /* since we are recycling buffers we should seldom need to alloc */
6920 /* alloc new page for storage */
6921 page
= dev_alloc_page();
6922 if (unlikely(!page
)) {
6923 rx_ring
->rx_stats
.alloc_failed
++;
6927 /* map page for use */
6928 dma
= dma_map_page(rx_ring
->dev
, page
, 0, PAGE_SIZE
, DMA_FROM_DEVICE
);
6930 /* if mapping failed free memory back to system since
6931 * there isn't much point in holding memory we can't use
6933 if (dma_mapping_error(rx_ring
->dev
, dma
)) {
6936 rx_ring
->rx_stats
.alloc_failed
++;
6942 bi
->page_offset
= 0;
6948 * igb_alloc_rx_buffers - Replace used receive buffers; packet split
6949 * @adapter: address of board private structure
6951 void igb_alloc_rx_buffers(struct igb_ring
*rx_ring
, u16 cleaned_count
)
6953 union e1000_adv_rx_desc
*rx_desc
;
6954 struct igb_rx_buffer
*bi
;
6955 u16 i
= rx_ring
->next_to_use
;
6961 rx_desc
= IGB_RX_DESC(rx_ring
, i
);
6962 bi
= &rx_ring
->rx_buffer_info
[i
];
6963 i
-= rx_ring
->count
;
6966 if (!igb_alloc_mapped_page(rx_ring
, bi
))
6969 /* Refresh the desc even if buffer_addrs didn't change
6970 * because each write-back erases this info.
6972 rx_desc
->read
.pkt_addr
= cpu_to_le64(bi
->dma
+ bi
->page_offset
);
6978 rx_desc
= IGB_RX_DESC(rx_ring
, 0);
6979 bi
= rx_ring
->rx_buffer_info
;
6980 i
-= rx_ring
->count
;
6983 /* clear the status bits for the next_to_use descriptor */
6984 rx_desc
->wb
.upper
.status_error
= 0;
6987 } while (cleaned_count
);
6989 i
+= rx_ring
->count
;
6991 if (rx_ring
->next_to_use
!= i
) {
6992 /* record the next descriptor to use */
6993 rx_ring
->next_to_use
= i
;
6995 /* update next to alloc since we have filled the ring */
6996 rx_ring
->next_to_alloc
= i
;
6998 /* Force memory writes to complete before letting h/w
6999 * know there are new descriptors to fetch. (Only
7000 * applicable for weak-ordered memory model archs,
7004 writel(i
, rx_ring
->tail
);
7014 static int igb_mii_ioctl(struct net_device
*netdev
, struct ifreq
*ifr
, int cmd
)
7016 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7017 struct mii_ioctl_data
*data
= if_mii(ifr
);
7019 if (adapter
->hw
.phy
.media_type
!= e1000_media_type_copper
)
7024 data
->phy_id
= adapter
->hw
.phy
.addr
;
7027 if (igb_read_phy_reg(&adapter
->hw
, data
->reg_num
& 0x1F,
7044 static int igb_ioctl(struct net_device
*netdev
, struct ifreq
*ifr
, int cmd
)
7050 return igb_mii_ioctl(netdev
, ifr
, cmd
);
7052 return igb_ptp_get_ts_config(netdev
, ifr
);
7054 return igb_ptp_set_ts_config(netdev
, ifr
);
7060 void igb_read_pci_cfg(struct e1000_hw
*hw
, u32 reg
, u16
*value
)
7062 struct igb_adapter
*adapter
= hw
->back
;
7064 pci_read_config_word(adapter
->pdev
, reg
, value
);
7067 void igb_write_pci_cfg(struct e1000_hw
*hw
, u32 reg
, u16
*value
)
7069 struct igb_adapter
*adapter
= hw
->back
;
7071 pci_write_config_word(adapter
->pdev
, reg
, *value
);
7074 s32
igb_read_pcie_cap_reg(struct e1000_hw
*hw
, u32 reg
, u16
*value
)
7076 struct igb_adapter
*adapter
= hw
->back
;
7078 if (pcie_capability_read_word(adapter
->pdev
, reg
, value
))
7079 return -E1000_ERR_CONFIG
;
7084 s32
igb_write_pcie_cap_reg(struct e1000_hw
*hw
, u32 reg
, u16
*value
)
7086 struct igb_adapter
*adapter
= hw
->back
;
7088 if (pcie_capability_write_word(adapter
->pdev
, reg
, *value
))
7089 return -E1000_ERR_CONFIG
;
7094 static void igb_vlan_mode(struct net_device
*netdev
, netdev_features_t features
)
7096 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7097 struct e1000_hw
*hw
= &adapter
->hw
;
7099 bool enable
= !!(features
& NETIF_F_HW_VLAN_CTAG_RX
);
7102 /* enable VLAN tag insert/strip */
7103 ctrl
= rd32(E1000_CTRL
);
7104 ctrl
|= E1000_CTRL_VME
;
7105 wr32(E1000_CTRL
, ctrl
);
7107 /* Disable CFI check */
7108 rctl
= rd32(E1000_RCTL
);
7109 rctl
&= ~E1000_RCTL_CFIEN
;
7110 wr32(E1000_RCTL
, rctl
);
7112 /* disable VLAN tag insert/strip */
7113 ctrl
= rd32(E1000_CTRL
);
7114 ctrl
&= ~E1000_CTRL_VME
;
7115 wr32(E1000_CTRL
, ctrl
);
7119 static int igb_vlan_rx_add_vid(struct net_device
*netdev
,
7120 __be16 proto
, u16 vid
)
7122 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7123 struct e1000_hw
*hw
= &adapter
->hw
;
7124 int pf_id
= adapter
->vfs_allocated_count
;
7126 /* add the filter since PF can receive vlans w/o entry in vlvf */
7127 igb_vfta_set(hw
, vid
, pf_id
, true, true);
7128 set_bit(vid
, adapter
->active_vlans
);
7133 static int igb_vlan_rx_kill_vid(struct net_device
*netdev
,
7134 __be16 proto
, u16 vid
)
7136 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7137 int pf_id
= adapter
->vfs_allocated_count
;
7138 struct e1000_hw
*hw
= &adapter
->hw
;
7140 /* remove VID from filter table */
7141 igb_vfta_set(hw
, vid
, pf_id
, false, true);
7143 clear_bit(vid
, adapter
->active_vlans
);
7148 static void igb_restore_vlan(struct igb_adapter
*adapter
)
7152 igb_vlan_mode(adapter
->netdev
, adapter
->netdev
->features
);
7153 igb_vlan_rx_add_vid(adapter
->netdev
, htons(ETH_P_8021Q
), 0);
7155 for_each_set_bit_from(vid
, adapter
->active_vlans
, VLAN_N_VID
)
7156 igb_vlan_rx_add_vid(adapter
->netdev
, htons(ETH_P_8021Q
), vid
);
7159 int igb_set_spd_dplx(struct igb_adapter
*adapter
, u32 spd
, u8 dplx
)
7161 struct pci_dev
*pdev
= adapter
->pdev
;
7162 struct e1000_mac_info
*mac
= &adapter
->hw
.mac
;
7166 /* Make sure dplx is at most 1 bit and lsb of speed is not set
7167 * for the switch() below to work
7169 if ((spd
& 1) || (dplx
& ~1))
7172 /* Fiber NIC's only allow 1000 gbps Full duplex
7173 * and 100Mbps Full duplex for 100baseFx sfp
7175 if (adapter
->hw
.phy
.media_type
== e1000_media_type_internal_serdes
) {
7176 switch (spd
+ dplx
) {
7177 case SPEED_10
+ DUPLEX_HALF
:
7178 case SPEED_10
+ DUPLEX_FULL
:
7179 case SPEED_100
+ DUPLEX_HALF
:
7186 switch (spd
+ dplx
) {
7187 case SPEED_10
+ DUPLEX_HALF
:
7188 mac
->forced_speed_duplex
= ADVERTISE_10_HALF
;
7190 case SPEED_10
+ DUPLEX_FULL
:
7191 mac
->forced_speed_duplex
= ADVERTISE_10_FULL
;
7193 case SPEED_100
+ DUPLEX_HALF
:
7194 mac
->forced_speed_duplex
= ADVERTISE_100_HALF
;
7196 case SPEED_100
+ DUPLEX_FULL
:
7197 mac
->forced_speed_duplex
= ADVERTISE_100_FULL
;
7199 case SPEED_1000
+ DUPLEX_FULL
:
7201 adapter
->hw
.phy
.autoneg_advertised
= ADVERTISE_1000_FULL
;
7203 case SPEED_1000
+ DUPLEX_HALF
: /* not supported */
7208 /* clear MDI, MDI(-X) override is only allowed when autoneg enabled */
7209 adapter
->hw
.phy
.mdix
= AUTO_ALL_MODES
;
7214 dev_err(&pdev
->dev
, "Unsupported Speed/Duplex configuration\n");
7218 static int __igb_shutdown(struct pci_dev
*pdev
, bool *enable_wake
,
7221 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7222 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7223 struct e1000_hw
*hw
= &adapter
->hw
;
7224 u32 ctrl
, rctl
, status
;
7225 u32 wufc
= runtime
? E1000_WUFC_LNKC
: adapter
->wol
;
7230 netif_device_detach(netdev
);
7232 if (netif_running(netdev
))
7233 __igb_close(netdev
, true);
7235 igb_clear_interrupt_scheme(adapter
);
7238 retval
= pci_save_state(pdev
);
7243 status
= rd32(E1000_STATUS
);
7244 if (status
& E1000_STATUS_LU
)
7245 wufc
&= ~E1000_WUFC_LNKC
;
7248 igb_setup_rctl(adapter
);
7249 igb_set_rx_mode(netdev
);
7251 /* turn on all-multi mode if wake on multicast is enabled */
7252 if (wufc
& E1000_WUFC_MC
) {
7253 rctl
= rd32(E1000_RCTL
);
7254 rctl
|= E1000_RCTL_MPE
;
7255 wr32(E1000_RCTL
, rctl
);
7258 ctrl
= rd32(E1000_CTRL
);
7259 /* advertise wake from D3Cold */
7260 #define E1000_CTRL_ADVD3WUC 0x00100000
7261 /* phy power management enable */
7262 #define E1000_CTRL_EN_PHY_PWR_MGMT 0x00200000
7263 ctrl
|= E1000_CTRL_ADVD3WUC
;
7264 wr32(E1000_CTRL
, ctrl
);
7266 /* Allow time for pending master requests to run */
7267 igb_disable_pcie_master(hw
);
7269 wr32(E1000_WUC
, E1000_WUC_PME_EN
);
7270 wr32(E1000_WUFC
, wufc
);
7273 wr32(E1000_WUFC
, 0);
7276 *enable_wake
= wufc
|| adapter
->en_mng_pt
;
7278 igb_power_down_link(adapter
);
7280 igb_power_up_link(adapter
);
7282 /* Release control of h/w to f/w. If f/w is AMT enabled, this
7283 * would have already happened in close and is redundant.
7285 igb_release_hw_control(adapter
);
7287 pci_disable_device(pdev
);
7293 #ifdef CONFIG_PM_SLEEP
7294 static int igb_suspend(struct device
*dev
)
7298 struct pci_dev
*pdev
= to_pci_dev(dev
);
7300 retval
= __igb_shutdown(pdev
, &wake
, 0);
7305 pci_prepare_to_sleep(pdev
);
7307 pci_wake_from_d3(pdev
, false);
7308 pci_set_power_state(pdev
, PCI_D3hot
);
7313 #endif /* CONFIG_PM_SLEEP */
7315 static int igb_resume(struct device
*dev
)
7317 struct pci_dev
*pdev
= to_pci_dev(dev
);
7318 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7319 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7320 struct e1000_hw
*hw
= &adapter
->hw
;
7323 pci_set_power_state(pdev
, PCI_D0
);
7324 pci_restore_state(pdev
);
7325 pci_save_state(pdev
);
7327 if (!pci_device_is_present(pdev
))
7329 err
= pci_enable_device_mem(pdev
);
7332 "igb: Cannot enable PCI device from suspend\n");
7335 pci_set_master(pdev
);
7337 pci_enable_wake(pdev
, PCI_D3hot
, 0);
7338 pci_enable_wake(pdev
, PCI_D3cold
, 0);
7340 if (igb_init_interrupt_scheme(adapter
, true)) {
7341 dev_err(&pdev
->dev
, "Unable to allocate memory for queues\n");
7348 /* let the f/w know that the h/w is now under the control of the
7351 igb_get_hw_control(adapter
);
7353 wr32(E1000_WUS
, ~0);
7355 if (netdev
->flags
& IFF_UP
) {
7357 err
= __igb_open(netdev
, true);
7363 netif_device_attach(netdev
);
7367 static int igb_runtime_idle(struct device
*dev
)
7369 struct pci_dev
*pdev
= to_pci_dev(dev
);
7370 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7371 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7373 if (!igb_has_link(adapter
))
7374 pm_schedule_suspend(dev
, MSEC_PER_SEC
* 5);
7379 static int igb_runtime_suspend(struct device
*dev
)
7381 struct pci_dev
*pdev
= to_pci_dev(dev
);
7385 retval
= __igb_shutdown(pdev
, &wake
, 1);
7390 pci_prepare_to_sleep(pdev
);
7392 pci_wake_from_d3(pdev
, false);
7393 pci_set_power_state(pdev
, PCI_D3hot
);
7399 static int igb_runtime_resume(struct device
*dev
)
7401 return igb_resume(dev
);
7403 #endif /* CONFIG_PM */
7405 static void igb_shutdown(struct pci_dev
*pdev
)
7409 __igb_shutdown(pdev
, &wake
, 0);
7411 if (system_state
== SYSTEM_POWER_OFF
) {
7412 pci_wake_from_d3(pdev
, wake
);
7413 pci_set_power_state(pdev
, PCI_D3hot
);
7417 #ifdef CONFIG_PCI_IOV
7418 static int igb_sriov_reinit(struct pci_dev
*dev
)
7420 struct net_device
*netdev
= pci_get_drvdata(dev
);
7421 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7422 struct pci_dev
*pdev
= adapter
->pdev
;
7426 if (netif_running(netdev
))
7431 igb_clear_interrupt_scheme(adapter
);
7433 igb_init_queue_configuration(adapter
);
7435 if (igb_init_interrupt_scheme(adapter
, true)) {
7437 dev_err(&pdev
->dev
, "Unable to allocate memory for queues\n");
7441 if (netif_running(netdev
))
7449 static int igb_pci_disable_sriov(struct pci_dev
*dev
)
7451 int err
= igb_disable_sriov(dev
);
7454 err
= igb_sriov_reinit(dev
);
7459 static int igb_pci_enable_sriov(struct pci_dev
*dev
, int num_vfs
)
7461 int err
= igb_enable_sriov(dev
, num_vfs
);
7466 err
= igb_sriov_reinit(dev
);
7475 static int igb_pci_sriov_configure(struct pci_dev
*dev
, int num_vfs
)
7477 #ifdef CONFIG_PCI_IOV
7479 return igb_pci_disable_sriov(dev
);
7481 return igb_pci_enable_sriov(dev
, num_vfs
);
7486 #ifdef CONFIG_NET_POLL_CONTROLLER
7487 /* Polling 'interrupt' - used by things like netconsole to send skbs
7488 * without having to re-enable interrupts. It's not called while
7489 * the interrupt routine is executing.
7491 static void igb_netpoll(struct net_device
*netdev
)
7493 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7494 struct e1000_hw
*hw
= &adapter
->hw
;
7495 struct igb_q_vector
*q_vector
;
7498 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
7499 q_vector
= adapter
->q_vector
[i
];
7500 if (adapter
->flags
& IGB_FLAG_HAS_MSIX
)
7501 wr32(E1000_EIMC
, q_vector
->eims_value
);
7503 igb_irq_disable(adapter
);
7504 napi_schedule(&q_vector
->napi
);
7507 #endif /* CONFIG_NET_POLL_CONTROLLER */
7510 * igb_io_error_detected - called when PCI error is detected
7511 * @pdev: Pointer to PCI device
7512 * @state: The current pci connection state
7514 * This function is called after a PCI bus error affecting
7515 * this device has been detected.
7517 static pci_ers_result_t
igb_io_error_detected(struct pci_dev
*pdev
,
7518 pci_channel_state_t state
)
7520 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7521 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7523 netif_device_detach(netdev
);
7525 if (state
== pci_channel_io_perm_failure
)
7526 return PCI_ERS_RESULT_DISCONNECT
;
7528 if (netif_running(netdev
))
7530 pci_disable_device(pdev
);
7532 /* Request a slot slot reset. */
7533 return PCI_ERS_RESULT_NEED_RESET
;
7537 * igb_io_slot_reset - called after the pci bus has been reset.
7538 * @pdev: Pointer to PCI device
7540 * Restart the card from scratch, as if from a cold-boot. Implementation
7541 * resembles the first-half of the igb_resume routine.
7543 static pci_ers_result_t
igb_io_slot_reset(struct pci_dev
*pdev
)
7545 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7546 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7547 struct e1000_hw
*hw
= &adapter
->hw
;
7548 pci_ers_result_t result
;
7551 if (pci_enable_device_mem(pdev
)) {
7553 "Cannot re-enable PCI device after reset.\n");
7554 result
= PCI_ERS_RESULT_DISCONNECT
;
7556 pci_set_master(pdev
);
7557 pci_restore_state(pdev
);
7558 pci_save_state(pdev
);
7560 pci_enable_wake(pdev
, PCI_D3hot
, 0);
7561 pci_enable_wake(pdev
, PCI_D3cold
, 0);
7564 wr32(E1000_WUS
, ~0);
7565 result
= PCI_ERS_RESULT_RECOVERED
;
7568 err
= pci_cleanup_aer_uncorrect_error_status(pdev
);
7571 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
7573 /* non-fatal, continue */
7580 * igb_io_resume - called when traffic can start flowing again.
7581 * @pdev: Pointer to PCI device
7583 * This callback is called when the error recovery driver tells us that
7584 * its OK to resume normal operation. Implementation resembles the
7585 * second-half of the igb_resume routine.
7587 static void igb_io_resume(struct pci_dev
*pdev
)
7589 struct net_device
*netdev
= pci_get_drvdata(pdev
);
7590 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7592 if (netif_running(netdev
)) {
7593 if (igb_up(adapter
)) {
7594 dev_err(&pdev
->dev
, "igb_up failed after reset\n");
7599 netif_device_attach(netdev
);
7601 /* let the f/w know that the h/w is now under the control of the
7604 igb_get_hw_control(adapter
);
7607 static void igb_rar_set_qsel(struct igb_adapter
*adapter
, u8
*addr
, u32 index
,
7610 struct e1000_hw
*hw
= &adapter
->hw
;
7611 u32 rar_low
, rar_high
;
7613 /* HW expects these in little endian so we reverse the byte order
7614 * from network order (big endian) to CPU endian
7616 rar_low
= le32_to_cpup((__be32
*)(addr
));
7617 rar_high
= le16_to_cpup((__be16
*)(addr
+ 4));
7619 /* Indicate to hardware the Address is Valid. */
7620 rar_high
|= E1000_RAH_AV
;
7622 if (hw
->mac
.type
== e1000_82575
)
7623 rar_high
|= E1000_RAH_POOL_1
* qsel
;
7625 rar_high
|= E1000_RAH_POOL_1
<< qsel
;
7627 wr32(E1000_RAL(index
), rar_low
);
7629 wr32(E1000_RAH(index
), rar_high
);
7633 static int igb_set_vf_mac(struct igb_adapter
*adapter
,
7634 int vf
, unsigned char *mac_addr
)
7636 struct e1000_hw
*hw
= &adapter
->hw
;
7637 /* VF MAC addresses start at end of receive addresses and moves
7638 * towards the first, as a result a collision should not be possible
7640 int rar_entry
= hw
->mac
.rar_entry_count
- (vf
+ 1);
7642 memcpy(adapter
->vf_data
[vf
].vf_mac_addresses
, mac_addr
, ETH_ALEN
);
7644 igb_rar_set_qsel(adapter
, mac_addr
, rar_entry
, vf
);
7649 static int igb_ndo_set_vf_mac(struct net_device
*netdev
, int vf
, u8
*mac
)
7651 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7652 if (!is_valid_ether_addr(mac
) || (vf
>= adapter
->vfs_allocated_count
))
7654 adapter
->vf_data
[vf
].flags
|= IGB_VF_FLAG_PF_SET_MAC
;
7655 dev_info(&adapter
->pdev
->dev
, "setting MAC %pM on VF %d\n", mac
, vf
);
7656 dev_info(&adapter
->pdev
->dev
,
7657 "Reload the VF driver to make this change effective.");
7658 if (test_bit(__IGB_DOWN
, &adapter
->state
)) {
7659 dev_warn(&adapter
->pdev
->dev
,
7660 "The VF MAC address has been set, but the PF device is not up.\n");
7661 dev_warn(&adapter
->pdev
->dev
,
7662 "Bring the PF device up before attempting to use the VF device.\n");
7664 return igb_set_vf_mac(adapter
, vf
, mac
);
7667 static int igb_link_mbps(int internal_link_speed
)
7669 switch (internal_link_speed
) {
7679 static void igb_set_vf_rate_limit(struct e1000_hw
*hw
, int vf
, int tx_rate
,
7686 /* Calculate the rate factor values to set */
7687 rf_int
= link_speed
/ tx_rate
;
7688 rf_dec
= (link_speed
- (rf_int
* tx_rate
));
7689 rf_dec
= (rf_dec
* (1 << E1000_RTTBCNRC_RF_INT_SHIFT
)) /
7692 bcnrc_val
= E1000_RTTBCNRC_RS_ENA
;
7693 bcnrc_val
|= ((rf_int
<< E1000_RTTBCNRC_RF_INT_SHIFT
) &
7694 E1000_RTTBCNRC_RF_INT_MASK
);
7695 bcnrc_val
|= (rf_dec
& E1000_RTTBCNRC_RF_DEC_MASK
);
7700 wr32(E1000_RTTDQSEL
, vf
); /* vf X uses queue X */
7701 /* Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
7702 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported.
7704 wr32(E1000_RTTBCNRM
, 0x14);
7705 wr32(E1000_RTTBCNRC
, bcnrc_val
);
7708 static void igb_check_vf_rate_limit(struct igb_adapter
*adapter
)
7710 int actual_link_speed
, i
;
7711 bool reset_rate
= false;
7713 /* VF TX rate limit was not set or not supported */
7714 if ((adapter
->vf_rate_link_speed
== 0) ||
7715 (adapter
->hw
.mac
.type
!= e1000_82576
))
7718 actual_link_speed
= igb_link_mbps(adapter
->link_speed
);
7719 if (actual_link_speed
!= adapter
->vf_rate_link_speed
) {
7721 adapter
->vf_rate_link_speed
= 0;
7722 dev_info(&adapter
->pdev
->dev
,
7723 "Link speed has been changed. VF Transmit rate is disabled\n");
7726 for (i
= 0; i
< adapter
->vfs_allocated_count
; i
++) {
7728 adapter
->vf_data
[i
].tx_rate
= 0;
7730 igb_set_vf_rate_limit(&adapter
->hw
, i
,
7731 adapter
->vf_data
[i
].tx_rate
,
7736 static int igb_ndo_set_vf_bw(struct net_device
*netdev
, int vf
,
7737 int min_tx_rate
, int max_tx_rate
)
7739 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7740 struct e1000_hw
*hw
= &adapter
->hw
;
7741 int actual_link_speed
;
7743 if (hw
->mac
.type
!= e1000_82576
)
7749 actual_link_speed
= igb_link_mbps(adapter
->link_speed
);
7750 if ((vf
>= adapter
->vfs_allocated_count
) ||
7751 (!(rd32(E1000_STATUS
) & E1000_STATUS_LU
)) ||
7752 (max_tx_rate
< 0) ||
7753 (max_tx_rate
> actual_link_speed
))
7756 adapter
->vf_rate_link_speed
= actual_link_speed
;
7757 adapter
->vf_data
[vf
].tx_rate
= (u16
)max_tx_rate
;
7758 igb_set_vf_rate_limit(hw
, vf
, max_tx_rate
, actual_link_speed
);
7763 static int igb_ndo_set_vf_spoofchk(struct net_device
*netdev
, int vf
,
7766 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7767 struct e1000_hw
*hw
= &adapter
->hw
;
7768 u32 reg_val
, reg_offset
;
7770 if (!adapter
->vfs_allocated_count
)
7773 if (vf
>= adapter
->vfs_allocated_count
)
7776 reg_offset
= (hw
->mac
.type
== e1000_82576
) ? E1000_DTXSWC
: E1000_TXSWC
;
7777 reg_val
= rd32(reg_offset
);
7779 reg_val
|= ((1 << vf
) |
7780 (1 << (vf
+ E1000_DTXSWC_VLAN_SPOOF_SHIFT
)));
7782 reg_val
&= ~((1 << vf
) |
7783 (1 << (vf
+ E1000_DTXSWC_VLAN_SPOOF_SHIFT
)));
7784 wr32(reg_offset
, reg_val
);
7786 adapter
->vf_data
[vf
].spoofchk_enabled
= setting
;
7790 static int igb_ndo_get_vf_config(struct net_device
*netdev
,
7791 int vf
, struct ifla_vf_info
*ivi
)
7793 struct igb_adapter
*adapter
= netdev_priv(netdev
);
7794 if (vf
>= adapter
->vfs_allocated_count
)
7797 memcpy(&ivi
->mac
, adapter
->vf_data
[vf
].vf_mac_addresses
, ETH_ALEN
);
7798 ivi
->max_tx_rate
= adapter
->vf_data
[vf
].tx_rate
;
7799 ivi
->min_tx_rate
= 0;
7800 ivi
->vlan
= adapter
->vf_data
[vf
].pf_vlan
;
7801 ivi
->qos
= adapter
->vf_data
[vf
].pf_qos
;
7802 ivi
->spoofchk
= adapter
->vf_data
[vf
].spoofchk_enabled
;
7806 static void igb_vmm_control(struct igb_adapter
*adapter
)
7808 struct e1000_hw
*hw
= &adapter
->hw
;
7811 switch (hw
->mac
.type
) {
7817 /* replication is not supported for 82575 */
7820 /* notify HW that the MAC is adding vlan tags */
7821 reg
= rd32(E1000_DTXCTL
);
7822 reg
|= E1000_DTXCTL_VLAN_ADDED
;
7823 wr32(E1000_DTXCTL
, reg
);
7826 /* enable replication vlan tag stripping */
7827 reg
= rd32(E1000_RPLOLR
);
7828 reg
|= E1000_RPLOLR_STRVLAN
;
7829 wr32(E1000_RPLOLR
, reg
);
7832 /* none of the above registers are supported by i350 */
7836 if (adapter
->vfs_allocated_count
) {
7837 igb_vmdq_set_loopback_pf(hw
, true);
7838 igb_vmdq_set_replication_pf(hw
, true);
7839 igb_vmdq_set_anti_spoofing_pf(hw
, true,
7840 adapter
->vfs_allocated_count
);
7842 igb_vmdq_set_loopback_pf(hw
, false);
7843 igb_vmdq_set_replication_pf(hw
, false);
7847 static void igb_init_dmac(struct igb_adapter
*adapter
, u32 pba
)
7849 struct e1000_hw
*hw
= &adapter
->hw
;
7853 if (hw
->mac
.type
> e1000_82580
) {
7854 if (adapter
->flags
& IGB_FLAG_DMAC
) {
7857 /* force threshold to 0. */
7858 wr32(E1000_DMCTXTH
, 0);
7860 /* DMA Coalescing high water mark needs to be greater
7861 * than the Rx threshold. Set hwm to PBA - max frame
7862 * size in 16B units, capping it at PBA - 6KB.
7864 hwm
= 64 * (pba
- 6);
7865 reg
= rd32(E1000_FCRTC
);
7866 reg
&= ~E1000_FCRTC_RTH_COAL_MASK
;
7867 reg
|= ((hwm
<< E1000_FCRTC_RTH_COAL_SHIFT
)
7868 & E1000_FCRTC_RTH_COAL_MASK
);
7869 wr32(E1000_FCRTC
, reg
);
7871 /* Set the DMA Coalescing Rx threshold to PBA - 2 * max
7872 * frame size, capping it at PBA - 10KB.
7874 dmac_thr
= pba
- 10;
7875 reg
= rd32(E1000_DMACR
);
7876 reg
&= ~E1000_DMACR_DMACTHR_MASK
;
7877 reg
|= ((dmac_thr
<< E1000_DMACR_DMACTHR_SHIFT
)
7878 & E1000_DMACR_DMACTHR_MASK
);
7880 /* transition to L0x or L1 if available..*/
7881 reg
|= (E1000_DMACR_DMAC_EN
| E1000_DMACR_DMAC_LX_MASK
);
7883 /* watchdog timer= +-1000 usec in 32usec intervals */
7886 /* Disable BMC-to-OS Watchdog Enable */
7887 if (hw
->mac
.type
!= e1000_i354
)
7888 reg
&= ~E1000_DMACR_DC_BMC2OSW_EN
;
7890 wr32(E1000_DMACR
, reg
);
7892 /* no lower threshold to disable
7893 * coalescing(smart fifb)-UTRESH=0
7895 wr32(E1000_DMCRTRH
, 0);
7897 reg
= (IGB_DMCTLX_DCFLUSH_DIS
| 0x4);
7899 wr32(E1000_DMCTLX
, reg
);
7901 /* free space in tx packet buffer to wake from
7904 wr32(E1000_DMCTXTH
, (IGB_MIN_TXPBSIZE
-
7905 (IGB_TX_BUF_4096
+ adapter
->max_frame_size
)) >> 6);
7907 /* make low power state decision controlled
7910 reg
= rd32(E1000_PCIEMISC
);
7911 reg
&= ~E1000_PCIEMISC_LX_DECISION
;
7912 wr32(E1000_PCIEMISC
, reg
);
7913 } /* endif adapter->dmac is not disabled */
7914 } else if (hw
->mac
.type
== e1000_82580
) {
7915 u32 reg
= rd32(E1000_PCIEMISC
);
7917 wr32(E1000_PCIEMISC
, reg
& ~E1000_PCIEMISC_LX_DECISION
);
7918 wr32(E1000_DMACR
, 0);
7923 * igb_read_i2c_byte - Reads 8 bit word over I2C
7924 * @hw: pointer to hardware structure
7925 * @byte_offset: byte offset to read
7926 * @dev_addr: device address
7929 * Performs byte read operation over I2C interface at
7930 * a specified device address.
7932 s32
igb_read_i2c_byte(struct e1000_hw
*hw
, u8 byte_offset
,
7933 u8 dev_addr
, u8
*data
)
7935 struct igb_adapter
*adapter
= container_of(hw
, struct igb_adapter
, hw
);
7936 struct i2c_client
*this_client
= adapter
->i2c_client
;
7941 return E1000_ERR_I2C
;
7943 swfw_mask
= E1000_SWFW_PHY0_SM
;
7945 if (hw
->mac
.ops
.acquire_swfw_sync(hw
, swfw_mask
))
7946 return E1000_ERR_SWFW_SYNC
;
7948 status
= i2c_smbus_read_byte_data(this_client
, byte_offset
);
7949 hw
->mac
.ops
.release_swfw_sync(hw
, swfw_mask
);
7952 return E1000_ERR_I2C
;
7960 * igb_write_i2c_byte - Writes 8 bit word over I2C
7961 * @hw: pointer to hardware structure
7962 * @byte_offset: byte offset to write
7963 * @dev_addr: device address
7964 * @data: value to write
7966 * Performs byte write operation over I2C interface at
7967 * a specified device address.
7969 s32
igb_write_i2c_byte(struct e1000_hw
*hw
, u8 byte_offset
,
7970 u8 dev_addr
, u8 data
)
7972 struct igb_adapter
*adapter
= container_of(hw
, struct igb_adapter
, hw
);
7973 struct i2c_client
*this_client
= adapter
->i2c_client
;
7975 u16 swfw_mask
= E1000_SWFW_PHY0_SM
;
7978 return E1000_ERR_I2C
;
7980 if (hw
->mac
.ops
.acquire_swfw_sync(hw
, swfw_mask
))
7981 return E1000_ERR_SWFW_SYNC
;
7982 status
= i2c_smbus_write_byte_data(this_client
, byte_offset
, data
);
7983 hw
->mac
.ops
.release_swfw_sync(hw
, swfw_mask
);
7986 return E1000_ERR_I2C
;
7992 int igb_reinit_queues(struct igb_adapter
*adapter
)
7994 struct net_device
*netdev
= adapter
->netdev
;
7995 struct pci_dev
*pdev
= adapter
->pdev
;
7998 if (netif_running(netdev
))
8001 igb_reset_interrupt_capability(adapter
);
8003 if (igb_init_interrupt_scheme(adapter
, true)) {
8004 dev_err(&pdev
->dev
, "Unable to allocate memory for queues\n");
8008 if (netif_running(netdev
))
8009 err
= igb_open(netdev
);