1 /*******************************************************************************
3 Intel 10 Gigabit PCI Express Linux driver
4 Copyright(c) 1999 - 2013 Intel Corporation.
6 This program is free software; you can redistribute it and/or modify it
7 under the terms and conditions of the GNU General Public License,
8 version 2, as published by the Free Software Foundation.
10 This program is distributed in the hope it will be useful, but WITHOUT
11 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
15 You should have received a copy of the GNU General Public License along with
16 this program; if not, write to the Free Software Foundation, Inc.,
17 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
19 The full GNU General Public License is included in this distribution in
20 the file called "COPYING".
23 e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
24 Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
26 *******************************************************************************/
28 #include <linux/types.h>
29 #include <linux/module.h>
30 #include <linux/pci.h>
31 #include <linux/netdevice.h>
32 #include <linux/vmalloc.h>
33 #include <linux/string.h>
35 #include <linux/interrupt.h>
37 #include <linux/tcp.h>
38 #include <linux/sctp.h>
39 #include <linux/pkt_sched.h>
40 #include <linux/ipv6.h>
41 #include <linux/slab.h>
42 #include <net/checksum.h>
43 #include <net/ip6_checksum.h>
44 #include <linux/ethtool.h>
46 #include <linux/if_vlan.h>
47 #include <linux/if_bridge.h>
48 #include <linux/prefetch.h>
49 #include <scsi/fc/fc_fcoe.h>
52 #include "ixgbe_common.h"
53 #include "ixgbe_dcb_82599.h"
54 #include "ixgbe_sriov.h"
56 char ixgbe_driver_name
[] = "ixgbe";
57 static const char ixgbe_driver_string
[] =
58 "Intel(R) 10 Gigabit PCI Express Network Driver";
60 char ixgbe_default_device_descr
[] =
61 "Intel(R) 10 Gigabit Network Connection";
63 static char ixgbe_default_device_descr
[] =
64 "Intel(R) 10 Gigabit Network Connection";
66 #define DRV_VERSION "3.13.10-k"
67 const char ixgbe_driver_version
[] = DRV_VERSION
;
68 static const char ixgbe_copyright
[] =
69 "Copyright (c) 1999-2013 Intel Corporation.";
71 static const struct ixgbe_info
*ixgbe_info_tbl
[] = {
72 [board_82598
] = &ixgbe_82598_info
,
73 [board_82599
] = &ixgbe_82599_info
,
74 [board_X540
] = &ixgbe_X540_info
,
77 /* ixgbe_pci_tbl - PCI Device ID Table
79 * Wildcard entries (PCI_ANY_ID) should come last
80 * Last entry must be all 0s
82 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
83 * Class, Class Mask, private data (not used) }
85 static DEFINE_PCI_DEVICE_TABLE(ixgbe_pci_tbl
) = {
86 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598
), board_82598
},
87 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AF_DUAL_PORT
), board_82598
},
88 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AF_SINGLE_PORT
), board_82598
},
89 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AT
), board_82598
},
90 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AT2
), board_82598
},
91 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598EB_CX4
), board_82598
},
92 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_CX4_DUAL_PORT
), board_82598
},
93 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_DA_DUAL_PORT
), board_82598
},
94 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM
), board_82598
},
95 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598EB_XF_LR
), board_82598
},
96 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598EB_SFP_LOM
), board_82598
},
97 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_BX
), board_82598
},
98 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_KX4
), board_82599
},
99 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_XAUI_LOM
), board_82599
},
100 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_KR
), board_82599
},
101 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP
), board_82599
},
102 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_EM
), board_82599
},
103 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_KX4_MEZZ
), board_82599
},
104 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_CX4
), board_82599
},
105 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_BACKPLANE_FCOE
), board_82599
},
106 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_FCOE
), board_82599
},
107 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_T3_LOM
), board_82599
},
108 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_COMBO_BACKPLANE
), board_82599
},
109 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X540T
), board_X540
},
110 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_SF2
), board_82599
},
111 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_LS
), board_82599
},
112 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599EN_SFP
), board_82599
},
113 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_SF_QP
), board_82599
},
114 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X540T1
), board_X540
},
115 /* required last entry */
118 MODULE_DEVICE_TABLE(pci
, ixgbe_pci_tbl
);
120 #ifdef CONFIG_IXGBE_DCA
121 static int ixgbe_notify_dca(struct notifier_block
*, unsigned long event
,
123 static struct notifier_block dca_notifier
= {
124 .notifier_call
= ixgbe_notify_dca
,
130 #ifdef CONFIG_PCI_IOV
131 static unsigned int max_vfs
;
132 module_param(max_vfs
, uint
, 0);
133 MODULE_PARM_DESC(max_vfs
,
134 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63");
135 #endif /* CONFIG_PCI_IOV */
137 static unsigned int allow_unsupported_sfp
;
138 module_param(allow_unsupported_sfp
, uint
, 0);
139 MODULE_PARM_DESC(allow_unsupported_sfp
,
140 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
142 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
143 static int debug
= -1;
144 module_param(debug
, int, 0);
145 MODULE_PARM_DESC(debug
, "Debug level (0=none,...,16=all)");
147 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
148 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
149 MODULE_LICENSE("GPL");
150 MODULE_VERSION(DRV_VERSION
);
152 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter
*adapter
,
156 struct pci_dev
*parent_dev
;
157 struct pci_bus
*parent_bus
;
159 parent_bus
= adapter
->pdev
->bus
->parent
;
163 parent_dev
= parent_bus
->self
;
167 pos
= pci_find_capability(parent_dev
, PCI_CAP_ID_EXP
);
171 pci_read_config_word(parent_dev
, pos
+ reg
, value
);
175 static s32
ixgbe_get_parent_bus_info(struct ixgbe_adapter
*adapter
)
177 struct ixgbe_hw
*hw
= &adapter
->hw
;
181 hw
->bus
.type
= ixgbe_bus_type_pci_express
;
183 /* Get the negotiated link width and speed from PCI config space of the
184 * parent, as this device is behind a switch
186 err
= ixgbe_read_pci_cfg_word_parent(adapter
, 18, &link_status
);
188 /* assume caller will handle error case */
192 hw
->bus
.width
= ixgbe_convert_bus_width(link_status
);
193 hw
->bus
.speed
= ixgbe_convert_bus_speed(link_status
);
198 static void ixgbe_service_event_schedule(struct ixgbe_adapter
*adapter
)
200 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
) &&
201 !test_and_set_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
))
202 schedule_work(&adapter
->service_task
);
205 static void ixgbe_service_event_complete(struct ixgbe_adapter
*adapter
)
207 BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
));
209 /* flush memory to make sure state is correct before next watchdog */
210 smp_mb__before_clear_bit();
211 clear_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
);
214 struct ixgbe_reg_info
{
219 static const struct ixgbe_reg_info ixgbe_reg_info_tbl
[] = {
221 /* General Registers */
222 {IXGBE_CTRL
, "CTRL"},
223 {IXGBE_STATUS
, "STATUS"},
224 {IXGBE_CTRL_EXT
, "CTRL_EXT"},
226 /* Interrupt Registers */
227 {IXGBE_EICR
, "EICR"},
230 {IXGBE_SRRCTL(0), "SRRCTL"},
231 {IXGBE_DCA_RXCTRL(0), "DRXCTL"},
232 {IXGBE_RDLEN(0), "RDLEN"},
233 {IXGBE_RDH(0), "RDH"},
234 {IXGBE_RDT(0), "RDT"},
235 {IXGBE_RXDCTL(0), "RXDCTL"},
236 {IXGBE_RDBAL(0), "RDBAL"},
237 {IXGBE_RDBAH(0), "RDBAH"},
240 {IXGBE_TDBAL(0), "TDBAL"},
241 {IXGBE_TDBAH(0), "TDBAH"},
242 {IXGBE_TDLEN(0), "TDLEN"},
243 {IXGBE_TDH(0), "TDH"},
244 {IXGBE_TDT(0), "TDT"},
245 {IXGBE_TXDCTL(0), "TXDCTL"},
247 /* List Terminator */
253 * ixgbe_regdump - register printout routine
255 static void ixgbe_regdump(struct ixgbe_hw
*hw
, struct ixgbe_reg_info
*reginfo
)
261 switch (reginfo
->ofs
) {
262 case IXGBE_SRRCTL(0):
263 for (i
= 0; i
< 64; i
++)
264 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_SRRCTL(i
));
266 case IXGBE_DCA_RXCTRL(0):
267 for (i
= 0; i
< 64; i
++)
268 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_DCA_RXCTRL(i
));
271 for (i
= 0; i
< 64; i
++)
272 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDLEN(i
));
275 for (i
= 0; i
< 64; i
++)
276 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDH(i
));
279 for (i
= 0; i
< 64; i
++)
280 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDT(i
));
282 case IXGBE_RXDCTL(0):
283 for (i
= 0; i
< 64; i
++)
284 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RXDCTL(i
));
287 for (i
= 0; i
< 64; i
++)
288 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDBAL(i
));
291 for (i
= 0; i
< 64; i
++)
292 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDBAH(i
));
295 for (i
= 0; i
< 64; i
++)
296 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDBAL(i
));
299 for (i
= 0; i
< 64; i
++)
300 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDBAH(i
));
303 for (i
= 0; i
< 64; i
++)
304 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDLEN(i
));
307 for (i
= 0; i
< 64; i
++)
308 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDH(i
));
311 for (i
= 0; i
< 64; i
++)
312 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDT(i
));
314 case IXGBE_TXDCTL(0):
315 for (i
= 0; i
< 64; i
++)
316 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TXDCTL(i
));
319 pr_info("%-15s %08x\n", reginfo
->name
,
320 IXGBE_READ_REG(hw
, reginfo
->ofs
));
324 for (i
= 0; i
< 8; i
++) {
325 snprintf(rname
, 16, "%s[%d-%d]", reginfo
->name
, i
*8, i
*8+7);
326 pr_err("%-15s", rname
);
327 for (j
= 0; j
< 8; j
++)
328 pr_cont(" %08x", regs
[i
*8+j
]);
335 * ixgbe_dump - Print registers, tx-rings and rx-rings
337 static void ixgbe_dump(struct ixgbe_adapter
*adapter
)
339 struct net_device
*netdev
= adapter
->netdev
;
340 struct ixgbe_hw
*hw
= &adapter
->hw
;
341 struct ixgbe_reg_info
*reginfo
;
343 struct ixgbe_ring
*tx_ring
;
344 struct ixgbe_tx_buffer
*tx_buffer
;
345 union ixgbe_adv_tx_desc
*tx_desc
;
346 struct my_u0
{ u64 a
; u64 b
; } *u0
;
347 struct ixgbe_ring
*rx_ring
;
348 union ixgbe_adv_rx_desc
*rx_desc
;
349 struct ixgbe_rx_buffer
*rx_buffer_info
;
353 if (!netif_msg_hw(adapter
))
356 /* Print netdevice Info */
358 dev_info(&adapter
->pdev
->dev
, "Net device Info\n");
359 pr_info("Device Name state "
360 "trans_start last_rx\n");
361 pr_info("%-15s %016lX %016lX %016lX\n",
368 /* Print Registers */
369 dev_info(&adapter
->pdev
->dev
, "Register Dump\n");
370 pr_info(" Register Name Value\n");
371 for (reginfo
= (struct ixgbe_reg_info
*)ixgbe_reg_info_tbl
;
372 reginfo
->name
; reginfo
++) {
373 ixgbe_regdump(hw
, reginfo
);
376 /* Print TX Ring Summary */
377 if (!netdev
|| !netif_running(netdev
))
380 dev_info(&adapter
->pdev
->dev
, "TX Rings Summary\n");
381 pr_info(" %s %s %s %s\n",
382 "Queue [NTU] [NTC] [bi(ntc)->dma ]",
383 "leng", "ntw", "timestamp");
384 for (n
= 0; n
< adapter
->num_tx_queues
; n
++) {
385 tx_ring
= adapter
->tx_ring
[n
];
386 tx_buffer
= &tx_ring
->tx_buffer_info
[tx_ring
->next_to_clean
];
387 pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
388 n
, tx_ring
->next_to_use
, tx_ring
->next_to_clean
,
389 (u64
)dma_unmap_addr(tx_buffer
, dma
),
390 dma_unmap_len(tx_buffer
, len
),
391 tx_buffer
->next_to_watch
,
392 (u64
)tx_buffer
->time_stamp
);
396 if (!netif_msg_tx_done(adapter
))
397 goto rx_ring_summary
;
399 dev_info(&adapter
->pdev
->dev
, "TX Rings Dump\n");
401 /* Transmit Descriptor Formats
403 * 82598 Advanced Transmit Descriptor
404 * +--------------------------------------------------------------+
405 * 0 | Buffer Address [63:0] |
406 * +--------------------------------------------------------------+
407 * 8 | PAYLEN | POPTS | IDX | STA | DCMD |DTYP | RSV | DTALEN |
408 * +--------------------------------------------------------------+
409 * 63 46 45 40 39 36 35 32 31 24 23 20 19 0
411 * 82598 Advanced Transmit Descriptor (Write-Back Format)
412 * +--------------------------------------------------------------+
414 * +--------------------------------------------------------------+
415 * 8 | RSV | STA | NXTSEQ |
416 * +--------------------------------------------------------------+
419 * 82599+ Advanced Transmit Descriptor
420 * +--------------------------------------------------------------+
421 * 0 | Buffer Address [63:0] |
422 * +--------------------------------------------------------------+
423 * 8 |PAYLEN |POPTS|CC|IDX |STA |DCMD |DTYP |MAC |RSV |DTALEN |
424 * +--------------------------------------------------------------+
425 * 63 46 45 40 39 38 36 35 32 31 24 23 20 19 18 17 16 15 0
427 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
428 * +--------------------------------------------------------------+
430 * +--------------------------------------------------------------+
431 * 8 | RSV | STA | RSV |
432 * +--------------------------------------------------------------+
436 for (n
= 0; n
< adapter
->num_tx_queues
; n
++) {
437 tx_ring
= adapter
->tx_ring
[n
];
438 pr_info("------------------------------------\n");
439 pr_info("TX QUEUE INDEX = %d\n", tx_ring
->queue_index
);
440 pr_info("------------------------------------\n");
441 pr_info("%s%s %s %s %s %s\n",
442 "T [desc] [address 63:0 ] ",
443 "[PlPOIdStDDt Ln] [bi->dma ] ",
444 "leng", "ntw", "timestamp", "bi->skb");
446 for (i
= 0; tx_ring
->desc
&& (i
< tx_ring
->count
); i
++) {
447 tx_desc
= IXGBE_TX_DESC(tx_ring
, i
);
448 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
449 u0
= (struct my_u0
*)tx_desc
;
450 if (dma_unmap_len(tx_buffer
, len
) > 0) {
451 pr_info("T [0x%03X] %016llX %016llX %016llX %08X %p %016llX %p",
455 (u64
)dma_unmap_addr(tx_buffer
, dma
),
456 dma_unmap_len(tx_buffer
, len
),
457 tx_buffer
->next_to_watch
,
458 (u64
)tx_buffer
->time_stamp
,
460 if (i
== tx_ring
->next_to_use
&&
461 i
== tx_ring
->next_to_clean
)
463 else if (i
== tx_ring
->next_to_use
)
465 else if (i
== tx_ring
->next_to_clean
)
470 if (netif_msg_pktdata(adapter
) &&
472 print_hex_dump(KERN_INFO
, "",
473 DUMP_PREFIX_ADDRESS
, 16, 1,
474 tx_buffer
->skb
->data
,
475 dma_unmap_len(tx_buffer
, len
),
481 /* Print RX Rings Summary */
483 dev_info(&adapter
->pdev
->dev
, "RX Rings Summary\n");
484 pr_info("Queue [NTU] [NTC]\n");
485 for (n
= 0; n
< adapter
->num_rx_queues
; n
++) {
486 rx_ring
= adapter
->rx_ring
[n
];
487 pr_info("%5d %5X %5X\n",
488 n
, rx_ring
->next_to_use
, rx_ring
->next_to_clean
);
492 if (!netif_msg_rx_status(adapter
))
495 dev_info(&adapter
->pdev
->dev
, "RX Rings Dump\n");
497 /* Receive Descriptor Formats
499 * 82598 Advanced Receive Descriptor (Read) Format
501 * +-----------------------------------------------------+
502 * 0 | Packet Buffer Address [63:1] |A0/NSE|
503 * +----------------------------------------------+------+
504 * 8 | Header Buffer Address [63:1] | DD |
505 * +-----------------------------------------------------+
508 * 82598 Advanced Receive Descriptor (Write-Back) Format
510 * 63 48 47 32 31 30 21 20 16 15 4 3 0
511 * +------------------------------------------------------+
512 * 0 | RSS Hash / |SPH| HDR_LEN | RSV |Packet| RSS |
513 * | Packet | IP | | | | Type | Type |
514 * | Checksum | Ident | | | | | |
515 * +------------------------------------------------------+
516 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
517 * +------------------------------------------------------+
518 * 63 48 47 32 31 20 19 0
520 * 82599+ Advanced Receive Descriptor (Read) Format
522 * +-----------------------------------------------------+
523 * 0 | Packet Buffer Address [63:1] |A0/NSE|
524 * +----------------------------------------------+------+
525 * 8 | Header Buffer Address [63:1] | DD |
526 * +-----------------------------------------------------+
529 * 82599+ Advanced Receive Descriptor (Write-Back) Format
531 * 63 48 47 32 31 30 21 20 17 16 4 3 0
532 * +------------------------------------------------------+
533 * 0 |RSS / Frag Checksum|SPH| HDR_LEN |RSC- |Packet| RSS |
534 * |/ RTT / PCoE_PARAM | | | CNT | Type | Type |
535 * |/ Flow Dir Flt ID | | | | | |
536 * +------------------------------------------------------+
537 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
538 * +------------------------------------------------------+
539 * 63 48 47 32 31 20 19 0
542 for (n
= 0; n
< adapter
->num_rx_queues
; n
++) {
543 rx_ring
= adapter
->rx_ring
[n
];
544 pr_info("------------------------------------\n");
545 pr_info("RX QUEUE INDEX = %d\n", rx_ring
->queue_index
);
546 pr_info("------------------------------------\n");
548 "R [desc] [ PktBuf A0] ",
549 "[ HeadBuf DD] [bi->dma ] [bi->skb ] ",
550 "<-- Adv Rx Read format\n");
552 "RWB[desc] [PcsmIpSHl PtRs] ",
553 "[vl er S cks ln] ---------------- [bi->skb ] ",
554 "<-- Adv Rx Write-Back format\n");
556 for (i
= 0; i
< rx_ring
->count
; i
++) {
557 rx_buffer_info
= &rx_ring
->rx_buffer_info
[i
];
558 rx_desc
= IXGBE_RX_DESC(rx_ring
, i
);
559 u0
= (struct my_u0
*)rx_desc
;
560 staterr
= le32_to_cpu(rx_desc
->wb
.upper
.status_error
);
561 if (staterr
& IXGBE_RXD_STAT_DD
) {
562 /* Descriptor Done */
563 pr_info("RWB[0x%03X] %016llX "
564 "%016llX ---------------- %p", i
,
567 rx_buffer_info
->skb
);
569 pr_info("R [0x%03X] %016llX "
570 "%016llX %016llX %p", i
,
573 (u64
)rx_buffer_info
->dma
,
574 rx_buffer_info
->skb
);
576 if (netif_msg_pktdata(adapter
) &&
577 rx_buffer_info
->dma
) {
578 print_hex_dump(KERN_INFO
, "",
579 DUMP_PREFIX_ADDRESS
, 16, 1,
580 page_address(rx_buffer_info
->page
) +
581 rx_buffer_info
->page_offset
,
582 ixgbe_rx_bufsz(rx_ring
), true);
586 if (i
== rx_ring
->next_to_use
)
588 else if (i
== rx_ring
->next_to_clean
)
600 static void ixgbe_release_hw_control(struct ixgbe_adapter
*adapter
)
604 /* Let firmware take over control of h/w */
605 ctrl_ext
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_CTRL_EXT
);
606 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_CTRL_EXT
,
607 ctrl_ext
& ~IXGBE_CTRL_EXT_DRV_LOAD
);
610 static void ixgbe_get_hw_control(struct ixgbe_adapter
*adapter
)
614 /* Let firmware know the driver has taken over */
615 ctrl_ext
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_CTRL_EXT
);
616 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_CTRL_EXT
,
617 ctrl_ext
| IXGBE_CTRL_EXT_DRV_LOAD
);
621 * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
622 * @adapter: pointer to adapter struct
623 * @direction: 0 for Rx, 1 for Tx, -1 for other causes
624 * @queue: queue to map the corresponding interrupt to
625 * @msix_vector: the vector to map to the corresponding queue
628 static void ixgbe_set_ivar(struct ixgbe_adapter
*adapter
, s8 direction
,
629 u8 queue
, u8 msix_vector
)
632 struct ixgbe_hw
*hw
= &adapter
->hw
;
633 switch (hw
->mac
.type
) {
634 case ixgbe_mac_82598EB
:
635 msix_vector
|= IXGBE_IVAR_ALLOC_VAL
;
638 index
= (((direction
* 64) + queue
) >> 2) & 0x1F;
639 ivar
= IXGBE_READ_REG(hw
, IXGBE_IVAR(index
));
640 ivar
&= ~(0xFF << (8 * (queue
& 0x3)));
641 ivar
|= (msix_vector
<< (8 * (queue
& 0x3)));
642 IXGBE_WRITE_REG(hw
, IXGBE_IVAR(index
), ivar
);
644 case ixgbe_mac_82599EB
:
646 if (direction
== -1) {
648 msix_vector
|= IXGBE_IVAR_ALLOC_VAL
;
649 index
= ((queue
& 1) * 8);
650 ivar
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_IVAR_MISC
);
651 ivar
&= ~(0xFF << index
);
652 ivar
|= (msix_vector
<< index
);
653 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_IVAR_MISC
, ivar
);
656 /* tx or rx causes */
657 msix_vector
|= IXGBE_IVAR_ALLOC_VAL
;
658 index
= ((16 * (queue
& 1)) + (8 * direction
));
659 ivar
= IXGBE_READ_REG(hw
, IXGBE_IVAR(queue
>> 1));
660 ivar
&= ~(0xFF << index
);
661 ivar
|= (msix_vector
<< index
);
662 IXGBE_WRITE_REG(hw
, IXGBE_IVAR(queue
>> 1), ivar
);
670 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter
*adapter
,
675 switch (adapter
->hw
.mac
.type
) {
676 case ixgbe_mac_82598EB
:
677 mask
= (IXGBE_EIMS_RTX_QUEUE
& qmask
);
678 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EICS
, mask
);
680 case ixgbe_mac_82599EB
:
682 mask
= (qmask
& 0xFFFFFFFF);
683 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EICS_EX(0), mask
);
684 mask
= (qmask
>> 32);
685 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EICS_EX(1), mask
);
692 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring
*ring
,
693 struct ixgbe_tx_buffer
*tx_buffer
)
695 if (tx_buffer
->skb
) {
696 dev_kfree_skb_any(tx_buffer
->skb
);
697 if (dma_unmap_len(tx_buffer
, len
))
698 dma_unmap_single(ring
->dev
,
699 dma_unmap_addr(tx_buffer
, dma
),
700 dma_unmap_len(tx_buffer
, len
),
702 } else if (dma_unmap_len(tx_buffer
, len
)) {
703 dma_unmap_page(ring
->dev
,
704 dma_unmap_addr(tx_buffer
, dma
),
705 dma_unmap_len(tx_buffer
, len
),
708 tx_buffer
->next_to_watch
= NULL
;
709 tx_buffer
->skb
= NULL
;
710 dma_unmap_len_set(tx_buffer
, len
, 0);
711 /* tx_buffer must be completely set up in the transmit path */
714 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter
*adapter
)
716 struct ixgbe_hw
*hw
= &adapter
->hw
;
717 struct ixgbe_hw_stats
*hwstats
= &adapter
->stats
;
721 if ((hw
->fc
.current_mode
!= ixgbe_fc_full
) &&
722 (hw
->fc
.current_mode
!= ixgbe_fc_rx_pause
))
725 switch (hw
->mac
.type
) {
726 case ixgbe_mac_82598EB
:
727 data
= IXGBE_READ_REG(hw
, IXGBE_LXOFFRXC
);
730 data
= IXGBE_READ_REG(hw
, IXGBE_LXOFFRXCNT
);
732 hwstats
->lxoffrxc
+= data
;
734 /* refill credits (no tx hang) if we received xoff */
738 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
739 clear_bit(__IXGBE_HANG_CHECK_ARMED
,
740 &adapter
->tx_ring
[i
]->state
);
743 static void ixgbe_update_xoff_received(struct ixgbe_adapter
*adapter
)
745 struct ixgbe_hw
*hw
= &adapter
->hw
;
746 struct ixgbe_hw_stats
*hwstats
= &adapter
->stats
;
750 bool pfc_en
= adapter
->dcb_cfg
.pfc_mode_enable
;
752 if (adapter
->ixgbe_ieee_pfc
)
753 pfc_en
|= !!(adapter
->ixgbe_ieee_pfc
->pfc_en
);
755 if (!(adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
) || !pfc_en
) {
756 ixgbe_update_xoff_rx_lfc(adapter
);
760 /* update stats for each tc, only valid with PFC enabled */
761 for (i
= 0; i
< MAX_TX_PACKET_BUFFERS
; i
++) {
764 switch (hw
->mac
.type
) {
765 case ixgbe_mac_82598EB
:
766 pxoffrxc
= IXGBE_READ_REG(hw
, IXGBE_PXOFFRXC(i
));
769 pxoffrxc
= IXGBE_READ_REG(hw
, IXGBE_PXOFFRXCNT(i
));
771 hwstats
->pxoffrxc
[i
] += pxoffrxc
;
772 /* Get the TC for given UP */
773 tc
= netdev_get_prio_tc_map(adapter
->netdev
, i
);
774 xoff
[tc
] += pxoffrxc
;
777 /* disarm tx queues that have received xoff frames */
778 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
779 struct ixgbe_ring
*tx_ring
= adapter
->tx_ring
[i
];
781 tc
= tx_ring
->dcb_tc
;
783 clear_bit(__IXGBE_HANG_CHECK_ARMED
, &tx_ring
->state
);
787 static u64
ixgbe_get_tx_completed(struct ixgbe_ring
*ring
)
789 return ring
->stats
.packets
;
792 static u64
ixgbe_get_tx_pending(struct ixgbe_ring
*ring
)
794 struct ixgbe_adapter
*adapter
= netdev_priv(ring
->netdev
);
795 struct ixgbe_hw
*hw
= &adapter
->hw
;
797 u32 head
= IXGBE_READ_REG(hw
, IXGBE_TDH(ring
->reg_idx
));
798 u32 tail
= IXGBE_READ_REG(hw
, IXGBE_TDT(ring
->reg_idx
));
801 return (head
< tail
) ?
802 tail
- head
: (tail
+ ring
->count
- head
);
807 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring
*tx_ring
)
809 u32 tx_done
= ixgbe_get_tx_completed(tx_ring
);
810 u32 tx_done_old
= tx_ring
->tx_stats
.tx_done_old
;
811 u32 tx_pending
= ixgbe_get_tx_pending(tx_ring
);
814 clear_check_for_tx_hang(tx_ring
);
817 * Check for a hung queue, but be thorough. This verifies
818 * that a transmit has been completed since the previous
819 * check AND there is at least one packet pending. The
820 * ARMED bit is set to indicate a potential hang. The
821 * bit is cleared if a pause frame is received to remove
822 * false hang detection due to PFC or 802.3x frames. By
823 * requiring this to fail twice we avoid races with
824 * pfc clearing the ARMED bit and conditions where we
825 * run the check_tx_hang logic with a transmit completion
826 * pending but without time to complete it yet.
828 if ((tx_done_old
== tx_done
) && tx_pending
) {
829 /* make sure it is true for two checks in a row */
830 ret
= test_and_set_bit(__IXGBE_HANG_CHECK_ARMED
,
833 /* update completed stats and continue */
834 tx_ring
->tx_stats
.tx_done_old
= tx_done
;
835 /* reset the countdown */
836 clear_bit(__IXGBE_HANG_CHECK_ARMED
, &tx_ring
->state
);
843 * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
844 * @adapter: driver private struct
846 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter
*adapter
)
849 /* Do the reset outside of interrupt context */
850 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
851 adapter
->flags2
|= IXGBE_FLAG2_RESET_REQUESTED
;
852 e_warn(drv
, "initiating reset due to tx timeout\n");
853 ixgbe_service_event_schedule(adapter
);
858 * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
859 * @q_vector: structure containing interrupt and ring information
860 * @tx_ring: tx ring to clean
862 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector
*q_vector
,
863 struct ixgbe_ring
*tx_ring
)
865 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
866 struct ixgbe_tx_buffer
*tx_buffer
;
867 union ixgbe_adv_tx_desc
*tx_desc
;
868 unsigned int total_bytes
= 0, total_packets
= 0;
869 unsigned int budget
= q_vector
->tx
.work_limit
;
870 unsigned int i
= tx_ring
->next_to_clean
;
872 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
875 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
876 tx_desc
= IXGBE_TX_DESC(tx_ring
, i
);
880 union ixgbe_adv_tx_desc
*eop_desc
= tx_buffer
->next_to_watch
;
882 /* if next_to_watch is not set then there is no work pending */
886 /* prevent any other reads prior to eop_desc */
887 read_barrier_depends();
889 /* if DD is not set pending work has not been completed */
890 if (!(eop_desc
->wb
.status
& cpu_to_le32(IXGBE_TXD_STAT_DD
)))
893 /* clear next_to_watch to prevent false hangs */
894 tx_buffer
->next_to_watch
= NULL
;
896 /* update the statistics for this packet */
897 total_bytes
+= tx_buffer
->bytecount
;
898 total_packets
+= tx_buffer
->gso_segs
;
901 dev_kfree_skb_any(tx_buffer
->skb
);
903 /* unmap skb header data */
904 dma_unmap_single(tx_ring
->dev
,
905 dma_unmap_addr(tx_buffer
, dma
),
906 dma_unmap_len(tx_buffer
, len
),
909 /* clear tx_buffer data */
910 tx_buffer
->skb
= NULL
;
911 dma_unmap_len_set(tx_buffer
, len
, 0);
913 /* unmap remaining buffers */
914 while (tx_desc
!= eop_desc
) {
920 tx_buffer
= tx_ring
->tx_buffer_info
;
921 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
924 /* unmap any remaining paged data */
925 if (dma_unmap_len(tx_buffer
, len
)) {
926 dma_unmap_page(tx_ring
->dev
,
927 dma_unmap_addr(tx_buffer
, dma
),
928 dma_unmap_len(tx_buffer
, len
),
930 dma_unmap_len_set(tx_buffer
, len
, 0);
934 /* move us one more past the eop_desc for start of next pkt */
940 tx_buffer
= tx_ring
->tx_buffer_info
;
941 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
944 /* issue prefetch for next Tx descriptor */
947 /* update budget accounting */
949 } while (likely(budget
));
952 tx_ring
->next_to_clean
= i
;
953 u64_stats_update_begin(&tx_ring
->syncp
);
954 tx_ring
->stats
.bytes
+= total_bytes
;
955 tx_ring
->stats
.packets
+= total_packets
;
956 u64_stats_update_end(&tx_ring
->syncp
);
957 q_vector
->tx
.total_bytes
+= total_bytes
;
958 q_vector
->tx
.total_packets
+= total_packets
;
960 if (check_for_tx_hang(tx_ring
) && ixgbe_check_tx_hang(tx_ring
)) {
961 /* schedule immediate reset if we believe we hung */
962 struct ixgbe_hw
*hw
= &adapter
->hw
;
963 e_err(drv
, "Detected Tx Unit Hang\n"
965 " TDH, TDT <%x>, <%x>\n"
966 " next_to_use <%x>\n"
967 " next_to_clean <%x>\n"
968 "tx_buffer_info[next_to_clean]\n"
969 " time_stamp <%lx>\n"
971 tx_ring
->queue_index
,
972 IXGBE_READ_REG(hw
, IXGBE_TDH(tx_ring
->reg_idx
)),
973 IXGBE_READ_REG(hw
, IXGBE_TDT(tx_ring
->reg_idx
)),
974 tx_ring
->next_to_use
, i
,
975 tx_ring
->tx_buffer_info
[i
].time_stamp
, jiffies
);
977 netif_stop_subqueue(tx_ring
->netdev
, tx_ring
->queue_index
);
980 "tx hang %d detected on queue %d, resetting adapter\n",
981 adapter
->tx_timeout_count
+ 1, tx_ring
->queue_index
);
983 /* schedule immediate reset if we believe we hung */
984 ixgbe_tx_timeout_reset(adapter
);
986 /* the adapter is about to reset, no point in enabling stuff */
990 netdev_tx_completed_queue(txring_txq(tx_ring
),
991 total_packets
, total_bytes
);
993 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
994 if (unlikely(total_packets
&& netif_carrier_ok(tx_ring
->netdev
) &&
995 (ixgbe_desc_unused(tx_ring
) >= TX_WAKE_THRESHOLD
))) {
996 /* Make sure that anybody stopping the queue after this
997 * sees the new next_to_clean.
1000 if (__netif_subqueue_stopped(tx_ring
->netdev
,
1001 tx_ring
->queue_index
)
1002 && !test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
1003 netif_wake_subqueue(tx_ring
->netdev
,
1004 tx_ring
->queue_index
);
1005 ++tx_ring
->tx_stats
.restart_queue
;
1012 #ifdef CONFIG_IXGBE_DCA
1013 static void ixgbe_update_tx_dca(struct ixgbe_adapter
*adapter
,
1014 struct ixgbe_ring
*tx_ring
,
1017 struct ixgbe_hw
*hw
= &adapter
->hw
;
1018 u32 txctrl
= dca3_get_tag(tx_ring
->dev
, cpu
);
1021 switch (hw
->mac
.type
) {
1022 case ixgbe_mac_82598EB
:
1023 reg_offset
= IXGBE_DCA_TXCTRL(tx_ring
->reg_idx
);
1025 case ixgbe_mac_82599EB
:
1026 case ixgbe_mac_X540
:
1027 reg_offset
= IXGBE_DCA_TXCTRL_82599(tx_ring
->reg_idx
);
1028 txctrl
<<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599
;
1031 /* for unknown hardware do not write register */
1036 * We can enable relaxed ordering for reads, but not writes when
1037 * DCA is enabled. This is due to a known issue in some chipsets
1038 * which will cause the DCA tag to be cleared.
1040 txctrl
|= IXGBE_DCA_TXCTRL_DESC_RRO_EN
|
1041 IXGBE_DCA_TXCTRL_DATA_RRO_EN
|
1042 IXGBE_DCA_TXCTRL_DESC_DCA_EN
;
1044 IXGBE_WRITE_REG(hw
, reg_offset
, txctrl
);
1047 static void ixgbe_update_rx_dca(struct ixgbe_adapter
*adapter
,
1048 struct ixgbe_ring
*rx_ring
,
1051 struct ixgbe_hw
*hw
= &adapter
->hw
;
1052 u32 rxctrl
= dca3_get_tag(rx_ring
->dev
, cpu
);
1053 u8 reg_idx
= rx_ring
->reg_idx
;
1056 switch (hw
->mac
.type
) {
1057 case ixgbe_mac_82599EB
:
1058 case ixgbe_mac_X540
:
1059 rxctrl
<<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599
;
1066 * We can enable relaxed ordering for reads, but not writes when
1067 * DCA is enabled. This is due to a known issue in some chipsets
1068 * which will cause the DCA tag to be cleared.
1070 rxctrl
|= IXGBE_DCA_RXCTRL_DESC_RRO_EN
|
1071 IXGBE_DCA_RXCTRL_DESC_DCA_EN
;
1073 IXGBE_WRITE_REG(hw
, IXGBE_DCA_RXCTRL(reg_idx
), rxctrl
);
1076 static void ixgbe_update_dca(struct ixgbe_q_vector
*q_vector
)
1078 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
1079 struct ixgbe_ring
*ring
;
1080 int cpu
= get_cpu();
1082 if (q_vector
->cpu
== cpu
)
1085 ixgbe_for_each_ring(ring
, q_vector
->tx
)
1086 ixgbe_update_tx_dca(adapter
, ring
, cpu
);
1088 ixgbe_for_each_ring(ring
, q_vector
->rx
)
1089 ixgbe_update_rx_dca(adapter
, ring
, cpu
);
1091 q_vector
->cpu
= cpu
;
1096 static void ixgbe_setup_dca(struct ixgbe_adapter
*adapter
)
1100 if (!(adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
))
1103 /* always use CB2 mode, difference is masked in the CB driver */
1104 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_DCA_CTRL
, 2);
1106 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
1107 adapter
->q_vector
[i
]->cpu
= -1;
1108 ixgbe_update_dca(adapter
->q_vector
[i
]);
1112 static int __ixgbe_notify_dca(struct device
*dev
, void *data
)
1114 struct ixgbe_adapter
*adapter
= dev_get_drvdata(dev
);
1115 unsigned long event
= *(unsigned long *)data
;
1117 if (!(adapter
->flags
& IXGBE_FLAG_DCA_CAPABLE
))
1121 case DCA_PROVIDER_ADD
:
1122 /* if we're already enabled, don't do it again */
1123 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
)
1125 if (dca_add_requester(dev
) == 0) {
1126 adapter
->flags
|= IXGBE_FLAG_DCA_ENABLED
;
1127 ixgbe_setup_dca(adapter
);
1130 /* Fall Through since DCA is disabled. */
1131 case DCA_PROVIDER_REMOVE
:
1132 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
) {
1133 dca_remove_requester(dev
);
1134 adapter
->flags
&= ~IXGBE_FLAG_DCA_ENABLED
;
1135 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_DCA_CTRL
, 1);
1143 #endif /* CONFIG_IXGBE_DCA */
1144 static inline void ixgbe_rx_hash(struct ixgbe_ring
*ring
,
1145 union ixgbe_adv_rx_desc
*rx_desc
,
1146 struct sk_buff
*skb
)
1148 if (ring
->netdev
->features
& NETIF_F_RXHASH
)
1149 skb
->rxhash
= le32_to_cpu(rx_desc
->wb
.lower
.hi_dword
.rss
);
1154 * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1155 * @ring: structure containing ring specific data
1156 * @rx_desc: advanced rx descriptor
1158 * Returns : true if it is FCoE pkt
1160 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring
*ring
,
1161 union ixgbe_adv_rx_desc
*rx_desc
)
1163 __le16 pkt_info
= rx_desc
->wb
.lower
.lo_dword
.hs_rss
.pkt_info
;
1165 return test_bit(__IXGBE_RX_FCOE
, &ring
->state
) &&
1166 ((pkt_info
& cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK
)) ==
1167 (cpu_to_le16(IXGBE_ETQF_FILTER_FCOE
<<
1168 IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT
)));
1171 #endif /* IXGBE_FCOE */
1173 * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1174 * @ring: structure containing ring specific data
1175 * @rx_desc: current Rx descriptor being processed
1176 * @skb: skb currently being received and modified
1178 static inline void ixgbe_rx_checksum(struct ixgbe_ring
*ring
,
1179 union ixgbe_adv_rx_desc
*rx_desc
,
1180 struct sk_buff
*skb
)
1182 skb_checksum_none_assert(skb
);
1184 /* Rx csum disabled */
1185 if (!(ring
->netdev
->features
& NETIF_F_RXCSUM
))
1188 /* if IP and error */
1189 if (ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_IPCS
) &&
1190 ixgbe_test_staterr(rx_desc
, IXGBE_RXDADV_ERR_IPE
)) {
1191 ring
->rx_stats
.csum_err
++;
1195 if (!ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_L4CS
))
1198 if (ixgbe_test_staterr(rx_desc
, IXGBE_RXDADV_ERR_TCPE
)) {
1199 __le16 pkt_info
= rx_desc
->wb
.lower
.lo_dword
.hs_rss
.pkt_info
;
1202 * 82599 errata, UDP frames with a 0 checksum can be marked as
1205 if ((pkt_info
& cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP
)) &&
1206 test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR
, &ring
->state
))
1209 ring
->rx_stats
.csum_err
++;
1213 /* It must be a TCP or UDP packet with a valid checksum */
1214 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1217 static inline void ixgbe_release_rx_desc(struct ixgbe_ring
*rx_ring
, u32 val
)
1219 rx_ring
->next_to_use
= val
;
1221 /* update next to alloc since we have filled the ring */
1222 rx_ring
->next_to_alloc
= val
;
1224 * Force memory writes to complete before letting h/w
1225 * know there are new descriptors to fetch. (Only
1226 * applicable for weak-ordered memory model archs,
1230 writel(val
, rx_ring
->tail
);
1233 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring
*rx_ring
,
1234 struct ixgbe_rx_buffer
*bi
)
1236 struct page
*page
= bi
->page
;
1237 dma_addr_t dma
= bi
->dma
;
1239 /* since we are recycling buffers we should seldom need to alloc */
1243 /* alloc new page for storage */
1244 if (likely(!page
)) {
1245 page
= __skb_alloc_pages(GFP_ATOMIC
| __GFP_COLD
| __GFP_COMP
,
1246 bi
->skb
, ixgbe_rx_pg_order(rx_ring
));
1247 if (unlikely(!page
)) {
1248 rx_ring
->rx_stats
.alloc_rx_page_failed
++;
1254 /* map page for use */
1255 dma
= dma_map_page(rx_ring
->dev
, page
, 0,
1256 ixgbe_rx_pg_size(rx_ring
), DMA_FROM_DEVICE
);
1259 * if mapping failed free memory back to system since
1260 * there isn't much point in holding memory we can't use
1262 if (dma_mapping_error(rx_ring
->dev
, dma
)) {
1263 __free_pages(page
, ixgbe_rx_pg_order(rx_ring
));
1266 rx_ring
->rx_stats
.alloc_rx_page_failed
++;
1271 bi
->page_offset
= 0;
1277 * ixgbe_alloc_rx_buffers - Replace used receive buffers
1278 * @rx_ring: ring to place buffers on
1279 * @cleaned_count: number of buffers to replace
1281 void ixgbe_alloc_rx_buffers(struct ixgbe_ring
*rx_ring
, u16 cleaned_count
)
1283 union ixgbe_adv_rx_desc
*rx_desc
;
1284 struct ixgbe_rx_buffer
*bi
;
1285 u16 i
= rx_ring
->next_to_use
;
1291 rx_desc
= IXGBE_RX_DESC(rx_ring
, i
);
1292 bi
= &rx_ring
->rx_buffer_info
[i
];
1293 i
-= rx_ring
->count
;
1296 if (!ixgbe_alloc_mapped_page(rx_ring
, bi
))
1300 * Refresh the desc even if buffer_addrs didn't change
1301 * because each write-back erases this info.
1303 rx_desc
->read
.pkt_addr
= cpu_to_le64(bi
->dma
+ bi
->page_offset
);
1309 rx_desc
= IXGBE_RX_DESC(rx_ring
, 0);
1310 bi
= rx_ring
->rx_buffer_info
;
1311 i
-= rx_ring
->count
;
1314 /* clear the hdr_addr for the next_to_use descriptor */
1315 rx_desc
->read
.hdr_addr
= 0;
1318 } while (cleaned_count
);
1320 i
+= rx_ring
->count
;
1322 if (rx_ring
->next_to_use
!= i
)
1323 ixgbe_release_rx_desc(rx_ring
, i
);
1327 * ixgbe_get_headlen - determine size of header for RSC/LRO/GRO/FCOE
1328 * @data: pointer to the start of the headers
1329 * @max_len: total length of section to find headers in
1331 * This function is meant to determine the length of headers that will
1332 * be recognized by hardware for LRO, GRO, and RSC offloads. The main
1333 * motivation of doing this is to only perform one pull for IPv4 TCP
1334 * packets so that we can do basic things like calculating the gso_size
1335 * based on the average data per packet.
1337 static unsigned int ixgbe_get_headlen(unsigned char *data
,
1338 unsigned int max_len
)
1341 unsigned char *network
;
1344 struct vlan_hdr
*vlan
;
1347 struct ipv6hdr
*ipv6
;
1350 u8 nexthdr
= 0; /* default to not TCP */
1353 /* this should never happen, but better safe than sorry */
1354 if (max_len
< ETH_HLEN
)
1357 /* initialize network frame pointer */
1360 /* set first protocol and move network header forward */
1361 protocol
= hdr
.eth
->h_proto
;
1362 hdr
.network
+= ETH_HLEN
;
1364 /* handle any vlan tag if present */
1365 if (protocol
== __constant_htons(ETH_P_8021Q
)) {
1366 if ((hdr
.network
- data
) > (max_len
- VLAN_HLEN
))
1369 protocol
= hdr
.vlan
->h_vlan_encapsulated_proto
;
1370 hdr
.network
+= VLAN_HLEN
;
1373 /* handle L3 protocols */
1374 if (protocol
== __constant_htons(ETH_P_IP
)) {
1375 if ((hdr
.network
- data
) > (max_len
- sizeof(struct iphdr
)))
1378 /* access ihl as a u8 to avoid unaligned access on ia64 */
1379 hlen
= (hdr
.network
[0] & 0x0F) << 2;
1381 /* verify hlen meets minimum size requirements */
1382 if (hlen
< sizeof(struct iphdr
))
1383 return hdr
.network
- data
;
1385 /* record next protocol if header is present */
1386 if (!(hdr
.ipv4
->frag_off
& htons(IP_OFFSET
)))
1387 nexthdr
= hdr
.ipv4
->protocol
;
1388 } else if (protocol
== __constant_htons(ETH_P_IPV6
)) {
1389 if ((hdr
.network
- data
) > (max_len
- sizeof(struct ipv6hdr
)))
1392 /* record next protocol */
1393 nexthdr
= hdr
.ipv6
->nexthdr
;
1394 hlen
= sizeof(struct ipv6hdr
);
1396 } else if (protocol
== __constant_htons(ETH_P_FCOE
)) {
1397 if ((hdr
.network
- data
) > (max_len
- FCOE_HEADER_LEN
))
1399 hlen
= FCOE_HEADER_LEN
;
1402 return hdr
.network
- data
;
1405 /* relocate pointer to start of L4 header */
1406 hdr
.network
+= hlen
;
1408 /* finally sort out TCP/UDP */
1409 if (nexthdr
== IPPROTO_TCP
) {
1410 if ((hdr
.network
- data
) > (max_len
- sizeof(struct tcphdr
)))
1413 /* access doff as a u8 to avoid unaligned access on ia64 */
1414 hlen
= (hdr
.network
[12] & 0xF0) >> 2;
1416 /* verify hlen meets minimum size requirements */
1417 if (hlen
< sizeof(struct tcphdr
))
1418 return hdr
.network
- data
;
1420 hdr
.network
+= hlen
;
1421 } else if (nexthdr
== IPPROTO_UDP
) {
1422 if ((hdr
.network
- data
) > (max_len
- sizeof(struct udphdr
)))
1425 hdr
.network
+= sizeof(struct udphdr
);
1429 * If everything has gone correctly hdr.network should be the
1430 * data section of the packet and will be the end of the header.
1431 * If not then it probably represents the end of the last recognized
1434 if ((hdr
.network
- data
) < max_len
)
1435 return hdr
.network
- data
;
1440 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring
*ring
,
1441 struct sk_buff
*skb
)
1443 u16 hdr_len
= skb_headlen(skb
);
1445 /* set gso_size to avoid messing up TCP MSS */
1446 skb_shinfo(skb
)->gso_size
= DIV_ROUND_UP((skb
->len
- hdr_len
),
1447 IXGBE_CB(skb
)->append_cnt
);
1448 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV4
;
1451 static void ixgbe_update_rsc_stats(struct ixgbe_ring
*rx_ring
,
1452 struct sk_buff
*skb
)
1454 /* if append_cnt is 0 then frame is not RSC */
1455 if (!IXGBE_CB(skb
)->append_cnt
)
1458 rx_ring
->rx_stats
.rsc_count
+= IXGBE_CB(skb
)->append_cnt
;
1459 rx_ring
->rx_stats
.rsc_flush
++;
1461 ixgbe_set_rsc_gso_size(rx_ring
, skb
);
1463 /* gso_size is computed using append_cnt so always clear it last */
1464 IXGBE_CB(skb
)->append_cnt
= 0;
1468 * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1469 * @rx_ring: rx descriptor ring packet is being transacted on
1470 * @rx_desc: pointer to the EOP Rx descriptor
1471 * @skb: pointer to current skb being populated
1473 * This function checks the ring, descriptor, and packet information in
1474 * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1475 * other fields within the skb.
1477 static void ixgbe_process_skb_fields(struct ixgbe_ring
*rx_ring
,
1478 union ixgbe_adv_rx_desc
*rx_desc
,
1479 struct sk_buff
*skb
)
1481 struct net_device
*dev
= rx_ring
->netdev
;
1483 ixgbe_update_rsc_stats(rx_ring
, skb
);
1485 ixgbe_rx_hash(rx_ring
, rx_desc
, skb
);
1487 ixgbe_rx_checksum(rx_ring
, rx_desc
, skb
);
1489 ixgbe_ptp_rx_hwtstamp(rx_ring
, rx_desc
, skb
);
1491 if ((dev
->features
& NETIF_F_HW_VLAN_CTAG_RX
) &&
1492 ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_VP
)) {
1493 u16 vid
= le16_to_cpu(rx_desc
->wb
.upper
.vlan
);
1494 __vlan_hwaccel_put_tag(skb
, vid
);
1497 skb_record_rx_queue(skb
, rx_ring
->queue_index
);
1499 skb
->protocol
= eth_type_trans(skb
, dev
);
1502 static void ixgbe_rx_skb(struct ixgbe_q_vector
*q_vector
,
1503 struct sk_buff
*skb
)
1505 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
1507 if (!(adapter
->flags
& IXGBE_FLAG_IN_NETPOLL
))
1508 napi_gro_receive(&q_vector
->napi
, skb
);
1514 * ixgbe_is_non_eop - process handling of non-EOP buffers
1515 * @rx_ring: Rx ring being processed
1516 * @rx_desc: Rx descriptor for current buffer
1517 * @skb: Current socket buffer containing buffer in progress
1519 * This function updates next to clean. If the buffer is an EOP buffer
1520 * this function exits returning false, otherwise it will place the
1521 * sk_buff in the next buffer to be chained and return true indicating
1522 * that this is in fact a non-EOP buffer.
1524 static bool ixgbe_is_non_eop(struct ixgbe_ring
*rx_ring
,
1525 union ixgbe_adv_rx_desc
*rx_desc
,
1526 struct sk_buff
*skb
)
1528 u32 ntc
= rx_ring
->next_to_clean
+ 1;
1530 /* fetch, update, and store next to clean */
1531 ntc
= (ntc
< rx_ring
->count
) ? ntc
: 0;
1532 rx_ring
->next_to_clean
= ntc
;
1534 prefetch(IXGBE_RX_DESC(rx_ring
, ntc
));
1536 /* update RSC append count if present */
1537 if (ring_is_rsc_enabled(rx_ring
)) {
1538 __le32 rsc_enabled
= rx_desc
->wb
.lower
.lo_dword
.data
&
1539 cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK
);
1541 if (unlikely(rsc_enabled
)) {
1542 u32 rsc_cnt
= le32_to_cpu(rsc_enabled
);
1544 rsc_cnt
>>= IXGBE_RXDADV_RSCCNT_SHIFT
;
1545 IXGBE_CB(skb
)->append_cnt
+= rsc_cnt
- 1;
1547 /* update ntc based on RSC value */
1548 ntc
= le32_to_cpu(rx_desc
->wb
.upper
.status_error
);
1549 ntc
&= IXGBE_RXDADV_NEXTP_MASK
;
1550 ntc
>>= IXGBE_RXDADV_NEXTP_SHIFT
;
1554 /* if we are the last buffer then there is nothing else to do */
1555 if (likely(ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_EOP
)))
1558 /* place skb in next buffer to be received */
1559 rx_ring
->rx_buffer_info
[ntc
].skb
= skb
;
1560 rx_ring
->rx_stats
.non_eop_descs
++;
1566 * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1567 * @rx_ring: rx descriptor ring packet is being transacted on
1568 * @skb: pointer to current skb being adjusted
1570 * This function is an ixgbe specific version of __pskb_pull_tail. The
1571 * main difference between this version and the original function is that
1572 * this function can make several assumptions about the state of things
1573 * that allow for significant optimizations versus the standard function.
1574 * As a result we can do things like drop a frag and maintain an accurate
1575 * truesize for the skb.
1577 static void ixgbe_pull_tail(struct ixgbe_ring
*rx_ring
,
1578 struct sk_buff
*skb
)
1580 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[0];
1582 unsigned int pull_len
;
1585 * it is valid to use page_address instead of kmap since we are
1586 * working with pages allocated out of the lomem pool per
1587 * alloc_page(GFP_ATOMIC)
1589 va
= skb_frag_address(frag
);
1592 * we need the header to contain the greater of either ETH_HLEN or
1593 * 60 bytes if the skb->len is less than 60 for skb_pad.
1595 pull_len
= ixgbe_get_headlen(va
, IXGBE_RX_HDR_SIZE
);
1597 /* align pull length to size of long to optimize memcpy performance */
1598 skb_copy_to_linear_data(skb
, va
, ALIGN(pull_len
, sizeof(long)));
1600 /* update all of the pointers */
1601 skb_frag_size_sub(frag
, pull_len
);
1602 frag
->page_offset
+= pull_len
;
1603 skb
->data_len
-= pull_len
;
1604 skb
->tail
+= pull_len
;
1608 * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1609 * @rx_ring: rx descriptor ring packet is being transacted on
1610 * @skb: pointer to current skb being updated
1612 * This function provides a basic DMA sync up for the first fragment of an
1613 * skb. The reason for doing this is that the first fragment cannot be
1614 * unmapped until we have reached the end of packet descriptor for a buffer
1617 static void ixgbe_dma_sync_frag(struct ixgbe_ring
*rx_ring
,
1618 struct sk_buff
*skb
)
1620 /* if the page was released unmap it, else just sync our portion */
1621 if (unlikely(IXGBE_CB(skb
)->page_released
)) {
1622 dma_unmap_page(rx_ring
->dev
, IXGBE_CB(skb
)->dma
,
1623 ixgbe_rx_pg_size(rx_ring
), DMA_FROM_DEVICE
);
1624 IXGBE_CB(skb
)->page_released
= false;
1626 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[0];
1628 dma_sync_single_range_for_cpu(rx_ring
->dev
,
1631 ixgbe_rx_bufsz(rx_ring
),
1634 IXGBE_CB(skb
)->dma
= 0;
1638 * ixgbe_cleanup_headers - Correct corrupted or empty headers
1639 * @rx_ring: rx descriptor ring packet is being transacted on
1640 * @rx_desc: pointer to the EOP Rx descriptor
1641 * @skb: pointer to current skb being fixed
1643 * Check for corrupted packet headers caused by senders on the local L2
1644 * embedded NIC switch not setting up their Tx Descriptors right. These
1645 * should be very rare.
1647 * Also address the case where we are pulling data in on pages only
1648 * and as such no data is present in the skb header.
1650 * In addition if skb is not at least 60 bytes we need to pad it so that
1651 * it is large enough to qualify as a valid Ethernet frame.
1653 * Returns true if an error was encountered and skb was freed.
1655 static bool ixgbe_cleanup_headers(struct ixgbe_ring
*rx_ring
,
1656 union ixgbe_adv_rx_desc
*rx_desc
,
1657 struct sk_buff
*skb
)
1659 struct net_device
*netdev
= rx_ring
->netdev
;
1661 /* verify that the packet does not have any known errors */
1662 if (unlikely(ixgbe_test_staterr(rx_desc
,
1663 IXGBE_RXDADV_ERR_FRAME_ERR_MASK
) &&
1664 !(netdev
->features
& NETIF_F_RXALL
))) {
1665 dev_kfree_skb_any(skb
);
1669 /* place header in linear portion of buffer */
1670 if (skb_is_nonlinear(skb
))
1671 ixgbe_pull_tail(rx_ring
, skb
);
1674 /* do not attempt to pad FCoE Frames as this will disrupt DDP */
1675 if (ixgbe_rx_is_fcoe(rx_ring
, rx_desc
))
1679 /* if skb_pad returns an error the skb was freed */
1680 if (unlikely(skb
->len
< 60)) {
1681 int pad_len
= 60 - skb
->len
;
1683 if (skb_pad(skb
, pad_len
))
1685 __skb_put(skb
, pad_len
);
1692 * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1693 * @rx_ring: rx descriptor ring to store buffers on
1694 * @old_buff: donor buffer to have page reused
1696 * Synchronizes page for reuse by the adapter
1698 static void ixgbe_reuse_rx_page(struct ixgbe_ring
*rx_ring
,
1699 struct ixgbe_rx_buffer
*old_buff
)
1701 struct ixgbe_rx_buffer
*new_buff
;
1702 u16 nta
= rx_ring
->next_to_alloc
;
1704 new_buff
= &rx_ring
->rx_buffer_info
[nta
];
1706 /* update, and store next to alloc */
1708 rx_ring
->next_to_alloc
= (nta
< rx_ring
->count
) ? nta
: 0;
1710 /* transfer page from old buffer to new buffer */
1711 new_buff
->page
= old_buff
->page
;
1712 new_buff
->dma
= old_buff
->dma
;
1713 new_buff
->page_offset
= old_buff
->page_offset
;
1715 /* sync the buffer for use by the device */
1716 dma_sync_single_range_for_device(rx_ring
->dev
, new_buff
->dma
,
1717 new_buff
->page_offset
,
1718 ixgbe_rx_bufsz(rx_ring
),
1723 * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1724 * @rx_ring: rx descriptor ring to transact packets on
1725 * @rx_buffer: buffer containing page to add
1726 * @rx_desc: descriptor containing length of buffer written by hardware
1727 * @skb: sk_buff to place the data into
1729 * This function will add the data contained in rx_buffer->page to the skb.
1730 * This is done either through a direct copy if the data in the buffer is
1731 * less than the skb header size, otherwise it will just attach the page as
1732 * a frag to the skb.
1734 * The function will then update the page offset if necessary and return
1735 * true if the buffer can be reused by the adapter.
1737 static bool ixgbe_add_rx_frag(struct ixgbe_ring
*rx_ring
,
1738 struct ixgbe_rx_buffer
*rx_buffer
,
1739 union ixgbe_adv_rx_desc
*rx_desc
,
1740 struct sk_buff
*skb
)
1742 struct page
*page
= rx_buffer
->page
;
1743 unsigned int size
= le16_to_cpu(rx_desc
->wb
.upper
.length
);
1744 #if (PAGE_SIZE < 8192)
1745 unsigned int truesize
= ixgbe_rx_bufsz(rx_ring
);
1747 unsigned int truesize
= ALIGN(size
, L1_CACHE_BYTES
);
1748 unsigned int last_offset
= ixgbe_rx_pg_size(rx_ring
) -
1749 ixgbe_rx_bufsz(rx_ring
);
1752 if ((size
<= IXGBE_RX_HDR_SIZE
) && !skb_is_nonlinear(skb
)) {
1753 unsigned char *va
= page_address(page
) + rx_buffer
->page_offset
;
1755 memcpy(__skb_put(skb
, size
), va
, ALIGN(size
, sizeof(long)));
1757 /* we can reuse buffer as-is, just make sure it is local */
1758 if (likely(page_to_nid(page
) == numa_node_id()))
1761 /* this page cannot be reused so discard it */
1766 skb_add_rx_frag(skb
, skb_shinfo(skb
)->nr_frags
, page
,
1767 rx_buffer
->page_offset
, size
, truesize
);
1769 /* avoid re-using remote pages */
1770 if (unlikely(page_to_nid(page
) != numa_node_id()))
1773 #if (PAGE_SIZE < 8192)
1774 /* if we are only owner of page we can reuse it */
1775 if (unlikely(page_count(page
) != 1))
1778 /* flip page offset to other buffer */
1779 rx_buffer
->page_offset
^= truesize
;
1782 * since we are the only owner of the page and we need to
1783 * increment it, just set the value to 2 in order to avoid
1784 * an unecessary locked operation
1786 atomic_set(&page
->_count
, 2);
1788 /* move offset up to the next cache line */
1789 rx_buffer
->page_offset
+= truesize
;
1791 if (rx_buffer
->page_offset
> last_offset
)
1794 /* bump ref count on page before it is given to the stack */
1801 static struct sk_buff
*ixgbe_fetch_rx_buffer(struct ixgbe_ring
*rx_ring
,
1802 union ixgbe_adv_rx_desc
*rx_desc
)
1804 struct ixgbe_rx_buffer
*rx_buffer
;
1805 struct sk_buff
*skb
;
1808 rx_buffer
= &rx_ring
->rx_buffer_info
[rx_ring
->next_to_clean
];
1809 page
= rx_buffer
->page
;
1812 skb
= rx_buffer
->skb
;
1815 void *page_addr
= page_address(page
) +
1816 rx_buffer
->page_offset
;
1818 /* prefetch first cache line of first page */
1819 prefetch(page_addr
);
1820 #if L1_CACHE_BYTES < 128
1821 prefetch(page_addr
+ L1_CACHE_BYTES
);
1824 /* allocate a skb to store the frags */
1825 skb
= netdev_alloc_skb_ip_align(rx_ring
->netdev
,
1827 if (unlikely(!skb
)) {
1828 rx_ring
->rx_stats
.alloc_rx_buff_failed
++;
1833 * we will be copying header into skb->data in
1834 * pskb_may_pull so it is in our interest to prefetch
1835 * it now to avoid a possible cache miss
1837 prefetchw(skb
->data
);
1840 * Delay unmapping of the first packet. It carries the
1841 * header information, HW may still access the header
1842 * after the writeback. Only unmap it when EOP is
1845 if (likely(ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_EOP
)))
1848 IXGBE_CB(skb
)->dma
= rx_buffer
->dma
;
1850 if (ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_EOP
))
1851 ixgbe_dma_sync_frag(rx_ring
, skb
);
1854 /* we are reusing so sync this buffer for CPU use */
1855 dma_sync_single_range_for_cpu(rx_ring
->dev
,
1857 rx_buffer
->page_offset
,
1858 ixgbe_rx_bufsz(rx_ring
),
1862 /* pull page into skb */
1863 if (ixgbe_add_rx_frag(rx_ring
, rx_buffer
, rx_desc
, skb
)) {
1864 /* hand second half of page back to the ring */
1865 ixgbe_reuse_rx_page(rx_ring
, rx_buffer
);
1866 } else if (IXGBE_CB(skb
)->dma
== rx_buffer
->dma
) {
1867 /* the page has been released from the ring */
1868 IXGBE_CB(skb
)->page_released
= true;
1870 /* we are not reusing the buffer so unmap it */
1871 dma_unmap_page(rx_ring
->dev
, rx_buffer
->dma
,
1872 ixgbe_rx_pg_size(rx_ring
),
1876 /* clear contents of buffer_info */
1877 rx_buffer
->skb
= NULL
;
1879 rx_buffer
->page
= NULL
;
1885 * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
1886 * @q_vector: structure containing interrupt and ring information
1887 * @rx_ring: rx descriptor ring to transact packets on
1888 * @budget: Total limit on number of packets to process
1890 * This function provides a "bounce buffer" approach to Rx interrupt
1891 * processing. The advantage to this is that on systems that have
1892 * expensive overhead for IOMMU access this provides a means of avoiding
1893 * it by maintaining the mapping of the page to the syste.
1895 * Returns true if all work is completed without reaching budget
1897 static bool ixgbe_clean_rx_irq(struct ixgbe_q_vector
*q_vector
,
1898 struct ixgbe_ring
*rx_ring
,
1901 unsigned int total_rx_bytes
= 0, total_rx_packets
= 0;
1903 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
1905 unsigned int mss
= 0;
1906 #endif /* IXGBE_FCOE */
1907 u16 cleaned_count
= ixgbe_desc_unused(rx_ring
);
1910 union ixgbe_adv_rx_desc
*rx_desc
;
1911 struct sk_buff
*skb
;
1913 /* return some buffers to hardware, one at a time is too slow */
1914 if (cleaned_count
>= IXGBE_RX_BUFFER_WRITE
) {
1915 ixgbe_alloc_rx_buffers(rx_ring
, cleaned_count
);
1919 rx_desc
= IXGBE_RX_DESC(rx_ring
, rx_ring
->next_to_clean
);
1921 if (!ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_DD
))
1925 * This memory barrier is needed to keep us from reading
1926 * any other fields out of the rx_desc until we know the
1927 * RXD_STAT_DD bit is set
1931 /* retrieve a buffer from the ring */
1932 skb
= ixgbe_fetch_rx_buffer(rx_ring
, rx_desc
);
1934 /* exit if we failed to retrieve a buffer */
1940 /* place incomplete frames back on ring for completion */
1941 if (ixgbe_is_non_eop(rx_ring
, rx_desc
, skb
))
1944 /* verify the packet layout is correct */
1945 if (ixgbe_cleanup_headers(rx_ring
, rx_desc
, skb
))
1948 /* probably a little skewed due to removing CRC */
1949 total_rx_bytes
+= skb
->len
;
1951 /* populate checksum, timestamp, VLAN, and protocol */
1952 ixgbe_process_skb_fields(rx_ring
, rx_desc
, skb
);
1955 /* if ddp, not passing to ULD unless for FCP_RSP or error */
1956 if (ixgbe_rx_is_fcoe(rx_ring
, rx_desc
)) {
1957 ddp_bytes
= ixgbe_fcoe_ddp(adapter
, rx_desc
, skb
);
1958 /* include DDPed FCoE data */
1959 if (ddp_bytes
> 0) {
1961 mss
= rx_ring
->netdev
->mtu
-
1962 sizeof(struct fcoe_hdr
) -
1963 sizeof(struct fc_frame_header
) -
1964 sizeof(struct fcoe_crc_eof
);
1968 total_rx_bytes
+= ddp_bytes
;
1969 total_rx_packets
+= DIV_ROUND_UP(ddp_bytes
,
1973 dev_kfree_skb_any(skb
);
1978 #endif /* IXGBE_FCOE */
1979 ixgbe_rx_skb(q_vector
, skb
);
1981 /* update budget accounting */
1983 } while (likely(total_rx_packets
< budget
));
1985 u64_stats_update_begin(&rx_ring
->syncp
);
1986 rx_ring
->stats
.packets
+= total_rx_packets
;
1987 rx_ring
->stats
.bytes
+= total_rx_bytes
;
1988 u64_stats_update_end(&rx_ring
->syncp
);
1989 q_vector
->rx
.total_packets
+= total_rx_packets
;
1990 q_vector
->rx
.total_bytes
+= total_rx_bytes
;
1993 ixgbe_alloc_rx_buffers(rx_ring
, cleaned_count
);
1995 return (total_rx_packets
< budget
);
1999 * ixgbe_configure_msix - Configure MSI-X hardware
2000 * @adapter: board private structure
2002 * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2005 static void ixgbe_configure_msix(struct ixgbe_adapter
*adapter
)
2007 struct ixgbe_q_vector
*q_vector
;
2011 /* Populate MSIX to EITR Select */
2012 if (adapter
->num_vfs
> 32) {
2013 u32 eitrsel
= (1 << (adapter
->num_vfs
- 32)) - 1;
2014 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EITRSEL
, eitrsel
);
2018 * Populate the IVAR table and set the ITR values to the
2019 * corresponding register.
2021 for (v_idx
= 0; v_idx
< adapter
->num_q_vectors
; v_idx
++) {
2022 struct ixgbe_ring
*ring
;
2023 q_vector
= adapter
->q_vector
[v_idx
];
2025 ixgbe_for_each_ring(ring
, q_vector
->rx
)
2026 ixgbe_set_ivar(adapter
, 0, ring
->reg_idx
, v_idx
);
2028 ixgbe_for_each_ring(ring
, q_vector
->tx
)
2029 ixgbe_set_ivar(adapter
, 1, ring
->reg_idx
, v_idx
);
2031 ixgbe_write_eitr(q_vector
);
2034 switch (adapter
->hw
.mac
.type
) {
2035 case ixgbe_mac_82598EB
:
2036 ixgbe_set_ivar(adapter
, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX
,
2039 case ixgbe_mac_82599EB
:
2040 case ixgbe_mac_X540
:
2041 ixgbe_set_ivar(adapter
, -1, 1, v_idx
);
2046 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EITR(v_idx
), 1950);
2048 /* set up to autoclear timer, and the vectors */
2049 mask
= IXGBE_EIMS_ENABLE_MASK
;
2050 mask
&= ~(IXGBE_EIMS_OTHER
|
2051 IXGBE_EIMS_MAILBOX
|
2054 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIAC
, mask
);
2057 enum latency_range
{
2061 latency_invalid
= 255
2065 * ixgbe_update_itr - update the dynamic ITR value based on statistics
2066 * @q_vector: structure containing interrupt and ring information
2067 * @ring_container: structure containing ring performance data
2069 * Stores a new ITR value based on packets and byte
2070 * counts during the last interrupt. The advantage of per interrupt
2071 * computation is faster updates and more accurate ITR for the current
2072 * traffic pattern. Constants in this function were computed
2073 * based on theoretical maximum wire speed and thresholds were set based
2074 * on testing data as well as attempting to minimize response time
2075 * while increasing bulk throughput.
2076 * this functionality is controlled by the InterruptThrottleRate module
2077 * parameter (see ixgbe_param.c)
2079 static void ixgbe_update_itr(struct ixgbe_q_vector
*q_vector
,
2080 struct ixgbe_ring_container
*ring_container
)
2082 int bytes
= ring_container
->total_bytes
;
2083 int packets
= ring_container
->total_packets
;
2086 u8 itr_setting
= ring_container
->itr
;
2091 /* simple throttlerate management
2092 * 0-10MB/s lowest (100000 ints/s)
2093 * 10-20MB/s low (20000 ints/s)
2094 * 20-1249MB/s bulk (8000 ints/s)
2096 /* what was last interrupt timeslice? */
2097 timepassed_us
= q_vector
->itr
>> 2;
2098 bytes_perint
= bytes
/ timepassed_us
; /* bytes/usec */
2100 switch (itr_setting
) {
2101 case lowest_latency
:
2102 if (bytes_perint
> 10)
2103 itr_setting
= low_latency
;
2106 if (bytes_perint
> 20)
2107 itr_setting
= bulk_latency
;
2108 else if (bytes_perint
<= 10)
2109 itr_setting
= lowest_latency
;
2112 if (bytes_perint
<= 20)
2113 itr_setting
= low_latency
;
2117 /* clear work counters since we have the values we need */
2118 ring_container
->total_bytes
= 0;
2119 ring_container
->total_packets
= 0;
2121 /* write updated itr to ring container */
2122 ring_container
->itr
= itr_setting
;
2126 * ixgbe_write_eitr - write EITR register in hardware specific way
2127 * @q_vector: structure containing interrupt and ring information
2129 * This function is made to be called by ethtool and by the driver
2130 * when it needs to update EITR registers at runtime. Hardware
2131 * specific quirks/differences are taken care of here.
2133 void ixgbe_write_eitr(struct ixgbe_q_vector
*q_vector
)
2135 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
2136 struct ixgbe_hw
*hw
= &adapter
->hw
;
2137 int v_idx
= q_vector
->v_idx
;
2138 u32 itr_reg
= q_vector
->itr
& IXGBE_MAX_EITR
;
2140 switch (adapter
->hw
.mac
.type
) {
2141 case ixgbe_mac_82598EB
:
2142 /* must write high and low 16 bits to reset counter */
2143 itr_reg
|= (itr_reg
<< 16);
2145 case ixgbe_mac_82599EB
:
2146 case ixgbe_mac_X540
:
2148 * set the WDIS bit to not clear the timer bits and cause an
2149 * immediate assertion of the interrupt
2151 itr_reg
|= IXGBE_EITR_CNT_WDIS
;
2156 IXGBE_WRITE_REG(hw
, IXGBE_EITR(v_idx
), itr_reg
);
2159 static void ixgbe_set_itr(struct ixgbe_q_vector
*q_vector
)
2161 u32 new_itr
= q_vector
->itr
;
2164 ixgbe_update_itr(q_vector
, &q_vector
->tx
);
2165 ixgbe_update_itr(q_vector
, &q_vector
->rx
);
2167 current_itr
= max(q_vector
->rx
.itr
, q_vector
->tx
.itr
);
2169 switch (current_itr
) {
2170 /* counts and packets in update_itr are dependent on these numbers */
2171 case lowest_latency
:
2172 new_itr
= IXGBE_100K_ITR
;
2175 new_itr
= IXGBE_20K_ITR
;
2178 new_itr
= IXGBE_8K_ITR
;
2184 if (new_itr
!= q_vector
->itr
) {
2185 /* do an exponential smoothing */
2186 new_itr
= (10 * new_itr
* q_vector
->itr
) /
2187 ((9 * new_itr
) + q_vector
->itr
);
2189 /* save the algorithm value here */
2190 q_vector
->itr
= new_itr
;
2192 ixgbe_write_eitr(q_vector
);
2197 * ixgbe_check_overtemp_subtask - check for over temperature
2198 * @adapter: pointer to adapter
2200 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter
*adapter
)
2202 struct ixgbe_hw
*hw
= &adapter
->hw
;
2203 u32 eicr
= adapter
->interrupt_event
;
2205 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
2208 if (!(adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
) &&
2209 !(adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_EVENT
))
2212 adapter
->flags2
&= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT
;
2214 switch (hw
->device_id
) {
2215 case IXGBE_DEV_ID_82599_T3_LOM
:
2217 * Since the warning interrupt is for both ports
2218 * we don't have to check if:
2219 * - This interrupt wasn't for our port.
2220 * - We may have missed the interrupt so always have to
2221 * check if we got a LSC
2223 if (!(eicr
& IXGBE_EICR_GPI_SDP0
) &&
2224 !(eicr
& IXGBE_EICR_LSC
))
2227 if (!(eicr
& IXGBE_EICR_LSC
) && hw
->mac
.ops
.check_link
) {
2229 bool link_up
= false;
2231 hw
->mac
.ops
.check_link(hw
, &speed
, &link_up
, false);
2237 /* Check if this is not due to overtemp */
2238 if (hw
->phy
.ops
.check_overtemp(hw
) != IXGBE_ERR_OVERTEMP
)
2243 if (!(eicr
& IXGBE_EICR_GPI_SDP0
))
2248 "Network adapter has been stopped because it has over heated. "
2249 "Restart the computer. If the problem persists, "
2250 "power off the system and replace the adapter\n");
2252 adapter
->interrupt_event
= 0;
2255 static void ixgbe_check_fan_failure(struct ixgbe_adapter
*adapter
, u32 eicr
)
2257 struct ixgbe_hw
*hw
= &adapter
->hw
;
2259 if ((adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
) &&
2260 (eicr
& IXGBE_EICR_GPI_SDP1
)) {
2261 e_crit(probe
, "Fan has stopped, replace the adapter\n");
2262 /* write to clear the interrupt */
2263 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, IXGBE_EICR_GPI_SDP1
);
2267 static void ixgbe_check_overtemp_event(struct ixgbe_adapter
*adapter
, u32 eicr
)
2269 if (!(adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
))
2272 switch (adapter
->hw
.mac
.type
) {
2273 case ixgbe_mac_82599EB
:
2275 * Need to check link state so complete overtemp check
2278 if (((eicr
& IXGBE_EICR_GPI_SDP0
) || (eicr
& IXGBE_EICR_LSC
)) &&
2279 (!test_bit(__IXGBE_DOWN
, &adapter
->state
))) {
2280 adapter
->interrupt_event
= eicr
;
2281 adapter
->flags2
|= IXGBE_FLAG2_TEMP_SENSOR_EVENT
;
2282 ixgbe_service_event_schedule(adapter
);
2286 case ixgbe_mac_X540
:
2287 if (!(eicr
& IXGBE_EICR_TS
))
2295 "Network adapter has been stopped because it has over heated. "
2296 "Restart the computer. If the problem persists, "
2297 "power off the system and replace the adapter\n");
2300 static void ixgbe_check_sfp_event(struct ixgbe_adapter
*adapter
, u32 eicr
)
2302 struct ixgbe_hw
*hw
= &adapter
->hw
;
2304 if (eicr
& IXGBE_EICR_GPI_SDP2
) {
2305 /* Clear the interrupt */
2306 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, IXGBE_EICR_GPI_SDP2
);
2307 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
2308 adapter
->flags2
|= IXGBE_FLAG2_SFP_NEEDS_RESET
;
2309 ixgbe_service_event_schedule(adapter
);
2313 if (eicr
& IXGBE_EICR_GPI_SDP1
) {
2314 /* Clear the interrupt */
2315 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, IXGBE_EICR_GPI_SDP1
);
2316 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
2317 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_CONFIG
;
2318 ixgbe_service_event_schedule(adapter
);
2323 static void ixgbe_check_lsc(struct ixgbe_adapter
*adapter
)
2325 struct ixgbe_hw
*hw
= &adapter
->hw
;
2328 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_UPDATE
;
2329 adapter
->link_check_timeout
= jiffies
;
2330 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
2331 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, IXGBE_EIMC_LSC
);
2332 IXGBE_WRITE_FLUSH(hw
);
2333 ixgbe_service_event_schedule(adapter
);
2337 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter
*adapter
,
2341 struct ixgbe_hw
*hw
= &adapter
->hw
;
2343 switch (hw
->mac
.type
) {
2344 case ixgbe_mac_82598EB
:
2345 mask
= (IXGBE_EIMS_RTX_QUEUE
& qmask
);
2346 IXGBE_WRITE_REG(hw
, IXGBE_EIMS
, mask
);
2348 case ixgbe_mac_82599EB
:
2349 case ixgbe_mac_X540
:
2350 mask
= (qmask
& 0xFFFFFFFF);
2352 IXGBE_WRITE_REG(hw
, IXGBE_EIMS_EX(0), mask
);
2353 mask
= (qmask
>> 32);
2355 IXGBE_WRITE_REG(hw
, IXGBE_EIMS_EX(1), mask
);
2360 /* skip the flush */
2363 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter
*adapter
,
2367 struct ixgbe_hw
*hw
= &adapter
->hw
;
2369 switch (hw
->mac
.type
) {
2370 case ixgbe_mac_82598EB
:
2371 mask
= (IXGBE_EIMS_RTX_QUEUE
& qmask
);
2372 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, mask
);
2374 case ixgbe_mac_82599EB
:
2375 case ixgbe_mac_X540
:
2376 mask
= (qmask
& 0xFFFFFFFF);
2378 IXGBE_WRITE_REG(hw
, IXGBE_EIMC_EX(0), mask
);
2379 mask
= (qmask
>> 32);
2381 IXGBE_WRITE_REG(hw
, IXGBE_EIMC_EX(1), mask
);
2386 /* skip the flush */
2390 * ixgbe_irq_enable - Enable default interrupt generation settings
2391 * @adapter: board private structure
2393 static inline void ixgbe_irq_enable(struct ixgbe_adapter
*adapter
, bool queues
,
2396 u32 mask
= (IXGBE_EIMS_ENABLE_MASK
& ~IXGBE_EIMS_RTX_QUEUE
);
2398 /* don't reenable LSC while waiting for link */
2399 if (adapter
->flags
& IXGBE_FLAG_NEED_LINK_UPDATE
)
2400 mask
&= ~IXGBE_EIMS_LSC
;
2402 if (adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
)
2403 switch (adapter
->hw
.mac
.type
) {
2404 case ixgbe_mac_82599EB
:
2405 mask
|= IXGBE_EIMS_GPI_SDP0
;
2407 case ixgbe_mac_X540
:
2408 mask
|= IXGBE_EIMS_TS
;
2413 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
)
2414 mask
|= IXGBE_EIMS_GPI_SDP1
;
2415 switch (adapter
->hw
.mac
.type
) {
2416 case ixgbe_mac_82599EB
:
2417 mask
|= IXGBE_EIMS_GPI_SDP1
;
2418 mask
|= IXGBE_EIMS_GPI_SDP2
;
2419 case ixgbe_mac_X540
:
2420 mask
|= IXGBE_EIMS_ECC
;
2421 mask
|= IXGBE_EIMS_MAILBOX
;
2427 if (adapter
->hw
.mac
.type
== ixgbe_mac_X540
)
2428 mask
|= IXGBE_EIMS_TIMESYNC
;
2430 if ((adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) &&
2431 !(adapter
->flags2
& IXGBE_FLAG2_FDIR_REQUIRES_REINIT
))
2432 mask
|= IXGBE_EIMS_FLOW_DIR
;
2434 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMS
, mask
);
2436 ixgbe_irq_enable_queues(adapter
, ~0);
2438 IXGBE_WRITE_FLUSH(&adapter
->hw
);
2441 static irqreturn_t
ixgbe_msix_other(int irq
, void *data
)
2443 struct ixgbe_adapter
*adapter
= data
;
2444 struct ixgbe_hw
*hw
= &adapter
->hw
;
2448 * Workaround for Silicon errata. Use clear-by-write instead
2449 * of clear-by-read. Reading with EICS will return the
2450 * interrupt causes without clearing, which later be done
2451 * with the write to EICR.
2453 eicr
= IXGBE_READ_REG(hw
, IXGBE_EICS
);
2454 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, eicr
);
2456 if (eicr
& IXGBE_EICR_LSC
)
2457 ixgbe_check_lsc(adapter
);
2459 if (eicr
& IXGBE_EICR_MAILBOX
)
2460 ixgbe_msg_task(adapter
);
2462 switch (hw
->mac
.type
) {
2463 case ixgbe_mac_82599EB
:
2464 case ixgbe_mac_X540
:
2465 if (eicr
& IXGBE_EICR_ECC
)
2466 e_info(link
, "Received unrecoverable ECC Err, please "
2468 /* Handle Flow Director Full threshold interrupt */
2469 if (eicr
& IXGBE_EICR_FLOW_DIR
) {
2470 int reinit_count
= 0;
2472 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
2473 struct ixgbe_ring
*ring
= adapter
->tx_ring
[i
];
2474 if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE
,
2479 /* no more flow director interrupts until after init */
2480 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, IXGBE_EIMC_FLOW_DIR
);
2481 adapter
->flags2
|= IXGBE_FLAG2_FDIR_REQUIRES_REINIT
;
2482 ixgbe_service_event_schedule(adapter
);
2485 ixgbe_check_sfp_event(adapter
, eicr
);
2486 ixgbe_check_overtemp_event(adapter
, eicr
);
2492 ixgbe_check_fan_failure(adapter
, eicr
);
2494 if (unlikely(eicr
& IXGBE_EICR_TIMESYNC
))
2495 ixgbe_ptp_check_pps_event(adapter
, eicr
);
2497 /* re-enable the original interrupt state, no lsc, no queues */
2498 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2499 ixgbe_irq_enable(adapter
, false, false);
2504 static irqreturn_t
ixgbe_msix_clean_rings(int irq
, void *data
)
2506 struct ixgbe_q_vector
*q_vector
= data
;
2508 /* EIAM disabled interrupts (on this vector) for us */
2510 if (q_vector
->rx
.ring
|| q_vector
->tx
.ring
)
2511 napi_schedule(&q_vector
->napi
);
2517 * ixgbe_poll - NAPI Rx polling callback
2518 * @napi: structure for representing this polling device
2519 * @budget: how many packets driver is allowed to clean
2521 * This function is used for legacy and MSI, NAPI mode
2523 int ixgbe_poll(struct napi_struct
*napi
, int budget
)
2525 struct ixgbe_q_vector
*q_vector
=
2526 container_of(napi
, struct ixgbe_q_vector
, napi
);
2527 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
2528 struct ixgbe_ring
*ring
;
2529 int per_ring_budget
;
2530 bool clean_complete
= true;
2532 #ifdef CONFIG_IXGBE_DCA
2533 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
)
2534 ixgbe_update_dca(q_vector
);
2537 ixgbe_for_each_ring(ring
, q_vector
->tx
)
2538 clean_complete
&= !!ixgbe_clean_tx_irq(q_vector
, ring
);
2540 /* attempt to distribute budget to each queue fairly, but don't allow
2541 * the budget to go below 1 because we'll exit polling */
2542 if (q_vector
->rx
.count
> 1)
2543 per_ring_budget
= max(budget
/q_vector
->rx
.count
, 1);
2545 per_ring_budget
= budget
;
2547 ixgbe_for_each_ring(ring
, q_vector
->rx
)
2548 clean_complete
&= ixgbe_clean_rx_irq(q_vector
, ring
,
2551 /* If all work not completed, return budget and keep polling */
2552 if (!clean_complete
)
2555 /* all work done, exit the polling mode */
2556 napi_complete(napi
);
2557 if (adapter
->rx_itr_setting
& 1)
2558 ixgbe_set_itr(q_vector
);
2559 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2560 ixgbe_irq_enable_queues(adapter
, ((u64
)1 << q_vector
->v_idx
));
2566 * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2567 * @adapter: board private structure
2569 * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2570 * interrupts from the kernel.
2572 static int ixgbe_request_msix_irqs(struct ixgbe_adapter
*adapter
)
2574 struct net_device
*netdev
= adapter
->netdev
;
2578 for (vector
= 0; vector
< adapter
->num_q_vectors
; vector
++) {
2579 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[vector
];
2580 struct msix_entry
*entry
= &adapter
->msix_entries
[vector
];
2582 if (q_vector
->tx
.ring
&& q_vector
->rx
.ring
) {
2583 snprintf(q_vector
->name
, sizeof(q_vector
->name
) - 1,
2584 "%s-%s-%d", netdev
->name
, "TxRx", ri
++);
2586 } else if (q_vector
->rx
.ring
) {
2587 snprintf(q_vector
->name
, sizeof(q_vector
->name
) - 1,
2588 "%s-%s-%d", netdev
->name
, "rx", ri
++);
2589 } else if (q_vector
->tx
.ring
) {
2590 snprintf(q_vector
->name
, sizeof(q_vector
->name
) - 1,
2591 "%s-%s-%d", netdev
->name
, "tx", ti
++);
2593 /* skip this unused q_vector */
2596 err
= request_irq(entry
->vector
, &ixgbe_msix_clean_rings
, 0,
2597 q_vector
->name
, q_vector
);
2599 e_err(probe
, "request_irq failed for MSIX interrupt "
2600 "Error: %d\n", err
);
2601 goto free_queue_irqs
;
2603 /* If Flow Director is enabled, set interrupt affinity */
2604 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) {
2605 /* assign the mask for this irq */
2606 irq_set_affinity_hint(entry
->vector
,
2607 &q_vector
->affinity_mask
);
2611 err
= request_irq(adapter
->msix_entries
[vector
].vector
,
2612 ixgbe_msix_other
, 0, netdev
->name
, adapter
);
2614 e_err(probe
, "request_irq for msix_other failed: %d\n", err
);
2615 goto free_queue_irqs
;
2623 irq_set_affinity_hint(adapter
->msix_entries
[vector
].vector
,
2625 free_irq(adapter
->msix_entries
[vector
].vector
,
2626 adapter
->q_vector
[vector
]);
2628 adapter
->flags
&= ~IXGBE_FLAG_MSIX_ENABLED
;
2629 pci_disable_msix(adapter
->pdev
);
2630 kfree(adapter
->msix_entries
);
2631 adapter
->msix_entries
= NULL
;
2636 * ixgbe_intr - legacy mode Interrupt Handler
2637 * @irq: interrupt number
2638 * @data: pointer to a network interface device structure
2640 static irqreturn_t
ixgbe_intr(int irq
, void *data
)
2642 struct ixgbe_adapter
*adapter
= data
;
2643 struct ixgbe_hw
*hw
= &adapter
->hw
;
2644 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[0];
2648 * Workaround for silicon errata #26 on 82598. Mask the interrupt
2649 * before the read of EICR.
2651 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, IXGBE_IRQ_CLEAR_MASK
);
2653 /* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2654 * therefore no explicit interrupt disable is necessary */
2655 eicr
= IXGBE_READ_REG(hw
, IXGBE_EICR
);
2658 * shared interrupt alert!
2659 * make sure interrupts are enabled because the read will
2660 * have disabled interrupts due to EIAM
2661 * finish the workaround of silicon errata on 82598. Unmask
2662 * the interrupt that we masked before the EICR read.
2664 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2665 ixgbe_irq_enable(adapter
, true, true);
2666 return IRQ_NONE
; /* Not our interrupt */
2669 if (eicr
& IXGBE_EICR_LSC
)
2670 ixgbe_check_lsc(adapter
);
2672 switch (hw
->mac
.type
) {
2673 case ixgbe_mac_82599EB
:
2674 ixgbe_check_sfp_event(adapter
, eicr
);
2676 case ixgbe_mac_X540
:
2677 if (eicr
& IXGBE_EICR_ECC
)
2678 e_info(link
, "Received unrecoverable ECC err, please "
2680 ixgbe_check_overtemp_event(adapter
, eicr
);
2686 ixgbe_check_fan_failure(adapter
, eicr
);
2687 if (unlikely(eicr
& IXGBE_EICR_TIMESYNC
))
2688 ixgbe_ptp_check_pps_event(adapter
, eicr
);
2690 /* would disable interrupts here but EIAM disabled it */
2691 napi_schedule(&q_vector
->napi
);
2694 * re-enable link(maybe) and non-queue interrupts, no flush.
2695 * ixgbe_poll will re-enable the queue interrupts
2697 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2698 ixgbe_irq_enable(adapter
, false, false);
2704 * ixgbe_request_irq - initialize interrupts
2705 * @adapter: board private structure
2707 * Attempts to configure interrupts using the best available
2708 * capabilities of the hardware and kernel.
2710 static int ixgbe_request_irq(struct ixgbe_adapter
*adapter
)
2712 struct net_device
*netdev
= adapter
->netdev
;
2715 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)
2716 err
= ixgbe_request_msix_irqs(adapter
);
2717 else if (adapter
->flags
& IXGBE_FLAG_MSI_ENABLED
)
2718 err
= request_irq(adapter
->pdev
->irq
, ixgbe_intr
, 0,
2719 netdev
->name
, adapter
);
2721 err
= request_irq(adapter
->pdev
->irq
, ixgbe_intr
, IRQF_SHARED
,
2722 netdev
->name
, adapter
);
2725 e_err(probe
, "request_irq failed, Error %d\n", err
);
2730 static void ixgbe_free_irq(struct ixgbe_adapter
*adapter
)
2734 if (!(adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)) {
2735 free_irq(adapter
->pdev
->irq
, adapter
);
2739 for (vector
= 0; vector
< adapter
->num_q_vectors
; vector
++) {
2740 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[vector
];
2741 struct msix_entry
*entry
= &adapter
->msix_entries
[vector
];
2743 /* free only the irqs that were actually requested */
2744 if (!q_vector
->rx
.ring
&& !q_vector
->tx
.ring
)
2747 /* clear the affinity_mask in the IRQ descriptor */
2748 irq_set_affinity_hint(entry
->vector
, NULL
);
2750 free_irq(entry
->vector
, q_vector
);
2753 free_irq(adapter
->msix_entries
[vector
++].vector
, adapter
);
2757 * ixgbe_irq_disable - Mask off interrupt generation on the NIC
2758 * @adapter: board private structure
2760 static inline void ixgbe_irq_disable(struct ixgbe_adapter
*adapter
)
2762 switch (adapter
->hw
.mac
.type
) {
2763 case ixgbe_mac_82598EB
:
2764 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC
, ~0);
2766 case ixgbe_mac_82599EB
:
2767 case ixgbe_mac_X540
:
2768 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC
, 0xFFFF0000);
2769 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC_EX(0), ~0);
2770 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC_EX(1), ~0);
2775 IXGBE_WRITE_FLUSH(&adapter
->hw
);
2776 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
) {
2779 for (vector
= 0; vector
< adapter
->num_q_vectors
; vector
++)
2780 synchronize_irq(adapter
->msix_entries
[vector
].vector
);
2782 synchronize_irq(adapter
->msix_entries
[vector
++].vector
);
2784 synchronize_irq(adapter
->pdev
->irq
);
2789 * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
2792 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter
*adapter
)
2794 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[0];
2796 ixgbe_write_eitr(q_vector
);
2798 ixgbe_set_ivar(adapter
, 0, 0, 0);
2799 ixgbe_set_ivar(adapter
, 1, 0, 0);
2801 e_info(hw
, "Legacy interrupt IVAR setup done\n");
2805 * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
2806 * @adapter: board private structure
2807 * @ring: structure containing ring specific data
2809 * Configure the Tx descriptor ring after a reset.
2811 void ixgbe_configure_tx_ring(struct ixgbe_adapter
*adapter
,
2812 struct ixgbe_ring
*ring
)
2814 struct ixgbe_hw
*hw
= &adapter
->hw
;
2815 u64 tdba
= ring
->dma
;
2817 u32 txdctl
= IXGBE_TXDCTL_ENABLE
;
2818 u8 reg_idx
= ring
->reg_idx
;
2820 /* disable queue to avoid issues while updating state */
2821 IXGBE_WRITE_REG(hw
, IXGBE_TXDCTL(reg_idx
), 0);
2822 IXGBE_WRITE_FLUSH(hw
);
2824 IXGBE_WRITE_REG(hw
, IXGBE_TDBAL(reg_idx
),
2825 (tdba
& DMA_BIT_MASK(32)));
2826 IXGBE_WRITE_REG(hw
, IXGBE_TDBAH(reg_idx
), (tdba
>> 32));
2827 IXGBE_WRITE_REG(hw
, IXGBE_TDLEN(reg_idx
),
2828 ring
->count
* sizeof(union ixgbe_adv_tx_desc
));
2829 IXGBE_WRITE_REG(hw
, IXGBE_TDH(reg_idx
), 0);
2830 IXGBE_WRITE_REG(hw
, IXGBE_TDT(reg_idx
), 0);
2831 ring
->tail
= hw
->hw_addr
+ IXGBE_TDT(reg_idx
);
2834 * set WTHRESH to encourage burst writeback, it should not be set
2835 * higher than 1 when:
2836 * - ITR is 0 as it could cause false TX hangs
2837 * - ITR is set to > 100k int/sec and BQL is enabled
2839 * In order to avoid issues WTHRESH + PTHRESH should always be equal
2840 * to or less than the number of on chip descriptors, which is
2843 #if IS_ENABLED(CONFIG_BQL)
2844 if (!ring
->q_vector
|| (ring
->q_vector
->itr
< IXGBE_100K_ITR
))
2846 if (!ring
->q_vector
|| (ring
->q_vector
->itr
< 8))
2848 txdctl
|= (1 << 16); /* WTHRESH = 1 */
2850 txdctl
|= (8 << 16); /* WTHRESH = 8 */
2853 * Setting PTHRESH to 32 both improves performance
2854 * and avoids a TX hang with DFP enabled
2856 txdctl
|= (1 << 8) | /* HTHRESH = 1 */
2857 32; /* PTHRESH = 32 */
2859 /* reinitialize flowdirector state */
2860 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) {
2861 ring
->atr_sample_rate
= adapter
->atr_sample_rate
;
2862 ring
->atr_count
= 0;
2863 set_bit(__IXGBE_TX_FDIR_INIT_DONE
, &ring
->state
);
2865 ring
->atr_sample_rate
= 0;
2868 /* initialize XPS */
2869 if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE
, &ring
->state
)) {
2870 struct ixgbe_q_vector
*q_vector
= ring
->q_vector
;
2873 netif_set_xps_queue(adapter
->netdev
,
2874 &q_vector
->affinity_mask
,
2878 clear_bit(__IXGBE_HANG_CHECK_ARMED
, &ring
->state
);
2881 IXGBE_WRITE_REG(hw
, IXGBE_TXDCTL(reg_idx
), txdctl
);
2883 /* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
2884 if (hw
->mac
.type
== ixgbe_mac_82598EB
&&
2885 !(IXGBE_READ_REG(hw
, IXGBE_LINKS
) & IXGBE_LINKS_UP
))
2888 /* poll to verify queue is enabled */
2890 usleep_range(1000, 2000);
2891 txdctl
= IXGBE_READ_REG(hw
, IXGBE_TXDCTL(reg_idx
));
2892 } while (--wait_loop
&& !(txdctl
& IXGBE_TXDCTL_ENABLE
));
2894 e_err(drv
, "Could not enable Tx Queue %d\n", reg_idx
);
2897 static void ixgbe_setup_mtqc(struct ixgbe_adapter
*adapter
)
2899 struct ixgbe_hw
*hw
= &adapter
->hw
;
2901 u8 tcs
= netdev_get_num_tc(adapter
->netdev
);
2903 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
2906 /* disable the arbiter while setting MTQC */
2907 rttdcs
= IXGBE_READ_REG(hw
, IXGBE_RTTDCS
);
2908 rttdcs
|= IXGBE_RTTDCS_ARBDIS
;
2909 IXGBE_WRITE_REG(hw
, IXGBE_RTTDCS
, rttdcs
);
2911 /* set transmit pool layout */
2912 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
2913 mtqc
= IXGBE_MTQC_VT_ENA
;
2915 mtqc
|= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_8TC_8TQ
;
2917 mtqc
|= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_4TC_4TQ
;
2918 else if (adapter
->ring_feature
[RING_F_RSS
].indices
== 4)
2919 mtqc
|= IXGBE_MTQC_32VF
;
2921 mtqc
|= IXGBE_MTQC_64VF
;
2924 mtqc
= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_8TC_8TQ
;
2926 mtqc
= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_4TC_4TQ
;
2928 mtqc
= IXGBE_MTQC_64Q_1PB
;
2931 IXGBE_WRITE_REG(hw
, IXGBE_MTQC
, mtqc
);
2933 /* Enable Security TX Buffer IFG for multiple pb */
2935 u32 sectx
= IXGBE_READ_REG(hw
, IXGBE_SECTXMINIFG
);
2936 sectx
|= IXGBE_SECTX_DCB
;
2937 IXGBE_WRITE_REG(hw
, IXGBE_SECTXMINIFG
, sectx
);
2940 /* re-enable the arbiter */
2941 rttdcs
&= ~IXGBE_RTTDCS_ARBDIS
;
2942 IXGBE_WRITE_REG(hw
, IXGBE_RTTDCS
, rttdcs
);
2946 * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
2947 * @adapter: board private structure
2949 * Configure the Tx unit of the MAC after a reset.
2951 static void ixgbe_configure_tx(struct ixgbe_adapter
*adapter
)
2953 struct ixgbe_hw
*hw
= &adapter
->hw
;
2957 ixgbe_setup_mtqc(adapter
);
2959 if (hw
->mac
.type
!= ixgbe_mac_82598EB
) {
2960 /* DMATXCTL.EN must be before Tx queues are enabled */
2961 dmatxctl
= IXGBE_READ_REG(hw
, IXGBE_DMATXCTL
);
2962 dmatxctl
|= IXGBE_DMATXCTL_TE
;
2963 IXGBE_WRITE_REG(hw
, IXGBE_DMATXCTL
, dmatxctl
);
2966 /* Setup the HW Tx Head and Tail descriptor pointers */
2967 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
2968 ixgbe_configure_tx_ring(adapter
, adapter
->tx_ring
[i
]);
2971 static void ixgbe_enable_rx_drop(struct ixgbe_adapter
*adapter
,
2972 struct ixgbe_ring
*ring
)
2974 struct ixgbe_hw
*hw
= &adapter
->hw
;
2975 u8 reg_idx
= ring
->reg_idx
;
2976 u32 srrctl
= IXGBE_READ_REG(hw
, IXGBE_SRRCTL(reg_idx
));
2978 srrctl
|= IXGBE_SRRCTL_DROP_EN
;
2980 IXGBE_WRITE_REG(hw
, IXGBE_SRRCTL(reg_idx
), srrctl
);
2983 static void ixgbe_disable_rx_drop(struct ixgbe_adapter
*adapter
,
2984 struct ixgbe_ring
*ring
)
2986 struct ixgbe_hw
*hw
= &adapter
->hw
;
2987 u8 reg_idx
= ring
->reg_idx
;
2988 u32 srrctl
= IXGBE_READ_REG(hw
, IXGBE_SRRCTL(reg_idx
));
2990 srrctl
&= ~IXGBE_SRRCTL_DROP_EN
;
2992 IXGBE_WRITE_REG(hw
, IXGBE_SRRCTL(reg_idx
), srrctl
);
2995 #ifdef CONFIG_IXGBE_DCB
2996 void ixgbe_set_rx_drop_en(struct ixgbe_adapter
*adapter
)
2998 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter
*adapter
)
3002 bool pfc_en
= adapter
->dcb_cfg
.pfc_mode_enable
;
3004 if (adapter
->ixgbe_ieee_pfc
)
3005 pfc_en
|= !!(adapter
->ixgbe_ieee_pfc
->pfc_en
);
3008 * We should set the drop enable bit if:
3011 * Number of Rx queues > 1 and flow control is disabled
3013 * This allows us to avoid head of line blocking for security
3014 * and performance reasons.
3016 if (adapter
->num_vfs
|| (adapter
->num_rx_queues
> 1 &&
3017 !(adapter
->hw
.fc
.current_mode
& ixgbe_fc_tx_pause
) && !pfc_en
)) {
3018 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3019 ixgbe_enable_rx_drop(adapter
, adapter
->rx_ring
[i
]);
3021 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3022 ixgbe_disable_rx_drop(adapter
, adapter
->rx_ring
[i
]);
3026 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3028 static void ixgbe_configure_srrctl(struct ixgbe_adapter
*adapter
,
3029 struct ixgbe_ring
*rx_ring
)
3031 struct ixgbe_hw
*hw
= &adapter
->hw
;
3033 u8 reg_idx
= rx_ring
->reg_idx
;
3035 if (hw
->mac
.type
== ixgbe_mac_82598EB
) {
3036 u16 mask
= adapter
->ring_feature
[RING_F_RSS
].mask
;
3039 * if VMDq is not active we must program one srrctl register
3040 * per RSS queue since we have enabled RDRXCTL.MVMEN
3045 /* configure header buffer length, needed for RSC */
3046 srrctl
= IXGBE_RX_HDR_SIZE
<< IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT
;
3048 /* configure the packet buffer length */
3049 srrctl
|= ixgbe_rx_bufsz(rx_ring
) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT
;
3051 /* configure descriptor type */
3052 srrctl
|= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF
;
3054 IXGBE_WRITE_REG(hw
, IXGBE_SRRCTL(reg_idx
), srrctl
);
3057 static void ixgbe_setup_mrqc(struct ixgbe_adapter
*adapter
)
3059 struct ixgbe_hw
*hw
= &adapter
->hw
;
3060 static const u32 seed
[10] = { 0xE291D73D, 0x1805EC6C, 0x2A94B30D,
3061 0xA54F2BEC, 0xEA49AF7C, 0xE214AD3D, 0xB855AABE,
3062 0x6A3E67EA, 0x14364D17, 0x3BED200D};
3063 u32 mrqc
= 0, reta
= 0;
3066 u16 rss_i
= adapter
->ring_feature
[RING_F_RSS
].indices
;
3069 * Program table for at least 2 queues w/ SR-IOV so that VFs can
3070 * make full use of any rings they may have. We will use the
3071 * PSRTYPE register to control how many rings we use within the PF.
3073 if ((adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) && (rss_i
< 2))
3076 /* Fill out hash function seeds */
3077 for (i
= 0; i
< 10; i
++)
3078 IXGBE_WRITE_REG(hw
, IXGBE_RSSRK(i
), seed
[i
]);
3080 /* Fill out redirection table */
3081 for (i
= 0, j
= 0; i
< 128; i
++, j
++) {
3084 /* reta = 4-byte sliding window of
3085 * 0x00..(indices-1)(indices-1)00..etc. */
3086 reta
= (reta
<< 8) | (j
* 0x11);
3088 IXGBE_WRITE_REG(hw
, IXGBE_RETA(i
>> 2), reta
);
3091 /* Disable indicating checksum in descriptor, enables RSS hash */
3092 rxcsum
= IXGBE_READ_REG(hw
, IXGBE_RXCSUM
);
3093 rxcsum
|= IXGBE_RXCSUM_PCSD
;
3094 IXGBE_WRITE_REG(hw
, IXGBE_RXCSUM
, rxcsum
);
3096 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
) {
3097 if (adapter
->ring_feature
[RING_F_RSS
].mask
)
3098 mrqc
= IXGBE_MRQC_RSSEN
;
3100 u8 tcs
= netdev_get_num_tc(adapter
->netdev
);
3102 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
3104 mrqc
= IXGBE_MRQC_VMDQRT8TCEN
; /* 8 TCs */
3106 mrqc
= IXGBE_MRQC_VMDQRT4TCEN
; /* 4 TCs */
3107 else if (adapter
->ring_feature
[RING_F_RSS
].indices
== 4)
3108 mrqc
= IXGBE_MRQC_VMDQRSS32EN
;
3110 mrqc
= IXGBE_MRQC_VMDQRSS64EN
;
3113 mrqc
= IXGBE_MRQC_RTRSS8TCEN
;
3115 mrqc
= IXGBE_MRQC_RTRSS4TCEN
;
3117 mrqc
= IXGBE_MRQC_RSSEN
;
3121 /* Perform hash on these packet types */
3122 mrqc
|= IXGBE_MRQC_RSS_FIELD_IPV4
|
3123 IXGBE_MRQC_RSS_FIELD_IPV4_TCP
|
3124 IXGBE_MRQC_RSS_FIELD_IPV6
|
3125 IXGBE_MRQC_RSS_FIELD_IPV6_TCP
;
3127 if (adapter
->flags2
& IXGBE_FLAG2_RSS_FIELD_IPV4_UDP
)
3128 mrqc
|= IXGBE_MRQC_RSS_FIELD_IPV4_UDP
;
3129 if (adapter
->flags2
& IXGBE_FLAG2_RSS_FIELD_IPV6_UDP
)
3130 mrqc
|= IXGBE_MRQC_RSS_FIELD_IPV6_UDP
;
3132 IXGBE_WRITE_REG(hw
, IXGBE_MRQC
, mrqc
);
3136 * ixgbe_configure_rscctl - enable RSC for the indicated ring
3137 * @adapter: address of board private structure
3138 * @index: index of ring to set
3140 static void ixgbe_configure_rscctl(struct ixgbe_adapter
*adapter
,
3141 struct ixgbe_ring
*ring
)
3143 struct ixgbe_hw
*hw
= &adapter
->hw
;
3145 u8 reg_idx
= ring
->reg_idx
;
3147 if (!ring_is_rsc_enabled(ring
))
3150 rscctrl
= IXGBE_READ_REG(hw
, IXGBE_RSCCTL(reg_idx
));
3151 rscctrl
|= IXGBE_RSCCTL_RSCEN
;
3153 * we must limit the number of descriptors so that the
3154 * total size of max desc * buf_len is not greater
3157 rscctrl
|= IXGBE_RSCCTL_MAXDESC_16
;
3158 IXGBE_WRITE_REG(hw
, IXGBE_RSCCTL(reg_idx
), rscctrl
);
3161 #define IXGBE_MAX_RX_DESC_POLL 10
3162 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter
*adapter
,
3163 struct ixgbe_ring
*ring
)
3165 struct ixgbe_hw
*hw
= &adapter
->hw
;
3166 int wait_loop
= IXGBE_MAX_RX_DESC_POLL
;
3168 u8 reg_idx
= ring
->reg_idx
;
3170 /* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3171 if (hw
->mac
.type
== ixgbe_mac_82598EB
&&
3172 !(IXGBE_READ_REG(hw
, IXGBE_LINKS
) & IXGBE_LINKS_UP
))
3176 usleep_range(1000, 2000);
3177 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3178 } while (--wait_loop
&& !(rxdctl
& IXGBE_RXDCTL_ENABLE
));
3181 e_err(drv
, "RXDCTL.ENABLE on Rx queue %d not set within "
3182 "the polling period\n", reg_idx
);
3186 void ixgbe_disable_rx_queue(struct ixgbe_adapter
*adapter
,
3187 struct ixgbe_ring
*ring
)
3189 struct ixgbe_hw
*hw
= &adapter
->hw
;
3190 int wait_loop
= IXGBE_MAX_RX_DESC_POLL
;
3192 u8 reg_idx
= ring
->reg_idx
;
3194 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3195 rxdctl
&= ~IXGBE_RXDCTL_ENABLE
;
3197 /* write value back with RXDCTL.ENABLE bit cleared */
3198 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(reg_idx
), rxdctl
);
3200 if (hw
->mac
.type
== ixgbe_mac_82598EB
&&
3201 !(IXGBE_READ_REG(hw
, IXGBE_LINKS
) & IXGBE_LINKS_UP
))
3204 /* the hardware may take up to 100us to really disable the rx queue */
3207 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3208 } while (--wait_loop
&& (rxdctl
& IXGBE_RXDCTL_ENABLE
));
3211 e_err(drv
, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3212 "the polling period\n", reg_idx
);
3216 void ixgbe_configure_rx_ring(struct ixgbe_adapter
*adapter
,
3217 struct ixgbe_ring
*ring
)
3219 struct ixgbe_hw
*hw
= &adapter
->hw
;
3220 u64 rdba
= ring
->dma
;
3222 u8 reg_idx
= ring
->reg_idx
;
3224 /* disable queue to avoid issues while updating state */
3225 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3226 ixgbe_disable_rx_queue(adapter
, ring
);
3228 IXGBE_WRITE_REG(hw
, IXGBE_RDBAL(reg_idx
), (rdba
& DMA_BIT_MASK(32)));
3229 IXGBE_WRITE_REG(hw
, IXGBE_RDBAH(reg_idx
), (rdba
>> 32));
3230 IXGBE_WRITE_REG(hw
, IXGBE_RDLEN(reg_idx
),
3231 ring
->count
* sizeof(union ixgbe_adv_rx_desc
));
3232 IXGBE_WRITE_REG(hw
, IXGBE_RDH(reg_idx
), 0);
3233 IXGBE_WRITE_REG(hw
, IXGBE_RDT(reg_idx
), 0);
3234 ring
->tail
= hw
->hw_addr
+ IXGBE_RDT(reg_idx
);
3236 ixgbe_configure_srrctl(adapter
, ring
);
3237 ixgbe_configure_rscctl(adapter
, ring
);
3239 if (hw
->mac
.type
== ixgbe_mac_82598EB
) {
3241 * enable cache line friendly hardware writes:
3242 * PTHRESH=32 descriptors (half the internal cache),
3243 * this also removes ugly rx_no_buffer_count increment
3244 * HTHRESH=4 descriptors (to minimize latency on fetch)
3245 * WTHRESH=8 burst writeback up to two cache lines
3247 rxdctl
&= ~0x3FFFFF;
3251 /* enable receive descriptor ring */
3252 rxdctl
|= IXGBE_RXDCTL_ENABLE
;
3253 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(reg_idx
), rxdctl
);
3255 ixgbe_rx_desc_queue_enable(adapter
, ring
);
3256 ixgbe_alloc_rx_buffers(ring
, ixgbe_desc_unused(ring
));
3259 static void ixgbe_setup_psrtype(struct ixgbe_adapter
*adapter
)
3261 struct ixgbe_hw
*hw
= &adapter
->hw
;
3262 int rss_i
= adapter
->ring_feature
[RING_F_RSS
].indices
;
3265 /* PSRTYPE must be initialized in non 82598 adapters */
3266 u32 psrtype
= IXGBE_PSRTYPE_TCPHDR
|
3267 IXGBE_PSRTYPE_UDPHDR
|
3268 IXGBE_PSRTYPE_IPV4HDR
|
3269 IXGBE_PSRTYPE_L2HDR
|
3270 IXGBE_PSRTYPE_IPV6HDR
;
3272 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
3280 for (p
= 0; p
< adapter
->num_rx_pools
; p
++)
3281 IXGBE_WRITE_REG(hw
, IXGBE_PSRTYPE(VMDQ_P(p
)),
3285 static void ixgbe_configure_virtualization(struct ixgbe_adapter
*adapter
)
3287 struct ixgbe_hw
*hw
= &adapter
->hw
;
3288 u32 reg_offset
, vf_shift
;
3289 u32 gcr_ext
, vmdctl
;
3292 if (!(adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
))
3295 vmdctl
= IXGBE_READ_REG(hw
, IXGBE_VT_CTL
);
3296 vmdctl
|= IXGBE_VMD_CTL_VMDQ_EN
;
3297 vmdctl
&= ~IXGBE_VT_CTL_POOL_MASK
;
3298 vmdctl
|= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT
;
3299 vmdctl
|= IXGBE_VT_CTL_REPLEN
;
3300 IXGBE_WRITE_REG(hw
, IXGBE_VT_CTL
, vmdctl
);
3302 vf_shift
= VMDQ_P(0) % 32;
3303 reg_offset
= (VMDQ_P(0) >= 32) ? 1 : 0;
3305 /* Enable only the PF's pool for Tx/Rx */
3306 IXGBE_WRITE_REG(hw
, IXGBE_VFRE(reg_offset
), (~0) << vf_shift
);
3307 IXGBE_WRITE_REG(hw
, IXGBE_VFRE(reg_offset
^ 1), reg_offset
- 1);
3308 IXGBE_WRITE_REG(hw
, IXGBE_VFTE(reg_offset
), (~0) << vf_shift
);
3309 IXGBE_WRITE_REG(hw
, IXGBE_VFTE(reg_offset
^ 1), reg_offset
- 1);
3310 if (adapter
->flags2
& IXGBE_FLAG2_BRIDGE_MODE_VEB
)
3311 IXGBE_WRITE_REG(hw
, IXGBE_PFDTXGSWC
, IXGBE_PFDTXGSWC_VT_LBEN
);
3313 /* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3314 hw
->mac
.ops
.set_vmdq(hw
, 0, VMDQ_P(0));
3317 * Set up VF register offsets for selected VT Mode,
3318 * i.e. 32 or 64 VFs for SR-IOV
3320 switch (adapter
->ring_feature
[RING_F_VMDQ
].mask
) {
3321 case IXGBE_82599_VMDQ_8Q_MASK
:
3322 gcr_ext
= IXGBE_GCR_EXT_VT_MODE_16
;
3324 case IXGBE_82599_VMDQ_4Q_MASK
:
3325 gcr_ext
= IXGBE_GCR_EXT_VT_MODE_32
;
3328 gcr_ext
= IXGBE_GCR_EXT_VT_MODE_64
;
3332 IXGBE_WRITE_REG(hw
, IXGBE_GCR_EXT
, gcr_ext
);
3335 /* Enable MAC Anti-Spoofing */
3336 hw
->mac
.ops
.set_mac_anti_spoofing(hw
, (adapter
->num_vfs
!= 0),
3338 /* For VFs that have spoof checking turned off */
3339 for (i
= 0; i
< adapter
->num_vfs
; i
++) {
3340 if (!adapter
->vfinfo
[i
].spoofchk_enabled
)
3341 ixgbe_ndo_set_vf_spoofchk(adapter
->netdev
, i
, false);
3345 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter
*adapter
)
3347 struct ixgbe_hw
*hw
= &adapter
->hw
;
3348 struct net_device
*netdev
= adapter
->netdev
;
3349 int max_frame
= netdev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
3350 struct ixgbe_ring
*rx_ring
;
3355 /* adjust max frame to be able to do baby jumbo for FCoE */
3356 if ((adapter
->flags
& IXGBE_FLAG_FCOE_ENABLED
) &&
3357 (max_frame
< IXGBE_FCOE_JUMBO_FRAME_SIZE
))
3358 max_frame
= IXGBE_FCOE_JUMBO_FRAME_SIZE
;
3360 #endif /* IXGBE_FCOE */
3362 /* adjust max frame to be at least the size of a standard frame */
3363 if (max_frame
< (ETH_FRAME_LEN
+ ETH_FCS_LEN
))
3364 max_frame
= (ETH_FRAME_LEN
+ ETH_FCS_LEN
);
3366 mhadd
= IXGBE_READ_REG(hw
, IXGBE_MHADD
);
3367 if (max_frame
!= (mhadd
>> IXGBE_MHADD_MFS_SHIFT
)) {
3368 mhadd
&= ~IXGBE_MHADD_MFS_MASK
;
3369 mhadd
|= max_frame
<< IXGBE_MHADD_MFS_SHIFT
;
3371 IXGBE_WRITE_REG(hw
, IXGBE_MHADD
, mhadd
);
3374 hlreg0
= IXGBE_READ_REG(hw
, IXGBE_HLREG0
);
3375 /* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3376 hlreg0
|= IXGBE_HLREG0_JUMBOEN
;
3377 IXGBE_WRITE_REG(hw
, IXGBE_HLREG0
, hlreg0
);
3380 * Setup the HW Rx Head and Tail Descriptor Pointers and
3381 * the Base and Length of the Rx Descriptor Ring
3383 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3384 rx_ring
= adapter
->rx_ring
[i
];
3385 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)
3386 set_ring_rsc_enabled(rx_ring
);
3388 clear_ring_rsc_enabled(rx_ring
);
3392 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter
*adapter
)
3394 struct ixgbe_hw
*hw
= &adapter
->hw
;
3395 u32 rdrxctl
= IXGBE_READ_REG(hw
, IXGBE_RDRXCTL
);
3397 switch (hw
->mac
.type
) {
3398 case ixgbe_mac_82598EB
:
3400 * For VMDq support of different descriptor types or
3401 * buffer sizes through the use of multiple SRRCTL
3402 * registers, RDRXCTL.MVMEN must be set to 1
3404 * also, the manual doesn't mention it clearly but DCA hints
3405 * will only use queue 0's tags unless this bit is set. Side
3406 * effects of setting this bit are only that SRRCTL must be
3407 * fully programmed [0..15]
3409 rdrxctl
|= IXGBE_RDRXCTL_MVMEN
;
3411 case ixgbe_mac_82599EB
:
3412 case ixgbe_mac_X540
:
3413 /* Disable RSC for ACK packets */
3414 IXGBE_WRITE_REG(hw
, IXGBE_RSCDBU
,
3415 (IXGBE_RSCDBU_RSCACKDIS
| IXGBE_READ_REG(hw
, IXGBE_RSCDBU
)));
3416 rdrxctl
&= ~IXGBE_RDRXCTL_RSCFRSTSIZE
;
3417 /* hardware requires some bits to be set by default */
3418 rdrxctl
|= (IXGBE_RDRXCTL_RSCACKC
| IXGBE_RDRXCTL_FCOE_WRFIX
);
3419 rdrxctl
|= IXGBE_RDRXCTL_CRCSTRIP
;
3422 /* We should do nothing since we don't know this hardware */
3426 IXGBE_WRITE_REG(hw
, IXGBE_RDRXCTL
, rdrxctl
);
3430 * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3431 * @adapter: board private structure
3433 * Configure the Rx unit of the MAC after a reset.
3435 static void ixgbe_configure_rx(struct ixgbe_adapter
*adapter
)
3437 struct ixgbe_hw
*hw
= &adapter
->hw
;
3441 /* disable receives while setting up the descriptors */
3442 rxctrl
= IXGBE_READ_REG(hw
, IXGBE_RXCTRL
);
3443 IXGBE_WRITE_REG(hw
, IXGBE_RXCTRL
, rxctrl
& ~IXGBE_RXCTRL_RXEN
);
3445 ixgbe_setup_psrtype(adapter
);
3446 ixgbe_setup_rdrxctl(adapter
);
3448 /* Program registers for the distribution of queues */
3449 ixgbe_setup_mrqc(adapter
);
3451 /* set_rx_buffer_len must be called before ring initialization */
3452 ixgbe_set_rx_buffer_len(adapter
);
3455 * Setup the HW Rx Head and Tail Descriptor Pointers and
3456 * the Base and Length of the Rx Descriptor Ring
3458 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3459 ixgbe_configure_rx_ring(adapter
, adapter
->rx_ring
[i
]);
3461 /* disable drop enable for 82598 parts */
3462 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
3463 rxctrl
|= IXGBE_RXCTRL_DMBYPS
;
3465 /* enable all receives */
3466 rxctrl
|= IXGBE_RXCTRL_RXEN
;
3467 hw
->mac
.ops
.enable_rx_dma(hw
, rxctrl
);
3470 static int ixgbe_vlan_rx_add_vid(struct net_device
*netdev
,
3471 __be16 proto
, u16 vid
)
3473 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
3474 struct ixgbe_hw
*hw
= &adapter
->hw
;
3476 /* add VID to filter table */
3477 hw
->mac
.ops
.set_vfta(&adapter
->hw
, vid
, VMDQ_P(0), true);
3478 set_bit(vid
, adapter
->active_vlans
);
3483 static int ixgbe_vlan_rx_kill_vid(struct net_device
*netdev
,
3484 __be16 proto
, u16 vid
)
3486 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
3487 struct ixgbe_hw
*hw
= &adapter
->hw
;
3489 /* remove VID from filter table */
3490 hw
->mac
.ops
.set_vfta(&adapter
->hw
, vid
, VMDQ_P(0), false);
3491 clear_bit(vid
, adapter
->active_vlans
);
3497 * ixgbe_vlan_filter_disable - helper to disable hw vlan filtering
3498 * @adapter: driver data
3500 static void ixgbe_vlan_filter_disable(struct ixgbe_adapter
*adapter
)
3502 struct ixgbe_hw
*hw
= &adapter
->hw
;
3505 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_VLNCTRL
);
3506 vlnctrl
&= ~(IXGBE_VLNCTRL_VFE
| IXGBE_VLNCTRL_CFIEN
);
3507 IXGBE_WRITE_REG(hw
, IXGBE_VLNCTRL
, vlnctrl
);
3511 * ixgbe_vlan_filter_enable - helper to enable hw vlan filtering
3512 * @adapter: driver data
3514 static void ixgbe_vlan_filter_enable(struct ixgbe_adapter
*adapter
)
3516 struct ixgbe_hw
*hw
= &adapter
->hw
;
3519 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_VLNCTRL
);
3520 vlnctrl
|= IXGBE_VLNCTRL_VFE
;
3521 vlnctrl
&= ~IXGBE_VLNCTRL_CFIEN
;
3522 IXGBE_WRITE_REG(hw
, IXGBE_VLNCTRL
, vlnctrl
);
3526 * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3527 * @adapter: driver data
3529 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter
*adapter
)
3531 struct ixgbe_hw
*hw
= &adapter
->hw
;
3535 switch (hw
->mac
.type
) {
3536 case ixgbe_mac_82598EB
:
3537 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_VLNCTRL
);
3538 vlnctrl
&= ~IXGBE_VLNCTRL_VME
;
3539 IXGBE_WRITE_REG(hw
, IXGBE_VLNCTRL
, vlnctrl
);
3541 case ixgbe_mac_82599EB
:
3542 case ixgbe_mac_X540
:
3543 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3544 j
= adapter
->rx_ring
[i
]->reg_idx
;
3545 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(j
));
3546 vlnctrl
&= ~IXGBE_RXDCTL_VME
;
3547 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(j
), vlnctrl
);
3556 * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3557 * @adapter: driver data
3559 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter
*adapter
)
3561 struct ixgbe_hw
*hw
= &adapter
->hw
;
3565 switch (hw
->mac
.type
) {
3566 case ixgbe_mac_82598EB
:
3567 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_VLNCTRL
);
3568 vlnctrl
|= IXGBE_VLNCTRL_VME
;
3569 IXGBE_WRITE_REG(hw
, IXGBE_VLNCTRL
, vlnctrl
);
3571 case ixgbe_mac_82599EB
:
3572 case ixgbe_mac_X540
:
3573 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3574 j
= adapter
->rx_ring
[i
]->reg_idx
;
3575 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(j
));
3576 vlnctrl
|= IXGBE_RXDCTL_VME
;
3577 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(j
), vlnctrl
);
3585 static void ixgbe_restore_vlan(struct ixgbe_adapter
*adapter
)
3589 ixgbe_vlan_rx_add_vid(adapter
->netdev
, htons(ETH_P_8021Q
), 0);
3591 for_each_set_bit(vid
, adapter
->active_vlans
, VLAN_N_VID
)
3592 ixgbe_vlan_rx_add_vid(adapter
->netdev
, htons(ETH_P_8021Q
), vid
);
3596 * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
3597 * @netdev: network interface device structure
3599 * Writes unicast address list to the RAR table.
3600 * Returns: -ENOMEM on failure/insufficient address space
3601 * 0 on no addresses written
3602 * X on writing X addresses to the RAR table
3604 static int ixgbe_write_uc_addr_list(struct net_device
*netdev
)
3606 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
3607 struct ixgbe_hw
*hw
= &adapter
->hw
;
3608 unsigned int rar_entries
= hw
->mac
.num_rar_entries
- 1;
3611 /* In SR-IOV mode significantly less RAR entries are available */
3612 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
3613 rar_entries
= IXGBE_MAX_PF_MACVLANS
- 1;
3615 /* return ENOMEM indicating insufficient memory for addresses */
3616 if (netdev_uc_count(netdev
) > rar_entries
)
3619 if (!netdev_uc_empty(netdev
)) {
3620 struct netdev_hw_addr
*ha
;
3621 /* return error if we do not support writing to RAR table */
3622 if (!hw
->mac
.ops
.set_rar
)
3625 netdev_for_each_uc_addr(ha
, netdev
) {
3628 hw
->mac
.ops
.set_rar(hw
, rar_entries
--, ha
->addr
,
3629 VMDQ_P(0), IXGBE_RAH_AV
);
3633 /* write the addresses in reverse order to avoid write combining */
3634 for (; rar_entries
> 0 ; rar_entries
--)
3635 hw
->mac
.ops
.clear_rar(hw
, rar_entries
);
3641 * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
3642 * @netdev: network interface device structure
3644 * The set_rx_method entry point is called whenever the unicast/multicast
3645 * address list or the network interface flags are updated. This routine is
3646 * responsible for configuring the hardware for proper unicast, multicast and
3649 void ixgbe_set_rx_mode(struct net_device
*netdev
)
3651 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
3652 struct ixgbe_hw
*hw
= &adapter
->hw
;
3653 u32 fctrl
, vmolr
= IXGBE_VMOLR_BAM
| IXGBE_VMOLR_AUPE
;
3656 /* Check for Promiscuous and All Multicast modes */
3658 fctrl
= IXGBE_READ_REG(hw
, IXGBE_FCTRL
);
3660 /* set all bits that we expect to always be set */
3661 fctrl
&= ~IXGBE_FCTRL_SBP
; /* disable store-bad-packets */
3662 fctrl
|= IXGBE_FCTRL_BAM
;
3663 fctrl
|= IXGBE_FCTRL_DPF
; /* discard pause frames when FC enabled */
3664 fctrl
|= IXGBE_FCTRL_PMCF
;
3666 /* clear the bits we are changing the status of */
3667 fctrl
&= ~(IXGBE_FCTRL_UPE
| IXGBE_FCTRL_MPE
);
3669 if (netdev
->flags
& IFF_PROMISC
) {
3670 hw
->addr_ctrl
.user_set_promisc
= true;
3671 fctrl
|= (IXGBE_FCTRL_UPE
| IXGBE_FCTRL_MPE
);
3672 vmolr
|= (IXGBE_VMOLR_ROPE
| IXGBE_VMOLR_MPE
);
3673 /* don't hardware filter vlans in promisc mode */
3674 ixgbe_vlan_filter_disable(adapter
);
3676 if (netdev
->flags
& IFF_ALLMULTI
) {
3677 fctrl
|= IXGBE_FCTRL_MPE
;
3678 vmolr
|= IXGBE_VMOLR_MPE
;
3681 * Write addresses to the MTA, if the attempt fails
3682 * then we should just turn on promiscuous mode so
3683 * that we can at least receive multicast traffic
3685 hw
->mac
.ops
.update_mc_addr_list(hw
, netdev
);
3686 vmolr
|= IXGBE_VMOLR_ROMPE
;
3688 ixgbe_vlan_filter_enable(adapter
);
3689 hw
->addr_ctrl
.user_set_promisc
= false;
3693 * Write addresses to available RAR registers, if there is not
3694 * sufficient space to store all the addresses then enable
3695 * unicast promiscuous mode
3697 count
= ixgbe_write_uc_addr_list(netdev
);
3699 fctrl
|= IXGBE_FCTRL_UPE
;
3700 vmolr
|= IXGBE_VMOLR_ROPE
;
3703 if (adapter
->num_vfs
)
3704 ixgbe_restore_vf_multicasts(adapter
);
3706 if (hw
->mac
.type
!= ixgbe_mac_82598EB
) {
3707 vmolr
|= IXGBE_READ_REG(hw
, IXGBE_VMOLR(VMDQ_P(0))) &
3708 ~(IXGBE_VMOLR_MPE
| IXGBE_VMOLR_ROMPE
|
3710 IXGBE_WRITE_REG(hw
, IXGBE_VMOLR(VMDQ_P(0)), vmolr
);
3713 /* This is useful for sniffing bad packets. */
3714 if (adapter
->netdev
->features
& NETIF_F_RXALL
) {
3715 /* UPE and MPE will be handled by normal PROMISC logic
3716 * in e1000e_set_rx_mode */
3717 fctrl
|= (IXGBE_FCTRL_SBP
| /* Receive bad packets */
3718 IXGBE_FCTRL_BAM
| /* RX All Bcast Pkts */
3719 IXGBE_FCTRL_PMCF
); /* RX All MAC Ctrl Pkts */
3721 fctrl
&= ~(IXGBE_FCTRL_DPF
);
3722 /* NOTE: VLAN filtering is disabled by setting PROMISC */
3725 IXGBE_WRITE_REG(hw
, IXGBE_FCTRL
, fctrl
);
3727 if (netdev
->features
& NETIF_F_HW_VLAN_CTAG_RX
)
3728 ixgbe_vlan_strip_enable(adapter
);
3730 ixgbe_vlan_strip_disable(adapter
);
3733 static void ixgbe_napi_enable_all(struct ixgbe_adapter
*adapter
)
3737 for (q_idx
= 0; q_idx
< adapter
->num_q_vectors
; q_idx
++)
3738 napi_enable(&adapter
->q_vector
[q_idx
]->napi
);
3741 static void ixgbe_napi_disable_all(struct ixgbe_adapter
*adapter
)
3745 for (q_idx
= 0; q_idx
< adapter
->num_q_vectors
; q_idx
++)
3746 napi_disable(&adapter
->q_vector
[q_idx
]->napi
);
3749 #ifdef CONFIG_IXGBE_DCB
3751 * ixgbe_configure_dcb - Configure DCB hardware
3752 * @adapter: ixgbe adapter struct
3754 * This is called by the driver on open to configure the DCB hardware.
3755 * This is also called by the gennetlink interface when reconfiguring
3758 static void ixgbe_configure_dcb(struct ixgbe_adapter
*adapter
)
3760 struct ixgbe_hw
*hw
= &adapter
->hw
;
3761 int max_frame
= adapter
->netdev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
3763 if (!(adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
)) {
3764 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
3765 netif_set_gso_max_size(adapter
->netdev
, 65536);
3769 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
3770 netif_set_gso_max_size(adapter
->netdev
, 32768);
3773 if (adapter
->netdev
->features
& NETIF_F_FCOE_MTU
)
3774 max_frame
= max(max_frame
, IXGBE_FCOE_JUMBO_FRAME_SIZE
);
3777 /* reconfigure the hardware */
3778 if (adapter
->dcbx_cap
& DCB_CAP_DCBX_VER_CEE
) {
3779 ixgbe_dcb_calculate_tc_credits(hw
, &adapter
->dcb_cfg
, max_frame
,
3781 ixgbe_dcb_calculate_tc_credits(hw
, &adapter
->dcb_cfg
, max_frame
,
3783 ixgbe_dcb_hw_config(hw
, &adapter
->dcb_cfg
);
3784 } else if (adapter
->ixgbe_ieee_ets
&& adapter
->ixgbe_ieee_pfc
) {
3785 ixgbe_dcb_hw_ets(&adapter
->hw
,
3786 adapter
->ixgbe_ieee_ets
,
3788 ixgbe_dcb_hw_pfc_config(&adapter
->hw
,
3789 adapter
->ixgbe_ieee_pfc
->pfc_en
,
3790 adapter
->ixgbe_ieee_ets
->prio_tc
);
3793 /* Enable RSS Hash per TC */
3794 if (hw
->mac
.type
!= ixgbe_mac_82598EB
) {
3796 u16 rss_i
= adapter
->ring_feature
[RING_F_RSS
].indices
- 1;
3803 /* write msb to all 8 TCs in one write */
3804 IXGBE_WRITE_REG(hw
, IXGBE_RQTC
, msb
* 0x11111111);
3809 /* Additional bittime to account for IXGBE framing */
3810 #define IXGBE_ETH_FRAMING 20
3813 * ixgbe_hpbthresh - calculate high water mark for flow control
3815 * @adapter: board private structure to calculate for
3816 * @pb: packet buffer to calculate
3818 static int ixgbe_hpbthresh(struct ixgbe_adapter
*adapter
, int pb
)
3820 struct ixgbe_hw
*hw
= &adapter
->hw
;
3821 struct net_device
*dev
= adapter
->netdev
;
3822 int link
, tc
, kb
, marker
;
3825 /* Calculate max LAN frame size */
3826 tc
= link
= dev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
+ IXGBE_ETH_FRAMING
;
3829 /* FCoE traffic class uses FCOE jumbo frames */
3830 if ((dev
->features
& NETIF_F_FCOE_MTU
) &&
3831 (tc
< IXGBE_FCOE_JUMBO_FRAME_SIZE
) &&
3832 (pb
== ixgbe_fcoe_get_tc(adapter
)))
3833 tc
= IXGBE_FCOE_JUMBO_FRAME_SIZE
;
3836 /* Calculate delay value for device */
3837 switch (hw
->mac
.type
) {
3838 case ixgbe_mac_X540
:
3839 dv_id
= IXGBE_DV_X540(link
, tc
);
3842 dv_id
= IXGBE_DV(link
, tc
);
3846 /* Loopback switch introduces additional latency */
3847 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
3848 dv_id
+= IXGBE_B2BT(tc
);
3850 /* Delay value is calculated in bit times convert to KB */
3851 kb
= IXGBE_BT2KB(dv_id
);
3852 rx_pba
= IXGBE_READ_REG(hw
, IXGBE_RXPBSIZE(pb
)) >> 10;
3854 marker
= rx_pba
- kb
;
3856 /* It is possible that the packet buffer is not large enough
3857 * to provide required headroom. In this case throw an error
3858 * to user and a do the best we can.
3861 e_warn(drv
, "Packet Buffer(%i) can not provide enough"
3862 "headroom to support flow control."
3863 "Decrease MTU or number of traffic classes\n", pb
);
3871 * ixgbe_lpbthresh - calculate low water mark for for flow control
3873 * @adapter: board private structure to calculate for
3874 * @pb: packet buffer to calculate
3876 static int ixgbe_lpbthresh(struct ixgbe_adapter
*adapter
)
3878 struct ixgbe_hw
*hw
= &adapter
->hw
;
3879 struct net_device
*dev
= adapter
->netdev
;
3883 /* Calculate max LAN frame size */
3884 tc
= dev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
3886 /* Calculate delay value for device */
3887 switch (hw
->mac
.type
) {
3888 case ixgbe_mac_X540
:
3889 dv_id
= IXGBE_LOW_DV_X540(tc
);
3892 dv_id
= IXGBE_LOW_DV(tc
);
3896 /* Delay value is calculated in bit times convert to KB */
3897 return IXGBE_BT2KB(dv_id
);
3901 * ixgbe_pbthresh_setup - calculate and setup high low water marks
3903 static void ixgbe_pbthresh_setup(struct ixgbe_adapter
*adapter
)
3905 struct ixgbe_hw
*hw
= &adapter
->hw
;
3906 int num_tc
= netdev_get_num_tc(adapter
->netdev
);
3912 hw
->fc
.low_water
= ixgbe_lpbthresh(adapter
);
3914 for (i
= 0; i
< num_tc
; i
++) {
3915 hw
->fc
.high_water
[i
] = ixgbe_hpbthresh(adapter
, i
);
3917 /* Low water marks must not be larger than high water marks */
3918 if (hw
->fc
.low_water
> hw
->fc
.high_water
[i
])
3919 hw
->fc
.low_water
= 0;
3923 static void ixgbe_configure_pb(struct ixgbe_adapter
*adapter
)
3925 struct ixgbe_hw
*hw
= &adapter
->hw
;
3927 u8 tc
= netdev_get_num_tc(adapter
->netdev
);
3929 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
||
3930 adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
)
3931 hdrm
= 32 << adapter
->fdir_pballoc
;
3935 hw
->mac
.ops
.set_rxpba(hw
, tc
, hdrm
, PBA_STRATEGY_EQUAL
);
3936 ixgbe_pbthresh_setup(adapter
);
3939 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter
*adapter
)
3941 struct ixgbe_hw
*hw
= &adapter
->hw
;
3942 struct hlist_node
*node2
;
3943 struct ixgbe_fdir_filter
*filter
;
3945 spin_lock(&adapter
->fdir_perfect_lock
);
3947 if (!hlist_empty(&adapter
->fdir_filter_list
))
3948 ixgbe_fdir_set_input_mask_82599(hw
, &adapter
->fdir_mask
);
3950 hlist_for_each_entry_safe(filter
, node2
,
3951 &adapter
->fdir_filter_list
, fdir_node
) {
3952 ixgbe_fdir_write_perfect_filter_82599(hw
,
3955 (filter
->action
== IXGBE_FDIR_DROP_QUEUE
) ?
3956 IXGBE_FDIR_DROP_QUEUE
:
3957 adapter
->rx_ring
[filter
->action
]->reg_idx
);
3960 spin_unlock(&adapter
->fdir_perfect_lock
);
3963 static void ixgbe_configure(struct ixgbe_adapter
*adapter
)
3965 struct ixgbe_hw
*hw
= &adapter
->hw
;
3967 ixgbe_configure_pb(adapter
);
3968 #ifdef CONFIG_IXGBE_DCB
3969 ixgbe_configure_dcb(adapter
);
3972 * We must restore virtualization before VLANs or else
3973 * the VLVF registers will not be populated
3975 ixgbe_configure_virtualization(adapter
);
3977 ixgbe_set_rx_mode(adapter
->netdev
);
3978 ixgbe_restore_vlan(adapter
);
3980 switch (hw
->mac
.type
) {
3981 case ixgbe_mac_82599EB
:
3982 case ixgbe_mac_X540
:
3983 hw
->mac
.ops
.disable_rx_buff(hw
);
3989 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) {
3990 ixgbe_init_fdir_signature_82599(&adapter
->hw
,
3991 adapter
->fdir_pballoc
);
3992 } else if (adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
) {
3993 ixgbe_init_fdir_perfect_82599(&adapter
->hw
,
3994 adapter
->fdir_pballoc
);
3995 ixgbe_fdir_filter_restore(adapter
);
3998 switch (hw
->mac
.type
) {
3999 case ixgbe_mac_82599EB
:
4000 case ixgbe_mac_X540
:
4001 hw
->mac
.ops
.enable_rx_buff(hw
);
4008 /* configure FCoE L2 filters, redirection table, and Rx control */
4009 ixgbe_configure_fcoe(adapter
);
4011 #endif /* IXGBE_FCOE */
4012 ixgbe_configure_tx(adapter
);
4013 ixgbe_configure_rx(adapter
);
4016 static inline bool ixgbe_is_sfp(struct ixgbe_hw
*hw
)
4018 switch (hw
->phy
.type
) {
4019 case ixgbe_phy_sfp_avago
:
4020 case ixgbe_phy_sfp_ftl
:
4021 case ixgbe_phy_sfp_intel
:
4022 case ixgbe_phy_sfp_unknown
:
4023 case ixgbe_phy_sfp_passive_tyco
:
4024 case ixgbe_phy_sfp_passive_unknown
:
4025 case ixgbe_phy_sfp_active_unknown
:
4026 case ixgbe_phy_sfp_ftl_active
:
4029 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
4037 * ixgbe_sfp_link_config - set up SFP+ link
4038 * @adapter: pointer to private adapter struct
4040 static void ixgbe_sfp_link_config(struct ixgbe_adapter
*adapter
)
4043 * We are assuming the worst case scenario here, and that
4044 * is that an SFP was inserted/removed after the reset
4045 * but before SFP detection was enabled. As such the best
4046 * solution is to just start searching as soon as we start
4048 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
)
4049 adapter
->flags2
|= IXGBE_FLAG2_SEARCH_FOR_SFP
;
4051 adapter
->flags2
|= IXGBE_FLAG2_SFP_NEEDS_RESET
;
4055 * ixgbe_non_sfp_link_config - set up non-SFP+ link
4056 * @hw: pointer to private hardware struct
4058 * Returns 0 on success, negative on failure
4060 static int ixgbe_non_sfp_link_config(struct ixgbe_hw
*hw
)
4063 bool autoneg
, link_up
= false;
4064 u32 ret
= IXGBE_ERR_LINK_SETUP
;
4066 if (hw
->mac
.ops
.check_link
)
4067 ret
= hw
->mac
.ops
.check_link(hw
, &speed
, &link_up
, false);
4072 speed
= hw
->phy
.autoneg_advertised
;
4073 if ((!speed
) && (hw
->mac
.ops
.get_link_capabilities
))
4074 ret
= hw
->mac
.ops
.get_link_capabilities(hw
, &speed
,
4079 if (hw
->mac
.ops
.setup_link
)
4080 ret
= hw
->mac
.ops
.setup_link(hw
, speed
, link_up
);
4085 static void ixgbe_setup_gpie(struct ixgbe_adapter
*adapter
)
4087 struct ixgbe_hw
*hw
= &adapter
->hw
;
4090 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
) {
4091 gpie
= IXGBE_GPIE_MSIX_MODE
| IXGBE_GPIE_PBA_SUPPORT
|
4093 gpie
|= IXGBE_GPIE_EIAME
;
4095 * use EIAM to auto-mask when MSI-X interrupt is asserted
4096 * this saves a register write for every interrupt
4098 switch (hw
->mac
.type
) {
4099 case ixgbe_mac_82598EB
:
4100 IXGBE_WRITE_REG(hw
, IXGBE_EIAM
, IXGBE_EICS_RTX_QUEUE
);
4102 case ixgbe_mac_82599EB
:
4103 case ixgbe_mac_X540
:
4105 IXGBE_WRITE_REG(hw
, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4106 IXGBE_WRITE_REG(hw
, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4110 /* legacy interrupts, use EIAM to auto-mask when reading EICR,
4111 * specifically only auto mask tx and rx interrupts */
4112 IXGBE_WRITE_REG(hw
, IXGBE_EIAM
, IXGBE_EICS_RTX_QUEUE
);
4115 /* XXX: to interrupt immediately for EICS writes, enable this */
4116 /* gpie |= IXGBE_GPIE_EIMEN; */
4118 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
4119 gpie
&= ~IXGBE_GPIE_VTMODE_MASK
;
4121 switch (adapter
->ring_feature
[RING_F_VMDQ
].mask
) {
4122 case IXGBE_82599_VMDQ_8Q_MASK
:
4123 gpie
|= IXGBE_GPIE_VTMODE_16
;
4125 case IXGBE_82599_VMDQ_4Q_MASK
:
4126 gpie
|= IXGBE_GPIE_VTMODE_32
;
4129 gpie
|= IXGBE_GPIE_VTMODE_64
;
4134 /* Enable Thermal over heat sensor interrupt */
4135 if (adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
) {
4136 switch (adapter
->hw
.mac
.type
) {
4137 case ixgbe_mac_82599EB
:
4138 gpie
|= IXGBE_SDP0_GPIEN
;
4140 case ixgbe_mac_X540
:
4141 gpie
|= IXGBE_EIMS_TS
;
4148 /* Enable fan failure interrupt */
4149 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
)
4150 gpie
|= IXGBE_SDP1_GPIEN
;
4152 if (hw
->mac
.type
== ixgbe_mac_82599EB
) {
4153 gpie
|= IXGBE_SDP1_GPIEN
;
4154 gpie
|= IXGBE_SDP2_GPIEN
;
4157 IXGBE_WRITE_REG(hw
, IXGBE_GPIE
, gpie
);
4160 static void ixgbe_up_complete(struct ixgbe_adapter
*adapter
)
4162 struct ixgbe_hw
*hw
= &adapter
->hw
;
4166 ixgbe_get_hw_control(adapter
);
4167 ixgbe_setup_gpie(adapter
);
4169 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)
4170 ixgbe_configure_msix(adapter
);
4172 ixgbe_configure_msi_and_legacy(adapter
);
4174 /* enable the optics for 82599 SFP+ fiber */
4175 if (hw
->mac
.ops
.enable_tx_laser
)
4176 hw
->mac
.ops
.enable_tx_laser(hw
);
4178 clear_bit(__IXGBE_DOWN
, &adapter
->state
);
4179 ixgbe_napi_enable_all(adapter
);
4181 if (ixgbe_is_sfp(hw
)) {
4182 ixgbe_sfp_link_config(adapter
);
4184 err
= ixgbe_non_sfp_link_config(hw
);
4186 e_err(probe
, "link_config FAILED %d\n", err
);
4189 /* clear any pending interrupts, may auto mask */
4190 IXGBE_READ_REG(hw
, IXGBE_EICR
);
4191 ixgbe_irq_enable(adapter
, true, true);
4194 * If this adapter has a fan, check to see if we had a failure
4195 * before we enabled the interrupt.
4197 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
) {
4198 u32 esdp
= IXGBE_READ_REG(hw
, IXGBE_ESDP
);
4199 if (esdp
& IXGBE_ESDP_SDP1
)
4200 e_crit(drv
, "Fan has stopped, replace the adapter\n");
4203 /* enable transmits */
4204 netif_tx_start_all_queues(adapter
->netdev
);
4206 /* bring the link up in the watchdog, this could race with our first
4207 * link up interrupt but shouldn't be a problem */
4208 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_UPDATE
;
4209 adapter
->link_check_timeout
= jiffies
;
4210 mod_timer(&adapter
->service_timer
, jiffies
);
4212 /* Set PF Reset Done bit so PF/VF Mail Ops can work */
4213 ctrl_ext
= IXGBE_READ_REG(hw
, IXGBE_CTRL_EXT
);
4214 ctrl_ext
|= IXGBE_CTRL_EXT_PFRSTD
;
4215 IXGBE_WRITE_REG(hw
, IXGBE_CTRL_EXT
, ctrl_ext
);
4218 void ixgbe_reinit_locked(struct ixgbe_adapter
*adapter
)
4220 WARN_ON(in_interrupt());
4221 /* put off any impending NetWatchDogTimeout */
4222 adapter
->netdev
->trans_start
= jiffies
;
4224 while (test_and_set_bit(__IXGBE_RESETTING
, &adapter
->state
))
4225 usleep_range(1000, 2000);
4226 ixgbe_down(adapter
);
4228 * If SR-IOV enabled then wait a bit before bringing the adapter
4229 * back up to give the VFs time to respond to the reset. The
4230 * two second wait is based upon the watchdog timer cycle in
4233 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
4236 clear_bit(__IXGBE_RESETTING
, &adapter
->state
);
4239 void ixgbe_up(struct ixgbe_adapter
*adapter
)
4241 /* hardware has been reset, we need to reload some things */
4242 ixgbe_configure(adapter
);
4244 ixgbe_up_complete(adapter
);
4247 void ixgbe_reset(struct ixgbe_adapter
*adapter
)
4249 struct ixgbe_hw
*hw
= &adapter
->hw
;
4252 /* lock SFP init bit to prevent race conditions with the watchdog */
4253 while (test_and_set_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
))
4254 usleep_range(1000, 2000);
4256 /* clear all SFP and link config related flags while holding SFP_INIT */
4257 adapter
->flags2
&= ~(IXGBE_FLAG2_SEARCH_FOR_SFP
|
4258 IXGBE_FLAG2_SFP_NEEDS_RESET
);
4259 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_CONFIG
;
4261 err
= hw
->mac
.ops
.init_hw(hw
);
4264 case IXGBE_ERR_SFP_NOT_PRESENT
:
4265 case IXGBE_ERR_SFP_NOT_SUPPORTED
:
4267 case IXGBE_ERR_MASTER_REQUESTS_PENDING
:
4268 e_dev_err("master disable timed out\n");
4270 case IXGBE_ERR_EEPROM_VERSION
:
4271 /* We are running on a pre-production device, log a warning */
4272 e_dev_warn("This device is a pre-production adapter/LOM. "
4273 "Please be aware there may be issues associated with "
4274 "your hardware. If you are experiencing problems "
4275 "please contact your Intel or hardware "
4276 "representative who provided you with this "
4280 e_dev_err("Hardware Error: %d\n", err
);
4283 clear_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
);
4285 /* reprogram the RAR[0] in case user changed it. */
4286 hw
->mac
.ops
.set_rar(hw
, 0, hw
->mac
.addr
, VMDQ_P(0), IXGBE_RAH_AV
);
4288 /* update SAN MAC vmdq pool selection */
4289 if (hw
->mac
.san_mac_rar_index
)
4290 hw
->mac
.ops
.set_vmdq_san_mac(hw
, VMDQ_P(0));
4292 if (adapter
->flags2
& IXGBE_FLAG2_PTP_ENABLED
)
4293 ixgbe_ptp_reset(adapter
);
4297 * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4298 * @rx_ring: ring to free buffers from
4300 static void ixgbe_clean_rx_ring(struct ixgbe_ring
*rx_ring
)
4302 struct device
*dev
= rx_ring
->dev
;
4306 /* ring already cleared, nothing to do */
4307 if (!rx_ring
->rx_buffer_info
)
4310 /* Free all the Rx ring sk_buffs */
4311 for (i
= 0; i
< rx_ring
->count
; i
++) {
4312 struct ixgbe_rx_buffer
*rx_buffer
;
4314 rx_buffer
= &rx_ring
->rx_buffer_info
[i
];
4315 if (rx_buffer
->skb
) {
4316 struct sk_buff
*skb
= rx_buffer
->skb
;
4317 if (IXGBE_CB(skb
)->page_released
) {
4320 ixgbe_rx_bufsz(rx_ring
),
4322 IXGBE_CB(skb
)->page_released
= false;
4326 rx_buffer
->skb
= NULL
;
4328 dma_unmap_page(dev
, rx_buffer
->dma
,
4329 ixgbe_rx_pg_size(rx_ring
),
4332 if (rx_buffer
->page
)
4333 __free_pages(rx_buffer
->page
,
4334 ixgbe_rx_pg_order(rx_ring
));
4335 rx_buffer
->page
= NULL
;
4338 size
= sizeof(struct ixgbe_rx_buffer
) * rx_ring
->count
;
4339 memset(rx_ring
->rx_buffer_info
, 0, size
);
4341 /* Zero out the descriptor ring */
4342 memset(rx_ring
->desc
, 0, rx_ring
->size
);
4344 rx_ring
->next_to_alloc
= 0;
4345 rx_ring
->next_to_clean
= 0;
4346 rx_ring
->next_to_use
= 0;
4350 * ixgbe_clean_tx_ring - Free Tx Buffers
4351 * @tx_ring: ring to be cleaned
4353 static void ixgbe_clean_tx_ring(struct ixgbe_ring
*tx_ring
)
4355 struct ixgbe_tx_buffer
*tx_buffer_info
;
4359 /* ring already cleared, nothing to do */
4360 if (!tx_ring
->tx_buffer_info
)
4363 /* Free all the Tx ring sk_buffs */
4364 for (i
= 0; i
< tx_ring
->count
; i
++) {
4365 tx_buffer_info
= &tx_ring
->tx_buffer_info
[i
];
4366 ixgbe_unmap_and_free_tx_resource(tx_ring
, tx_buffer_info
);
4369 netdev_tx_reset_queue(txring_txq(tx_ring
));
4371 size
= sizeof(struct ixgbe_tx_buffer
) * tx_ring
->count
;
4372 memset(tx_ring
->tx_buffer_info
, 0, size
);
4374 /* Zero out the descriptor ring */
4375 memset(tx_ring
->desc
, 0, tx_ring
->size
);
4377 tx_ring
->next_to_use
= 0;
4378 tx_ring
->next_to_clean
= 0;
4382 * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
4383 * @adapter: board private structure
4385 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter
*adapter
)
4389 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
4390 ixgbe_clean_rx_ring(adapter
->rx_ring
[i
]);
4394 * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
4395 * @adapter: board private structure
4397 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter
*adapter
)
4401 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
4402 ixgbe_clean_tx_ring(adapter
->tx_ring
[i
]);
4405 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter
*adapter
)
4407 struct hlist_node
*node2
;
4408 struct ixgbe_fdir_filter
*filter
;
4410 spin_lock(&adapter
->fdir_perfect_lock
);
4412 hlist_for_each_entry_safe(filter
, node2
,
4413 &adapter
->fdir_filter_list
, fdir_node
) {
4414 hlist_del(&filter
->fdir_node
);
4417 adapter
->fdir_filter_count
= 0;
4419 spin_unlock(&adapter
->fdir_perfect_lock
);
4422 void ixgbe_down(struct ixgbe_adapter
*adapter
)
4424 struct net_device
*netdev
= adapter
->netdev
;
4425 struct ixgbe_hw
*hw
= &adapter
->hw
;
4429 /* signal that we are down to the interrupt handler */
4430 set_bit(__IXGBE_DOWN
, &adapter
->state
);
4432 /* disable receives */
4433 rxctrl
= IXGBE_READ_REG(hw
, IXGBE_RXCTRL
);
4434 IXGBE_WRITE_REG(hw
, IXGBE_RXCTRL
, rxctrl
& ~IXGBE_RXCTRL_RXEN
);
4436 /* disable all enabled rx queues */
4437 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
4438 /* this call also flushes the previous write */
4439 ixgbe_disable_rx_queue(adapter
, adapter
->rx_ring
[i
]);
4441 usleep_range(10000, 20000);
4443 netif_tx_stop_all_queues(netdev
);
4445 /* call carrier off first to avoid false dev_watchdog timeouts */
4446 netif_carrier_off(netdev
);
4447 netif_tx_disable(netdev
);
4449 ixgbe_irq_disable(adapter
);
4451 ixgbe_napi_disable_all(adapter
);
4453 adapter
->flags2
&= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT
|
4454 IXGBE_FLAG2_RESET_REQUESTED
);
4455 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_UPDATE
;
4457 del_timer_sync(&adapter
->service_timer
);
4459 if (adapter
->num_vfs
) {
4460 /* Clear EITR Select mapping */
4461 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EITRSEL
, 0);
4463 /* Mark all the VFs as inactive */
4464 for (i
= 0 ; i
< adapter
->num_vfs
; i
++)
4465 adapter
->vfinfo
[i
].clear_to_send
= false;
4467 /* ping all the active vfs to let them know we are going down */
4468 ixgbe_ping_all_vfs(adapter
);
4470 /* Disable all VFTE/VFRE TX/RX */
4471 ixgbe_disable_tx_rx(adapter
);
4474 /* disable transmits in the hardware now that interrupts are off */
4475 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
4476 u8 reg_idx
= adapter
->tx_ring
[i
]->reg_idx
;
4477 IXGBE_WRITE_REG(hw
, IXGBE_TXDCTL(reg_idx
), IXGBE_TXDCTL_SWFLSH
);
4480 /* Disable the Tx DMA engine on 82599 and X540 */
4481 switch (hw
->mac
.type
) {
4482 case ixgbe_mac_82599EB
:
4483 case ixgbe_mac_X540
:
4484 IXGBE_WRITE_REG(hw
, IXGBE_DMATXCTL
,
4485 (IXGBE_READ_REG(hw
, IXGBE_DMATXCTL
) &
4486 ~IXGBE_DMATXCTL_TE
));
4492 if (!pci_channel_offline(adapter
->pdev
))
4493 ixgbe_reset(adapter
);
4495 /* power down the optics for 82599 SFP+ fiber */
4496 if (hw
->mac
.ops
.disable_tx_laser
)
4497 hw
->mac
.ops
.disable_tx_laser(hw
);
4499 ixgbe_clean_all_tx_rings(adapter
);
4500 ixgbe_clean_all_rx_rings(adapter
);
4502 #ifdef CONFIG_IXGBE_DCA
4503 /* since we reset the hardware DCA settings were cleared */
4504 ixgbe_setup_dca(adapter
);
4509 * ixgbe_tx_timeout - Respond to a Tx Hang
4510 * @netdev: network interface device structure
4512 static void ixgbe_tx_timeout(struct net_device
*netdev
)
4514 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
4516 /* Do the reset outside of interrupt context */
4517 ixgbe_tx_timeout_reset(adapter
);
4521 * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
4522 * @adapter: board private structure to initialize
4524 * ixgbe_sw_init initializes the Adapter private data structure.
4525 * Fields are initialized based on PCI device information and
4526 * OS network device settings (MTU size).
4528 static int ixgbe_sw_init(struct ixgbe_adapter
*adapter
)
4530 struct ixgbe_hw
*hw
= &adapter
->hw
;
4531 struct pci_dev
*pdev
= adapter
->pdev
;
4532 unsigned int rss
, fdir
;
4534 #ifdef CONFIG_IXGBE_DCB
4536 struct tc_configuration
*tc
;
4539 /* PCI config space info */
4541 hw
->vendor_id
= pdev
->vendor
;
4542 hw
->device_id
= pdev
->device
;
4543 hw
->revision_id
= pdev
->revision
;
4544 hw
->subsystem_vendor_id
= pdev
->subsystem_vendor
;
4545 hw
->subsystem_device_id
= pdev
->subsystem_device
;
4547 /* Set common capability flags and settings */
4548 rss
= min_t(int, IXGBE_MAX_RSS_INDICES
, num_online_cpus());
4549 adapter
->ring_feature
[RING_F_RSS
].limit
= rss
;
4550 adapter
->flags2
|= IXGBE_FLAG2_RSC_CAPABLE
;
4551 adapter
->flags2
|= IXGBE_FLAG2_RSC_ENABLED
;
4552 adapter
->max_q_vectors
= MAX_Q_VECTORS_82599
;
4553 adapter
->atr_sample_rate
= 20;
4554 fdir
= min_t(int, IXGBE_MAX_FDIR_INDICES
, num_online_cpus());
4555 adapter
->ring_feature
[RING_F_FDIR
].limit
= fdir
;
4556 adapter
->fdir_pballoc
= IXGBE_FDIR_PBALLOC_64K
;
4557 #ifdef CONFIG_IXGBE_DCA
4558 adapter
->flags
|= IXGBE_FLAG_DCA_CAPABLE
;
4561 adapter
->flags
|= IXGBE_FLAG_FCOE_CAPABLE
;
4562 adapter
->flags
&= ~IXGBE_FLAG_FCOE_ENABLED
;
4563 #ifdef CONFIG_IXGBE_DCB
4564 /* Default traffic class to use for FCoE */
4565 adapter
->fcoe
.up
= IXGBE_FCOE_DEFTC
;
4566 #endif /* CONFIG_IXGBE_DCB */
4567 #endif /* IXGBE_FCOE */
4569 /* Set MAC specific capability flags and exceptions */
4570 switch (hw
->mac
.type
) {
4571 case ixgbe_mac_82598EB
:
4572 adapter
->flags2
&= ~IXGBE_FLAG2_RSC_CAPABLE
;
4573 adapter
->flags2
&= ~IXGBE_FLAG2_RSC_ENABLED
;
4575 if (hw
->device_id
== IXGBE_DEV_ID_82598AT
)
4576 adapter
->flags
|= IXGBE_FLAG_FAN_FAIL_CAPABLE
;
4578 adapter
->max_q_vectors
= MAX_Q_VECTORS_82598
;
4579 adapter
->ring_feature
[RING_F_FDIR
].limit
= 0;
4580 adapter
->atr_sample_rate
= 0;
4581 adapter
->fdir_pballoc
= 0;
4583 adapter
->flags
&= ~IXGBE_FLAG_FCOE_CAPABLE
;
4584 adapter
->flags
&= ~IXGBE_FLAG_FCOE_ENABLED
;
4585 #ifdef CONFIG_IXGBE_DCB
4586 adapter
->fcoe
.up
= 0;
4587 #endif /* IXGBE_DCB */
4588 #endif /* IXGBE_FCOE */
4590 case ixgbe_mac_82599EB
:
4591 if (hw
->device_id
== IXGBE_DEV_ID_82599_T3_LOM
)
4592 adapter
->flags2
|= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
;
4594 case ixgbe_mac_X540
:
4595 fwsm
= IXGBE_READ_REG(hw
, IXGBE_FWSM
);
4596 if (fwsm
& IXGBE_FWSM_TS_ENABLED
)
4597 adapter
->flags2
|= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
;
4604 /* FCoE support exists, always init the FCoE lock */
4605 spin_lock_init(&adapter
->fcoe
.lock
);
4608 /* n-tuple support exists, always init our spinlock */
4609 spin_lock_init(&adapter
->fdir_perfect_lock
);
4611 #ifdef CONFIG_IXGBE_DCB
4612 switch (hw
->mac
.type
) {
4613 case ixgbe_mac_X540
:
4614 adapter
->dcb_cfg
.num_tcs
.pg_tcs
= X540_TRAFFIC_CLASS
;
4615 adapter
->dcb_cfg
.num_tcs
.pfc_tcs
= X540_TRAFFIC_CLASS
;
4618 adapter
->dcb_cfg
.num_tcs
.pg_tcs
= MAX_TRAFFIC_CLASS
;
4619 adapter
->dcb_cfg
.num_tcs
.pfc_tcs
= MAX_TRAFFIC_CLASS
;
4623 /* Configure DCB traffic classes */
4624 for (j
= 0; j
< MAX_TRAFFIC_CLASS
; j
++) {
4625 tc
= &adapter
->dcb_cfg
.tc_config
[j
];
4626 tc
->path
[DCB_TX_CONFIG
].bwg_id
= 0;
4627 tc
->path
[DCB_TX_CONFIG
].bwg_percent
= 12 + (j
& 1);
4628 tc
->path
[DCB_RX_CONFIG
].bwg_id
= 0;
4629 tc
->path
[DCB_RX_CONFIG
].bwg_percent
= 12 + (j
& 1);
4630 tc
->dcb_pfc
= pfc_disabled
;
4633 /* Initialize default user to priority mapping, UPx->TC0 */
4634 tc
= &adapter
->dcb_cfg
.tc_config
[0];
4635 tc
->path
[DCB_TX_CONFIG
].up_to_tc_bitmap
= 0xFF;
4636 tc
->path
[DCB_RX_CONFIG
].up_to_tc_bitmap
= 0xFF;
4638 adapter
->dcb_cfg
.bw_percentage
[DCB_TX_CONFIG
][0] = 100;
4639 adapter
->dcb_cfg
.bw_percentage
[DCB_RX_CONFIG
][0] = 100;
4640 adapter
->dcb_cfg
.pfc_mode_enable
= false;
4641 adapter
->dcb_set_bitmap
= 0x00;
4642 adapter
->dcbx_cap
= DCB_CAP_DCBX_HOST
| DCB_CAP_DCBX_VER_CEE
;
4643 memcpy(&adapter
->temp_dcb_cfg
, &adapter
->dcb_cfg
,
4644 sizeof(adapter
->temp_dcb_cfg
));
4648 /* default flow control settings */
4649 hw
->fc
.requested_mode
= ixgbe_fc_full
;
4650 hw
->fc
.current_mode
= ixgbe_fc_full
; /* init for ethtool output */
4651 ixgbe_pbthresh_setup(adapter
);
4652 hw
->fc
.pause_time
= IXGBE_DEFAULT_FCPAUSE
;
4653 hw
->fc
.send_xon
= true;
4654 hw
->fc
.disable_fc_autoneg
=
4655 (ixgbe_device_supports_autoneg_fc(hw
) == 0) ? false : true;
4657 #ifdef CONFIG_PCI_IOV
4658 /* assign number of SR-IOV VFs */
4659 if (hw
->mac
.type
!= ixgbe_mac_82598EB
)
4660 adapter
->num_vfs
= (max_vfs
> 63) ? 0 : max_vfs
;
4663 /* enable itr by default in dynamic mode */
4664 adapter
->rx_itr_setting
= 1;
4665 adapter
->tx_itr_setting
= 1;
4667 /* set default ring sizes */
4668 adapter
->tx_ring_count
= IXGBE_DEFAULT_TXD
;
4669 adapter
->rx_ring_count
= IXGBE_DEFAULT_RXD
;
4671 /* set default work limits */
4672 adapter
->tx_work_limit
= IXGBE_DEFAULT_TX_WORK
;
4674 /* initialize eeprom parameters */
4675 if (ixgbe_init_eeprom_params_generic(hw
)) {
4676 e_dev_err("EEPROM initialization failed\n");
4680 set_bit(__IXGBE_DOWN
, &adapter
->state
);
4686 * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
4687 * @tx_ring: tx descriptor ring (for a specific queue) to setup
4689 * Return 0 on success, negative on failure
4691 int ixgbe_setup_tx_resources(struct ixgbe_ring
*tx_ring
)
4693 struct device
*dev
= tx_ring
->dev
;
4694 int orig_node
= dev_to_node(dev
);
4698 size
= sizeof(struct ixgbe_tx_buffer
) * tx_ring
->count
;
4700 if (tx_ring
->q_vector
)
4701 numa_node
= tx_ring
->q_vector
->numa_node
;
4703 tx_ring
->tx_buffer_info
= vzalloc_node(size
, numa_node
);
4704 if (!tx_ring
->tx_buffer_info
)
4705 tx_ring
->tx_buffer_info
= vzalloc(size
);
4706 if (!tx_ring
->tx_buffer_info
)
4709 /* round up to nearest 4K */
4710 tx_ring
->size
= tx_ring
->count
* sizeof(union ixgbe_adv_tx_desc
);
4711 tx_ring
->size
= ALIGN(tx_ring
->size
, 4096);
4713 set_dev_node(dev
, numa_node
);
4714 tx_ring
->desc
= dma_alloc_coherent(dev
,
4718 set_dev_node(dev
, orig_node
);
4720 tx_ring
->desc
= dma_alloc_coherent(dev
, tx_ring
->size
,
4721 &tx_ring
->dma
, GFP_KERNEL
);
4725 tx_ring
->next_to_use
= 0;
4726 tx_ring
->next_to_clean
= 0;
4730 vfree(tx_ring
->tx_buffer_info
);
4731 tx_ring
->tx_buffer_info
= NULL
;
4732 dev_err(dev
, "Unable to allocate memory for the Tx descriptor ring\n");
4737 * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
4738 * @adapter: board private structure
4740 * If this function returns with an error, then it's possible one or
4741 * more of the rings is populated (while the rest are not). It is the
4742 * callers duty to clean those orphaned rings.
4744 * Return 0 on success, negative on failure
4746 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter
*adapter
)
4750 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
4751 err
= ixgbe_setup_tx_resources(adapter
->tx_ring
[i
]);
4755 e_err(probe
, "Allocation for Tx Queue %u failed\n", i
);
4761 /* rewind the index freeing the rings as we go */
4763 ixgbe_free_tx_resources(adapter
->tx_ring
[i
]);
4768 * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
4769 * @rx_ring: rx descriptor ring (for a specific queue) to setup
4771 * Returns 0 on success, negative on failure
4773 int ixgbe_setup_rx_resources(struct ixgbe_ring
*rx_ring
)
4775 struct device
*dev
= rx_ring
->dev
;
4776 int orig_node
= dev_to_node(dev
);
4780 size
= sizeof(struct ixgbe_rx_buffer
) * rx_ring
->count
;
4782 if (rx_ring
->q_vector
)
4783 numa_node
= rx_ring
->q_vector
->numa_node
;
4785 rx_ring
->rx_buffer_info
= vzalloc_node(size
, numa_node
);
4786 if (!rx_ring
->rx_buffer_info
)
4787 rx_ring
->rx_buffer_info
= vzalloc(size
);
4788 if (!rx_ring
->rx_buffer_info
)
4791 /* Round up to nearest 4K */
4792 rx_ring
->size
= rx_ring
->count
* sizeof(union ixgbe_adv_rx_desc
);
4793 rx_ring
->size
= ALIGN(rx_ring
->size
, 4096);
4795 set_dev_node(dev
, numa_node
);
4796 rx_ring
->desc
= dma_alloc_coherent(dev
,
4800 set_dev_node(dev
, orig_node
);
4802 rx_ring
->desc
= dma_alloc_coherent(dev
, rx_ring
->size
,
4803 &rx_ring
->dma
, GFP_KERNEL
);
4807 rx_ring
->next_to_clean
= 0;
4808 rx_ring
->next_to_use
= 0;
4812 vfree(rx_ring
->rx_buffer_info
);
4813 rx_ring
->rx_buffer_info
= NULL
;
4814 dev_err(dev
, "Unable to allocate memory for the Rx descriptor ring\n");
4819 * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
4820 * @adapter: board private structure
4822 * If this function returns with an error, then it's possible one or
4823 * more of the rings is populated (while the rest are not). It is the
4824 * callers duty to clean those orphaned rings.
4826 * Return 0 on success, negative on failure
4828 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter
*adapter
)
4832 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
4833 err
= ixgbe_setup_rx_resources(adapter
->rx_ring
[i
]);
4837 e_err(probe
, "Allocation for Rx Queue %u failed\n", i
);
4842 err
= ixgbe_setup_fcoe_ddp_resources(adapter
);
4847 /* rewind the index freeing the rings as we go */
4849 ixgbe_free_rx_resources(adapter
->rx_ring
[i
]);
4854 * ixgbe_free_tx_resources - Free Tx Resources per Queue
4855 * @tx_ring: Tx descriptor ring for a specific queue
4857 * Free all transmit software resources
4859 void ixgbe_free_tx_resources(struct ixgbe_ring
*tx_ring
)
4861 ixgbe_clean_tx_ring(tx_ring
);
4863 vfree(tx_ring
->tx_buffer_info
);
4864 tx_ring
->tx_buffer_info
= NULL
;
4866 /* if not set, then don't free */
4870 dma_free_coherent(tx_ring
->dev
, tx_ring
->size
,
4871 tx_ring
->desc
, tx_ring
->dma
);
4873 tx_ring
->desc
= NULL
;
4877 * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
4878 * @adapter: board private structure
4880 * Free all transmit software resources
4882 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter
*adapter
)
4886 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
4887 if (adapter
->tx_ring
[i
]->desc
)
4888 ixgbe_free_tx_resources(adapter
->tx_ring
[i
]);
4892 * ixgbe_free_rx_resources - Free Rx Resources
4893 * @rx_ring: ring to clean the resources from
4895 * Free all receive software resources
4897 void ixgbe_free_rx_resources(struct ixgbe_ring
*rx_ring
)
4899 ixgbe_clean_rx_ring(rx_ring
);
4901 vfree(rx_ring
->rx_buffer_info
);
4902 rx_ring
->rx_buffer_info
= NULL
;
4904 /* if not set, then don't free */
4908 dma_free_coherent(rx_ring
->dev
, rx_ring
->size
,
4909 rx_ring
->desc
, rx_ring
->dma
);
4911 rx_ring
->desc
= NULL
;
4915 * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
4916 * @adapter: board private structure
4918 * Free all receive software resources
4920 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter
*adapter
)
4925 ixgbe_free_fcoe_ddp_resources(adapter
);
4928 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
4929 if (adapter
->rx_ring
[i
]->desc
)
4930 ixgbe_free_rx_resources(adapter
->rx_ring
[i
]);
4934 * ixgbe_change_mtu - Change the Maximum Transfer Unit
4935 * @netdev: network interface device structure
4936 * @new_mtu: new value for maximum frame size
4938 * Returns 0 on success, negative on failure
4940 static int ixgbe_change_mtu(struct net_device
*netdev
, int new_mtu
)
4942 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
4943 int max_frame
= new_mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
4945 /* MTU < 68 is an error and causes problems on some kernels */
4946 if ((new_mtu
< 68) || (max_frame
> IXGBE_MAX_JUMBO_FRAME_SIZE
))
4950 * For 82599EB we cannot allow legacy VFs to enable their receive
4951 * paths when MTU greater than 1500 is configured. So display a
4952 * warning that legacy VFs will be disabled.
4954 if ((adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) &&
4955 (adapter
->hw
.mac
.type
== ixgbe_mac_82599EB
) &&
4956 (max_frame
> (ETH_FRAME_LEN
+ ETH_FCS_LEN
)))
4957 e_warn(probe
, "Setting MTU > 1500 will disable legacy VFs\n");
4959 e_info(probe
, "changing MTU from %d to %d\n", netdev
->mtu
, new_mtu
);
4961 /* must set new MTU before calling down or up */
4962 netdev
->mtu
= new_mtu
;
4964 if (netif_running(netdev
))
4965 ixgbe_reinit_locked(adapter
);
4971 * ixgbe_open - Called when a network interface is made active
4972 * @netdev: network interface device structure
4974 * Returns 0 on success, negative value on failure
4976 * The open entry point is called when a network interface is made
4977 * active by the system (IFF_UP). At this point all resources needed
4978 * for transmit and receive operations are allocated, the interrupt
4979 * handler is registered with the OS, the watchdog timer is started,
4980 * and the stack is notified that the interface is ready.
4982 static int ixgbe_open(struct net_device
*netdev
)
4984 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
4987 /* disallow open during test */
4988 if (test_bit(__IXGBE_TESTING
, &adapter
->state
))
4991 netif_carrier_off(netdev
);
4993 /* allocate transmit descriptors */
4994 err
= ixgbe_setup_all_tx_resources(adapter
);
4998 /* allocate receive descriptors */
4999 err
= ixgbe_setup_all_rx_resources(adapter
);
5003 ixgbe_configure(adapter
);
5005 err
= ixgbe_request_irq(adapter
);
5009 /* Notify the stack of the actual queue counts. */
5010 err
= netif_set_real_num_tx_queues(netdev
,
5011 adapter
->num_rx_pools
> 1 ? 1 :
5012 adapter
->num_tx_queues
);
5014 goto err_set_queues
;
5017 err
= netif_set_real_num_rx_queues(netdev
,
5018 adapter
->num_rx_pools
> 1 ? 1 :
5019 adapter
->num_rx_queues
);
5021 goto err_set_queues
;
5023 ixgbe_ptp_init(adapter
);
5025 ixgbe_up_complete(adapter
);
5030 ixgbe_free_irq(adapter
);
5032 ixgbe_free_all_rx_resources(adapter
);
5034 ixgbe_free_all_tx_resources(adapter
);
5036 ixgbe_reset(adapter
);
5042 * ixgbe_close - Disables a network interface
5043 * @netdev: network interface device structure
5045 * Returns 0, this is not allowed to fail
5047 * The close entry point is called when an interface is de-activated
5048 * by the OS. The hardware is still under the drivers control, but
5049 * needs to be disabled. A global MAC reset is issued to stop the
5050 * hardware, and all transmit and receive resources are freed.
5052 static int ixgbe_close(struct net_device
*netdev
)
5054 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
5056 ixgbe_ptp_stop(adapter
);
5058 ixgbe_down(adapter
);
5059 ixgbe_free_irq(adapter
);
5061 ixgbe_fdir_filter_exit(adapter
);
5063 ixgbe_free_all_tx_resources(adapter
);
5064 ixgbe_free_all_rx_resources(adapter
);
5066 ixgbe_release_hw_control(adapter
);
5072 static int ixgbe_resume(struct pci_dev
*pdev
)
5074 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
5075 struct net_device
*netdev
= adapter
->netdev
;
5078 pci_set_power_state(pdev
, PCI_D0
);
5079 pci_restore_state(pdev
);
5081 * pci_restore_state clears dev->state_saved so call
5082 * pci_save_state to restore it.
5084 pci_save_state(pdev
);
5086 err
= pci_enable_device_mem(pdev
);
5088 e_dev_err("Cannot enable PCI device from suspend\n");
5091 pci_set_master(pdev
);
5093 pci_wake_from_d3(pdev
, false);
5095 ixgbe_reset(adapter
);
5097 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_WUS
, ~0);
5100 err
= ixgbe_init_interrupt_scheme(adapter
);
5101 if (!err
&& netif_running(netdev
))
5102 err
= ixgbe_open(netdev
);
5109 netif_device_attach(netdev
);
5113 #endif /* CONFIG_PM */
5115 static int __ixgbe_shutdown(struct pci_dev
*pdev
, bool *enable_wake
)
5117 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
5118 struct net_device
*netdev
= adapter
->netdev
;
5119 struct ixgbe_hw
*hw
= &adapter
->hw
;
5121 u32 wufc
= adapter
->wol
;
5126 netif_device_detach(netdev
);
5129 if (netif_running(netdev
)) {
5130 ixgbe_down(adapter
);
5131 ixgbe_free_irq(adapter
);
5132 ixgbe_free_all_tx_resources(adapter
);
5133 ixgbe_free_all_rx_resources(adapter
);
5137 ixgbe_clear_interrupt_scheme(adapter
);
5140 retval
= pci_save_state(pdev
);
5146 ixgbe_set_rx_mode(netdev
);
5148 /* enable the optics for 82599 SFP+ fiber as we can WoL */
5149 if (hw
->mac
.ops
.enable_tx_laser
)
5150 hw
->mac
.ops
.enable_tx_laser(hw
);
5152 /* turn on all-multi mode if wake on multicast is enabled */
5153 if (wufc
& IXGBE_WUFC_MC
) {
5154 fctrl
= IXGBE_READ_REG(hw
, IXGBE_FCTRL
);
5155 fctrl
|= IXGBE_FCTRL_MPE
;
5156 IXGBE_WRITE_REG(hw
, IXGBE_FCTRL
, fctrl
);
5159 ctrl
= IXGBE_READ_REG(hw
, IXGBE_CTRL
);
5160 ctrl
|= IXGBE_CTRL_GIO_DIS
;
5161 IXGBE_WRITE_REG(hw
, IXGBE_CTRL
, ctrl
);
5163 IXGBE_WRITE_REG(hw
, IXGBE_WUFC
, wufc
);
5165 IXGBE_WRITE_REG(hw
, IXGBE_WUC
, 0);
5166 IXGBE_WRITE_REG(hw
, IXGBE_WUFC
, 0);
5169 switch (hw
->mac
.type
) {
5170 case ixgbe_mac_82598EB
:
5171 pci_wake_from_d3(pdev
, false);
5173 case ixgbe_mac_82599EB
:
5174 case ixgbe_mac_X540
:
5175 pci_wake_from_d3(pdev
, !!wufc
);
5181 *enable_wake
= !!wufc
;
5183 ixgbe_release_hw_control(adapter
);
5185 pci_disable_device(pdev
);
5191 static int ixgbe_suspend(struct pci_dev
*pdev
, pm_message_t state
)
5196 retval
= __ixgbe_shutdown(pdev
, &wake
);
5201 pci_prepare_to_sleep(pdev
);
5203 pci_wake_from_d3(pdev
, false);
5204 pci_set_power_state(pdev
, PCI_D3hot
);
5209 #endif /* CONFIG_PM */
5211 static void ixgbe_shutdown(struct pci_dev
*pdev
)
5215 __ixgbe_shutdown(pdev
, &wake
);
5217 if (system_state
== SYSTEM_POWER_OFF
) {
5218 pci_wake_from_d3(pdev
, wake
);
5219 pci_set_power_state(pdev
, PCI_D3hot
);
5224 * ixgbe_update_stats - Update the board statistics counters.
5225 * @adapter: board private structure
5227 void ixgbe_update_stats(struct ixgbe_adapter
*adapter
)
5229 struct net_device
*netdev
= adapter
->netdev
;
5230 struct ixgbe_hw
*hw
= &adapter
->hw
;
5231 struct ixgbe_hw_stats
*hwstats
= &adapter
->stats
;
5233 u32 i
, missed_rx
= 0, mpc
, bprc
, lxon
, lxoff
, xon_off_tot
;
5234 u64 non_eop_descs
= 0, restart_queue
= 0, tx_busy
= 0;
5235 u64 alloc_rx_page_failed
= 0, alloc_rx_buff_failed
= 0;
5236 u64 bytes
= 0, packets
= 0, hw_csum_rx_error
= 0;
5238 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
5239 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
5242 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
) {
5245 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
5246 rsc_count
+= adapter
->rx_ring
[i
]->rx_stats
.rsc_count
;
5247 rsc_flush
+= adapter
->rx_ring
[i
]->rx_stats
.rsc_flush
;
5249 adapter
->rsc_total_count
= rsc_count
;
5250 adapter
->rsc_total_flush
= rsc_flush
;
5253 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
5254 struct ixgbe_ring
*rx_ring
= adapter
->rx_ring
[i
];
5255 non_eop_descs
+= rx_ring
->rx_stats
.non_eop_descs
;
5256 alloc_rx_page_failed
+= rx_ring
->rx_stats
.alloc_rx_page_failed
;
5257 alloc_rx_buff_failed
+= rx_ring
->rx_stats
.alloc_rx_buff_failed
;
5258 hw_csum_rx_error
+= rx_ring
->rx_stats
.csum_err
;
5259 bytes
+= rx_ring
->stats
.bytes
;
5260 packets
+= rx_ring
->stats
.packets
;
5262 adapter
->non_eop_descs
= non_eop_descs
;
5263 adapter
->alloc_rx_page_failed
= alloc_rx_page_failed
;
5264 adapter
->alloc_rx_buff_failed
= alloc_rx_buff_failed
;
5265 adapter
->hw_csum_rx_error
= hw_csum_rx_error
;
5266 netdev
->stats
.rx_bytes
= bytes
;
5267 netdev
->stats
.rx_packets
= packets
;
5271 /* gather some stats to the adapter struct that are per queue */
5272 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
5273 struct ixgbe_ring
*tx_ring
= adapter
->tx_ring
[i
];
5274 restart_queue
+= tx_ring
->tx_stats
.restart_queue
;
5275 tx_busy
+= tx_ring
->tx_stats
.tx_busy
;
5276 bytes
+= tx_ring
->stats
.bytes
;
5277 packets
+= tx_ring
->stats
.packets
;
5279 adapter
->restart_queue
= restart_queue
;
5280 adapter
->tx_busy
= tx_busy
;
5281 netdev
->stats
.tx_bytes
= bytes
;
5282 netdev
->stats
.tx_packets
= packets
;
5284 hwstats
->crcerrs
+= IXGBE_READ_REG(hw
, IXGBE_CRCERRS
);
5286 /* 8 register reads */
5287 for (i
= 0; i
< 8; i
++) {
5288 /* for packet buffers not used, the register should read 0 */
5289 mpc
= IXGBE_READ_REG(hw
, IXGBE_MPC(i
));
5291 hwstats
->mpc
[i
] += mpc
;
5292 total_mpc
+= hwstats
->mpc
[i
];
5293 hwstats
->pxontxc
[i
] += IXGBE_READ_REG(hw
, IXGBE_PXONTXC(i
));
5294 hwstats
->pxofftxc
[i
] += IXGBE_READ_REG(hw
, IXGBE_PXOFFTXC(i
));
5295 switch (hw
->mac
.type
) {
5296 case ixgbe_mac_82598EB
:
5297 hwstats
->rnbc
[i
] += IXGBE_READ_REG(hw
, IXGBE_RNBC(i
));
5298 hwstats
->qbtc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBTC(i
));
5299 hwstats
->qbrc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBRC(i
));
5300 hwstats
->pxonrxc
[i
] +=
5301 IXGBE_READ_REG(hw
, IXGBE_PXONRXC(i
));
5303 case ixgbe_mac_82599EB
:
5304 case ixgbe_mac_X540
:
5305 hwstats
->pxonrxc
[i
] +=
5306 IXGBE_READ_REG(hw
, IXGBE_PXONRXCNT(i
));
5313 /*16 register reads */
5314 for (i
= 0; i
< 16; i
++) {
5315 hwstats
->qptc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QPTC(i
));
5316 hwstats
->qprc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QPRC(i
));
5317 if ((hw
->mac
.type
== ixgbe_mac_82599EB
) ||
5318 (hw
->mac
.type
== ixgbe_mac_X540
)) {
5319 hwstats
->qbtc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBTC_L(i
));
5320 IXGBE_READ_REG(hw
, IXGBE_QBTC_H(i
)); /* to clear */
5321 hwstats
->qbrc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBRC_L(i
));
5322 IXGBE_READ_REG(hw
, IXGBE_QBRC_H(i
)); /* to clear */
5326 hwstats
->gprc
+= IXGBE_READ_REG(hw
, IXGBE_GPRC
);
5327 /* work around hardware counting issue */
5328 hwstats
->gprc
-= missed_rx
;
5330 ixgbe_update_xoff_received(adapter
);
5332 /* 82598 hardware only has a 32 bit counter in the high register */
5333 switch (hw
->mac
.type
) {
5334 case ixgbe_mac_82598EB
:
5335 hwstats
->lxonrxc
+= IXGBE_READ_REG(hw
, IXGBE_LXONRXC
);
5336 hwstats
->gorc
+= IXGBE_READ_REG(hw
, IXGBE_GORCH
);
5337 hwstats
->gotc
+= IXGBE_READ_REG(hw
, IXGBE_GOTCH
);
5338 hwstats
->tor
+= IXGBE_READ_REG(hw
, IXGBE_TORH
);
5340 case ixgbe_mac_X540
:
5341 /* OS2BMC stats are X540 only*/
5342 hwstats
->o2bgptc
+= IXGBE_READ_REG(hw
, IXGBE_O2BGPTC
);
5343 hwstats
->o2bspc
+= IXGBE_READ_REG(hw
, IXGBE_O2BSPC
);
5344 hwstats
->b2ospc
+= IXGBE_READ_REG(hw
, IXGBE_B2OSPC
);
5345 hwstats
->b2ogprc
+= IXGBE_READ_REG(hw
, IXGBE_B2OGPRC
);
5346 case ixgbe_mac_82599EB
:
5347 for (i
= 0; i
< 16; i
++)
5348 adapter
->hw_rx_no_dma_resources
+=
5349 IXGBE_READ_REG(hw
, IXGBE_QPRDC(i
));
5350 hwstats
->gorc
+= IXGBE_READ_REG(hw
, IXGBE_GORCL
);
5351 IXGBE_READ_REG(hw
, IXGBE_GORCH
); /* to clear */
5352 hwstats
->gotc
+= IXGBE_READ_REG(hw
, IXGBE_GOTCL
);
5353 IXGBE_READ_REG(hw
, IXGBE_GOTCH
); /* to clear */
5354 hwstats
->tor
+= IXGBE_READ_REG(hw
, IXGBE_TORL
);
5355 IXGBE_READ_REG(hw
, IXGBE_TORH
); /* to clear */
5356 hwstats
->lxonrxc
+= IXGBE_READ_REG(hw
, IXGBE_LXONRXCNT
);
5357 hwstats
->fdirmatch
+= IXGBE_READ_REG(hw
, IXGBE_FDIRMATCH
);
5358 hwstats
->fdirmiss
+= IXGBE_READ_REG(hw
, IXGBE_FDIRMISS
);
5360 hwstats
->fccrc
+= IXGBE_READ_REG(hw
, IXGBE_FCCRC
);
5361 hwstats
->fcoerpdc
+= IXGBE_READ_REG(hw
, IXGBE_FCOERPDC
);
5362 hwstats
->fcoeprc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEPRC
);
5363 hwstats
->fcoeptc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEPTC
);
5364 hwstats
->fcoedwrc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEDWRC
);
5365 hwstats
->fcoedwtc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEDWTC
);
5366 /* Add up per cpu counters for total ddp aloc fail */
5367 if (adapter
->fcoe
.ddp_pool
) {
5368 struct ixgbe_fcoe
*fcoe
= &adapter
->fcoe
;
5369 struct ixgbe_fcoe_ddp_pool
*ddp_pool
;
5371 u64 noddp
= 0, noddp_ext_buff
= 0;
5372 for_each_possible_cpu(cpu
) {
5373 ddp_pool
= per_cpu_ptr(fcoe
->ddp_pool
, cpu
);
5374 noddp
+= ddp_pool
->noddp
;
5375 noddp_ext_buff
+= ddp_pool
->noddp_ext_buff
;
5377 hwstats
->fcoe_noddp
= noddp
;
5378 hwstats
->fcoe_noddp_ext_buff
= noddp_ext_buff
;
5380 #endif /* IXGBE_FCOE */
5385 bprc
= IXGBE_READ_REG(hw
, IXGBE_BPRC
);
5386 hwstats
->bprc
+= bprc
;
5387 hwstats
->mprc
+= IXGBE_READ_REG(hw
, IXGBE_MPRC
);
5388 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
5389 hwstats
->mprc
-= bprc
;
5390 hwstats
->roc
+= IXGBE_READ_REG(hw
, IXGBE_ROC
);
5391 hwstats
->prc64
+= IXGBE_READ_REG(hw
, IXGBE_PRC64
);
5392 hwstats
->prc127
+= IXGBE_READ_REG(hw
, IXGBE_PRC127
);
5393 hwstats
->prc255
+= IXGBE_READ_REG(hw
, IXGBE_PRC255
);
5394 hwstats
->prc511
+= IXGBE_READ_REG(hw
, IXGBE_PRC511
);
5395 hwstats
->prc1023
+= IXGBE_READ_REG(hw
, IXGBE_PRC1023
);
5396 hwstats
->prc1522
+= IXGBE_READ_REG(hw
, IXGBE_PRC1522
);
5397 hwstats
->rlec
+= IXGBE_READ_REG(hw
, IXGBE_RLEC
);
5398 lxon
= IXGBE_READ_REG(hw
, IXGBE_LXONTXC
);
5399 hwstats
->lxontxc
+= lxon
;
5400 lxoff
= IXGBE_READ_REG(hw
, IXGBE_LXOFFTXC
);
5401 hwstats
->lxofftxc
+= lxoff
;
5402 hwstats
->gptc
+= IXGBE_READ_REG(hw
, IXGBE_GPTC
);
5403 hwstats
->mptc
+= IXGBE_READ_REG(hw
, IXGBE_MPTC
);
5405 * 82598 errata - tx of flow control packets is included in tx counters
5407 xon_off_tot
= lxon
+ lxoff
;
5408 hwstats
->gptc
-= xon_off_tot
;
5409 hwstats
->mptc
-= xon_off_tot
;
5410 hwstats
->gotc
-= (xon_off_tot
* (ETH_ZLEN
+ ETH_FCS_LEN
));
5411 hwstats
->ruc
+= IXGBE_READ_REG(hw
, IXGBE_RUC
);
5412 hwstats
->rfc
+= IXGBE_READ_REG(hw
, IXGBE_RFC
);
5413 hwstats
->rjc
+= IXGBE_READ_REG(hw
, IXGBE_RJC
);
5414 hwstats
->tpr
+= IXGBE_READ_REG(hw
, IXGBE_TPR
);
5415 hwstats
->ptc64
+= IXGBE_READ_REG(hw
, IXGBE_PTC64
);
5416 hwstats
->ptc64
-= xon_off_tot
;
5417 hwstats
->ptc127
+= IXGBE_READ_REG(hw
, IXGBE_PTC127
);
5418 hwstats
->ptc255
+= IXGBE_READ_REG(hw
, IXGBE_PTC255
);
5419 hwstats
->ptc511
+= IXGBE_READ_REG(hw
, IXGBE_PTC511
);
5420 hwstats
->ptc1023
+= IXGBE_READ_REG(hw
, IXGBE_PTC1023
);
5421 hwstats
->ptc1522
+= IXGBE_READ_REG(hw
, IXGBE_PTC1522
);
5422 hwstats
->bptc
+= IXGBE_READ_REG(hw
, IXGBE_BPTC
);
5424 /* Fill out the OS statistics structure */
5425 netdev
->stats
.multicast
= hwstats
->mprc
;
5428 netdev
->stats
.rx_errors
= hwstats
->crcerrs
+ hwstats
->rlec
;
5429 netdev
->stats
.rx_dropped
= 0;
5430 netdev
->stats
.rx_length_errors
= hwstats
->rlec
;
5431 netdev
->stats
.rx_crc_errors
= hwstats
->crcerrs
;
5432 netdev
->stats
.rx_missed_errors
= total_mpc
;
5436 * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
5437 * @adapter: pointer to the device adapter structure
5439 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter
*adapter
)
5441 struct ixgbe_hw
*hw
= &adapter
->hw
;
5444 if (!(adapter
->flags2
& IXGBE_FLAG2_FDIR_REQUIRES_REINIT
))
5447 adapter
->flags2
&= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT
;
5449 /* if interface is down do nothing */
5450 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
5453 /* do nothing if we are not using signature filters */
5454 if (!(adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
))
5457 adapter
->fdir_overflow
++;
5459 if (ixgbe_reinit_fdir_tables_82599(hw
) == 0) {
5460 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
5461 set_bit(__IXGBE_TX_FDIR_INIT_DONE
,
5462 &(adapter
->tx_ring
[i
]->state
));
5463 /* re-enable flow director interrupts */
5464 IXGBE_WRITE_REG(hw
, IXGBE_EIMS
, IXGBE_EIMS_FLOW_DIR
);
5466 e_err(probe
, "failed to finish FDIR re-initialization, "
5467 "ignored adding FDIR ATR filters\n");
5472 * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
5473 * @adapter: pointer to the device adapter structure
5475 * This function serves two purposes. First it strobes the interrupt lines
5476 * in order to make certain interrupts are occurring. Secondly it sets the
5477 * bits needed to check for TX hangs. As a result we should immediately
5478 * determine if a hang has occurred.
5480 static void ixgbe_check_hang_subtask(struct ixgbe_adapter
*adapter
)
5482 struct ixgbe_hw
*hw
= &adapter
->hw
;
5486 /* If we're down or resetting, just bail */
5487 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
5488 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
5491 /* Force detection of hung controller */
5492 if (netif_carrier_ok(adapter
->netdev
)) {
5493 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
5494 set_check_for_tx_hang(adapter
->tx_ring
[i
]);
5497 if (!(adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)) {
5499 * for legacy and MSI interrupts don't set any bits
5500 * that are enabled for EIAM, because this operation
5501 * would set *both* EIMS and EICS for any bit in EIAM
5503 IXGBE_WRITE_REG(hw
, IXGBE_EICS
,
5504 (IXGBE_EICS_TCP_TIMER
| IXGBE_EICS_OTHER
));
5506 /* get one bit for every active tx/rx interrupt vector */
5507 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
5508 struct ixgbe_q_vector
*qv
= adapter
->q_vector
[i
];
5509 if (qv
->rx
.ring
|| qv
->tx
.ring
)
5510 eics
|= ((u64
)1 << i
);
5514 /* Cause software interrupt to ensure rings are cleaned */
5515 ixgbe_irq_rearm_queues(adapter
, eics
);
5520 * ixgbe_watchdog_update_link - update the link status
5521 * @adapter: pointer to the device adapter structure
5522 * @link_speed: pointer to a u32 to store the link_speed
5524 static void ixgbe_watchdog_update_link(struct ixgbe_adapter
*adapter
)
5526 struct ixgbe_hw
*hw
= &adapter
->hw
;
5527 u32 link_speed
= adapter
->link_speed
;
5528 bool link_up
= adapter
->link_up
;
5529 bool pfc_en
= adapter
->dcb_cfg
.pfc_mode_enable
;
5531 if (!(adapter
->flags
& IXGBE_FLAG_NEED_LINK_UPDATE
))
5534 if (hw
->mac
.ops
.check_link
) {
5535 hw
->mac
.ops
.check_link(hw
, &link_speed
, &link_up
, false);
5537 /* always assume link is up, if no check link function */
5538 link_speed
= IXGBE_LINK_SPEED_10GB_FULL
;
5542 if (adapter
->ixgbe_ieee_pfc
)
5543 pfc_en
|= !!(adapter
->ixgbe_ieee_pfc
->pfc_en
);
5545 if (link_up
&& !((adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
) && pfc_en
)) {
5546 hw
->mac
.ops
.fc_enable(hw
);
5547 ixgbe_set_rx_drop_en(adapter
);
5551 time_after(jiffies
, (adapter
->link_check_timeout
+
5552 IXGBE_TRY_LINK_TIMEOUT
))) {
5553 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_UPDATE
;
5554 IXGBE_WRITE_REG(hw
, IXGBE_EIMS
, IXGBE_EIMC_LSC
);
5555 IXGBE_WRITE_FLUSH(hw
);
5558 adapter
->link_up
= link_up
;
5559 adapter
->link_speed
= link_speed
;
5562 static void ixgbe_update_default_up(struct ixgbe_adapter
*adapter
)
5564 #ifdef CONFIG_IXGBE_DCB
5565 struct net_device
*netdev
= adapter
->netdev
;
5566 struct dcb_app app
= {
5567 .selector
= IEEE_8021QAZ_APP_SEL_ETHERTYPE
,
5572 if (adapter
->dcbx_cap
& DCB_CAP_DCBX_VER_IEEE
)
5573 up
= dcb_ieee_getapp_mask(netdev
, &app
);
5575 adapter
->default_up
= (up
> 1) ? (ffs(up
) - 1) : 0;
5580 * ixgbe_watchdog_link_is_up - update netif_carrier status and
5581 * print link up message
5582 * @adapter: pointer to the device adapter structure
5584 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter
*adapter
)
5586 struct net_device
*netdev
= adapter
->netdev
;
5587 struct ixgbe_hw
*hw
= &adapter
->hw
;
5588 u32 link_speed
= adapter
->link_speed
;
5589 bool flow_rx
, flow_tx
;
5591 /* only continue if link was previously down */
5592 if (netif_carrier_ok(netdev
))
5595 adapter
->flags2
&= ~IXGBE_FLAG2_SEARCH_FOR_SFP
;
5597 switch (hw
->mac
.type
) {
5598 case ixgbe_mac_82598EB
: {
5599 u32 frctl
= IXGBE_READ_REG(hw
, IXGBE_FCTRL
);
5600 u32 rmcs
= IXGBE_READ_REG(hw
, IXGBE_RMCS
);
5601 flow_rx
= !!(frctl
& IXGBE_FCTRL_RFCE
);
5602 flow_tx
= !!(rmcs
& IXGBE_RMCS_TFCE_802_3X
);
5605 case ixgbe_mac_X540
:
5606 case ixgbe_mac_82599EB
: {
5607 u32 mflcn
= IXGBE_READ_REG(hw
, IXGBE_MFLCN
);
5608 u32 fccfg
= IXGBE_READ_REG(hw
, IXGBE_FCCFG
);
5609 flow_rx
= !!(mflcn
& IXGBE_MFLCN_RFCE
);
5610 flow_tx
= !!(fccfg
& IXGBE_FCCFG_TFCE_802_3X
);
5619 adapter
->last_rx_ptp_check
= jiffies
;
5621 if (adapter
->flags2
& IXGBE_FLAG2_PTP_ENABLED
)
5622 ixgbe_ptp_start_cyclecounter(adapter
);
5624 e_info(drv
, "NIC Link is Up %s, Flow Control: %s\n",
5625 (link_speed
== IXGBE_LINK_SPEED_10GB_FULL
?
5627 (link_speed
== IXGBE_LINK_SPEED_1GB_FULL
?
5629 (link_speed
== IXGBE_LINK_SPEED_100_FULL
?
5632 ((flow_rx
&& flow_tx
) ? "RX/TX" :
5634 (flow_tx
? "TX" : "None"))));
5636 netif_carrier_on(netdev
);
5637 ixgbe_check_vf_rate_limit(adapter
);
5639 /* update the default user priority for VFs */
5640 ixgbe_update_default_up(adapter
);
5642 /* ping all the active vfs to let them know link has changed */
5643 ixgbe_ping_all_vfs(adapter
);
5647 * ixgbe_watchdog_link_is_down - update netif_carrier status and
5648 * print link down message
5649 * @adapter: pointer to the adapter structure
5651 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter
*adapter
)
5653 struct net_device
*netdev
= adapter
->netdev
;
5654 struct ixgbe_hw
*hw
= &adapter
->hw
;
5656 adapter
->link_up
= false;
5657 adapter
->link_speed
= 0;
5659 /* only continue if link was up previously */
5660 if (!netif_carrier_ok(netdev
))
5663 /* poll for SFP+ cable when link is down */
5664 if (ixgbe_is_sfp(hw
) && hw
->mac
.type
== ixgbe_mac_82598EB
)
5665 adapter
->flags2
|= IXGBE_FLAG2_SEARCH_FOR_SFP
;
5667 if (adapter
->flags2
& IXGBE_FLAG2_PTP_ENABLED
)
5668 ixgbe_ptp_start_cyclecounter(adapter
);
5670 e_info(drv
, "NIC Link is Down\n");
5671 netif_carrier_off(netdev
);
5673 /* ping all the active vfs to let them know link has changed */
5674 ixgbe_ping_all_vfs(adapter
);
5678 * ixgbe_watchdog_flush_tx - flush queues on link down
5679 * @adapter: pointer to the device adapter structure
5681 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter
*adapter
)
5684 int some_tx_pending
= 0;
5686 if (!netif_carrier_ok(adapter
->netdev
)) {
5687 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
5688 struct ixgbe_ring
*tx_ring
= adapter
->tx_ring
[i
];
5689 if (tx_ring
->next_to_use
!= tx_ring
->next_to_clean
) {
5690 some_tx_pending
= 1;
5695 if (some_tx_pending
) {
5696 /* We've lost link, so the controller stops DMA,
5697 * but we've got queued Tx work that's never going
5698 * to get done, so reset controller to flush Tx.
5699 * (Do the reset outside of interrupt context).
5701 e_warn(drv
, "initiating reset to clear Tx work after link loss\n");
5702 adapter
->flags2
|= IXGBE_FLAG2_RESET_REQUESTED
;
5707 static void ixgbe_spoof_check(struct ixgbe_adapter
*adapter
)
5711 /* Do not perform spoof check for 82598 or if not in IOV mode */
5712 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
||
5713 adapter
->num_vfs
== 0)
5716 ssvpc
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_SSVPC
);
5719 * ssvpc register is cleared on read, if zero then no
5720 * spoofed packets in the last interval.
5725 e_warn(drv
, "%u Spoofed packets detected\n", ssvpc
);
5729 * ixgbe_watchdog_subtask - check and bring link up
5730 * @adapter: pointer to the device adapter structure
5732 static void ixgbe_watchdog_subtask(struct ixgbe_adapter
*adapter
)
5734 /* if interface is down do nothing */
5735 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
5736 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
5739 ixgbe_watchdog_update_link(adapter
);
5741 if (adapter
->link_up
)
5742 ixgbe_watchdog_link_is_up(adapter
);
5744 ixgbe_watchdog_link_is_down(adapter
);
5746 ixgbe_spoof_check(adapter
);
5747 ixgbe_update_stats(adapter
);
5749 ixgbe_watchdog_flush_tx(adapter
);
5753 * ixgbe_sfp_detection_subtask - poll for SFP+ cable
5754 * @adapter: the ixgbe adapter structure
5756 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter
*adapter
)
5758 struct ixgbe_hw
*hw
= &adapter
->hw
;
5761 /* not searching for SFP so there is nothing to do here */
5762 if (!(adapter
->flags2
& IXGBE_FLAG2_SEARCH_FOR_SFP
) &&
5763 !(adapter
->flags2
& IXGBE_FLAG2_SFP_NEEDS_RESET
))
5766 /* concurent i2c reads are not supported */
5767 if (test_bit(__IXGBE_READ_I2C
, &adapter
->state
))
5770 /* someone else is in init, wait until next service event */
5771 if (test_and_set_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
))
5774 err
= hw
->phy
.ops
.identify_sfp(hw
);
5775 if (err
== IXGBE_ERR_SFP_NOT_SUPPORTED
)
5778 if (err
== IXGBE_ERR_SFP_NOT_PRESENT
) {
5779 /* If no cable is present, then we need to reset
5780 * the next time we find a good cable. */
5781 adapter
->flags2
|= IXGBE_FLAG2_SFP_NEEDS_RESET
;
5788 /* exit if reset not needed */
5789 if (!(adapter
->flags2
& IXGBE_FLAG2_SFP_NEEDS_RESET
))
5792 adapter
->flags2
&= ~IXGBE_FLAG2_SFP_NEEDS_RESET
;
5795 * A module may be identified correctly, but the EEPROM may not have
5796 * support for that module. setup_sfp() will fail in that case, so
5797 * we should not allow that module to load.
5799 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
5800 err
= hw
->phy
.ops
.reset(hw
);
5802 err
= hw
->mac
.ops
.setup_sfp(hw
);
5804 if (err
== IXGBE_ERR_SFP_NOT_SUPPORTED
)
5807 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_CONFIG
;
5808 e_info(probe
, "detected SFP+: %d\n", hw
->phy
.sfp_type
);
5811 clear_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
);
5813 if ((err
== IXGBE_ERR_SFP_NOT_SUPPORTED
) &&
5814 (adapter
->netdev
->reg_state
== NETREG_REGISTERED
)) {
5815 e_dev_err("failed to initialize because an unsupported "
5816 "SFP+ module type was detected.\n");
5817 e_dev_err("Reload the driver after installing a "
5818 "supported module.\n");
5819 unregister_netdev(adapter
->netdev
);
5824 * ixgbe_sfp_link_config_subtask - set up link SFP after module install
5825 * @adapter: the ixgbe adapter structure
5827 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter
*adapter
)
5829 struct ixgbe_hw
*hw
= &adapter
->hw
;
5831 bool autoneg
= false;
5833 if (!(adapter
->flags
& IXGBE_FLAG_NEED_LINK_CONFIG
))
5836 /* someone else is in init, wait until next service event */
5837 if (test_and_set_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
))
5840 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_CONFIG
;
5842 speed
= hw
->phy
.autoneg_advertised
;
5843 if ((!speed
) && (hw
->mac
.ops
.get_link_capabilities
))
5844 hw
->mac
.ops
.get_link_capabilities(hw
, &speed
, &autoneg
);
5845 if (hw
->mac
.ops
.setup_link
)
5846 hw
->mac
.ops
.setup_link(hw
, speed
, true);
5848 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_UPDATE
;
5849 adapter
->link_check_timeout
= jiffies
;
5850 clear_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
);
5853 #ifdef CONFIG_PCI_IOV
5854 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter
*adapter
)
5857 struct ixgbe_hw
*hw
= &adapter
->hw
;
5858 struct net_device
*netdev
= adapter
->netdev
;
5862 gpc
= IXGBE_READ_REG(hw
, IXGBE_TXDGPC
);
5863 if (gpc
) /* If incrementing then no need for the check below */
5866 * Check to see if a bad DMA write target from an errant or
5867 * malicious VF has caused a PCIe error. If so then we can
5868 * issue a VFLR to the offending VF(s) and then resume without
5869 * requesting a full slot reset.
5872 for (vf
= 0; vf
< adapter
->num_vfs
; vf
++) {
5873 ciaa
= (vf
<< 16) | 0x80000000;
5874 /* 32 bit read so align, we really want status at offset 6 */
5875 ciaa
|= PCI_COMMAND
;
5876 IXGBE_WRITE_REG(hw
, IXGBE_CIAA_82599
, ciaa
);
5877 ciad
= IXGBE_READ_REG(hw
, IXGBE_CIAD_82599
);
5879 /* disable debug mode asap after reading data */
5880 IXGBE_WRITE_REG(hw
, IXGBE_CIAA_82599
, ciaa
);
5881 /* Get the upper 16 bits which will be the PCI status reg */
5883 if (ciad
& PCI_STATUS_REC_MASTER_ABORT
) {
5884 netdev_err(netdev
, "VF %d Hung DMA\n", vf
);
5886 ciaa
= (vf
<< 16) | 0x80000000;
5888 IXGBE_WRITE_REG(hw
, IXGBE_CIAA_82599
, ciaa
);
5889 ciad
= 0x00008000; /* VFLR */
5890 IXGBE_WRITE_REG(hw
, IXGBE_CIAD_82599
, ciad
);
5892 IXGBE_WRITE_REG(hw
, IXGBE_CIAA_82599
, ciaa
);
5899 * ixgbe_service_timer - Timer Call-back
5900 * @data: pointer to adapter cast into an unsigned long
5902 static void ixgbe_service_timer(unsigned long data
)
5904 struct ixgbe_adapter
*adapter
= (struct ixgbe_adapter
*)data
;
5905 unsigned long next_event_offset
;
5908 /* poll faster when waiting for link */
5909 if (adapter
->flags
& IXGBE_FLAG_NEED_LINK_UPDATE
)
5910 next_event_offset
= HZ
/ 10;
5912 next_event_offset
= HZ
* 2;
5914 #ifdef CONFIG_PCI_IOV
5916 * don't bother with SR-IOV VF DMA hang check if there are
5917 * no VFs or the link is down
5919 if (!adapter
->num_vfs
||
5920 (adapter
->flags
& IXGBE_FLAG_NEED_LINK_UPDATE
))
5921 goto normal_timer_service
;
5923 /* If we have VFs allocated then we must check for DMA hangs */
5924 ixgbe_check_for_bad_vf(adapter
);
5925 next_event_offset
= HZ
/ 50;
5926 adapter
->timer_event_accumulator
++;
5928 if (adapter
->timer_event_accumulator
>= 100)
5929 adapter
->timer_event_accumulator
= 0;
5933 normal_timer_service
:
5935 /* Reset the timer */
5936 mod_timer(&adapter
->service_timer
, next_event_offset
+ jiffies
);
5939 ixgbe_service_event_schedule(adapter
);
5942 static void ixgbe_reset_subtask(struct ixgbe_adapter
*adapter
)
5944 if (!(adapter
->flags2
& IXGBE_FLAG2_RESET_REQUESTED
))
5947 adapter
->flags2
&= ~IXGBE_FLAG2_RESET_REQUESTED
;
5949 /* If we're already down or resetting, just bail */
5950 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
5951 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
5954 ixgbe_dump(adapter
);
5955 netdev_err(adapter
->netdev
, "Reset adapter\n");
5956 adapter
->tx_timeout_count
++;
5958 ixgbe_reinit_locked(adapter
);
5962 * ixgbe_service_task - manages and runs subtasks
5963 * @work: pointer to work_struct containing our data
5965 static void ixgbe_service_task(struct work_struct
*work
)
5967 struct ixgbe_adapter
*adapter
= container_of(work
,
5968 struct ixgbe_adapter
,
5970 ixgbe_reset_subtask(adapter
);
5971 ixgbe_sfp_detection_subtask(adapter
);
5972 ixgbe_sfp_link_config_subtask(adapter
);
5973 ixgbe_check_overtemp_subtask(adapter
);
5974 ixgbe_watchdog_subtask(adapter
);
5975 ixgbe_fdir_reinit_subtask(adapter
);
5976 ixgbe_check_hang_subtask(adapter
);
5978 if (adapter
->flags2
& IXGBE_FLAG2_PTP_ENABLED
) {
5979 ixgbe_ptp_overflow_check(adapter
);
5980 ixgbe_ptp_rx_hang(adapter
);
5983 ixgbe_service_event_complete(adapter
);
5986 static int ixgbe_tso(struct ixgbe_ring
*tx_ring
,
5987 struct ixgbe_tx_buffer
*first
,
5990 struct sk_buff
*skb
= first
->skb
;
5991 u32 vlan_macip_lens
, type_tucmd
;
5992 u32 mss_l4len_idx
, l4len
;
5994 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
5997 if (!skb_is_gso(skb
))
6000 if (skb_header_cloned(skb
)) {
6001 int err
= pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
);
6006 /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
6007 type_tucmd
= IXGBE_ADVTXD_TUCMD_L4T_TCP
;
6009 if (first
->protocol
== __constant_htons(ETH_P_IP
)) {
6010 struct iphdr
*iph
= ip_hdr(skb
);
6013 tcp_hdr(skb
)->check
= ~csum_tcpudp_magic(iph
->saddr
,
6017 type_tucmd
|= IXGBE_ADVTXD_TUCMD_IPV4
;
6018 first
->tx_flags
|= IXGBE_TX_FLAGS_TSO
|
6019 IXGBE_TX_FLAGS_CSUM
|
6020 IXGBE_TX_FLAGS_IPV4
;
6021 } else if (skb_is_gso_v6(skb
)) {
6022 ipv6_hdr(skb
)->payload_len
= 0;
6023 tcp_hdr(skb
)->check
=
6024 ~csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
6025 &ipv6_hdr(skb
)->daddr
,
6027 first
->tx_flags
|= IXGBE_TX_FLAGS_TSO
|
6028 IXGBE_TX_FLAGS_CSUM
;
6031 /* compute header lengths */
6032 l4len
= tcp_hdrlen(skb
);
6033 *hdr_len
= skb_transport_offset(skb
) + l4len
;
6035 /* update gso size and bytecount with header size */
6036 first
->gso_segs
= skb_shinfo(skb
)->gso_segs
;
6037 first
->bytecount
+= (first
->gso_segs
- 1) * *hdr_len
;
6039 /* mss_l4len_id: use 0 as index for TSO */
6040 mss_l4len_idx
= l4len
<< IXGBE_ADVTXD_L4LEN_SHIFT
;
6041 mss_l4len_idx
|= skb_shinfo(skb
)->gso_size
<< IXGBE_ADVTXD_MSS_SHIFT
;
6043 /* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
6044 vlan_macip_lens
= skb_network_header_len(skb
);
6045 vlan_macip_lens
|= skb_network_offset(skb
) << IXGBE_ADVTXD_MACLEN_SHIFT
;
6046 vlan_macip_lens
|= first
->tx_flags
& IXGBE_TX_FLAGS_VLAN_MASK
;
6048 ixgbe_tx_ctxtdesc(tx_ring
, vlan_macip_lens
, 0, type_tucmd
,
6054 static void ixgbe_tx_csum(struct ixgbe_ring
*tx_ring
,
6055 struct ixgbe_tx_buffer
*first
)
6057 struct sk_buff
*skb
= first
->skb
;
6058 u32 vlan_macip_lens
= 0;
6059 u32 mss_l4len_idx
= 0;
6062 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
6063 if (!(first
->tx_flags
& IXGBE_TX_FLAGS_HW_VLAN
) &&
6064 !(first
->tx_flags
& IXGBE_TX_FLAGS_CC
))
6068 switch (first
->protocol
) {
6069 case __constant_htons(ETH_P_IP
):
6070 vlan_macip_lens
|= skb_network_header_len(skb
);
6071 type_tucmd
|= IXGBE_ADVTXD_TUCMD_IPV4
;
6072 l4_hdr
= ip_hdr(skb
)->protocol
;
6074 case __constant_htons(ETH_P_IPV6
):
6075 vlan_macip_lens
|= skb_network_header_len(skb
);
6076 l4_hdr
= ipv6_hdr(skb
)->nexthdr
;
6079 if (unlikely(net_ratelimit())) {
6080 dev_warn(tx_ring
->dev
,
6081 "partial checksum but proto=%x!\n",
6089 type_tucmd
|= IXGBE_ADVTXD_TUCMD_L4T_TCP
;
6090 mss_l4len_idx
= tcp_hdrlen(skb
) <<
6091 IXGBE_ADVTXD_L4LEN_SHIFT
;
6094 type_tucmd
|= IXGBE_ADVTXD_TUCMD_L4T_SCTP
;
6095 mss_l4len_idx
= sizeof(struct sctphdr
) <<
6096 IXGBE_ADVTXD_L4LEN_SHIFT
;
6099 mss_l4len_idx
= sizeof(struct udphdr
) <<
6100 IXGBE_ADVTXD_L4LEN_SHIFT
;
6103 if (unlikely(net_ratelimit())) {
6104 dev_warn(tx_ring
->dev
,
6105 "partial checksum but l4 proto=%x!\n",
6111 /* update TX checksum flag */
6112 first
->tx_flags
|= IXGBE_TX_FLAGS_CSUM
;
6115 /* vlan_macip_lens: MACLEN, VLAN tag */
6116 vlan_macip_lens
|= skb_network_offset(skb
) << IXGBE_ADVTXD_MACLEN_SHIFT
;
6117 vlan_macip_lens
|= first
->tx_flags
& IXGBE_TX_FLAGS_VLAN_MASK
;
6119 ixgbe_tx_ctxtdesc(tx_ring
, vlan_macip_lens
, 0,
6120 type_tucmd
, mss_l4len_idx
);
6123 #define IXGBE_SET_FLAG(_input, _flag, _result) \
6124 ((_flag <= _result) ? \
6125 ((u32)(_input & _flag) * (_result / _flag)) : \
6126 ((u32)(_input & _flag) / (_flag / _result)))
6128 static u32
ixgbe_tx_cmd_type(struct sk_buff
*skb
, u32 tx_flags
)
6130 /* set type for advanced descriptor with frame checksum insertion */
6131 u32 cmd_type
= IXGBE_ADVTXD_DTYP_DATA
|
6132 IXGBE_ADVTXD_DCMD_DEXT
|
6133 IXGBE_ADVTXD_DCMD_IFCS
;
6135 /* set HW vlan bit if vlan is present */
6136 cmd_type
|= IXGBE_SET_FLAG(tx_flags
, IXGBE_TX_FLAGS_HW_VLAN
,
6137 IXGBE_ADVTXD_DCMD_VLE
);
6139 /* set segmentation enable bits for TSO/FSO */
6140 cmd_type
|= IXGBE_SET_FLAG(tx_flags
, IXGBE_TX_FLAGS_TSO
,
6141 IXGBE_ADVTXD_DCMD_TSE
);
6143 /* set timestamp bit if present */
6144 cmd_type
|= IXGBE_SET_FLAG(tx_flags
, IXGBE_TX_FLAGS_TSTAMP
,
6145 IXGBE_ADVTXD_MAC_TSTAMP
);
6147 /* insert frame checksum */
6148 cmd_type
^= IXGBE_SET_FLAG(skb
->no_fcs
, 1, IXGBE_ADVTXD_DCMD_IFCS
);
6153 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc
*tx_desc
,
6154 u32 tx_flags
, unsigned int paylen
)
6156 u32 olinfo_status
= paylen
<< IXGBE_ADVTXD_PAYLEN_SHIFT
;
6158 /* enable L4 checksum for TSO and TX checksum offload */
6159 olinfo_status
|= IXGBE_SET_FLAG(tx_flags
,
6160 IXGBE_TX_FLAGS_CSUM
,
6161 IXGBE_ADVTXD_POPTS_TXSM
);
6163 /* enble IPv4 checksum for TSO */
6164 olinfo_status
|= IXGBE_SET_FLAG(tx_flags
,
6165 IXGBE_TX_FLAGS_IPV4
,
6166 IXGBE_ADVTXD_POPTS_IXSM
);
6169 * Check Context must be set if Tx switch is enabled, which it
6170 * always is for case where virtual functions are running
6172 olinfo_status
|= IXGBE_SET_FLAG(tx_flags
,
6176 tx_desc
->read
.olinfo_status
= cpu_to_le32(olinfo_status
);
6179 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
6182 static void ixgbe_tx_map(struct ixgbe_ring
*tx_ring
,
6183 struct ixgbe_tx_buffer
*first
,
6186 struct sk_buff
*skb
= first
->skb
;
6187 struct ixgbe_tx_buffer
*tx_buffer
;
6188 union ixgbe_adv_tx_desc
*tx_desc
;
6189 struct skb_frag_struct
*frag
;
6191 unsigned int data_len
, size
;
6192 u32 tx_flags
= first
->tx_flags
;
6193 u32 cmd_type
= ixgbe_tx_cmd_type(skb
, tx_flags
);
6194 u16 i
= tx_ring
->next_to_use
;
6196 tx_desc
= IXGBE_TX_DESC(tx_ring
, i
);
6198 ixgbe_tx_olinfo_status(tx_desc
, tx_flags
, skb
->len
- hdr_len
);
6200 size
= skb_headlen(skb
);
6201 data_len
= skb
->data_len
;
6204 if (tx_flags
& IXGBE_TX_FLAGS_FCOE
) {
6205 if (data_len
< sizeof(struct fcoe_crc_eof
)) {
6206 size
-= sizeof(struct fcoe_crc_eof
) - data_len
;
6209 data_len
-= sizeof(struct fcoe_crc_eof
);
6214 dma
= dma_map_single(tx_ring
->dev
, skb
->data
, size
, DMA_TO_DEVICE
);
6218 for (frag
= &skb_shinfo(skb
)->frags
[0];; frag
++) {
6219 if (dma_mapping_error(tx_ring
->dev
, dma
))
6222 /* record length, and DMA address */
6223 dma_unmap_len_set(tx_buffer
, len
, size
);
6224 dma_unmap_addr_set(tx_buffer
, dma
, dma
);
6226 tx_desc
->read
.buffer_addr
= cpu_to_le64(dma
);
6228 while (unlikely(size
> IXGBE_MAX_DATA_PER_TXD
)) {
6229 tx_desc
->read
.cmd_type_len
=
6230 cpu_to_le32(cmd_type
^ IXGBE_MAX_DATA_PER_TXD
);
6234 if (i
== tx_ring
->count
) {
6235 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
6238 tx_desc
->read
.olinfo_status
= 0;
6240 dma
+= IXGBE_MAX_DATA_PER_TXD
;
6241 size
-= IXGBE_MAX_DATA_PER_TXD
;
6243 tx_desc
->read
.buffer_addr
= cpu_to_le64(dma
);
6246 if (likely(!data_len
))
6249 tx_desc
->read
.cmd_type_len
= cpu_to_le32(cmd_type
^ size
);
6253 if (i
== tx_ring
->count
) {
6254 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
6257 tx_desc
->read
.olinfo_status
= 0;
6260 size
= min_t(unsigned int, data_len
, skb_frag_size(frag
));
6262 size
= skb_frag_size(frag
);
6266 dma
= skb_frag_dma_map(tx_ring
->dev
, frag
, 0, size
,
6269 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
6272 /* write last descriptor with RS and EOP bits */
6273 cmd_type
|= size
| IXGBE_TXD_CMD
;
6274 tx_desc
->read
.cmd_type_len
= cpu_to_le32(cmd_type
);
6276 netdev_tx_sent_queue(txring_txq(tx_ring
), first
->bytecount
);
6278 /* set the timestamp */
6279 first
->time_stamp
= jiffies
;
6282 * Force memory writes to complete before letting h/w know there
6283 * are new descriptors to fetch. (Only applicable for weak-ordered
6284 * memory model archs, such as IA-64).
6286 * We also need this memory barrier to make certain all of the
6287 * status bits have been updated before next_to_watch is written.
6291 /* set next_to_watch value indicating a packet is present */
6292 first
->next_to_watch
= tx_desc
;
6295 if (i
== tx_ring
->count
)
6298 tx_ring
->next_to_use
= i
;
6300 /* notify HW of packet */
6301 writel(i
, tx_ring
->tail
);
6305 dev_err(tx_ring
->dev
, "TX DMA map failed\n");
6307 /* clear dma mappings for failed tx_buffer_info map */
6309 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
6310 ixgbe_unmap_and_free_tx_resource(tx_ring
, tx_buffer
);
6311 if (tx_buffer
== first
)
6318 tx_ring
->next_to_use
= i
;
6321 static void ixgbe_atr(struct ixgbe_ring
*ring
,
6322 struct ixgbe_tx_buffer
*first
)
6324 struct ixgbe_q_vector
*q_vector
= ring
->q_vector
;
6325 union ixgbe_atr_hash_dword input
= { .dword
= 0 };
6326 union ixgbe_atr_hash_dword common
= { .dword
= 0 };
6328 unsigned char *network
;
6330 struct ipv6hdr
*ipv6
;
6335 /* if ring doesn't have a interrupt vector, cannot perform ATR */
6339 /* do nothing if sampling is disabled */
6340 if (!ring
->atr_sample_rate
)
6345 /* snag network header to get L4 type and address */
6346 hdr
.network
= skb_network_header(first
->skb
);
6348 /* Currently only IPv4/IPv6 with TCP is supported */
6349 if ((first
->protocol
!= __constant_htons(ETH_P_IPV6
) ||
6350 hdr
.ipv6
->nexthdr
!= IPPROTO_TCP
) &&
6351 (first
->protocol
!= __constant_htons(ETH_P_IP
) ||
6352 hdr
.ipv4
->protocol
!= IPPROTO_TCP
))
6355 th
= tcp_hdr(first
->skb
);
6357 /* skip this packet since it is invalid or the socket is closing */
6361 /* sample on all syn packets or once every atr sample count */
6362 if (!th
->syn
&& (ring
->atr_count
< ring
->atr_sample_rate
))
6365 /* reset sample count */
6366 ring
->atr_count
= 0;
6368 vlan_id
= htons(first
->tx_flags
>> IXGBE_TX_FLAGS_VLAN_SHIFT
);
6371 * src and dst are inverted, think how the receiver sees them
6373 * The input is broken into two sections, a non-compressed section
6374 * containing vm_pool, vlan_id, and flow_type. The rest of the data
6375 * is XORed together and stored in the compressed dword.
6377 input
.formatted
.vlan_id
= vlan_id
;
6380 * since src port and flex bytes occupy the same word XOR them together
6381 * and write the value to source port portion of compressed dword
6383 if (first
->tx_flags
& (IXGBE_TX_FLAGS_SW_VLAN
| IXGBE_TX_FLAGS_HW_VLAN
))
6384 common
.port
.src
^= th
->dest
^ __constant_htons(ETH_P_8021Q
);
6386 common
.port
.src
^= th
->dest
^ first
->protocol
;
6387 common
.port
.dst
^= th
->source
;
6389 if (first
->protocol
== __constant_htons(ETH_P_IP
)) {
6390 input
.formatted
.flow_type
= IXGBE_ATR_FLOW_TYPE_TCPV4
;
6391 common
.ip
^= hdr
.ipv4
->saddr
^ hdr
.ipv4
->daddr
;
6393 input
.formatted
.flow_type
= IXGBE_ATR_FLOW_TYPE_TCPV6
;
6394 common
.ip
^= hdr
.ipv6
->saddr
.s6_addr32
[0] ^
6395 hdr
.ipv6
->saddr
.s6_addr32
[1] ^
6396 hdr
.ipv6
->saddr
.s6_addr32
[2] ^
6397 hdr
.ipv6
->saddr
.s6_addr32
[3] ^
6398 hdr
.ipv6
->daddr
.s6_addr32
[0] ^
6399 hdr
.ipv6
->daddr
.s6_addr32
[1] ^
6400 hdr
.ipv6
->daddr
.s6_addr32
[2] ^
6401 hdr
.ipv6
->daddr
.s6_addr32
[3];
6404 /* This assumes the Rx queue and Tx queue are bound to the same CPU */
6405 ixgbe_fdir_add_signature_filter_82599(&q_vector
->adapter
->hw
,
6406 input
, common
, ring
->queue_index
);
6409 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring
*tx_ring
, u16 size
)
6411 netif_stop_subqueue(tx_ring
->netdev
, tx_ring
->queue_index
);
6412 /* Herbert's original patch had:
6413 * smp_mb__after_netif_stop_queue();
6414 * but since that doesn't exist yet, just open code it. */
6417 /* We need to check again in a case another CPU has just
6418 * made room available. */
6419 if (likely(ixgbe_desc_unused(tx_ring
) < size
))
6422 /* A reprieve! - use start_queue because it doesn't call schedule */
6423 netif_start_subqueue(tx_ring
->netdev
, tx_ring
->queue_index
);
6424 ++tx_ring
->tx_stats
.restart_queue
;
6428 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring
*tx_ring
, u16 size
)
6430 if (likely(ixgbe_desc_unused(tx_ring
) >= size
))
6432 return __ixgbe_maybe_stop_tx(tx_ring
, size
);
6436 static u16
ixgbe_select_queue(struct net_device
*dev
, struct sk_buff
*skb
)
6438 struct ixgbe_adapter
*adapter
;
6439 struct ixgbe_ring_feature
*f
;
6443 * only execute the code below if protocol is FCoE
6444 * or FIP and we have FCoE enabled on the adapter
6446 switch (vlan_get_protocol(skb
)) {
6447 case __constant_htons(ETH_P_FCOE
):
6448 case __constant_htons(ETH_P_FIP
):
6449 adapter
= netdev_priv(dev
);
6451 if (adapter
->flags
& IXGBE_FLAG_FCOE_ENABLED
)
6454 return __netdev_pick_tx(dev
, skb
);
6457 f
= &adapter
->ring_feature
[RING_F_FCOE
];
6459 txq
= skb_rx_queue_recorded(skb
) ? skb_get_rx_queue(skb
) :
6462 while (txq
>= f
->indices
)
6465 return txq
+ f
->offset
;
6469 netdev_tx_t
ixgbe_xmit_frame_ring(struct sk_buff
*skb
,
6470 struct ixgbe_adapter
*adapter
,
6471 struct ixgbe_ring
*tx_ring
)
6473 struct ixgbe_tx_buffer
*first
;
6477 u16 count
= TXD_USE_COUNT(skb_headlen(skb
));
6478 __be16 protocol
= skb
->protocol
;
6482 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
6483 * + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
6484 * + 2 desc gap to keep tail from touching head,
6485 * + 1 desc for context descriptor,
6486 * otherwise try next time
6488 for (f
= 0; f
< skb_shinfo(skb
)->nr_frags
; f
++)
6489 count
+= TXD_USE_COUNT(skb_shinfo(skb
)->frags
[f
].size
);
6491 if (ixgbe_maybe_stop_tx(tx_ring
, count
+ 3)) {
6492 tx_ring
->tx_stats
.tx_busy
++;
6493 return NETDEV_TX_BUSY
;
6496 /* record the location of the first descriptor for this packet */
6497 first
= &tx_ring
->tx_buffer_info
[tx_ring
->next_to_use
];
6499 first
->bytecount
= skb
->len
;
6500 first
->gso_segs
= 1;
6502 /* if we have a HW VLAN tag being added default to the HW one */
6503 if (vlan_tx_tag_present(skb
)) {
6504 tx_flags
|= vlan_tx_tag_get(skb
) << IXGBE_TX_FLAGS_VLAN_SHIFT
;
6505 tx_flags
|= IXGBE_TX_FLAGS_HW_VLAN
;
6506 /* else if it is a SW VLAN check the next protocol and store the tag */
6507 } else if (protocol
== __constant_htons(ETH_P_8021Q
)) {
6508 struct vlan_hdr
*vhdr
, _vhdr
;
6509 vhdr
= skb_header_pointer(skb
, ETH_HLEN
, sizeof(_vhdr
), &_vhdr
);
6513 protocol
= vhdr
->h_vlan_encapsulated_proto
;
6514 tx_flags
|= ntohs(vhdr
->h_vlan_TCI
) <<
6515 IXGBE_TX_FLAGS_VLAN_SHIFT
;
6516 tx_flags
|= IXGBE_TX_FLAGS_SW_VLAN
;
6519 skb_tx_timestamp(skb
);
6521 if (unlikely(skb_shinfo(skb
)->tx_flags
& SKBTX_HW_TSTAMP
)) {
6522 skb_shinfo(skb
)->tx_flags
|= SKBTX_IN_PROGRESS
;
6523 tx_flags
|= IXGBE_TX_FLAGS_TSTAMP
;
6525 /* schedule check for Tx timestamp */
6526 adapter
->ptp_tx_skb
= skb_get(skb
);
6527 adapter
->ptp_tx_start
= jiffies
;
6528 schedule_work(&adapter
->ptp_tx_work
);
6531 #ifdef CONFIG_PCI_IOV
6533 * Use the l2switch_enable flag - would be false if the DMA
6534 * Tx switch had been disabled.
6536 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
6537 tx_flags
|= IXGBE_TX_FLAGS_CC
;
6540 /* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
6541 if ((adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
) &&
6542 ((tx_flags
& (IXGBE_TX_FLAGS_HW_VLAN
| IXGBE_TX_FLAGS_SW_VLAN
)) ||
6543 (skb
->priority
!= TC_PRIO_CONTROL
))) {
6544 tx_flags
&= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK
;
6545 tx_flags
|= (skb
->priority
& 0x7) <<
6546 IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT
;
6547 if (tx_flags
& IXGBE_TX_FLAGS_SW_VLAN
) {
6548 struct vlan_ethhdr
*vhdr
;
6549 if (skb_header_cloned(skb
) &&
6550 pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
))
6552 vhdr
= (struct vlan_ethhdr
*)skb
->data
;
6553 vhdr
->h_vlan_TCI
= htons(tx_flags
>>
6554 IXGBE_TX_FLAGS_VLAN_SHIFT
);
6556 tx_flags
|= IXGBE_TX_FLAGS_HW_VLAN
;
6560 /* record initial flags and protocol */
6561 first
->tx_flags
= tx_flags
;
6562 first
->protocol
= protocol
;
6565 /* setup tx offload for FCoE */
6566 if ((protocol
== __constant_htons(ETH_P_FCOE
)) &&
6567 (tx_ring
->netdev
->features
& (NETIF_F_FSO
| NETIF_F_FCOE_CRC
))) {
6568 tso
= ixgbe_fso(tx_ring
, first
, &hdr_len
);
6575 #endif /* IXGBE_FCOE */
6576 tso
= ixgbe_tso(tx_ring
, first
, &hdr_len
);
6580 ixgbe_tx_csum(tx_ring
, first
);
6582 /* add the ATR filter if ATR is on */
6583 if (test_bit(__IXGBE_TX_FDIR_INIT_DONE
, &tx_ring
->state
))
6584 ixgbe_atr(tx_ring
, first
);
6588 #endif /* IXGBE_FCOE */
6589 ixgbe_tx_map(tx_ring
, first
, hdr_len
);
6591 ixgbe_maybe_stop_tx(tx_ring
, DESC_NEEDED
);
6593 return NETDEV_TX_OK
;
6596 dev_kfree_skb_any(first
->skb
);
6599 return NETDEV_TX_OK
;
6602 static netdev_tx_t
ixgbe_xmit_frame(struct sk_buff
*skb
,
6603 struct net_device
*netdev
)
6605 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6606 struct ixgbe_ring
*tx_ring
;
6609 * The minimum packet size for olinfo paylen is 17 so pad the skb
6610 * in order to meet this minimum size requirement.
6612 if (unlikely(skb
->len
< 17)) {
6613 if (skb_pad(skb
, 17 - skb
->len
))
6614 return NETDEV_TX_OK
;
6616 skb_set_tail_pointer(skb
, 17);
6619 tx_ring
= adapter
->tx_ring
[skb
->queue_mapping
];
6620 return ixgbe_xmit_frame_ring(skb
, adapter
, tx_ring
);
6624 * ixgbe_set_mac - Change the Ethernet Address of the NIC
6625 * @netdev: network interface device structure
6626 * @p: pointer to an address structure
6628 * Returns 0 on success, negative on failure
6630 static int ixgbe_set_mac(struct net_device
*netdev
, void *p
)
6632 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6633 struct ixgbe_hw
*hw
= &adapter
->hw
;
6634 struct sockaddr
*addr
= p
;
6636 if (!is_valid_ether_addr(addr
->sa_data
))
6637 return -EADDRNOTAVAIL
;
6639 memcpy(netdev
->dev_addr
, addr
->sa_data
, netdev
->addr_len
);
6640 memcpy(hw
->mac
.addr
, addr
->sa_data
, netdev
->addr_len
);
6642 hw
->mac
.ops
.set_rar(hw
, 0, hw
->mac
.addr
, VMDQ_P(0), IXGBE_RAH_AV
);
6648 ixgbe_mdio_read(struct net_device
*netdev
, int prtad
, int devad
, u16 addr
)
6650 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6651 struct ixgbe_hw
*hw
= &adapter
->hw
;
6655 if (prtad
!= hw
->phy
.mdio
.prtad
)
6657 rc
= hw
->phy
.ops
.read_reg(hw
, addr
, devad
, &value
);
6663 static int ixgbe_mdio_write(struct net_device
*netdev
, int prtad
, int devad
,
6664 u16 addr
, u16 value
)
6666 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6667 struct ixgbe_hw
*hw
= &adapter
->hw
;
6669 if (prtad
!= hw
->phy
.mdio
.prtad
)
6671 return hw
->phy
.ops
.write_reg(hw
, addr
, devad
, value
);
6674 static int ixgbe_ioctl(struct net_device
*netdev
, struct ifreq
*req
, int cmd
)
6676 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6680 return ixgbe_ptp_hwtstamp_ioctl(adapter
, req
, cmd
);
6682 return mdio_mii_ioctl(&adapter
->hw
.phy
.mdio
, if_mii(req
), cmd
);
6687 * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
6689 * @netdev: network interface device structure
6691 * Returns non-zero on failure
6693 static int ixgbe_add_sanmac_netdev(struct net_device
*dev
)
6696 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
6697 struct ixgbe_hw
*hw
= &adapter
->hw
;
6699 if (is_valid_ether_addr(hw
->mac
.san_addr
)) {
6701 err
= dev_addr_add(dev
, hw
->mac
.san_addr
, NETDEV_HW_ADDR_T_SAN
);
6704 /* update SAN MAC vmdq pool selection */
6705 hw
->mac
.ops
.set_vmdq_san_mac(hw
, VMDQ_P(0));
6711 * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
6713 * @netdev: network interface device structure
6715 * Returns non-zero on failure
6717 static int ixgbe_del_sanmac_netdev(struct net_device
*dev
)
6720 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
6721 struct ixgbe_mac_info
*mac
= &adapter
->hw
.mac
;
6723 if (is_valid_ether_addr(mac
->san_addr
)) {
6725 err
= dev_addr_del(dev
, mac
->san_addr
, NETDEV_HW_ADDR_T_SAN
);
6731 #ifdef CONFIG_NET_POLL_CONTROLLER
6733 * Polling 'interrupt' - used by things like netconsole to send skbs
6734 * without having to re-enable interrupts. It's not called while
6735 * the interrupt routine is executing.
6737 static void ixgbe_netpoll(struct net_device
*netdev
)
6739 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6742 /* if interface is down do nothing */
6743 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
6746 adapter
->flags
|= IXGBE_FLAG_IN_NETPOLL
;
6747 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
) {
6748 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
6749 ixgbe_msix_clean_rings(0, adapter
->q_vector
[i
]);
6751 ixgbe_intr(adapter
->pdev
->irq
, netdev
);
6753 adapter
->flags
&= ~IXGBE_FLAG_IN_NETPOLL
;
6757 static struct rtnl_link_stats64
*ixgbe_get_stats64(struct net_device
*netdev
,
6758 struct rtnl_link_stats64
*stats
)
6760 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6764 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
6765 struct ixgbe_ring
*ring
= ACCESS_ONCE(adapter
->rx_ring
[i
]);
6771 start
= u64_stats_fetch_begin_bh(&ring
->syncp
);
6772 packets
= ring
->stats
.packets
;
6773 bytes
= ring
->stats
.bytes
;
6774 } while (u64_stats_fetch_retry_bh(&ring
->syncp
, start
));
6775 stats
->rx_packets
+= packets
;
6776 stats
->rx_bytes
+= bytes
;
6780 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
6781 struct ixgbe_ring
*ring
= ACCESS_ONCE(adapter
->tx_ring
[i
]);
6787 start
= u64_stats_fetch_begin_bh(&ring
->syncp
);
6788 packets
= ring
->stats
.packets
;
6789 bytes
= ring
->stats
.bytes
;
6790 } while (u64_stats_fetch_retry_bh(&ring
->syncp
, start
));
6791 stats
->tx_packets
+= packets
;
6792 stats
->tx_bytes
+= bytes
;
6796 /* following stats updated by ixgbe_watchdog_task() */
6797 stats
->multicast
= netdev
->stats
.multicast
;
6798 stats
->rx_errors
= netdev
->stats
.rx_errors
;
6799 stats
->rx_length_errors
= netdev
->stats
.rx_length_errors
;
6800 stats
->rx_crc_errors
= netdev
->stats
.rx_crc_errors
;
6801 stats
->rx_missed_errors
= netdev
->stats
.rx_missed_errors
;
6805 #ifdef CONFIG_IXGBE_DCB
6807 * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
6808 * @adapter: pointer to ixgbe_adapter
6809 * @tc: number of traffic classes currently enabled
6811 * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
6812 * 802.1Q priority maps to a packet buffer that exists.
6814 static void ixgbe_validate_rtr(struct ixgbe_adapter
*adapter
, u8 tc
)
6816 struct ixgbe_hw
*hw
= &adapter
->hw
;
6820 /* 82598 have a static priority to TC mapping that can not
6821 * be changed so no validation is needed.
6823 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
6826 reg
= IXGBE_READ_REG(hw
, IXGBE_RTRUP2TC
);
6829 for (i
= 0; i
< MAX_TRAFFIC_CLASS
; i
++) {
6830 u8 up2tc
= reg
>> (i
* IXGBE_RTRUP2TC_UP_SHIFT
);
6832 /* If up2tc is out of bounds default to zero */
6834 reg
&= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT
);
6838 IXGBE_WRITE_REG(hw
, IXGBE_RTRUP2TC
, reg
);
6844 * ixgbe_set_prio_tc_map - Configure netdev prio tc map
6845 * @adapter: Pointer to adapter struct
6847 * Populate the netdev user priority to tc map
6849 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter
*adapter
)
6851 struct net_device
*dev
= adapter
->netdev
;
6852 struct ixgbe_dcb_config
*dcb_cfg
= &adapter
->dcb_cfg
;
6853 struct ieee_ets
*ets
= adapter
->ixgbe_ieee_ets
;
6856 for (prio
= 0; prio
< MAX_USER_PRIORITY
; prio
++) {
6859 if (adapter
->dcbx_cap
& DCB_CAP_DCBX_VER_CEE
)
6860 tc
= ixgbe_dcb_get_tc_from_up(dcb_cfg
, 0, prio
);
6862 tc
= ets
->prio_tc
[prio
];
6864 netdev_set_prio_tc_map(dev
, prio
, tc
);
6868 #endif /* CONFIG_IXGBE_DCB */
6870 * ixgbe_setup_tc - configure net_device for multiple traffic classes
6872 * @netdev: net device to configure
6873 * @tc: number of traffic classes to enable
6875 int ixgbe_setup_tc(struct net_device
*dev
, u8 tc
)
6877 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
6878 struct ixgbe_hw
*hw
= &adapter
->hw
;
6880 /* Hardware supports up to 8 traffic classes */
6881 if (tc
> adapter
->dcb_cfg
.num_tcs
.pg_tcs
||
6882 (hw
->mac
.type
== ixgbe_mac_82598EB
&&
6883 tc
< MAX_TRAFFIC_CLASS
))
6886 /* Hardware has to reinitialize queues and interrupts to
6887 * match packet buffer alignment. Unfortunately, the
6888 * hardware is not flexible enough to do this dynamically.
6890 if (netif_running(dev
))
6892 ixgbe_clear_interrupt_scheme(adapter
);
6894 #ifdef CONFIG_IXGBE_DCB
6896 netdev_set_num_tc(dev
, tc
);
6897 ixgbe_set_prio_tc_map(adapter
);
6899 adapter
->flags
|= IXGBE_FLAG_DCB_ENABLED
;
6901 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
) {
6902 adapter
->last_lfc_mode
= adapter
->hw
.fc
.requested_mode
;
6903 adapter
->hw
.fc
.requested_mode
= ixgbe_fc_none
;
6906 netdev_reset_tc(dev
);
6908 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
)
6909 adapter
->hw
.fc
.requested_mode
= adapter
->last_lfc_mode
;
6911 adapter
->flags
&= ~IXGBE_FLAG_DCB_ENABLED
;
6913 adapter
->temp_dcb_cfg
.pfc_mode_enable
= false;
6914 adapter
->dcb_cfg
.pfc_mode_enable
= false;
6917 ixgbe_validate_rtr(adapter
, tc
);
6919 #endif /* CONFIG_IXGBE_DCB */
6920 ixgbe_init_interrupt_scheme(adapter
);
6922 if (netif_running(dev
))
6923 return ixgbe_open(dev
);
6928 #ifdef CONFIG_PCI_IOV
6929 void ixgbe_sriov_reinit(struct ixgbe_adapter
*adapter
)
6931 struct net_device
*netdev
= adapter
->netdev
;
6934 ixgbe_setup_tc(netdev
, netdev_get_num_tc(netdev
));
6939 void ixgbe_do_reset(struct net_device
*netdev
)
6941 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6943 if (netif_running(netdev
))
6944 ixgbe_reinit_locked(adapter
);
6946 ixgbe_reset(adapter
);
6949 static netdev_features_t
ixgbe_fix_features(struct net_device
*netdev
,
6950 netdev_features_t features
)
6952 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6954 /* If Rx checksum is disabled, then RSC/LRO should also be disabled */
6955 if (!(features
& NETIF_F_RXCSUM
))
6956 features
&= ~NETIF_F_LRO
;
6958 /* Turn off LRO if not RSC capable */
6959 if (!(adapter
->flags2
& IXGBE_FLAG2_RSC_CAPABLE
))
6960 features
&= ~NETIF_F_LRO
;
6965 static int ixgbe_set_features(struct net_device
*netdev
,
6966 netdev_features_t features
)
6968 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
6969 netdev_features_t changed
= netdev
->features
^ features
;
6970 bool need_reset
= false;
6972 /* Make sure RSC matches LRO, reset if change */
6973 if (!(features
& NETIF_F_LRO
)) {
6974 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)
6976 adapter
->flags2
&= ~IXGBE_FLAG2_RSC_ENABLED
;
6977 } else if ((adapter
->flags2
& IXGBE_FLAG2_RSC_CAPABLE
) &&
6978 !(adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)) {
6979 if (adapter
->rx_itr_setting
== 1 ||
6980 adapter
->rx_itr_setting
> IXGBE_MIN_RSC_ITR
) {
6981 adapter
->flags2
|= IXGBE_FLAG2_RSC_ENABLED
;
6983 } else if ((changed
^ features
) & NETIF_F_LRO
) {
6984 e_info(probe
, "rx-usecs set too low, "
6990 * Check if Flow Director n-tuple support was enabled or disabled. If
6991 * the state changed, we need to reset.
6993 switch (features
& NETIF_F_NTUPLE
) {
6994 case NETIF_F_NTUPLE
:
6995 /* turn off ATR, enable perfect filters and reset */
6996 if (!(adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
))
6999 adapter
->flags
&= ~IXGBE_FLAG_FDIR_HASH_CAPABLE
;
7000 adapter
->flags
|= IXGBE_FLAG_FDIR_PERFECT_CAPABLE
;
7003 /* turn off perfect filters, enable ATR and reset */
7004 if (adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
)
7007 adapter
->flags
&= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE
;
7009 /* We cannot enable ATR if SR-IOV is enabled */
7010 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
7013 /* We cannot enable ATR if we have 2 or more traffic classes */
7014 if (netdev_get_num_tc(netdev
) > 1)
7017 /* We cannot enable ATR if RSS is disabled */
7018 if (adapter
->ring_feature
[RING_F_RSS
].limit
<= 1)
7021 /* A sample rate of 0 indicates ATR disabled */
7022 if (!adapter
->atr_sample_rate
)
7025 adapter
->flags
|= IXGBE_FLAG_FDIR_HASH_CAPABLE
;
7029 if (features
& NETIF_F_HW_VLAN_CTAG_RX
)
7030 ixgbe_vlan_strip_enable(adapter
);
7032 ixgbe_vlan_strip_disable(adapter
);
7034 if (changed
& NETIF_F_RXALL
)
7037 netdev
->features
= features
;
7039 ixgbe_do_reset(netdev
);
7044 static int ixgbe_ndo_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
7045 struct net_device
*dev
,
7046 const unsigned char *addr
,
7049 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
7052 if (!(adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
))
7053 return ndo_dflt_fdb_add(ndm
, tb
, dev
, addr
, flags
);
7055 /* Hardware does not support aging addresses so if a
7056 * ndm_state is given only allow permanent addresses
7058 if (ndm
->ndm_state
&& !(ndm
->ndm_state
& NUD_PERMANENT
)) {
7059 pr_info("%s: FDB only supports static addresses\n",
7064 if (is_unicast_ether_addr(addr
) || is_link_local_ether_addr(addr
)) {
7065 u32 rar_uc_entries
= IXGBE_MAX_PF_MACVLANS
;
7067 if (netdev_uc_count(dev
) < rar_uc_entries
)
7068 err
= dev_uc_add_excl(dev
, addr
);
7071 } else if (is_multicast_ether_addr(addr
)) {
7072 err
= dev_mc_add_excl(dev
, addr
);
7077 /* Only return duplicate errors if NLM_F_EXCL is set */
7078 if (err
== -EEXIST
&& !(flags
& NLM_F_EXCL
))
7084 static int ixgbe_ndo_bridge_setlink(struct net_device
*dev
,
7085 struct nlmsghdr
*nlh
)
7087 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
7088 struct nlattr
*attr
, *br_spec
;
7091 if (!(adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
))
7094 br_spec
= nlmsg_find_attr(nlh
, sizeof(struct ifinfomsg
), IFLA_AF_SPEC
);
7096 nla_for_each_nested(attr
, br_spec
, rem
) {
7100 if (nla_type(attr
) != IFLA_BRIDGE_MODE
)
7103 mode
= nla_get_u16(attr
);
7104 if (mode
== BRIDGE_MODE_VEPA
) {
7106 adapter
->flags2
&= ~IXGBE_FLAG2_BRIDGE_MODE_VEB
;
7107 } else if (mode
== BRIDGE_MODE_VEB
) {
7108 reg
= IXGBE_PFDTXGSWC_VT_LBEN
;
7109 adapter
->flags2
|= IXGBE_FLAG2_BRIDGE_MODE_VEB
;
7113 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_PFDTXGSWC
, reg
);
7115 e_info(drv
, "enabling bridge mode: %s\n",
7116 mode
== BRIDGE_MODE_VEPA
? "VEPA" : "VEB");
7122 static int ixgbe_ndo_bridge_getlink(struct sk_buff
*skb
, u32 pid
, u32 seq
,
7123 struct net_device
*dev
,
7126 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
7129 if (!(adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
))
7132 if (adapter
->flags2
& IXGBE_FLAG2_BRIDGE_MODE_VEB
)
7133 mode
= BRIDGE_MODE_VEB
;
7135 mode
= BRIDGE_MODE_VEPA
;
7137 return ndo_dflt_bridge_getlink(skb
, pid
, seq
, dev
, mode
);
7140 static const struct net_device_ops ixgbe_netdev_ops
= {
7141 .ndo_open
= ixgbe_open
,
7142 .ndo_stop
= ixgbe_close
,
7143 .ndo_start_xmit
= ixgbe_xmit_frame
,
7145 .ndo_select_queue
= ixgbe_select_queue
,
7147 .ndo_set_rx_mode
= ixgbe_set_rx_mode
,
7148 .ndo_validate_addr
= eth_validate_addr
,
7149 .ndo_set_mac_address
= ixgbe_set_mac
,
7150 .ndo_change_mtu
= ixgbe_change_mtu
,
7151 .ndo_tx_timeout
= ixgbe_tx_timeout
,
7152 .ndo_vlan_rx_add_vid
= ixgbe_vlan_rx_add_vid
,
7153 .ndo_vlan_rx_kill_vid
= ixgbe_vlan_rx_kill_vid
,
7154 .ndo_do_ioctl
= ixgbe_ioctl
,
7155 .ndo_set_vf_mac
= ixgbe_ndo_set_vf_mac
,
7156 .ndo_set_vf_vlan
= ixgbe_ndo_set_vf_vlan
,
7157 .ndo_set_vf_tx_rate
= ixgbe_ndo_set_vf_bw
,
7158 .ndo_set_vf_spoofchk
= ixgbe_ndo_set_vf_spoofchk
,
7159 .ndo_get_vf_config
= ixgbe_ndo_get_vf_config
,
7160 .ndo_get_stats64
= ixgbe_get_stats64
,
7161 #ifdef CONFIG_IXGBE_DCB
7162 .ndo_setup_tc
= ixgbe_setup_tc
,
7164 #ifdef CONFIG_NET_POLL_CONTROLLER
7165 .ndo_poll_controller
= ixgbe_netpoll
,
7168 .ndo_fcoe_ddp_setup
= ixgbe_fcoe_ddp_get
,
7169 .ndo_fcoe_ddp_target
= ixgbe_fcoe_ddp_target
,
7170 .ndo_fcoe_ddp_done
= ixgbe_fcoe_ddp_put
,
7171 .ndo_fcoe_enable
= ixgbe_fcoe_enable
,
7172 .ndo_fcoe_disable
= ixgbe_fcoe_disable
,
7173 .ndo_fcoe_get_wwn
= ixgbe_fcoe_get_wwn
,
7174 .ndo_fcoe_get_hbainfo
= ixgbe_fcoe_get_hbainfo
,
7175 #endif /* IXGBE_FCOE */
7176 .ndo_set_features
= ixgbe_set_features
,
7177 .ndo_fix_features
= ixgbe_fix_features
,
7178 .ndo_fdb_add
= ixgbe_ndo_fdb_add
,
7179 .ndo_bridge_setlink
= ixgbe_ndo_bridge_setlink
,
7180 .ndo_bridge_getlink
= ixgbe_ndo_bridge_getlink
,
7184 * ixgbe_wol_supported - Check whether device supports WoL
7185 * @hw: hw specific details
7186 * @device_id: the device ID
7187 * @subdev_id: the subsystem device ID
7189 * This function is used by probe and ethtool to determine
7190 * which devices have WoL support
7193 int ixgbe_wol_supported(struct ixgbe_adapter
*adapter
, u16 device_id
,
7196 struct ixgbe_hw
*hw
= &adapter
->hw
;
7197 u16 wol_cap
= adapter
->eeprom_cap
& IXGBE_DEVICE_CAPS_WOL_MASK
;
7198 int is_wol_supported
= 0;
7200 switch (device_id
) {
7201 case IXGBE_DEV_ID_82599_SFP
:
7202 /* Only these subdevices could supports WOL */
7203 switch (subdevice_id
) {
7204 case IXGBE_SUBDEV_ID_82599_560FLR
:
7205 /* only support first port */
7206 if (hw
->bus
.func
!= 0)
7208 case IXGBE_SUBDEV_ID_82599_SFP
:
7209 case IXGBE_SUBDEV_ID_82599_RNDC
:
7210 case IXGBE_SUBDEV_ID_82599_ECNA_DP
:
7211 case IXGBE_SUBDEV_ID_82599_LOM_SFP
:
7212 is_wol_supported
= 1;
7216 case IXGBE_DEV_ID_82599_COMBO_BACKPLANE
:
7217 /* All except this subdevice support WOL */
7218 if (subdevice_id
!= IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ
)
7219 is_wol_supported
= 1;
7221 case IXGBE_DEV_ID_82599_KX4
:
7222 is_wol_supported
= 1;
7224 case IXGBE_DEV_ID_X540T
:
7225 case IXGBE_DEV_ID_X540T1
:
7226 /* check eeprom to see if enabled wol */
7227 if ((wol_cap
== IXGBE_DEVICE_CAPS_WOL_PORT0_1
) ||
7228 ((wol_cap
== IXGBE_DEVICE_CAPS_WOL_PORT0
) &&
7229 (hw
->bus
.func
== 0))) {
7230 is_wol_supported
= 1;
7235 return is_wol_supported
;
7239 * ixgbe_probe - Device Initialization Routine
7240 * @pdev: PCI device information struct
7241 * @ent: entry in ixgbe_pci_tbl
7243 * Returns 0 on success, negative on failure
7245 * ixgbe_probe initializes an adapter identified by a pci_dev structure.
7246 * The OS initialization, configuring of the adapter private structure,
7247 * and a hardware reset occur.
7249 static int ixgbe_probe(struct pci_dev
*pdev
, const struct pci_device_id
*ent
)
7251 struct net_device
*netdev
;
7252 struct ixgbe_adapter
*adapter
= NULL
;
7253 struct ixgbe_hw
*hw
;
7254 const struct ixgbe_info
*ii
= ixgbe_info_tbl
[ent
->driver_data
];
7255 static int cards_found
;
7256 int i
, err
, pci_using_dac
;
7257 unsigned int indices
= MAX_TX_QUEUES
;
7258 u8 part_str
[IXGBE_PBANUM_LENGTH
];
7264 /* Catch broken hardware that put the wrong VF device ID in
7265 * the PCIe SR-IOV capability.
7267 if (pdev
->is_virtfn
) {
7268 WARN(1, KERN_ERR
"%s (%hx:%hx) should not be a VF!\n",
7269 pci_name(pdev
), pdev
->vendor
, pdev
->device
);
7273 err
= pci_enable_device_mem(pdev
);
7277 if (!dma_set_mask(&pdev
->dev
, DMA_BIT_MASK(64)) &&
7278 !dma_set_coherent_mask(&pdev
->dev
, DMA_BIT_MASK(64))) {
7281 err
= dma_set_mask(&pdev
->dev
, DMA_BIT_MASK(32));
7283 err
= dma_set_coherent_mask(&pdev
->dev
,
7287 "No usable DMA configuration, aborting\n");
7294 err
= pci_request_selected_regions(pdev
, pci_select_bars(pdev
,
7295 IORESOURCE_MEM
), ixgbe_driver_name
);
7298 "pci_request_selected_regions failed 0x%x\n", err
);
7302 pci_enable_pcie_error_reporting(pdev
);
7304 pci_set_master(pdev
);
7305 pci_save_state(pdev
);
7307 if (ii
->mac
== ixgbe_mac_82598EB
) {
7308 #ifdef CONFIG_IXGBE_DCB
7309 /* 8 TC w/ 4 queues per TC */
7310 indices
= 4 * MAX_TRAFFIC_CLASS
;
7312 indices
= IXGBE_MAX_RSS_INDICES
;
7316 netdev
= alloc_etherdev_mq(sizeof(struct ixgbe_adapter
), indices
);
7319 goto err_alloc_etherdev
;
7322 SET_NETDEV_DEV(netdev
, &pdev
->dev
);
7324 adapter
= netdev_priv(netdev
);
7325 pci_set_drvdata(pdev
, adapter
);
7327 adapter
->netdev
= netdev
;
7328 adapter
->pdev
= pdev
;
7331 adapter
->msg_enable
= netif_msg_init(debug
, DEFAULT_MSG_ENABLE
);
7333 hw
->hw_addr
= ioremap(pci_resource_start(pdev
, 0),
7334 pci_resource_len(pdev
, 0));
7340 netdev
->netdev_ops
= &ixgbe_netdev_ops
;
7341 ixgbe_set_ethtool_ops(netdev
);
7342 netdev
->watchdog_timeo
= 5 * HZ
;
7343 strncpy(netdev
->name
, pci_name(pdev
), sizeof(netdev
->name
) - 1);
7345 adapter
->bd_number
= cards_found
;
7348 memcpy(&hw
->mac
.ops
, ii
->mac_ops
, sizeof(hw
->mac
.ops
));
7349 hw
->mac
.type
= ii
->mac
;
7352 memcpy(&hw
->eeprom
.ops
, ii
->eeprom_ops
, sizeof(hw
->eeprom
.ops
));
7353 eec
= IXGBE_READ_REG(hw
, IXGBE_EEC
);
7354 /* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
7355 if (!(eec
& (1 << 8)))
7356 hw
->eeprom
.ops
.read
= &ixgbe_read_eeprom_bit_bang_generic
;
7359 memcpy(&hw
->phy
.ops
, ii
->phy_ops
, sizeof(hw
->phy
.ops
));
7360 hw
->phy
.sfp_type
= ixgbe_sfp_type_unknown
;
7361 /* ixgbe_identify_phy_generic will set prtad and mmds properly */
7362 hw
->phy
.mdio
.prtad
= MDIO_PRTAD_NONE
;
7363 hw
->phy
.mdio
.mmds
= 0;
7364 hw
->phy
.mdio
.mode_support
= MDIO_SUPPORTS_C45
| MDIO_EMULATE_C22
;
7365 hw
->phy
.mdio
.dev
= netdev
;
7366 hw
->phy
.mdio
.mdio_read
= ixgbe_mdio_read
;
7367 hw
->phy
.mdio
.mdio_write
= ixgbe_mdio_write
;
7369 ii
->get_invariants(hw
);
7371 /* setup the private structure */
7372 err
= ixgbe_sw_init(adapter
);
7376 /* Cache if MNG FW is up so we don't have to read the REG later */
7377 if (hw
->mac
.ops
.mng_fw_enabled
)
7378 hw
->mng_fw_enabled
= hw
->mac
.ops
.mng_fw_enabled(hw
);
7380 /* Make it possible the adapter to be woken up via WOL */
7381 switch (adapter
->hw
.mac
.type
) {
7382 case ixgbe_mac_82599EB
:
7383 case ixgbe_mac_X540
:
7384 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_WUS
, ~0);
7391 * If there is a fan on this device and it has failed log the
7394 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
) {
7395 u32 esdp
= IXGBE_READ_REG(hw
, IXGBE_ESDP
);
7396 if (esdp
& IXGBE_ESDP_SDP1
)
7397 e_crit(probe
, "Fan has stopped, replace the adapter\n");
7400 if (allow_unsupported_sfp
)
7401 hw
->allow_unsupported_sfp
= allow_unsupported_sfp
;
7403 /* reset_hw fills in the perm_addr as well */
7404 hw
->phy
.reset_if_overtemp
= true;
7405 err
= hw
->mac
.ops
.reset_hw(hw
);
7406 hw
->phy
.reset_if_overtemp
= false;
7407 if (err
== IXGBE_ERR_SFP_NOT_PRESENT
&&
7408 hw
->mac
.type
== ixgbe_mac_82598EB
) {
7410 } else if (err
== IXGBE_ERR_SFP_NOT_SUPPORTED
) {
7411 e_dev_err("failed to load because an unsupported SFP+ "
7412 "module type was detected.\n");
7413 e_dev_err("Reload the driver after installing a supported "
7417 e_dev_err("HW Init failed: %d\n", err
);
7421 #ifdef CONFIG_PCI_IOV
7422 /* SR-IOV not supported on the 82598 */
7423 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
)
7426 ixgbe_init_mbx_params_pf(hw
);
7427 memcpy(&hw
->mbx
.ops
, ii
->mbx_ops
, sizeof(hw
->mbx
.ops
));
7428 ixgbe_enable_sriov(adapter
);
7429 pci_sriov_set_totalvfs(pdev
, 63);
7433 netdev
->features
= NETIF_F_SG
|
7436 NETIF_F_HW_VLAN_CTAG_TX
|
7437 NETIF_F_HW_VLAN_CTAG_RX
|
7438 NETIF_F_HW_VLAN_CTAG_FILTER
|
7444 netdev
->hw_features
= netdev
->features
;
7446 switch (adapter
->hw
.mac
.type
) {
7447 case ixgbe_mac_82599EB
:
7448 case ixgbe_mac_X540
:
7449 netdev
->features
|= NETIF_F_SCTP_CSUM
;
7450 netdev
->hw_features
|= NETIF_F_SCTP_CSUM
|
7457 netdev
->hw_features
|= NETIF_F_RXALL
;
7459 netdev
->vlan_features
|= NETIF_F_TSO
;
7460 netdev
->vlan_features
|= NETIF_F_TSO6
;
7461 netdev
->vlan_features
|= NETIF_F_IP_CSUM
;
7462 netdev
->vlan_features
|= NETIF_F_IPV6_CSUM
;
7463 netdev
->vlan_features
|= NETIF_F_SG
;
7465 netdev
->priv_flags
|= IFF_UNICAST_FLT
;
7466 netdev
->priv_flags
|= IFF_SUPP_NOFCS
;
7468 #ifdef CONFIG_IXGBE_DCB
7469 netdev
->dcbnl_ops
= &dcbnl_ops
;
7473 if (adapter
->flags
& IXGBE_FLAG_FCOE_CAPABLE
) {
7474 unsigned int fcoe_l
;
7476 if (hw
->mac
.ops
.get_device_caps
) {
7477 hw
->mac
.ops
.get_device_caps(hw
, &device_caps
);
7478 if (device_caps
& IXGBE_DEVICE_CAPS_FCOE_OFFLOADS
)
7479 adapter
->flags
&= ~IXGBE_FLAG_FCOE_CAPABLE
;
7483 fcoe_l
= min_t(int, IXGBE_FCRETA_SIZE
, num_online_cpus());
7484 adapter
->ring_feature
[RING_F_FCOE
].limit
= fcoe_l
;
7486 netdev
->features
|= NETIF_F_FSO
|
7489 netdev
->vlan_features
|= NETIF_F_FSO
|
7493 #endif /* IXGBE_FCOE */
7494 if (pci_using_dac
) {
7495 netdev
->features
|= NETIF_F_HIGHDMA
;
7496 netdev
->vlan_features
|= NETIF_F_HIGHDMA
;
7499 if (adapter
->flags2
& IXGBE_FLAG2_RSC_CAPABLE
)
7500 netdev
->hw_features
|= NETIF_F_LRO
;
7501 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)
7502 netdev
->features
|= NETIF_F_LRO
;
7504 /* make sure the EEPROM is good */
7505 if (hw
->eeprom
.ops
.validate_checksum(hw
, NULL
) < 0) {
7506 e_dev_err("The EEPROM Checksum Is Not Valid\n");
7511 memcpy(netdev
->dev_addr
, hw
->mac
.perm_addr
, netdev
->addr_len
);
7513 if (!is_valid_ether_addr(netdev
->dev_addr
)) {
7514 e_dev_err("invalid MAC address\n");
7519 setup_timer(&adapter
->service_timer
, &ixgbe_service_timer
,
7520 (unsigned long) adapter
);
7522 INIT_WORK(&adapter
->service_task
, ixgbe_service_task
);
7523 clear_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
);
7525 err
= ixgbe_init_interrupt_scheme(adapter
);
7529 /* WOL not supported for all devices */
7531 hw
->eeprom
.ops
.read(hw
, 0x2c, &adapter
->eeprom_cap
);
7532 hw
->wol_supported
= ixgbe_wol_supported(adapter
, pdev
->device
,
7533 pdev
->subsystem_device
);
7534 if (hw
->wol_supported
)
7535 adapter
->wol
= IXGBE_WUFC_MAG
;
7537 device_set_wakeup_enable(&adapter
->pdev
->dev
, adapter
->wol
);
7539 /* save off EEPROM version number */
7540 hw
->eeprom
.ops
.read(hw
, 0x2e, &adapter
->eeprom_verh
);
7541 hw
->eeprom
.ops
.read(hw
, 0x2d, &adapter
->eeprom_verl
);
7543 /* pick up the PCI bus settings for reporting later */
7544 hw
->mac
.ops
.get_bus_info(hw
);
7545 if (hw
->device_id
== IXGBE_DEV_ID_82599_SFP_SF_QP
)
7546 ixgbe_get_parent_bus_info(adapter
);
7548 /* print bus type/speed/width info */
7549 e_dev_info("(PCI Express:%s:%s) %pM\n",
7550 (hw
->bus
.speed
== ixgbe_bus_speed_8000
? "8.0GT/s" :
7551 hw
->bus
.speed
== ixgbe_bus_speed_5000
? "5.0GT/s" :
7552 hw
->bus
.speed
== ixgbe_bus_speed_2500
? "2.5GT/s" :
7554 (hw
->bus
.width
== ixgbe_bus_width_pcie_x8
? "Width x8" :
7555 hw
->bus
.width
== ixgbe_bus_width_pcie_x4
? "Width x4" :
7556 hw
->bus
.width
== ixgbe_bus_width_pcie_x1
? "Width x1" :
7560 err
= ixgbe_read_pba_string_generic(hw
, part_str
, IXGBE_PBANUM_LENGTH
);
7562 strncpy(part_str
, "Unknown", IXGBE_PBANUM_LENGTH
);
7563 if (ixgbe_is_sfp(hw
) && hw
->phy
.sfp_type
!= ixgbe_sfp_type_not_present
)
7564 e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
7565 hw
->mac
.type
, hw
->phy
.type
, hw
->phy
.sfp_type
,
7568 e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
7569 hw
->mac
.type
, hw
->phy
.type
, part_str
);
7571 if (hw
->bus
.width
<= ixgbe_bus_width_pcie_x4
) {
7572 e_dev_warn("PCI-Express bandwidth available for this card is "
7573 "not sufficient for optimal performance.\n");
7574 e_dev_warn("For optimal performance a x8 PCI-Express slot "
7578 /* reset the hardware with the new settings */
7579 err
= hw
->mac
.ops
.start_hw(hw
);
7580 if (err
== IXGBE_ERR_EEPROM_VERSION
) {
7581 /* We are running on a pre-production device, log a warning */
7582 e_dev_warn("This device is a pre-production adapter/LOM. "
7583 "Please be aware there may be issues associated "
7584 "with your hardware. If you are experiencing "
7585 "problems please contact your Intel or hardware "
7586 "representative who provided you with this "
7589 strcpy(netdev
->name
, "eth%d");
7590 err
= register_netdev(netdev
);
7594 /* power down the optics for 82599 SFP+ fiber */
7595 if (hw
->mac
.ops
.disable_tx_laser
)
7596 hw
->mac
.ops
.disable_tx_laser(hw
);
7598 /* carrier off reporting is important to ethtool even BEFORE open */
7599 netif_carrier_off(netdev
);
7601 #ifdef CONFIG_IXGBE_DCA
7602 if (dca_add_requester(&pdev
->dev
) == 0) {
7603 adapter
->flags
|= IXGBE_FLAG_DCA_ENABLED
;
7604 ixgbe_setup_dca(adapter
);
7607 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
7608 e_info(probe
, "IOV is enabled with %d VFs\n", adapter
->num_vfs
);
7609 for (i
= 0; i
< adapter
->num_vfs
; i
++)
7610 ixgbe_vf_configuration(pdev
, (i
| 0x10000000));
7613 /* firmware requires driver version to be 0xFFFFFFFF
7614 * since os does not support feature
7616 if (hw
->mac
.ops
.set_fw_drv_ver
)
7617 hw
->mac
.ops
.set_fw_drv_ver(hw
, 0xFF, 0xFF, 0xFF,
7620 /* add san mac addr to netdev */
7621 ixgbe_add_sanmac_netdev(netdev
);
7623 e_dev_info("%s\n", ixgbe_default_device_descr
);
7626 #ifdef CONFIG_IXGBE_HWMON
7627 if (ixgbe_sysfs_init(adapter
))
7628 e_err(probe
, "failed to allocate sysfs resources\n");
7629 #endif /* CONFIG_IXGBE_HWMON */
7631 ixgbe_dbg_adapter_init(adapter
);
7633 /* Need link setup for MNG FW, else wait for IXGBE_UP */
7634 if (hw
->mng_fw_enabled
&& hw
->mac
.ops
.setup_link
)
7635 hw
->mac
.ops
.setup_link(hw
,
7636 IXGBE_LINK_SPEED_10GB_FULL
| IXGBE_LINK_SPEED_1GB_FULL
,
7642 ixgbe_release_hw_control(adapter
);
7643 ixgbe_clear_interrupt_scheme(adapter
);
7645 ixgbe_disable_sriov(adapter
);
7646 adapter
->flags2
&= ~IXGBE_FLAG2_SEARCH_FOR_SFP
;
7647 iounmap(hw
->hw_addr
);
7649 free_netdev(netdev
);
7651 pci_release_selected_regions(pdev
,
7652 pci_select_bars(pdev
, IORESOURCE_MEM
));
7655 pci_disable_device(pdev
);
7660 * ixgbe_remove - Device Removal Routine
7661 * @pdev: PCI device information struct
7663 * ixgbe_remove is called by the PCI subsystem to alert the driver
7664 * that it should release a PCI device. The could be caused by a
7665 * Hot-Plug event, or because the driver is going to be removed from
7668 static void ixgbe_remove(struct pci_dev
*pdev
)
7670 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
7671 struct net_device
*netdev
= adapter
->netdev
;
7673 ixgbe_dbg_adapter_exit(adapter
);
7675 set_bit(__IXGBE_DOWN
, &adapter
->state
);
7676 cancel_work_sync(&adapter
->service_task
);
7679 #ifdef CONFIG_IXGBE_DCA
7680 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
) {
7681 adapter
->flags
&= ~IXGBE_FLAG_DCA_ENABLED
;
7682 dca_remove_requester(&pdev
->dev
);
7683 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_DCA_CTRL
, 1);
7687 #ifdef CONFIG_IXGBE_HWMON
7688 ixgbe_sysfs_exit(adapter
);
7689 #endif /* CONFIG_IXGBE_HWMON */
7691 /* remove the added san mac */
7692 ixgbe_del_sanmac_netdev(netdev
);
7694 if (netdev
->reg_state
== NETREG_REGISTERED
)
7695 unregister_netdev(netdev
);
7697 #ifdef CONFIG_PCI_IOV
7699 * Only disable SR-IOV on unload if the user specified the now
7700 * deprecated max_vfs module parameter.
7703 ixgbe_disable_sriov(adapter
);
7705 ixgbe_clear_interrupt_scheme(adapter
);
7707 ixgbe_release_hw_control(adapter
);
7710 kfree(adapter
->ixgbe_ieee_pfc
);
7711 kfree(adapter
->ixgbe_ieee_ets
);
7714 iounmap(adapter
->hw
.hw_addr
);
7715 pci_release_selected_regions(pdev
, pci_select_bars(pdev
,
7718 e_dev_info("complete\n");
7720 free_netdev(netdev
);
7722 pci_disable_pcie_error_reporting(pdev
);
7724 pci_disable_device(pdev
);
7728 * ixgbe_io_error_detected - called when PCI error is detected
7729 * @pdev: Pointer to PCI device
7730 * @state: The current pci connection state
7732 * This function is called after a PCI bus error affecting
7733 * this device has been detected.
7735 static pci_ers_result_t
ixgbe_io_error_detected(struct pci_dev
*pdev
,
7736 pci_channel_state_t state
)
7738 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
7739 struct net_device
*netdev
= adapter
->netdev
;
7741 #ifdef CONFIG_PCI_IOV
7742 struct pci_dev
*bdev
, *vfdev
;
7743 u32 dw0
, dw1
, dw2
, dw3
;
7745 u16 req_id
, pf_func
;
7747 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
||
7748 adapter
->num_vfs
== 0)
7749 goto skip_bad_vf_detection
;
7751 bdev
= pdev
->bus
->self
;
7752 while (bdev
&& (pci_pcie_type(bdev
) != PCI_EXP_TYPE_ROOT_PORT
))
7753 bdev
= bdev
->bus
->self
;
7756 goto skip_bad_vf_detection
;
7758 pos
= pci_find_ext_capability(bdev
, PCI_EXT_CAP_ID_ERR
);
7760 goto skip_bad_vf_detection
;
7762 pci_read_config_dword(bdev
, pos
+ PCI_ERR_HEADER_LOG
, &dw0
);
7763 pci_read_config_dword(bdev
, pos
+ PCI_ERR_HEADER_LOG
+ 4, &dw1
);
7764 pci_read_config_dword(bdev
, pos
+ PCI_ERR_HEADER_LOG
+ 8, &dw2
);
7765 pci_read_config_dword(bdev
, pos
+ PCI_ERR_HEADER_LOG
+ 12, &dw3
);
7768 /* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
7769 if (!(req_id
& 0x0080))
7770 goto skip_bad_vf_detection
;
7772 pf_func
= req_id
& 0x01;
7773 if ((pf_func
& 1) == (pdev
->devfn
& 1)) {
7774 unsigned int device_id
;
7776 vf
= (req_id
& 0x7F) >> 1;
7777 e_dev_err("VF %d has caused a PCIe error\n", vf
);
7778 e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
7779 "%8.8x\tdw3: %8.8x\n",
7780 dw0
, dw1
, dw2
, dw3
);
7781 switch (adapter
->hw
.mac
.type
) {
7782 case ixgbe_mac_82599EB
:
7783 device_id
= IXGBE_82599_VF_DEVICE_ID
;
7785 case ixgbe_mac_X540
:
7786 device_id
= IXGBE_X540_VF_DEVICE_ID
;
7793 /* Find the pci device of the offending VF */
7794 vfdev
= pci_get_device(PCI_VENDOR_ID_INTEL
, device_id
, NULL
);
7796 if (vfdev
->devfn
== (req_id
& 0xFF))
7798 vfdev
= pci_get_device(PCI_VENDOR_ID_INTEL
,
7802 * There's a slim chance the VF could have been hot plugged,
7803 * so if it is no longer present we don't need to issue the
7804 * VFLR. Just clean up the AER in that case.
7807 e_dev_err("Issuing VFLR to VF %d\n", vf
);
7808 pci_write_config_dword(vfdev
, 0xA8, 0x00008000);
7809 /* Free device reference count */
7813 pci_cleanup_aer_uncorrect_error_status(pdev
);
7817 * Even though the error may have occurred on the other port
7818 * we still need to increment the vf error reference count for
7819 * both ports because the I/O resume function will be called
7822 adapter
->vferr_refcount
++;
7824 return PCI_ERS_RESULT_RECOVERED
;
7826 skip_bad_vf_detection
:
7827 #endif /* CONFIG_PCI_IOV */
7828 netif_device_detach(netdev
);
7830 if (state
== pci_channel_io_perm_failure
)
7831 return PCI_ERS_RESULT_DISCONNECT
;
7833 if (netif_running(netdev
))
7834 ixgbe_down(adapter
);
7835 pci_disable_device(pdev
);
7837 /* Request a slot reset. */
7838 return PCI_ERS_RESULT_NEED_RESET
;
7842 * ixgbe_io_slot_reset - called after the pci bus has been reset.
7843 * @pdev: Pointer to PCI device
7845 * Restart the card from scratch, as if from a cold-boot.
7847 static pci_ers_result_t
ixgbe_io_slot_reset(struct pci_dev
*pdev
)
7849 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
7850 pci_ers_result_t result
;
7853 if (pci_enable_device_mem(pdev
)) {
7854 e_err(probe
, "Cannot re-enable PCI device after reset.\n");
7855 result
= PCI_ERS_RESULT_DISCONNECT
;
7857 pci_set_master(pdev
);
7858 pci_restore_state(pdev
);
7859 pci_save_state(pdev
);
7861 pci_wake_from_d3(pdev
, false);
7863 ixgbe_reset(adapter
);
7864 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_WUS
, ~0);
7865 result
= PCI_ERS_RESULT_RECOVERED
;
7868 err
= pci_cleanup_aer_uncorrect_error_status(pdev
);
7870 e_dev_err("pci_cleanup_aer_uncorrect_error_status "
7871 "failed 0x%0x\n", err
);
7872 /* non-fatal, continue */
7879 * ixgbe_io_resume - called when traffic can start flowing again.
7880 * @pdev: Pointer to PCI device
7882 * This callback is called when the error recovery driver tells us that
7883 * its OK to resume normal operation.
7885 static void ixgbe_io_resume(struct pci_dev
*pdev
)
7887 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
7888 struct net_device
*netdev
= adapter
->netdev
;
7890 #ifdef CONFIG_PCI_IOV
7891 if (adapter
->vferr_refcount
) {
7892 e_info(drv
, "Resuming after VF err\n");
7893 adapter
->vferr_refcount
--;
7898 if (netif_running(netdev
))
7901 netif_device_attach(netdev
);
7904 static const struct pci_error_handlers ixgbe_err_handler
= {
7905 .error_detected
= ixgbe_io_error_detected
,
7906 .slot_reset
= ixgbe_io_slot_reset
,
7907 .resume
= ixgbe_io_resume
,
7910 static struct pci_driver ixgbe_driver
= {
7911 .name
= ixgbe_driver_name
,
7912 .id_table
= ixgbe_pci_tbl
,
7913 .probe
= ixgbe_probe
,
7914 .remove
= ixgbe_remove
,
7916 .suspend
= ixgbe_suspend
,
7917 .resume
= ixgbe_resume
,
7919 .shutdown
= ixgbe_shutdown
,
7920 .sriov_configure
= ixgbe_pci_sriov_configure
,
7921 .err_handler
= &ixgbe_err_handler
7925 * ixgbe_init_module - Driver Registration Routine
7927 * ixgbe_init_module is the first routine called when the driver is
7928 * loaded. All it does is register with the PCI subsystem.
7930 static int __init
ixgbe_init_module(void)
7933 pr_info("%s - version %s\n", ixgbe_driver_string
, ixgbe_driver_version
);
7934 pr_info("%s\n", ixgbe_copyright
);
7938 ret
= pci_register_driver(&ixgbe_driver
);
7944 #ifdef CONFIG_IXGBE_DCA
7945 dca_register_notify(&dca_notifier
);
7951 module_init(ixgbe_init_module
);
7954 * ixgbe_exit_module - Driver Exit Cleanup Routine
7956 * ixgbe_exit_module is called just before the driver is removed
7959 static void __exit
ixgbe_exit_module(void)
7961 #ifdef CONFIG_IXGBE_DCA
7962 dca_unregister_notify(&dca_notifier
);
7964 pci_unregister_driver(&ixgbe_driver
);
7968 rcu_barrier(); /* Wait for completion of call_rcu()'s */
7971 #ifdef CONFIG_IXGBE_DCA
7972 static int ixgbe_notify_dca(struct notifier_block
*nb
, unsigned long event
,
7977 ret_val
= driver_for_each_device(&ixgbe_driver
.driver
, NULL
, &event
,
7978 __ixgbe_notify_dca
);
7980 return ret_val
? NOTIFY_BAD
: NOTIFY_DONE
;
7983 #endif /* CONFIG_IXGBE_DCA */
7985 module_exit(ixgbe_exit_module
);