]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - drivers/net/ethernet/ibm/ibmvnic.c
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:


352ffe982d849b2bef82968d8c34fee390235274
[mirror_ubuntu-jammy-kernel.git] / drivers / net / ethernet / ibm / ibmvnic.c
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /**************************************************************************/
3 /* */
4 /* IBM System i and System p Virtual NIC Device Driver */
5 /* Copyright (C) 2014 IBM Corp. */
6 /* Santiago Leon (santi_leon@yahoo.com) */
7 /* Thomas Falcon (tlfalcon@linux.vnet.ibm.com) */
8 /* John Allen (jallen@linux.vnet.ibm.com) */
9 /* */
10 /* */
11 /* This module contains the implementation of a virtual ethernet device */
12 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN */
13 /* option of the RS/6000 Platform Architecture to interface with virtual */
14 /* ethernet NICs that are presented to the partition by the hypervisor. */
15 /* */
16 /* Messages are passed between the VNIC driver and the VNIC server using */
17 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to */
18 /* issue and receive commands that initiate communication with the server */
19 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but */
20 /* are used by the driver to notify the server that a packet is */
21 /* ready for transmission or that a buffer has been added to receive a */
22 /* packet. Subsequently, sCRQs are used by the server to notify the */
23 /* driver that a packet transmission has been completed or that a packet */
24 /* has been received and placed in a waiting buffer. */
25 /* */
26 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in */
27 /* which skbs are DMA mapped and immediately unmapped when the transmit */
28 /* or receive has been completed, the VNIC driver is required to use */
29 /* "long term mapping". This entails that large, continuous DMA mapped */
30 /* buffers are allocated on driver initialization and these buffers are */
31 /* then continuously reused to pass skbs to and from the VNIC server. */
32 /* */
33 /**************************************************************************/
34
35 #include <linux/module.h>
36 #include <linux/moduleparam.h>
37 #include <linux/types.h>
38 #include <linux/errno.h>
39 #include <linux/completion.h>
40 #include <linux/ioport.h>
41 #include <linux/dma-mapping.h>
42 #include <linux/kernel.h>
43 #include <linux/netdevice.h>
44 #include <linux/etherdevice.h>
45 #include <linux/skbuff.h>
46 #include <linux/init.h>
47 #include <linux/delay.h>
48 #include <linux/mm.h>
49 #include <linux/ethtool.h>
50 #include <linux/proc_fs.h>
51 #include <linux/if_arp.h>
52 #include <linux/in.h>
53 #include <linux/ip.h>
54 #include <linux/ipv6.h>
55 #include <linux/irq.h>
56 #include <linux/kthread.h>
57 #include <linux/seq_file.h>
58 #include <linux/interrupt.h>
59 #include <net/net_namespace.h>
60 #include <asm/hvcall.h>
61 #include <linux/atomic.h>
62 #include <asm/vio.h>
63 #include <asm/iommu.h>
64 #include <linux/uaccess.h>
65 #include <asm/firmware.h>
66 #include <linux/workqueue.h>
67 #include <linux/if_vlan.h>
68 #include <linux/utsname.h>
69
70 #include "ibmvnic.h"
71
72 static const char ibmvnic_driver_name[] = "ibmvnic";
73 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
74
75 MODULE_AUTHOR("Santiago Leon");
76 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
77 MODULE_LICENSE("GPL");
78 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
79
80 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
81 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
82 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
83 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
84 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
85 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
86 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
87 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
88 static int enable_scrq_irq(struct ibmvnic_adapter *,
89 struct ibmvnic_sub_crq_queue *);
90 static int disable_scrq_irq(struct ibmvnic_adapter *,
91 struct ibmvnic_sub_crq_queue *);
92 static int pending_scrq(struct ibmvnic_adapter *,
93 struct ibmvnic_sub_crq_queue *);
94 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
95 struct ibmvnic_sub_crq_queue *);
96 static int ibmvnic_poll(struct napi_struct *napi, int data);
97 static void send_query_map(struct ibmvnic_adapter *adapter);
98 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, u32, u8);
99 static int send_request_unmap(struct ibmvnic_adapter *, u8);
100 static int send_login(struct ibmvnic_adapter *adapter);
101 static void send_query_cap(struct ibmvnic_adapter *adapter);
102 static int init_sub_crqs(struct ibmvnic_adapter *);
103 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
104 static int ibmvnic_reset_init(struct ibmvnic_adapter *, bool reset);
105 static void release_crq_queue(struct ibmvnic_adapter *);
106 static int __ibmvnic_set_mac(struct net_device *, u8 *);
107 static int init_crq_queue(struct ibmvnic_adapter *adapter);
108 static int send_query_phys_parms(struct ibmvnic_adapter *adapter);
109 static void ibmvnic_tx_scrq_clean_buffer(struct ibmvnic_adapter *adapter,
110 struct ibmvnic_sub_crq_queue *tx_scrq);
111
112 struct ibmvnic_stat {
113 char name[ETH_GSTRING_LEN];
114 int offset;
115 };
116
117 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
118 offsetof(struct ibmvnic_statistics, stat))
119 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + (off))))
120
121 static const struct ibmvnic_stat ibmvnic_stats[] = {
122 {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
123 {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
124 {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
125 {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
126 {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
127 {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
128 {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
129 {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
130 {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
131 {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
132 {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
133 {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
134 {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
135 {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
136 {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
137 {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
138 {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
139 {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
140 {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
141 {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
142 {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
143 {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
144 };
145
146 static int send_crq_init_complete(struct ibmvnic_adapter *adapter)
147 {
148 union ibmvnic_crq crq;
149
150 memset(&crq, 0, sizeof(crq));
151 crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
152 crq.generic.cmd = IBMVNIC_CRQ_INIT_COMPLETE;
153
154 return ibmvnic_send_crq(adapter, &crq);
155 }
156
157 static int send_version_xchg(struct ibmvnic_adapter *adapter)
158 {
159 union ibmvnic_crq crq;
160
161 memset(&crq, 0, sizeof(crq));
162 crq.version_exchange.first = IBMVNIC_CRQ_CMD;
163 crq.version_exchange.cmd = VERSION_EXCHANGE;
164 crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
165
166 return ibmvnic_send_crq(adapter, &crq);
167 }
168
169 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
170 unsigned long length, unsigned long *number,
171 unsigned long *irq)
172 {
173 unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
174 long rc;
175
176 rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
177 *number = retbuf[0];
178 *irq = retbuf[1];
179
180 return rc;
181 }
182
183 /**
184 * ibmvnic_wait_for_completion - Check device state and wait for completion
185 * @adapter: private device data
186 * @comp_done: completion structure to wait for
187 * @timeout: time to wait in milliseconds
188 *
189 * Wait for a completion signal or until the timeout limit is reached
190 * while checking that the device is still active.
191 */
192 static int ibmvnic_wait_for_completion(struct ibmvnic_adapter *adapter,
193 struct completion *comp_done,
194 unsigned long timeout)
195 {
196 struct net_device *netdev;
197 unsigned long div_timeout;
198 u8 retry;
199
200 netdev = adapter->netdev;
201 retry = 5;
202 div_timeout = msecs_to_jiffies(timeout / retry);
203 while (true) {
204 if (!adapter->crq.active) {
205 netdev_err(netdev, "Device down!\n");
206 return -ENODEV;
207 }
208 if (!retry--)
209 break;
210 if (wait_for_completion_timeout(comp_done, div_timeout))
211 return 0;
212 }
213 netdev_err(netdev, "Operation timed out.\n");
214 return -ETIMEDOUT;
215 }
216
217 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
218 struct ibmvnic_long_term_buff *ltb, int size)
219 {
220 struct device *dev = &adapter->vdev->dev;
221 int rc;
222
223 ltb->size = size;
224 ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
225 GFP_KERNEL);
226
227 if (!ltb->buff) {
228 dev_err(dev, "Couldn't alloc long term buffer\n");
229 return -ENOMEM;
230 }
231 ltb->map_id = adapter->map_id;
232 adapter->map_id++;
233
234 mutex_lock(&adapter->fw_lock);
235 adapter->fw_done_rc = 0;
236 reinit_completion(&adapter->fw_done);
237
238 rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
239 if (rc) {
240 dev_err(dev, "send_request_map failed, rc = %d\n", rc);
241 goto out;
242 }
243
244 rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
245 if (rc) {
246 dev_err(dev,
247 "Long term map request aborted or timed out,rc = %d\n",
248 rc);
249 goto out;
250 }
251
252 if (adapter->fw_done_rc) {
253 dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
254 adapter->fw_done_rc);
255 rc = -1;
256 goto out;
257 }
258 rc = 0;
259 out:
260 if (rc) {
261 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
262 ltb->buff = NULL;
263 }
264 mutex_unlock(&adapter->fw_lock);
265 return rc;
266 }
267
268 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
269 struct ibmvnic_long_term_buff *ltb)
270 {
271 struct device *dev = &adapter->vdev->dev;
272
273 if (!ltb->buff)
274 return;
275
276 /* VIOS automatically unmaps the long term buffer at remote
277 * end for the following resets:
278 * FAILOVER, MOBILITY, TIMEOUT.
279 */
280 if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
281 adapter->reset_reason != VNIC_RESET_MOBILITY &&
282 adapter->reset_reason != VNIC_RESET_TIMEOUT)
283 send_request_unmap(adapter, ltb->map_id);
284 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
285 ltb->buff = NULL;
286 ltb->map_id = 0;
287 }
288
289 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
290 struct ibmvnic_long_term_buff *ltb)
291 {
292 struct device *dev = &adapter->vdev->dev;
293 int rc;
294
295 memset(ltb->buff, 0, ltb->size);
296
297 mutex_lock(&adapter->fw_lock);
298 adapter->fw_done_rc = 0;
299
300 reinit_completion(&adapter->fw_done);
301 rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
302 if (rc) {
303 mutex_unlock(&adapter->fw_lock);
304 return rc;
305 }
306
307 rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
308 if (rc) {
309 dev_info(dev,
310 "Reset failed, long term map request timed out or aborted\n");
311 mutex_unlock(&adapter->fw_lock);
312 return rc;
313 }
314
315 if (adapter->fw_done_rc) {
316 dev_info(dev,
317 "Reset failed, attempting to free and reallocate buffer\n");
318 free_long_term_buff(adapter, ltb);
319 mutex_unlock(&adapter->fw_lock);
320 return alloc_long_term_buff(adapter, ltb, ltb->size);
321 }
322 mutex_unlock(&adapter->fw_lock);
323 return 0;
324 }
325
326 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
327 {
328 int i;
329
330 for (i = 0; i < adapter->num_active_rx_pools; i++)
331 adapter->rx_pool[i].active = 0;
332 }
333
334 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
335 struct ibmvnic_rx_pool *pool)
336 {
337 int count = pool->size - atomic_read(&pool->available);
338 u64 handle = adapter->rx_scrq[pool->index]->handle;
339 struct device *dev = &adapter->vdev->dev;
340 struct ibmvnic_ind_xmit_queue *ind_bufp;
341 struct ibmvnic_sub_crq_queue *rx_scrq;
342 union sub_crq *sub_crq;
343 int buffers_added = 0;
344 unsigned long lpar_rc;
345 struct sk_buff *skb;
346 unsigned int offset;
347 dma_addr_t dma_addr;
348 unsigned char *dst;
349 int shift = 0;
350 int index;
351 int i;
352
353 if (!pool->active)
354 return;
355
356 rx_scrq = adapter->rx_scrq[pool->index];
357 ind_bufp = &rx_scrq->ind_buf;
358
359 /* netdev_skb_alloc() could have failed after we saved a few skbs
360 * in the indir_buf and we would not have sent them to VIOS yet.
361 * To account for them, start the loop at ind_bufp->index rather
362 * than 0. If we pushed all the skbs to VIOS, ind_bufp->index will
363 * be 0.
364 */
365 for (i = ind_bufp->index; i < count; ++i) {
366 skb = netdev_alloc_skb(adapter->netdev, pool->buff_size);
367 if (!skb) {
368 dev_err(dev, "Couldn't replenish rx buff\n");
369 adapter->replenish_no_mem++;
370 break;
371 }
372
373 index = pool->free_map[pool->next_free];
374
375 if (pool->rx_buff[index].skb)
376 dev_err(dev, "Inconsistent free_map!\n");
377
378 /* Copy the skb to the long term mapped DMA buffer */
379 offset = index * pool->buff_size;
380 dst = pool->long_term_buff.buff + offset;
381 memset(dst, 0, pool->buff_size);
382 dma_addr = pool->long_term_buff.addr + offset;
383 pool->rx_buff[index].data = dst;
384
385 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
386 pool->rx_buff[index].dma = dma_addr;
387 pool->rx_buff[index].skb = skb;
388 pool->rx_buff[index].pool_index = pool->index;
389 pool->rx_buff[index].size = pool->buff_size;
390
391 sub_crq = &ind_bufp->indir_arr[ind_bufp->index++];
392 memset(sub_crq, 0, sizeof(*sub_crq));
393 sub_crq->rx_add.first = IBMVNIC_CRQ_CMD;
394 sub_crq->rx_add.correlator =
395 cpu_to_be64((u64)&pool->rx_buff[index]);
396 sub_crq->rx_add.ioba = cpu_to_be32(dma_addr);
397 sub_crq->rx_add.map_id = pool->long_term_buff.map_id;
398
399 /* The length field of the sCRQ is defined to be 24 bits so the
400 * buffer size needs to be left shifted by a byte before it is
401 * converted to big endian to prevent the last byte from being
402 * truncated.
403 */
404 #ifdef __LITTLE_ENDIAN__
405 shift = 8;
406 #endif
407 sub_crq->rx_add.len = cpu_to_be32(pool->buff_size << shift);
408 pool->next_free = (pool->next_free + 1) % pool->size;
409 if (ind_bufp->index == IBMVNIC_MAX_IND_DESCS ||
410 i == count - 1) {
411 lpar_rc =
412 send_subcrq_indirect(adapter, handle,
413 (u64)ind_bufp->indir_dma,
414 (u64)ind_bufp->index);
415 if (lpar_rc != H_SUCCESS)
416 goto failure;
417 buffers_added += ind_bufp->index;
418 adapter->replenish_add_buff_success += ind_bufp->index;
419 ind_bufp->index = 0;
420 }
421 }
422 atomic_add(buffers_added, &pool->available);
423 return;
424
425 failure:
426 if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
427 dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
428 for (i = ind_bufp->index - 1; i >= 0; --i) {
429 struct ibmvnic_rx_buff *rx_buff;
430
431 pool->next_free = pool->next_free == 0 ?
432 pool->size - 1 : pool->next_free - 1;
433 sub_crq = &ind_bufp->indir_arr[i];
434 rx_buff = (struct ibmvnic_rx_buff *)
435 be64_to_cpu(sub_crq->rx_add.correlator);
436 index = (int)(rx_buff - pool->rx_buff);
437 pool->free_map[pool->next_free] = index;
438 dev_kfree_skb_any(pool->rx_buff[index].skb);
439 pool->rx_buff[index].skb = NULL;
440 }
441 adapter->replenish_add_buff_failure += ind_bufp->index;
442 atomic_add(buffers_added, &pool->available);
443 ind_bufp->index = 0;
444 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
445 /* Disable buffer pool replenishment and report carrier off if
446 * queue is closed or pending failover.
447 * Firmware guarantees that a signal will be sent to the
448 * driver, triggering a reset.
449 */
450 deactivate_rx_pools(adapter);
451 netif_carrier_off(adapter->netdev);
452 }
453 }
454
455 static void replenish_pools(struct ibmvnic_adapter *adapter)
456 {
457 int i;
458
459 adapter->replenish_task_cycles++;
460 for (i = 0; i < adapter->num_active_rx_pools; i++) {
461 if (adapter->rx_pool[i].active)
462 replenish_rx_pool(adapter, &adapter->rx_pool[i]);
463 }
464
465 netdev_dbg(adapter->netdev, "Replenished %d pools\n", i);
466 }
467
468 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
469 {
470 kfree(adapter->tx_stats_buffers);
471 kfree(adapter->rx_stats_buffers);
472 adapter->tx_stats_buffers = NULL;
473 adapter->rx_stats_buffers = NULL;
474 }
475
476 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
477 {
478 adapter->tx_stats_buffers =
479 kcalloc(IBMVNIC_MAX_QUEUES,
480 sizeof(struct ibmvnic_tx_queue_stats),
481 GFP_KERNEL);
482 if (!adapter->tx_stats_buffers)
483 return -ENOMEM;
484
485 adapter->rx_stats_buffers =
486 kcalloc(IBMVNIC_MAX_QUEUES,
487 sizeof(struct ibmvnic_rx_queue_stats),
488 GFP_KERNEL);
489 if (!adapter->rx_stats_buffers)
490 return -ENOMEM;
491
492 return 0;
493 }
494
495 static void release_stats_token(struct ibmvnic_adapter *adapter)
496 {
497 struct device *dev = &adapter->vdev->dev;
498
499 if (!adapter->stats_token)
500 return;
501
502 dma_unmap_single(dev, adapter->stats_token,
503 sizeof(struct ibmvnic_statistics),
504 DMA_FROM_DEVICE);
505 adapter->stats_token = 0;
506 }
507
508 static int init_stats_token(struct ibmvnic_adapter *adapter)
509 {
510 struct device *dev = &adapter->vdev->dev;
511 dma_addr_t stok;
512
513 stok = dma_map_single(dev, &adapter->stats,
514 sizeof(struct ibmvnic_statistics),
515 DMA_FROM_DEVICE);
516 if (dma_mapping_error(dev, stok)) {
517 dev_err(dev, "Couldn't map stats buffer\n");
518 return -1;
519 }
520
521 adapter->stats_token = stok;
522 netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
523 return 0;
524 }
525
526 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
527 {
528 struct ibmvnic_rx_pool *rx_pool;
529 u64 buff_size;
530 int rx_scrqs;
531 int i, j, rc;
532
533 if (!adapter->rx_pool)
534 return -1;
535
536 buff_size = adapter->cur_rx_buf_sz;
537 rx_scrqs = adapter->num_active_rx_pools;
538 for (i = 0; i < rx_scrqs; i++) {
539 rx_pool = &adapter->rx_pool[i];
540
541 netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
542
543 if (rx_pool->buff_size != buff_size) {
544 free_long_term_buff(adapter, &rx_pool->long_term_buff);
545 rx_pool->buff_size = ALIGN(buff_size, L1_CACHE_BYTES);
546 rc = alloc_long_term_buff(adapter,
547 &rx_pool->long_term_buff,
548 rx_pool->size *
549 rx_pool->buff_size);
550 } else {
551 rc = reset_long_term_buff(adapter,
552 &rx_pool->long_term_buff);
553 }
554
555 if (rc)
556 return rc;
557
558 for (j = 0; j < rx_pool->size; j++)
559 rx_pool->free_map[j] = j;
560
561 memset(rx_pool->rx_buff, 0,
562 rx_pool->size * sizeof(struct ibmvnic_rx_buff));
563
564 atomic_set(&rx_pool->available, 0);
565 rx_pool->next_alloc = 0;
566 rx_pool->next_free = 0;
567 rx_pool->active = 1;
568 }
569
570 return 0;
571 }
572
573 static void release_rx_pools(struct ibmvnic_adapter *adapter)
574 {
575 struct ibmvnic_rx_pool *rx_pool;
576 int i, j;
577
578 if (!adapter->rx_pool)
579 return;
580
581 for (i = 0; i < adapter->num_active_rx_pools; i++) {
582 rx_pool = &adapter->rx_pool[i];
583
584 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
585
586 kfree(rx_pool->free_map);
587 free_long_term_buff(adapter, &rx_pool->long_term_buff);
588
589 if (!rx_pool->rx_buff)
590 continue;
591
592 for (j = 0; j < rx_pool->size; j++) {
593 if (rx_pool->rx_buff[j].skb) {
594 dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
595 rx_pool->rx_buff[j].skb = NULL;
596 }
597 }
598
599 kfree(rx_pool->rx_buff);
600 }
601
602 kfree(adapter->rx_pool);
603 adapter->rx_pool = NULL;
604 adapter->num_active_rx_pools = 0;
605 }
606
607 static int init_rx_pools(struct net_device *netdev)
608 {
609 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
610 struct device *dev = &adapter->vdev->dev;
611 struct ibmvnic_rx_pool *rx_pool;
612 int rxadd_subcrqs;
613 u64 buff_size;
614 int i, j;
615
616 rxadd_subcrqs = adapter->num_active_rx_scrqs;
617 buff_size = adapter->cur_rx_buf_sz;
618
619 adapter->rx_pool = kcalloc(rxadd_subcrqs,
620 sizeof(struct ibmvnic_rx_pool),
621 GFP_KERNEL);
622 if (!adapter->rx_pool) {
623 dev_err(dev, "Failed to allocate rx pools\n");
624 return -1;
625 }
626
627 adapter->num_active_rx_pools = rxadd_subcrqs;
628
629 for (i = 0; i < rxadd_subcrqs; i++) {
630 rx_pool = &adapter->rx_pool[i];
631
632 netdev_dbg(adapter->netdev,
633 "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
634 i, adapter->req_rx_add_entries_per_subcrq,
635 buff_size);
636
637 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
638 rx_pool->index = i;
639 rx_pool->buff_size = ALIGN(buff_size, L1_CACHE_BYTES);
640 rx_pool->active = 1;
641
642 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
643 GFP_KERNEL);
644 if (!rx_pool->free_map) {
645 release_rx_pools(adapter);
646 return -1;
647 }
648
649 rx_pool->rx_buff = kcalloc(rx_pool->size,
650 sizeof(struct ibmvnic_rx_buff),
651 GFP_KERNEL);
652 if (!rx_pool->rx_buff) {
653 dev_err(dev, "Couldn't alloc rx buffers\n");
654 release_rx_pools(adapter);
655 return -1;
656 }
657
658 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
659 rx_pool->size * rx_pool->buff_size)) {
660 release_rx_pools(adapter);
661 return -1;
662 }
663
664 for (j = 0; j < rx_pool->size; ++j)
665 rx_pool->free_map[j] = j;
666
667 atomic_set(&rx_pool->available, 0);
668 rx_pool->next_alloc = 0;
669 rx_pool->next_free = 0;
670 }
671
672 return 0;
673 }
674
675 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
676 struct ibmvnic_tx_pool *tx_pool)
677 {
678 int rc, i;
679
680 rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
681 if (rc)
682 return rc;
683
684 memset(tx_pool->tx_buff, 0,
685 tx_pool->num_buffers *
686 sizeof(struct ibmvnic_tx_buff));
687
688 for (i = 0; i < tx_pool->num_buffers; i++)
689 tx_pool->free_map[i] = i;
690
691 tx_pool->consumer_index = 0;
692 tx_pool->producer_index = 0;
693
694 return 0;
695 }
696
697 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
698 {
699 int tx_scrqs;
700 int i, rc;
701
702 if (!adapter->tx_pool)
703 return -1;
704
705 tx_scrqs = adapter->num_active_tx_pools;
706 for (i = 0; i < tx_scrqs; i++) {
707 ibmvnic_tx_scrq_clean_buffer(adapter, adapter->tx_scrq[i]);
708 rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
709 if (rc)
710 return rc;
711 rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
712 if (rc)
713 return rc;
714 }
715
716 return 0;
717 }
718
719 static void release_vpd_data(struct ibmvnic_adapter *adapter)
720 {
721 if (!adapter->vpd)
722 return;
723
724 kfree(adapter->vpd->buff);
725 kfree(adapter->vpd);
726
727 adapter->vpd = NULL;
728 }
729
730 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
731 struct ibmvnic_tx_pool *tx_pool)
732 {
733 kfree(tx_pool->tx_buff);
734 kfree(tx_pool->free_map);
735 free_long_term_buff(adapter, &tx_pool->long_term_buff);
736 }
737
738 static void release_tx_pools(struct ibmvnic_adapter *adapter)
739 {
740 int i;
741
742 if (!adapter->tx_pool)
743 return;
744
745 for (i = 0; i < adapter->num_active_tx_pools; i++) {
746 release_one_tx_pool(adapter, &adapter->tx_pool[i]);
747 release_one_tx_pool(adapter, &adapter->tso_pool[i]);
748 }
749
750 kfree(adapter->tx_pool);
751 adapter->tx_pool = NULL;
752 kfree(adapter->tso_pool);
753 adapter->tso_pool = NULL;
754 adapter->num_active_tx_pools = 0;
755 }
756
757 static int init_one_tx_pool(struct net_device *netdev,
758 struct ibmvnic_tx_pool *tx_pool,
759 int num_entries, int buf_size)
760 {
761 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
762 int i;
763
764 tx_pool->tx_buff = kcalloc(num_entries,
765 sizeof(struct ibmvnic_tx_buff),
766 GFP_KERNEL);
767 if (!tx_pool->tx_buff)
768 return -1;
769
770 if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
771 num_entries * buf_size))
772 return -1;
773
774 tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
775 if (!tx_pool->free_map)
776 return -1;
777
778 for (i = 0; i < num_entries; i++)
779 tx_pool->free_map[i] = i;
780
781 tx_pool->consumer_index = 0;
782 tx_pool->producer_index = 0;
783 tx_pool->num_buffers = num_entries;
784 tx_pool->buf_size = buf_size;
785
786 return 0;
787 }
788
789 static int init_tx_pools(struct net_device *netdev)
790 {
791 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
792 int tx_subcrqs;
793 u64 buff_size;
794 int i, rc;
795
796 tx_subcrqs = adapter->num_active_tx_scrqs;
797 adapter->tx_pool = kcalloc(tx_subcrqs,
798 sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
799 if (!adapter->tx_pool)
800 return -1;
801
802 adapter->tso_pool = kcalloc(tx_subcrqs,
803 sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
804 if (!adapter->tso_pool) {
805 kfree(adapter->tx_pool);
806 adapter->tx_pool = NULL;
807 return -1;
808 }
809
810 adapter->num_active_tx_pools = tx_subcrqs;
811
812 for (i = 0; i < tx_subcrqs; i++) {
813 buff_size = adapter->req_mtu + VLAN_HLEN;
814 buff_size = ALIGN(buff_size, L1_CACHE_BYTES);
815 rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
816 adapter->req_tx_entries_per_subcrq,
817 buff_size);
818 if (rc) {
819 release_tx_pools(adapter);
820 return rc;
821 }
822
823 rc = init_one_tx_pool(netdev, &adapter->tso_pool[i],
824 IBMVNIC_TSO_BUFS,
825 IBMVNIC_TSO_BUF_SZ);
826 if (rc) {
827 release_tx_pools(adapter);
828 return rc;
829 }
830 }
831
832 return 0;
833 }
834
835 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
836 {
837 int i;
838
839 if (adapter->napi_enabled)
840 return;
841
842 for (i = 0; i < adapter->req_rx_queues; i++)
843 napi_enable(&adapter->napi[i]);
844
845 adapter->napi_enabled = true;
846 }
847
848 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
849 {
850 int i;
851
852 if (!adapter->napi_enabled)
853 return;
854
855 for (i = 0; i < adapter->req_rx_queues; i++) {
856 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
857 napi_disable(&adapter->napi[i]);
858 }
859
860 adapter->napi_enabled = false;
861 }
862
863 static int init_napi(struct ibmvnic_adapter *adapter)
864 {
865 int i;
866
867 adapter->napi = kcalloc(adapter->req_rx_queues,
868 sizeof(struct napi_struct), GFP_KERNEL);
869 if (!adapter->napi)
870 return -ENOMEM;
871
872 for (i = 0; i < adapter->req_rx_queues; i++) {
873 netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
874 netif_napi_add(adapter->netdev, &adapter->napi[i],
875 ibmvnic_poll, NAPI_POLL_WEIGHT);
876 }
877
878 adapter->num_active_rx_napi = adapter->req_rx_queues;
879 return 0;
880 }
881
882 static void release_napi(struct ibmvnic_adapter *adapter)
883 {
884 int i;
885
886 if (!adapter->napi)
887 return;
888
889 for (i = 0; i < adapter->num_active_rx_napi; i++) {
890 netdev_dbg(adapter->netdev, "Releasing napi[%d]\n", i);
891 netif_napi_del(&adapter->napi[i]);
892 }
893
894 kfree(adapter->napi);
895 adapter->napi = NULL;
896 adapter->num_active_rx_napi = 0;
897 adapter->napi_enabled = false;
898 }
899
900 static const char *adapter_state_to_string(enum vnic_state state)
901 {
902 switch (state) {
903 case VNIC_PROBING:
904 return "PROBING";
905 case VNIC_PROBED:
906 return "PROBED";
907 case VNIC_OPENING:
908 return "OPENING";
909 case VNIC_OPEN:
910 return "OPEN";
911 case VNIC_CLOSING:
912 return "CLOSING";
913 case VNIC_CLOSED:
914 return "CLOSED";
915 case VNIC_REMOVING:
916 return "REMOVING";
917 case VNIC_REMOVED:
918 return "REMOVED";
919 case VNIC_DOWN:
920 return "DOWN";
921 }
922 return "UNKNOWN";
923 }
924
925 static int ibmvnic_login(struct net_device *netdev)
926 {
927 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
928 unsigned long timeout = msecs_to_jiffies(20000);
929 int retry_count = 0;
930 int retries = 10;
931 bool retry;
932 int rc;
933
934 do {
935 retry = false;
936 if (retry_count > retries) {
937 netdev_warn(netdev, "Login attempts exceeded\n");
938 return -1;
939 }
940
941 adapter->init_done_rc = 0;
942 reinit_completion(&adapter->init_done);
943 rc = send_login(adapter);
944 if (rc)
945 return rc;
946
947 if (!wait_for_completion_timeout(&adapter->init_done,
948 timeout)) {
949 netdev_warn(netdev, "Login timed out, retrying...\n");
950 retry = true;
951 adapter->init_done_rc = 0;
952 retry_count++;
953 continue;
954 }
955
956 if (adapter->init_done_rc == ABORTED) {
957 netdev_warn(netdev, "Login aborted, retrying...\n");
958 retry = true;
959 adapter->init_done_rc = 0;
960 retry_count++;
961 /* FW or device may be busy, so
962 * wait a bit before retrying login
963 */
964 msleep(500);
965 } else if (adapter->init_done_rc == PARTIALSUCCESS) {
966 retry_count++;
967 release_sub_crqs(adapter, 1);
968
969 retry = true;
970 netdev_dbg(netdev,
971 "Received partial success, retrying...\n");
972 adapter->init_done_rc = 0;
973 reinit_completion(&adapter->init_done);
974 send_query_cap(adapter);
975 if (!wait_for_completion_timeout(&adapter->init_done,
976 timeout)) {
977 netdev_warn(netdev,
978 "Capabilities query timed out\n");
979 return -1;
980 }
981
982 rc = init_sub_crqs(adapter);
983 if (rc) {
984 netdev_warn(netdev,
985 "SCRQ initialization failed\n");
986 return -1;
987 }
988
989 rc = init_sub_crq_irqs(adapter);
990 if (rc) {
991 netdev_warn(netdev,
992 "SCRQ irq initialization failed\n");
993 return -1;
994 }
995 } else if (adapter->init_done_rc) {
996 netdev_warn(netdev, "Adapter login failed\n");
997 return -1;
998 }
999 } while (retry);
1000
1001 __ibmvnic_set_mac(netdev, adapter->mac_addr);
1002
1003 netdev_dbg(netdev, "[S:%s] Login succeeded\n", adapter_state_to_string(adapter->state));
1004 return 0;
1005 }
1006
1007 static void release_login_buffer(struct ibmvnic_adapter *adapter)
1008 {
1009 kfree(adapter->login_buf);
1010 adapter->login_buf = NULL;
1011 }
1012
1013 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
1014 {
1015 kfree(adapter->login_rsp_buf);
1016 adapter->login_rsp_buf = NULL;
1017 }
1018
1019 static void release_resources(struct ibmvnic_adapter *adapter)
1020 {
1021 release_vpd_data(adapter);
1022
1023 release_tx_pools(adapter);
1024 release_rx_pools(adapter);
1025
1026 release_napi(adapter);
1027 release_login_buffer(adapter);
1028 release_login_rsp_buffer(adapter);
1029 }
1030
1031 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
1032 {
1033 struct net_device *netdev = adapter->netdev;
1034 unsigned long timeout = msecs_to_jiffies(20000);
1035 union ibmvnic_crq crq;
1036 bool resend;
1037 int rc;
1038
1039 netdev_dbg(netdev, "setting link state %d\n", link_state);
1040
1041 memset(&crq, 0, sizeof(crq));
1042 crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
1043 crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
1044 crq.logical_link_state.link_state = link_state;
1045
1046 do {
1047 resend = false;
1048
1049 reinit_completion(&adapter->init_done);
1050 rc = ibmvnic_send_crq(adapter, &crq);
1051 if (rc) {
1052 netdev_err(netdev, "Failed to set link state\n");
1053 return rc;
1054 }
1055
1056 if (!wait_for_completion_timeout(&adapter->init_done,
1057 timeout)) {
1058 netdev_err(netdev, "timeout setting link state\n");
1059 return -1;
1060 }
1061
1062 if (adapter->init_done_rc == PARTIALSUCCESS) {
1063 /* Partuial success, delay and re-send */
1064 mdelay(1000);
1065 resend = true;
1066 } else if (adapter->init_done_rc) {
1067 netdev_warn(netdev, "Unable to set link state, rc=%d\n",
1068 adapter->init_done_rc);
1069 return adapter->init_done_rc;
1070 }
1071 } while (resend);
1072
1073 return 0;
1074 }
1075
1076 static int set_real_num_queues(struct net_device *netdev)
1077 {
1078 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1079 int rc;
1080
1081 netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
1082 adapter->req_tx_queues, adapter->req_rx_queues);
1083
1084 rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
1085 if (rc) {
1086 netdev_err(netdev, "failed to set the number of tx queues\n");
1087 return rc;
1088 }
1089
1090 rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
1091 if (rc)
1092 netdev_err(netdev, "failed to set the number of rx queues\n");
1093
1094 return rc;
1095 }
1096
1097 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
1098 {
1099 struct device *dev = &adapter->vdev->dev;
1100 union ibmvnic_crq crq;
1101 int len = 0;
1102 int rc;
1103
1104 if (adapter->vpd->buff)
1105 len = adapter->vpd->len;
1106
1107 mutex_lock(&adapter->fw_lock);
1108 adapter->fw_done_rc = 0;
1109 reinit_completion(&adapter->fw_done);
1110
1111 crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
1112 crq.get_vpd_size.cmd = GET_VPD_SIZE;
1113 rc = ibmvnic_send_crq(adapter, &crq);
1114 if (rc) {
1115 mutex_unlock(&adapter->fw_lock);
1116 return rc;
1117 }
1118
1119 rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1120 if (rc) {
1121 dev_err(dev, "Could not retrieve VPD size, rc = %d\n", rc);
1122 mutex_unlock(&adapter->fw_lock);
1123 return rc;
1124 }
1125 mutex_unlock(&adapter->fw_lock);
1126
1127 if (!adapter->vpd->len)
1128 return -ENODATA;
1129
1130 if (!adapter->vpd->buff)
1131 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
1132 else if (adapter->vpd->len != len)
1133 adapter->vpd->buff =
1134 krealloc(adapter->vpd->buff,
1135 adapter->vpd->len, GFP_KERNEL);
1136
1137 if (!adapter->vpd->buff) {
1138 dev_err(dev, "Could allocate VPD buffer\n");
1139 return -ENOMEM;
1140 }
1141
1142 adapter->vpd->dma_addr =
1143 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
1144 DMA_FROM_DEVICE);
1145 if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
1146 dev_err(dev, "Could not map VPD buffer\n");
1147 kfree(adapter->vpd->buff);
1148 adapter->vpd->buff = NULL;
1149 return -ENOMEM;
1150 }
1151
1152 mutex_lock(&adapter->fw_lock);
1153 adapter->fw_done_rc = 0;
1154 reinit_completion(&adapter->fw_done);
1155
1156 crq.get_vpd.first = IBMVNIC_CRQ_CMD;
1157 crq.get_vpd.cmd = GET_VPD;
1158 crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
1159 crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1160 rc = ibmvnic_send_crq(adapter, &crq);
1161 if (rc) {
1162 kfree(adapter->vpd->buff);
1163 adapter->vpd->buff = NULL;
1164 mutex_unlock(&adapter->fw_lock);
1165 return rc;
1166 }
1167
1168 rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1169 if (rc) {
1170 dev_err(dev, "Unable to retrieve VPD, rc = %d\n", rc);
1171 kfree(adapter->vpd->buff);
1172 adapter->vpd->buff = NULL;
1173 mutex_unlock(&adapter->fw_lock);
1174 return rc;
1175 }
1176
1177 mutex_unlock(&adapter->fw_lock);
1178 return 0;
1179 }
1180
1181 static int init_resources(struct ibmvnic_adapter *adapter)
1182 {
1183 struct net_device *netdev = adapter->netdev;
1184 int rc;
1185
1186 rc = set_real_num_queues(netdev);
1187 if (rc)
1188 return rc;
1189
1190 adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1191 if (!adapter->vpd)
1192 return -ENOMEM;
1193
1194 /* Vital Product Data (VPD) */
1195 rc = ibmvnic_get_vpd(adapter);
1196 if (rc) {
1197 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1198 return rc;
1199 }
1200
1201 adapter->map_id = 1;
1202
1203 rc = init_napi(adapter);
1204 if (rc)
1205 return rc;
1206
1207 send_query_map(adapter);
1208
1209 rc = init_rx_pools(netdev);
1210 if (rc)
1211 return rc;
1212
1213 rc = init_tx_pools(netdev);
1214 return rc;
1215 }
1216
1217 static int __ibmvnic_open(struct net_device *netdev)
1218 {
1219 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1220 enum vnic_state prev_state = adapter->state;
1221 int i, rc;
1222
1223 adapter->state = VNIC_OPENING;
1224 replenish_pools(adapter);
1225 ibmvnic_napi_enable(adapter);
1226
1227 /* We're ready to receive frames, enable the sub-crq interrupts and
1228 * set the logical link state to up
1229 */
1230 for (i = 0; i < adapter->req_rx_queues; i++) {
1231 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1232 if (prev_state == VNIC_CLOSED)
1233 enable_irq(adapter->rx_scrq[i]->irq);
1234 enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1235 }
1236
1237 for (i = 0; i < adapter->req_tx_queues; i++) {
1238 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1239 if (prev_state == VNIC_CLOSED)
1240 enable_irq(adapter->tx_scrq[i]->irq);
1241 enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1242 netdev_tx_reset_queue(netdev_get_tx_queue(netdev, i));
1243 }
1244
1245 rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1246 if (rc) {
1247 ibmvnic_napi_disable(adapter);
1248 release_resources(adapter);
1249 return rc;
1250 }
1251
1252 netif_tx_start_all_queues(netdev);
1253
1254 if (prev_state == VNIC_CLOSED) {
1255 for (i = 0; i < adapter->req_rx_queues; i++)
1256 napi_schedule(&adapter->napi[i]);
1257 }
1258
1259 adapter->state = VNIC_OPEN;
1260 return rc;
1261 }
1262
1263 static int ibmvnic_open(struct net_device *netdev)
1264 {
1265 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1266 int rc;
1267
1268 ASSERT_RTNL();
1269
1270 /* If device failover is pending or we are about to reset, just set
1271 * device state and return. Device operation will be handled by reset
1272 * routine.
1273 *
1274 * It should be safe to overwrite the adapter->state here. Since
1275 * we hold the rtnl, either the reset has not actually started or
1276 * the rtnl got dropped during the set_link_state() in do_reset().
1277 * In the former case, no one else is changing the state (again we
1278 * have the rtnl) and in the latter case, do_reset() will detect and
1279 * honor our setting below.
1280 */
1281 if (adapter->failover_pending || (test_bit(0, &adapter->resetting))) {
1282 netdev_dbg(netdev, "[S:%s FOP:%d] Resetting, deferring open\n",
1283 adapter_state_to_string(adapter->state),
1284 adapter->failover_pending);
1285 adapter->state = VNIC_OPEN;
1286 rc = 0;
1287 goto out;
1288 }
1289
1290 if (adapter->state != VNIC_CLOSED) {
1291 rc = ibmvnic_login(netdev);
1292 if (rc)
1293 goto out;
1294
1295 rc = init_resources(adapter);
1296 if (rc) {
1297 netdev_err(netdev, "failed to initialize resources\n");
1298 release_resources(adapter);
1299 goto out;
1300 }
1301 }
1302
1303 rc = __ibmvnic_open(netdev);
1304
1305 out:
1306 /* If open failed and there is a pending failover or in-progress reset,
1307 * set device state and return. Device operation will be handled by
1308 * reset routine. See also comments above regarding rtnl.
1309 */
1310 if (rc &&
1311 (adapter->failover_pending || (test_bit(0, &adapter->resetting)))) {
1312 adapter->state = VNIC_OPEN;
1313 rc = 0;
1314 }
1315 return rc;
1316 }
1317
1318 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1319 {
1320 struct ibmvnic_rx_pool *rx_pool;
1321 struct ibmvnic_rx_buff *rx_buff;
1322 u64 rx_entries;
1323 int rx_scrqs;
1324 int i, j;
1325
1326 if (!adapter->rx_pool)
1327 return;
1328
1329 rx_scrqs = adapter->num_active_rx_pools;
1330 rx_entries = adapter->req_rx_add_entries_per_subcrq;
1331
1332 /* Free any remaining skbs in the rx buffer pools */
1333 for (i = 0; i < rx_scrqs; i++) {
1334 rx_pool = &adapter->rx_pool[i];
1335 if (!rx_pool || !rx_pool->rx_buff)
1336 continue;
1337
1338 netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1339 for (j = 0; j < rx_entries; j++) {
1340 rx_buff = &rx_pool->rx_buff[j];
1341 if (rx_buff && rx_buff->skb) {
1342 dev_kfree_skb_any(rx_buff->skb);
1343 rx_buff->skb = NULL;
1344 }
1345 }
1346 }
1347 }
1348
1349 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1350 struct ibmvnic_tx_pool *tx_pool)
1351 {
1352 struct ibmvnic_tx_buff *tx_buff;
1353 u64 tx_entries;
1354 int i;
1355
1356 if (!tx_pool || !tx_pool->tx_buff)
1357 return;
1358
1359 tx_entries = tx_pool->num_buffers;
1360
1361 for (i = 0; i < tx_entries; i++) {
1362 tx_buff = &tx_pool->tx_buff[i];
1363 if (tx_buff && tx_buff->skb) {
1364 dev_kfree_skb_any(tx_buff->skb);
1365 tx_buff->skb = NULL;
1366 }
1367 }
1368 }
1369
1370 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1371 {
1372 int tx_scrqs;
1373 int i;
1374
1375 if (!adapter->tx_pool || !adapter->tso_pool)
1376 return;
1377
1378 tx_scrqs = adapter->num_active_tx_pools;
1379
1380 /* Free any remaining skbs in the tx buffer pools */
1381 for (i = 0; i < tx_scrqs; i++) {
1382 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1383 clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1384 clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1385 }
1386 }
1387
1388 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1389 {
1390 struct net_device *netdev = adapter->netdev;
1391 int i;
1392
1393 if (adapter->tx_scrq) {
1394 for (i = 0; i < adapter->req_tx_queues; i++)
1395 if (adapter->tx_scrq[i]->irq) {
1396 netdev_dbg(netdev,
1397 "Disabling tx_scrq[%d] irq\n", i);
1398 disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1399 disable_irq(adapter->tx_scrq[i]->irq);
1400 }
1401 }
1402
1403 if (adapter->rx_scrq) {
1404 for (i = 0; i < adapter->req_rx_queues; i++) {
1405 if (adapter->rx_scrq[i]->irq) {
1406 netdev_dbg(netdev,
1407 "Disabling rx_scrq[%d] irq\n", i);
1408 disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1409 disable_irq(adapter->rx_scrq[i]->irq);
1410 }
1411 }
1412 }
1413 }
1414
1415 static void ibmvnic_cleanup(struct net_device *netdev)
1416 {
1417 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1418
1419 /* ensure that transmissions are stopped if called by do_reset */
1420 if (test_bit(0, &adapter->resetting))
1421 netif_tx_disable(netdev);
1422 else
1423 netif_tx_stop_all_queues(netdev);
1424
1425 ibmvnic_napi_disable(adapter);
1426 ibmvnic_disable_irqs(adapter);
1427
1428 clean_rx_pools(adapter);
1429 clean_tx_pools(adapter);
1430 }
1431
1432 static int __ibmvnic_close(struct net_device *netdev)
1433 {
1434 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1435 int rc = 0;
1436
1437 adapter->state = VNIC_CLOSING;
1438 rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1439 adapter->state = VNIC_CLOSED;
1440 return rc;
1441 }
1442
1443 static int ibmvnic_close(struct net_device *netdev)
1444 {
1445 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1446 int rc;
1447
1448 netdev_dbg(netdev, "[S:%s FOP:%d FRR:%d] Closing\n",
1449 adapter_state_to_string(adapter->state),
1450 adapter->failover_pending,
1451 adapter->force_reset_recovery);
1452
1453 /* If device failover is pending, just set device state and return.
1454 * Device operation will be handled by reset routine.
1455 */
1456 if (adapter->failover_pending) {
1457 adapter->state = VNIC_CLOSED;
1458 return 0;
1459 }
1460
1461 rc = __ibmvnic_close(netdev);
1462 ibmvnic_cleanup(netdev);
1463
1464 return rc;
1465 }
1466
1467 /**
1468 * build_hdr_data - creates L2/L3/L4 header data buffer
1469 * @hdr_field: bitfield determining needed headers
1470 * @skb: socket buffer
1471 * @hdr_len: array of header lengths
1472 * @hdr_data: buffer to write the header to
1473 *
1474 * Reads hdr_field to determine which headers are needed by firmware.
1475 * Builds a buffer containing these headers. Saves individual header
1476 * lengths and total buffer length to be used to build descriptors.
1477 */
1478 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1479 int *hdr_len, u8 *hdr_data)
1480 {
1481 int len = 0;
1482 u8 *hdr;
1483
1484 if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1485 hdr_len[0] = sizeof(struct vlan_ethhdr);
1486 else
1487 hdr_len[0] = sizeof(struct ethhdr);
1488
1489 if (skb->protocol == htons(ETH_P_IP)) {
1490 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1491 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1492 hdr_len[2] = tcp_hdrlen(skb);
1493 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1494 hdr_len[2] = sizeof(struct udphdr);
1495 } else if (skb->protocol == htons(ETH_P_IPV6)) {
1496 hdr_len[1] = sizeof(struct ipv6hdr);
1497 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1498 hdr_len[2] = tcp_hdrlen(skb);
1499 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1500 hdr_len[2] = sizeof(struct udphdr);
1501 } else if (skb->protocol == htons(ETH_P_ARP)) {
1502 hdr_len[1] = arp_hdr_len(skb->dev);
1503 hdr_len[2] = 0;
1504 }
1505
1506 memset(hdr_data, 0, 120);
1507 if ((hdr_field >> 6) & 1) {
1508 hdr = skb_mac_header(skb);
1509 memcpy(hdr_data, hdr, hdr_len[0]);
1510 len += hdr_len[0];
1511 }
1512
1513 if ((hdr_field >> 5) & 1) {
1514 hdr = skb_network_header(skb);
1515 memcpy(hdr_data + len, hdr, hdr_len[1]);
1516 len += hdr_len[1];
1517 }
1518
1519 if ((hdr_field >> 4) & 1) {
1520 hdr = skb_transport_header(skb);
1521 memcpy(hdr_data + len, hdr, hdr_len[2]);
1522 len += hdr_len[2];
1523 }
1524 return len;
1525 }
1526
1527 /**
1528 * create_hdr_descs - create header and header extension descriptors
1529 * @hdr_field: bitfield determining needed headers
1530 * @hdr_data: buffer containing header data
1531 * @len: length of data buffer
1532 * @hdr_len: array of individual header lengths
1533 * @scrq_arr: descriptor array
1534 *
1535 * Creates header and, if needed, header extension descriptors and
1536 * places them in a descriptor array, scrq_arr
1537 */
1538
1539 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1540 union sub_crq *scrq_arr)
1541 {
1542 union sub_crq hdr_desc;
1543 int tmp_len = len;
1544 int num_descs = 0;
1545 u8 *data, *cur;
1546 int tmp;
1547
1548 while (tmp_len > 0) {
1549 cur = hdr_data + len - tmp_len;
1550
1551 memset(&hdr_desc, 0, sizeof(hdr_desc));
1552 if (cur != hdr_data) {
1553 data = hdr_desc.hdr_ext.data;
1554 tmp = tmp_len > 29 ? 29 : tmp_len;
1555 hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1556 hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1557 hdr_desc.hdr_ext.len = tmp;
1558 } else {
1559 data = hdr_desc.hdr.data;
1560 tmp = tmp_len > 24 ? 24 : tmp_len;
1561 hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1562 hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1563 hdr_desc.hdr.len = tmp;
1564 hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1565 hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1566 hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1567 hdr_desc.hdr.flag = hdr_field << 1;
1568 }
1569 memcpy(data, cur, tmp);
1570 tmp_len -= tmp;
1571 *scrq_arr = hdr_desc;
1572 scrq_arr++;
1573 num_descs++;
1574 }
1575
1576 return num_descs;
1577 }
1578
1579 /**
1580 * build_hdr_descs_arr - build a header descriptor array
1581 * @skb: tx socket buffer
1582 * @indir_arr: indirect array
1583 * @num_entries: number of descriptors to be sent
1584 * @hdr_field: bit field determining which headers will be sent
1585 *
1586 * This function will build a TX descriptor array with applicable
1587 * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1588 */
1589
1590 static void build_hdr_descs_arr(struct sk_buff *skb,
1591 union sub_crq *indir_arr,
1592 int *num_entries, u8 hdr_field)
1593 {
1594 int hdr_len[3] = {0, 0, 0};
1595 u8 hdr_data[140] = {0};
1596 int tot_len;
1597
1598 tot_len = build_hdr_data(hdr_field, skb, hdr_len,
1599 hdr_data);
1600 *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1601 indir_arr + 1);
1602 }
1603
1604 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1605 struct net_device *netdev)
1606 {
1607 /* For some backing devices, mishandling of small packets
1608 * can result in a loss of connection or TX stall. Device
1609 * architects recommend that no packet should be smaller
1610 * than the minimum MTU value provided to the driver, so
1611 * pad any packets to that length
1612 */
1613 if (skb->len < netdev->min_mtu)
1614 return skb_put_padto(skb, netdev->min_mtu);
1615
1616 return 0;
1617 }
1618
1619 static void ibmvnic_tx_scrq_clean_buffer(struct ibmvnic_adapter *adapter,
1620 struct ibmvnic_sub_crq_queue *tx_scrq)
1621 {
1622 struct ibmvnic_ind_xmit_queue *ind_bufp;
1623 struct ibmvnic_tx_buff *tx_buff;
1624 struct ibmvnic_tx_pool *tx_pool;
1625 union sub_crq tx_scrq_entry;
1626 int queue_num;
1627 int entries;
1628 int index;
1629 int i;
1630
1631 ind_bufp = &tx_scrq->ind_buf;
1632 entries = (u64)ind_bufp->index;
1633 queue_num = tx_scrq->pool_index;
1634
1635 for (i = entries - 1; i >= 0; --i) {
1636 tx_scrq_entry = ind_bufp->indir_arr[i];
1637 if (tx_scrq_entry.v1.type != IBMVNIC_TX_DESC)
1638 continue;
1639 index = be32_to_cpu(tx_scrq_entry.v1.correlator);
1640 if (index & IBMVNIC_TSO_POOL_MASK) {
1641 tx_pool = &adapter->tso_pool[queue_num];
1642 index &= ~IBMVNIC_TSO_POOL_MASK;
1643 } else {
1644 tx_pool = &adapter->tx_pool[queue_num];
1645 }
1646 tx_pool->free_map[tx_pool->consumer_index] = index;
1647 tx_pool->consumer_index = tx_pool->consumer_index == 0 ?
1648 tx_pool->num_buffers - 1 :
1649 tx_pool->consumer_index - 1;
1650 tx_buff = &tx_pool->tx_buff[index];
1651 adapter->netdev->stats.tx_packets--;
1652 adapter->netdev->stats.tx_bytes -= tx_buff->skb->len;
1653 adapter->tx_stats_buffers[queue_num].packets--;
1654 adapter->tx_stats_buffers[queue_num].bytes -=
1655 tx_buff->skb->len;
1656 dev_kfree_skb_any(tx_buff->skb);
1657 tx_buff->skb = NULL;
1658 adapter->netdev->stats.tx_dropped++;
1659 }
1660 ind_bufp->index = 0;
1661 if (atomic_sub_return(entries, &tx_scrq->used) <=
1662 (adapter->req_tx_entries_per_subcrq / 2) &&
1663 __netif_subqueue_stopped(adapter->netdev, queue_num) &&
1664 !test_bit(0, &adapter->resetting)) {
1665 netif_wake_subqueue(adapter->netdev, queue_num);
1666 netdev_dbg(adapter->netdev, "Started queue %d\n",
1667 queue_num);
1668 }
1669 }
1670
1671 static int ibmvnic_tx_scrq_flush(struct ibmvnic_adapter *adapter,
1672 struct ibmvnic_sub_crq_queue *tx_scrq)
1673 {
1674 struct ibmvnic_ind_xmit_queue *ind_bufp;
1675 u64 dma_addr;
1676 u64 entries;
1677 u64 handle;
1678 int rc;
1679
1680 ind_bufp = &tx_scrq->ind_buf;
1681 dma_addr = (u64)ind_bufp->indir_dma;
1682 entries = (u64)ind_bufp->index;
1683 handle = tx_scrq->handle;
1684
1685 if (!entries)
1686 return 0;
1687 rc = send_subcrq_indirect(adapter, handle, dma_addr, entries);
1688 if (rc)
1689 ibmvnic_tx_scrq_clean_buffer(adapter, tx_scrq);
1690 else
1691 ind_bufp->index = 0;
1692 return 0;
1693 }
1694
1695 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1696 {
1697 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1698 int queue_num = skb_get_queue_mapping(skb);
1699 u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1700 struct device *dev = &adapter->vdev->dev;
1701 struct ibmvnic_ind_xmit_queue *ind_bufp;
1702 struct ibmvnic_tx_buff *tx_buff = NULL;
1703 struct ibmvnic_sub_crq_queue *tx_scrq;
1704 struct ibmvnic_tx_pool *tx_pool;
1705 unsigned int tx_send_failed = 0;
1706 netdev_tx_t ret = NETDEV_TX_OK;
1707 unsigned int tx_map_failed = 0;
1708 union sub_crq indir_arr[16];
1709 unsigned int tx_dropped = 0;
1710 unsigned int tx_packets = 0;
1711 unsigned int tx_bytes = 0;
1712 dma_addr_t data_dma_addr;
1713 struct netdev_queue *txq;
1714 unsigned long lpar_rc;
1715 union sub_crq tx_crq;
1716 unsigned int offset;
1717 int num_entries = 1;
1718 unsigned char *dst;
1719 int index = 0;
1720 u8 proto = 0;
1721
1722 tx_scrq = adapter->tx_scrq[queue_num];
1723 txq = netdev_get_tx_queue(netdev, queue_num);
1724 ind_bufp = &tx_scrq->ind_buf;
1725
1726 if (test_bit(0, &adapter->resetting)) {
1727 dev_kfree_skb_any(skb);
1728
1729 tx_send_failed++;
1730 tx_dropped++;
1731 ret = NETDEV_TX_OK;
1732 goto out;
1733 }
1734
1735 if (ibmvnic_xmit_workarounds(skb, netdev)) {
1736 tx_dropped++;
1737 tx_send_failed++;
1738 ret = NETDEV_TX_OK;
1739 ibmvnic_tx_scrq_flush(adapter, tx_scrq);
1740 goto out;
1741 }
1742 if (skb_is_gso(skb))
1743 tx_pool = &adapter->tso_pool[queue_num];
1744 else
1745 tx_pool = &adapter->tx_pool[queue_num];
1746
1747 index = tx_pool->free_map[tx_pool->consumer_index];
1748
1749 if (index == IBMVNIC_INVALID_MAP) {
1750 dev_kfree_skb_any(skb);
1751 tx_send_failed++;
1752 tx_dropped++;
1753 ibmvnic_tx_scrq_flush(adapter, tx_scrq);
1754 ret = NETDEV_TX_OK;
1755 goto out;
1756 }
1757
1758 tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1759
1760 offset = index * tx_pool->buf_size;
1761 dst = tx_pool->long_term_buff.buff + offset;
1762 memset(dst, 0, tx_pool->buf_size);
1763 data_dma_addr = tx_pool->long_term_buff.addr + offset;
1764
1765 if (skb_shinfo(skb)->nr_frags) {
1766 int cur, i;
1767
1768 /* Copy the head */
1769 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1770 cur = skb_headlen(skb);
1771
1772 /* Copy the frags */
1773 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1774 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1775
1776 memcpy(dst + cur, skb_frag_address(frag),
1777 skb_frag_size(frag));
1778 cur += skb_frag_size(frag);
1779 }
1780 } else {
1781 skb_copy_from_linear_data(skb, dst, skb->len);
1782 }
1783
1784 /* post changes to long_term_buff *dst before VIOS accessing it */
1785 dma_wmb();
1786
1787 tx_pool->consumer_index =
1788 (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1789
1790 tx_buff = &tx_pool->tx_buff[index];
1791 tx_buff->skb = skb;
1792 tx_buff->index = index;
1793 tx_buff->pool_index = queue_num;
1794
1795 memset(&tx_crq, 0, sizeof(tx_crq));
1796 tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1797 tx_crq.v1.type = IBMVNIC_TX_DESC;
1798 tx_crq.v1.n_crq_elem = 1;
1799 tx_crq.v1.n_sge = 1;
1800 tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1801
1802 if (skb_is_gso(skb))
1803 tx_crq.v1.correlator =
1804 cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1805 else
1806 tx_crq.v1.correlator = cpu_to_be32(index);
1807 tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1808 tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1809 tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1810
1811 if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1812 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1813 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1814 }
1815
1816 if (skb->protocol == htons(ETH_P_IP)) {
1817 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1818 proto = ip_hdr(skb)->protocol;
1819 } else if (skb->protocol == htons(ETH_P_IPV6)) {
1820 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1821 proto = ipv6_hdr(skb)->nexthdr;
1822 }
1823
1824 if (proto == IPPROTO_TCP)
1825 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1826 else if (proto == IPPROTO_UDP)
1827 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1828
1829 if (skb->ip_summed == CHECKSUM_PARTIAL) {
1830 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1831 hdrs += 2;
1832 }
1833 if (skb_is_gso(skb)) {
1834 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1835 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1836 hdrs += 2;
1837 }
1838
1839 if ((*hdrs >> 7) & 1)
1840 build_hdr_descs_arr(skb, indir_arr, &num_entries, *hdrs);
1841
1842 tx_crq.v1.n_crq_elem = num_entries;
1843 tx_buff->num_entries = num_entries;
1844 /* flush buffer if current entry can not fit */
1845 if (num_entries + ind_bufp->index > IBMVNIC_MAX_IND_DESCS) {
1846 lpar_rc = ibmvnic_tx_scrq_flush(adapter, tx_scrq);
1847 if (lpar_rc != H_SUCCESS)
1848 goto tx_flush_err;
1849 }
1850
1851 indir_arr[0] = tx_crq;
1852 memcpy(&ind_bufp->indir_arr[ind_bufp->index], &indir_arr[0],
1853 num_entries * sizeof(struct ibmvnic_generic_scrq));
1854 ind_bufp->index += num_entries;
1855 if (__netdev_tx_sent_queue(txq, skb->len,
1856 netdev_xmit_more() &&
1857 ind_bufp->index < IBMVNIC_MAX_IND_DESCS)) {
1858 lpar_rc = ibmvnic_tx_scrq_flush(adapter, tx_scrq);
1859 if (lpar_rc != H_SUCCESS)
1860 goto tx_err;
1861 }
1862
1863 if (atomic_add_return(num_entries, &tx_scrq->used)
1864 >= adapter->req_tx_entries_per_subcrq) {
1865 netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1866 netif_stop_subqueue(netdev, queue_num);
1867 }
1868
1869 tx_packets++;
1870 tx_bytes += skb->len;
1871 txq->trans_start = jiffies;
1872 ret = NETDEV_TX_OK;
1873 goto out;
1874
1875 tx_flush_err:
1876 dev_kfree_skb_any(skb);
1877 tx_buff->skb = NULL;
1878 tx_pool->consumer_index = tx_pool->consumer_index == 0 ?
1879 tx_pool->num_buffers - 1 :
1880 tx_pool->consumer_index - 1;
1881 tx_dropped++;
1882 tx_err:
1883 if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1884 dev_err_ratelimited(dev, "tx: send failed\n");
1885
1886 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1887 /* Disable TX and report carrier off if queue is closed
1888 * or pending failover.
1889 * Firmware guarantees that a signal will be sent to the
1890 * driver, triggering a reset or some other action.
1891 */
1892 netif_tx_stop_all_queues(netdev);
1893 netif_carrier_off(netdev);
1894 }
1895 out:
1896 netdev->stats.tx_dropped += tx_dropped;
1897 netdev->stats.tx_bytes += tx_bytes;
1898 netdev->stats.tx_packets += tx_packets;
1899 adapter->tx_send_failed += tx_send_failed;
1900 adapter->tx_map_failed += tx_map_failed;
1901 adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1902 adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1903 adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1904
1905 return ret;
1906 }
1907
1908 static void ibmvnic_set_multi(struct net_device *netdev)
1909 {
1910 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1911 struct netdev_hw_addr *ha;
1912 union ibmvnic_crq crq;
1913
1914 memset(&crq, 0, sizeof(crq));
1915 crq.request_capability.first = IBMVNIC_CRQ_CMD;
1916 crq.request_capability.cmd = REQUEST_CAPABILITY;
1917
1918 if (netdev->flags & IFF_PROMISC) {
1919 if (!adapter->promisc_supported)
1920 return;
1921 } else {
1922 if (netdev->flags & IFF_ALLMULTI) {
1923 /* Accept all multicast */
1924 memset(&crq, 0, sizeof(crq));
1925 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1926 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1927 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1928 ibmvnic_send_crq(adapter, &crq);
1929 } else if (netdev_mc_empty(netdev)) {
1930 /* Reject all multicast */
1931 memset(&crq, 0, sizeof(crq));
1932 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1933 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1934 crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1935 ibmvnic_send_crq(adapter, &crq);
1936 } else {
1937 /* Accept one or more multicast(s) */
1938 netdev_for_each_mc_addr(ha, netdev) {
1939 memset(&crq, 0, sizeof(crq));
1940 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1941 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1942 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1943 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1944 ha->addr);
1945 ibmvnic_send_crq(adapter, &crq);
1946 }
1947 }
1948 }
1949 }
1950
1951 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1952 {
1953 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1954 union ibmvnic_crq crq;
1955 int rc;
1956
1957 if (!is_valid_ether_addr(dev_addr)) {
1958 rc = -EADDRNOTAVAIL;
1959 goto err;
1960 }
1961
1962 memset(&crq, 0, sizeof(crq));
1963 crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1964 crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1965 ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1966
1967 mutex_lock(&adapter->fw_lock);
1968 adapter->fw_done_rc = 0;
1969 reinit_completion(&adapter->fw_done);
1970
1971 rc = ibmvnic_send_crq(adapter, &crq);
1972 if (rc) {
1973 rc = -EIO;
1974 mutex_unlock(&adapter->fw_lock);
1975 goto err;
1976 }
1977
1978 rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1979 /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1980 if (rc || adapter->fw_done_rc) {
1981 rc = -EIO;
1982 mutex_unlock(&adapter->fw_lock);
1983 goto err;
1984 }
1985 mutex_unlock(&adapter->fw_lock);
1986 return 0;
1987 err:
1988 ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1989 return rc;
1990 }
1991
1992 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1993 {
1994 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1995 struct sockaddr *addr = p;
1996 int rc;
1997
1998 rc = 0;
1999 if (!is_valid_ether_addr(addr->sa_data))
2000 return -EADDRNOTAVAIL;
2001
2002 ether_addr_copy(adapter->mac_addr, addr->sa_data);
2003 if (adapter->state != VNIC_PROBED)
2004 rc = __ibmvnic_set_mac(netdev, addr->sa_data);
2005
2006 return rc;
2007 }
2008
2009 static const char *reset_reason_to_string(enum ibmvnic_reset_reason reason)
2010 {
2011 switch (reason) {
2012 case VNIC_RESET_FAILOVER:
2013 return "FAILOVER";
2014 case VNIC_RESET_MOBILITY:
2015 return "MOBILITY";
2016 case VNIC_RESET_FATAL:
2017 return "FATAL";
2018 case VNIC_RESET_NON_FATAL:
2019 return "NON_FATAL";
2020 case VNIC_RESET_TIMEOUT:
2021 return "TIMEOUT";
2022 case VNIC_RESET_CHANGE_PARAM:
2023 return "CHANGE_PARAM";
2024 case VNIC_RESET_PASSIVE_INIT:
2025 return "PASSIVE_INIT";
2026 }
2027 return "UNKNOWN";
2028 }
2029
2030 /*
2031 * do_reset returns zero if we are able to keep processing reset events, or
2032 * non-zero if we hit a fatal error and must halt.
2033 */
2034 static int do_reset(struct ibmvnic_adapter *adapter,
2035 struct ibmvnic_rwi *rwi, u32 reset_state)
2036 {
2037 u64 old_num_rx_queues, old_num_tx_queues;
2038 u64 old_num_rx_slots, old_num_tx_slots;
2039 struct net_device *netdev = adapter->netdev;
2040 int rc;
2041
2042 netdev_dbg(adapter->netdev,
2043 "[S:%s FOP:%d] Reset reason: %s, reset_state: %s\n",
2044 adapter_state_to_string(adapter->state),
2045 adapter->failover_pending,
2046 reset_reason_to_string(rwi->reset_reason),
2047 adapter_state_to_string(reset_state));
2048
2049 adapter->reset_reason = rwi->reset_reason;
2050 /* requestor of VNIC_RESET_CHANGE_PARAM already has the rtnl lock */
2051 if (!(adapter->reset_reason == VNIC_RESET_CHANGE_PARAM))
2052 rtnl_lock();
2053
2054 /* Now that we have the rtnl lock, clear any pending failover.
2055 * This will ensure ibmvnic_open() has either completed or will
2056 * block until failover is complete.
2057 */
2058 if (rwi->reset_reason == VNIC_RESET_FAILOVER)
2059 adapter->failover_pending = false;
2060
2061 /* read the state and check (again) after getting rtnl */
2062 reset_state = adapter->state;
2063
2064 if (reset_state == VNIC_REMOVING || reset_state == VNIC_REMOVED) {
2065 rc = -EBUSY;
2066 goto out;
2067 }
2068
2069 netif_carrier_off(netdev);
2070
2071 old_num_rx_queues = adapter->req_rx_queues;
2072 old_num_tx_queues = adapter->req_tx_queues;
2073 old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
2074 old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
2075
2076 ibmvnic_cleanup(netdev);
2077
2078 if (reset_state == VNIC_OPEN &&
2079 adapter->reset_reason != VNIC_RESET_MOBILITY &&
2080 adapter->reset_reason != VNIC_RESET_FAILOVER) {
2081 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM) {
2082 rc = __ibmvnic_close(netdev);
2083 if (rc)
2084 goto out;
2085 } else {
2086 adapter->state = VNIC_CLOSING;
2087
2088 /* Release the RTNL lock before link state change and
2089 * re-acquire after the link state change to allow
2090 * linkwatch_event to grab the RTNL lock and run during
2091 * a reset.
2092 */
2093 rtnl_unlock();
2094 rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
2095 rtnl_lock();
2096 if (rc)
2097 goto out;
2098
2099 if (adapter->state == VNIC_OPEN) {
2100 /* When we dropped rtnl, ibmvnic_open() got
2101 * it and noticed that we are resetting and
2102 * set the adapter state to OPEN. Update our
2103 * new "target" state, and resume the reset
2104 * from VNIC_CLOSING state.
2105 */
2106 netdev_dbg(netdev,
2107 "Open changed state from %s, updating.\n",
2108 adapter_state_to_string(reset_state));
2109 reset_state = VNIC_OPEN;
2110 adapter->state = VNIC_CLOSING;
2111 }
2112
2113 if (adapter->state != VNIC_CLOSING) {
2114 /* If someone else changed the adapter state
2115 * when we dropped the rtnl, fail the reset
2116 */
2117 rc = -1;
2118 goto out;
2119 }
2120 adapter->state = VNIC_CLOSED;
2121 }
2122 }
2123
2124 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM) {
2125 release_resources(adapter);
2126 release_sub_crqs(adapter, 1);
2127 release_crq_queue(adapter);
2128 }
2129
2130 if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
2131 /* remove the closed state so when we call open it appears
2132 * we are coming from the probed state.
2133 */
2134 adapter->state = VNIC_PROBED;
2135
2136 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM) {
2137 rc = init_crq_queue(adapter);
2138 } else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
2139 rc = ibmvnic_reenable_crq_queue(adapter);
2140 release_sub_crqs(adapter, 1);
2141 } else {
2142 rc = ibmvnic_reset_crq(adapter);
2143 if (rc == H_CLOSED || rc == H_SUCCESS) {
2144 rc = vio_enable_interrupts(adapter->vdev);
2145 if (rc)
2146 netdev_err(adapter->netdev,
2147 "Reset failed to enable interrupts. rc=%d\n",
2148 rc);
2149 }
2150 }
2151
2152 if (rc) {
2153 netdev_err(adapter->netdev,
2154 "Reset couldn't initialize crq. rc=%d\n", rc);
2155 goto out;
2156 }
2157
2158 rc = ibmvnic_reset_init(adapter, true);
2159 if (rc) {
2160 rc = IBMVNIC_INIT_FAILED;
2161 goto out;
2162 }
2163
2164 /* If the adapter was in PROBE or DOWN state prior to the reset,
2165 * exit here.
2166 */
2167 if (reset_state == VNIC_PROBED || reset_state == VNIC_DOWN) {
2168 rc = 0;
2169 goto out;
2170 }
2171
2172 rc = ibmvnic_login(netdev);
2173 if (rc)
2174 goto out;
2175
2176 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM) {
2177 rc = init_resources(adapter);
2178 if (rc)
2179 goto out;
2180 } else if (adapter->req_rx_queues != old_num_rx_queues ||
2181 adapter->req_tx_queues != old_num_tx_queues ||
2182 adapter->req_rx_add_entries_per_subcrq !=
2183 old_num_rx_slots ||
2184 adapter->req_tx_entries_per_subcrq !=
2185 old_num_tx_slots ||
2186 !adapter->rx_pool ||
2187 !adapter->tso_pool ||
2188 !adapter->tx_pool) {
2189 release_rx_pools(adapter);
2190 release_tx_pools(adapter);
2191 release_napi(adapter);
2192 release_vpd_data(adapter);
2193
2194 rc = init_resources(adapter);
2195 if (rc)
2196 goto out;
2197
2198 } else {
2199 rc = reset_tx_pools(adapter);
2200 if (rc) {
2201 netdev_dbg(adapter->netdev, "reset tx pools failed (%d)\n",
2202 rc);
2203 goto out;
2204 }
2205
2206 rc = reset_rx_pools(adapter);
2207 if (rc) {
2208 netdev_dbg(adapter->netdev, "reset rx pools failed (%d)\n",
2209 rc);
2210 goto out;
2211 }
2212 }
2213 ibmvnic_disable_irqs(adapter);
2214 }
2215 adapter->state = VNIC_CLOSED;
2216
2217 if (reset_state == VNIC_CLOSED) {
2218 rc = 0;
2219 goto out;
2220 }
2221
2222 rc = __ibmvnic_open(netdev);
2223 if (rc) {
2224 rc = IBMVNIC_OPEN_FAILED;
2225 goto out;
2226 }
2227
2228 /* refresh device's multicast list */
2229 ibmvnic_set_multi(netdev);
2230
2231 if (adapter->reset_reason == VNIC_RESET_FAILOVER ||
2232 adapter->reset_reason == VNIC_RESET_MOBILITY)
2233 __netdev_notify_peers(netdev);
2234
2235 rc = 0;
2236
2237 out:
2238 /* restore the adapter state if reset failed */
2239 if (rc)
2240 adapter->state = reset_state;
2241 /* requestor of VNIC_RESET_CHANGE_PARAM should still hold the rtnl lock */
2242 if (!(adapter->reset_reason == VNIC_RESET_CHANGE_PARAM))
2243 rtnl_unlock();
2244
2245 netdev_dbg(adapter->netdev, "[S:%s FOP:%d] Reset done, rc %d\n",
2246 adapter_state_to_string(adapter->state),
2247 adapter->failover_pending, rc);
2248 return rc;
2249 }
2250
2251 static int do_hard_reset(struct ibmvnic_adapter *adapter,
2252 struct ibmvnic_rwi *rwi, u32 reset_state)
2253 {
2254 struct net_device *netdev = adapter->netdev;
2255 int rc;
2256
2257 netdev_dbg(adapter->netdev, "Hard resetting driver (%s)\n",
2258 reset_reason_to_string(rwi->reset_reason));
2259
2260 /* read the state and check (again) after getting rtnl */
2261 reset_state = adapter->state;
2262
2263 if (reset_state == VNIC_REMOVING || reset_state == VNIC_REMOVED) {
2264 rc = -EBUSY;
2265 goto out;
2266 }
2267
2268 netif_carrier_off(netdev);
2269 adapter->reset_reason = rwi->reset_reason;
2270
2271 ibmvnic_cleanup(netdev);
2272 release_resources(adapter);
2273 release_sub_crqs(adapter, 0);
2274 release_crq_queue(adapter);
2275
2276 /* remove the closed state so when we call open it appears
2277 * we are coming from the probed state.
2278 */
2279 adapter->state = VNIC_PROBED;
2280
2281 reinit_completion(&adapter->init_done);
2282 rc = init_crq_queue(adapter);
2283 if (rc) {
2284 netdev_err(adapter->netdev,
2285 "Couldn't initialize crq. rc=%d\n", rc);
2286 goto out;
2287 }
2288
2289 rc = ibmvnic_reset_init(adapter, false);
2290 if (rc)
2291 goto out;
2292
2293 /* If the adapter was in PROBE or DOWN state prior to the reset,
2294 * exit here.
2295 */
2296 if (reset_state == VNIC_PROBED || reset_state == VNIC_DOWN)
2297 goto out;
2298
2299 rc = ibmvnic_login(netdev);
2300 if (rc)
2301 goto out;
2302
2303 rc = init_resources(adapter);
2304 if (rc)
2305 goto out;
2306
2307 ibmvnic_disable_irqs(adapter);
2308 adapter->state = VNIC_CLOSED;
2309
2310 if (reset_state == VNIC_CLOSED)
2311 goto out;
2312
2313 rc = __ibmvnic_open(netdev);
2314 if (rc) {
2315 rc = IBMVNIC_OPEN_FAILED;
2316 goto out;
2317 }
2318
2319 __netdev_notify_peers(netdev);
2320 out:
2321 /* restore adapter state if reset failed */
2322 if (rc)
2323 adapter->state = reset_state;
2324 netdev_dbg(adapter->netdev, "[S:%s FOP:%d] Hard reset done, rc %d\n",
2325 adapter_state_to_string(adapter->state),
2326 adapter->failover_pending, rc);
2327 return rc;
2328 }
2329
2330 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
2331 {
2332 struct ibmvnic_rwi *rwi;
2333 unsigned long flags;
2334
2335 spin_lock_irqsave(&adapter->rwi_lock, flags);
2336
2337 if (!list_empty(&adapter->rwi_list)) {
2338 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
2339 list);
2340 list_del(&rwi->list);
2341 } else {
2342 rwi = NULL;
2343 }
2344
2345 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2346 return rwi;
2347 }
2348
2349 /**
2350 * do_passive_init - complete probing when partner device is detected.
2351 * @adapter: ibmvnic_adapter struct
2352 *
2353 * If the ibmvnic device does not have a partner device to communicate with at boot
2354 * and that partner device comes online at a later time, this function is called
2355 * to complete the initialization process of ibmvnic device.
2356 * Caller is expected to hold rtnl_lock().
2357 *
2358 * Returns non-zero if sub-CRQs are not initialized properly leaving the device
2359 * in the down state.
2360 * Returns 0 upon success and the device is in PROBED state.
2361 */
2362
2363 static int do_passive_init(struct ibmvnic_adapter *adapter)
2364 {
2365 unsigned long timeout = msecs_to_jiffies(30000);
2366 struct net_device *netdev = adapter->netdev;
2367 struct device *dev = &adapter->vdev->dev;
2368 int rc;
2369
2370 netdev_dbg(netdev, "Partner device found, probing.\n");
2371
2372 adapter->state = VNIC_PROBING;
2373 reinit_completion(&adapter->init_done);
2374 adapter->init_done_rc = 0;
2375 adapter->crq.active = true;
2376
2377 rc = send_crq_init_complete(adapter);
2378 if (rc)
2379 goto out;
2380
2381 rc = send_version_xchg(adapter);
2382 if (rc)
2383 netdev_dbg(adapter->netdev, "send_version_xchg failed, rc=%d\n", rc);
2384
2385 if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
2386 dev_err(dev, "Initialization sequence timed out\n");
2387 rc = -ETIMEDOUT;
2388 goto out;
2389 }
2390
2391 rc = init_sub_crqs(adapter);
2392 if (rc) {
2393 dev_err(dev, "Initialization of sub crqs failed, rc=%d\n", rc);
2394 goto out;
2395 }
2396
2397 rc = init_sub_crq_irqs(adapter);
2398 if (rc) {
2399 dev_err(dev, "Failed to initialize sub crq irqs\n, rc=%d", rc);
2400 goto init_failed;
2401 }
2402
2403 netdev->mtu = adapter->req_mtu - ETH_HLEN;
2404 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
2405 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
2406
2407 adapter->state = VNIC_PROBED;
2408 netdev_dbg(netdev, "Probed successfully. Waiting for signal from partner device.\n");
2409
2410 return 0;
2411
2412 init_failed:
2413 release_sub_crqs(adapter, 1);
2414 out:
2415 adapter->state = VNIC_DOWN;
2416 return rc;
2417 }
2418
2419 static void __ibmvnic_reset(struct work_struct *work)
2420 {
2421 struct ibmvnic_adapter *adapter;
2422 bool saved_state = false;
2423 struct ibmvnic_rwi *tmprwi;
2424 struct ibmvnic_rwi *rwi;
2425 unsigned long flags;
2426 u32 reset_state;
2427 int rc = 0;
2428
2429 adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
2430
2431 if (test_and_set_bit_lock(0, &adapter->resetting)) {
2432 queue_delayed_work(system_long_wq,
2433 &adapter->ibmvnic_delayed_reset,
2434 IBMVNIC_RESET_DELAY);
2435 return;
2436 }
2437
2438 rwi = get_next_rwi(adapter);
2439 while (rwi) {
2440 spin_lock_irqsave(&adapter->state_lock, flags);
2441
2442 if (adapter->state == VNIC_REMOVING ||
2443 adapter->state == VNIC_REMOVED) {
2444 spin_unlock_irqrestore(&adapter->state_lock, flags);
2445 kfree(rwi);
2446 rc = EBUSY;
2447 break;
2448 }
2449
2450 if (!saved_state) {
2451 reset_state = adapter->state;
2452 saved_state = true;
2453 }
2454 spin_unlock_irqrestore(&adapter->state_lock, flags);
2455
2456 if (rwi->reset_reason == VNIC_RESET_PASSIVE_INIT) {
2457 rtnl_lock();
2458 rc = do_passive_init(adapter);
2459 rtnl_unlock();
2460 if (!rc)
2461 netif_carrier_on(adapter->netdev);
2462 } else if (adapter->force_reset_recovery) {
2463 /* Since we are doing a hard reset now, clear the
2464 * failover_pending flag so we don't ignore any
2465 * future MOBILITY or other resets.
2466 */
2467 adapter->failover_pending = false;
2468
2469 /* Transport event occurred during previous reset */
2470 if (adapter->wait_for_reset) {
2471 /* Previous was CHANGE_PARAM; caller locked */
2472 adapter->force_reset_recovery = false;
2473 rc = do_hard_reset(adapter, rwi, reset_state);
2474 } else {
2475 rtnl_lock();
2476 adapter->force_reset_recovery = false;
2477 rc = do_hard_reset(adapter, rwi, reset_state);
2478 rtnl_unlock();
2479 }
2480 if (rc) {
2481 /* give backing device time to settle down */
2482 netdev_dbg(adapter->netdev,
2483 "[S:%s] Hard reset failed, waiting 60 secs\n",
2484 adapter_state_to_string(adapter->state));
2485 set_current_state(TASK_UNINTERRUPTIBLE);
2486 schedule_timeout(60 * HZ);
2487 }
2488 } else {
2489 rc = do_reset(adapter, rwi, reset_state);
2490 }
2491 tmprwi = rwi;
2492 adapter->last_reset_time = jiffies;
2493
2494 if (rc)
2495 netdev_dbg(adapter->netdev, "Reset failed, rc=%d\n", rc);
2496
2497 rwi = get_next_rwi(adapter);
2498
2499 /*
2500 * If there is another reset queued, free the previous rwi
2501 * and process the new reset even if previous reset failed
2502 * (the previous reset could have failed because of a fail
2503 * over for instance, so process the fail over).
2504 *
2505 * If there are no resets queued and the previous reset failed,
2506 * the adapter would be in an undefined state. So retry the
2507 * previous reset as a hard reset.
2508 */
2509 if (rwi)
2510 kfree(tmprwi);
2511 else if (rc)
2512 rwi = tmprwi;
2513
2514 if (rwi && (rwi->reset_reason == VNIC_RESET_FAILOVER ||
2515 rwi->reset_reason == VNIC_RESET_MOBILITY || rc))
2516 adapter->force_reset_recovery = true;
2517 }
2518
2519 if (adapter->wait_for_reset) {
2520 adapter->reset_done_rc = rc;
2521 complete(&adapter->reset_done);
2522 }
2523
2524 clear_bit_unlock(0, &adapter->resetting);
2525
2526 netdev_dbg(adapter->netdev,
2527 "[S:%s FRR:%d WFR:%d] Done processing resets\n",
2528 adapter_state_to_string(adapter->state),
2529 adapter->force_reset_recovery,
2530 adapter->wait_for_reset);
2531 }
2532
2533 static void __ibmvnic_delayed_reset(struct work_struct *work)
2534 {
2535 struct ibmvnic_adapter *adapter;
2536
2537 adapter = container_of(work, struct ibmvnic_adapter,
2538 ibmvnic_delayed_reset.work);
2539 __ibmvnic_reset(&adapter->ibmvnic_reset);
2540 }
2541
2542 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2543 enum ibmvnic_reset_reason reason)
2544 {
2545 struct list_head *entry, *tmp_entry;
2546 struct ibmvnic_rwi *rwi, *tmp;
2547 struct net_device *netdev = adapter->netdev;
2548 unsigned long flags;
2549 int ret;
2550
2551 spin_lock_irqsave(&adapter->rwi_lock, flags);
2552
2553 /* If failover is pending don't schedule any other reset.
2554 * Instead let the failover complete. If there is already a
2555 * a failover reset scheduled, we will detect and drop the
2556 * duplicate reset when walking the ->rwi_list below.
2557 */
2558 if (adapter->state == VNIC_REMOVING ||
2559 adapter->state == VNIC_REMOVED ||
2560 (adapter->failover_pending && reason != VNIC_RESET_FAILOVER)) {
2561 ret = EBUSY;
2562 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2563 goto err;
2564 }
2565
2566 if (adapter->state == VNIC_PROBING) {
2567 netdev_warn(netdev, "Adapter reset during probe\n");
2568 adapter->init_done_rc = -EAGAIN;
2569 ret = EAGAIN;
2570 goto err;
2571 }
2572
2573 list_for_each_entry(tmp, &adapter->rwi_list, list) {
2574 if (tmp->reset_reason == reason) {
2575 netdev_dbg(netdev, "Skipping matching reset, reason=%s\n",
2576 reset_reason_to_string(reason));
2577 ret = EBUSY;
2578 goto err;
2579 }
2580 }
2581
2582 rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2583 if (!rwi) {
2584 ret = ENOMEM;
2585 goto err;
2586 }
2587 /* if we just received a transport event,
2588 * flush reset queue and process this reset
2589 */
2590 if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2591 list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2592 list_del(entry);
2593 }
2594 rwi->reset_reason = reason;
2595 list_add_tail(&rwi->list, &adapter->rwi_list);
2596 netdev_dbg(adapter->netdev, "Scheduling reset (reason %s)\n",
2597 reset_reason_to_string(reason));
2598 queue_work(system_long_wq, &adapter->ibmvnic_reset);
2599
2600 ret = 0;
2601 err:
2602 /* ibmvnic_close() below can block, so drop the lock first */
2603 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2604
2605 if (ret == ENOMEM)
2606 ibmvnic_close(netdev);
2607
2608 return -ret;
2609 }
2610
2611 static void ibmvnic_tx_timeout(struct net_device *dev, unsigned int txqueue)
2612 {
2613 struct ibmvnic_adapter *adapter = netdev_priv(dev);
2614
2615 if (test_bit(0, &adapter->resetting)) {
2616 netdev_err(adapter->netdev,
2617 "Adapter is resetting, skip timeout reset\n");
2618 return;
2619 }
2620 /* No queuing up reset until at least 5 seconds (default watchdog val)
2621 * after last reset
2622 */
2623 if (time_before(jiffies, (adapter->last_reset_time + dev->watchdog_timeo))) {
2624 netdev_dbg(dev, "Not yet time to tx timeout.\n");
2625 return;
2626 }
2627 ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2628 }
2629
2630 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2631 struct ibmvnic_rx_buff *rx_buff)
2632 {
2633 struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2634
2635 rx_buff->skb = NULL;
2636
2637 pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2638 pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2639
2640 atomic_dec(&pool->available);
2641 }
2642
2643 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2644 {
2645 struct ibmvnic_sub_crq_queue *rx_scrq;
2646 struct ibmvnic_adapter *adapter;
2647 struct net_device *netdev;
2648 int frames_processed;
2649 int scrq_num;
2650
2651 netdev = napi->dev;
2652 adapter = netdev_priv(netdev);
2653 scrq_num = (int)(napi - adapter->napi);
2654 frames_processed = 0;
2655 rx_scrq = adapter->rx_scrq[scrq_num];
2656
2657 restart_poll:
2658 while (frames_processed < budget) {
2659 struct sk_buff *skb;
2660 struct ibmvnic_rx_buff *rx_buff;
2661 union sub_crq *next;
2662 u32 length;
2663 u16 offset;
2664 u8 flags = 0;
2665
2666 if (unlikely(test_bit(0, &adapter->resetting) &&
2667 adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2668 enable_scrq_irq(adapter, rx_scrq);
2669 napi_complete_done(napi, frames_processed);
2670 return frames_processed;
2671 }
2672
2673 if (!pending_scrq(adapter, rx_scrq))
2674 break;
2675 next = ibmvnic_next_scrq(adapter, rx_scrq);
2676 rx_buff = (struct ibmvnic_rx_buff *)
2677 be64_to_cpu(next->rx_comp.correlator);
2678 /* do error checking */
2679 if (next->rx_comp.rc) {
2680 netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2681 be16_to_cpu(next->rx_comp.rc));
2682 /* free the entry */
2683 next->rx_comp.first = 0;
2684 dev_kfree_skb_any(rx_buff->skb);
2685 remove_buff_from_pool(adapter, rx_buff);
2686 continue;
2687 } else if (!rx_buff->skb) {
2688 /* free the entry */
2689 next->rx_comp.first = 0;
2690 remove_buff_from_pool(adapter, rx_buff);
2691 continue;
2692 }
2693
2694 length = be32_to_cpu(next->rx_comp.len);
2695 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2696 flags = next->rx_comp.flags;
2697 skb = rx_buff->skb;
2698 /* load long_term_buff before copying to skb */
2699 dma_rmb();
2700 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2701 length);
2702
2703 /* VLAN Header has been stripped by the system firmware and
2704 * needs to be inserted by the driver
2705 */
2706 if (adapter->rx_vlan_header_insertion &&
2707 (flags & IBMVNIC_VLAN_STRIPPED))
2708 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2709 ntohs(next->rx_comp.vlan_tci));
2710
2711 /* free the entry */
2712 next->rx_comp.first = 0;
2713 remove_buff_from_pool(adapter, rx_buff);
2714
2715 skb_put(skb, length);
2716 skb->protocol = eth_type_trans(skb, netdev);
2717 skb_record_rx_queue(skb, scrq_num);
2718
2719 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2720 flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2721 skb->ip_summed = CHECKSUM_UNNECESSARY;
2722 }
2723
2724 length = skb->len;
2725 napi_gro_receive(napi, skb); /* send it up */
2726 netdev->stats.rx_packets++;
2727 netdev->stats.rx_bytes += length;
2728 adapter->rx_stats_buffers[scrq_num].packets++;
2729 adapter->rx_stats_buffers[scrq_num].bytes += length;
2730 frames_processed++;
2731 }
2732
2733 if (adapter->state != VNIC_CLOSING &&
2734 ((atomic_read(&adapter->rx_pool[scrq_num].available) <
2735 adapter->req_rx_add_entries_per_subcrq / 2) ||
2736 frames_processed < budget))
2737 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2738 if (frames_processed < budget) {
2739 if (napi_complete_done(napi, frames_processed)) {
2740 enable_scrq_irq(adapter, rx_scrq);
2741 if (pending_scrq(adapter, rx_scrq)) {
2742 if (napi_reschedule(napi)) {
2743 disable_scrq_irq(adapter, rx_scrq);
2744 goto restart_poll;
2745 }
2746 }
2747 }
2748 }
2749 return frames_processed;
2750 }
2751
2752 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2753 {
2754 int rc, ret;
2755
2756 adapter->fallback.mtu = adapter->req_mtu;
2757 adapter->fallback.rx_queues = adapter->req_rx_queues;
2758 adapter->fallback.tx_queues = adapter->req_tx_queues;
2759 adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2760 adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2761
2762 reinit_completion(&adapter->reset_done);
2763 adapter->wait_for_reset = true;
2764 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2765
2766 if (rc) {
2767 ret = rc;
2768 goto out;
2769 }
2770 rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done, 60000);
2771 if (rc) {
2772 ret = -ENODEV;
2773 goto out;
2774 }
2775
2776 ret = 0;
2777 if (adapter->reset_done_rc) {
2778 ret = -EIO;
2779 adapter->desired.mtu = adapter->fallback.mtu;
2780 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2781 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2782 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2783 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2784
2785 reinit_completion(&adapter->reset_done);
2786 adapter->wait_for_reset = true;
2787 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2788 if (rc) {
2789 ret = rc;
2790 goto out;
2791 }
2792 rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done,
2793 60000);
2794 if (rc) {
2795 ret = -ENODEV;
2796 goto out;
2797 }
2798 }
2799 out:
2800 adapter->wait_for_reset = false;
2801
2802 return ret;
2803 }
2804
2805 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2806 {
2807 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2808
2809 adapter->desired.mtu = new_mtu + ETH_HLEN;
2810
2811 return wait_for_reset(adapter);
2812 }
2813
2814 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2815 struct net_device *dev,
2816 netdev_features_t features)
2817 {
2818 /* Some backing hardware adapters can not
2819 * handle packets with a MSS less than 224
2820 * or with only one segment.
2821 */
2822 if (skb_is_gso(skb)) {
2823 if (skb_shinfo(skb)->gso_size < 224 ||
2824 skb_shinfo(skb)->gso_segs == 1)
2825 features &= ~NETIF_F_GSO_MASK;
2826 }
2827
2828 return features;
2829 }
2830
2831 static const struct net_device_ops ibmvnic_netdev_ops = {
2832 .ndo_open = ibmvnic_open,
2833 .ndo_stop = ibmvnic_close,
2834 .ndo_start_xmit = ibmvnic_xmit,
2835 .ndo_set_rx_mode = ibmvnic_set_multi,
2836 .ndo_set_mac_address = ibmvnic_set_mac,
2837 .ndo_validate_addr = eth_validate_addr,
2838 .ndo_tx_timeout = ibmvnic_tx_timeout,
2839 .ndo_change_mtu = ibmvnic_change_mtu,
2840 .ndo_features_check = ibmvnic_features_check,
2841 };
2842
2843 /* ethtool functions */
2844
2845 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2846 struct ethtool_link_ksettings *cmd)
2847 {
2848 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2849 int rc;
2850
2851 rc = send_query_phys_parms(adapter);
2852 if (rc) {
2853 adapter->speed = SPEED_UNKNOWN;
2854 adapter->duplex = DUPLEX_UNKNOWN;
2855 }
2856 cmd->base.speed = adapter->speed;
2857 cmd->base.duplex = adapter->duplex;
2858 cmd->base.port = PORT_FIBRE;
2859 cmd->base.phy_address = 0;
2860 cmd->base.autoneg = AUTONEG_ENABLE;
2861
2862 return 0;
2863 }
2864
2865 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2866 struct ethtool_drvinfo *info)
2867 {
2868 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2869
2870 strscpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2871 strscpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2872 strscpy(info->fw_version, adapter->fw_version,
2873 sizeof(info->fw_version));
2874 }
2875
2876 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2877 {
2878 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2879
2880 return adapter->msg_enable;
2881 }
2882
2883 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2884 {
2885 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2886
2887 adapter->msg_enable = data;
2888 }
2889
2890 static u32 ibmvnic_get_link(struct net_device *netdev)
2891 {
2892 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2893
2894 /* Don't need to send a query because we request a logical link up at
2895 * init and then we wait for link state indications
2896 */
2897 return adapter->logical_link_state;
2898 }
2899
2900 static void ibmvnic_get_ringparam(struct net_device *netdev,
2901 struct ethtool_ringparam *ring)
2902 {
2903 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2904
2905 if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2906 ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2907 ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2908 } else {
2909 ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2910 ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2911 }
2912 ring->rx_mini_max_pending = 0;
2913 ring->rx_jumbo_max_pending = 0;
2914 ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2915 ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2916 ring->rx_mini_pending = 0;
2917 ring->rx_jumbo_pending = 0;
2918 }
2919
2920 static int ibmvnic_set_ringparam(struct net_device *netdev,
2921 struct ethtool_ringparam *ring)
2922 {
2923 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2924 int ret;
2925
2926 ret = 0;
2927 adapter->desired.rx_entries = ring->rx_pending;
2928 adapter->desired.tx_entries = ring->tx_pending;
2929
2930 ret = wait_for_reset(adapter);
2931
2932 if (!ret &&
2933 (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2934 adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2935 netdev_info(netdev,
2936 "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2937 ring->rx_pending, ring->tx_pending,
2938 adapter->req_rx_add_entries_per_subcrq,
2939 adapter->req_tx_entries_per_subcrq);
2940 return ret;
2941 }
2942
2943 static void ibmvnic_get_channels(struct net_device *netdev,
2944 struct ethtool_channels *channels)
2945 {
2946 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2947
2948 if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2949 channels->max_rx = adapter->max_rx_queues;
2950 channels->max_tx = adapter->max_tx_queues;
2951 } else {
2952 channels->max_rx = IBMVNIC_MAX_QUEUES;
2953 channels->max_tx = IBMVNIC_MAX_QUEUES;
2954 }
2955
2956 channels->max_other = 0;
2957 channels->max_combined = 0;
2958 channels->rx_count = adapter->req_rx_queues;
2959 channels->tx_count = adapter->req_tx_queues;
2960 channels->other_count = 0;
2961 channels->combined_count = 0;
2962 }
2963
2964 static int ibmvnic_set_channels(struct net_device *netdev,
2965 struct ethtool_channels *channels)
2966 {
2967 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2968 int ret;
2969
2970 ret = 0;
2971 adapter->desired.rx_queues = channels->rx_count;
2972 adapter->desired.tx_queues = channels->tx_count;
2973
2974 ret = wait_for_reset(adapter);
2975
2976 if (!ret &&
2977 (adapter->req_rx_queues != channels->rx_count ||
2978 adapter->req_tx_queues != channels->tx_count))
2979 netdev_info(netdev,
2980 "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2981 channels->rx_count, channels->tx_count,
2982 adapter->req_rx_queues, adapter->req_tx_queues);
2983 return ret;
2984 }
2985
2986 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2987 {
2988 struct ibmvnic_adapter *adapter = netdev_priv(dev);
2989 int i;
2990
2991 switch (stringset) {
2992 case ETH_SS_STATS:
2993 for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2994 i++, data += ETH_GSTRING_LEN)
2995 memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2996
2997 for (i = 0; i < adapter->req_tx_queues; i++) {
2998 snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2999 data += ETH_GSTRING_LEN;
3000
3001 snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
3002 data += ETH_GSTRING_LEN;
3003
3004 snprintf(data, ETH_GSTRING_LEN,
3005 "tx%d_dropped_packets", i);
3006 data += ETH_GSTRING_LEN;
3007 }
3008
3009 for (i = 0; i < adapter->req_rx_queues; i++) {
3010 snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
3011 data += ETH_GSTRING_LEN;
3012
3013 snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
3014 data += ETH_GSTRING_LEN;
3015
3016 snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
3017 data += ETH_GSTRING_LEN;
3018 }
3019 break;
3020
3021 case ETH_SS_PRIV_FLAGS:
3022 for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
3023 strcpy(data + i * ETH_GSTRING_LEN,
3024 ibmvnic_priv_flags[i]);
3025 break;
3026 default:
3027 return;
3028 }
3029 }
3030
3031 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
3032 {
3033 struct ibmvnic_adapter *adapter = netdev_priv(dev);
3034
3035 switch (sset) {
3036 case ETH_SS_STATS:
3037 return ARRAY_SIZE(ibmvnic_stats) +
3038 adapter->req_tx_queues * NUM_TX_STATS +
3039 adapter->req_rx_queues * NUM_RX_STATS;
3040 case ETH_SS_PRIV_FLAGS:
3041 return ARRAY_SIZE(ibmvnic_priv_flags);
3042 default:
3043 return -EOPNOTSUPP;
3044 }
3045 }
3046
3047 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
3048 struct ethtool_stats *stats, u64 *data)
3049 {
3050 struct ibmvnic_adapter *adapter = netdev_priv(dev);
3051 union ibmvnic_crq crq;
3052 int i, j;
3053 int rc;
3054
3055 memset(&crq, 0, sizeof(crq));
3056 crq.request_statistics.first = IBMVNIC_CRQ_CMD;
3057 crq.request_statistics.cmd = REQUEST_STATISTICS;
3058 crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
3059 crq.request_statistics.len =
3060 cpu_to_be32(sizeof(struct ibmvnic_statistics));
3061
3062 /* Wait for data to be written */
3063 reinit_completion(&adapter->stats_done);
3064 rc = ibmvnic_send_crq(adapter, &crq);
3065 if (rc)
3066 return;
3067 rc = ibmvnic_wait_for_completion(adapter, &adapter->stats_done, 10000);
3068 if (rc)
3069 return;
3070
3071 for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
3072 data[i] = be64_to_cpu(IBMVNIC_GET_STAT
3073 (adapter, ibmvnic_stats[i].offset));
3074
3075 for (j = 0; j < adapter->req_tx_queues; j++) {
3076 data[i] = adapter->tx_stats_buffers[j].packets;
3077 i++;
3078 data[i] = adapter->tx_stats_buffers[j].bytes;
3079 i++;
3080 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
3081 i++;
3082 }
3083
3084 for (j = 0; j < adapter->req_rx_queues; j++) {
3085 data[i] = adapter->rx_stats_buffers[j].packets;
3086 i++;
3087 data[i] = adapter->rx_stats_buffers[j].bytes;
3088 i++;
3089 data[i] = adapter->rx_stats_buffers[j].interrupts;
3090 i++;
3091 }
3092 }
3093
3094 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
3095 {
3096 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
3097
3098 return adapter->priv_flags;
3099 }
3100
3101 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
3102 {
3103 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
3104 bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
3105
3106 if (which_maxes)
3107 adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
3108 else
3109 adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
3110
3111 return 0;
3112 }
3113
3114 static const struct ethtool_ops ibmvnic_ethtool_ops = {
3115 .get_drvinfo = ibmvnic_get_drvinfo,
3116 .get_msglevel = ibmvnic_get_msglevel,
3117 .set_msglevel = ibmvnic_set_msglevel,
3118 .get_link = ibmvnic_get_link,
3119 .get_ringparam = ibmvnic_get_ringparam,
3120 .set_ringparam = ibmvnic_set_ringparam,
3121 .get_channels = ibmvnic_get_channels,
3122 .set_channels = ibmvnic_set_channels,
3123 .get_strings = ibmvnic_get_strings,
3124 .get_sset_count = ibmvnic_get_sset_count,
3125 .get_ethtool_stats = ibmvnic_get_ethtool_stats,
3126 .get_link_ksettings = ibmvnic_get_link_ksettings,
3127 .get_priv_flags = ibmvnic_get_priv_flags,
3128 .set_priv_flags = ibmvnic_set_priv_flags,
3129 };
3130
3131 /* Routines for managing CRQs/sCRQs */
3132
3133 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
3134 struct ibmvnic_sub_crq_queue *scrq)
3135 {
3136 int rc;
3137
3138 if (!scrq) {
3139 netdev_dbg(adapter->netdev, "Invalid scrq reset.\n");
3140 return -EINVAL;
3141 }
3142
3143 if (scrq->irq) {
3144 free_irq(scrq->irq, scrq);
3145 irq_dispose_mapping(scrq->irq);
3146 scrq->irq = 0;
3147 }
3148
3149 if (scrq->msgs) {
3150 memset(scrq->msgs, 0, 4 * PAGE_SIZE);
3151 atomic_set(&scrq->used, 0);
3152 scrq->cur = 0;
3153 scrq->ind_buf.index = 0;
3154 } else {
3155 netdev_dbg(adapter->netdev, "Invalid scrq reset\n");
3156 return -EINVAL;
3157 }
3158
3159 rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
3160 4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
3161 return rc;
3162 }
3163
3164 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
3165 {
3166 int i, rc;
3167
3168 if (!adapter->tx_scrq || !adapter->rx_scrq)
3169 return -EINVAL;
3170
3171 for (i = 0; i < adapter->req_tx_queues; i++) {
3172 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
3173 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
3174 if (rc)
3175 return rc;
3176 }
3177
3178 for (i = 0; i < adapter->req_rx_queues; i++) {
3179 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
3180 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
3181 if (rc)
3182 return rc;
3183 }
3184
3185 return rc;
3186 }
3187
3188 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
3189 struct ibmvnic_sub_crq_queue *scrq,
3190 bool do_h_free)
3191 {
3192 struct device *dev = &adapter->vdev->dev;
3193 long rc;
3194
3195 netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
3196
3197 if (do_h_free) {
3198 /* Close the sub-crqs */
3199 do {
3200 rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
3201 adapter->vdev->unit_address,
3202 scrq->crq_num);
3203 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3204
3205 if (rc) {
3206 netdev_err(adapter->netdev,
3207 "Failed to release sub-CRQ %16lx, rc = %ld\n",
3208 scrq->crq_num, rc);
3209 }
3210 }
3211
3212 dma_free_coherent(dev,
3213 IBMVNIC_IND_ARR_SZ,
3214 scrq->ind_buf.indir_arr,
3215 scrq->ind_buf.indir_dma);
3216
3217 dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
3218 DMA_BIDIRECTIONAL);
3219 free_pages((unsigned long)scrq->msgs, 2);
3220 kfree(scrq);
3221 }
3222
3223 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
3224 *adapter)
3225 {
3226 struct device *dev = &adapter->vdev->dev;
3227 struct ibmvnic_sub_crq_queue *scrq;
3228 int rc;
3229
3230 scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
3231 if (!scrq)
3232 return NULL;
3233
3234 scrq->msgs =
3235 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
3236 if (!scrq->msgs) {
3237 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
3238 goto zero_page_failed;
3239 }
3240
3241 scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
3242 DMA_BIDIRECTIONAL);
3243 if (dma_mapping_error(dev, scrq->msg_token)) {
3244 dev_warn(dev, "Couldn't map crq queue messages page\n");
3245 goto map_failed;
3246 }
3247
3248 rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
3249 4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
3250
3251 if (rc == H_RESOURCE)
3252 rc = ibmvnic_reset_crq(adapter);
3253
3254 if (rc == H_CLOSED) {
3255 dev_warn(dev, "Partner adapter not ready, waiting.\n");
3256 } else if (rc) {
3257 dev_warn(dev, "Error %d registering sub-crq\n", rc);
3258 goto reg_failed;
3259 }
3260
3261 scrq->adapter = adapter;
3262 scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
3263 scrq->ind_buf.index = 0;
3264
3265 scrq->ind_buf.indir_arr =
3266 dma_alloc_coherent(dev,
3267 IBMVNIC_IND_ARR_SZ,
3268 &scrq->ind_buf.indir_dma,
3269 GFP_KERNEL);
3270
3271 if (!scrq->ind_buf.indir_arr)
3272 goto indir_failed;
3273
3274 spin_lock_init(&scrq->lock);
3275
3276 netdev_dbg(adapter->netdev,
3277 "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
3278 scrq->crq_num, scrq->hw_irq, scrq->irq);
3279
3280 return scrq;
3281
3282 indir_failed:
3283 do {
3284 rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
3285 adapter->vdev->unit_address,
3286 scrq->crq_num);
3287 } while (rc == H_BUSY || rc == H_IS_LONG_BUSY(rc));
3288 reg_failed:
3289 dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
3290 DMA_BIDIRECTIONAL);
3291 map_failed:
3292 free_pages((unsigned long)scrq->msgs, 2);
3293 zero_page_failed:
3294 kfree(scrq);
3295
3296 return NULL;
3297 }
3298
3299 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
3300 {
3301 int i;
3302
3303 if (adapter->tx_scrq) {
3304 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
3305 if (!adapter->tx_scrq[i])
3306 continue;
3307
3308 netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
3309 i);
3310 ibmvnic_tx_scrq_clean_buffer(adapter, adapter->tx_scrq[i]);
3311 if (adapter->tx_scrq[i]->irq) {
3312 free_irq(adapter->tx_scrq[i]->irq,
3313 adapter->tx_scrq[i]);
3314 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
3315 adapter->tx_scrq[i]->irq = 0;
3316 }
3317
3318 release_sub_crq_queue(adapter, adapter->tx_scrq[i],
3319 do_h_free);
3320 }
3321
3322 kfree(adapter->tx_scrq);
3323 adapter->tx_scrq = NULL;
3324 adapter->num_active_tx_scrqs = 0;
3325 }
3326
3327 if (adapter->rx_scrq) {
3328 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
3329 if (!adapter->rx_scrq[i])
3330 continue;
3331
3332 netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
3333 i);
3334 if (adapter->rx_scrq[i]->irq) {
3335 free_irq(adapter->rx_scrq[i]->irq,
3336 adapter->rx_scrq[i]);
3337 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
3338 adapter->rx_scrq[i]->irq = 0;
3339 }
3340
3341 release_sub_crq_queue(adapter, adapter->rx_scrq[i],
3342 do_h_free);
3343 }
3344
3345 kfree(adapter->rx_scrq);
3346 adapter->rx_scrq = NULL;
3347 adapter->num_active_rx_scrqs = 0;
3348 }
3349 }
3350
3351 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
3352 struct ibmvnic_sub_crq_queue *scrq)
3353 {
3354 struct device *dev = &adapter->vdev->dev;
3355 unsigned long rc;
3356
3357 rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3358 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3359 if (rc)
3360 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
3361 scrq->hw_irq, rc);
3362 return rc;
3363 }
3364
3365 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
3366 struct ibmvnic_sub_crq_queue *scrq)
3367 {
3368 struct device *dev = &adapter->vdev->dev;
3369 unsigned long rc;
3370
3371 if (scrq->hw_irq > 0x100000000ULL) {
3372 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
3373 return 1;
3374 }
3375
3376 if (test_bit(0, &adapter->resetting) &&
3377 adapter->reset_reason == VNIC_RESET_MOBILITY) {
3378 u64 val = (0xff000000) | scrq->hw_irq;
3379
3380 rc = plpar_hcall_norets(H_EOI, val);
3381 /* H_EOI would fail with rc = H_FUNCTION when running
3382 * in XIVE mode which is expected, but not an error.
3383 */
3384 if (rc && (rc != H_FUNCTION))
3385 dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
3386 val, rc);
3387 }
3388
3389 rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3390 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3391 if (rc)
3392 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
3393 scrq->hw_irq, rc);
3394 return rc;
3395 }
3396
3397 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
3398 struct ibmvnic_sub_crq_queue *scrq)
3399 {
3400 struct device *dev = &adapter->vdev->dev;
3401 struct ibmvnic_tx_pool *tx_pool;
3402 struct ibmvnic_tx_buff *txbuff;
3403 struct netdev_queue *txq;
3404 union sub_crq *next;
3405 int index;
3406 int i;
3407
3408 restart_loop:
3409 while (pending_scrq(adapter, scrq)) {
3410 unsigned int pool = scrq->pool_index;
3411 int num_entries = 0;
3412 int total_bytes = 0;
3413 int num_packets = 0;
3414
3415 next = ibmvnic_next_scrq(adapter, scrq);
3416 for (i = 0; i < next->tx_comp.num_comps; i++) {
3417 index = be32_to_cpu(next->tx_comp.correlators[i]);
3418 if (index & IBMVNIC_TSO_POOL_MASK) {
3419 tx_pool = &adapter->tso_pool[pool];
3420 index &= ~IBMVNIC_TSO_POOL_MASK;
3421 } else {
3422 tx_pool = &adapter->tx_pool[pool];
3423 }
3424
3425 txbuff = &tx_pool->tx_buff[index];
3426 num_packets++;
3427 num_entries += txbuff->num_entries;
3428 if (txbuff->skb) {
3429 total_bytes += txbuff->skb->len;
3430 if (next->tx_comp.rcs[i]) {
3431 dev_err(dev, "tx error %x\n",
3432 next->tx_comp.rcs[i]);
3433 dev_kfree_skb_irq(txbuff->skb);
3434 } else {
3435 dev_consume_skb_irq(txbuff->skb);
3436 }
3437 txbuff->skb = NULL;
3438 } else {
3439 netdev_warn(adapter->netdev,
3440 "TX completion received with NULL socket buffer\n");
3441 }
3442 tx_pool->free_map[tx_pool->producer_index] = index;
3443 tx_pool->producer_index =
3444 (tx_pool->producer_index + 1) %
3445 tx_pool->num_buffers;
3446 }
3447 /* remove tx_comp scrq*/
3448 next->tx_comp.first = 0;
3449
3450 txq = netdev_get_tx_queue(adapter->netdev, scrq->pool_index);
3451 netdev_tx_completed_queue(txq, num_packets, total_bytes);
3452
3453 if (atomic_sub_return(num_entries, &scrq->used) <=
3454 (adapter->req_tx_entries_per_subcrq / 2) &&
3455 __netif_subqueue_stopped(adapter->netdev,
3456 scrq->pool_index)) {
3457 netif_wake_subqueue(adapter->netdev, scrq->pool_index);
3458 netdev_dbg(adapter->netdev, "Started queue %d\n",
3459 scrq->pool_index);
3460 }
3461 }
3462
3463 enable_scrq_irq(adapter, scrq);
3464
3465 if (pending_scrq(adapter, scrq)) {
3466 disable_scrq_irq(adapter, scrq);
3467 goto restart_loop;
3468 }
3469
3470 return 0;
3471 }
3472
3473 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
3474 {
3475 struct ibmvnic_sub_crq_queue *scrq = instance;
3476 struct ibmvnic_adapter *adapter = scrq->adapter;
3477
3478 disable_scrq_irq(adapter, scrq);
3479 ibmvnic_complete_tx(adapter, scrq);
3480
3481 return IRQ_HANDLED;
3482 }
3483
3484 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
3485 {
3486 struct ibmvnic_sub_crq_queue *scrq = instance;
3487 struct ibmvnic_adapter *adapter = scrq->adapter;
3488
3489 /* When booting a kdump kernel we can hit pending interrupts
3490 * prior to completing driver initialization.
3491 */
3492 if (unlikely(adapter->state != VNIC_OPEN))
3493 return IRQ_NONE;
3494
3495 adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
3496
3497 if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
3498 disable_scrq_irq(adapter, scrq);
3499 __napi_schedule(&adapter->napi[scrq->scrq_num]);
3500 }
3501
3502 return IRQ_HANDLED;
3503 }
3504
3505 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
3506 {
3507 struct device *dev = &adapter->vdev->dev;
3508 struct ibmvnic_sub_crq_queue *scrq;
3509 int i = 0, j = 0;
3510 int rc = 0;
3511
3512 for (i = 0; i < adapter->req_tx_queues; i++) {
3513 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
3514 i);
3515 scrq = adapter->tx_scrq[i];
3516 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3517
3518 if (!scrq->irq) {
3519 rc = -EINVAL;
3520 dev_err(dev, "Error mapping irq\n");
3521 goto req_tx_irq_failed;
3522 }
3523
3524 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
3525 adapter->vdev->unit_address, i);
3526 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
3527 0, scrq->name, scrq);
3528
3529 if (rc) {
3530 dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
3531 scrq->irq, rc);
3532 irq_dispose_mapping(scrq->irq);
3533 goto req_tx_irq_failed;
3534 }
3535 }
3536
3537 for (i = 0; i < adapter->req_rx_queues; i++) {
3538 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
3539 i);
3540 scrq = adapter->rx_scrq[i];
3541 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3542 if (!scrq->irq) {
3543 rc = -EINVAL;
3544 dev_err(dev, "Error mapping irq\n");
3545 goto req_rx_irq_failed;
3546 }
3547 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
3548 adapter->vdev->unit_address, i);
3549 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
3550 0, scrq->name, scrq);
3551 if (rc) {
3552 dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
3553 scrq->irq, rc);
3554 irq_dispose_mapping(scrq->irq);
3555 goto req_rx_irq_failed;
3556 }
3557 }
3558 return rc;
3559
3560 req_rx_irq_failed:
3561 for (j = 0; j < i; j++) {
3562 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
3563 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
3564 }
3565 i = adapter->req_tx_queues;
3566 req_tx_irq_failed:
3567 for (j = 0; j < i; j++) {
3568 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
3569 irq_dispose_mapping(adapter->tx_scrq[j]->irq);
3570 }
3571 release_sub_crqs(adapter, 1);
3572 return rc;
3573 }
3574
3575 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
3576 {
3577 struct device *dev = &adapter->vdev->dev;
3578 struct ibmvnic_sub_crq_queue **allqueues;
3579 int registered_queues = 0;
3580 int total_queues;
3581 int more = 0;
3582 int i;
3583
3584 total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
3585
3586 allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
3587 if (!allqueues)
3588 return -1;
3589
3590 for (i = 0; i < total_queues; i++) {
3591 allqueues[i] = init_sub_crq_queue(adapter);
3592 if (!allqueues[i]) {
3593 dev_warn(dev, "Couldn't allocate all sub-crqs\n");
3594 break;
3595 }
3596 registered_queues++;
3597 }
3598
3599 /* Make sure we were able to register the minimum number of queues */
3600 if (registered_queues <
3601 adapter->min_tx_queues + adapter->min_rx_queues) {
3602 dev_err(dev, "Fatal: Couldn't init min number of sub-crqs\n");
3603 goto tx_failed;
3604 }
3605
3606 /* Distribute the failed allocated queues*/
3607 for (i = 0; i < total_queues - registered_queues + more ; i++) {
3608 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3609 switch (i % 3) {
3610 case 0:
3611 if (adapter->req_rx_queues > adapter->min_rx_queues)
3612 adapter->req_rx_queues--;
3613 else
3614 more++;
3615 break;
3616 case 1:
3617 if (adapter->req_tx_queues > adapter->min_tx_queues)
3618 adapter->req_tx_queues--;
3619 else
3620 more++;
3621 break;
3622 }
3623 }
3624
3625 adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3626 sizeof(*adapter->tx_scrq), GFP_KERNEL);
3627 if (!adapter->tx_scrq)
3628 goto tx_failed;
3629
3630 for (i = 0; i < adapter->req_tx_queues; i++) {
3631 adapter->tx_scrq[i] = allqueues[i];
3632 adapter->tx_scrq[i]->pool_index = i;
3633 adapter->num_active_tx_scrqs++;
3634 }
3635
3636 adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3637 sizeof(*adapter->rx_scrq), GFP_KERNEL);
3638 if (!adapter->rx_scrq)
3639 goto rx_failed;
3640
3641 for (i = 0; i < adapter->req_rx_queues; i++) {
3642 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3643 adapter->rx_scrq[i]->scrq_num = i;
3644 adapter->num_active_rx_scrqs++;
3645 }
3646
3647 kfree(allqueues);
3648 return 0;
3649
3650 rx_failed:
3651 kfree(adapter->tx_scrq);
3652 adapter->tx_scrq = NULL;
3653 tx_failed:
3654 for (i = 0; i < registered_queues; i++)
3655 release_sub_crq_queue(adapter, allqueues[i], 1);
3656 kfree(allqueues);
3657 return -1;
3658 }
3659
3660 static void send_request_cap(struct ibmvnic_adapter *adapter, int retry)
3661 {
3662 struct device *dev = &adapter->vdev->dev;
3663 union ibmvnic_crq crq;
3664 int max_entries;
3665
3666 if (!retry) {
3667 /* Sub-CRQ entries are 32 byte long */
3668 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3669
3670 if (adapter->min_tx_entries_per_subcrq > entries_page ||
3671 adapter->min_rx_add_entries_per_subcrq > entries_page) {
3672 dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3673 return;
3674 }
3675
3676 if (adapter->desired.mtu)
3677 adapter->req_mtu = adapter->desired.mtu;
3678 else
3679 adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3680
3681 if (!adapter->desired.tx_entries)
3682 adapter->desired.tx_entries =
3683 adapter->max_tx_entries_per_subcrq;
3684 if (!adapter->desired.rx_entries)
3685 adapter->desired.rx_entries =
3686 adapter->max_rx_add_entries_per_subcrq;
3687
3688 max_entries = IBMVNIC_MAX_LTB_SIZE /
3689 (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3690
3691 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3692 adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3693 adapter->desired.tx_entries = max_entries;
3694 }
3695
3696 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3697 adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3698 adapter->desired.rx_entries = max_entries;
3699 }
3700
3701 if (adapter->desired.tx_entries)
3702 adapter->req_tx_entries_per_subcrq =
3703 adapter->desired.tx_entries;
3704 else
3705 adapter->req_tx_entries_per_subcrq =
3706 adapter->max_tx_entries_per_subcrq;
3707
3708 if (adapter->desired.rx_entries)
3709 adapter->req_rx_add_entries_per_subcrq =
3710 adapter->desired.rx_entries;
3711 else
3712 adapter->req_rx_add_entries_per_subcrq =
3713 adapter->max_rx_add_entries_per_subcrq;
3714
3715 if (adapter->desired.tx_queues)
3716 adapter->req_tx_queues =
3717 adapter->desired.tx_queues;
3718 else
3719 adapter->req_tx_queues =
3720 adapter->opt_tx_comp_sub_queues;
3721
3722 if (adapter->desired.rx_queues)
3723 adapter->req_rx_queues =
3724 adapter->desired.rx_queues;
3725 else
3726 adapter->req_rx_queues =
3727 adapter->opt_rx_comp_queues;
3728
3729 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3730 }
3731
3732 memset(&crq, 0, sizeof(crq));
3733 crq.request_capability.first = IBMVNIC_CRQ_CMD;
3734 crq.request_capability.cmd = REQUEST_CAPABILITY;
3735
3736 crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3737 crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3738 atomic_inc(&adapter->running_cap_crqs);
3739 ibmvnic_send_crq(adapter, &crq);
3740
3741 crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3742 crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3743 atomic_inc(&adapter->running_cap_crqs);
3744 ibmvnic_send_crq(adapter, &crq);
3745
3746 crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3747 crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3748 atomic_inc(&adapter->running_cap_crqs);
3749 ibmvnic_send_crq(adapter, &crq);
3750
3751 crq.request_capability.capability =
3752 cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3753 crq.request_capability.number =
3754 cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3755 atomic_inc(&adapter->running_cap_crqs);
3756 ibmvnic_send_crq(adapter, &crq);
3757
3758 crq.request_capability.capability =
3759 cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3760 crq.request_capability.number =
3761 cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3762 atomic_inc(&adapter->running_cap_crqs);
3763 ibmvnic_send_crq(adapter, &crq);
3764
3765 crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3766 crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3767 atomic_inc(&adapter->running_cap_crqs);
3768 ibmvnic_send_crq(adapter, &crq);
3769
3770 if (adapter->netdev->flags & IFF_PROMISC) {
3771 if (adapter->promisc_supported) {
3772 crq.request_capability.capability =
3773 cpu_to_be16(PROMISC_REQUESTED);
3774 crq.request_capability.number = cpu_to_be64(1);
3775 atomic_inc(&adapter->running_cap_crqs);
3776 ibmvnic_send_crq(adapter, &crq);
3777 }
3778 } else {
3779 crq.request_capability.capability =
3780 cpu_to_be16(PROMISC_REQUESTED);
3781 crq.request_capability.number = cpu_to_be64(0);
3782 atomic_inc(&adapter->running_cap_crqs);
3783 ibmvnic_send_crq(adapter, &crq);
3784 }
3785 }
3786
3787 static int pending_scrq(struct ibmvnic_adapter *adapter,
3788 struct ibmvnic_sub_crq_queue *scrq)
3789 {
3790 union sub_crq *entry = &scrq->msgs[scrq->cur];
3791 int rc;
3792
3793 rc = !!(entry->generic.first & IBMVNIC_CRQ_CMD_RSP);
3794
3795 /* Ensure that the SCRQ valid flag is loaded prior to loading the
3796 * contents of the SCRQ descriptor
3797 */
3798 dma_rmb();
3799
3800 return rc;
3801 }
3802
3803 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3804 struct ibmvnic_sub_crq_queue *scrq)
3805 {
3806 union sub_crq *entry;
3807 unsigned long flags;
3808
3809 spin_lock_irqsave(&scrq->lock, flags);
3810 entry = &scrq->msgs[scrq->cur];
3811 if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3812 if (++scrq->cur == scrq->size)
3813 scrq->cur = 0;
3814 } else {
3815 entry = NULL;
3816 }
3817 spin_unlock_irqrestore(&scrq->lock, flags);
3818
3819 /* Ensure that the SCRQ valid flag is loaded prior to loading the
3820 * contents of the SCRQ descriptor
3821 */
3822 dma_rmb();
3823
3824 return entry;
3825 }
3826
3827 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3828 {
3829 struct ibmvnic_crq_queue *queue = &adapter->crq;
3830 union ibmvnic_crq *crq;
3831
3832 crq = &queue->msgs[queue->cur];
3833 if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3834 if (++queue->cur == queue->size)
3835 queue->cur = 0;
3836 } else {
3837 crq = NULL;
3838 }
3839
3840 return crq;
3841 }
3842
3843 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3844 {
3845 switch (rc) {
3846 case H_PARAMETER:
3847 dev_warn_ratelimited(dev,
3848 "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3849 func, rc);
3850 break;
3851 case H_CLOSED:
3852 dev_warn_ratelimited(dev,
3853 "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3854 func, rc);
3855 break;
3856 default:
3857 dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3858 break;
3859 }
3860 }
3861
3862 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3863 u64 remote_handle, u64 ioba, u64 num_entries)
3864 {
3865 unsigned int ua = adapter->vdev->unit_address;
3866 struct device *dev = &adapter->vdev->dev;
3867 int rc;
3868
3869 /* Make sure the hypervisor sees the complete request */
3870 dma_wmb();
3871 rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3872 cpu_to_be64(remote_handle),
3873 ioba, num_entries);
3874
3875 if (rc)
3876 print_subcrq_error(dev, rc, __func__);
3877
3878 return rc;
3879 }
3880
3881 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3882 union ibmvnic_crq *crq)
3883 {
3884 unsigned int ua = adapter->vdev->unit_address;
3885 struct device *dev = &adapter->vdev->dev;
3886 u64 *u64_crq = (u64 *)crq;
3887 int rc;
3888
3889 netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3890 (unsigned long)cpu_to_be64(u64_crq[0]),
3891 (unsigned long)cpu_to_be64(u64_crq[1]));
3892
3893 if (!adapter->crq.active &&
3894 crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3895 dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3896 return -EINVAL;
3897 }
3898
3899 /* Make sure the hypervisor sees the complete request */
3900 dma_wmb();
3901
3902 rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3903 cpu_to_be64(u64_crq[0]),
3904 cpu_to_be64(u64_crq[1]));
3905
3906 if (rc) {
3907 if (rc == H_CLOSED) {
3908 dev_warn(dev, "CRQ Queue closed\n");
3909 /* do not reset, report the fail, wait for passive init from server */
3910 }
3911
3912 dev_warn(dev, "Send error (rc=%d)\n", rc);
3913 }
3914
3915 return rc;
3916 }
3917
3918 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3919 {
3920 struct device *dev = &adapter->vdev->dev;
3921 union ibmvnic_crq crq;
3922 int retries = 100;
3923 int rc;
3924
3925 memset(&crq, 0, sizeof(crq));
3926 crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3927 crq.generic.cmd = IBMVNIC_CRQ_INIT;
3928 netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3929
3930 do {
3931 rc = ibmvnic_send_crq(adapter, &crq);
3932 if (rc != H_CLOSED)
3933 break;
3934 retries--;
3935 msleep(50);
3936
3937 } while (retries > 0);
3938
3939 if (rc) {
3940 dev_err(dev, "Failed to send init request, rc = %d\n", rc);
3941 return rc;
3942 }
3943
3944 return 0;
3945 }
3946
3947 struct vnic_login_client_data {
3948 u8 type;
3949 __be16 len;
3950 char name[];
3951 } __packed;
3952
3953 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3954 {
3955 int len;
3956
3957 /* Calculate the amount of buffer space needed for the
3958 * vnic client data in the login buffer. There are four entries,
3959 * OS name, LPAR name, device name, and a null last entry.
3960 */
3961 len = 4 * sizeof(struct vnic_login_client_data);
3962 len += 6; /* "Linux" plus NULL */
3963 len += strlen(utsname()->nodename) + 1;
3964 len += strlen(adapter->netdev->name) + 1;
3965
3966 return len;
3967 }
3968
3969 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3970 struct vnic_login_client_data *vlcd)
3971 {
3972 const char *os_name = "Linux";
3973 int len;
3974
3975 /* Type 1 - LPAR OS */
3976 vlcd->type = 1;
3977 len = strlen(os_name) + 1;
3978 vlcd->len = cpu_to_be16(len);
3979 strscpy(vlcd->name, os_name, len);
3980 vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3981
3982 /* Type 2 - LPAR name */
3983 vlcd->type = 2;
3984 len = strlen(utsname()->nodename) + 1;
3985 vlcd->len = cpu_to_be16(len);
3986 strscpy(vlcd->name, utsname()->nodename, len);
3987 vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3988
3989 /* Type 3 - device name */
3990 vlcd->type = 3;
3991 len = strlen(adapter->netdev->name) + 1;
3992 vlcd->len = cpu_to_be16(len);
3993 strscpy(vlcd->name, adapter->netdev->name, len);
3994 }
3995
3996 static int send_login(struct ibmvnic_adapter *adapter)
3997 {
3998 struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3999 struct ibmvnic_login_buffer *login_buffer;
4000 struct device *dev = &adapter->vdev->dev;
4001 struct vnic_login_client_data *vlcd;
4002 dma_addr_t rsp_buffer_token;
4003 dma_addr_t buffer_token;
4004 size_t rsp_buffer_size;
4005 union ibmvnic_crq crq;
4006 int client_data_len;
4007 size_t buffer_size;
4008 __be64 *tx_list_p;
4009 __be64 *rx_list_p;
4010 int rc;
4011 int i;
4012
4013 if (!adapter->tx_scrq || !adapter->rx_scrq) {
4014 netdev_err(adapter->netdev,
4015 "RX or TX queues are not allocated, device login failed\n");
4016 return -1;
4017 }
4018
4019 release_login_buffer(adapter);
4020 release_login_rsp_buffer(adapter);
4021
4022 client_data_len = vnic_client_data_len(adapter);
4023
4024 buffer_size =
4025 sizeof(struct ibmvnic_login_buffer) +
4026 sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
4027 client_data_len;
4028
4029 login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
4030 if (!login_buffer)
4031 goto buf_alloc_failed;
4032
4033 buffer_token = dma_map_single(dev, login_buffer, buffer_size,
4034 DMA_TO_DEVICE);
4035 if (dma_mapping_error(dev, buffer_token)) {
4036 dev_err(dev, "Couldn't map login buffer\n");
4037 goto buf_map_failed;
4038 }
4039
4040 rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
4041 sizeof(u64) * adapter->req_tx_queues +
4042 sizeof(u64) * adapter->req_rx_queues +
4043 sizeof(u64) * adapter->req_rx_queues +
4044 sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
4045
4046 login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
4047 if (!login_rsp_buffer)
4048 goto buf_rsp_alloc_failed;
4049
4050 rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
4051 rsp_buffer_size, DMA_FROM_DEVICE);
4052 if (dma_mapping_error(dev, rsp_buffer_token)) {
4053 dev_err(dev, "Couldn't map login rsp buffer\n");
4054 goto buf_rsp_map_failed;
4055 }
4056
4057 adapter->login_buf = login_buffer;
4058 adapter->login_buf_token = buffer_token;
4059 adapter->login_buf_sz = buffer_size;
4060 adapter->login_rsp_buf = login_rsp_buffer;
4061 adapter->login_rsp_buf_token = rsp_buffer_token;
4062 adapter->login_rsp_buf_sz = rsp_buffer_size;
4063
4064 login_buffer->len = cpu_to_be32(buffer_size);
4065 login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
4066 login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
4067 login_buffer->off_txcomp_subcrqs =
4068 cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
4069 login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
4070 login_buffer->off_rxcomp_subcrqs =
4071 cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
4072 sizeof(u64) * adapter->req_tx_queues);
4073 login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
4074 login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
4075
4076 tx_list_p = (__be64 *)((char *)login_buffer +
4077 sizeof(struct ibmvnic_login_buffer));
4078 rx_list_p = (__be64 *)((char *)login_buffer +
4079 sizeof(struct ibmvnic_login_buffer) +
4080 sizeof(u64) * adapter->req_tx_queues);
4081
4082 for (i = 0; i < adapter->req_tx_queues; i++) {
4083 if (adapter->tx_scrq[i]) {
4084 tx_list_p[i] =
4085 cpu_to_be64(adapter->tx_scrq[i]->crq_num);
4086 }
4087 }
4088
4089 for (i = 0; i < adapter->req_rx_queues; i++) {
4090 if (adapter->rx_scrq[i]) {
4091 rx_list_p[i] =
4092 cpu_to_be64(adapter->rx_scrq[i]->crq_num);
4093 }
4094 }
4095
4096 /* Insert vNIC login client data */
4097 vlcd = (struct vnic_login_client_data *)
4098 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
4099 login_buffer->client_data_offset =
4100 cpu_to_be32((char *)vlcd - (char *)login_buffer);
4101 login_buffer->client_data_len = cpu_to_be32(client_data_len);
4102
4103 vnic_add_client_data(adapter, vlcd);
4104
4105 netdev_dbg(adapter->netdev, "Login Buffer:\n");
4106 for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
4107 netdev_dbg(adapter->netdev, "%016lx\n",
4108 ((unsigned long *)(adapter->login_buf))[i]);
4109 }
4110
4111 memset(&crq, 0, sizeof(crq));
4112 crq.login.first = IBMVNIC_CRQ_CMD;
4113 crq.login.cmd = LOGIN;
4114 crq.login.ioba = cpu_to_be32(buffer_token);
4115 crq.login.len = cpu_to_be32(buffer_size);
4116
4117 adapter->login_pending = true;
4118 rc = ibmvnic_send_crq(adapter, &crq);
4119 if (rc) {
4120 adapter->login_pending = false;
4121 netdev_err(adapter->netdev, "Failed to send login, rc=%d\n", rc);
4122 goto buf_rsp_map_failed;
4123 }
4124
4125 return 0;
4126
4127 buf_rsp_map_failed:
4128 kfree(login_rsp_buffer);
4129 adapter->login_rsp_buf = NULL;
4130 buf_rsp_alloc_failed:
4131 dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
4132 buf_map_failed:
4133 kfree(login_buffer);
4134 adapter->login_buf = NULL;
4135 buf_alloc_failed:
4136 return -1;
4137 }
4138
4139 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
4140 u32 len, u8 map_id)
4141 {
4142 union ibmvnic_crq crq;
4143
4144 memset(&crq, 0, sizeof(crq));
4145 crq.request_map.first = IBMVNIC_CRQ_CMD;
4146 crq.request_map.cmd = REQUEST_MAP;
4147 crq.request_map.map_id = map_id;
4148 crq.request_map.ioba = cpu_to_be32(addr);
4149 crq.request_map.len = cpu_to_be32(len);
4150 return ibmvnic_send_crq(adapter, &crq);
4151 }
4152
4153 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
4154 {
4155 union ibmvnic_crq crq;
4156
4157 memset(&crq, 0, sizeof(crq));
4158 crq.request_unmap.first = IBMVNIC_CRQ_CMD;
4159 crq.request_unmap.cmd = REQUEST_UNMAP;
4160 crq.request_unmap.map_id = map_id;
4161 return ibmvnic_send_crq(adapter, &crq);
4162 }
4163
4164 static void send_query_map(struct ibmvnic_adapter *adapter)
4165 {
4166 union ibmvnic_crq crq;
4167
4168 memset(&crq, 0, sizeof(crq));
4169 crq.query_map.first = IBMVNIC_CRQ_CMD;
4170 crq.query_map.cmd = QUERY_MAP;
4171 ibmvnic_send_crq(adapter, &crq);
4172 }
4173
4174 /* Send a series of CRQs requesting various capabilities of the VNIC server */
4175 static void send_query_cap(struct ibmvnic_adapter *adapter)
4176 {
4177 union ibmvnic_crq crq;
4178
4179 atomic_set(&adapter->running_cap_crqs, 0);
4180 memset(&crq, 0, sizeof(crq));
4181 crq.query_capability.first = IBMVNIC_CRQ_CMD;
4182 crq.query_capability.cmd = QUERY_CAPABILITY;
4183
4184 crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
4185 atomic_inc(&adapter->running_cap_crqs);
4186 ibmvnic_send_crq(adapter, &crq);
4187
4188 crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
4189 atomic_inc(&adapter->running_cap_crqs);
4190 ibmvnic_send_crq(adapter, &crq);
4191
4192 crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
4193 atomic_inc(&adapter->running_cap_crqs);
4194 ibmvnic_send_crq(adapter, &crq);
4195
4196 crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
4197 atomic_inc(&adapter->running_cap_crqs);
4198 ibmvnic_send_crq(adapter, &crq);
4199
4200 crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
4201 atomic_inc(&adapter->running_cap_crqs);
4202 ibmvnic_send_crq(adapter, &crq);
4203
4204 crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
4205 atomic_inc(&adapter->running_cap_crqs);
4206 ibmvnic_send_crq(adapter, &crq);
4207
4208 crq.query_capability.capability =
4209 cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
4210 atomic_inc(&adapter->running_cap_crqs);
4211 ibmvnic_send_crq(adapter, &crq);
4212
4213 crq.query_capability.capability =
4214 cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
4215 atomic_inc(&adapter->running_cap_crqs);
4216 ibmvnic_send_crq(adapter, &crq);
4217
4218 crq.query_capability.capability =
4219 cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
4220 atomic_inc(&adapter->running_cap_crqs);
4221 ibmvnic_send_crq(adapter, &crq);
4222
4223 crq.query_capability.capability =
4224 cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
4225 atomic_inc(&adapter->running_cap_crqs);
4226 ibmvnic_send_crq(adapter, &crq);
4227
4228 crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
4229 atomic_inc(&adapter->running_cap_crqs);
4230 ibmvnic_send_crq(adapter, &crq);
4231
4232 crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
4233 atomic_inc(&adapter->running_cap_crqs);
4234 ibmvnic_send_crq(adapter, &crq);
4235
4236 crq.query_capability.capability = cpu_to_be16(MIN_MTU);
4237 atomic_inc(&adapter->running_cap_crqs);
4238 ibmvnic_send_crq(adapter, &crq);
4239
4240 crq.query_capability.capability = cpu_to_be16(MAX_MTU);
4241 atomic_inc(&adapter->running_cap_crqs);
4242 ibmvnic_send_crq(adapter, &crq);
4243
4244 crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
4245 atomic_inc(&adapter->running_cap_crqs);
4246 ibmvnic_send_crq(adapter, &crq);
4247
4248 crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
4249 atomic_inc(&adapter->running_cap_crqs);
4250 ibmvnic_send_crq(adapter, &crq);
4251
4252 crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
4253 atomic_inc(&adapter->running_cap_crqs);
4254 ibmvnic_send_crq(adapter, &crq);
4255
4256 crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
4257 atomic_inc(&adapter->running_cap_crqs);
4258 ibmvnic_send_crq(adapter, &crq);
4259
4260 crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
4261 atomic_inc(&adapter->running_cap_crqs);
4262 ibmvnic_send_crq(adapter, &crq);
4263
4264 crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
4265 atomic_inc(&adapter->running_cap_crqs);
4266 ibmvnic_send_crq(adapter, &crq);
4267
4268 crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
4269 atomic_inc(&adapter->running_cap_crqs);
4270 ibmvnic_send_crq(adapter, &crq);
4271
4272 crq.query_capability.capability =
4273 cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
4274 atomic_inc(&adapter->running_cap_crqs);
4275 ibmvnic_send_crq(adapter, &crq);
4276
4277 crq.query_capability.capability =
4278 cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
4279 atomic_inc(&adapter->running_cap_crqs);
4280 ibmvnic_send_crq(adapter, &crq);
4281
4282 crq.query_capability.capability =
4283 cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
4284 atomic_inc(&adapter->running_cap_crqs);
4285 ibmvnic_send_crq(adapter, &crq);
4286
4287 crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
4288 atomic_inc(&adapter->running_cap_crqs);
4289 ibmvnic_send_crq(adapter, &crq);
4290 }
4291
4292 static void send_query_ip_offload(struct ibmvnic_adapter *adapter)
4293 {
4294 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4295 struct device *dev = &adapter->vdev->dev;
4296 union ibmvnic_crq crq;
4297
4298 adapter->ip_offload_tok =
4299 dma_map_single(dev,
4300 &adapter->ip_offload_buf,
4301 buf_sz,
4302 DMA_FROM_DEVICE);
4303
4304 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4305 if (!firmware_has_feature(FW_FEATURE_CMO))
4306 dev_err(dev, "Couldn't map offload buffer\n");
4307 return;
4308 }
4309
4310 memset(&crq, 0, sizeof(crq));
4311 crq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4312 crq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4313 crq.query_ip_offload.len = cpu_to_be32(buf_sz);
4314 crq.query_ip_offload.ioba =
4315 cpu_to_be32(adapter->ip_offload_tok);
4316
4317 ibmvnic_send_crq(adapter, &crq);
4318 }
4319
4320 static void send_control_ip_offload(struct ibmvnic_adapter *adapter)
4321 {
4322 struct ibmvnic_control_ip_offload_buffer *ctrl_buf = &adapter->ip_offload_ctrl;
4323 struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
4324 struct device *dev = &adapter->vdev->dev;
4325 netdev_features_t old_hw_features = 0;
4326 union ibmvnic_crq crq;
4327
4328 adapter->ip_offload_ctrl_tok =
4329 dma_map_single(dev,
4330 ctrl_buf,
4331 sizeof(adapter->ip_offload_ctrl),
4332 DMA_TO_DEVICE);
4333
4334 if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
4335 dev_err(dev, "Couldn't map ip offload control buffer\n");
4336 return;
4337 }
4338
4339 ctrl_buf->len = cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4340 ctrl_buf->version = cpu_to_be32(INITIAL_VERSION_IOB);
4341 ctrl_buf->ipv4_chksum = buf->ipv4_chksum;
4342 ctrl_buf->ipv6_chksum = buf->ipv6_chksum;
4343 ctrl_buf->tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
4344 ctrl_buf->udp_ipv4_chksum = buf->udp_ipv4_chksum;
4345 ctrl_buf->tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
4346 ctrl_buf->udp_ipv6_chksum = buf->udp_ipv6_chksum;
4347 ctrl_buf->large_tx_ipv4 = buf->large_tx_ipv4;
4348 ctrl_buf->large_tx_ipv6 = buf->large_tx_ipv6;
4349
4350 /* large_rx disabled for now, additional features needed */
4351 ctrl_buf->large_rx_ipv4 = 0;
4352 ctrl_buf->large_rx_ipv6 = 0;
4353
4354 if (adapter->state != VNIC_PROBING) {
4355 old_hw_features = adapter->netdev->hw_features;
4356 adapter->netdev->hw_features = 0;
4357 }
4358
4359 adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
4360
4361 if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
4362 adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
4363
4364 if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
4365 adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
4366
4367 if ((adapter->netdev->features &
4368 (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
4369 adapter->netdev->hw_features |= NETIF_F_RXCSUM;
4370
4371 if (buf->large_tx_ipv4)
4372 adapter->netdev->hw_features |= NETIF_F_TSO;
4373 if (buf->large_tx_ipv6)
4374 adapter->netdev->hw_features |= NETIF_F_TSO6;
4375
4376 if (adapter->state == VNIC_PROBING) {
4377 adapter->netdev->features |= adapter->netdev->hw_features;
4378 } else if (old_hw_features != adapter->netdev->hw_features) {
4379 netdev_features_t tmp = 0;
4380
4381 /* disable features no longer supported */
4382 adapter->netdev->features &= adapter->netdev->hw_features;
4383 /* turn on features now supported if previously enabled */
4384 tmp = (old_hw_features ^ adapter->netdev->hw_features) &
4385 adapter->netdev->hw_features;
4386 adapter->netdev->features |=
4387 tmp & adapter->netdev->wanted_features;
4388 }
4389
4390 memset(&crq, 0, sizeof(crq));
4391 crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
4392 crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
4393 crq.control_ip_offload.len =
4394 cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4395 crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
4396 ibmvnic_send_crq(adapter, &crq);
4397 }
4398
4399 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
4400 struct ibmvnic_adapter *adapter)
4401 {
4402 struct device *dev = &adapter->vdev->dev;
4403
4404 if (crq->get_vpd_size_rsp.rc.code) {
4405 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
4406 crq->get_vpd_size_rsp.rc.code);
4407 complete(&adapter->fw_done);
4408 return;
4409 }
4410
4411 adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
4412 complete(&adapter->fw_done);
4413 }
4414
4415 static void handle_vpd_rsp(union ibmvnic_crq *crq,
4416 struct ibmvnic_adapter *adapter)
4417 {
4418 struct device *dev = &adapter->vdev->dev;
4419 unsigned char *substr = NULL;
4420 u8 fw_level_len = 0;
4421
4422 memset(adapter->fw_version, 0, 32);
4423
4424 dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
4425 DMA_FROM_DEVICE);
4426
4427 if (crq->get_vpd_rsp.rc.code) {
4428 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
4429 crq->get_vpd_rsp.rc.code);
4430 goto complete;
4431 }
4432
4433 /* get the position of the firmware version info
4434 * located after the ASCII 'RM' substring in the buffer
4435 */
4436 substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
4437 if (!substr) {
4438 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
4439 goto complete;
4440 }
4441
4442 /* get length of firmware level ASCII substring */
4443 if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
4444 fw_level_len = *(substr + 2);
4445 } else {
4446 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
4447 goto complete;
4448 }
4449
4450 /* copy firmware version string from vpd into adapter */
4451 if ((substr + 3 + fw_level_len) <
4452 (adapter->vpd->buff + adapter->vpd->len)) {
4453 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
4454 } else {
4455 dev_info(dev, "FW substr extrapolated VPD buff\n");
4456 }
4457
4458 complete:
4459 if (adapter->fw_version[0] == '\0')
4460 strscpy((char *)adapter->fw_version, "N/A", sizeof(adapter->fw_version));
4461 complete(&adapter->fw_done);
4462 }
4463
4464 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
4465 {
4466 struct device *dev = &adapter->vdev->dev;
4467 struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
4468 int i;
4469
4470 dma_unmap_single(dev, adapter->ip_offload_tok,
4471 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
4472
4473 netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
4474 for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
4475 netdev_dbg(adapter->netdev, "%016lx\n",
4476 ((unsigned long *)(buf))[i]);
4477
4478 netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
4479 netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
4480 netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
4481 buf->tcp_ipv4_chksum);
4482 netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
4483 buf->tcp_ipv6_chksum);
4484 netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
4485 buf->udp_ipv4_chksum);
4486 netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
4487 buf->udp_ipv6_chksum);
4488 netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
4489 buf->large_tx_ipv4);
4490 netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
4491 buf->large_tx_ipv6);
4492 netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
4493 buf->large_rx_ipv4);
4494 netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
4495 buf->large_rx_ipv6);
4496 netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
4497 buf->max_ipv4_header_size);
4498 netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
4499 buf->max_ipv6_header_size);
4500 netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
4501 buf->max_tcp_header_size);
4502 netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
4503 buf->max_udp_header_size);
4504 netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
4505 buf->max_large_tx_size);
4506 netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
4507 buf->max_large_rx_size);
4508 netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
4509 buf->ipv6_extension_header);
4510 netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
4511 buf->tcp_pseudosum_req);
4512 netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
4513 buf->num_ipv6_ext_headers);
4514 netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
4515 buf->off_ipv6_ext_headers);
4516
4517 send_control_ip_offload(adapter);
4518 }
4519
4520 static const char *ibmvnic_fw_err_cause(u16 cause)
4521 {
4522 switch (cause) {
4523 case ADAPTER_PROBLEM:
4524 return "adapter problem";
4525 case BUS_PROBLEM:
4526 return "bus problem";
4527 case FW_PROBLEM:
4528 return "firmware problem";
4529 case DD_PROBLEM:
4530 return "device driver problem";
4531 case EEH_RECOVERY:
4532 return "EEH recovery";
4533 case FW_UPDATED:
4534 return "firmware updated";
4535 case LOW_MEMORY:
4536 return "low Memory";
4537 default:
4538 return "unknown";
4539 }
4540 }
4541
4542 static void handle_error_indication(union ibmvnic_crq *crq,
4543 struct ibmvnic_adapter *adapter)
4544 {
4545 struct device *dev = &adapter->vdev->dev;
4546 u16 cause;
4547
4548 cause = be16_to_cpu(crq->error_indication.error_cause);
4549
4550 dev_warn_ratelimited(dev,
4551 "Firmware reports %serror, cause: %s. Starting recovery...\n",
4552 crq->error_indication.flags
4553 & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
4554 ibmvnic_fw_err_cause(cause));
4555
4556 if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
4557 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4558 else
4559 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
4560 }
4561
4562 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
4563 struct ibmvnic_adapter *adapter)
4564 {
4565 struct net_device *netdev = adapter->netdev;
4566 struct device *dev = &adapter->vdev->dev;
4567 long rc;
4568
4569 rc = crq->change_mac_addr_rsp.rc.code;
4570 if (rc) {
4571 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
4572 goto out;
4573 }
4574 /* crq->change_mac_addr.mac_addr is the requested one
4575 * crq->change_mac_addr_rsp.mac_addr is the returned valid one.
4576 */
4577 ether_addr_copy(netdev->dev_addr,
4578 &crq->change_mac_addr_rsp.mac_addr[0]);
4579 ether_addr_copy(adapter->mac_addr,
4580 &crq->change_mac_addr_rsp.mac_addr[0]);
4581 out:
4582 complete(&adapter->fw_done);
4583 return rc;
4584 }
4585
4586 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
4587 struct ibmvnic_adapter *adapter)
4588 {
4589 struct device *dev = &adapter->vdev->dev;
4590 u64 *req_value;
4591 char *name;
4592
4593 atomic_dec(&adapter->running_cap_crqs);
4594 switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
4595 case REQ_TX_QUEUES:
4596 req_value = &adapter->req_tx_queues;
4597 name = "tx";
4598 break;
4599 case REQ_RX_QUEUES:
4600 req_value = &adapter->req_rx_queues;
4601 name = "rx";
4602 break;
4603 case REQ_RX_ADD_QUEUES:
4604 req_value = &adapter->req_rx_add_queues;
4605 name = "rx_add";
4606 break;
4607 case REQ_TX_ENTRIES_PER_SUBCRQ:
4608 req_value = &adapter->req_tx_entries_per_subcrq;
4609 name = "tx_entries_per_subcrq";
4610 break;
4611 case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
4612 req_value = &adapter->req_rx_add_entries_per_subcrq;
4613 name = "rx_add_entries_per_subcrq";
4614 break;
4615 case REQ_MTU:
4616 req_value = &adapter->req_mtu;
4617 name = "mtu";
4618 break;
4619 case PROMISC_REQUESTED:
4620 req_value = &adapter->promisc;
4621 name = "promisc";
4622 break;
4623 default:
4624 dev_err(dev, "Got invalid cap request rsp %d\n",
4625 crq->request_capability.capability);
4626 return;
4627 }
4628
4629 switch (crq->request_capability_rsp.rc.code) {
4630 case SUCCESS:
4631 break;
4632 case PARTIALSUCCESS:
4633 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
4634 *req_value,
4635 (long)be64_to_cpu(crq->request_capability_rsp.number),
4636 name);
4637
4638 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
4639 REQ_MTU) {
4640 pr_err("mtu of %llu is not supported. Reverting.\n",
4641 *req_value);
4642 *req_value = adapter->fallback.mtu;
4643 } else {
4644 *req_value =
4645 be64_to_cpu(crq->request_capability_rsp.number);
4646 }
4647
4648 send_request_cap(adapter, 1);
4649 return;
4650 default:
4651 dev_err(dev, "Error %d in request cap rsp\n",
4652 crq->request_capability_rsp.rc.code);
4653 return;
4654 }
4655
4656 /* Done receiving requested capabilities, query IP offload support */
4657 if (atomic_read(&adapter->running_cap_crqs) == 0) {
4658 adapter->wait_capability = false;
4659 send_query_ip_offload(adapter);
4660 }
4661 }
4662
4663 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4664 struct ibmvnic_adapter *adapter)
4665 {
4666 struct device *dev = &adapter->vdev->dev;
4667 struct net_device *netdev = adapter->netdev;
4668 struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4669 struct ibmvnic_login_buffer *login = adapter->login_buf;
4670 u64 *tx_handle_array;
4671 u64 *rx_handle_array;
4672 int num_tx_pools;
4673 int num_rx_pools;
4674 u64 *size_array;
4675 int i;
4676
4677 /* CHECK: Test/set of login_pending does not need to be atomic
4678 * because only ibmvnic_tasklet tests/clears this.
4679 */
4680 if (!adapter->login_pending) {
4681 netdev_warn(netdev, "Ignoring unexpected login response\n");
4682 return 0;
4683 }
4684 adapter->login_pending = false;
4685
4686 dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4687 DMA_TO_DEVICE);
4688 dma_unmap_single(dev, adapter->login_rsp_buf_token,
4689 adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4690
4691 /* If the number of queues requested can't be allocated by the
4692 * server, the login response will return with code 1. We will need
4693 * to resend the login buffer with fewer queues requested.
4694 */
4695 if (login_rsp_crq->generic.rc.code) {
4696 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4697 complete(&adapter->init_done);
4698 return 0;
4699 }
4700
4701 if (adapter->failover_pending) {
4702 adapter->init_done_rc = -EAGAIN;
4703 netdev_dbg(netdev, "Failover pending, ignoring login response\n");
4704 complete(&adapter->init_done);
4705 /* login response buffer will be released on reset */
4706 return 0;
4707 }
4708
4709 netdev->mtu = adapter->req_mtu - ETH_HLEN;
4710
4711 netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4712 for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4713 netdev_dbg(adapter->netdev, "%016lx\n",
4714 ((unsigned long *)(adapter->login_rsp_buf))[i]);
4715 }
4716
4717 /* Sanity checks */
4718 if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4719 (be32_to_cpu(login->num_rxcomp_subcrqs) *
4720 adapter->req_rx_add_queues !=
4721 be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4722 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4723 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4724 return -EIO;
4725 }
4726 size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4727 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
4728 /* variable buffer sizes are not supported, so just read the
4729 * first entry.
4730 */
4731 adapter->cur_rx_buf_sz = be64_to_cpu(size_array[0]);
4732
4733 num_tx_pools = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
4734 num_rx_pools = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4735
4736 tx_handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4737 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
4738 rx_handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4739 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_subcrqs));
4740
4741 for (i = 0; i < num_tx_pools; i++)
4742 adapter->tx_scrq[i]->handle = tx_handle_array[i];
4743
4744 for (i = 0; i < num_rx_pools; i++)
4745 adapter->rx_scrq[i]->handle = rx_handle_array[i];
4746
4747 adapter->num_active_tx_scrqs = num_tx_pools;
4748 adapter->num_active_rx_scrqs = num_rx_pools;
4749 release_login_rsp_buffer(adapter);
4750 release_login_buffer(adapter);
4751 complete(&adapter->init_done);
4752
4753 return 0;
4754 }
4755
4756 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4757 struct ibmvnic_adapter *adapter)
4758 {
4759 struct device *dev = &adapter->vdev->dev;
4760 long rc;
4761
4762 rc = crq->request_unmap_rsp.rc.code;
4763 if (rc)
4764 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4765 }
4766
4767 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4768 struct ibmvnic_adapter *adapter)
4769 {
4770 struct net_device *netdev = adapter->netdev;
4771 struct device *dev = &adapter->vdev->dev;
4772 long rc;
4773
4774 rc = crq->query_map_rsp.rc.code;
4775 if (rc) {
4776 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4777 return;
4778 }
4779 netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4780 crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4781 crq->query_map_rsp.free_pages);
4782 }
4783
4784 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4785 struct ibmvnic_adapter *adapter)
4786 {
4787 struct net_device *netdev = adapter->netdev;
4788 struct device *dev = &adapter->vdev->dev;
4789 long rc;
4790
4791 atomic_dec(&adapter->running_cap_crqs);
4792 netdev_dbg(netdev, "Outstanding queries: %d\n",
4793 atomic_read(&adapter->running_cap_crqs));
4794 rc = crq->query_capability.rc.code;
4795 if (rc) {
4796 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4797 goto out;
4798 }
4799
4800 switch (be16_to_cpu(crq->query_capability.capability)) {
4801 case MIN_TX_QUEUES:
4802 adapter->min_tx_queues =
4803 be64_to_cpu(crq->query_capability.number);
4804 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4805 adapter->min_tx_queues);
4806 break;
4807 case MIN_RX_QUEUES:
4808 adapter->min_rx_queues =
4809 be64_to_cpu(crq->query_capability.number);
4810 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4811 adapter->min_rx_queues);
4812 break;
4813 case MIN_RX_ADD_QUEUES:
4814 adapter->min_rx_add_queues =
4815 be64_to_cpu(crq->query_capability.number);
4816 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4817 adapter->min_rx_add_queues);
4818 break;
4819 case MAX_TX_QUEUES:
4820 adapter->max_tx_queues =
4821 be64_to_cpu(crq->query_capability.number);
4822 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4823 adapter->max_tx_queues);
4824 break;
4825 case MAX_RX_QUEUES:
4826 adapter->max_rx_queues =
4827 be64_to_cpu(crq->query_capability.number);
4828 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4829 adapter->max_rx_queues);
4830 break;
4831 case MAX_RX_ADD_QUEUES:
4832 adapter->max_rx_add_queues =
4833 be64_to_cpu(crq->query_capability.number);
4834 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4835 adapter->max_rx_add_queues);
4836 break;
4837 case MIN_TX_ENTRIES_PER_SUBCRQ:
4838 adapter->min_tx_entries_per_subcrq =
4839 be64_to_cpu(crq->query_capability.number);
4840 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4841 adapter->min_tx_entries_per_subcrq);
4842 break;
4843 case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4844 adapter->min_rx_add_entries_per_subcrq =
4845 be64_to_cpu(crq->query_capability.number);
4846 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4847 adapter->min_rx_add_entries_per_subcrq);
4848 break;
4849 case MAX_TX_ENTRIES_PER_SUBCRQ:
4850 adapter->max_tx_entries_per_subcrq =
4851 be64_to_cpu(crq->query_capability.number);
4852 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4853 adapter->max_tx_entries_per_subcrq);
4854 break;
4855 case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4856 adapter->max_rx_add_entries_per_subcrq =
4857 be64_to_cpu(crq->query_capability.number);
4858 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4859 adapter->max_rx_add_entries_per_subcrq);
4860 break;
4861 case TCP_IP_OFFLOAD:
4862 adapter->tcp_ip_offload =
4863 be64_to_cpu(crq->query_capability.number);
4864 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4865 adapter->tcp_ip_offload);
4866 break;
4867 case PROMISC_SUPPORTED:
4868 adapter->promisc_supported =
4869 be64_to_cpu(crq->query_capability.number);
4870 netdev_dbg(netdev, "promisc_supported = %lld\n",
4871 adapter->promisc_supported);
4872 break;
4873 case MIN_MTU:
4874 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4875 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4876 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4877 break;
4878 case MAX_MTU:
4879 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4880 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4881 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4882 break;
4883 case MAX_MULTICAST_FILTERS:
4884 adapter->max_multicast_filters =
4885 be64_to_cpu(crq->query_capability.number);
4886 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4887 adapter->max_multicast_filters);
4888 break;
4889 case VLAN_HEADER_INSERTION:
4890 adapter->vlan_header_insertion =
4891 be64_to_cpu(crq->query_capability.number);
4892 if (adapter->vlan_header_insertion)
4893 netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4894 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4895 adapter->vlan_header_insertion);
4896 break;
4897 case RX_VLAN_HEADER_INSERTION:
4898 adapter->rx_vlan_header_insertion =
4899 be64_to_cpu(crq->query_capability.number);
4900 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4901 adapter->rx_vlan_header_insertion);
4902 break;
4903 case MAX_TX_SG_ENTRIES:
4904 adapter->max_tx_sg_entries =
4905 be64_to_cpu(crq->query_capability.number);
4906 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4907 adapter->max_tx_sg_entries);
4908 break;
4909 case RX_SG_SUPPORTED:
4910 adapter->rx_sg_supported =
4911 be64_to_cpu(crq->query_capability.number);
4912 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4913 adapter->rx_sg_supported);
4914 break;
4915 case OPT_TX_COMP_SUB_QUEUES:
4916 adapter->opt_tx_comp_sub_queues =
4917 be64_to_cpu(crq->query_capability.number);
4918 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4919 adapter->opt_tx_comp_sub_queues);
4920 break;
4921 case OPT_RX_COMP_QUEUES:
4922 adapter->opt_rx_comp_queues =
4923 be64_to_cpu(crq->query_capability.number);
4924 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4925 adapter->opt_rx_comp_queues);
4926 break;
4927 case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4928 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4929 be64_to_cpu(crq->query_capability.number);
4930 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4931 adapter->opt_rx_bufadd_q_per_rx_comp_q);
4932 break;
4933 case OPT_TX_ENTRIES_PER_SUBCRQ:
4934 adapter->opt_tx_entries_per_subcrq =
4935 be64_to_cpu(crq->query_capability.number);
4936 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4937 adapter->opt_tx_entries_per_subcrq);
4938 break;
4939 case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4940 adapter->opt_rxba_entries_per_subcrq =
4941 be64_to_cpu(crq->query_capability.number);
4942 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4943 adapter->opt_rxba_entries_per_subcrq);
4944 break;
4945 case TX_RX_DESC_REQ:
4946 adapter->tx_rx_desc_req = crq->query_capability.number;
4947 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4948 adapter->tx_rx_desc_req);
4949 break;
4950
4951 default:
4952 netdev_err(netdev, "Got invalid cap rsp %d\n",
4953 crq->query_capability.capability);
4954 }
4955
4956 out:
4957 if (atomic_read(&adapter->running_cap_crqs) == 0) {
4958 adapter->wait_capability = false;
4959 send_request_cap(adapter, 0);
4960 }
4961 }
4962
4963 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4964 {
4965 union ibmvnic_crq crq;
4966 int rc;
4967
4968 memset(&crq, 0, sizeof(crq));
4969 crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4970 crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4971
4972 mutex_lock(&adapter->fw_lock);
4973 adapter->fw_done_rc = 0;
4974 reinit_completion(&adapter->fw_done);
4975
4976 rc = ibmvnic_send_crq(adapter, &crq);
4977 if (rc) {
4978 mutex_unlock(&adapter->fw_lock);
4979 return rc;
4980 }
4981
4982 rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
4983 if (rc) {
4984 mutex_unlock(&adapter->fw_lock);
4985 return rc;
4986 }
4987
4988 mutex_unlock(&adapter->fw_lock);
4989 return adapter->fw_done_rc ? -EIO : 0;
4990 }
4991
4992 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4993 struct ibmvnic_adapter *adapter)
4994 {
4995 struct net_device *netdev = adapter->netdev;
4996 int rc;
4997 __be32 rspeed = cpu_to_be32(crq->query_phys_parms_rsp.speed);
4998
4999 rc = crq->query_phys_parms_rsp.rc.code;
5000 if (rc) {
5001 netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
5002 return rc;
5003 }
5004 switch (rspeed) {
5005 case IBMVNIC_10MBPS:
5006 adapter->speed = SPEED_10;
5007 break;
5008 case IBMVNIC_100MBPS:
5009 adapter->speed = SPEED_100;
5010 break;
5011 case IBMVNIC_1GBPS:
5012 adapter->speed = SPEED_1000;
5013 break;
5014 case IBMVNIC_10GBPS:
5015 adapter->speed = SPEED_10000;
5016 break;
5017 case IBMVNIC_25GBPS:
5018 adapter->speed = SPEED_25000;
5019 break;
5020 case IBMVNIC_40GBPS:
5021 adapter->speed = SPEED_40000;
5022 break;
5023 case IBMVNIC_50GBPS:
5024 adapter->speed = SPEED_50000;
5025 break;
5026 case IBMVNIC_100GBPS:
5027 adapter->speed = SPEED_100000;
5028 break;
5029 case IBMVNIC_200GBPS:
5030 adapter->speed = SPEED_200000;
5031 break;
5032 default:
5033 if (netif_carrier_ok(netdev))
5034 netdev_warn(netdev, "Unknown speed 0x%08x\n", rspeed);
5035 adapter->speed = SPEED_UNKNOWN;
5036 }
5037 if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
5038 adapter->duplex = DUPLEX_FULL;
5039 else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
5040 adapter->duplex = DUPLEX_HALF;
5041 else
5042 adapter->duplex = DUPLEX_UNKNOWN;
5043
5044 return rc;
5045 }
5046
5047 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
5048 struct ibmvnic_adapter *adapter)
5049 {
5050 struct ibmvnic_generic_crq *gen_crq = &crq->generic;
5051 struct net_device *netdev = adapter->netdev;
5052 struct device *dev = &adapter->vdev->dev;
5053 u64 *u64_crq = (u64 *)crq;
5054 long rc;
5055
5056 netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
5057 (unsigned long)cpu_to_be64(u64_crq[0]),
5058 (unsigned long)cpu_to_be64(u64_crq[1]));
5059 switch (gen_crq->first) {
5060 case IBMVNIC_CRQ_INIT_RSP:
5061 switch (gen_crq->cmd) {
5062 case IBMVNIC_CRQ_INIT:
5063 dev_info(dev, "Partner initialized\n");
5064 adapter->from_passive_init = true;
5065 /* Discard any stale login responses from prev reset.
5066 * CHECK: should we clear even on INIT_COMPLETE?
5067 */
5068 adapter->login_pending = false;
5069
5070 if (adapter->state == VNIC_DOWN)
5071 rc = ibmvnic_reset(adapter, VNIC_RESET_PASSIVE_INIT);
5072 else
5073 rc = ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
5074
5075 if (rc && rc != -EBUSY) {
5076 /* We were unable to schedule the failover
5077 * reset either because the adapter was still
5078 * probing (eg: during kexec) or we could not
5079 * allocate memory. Clear the failover_pending
5080 * flag since no one else will. We ignore
5081 * EBUSY because it means either FAILOVER reset
5082 * is already scheduled or the adapter is
5083 * being removed.
5084 */
5085 netdev_err(netdev,
5086 "Error %ld scheduling failover reset\n",
5087 rc);
5088 adapter->failover_pending = false;
5089 }
5090
5091 if (!completion_done(&adapter->init_done)) {
5092 complete(&adapter->init_done);
5093 if (!adapter->init_done_rc)
5094 adapter->init_done_rc = -EAGAIN;
5095 }
5096
5097 break;
5098 case IBMVNIC_CRQ_INIT_COMPLETE:
5099 dev_info(dev, "Partner initialization complete\n");
5100 adapter->crq.active = true;
5101 send_version_xchg(adapter);
5102 break;
5103 default:
5104 dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
5105 }
5106 return;
5107 case IBMVNIC_CRQ_XPORT_EVENT:
5108 netif_carrier_off(netdev);
5109 adapter->crq.active = false;
5110 /* terminate any thread waiting for a response
5111 * from the device
5112 */
5113 if (!completion_done(&adapter->fw_done)) {
5114 adapter->fw_done_rc = -EIO;
5115 complete(&adapter->fw_done);
5116 }
5117 if (!completion_done(&adapter->stats_done))
5118 complete(&adapter->stats_done);
5119 if (test_bit(0, &adapter->resetting))
5120 adapter->force_reset_recovery = true;
5121 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
5122 dev_info(dev, "Migrated, re-enabling adapter\n");
5123 ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
5124 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
5125 dev_info(dev, "Backing device failover detected\n");
5126 adapter->failover_pending = true;
5127 } else {
5128 /* The adapter lost the connection */
5129 dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
5130 gen_crq->cmd);
5131 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
5132 }
5133 return;
5134 case IBMVNIC_CRQ_CMD_RSP:
5135 break;
5136 default:
5137 dev_err(dev, "Got an invalid msg type 0x%02x\n",
5138 gen_crq->first);
5139 return;
5140 }
5141
5142 switch (gen_crq->cmd) {
5143 case VERSION_EXCHANGE_RSP:
5144 rc = crq->version_exchange_rsp.rc.code;
5145 if (rc) {
5146 dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
5147 break;
5148 }
5149 ibmvnic_version =
5150 be16_to_cpu(crq->version_exchange_rsp.version);
5151 dev_info(dev, "Partner protocol version is %d\n",
5152 ibmvnic_version);
5153 send_query_cap(adapter);
5154 break;
5155 case QUERY_CAPABILITY_RSP:
5156 handle_query_cap_rsp(crq, adapter);
5157 break;
5158 case QUERY_MAP_RSP:
5159 handle_query_map_rsp(crq, adapter);
5160 break;
5161 case REQUEST_MAP_RSP:
5162 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
5163 complete(&adapter->fw_done);
5164 break;
5165 case REQUEST_UNMAP_RSP:
5166 handle_request_unmap_rsp(crq, adapter);
5167 break;
5168 case REQUEST_CAPABILITY_RSP:
5169 handle_request_cap_rsp(crq, adapter);
5170 break;
5171 case LOGIN_RSP:
5172 netdev_dbg(netdev, "Got Login Response\n");
5173 handle_login_rsp(crq, adapter);
5174 break;
5175 case LOGICAL_LINK_STATE_RSP:
5176 netdev_dbg(netdev,
5177 "Got Logical Link State Response, state: %d rc: %d\n",
5178 crq->logical_link_state_rsp.link_state,
5179 crq->logical_link_state_rsp.rc.code);
5180 adapter->logical_link_state =
5181 crq->logical_link_state_rsp.link_state;
5182 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
5183 complete(&adapter->init_done);
5184 break;
5185 case LINK_STATE_INDICATION:
5186 netdev_dbg(netdev, "Got Logical Link State Indication\n");
5187 adapter->phys_link_state =
5188 crq->link_state_indication.phys_link_state;
5189 adapter->logical_link_state =
5190 crq->link_state_indication.logical_link_state;
5191 if (adapter->phys_link_state && adapter->logical_link_state)
5192 netif_carrier_on(netdev);
5193 else
5194 netif_carrier_off(netdev);
5195 break;
5196 case CHANGE_MAC_ADDR_RSP:
5197 netdev_dbg(netdev, "Got MAC address change Response\n");
5198 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
5199 break;
5200 case ERROR_INDICATION:
5201 netdev_dbg(netdev, "Got Error Indication\n");
5202 handle_error_indication(crq, adapter);
5203 break;
5204 case REQUEST_STATISTICS_RSP:
5205 netdev_dbg(netdev, "Got Statistics Response\n");
5206 complete(&adapter->stats_done);
5207 break;
5208 case QUERY_IP_OFFLOAD_RSP:
5209 netdev_dbg(netdev, "Got Query IP offload Response\n");
5210 handle_query_ip_offload_rsp(adapter);
5211 break;
5212 case MULTICAST_CTRL_RSP:
5213 netdev_dbg(netdev, "Got multicast control Response\n");
5214 break;
5215 case CONTROL_IP_OFFLOAD_RSP:
5216 netdev_dbg(netdev, "Got Control IP offload Response\n");
5217 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
5218 sizeof(adapter->ip_offload_ctrl),
5219 DMA_TO_DEVICE);
5220 complete(&adapter->init_done);
5221 break;
5222 case COLLECT_FW_TRACE_RSP:
5223 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
5224 complete(&adapter->fw_done);
5225 break;
5226 case GET_VPD_SIZE_RSP:
5227 handle_vpd_size_rsp(crq, adapter);
5228 break;
5229 case GET_VPD_RSP:
5230 handle_vpd_rsp(crq, adapter);
5231 break;
5232 case QUERY_PHYS_PARMS_RSP:
5233 adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
5234 complete(&adapter->fw_done);
5235 break;
5236 default:
5237 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
5238 gen_crq->cmd);
5239 }
5240 }
5241
5242 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
5243 {
5244 struct ibmvnic_adapter *adapter = instance;
5245
5246 tasklet_schedule(&adapter->tasklet);
5247 return IRQ_HANDLED;
5248 }
5249
5250 static void ibmvnic_tasklet(struct tasklet_struct *t)
5251 {
5252 struct ibmvnic_adapter *adapter = from_tasklet(adapter, t, tasklet);
5253 struct ibmvnic_crq_queue *queue = &adapter->crq;
5254 union ibmvnic_crq *crq;
5255 unsigned long flags;
5256 bool done = false;
5257
5258 spin_lock_irqsave(&queue->lock, flags);
5259 while (!done) {
5260 /* Pull all the valid messages off the CRQ */
5261 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
5262 /* This barrier makes sure ibmvnic_next_crq()'s
5263 * crq->generic.first & IBMVNIC_CRQ_CMD_RSP is loaded
5264 * before ibmvnic_handle_crq()'s
5265 * switch(gen_crq->first) and switch(gen_crq->cmd).
5266 */
5267 dma_rmb();
5268 ibmvnic_handle_crq(crq, adapter);
5269 crq->generic.first = 0;
5270 }
5271
5272 /* remain in tasklet until all
5273 * capabilities responses are received
5274 */
5275 if (!adapter->wait_capability)
5276 done = true;
5277 }
5278 /* if capabilities CRQ's were sent in this tasklet, the following
5279 * tasklet must wait until all responses are received
5280 */
5281 if (atomic_read(&adapter->running_cap_crqs) != 0)
5282 adapter->wait_capability = true;
5283 spin_unlock_irqrestore(&queue->lock, flags);
5284 }
5285
5286 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
5287 {
5288 struct vio_dev *vdev = adapter->vdev;
5289 int rc;
5290
5291 do {
5292 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
5293 } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
5294
5295 if (rc)
5296 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
5297
5298 return rc;
5299 }
5300
5301 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
5302 {
5303 struct ibmvnic_crq_queue *crq = &adapter->crq;
5304 struct device *dev = &adapter->vdev->dev;
5305 struct vio_dev *vdev = adapter->vdev;
5306 int rc;
5307
5308 /* Close the CRQ */
5309 do {
5310 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
5311 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
5312
5313 /* Clean out the queue */
5314 if (!crq->msgs)
5315 return -EINVAL;
5316
5317 memset(crq->msgs, 0, PAGE_SIZE);
5318 crq->cur = 0;
5319 crq->active = false;
5320
5321 /* And re-open it again */
5322 rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
5323 crq->msg_token, PAGE_SIZE);
5324
5325 if (rc == H_CLOSED)
5326 /* Adapter is good, but other end is not ready */
5327 dev_warn(dev, "Partner adapter not ready\n");
5328 else if (rc != 0)
5329 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
5330
5331 return rc;
5332 }
5333
5334 static void release_crq_queue(struct ibmvnic_adapter *adapter)
5335 {
5336 struct ibmvnic_crq_queue *crq = &adapter->crq;
5337 struct vio_dev *vdev = adapter->vdev;
5338 long rc;
5339
5340 if (!crq->msgs)
5341 return;
5342
5343 netdev_dbg(adapter->netdev, "Releasing CRQ\n");
5344 free_irq(vdev->irq, adapter);
5345 tasklet_kill(&adapter->tasklet);
5346 do {
5347 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
5348 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
5349
5350 dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
5351 DMA_BIDIRECTIONAL);
5352 free_page((unsigned long)crq->msgs);
5353 crq->msgs = NULL;
5354 crq->active = false;
5355 }
5356
5357 static int init_crq_queue(struct ibmvnic_adapter *adapter)
5358 {
5359 struct ibmvnic_crq_queue *crq = &adapter->crq;
5360 struct device *dev = &adapter->vdev->dev;
5361 struct vio_dev *vdev = adapter->vdev;
5362 int rc, retrc = -ENOMEM;
5363
5364 if (crq->msgs)
5365 return 0;
5366
5367 crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
5368 /* Should we allocate more than one page? */
5369
5370 if (!crq->msgs)
5371 return -ENOMEM;
5372
5373 crq->size = PAGE_SIZE / sizeof(*crq->msgs);
5374 crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
5375 DMA_BIDIRECTIONAL);
5376 if (dma_mapping_error(dev, crq->msg_token))
5377 goto map_failed;
5378
5379 rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
5380 crq->msg_token, PAGE_SIZE);
5381
5382 if (rc == H_RESOURCE)
5383 /* maybe kexecing and resource is busy. try a reset */
5384 rc = ibmvnic_reset_crq(adapter);
5385 retrc = rc;
5386
5387 if (rc == H_CLOSED) {
5388 dev_warn(dev, "Partner adapter not ready\n");
5389 } else if (rc) {
5390 dev_warn(dev, "Error %d opening adapter\n", rc);
5391 goto reg_crq_failed;
5392 }
5393
5394 retrc = 0;
5395
5396 tasklet_setup(&adapter->tasklet, (void *)ibmvnic_tasklet);
5397
5398 netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
5399 snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
5400 adapter->vdev->unit_address);
5401 rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
5402 if (rc) {
5403 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
5404 vdev->irq, rc);
5405 goto req_irq_failed;
5406 }
5407
5408 rc = vio_enable_interrupts(vdev);
5409 if (rc) {
5410 dev_err(dev, "Error %d enabling interrupts\n", rc);
5411 goto req_irq_failed;
5412 }
5413
5414 crq->cur = 0;
5415 spin_lock_init(&crq->lock);
5416
5417 /* process any CRQs that were queued before we enabled interrupts */
5418 tasklet_schedule(&adapter->tasklet);
5419
5420 return retrc;
5421
5422 req_irq_failed:
5423 tasklet_kill(&adapter->tasklet);
5424 do {
5425 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
5426 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
5427 reg_crq_failed:
5428 dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
5429 map_failed:
5430 free_page((unsigned long)crq->msgs);
5431 crq->msgs = NULL;
5432 return retrc;
5433 }
5434
5435 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter, bool reset)
5436 {
5437 struct device *dev = &adapter->vdev->dev;
5438 unsigned long timeout = msecs_to_jiffies(20000);
5439 u64 old_num_rx_queues = adapter->req_rx_queues;
5440 u64 old_num_tx_queues = adapter->req_tx_queues;
5441 int rc;
5442
5443 adapter->from_passive_init = false;
5444
5445 if (reset)
5446 reinit_completion(&adapter->init_done);
5447
5448 adapter->init_done_rc = 0;
5449 rc = ibmvnic_send_crq_init(adapter);
5450 if (rc) {
5451 dev_err(dev, "Send crq init failed with error %d\n", rc);
5452 return rc;
5453 }
5454
5455 if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
5456 dev_err(dev, "Initialization sequence timed out\n");
5457 return -1;
5458 }
5459
5460 if (adapter->init_done_rc) {
5461 release_crq_queue(adapter);
5462 return adapter->init_done_rc;
5463 }
5464
5465 if (adapter->from_passive_init) {
5466 adapter->state = VNIC_OPEN;
5467 adapter->from_passive_init = false;
5468 return -1;
5469 }
5470
5471 if (reset &&
5472 test_bit(0, &adapter->resetting) && !adapter->wait_for_reset &&
5473 adapter->reset_reason != VNIC_RESET_MOBILITY) {
5474 if (adapter->req_rx_queues != old_num_rx_queues ||
5475 adapter->req_tx_queues != old_num_tx_queues) {
5476 release_sub_crqs(adapter, 0);
5477 rc = init_sub_crqs(adapter);
5478 } else {
5479 rc = reset_sub_crq_queues(adapter);
5480 }
5481 } else {
5482 rc = init_sub_crqs(adapter);
5483 }
5484
5485 if (rc) {
5486 dev_err(dev, "Initialization of sub crqs failed\n");
5487 release_crq_queue(adapter);
5488 return rc;
5489 }
5490
5491 rc = init_sub_crq_irqs(adapter);
5492 if (rc) {
5493 dev_err(dev, "Failed to initialize sub crq irqs\n");
5494 release_crq_queue(adapter);
5495 }
5496
5497 return rc;
5498 }
5499
5500 static struct device_attribute dev_attr_failover;
5501
5502 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
5503 {
5504 struct ibmvnic_adapter *adapter;
5505 struct net_device *netdev;
5506 unsigned char *mac_addr_p;
5507 bool init_success;
5508 int rc;
5509
5510 dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
5511 dev->unit_address);
5512
5513 mac_addr_p = (unsigned char *)vio_get_attribute(dev,
5514 VETH_MAC_ADDR, NULL);
5515 if (!mac_addr_p) {
5516 dev_err(&dev->dev,
5517 "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
5518 __FILE__, __LINE__);
5519 return 0;
5520 }
5521
5522 netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
5523 IBMVNIC_MAX_QUEUES);
5524 if (!netdev)
5525 return -ENOMEM;
5526
5527 adapter = netdev_priv(netdev);
5528 adapter->state = VNIC_PROBING;
5529 dev_set_drvdata(&dev->dev, netdev);
5530 adapter->vdev = dev;
5531 adapter->netdev = netdev;
5532 adapter->login_pending = false;
5533
5534 ether_addr_copy(adapter->mac_addr, mac_addr_p);
5535 ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
5536 netdev->irq = dev->irq;
5537 netdev->netdev_ops = &ibmvnic_netdev_ops;
5538 netdev->ethtool_ops = &ibmvnic_ethtool_ops;
5539 SET_NETDEV_DEV(netdev, &dev->dev);
5540
5541 INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
5542 INIT_DELAYED_WORK(&adapter->ibmvnic_delayed_reset,
5543 __ibmvnic_delayed_reset);
5544 INIT_LIST_HEAD(&adapter->rwi_list);
5545 spin_lock_init(&adapter->rwi_lock);
5546 spin_lock_init(&adapter->state_lock);
5547 mutex_init(&adapter->fw_lock);
5548 init_completion(&adapter->init_done);
5549 init_completion(&adapter->fw_done);
5550 init_completion(&adapter->reset_done);
5551 init_completion(&adapter->stats_done);
5552 clear_bit(0, &adapter->resetting);
5553
5554 init_success = false;
5555 do {
5556 rc = init_crq_queue(adapter);
5557 if (rc) {
5558 dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
5559 rc);
5560 goto ibmvnic_init_fail;
5561 }
5562
5563 rc = ibmvnic_reset_init(adapter, false);
5564 } while (rc == -EAGAIN);
5565
5566 /* We are ignoring the error from ibmvnic_reset_init() assuming that the
5567 * partner is not ready. CRQ is not active. When the partner becomes
5568 * ready, we will do the passive init reset.
5569 */
5570
5571 if (!rc)
5572 init_success = true;
5573
5574 rc = init_stats_buffers(adapter);
5575 if (rc)
5576 goto ibmvnic_init_fail;
5577
5578 rc = init_stats_token(adapter);
5579 if (rc)
5580 goto ibmvnic_stats_fail;
5581
5582 rc = device_create_file(&dev->dev, &dev_attr_failover);
5583 if (rc)
5584 goto ibmvnic_dev_file_err;
5585
5586 netif_carrier_off(netdev);
5587 rc = register_netdev(netdev);
5588 if (rc) {
5589 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
5590 goto ibmvnic_register_fail;
5591 }
5592 dev_info(&dev->dev, "ibmvnic registered\n");
5593
5594 if (init_success) {
5595 adapter->state = VNIC_PROBED;
5596 netdev->mtu = adapter->req_mtu - ETH_HLEN;
5597 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
5598 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
5599 } else {
5600 adapter->state = VNIC_DOWN;
5601 }
5602
5603 adapter->wait_for_reset = false;
5604 adapter->last_reset_time = jiffies;
5605 return 0;
5606
5607 ibmvnic_register_fail:
5608 device_remove_file(&dev->dev, &dev_attr_failover);
5609
5610 ibmvnic_dev_file_err:
5611 release_stats_token(adapter);
5612
5613 ibmvnic_stats_fail:
5614 release_stats_buffers(adapter);
5615
5616 ibmvnic_init_fail:
5617 release_sub_crqs(adapter, 1);
5618 release_crq_queue(adapter);
5619 mutex_destroy(&adapter->fw_lock);
5620 free_netdev(netdev);
5621
5622 return rc;
5623 }
5624
5625 static void ibmvnic_remove(struct vio_dev *dev)
5626 {
5627 struct net_device *netdev = dev_get_drvdata(&dev->dev);
5628 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5629 unsigned long flags;
5630
5631 spin_lock_irqsave(&adapter->state_lock, flags);
5632
5633 /* If ibmvnic_reset() is scheduling a reset, wait for it to
5634 * finish. Then, set the state to REMOVING to prevent it from
5635 * scheduling any more work and to have reset functions ignore
5636 * any resets that have already been scheduled. Drop the lock
5637 * after setting state, so __ibmvnic_reset() which is called
5638 * from the flush_work() below, can make progress.
5639 */
5640 spin_lock(&adapter->rwi_lock);
5641 adapter->state = VNIC_REMOVING;
5642 spin_unlock(&adapter->rwi_lock);
5643
5644 spin_unlock_irqrestore(&adapter->state_lock, flags);
5645
5646 flush_work(&adapter->ibmvnic_reset);
5647 flush_delayed_work(&adapter->ibmvnic_delayed_reset);
5648
5649 rtnl_lock();
5650 unregister_netdevice(netdev);
5651
5652 release_resources(adapter);
5653 release_sub_crqs(adapter, 1);
5654 release_crq_queue(adapter);
5655
5656 release_stats_token(adapter);
5657 release_stats_buffers(adapter);
5658
5659 adapter->state = VNIC_REMOVED;
5660
5661 rtnl_unlock();
5662 mutex_destroy(&adapter->fw_lock);
5663 device_remove_file(&dev->dev, &dev_attr_failover);
5664 free_netdev(netdev);
5665 dev_set_drvdata(&dev->dev, NULL);
5666 }
5667
5668 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
5669 const char *buf, size_t count)
5670 {
5671 struct net_device *netdev = dev_get_drvdata(dev);
5672 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5673 unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
5674 __be64 session_token;
5675 long rc;
5676
5677 if (!sysfs_streq(buf, "1"))
5678 return -EINVAL;
5679
5680 rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
5681 H_GET_SESSION_TOKEN, 0, 0, 0);
5682 if (rc) {
5683 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
5684 rc);
5685 goto last_resort;
5686 }
5687
5688 session_token = (__be64)retbuf[0];
5689 netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
5690 be64_to_cpu(session_token));
5691 rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
5692 H_SESSION_ERR_DETECTED, session_token, 0, 0);
5693 if (rc)
5694 netdev_err(netdev,
5695 "H_VIOCTL initiated failover failed, rc %ld\n",
5696 rc);
5697
5698 last_resort:
5699 netdev_dbg(netdev, "Trying to send CRQ_CMD, the last resort\n");
5700 ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
5701
5702 return count;
5703 }
5704 static DEVICE_ATTR_WO(failover);
5705
5706 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
5707 {
5708 struct net_device *netdev = dev_get_drvdata(&vdev->dev);
5709 struct ibmvnic_adapter *adapter;
5710 struct iommu_table *tbl;
5711 unsigned long ret = 0;
5712 int i;
5713
5714 tbl = get_iommu_table_base(&vdev->dev);
5715
5716 /* netdev inits at probe time along with the structures we need below*/
5717 if (!netdev)
5718 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
5719
5720 adapter = netdev_priv(netdev);
5721
5722 ret += PAGE_SIZE; /* the crq message queue */
5723 ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
5724
5725 for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
5726 ret += 4 * PAGE_SIZE; /* the scrq message queue */
5727
5728 for (i = 0; i < adapter->num_active_rx_pools; i++)
5729 ret += adapter->rx_pool[i].size *
5730 IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
5731
5732 return ret;
5733 }
5734
5735 static int ibmvnic_resume(struct device *dev)
5736 {
5737 struct net_device *netdev = dev_get_drvdata(dev);
5738 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5739
5740 if (adapter->state != VNIC_OPEN)
5741 return 0;
5742
5743 tasklet_schedule(&adapter->tasklet);
5744
5745 return 0;
5746 }
5747
5748 static const struct vio_device_id ibmvnic_device_table[] = {
5749 {"network", "IBM,vnic"},
5750 {"", "" }
5751 };
5752 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5753
5754 static const struct dev_pm_ops ibmvnic_pm_ops = {
5755 .resume = ibmvnic_resume
5756 };
5757
5758 static struct vio_driver ibmvnic_driver = {
5759 .id_table = ibmvnic_device_table,
5760 .probe = ibmvnic_probe,
5761 .remove = ibmvnic_remove,
5762 .get_desired_dma = ibmvnic_get_desired_dma,
5763 .name = ibmvnic_driver_name,
5764 .pm = &ibmvnic_pm_ops,
5765 };
5766
5767 /* module functions */
5768 static int __init ibmvnic_module_init(void)
5769 {
5770 pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5771 IBMVNIC_DRIVER_VERSION);
5772
5773 return vio_register_driver(&ibmvnic_driver);
5774 }
5775
5776 static void __exit ibmvnic_module_exit(void)
5777 {
5778 vio_unregister_driver(&ibmvnic_driver);
5779 }
5780
5781 module_init(ibmvnic_module_init);
5782 module_exit(ibmvnic_module_exit);
Ubuntu Jammy Linux kernel mirror