treewide: remove redundant IS_ERR() before error code check
[linux/fpc-iii.git] / drivers / net / ethernet / ibm / ibmvnic.c
blobc75239d8820f9519089d2caa74fcea6ab2c02cc1
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 /**************************************************************************/
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>
70 #include "ibmvnic.h"
72 static const char ibmvnic_driver_name[] = "ibmvnic";
73 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
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);
80 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
81 static int ibmvnic_remove(struct vio_dev *);
82 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
83 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
84 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
85 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
86 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
87 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
88 union sub_crq *sub_crq);
89 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
90 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
91 static int enable_scrq_irq(struct ibmvnic_adapter *,
92 struct ibmvnic_sub_crq_queue *);
93 static int disable_scrq_irq(struct ibmvnic_adapter *,
94 struct ibmvnic_sub_crq_queue *);
95 static int pending_scrq(struct ibmvnic_adapter *,
96 struct ibmvnic_sub_crq_queue *);
97 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
98 struct ibmvnic_sub_crq_queue *);
99 static int ibmvnic_poll(struct napi_struct *napi, int data);
100 static void send_map_query(struct ibmvnic_adapter *adapter);
101 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
102 static int send_request_unmap(struct ibmvnic_adapter *, u8);
103 static int send_login(struct ibmvnic_adapter *adapter);
104 static void send_cap_queries(struct ibmvnic_adapter *adapter);
105 static int init_sub_crqs(struct ibmvnic_adapter *);
106 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
107 static int ibmvnic_init(struct ibmvnic_adapter *);
108 static int ibmvnic_reset_init(struct ibmvnic_adapter *);
109 static void release_crq_queue(struct ibmvnic_adapter *);
110 static int __ibmvnic_set_mac(struct net_device *, u8 *);
111 static int init_crq_queue(struct ibmvnic_adapter *adapter);
112 static int send_query_phys_parms(struct ibmvnic_adapter *adapter);
114 struct ibmvnic_stat {
115 char name[ETH_GSTRING_LEN];
116 int offset;
119 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
120 offsetof(struct ibmvnic_statistics, stat))
121 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
123 static const struct ibmvnic_stat ibmvnic_stats[] = {
124 {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
125 {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
126 {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
127 {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
128 {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
129 {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
130 {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
131 {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
132 {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
133 {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
134 {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
135 {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
136 {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
137 {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
138 {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
139 {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
140 {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
141 {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
142 {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
143 {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
144 {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
145 {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
148 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
149 unsigned long length, unsigned long *number,
150 unsigned long *irq)
152 unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
153 long rc;
155 rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
156 *number = retbuf[0];
157 *irq = retbuf[1];
159 return rc;
163 * ibmvnic_wait_for_completion - Check device state and wait for completion
164 * @adapter: private device data
165 * @comp_done: completion structure to wait for
166 * @timeout: time to wait in milliseconds
168 * Wait for a completion signal or until the timeout limit is reached
169 * while checking that the device is still active.
171 static int ibmvnic_wait_for_completion(struct ibmvnic_adapter *adapter,
172 struct completion *comp_done,
173 unsigned long timeout)
175 struct net_device *netdev;
176 unsigned long div_timeout;
177 u8 retry;
179 netdev = adapter->netdev;
180 retry = 5;
181 div_timeout = msecs_to_jiffies(timeout / retry);
182 while (true) {
183 if (!adapter->crq.active) {
184 netdev_err(netdev, "Device down!\n");
185 return -ENODEV;
187 if (!retry--)
188 break;
189 if (wait_for_completion_timeout(comp_done, div_timeout))
190 return 0;
192 netdev_err(netdev, "Operation timed out.\n");
193 return -ETIMEDOUT;
196 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
197 struct ibmvnic_long_term_buff *ltb, int size)
199 struct device *dev = &adapter->vdev->dev;
200 int rc;
202 ltb->size = size;
203 ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
204 GFP_KERNEL);
206 if (!ltb->buff) {
207 dev_err(dev, "Couldn't alloc long term buffer\n");
208 return -ENOMEM;
210 ltb->map_id = adapter->map_id;
211 adapter->map_id++;
213 mutex_lock(&adapter->fw_lock);
214 adapter->fw_done_rc = 0;
215 reinit_completion(&adapter->fw_done);
216 rc = send_request_map(adapter, ltb->addr,
217 ltb->size, ltb->map_id);
218 if (rc) {
219 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
220 mutex_unlock(&adapter->fw_lock);
221 return rc;
224 rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
225 if (rc) {
226 dev_err(dev,
227 "Long term map request aborted or timed out,rc = %d\n",
228 rc);
229 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
230 mutex_unlock(&adapter->fw_lock);
231 return rc;
234 if (adapter->fw_done_rc) {
235 dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
236 adapter->fw_done_rc);
237 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
238 mutex_unlock(&adapter->fw_lock);
239 return -1;
241 mutex_unlock(&adapter->fw_lock);
242 return 0;
245 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
246 struct ibmvnic_long_term_buff *ltb)
248 struct device *dev = &adapter->vdev->dev;
250 if (!ltb->buff)
251 return;
253 if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
254 adapter->reset_reason != VNIC_RESET_MOBILITY)
255 send_request_unmap(adapter, ltb->map_id);
256 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
259 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
260 struct ibmvnic_long_term_buff *ltb)
262 struct device *dev = &adapter->vdev->dev;
263 int rc;
265 memset(ltb->buff, 0, ltb->size);
267 mutex_lock(&adapter->fw_lock);
268 adapter->fw_done_rc = 0;
270 reinit_completion(&adapter->fw_done);
271 rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
272 if (rc) {
273 mutex_unlock(&adapter->fw_lock);
274 return rc;
277 rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
278 if (rc) {
279 dev_info(dev,
280 "Reset failed, long term map request timed out or aborted\n");
281 mutex_unlock(&adapter->fw_lock);
282 return rc;
285 if (adapter->fw_done_rc) {
286 dev_info(dev,
287 "Reset failed, attempting to free and reallocate buffer\n");
288 free_long_term_buff(adapter, ltb);
289 mutex_unlock(&adapter->fw_lock);
290 return alloc_long_term_buff(adapter, ltb, ltb->size);
292 mutex_unlock(&adapter->fw_lock);
293 return 0;
296 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
298 int i;
300 for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
301 i++)
302 adapter->rx_pool[i].active = 0;
305 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
306 struct ibmvnic_rx_pool *pool)
308 int count = pool->size - atomic_read(&pool->available);
309 struct device *dev = &adapter->vdev->dev;
310 int buffers_added = 0;
311 unsigned long lpar_rc;
312 union sub_crq sub_crq;
313 struct sk_buff *skb;
314 unsigned int offset;
315 dma_addr_t dma_addr;
316 unsigned char *dst;
317 u64 *handle_array;
318 int shift = 0;
319 int index;
320 int i;
322 if (!pool->active)
323 return;
325 handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
326 be32_to_cpu(adapter->login_rsp_buf->
327 off_rxadd_subcrqs));
329 for (i = 0; i < count; ++i) {
330 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
331 if (!skb) {
332 dev_err(dev, "Couldn't replenish rx buff\n");
333 adapter->replenish_no_mem++;
334 break;
337 index = pool->free_map[pool->next_free];
339 if (pool->rx_buff[index].skb)
340 dev_err(dev, "Inconsistent free_map!\n");
342 /* Copy the skb to the long term mapped DMA buffer */
343 offset = index * pool->buff_size;
344 dst = pool->long_term_buff.buff + offset;
345 memset(dst, 0, pool->buff_size);
346 dma_addr = pool->long_term_buff.addr + offset;
347 pool->rx_buff[index].data = dst;
349 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
350 pool->rx_buff[index].dma = dma_addr;
351 pool->rx_buff[index].skb = skb;
352 pool->rx_buff[index].pool_index = pool->index;
353 pool->rx_buff[index].size = pool->buff_size;
355 memset(&sub_crq, 0, sizeof(sub_crq));
356 sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
357 sub_crq.rx_add.correlator =
358 cpu_to_be64((u64)&pool->rx_buff[index]);
359 sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
360 sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
362 /* The length field of the sCRQ is defined to be 24 bits so the
363 * buffer size needs to be left shifted by a byte before it is
364 * converted to big endian to prevent the last byte from being
365 * truncated.
367 #ifdef __LITTLE_ENDIAN__
368 shift = 8;
369 #endif
370 sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
372 lpar_rc = send_subcrq(adapter, handle_array[pool->index],
373 &sub_crq);
374 if (lpar_rc != H_SUCCESS)
375 goto failure;
377 buffers_added++;
378 adapter->replenish_add_buff_success++;
379 pool->next_free = (pool->next_free + 1) % pool->size;
381 atomic_add(buffers_added, &pool->available);
382 return;
384 failure:
385 if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
386 dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
387 pool->free_map[pool->next_free] = index;
388 pool->rx_buff[index].skb = NULL;
390 dev_kfree_skb_any(skb);
391 adapter->replenish_add_buff_failure++;
392 atomic_add(buffers_added, &pool->available);
394 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
395 /* Disable buffer pool replenishment and report carrier off if
396 * queue is closed or pending failover.
397 * Firmware guarantees that a signal will be sent to the
398 * driver, triggering a reset.
400 deactivate_rx_pools(adapter);
401 netif_carrier_off(adapter->netdev);
405 static void replenish_pools(struct ibmvnic_adapter *adapter)
407 int i;
409 adapter->replenish_task_cycles++;
410 for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
411 i++) {
412 if (adapter->rx_pool[i].active)
413 replenish_rx_pool(adapter, &adapter->rx_pool[i]);
417 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
419 kfree(adapter->tx_stats_buffers);
420 kfree(adapter->rx_stats_buffers);
421 adapter->tx_stats_buffers = NULL;
422 adapter->rx_stats_buffers = NULL;
425 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
427 adapter->tx_stats_buffers =
428 kcalloc(IBMVNIC_MAX_QUEUES,
429 sizeof(struct ibmvnic_tx_queue_stats),
430 GFP_KERNEL);
431 if (!adapter->tx_stats_buffers)
432 return -ENOMEM;
434 adapter->rx_stats_buffers =
435 kcalloc(IBMVNIC_MAX_QUEUES,
436 sizeof(struct ibmvnic_rx_queue_stats),
437 GFP_KERNEL);
438 if (!adapter->rx_stats_buffers)
439 return -ENOMEM;
441 return 0;
444 static void release_stats_token(struct ibmvnic_adapter *adapter)
446 struct device *dev = &adapter->vdev->dev;
448 if (!adapter->stats_token)
449 return;
451 dma_unmap_single(dev, adapter->stats_token,
452 sizeof(struct ibmvnic_statistics),
453 DMA_FROM_DEVICE);
454 adapter->stats_token = 0;
457 static int init_stats_token(struct ibmvnic_adapter *adapter)
459 struct device *dev = &adapter->vdev->dev;
460 dma_addr_t stok;
462 stok = dma_map_single(dev, &adapter->stats,
463 sizeof(struct ibmvnic_statistics),
464 DMA_FROM_DEVICE);
465 if (dma_mapping_error(dev, stok)) {
466 dev_err(dev, "Couldn't map stats buffer\n");
467 return -1;
470 adapter->stats_token = stok;
471 netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
472 return 0;
475 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
477 struct ibmvnic_rx_pool *rx_pool;
478 int rx_scrqs;
479 int i, j, rc;
480 u64 *size_array;
482 size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
483 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
485 rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
486 for (i = 0; i < rx_scrqs; i++) {
487 rx_pool = &adapter->rx_pool[i];
489 netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
491 if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
492 free_long_term_buff(adapter, &rx_pool->long_term_buff);
493 rx_pool->buff_size = be64_to_cpu(size_array[i]);
494 rc = alloc_long_term_buff(adapter,
495 &rx_pool->long_term_buff,
496 rx_pool->size *
497 rx_pool->buff_size);
498 } else {
499 rc = reset_long_term_buff(adapter,
500 &rx_pool->long_term_buff);
503 if (rc)
504 return rc;
506 for (j = 0; j < rx_pool->size; j++)
507 rx_pool->free_map[j] = j;
509 memset(rx_pool->rx_buff, 0,
510 rx_pool->size * sizeof(struct ibmvnic_rx_buff));
512 atomic_set(&rx_pool->available, 0);
513 rx_pool->next_alloc = 0;
514 rx_pool->next_free = 0;
515 rx_pool->active = 1;
518 return 0;
521 static void release_rx_pools(struct ibmvnic_adapter *adapter)
523 struct ibmvnic_rx_pool *rx_pool;
524 int i, j;
526 if (!adapter->rx_pool)
527 return;
529 for (i = 0; i < adapter->num_active_rx_pools; i++) {
530 rx_pool = &adapter->rx_pool[i];
532 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
534 kfree(rx_pool->free_map);
535 free_long_term_buff(adapter, &rx_pool->long_term_buff);
537 if (!rx_pool->rx_buff)
538 continue;
540 for (j = 0; j < rx_pool->size; j++) {
541 if (rx_pool->rx_buff[j].skb) {
542 dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
543 rx_pool->rx_buff[j].skb = NULL;
547 kfree(rx_pool->rx_buff);
550 kfree(adapter->rx_pool);
551 adapter->rx_pool = NULL;
552 adapter->num_active_rx_pools = 0;
555 static int init_rx_pools(struct net_device *netdev)
557 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
558 struct device *dev = &adapter->vdev->dev;
559 struct ibmvnic_rx_pool *rx_pool;
560 int rxadd_subcrqs;
561 u64 *size_array;
562 int i, j;
564 rxadd_subcrqs =
565 be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
566 size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
567 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
569 adapter->rx_pool = kcalloc(rxadd_subcrqs,
570 sizeof(struct ibmvnic_rx_pool),
571 GFP_KERNEL);
572 if (!adapter->rx_pool) {
573 dev_err(dev, "Failed to allocate rx pools\n");
574 return -1;
577 adapter->num_active_rx_pools = rxadd_subcrqs;
579 for (i = 0; i < rxadd_subcrqs; i++) {
580 rx_pool = &adapter->rx_pool[i];
582 netdev_dbg(adapter->netdev,
583 "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
584 i, adapter->req_rx_add_entries_per_subcrq,
585 be64_to_cpu(size_array[i]));
587 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
588 rx_pool->index = i;
589 rx_pool->buff_size = be64_to_cpu(size_array[i]);
590 rx_pool->active = 1;
592 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
593 GFP_KERNEL);
594 if (!rx_pool->free_map) {
595 release_rx_pools(adapter);
596 return -1;
599 rx_pool->rx_buff = kcalloc(rx_pool->size,
600 sizeof(struct ibmvnic_rx_buff),
601 GFP_KERNEL);
602 if (!rx_pool->rx_buff) {
603 dev_err(dev, "Couldn't alloc rx buffers\n");
604 release_rx_pools(adapter);
605 return -1;
608 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
609 rx_pool->size * rx_pool->buff_size)) {
610 release_rx_pools(adapter);
611 return -1;
614 for (j = 0; j < rx_pool->size; ++j)
615 rx_pool->free_map[j] = j;
617 atomic_set(&rx_pool->available, 0);
618 rx_pool->next_alloc = 0;
619 rx_pool->next_free = 0;
622 return 0;
625 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
626 struct ibmvnic_tx_pool *tx_pool)
628 int rc, i;
630 rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
631 if (rc)
632 return rc;
634 memset(tx_pool->tx_buff, 0,
635 tx_pool->num_buffers *
636 sizeof(struct ibmvnic_tx_buff));
638 for (i = 0; i < tx_pool->num_buffers; i++)
639 tx_pool->free_map[i] = i;
641 tx_pool->consumer_index = 0;
642 tx_pool->producer_index = 0;
644 return 0;
647 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
649 int tx_scrqs;
650 int i, rc;
652 tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
653 for (i = 0; i < tx_scrqs; i++) {
654 rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
655 if (rc)
656 return rc;
657 rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
658 if (rc)
659 return rc;
662 return 0;
665 static void release_vpd_data(struct ibmvnic_adapter *adapter)
667 if (!adapter->vpd)
668 return;
670 kfree(adapter->vpd->buff);
671 kfree(adapter->vpd);
673 adapter->vpd = NULL;
676 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
677 struct ibmvnic_tx_pool *tx_pool)
679 kfree(tx_pool->tx_buff);
680 kfree(tx_pool->free_map);
681 free_long_term_buff(adapter, &tx_pool->long_term_buff);
684 static void release_tx_pools(struct ibmvnic_adapter *adapter)
686 int i;
688 if (!adapter->tx_pool)
689 return;
691 for (i = 0; i < adapter->num_active_tx_pools; i++) {
692 release_one_tx_pool(adapter, &adapter->tx_pool[i]);
693 release_one_tx_pool(adapter, &adapter->tso_pool[i]);
696 kfree(adapter->tx_pool);
697 adapter->tx_pool = NULL;
698 kfree(adapter->tso_pool);
699 adapter->tso_pool = NULL;
700 adapter->num_active_tx_pools = 0;
703 static int init_one_tx_pool(struct net_device *netdev,
704 struct ibmvnic_tx_pool *tx_pool,
705 int num_entries, int buf_size)
707 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
708 int i;
710 tx_pool->tx_buff = kcalloc(num_entries,
711 sizeof(struct ibmvnic_tx_buff),
712 GFP_KERNEL);
713 if (!tx_pool->tx_buff)
714 return -1;
716 if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
717 num_entries * buf_size))
718 return -1;
720 tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
721 if (!tx_pool->free_map)
722 return -1;
724 for (i = 0; i < num_entries; i++)
725 tx_pool->free_map[i] = i;
727 tx_pool->consumer_index = 0;
728 tx_pool->producer_index = 0;
729 tx_pool->num_buffers = num_entries;
730 tx_pool->buf_size = buf_size;
732 return 0;
735 static int init_tx_pools(struct net_device *netdev)
737 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
738 int tx_subcrqs;
739 int i, rc;
741 tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
742 adapter->tx_pool = kcalloc(tx_subcrqs,
743 sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
744 if (!adapter->tx_pool)
745 return -1;
747 adapter->tso_pool = kcalloc(tx_subcrqs,
748 sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
749 if (!adapter->tso_pool)
750 return -1;
752 adapter->num_active_tx_pools = tx_subcrqs;
754 for (i = 0; i < tx_subcrqs; i++) {
755 rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
756 adapter->req_tx_entries_per_subcrq,
757 adapter->req_mtu + VLAN_HLEN);
758 if (rc) {
759 release_tx_pools(adapter);
760 return rc;
763 rc = init_one_tx_pool(netdev, &adapter->tso_pool[i],
764 IBMVNIC_TSO_BUFS,
765 IBMVNIC_TSO_BUF_SZ);
766 if (rc) {
767 release_tx_pools(adapter);
768 return rc;
772 return 0;
775 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
777 int i;
779 if (adapter->napi_enabled)
780 return;
782 for (i = 0; i < adapter->req_rx_queues; i++)
783 napi_enable(&adapter->napi[i]);
785 adapter->napi_enabled = true;
788 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
790 int i;
792 if (!adapter->napi_enabled)
793 return;
795 for (i = 0; i < adapter->req_rx_queues; i++) {
796 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
797 napi_disable(&adapter->napi[i]);
800 adapter->napi_enabled = false;
803 static int init_napi(struct ibmvnic_adapter *adapter)
805 int i;
807 adapter->napi = kcalloc(adapter->req_rx_queues,
808 sizeof(struct napi_struct), GFP_KERNEL);
809 if (!adapter->napi)
810 return -ENOMEM;
812 for (i = 0; i < adapter->req_rx_queues; i++) {
813 netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
814 netif_napi_add(adapter->netdev, &adapter->napi[i],
815 ibmvnic_poll, NAPI_POLL_WEIGHT);
818 adapter->num_active_rx_napi = adapter->req_rx_queues;
819 return 0;
822 static void release_napi(struct ibmvnic_adapter *adapter)
824 int i;
826 if (!adapter->napi)
827 return;
829 for (i = 0; i < adapter->num_active_rx_napi; i++) {
830 netdev_dbg(adapter->netdev, "Releasing napi[%d]\n", i);
831 netif_napi_del(&adapter->napi[i]);
834 kfree(adapter->napi);
835 adapter->napi = NULL;
836 adapter->num_active_rx_napi = 0;
837 adapter->napi_enabled = false;
840 static int ibmvnic_login(struct net_device *netdev)
842 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
843 unsigned long timeout = msecs_to_jiffies(30000);
844 int retry_count = 0;
845 bool retry;
846 int rc;
848 do {
849 retry = false;
850 if (retry_count > IBMVNIC_MAX_QUEUES) {
851 netdev_warn(netdev, "Login attempts exceeded\n");
852 return -1;
855 adapter->init_done_rc = 0;
856 reinit_completion(&adapter->init_done);
857 rc = send_login(adapter);
858 if (rc) {
859 netdev_warn(netdev, "Unable to login\n");
860 return rc;
863 if (!wait_for_completion_timeout(&adapter->init_done,
864 timeout)) {
865 netdev_warn(netdev, "Login timed out\n");
866 return -1;
869 if (adapter->init_done_rc == PARTIALSUCCESS) {
870 retry_count++;
871 release_sub_crqs(adapter, 1);
873 retry = true;
874 netdev_dbg(netdev,
875 "Received partial success, retrying...\n");
876 adapter->init_done_rc = 0;
877 reinit_completion(&adapter->init_done);
878 send_cap_queries(adapter);
879 if (!wait_for_completion_timeout(&adapter->init_done,
880 timeout)) {
881 netdev_warn(netdev,
882 "Capabilities query timed out\n");
883 return -1;
886 rc = init_sub_crqs(adapter);
887 if (rc) {
888 netdev_warn(netdev,
889 "SCRQ initialization failed\n");
890 return -1;
893 rc = init_sub_crq_irqs(adapter);
894 if (rc) {
895 netdev_warn(netdev,
896 "SCRQ irq initialization failed\n");
897 return -1;
899 } else if (adapter->init_done_rc) {
900 netdev_warn(netdev, "Adapter login failed\n");
901 return -1;
903 } while (retry);
905 __ibmvnic_set_mac(netdev, adapter->mac_addr);
907 return 0;
910 static void release_login_buffer(struct ibmvnic_adapter *adapter)
912 kfree(adapter->login_buf);
913 adapter->login_buf = NULL;
916 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
918 kfree(adapter->login_rsp_buf);
919 adapter->login_rsp_buf = NULL;
922 static void release_resources(struct ibmvnic_adapter *adapter)
924 release_vpd_data(adapter);
926 release_tx_pools(adapter);
927 release_rx_pools(adapter);
929 release_napi(adapter);
930 release_login_rsp_buffer(adapter);
933 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
935 struct net_device *netdev = adapter->netdev;
936 unsigned long timeout = msecs_to_jiffies(30000);
937 union ibmvnic_crq crq;
938 bool resend;
939 int rc;
941 netdev_dbg(netdev, "setting link state %d\n", link_state);
943 memset(&crq, 0, sizeof(crq));
944 crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
945 crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
946 crq.logical_link_state.link_state = link_state;
948 do {
949 resend = false;
951 reinit_completion(&adapter->init_done);
952 rc = ibmvnic_send_crq(adapter, &crq);
953 if (rc) {
954 netdev_err(netdev, "Failed to set link state\n");
955 return rc;
958 if (!wait_for_completion_timeout(&adapter->init_done,
959 timeout)) {
960 netdev_err(netdev, "timeout setting link state\n");
961 return -1;
964 if (adapter->init_done_rc == 1) {
965 /* Partuial success, delay and re-send */
966 mdelay(1000);
967 resend = true;
968 } else if (adapter->init_done_rc) {
969 netdev_warn(netdev, "Unable to set link state, rc=%d\n",
970 adapter->init_done_rc);
971 return adapter->init_done_rc;
973 } while (resend);
975 return 0;
978 static int set_real_num_queues(struct net_device *netdev)
980 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
981 int rc;
983 netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
984 adapter->req_tx_queues, adapter->req_rx_queues);
986 rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
987 if (rc) {
988 netdev_err(netdev, "failed to set the number of tx queues\n");
989 return rc;
992 rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
993 if (rc)
994 netdev_err(netdev, "failed to set the number of rx queues\n");
996 return rc;
999 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
1001 struct device *dev = &adapter->vdev->dev;
1002 union ibmvnic_crq crq;
1003 int len = 0;
1004 int rc;
1006 if (adapter->vpd->buff)
1007 len = adapter->vpd->len;
1009 mutex_lock(&adapter->fw_lock);
1010 adapter->fw_done_rc = 0;
1011 reinit_completion(&adapter->fw_done);
1013 crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
1014 crq.get_vpd_size.cmd = GET_VPD_SIZE;
1015 rc = ibmvnic_send_crq(adapter, &crq);
1016 if (rc) {
1017 mutex_unlock(&adapter->fw_lock);
1018 return rc;
1021 rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1022 if (rc) {
1023 dev_err(dev, "Could not retrieve VPD size, rc = %d\n", rc);
1024 mutex_unlock(&adapter->fw_lock);
1025 return rc;
1027 mutex_unlock(&adapter->fw_lock);
1029 if (!adapter->vpd->len)
1030 return -ENODATA;
1032 if (!adapter->vpd->buff)
1033 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
1034 else if (adapter->vpd->len != len)
1035 adapter->vpd->buff =
1036 krealloc(adapter->vpd->buff,
1037 adapter->vpd->len, GFP_KERNEL);
1039 if (!adapter->vpd->buff) {
1040 dev_err(dev, "Could allocate VPD buffer\n");
1041 return -ENOMEM;
1044 adapter->vpd->dma_addr =
1045 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
1046 DMA_FROM_DEVICE);
1047 if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
1048 dev_err(dev, "Could not map VPD buffer\n");
1049 kfree(adapter->vpd->buff);
1050 adapter->vpd->buff = NULL;
1051 return -ENOMEM;
1054 mutex_lock(&adapter->fw_lock);
1055 adapter->fw_done_rc = 0;
1056 reinit_completion(&adapter->fw_done);
1058 crq.get_vpd.first = IBMVNIC_CRQ_CMD;
1059 crq.get_vpd.cmd = GET_VPD;
1060 crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
1061 crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1062 rc = ibmvnic_send_crq(adapter, &crq);
1063 if (rc) {
1064 kfree(adapter->vpd->buff);
1065 adapter->vpd->buff = NULL;
1066 mutex_unlock(&adapter->fw_lock);
1067 return rc;
1070 rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1071 if (rc) {
1072 dev_err(dev, "Unable to retrieve VPD, rc = %d\n", rc);
1073 kfree(adapter->vpd->buff);
1074 adapter->vpd->buff = NULL;
1075 mutex_unlock(&adapter->fw_lock);
1076 return rc;
1079 mutex_unlock(&adapter->fw_lock);
1080 return 0;
1083 static int init_resources(struct ibmvnic_adapter *adapter)
1085 struct net_device *netdev = adapter->netdev;
1086 int rc;
1088 rc = set_real_num_queues(netdev);
1089 if (rc)
1090 return rc;
1092 adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1093 if (!adapter->vpd)
1094 return -ENOMEM;
1096 /* Vital Product Data (VPD) */
1097 rc = ibmvnic_get_vpd(adapter);
1098 if (rc) {
1099 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1100 return rc;
1103 adapter->map_id = 1;
1105 rc = init_napi(adapter);
1106 if (rc)
1107 return rc;
1109 send_map_query(adapter);
1111 rc = init_rx_pools(netdev);
1112 if (rc)
1113 return rc;
1115 rc = init_tx_pools(netdev);
1116 return rc;
1119 static int __ibmvnic_open(struct net_device *netdev)
1121 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1122 enum vnic_state prev_state = adapter->state;
1123 int i, rc;
1125 adapter->state = VNIC_OPENING;
1126 replenish_pools(adapter);
1127 ibmvnic_napi_enable(adapter);
1129 /* We're ready to receive frames, enable the sub-crq interrupts and
1130 * set the logical link state to up
1132 for (i = 0; i < adapter->req_rx_queues; i++) {
1133 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1134 if (prev_state == VNIC_CLOSED)
1135 enable_irq(adapter->rx_scrq[i]->irq);
1136 enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1139 for (i = 0; i < adapter->req_tx_queues; i++) {
1140 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1141 if (prev_state == VNIC_CLOSED)
1142 enable_irq(adapter->tx_scrq[i]->irq);
1143 enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1146 rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1147 if (rc) {
1148 for (i = 0; i < adapter->req_rx_queues; i++)
1149 napi_disable(&adapter->napi[i]);
1150 release_resources(adapter);
1151 return rc;
1154 netif_tx_start_all_queues(netdev);
1156 if (prev_state == VNIC_CLOSED) {
1157 for (i = 0; i < adapter->req_rx_queues; i++)
1158 napi_schedule(&adapter->napi[i]);
1161 adapter->state = VNIC_OPEN;
1162 return rc;
1165 static int ibmvnic_open(struct net_device *netdev)
1167 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1168 int rc;
1170 /* If device failover is pending, just set device state and return.
1171 * Device operation will be handled by reset routine.
1173 if (adapter->failover_pending) {
1174 adapter->state = VNIC_OPEN;
1175 return 0;
1178 if (adapter->state != VNIC_CLOSED) {
1179 rc = ibmvnic_login(netdev);
1180 if (rc)
1181 return rc;
1183 rc = init_resources(adapter);
1184 if (rc) {
1185 netdev_err(netdev, "failed to initialize resources\n");
1186 release_resources(adapter);
1187 return rc;
1191 rc = __ibmvnic_open(netdev);
1193 return rc;
1196 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1198 struct ibmvnic_rx_pool *rx_pool;
1199 struct ibmvnic_rx_buff *rx_buff;
1200 u64 rx_entries;
1201 int rx_scrqs;
1202 int i, j;
1204 if (!adapter->rx_pool)
1205 return;
1207 rx_scrqs = adapter->num_active_rx_pools;
1208 rx_entries = adapter->req_rx_add_entries_per_subcrq;
1210 /* Free any remaining skbs in the rx buffer pools */
1211 for (i = 0; i < rx_scrqs; i++) {
1212 rx_pool = &adapter->rx_pool[i];
1213 if (!rx_pool || !rx_pool->rx_buff)
1214 continue;
1216 netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1217 for (j = 0; j < rx_entries; j++) {
1218 rx_buff = &rx_pool->rx_buff[j];
1219 if (rx_buff && rx_buff->skb) {
1220 dev_kfree_skb_any(rx_buff->skb);
1221 rx_buff->skb = NULL;
1227 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1228 struct ibmvnic_tx_pool *tx_pool)
1230 struct ibmvnic_tx_buff *tx_buff;
1231 u64 tx_entries;
1232 int i;
1234 if (!tx_pool || !tx_pool->tx_buff)
1235 return;
1237 tx_entries = tx_pool->num_buffers;
1239 for (i = 0; i < tx_entries; i++) {
1240 tx_buff = &tx_pool->tx_buff[i];
1241 if (tx_buff && tx_buff->skb) {
1242 dev_kfree_skb_any(tx_buff->skb);
1243 tx_buff->skb = NULL;
1248 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1250 int tx_scrqs;
1251 int i;
1253 if (!adapter->tx_pool || !adapter->tso_pool)
1254 return;
1256 tx_scrqs = adapter->num_active_tx_pools;
1258 /* Free any remaining skbs in the tx buffer pools */
1259 for (i = 0; i < tx_scrqs; i++) {
1260 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1261 clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1262 clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1266 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1268 struct net_device *netdev = adapter->netdev;
1269 int i;
1271 if (adapter->tx_scrq) {
1272 for (i = 0; i < adapter->req_tx_queues; i++)
1273 if (adapter->tx_scrq[i]->irq) {
1274 netdev_dbg(netdev,
1275 "Disabling tx_scrq[%d] irq\n", i);
1276 disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1277 disable_irq(adapter->tx_scrq[i]->irq);
1281 if (adapter->rx_scrq) {
1282 for (i = 0; i < adapter->req_rx_queues; i++) {
1283 if (adapter->rx_scrq[i]->irq) {
1284 netdev_dbg(netdev,
1285 "Disabling rx_scrq[%d] irq\n", i);
1286 disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1287 disable_irq(adapter->rx_scrq[i]->irq);
1293 static void ibmvnic_cleanup(struct net_device *netdev)
1295 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1297 /* ensure that transmissions are stopped if called by do_reset */
1298 if (test_bit(0, &adapter->resetting))
1299 netif_tx_disable(netdev);
1300 else
1301 netif_tx_stop_all_queues(netdev);
1303 ibmvnic_napi_disable(adapter);
1304 ibmvnic_disable_irqs(adapter);
1306 clean_rx_pools(adapter);
1307 clean_tx_pools(adapter);
1310 static int __ibmvnic_close(struct net_device *netdev)
1312 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1313 int rc = 0;
1315 adapter->state = VNIC_CLOSING;
1316 rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1317 if (rc)
1318 return rc;
1319 adapter->state = VNIC_CLOSED;
1320 return 0;
1323 static int ibmvnic_close(struct net_device *netdev)
1325 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1326 int rc;
1328 /* If device failover is pending, just set device state and return.
1329 * Device operation will be handled by reset routine.
1331 if (adapter->failover_pending) {
1332 adapter->state = VNIC_CLOSED;
1333 return 0;
1336 rc = __ibmvnic_close(netdev);
1337 ibmvnic_cleanup(netdev);
1339 return rc;
1343 * build_hdr_data - creates L2/L3/L4 header data buffer
1344 * @hdr_field - bitfield determining needed headers
1345 * @skb - socket buffer
1346 * @hdr_len - array of header lengths
1347 * @tot_len - total length of data
1349 * Reads hdr_field to determine which headers are needed by firmware.
1350 * Builds a buffer containing these headers. Saves individual header
1351 * lengths and total buffer length to be used to build descriptors.
1353 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1354 int *hdr_len, u8 *hdr_data)
1356 int len = 0;
1357 u8 *hdr;
1359 if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1360 hdr_len[0] = sizeof(struct vlan_ethhdr);
1361 else
1362 hdr_len[0] = sizeof(struct ethhdr);
1364 if (skb->protocol == htons(ETH_P_IP)) {
1365 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1366 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1367 hdr_len[2] = tcp_hdrlen(skb);
1368 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1369 hdr_len[2] = sizeof(struct udphdr);
1370 } else if (skb->protocol == htons(ETH_P_IPV6)) {
1371 hdr_len[1] = sizeof(struct ipv6hdr);
1372 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1373 hdr_len[2] = tcp_hdrlen(skb);
1374 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1375 hdr_len[2] = sizeof(struct udphdr);
1376 } else if (skb->protocol == htons(ETH_P_ARP)) {
1377 hdr_len[1] = arp_hdr_len(skb->dev);
1378 hdr_len[2] = 0;
1381 memset(hdr_data, 0, 120);
1382 if ((hdr_field >> 6) & 1) {
1383 hdr = skb_mac_header(skb);
1384 memcpy(hdr_data, hdr, hdr_len[0]);
1385 len += hdr_len[0];
1388 if ((hdr_field >> 5) & 1) {
1389 hdr = skb_network_header(skb);
1390 memcpy(hdr_data + len, hdr, hdr_len[1]);
1391 len += hdr_len[1];
1394 if ((hdr_field >> 4) & 1) {
1395 hdr = skb_transport_header(skb);
1396 memcpy(hdr_data + len, hdr, hdr_len[2]);
1397 len += hdr_len[2];
1399 return len;
1403 * create_hdr_descs - create header and header extension descriptors
1404 * @hdr_field - bitfield determining needed headers
1405 * @data - buffer containing header data
1406 * @len - length of data buffer
1407 * @hdr_len - array of individual header lengths
1408 * @scrq_arr - descriptor array
1410 * Creates header and, if needed, header extension descriptors and
1411 * places them in a descriptor array, scrq_arr
1414 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1415 union sub_crq *scrq_arr)
1417 union sub_crq hdr_desc;
1418 int tmp_len = len;
1419 int num_descs = 0;
1420 u8 *data, *cur;
1421 int tmp;
1423 while (tmp_len > 0) {
1424 cur = hdr_data + len - tmp_len;
1426 memset(&hdr_desc, 0, sizeof(hdr_desc));
1427 if (cur != hdr_data) {
1428 data = hdr_desc.hdr_ext.data;
1429 tmp = tmp_len > 29 ? 29 : tmp_len;
1430 hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1431 hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1432 hdr_desc.hdr_ext.len = tmp;
1433 } else {
1434 data = hdr_desc.hdr.data;
1435 tmp = tmp_len > 24 ? 24 : tmp_len;
1436 hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1437 hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1438 hdr_desc.hdr.len = tmp;
1439 hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1440 hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1441 hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1442 hdr_desc.hdr.flag = hdr_field << 1;
1444 memcpy(data, cur, tmp);
1445 tmp_len -= tmp;
1446 *scrq_arr = hdr_desc;
1447 scrq_arr++;
1448 num_descs++;
1451 return num_descs;
1455 * build_hdr_descs_arr - build a header descriptor array
1456 * @skb - socket buffer
1457 * @num_entries - number of descriptors to be sent
1458 * @subcrq - first TX descriptor
1459 * @hdr_field - bit field determining which headers will be sent
1461 * This function will build a TX descriptor array with applicable
1462 * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1465 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1466 int *num_entries, u8 hdr_field)
1468 int hdr_len[3] = {0, 0, 0};
1469 int tot_len;
1470 u8 *hdr_data = txbuff->hdr_data;
1472 tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1473 txbuff->hdr_data);
1474 *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1475 txbuff->indir_arr + 1);
1478 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1479 struct net_device *netdev)
1481 /* For some backing devices, mishandling of small packets
1482 * can result in a loss of connection or TX stall. Device
1483 * architects recommend that no packet should be smaller
1484 * than the minimum MTU value provided to the driver, so
1485 * pad any packets to that length
1487 if (skb->len < netdev->min_mtu)
1488 return skb_put_padto(skb, netdev->min_mtu);
1490 return 0;
1493 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1495 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1496 int queue_num = skb_get_queue_mapping(skb);
1497 u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1498 struct device *dev = &adapter->vdev->dev;
1499 struct ibmvnic_tx_buff *tx_buff = NULL;
1500 struct ibmvnic_sub_crq_queue *tx_scrq;
1501 struct ibmvnic_tx_pool *tx_pool;
1502 unsigned int tx_send_failed = 0;
1503 unsigned int tx_map_failed = 0;
1504 unsigned int tx_dropped = 0;
1505 unsigned int tx_packets = 0;
1506 unsigned int tx_bytes = 0;
1507 dma_addr_t data_dma_addr;
1508 struct netdev_queue *txq;
1509 unsigned long lpar_rc;
1510 union sub_crq tx_crq;
1511 unsigned int offset;
1512 int num_entries = 1;
1513 unsigned char *dst;
1514 u64 *handle_array;
1515 int index = 0;
1516 u8 proto = 0;
1517 netdev_tx_t ret = NETDEV_TX_OK;
1519 if (test_bit(0, &adapter->resetting)) {
1520 if (!netif_subqueue_stopped(netdev, skb))
1521 netif_stop_subqueue(netdev, queue_num);
1522 dev_kfree_skb_any(skb);
1524 tx_send_failed++;
1525 tx_dropped++;
1526 ret = NETDEV_TX_OK;
1527 goto out;
1530 if (ibmvnic_xmit_workarounds(skb, netdev)) {
1531 tx_dropped++;
1532 tx_send_failed++;
1533 ret = NETDEV_TX_OK;
1534 goto out;
1536 if (skb_is_gso(skb))
1537 tx_pool = &adapter->tso_pool[queue_num];
1538 else
1539 tx_pool = &adapter->tx_pool[queue_num];
1541 tx_scrq = adapter->tx_scrq[queue_num];
1542 txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1543 handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1544 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1546 index = tx_pool->free_map[tx_pool->consumer_index];
1548 if (index == IBMVNIC_INVALID_MAP) {
1549 dev_kfree_skb_any(skb);
1550 tx_send_failed++;
1551 tx_dropped++;
1552 ret = NETDEV_TX_OK;
1553 goto out;
1556 tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1558 offset = index * tx_pool->buf_size;
1559 dst = tx_pool->long_term_buff.buff + offset;
1560 memset(dst, 0, tx_pool->buf_size);
1561 data_dma_addr = tx_pool->long_term_buff.addr + offset;
1563 if (skb_shinfo(skb)->nr_frags) {
1564 int cur, i;
1566 /* Copy the head */
1567 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1568 cur = skb_headlen(skb);
1570 /* Copy the frags */
1571 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1572 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1574 memcpy(dst + cur,
1575 page_address(skb_frag_page(frag)) +
1576 skb_frag_off(frag), skb_frag_size(frag));
1577 cur += skb_frag_size(frag);
1579 } else {
1580 skb_copy_from_linear_data(skb, dst, skb->len);
1583 tx_pool->consumer_index =
1584 (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1586 tx_buff = &tx_pool->tx_buff[index];
1587 tx_buff->skb = skb;
1588 tx_buff->data_dma[0] = data_dma_addr;
1589 tx_buff->data_len[0] = skb->len;
1590 tx_buff->index = index;
1591 tx_buff->pool_index = queue_num;
1592 tx_buff->last_frag = true;
1594 memset(&tx_crq, 0, sizeof(tx_crq));
1595 tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1596 tx_crq.v1.type = IBMVNIC_TX_DESC;
1597 tx_crq.v1.n_crq_elem = 1;
1598 tx_crq.v1.n_sge = 1;
1599 tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1601 if (skb_is_gso(skb))
1602 tx_crq.v1.correlator =
1603 cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1604 else
1605 tx_crq.v1.correlator = cpu_to_be32(index);
1606 tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1607 tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1608 tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1610 if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1611 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1612 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1615 if (skb->protocol == htons(ETH_P_IP)) {
1616 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1617 proto = ip_hdr(skb)->protocol;
1618 } else if (skb->protocol == htons(ETH_P_IPV6)) {
1619 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1620 proto = ipv6_hdr(skb)->nexthdr;
1623 if (proto == IPPROTO_TCP)
1624 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1625 else if (proto == IPPROTO_UDP)
1626 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1628 if (skb->ip_summed == CHECKSUM_PARTIAL) {
1629 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1630 hdrs += 2;
1632 if (skb_is_gso(skb)) {
1633 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1634 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1635 hdrs += 2;
1637 /* determine if l2/3/4 headers are sent to firmware */
1638 if ((*hdrs >> 7) & 1) {
1639 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1640 tx_crq.v1.n_crq_elem = num_entries;
1641 tx_buff->num_entries = num_entries;
1642 tx_buff->indir_arr[0] = tx_crq;
1643 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1644 sizeof(tx_buff->indir_arr),
1645 DMA_TO_DEVICE);
1646 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1647 dev_kfree_skb_any(skb);
1648 tx_buff->skb = NULL;
1649 if (!firmware_has_feature(FW_FEATURE_CMO))
1650 dev_err(dev, "tx: unable to map descriptor array\n");
1651 tx_map_failed++;
1652 tx_dropped++;
1653 ret = NETDEV_TX_OK;
1654 goto tx_err_out;
1656 lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1657 (u64)tx_buff->indir_dma,
1658 (u64)num_entries);
1659 dma_unmap_single(dev, tx_buff->indir_dma,
1660 sizeof(tx_buff->indir_arr), DMA_TO_DEVICE);
1661 } else {
1662 tx_buff->num_entries = num_entries;
1663 lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1664 &tx_crq);
1666 if (lpar_rc != H_SUCCESS) {
1667 if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1668 dev_err_ratelimited(dev, "tx: send failed\n");
1669 dev_kfree_skb_any(skb);
1670 tx_buff->skb = NULL;
1672 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1673 /* Disable TX and report carrier off if queue is closed
1674 * or pending failover.
1675 * Firmware guarantees that a signal will be sent to the
1676 * driver, triggering a reset or some other action.
1678 netif_tx_stop_all_queues(netdev);
1679 netif_carrier_off(netdev);
1682 tx_send_failed++;
1683 tx_dropped++;
1684 ret = NETDEV_TX_OK;
1685 goto tx_err_out;
1688 if (atomic_add_return(num_entries, &tx_scrq->used)
1689 >= adapter->req_tx_entries_per_subcrq) {
1690 netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1691 netif_stop_subqueue(netdev, queue_num);
1694 tx_packets++;
1695 tx_bytes += skb->len;
1696 txq->trans_start = jiffies;
1697 ret = NETDEV_TX_OK;
1698 goto out;
1700 tx_err_out:
1701 /* roll back consumer index and map array*/
1702 if (tx_pool->consumer_index == 0)
1703 tx_pool->consumer_index =
1704 tx_pool->num_buffers - 1;
1705 else
1706 tx_pool->consumer_index--;
1707 tx_pool->free_map[tx_pool->consumer_index] = index;
1708 out:
1709 netdev->stats.tx_dropped += tx_dropped;
1710 netdev->stats.tx_bytes += tx_bytes;
1711 netdev->stats.tx_packets += tx_packets;
1712 adapter->tx_send_failed += tx_send_failed;
1713 adapter->tx_map_failed += tx_map_failed;
1714 adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1715 adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1716 adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1718 return ret;
1721 static void ibmvnic_set_multi(struct net_device *netdev)
1723 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1724 struct netdev_hw_addr *ha;
1725 union ibmvnic_crq crq;
1727 memset(&crq, 0, sizeof(crq));
1728 crq.request_capability.first = IBMVNIC_CRQ_CMD;
1729 crq.request_capability.cmd = REQUEST_CAPABILITY;
1731 if (netdev->flags & IFF_PROMISC) {
1732 if (!adapter->promisc_supported)
1733 return;
1734 } else {
1735 if (netdev->flags & IFF_ALLMULTI) {
1736 /* Accept all multicast */
1737 memset(&crq, 0, sizeof(crq));
1738 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1739 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1740 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1741 ibmvnic_send_crq(adapter, &crq);
1742 } else if (netdev_mc_empty(netdev)) {
1743 /* Reject all multicast */
1744 memset(&crq, 0, sizeof(crq));
1745 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1746 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1747 crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1748 ibmvnic_send_crq(adapter, &crq);
1749 } else {
1750 /* Accept one or more multicast(s) */
1751 netdev_for_each_mc_addr(ha, netdev) {
1752 memset(&crq, 0, sizeof(crq));
1753 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1754 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1755 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1756 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1757 ha->addr);
1758 ibmvnic_send_crq(adapter, &crq);
1764 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1766 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1767 union ibmvnic_crq crq;
1768 int rc;
1770 if (!is_valid_ether_addr(dev_addr)) {
1771 rc = -EADDRNOTAVAIL;
1772 goto err;
1775 memset(&crq, 0, sizeof(crq));
1776 crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1777 crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1778 ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1780 mutex_lock(&adapter->fw_lock);
1781 adapter->fw_done_rc = 0;
1782 reinit_completion(&adapter->fw_done);
1784 rc = ibmvnic_send_crq(adapter, &crq);
1785 if (rc) {
1786 rc = -EIO;
1787 mutex_unlock(&adapter->fw_lock);
1788 goto err;
1791 rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1792 /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1793 if (rc || adapter->fw_done_rc) {
1794 rc = -EIO;
1795 mutex_unlock(&adapter->fw_lock);
1796 goto err;
1798 mutex_unlock(&adapter->fw_lock);
1799 return 0;
1800 err:
1801 ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1802 return rc;
1805 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1807 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1808 struct sockaddr *addr = p;
1809 int rc;
1811 rc = 0;
1812 ether_addr_copy(adapter->mac_addr, addr->sa_data);
1813 if (adapter->state != VNIC_PROBED)
1814 rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1816 return rc;
1820 * do_change_param_reset returns zero if we are able to keep processing reset
1821 * events, or non-zero if we hit a fatal error and must halt.
1823 static int do_change_param_reset(struct ibmvnic_adapter *adapter,
1824 struct ibmvnic_rwi *rwi,
1825 u32 reset_state)
1827 struct net_device *netdev = adapter->netdev;
1828 int i, rc;
1830 netdev_dbg(adapter->netdev, "Change param resetting driver (%d)\n",
1831 rwi->reset_reason);
1833 netif_carrier_off(netdev);
1834 adapter->reset_reason = rwi->reset_reason;
1836 ibmvnic_cleanup(netdev);
1838 if (reset_state == VNIC_OPEN) {
1839 rc = __ibmvnic_close(netdev);
1840 if (rc)
1841 return rc;
1844 release_resources(adapter);
1845 release_sub_crqs(adapter, 1);
1846 release_crq_queue(adapter);
1848 adapter->state = VNIC_PROBED;
1850 rc = init_crq_queue(adapter);
1852 if (rc) {
1853 netdev_err(adapter->netdev,
1854 "Couldn't initialize crq. rc=%d\n", rc);
1855 return rc;
1858 rc = ibmvnic_reset_init(adapter);
1859 if (rc)
1860 return IBMVNIC_INIT_FAILED;
1862 /* If the adapter was in PROBE state prior to the reset,
1863 * exit here.
1865 if (reset_state == VNIC_PROBED)
1866 return 0;
1868 rc = ibmvnic_login(netdev);
1869 if (rc) {
1870 adapter->state = reset_state;
1871 return rc;
1874 rc = init_resources(adapter);
1875 if (rc)
1876 return rc;
1878 ibmvnic_disable_irqs(adapter);
1880 adapter->state = VNIC_CLOSED;
1882 if (reset_state == VNIC_CLOSED)
1883 return 0;
1885 rc = __ibmvnic_open(netdev);
1886 if (rc)
1887 return IBMVNIC_OPEN_FAILED;
1889 /* refresh device's multicast list */
1890 ibmvnic_set_multi(netdev);
1892 /* kick napi */
1893 for (i = 0; i < adapter->req_rx_queues; i++)
1894 napi_schedule(&adapter->napi[i]);
1896 return 0;
1900 * do_reset returns zero if we are able to keep processing reset events, or
1901 * non-zero if we hit a fatal error and must halt.
1903 static int do_reset(struct ibmvnic_adapter *adapter,
1904 struct ibmvnic_rwi *rwi, u32 reset_state)
1906 u64 old_num_rx_queues, old_num_tx_queues;
1907 u64 old_num_rx_slots, old_num_tx_slots;
1908 struct net_device *netdev = adapter->netdev;
1909 int i, rc;
1911 netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1912 rwi->reset_reason);
1914 rtnl_lock();
1916 netif_carrier_off(netdev);
1917 adapter->reset_reason = rwi->reset_reason;
1919 old_num_rx_queues = adapter->req_rx_queues;
1920 old_num_tx_queues = adapter->req_tx_queues;
1921 old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1922 old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1924 ibmvnic_cleanup(netdev);
1926 if (reset_state == VNIC_OPEN &&
1927 adapter->reset_reason != VNIC_RESET_MOBILITY &&
1928 adapter->reset_reason != VNIC_RESET_FAILOVER) {
1929 adapter->state = VNIC_CLOSING;
1931 /* Release the RTNL lock before link state change and
1932 * re-acquire after the link state change to allow
1933 * linkwatch_event to grab the RTNL lock and run during
1934 * a reset.
1936 rtnl_unlock();
1937 rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1938 rtnl_lock();
1939 if (rc)
1940 goto out;
1942 if (adapter->state != VNIC_CLOSING) {
1943 rc = -1;
1944 goto out;
1947 adapter->state = VNIC_CLOSED;
1950 if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1951 /* remove the closed state so when we call open it appears
1952 * we are coming from the probed state.
1954 adapter->state = VNIC_PROBED;
1956 if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1957 rc = ibmvnic_reenable_crq_queue(adapter);
1958 release_sub_crqs(adapter, 1);
1959 } else {
1960 rc = ibmvnic_reset_crq(adapter);
1961 if (!rc)
1962 rc = vio_enable_interrupts(adapter->vdev);
1965 if (rc) {
1966 netdev_err(adapter->netdev,
1967 "Couldn't initialize crq. rc=%d\n", rc);
1968 goto out;
1971 rc = ibmvnic_reset_init(adapter);
1972 if (rc) {
1973 rc = IBMVNIC_INIT_FAILED;
1974 goto out;
1977 /* If the adapter was in PROBE state prior to the reset,
1978 * exit here.
1980 if (reset_state == VNIC_PROBED) {
1981 rc = 0;
1982 goto out;
1985 rc = ibmvnic_login(netdev);
1986 if (rc) {
1987 adapter->state = reset_state;
1988 goto out;
1991 if (adapter->req_rx_queues != old_num_rx_queues ||
1992 adapter->req_tx_queues != old_num_tx_queues ||
1993 adapter->req_rx_add_entries_per_subcrq !=
1994 old_num_rx_slots ||
1995 adapter->req_tx_entries_per_subcrq !=
1996 old_num_tx_slots) {
1997 release_rx_pools(adapter);
1998 release_tx_pools(adapter);
1999 release_napi(adapter);
2000 release_vpd_data(adapter);
2002 rc = init_resources(adapter);
2003 if (rc)
2004 goto out;
2006 } else {
2007 rc = reset_tx_pools(adapter);
2008 if (rc)
2009 goto out;
2011 rc = reset_rx_pools(adapter);
2012 if (rc)
2013 goto out;
2015 ibmvnic_disable_irqs(adapter);
2017 adapter->state = VNIC_CLOSED;
2019 if (reset_state == VNIC_CLOSED) {
2020 rc = 0;
2021 goto out;
2024 rc = __ibmvnic_open(netdev);
2025 if (rc) {
2026 rc = IBMVNIC_OPEN_FAILED;
2027 goto out;
2030 /* refresh device's multicast list */
2031 ibmvnic_set_multi(netdev);
2033 /* kick napi */
2034 for (i = 0; i < adapter->req_rx_queues; i++)
2035 napi_schedule(&adapter->napi[i]);
2037 if (adapter->reset_reason != VNIC_RESET_FAILOVER)
2038 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
2040 rc = 0;
2042 out:
2043 rtnl_unlock();
2045 return rc;
2048 static int do_hard_reset(struct ibmvnic_adapter *adapter,
2049 struct ibmvnic_rwi *rwi, u32 reset_state)
2051 struct net_device *netdev = adapter->netdev;
2052 int rc;
2054 netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
2055 rwi->reset_reason);
2057 netif_carrier_off(netdev);
2058 adapter->reset_reason = rwi->reset_reason;
2060 ibmvnic_cleanup(netdev);
2061 release_resources(adapter);
2062 release_sub_crqs(adapter, 0);
2063 release_crq_queue(adapter);
2065 /* remove the closed state so when we call open it appears
2066 * we are coming from the probed state.
2068 adapter->state = VNIC_PROBED;
2070 reinit_completion(&adapter->init_done);
2071 rc = init_crq_queue(adapter);
2072 if (rc) {
2073 netdev_err(adapter->netdev,
2074 "Couldn't initialize crq. rc=%d\n", rc);
2075 return rc;
2078 rc = ibmvnic_init(adapter);
2079 if (rc)
2080 return rc;
2082 /* If the adapter was in PROBE state prior to the reset,
2083 * exit here.
2085 if (reset_state == VNIC_PROBED)
2086 return 0;
2088 rc = ibmvnic_login(netdev);
2089 if (rc) {
2090 adapter->state = VNIC_PROBED;
2091 return 0;
2094 rc = init_resources(adapter);
2095 if (rc)
2096 return rc;
2098 ibmvnic_disable_irqs(adapter);
2099 adapter->state = VNIC_CLOSED;
2101 if (reset_state == VNIC_CLOSED)
2102 return 0;
2104 rc = __ibmvnic_open(netdev);
2105 if (rc)
2106 return IBMVNIC_OPEN_FAILED;
2108 return 0;
2111 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
2113 struct ibmvnic_rwi *rwi;
2114 unsigned long flags;
2116 spin_lock_irqsave(&adapter->rwi_lock, flags);
2118 if (!list_empty(&adapter->rwi_list)) {
2119 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
2120 list);
2121 list_del(&rwi->list);
2122 } else {
2123 rwi = NULL;
2126 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2127 return rwi;
2130 static void free_all_rwi(struct ibmvnic_adapter *adapter)
2132 struct ibmvnic_rwi *rwi;
2134 rwi = get_next_rwi(adapter);
2135 while (rwi) {
2136 kfree(rwi);
2137 rwi = get_next_rwi(adapter);
2141 static void __ibmvnic_reset(struct work_struct *work)
2143 struct ibmvnic_rwi *rwi;
2144 struct ibmvnic_adapter *adapter;
2145 u32 reset_state;
2146 int rc = 0;
2148 adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
2150 if (test_and_set_bit_lock(0, &adapter->resetting)) {
2151 schedule_delayed_work(&adapter->ibmvnic_delayed_reset,
2152 IBMVNIC_RESET_DELAY);
2153 return;
2156 reset_state = adapter->state;
2158 rwi = get_next_rwi(adapter);
2159 while (rwi) {
2160 if (adapter->state == VNIC_REMOVING ||
2161 adapter->state == VNIC_REMOVED) {
2162 kfree(rwi);
2163 rc = EBUSY;
2164 break;
2167 if (rwi->reset_reason == VNIC_RESET_CHANGE_PARAM) {
2168 /* CHANGE_PARAM requestor holds rtnl_lock */
2169 rc = do_change_param_reset(adapter, rwi, reset_state);
2170 } else if (adapter->force_reset_recovery) {
2171 /* Transport event occurred during previous reset */
2172 if (adapter->wait_for_reset) {
2173 /* Previous was CHANGE_PARAM; caller locked */
2174 adapter->force_reset_recovery = false;
2175 rc = do_hard_reset(adapter, rwi, reset_state);
2176 } else {
2177 rtnl_lock();
2178 adapter->force_reset_recovery = false;
2179 rc = do_hard_reset(adapter, rwi, reset_state);
2180 rtnl_unlock();
2182 } else {
2183 rc = do_reset(adapter, rwi, reset_state);
2185 kfree(rwi);
2186 if (rc == IBMVNIC_OPEN_FAILED) {
2187 if (list_empty(&adapter->rwi_list))
2188 adapter->state = VNIC_CLOSED;
2189 else
2190 adapter->state = reset_state;
2191 rc = 0;
2192 } else if (rc && rc != IBMVNIC_INIT_FAILED &&
2193 !adapter->force_reset_recovery)
2194 break;
2196 rwi = get_next_rwi(adapter);
2198 if (rwi && (rwi->reset_reason == VNIC_RESET_FAILOVER ||
2199 rwi->reset_reason == VNIC_RESET_MOBILITY))
2200 adapter->force_reset_recovery = true;
2203 if (adapter->wait_for_reset) {
2204 adapter->reset_done_rc = rc;
2205 complete(&adapter->reset_done);
2208 if (rc) {
2209 netdev_dbg(adapter->netdev, "Reset failed\n");
2210 free_all_rwi(adapter);
2213 clear_bit_unlock(0, &adapter->resetting);
2216 static void __ibmvnic_delayed_reset(struct work_struct *work)
2218 struct ibmvnic_adapter *adapter;
2220 adapter = container_of(work, struct ibmvnic_adapter,
2221 ibmvnic_delayed_reset.work);
2222 __ibmvnic_reset(&adapter->ibmvnic_reset);
2225 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2226 enum ibmvnic_reset_reason reason)
2228 struct list_head *entry, *tmp_entry;
2229 struct ibmvnic_rwi *rwi, *tmp;
2230 struct net_device *netdev = adapter->netdev;
2231 unsigned long flags;
2232 int ret;
2234 if (adapter->state == VNIC_REMOVING ||
2235 adapter->state == VNIC_REMOVED ||
2236 adapter->failover_pending) {
2237 ret = EBUSY;
2238 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2239 goto err;
2242 if (adapter->state == VNIC_PROBING) {
2243 netdev_warn(netdev, "Adapter reset during probe\n");
2244 ret = adapter->init_done_rc = EAGAIN;
2245 goto err;
2248 spin_lock_irqsave(&adapter->rwi_lock, flags);
2250 list_for_each(entry, &adapter->rwi_list) {
2251 tmp = list_entry(entry, struct ibmvnic_rwi, list);
2252 if (tmp->reset_reason == reason) {
2253 netdev_dbg(netdev, "Skipping matching reset\n");
2254 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2255 ret = EBUSY;
2256 goto err;
2260 rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2261 if (!rwi) {
2262 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2263 ibmvnic_close(netdev);
2264 ret = ENOMEM;
2265 goto err;
2267 /* if we just received a transport event,
2268 * flush reset queue and process this reset
2270 if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2271 list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2272 list_del(entry);
2274 rwi->reset_reason = reason;
2275 list_add_tail(&rwi->list, &adapter->rwi_list);
2276 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2277 netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2278 schedule_work(&adapter->ibmvnic_reset);
2280 return 0;
2281 err:
2282 return -ret;
2285 static void ibmvnic_tx_timeout(struct net_device *dev, unsigned int txqueue)
2287 struct ibmvnic_adapter *adapter = netdev_priv(dev);
2289 ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2292 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2293 struct ibmvnic_rx_buff *rx_buff)
2295 struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2297 rx_buff->skb = NULL;
2299 pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2300 pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2302 atomic_dec(&pool->available);
2305 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2307 struct net_device *netdev = napi->dev;
2308 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2309 int scrq_num = (int)(napi - adapter->napi);
2310 int frames_processed = 0;
2312 restart_poll:
2313 while (frames_processed < budget) {
2314 struct sk_buff *skb;
2315 struct ibmvnic_rx_buff *rx_buff;
2316 union sub_crq *next;
2317 u32 length;
2318 u16 offset;
2319 u8 flags = 0;
2321 if (unlikely(test_bit(0, &adapter->resetting) &&
2322 adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2323 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2324 napi_complete_done(napi, frames_processed);
2325 return frames_processed;
2328 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2329 break;
2330 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2331 rx_buff =
2332 (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2333 rx_comp.correlator);
2334 /* do error checking */
2335 if (next->rx_comp.rc) {
2336 netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2337 be16_to_cpu(next->rx_comp.rc));
2338 /* free the entry */
2339 next->rx_comp.first = 0;
2340 dev_kfree_skb_any(rx_buff->skb);
2341 remove_buff_from_pool(adapter, rx_buff);
2342 continue;
2343 } else if (!rx_buff->skb) {
2344 /* free the entry */
2345 next->rx_comp.first = 0;
2346 remove_buff_from_pool(adapter, rx_buff);
2347 continue;
2350 length = be32_to_cpu(next->rx_comp.len);
2351 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2352 flags = next->rx_comp.flags;
2353 skb = rx_buff->skb;
2354 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2355 length);
2357 /* VLAN Header has been stripped by the system firmware and
2358 * needs to be inserted by the driver
2360 if (adapter->rx_vlan_header_insertion &&
2361 (flags & IBMVNIC_VLAN_STRIPPED))
2362 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2363 ntohs(next->rx_comp.vlan_tci));
2365 /* free the entry */
2366 next->rx_comp.first = 0;
2367 remove_buff_from_pool(adapter, rx_buff);
2369 skb_put(skb, length);
2370 skb->protocol = eth_type_trans(skb, netdev);
2371 skb_record_rx_queue(skb, scrq_num);
2373 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2374 flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2375 skb->ip_summed = CHECKSUM_UNNECESSARY;
2378 length = skb->len;
2379 napi_gro_receive(napi, skb); /* send it up */
2380 netdev->stats.rx_packets++;
2381 netdev->stats.rx_bytes += length;
2382 adapter->rx_stats_buffers[scrq_num].packets++;
2383 adapter->rx_stats_buffers[scrq_num].bytes += length;
2384 frames_processed++;
2387 if (adapter->state != VNIC_CLOSING)
2388 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2390 if (frames_processed < budget) {
2391 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2392 napi_complete_done(napi, frames_processed);
2393 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2394 napi_reschedule(napi)) {
2395 disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2396 goto restart_poll;
2399 return frames_processed;
2402 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2404 int rc, ret;
2406 adapter->fallback.mtu = adapter->req_mtu;
2407 adapter->fallback.rx_queues = adapter->req_rx_queues;
2408 adapter->fallback.tx_queues = adapter->req_tx_queues;
2409 adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2410 adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2412 reinit_completion(&adapter->reset_done);
2413 adapter->wait_for_reset = true;
2414 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2416 if (rc) {
2417 ret = rc;
2418 goto out;
2420 rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done, 60000);
2421 if (rc) {
2422 ret = -ENODEV;
2423 goto out;
2426 ret = 0;
2427 if (adapter->reset_done_rc) {
2428 ret = -EIO;
2429 adapter->desired.mtu = adapter->fallback.mtu;
2430 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2431 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2432 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2433 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2435 reinit_completion(&adapter->reset_done);
2436 adapter->wait_for_reset = true;
2437 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2438 if (rc) {
2439 ret = rc;
2440 goto out;
2442 rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done,
2443 60000);
2444 if (rc) {
2445 ret = -ENODEV;
2446 goto out;
2449 out:
2450 adapter->wait_for_reset = false;
2452 return ret;
2455 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2457 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2459 adapter->desired.mtu = new_mtu + ETH_HLEN;
2461 return wait_for_reset(adapter);
2464 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2465 struct net_device *dev,
2466 netdev_features_t features)
2468 /* Some backing hardware adapters can not
2469 * handle packets with a MSS less than 224
2470 * or with only one segment.
2472 if (skb_is_gso(skb)) {
2473 if (skb_shinfo(skb)->gso_size < 224 ||
2474 skb_shinfo(skb)->gso_segs == 1)
2475 features &= ~NETIF_F_GSO_MASK;
2478 return features;
2481 static const struct net_device_ops ibmvnic_netdev_ops = {
2482 .ndo_open = ibmvnic_open,
2483 .ndo_stop = ibmvnic_close,
2484 .ndo_start_xmit = ibmvnic_xmit,
2485 .ndo_set_rx_mode = ibmvnic_set_multi,
2486 .ndo_set_mac_address = ibmvnic_set_mac,
2487 .ndo_validate_addr = eth_validate_addr,
2488 .ndo_tx_timeout = ibmvnic_tx_timeout,
2489 .ndo_change_mtu = ibmvnic_change_mtu,
2490 .ndo_features_check = ibmvnic_features_check,
2493 /* ethtool functions */
2495 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2496 struct ethtool_link_ksettings *cmd)
2498 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2499 int rc;
2501 rc = send_query_phys_parms(adapter);
2502 if (rc) {
2503 adapter->speed = SPEED_UNKNOWN;
2504 adapter->duplex = DUPLEX_UNKNOWN;
2506 cmd->base.speed = adapter->speed;
2507 cmd->base.duplex = adapter->duplex;
2508 cmd->base.port = PORT_FIBRE;
2509 cmd->base.phy_address = 0;
2510 cmd->base.autoneg = AUTONEG_ENABLE;
2512 return 0;
2515 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2516 struct ethtool_drvinfo *info)
2518 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2520 strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2521 strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2522 strlcpy(info->fw_version, adapter->fw_version,
2523 sizeof(info->fw_version));
2526 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2528 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2530 return adapter->msg_enable;
2533 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2535 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2537 adapter->msg_enable = data;
2540 static u32 ibmvnic_get_link(struct net_device *netdev)
2542 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2544 /* Don't need to send a query because we request a logical link up at
2545 * init and then we wait for link state indications
2547 return adapter->logical_link_state;
2550 static void ibmvnic_get_ringparam(struct net_device *netdev,
2551 struct ethtool_ringparam *ring)
2553 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2555 if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2556 ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2557 ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2558 } else {
2559 ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2560 ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2562 ring->rx_mini_max_pending = 0;
2563 ring->rx_jumbo_max_pending = 0;
2564 ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2565 ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2566 ring->rx_mini_pending = 0;
2567 ring->rx_jumbo_pending = 0;
2570 static int ibmvnic_set_ringparam(struct net_device *netdev,
2571 struct ethtool_ringparam *ring)
2573 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2574 int ret;
2576 ret = 0;
2577 adapter->desired.rx_entries = ring->rx_pending;
2578 adapter->desired.tx_entries = ring->tx_pending;
2580 ret = wait_for_reset(adapter);
2582 if (!ret &&
2583 (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2584 adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2585 netdev_info(netdev,
2586 "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2587 ring->rx_pending, ring->tx_pending,
2588 adapter->req_rx_add_entries_per_subcrq,
2589 adapter->req_tx_entries_per_subcrq);
2590 return ret;
2593 static void ibmvnic_get_channels(struct net_device *netdev,
2594 struct ethtool_channels *channels)
2596 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2598 if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2599 channels->max_rx = adapter->max_rx_queues;
2600 channels->max_tx = adapter->max_tx_queues;
2601 } else {
2602 channels->max_rx = IBMVNIC_MAX_QUEUES;
2603 channels->max_tx = IBMVNIC_MAX_QUEUES;
2606 channels->max_other = 0;
2607 channels->max_combined = 0;
2608 channels->rx_count = adapter->req_rx_queues;
2609 channels->tx_count = adapter->req_tx_queues;
2610 channels->other_count = 0;
2611 channels->combined_count = 0;
2614 static int ibmvnic_set_channels(struct net_device *netdev,
2615 struct ethtool_channels *channels)
2617 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2618 int ret;
2620 ret = 0;
2621 adapter->desired.rx_queues = channels->rx_count;
2622 adapter->desired.tx_queues = channels->tx_count;
2624 ret = wait_for_reset(adapter);
2626 if (!ret &&
2627 (adapter->req_rx_queues != channels->rx_count ||
2628 adapter->req_tx_queues != channels->tx_count))
2629 netdev_info(netdev,
2630 "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2631 channels->rx_count, channels->tx_count,
2632 adapter->req_rx_queues, adapter->req_tx_queues);
2633 return ret;
2637 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2639 struct ibmvnic_adapter *adapter = netdev_priv(dev);
2640 int i;
2642 switch (stringset) {
2643 case ETH_SS_STATS:
2644 for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2645 i++, data += ETH_GSTRING_LEN)
2646 memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2648 for (i = 0; i < adapter->req_tx_queues; i++) {
2649 snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2650 data += ETH_GSTRING_LEN;
2652 snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2653 data += ETH_GSTRING_LEN;
2655 snprintf(data, ETH_GSTRING_LEN,
2656 "tx%d_dropped_packets", i);
2657 data += ETH_GSTRING_LEN;
2660 for (i = 0; i < adapter->req_rx_queues; i++) {
2661 snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2662 data += ETH_GSTRING_LEN;
2664 snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2665 data += ETH_GSTRING_LEN;
2667 snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2668 data += ETH_GSTRING_LEN;
2670 break;
2672 case ETH_SS_PRIV_FLAGS:
2673 for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2674 strcpy(data + i * ETH_GSTRING_LEN,
2675 ibmvnic_priv_flags[i]);
2676 break;
2677 default:
2678 return;
2682 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2684 struct ibmvnic_adapter *adapter = netdev_priv(dev);
2686 switch (sset) {
2687 case ETH_SS_STATS:
2688 return ARRAY_SIZE(ibmvnic_stats) +
2689 adapter->req_tx_queues * NUM_TX_STATS +
2690 adapter->req_rx_queues * NUM_RX_STATS;
2691 case ETH_SS_PRIV_FLAGS:
2692 return ARRAY_SIZE(ibmvnic_priv_flags);
2693 default:
2694 return -EOPNOTSUPP;
2698 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2699 struct ethtool_stats *stats, u64 *data)
2701 struct ibmvnic_adapter *adapter = netdev_priv(dev);
2702 union ibmvnic_crq crq;
2703 int i, j;
2704 int rc;
2706 memset(&crq, 0, sizeof(crq));
2707 crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2708 crq.request_statistics.cmd = REQUEST_STATISTICS;
2709 crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2710 crq.request_statistics.len =
2711 cpu_to_be32(sizeof(struct ibmvnic_statistics));
2713 /* Wait for data to be written */
2714 reinit_completion(&adapter->stats_done);
2715 rc = ibmvnic_send_crq(adapter, &crq);
2716 if (rc)
2717 return;
2718 rc = ibmvnic_wait_for_completion(adapter, &adapter->stats_done, 10000);
2719 if (rc)
2720 return;
2722 for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2723 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2724 ibmvnic_stats[i].offset));
2726 for (j = 0; j < adapter->req_tx_queues; j++) {
2727 data[i] = adapter->tx_stats_buffers[j].packets;
2728 i++;
2729 data[i] = adapter->tx_stats_buffers[j].bytes;
2730 i++;
2731 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2732 i++;
2735 for (j = 0; j < adapter->req_rx_queues; j++) {
2736 data[i] = adapter->rx_stats_buffers[j].packets;
2737 i++;
2738 data[i] = adapter->rx_stats_buffers[j].bytes;
2739 i++;
2740 data[i] = adapter->rx_stats_buffers[j].interrupts;
2741 i++;
2745 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2747 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2749 return adapter->priv_flags;
2752 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2754 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2755 bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2757 if (which_maxes)
2758 adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2759 else
2760 adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2762 return 0;
2764 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2765 .get_drvinfo = ibmvnic_get_drvinfo,
2766 .get_msglevel = ibmvnic_get_msglevel,
2767 .set_msglevel = ibmvnic_set_msglevel,
2768 .get_link = ibmvnic_get_link,
2769 .get_ringparam = ibmvnic_get_ringparam,
2770 .set_ringparam = ibmvnic_set_ringparam,
2771 .get_channels = ibmvnic_get_channels,
2772 .set_channels = ibmvnic_set_channels,
2773 .get_strings = ibmvnic_get_strings,
2774 .get_sset_count = ibmvnic_get_sset_count,
2775 .get_ethtool_stats = ibmvnic_get_ethtool_stats,
2776 .get_link_ksettings = ibmvnic_get_link_ksettings,
2777 .get_priv_flags = ibmvnic_get_priv_flags,
2778 .set_priv_flags = ibmvnic_set_priv_flags,
2781 /* Routines for managing CRQs/sCRQs */
2783 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2784 struct ibmvnic_sub_crq_queue *scrq)
2786 int rc;
2788 if (scrq->irq) {
2789 free_irq(scrq->irq, scrq);
2790 irq_dispose_mapping(scrq->irq);
2791 scrq->irq = 0;
2794 memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2795 atomic_set(&scrq->used, 0);
2796 scrq->cur = 0;
2798 rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2799 4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2800 return rc;
2803 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2805 int i, rc;
2807 for (i = 0; i < adapter->req_tx_queues; i++) {
2808 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2809 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2810 if (rc)
2811 return rc;
2814 for (i = 0; i < adapter->req_rx_queues; i++) {
2815 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2816 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2817 if (rc)
2818 return rc;
2821 return rc;
2824 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2825 struct ibmvnic_sub_crq_queue *scrq,
2826 bool do_h_free)
2828 struct device *dev = &adapter->vdev->dev;
2829 long rc;
2831 netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2833 if (do_h_free) {
2834 /* Close the sub-crqs */
2835 do {
2836 rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2837 adapter->vdev->unit_address,
2838 scrq->crq_num);
2839 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2841 if (rc) {
2842 netdev_err(adapter->netdev,
2843 "Failed to release sub-CRQ %16lx, rc = %ld\n",
2844 scrq->crq_num, rc);
2848 dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2849 DMA_BIDIRECTIONAL);
2850 free_pages((unsigned long)scrq->msgs, 2);
2851 kfree(scrq);
2854 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2855 *adapter)
2857 struct device *dev = &adapter->vdev->dev;
2858 struct ibmvnic_sub_crq_queue *scrq;
2859 int rc;
2861 scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2862 if (!scrq)
2863 return NULL;
2865 scrq->msgs =
2866 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2867 if (!scrq->msgs) {
2868 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2869 goto zero_page_failed;
2872 scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2873 DMA_BIDIRECTIONAL);
2874 if (dma_mapping_error(dev, scrq->msg_token)) {
2875 dev_warn(dev, "Couldn't map crq queue messages page\n");
2876 goto map_failed;
2879 rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2880 4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2882 if (rc == H_RESOURCE)
2883 rc = ibmvnic_reset_crq(adapter);
2885 if (rc == H_CLOSED) {
2886 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2887 } else if (rc) {
2888 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2889 goto reg_failed;
2892 scrq->adapter = adapter;
2893 scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2894 spin_lock_init(&scrq->lock);
2896 netdev_dbg(adapter->netdev,
2897 "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2898 scrq->crq_num, scrq->hw_irq, scrq->irq);
2900 return scrq;
2902 reg_failed:
2903 dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2904 DMA_BIDIRECTIONAL);
2905 map_failed:
2906 free_pages((unsigned long)scrq->msgs, 2);
2907 zero_page_failed:
2908 kfree(scrq);
2910 return NULL;
2913 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2915 int i;
2917 if (adapter->tx_scrq) {
2918 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2919 if (!adapter->tx_scrq[i])
2920 continue;
2922 netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2924 if (adapter->tx_scrq[i]->irq) {
2925 free_irq(adapter->tx_scrq[i]->irq,
2926 adapter->tx_scrq[i]);
2927 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2928 adapter->tx_scrq[i]->irq = 0;
2931 release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2932 do_h_free);
2935 kfree(adapter->tx_scrq);
2936 adapter->tx_scrq = NULL;
2937 adapter->num_active_tx_scrqs = 0;
2940 if (adapter->rx_scrq) {
2941 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2942 if (!adapter->rx_scrq[i])
2943 continue;
2945 netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2947 if (adapter->rx_scrq[i]->irq) {
2948 free_irq(adapter->rx_scrq[i]->irq,
2949 adapter->rx_scrq[i]);
2950 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2951 adapter->rx_scrq[i]->irq = 0;
2954 release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2955 do_h_free);
2958 kfree(adapter->rx_scrq);
2959 adapter->rx_scrq = NULL;
2960 adapter->num_active_rx_scrqs = 0;
2964 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2965 struct ibmvnic_sub_crq_queue *scrq)
2967 struct device *dev = &adapter->vdev->dev;
2968 unsigned long rc;
2970 rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2971 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2972 if (rc)
2973 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2974 scrq->hw_irq, rc);
2975 return rc;
2978 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2979 struct ibmvnic_sub_crq_queue *scrq)
2981 struct device *dev = &adapter->vdev->dev;
2982 unsigned long rc;
2984 if (scrq->hw_irq > 0x100000000ULL) {
2985 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2986 return 1;
2989 if (test_bit(0, &adapter->resetting) &&
2990 adapter->reset_reason == VNIC_RESET_MOBILITY) {
2991 u64 val = (0xff000000) | scrq->hw_irq;
2993 rc = plpar_hcall_norets(H_EOI, val);
2994 /* H_EOI would fail with rc = H_FUNCTION when running
2995 * in XIVE mode which is expected, but not an error.
2997 if (rc && (rc != H_FUNCTION))
2998 dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2999 val, rc);
3002 rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3003 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3004 if (rc)
3005 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
3006 scrq->hw_irq, rc);
3007 return rc;
3010 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
3011 struct ibmvnic_sub_crq_queue *scrq)
3013 struct device *dev = &adapter->vdev->dev;
3014 struct ibmvnic_tx_pool *tx_pool;
3015 struct ibmvnic_tx_buff *txbuff;
3016 union sub_crq *next;
3017 int index;
3018 int i, j;
3020 restart_loop:
3021 while (pending_scrq(adapter, scrq)) {
3022 unsigned int pool = scrq->pool_index;
3023 int num_entries = 0;
3025 next = ibmvnic_next_scrq(adapter, scrq);
3026 for (i = 0; i < next->tx_comp.num_comps; i++) {
3027 if (next->tx_comp.rcs[i]) {
3028 dev_err(dev, "tx error %x\n",
3029 next->tx_comp.rcs[i]);
3030 continue;
3032 index = be32_to_cpu(next->tx_comp.correlators[i]);
3033 if (index & IBMVNIC_TSO_POOL_MASK) {
3034 tx_pool = &adapter->tso_pool[pool];
3035 index &= ~IBMVNIC_TSO_POOL_MASK;
3036 } else {
3037 tx_pool = &adapter->tx_pool[pool];
3040 txbuff = &tx_pool->tx_buff[index];
3042 for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
3043 if (!txbuff->data_dma[j])
3044 continue;
3046 txbuff->data_dma[j] = 0;
3049 if (txbuff->last_frag) {
3050 dev_kfree_skb_any(txbuff->skb);
3051 txbuff->skb = NULL;
3054 num_entries += txbuff->num_entries;
3056 tx_pool->free_map[tx_pool->producer_index] = index;
3057 tx_pool->producer_index =
3058 (tx_pool->producer_index + 1) %
3059 tx_pool->num_buffers;
3061 /* remove tx_comp scrq*/
3062 next->tx_comp.first = 0;
3064 if (atomic_sub_return(num_entries, &scrq->used) <=
3065 (adapter->req_tx_entries_per_subcrq / 2) &&
3066 __netif_subqueue_stopped(adapter->netdev,
3067 scrq->pool_index)) {
3068 netif_wake_subqueue(adapter->netdev, scrq->pool_index);
3069 netdev_dbg(adapter->netdev, "Started queue %d\n",
3070 scrq->pool_index);
3074 enable_scrq_irq(adapter, scrq);
3076 if (pending_scrq(adapter, scrq)) {
3077 disable_scrq_irq(adapter, scrq);
3078 goto restart_loop;
3081 return 0;
3084 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
3086 struct ibmvnic_sub_crq_queue *scrq = instance;
3087 struct ibmvnic_adapter *adapter = scrq->adapter;
3089 disable_scrq_irq(adapter, scrq);
3090 ibmvnic_complete_tx(adapter, scrq);
3092 return IRQ_HANDLED;
3095 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
3097 struct ibmvnic_sub_crq_queue *scrq = instance;
3098 struct ibmvnic_adapter *adapter = scrq->adapter;
3100 /* When booting a kdump kernel we can hit pending interrupts
3101 * prior to completing driver initialization.
3103 if (unlikely(adapter->state != VNIC_OPEN))
3104 return IRQ_NONE;
3106 adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
3108 if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
3109 disable_scrq_irq(adapter, scrq);
3110 __napi_schedule(&adapter->napi[scrq->scrq_num]);
3113 return IRQ_HANDLED;
3116 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
3118 struct device *dev = &adapter->vdev->dev;
3119 struct ibmvnic_sub_crq_queue *scrq;
3120 int i = 0, j = 0;
3121 int rc = 0;
3123 for (i = 0; i < adapter->req_tx_queues; i++) {
3124 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
3126 scrq = adapter->tx_scrq[i];
3127 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3129 if (!scrq->irq) {
3130 rc = -EINVAL;
3131 dev_err(dev, "Error mapping irq\n");
3132 goto req_tx_irq_failed;
3135 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
3136 adapter->vdev->unit_address, i);
3137 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
3138 0, scrq->name, scrq);
3140 if (rc) {
3141 dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
3142 scrq->irq, rc);
3143 irq_dispose_mapping(scrq->irq);
3144 goto req_tx_irq_failed;
3148 for (i = 0; i < adapter->req_rx_queues; i++) {
3149 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
3151 scrq = adapter->rx_scrq[i];
3152 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3153 if (!scrq->irq) {
3154 rc = -EINVAL;
3155 dev_err(dev, "Error mapping irq\n");
3156 goto req_rx_irq_failed;
3158 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
3159 adapter->vdev->unit_address, i);
3160 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
3161 0, scrq->name, scrq);
3162 if (rc) {
3163 dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
3164 scrq->irq, rc);
3165 irq_dispose_mapping(scrq->irq);
3166 goto req_rx_irq_failed;
3169 return rc;
3171 req_rx_irq_failed:
3172 for (j = 0; j < i; j++) {
3173 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
3174 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
3176 i = adapter->req_tx_queues;
3177 req_tx_irq_failed:
3178 for (j = 0; j < i; j++) {
3179 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
3180 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
3182 release_sub_crqs(adapter, 1);
3183 return rc;
3186 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
3188 struct device *dev = &adapter->vdev->dev;
3189 struct ibmvnic_sub_crq_queue **allqueues;
3190 int registered_queues = 0;
3191 int total_queues;
3192 int more = 0;
3193 int i;
3195 total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
3197 allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
3198 if (!allqueues)
3199 return -1;
3201 for (i = 0; i < total_queues; i++) {
3202 allqueues[i] = init_sub_crq_queue(adapter);
3203 if (!allqueues[i]) {
3204 dev_warn(dev, "Couldn't allocate all sub-crqs\n");
3205 break;
3207 registered_queues++;
3210 /* Make sure we were able to register the minimum number of queues */
3211 if (registered_queues <
3212 adapter->min_tx_queues + adapter->min_rx_queues) {
3213 dev_err(dev, "Fatal: Couldn't init min number of sub-crqs\n");
3214 goto tx_failed;
3217 /* Distribute the failed allocated queues*/
3218 for (i = 0; i < total_queues - registered_queues + more ; i++) {
3219 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3220 switch (i % 3) {
3221 case 0:
3222 if (adapter->req_rx_queues > adapter->min_rx_queues)
3223 adapter->req_rx_queues--;
3224 else
3225 more++;
3226 break;
3227 case 1:
3228 if (adapter->req_tx_queues > adapter->min_tx_queues)
3229 adapter->req_tx_queues--;
3230 else
3231 more++;
3232 break;
3236 adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3237 sizeof(*adapter->tx_scrq), GFP_KERNEL);
3238 if (!adapter->tx_scrq)
3239 goto tx_failed;
3241 for (i = 0; i < adapter->req_tx_queues; i++) {
3242 adapter->tx_scrq[i] = allqueues[i];
3243 adapter->tx_scrq[i]->pool_index = i;
3244 adapter->num_active_tx_scrqs++;
3247 adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3248 sizeof(*adapter->rx_scrq), GFP_KERNEL);
3249 if (!adapter->rx_scrq)
3250 goto rx_failed;
3252 for (i = 0; i < adapter->req_rx_queues; i++) {
3253 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3254 adapter->rx_scrq[i]->scrq_num = i;
3255 adapter->num_active_rx_scrqs++;
3258 kfree(allqueues);
3259 return 0;
3261 rx_failed:
3262 kfree(adapter->tx_scrq);
3263 adapter->tx_scrq = NULL;
3264 tx_failed:
3265 for (i = 0; i < registered_queues; i++)
3266 release_sub_crq_queue(adapter, allqueues[i], 1);
3267 kfree(allqueues);
3268 return -1;
3271 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3273 struct device *dev = &adapter->vdev->dev;
3274 union ibmvnic_crq crq;
3275 int max_entries;
3277 if (!retry) {
3278 /* Sub-CRQ entries are 32 byte long */
3279 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3281 if (adapter->min_tx_entries_per_subcrq > entries_page ||
3282 adapter->min_rx_add_entries_per_subcrq > entries_page) {
3283 dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3284 return;
3287 if (adapter->desired.mtu)
3288 adapter->req_mtu = adapter->desired.mtu;
3289 else
3290 adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3292 if (!adapter->desired.tx_entries)
3293 adapter->desired.tx_entries =
3294 adapter->max_tx_entries_per_subcrq;
3295 if (!adapter->desired.rx_entries)
3296 adapter->desired.rx_entries =
3297 adapter->max_rx_add_entries_per_subcrq;
3299 max_entries = IBMVNIC_MAX_LTB_SIZE /
3300 (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3302 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3303 adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3304 adapter->desired.tx_entries = max_entries;
3307 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3308 adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3309 adapter->desired.rx_entries = max_entries;
3312 if (adapter->desired.tx_entries)
3313 adapter->req_tx_entries_per_subcrq =
3314 adapter->desired.tx_entries;
3315 else
3316 adapter->req_tx_entries_per_subcrq =
3317 adapter->max_tx_entries_per_subcrq;
3319 if (adapter->desired.rx_entries)
3320 adapter->req_rx_add_entries_per_subcrq =
3321 adapter->desired.rx_entries;
3322 else
3323 adapter->req_rx_add_entries_per_subcrq =
3324 adapter->max_rx_add_entries_per_subcrq;
3326 if (adapter->desired.tx_queues)
3327 adapter->req_tx_queues =
3328 adapter->desired.tx_queues;
3329 else
3330 adapter->req_tx_queues =
3331 adapter->opt_tx_comp_sub_queues;
3333 if (adapter->desired.rx_queues)
3334 adapter->req_rx_queues =
3335 adapter->desired.rx_queues;
3336 else
3337 adapter->req_rx_queues =
3338 adapter->opt_rx_comp_queues;
3340 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3343 memset(&crq, 0, sizeof(crq));
3344 crq.request_capability.first = IBMVNIC_CRQ_CMD;
3345 crq.request_capability.cmd = REQUEST_CAPABILITY;
3347 crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3348 crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3349 atomic_inc(&adapter->running_cap_crqs);
3350 ibmvnic_send_crq(adapter, &crq);
3352 crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3353 crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3354 atomic_inc(&adapter->running_cap_crqs);
3355 ibmvnic_send_crq(adapter, &crq);
3357 crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3358 crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3359 atomic_inc(&adapter->running_cap_crqs);
3360 ibmvnic_send_crq(adapter, &crq);
3362 crq.request_capability.capability =
3363 cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3364 crq.request_capability.number =
3365 cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3366 atomic_inc(&adapter->running_cap_crqs);
3367 ibmvnic_send_crq(adapter, &crq);
3369 crq.request_capability.capability =
3370 cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3371 crq.request_capability.number =
3372 cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3373 atomic_inc(&adapter->running_cap_crqs);
3374 ibmvnic_send_crq(adapter, &crq);
3376 crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3377 crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3378 atomic_inc(&adapter->running_cap_crqs);
3379 ibmvnic_send_crq(adapter, &crq);
3381 if (adapter->netdev->flags & IFF_PROMISC) {
3382 if (adapter->promisc_supported) {
3383 crq.request_capability.capability =
3384 cpu_to_be16(PROMISC_REQUESTED);
3385 crq.request_capability.number = cpu_to_be64(1);
3386 atomic_inc(&adapter->running_cap_crqs);
3387 ibmvnic_send_crq(adapter, &crq);
3389 } else {
3390 crq.request_capability.capability =
3391 cpu_to_be16(PROMISC_REQUESTED);
3392 crq.request_capability.number = cpu_to_be64(0);
3393 atomic_inc(&adapter->running_cap_crqs);
3394 ibmvnic_send_crq(adapter, &crq);
3398 static int pending_scrq(struct ibmvnic_adapter *adapter,
3399 struct ibmvnic_sub_crq_queue *scrq)
3401 union sub_crq *entry = &scrq->msgs[scrq->cur];
3403 if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3404 return 1;
3405 else
3406 return 0;
3409 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3410 struct ibmvnic_sub_crq_queue *scrq)
3412 union sub_crq *entry;
3413 unsigned long flags;
3415 spin_lock_irqsave(&scrq->lock, flags);
3416 entry = &scrq->msgs[scrq->cur];
3417 if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3418 if (++scrq->cur == scrq->size)
3419 scrq->cur = 0;
3420 } else {
3421 entry = NULL;
3423 spin_unlock_irqrestore(&scrq->lock, flags);
3425 return entry;
3428 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3430 struct ibmvnic_crq_queue *queue = &adapter->crq;
3431 union ibmvnic_crq *crq;
3433 crq = &queue->msgs[queue->cur];
3434 if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3435 if (++queue->cur == queue->size)
3436 queue->cur = 0;
3437 } else {
3438 crq = NULL;
3441 return crq;
3444 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3446 switch (rc) {
3447 case H_PARAMETER:
3448 dev_warn_ratelimited(dev,
3449 "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3450 func, rc);
3451 break;
3452 case H_CLOSED:
3453 dev_warn_ratelimited(dev,
3454 "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3455 func, rc);
3456 break;
3457 default:
3458 dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3459 break;
3463 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3464 union sub_crq *sub_crq)
3466 unsigned int ua = adapter->vdev->unit_address;
3467 struct device *dev = &adapter->vdev->dev;
3468 u64 *u64_crq = (u64 *)sub_crq;
3469 int rc;
3471 netdev_dbg(adapter->netdev,
3472 "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3473 (unsigned long int)cpu_to_be64(remote_handle),
3474 (unsigned long int)cpu_to_be64(u64_crq[0]),
3475 (unsigned long int)cpu_to_be64(u64_crq[1]),
3476 (unsigned long int)cpu_to_be64(u64_crq[2]),
3477 (unsigned long int)cpu_to_be64(u64_crq[3]));
3479 /* Make sure the hypervisor sees the complete request */
3480 mb();
3482 rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3483 cpu_to_be64(remote_handle),
3484 cpu_to_be64(u64_crq[0]),
3485 cpu_to_be64(u64_crq[1]),
3486 cpu_to_be64(u64_crq[2]),
3487 cpu_to_be64(u64_crq[3]));
3489 if (rc)
3490 print_subcrq_error(dev, rc, __func__);
3492 return rc;
3495 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3496 u64 remote_handle, u64 ioba, u64 num_entries)
3498 unsigned int ua = adapter->vdev->unit_address;
3499 struct device *dev = &adapter->vdev->dev;
3500 int rc;
3502 /* Make sure the hypervisor sees the complete request */
3503 mb();
3504 rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3505 cpu_to_be64(remote_handle),
3506 ioba, num_entries);
3508 if (rc)
3509 print_subcrq_error(dev, rc, __func__);
3511 return rc;
3514 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3515 union ibmvnic_crq *crq)
3517 unsigned int ua = adapter->vdev->unit_address;
3518 struct device *dev = &adapter->vdev->dev;
3519 u64 *u64_crq = (u64 *)crq;
3520 int rc;
3522 netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3523 (unsigned long int)cpu_to_be64(u64_crq[0]),
3524 (unsigned long int)cpu_to_be64(u64_crq[1]));
3526 if (!adapter->crq.active &&
3527 crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3528 dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3529 return -EINVAL;
3532 /* Make sure the hypervisor sees the complete request */
3533 mb();
3535 rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3536 cpu_to_be64(u64_crq[0]),
3537 cpu_to_be64(u64_crq[1]));
3539 if (rc) {
3540 if (rc == H_CLOSED) {
3541 dev_warn(dev, "CRQ Queue closed\n");
3542 if (test_bit(0, &adapter->resetting))
3543 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3546 dev_warn(dev, "Send error (rc=%d)\n", rc);
3549 return rc;
3552 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3554 union ibmvnic_crq crq;
3556 memset(&crq, 0, sizeof(crq));
3557 crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3558 crq.generic.cmd = IBMVNIC_CRQ_INIT;
3559 netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3561 return ibmvnic_send_crq(adapter, &crq);
3564 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3566 union ibmvnic_crq crq;
3568 memset(&crq, 0, sizeof(crq));
3569 crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3570 crq.version_exchange.cmd = VERSION_EXCHANGE;
3571 crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3573 return ibmvnic_send_crq(adapter, &crq);
3576 struct vnic_login_client_data {
3577 u8 type;
3578 __be16 len;
3579 char name[];
3580 } __packed;
3582 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3584 int len;
3586 /* Calculate the amount of buffer space needed for the
3587 * vnic client data in the login buffer. There are four entries,
3588 * OS name, LPAR name, device name, and a null last entry.
3590 len = 4 * sizeof(struct vnic_login_client_data);
3591 len += 6; /* "Linux" plus NULL */
3592 len += strlen(utsname()->nodename) + 1;
3593 len += strlen(adapter->netdev->name) + 1;
3595 return len;
3598 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3599 struct vnic_login_client_data *vlcd)
3601 const char *os_name = "Linux";
3602 int len;
3604 /* Type 1 - LPAR OS */
3605 vlcd->type = 1;
3606 len = strlen(os_name) + 1;
3607 vlcd->len = cpu_to_be16(len);
3608 strncpy(vlcd->name, os_name, len);
3609 vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3611 /* Type 2 - LPAR name */
3612 vlcd->type = 2;
3613 len = strlen(utsname()->nodename) + 1;
3614 vlcd->len = cpu_to_be16(len);
3615 strncpy(vlcd->name, utsname()->nodename, len);
3616 vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3618 /* Type 3 - device name */
3619 vlcd->type = 3;
3620 len = strlen(adapter->netdev->name) + 1;
3621 vlcd->len = cpu_to_be16(len);
3622 strncpy(vlcd->name, adapter->netdev->name, len);
3625 static int send_login(struct ibmvnic_adapter *adapter)
3627 struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3628 struct ibmvnic_login_buffer *login_buffer;
3629 struct device *dev = &adapter->vdev->dev;
3630 dma_addr_t rsp_buffer_token;
3631 dma_addr_t buffer_token;
3632 size_t rsp_buffer_size;
3633 union ibmvnic_crq crq;
3634 size_t buffer_size;
3635 __be64 *tx_list_p;
3636 __be64 *rx_list_p;
3637 int client_data_len;
3638 struct vnic_login_client_data *vlcd;
3639 int i;
3641 if (!adapter->tx_scrq || !adapter->rx_scrq) {
3642 netdev_err(adapter->netdev,
3643 "RX or TX queues are not allocated, device login failed\n");
3644 return -1;
3647 release_login_rsp_buffer(adapter);
3648 client_data_len = vnic_client_data_len(adapter);
3650 buffer_size =
3651 sizeof(struct ibmvnic_login_buffer) +
3652 sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3653 client_data_len;
3655 login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3656 if (!login_buffer)
3657 goto buf_alloc_failed;
3659 buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3660 DMA_TO_DEVICE);
3661 if (dma_mapping_error(dev, buffer_token)) {
3662 dev_err(dev, "Couldn't map login buffer\n");
3663 goto buf_map_failed;
3666 rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3667 sizeof(u64) * adapter->req_tx_queues +
3668 sizeof(u64) * adapter->req_rx_queues +
3669 sizeof(u64) * adapter->req_rx_queues +
3670 sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3672 login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3673 if (!login_rsp_buffer)
3674 goto buf_rsp_alloc_failed;
3676 rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3677 rsp_buffer_size, DMA_FROM_DEVICE);
3678 if (dma_mapping_error(dev, rsp_buffer_token)) {
3679 dev_err(dev, "Couldn't map login rsp buffer\n");
3680 goto buf_rsp_map_failed;
3683 adapter->login_buf = login_buffer;
3684 adapter->login_buf_token = buffer_token;
3685 adapter->login_buf_sz = buffer_size;
3686 adapter->login_rsp_buf = login_rsp_buffer;
3687 adapter->login_rsp_buf_token = rsp_buffer_token;
3688 adapter->login_rsp_buf_sz = rsp_buffer_size;
3690 login_buffer->len = cpu_to_be32(buffer_size);
3691 login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3692 login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3693 login_buffer->off_txcomp_subcrqs =
3694 cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3695 login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3696 login_buffer->off_rxcomp_subcrqs =
3697 cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3698 sizeof(u64) * adapter->req_tx_queues);
3699 login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3700 login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3702 tx_list_p = (__be64 *)((char *)login_buffer +
3703 sizeof(struct ibmvnic_login_buffer));
3704 rx_list_p = (__be64 *)((char *)login_buffer +
3705 sizeof(struct ibmvnic_login_buffer) +
3706 sizeof(u64) * adapter->req_tx_queues);
3708 for (i = 0; i < adapter->req_tx_queues; i++) {
3709 if (adapter->tx_scrq[i]) {
3710 tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3711 crq_num);
3715 for (i = 0; i < adapter->req_rx_queues; i++) {
3716 if (adapter->rx_scrq[i]) {
3717 rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3718 crq_num);
3722 /* Insert vNIC login client data */
3723 vlcd = (struct vnic_login_client_data *)
3724 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3725 login_buffer->client_data_offset =
3726 cpu_to_be32((char *)vlcd - (char *)login_buffer);
3727 login_buffer->client_data_len = cpu_to_be32(client_data_len);
3729 vnic_add_client_data(adapter, vlcd);
3731 netdev_dbg(adapter->netdev, "Login Buffer:\n");
3732 for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3733 netdev_dbg(adapter->netdev, "%016lx\n",
3734 ((unsigned long int *)(adapter->login_buf))[i]);
3737 memset(&crq, 0, sizeof(crq));
3738 crq.login.first = IBMVNIC_CRQ_CMD;
3739 crq.login.cmd = LOGIN;
3740 crq.login.ioba = cpu_to_be32(buffer_token);
3741 crq.login.len = cpu_to_be32(buffer_size);
3742 ibmvnic_send_crq(adapter, &crq);
3744 return 0;
3746 buf_rsp_map_failed:
3747 kfree(login_rsp_buffer);
3748 buf_rsp_alloc_failed:
3749 dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3750 buf_map_failed:
3751 kfree(login_buffer);
3752 buf_alloc_failed:
3753 return -1;
3756 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3757 u32 len, u8 map_id)
3759 union ibmvnic_crq crq;
3761 memset(&crq, 0, sizeof(crq));
3762 crq.request_map.first = IBMVNIC_CRQ_CMD;
3763 crq.request_map.cmd = REQUEST_MAP;
3764 crq.request_map.map_id = map_id;
3765 crq.request_map.ioba = cpu_to_be32(addr);
3766 crq.request_map.len = cpu_to_be32(len);
3767 return ibmvnic_send_crq(adapter, &crq);
3770 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3772 union ibmvnic_crq crq;
3774 memset(&crq, 0, sizeof(crq));
3775 crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3776 crq.request_unmap.cmd = REQUEST_UNMAP;
3777 crq.request_unmap.map_id = map_id;
3778 return ibmvnic_send_crq(adapter, &crq);
3781 static void send_map_query(struct ibmvnic_adapter *adapter)
3783 union ibmvnic_crq crq;
3785 memset(&crq, 0, sizeof(crq));
3786 crq.query_map.first = IBMVNIC_CRQ_CMD;
3787 crq.query_map.cmd = QUERY_MAP;
3788 ibmvnic_send_crq(adapter, &crq);
3791 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3792 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3794 union ibmvnic_crq crq;
3796 atomic_set(&adapter->running_cap_crqs, 0);
3797 memset(&crq, 0, sizeof(crq));
3798 crq.query_capability.first = IBMVNIC_CRQ_CMD;
3799 crq.query_capability.cmd = QUERY_CAPABILITY;
3801 crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3802 atomic_inc(&adapter->running_cap_crqs);
3803 ibmvnic_send_crq(adapter, &crq);
3805 crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3806 atomic_inc(&adapter->running_cap_crqs);
3807 ibmvnic_send_crq(adapter, &crq);
3809 crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3810 atomic_inc(&adapter->running_cap_crqs);
3811 ibmvnic_send_crq(adapter, &crq);
3813 crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3814 atomic_inc(&adapter->running_cap_crqs);
3815 ibmvnic_send_crq(adapter, &crq);
3817 crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3818 atomic_inc(&adapter->running_cap_crqs);
3819 ibmvnic_send_crq(adapter, &crq);
3821 crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3822 atomic_inc(&adapter->running_cap_crqs);
3823 ibmvnic_send_crq(adapter, &crq);
3825 crq.query_capability.capability =
3826 cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3827 atomic_inc(&adapter->running_cap_crqs);
3828 ibmvnic_send_crq(adapter, &crq);
3830 crq.query_capability.capability =
3831 cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3832 atomic_inc(&adapter->running_cap_crqs);
3833 ibmvnic_send_crq(adapter, &crq);
3835 crq.query_capability.capability =
3836 cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3837 atomic_inc(&adapter->running_cap_crqs);
3838 ibmvnic_send_crq(adapter, &crq);
3840 crq.query_capability.capability =
3841 cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3842 atomic_inc(&adapter->running_cap_crqs);
3843 ibmvnic_send_crq(adapter, &crq);
3845 crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3846 atomic_inc(&adapter->running_cap_crqs);
3847 ibmvnic_send_crq(adapter, &crq);
3849 crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3850 atomic_inc(&adapter->running_cap_crqs);
3851 ibmvnic_send_crq(adapter, &crq);
3853 crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3854 atomic_inc(&adapter->running_cap_crqs);
3855 ibmvnic_send_crq(adapter, &crq);
3857 crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3858 atomic_inc(&adapter->running_cap_crqs);
3859 ibmvnic_send_crq(adapter, &crq);
3861 crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3862 atomic_inc(&adapter->running_cap_crqs);
3863 ibmvnic_send_crq(adapter, &crq);
3865 crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3866 atomic_inc(&adapter->running_cap_crqs);
3867 ibmvnic_send_crq(adapter, &crq);
3869 crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3870 atomic_inc(&adapter->running_cap_crqs);
3871 ibmvnic_send_crq(adapter, &crq);
3873 crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3874 atomic_inc(&adapter->running_cap_crqs);
3875 ibmvnic_send_crq(adapter, &crq);
3877 crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3878 atomic_inc(&adapter->running_cap_crqs);
3879 ibmvnic_send_crq(adapter, &crq);
3881 crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3882 atomic_inc(&adapter->running_cap_crqs);
3883 ibmvnic_send_crq(adapter, &crq);
3885 crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3886 atomic_inc(&adapter->running_cap_crqs);
3887 ibmvnic_send_crq(adapter, &crq);
3889 crq.query_capability.capability =
3890 cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3891 atomic_inc(&adapter->running_cap_crqs);
3892 ibmvnic_send_crq(adapter, &crq);
3894 crq.query_capability.capability =
3895 cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3896 atomic_inc(&adapter->running_cap_crqs);
3897 ibmvnic_send_crq(adapter, &crq);
3899 crq.query_capability.capability =
3900 cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3901 atomic_inc(&adapter->running_cap_crqs);
3902 ibmvnic_send_crq(adapter, &crq);
3904 crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3905 atomic_inc(&adapter->running_cap_crqs);
3906 ibmvnic_send_crq(adapter, &crq);
3909 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3910 struct ibmvnic_adapter *adapter)
3912 struct device *dev = &adapter->vdev->dev;
3914 if (crq->get_vpd_size_rsp.rc.code) {
3915 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3916 crq->get_vpd_size_rsp.rc.code);
3917 complete(&adapter->fw_done);
3918 return;
3921 adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3922 complete(&adapter->fw_done);
3925 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3926 struct ibmvnic_adapter *adapter)
3928 struct device *dev = &adapter->vdev->dev;
3929 unsigned char *substr = NULL;
3930 u8 fw_level_len = 0;
3932 memset(adapter->fw_version, 0, 32);
3934 dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3935 DMA_FROM_DEVICE);
3937 if (crq->get_vpd_rsp.rc.code) {
3938 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3939 crq->get_vpd_rsp.rc.code);
3940 goto complete;
3943 /* get the position of the firmware version info
3944 * located after the ASCII 'RM' substring in the buffer
3946 substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3947 if (!substr) {
3948 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3949 goto complete;
3952 /* get length of firmware level ASCII substring */
3953 if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3954 fw_level_len = *(substr + 2);
3955 } else {
3956 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3957 goto complete;
3960 /* copy firmware version string from vpd into adapter */
3961 if ((substr + 3 + fw_level_len) <
3962 (adapter->vpd->buff + adapter->vpd->len)) {
3963 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3964 } else {
3965 dev_info(dev, "FW substr extrapolated VPD buff\n");
3968 complete:
3969 if (adapter->fw_version[0] == '\0')
3970 strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3971 complete(&adapter->fw_done);
3974 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3976 struct device *dev = &adapter->vdev->dev;
3977 struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3978 netdev_features_t old_hw_features = 0;
3979 union ibmvnic_crq crq;
3980 int i;
3982 dma_unmap_single(dev, adapter->ip_offload_tok,
3983 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3985 netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3986 for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3987 netdev_dbg(adapter->netdev, "%016lx\n",
3988 ((unsigned long int *)(buf))[i]);
3990 netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3991 netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3992 netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3993 buf->tcp_ipv4_chksum);
3994 netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3995 buf->tcp_ipv6_chksum);
3996 netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3997 buf->udp_ipv4_chksum);
3998 netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3999 buf->udp_ipv6_chksum);
4000 netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
4001 buf->large_tx_ipv4);
4002 netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
4003 buf->large_tx_ipv6);
4004 netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
4005 buf->large_rx_ipv4);
4006 netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
4007 buf->large_rx_ipv6);
4008 netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
4009 buf->max_ipv4_header_size);
4010 netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
4011 buf->max_ipv6_header_size);
4012 netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
4013 buf->max_tcp_header_size);
4014 netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
4015 buf->max_udp_header_size);
4016 netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
4017 buf->max_large_tx_size);
4018 netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
4019 buf->max_large_rx_size);
4020 netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
4021 buf->ipv6_extension_header);
4022 netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
4023 buf->tcp_pseudosum_req);
4024 netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
4025 buf->num_ipv6_ext_headers);
4026 netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
4027 buf->off_ipv6_ext_headers);
4029 adapter->ip_offload_ctrl_tok =
4030 dma_map_single(dev, &adapter->ip_offload_ctrl,
4031 sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
4033 if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
4034 dev_err(dev, "Couldn't map ip offload control buffer\n");
4035 return;
4038 adapter->ip_offload_ctrl.len =
4039 cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4040 adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
4041 adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
4042 adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
4043 adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
4044 adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
4045 adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
4046 adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
4047 adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
4048 adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
4050 /* large_rx disabled for now, additional features needed */
4051 adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
4052 adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
4054 if (adapter->state != VNIC_PROBING) {
4055 old_hw_features = adapter->netdev->hw_features;
4056 adapter->netdev->hw_features = 0;
4059 adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
4061 if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
4062 adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
4064 if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
4065 adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
4067 if ((adapter->netdev->features &
4068 (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
4069 adapter->netdev->hw_features |= NETIF_F_RXCSUM;
4071 if (buf->large_tx_ipv4)
4072 adapter->netdev->hw_features |= NETIF_F_TSO;
4073 if (buf->large_tx_ipv6)
4074 adapter->netdev->hw_features |= NETIF_F_TSO6;
4076 if (adapter->state == VNIC_PROBING) {
4077 adapter->netdev->features |= adapter->netdev->hw_features;
4078 } else if (old_hw_features != adapter->netdev->hw_features) {
4079 netdev_features_t tmp = 0;
4081 /* disable features no longer supported */
4082 adapter->netdev->features &= adapter->netdev->hw_features;
4083 /* turn on features now supported if previously enabled */
4084 tmp = (old_hw_features ^ adapter->netdev->hw_features) &
4085 adapter->netdev->hw_features;
4086 adapter->netdev->features |=
4087 tmp & adapter->netdev->wanted_features;
4090 memset(&crq, 0, sizeof(crq));
4091 crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
4092 crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
4093 crq.control_ip_offload.len =
4094 cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4095 crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
4096 ibmvnic_send_crq(adapter, &crq);
4099 static const char *ibmvnic_fw_err_cause(u16 cause)
4101 switch (cause) {
4102 case ADAPTER_PROBLEM:
4103 return "adapter problem";
4104 case BUS_PROBLEM:
4105 return "bus problem";
4106 case FW_PROBLEM:
4107 return "firmware problem";
4108 case DD_PROBLEM:
4109 return "device driver problem";
4110 case EEH_RECOVERY:
4111 return "EEH recovery";
4112 case FW_UPDATED:
4113 return "firmware updated";
4114 case LOW_MEMORY:
4115 return "low Memory";
4116 default:
4117 return "unknown";
4121 static void handle_error_indication(union ibmvnic_crq *crq,
4122 struct ibmvnic_adapter *adapter)
4124 struct device *dev = &adapter->vdev->dev;
4125 u16 cause;
4127 cause = be16_to_cpu(crq->error_indication.error_cause);
4129 dev_warn_ratelimited(dev,
4130 "Firmware reports %serror, cause: %s. Starting recovery...\n",
4131 crq->error_indication.flags
4132 & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
4133 ibmvnic_fw_err_cause(cause));
4135 if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
4136 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4137 else
4138 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
4141 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
4142 struct ibmvnic_adapter *adapter)
4144 struct net_device *netdev = adapter->netdev;
4145 struct device *dev = &adapter->vdev->dev;
4146 long rc;
4148 rc = crq->change_mac_addr_rsp.rc.code;
4149 if (rc) {
4150 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
4151 goto out;
4153 ether_addr_copy(netdev->dev_addr,
4154 &crq->change_mac_addr_rsp.mac_addr[0]);
4155 out:
4156 complete(&adapter->fw_done);
4157 return rc;
4160 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
4161 struct ibmvnic_adapter *adapter)
4163 struct device *dev = &adapter->vdev->dev;
4164 u64 *req_value;
4165 char *name;
4167 atomic_dec(&adapter->running_cap_crqs);
4168 switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
4169 case REQ_TX_QUEUES:
4170 req_value = &adapter->req_tx_queues;
4171 name = "tx";
4172 break;
4173 case REQ_RX_QUEUES:
4174 req_value = &adapter->req_rx_queues;
4175 name = "rx";
4176 break;
4177 case REQ_RX_ADD_QUEUES:
4178 req_value = &adapter->req_rx_add_queues;
4179 name = "rx_add";
4180 break;
4181 case REQ_TX_ENTRIES_PER_SUBCRQ:
4182 req_value = &adapter->req_tx_entries_per_subcrq;
4183 name = "tx_entries_per_subcrq";
4184 break;
4185 case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
4186 req_value = &adapter->req_rx_add_entries_per_subcrq;
4187 name = "rx_add_entries_per_subcrq";
4188 break;
4189 case REQ_MTU:
4190 req_value = &adapter->req_mtu;
4191 name = "mtu";
4192 break;
4193 case PROMISC_REQUESTED:
4194 req_value = &adapter->promisc;
4195 name = "promisc";
4196 break;
4197 default:
4198 dev_err(dev, "Got invalid cap request rsp %d\n",
4199 crq->request_capability.capability);
4200 return;
4203 switch (crq->request_capability_rsp.rc.code) {
4204 case SUCCESS:
4205 break;
4206 case PARTIALSUCCESS:
4207 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
4208 *req_value,
4209 (long int)be64_to_cpu(crq->request_capability_rsp.
4210 number), name);
4212 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
4213 REQ_MTU) {
4214 pr_err("mtu of %llu is not supported. Reverting.\n",
4215 *req_value);
4216 *req_value = adapter->fallback.mtu;
4217 } else {
4218 *req_value =
4219 be64_to_cpu(crq->request_capability_rsp.number);
4222 ibmvnic_send_req_caps(adapter, 1);
4223 return;
4224 default:
4225 dev_err(dev, "Error %d in request cap rsp\n",
4226 crq->request_capability_rsp.rc.code);
4227 return;
4230 /* Done receiving requested capabilities, query IP offload support */
4231 if (atomic_read(&adapter->running_cap_crqs) == 0) {
4232 union ibmvnic_crq newcrq;
4233 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4234 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
4235 &adapter->ip_offload_buf;
4237 adapter->wait_capability = false;
4238 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
4239 buf_sz,
4240 DMA_FROM_DEVICE);
4242 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4243 if (!firmware_has_feature(FW_FEATURE_CMO))
4244 dev_err(dev, "Couldn't map offload buffer\n");
4245 return;
4248 memset(&newcrq, 0, sizeof(newcrq));
4249 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4250 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4251 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
4252 newcrq.query_ip_offload.ioba =
4253 cpu_to_be32(adapter->ip_offload_tok);
4255 ibmvnic_send_crq(adapter, &newcrq);
4259 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4260 struct ibmvnic_adapter *adapter)
4262 struct device *dev = &adapter->vdev->dev;
4263 struct net_device *netdev = adapter->netdev;
4264 struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4265 struct ibmvnic_login_buffer *login = adapter->login_buf;
4266 int i;
4268 dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4269 DMA_TO_DEVICE);
4270 dma_unmap_single(dev, adapter->login_rsp_buf_token,
4271 adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4273 /* If the number of queues requested can't be allocated by the
4274 * server, the login response will return with code 1. We will need
4275 * to resend the login buffer with fewer queues requested.
4277 if (login_rsp_crq->generic.rc.code) {
4278 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4279 complete(&adapter->init_done);
4280 return 0;
4283 netdev->mtu = adapter->req_mtu - ETH_HLEN;
4285 netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4286 for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4287 netdev_dbg(adapter->netdev, "%016lx\n",
4288 ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4291 /* Sanity checks */
4292 if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4293 (be32_to_cpu(login->num_rxcomp_subcrqs) *
4294 adapter->req_rx_add_queues !=
4295 be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4296 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4297 ibmvnic_remove(adapter->vdev);
4298 return -EIO;
4300 release_login_buffer(adapter);
4301 complete(&adapter->init_done);
4303 return 0;
4306 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4307 struct ibmvnic_adapter *adapter)
4309 struct device *dev = &adapter->vdev->dev;
4310 long rc;
4312 rc = crq->request_unmap_rsp.rc.code;
4313 if (rc)
4314 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4317 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4318 struct ibmvnic_adapter *adapter)
4320 struct net_device *netdev = adapter->netdev;
4321 struct device *dev = &adapter->vdev->dev;
4322 long rc;
4324 rc = crq->query_map_rsp.rc.code;
4325 if (rc) {
4326 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4327 return;
4329 netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4330 crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4331 crq->query_map_rsp.free_pages);
4334 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4335 struct ibmvnic_adapter *adapter)
4337 struct net_device *netdev = adapter->netdev;
4338 struct device *dev = &adapter->vdev->dev;
4339 long rc;
4341 atomic_dec(&adapter->running_cap_crqs);
4342 netdev_dbg(netdev, "Outstanding queries: %d\n",
4343 atomic_read(&adapter->running_cap_crqs));
4344 rc = crq->query_capability.rc.code;
4345 if (rc) {
4346 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4347 goto out;
4350 switch (be16_to_cpu(crq->query_capability.capability)) {
4351 case MIN_TX_QUEUES:
4352 adapter->min_tx_queues =
4353 be64_to_cpu(crq->query_capability.number);
4354 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4355 adapter->min_tx_queues);
4356 break;
4357 case MIN_RX_QUEUES:
4358 adapter->min_rx_queues =
4359 be64_to_cpu(crq->query_capability.number);
4360 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4361 adapter->min_rx_queues);
4362 break;
4363 case MIN_RX_ADD_QUEUES:
4364 adapter->min_rx_add_queues =
4365 be64_to_cpu(crq->query_capability.number);
4366 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4367 adapter->min_rx_add_queues);
4368 break;
4369 case MAX_TX_QUEUES:
4370 adapter->max_tx_queues =
4371 be64_to_cpu(crq->query_capability.number);
4372 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4373 adapter->max_tx_queues);
4374 break;
4375 case MAX_RX_QUEUES:
4376 adapter->max_rx_queues =
4377 be64_to_cpu(crq->query_capability.number);
4378 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4379 adapter->max_rx_queues);
4380 break;
4381 case MAX_RX_ADD_QUEUES:
4382 adapter->max_rx_add_queues =
4383 be64_to_cpu(crq->query_capability.number);
4384 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4385 adapter->max_rx_add_queues);
4386 break;
4387 case MIN_TX_ENTRIES_PER_SUBCRQ:
4388 adapter->min_tx_entries_per_subcrq =
4389 be64_to_cpu(crq->query_capability.number);
4390 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4391 adapter->min_tx_entries_per_subcrq);
4392 break;
4393 case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4394 adapter->min_rx_add_entries_per_subcrq =
4395 be64_to_cpu(crq->query_capability.number);
4396 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4397 adapter->min_rx_add_entries_per_subcrq);
4398 break;
4399 case MAX_TX_ENTRIES_PER_SUBCRQ:
4400 adapter->max_tx_entries_per_subcrq =
4401 be64_to_cpu(crq->query_capability.number);
4402 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4403 adapter->max_tx_entries_per_subcrq);
4404 break;
4405 case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4406 adapter->max_rx_add_entries_per_subcrq =
4407 be64_to_cpu(crq->query_capability.number);
4408 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4409 adapter->max_rx_add_entries_per_subcrq);
4410 break;
4411 case TCP_IP_OFFLOAD:
4412 adapter->tcp_ip_offload =
4413 be64_to_cpu(crq->query_capability.number);
4414 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4415 adapter->tcp_ip_offload);
4416 break;
4417 case PROMISC_SUPPORTED:
4418 adapter->promisc_supported =
4419 be64_to_cpu(crq->query_capability.number);
4420 netdev_dbg(netdev, "promisc_supported = %lld\n",
4421 adapter->promisc_supported);
4422 break;
4423 case MIN_MTU:
4424 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4425 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4426 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4427 break;
4428 case MAX_MTU:
4429 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4430 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4431 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4432 break;
4433 case MAX_MULTICAST_FILTERS:
4434 adapter->max_multicast_filters =
4435 be64_to_cpu(crq->query_capability.number);
4436 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4437 adapter->max_multicast_filters);
4438 break;
4439 case VLAN_HEADER_INSERTION:
4440 adapter->vlan_header_insertion =
4441 be64_to_cpu(crq->query_capability.number);
4442 if (adapter->vlan_header_insertion)
4443 netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4444 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4445 adapter->vlan_header_insertion);
4446 break;
4447 case RX_VLAN_HEADER_INSERTION:
4448 adapter->rx_vlan_header_insertion =
4449 be64_to_cpu(crq->query_capability.number);
4450 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4451 adapter->rx_vlan_header_insertion);
4452 break;
4453 case MAX_TX_SG_ENTRIES:
4454 adapter->max_tx_sg_entries =
4455 be64_to_cpu(crq->query_capability.number);
4456 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4457 adapter->max_tx_sg_entries);
4458 break;
4459 case RX_SG_SUPPORTED:
4460 adapter->rx_sg_supported =
4461 be64_to_cpu(crq->query_capability.number);
4462 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4463 adapter->rx_sg_supported);
4464 break;
4465 case OPT_TX_COMP_SUB_QUEUES:
4466 adapter->opt_tx_comp_sub_queues =
4467 be64_to_cpu(crq->query_capability.number);
4468 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4469 adapter->opt_tx_comp_sub_queues);
4470 break;
4471 case OPT_RX_COMP_QUEUES:
4472 adapter->opt_rx_comp_queues =
4473 be64_to_cpu(crq->query_capability.number);
4474 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4475 adapter->opt_rx_comp_queues);
4476 break;
4477 case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4478 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4479 be64_to_cpu(crq->query_capability.number);
4480 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4481 adapter->opt_rx_bufadd_q_per_rx_comp_q);
4482 break;
4483 case OPT_TX_ENTRIES_PER_SUBCRQ:
4484 adapter->opt_tx_entries_per_subcrq =
4485 be64_to_cpu(crq->query_capability.number);
4486 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4487 adapter->opt_tx_entries_per_subcrq);
4488 break;
4489 case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4490 adapter->opt_rxba_entries_per_subcrq =
4491 be64_to_cpu(crq->query_capability.number);
4492 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4493 adapter->opt_rxba_entries_per_subcrq);
4494 break;
4495 case TX_RX_DESC_REQ:
4496 adapter->tx_rx_desc_req = crq->query_capability.number;
4497 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4498 adapter->tx_rx_desc_req);
4499 break;
4501 default:
4502 netdev_err(netdev, "Got invalid cap rsp %d\n",
4503 crq->query_capability.capability);
4506 out:
4507 if (atomic_read(&adapter->running_cap_crqs) == 0) {
4508 adapter->wait_capability = false;
4509 ibmvnic_send_req_caps(adapter, 0);
4513 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4515 union ibmvnic_crq crq;
4516 int rc;
4518 memset(&crq, 0, sizeof(crq));
4519 crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4520 crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4522 mutex_lock(&adapter->fw_lock);
4523 adapter->fw_done_rc = 0;
4524 reinit_completion(&adapter->fw_done);
4526 rc = ibmvnic_send_crq(adapter, &crq);
4527 if (rc) {
4528 mutex_unlock(&adapter->fw_lock);
4529 return rc;
4532 rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
4533 if (rc) {
4534 mutex_unlock(&adapter->fw_lock);
4535 return rc;
4538 mutex_unlock(&adapter->fw_lock);
4539 return adapter->fw_done_rc ? -EIO : 0;
4542 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4543 struct ibmvnic_adapter *adapter)
4545 struct net_device *netdev = adapter->netdev;
4546 int rc;
4547 __be32 rspeed = cpu_to_be32(crq->query_phys_parms_rsp.speed);
4549 rc = crq->query_phys_parms_rsp.rc.code;
4550 if (rc) {
4551 netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4552 return rc;
4554 switch (rspeed) {
4555 case IBMVNIC_10MBPS:
4556 adapter->speed = SPEED_10;
4557 break;
4558 case IBMVNIC_100MBPS:
4559 adapter->speed = SPEED_100;
4560 break;
4561 case IBMVNIC_1GBPS:
4562 adapter->speed = SPEED_1000;
4563 break;
4564 case IBMVNIC_10GBP:
4565 adapter->speed = SPEED_10000;
4566 break;
4567 case IBMVNIC_25GBPS:
4568 adapter->speed = SPEED_25000;
4569 break;
4570 case IBMVNIC_40GBPS:
4571 adapter->speed = SPEED_40000;
4572 break;
4573 case IBMVNIC_50GBPS:
4574 adapter->speed = SPEED_50000;
4575 break;
4576 case IBMVNIC_100GBPS:
4577 adapter->speed = SPEED_100000;
4578 break;
4579 default:
4580 if (netif_carrier_ok(netdev))
4581 netdev_warn(netdev, "Unknown speed 0x%08x\n", rspeed);
4582 adapter->speed = SPEED_UNKNOWN;
4584 if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4585 adapter->duplex = DUPLEX_FULL;
4586 else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4587 adapter->duplex = DUPLEX_HALF;
4588 else
4589 adapter->duplex = DUPLEX_UNKNOWN;
4591 return rc;
4594 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4595 struct ibmvnic_adapter *adapter)
4597 struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4598 struct net_device *netdev = adapter->netdev;
4599 struct device *dev = &adapter->vdev->dev;
4600 u64 *u64_crq = (u64 *)crq;
4601 long rc;
4603 netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4604 (unsigned long int)cpu_to_be64(u64_crq[0]),
4605 (unsigned long int)cpu_to_be64(u64_crq[1]));
4606 switch (gen_crq->first) {
4607 case IBMVNIC_CRQ_INIT_RSP:
4608 switch (gen_crq->cmd) {
4609 case IBMVNIC_CRQ_INIT:
4610 dev_info(dev, "Partner initialized\n");
4611 adapter->from_passive_init = true;
4612 adapter->failover_pending = false;
4613 if (!completion_done(&adapter->init_done)) {
4614 complete(&adapter->init_done);
4615 adapter->init_done_rc = -EIO;
4617 ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4618 break;
4619 case IBMVNIC_CRQ_INIT_COMPLETE:
4620 dev_info(dev, "Partner initialization complete\n");
4621 adapter->crq.active = true;
4622 send_version_xchg(adapter);
4623 break;
4624 default:
4625 dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4627 return;
4628 case IBMVNIC_CRQ_XPORT_EVENT:
4629 netif_carrier_off(netdev);
4630 adapter->crq.active = false;
4631 /* terminate any thread waiting for a response
4632 * from the device
4634 if (!completion_done(&adapter->fw_done)) {
4635 adapter->fw_done_rc = -EIO;
4636 complete(&adapter->fw_done);
4638 if (!completion_done(&adapter->stats_done))
4639 complete(&adapter->stats_done);
4640 if (test_bit(0, &adapter->resetting))
4641 adapter->force_reset_recovery = true;
4642 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4643 dev_info(dev, "Migrated, re-enabling adapter\n");
4644 ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4645 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4646 dev_info(dev, "Backing device failover detected\n");
4647 adapter->failover_pending = true;
4648 } else {
4649 /* The adapter lost the connection */
4650 dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4651 gen_crq->cmd);
4652 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4654 return;
4655 case IBMVNIC_CRQ_CMD_RSP:
4656 break;
4657 default:
4658 dev_err(dev, "Got an invalid msg type 0x%02x\n",
4659 gen_crq->first);
4660 return;
4663 switch (gen_crq->cmd) {
4664 case VERSION_EXCHANGE_RSP:
4665 rc = crq->version_exchange_rsp.rc.code;
4666 if (rc) {
4667 dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4668 break;
4670 dev_info(dev, "Partner protocol version is %d\n",
4671 crq->version_exchange_rsp.version);
4672 if (be16_to_cpu(crq->version_exchange_rsp.version) <
4673 ibmvnic_version)
4674 ibmvnic_version =
4675 be16_to_cpu(crq->version_exchange_rsp.version);
4676 send_cap_queries(adapter);
4677 break;
4678 case QUERY_CAPABILITY_RSP:
4679 handle_query_cap_rsp(crq, adapter);
4680 break;
4681 case QUERY_MAP_RSP:
4682 handle_query_map_rsp(crq, adapter);
4683 break;
4684 case REQUEST_MAP_RSP:
4685 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4686 complete(&adapter->fw_done);
4687 break;
4688 case REQUEST_UNMAP_RSP:
4689 handle_request_unmap_rsp(crq, adapter);
4690 break;
4691 case REQUEST_CAPABILITY_RSP:
4692 handle_request_cap_rsp(crq, adapter);
4693 break;
4694 case LOGIN_RSP:
4695 netdev_dbg(netdev, "Got Login Response\n");
4696 handle_login_rsp(crq, adapter);
4697 break;
4698 case LOGICAL_LINK_STATE_RSP:
4699 netdev_dbg(netdev,
4700 "Got Logical Link State Response, state: %d rc: %d\n",
4701 crq->logical_link_state_rsp.link_state,
4702 crq->logical_link_state_rsp.rc.code);
4703 adapter->logical_link_state =
4704 crq->logical_link_state_rsp.link_state;
4705 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4706 complete(&adapter->init_done);
4707 break;
4708 case LINK_STATE_INDICATION:
4709 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4710 adapter->phys_link_state =
4711 crq->link_state_indication.phys_link_state;
4712 adapter->logical_link_state =
4713 crq->link_state_indication.logical_link_state;
4714 if (adapter->phys_link_state && adapter->logical_link_state)
4715 netif_carrier_on(netdev);
4716 else
4717 netif_carrier_off(netdev);
4718 break;
4719 case CHANGE_MAC_ADDR_RSP:
4720 netdev_dbg(netdev, "Got MAC address change Response\n");
4721 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4722 break;
4723 case ERROR_INDICATION:
4724 netdev_dbg(netdev, "Got Error Indication\n");
4725 handle_error_indication(crq, adapter);
4726 break;
4727 case REQUEST_STATISTICS_RSP:
4728 netdev_dbg(netdev, "Got Statistics Response\n");
4729 complete(&adapter->stats_done);
4730 break;
4731 case QUERY_IP_OFFLOAD_RSP:
4732 netdev_dbg(netdev, "Got Query IP offload Response\n");
4733 handle_query_ip_offload_rsp(adapter);
4734 break;
4735 case MULTICAST_CTRL_RSP:
4736 netdev_dbg(netdev, "Got multicast control Response\n");
4737 break;
4738 case CONTROL_IP_OFFLOAD_RSP:
4739 netdev_dbg(netdev, "Got Control IP offload Response\n");
4740 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4741 sizeof(adapter->ip_offload_ctrl),
4742 DMA_TO_DEVICE);
4743 complete(&adapter->init_done);
4744 break;
4745 case COLLECT_FW_TRACE_RSP:
4746 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4747 complete(&adapter->fw_done);
4748 break;
4749 case GET_VPD_SIZE_RSP:
4750 handle_vpd_size_rsp(crq, adapter);
4751 break;
4752 case GET_VPD_RSP:
4753 handle_vpd_rsp(crq, adapter);
4754 break;
4755 case QUERY_PHYS_PARMS_RSP:
4756 adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4757 complete(&adapter->fw_done);
4758 break;
4759 default:
4760 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4761 gen_crq->cmd);
4765 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4767 struct ibmvnic_adapter *adapter = instance;
4769 tasklet_schedule(&adapter->tasklet);
4770 return IRQ_HANDLED;
4773 static void ibmvnic_tasklet(void *data)
4775 struct ibmvnic_adapter *adapter = data;
4776 struct ibmvnic_crq_queue *queue = &adapter->crq;
4777 union ibmvnic_crq *crq;
4778 unsigned long flags;
4779 bool done = false;
4781 spin_lock_irqsave(&queue->lock, flags);
4782 while (!done) {
4783 /* Pull all the valid messages off the CRQ */
4784 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4785 ibmvnic_handle_crq(crq, adapter);
4786 crq->generic.first = 0;
4789 /* remain in tasklet until all
4790 * capabilities responses are received
4792 if (!adapter->wait_capability)
4793 done = true;
4795 /* if capabilities CRQ's were sent in this tasklet, the following
4796 * tasklet must wait until all responses are received
4798 if (atomic_read(&adapter->running_cap_crqs) != 0)
4799 adapter->wait_capability = true;
4800 spin_unlock_irqrestore(&queue->lock, flags);
4803 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4805 struct vio_dev *vdev = adapter->vdev;
4806 int rc;
4808 do {
4809 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4810 } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4812 if (rc)
4813 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4815 return rc;
4818 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4820 struct ibmvnic_crq_queue *crq = &adapter->crq;
4821 struct device *dev = &adapter->vdev->dev;
4822 struct vio_dev *vdev = adapter->vdev;
4823 int rc;
4825 /* Close the CRQ */
4826 do {
4827 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4828 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4830 /* Clean out the queue */
4831 memset(crq->msgs, 0, PAGE_SIZE);
4832 crq->cur = 0;
4833 crq->active = false;
4835 /* And re-open it again */
4836 rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4837 crq->msg_token, PAGE_SIZE);
4839 if (rc == H_CLOSED)
4840 /* Adapter is good, but other end is not ready */
4841 dev_warn(dev, "Partner adapter not ready\n");
4842 else if (rc != 0)
4843 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4845 return rc;
4848 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4850 struct ibmvnic_crq_queue *crq = &adapter->crq;
4851 struct vio_dev *vdev = adapter->vdev;
4852 long rc;
4854 if (!crq->msgs)
4855 return;
4857 netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4858 free_irq(vdev->irq, adapter);
4859 tasklet_kill(&adapter->tasklet);
4860 do {
4861 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4862 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4864 dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4865 DMA_BIDIRECTIONAL);
4866 free_page((unsigned long)crq->msgs);
4867 crq->msgs = NULL;
4868 crq->active = false;
4871 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4873 struct ibmvnic_crq_queue *crq = &adapter->crq;
4874 struct device *dev = &adapter->vdev->dev;
4875 struct vio_dev *vdev = adapter->vdev;
4876 int rc, retrc = -ENOMEM;
4878 if (crq->msgs)
4879 return 0;
4881 crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4882 /* Should we allocate more than one page? */
4884 if (!crq->msgs)
4885 return -ENOMEM;
4887 crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4888 crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4889 DMA_BIDIRECTIONAL);
4890 if (dma_mapping_error(dev, crq->msg_token))
4891 goto map_failed;
4893 rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4894 crq->msg_token, PAGE_SIZE);
4896 if (rc == H_RESOURCE)
4897 /* maybe kexecing and resource is busy. try a reset */
4898 rc = ibmvnic_reset_crq(adapter);
4899 retrc = rc;
4901 if (rc == H_CLOSED) {
4902 dev_warn(dev, "Partner adapter not ready\n");
4903 } else if (rc) {
4904 dev_warn(dev, "Error %d opening adapter\n", rc);
4905 goto reg_crq_failed;
4908 retrc = 0;
4910 tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4911 (unsigned long)adapter);
4913 netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4914 snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
4915 adapter->vdev->unit_address);
4916 rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
4917 if (rc) {
4918 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4919 vdev->irq, rc);
4920 goto req_irq_failed;
4923 rc = vio_enable_interrupts(vdev);
4924 if (rc) {
4925 dev_err(dev, "Error %d enabling interrupts\n", rc);
4926 goto req_irq_failed;
4929 crq->cur = 0;
4930 spin_lock_init(&crq->lock);
4932 return retrc;
4934 req_irq_failed:
4935 tasklet_kill(&adapter->tasklet);
4936 do {
4937 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4938 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4939 reg_crq_failed:
4940 dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4941 map_failed:
4942 free_page((unsigned long)crq->msgs);
4943 crq->msgs = NULL;
4944 return retrc;
4947 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4949 struct device *dev = &adapter->vdev->dev;
4950 unsigned long timeout = msecs_to_jiffies(30000);
4951 u64 old_num_rx_queues, old_num_tx_queues;
4952 int rc;
4954 adapter->from_passive_init = false;
4956 old_num_rx_queues = adapter->req_rx_queues;
4957 old_num_tx_queues = adapter->req_tx_queues;
4959 reinit_completion(&adapter->init_done);
4960 adapter->init_done_rc = 0;
4961 ibmvnic_send_crq_init(adapter);
4962 if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4963 dev_err(dev, "Initialization sequence timed out\n");
4964 return -1;
4967 if (adapter->init_done_rc) {
4968 release_crq_queue(adapter);
4969 return adapter->init_done_rc;
4972 if (adapter->from_passive_init) {
4973 adapter->state = VNIC_OPEN;
4974 adapter->from_passive_init = false;
4975 return -1;
4978 if (test_bit(0, &adapter->resetting) && !adapter->wait_for_reset &&
4979 adapter->reset_reason != VNIC_RESET_MOBILITY) {
4980 if (adapter->req_rx_queues != old_num_rx_queues ||
4981 adapter->req_tx_queues != old_num_tx_queues) {
4982 release_sub_crqs(adapter, 0);
4983 rc = init_sub_crqs(adapter);
4984 } else {
4985 rc = reset_sub_crq_queues(adapter);
4987 } else {
4988 rc = init_sub_crqs(adapter);
4991 if (rc) {
4992 dev_err(dev, "Initialization of sub crqs failed\n");
4993 release_crq_queue(adapter);
4994 return rc;
4997 rc = init_sub_crq_irqs(adapter);
4998 if (rc) {
4999 dev_err(dev, "Failed to initialize sub crq irqs\n");
5000 release_crq_queue(adapter);
5003 return rc;
5006 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
5008 struct device *dev = &adapter->vdev->dev;
5009 unsigned long timeout = msecs_to_jiffies(30000);
5010 int rc;
5012 adapter->from_passive_init = false;
5014 adapter->init_done_rc = 0;
5015 ibmvnic_send_crq_init(adapter);
5016 if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
5017 dev_err(dev, "Initialization sequence timed out\n");
5018 return -1;
5021 if (adapter->init_done_rc) {
5022 release_crq_queue(adapter);
5023 return adapter->init_done_rc;
5026 if (adapter->from_passive_init) {
5027 adapter->state = VNIC_OPEN;
5028 adapter->from_passive_init = false;
5029 return -1;
5032 rc = init_sub_crqs(adapter);
5033 if (rc) {
5034 dev_err(dev, "Initialization of sub crqs failed\n");
5035 release_crq_queue(adapter);
5036 return rc;
5039 rc = init_sub_crq_irqs(adapter);
5040 if (rc) {
5041 dev_err(dev, "Failed to initialize sub crq irqs\n");
5042 release_crq_queue(adapter);
5045 return rc;
5048 static struct device_attribute dev_attr_failover;
5050 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
5052 struct ibmvnic_adapter *adapter;
5053 struct net_device *netdev;
5054 unsigned char *mac_addr_p;
5055 int rc;
5057 dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
5058 dev->unit_address);
5060 mac_addr_p = (unsigned char *)vio_get_attribute(dev,
5061 VETH_MAC_ADDR, NULL);
5062 if (!mac_addr_p) {
5063 dev_err(&dev->dev,
5064 "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
5065 __FILE__, __LINE__);
5066 return 0;
5069 netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
5070 IBMVNIC_MAX_QUEUES);
5071 if (!netdev)
5072 return -ENOMEM;
5074 adapter = netdev_priv(netdev);
5075 adapter->state = VNIC_PROBING;
5076 dev_set_drvdata(&dev->dev, netdev);
5077 adapter->vdev = dev;
5078 adapter->netdev = netdev;
5080 ether_addr_copy(adapter->mac_addr, mac_addr_p);
5081 ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
5082 netdev->irq = dev->irq;
5083 netdev->netdev_ops = &ibmvnic_netdev_ops;
5084 netdev->ethtool_ops = &ibmvnic_ethtool_ops;
5085 SET_NETDEV_DEV(netdev, &dev->dev);
5087 spin_lock_init(&adapter->stats_lock);
5089 INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
5090 INIT_DELAYED_WORK(&adapter->ibmvnic_delayed_reset,
5091 __ibmvnic_delayed_reset);
5092 INIT_LIST_HEAD(&adapter->rwi_list);
5093 spin_lock_init(&adapter->rwi_lock);
5094 mutex_init(&adapter->fw_lock);
5095 init_completion(&adapter->init_done);
5096 init_completion(&adapter->fw_done);
5097 init_completion(&adapter->reset_done);
5098 init_completion(&adapter->stats_done);
5099 clear_bit(0, &adapter->resetting);
5101 do {
5102 rc = init_crq_queue(adapter);
5103 if (rc) {
5104 dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
5105 rc);
5106 goto ibmvnic_init_fail;
5109 rc = ibmvnic_init(adapter);
5110 if (rc && rc != EAGAIN)
5111 goto ibmvnic_init_fail;
5112 } while (rc == EAGAIN);
5114 rc = init_stats_buffers(adapter);
5115 if (rc)
5116 goto ibmvnic_init_fail;
5118 rc = init_stats_token(adapter);
5119 if (rc)
5120 goto ibmvnic_stats_fail;
5122 netdev->mtu = adapter->req_mtu - ETH_HLEN;
5123 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
5124 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
5126 rc = device_create_file(&dev->dev, &dev_attr_failover);
5127 if (rc)
5128 goto ibmvnic_dev_file_err;
5130 netif_carrier_off(netdev);
5131 rc = register_netdev(netdev);
5132 if (rc) {
5133 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
5134 goto ibmvnic_register_fail;
5136 dev_info(&dev->dev, "ibmvnic registered\n");
5138 adapter->state = VNIC_PROBED;
5140 adapter->wait_for_reset = false;
5142 return 0;
5144 ibmvnic_register_fail:
5145 device_remove_file(&dev->dev, &dev_attr_failover);
5147 ibmvnic_dev_file_err:
5148 release_stats_token(adapter);
5150 ibmvnic_stats_fail:
5151 release_stats_buffers(adapter);
5153 ibmvnic_init_fail:
5154 release_sub_crqs(adapter, 1);
5155 release_crq_queue(adapter);
5156 mutex_destroy(&adapter->fw_lock);
5157 free_netdev(netdev);
5159 return rc;
5162 static int ibmvnic_remove(struct vio_dev *dev)
5164 struct net_device *netdev = dev_get_drvdata(&dev->dev);
5165 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5167 adapter->state = VNIC_REMOVING;
5168 rtnl_lock();
5169 unregister_netdevice(netdev);
5171 release_resources(adapter);
5172 release_sub_crqs(adapter, 1);
5173 release_crq_queue(adapter);
5175 release_stats_token(adapter);
5176 release_stats_buffers(adapter);
5178 adapter->state = VNIC_REMOVED;
5180 rtnl_unlock();
5181 mutex_destroy(&adapter->fw_lock);
5182 device_remove_file(&dev->dev, &dev_attr_failover);
5183 free_netdev(netdev);
5184 dev_set_drvdata(&dev->dev, NULL);
5186 return 0;
5189 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
5190 const char *buf, size_t count)
5192 struct net_device *netdev = dev_get_drvdata(dev);
5193 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5194 unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
5195 __be64 session_token;
5196 long rc;
5198 if (!sysfs_streq(buf, "1"))
5199 return -EINVAL;
5201 rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
5202 H_GET_SESSION_TOKEN, 0, 0, 0);
5203 if (rc) {
5204 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
5205 rc);
5206 return -EINVAL;
5209 session_token = (__be64)retbuf[0];
5210 netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
5211 be64_to_cpu(session_token));
5212 rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
5213 H_SESSION_ERR_DETECTED, session_token, 0, 0);
5214 if (rc) {
5215 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
5216 rc);
5217 return -EINVAL;
5220 return count;
5223 static DEVICE_ATTR_WO(failover);
5225 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
5227 struct net_device *netdev = dev_get_drvdata(&vdev->dev);
5228 struct ibmvnic_adapter *adapter;
5229 struct iommu_table *tbl;
5230 unsigned long ret = 0;
5231 int i;
5233 tbl = get_iommu_table_base(&vdev->dev);
5235 /* netdev inits at probe time along with the structures we need below*/
5236 if (!netdev)
5237 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
5239 adapter = netdev_priv(netdev);
5241 ret += PAGE_SIZE; /* the crq message queue */
5242 ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
5244 for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
5245 ret += 4 * PAGE_SIZE; /* the scrq message queue */
5247 for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
5248 i++)
5249 ret += adapter->rx_pool[i].size *
5250 IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
5252 return ret;
5255 static int ibmvnic_resume(struct device *dev)
5257 struct net_device *netdev = dev_get_drvdata(dev);
5258 struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5260 if (adapter->state != VNIC_OPEN)
5261 return 0;
5263 tasklet_schedule(&adapter->tasklet);
5265 return 0;
5268 static const struct vio_device_id ibmvnic_device_table[] = {
5269 {"network", "IBM,vnic"},
5270 {"", "" }
5272 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5274 static const struct dev_pm_ops ibmvnic_pm_ops = {
5275 .resume = ibmvnic_resume
5278 static struct vio_driver ibmvnic_driver = {
5279 .id_table = ibmvnic_device_table,
5280 .probe = ibmvnic_probe,
5281 .remove = ibmvnic_remove,
5282 .get_desired_dma = ibmvnic_get_desired_dma,
5283 .name = ibmvnic_driver_name,
5284 .pm = &ibmvnic_pm_ops,
5287 /* module functions */
5288 static int __init ibmvnic_module_init(void)
5290 pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5291 IBMVNIC_DRIVER_VERSION);
5293 return vio_register_driver(&ibmvnic_driver);
5296 static void __exit ibmvnic_module_exit(void)
5298 vio_unregister_driver(&ibmvnic_driver);
5301 module_init(ibmvnic_module_init);
5302 module_exit(ibmvnic_module_exit);