xref: /openbmc/linux/drivers/net/ethernet/ibm/ibmvnic.c (revision 04295878beac396dae47ba93141cae0d9386e7ef)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /**************************************************************************/
3 /*                                                                        */
4 /*  IBM System i and System p Virtual NIC Device Driver                   */
5 /*  Copyright (C) 2014 IBM Corp.                                          */
6 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
7 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
8 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
9 /*                                                                        */
10 /*                                                                        */
11 /* This module contains the implementation of a virtual ethernet device   */
12 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
13 /* option of the RS/6000 Platform Architecture to interface with virtual  */
14 /* ethernet NICs that are presented to the partition by the hypervisor.   */
15 /*									   */
16 /* Messages are passed between the VNIC driver and the VNIC server using  */
17 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
18 /* issue and receive commands that initiate communication with the server */
19 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
20 /* are used by the driver to notify the server that a packet is           */
21 /* ready for transmission or that a buffer has been added to receive a    */
22 /* packet. Subsequently, sCRQs are used by the server to notify the       */
23 /* driver that a packet transmission has been completed or that a packet  */
24 /* has been received and placed in a waiting buffer.                      */
25 /*                                                                        */
26 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
27 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
28 /* or receive has been completed, the VNIC driver is required to use      */
29 /* "long term mapping". This entails that large, continuous DMA mapped    */
30 /* buffers are allocated on driver initialization and these buffers are   */
31 /* then continuously reused to pass skbs to and from the VNIC server.     */
32 /*                                                                        */
33 /**************************************************************************/
34 
35 #include <linux/module.h>
36 #include <linux/moduleparam.h>
37 #include <linux/types.h>
38 #include <linux/errno.h>
39 #include <linux/completion.h>
40 #include <linux/ioport.h>
41 #include <linux/dma-mapping.h>
42 #include <linux/kernel.h>
43 #include <linux/netdevice.h>
44 #include <linux/etherdevice.h>
45 #include <linux/skbuff.h>
46 #include <linux/init.h>
47 #include <linux/delay.h>
48 #include <linux/mm.h>
49 #include <linux/ethtool.h>
50 #include <linux/proc_fs.h>
51 #include <linux/if_arp.h>
52 #include <linux/in.h>
53 #include <linux/ip.h>
54 #include <linux/ipv6.h>
55 #include <linux/irq.h>
56 #include <linux/kthread.h>
57 #include <linux/seq_file.h>
58 #include <linux/interrupt.h>
59 #include <net/net_namespace.h>
60 #include <asm/hvcall.h>
61 #include <linux/atomic.h>
62 #include <asm/vio.h>
63 #include <asm/iommu.h>
64 #include <linux/uaccess.h>
65 #include <asm/firmware.h>
66 #include <linux/workqueue.h>
67 #include <linux/if_vlan.h>
68 #include <linux/utsname.h>
69 
70 #include "ibmvnic.h"
71 
72 static const char ibmvnic_driver_name[] = "ibmvnic";
73 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
74 
75 MODULE_AUTHOR("Santiago Leon");
76 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
77 MODULE_LICENSE("GPL");
78 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
79 
80 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
81 static 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_indirect(struct ibmvnic_adapter *, u64, u64, u64);
88 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
89 static int enable_scrq_irq(struct ibmvnic_adapter *,
90 			   struct ibmvnic_sub_crq_queue *);
91 static int disable_scrq_irq(struct ibmvnic_adapter *,
92 			    struct ibmvnic_sub_crq_queue *);
93 static int pending_scrq(struct ibmvnic_adapter *,
94 			struct ibmvnic_sub_crq_queue *);
95 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
96 					struct ibmvnic_sub_crq_queue *);
97 static int ibmvnic_poll(struct napi_struct *napi, int data);
98 static void send_query_map(struct ibmvnic_adapter *adapter);
99 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
100 static int send_request_unmap(struct ibmvnic_adapter *, u8);
101 static int send_login(struct ibmvnic_adapter *adapter);
102 static void send_query_cap(struct ibmvnic_adapter *adapter);
103 static int init_sub_crqs(struct ibmvnic_adapter *);
104 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
105 static int ibmvnic_reset_init(struct ibmvnic_adapter *, bool reset);
106 static void release_crq_queue(struct ibmvnic_adapter *);
107 static int __ibmvnic_set_mac(struct net_device *, u8 *);
108 static int init_crq_queue(struct ibmvnic_adapter *adapter);
109 static int send_query_phys_parms(struct ibmvnic_adapter *adapter);
110 
111 struct ibmvnic_stat {
112 	char name[ETH_GSTRING_LEN];
113 	int offset;
114 };
115 
116 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
117 			     offsetof(struct ibmvnic_statistics, stat))
118 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
119 
120 static const struct ibmvnic_stat ibmvnic_stats[] = {
121 	{"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
122 	{"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
123 	{"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
124 	{"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
125 	{"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
126 	{"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
127 	{"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
128 	{"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
129 	{"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
130 	{"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
131 	{"align_errors", IBMVNIC_STAT_OFF(align_errors)},
132 	{"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
133 	{"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
134 	{"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
135 	{"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
136 	{"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
137 	{"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
138 	{"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
139 	{"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
140 	{"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
141 	{"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
142 	{"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
143 };
144 
145 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
146 			  unsigned long length, unsigned long *number,
147 			  unsigned long *irq)
148 {
149 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
150 	long rc;
151 
152 	rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
153 	*number = retbuf[0];
154 	*irq = retbuf[1];
155 
156 	return rc;
157 }
158 
159 /**
160  * ibmvnic_wait_for_completion - Check device state and wait for completion
161  * @adapter: private device data
162  * @comp_done: completion structure to wait for
163  * @timeout: time to wait in milliseconds
164  *
165  * Wait for a completion signal or until the timeout limit is reached
166  * while checking that the device is still active.
167  */
168 static int ibmvnic_wait_for_completion(struct ibmvnic_adapter *adapter,
169 				       struct completion *comp_done,
170 				       unsigned long timeout)
171 {
172 	struct net_device *netdev;
173 	unsigned long div_timeout;
174 	u8 retry;
175 
176 	netdev = adapter->netdev;
177 	retry = 5;
178 	div_timeout = msecs_to_jiffies(timeout / retry);
179 	while (true) {
180 		if (!adapter->crq.active) {
181 			netdev_err(netdev, "Device down!\n");
182 			return -ENODEV;
183 		}
184 		if (!retry--)
185 			break;
186 		if (wait_for_completion_timeout(comp_done, div_timeout))
187 			return 0;
188 	}
189 	netdev_err(netdev, "Operation timed out.\n");
190 	return -ETIMEDOUT;
191 }
192 
193 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
194 				struct ibmvnic_long_term_buff *ltb, int size)
195 {
196 	struct device *dev = &adapter->vdev->dev;
197 	int rc;
198 
199 	ltb->size = size;
200 	ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
201 				       GFP_KERNEL);
202 
203 	if (!ltb->buff) {
204 		dev_err(dev, "Couldn't alloc long term buffer\n");
205 		return -ENOMEM;
206 	}
207 	ltb->map_id = adapter->map_id;
208 	adapter->map_id++;
209 
210 	mutex_lock(&adapter->fw_lock);
211 	adapter->fw_done_rc = 0;
212 	reinit_completion(&adapter->fw_done);
213 	rc = send_request_map(adapter, ltb->addr,
214 			      ltb->size, ltb->map_id);
215 	if (rc) {
216 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
217 		mutex_unlock(&adapter->fw_lock);
218 		return rc;
219 	}
220 
221 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
222 	if (rc) {
223 		dev_err(dev,
224 			"Long term map request aborted or timed out,rc = %d\n",
225 			rc);
226 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
227 		mutex_unlock(&adapter->fw_lock);
228 		return rc;
229 	}
230 
231 	if (adapter->fw_done_rc) {
232 		dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
233 			adapter->fw_done_rc);
234 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
235 		mutex_unlock(&adapter->fw_lock);
236 		return -1;
237 	}
238 	mutex_unlock(&adapter->fw_lock);
239 	return 0;
240 }
241 
242 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
243 				struct ibmvnic_long_term_buff *ltb)
244 {
245 	struct device *dev = &adapter->vdev->dev;
246 
247 	if (!ltb->buff)
248 		return;
249 
250 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
251 	    adapter->reset_reason != VNIC_RESET_MOBILITY)
252 		send_request_unmap(adapter, ltb->map_id);
253 	dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
254 }
255 
256 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
257 				struct ibmvnic_long_term_buff *ltb)
258 {
259 	struct device *dev = &adapter->vdev->dev;
260 	int rc;
261 
262 	memset(ltb->buff, 0, ltb->size);
263 
264 	mutex_lock(&adapter->fw_lock);
265 	adapter->fw_done_rc = 0;
266 
267 	reinit_completion(&adapter->fw_done);
268 	rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
269 	if (rc) {
270 		mutex_unlock(&adapter->fw_lock);
271 		return rc;
272 	}
273 
274 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
275 	if (rc) {
276 		dev_info(dev,
277 			 "Reset failed, long term map request timed out or aborted\n");
278 		mutex_unlock(&adapter->fw_lock);
279 		return rc;
280 	}
281 
282 	if (adapter->fw_done_rc) {
283 		dev_info(dev,
284 			 "Reset failed, attempting to free and reallocate buffer\n");
285 		free_long_term_buff(adapter, ltb);
286 		mutex_unlock(&adapter->fw_lock);
287 		return alloc_long_term_buff(adapter, ltb, ltb->size);
288 	}
289 	mutex_unlock(&adapter->fw_lock);
290 	return 0;
291 }
292 
293 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
294 {
295 	int i;
296 
297 	for (i = 0; i < adapter->num_active_rx_pools; i++)
298 		adapter->rx_pool[i].active = 0;
299 }
300 
301 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
302 			      struct ibmvnic_rx_pool *pool)
303 {
304 	int count = pool->size - atomic_read(&pool->available);
305 	u64 handle = adapter->rx_scrq[pool->index]->handle;
306 	struct device *dev = &adapter->vdev->dev;
307 	struct ibmvnic_ind_xmit_queue *ind_bufp;
308 	struct ibmvnic_sub_crq_queue *rx_scrq;
309 	union sub_crq *sub_crq;
310 	int buffers_added = 0;
311 	unsigned long lpar_rc;
312 	struct sk_buff *skb;
313 	unsigned int offset;
314 	dma_addr_t dma_addr;
315 	unsigned char *dst;
316 	int shift = 0;
317 	int index;
318 	int i;
319 
320 	if (!pool->active)
321 		return;
322 
323 	rx_scrq = adapter->rx_scrq[pool->index];
324 	ind_bufp = &rx_scrq->ind_buf;
325 	for (i = 0; i < count; ++i) {
326 		skb = netdev_alloc_skb(adapter->netdev, pool->buff_size);
327 		if (!skb) {
328 			dev_err(dev, "Couldn't replenish rx buff\n");
329 			adapter->replenish_no_mem++;
330 			break;
331 		}
332 
333 		index = pool->free_map[pool->next_free];
334 
335 		if (pool->rx_buff[index].skb)
336 			dev_err(dev, "Inconsistent free_map!\n");
337 
338 		/* Copy the skb to the long term mapped DMA buffer */
339 		offset = index * pool->buff_size;
340 		dst = pool->long_term_buff.buff + offset;
341 		memset(dst, 0, pool->buff_size);
342 		dma_addr = pool->long_term_buff.addr + offset;
343 		pool->rx_buff[index].data = dst;
344 
345 		pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
346 		pool->rx_buff[index].dma = dma_addr;
347 		pool->rx_buff[index].skb = skb;
348 		pool->rx_buff[index].pool_index = pool->index;
349 		pool->rx_buff[index].size = pool->buff_size;
350 
351 		sub_crq = &ind_bufp->indir_arr[ind_bufp->index++];
352 		memset(sub_crq, 0, sizeof(*sub_crq));
353 		sub_crq->rx_add.first = IBMVNIC_CRQ_CMD;
354 		sub_crq->rx_add.correlator =
355 		    cpu_to_be64((u64)&pool->rx_buff[index]);
356 		sub_crq->rx_add.ioba = cpu_to_be32(dma_addr);
357 		sub_crq->rx_add.map_id = pool->long_term_buff.map_id;
358 
359 		/* The length field of the sCRQ is defined to be 24 bits so the
360 		 * buffer size needs to be left shifted by a byte before it is
361 		 * converted to big endian to prevent the last byte from being
362 		 * truncated.
363 		 */
364 #ifdef __LITTLE_ENDIAN__
365 		shift = 8;
366 #endif
367 		sub_crq->rx_add.len = cpu_to_be32(pool->buff_size << shift);
368 		pool->next_free = (pool->next_free + 1) % pool->size;
369 		if (ind_bufp->index == IBMVNIC_MAX_IND_DESCS ||
370 		    i == count - 1) {
371 			lpar_rc =
372 				send_subcrq_indirect(adapter, handle,
373 						     (u64)ind_bufp->indir_dma,
374 						     (u64)ind_bufp->index);
375 			if (lpar_rc != H_SUCCESS)
376 				goto failure;
377 			buffers_added += ind_bufp->index;
378 			adapter->replenish_add_buff_success += ind_bufp->index;
379 			ind_bufp->index = 0;
380 		}
381 	}
382 	atomic_add(buffers_added, &pool->available);
383 	return;
384 
385 failure:
386 	if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
387 		dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
388 	for (i = ind_bufp->index - 1; i >= 0; --i) {
389 		struct ibmvnic_rx_buff *rx_buff;
390 
391 		pool->next_free = pool->next_free == 0 ?
392 				  pool->size - 1 : pool->next_free - 1;
393 		sub_crq = &ind_bufp->indir_arr[i];
394 		rx_buff = (struct ibmvnic_rx_buff *)
395 				be64_to_cpu(sub_crq->rx_add.correlator);
396 		index = (int)(rx_buff - pool->rx_buff);
397 		pool->free_map[pool->next_free] = index;
398 		dev_kfree_skb_any(pool->rx_buff[index].skb);
399 		pool->rx_buff[index].skb = NULL;
400 	}
401 	ind_bufp->index = 0;
402 	if (lpar_rc == H_CLOSED || adapter->failover_pending) {
403 		/* Disable buffer pool replenishment and report carrier off if
404 		 * queue is closed or pending failover.
405 		 * Firmware guarantees that a signal will be sent to the
406 		 * driver, triggering a reset.
407 		 */
408 		deactivate_rx_pools(adapter);
409 		netif_carrier_off(adapter->netdev);
410 	}
411 }
412 
413 static void replenish_pools(struct ibmvnic_adapter *adapter)
414 {
415 	int i;
416 
417 	adapter->replenish_task_cycles++;
418 	for (i = 0; i < adapter->num_active_rx_pools; i++) {
419 		if (adapter->rx_pool[i].active)
420 			replenish_rx_pool(adapter, &adapter->rx_pool[i]);
421 	}
422 }
423 
424 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
425 {
426 	kfree(adapter->tx_stats_buffers);
427 	kfree(adapter->rx_stats_buffers);
428 	adapter->tx_stats_buffers = NULL;
429 	adapter->rx_stats_buffers = NULL;
430 }
431 
432 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
433 {
434 	adapter->tx_stats_buffers =
435 				kcalloc(IBMVNIC_MAX_QUEUES,
436 					sizeof(struct ibmvnic_tx_queue_stats),
437 					GFP_KERNEL);
438 	if (!adapter->tx_stats_buffers)
439 		return -ENOMEM;
440 
441 	adapter->rx_stats_buffers =
442 				kcalloc(IBMVNIC_MAX_QUEUES,
443 					sizeof(struct ibmvnic_rx_queue_stats),
444 					GFP_KERNEL);
445 	if (!adapter->rx_stats_buffers)
446 		return -ENOMEM;
447 
448 	return 0;
449 }
450 
451 static void release_stats_token(struct ibmvnic_adapter *adapter)
452 {
453 	struct device *dev = &adapter->vdev->dev;
454 
455 	if (!adapter->stats_token)
456 		return;
457 
458 	dma_unmap_single(dev, adapter->stats_token,
459 			 sizeof(struct ibmvnic_statistics),
460 			 DMA_FROM_DEVICE);
461 	adapter->stats_token = 0;
462 }
463 
464 static int init_stats_token(struct ibmvnic_adapter *adapter)
465 {
466 	struct device *dev = &adapter->vdev->dev;
467 	dma_addr_t stok;
468 
469 	stok = dma_map_single(dev, &adapter->stats,
470 			      sizeof(struct ibmvnic_statistics),
471 			      DMA_FROM_DEVICE);
472 	if (dma_mapping_error(dev, stok)) {
473 		dev_err(dev, "Couldn't map stats buffer\n");
474 		return -1;
475 	}
476 
477 	adapter->stats_token = stok;
478 	netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
479 	return 0;
480 }
481 
482 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
483 {
484 	struct ibmvnic_rx_pool *rx_pool;
485 	u64 buff_size;
486 	int rx_scrqs;
487 	int i, j, rc;
488 
489 	if (!adapter->rx_pool)
490 		return -1;
491 
492 	buff_size = adapter->cur_rx_buf_sz;
493 	rx_scrqs = adapter->num_active_rx_pools;
494 	for (i = 0; i < rx_scrqs; i++) {
495 		rx_pool = &adapter->rx_pool[i];
496 
497 		netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
498 
499 		if (rx_pool->buff_size != buff_size) {
500 			free_long_term_buff(adapter, &rx_pool->long_term_buff);
501 			rx_pool->buff_size = ALIGN(buff_size, L1_CACHE_BYTES);
502 			rc = alloc_long_term_buff(adapter,
503 						  &rx_pool->long_term_buff,
504 						  rx_pool->size *
505 						  rx_pool->buff_size);
506 		} else {
507 			rc = reset_long_term_buff(adapter,
508 						  &rx_pool->long_term_buff);
509 		}
510 
511 		if (rc)
512 			return rc;
513 
514 		for (j = 0; j < rx_pool->size; j++)
515 			rx_pool->free_map[j] = j;
516 
517 		memset(rx_pool->rx_buff, 0,
518 		       rx_pool->size * sizeof(struct ibmvnic_rx_buff));
519 
520 		atomic_set(&rx_pool->available, 0);
521 		rx_pool->next_alloc = 0;
522 		rx_pool->next_free = 0;
523 		rx_pool->active = 1;
524 	}
525 
526 	return 0;
527 }
528 
529 static void release_rx_pools(struct ibmvnic_adapter *adapter)
530 {
531 	struct ibmvnic_rx_pool *rx_pool;
532 	int i, j;
533 
534 	if (!adapter->rx_pool)
535 		return;
536 
537 	for (i = 0; i < adapter->num_active_rx_pools; i++) {
538 		rx_pool = &adapter->rx_pool[i];
539 
540 		netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
541 
542 		kfree(rx_pool->free_map);
543 		free_long_term_buff(adapter, &rx_pool->long_term_buff);
544 
545 		if (!rx_pool->rx_buff)
546 			continue;
547 
548 		for (j = 0; j < rx_pool->size; j++) {
549 			if (rx_pool->rx_buff[j].skb) {
550 				dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
551 				rx_pool->rx_buff[j].skb = NULL;
552 			}
553 		}
554 
555 		kfree(rx_pool->rx_buff);
556 	}
557 
558 	kfree(adapter->rx_pool);
559 	adapter->rx_pool = NULL;
560 	adapter->num_active_rx_pools = 0;
561 }
562 
563 static int init_rx_pools(struct net_device *netdev)
564 {
565 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
566 	struct device *dev = &adapter->vdev->dev;
567 	struct ibmvnic_rx_pool *rx_pool;
568 	int rxadd_subcrqs;
569 	u64 buff_size;
570 	int i, j;
571 
572 	rxadd_subcrqs = adapter->num_active_rx_scrqs;
573 	buff_size = adapter->cur_rx_buf_sz;
574 
575 	adapter->rx_pool = kcalloc(rxadd_subcrqs,
576 				   sizeof(struct ibmvnic_rx_pool),
577 				   GFP_KERNEL);
578 	if (!adapter->rx_pool) {
579 		dev_err(dev, "Failed to allocate rx pools\n");
580 		return -1;
581 	}
582 
583 	adapter->num_active_rx_pools = rxadd_subcrqs;
584 
585 	for (i = 0; i < rxadd_subcrqs; i++) {
586 		rx_pool = &adapter->rx_pool[i];
587 
588 		netdev_dbg(adapter->netdev,
589 			   "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
590 			   i, adapter->req_rx_add_entries_per_subcrq,
591 			   buff_size);
592 
593 		rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
594 		rx_pool->index = i;
595 		rx_pool->buff_size = ALIGN(buff_size, L1_CACHE_BYTES);
596 		rx_pool->active = 1;
597 
598 		rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
599 					    GFP_KERNEL);
600 		if (!rx_pool->free_map) {
601 			release_rx_pools(adapter);
602 			return -1;
603 		}
604 
605 		rx_pool->rx_buff = kcalloc(rx_pool->size,
606 					   sizeof(struct ibmvnic_rx_buff),
607 					   GFP_KERNEL);
608 		if (!rx_pool->rx_buff) {
609 			dev_err(dev, "Couldn't alloc rx buffers\n");
610 			release_rx_pools(adapter);
611 			return -1;
612 		}
613 
614 		if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
615 					 rx_pool->size * rx_pool->buff_size)) {
616 			release_rx_pools(adapter);
617 			return -1;
618 		}
619 
620 		for (j = 0; j < rx_pool->size; ++j)
621 			rx_pool->free_map[j] = j;
622 
623 		atomic_set(&rx_pool->available, 0);
624 		rx_pool->next_alloc = 0;
625 		rx_pool->next_free = 0;
626 	}
627 
628 	return 0;
629 }
630 
631 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
632 			     struct ibmvnic_tx_pool *tx_pool)
633 {
634 	int rc, i;
635 
636 	rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
637 	if (rc)
638 		return rc;
639 
640 	memset(tx_pool->tx_buff, 0,
641 	       tx_pool->num_buffers *
642 	       sizeof(struct ibmvnic_tx_buff));
643 
644 	for (i = 0; i < tx_pool->num_buffers; i++)
645 		tx_pool->free_map[i] = i;
646 
647 	tx_pool->consumer_index = 0;
648 	tx_pool->producer_index = 0;
649 
650 	return 0;
651 }
652 
653 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
654 {
655 	int tx_scrqs;
656 	int i, rc;
657 
658 	if (!adapter->tx_pool)
659 		return -1;
660 
661 	tx_scrqs = adapter->num_active_tx_pools;
662 	for (i = 0; i < tx_scrqs; i++) {
663 		rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
664 		if (rc)
665 			return rc;
666 		rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
667 		if (rc)
668 			return rc;
669 	}
670 
671 	return 0;
672 }
673 
674 static void release_vpd_data(struct ibmvnic_adapter *adapter)
675 {
676 	if (!adapter->vpd)
677 		return;
678 
679 	kfree(adapter->vpd->buff);
680 	kfree(adapter->vpd);
681 
682 	adapter->vpd = NULL;
683 }
684 
685 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
686 				struct ibmvnic_tx_pool *tx_pool)
687 {
688 	kfree(tx_pool->tx_buff);
689 	kfree(tx_pool->free_map);
690 	free_long_term_buff(adapter, &tx_pool->long_term_buff);
691 }
692 
693 static void release_tx_pools(struct ibmvnic_adapter *adapter)
694 {
695 	int i;
696 
697 	if (!adapter->tx_pool)
698 		return;
699 
700 	for (i = 0; i < adapter->num_active_tx_pools; i++) {
701 		release_one_tx_pool(adapter, &adapter->tx_pool[i]);
702 		release_one_tx_pool(adapter, &adapter->tso_pool[i]);
703 	}
704 
705 	kfree(adapter->tx_pool);
706 	adapter->tx_pool = NULL;
707 	kfree(adapter->tso_pool);
708 	adapter->tso_pool = NULL;
709 	adapter->num_active_tx_pools = 0;
710 }
711 
712 static int init_one_tx_pool(struct net_device *netdev,
713 			    struct ibmvnic_tx_pool *tx_pool,
714 			    int num_entries, int buf_size)
715 {
716 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
717 	int i;
718 
719 	tx_pool->tx_buff = kcalloc(num_entries,
720 				   sizeof(struct ibmvnic_tx_buff),
721 				   GFP_KERNEL);
722 	if (!tx_pool->tx_buff)
723 		return -1;
724 
725 	if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
726 				 num_entries * buf_size))
727 		return -1;
728 
729 	tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
730 	if (!tx_pool->free_map)
731 		return -1;
732 
733 	for (i = 0; i < num_entries; i++)
734 		tx_pool->free_map[i] = i;
735 
736 	tx_pool->consumer_index = 0;
737 	tx_pool->producer_index = 0;
738 	tx_pool->num_buffers = num_entries;
739 	tx_pool->buf_size = buf_size;
740 
741 	return 0;
742 }
743 
744 static int init_tx_pools(struct net_device *netdev)
745 {
746 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
747 	int tx_subcrqs;
748 	u64 buff_size;
749 	int i, rc;
750 
751 	tx_subcrqs = adapter->num_active_tx_scrqs;
752 	adapter->tx_pool = kcalloc(tx_subcrqs,
753 				   sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
754 	if (!adapter->tx_pool)
755 		return -1;
756 
757 	adapter->tso_pool = kcalloc(tx_subcrqs,
758 				    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
759 	if (!adapter->tso_pool)
760 		return -1;
761 
762 	adapter->num_active_tx_pools = tx_subcrqs;
763 
764 	for (i = 0; i < tx_subcrqs; i++) {
765 		buff_size = adapter->req_mtu + VLAN_HLEN;
766 		buff_size = ALIGN(buff_size, L1_CACHE_BYTES);
767 		rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
768 				      adapter->req_tx_entries_per_subcrq,
769 				      buff_size);
770 		if (rc) {
771 			release_tx_pools(adapter);
772 			return rc;
773 		}
774 
775 		rc = init_one_tx_pool(netdev, &adapter->tso_pool[i],
776 				      IBMVNIC_TSO_BUFS,
777 				      IBMVNIC_TSO_BUF_SZ);
778 		if (rc) {
779 			release_tx_pools(adapter);
780 			return rc;
781 		}
782 	}
783 
784 	return 0;
785 }
786 
787 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
788 {
789 	int i;
790 
791 	if (adapter->napi_enabled)
792 		return;
793 
794 	for (i = 0; i < adapter->req_rx_queues; i++)
795 		napi_enable(&adapter->napi[i]);
796 
797 	adapter->napi_enabled = true;
798 }
799 
800 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
801 {
802 	int i;
803 
804 	if (!adapter->napi_enabled)
805 		return;
806 
807 	for (i = 0; i < adapter->req_rx_queues; i++) {
808 		netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
809 		napi_disable(&adapter->napi[i]);
810 	}
811 
812 	adapter->napi_enabled = false;
813 }
814 
815 static int init_napi(struct ibmvnic_adapter *adapter)
816 {
817 	int i;
818 
819 	adapter->napi = kcalloc(adapter->req_rx_queues,
820 				sizeof(struct napi_struct), GFP_KERNEL);
821 	if (!adapter->napi)
822 		return -ENOMEM;
823 
824 	for (i = 0; i < adapter->req_rx_queues; i++) {
825 		netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
826 		netif_napi_add(adapter->netdev, &adapter->napi[i],
827 			       ibmvnic_poll, NAPI_POLL_WEIGHT);
828 	}
829 
830 	adapter->num_active_rx_napi = adapter->req_rx_queues;
831 	return 0;
832 }
833 
834 static void release_napi(struct ibmvnic_adapter *adapter)
835 {
836 	int i;
837 
838 	if (!adapter->napi)
839 		return;
840 
841 	for (i = 0; i < adapter->num_active_rx_napi; i++) {
842 		netdev_dbg(adapter->netdev, "Releasing napi[%d]\n", i);
843 		netif_napi_del(&adapter->napi[i]);
844 	}
845 
846 	kfree(adapter->napi);
847 	adapter->napi = NULL;
848 	adapter->num_active_rx_napi = 0;
849 	adapter->napi_enabled = false;
850 }
851 
852 static int ibmvnic_login(struct net_device *netdev)
853 {
854 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
855 	unsigned long timeout = msecs_to_jiffies(30000);
856 	int retry_count = 0;
857 	int retries = 10;
858 	bool retry;
859 	int rc;
860 
861 	do {
862 		retry = false;
863 		if (retry_count > retries) {
864 			netdev_warn(netdev, "Login attempts exceeded\n");
865 			return -1;
866 		}
867 
868 		adapter->init_done_rc = 0;
869 		reinit_completion(&adapter->init_done);
870 		rc = send_login(adapter);
871 		if (rc) {
872 			netdev_warn(netdev, "Unable to login\n");
873 			return rc;
874 		}
875 
876 		if (!wait_for_completion_timeout(&adapter->init_done,
877 						 timeout)) {
878 			netdev_warn(netdev, "Login timed out, retrying...\n");
879 			retry = true;
880 			adapter->init_done_rc = 0;
881 			retry_count++;
882 			continue;
883 		}
884 
885 		if (adapter->init_done_rc == ABORTED) {
886 			netdev_warn(netdev, "Login aborted, retrying...\n");
887 			retry = true;
888 			adapter->init_done_rc = 0;
889 			retry_count++;
890 			/* FW or device may be busy, so
891 			 * wait a bit before retrying login
892 			 */
893 			msleep(500);
894 		} else if (adapter->init_done_rc == PARTIALSUCCESS) {
895 			retry_count++;
896 			release_sub_crqs(adapter, 1);
897 
898 			retry = true;
899 			netdev_dbg(netdev,
900 				   "Received partial success, retrying...\n");
901 			adapter->init_done_rc = 0;
902 			reinit_completion(&adapter->init_done);
903 			send_query_cap(adapter);
904 			if (!wait_for_completion_timeout(&adapter->init_done,
905 							 timeout)) {
906 				netdev_warn(netdev,
907 					    "Capabilities query timed out\n");
908 				return -1;
909 			}
910 
911 			rc = init_sub_crqs(adapter);
912 			if (rc) {
913 				netdev_warn(netdev,
914 					    "SCRQ initialization failed\n");
915 				return -1;
916 			}
917 
918 			rc = init_sub_crq_irqs(adapter);
919 			if (rc) {
920 				netdev_warn(netdev,
921 					    "SCRQ irq initialization failed\n");
922 				return -1;
923 			}
924 		} else if (adapter->init_done_rc) {
925 			netdev_warn(netdev, "Adapter login failed\n");
926 			return -1;
927 		}
928 	} while (retry);
929 
930 	__ibmvnic_set_mac(netdev, adapter->mac_addr);
931 
932 	return 0;
933 }
934 
935 static void release_login_buffer(struct ibmvnic_adapter *adapter)
936 {
937 	kfree(adapter->login_buf);
938 	adapter->login_buf = NULL;
939 }
940 
941 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
942 {
943 	kfree(adapter->login_rsp_buf);
944 	adapter->login_rsp_buf = NULL;
945 }
946 
947 static void release_resources(struct ibmvnic_adapter *adapter)
948 {
949 	release_vpd_data(adapter);
950 
951 	release_tx_pools(adapter);
952 	release_rx_pools(adapter);
953 
954 	release_napi(adapter);
955 	release_login_rsp_buffer(adapter);
956 }
957 
958 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
959 {
960 	struct net_device *netdev = adapter->netdev;
961 	unsigned long timeout = msecs_to_jiffies(30000);
962 	union ibmvnic_crq crq;
963 	bool resend;
964 	int rc;
965 
966 	netdev_dbg(netdev, "setting link state %d\n", link_state);
967 
968 	memset(&crq, 0, sizeof(crq));
969 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
970 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
971 	crq.logical_link_state.link_state = link_state;
972 
973 	do {
974 		resend = false;
975 
976 		reinit_completion(&adapter->init_done);
977 		rc = ibmvnic_send_crq(adapter, &crq);
978 		if (rc) {
979 			netdev_err(netdev, "Failed to set link state\n");
980 			return rc;
981 		}
982 
983 		if (!wait_for_completion_timeout(&adapter->init_done,
984 						 timeout)) {
985 			netdev_err(netdev, "timeout setting link state\n");
986 			return -1;
987 		}
988 
989 		if (adapter->init_done_rc == PARTIALSUCCESS) {
990 			/* Partuial success, delay and re-send */
991 			mdelay(1000);
992 			resend = true;
993 		} else if (adapter->init_done_rc) {
994 			netdev_warn(netdev, "Unable to set link state, rc=%d\n",
995 				    adapter->init_done_rc);
996 			return adapter->init_done_rc;
997 		}
998 	} while (resend);
999 
1000 	return 0;
1001 }
1002 
1003 static int set_real_num_queues(struct net_device *netdev)
1004 {
1005 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1006 	int rc;
1007 
1008 	netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
1009 		   adapter->req_tx_queues, adapter->req_rx_queues);
1010 
1011 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
1012 	if (rc) {
1013 		netdev_err(netdev, "failed to set the number of tx queues\n");
1014 		return rc;
1015 	}
1016 
1017 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
1018 	if (rc)
1019 		netdev_err(netdev, "failed to set the number of rx queues\n");
1020 
1021 	return rc;
1022 }
1023 
1024 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
1025 {
1026 	struct device *dev = &adapter->vdev->dev;
1027 	union ibmvnic_crq crq;
1028 	int len = 0;
1029 	int rc;
1030 
1031 	if (adapter->vpd->buff)
1032 		len = adapter->vpd->len;
1033 
1034 	mutex_lock(&adapter->fw_lock);
1035 	adapter->fw_done_rc = 0;
1036 	reinit_completion(&adapter->fw_done);
1037 
1038 	crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
1039 	crq.get_vpd_size.cmd = GET_VPD_SIZE;
1040 	rc = ibmvnic_send_crq(adapter, &crq);
1041 	if (rc) {
1042 		mutex_unlock(&adapter->fw_lock);
1043 		return rc;
1044 	}
1045 
1046 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1047 	if (rc) {
1048 		dev_err(dev, "Could not retrieve VPD size, rc = %d\n", rc);
1049 		mutex_unlock(&adapter->fw_lock);
1050 		return rc;
1051 	}
1052 	mutex_unlock(&adapter->fw_lock);
1053 
1054 	if (!adapter->vpd->len)
1055 		return -ENODATA;
1056 
1057 	if (!adapter->vpd->buff)
1058 		adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
1059 	else if (adapter->vpd->len != len)
1060 		adapter->vpd->buff =
1061 			krealloc(adapter->vpd->buff,
1062 				 adapter->vpd->len, GFP_KERNEL);
1063 
1064 	if (!adapter->vpd->buff) {
1065 		dev_err(dev, "Could allocate VPD buffer\n");
1066 		return -ENOMEM;
1067 	}
1068 
1069 	adapter->vpd->dma_addr =
1070 		dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
1071 			       DMA_FROM_DEVICE);
1072 	if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
1073 		dev_err(dev, "Could not map VPD buffer\n");
1074 		kfree(adapter->vpd->buff);
1075 		adapter->vpd->buff = NULL;
1076 		return -ENOMEM;
1077 	}
1078 
1079 	mutex_lock(&adapter->fw_lock);
1080 	adapter->fw_done_rc = 0;
1081 	reinit_completion(&adapter->fw_done);
1082 
1083 	crq.get_vpd.first = IBMVNIC_CRQ_CMD;
1084 	crq.get_vpd.cmd = GET_VPD;
1085 	crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
1086 	crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1087 	rc = ibmvnic_send_crq(adapter, &crq);
1088 	if (rc) {
1089 		kfree(adapter->vpd->buff);
1090 		adapter->vpd->buff = NULL;
1091 		mutex_unlock(&adapter->fw_lock);
1092 		return rc;
1093 	}
1094 
1095 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1096 	if (rc) {
1097 		dev_err(dev, "Unable to retrieve VPD, rc = %d\n", rc);
1098 		kfree(adapter->vpd->buff);
1099 		adapter->vpd->buff = NULL;
1100 		mutex_unlock(&adapter->fw_lock);
1101 		return rc;
1102 	}
1103 
1104 	mutex_unlock(&adapter->fw_lock);
1105 	return 0;
1106 }
1107 
1108 static int init_resources(struct ibmvnic_adapter *adapter)
1109 {
1110 	struct net_device *netdev = adapter->netdev;
1111 	int rc;
1112 
1113 	rc = set_real_num_queues(netdev);
1114 	if (rc)
1115 		return rc;
1116 
1117 	adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1118 	if (!adapter->vpd)
1119 		return -ENOMEM;
1120 
1121 	/* Vital Product Data (VPD) */
1122 	rc = ibmvnic_get_vpd(adapter);
1123 	if (rc) {
1124 		netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1125 		return rc;
1126 	}
1127 
1128 	adapter->map_id = 1;
1129 
1130 	rc = init_napi(adapter);
1131 	if (rc)
1132 		return rc;
1133 
1134 	send_query_map(adapter);
1135 
1136 	rc = init_rx_pools(netdev);
1137 	if (rc)
1138 		return rc;
1139 
1140 	rc = init_tx_pools(netdev);
1141 	return rc;
1142 }
1143 
1144 static int __ibmvnic_open(struct net_device *netdev)
1145 {
1146 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1147 	enum vnic_state prev_state = adapter->state;
1148 	int i, rc;
1149 
1150 	adapter->state = VNIC_OPENING;
1151 	replenish_pools(adapter);
1152 	ibmvnic_napi_enable(adapter);
1153 
1154 	/* We're ready to receive frames, enable the sub-crq interrupts and
1155 	 * set the logical link state to up
1156 	 */
1157 	for (i = 0; i < adapter->req_rx_queues; i++) {
1158 		netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1159 		if (prev_state == VNIC_CLOSED)
1160 			enable_irq(adapter->rx_scrq[i]->irq);
1161 		enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1162 	}
1163 
1164 	for (i = 0; i < adapter->req_tx_queues; i++) {
1165 		netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1166 		if (prev_state == VNIC_CLOSED)
1167 			enable_irq(adapter->tx_scrq[i]->irq);
1168 		enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1169 		netdev_tx_reset_queue(netdev_get_tx_queue(netdev, i));
1170 	}
1171 
1172 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1173 	if (rc) {
1174 		for (i = 0; i < adapter->req_rx_queues; i++)
1175 			napi_disable(&adapter->napi[i]);
1176 		release_resources(adapter);
1177 		return rc;
1178 	}
1179 
1180 	netif_tx_start_all_queues(netdev);
1181 
1182 	if (prev_state == VNIC_CLOSED) {
1183 		for (i = 0; i < adapter->req_rx_queues; i++)
1184 			napi_schedule(&adapter->napi[i]);
1185 	}
1186 
1187 	adapter->state = VNIC_OPEN;
1188 	return rc;
1189 }
1190 
1191 static int ibmvnic_open(struct net_device *netdev)
1192 {
1193 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1194 	int rc;
1195 
1196 	/* If device failover is pending, just set device state and return.
1197 	 * Device operation will be handled by reset routine.
1198 	 */
1199 	if (adapter->failover_pending) {
1200 		adapter->state = VNIC_OPEN;
1201 		return 0;
1202 	}
1203 
1204 	if (adapter->state != VNIC_CLOSED) {
1205 		rc = ibmvnic_login(netdev);
1206 		if (rc)
1207 			goto out;
1208 
1209 		rc = init_resources(adapter);
1210 		if (rc) {
1211 			netdev_err(netdev, "failed to initialize resources\n");
1212 			release_resources(adapter);
1213 			goto out;
1214 		}
1215 	}
1216 
1217 	rc = __ibmvnic_open(netdev);
1218 
1219 out:
1220 	/*
1221 	 * If open fails due to a pending failover, set device state and
1222 	 * return. Device operation will be handled by reset routine.
1223 	 */
1224 	if (rc && adapter->failover_pending) {
1225 		adapter->state = VNIC_OPEN;
1226 		rc = 0;
1227 	}
1228 	return rc;
1229 }
1230 
1231 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1232 {
1233 	struct ibmvnic_rx_pool *rx_pool;
1234 	struct ibmvnic_rx_buff *rx_buff;
1235 	u64 rx_entries;
1236 	int rx_scrqs;
1237 	int i, j;
1238 
1239 	if (!adapter->rx_pool)
1240 		return;
1241 
1242 	rx_scrqs = adapter->num_active_rx_pools;
1243 	rx_entries = adapter->req_rx_add_entries_per_subcrq;
1244 
1245 	/* Free any remaining skbs in the rx buffer pools */
1246 	for (i = 0; i < rx_scrqs; i++) {
1247 		rx_pool = &adapter->rx_pool[i];
1248 		if (!rx_pool || !rx_pool->rx_buff)
1249 			continue;
1250 
1251 		netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1252 		for (j = 0; j < rx_entries; j++) {
1253 			rx_buff = &rx_pool->rx_buff[j];
1254 			if (rx_buff && rx_buff->skb) {
1255 				dev_kfree_skb_any(rx_buff->skb);
1256 				rx_buff->skb = NULL;
1257 			}
1258 		}
1259 	}
1260 }
1261 
1262 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1263 			      struct ibmvnic_tx_pool *tx_pool)
1264 {
1265 	struct ibmvnic_tx_buff *tx_buff;
1266 	u64 tx_entries;
1267 	int i;
1268 
1269 	if (!tx_pool || !tx_pool->tx_buff)
1270 		return;
1271 
1272 	tx_entries = tx_pool->num_buffers;
1273 
1274 	for (i = 0; i < tx_entries; i++) {
1275 		tx_buff = &tx_pool->tx_buff[i];
1276 		if (tx_buff && tx_buff->skb) {
1277 			dev_kfree_skb_any(tx_buff->skb);
1278 			tx_buff->skb = NULL;
1279 		}
1280 	}
1281 }
1282 
1283 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1284 {
1285 	int tx_scrqs;
1286 	int i;
1287 
1288 	if (!adapter->tx_pool || !adapter->tso_pool)
1289 		return;
1290 
1291 	tx_scrqs = adapter->num_active_tx_pools;
1292 
1293 	/* Free any remaining skbs in the tx buffer pools */
1294 	for (i = 0; i < tx_scrqs; i++) {
1295 		netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1296 		clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1297 		clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1298 	}
1299 }
1300 
1301 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1302 {
1303 	struct net_device *netdev = adapter->netdev;
1304 	int i;
1305 
1306 	if (adapter->tx_scrq) {
1307 		for (i = 0; i < adapter->req_tx_queues; i++)
1308 			if (adapter->tx_scrq[i]->irq) {
1309 				netdev_dbg(netdev,
1310 					   "Disabling tx_scrq[%d] irq\n", i);
1311 				disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1312 				disable_irq(adapter->tx_scrq[i]->irq);
1313 			}
1314 	}
1315 
1316 	if (adapter->rx_scrq) {
1317 		for (i = 0; i < adapter->req_rx_queues; i++) {
1318 			if (adapter->rx_scrq[i]->irq) {
1319 				netdev_dbg(netdev,
1320 					   "Disabling rx_scrq[%d] irq\n", i);
1321 				disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1322 				disable_irq(adapter->rx_scrq[i]->irq);
1323 			}
1324 		}
1325 	}
1326 }
1327 
1328 static void ibmvnic_cleanup(struct net_device *netdev)
1329 {
1330 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1331 
1332 	/* ensure that transmissions are stopped if called by do_reset */
1333 	if (test_bit(0, &adapter->resetting))
1334 		netif_tx_disable(netdev);
1335 	else
1336 		netif_tx_stop_all_queues(netdev);
1337 
1338 	ibmvnic_napi_disable(adapter);
1339 	ibmvnic_disable_irqs(adapter);
1340 
1341 	clean_rx_pools(adapter);
1342 	clean_tx_pools(adapter);
1343 }
1344 
1345 static int __ibmvnic_close(struct net_device *netdev)
1346 {
1347 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1348 	int rc = 0;
1349 
1350 	adapter->state = VNIC_CLOSING;
1351 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1352 	if (rc)
1353 		return rc;
1354 	adapter->state = VNIC_CLOSED;
1355 	return 0;
1356 }
1357 
1358 static int ibmvnic_close(struct net_device *netdev)
1359 {
1360 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1361 	int rc;
1362 
1363 	/* If device failover is pending, just set device state and return.
1364 	 * Device operation will be handled by reset routine.
1365 	 */
1366 	if (adapter->failover_pending) {
1367 		adapter->state = VNIC_CLOSED;
1368 		return 0;
1369 	}
1370 
1371 	rc = __ibmvnic_close(netdev);
1372 	ibmvnic_cleanup(netdev);
1373 
1374 	return rc;
1375 }
1376 
1377 /**
1378  * build_hdr_data - creates L2/L3/L4 header data buffer
1379  * @hdr_field - bitfield determining needed headers
1380  * @skb - socket buffer
1381  * @hdr_len - array of header lengths
1382  * @tot_len - total length of data
1383  *
1384  * Reads hdr_field to determine which headers are needed by firmware.
1385  * Builds a buffer containing these headers.  Saves individual header
1386  * lengths and total buffer length to be used to build descriptors.
1387  */
1388 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1389 			  int *hdr_len, u8 *hdr_data)
1390 {
1391 	int len = 0;
1392 	u8 *hdr;
1393 
1394 	if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1395 		hdr_len[0] = sizeof(struct vlan_ethhdr);
1396 	else
1397 		hdr_len[0] = sizeof(struct ethhdr);
1398 
1399 	if (skb->protocol == htons(ETH_P_IP)) {
1400 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
1401 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1402 			hdr_len[2] = tcp_hdrlen(skb);
1403 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1404 			hdr_len[2] = sizeof(struct udphdr);
1405 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1406 		hdr_len[1] = sizeof(struct ipv6hdr);
1407 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1408 			hdr_len[2] = tcp_hdrlen(skb);
1409 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1410 			hdr_len[2] = sizeof(struct udphdr);
1411 	} else if (skb->protocol == htons(ETH_P_ARP)) {
1412 		hdr_len[1] = arp_hdr_len(skb->dev);
1413 		hdr_len[2] = 0;
1414 	}
1415 
1416 	memset(hdr_data, 0, 120);
1417 	if ((hdr_field >> 6) & 1) {
1418 		hdr = skb_mac_header(skb);
1419 		memcpy(hdr_data, hdr, hdr_len[0]);
1420 		len += hdr_len[0];
1421 	}
1422 
1423 	if ((hdr_field >> 5) & 1) {
1424 		hdr = skb_network_header(skb);
1425 		memcpy(hdr_data + len, hdr, hdr_len[1]);
1426 		len += hdr_len[1];
1427 	}
1428 
1429 	if ((hdr_field >> 4) & 1) {
1430 		hdr = skb_transport_header(skb);
1431 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1432 		len += hdr_len[2];
1433 	}
1434 	return len;
1435 }
1436 
1437 /**
1438  * create_hdr_descs - create header and header extension descriptors
1439  * @hdr_field - bitfield determining needed headers
1440  * @data - buffer containing header data
1441  * @len - length of data buffer
1442  * @hdr_len - array of individual header lengths
1443  * @scrq_arr - descriptor array
1444  *
1445  * Creates header and, if needed, header extension descriptors and
1446  * places them in a descriptor array, scrq_arr
1447  */
1448 
1449 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1450 			    union sub_crq *scrq_arr)
1451 {
1452 	union sub_crq hdr_desc;
1453 	int tmp_len = len;
1454 	int num_descs = 0;
1455 	u8 *data, *cur;
1456 	int tmp;
1457 
1458 	while (tmp_len > 0) {
1459 		cur = hdr_data + len - tmp_len;
1460 
1461 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1462 		if (cur != hdr_data) {
1463 			data = hdr_desc.hdr_ext.data;
1464 			tmp = tmp_len > 29 ? 29 : tmp_len;
1465 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1466 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1467 			hdr_desc.hdr_ext.len = tmp;
1468 		} else {
1469 			data = hdr_desc.hdr.data;
1470 			tmp = tmp_len > 24 ? 24 : tmp_len;
1471 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1472 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1473 			hdr_desc.hdr.len = tmp;
1474 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1475 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1476 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1477 			hdr_desc.hdr.flag = hdr_field << 1;
1478 		}
1479 		memcpy(data, cur, tmp);
1480 		tmp_len -= tmp;
1481 		*scrq_arr = hdr_desc;
1482 		scrq_arr++;
1483 		num_descs++;
1484 	}
1485 
1486 	return num_descs;
1487 }
1488 
1489 /**
1490  * build_hdr_descs_arr - build a header descriptor array
1491  * @skb - socket buffer
1492  * @num_entries - number of descriptors to be sent
1493  * @subcrq - first TX descriptor
1494  * @hdr_field - bit field determining which headers will be sent
1495  *
1496  * This function will build a TX descriptor array with applicable
1497  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1498  */
1499 
1500 static void build_hdr_descs_arr(struct sk_buff *skb,
1501 				union sub_crq *indir_arr,
1502 				int *num_entries, u8 hdr_field)
1503 {
1504 	int hdr_len[3] = {0, 0, 0};
1505 	u8 hdr_data[140] = {0};
1506 	int tot_len;
1507 
1508 	tot_len = build_hdr_data(hdr_field, skb, hdr_len,
1509 				 hdr_data);
1510 	*num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1511 					 indir_arr + 1);
1512 }
1513 
1514 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1515 				    struct net_device *netdev)
1516 {
1517 	/* For some backing devices, mishandling of small packets
1518 	 * can result in a loss of connection or TX stall. Device
1519 	 * architects recommend that no packet should be smaller
1520 	 * than the minimum MTU value provided to the driver, so
1521 	 * pad any packets to that length
1522 	 */
1523 	if (skb->len < netdev->min_mtu)
1524 		return skb_put_padto(skb, netdev->min_mtu);
1525 
1526 	return 0;
1527 }
1528 
1529 static void ibmvnic_tx_scrq_clean_buffer(struct ibmvnic_adapter *adapter,
1530 					 struct ibmvnic_sub_crq_queue *tx_scrq)
1531 {
1532 	struct ibmvnic_ind_xmit_queue *ind_bufp;
1533 	struct ibmvnic_tx_buff *tx_buff;
1534 	struct ibmvnic_tx_pool *tx_pool;
1535 	union sub_crq tx_scrq_entry;
1536 	int queue_num;
1537 	int entries;
1538 	int index;
1539 	int i;
1540 
1541 	ind_bufp = &tx_scrq->ind_buf;
1542 	entries = (u64)ind_bufp->index;
1543 	queue_num = tx_scrq->pool_index;
1544 
1545 	for (i = entries - 1; i >= 0; --i) {
1546 		tx_scrq_entry = ind_bufp->indir_arr[i];
1547 		if (tx_scrq_entry.v1.type != IBMVNIC_TX_DESC)
1548 			continue;
1549 		index = be32_to_cpu(tx_scrq_entry.v1.correlator);
1550 		if (index & IBMVNIC_TSO_POOL_MASK) {
1551 			tx_pool = &adapter->tso_pool[queue_num];
1552 			index &= ~IBMVNIC_TSO_POOL_MASK;
1553 		} else {
1554 			tx_pool = &adapter->tx_pool[queue_num];
1555 		}
1556 		tx_pool->free_map[tx_pool->consumer_index] = index;
1557 		tx_pool->consumer_index = tx_pool->consumer_index == 0 ?
1558 					  tx_pool->num_buffers - 1 :
1559 					  tx_pool->consumer_index - 1;
1560 		tx_buff = &tx_pool->tx_buff[index];
1561 		adapter->netdev->stats.tx_packets--;
1562 		adapter->netdev->stats.tx_bytes -= tx_buff->skb->len;
1563 		adapter->tx_stats_buffers[queue_num].packets--;
1564 		adapter->tx_stats_buffers[queue_num].bytes -=
1565 						tx_buff->skb->len;
1566 		dev_kfree_skb_any(tx_buff->skb);
1567 		tx_buff->skb = NULL;
1568 		adapter->netdev->stats.tx_dropped++;
1569 	}
1570 	ind_bufp->index = 0;
1571 	if (atomic_sub_return(entries, &tx_scrq->used) <=
1572 	    (adapter->req_tx_entries_per_subcrq / 2) &&
1573 	    __netif_subqueue_stopped(adapter->netdev, queue_num)) {
1574 		netif_wake_subqueue(adapter->netdev, queue_num);
1575 		netdev_dbg(adapter->netdev, "Started queue %d\n",
1576 			   queue_num);
1577 	}
1578 }
1579 
1580 static int ibmvnic_tx_scrq_flush(struct ibmvnic_adapter *adapter,
1581 				 struct ibmvnic_sub_crq_queue *tx_scrq)
1582 {
1583 	struct ibmvnic_ind_xmit_queue *ind_bufp;
1584 	u64 dma_addr;
1585 	u64 entries;
1586 	u64 handle;
1587 	int rc;
1588 
1589 	ind_bufp = &tx_scrq->ind_buf;
1590 	dma_addr = (u64)ind_bufp->indir_dma;
1591 	entries = (u64)ind_bufp->index;
1592 	handle = tx_scrq->handle;
1593 
1594 	if (!entries)
1595 		return 0;
1596 	rc = send_subcrq_indirect(adapter, handle, dma_addr, entries);
1597 	if (rc)
1598 		ibmvnic_tx_scrq_clean_buffer(adapter, tx_scrq);
1599 	else
1600 		ind_bufp->index = 0;
1601 	return 0;
1602 }
1603 
1604 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1605 {
1606 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1607 	int queue_num = skb_get_queue_mapping(skb);
1608 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1609 	struct device *dev = &adapter->vdev->dev;
1610 	struct ibmvnic_ind_xmit_queue *ind_bufp;
1611 	struct ibmvnic_tx_buff *tx_buff = NULL;
1612 	struct ibmvnic_sub_crq_queue *tx_scrq;
1613 	struct ibmvnic_tx_pool *tx_pool;
1614 	unsigned int tx_send_failed = 0;
1615 	netdev_tx_t ret = NETDEV_TX_OK;
1616 	unsigned int tx_map_failed = 0;
1617 	union sub_crq indir_arr[16];
1618 	unsigned int tx_dropped = 0;
1619 	unsigned int tx_packets = 0;
1620 	unsigned int tx_bytes = 0;
1621 	dma_addr_t data_dma_addr;
1622 	struct netdev_queue *txq;
1623 	unsigned long lpar_rc;
1624 	union sub_crq tx_crq;
1625 	unsigned int offset;
1626 	int num_entries = 1;
1627 	unsigned char *dst;
1628 	int index = 0;
1629 	u8 proto = 0;
1630 
1631 	tx_scrq = adapter->tx_scrq[queue_num];
1632 	txq = netdev_get_tx_queue(netdev, queue_num);
1633 	ind_bufp = &tx_scrq->ind_buf;
1634 
1635 	if (test_bit(0, &adapter->resetting)) {
1636 		if (!netif_subqueue_stopped(netdev, skb))
1637 			netif_stop_subqueue(netdev, queue_num);
1638 		dev_kfree_skb_any(skb);
1639 
1640 		tx_send_failed++;
1641 		tx_dropped++;
1642 		ret = NETDEV_TX_OK;
1643 		ibmvnic_tx_scrq_flush(adapter, tx_scrq);
1644 		goto out;
1645 	}
1646 
1647 	if (ibmvnic_xmit_workarounds(skb, netdev)) {
1648 		tx_dropped++;
1649 		tx_send_failed++;
1650 		ret = NETDEV_TX_OK;
1651 		ibmvnic_tx_scrq_flush(adapter, tx_scrq);
1652 		goto out;
1653 	}
1654 	if (skb_is_gso(skb))
1655 		tx_pool = &adapter->tso_pool[queue_num];
1656 	else
1657 		tx_pool = &adapter->tx_pool[queue_num];
1658 
1659 	index = tx_pool->free_map[tx_pool->consumer_index];
1660 
1661 	if (index == IBMVNIC_INVALID_MAP) {
1662 		dev_kfree_skb_any(skb);
1663 		tx_send_failed++;
1664 		tx_dropped++;
1665 		ret = NETDEV_TX_OK;
1666 		ibmvnic_tx_scrq_flush(adapter, tx_scrq);
1667 		goto out;
1668 	}
1669 
1670 	tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1671 
1672 	offset = index * tx_pool->buf_size;
1673 	dst = tx_pool->long_term_buff.buff + offset;
1674 	memset(dst, 0, tx_pool->buf_size);
1675 	data_dma_addr = tx_pool->long_term_buff.addr + offset;
1676 
1677 	if (skb_shinfo(skb)->nr_frags) {
1678 		int cur, i;
1679 
1680 		/* Copy the head */
1681 		skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1682 		cur = skb_headlen(skb);
1683 
1684 		/* Copy the frags */
1685 		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1686 			const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1687 
1688 			memcpy(dst + cur,
1689 			       page_address(skb_frag_page(frag)) +
1690 			       skb_frag_off(frag), skb_frag_size(frag));
1691 			cur += skb_frag_size(frag);
1692 		}
1693 	} else {
1694 		skb_copy_from_linear_data(skb, dst, skb->len);
1695 	}
1696 
1697 	tx_pool->consumer_index =
1698 	    (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1699 
1700 	tx_buff = &tx_pool->tx_buff[index];
1701 	tx_buff->skb = skb;
1702 	tx_buff->index = index;
1703 	tx_buff->pool_index = queue_num;
1704 
1705 	memset(&tx_crq, 0, sizeof(tx_crq));
1706 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1707 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1708 	tx_crq.v1.n_crq_elem = 1;
1709 	tx_crq.v1.n_sge = 1;
1710 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1711 
1712 	if (skb_is_gso(skb))
1713 		tx_crq.v1.correlator =
1714 			cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1715 	else
1716 		tx_crq.v1.correlator = cpu_to_be32(index);
1717 	tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1718 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1719 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1720 
1721 	if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1722 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1723 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1724 	}
1725 
1726 	if (skb->protocol == htons(ETH_P_IP)) {
1727 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1728 		proto = ip_hdr(skb)->protocol;
1729 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1730 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1731 		proto = ipv6_hdr(skb)->nexthdr;
1732 	}
1733 
1734 	if (proto == IPPROTO_TCP)
1735 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1736 	else if (proto == IPPROTO_UDP)
1737 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1738 
1739 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1740 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1741 		hdrs += 2;
1742 	}
1743 	if (skb_is_gso(skb)) {
1744 		tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1745 		tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1746 		hdrs += 2;
1747 	}
1748 
1749 	if ((*hdrs >> 7) & 1)
1750 		build_hdr_descs_arr(skb, indir_arr, &num_entries, *hdrs);
1751 
1752 	tx_crq.v1.n_crq_elem = num_entries;
1753 	tx_buff->num_entries = num_entries;
1754 	/* flush buffer if current entry can not fit */
1755 	if (num_entries + ind_bufp->index > IBMVNIC_MAX_IND_DESCS) {
1756 		lpar_rc = ibmvnic_tx_scrq_flush(adapter, tx_scrq);
1757 		if (lpar_rc != H_SUCCESS)
1758 			goto tx_flush_err;
1759 	}
1760 
1761 	indir_arr[0] = tx_crq;
1762 	memcpy(&ind_bufp->indir_arr[ind_bufp->index], &indir_arr[0],
1763 	       num_entries * sizeof(struct ibmvnic_generic_scrq));
1764 	ind_bufp->index += num_entries;
1765 	if (__netdev_tx_sent_queue(txq, skb->len,
1766 				   netdev_xmit_more() &&
1767 				   ind_bufp->index < IBMVNIC_MAX_IND_DESCS)) {
1768 		lpar_rc = ibmvnic_tx_scrq_flush(adapter, tx_scrq);
1769 		if (lpar_rc != H_SUCCESS)
1770 			goto tx_err;
1771 	}
1772 
1773 	if (atomic_add_return(num_entries, &tx_scrq->used)
1774 					>= adapter->req_tx_entries_per_subcrq) {
1775 		netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1776 		netif_stop_subqueue(netdev, queue_num);
1777 	}
1778 
1779 	tx_packets++;
1780 	tx_bytes += skb->len;
1781 	txq->trans_start = jiffies;
1782 	ret = NETDEV_TX_OK;
1783 	goto out;
1784 
1785 tx_flush_err:
1786 	dev_kfree_skb_any(skb);
1787 	tx_buff->skb = NULL;
1788 	tx_pool->consumer_index = tx_pool->consumer_index == 0 ?
1789 				  tx_pool->num_buffers - 1 :
1790 				  tx_pool->consumer_index - 1;
1791 	tx_dropped++;
1792 tx_err:
1793 	if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1794 		dev_err_ratelimited(dev, "tx: send failed\n");
1795 
1796 	if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1797 		/* Disable TX and report carrier off if queue is closed
1798 		 * or pending failover.
1799 		 * Firmware guarantees that a signal will be sent to the
1800 		 * driver, triggering a reset or some other action.
1801 		 */
1802 		netif_tx_stop_all_queues(netdev);
1803 		netif_carrier_off(netdev);
1804 	}
1805 out:
1806 	netdev->stats.tx_dropped += tx_dropped;
1807 	netdev->stats.tx_bytes += tx_bytes;
1808 	netdev->stats.tx_packets += tx_packets;
1809 	adapter->tx_send_failed += tx_send_failed;
1810 	adapter->tx_map_failed += tx_map_failed;
1811 	adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1812 	adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1813 	adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1814 
1815 	return ret;
1816 }
1817 
1818 static void ibmvnic_set_multi(struct net_device *netdev)
1819 {
1820 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1821 	struct netdev_hw_addr *ha;
1822 	union ibmvnic_crq crq;
1823 
1824 	memset(&crq, 0, sizeof(crq));
1825 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1826 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1827 
1828 	if (netdev->flags & IFF_PROMISC) {
1829 		if (!adapter->promisc_supported)
1830 			return;
1831 	} else {
1832 		if (netdev->flags & IFF_ALLMULTI) {
1833 			/* Accept all multicast */
1834 			memset(&crq, 0, sizeof(crq));
1835 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1836 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1837 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1838 			ibmvnic_send_crq(adapter, &crq);
1839 		} else if (netdev_mc_empty(netdev)) {
1840 			/* Reject all multicast */
1841 			memset(&crq, 0, sizeof(crq));
1842 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1843 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1844 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1845 			ibmvnic_send_crq(adapter, &crq);
1846 		} else {
1847 			/* Accept one or more multicast(s) */
1848 			netdev_for_each_mc_addr(ha, netdev) {
1849 				memset(&crq, 0, sizeof(crq));
1850 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1851 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1852 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1853 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1854 						ha->addr);
1855 				ibmvnic_send_crq(adapter, &crq);
1856 			}
1857 		}
1858 	}
1859 }
1860 
1861 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1862 {
1863 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1864 	union ibmvnic_crq crq;
1865 	int rc;
1866 
1867 	if (!is_valid_ether_addr(dev_addr)) {
1868 		rc = -EADDRNOTAVAIL;
1869 		goto err;
1870 	}
1871 
1872 	memset(&crq, 0, sizeof(crq));
1873 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1874 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1875 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1876 
1877 	mutex_lock(&adapter->fw_lock);
1878 	adapter->fw_done_rc = 0;
1879 	reinit_completion(&adapter->fw_done);
1880 
1881 	rc = ibmvnic_send_crq(adapter, &crq);
1882 	if (rc) {
1883 		rc = -EIO;
1884 		mutex_unlock(&adapter->fw_lock);
1885 		goto err;
1886 	}
1887 
1888 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1889 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1890 	if (rc || adapter->fw_done_rc) {
1891 		rc = -EIO;
1892 		mutex_unlock(&adapter->fw_lock);
1893 		goto err;
1894 	}
1895 	mutex_unlock(&adapter->fw_lock);
1896 	return 0;
1897 err:
1898 	ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1899 	return rc;
1900 }
1901 
1902 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1903 {
1904 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1905 	struct sockaddr *addr = p;
1906 	int rc;
1907 
1908 	rc = 0;
1909 	if (!is_valid_ether_addr(addr->sa_data))
1910 		return -EADDRNOTAVAIL;
1911 
1912 	if (adapter->state != VNIC_PROBED) {
1913 		ether_addr_copy(adapter->mac_addr, addr->sa_data);
1914 		rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1915 	}
1916 
1917 	return rc;
1918 }
1919 
1920 /**
1921  * do_change_param_reset returns zero if we are able to keep processing reset
1922  * events, or non-zero if we hit a fatal error and must halt.
1923  */
1924 static int do_change_param_reset(struct ibmvnic_adapter *adapter,
1925 				 struct ibmvnic_rwi *rwi,
1926 				 u32 reset_state)
1927 {
1928 	struct net_device *netdev = adapter->netdev;
1929 	int i, rc;
1930 
1931 	netdev_dbg(adapter->netdev, "Change param resetting driver (%d)\n",
1932 		   rwi->reset_reason);
1933 
1934 	netif_carrier_off(netdev);
1935 	adapter->reset_reason = rwi->reset_reason;
1936 
1937 	ibmvnic_cleanup(netdev);
1938 
1939 	if (reset_state == VNIC_OPEN) {
1940 		rc = __ibmvnic_close(netdev);
1941 		if (rc)
1942 			return rc;
1943 	}
1944 
1945 	release_resources(adapter);
1946 	release_sub_crqs(adapter, 1);
1947 	release_crq_queue(adapter);
1948 
1949 	adapter->state = VNIC_PROBED;
1950 
1951 	rc = init_crq_queue(adapter);
1952 
1953 	if (rc) {
1954 		netdev_err(adapter->netdev,
1955 			   "Couldn't initialize crq. rc=%d\n", rc);
1956 		return rc;
1957 	}
1958 
1959 	rc = ibmvnic_reset_init(adapter, true);
1960 	if (rc)
1961 		return IBMVNIC_INIT_FAILED;
1962 
1963 	/* If the adapter was in PROBE state prior to the reset,
1964 	 * exit here.
1965 	 */
1966 	if (reset_state == VNIC_PROBED)
1967 		return 0;
1968 
1969 	rc = ibmvnic_login(netdev);
1970 	if (rc) {
1971 		adapter->state = reset_state;
1972 		return rc;
1973 	}
1974 
1975 	rc = init_resources(adapter);
1976 	if (rc)
1977 		return rc;
1978 
1979 	ibmvnic_disable_irqs(adapter);
1980 
1981 	adapter->state = VNIC_CLOSED;
1982 
1983 	if (reset_state == VNIC_CLOSED)
1984 		return 0;
1985 
1986 	rc = __ibmvnic_open(netdev);
1987 	if (rc)
1988 		return IBMVNIC_OPEN_FAILED;
1989 
1990 	/* refresh device's multicast list */
1991 	ibmvnic_set_multi(netdev);
1992 
1993 	/* kick napi */
1994 	for (i = 0; i < adapter->req_rx_queues; i++)
1995 		napi_schedule(&adapter->napi[i]);
1996 
1997 	return 0;
1998 }
1999 
2000 /**
2001  * do_reset returns zero if we are able to keep processing reset events, or
2002  * non-zero if we hit a fatal error and must halt.
2003  */
2004 static int do_reset(struct ibmvnic_adapter *adapter,
2005 		    struct ibmvnic_rwi *rwi, u32 reset_state)
2006 {
2007 	u64 old_num_rx_queues, old_num_tx_queues;
2008 	u64 old_num_rx_slots, old_num_tx_slots;
2009 	struct net_device *netdev = adapter->netdev;
2010 	int i, rc;
2011 
2012 	netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
2013 		   rwi->reset_reason);
2014 
2015 	rtnl_lock();
2016 	/*
2017 	 * Now that we have the rtnl lock, clear any pending failover.
2018 	 * This will ensure ibmvnic_open() has either completed or will
2019 	 * block until failover is complete.
2020 	 */
2021 	if (rwi->reset_reason == VNIC_RESET_FAILOVER)
2022 		adapter->failover_pending = false;
2023 
2024 	netif_carrier_off(netdev);
2025 	adapter->reset_reason = rwi->reset_reason;
2026 
2027 	old_num_rx_queues = adapter->req_rx_queues;
2028 	old_num_tx_queues = adapter->req_tx_queues;
2029 	old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
2030 	old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
2031 
2032 	ibmvnic_cleanup(netdev);
2033 
2034 	if (reset_state == VNIC_OPEN &&
2035 	    adapter->reset_reason != VNIC_RESET_MOBILITY &&
2036 	    adapter->reset_reason != VNIC_RESET_FAILOVER) {
2037 		adapter->state = VNIC_CLOSING;
2038 
2039 		/* Release the RTNL lock before link state change and
2040 		 * re-acquire after the link state change to allow
2041 		 * linkwatch_event to grab the RTNL lock and run during
2042 		 * a reset.
2043 		 */
2044 		rtnl_unlock();
2045 		rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
2046 		rtnl_lock();
2047 		if (rc)
2048 			goto out;
2049 
2050 		if (adapter->state != VNIC_CLOSING) {
2051 			rc = -1;
2052 			goto out;
2053 		}
2054 
2055 		adapter->state = VNIC_CLOSED;
2056 	}
2057 
2058 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
2059 		/* remove the closed state so when we call open it appears
2060 		 * we are coming from the probed state.
2061 		 */
2062 		adapter->state = VNIC_PROBED;
2063 
2064 		if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
2065 			rc = ibmvnic_reenable_crq_queue(adapter);
2066 			release_sub_crqs(adapter, 1);
2067 		} else {
2068 			rc = ibmvnic_reset_crq(adapter);
2069 			if (rc == H_CLOSED || rc == H_SUCCESS) {
2070 				rc = vio_enable_interrupts(adapter->vdev);
2071 				if (rc)
2072 					netdev_err(adapter->netdev,
2073 						   "Reset failed to enable interrupts. rc=%d\n",
2074 						   rc);
2075 			}
2076 		}
2077 
2078 		if (rc) {
2079 			netdev_err(adapter->netdev,
2080 				   "Reset couldn't initialize crq. rc=%d\n", rc);
2081 			goto out;
2082 		}
2083 
2084 		rc = ibmvnic_reset_init(adapter, true);
2085 		if (rc) {
2086 			rc = IBMVNIC_INIT_FAILED;
2087 			goto out;
2088 		}
2089 
2090 		/* If the adapter was in PROBE state prior to the reset,
2091 		 * exit here.
2092 		 */
2093 		if (reset_state == VNIC_PROBED) {
2094 			rc = 0;
2095 			goto out;
2096 		}
2097 
2098 		rc = ibmvnic_login(netdev);
2099 		if (rc) {
2100 			adapter->state = reset_state;
2101 			goto out;
2102 		}
2103 
2104 		if (adapter->req_rx_queues != old_num_rx_queues ||
2105 		    adapter->req_tx_queues != old_num_tx_queues ||
2106 		    adapter->req_rx_add_entries_per_subcrq !=
2107 		    old_num_rx_slots ||
2108 		    adapter->req_tx_entries_per_subcrq !=
2109 		    old_num_tx_slots ||
2110 		    !adapter->rx_pool ||
2111 		    !adapter->tso_pool ||
2112 		    !adapter->tx_pool) {
2113 			release_rx_pools(adapter);
2114 			release_tx_pools(adapter);
2115 			release_napi(adapter);
2116 			release_vpd_data(adapter);
2117 
2118 			rc = init_resources(adapter);
2119 			if (rc)
2120 				goto out;
2121 
2122 		} else {
2123 			rc = reset_tx_pools(adapter);
2124 			if (rc) {
2125 				netdev_dbg(adapter->netdev, "reset tx pools failed (%d)\n",
2126 						rc);
2127 				goto out;
2128 			}
2129 
2130 			rc = reset_rx_pools(adapter);
2131 			if (rc) {
2132 				netdev_dbg(adapter->netdev, "reset rx pools failed (%d)\n",
2133 						rc);
2134 				goto out;
2135 			}
2136 		}
2137 		ibmvnic_disable_irqs(adapter);
2138 	}
2139 	adapter->state = VNIC_CLOSED;
2140 
2141 	if (reset_state == VNIC_CLOSED) {
2142 		rc = 0;
2143 		goto out;
2144 	}
2145 
2146 	rc = __ibmvnic_open(netdev);
2147 	if (rc) {
2148 		rc = IBMVNIC_OPEN_FAILED;
2149 		goto out;
2150 	}
2151 
2152 	/* refresh device's multicast list */
2153 	ibmvnic_set_multi(netdev);
2154 
2155 	/* kick napi */
2156 	for (i = 0; i < adapter->req_rx_queues; i++)
2157 		napi_schedule(&adapter->napi[i]);
2158 
2159 	if (adapter->reset_reason != VNIC_RESET_FAILOVER)
2160 		call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
2161 
2162 	rc = 0;
2163 
2164 out:
2165 	rtnl_unlock();
2166 
2167 	return rc;
2168 }
2169 
2170 static int do_hard_reset(struct ibmvnic_adapter *adapter,
2171 			 struct ibmvnic_rwi *rwi, u32 reset_state)
2172 {
2173 	struct net_device *netdev = adapter->netdev;
2174 	int rc;
2175 
2176 	netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
2177 		   rwi->reset_reason);
2178 
2179 	netif_carrier_off(netdev);
2180 	adapter->reset_reason = rwi->reset_reason;
2181 
2182 	ibmvnic_cleanup(netdev);
2183 	release_resources(adapter);
2184 	release_sub_crqs(adapter, 0);
2185 	release_crq_queue(adapter);
2186 
2187 	/* remove the closed state so when we call open it appears
2188 	 * we are coming from the probed state.
2189 	 */
2190 	adapter->state = VNIC_PROBED;
2191 
2192 	reinit_completion(&adapter->init_done);
2193 	rc = init_crq_queue(adapter);
2194 	if (rc) {
2195 		netdev_err(adapter->netdev,
2196 			   "Couldn't initialize crq. rc=%d\n", rc);
2197 		return rc;
2198 	}
2199 
2200 	rc = ibmvnic_reset_init(adapter, false);
2201 	if (rc)
2202 		return rc;
2203 
2204 	/* If the adapter was in PROBE state prior to the reset,
2205 	 * exit here.
2206 	 */
2207 	if (reset_state == VNIC_PROBED)
2208 		return 0;
2209 
2210 	rc = ibmvnic_login(netdev);
2211 	if (rc) {
2212 		adapter->state = VNIC_PROBED;
2213 		return 0;
2214 	}
2215 
2216 	rc = init_resources(adapter);
2217 	if (rc)
2218 		return rc;
2219 
2220 	ibmvnic_disable_irqs(adapter);
2221 	adapter->state = VNIC_CLOSED;
2222 
2223 	if (reset_state == VNIC_CLOSED)
2224 		return 0;
2225 
2226 	rc = __ibmvnic_open(netdev);
2227 	if (rc)
2228 		return IBMVNIC_OPEN_FAILED;
2229 
2230 	return 0;
2231 }
2232 
2233 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
2234 {
2235 	struct ibmvnic_rwi *rwi;
2236 	unsigned long flags;
2237 
2238 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2239 
2240 	if (!list_empty(&adapter->rwi_list)) {
2241 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
2242 				       list);
2243 		list_del(&rwi->list);
2244 	} else {
2245 		rwi = NULL;
2246 	}
2247 
2248 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2249 	return rwi;
2250 }
2251 
2252 static void free_all_rwi(struct ibmvnic_adapter *adapter)
2253 {
2254 	struct ibmvnic_rwi *rwi;
2255 
2256 	rwi = get_next_rwi(adapter);
2257 	while (rwi) {
2258 		kfree(rwi);
2259 		rwi = get_next_rwi(adapter);
2260 	}
2261 }
2262 
2263 static void __ibmvnic_reset(struct work_struct *work)
2264 {
2265 	struct ibmvnic_rwi *rwi;
2266 	struct ibmvnic_adapter *adapter;
2267 	bool saved_state = false;
2268 	unsigned long flags;
2269 	u32 reset_state;
2270 	int rc = 0;
2271 
2272 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
2273 
2274 	if (test_and_set_bit_lock(0, &adapter->resetting)) {
2275 		schedule_delayed_work(&adapter->ibmvnic_delayed_reset,
2276 				      IBMVNIC_RESET_DELAY);
2277 		return;
2278 	}
2279 
2280 	rwi = get_next_rwi(adapter);
2281 	while (rwi) {
2282 		spin_lock_irqsave(&adapter->state_lock, flags);
2283 
2284 		if (adapter->state == VNIC_REMOVING ||
2285 		    adapter->state == VNIC_REMOVED) {
2286 			spin_unlock_irqrestore(&adapter->state_lock, flags);
2287 			kfree(rwi);
2288 			rc = EBUSY;
2289 			break;
2290 		}
2291 
2292 		if (!saved_state) {
2293 			reset_state = adapter->state;
2294 			adapter->state = VNIC_RESETTING;
2295 			saved_state = true;
2296 		}
2297 		spin_unlock_irqrestore(&adapter->state_lock, flags);
2298 
2299 		if (rwi->reset_reason == VNIC_RESET_CHANGE_PARAM) {
2300 			/* CHANGE_PARAM requestor holds rtnl_lock */
2301 			rc = do_change_param_reset(adapter, rwi, reset_state);
2302 		} else if (adapter->force_reset_recovery) {
2303 			/*
2304 			 * Since we are doing a hard reset now, clear the
2305 			 * failover_pending flag so we don't ignore any
2306 			 * future MOBILITY or other resets.
2307 			 */
2308 			adapter->failover_pending = false;
2309 
2310 			/* Transport event occurred during previous reset */
2311 			if (adapter->wait_for_reset) {
2312 				/* Previous was CHANGE_PARAM; caller locked */
2313 				adapter->force_reset_recovery = false;
2314 				rc = do_hard_reset(adapter, rwi, reset_state);
2315 			} else {
2316 				rtnl_lock();
2317 				adapter->force_reset_recovery = false;
2318 				rc = do_hard_reset(adapter, rwi, reset_state);
2319 				rtnl_unlock();
2320 			}
2321 		} else if (!(rwi->reset_reason == VNIC_RESET_FATAL &&
2322 				adapter->from_passive_init)) {
2323 			rc = do_reset(adapter, rwi, reset_state);
2324 		}
2325 		kfree(rwi);
2326 		if (rc == IBMVNIC_OPEN_FAILED) {
2327 			if (list_empty(&adapter->rwi_list))
2328 				adapter->state = VNIC_CLOSED;
2329 			else
2330 				adapter->state = reset_state;
2331 			rc = 0;
2332 		} else if (rc && rc != IBMVNIC_INIT_FAILED &&
2333 		    !adapter->force_reset_recovery)
2334 			break;
2335 
2336 		rwi = get_next_rwi(adapter);
2337 
2338 		if (rwi && (rwi->reset_reason == VNIC_RESET_FAILOVER ||
2339 			    rwi->reset_reason == VNIC_RESET_MOBILITY))
2340 			adapter->force_reset_recovery = true;
2341 	}
2342 
2343 	if (adapter->wait_for_reset) {
2344 		adapter->reset_done_rc = rc;
2345 		complete(&adapter->reset_done);
2346 	}
2347 
2348 	if (rc) {
2349 		netdev_dbg(adapter->netdev, "Reset failed\n");
2350 		free_all_rwi(adapter);
2351 	}
2352 
2353 	clear_bit_unlock(0, &adapter->resetting);
2354 }
2355 
2356 static void __ibmvnic_delayed_reset(struct work_struct *work)
2357 {
2358 	struct ibmvnic_adapter *adapter;
2359 
2360 	adapter = container_of(work, struct ibmvnic_adapter,
2361 			       ibmvnic_delayed_reset.work);
2362 	__ibmvnic_reset(&adapter->ibmvnic_reset);
2363 }
2364 
2365 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2366 			 enum ibmvnic_reset_reason reason)
2367 {
2368 	struct list_head *entry, *tmp_entry;
2369 	struct ibmvnic_rwi *rwi, *tmp;
2370 	struct net_device *netdev = adapter->netdev;
2371 	unsigned long flags;
2372 	int ret;
2373 
2374 	/*
2375 	 * If failover is pending don't schedule any other reset.
2376 	 * Instead let the failover complete. If there is already a
2377 	 * a failover reset scheduled, we will detect and drop the
2378 	 * duplicate reset when walking the ->rwi_list below.
2379 	 */
2380 	if (adapter->state == VNIC_REMOVING ||
2381 	    adapter->state == VNIC_REMOVED ||
2382 	    (adapter->failover_pending && reason != VNIC_RESET_FAILOVER)) {
2383 		ret = EBUSY;
2384 		netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2385 		goto err;
2386 	}
2387 
2388 	if (adapter->state == VNIC_PROBING) {
2389 		netdev_warn(netdev, "Adapter reset during probe\n");
2390 		ret = adapter->init_done_rc = EAGAIN;
2391 		goto err;
2392 	}
2393 
2394 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2395 
2396 	list_for_each(entry, &adapter->rwi_list) {
2397 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
2398 		if (tmp->reset_reason == reason) {
2399 			netdev_dbg(netdev, "Skipping matching reset\n");
2400 			spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2401 			ret = EBUSY;
2402 			goto err;
2403 		}
2404 	}
2405 
2406 	rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2407 	if (!rwi) {
2408 		spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2409 		ibmvnic_close(netdev);
2410 		ret = ENOMEM;
2411 		goto err;
2412 	}
2413 	/* if we just received a transport event,
2414 	 * flush reset queue and process this reset
2415 	 */
2416 	if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2417 		list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2418 			list_del(entry);
2419 	}
2420 	rwi->reset_reason = reason;
2421 	list_add_tail(&rwi->list, &adapter->rwi_list);
2422 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2423 	netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2424 	schedule_work(&adapter->ibmvnic_reset);
2425 
2426 	return 0;
2427 err:
2428 	return -ret;
2429 }
2430 
2431 static void ibmvnic_tx_timeout(struct net_device *dev, unsigned int txqueue)
2432 {
2433 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2434 
2435 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2436 }
2437 
2438 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2439 				  struct ibmvnic_rx_buff *rx_buff)
2440 {
2441 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2442 
2443 	rx_buff->skb = NULL;
2444 
2445 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2446 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2447 
2448 	atomic_dec(&pool->available);
2449 }
2450 
2451 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2452 {
2453 	struct ibmvnic_sub_crq_queue *rx_scrq;
2454 	struct ibmvnic_adapter *adapter;
2455 	struct net_device *netdev;
2456 	int frames_processed;
2457 	int scrq_num;
2458 
2459 	netdev = napi->dev;
2460 	adapter = netdev_priv(netdev);
2461 	scrq_num = (int)(napi - adapter->napi);
2462 	frames_processed = 0;
2463 	rx_scrq = adapter->rx_scrq[scrq_num];
2464 
2465 restart_poll:
2466 	while (frames_processed < budget) {
2467 		struct sk_buff *skb;
2468 		struct ibmvnic_rx_buff *rx_buff;
2469 		union sub_crq *next;
2470 		u32 length;
2471 		u16 offset;
2472 		u8 flags = 0;
2473 
2474 		if (unlikely(test_bit(0, &adapter->resetting) &&
2475 			     adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2476 			enable_scrq_irq(adapter, rx_scrq);
2477 			napi_complete_done(napi, frames_processed);
2478 			return frames_processed;
2479 		}
2480 
2481 		if (!pending_scrq(adapter, rx_scrq))
2482 			break;
2483 		next = ibmvnic_next_scrq(adapter, rx_scrq);
2484 		rx_buff =
2485 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2486 							  rx_comp.correlator);
2487 		/* do error checking */
2488 		if (next->rx_comp.rc) {
2489 			netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2490 				   be16_to_cpu(next->rx_comp.rc));
2491 			/* free the entry */
2492 			next->rx_comp.first = 0;
2493 			dev_kfree_skb_any(rx_buff->skb);
2494 			remove_buff_from_pool(adapter, rx_buff);
2495 			continue;
2496 		} else if (!rx_buff->skb) {
2497 			/* free the entry */
2498 			next->rx_comp.first = 0;
2499 			remove_buff_from_pool(adapter, rx_buff);
2500 			continue;
2501 		}
2502 
2503 		length = be32_to_cpu(next->rx_comp.len);
2504 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
2505 		flags = next->rx_comp.flags;
2506 		skb = rx_buff->skb;
2507 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
2508 					length);
2509 
2510 		/* VLAN Header has been stripped by the system firmware and
2511 		 * needs to be inserted by the driver
2512 		 */
2513 		if (adapter->rx_vlan_header_insertion &&
2514 		    (flags & IBMVNIC_VLAN_STRIPPED))
2515 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2516 					       ntohs(next->rx_comp.vlan_tci));
2517 
2518 		/* free the entry */
2519 		next->rx_comp.first = 0;
2520 		remove_buff_from_pool(adapter, rx_buff);
2521 
2522 		skb_put(skb, length);
2523 		skb->protocol = eth_type_trans(skb, netdev);
2524 		skb_record_rx_queue(skb, scrq_num);
2525 
2526 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2527 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2528 			skb->ip_summed = CHECKSUM_UNNECESSARY;
2529 		}
2530 
2531 		length = skb->len;
2532 		napi_gro_receive(napi, skb); /* send it up */
2533 		netdev->stats.rx_packets++;
2534 		netdev->stats.rx_bytes += length;
2535 		adapter->rx_stats_buffers[scrq_num].packets++;
2536 		adapter->rx_stats_buffers[scrq_num].bytes += length;
2537 		frames_processed++;
2538 	}
2539 
2540 	if (adapter->state != VNIC_CLOSING &&
2541 	    ((atomic_read(&adapter->rx_pool[scrq_num].available) <
2542 	      adapter->req_rx_add_entries_per_subcrq / 2) ||
2543 	      frames_processed < budget))
2544 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2545 	if (frames_processed < budget) {
2546 		if (napi_complete_done(napi, frames_processed)) {
2547 			enable_scrq_irq(adapter, rx_scrq);
2548 			if (pending_scrq(adapter, rx_scrq)) {
2549 				rmb();
2550 				if (napi_reschedule(napi)) {
2551 					disable_scrq_irq(adapter, rx_scrq);
2552 					goto restart_poll;
2553 				}
2554 			}
2555 		}
2556 	}
2557 	return frames_processed;
2558 }
2559 
2560 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2561 {
2562 	int rc, ret;
2563 
2564 	adapter->fallback.mtu = adapter->req_mtu;
2565 	adapter->fallback.rx_queues = adapter->req_rx_queues;
2566 	adapter->fallback.tx_queues = adapter->req_tx_queues;
2567 	adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2568 	adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2569 
2570 	reinit_completion(&adapter->reset_done);
2571 	adapter->wait_for_reset = true;
2572 	rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2573 
2574 	if (rc) {
2575 		ret = rc;
2576 		goto out;
2577 	}
2578 	rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done, 60000);
2579 	if (rc) {
2580 		ret = -ENODEV;
2581 		goto out;
2582 	}
2583 
2584 	ret = 0;
2585 	if (adapter->reset_done_rc) {
2586 		ret = -EIO;
2587 		adapter->desired.mtu = adapter->fallback.mtu;
2588 		adapter->desired.rx_queues = adapter->fallback.rx_queues;
2589 		adapter->desired.tx_queues = adapter->fallback.tx_queues;
2590 		adapter->desired.rx_entries = adapter->fallback.rx_entries;
2591 		adapter->desired.tx_entries = adapter->fallback.tx_entries;
2592 
2593 		reinit_completion(&adapter->reset_done);
2594 		adapter->wait_for_reset = true;
2595 		rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2596 		if (rc) {
2597 			ret = rc;
2598 			goto out;
2599 		}
2600 		rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done,
2601 						 60000);
2602 		if (rc) {
2603 			ret = -ENODEV;
2604 			goto out;
2605 		}
2606 	}
2607 out:
2608 	adapter->wait_for_reset = false;
2609 
2610 	return ret;
2611 }
2612 
2613 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2614 {
2615 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2616 
2617 	adapter->desired.mtu = new_mtu + ETH_HLEN;
2618 
2619 	return wait_for_reset(adapter);
2620 }
2621 
2622 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2623 						struct net_device *dev,
2624 						netdev_features_t features)
2625 {
2626 	/* Some backing hardware adapters can not
2627 	 * handle packets with a MSS less than 224
2628 	 * or with only one segment.
2629 	 */
2630 	if (skb_is_gso(skb)) {
2631 		if (skb_shinfo(skb)->gso_size < 224 ||
2632 		    skb_shinfo(skb)->gso_segs == 1)
2633 			features &= ~NETIF_F_GSO_MASK;
2634 	}
2635 
2636 	return features;
2637 }
2638 
2639 static const struct net_device_ops ibmvnic_netdev_ops = {
2640 	.ndo_open		= ibmvnic_open,
2641 	.ndo_stop		= ibmvnic_close,
2642 	.ndo_start_xmit		= ibmvnic_xmit,
2643 	.ndo_set_rx_mode	= ibmvnic_set_multi,
2644 	.ndo_set_mac_address	= ibmvnic_set_mac,
2645 	.ndo_validate_addr	= eth_validate_addr,
2646 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
2647 	.ndo_change_mtu		= ibmvnic_change_mtu,
2648 	.ndo_features_check     = ibmvnic_features_check,
2649 };
2650 
2651 /* ethtool functions */
2652 
2653 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2654 				      struct ethtool_link_ksettings *cmd)
2655 {
2656 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2657 	int rc;
2658 
2659 	rc = send_query_phys_parms(adapter);
2660 	if (rc) {
2661 		adapter->speed = SPEED_UNKNOWN;
2662 		adapter->duplex = DUPLEX_UNKNOWN;
2663 	}
2664 	cmd->base.speed = adapter->speed;
2665 	cmd->base.duplex = adapter->duplex;
2666 	cmd->base.port = PORT_FIBRE;
2667 	cmd->base.phy_address = 0;
2668 	cmd->base.autoneg = AUTONEG_ENABLE;
2669 
2670 	return 0;
2671 }
2672 
2673 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2674 				struct ethtool_drvinfo *info)
2675 {
2676 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2677 
2678 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2679 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2680 	strlcpy(info->fw_version, adapter->fw_version,
2681 		sizeof(info->fw_version));
2682 }
2683 
2684 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2685 {
2686 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2687 
2688 	return adapter->msg_enable;
2689 }
2690 
2691 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2692 {
2693 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2694 
2695 	adapter->msg_enable = data;
2696 }
2697 
2698 static u32 ibmvnic_get_link(struct net_device *netdev)
2699 {
2700 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2701 
2702 	/* Don't need to send a query because we request a logical link up at
2703 	 * init and then we wait for link state indications
2704 	 */
2705 	return adapter->logical_link_state;
2706 }
2707 
2708 static void ibmvnic_get_ringparam(struct net_device *netdev,
2709 				  struct ethtool_ringparam *ring)
2710 {
2711 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2712 
2713 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2714 		ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2715 		ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2716 	} else {
2717 		ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2718 		ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2719 	}
2720 	ring->rx_mini_max_pending = 0;
2721 	ring->rx_jumbo_max_pending = 0;
2722 	ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2723 	ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2724 	ring->rx_mini_pending = 0;
2725 	ring->rx_jumbo_pending = 0;
2726 }
2727 
2728 static int ibmvnic_set_ringparam(struct net_device *netdev,
2729 				 struct ethtool_ringparam *ring)
2730 {
2731 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2732 	int ret;
2733 
2734 	ret = 0;
2735 	adapter->desired.rx_entries = ring->rx_pending;
2736 	adapter->desired.tx_entries = ring->tx_pending;
2737 
2738 	ret = wait_for_reset(adapter);
2739 
2740 	if (!ret &&
2741 	    (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2742 	     adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2743 		netdev_info(netdev,
2744 			    "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2745 			    ring->rx_pending, ring->tx_pending,
2746 			    adapter->req_rx_add_entries_per_subcrq,
2747 			    adapter->req_tx_entries_per_subcrq);
2748 	return ret;
2749 }
2750 
2751 static void ibmvnic_get_channels(struct net_device *netdev,
2752 				 struct ethtool_channels *channels)
2753 {
2754 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2755 
2756 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2757 		channels->max_rx = adapter->max_rx_queues;
2758 		channels->max_tx = adapter->max_tx_queues;
2759 	} else {
2760 		channels->max_rx = IBMVNIC_MAX_QUEUES;
2761 		channels->max_tx = IBMVNIC_MAX_QUEUES;
2762 	}
2763 
2764 	channels->max_other = 0;
2765 	channels->max_combined = 0;
2766 	channels->rx_count = adapter->req_rx_queues;
2767 	channels->tx_count = adapter->req_tx_queues;
2768 	channels->other_count = 0;
2769 	channels->combined_count = 0;
2770 }
2771 
2772 static int ibmvnic_set_channels(struct net_device *netdev,
2773 				struct ethtool_channels *channels)
2774 {
2775 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2776 	int ret;
2777 
2778 	ret = 0;
2779 	adapter->desired.rx_queues = channels->rx_count;
2780 	adapter->desired.tx_queues = channels->tx_count;
2781 
2782 	ret = wait_for_reset(adapter);
2783 
2784 	if (!ret &&
2785 	    (adapter->req_rx_queues != channels->rx_count ||
2786 	     adapter->req_tx_queues != channels->tx_count))
2787 		netdev_info(netdev,
2788 			    "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2789 			    channels->rx_count, channels->tx_count,
2790 			    adapter->req_rx_queues, adapter->req_tx_queues);
2791 	return ret;
2792 
2793 }
2794 
2795 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2796 {
2797 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2798 	int i;
2799 
2800 	switch (stringset) {
2801 	case ETH_SS_STATS:
2802 		for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2803 				i++, data += ETH_GSTRING_LEN)
2804 			memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2805 
2806 		for (i = 0; i < adapter->req_tx_queues; i++) {
2807 			snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2808 			data += ETH_GSTRING_LEN;
2809 
2810 			snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2811 			data += ETH_GSTRING_LEN;
2812 
2813 			snprintf(data, ETH_GSTRING_LEN,
2814 				 "tx%d_dropped_packets", i);
2815 			data += ETH_GSTRING_LEN;
2816 		}
2817 
2818 		for (i = 0; i < adapter->req_rx_queues; i++) {
2819 			snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2820 			data += ETH_GSTRING_LEN;
2821 
2822 			snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2823 			data += ETH_GSTRING_LEN;
2824 
2825 			snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2826 			data += ETH_GSTRING_LEN;
2827 		}
2828 		break;
2829 
2830 	case ETH_SS_PRIV_FLAGS:
2831 		for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2832 			strcpy(data + i * ETH_GSTRING_LEN,
2833 			       ibmvnic_priv_flags[i]);
2834 		break;
2835 	default:
2836 		return;
2837 	}
2838 }
2839 
2840 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2841 {
2842 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2843 
2844 	switch (sset) {
2845 	case ETH_SS_STATS:
2846 		return ARRAY_SIZE(ibmvnic_stats) +
2847 		       adapter->req_tx_queues * NUM_TX_STATS +
2848 		       adapter->req_rx_queues * NUM_RX_STATS;
2849 	case ETH_SS_PRIV_FLAGS:
2850 		return ARRAY_SIZE(ibmvnic_priv_flags);
2851 	default:
2852 		return -EOPNOTSUPP;
2853 	}
2854 }
2855 
2856 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2857 				      struct ethtool_stats *stats, u64 *data)
2858 {
2859 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2860 	union ibmvnic_crq crq;
2861 	int i, j;
2862 	int rc;
2863 
2864 	memset(&crq, 0, sizeof(crq));
2865 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2866 	crq.request_statistics.cmd = REQUEST_STATISTICS;
2867 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2868 	crq.request_statistics.len =
2869 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
2870 
2871 	/* Wait for data to be written */
2872 	reinit_completion(&adapter->stats_done);
2873 	rc = ibmvnic_send_crq(adapter, &crq);
2874 	if (rc)
2875 		return;
2876 	rc = ibmvnic_wait_for_completion(adapter, &adapter->stats_done, 10000);
2877 	if (rc)
2878 		return;
2879 
2880 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2881 		data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2882 						ibmvnic_stats[i].offset));
2883 
2884 	for (j = 0; j < adapter->req_tx_queues; j++) {
2885 		data[i] = adapter->tx_stats_buffers[j].packets;
2886 		i++;
2887 		data[i] = adapter->tx_stats_buffers[j].bytes;
2888 		i++;
2889 		data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2890 		i++;
2891 	}
2892 
2893 	for (j = 0; j < adapter->req_rx_queues; j++) {
2894 		data[i] = adapter->rx_stats_buffers[j].packets;
2895 		i++;
2896 		data[i] = adapter->rx_stats_buffers[j].bytes;
2897 		i++;
2898 		data[i] = adapter->rx_stats_buffers[j].interrupts;
2899 		i++;
2900 	}
2901 }
2902 
2903 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2904 {
2905 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2906 
2907 	return adapter->priv_flags;
2908 }
2909 
2910 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2911 {
2912 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2913 	bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2914 
2915 	if (which_maxes)
2916 		adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2917 	else
2918 		adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2919 
2920 	return 0;
2921 }
2922 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2923 	.get_drvinfo		= ibmvnic_get_drvinfo,
2924 	.get_msglevel		= ibmvnic_get_msglevel,
2925 	.set_msglevel		= ibmvnic_set_msglevel,
2926 	.get_link		= ibmvnic_get_link,
2927 	.get_ringparam		= ibmvnic_get_ringparam,
2928 	.set_ringparam		= ibmvnic_set_ringparam,
2929 	.get_channels		= ibmvnic_get_channels,
2930 	.set_channels		= ibmvnic_set_channels,
2931 	.get_strings            = ibmvnic_get_strings,
2932 	.get_sset_count         = ibmvnic_get_sset_count,
2933 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
2934 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
2935 	.get_priv_flags		= ibmvnic_get_priv_flags,
2936 	.set_priv_flags		= ibmvnic_set_priv_flags,
2937 };
2938 
2939 /* Routines for managing CRQs/sCRQs  */
2940 
2941 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2942 				   struct ibmvnic_sub_crq_queue *scrq)
2943 {
2944 	int rc;
2945 
2946 	if (scrq->irq) {
2947 		free_irq(scrq->irq, scrq);
2948 		irq_dispose_mapping(scrq->irq);
2949 		scrq->irq = 0;
2950 	}
2951 
2952 	memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2953 	atomic_set(&scrq->used, 0);
2954 	scrq->cur = 0;
2955 	scrq->ind_buf.index = 0;
2956 
2957 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2958 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2959 	return rc;
2960 }
2961 
2962 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2963 {
2964 	int i, rc;
2965 
2966 	for (i = 0; i < adapter->req_tx_queues; i++) {
2967 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2968 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2969 		if (rc)
2970 			return rc;
2971 	}
2972 
2973 	for (i = 0; i < adapter->req_rx_queues; i++) {
2974 		netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2975 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2976 		if (rc)
2977 			return rc;
2978 	}
2979 
2980 	return rc;
2981 }
2982 
2983 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2984 				  struct ibmvnic_sub_crq_queue *scrq,
2985 				  bool do_h_free)
2986 {
2987 	struct device *dev = &adapter->vdev->dev;
2988 	long rc;
2989 
2990 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2991 
2992 	if (do_h_free) {
2993 		/* Close the sub-crqs */
2994 		do {
2995 			rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2996 						adapter->vdev->unit_address,
2997 						scrq->crq_num);
2998 		} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2999 
3000 		if (rc) {
3001 			netdev_err(adapter->netdev,
3002 				   "Failed to release sub-CRQ %16lx, rc = %ld\n",
3003 				   scrq->crq_num, rc);
3004 		}
3005 	}
3006 
3007 	dma_free_coherent(dev,
3008 			  IBMVNIC_IND_ARR_SZ,
3009 			  scrq->ind_buf.indir_arr,
3010 			  scrq->ind_buf.indir_dma);
3011 
3012 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
3013 			 DMA_BIDIRECTIONAL);
3014 	free_pages((unsigned long)scrq->msgs, 2);
3015 	kfree(scrq);
3016 }
3017 
3018 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
3019 							*adapter)
3020 {
3021 	struct device *dev = &adapter->vdev->dev;
3022 	struct ibmvnic_sub_crq_queue *scrq;
3023 	int rc;
3024 
3025 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
3026 	if (!scrq)
3027 		return NULL;
3028 
3029 	scrq->msgs =
3030 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
3031 	if (!scrq->msgs) {
3032 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
3033 		goto zero_page_failed;
3034 	}
3035 
3036 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
3037 					 DMA_BIDIRECTIONAL);
3038 	if (dma_mapping_error(dev, scrq->msg_token)) {
3039 		dev_warn(dev, "Couldn't map crq queue messages page\n");
3040 		goto map_failed;
3041 	}
3042 
3043 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
3044 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
3045 
3046 	if (rc == H_RESOURCE)
3047 		rc = ibmvnic_reset_crq(adapter);
3048 
3049 	if (rc == H_CLOSED) {
3050 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
3051 	} else if (rc) {
3052 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
3053 		goto reg_failed;
3054 	}
3055 
3056 	scrq->adapter = adapter;
3057 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
3058 	scrq->ind_buf.index = 0;
3059 
3060 	scrq->ind_buf.indir_arr =
3061 		dma_alloc_coherent(dev,
3062 				   IBMVNIC_IND_ARR_SZ,
3063 				   &scrq->ind_buf.indir_dma,
3064 				   GFP_KERNEL);
3065 
3066 	if (!scrq->ind_buf.indir_arr)
3067 		goto indir_failed;
3068 
3069 	spin_lock_init(&scrq->lock);
3070 
3071 	netdev_dbg(adapter->netdev,
3072 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
3073 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
3074 
3075 	return scrq;
3076 
3077 indir_failed:
3078 	do {
3079 		rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
3080 					adapter->vdev->unit_address,
3081 					scrq->crq_num);
3082 	} while (rc == H_BUSY || rc == H_IS_LONG_BUSY(rc));
3083 reg_failed:
3084 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
3085 			 DMA_BIDIRECTIONAL);
3086 map_failed:
3087 	free_pages((unsigned long)scrq->msgs, 2);
3088 zero_page_failed:
3089 	kfree(scrq);
3090 
3091 	return NULL;
3092 }
3093 
3094 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
3095 {
3096 	int i;
3097 
3098 	if (adapter->tx_scrq) {
3099 		for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
3100 			if (!adapter->tx_scrq[i])
3101 				continue;
3102 
3103 			netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
3104 				   i);
3105 			if (adapter->tx_scrq[i]->irq) {
3106 				free_irq(adapter->tx_scrq[i]->irq,
3107 					 adapter->tx_scrq[i]);
3108 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
3109 				adapter->tx_scrq[i]->irq = 0;
3110 			}
3111 
3112 			release_sub_crq_queue(adapter, adapter->tx_scrq[i],
3113 					      do_h_free);
3114 		}
3115 
3116 		kfree(adapter->tx_scrq);
3117 		adapter->tx_scrq = NULL;
3118 		adapter->num_active_tx_scrqs = 0;
3119 	}
3120 
3121 	if (adapter->rx_scrq) {
3122 		for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
3123 			if (!adapter->rx_scrq[i])
3124 				continue;
3125 
3126 			netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
3127 				   i);
3128 			if (adapter->rx_scrq[i]->irq) {
3129 				free_irq(adapter->rx_scrq[i]->irq,
3130 					 adapter->rx_scrq[i]);
3131 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
3132 				adapter->rx_scrq[i]->irq = 0;
3133 			}
3134 
3135 			release_sub_crq_queue(adapter, adapter->rx_scrq[i],
3136 					      do_h_free);
3137 		}
3138 
3139 		kfree(adapter->rx_scrq);
3140 		adapter->rx_scrq = NULL;
3141 		adapter->num_active_rx_scrqs = 0;
3142 	}
3143 }
3144 
3145 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
3146 			    struct ibmvnic_sub_crq_queue *scrq)
3147 {
3148 	struct device *dev = &adapter->vdev->dev;
3149 	unsigned long rc;
3150 
3151 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3152 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3153 	if (rc)
3154 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
3155 			scrq->hw_irq, rc);
3156 	return rc;
3157 }
3158 
3159 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
3160 			   struct ibmvnic_sub_crq_queue *scrq)
3161 {
3162 	struct device *dev = &adapter->vdev->dev;
3163 	unsigned long rc;
3164 
3165 	if (scrq->hw_irq > 0x100000000ULL) {
3166 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
3167 		return 1;
3168 	}
3169 
3170 	if (test_bit(0, &adapter->resetting) &&
3171 	    adapter->reset_reason == VNIC_RESET_MOBILITY) {
3172 		u64 val = (0xff000000) | scrq->hw_irq;
3173 
3174 		rc = plpar_hcall_norets(H_EOI, val);
3175 		/* H_EOI would fail with rc = H_FUNCTION when running
3176 		 * in XIVE mode which is expected, but not an error.
3177 		 */
3178 		if (rc && (rc != H_FUNCTION))
3179 			dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
3180 				val, rc);
3181 	}
3182 
3183 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3184 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3185 	if (rc)
3186 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
3187 			scrq->hw_irq, rc);
3188 	return rc;
3189 }
3190 
3191 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
3192 			       struct ibmvnic_sub_crq_queue *scrq)
3193 {
3194 	struct device *dev = &adapter->vdev->dev;
3195 	struct ibmvnic_tx_pool *tx_pool;
3196 	struct ibmvnic_tx_buff *txbuff;
3197 	struct netdev_queue *txq;
3198 	union sub_crq *next;
3199 	int index;
3200 	int i;
3201 
3202 restart_loop:
3203 	while (pending_scrq(adapter, scrq)) {
3204 		unsigned int pool = scrq->pool_index;
3205 		int num_entries = 0;
3206 		int total_bytes = 0;
3207 		int num_packets = 0;
3208 
3209 		next = ibmvnic_next_scrq(adapter, scrq);
3210 		for (i = 0; i < next->tx_comp.num_comps; i++) {
3211 			if (next->tx_comp.rcs[i]) {
3212 				dev_err(dev, "tx error %x\n",
3213 					next->tx_comp.rcs[i]);
3214 				continue;
3215 			}
3216 			index = be32_to_cpu(next->tx_comp.correlators[i]);
3217 			if (index & IBMVNIC_TSO_POOL_MASK) {
3218 				tx_pool = &adapter->tso_pool[pool];
3219 				index &= ~IBMVNIC_TSO_POOL_MASK;
3220 			} else {
3221 				tx_pool = &adapter->tx_pool[pool];
3222 			}
3223 
3224 			txbuff = &tx_pool->tx_buff[index];
3225 			num_packets++;
3226 			num_entries += txbuff->num_entries;
3227 			if (txbuff->skb) {
3228 				total_bytes += txbuff->skb->len;
3229 				dev_consume_skb_irq(txbuff->skb);
3230 				txbuff->skb = NULL;
3231 			} else {
3232 				netdev_warn(adapter->netdev,
3233 					    "TX completion received with NULL socket buffer\n");
3234 			}
3235 			tx_pool->free_map[tx_pool->producer_index] = index;
3236 			tx_pool->producer_index =
3237 				(tx_pool->producer_index + 1) %
3238 					tx_pool->num_buffers;
3239 		}
3240 		/* remove tx_comp scrq*/
3241 		next->tx_comp.first = 0;
3242 
3243 		txq = netdev_get_tx_queue(adapter->netdev, scrq->pool_index);
3244 		netdev_tx_completed_queue(txq, num_packets, total_bytes);
3245 
3246 		if (atomic_sub_return(num_entries, &scrq->used) <=
3247 		    (adapter->req_tx_entries_per_subcrq / 2) &&
3248 		    __netif_subqueue_stopped(adapter->netdev,
3249 					     scrq->pool_index)) {
3250 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
3251 			netdev_dbg(adapter->netdev, "Started queue %d\n",
3252 				   scrq->pool_index);
3253 		}
3254 	}
3255 
3256 	enable_scrq_irq(adapter, scrq);
3257 
3258 	if (pending_scrq(adapter, scrq)) {
3259 		disable_scrq_irq(adapter, scrq);
3260 		goto restart_loop;
3261 	}
3262 
3263 	return 0;
3264 }
3265 
3266 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
3267 {
3268 	struct ibmvnic_sub_crq_queue *scrq = instance;
3269 	struct ibmvnic_adapter *adapter = scrq->adapter;
3270 
3271 	disable_scrq_irq(adapter, scrq);
3272 	ibmvnic_complete_tx(adapter, scrq);
3273 
3274 	return IRQ_HANDLED;
3275 }
3276 
3277 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
3278 {
3279 	struct ibmvnic_sub_crq_queue *scrq = instance;
3280 	struct ibmvnic_adapter *adapter = scrq->adapter;
3281 
3282 	/* When booting a kdump kernel we can hit pending interrupts
3283 	 * prior to completing driver initialization.
3284 	 */
3285 	if (unlikely(adapter->state != VNIC_OPEN))
3286 		return IRQ_NONE;
3287 
3288 	adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
3289 
3290 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
3291 		disable_scrq_irq(adapter, scrq);
3292 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
3293 	}
3294 
3295 	return IRQ_HANDLED;
3296 }
3297 
3298 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
3299 {
3300 	struct device *dev = &adapter->vdev->dev;
3301 	struct ibmvnic_sub_crq_queue *scrq;
3302 	int i = 0, j = 0;
3303 	int rc = 0;
3304 
3305 	for (i = 0; i < adapter->req_tx_queues; i++) {
3306 		netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
3307 			   i);
3308 		scrq = adapter->tx_scrq[i];
3309 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3310 
3311 		if (!scrq->irq) {
3312 			rc = -EINVAL;
3313 			dev_err(dev, "Error mapping irq\n");
3314 			goto req_tx_irq_failed;
3315 		}
3316 
3317 		snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
3318 			 adapter->vdev->unit_address, i);
3319 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
3320 				 0, scrq->name, scrq);
3321 
3322 		if (rc) {
3323 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
3324 				scrq->irq, rc);
3325 			irq_dispose_mapping(scrq->irq);
3326 			goto req_tx_irq_failed;
3327 		}
3328 	}
3329 
3330 	for (i = 0; i < adapter->req_rx_queues; i++) {
3331 		netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
3332 			   i);
3333 		scrq = adapter->rx_scrq[i];
3334 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3335 		if (!scrq->irq) {
3336 			rc = -EINVAL;
3337 			dev_err(dev, "Error mapping irq\n");
3338 			goto req_rx_irq_failed;
3339 		}
3340 		snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
3341 			 adapter->vdev->unit_address, i);
3342 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
3343 				 0, scrq->name, scrq);
3344 		if (rc) {
3345 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
3346 				scrq->irq, rc);
3347 			irq_dispose_mapping(scrq->irq);
3348 			goto req_rx_irq_failed;
3349 		}
3350 	}
3351 	return rc;
3352 
3353 req_rx_irq_failed:
3354 	for (j = 0; j < i; j++) {
3355 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
3356 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
3357 	}
3358 	i = adapter->req_tx_queues;
3359 req_tx_irq_failed:
3360 	for (j = 0; j < i; j++) {
3361 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
3362 		irq_dispose_mapping(adapter->tx_scrq[j]->irq);
3363 	}
3364 	release_sub_crqs(adapter, 1);
3365 	return rc;
3366 }
3367 
3368 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
3369 {
3370 	struct device *dev = &adapter->vdev->dev;
3371 	struct ibmvnic_sub_crq_queue **allqueues;
3372 	int registered_queues = 0;
3373 	int total_queues;
3374 	int more = 0;
3375 	int i;
3376 
3377 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
3378 
3379 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
3380 	if (!allqueues)
3381 		return -1;
3382 
3383 	for (i = 0; i < total_queues; i++) {
3384 		allqueues[i] = init_sub_crq_queue(adapter);
3385 		if (!allqueues[i]) {
3386 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
3387 			break;
3388 		}
3389 		registered_queues++;
3390 	}
3391 
3392 	/* Make sure we were able to register the minimum number of queues */
3393 	if (registered_queues <
3394 	    adapter->min_tx_queues + adapter->min_rx_queues) {
3395 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
3396 		goto tx_failed;
3397 	}
3398 
3399 	/* Distribute the failed allocated queues*/
3400 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
3401 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3402 		switch (i % 3) {
3403 		case 0:
3404 			if (adapter->req_rx_queues > adapter->min_rx_queues)
3405 				adapter->req_rx_queues--;
3406 			else
3407 				more++;
3408 			break;
3409 		case 1:
3410 			if (adapter->req_tx_queues > adapter->min_tx_queues)
3411 				adapter->req_tx_queues--;
3412 			else
3413 				more++;
3414 			break;
3415 		}
3416 	}
3417 
3418 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3419 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
3420 	if (!adapter->tx_scrq)
3421 		goto tx_failed;
3422 
3423 	for (i = 0; i < adapter->req_tx_queues; i++) {
3424 		adapter->tx_scrq[i] = allqueues[i];
3425 		adapter->tx_scrq[i]->pool_index = i;
3426 		adapter->num_active_tx_scrqs++;
3427 	}
3428 
3429 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3430 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
3431 	if (!adapter->rx_scrq)
3432 		goto rx_failed;
3433 
3434 	for (i = 0; i < adapter->req_rx_queues; i++) {
3435 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3436 		adapter->rx_scrq[i]->scrq_num = i;
3437 		adapter->num_active_rx_scrqs++;
3438 	}
3439 
3440 	kfree(allqueues);
3441 	return 0;
3442 
3443 rx_failed:
3444 	kfree(adapter->tx_scrq);
3445 	adapter->tx_scrq = NULL;
3446 tx_failed:
3447 	for (i = 0; i < registered_queues; i++)
3448 		release_sub_crq_queue(adapter, allqueues[i], 1);
3449 	kfree(allqueues);
3450 	return -1;
3451 }
3452 
3453 static void send_request_cap(struct ibmvnic_adapter *adapter, int retry)
3454 {
3455 	struct device *dev = &adapter->vdev->dev;
3456 	union ibmvnic_crq crq;
3457 	int max_entries;
3458 
3459 	if (!retry) {
3460 		/* Sub-CRQ entries are 32 byte long */
3461 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3462 
3463 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
3464 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
3465 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3466 			return;
3467 		}
3468 
3469 		if (adapter->desired.mtu)
3470 			adapter->req_mtu = adapter->desired.mtu;
3471 		else
3472 			adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3473 
3474 		if (!adapter->desired.tx_entries)
3475 			adapter->desired.tx_entries =
3476 					adapter->max_tx_entries_per_subcrq;
3477 		if (!adapter->desired.rx_entries)
3478 			adapter->desired.rx_entries =
3479 					adapter->max_rx_add_entries_per_subcrq;
3480 
3481 		max_entries = IBMVNIC_MAX_LTB_SIZE /
3482 			      (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3483 
3484 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3485 			adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3486 			adapter->desired.tx_entries = max_entries;
3487 		}
3488 
3489 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3490 			adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3491 			adapter->desired.rx_entries = max_entries;
3492 		}
3493 
3494 		if (adapter->desired.tx_entries)
3495 			adapter->req_tx_entries_per_subcrq =
3496 					adapter->desired.tx_entries;
3497 		else
3498 			adapter->req_tx_entries_per_subcrq =
3499 					adapter->max_tx_entries_per_subcrq;
3500 
3501 		if (adapter->desired.rx_entries)
3502 			adapter->req_rx_add_entries_per_subcrq =
3503 					adapter->desired.rx_entries;
3504 		else
3505 			adapter->req_rx_add_entries_per_subcrq =
3506 					adapter->max_rx_add_entries_per_subcrq;
3507 
3508 		if (adapter->desired.tx_queues)
3509 			adapter->req_tx_queues =
3510 					adapter->desired.tx_queues;
3511 		else
3512 			adapter->req_tx_queues =
3513 					adapter->opt_tx_comp_sub_queues;
3514 
3515 		if (adapter->desired.rx_queues)
3516 			adapter->req_rx_queues =
3517 					adapter->desired.rx_queues;
3518 		else
3519 			adapter->req_rx_queues =
3520 					adapter->opt_rx_comp_queues;
3521 
3522 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3523 	}
3524 
3525 	memset(&crq, 0, sizeof(crq));
3526 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
3527 	crq.request_capability.cmd = REQUEST_CAPABILITY;
3528 
3529 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3530 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3531 	atomic_inc(&adapter->running_cap_crqs);
3532 	ibmvnic_send_crq(adapter, &crq);
3533 
3534 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3535 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3536 	atomic_inc(&adapter->running_cap_crqs);
3537 	ibmvnic_send_crq(adapter, &crq);
3538 
3539 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3540 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3541 	atomic_inc(&adapter->running_cap_crqs);
3542 	ibmvnic_send_crq(adapter, &crq);
3543 
3544 	crq.request_capability.capability =
3545 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3546 	crq.request_capability.number =
3547 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3548 	atomic_inc(&adapter->running_cap_crqs);
3549 	ibmvnic_send_crq(adapter, &crq);
3550 
3551 	crq.request_capability.capability =
3552 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3553 	crq.request_capability.number =
3554 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3555 	atomic_inc(&adapter->running_cap_crqs);
3556 	ibmvnic_send_crq(adapter, &crq);
3557 
3558 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3559 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3560 	atomic_inc(&adapter->running_cap_crqs);
3561 	ibmvnic_send_crq(adapter, &crq);
3562 
3563 	if (adapter->netdev->flags & IFF_PROMISC) {
3564 		if (adapter->promisc_supported) {
3565 			crq.request_capability.capability =
3566 			    cpu_to_be16(PROMISC_REQUESTED);
3567 			crq.request_capability.number = cpu_to_be64(1);
3568 			atomic_inc(&adapter->running_cap_crqs);
3569 			ibmvnic_send_crq(adapter, &crq);
3570 		}
3571 	} else {
3572 		crq.request_capability.capability =
3573 		    cpu_to_be16(PROMISC_REQUESTED);
3574 		crq.request_capability.number = cpu_to_be64(0);
3575 		atomic_inc(&adapter->running_cap_crqs);
3576 		ibmvnic_send_crq(adapter, &crq);
3577 	}
3578 }
3579 
3580 static int pending_scrq(struct ibmvnic_adapter *adapter,
3581 			struct ibmvnic_sub_crq_queue *scrq)
3582 {
3583 	union sub_crq *entry = &scrq->msgs[scrq->cur];
3584 
3585 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3586 		return 1;
3587 	else
3588 		return 0;
3589 }
3590 
3591 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3592 					struct ibmvnic_sub_crq_queue *scrq)
3593 {
3594 	union sub_crq *entry;
3595 	unsigned long flags;
3596 
3597 	spin_lock_irqsave(&scrq->lock, flags);
3598 	entry = &scrq->msgs[scrq->cur];
3599 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3600 		if (++scrq->cur == scrq->size)
3601 			scrq->cur = 0;
3602 	} else {
3603 		entry = NULL;
3604 	}
3605 	spin_unlock_irqrestore(&scrq->lock, flags);
3606 
3607 	return entry;
3608 }
3609 
3610 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3611 {
3612 	struct ibmvnic_crq_queue *queue = &adapter->crq;
3613 	union ibmvnic_crq *crq;
3614 
3615 	crq = &queue->msgs[queue->cur];
3616 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3617 		if (++queue->cur == queue->size)
3618 			queue->cur = 0;
3619 	} else {
3620 		crq = NULL;
3621 	}
3622 
3623 	return crq;
3624 }
3625 
3626 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3627 {
3628 	switch (rc) {
3629 	case H_PARAMETER:
3630 		dev_warn_ratelimited(dev,
3631 				     "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3632 				     func, rc);
3633 		break;
3634 	case H_CLOSED:
3635 		dev_warn_ratelimited(dev,
3636 				     "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3637 				     func, rc);
3638 		break;
3639 	default:
3640 		dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3641 		break;
3642 	}
3643 }
3644 
3645 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3646 				u64 remote_handle, u64 ioba, u64 num_entries)
3647 {
3648 	unsigned int ua = adapter->vdev->unit_address;
3649 	struct device *dev = &adapter->vdev->dev;
3650 	int rc;
3651 
3652 	/* Make sure the hypervisor sees the complete request */
3653 	mb();
3654 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3655 				cpu_to_be64(remote_handle),
3656 				ioba, num_entries);
3657 
3658 	if (rc)
3659 		print_subcrq_error(dev, rc, __func__);
3660 
3661 	return rc;
3662 }
3663 
3664 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3665 			    union ibmvnic_crq *crq)
3666 {
3667 	unsigned int ua = adapter->vdev->unit_address;
3668 	struct device *dev = &adapter->vdev->dev;
3669 	u64 *u64_crq = (u64 *)crq;
3670 	int rc;
3671 
3672 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3673 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3674 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
3675 
3676 	if (!adapter->crq.active &&
3677 	    crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3678 		dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3679 		return -EINVAL;
3680 	}
3681 
3682 	/* Make sure the hypervisor sees the complete request */
3683 	mb();
3684 
3685 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3686 				cpu_to_be64(u64_crq[0]),
3687 				cpu_to_be64(u64_crq[1]));
3688 
3689 	if (rc) {
3690 		if (rc == H_CLOSED) {
3691 			dev_warn(dev, "CRQ Queue closed\n");
3692 			/* do not reset, report the fail, wait for passive init from server */
3693 		}
3694 
3695 		dev_warn(dev, "Send error (rc=%d)\n", rc);
3696 	}
3697 
3698 	return rc;
3699 }
3700 
3701 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3702 {
3703 	struct device *dev = &adapter->vdev->dev;
3704 	union ibmvnic_crq crq;
3705 	int retries = 100;
3706 	int rc;
3707 
3708 	memset(&crq, 0, sizeof(crq));
3709 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3710 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
3711 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3712 
3713 	do {
3714 		rc = ibmvnic_send_crq(adapter, &crq);
3715 		if (rc != H_CLOSED)
3716 			break;
3717 		retries--;
3718 		msleep(50);
3719 
3720 	} while (retries > 0);
3721 
3722 	if (rc) {
3723 		dev_err(dev, "Failed to send init request, rc = %d\n", rc);
3724 		return rc;
3725 	}
3726 
3727 	return 0;
3728 }
3729 
3730 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3731 {
3732 	union ibmvnic_crq crq;
3733 
3734 	memset(&crq, 0, sizeof(crq));
3735 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3736 	crq.version_exchange.cmd = VERSION_EXCHANGE;
3737 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3738 
3739 	return ibmvnic_send_crq(adapter, &crq);
3740 }
3741 
3742 struct vnic_login_client_data {
3743 	u8	type;
3744 	__be16	len;
3745 	char	name[];
3746 } __packed;
3747 
3748 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3749 {
3750 	int len;
3751 
3752 	/* Calculate the amount of buffer space needed for the
3753 	 * vnic client data in the login buffer. There are four entries,
3754 	 * OS name, LPAR name, device name, and a null last entry.
3755 	 */
3756 	len = 4 * sizeof(struct vnic_login_client_data);
3757 	len += 6; /* "Linux" plus NULL */
3758 	len += strlen(utsname()->nodename) + 1;
3759 	len += strlen(adapter->netdev->name) + 1;
3760 
3761 	return len;
3762 }
3763 
3764 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3765 				 struct vnic_login_client_data *vlcd)
3766 {
3767 	const char *os_name = "Linux";
3768 	int len;
3769 
3770 	/* Type 1 - LPAR OS */
3771 	vlcd->type = 1;
3772 	len = strlen(os_name) + 1;
3773 	vlcd->len = cpu_to_be16(len);
3774 	strncpy(vlcd->name, os_name, len);
3775 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3776 
3777 	/* Type 2 - LPAR name */
3778 	vlcd->type = 2;
3779 	len = strlen(utsname()->nodename) + 1;
3780 	vlcd->len = cpu_to_be16(len);
3781 	strncpy(vlcd->name, utsname()->nodename, len);
3782 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3783 
3784 	/* Type 3 - device name */
3785 	vlcd->type = 3;
3786 	len = strlen(adapter->netdev->name) + 1;
3787 	vlcd->len = cpu_to_be16(len);
3788 	strncpy(vlcd->name, adapter->netdev->name, len);
3789 }
3790 
3791 static int send_login(struct ibmvnic_adapter *adapter)
3792 {
3793 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3794 	struct ibmvnic_login_buffer *login_buffer;
3795 	struct device *dev = &adapter->vdev->dev;
3796 	dma_addr_t rsp_buffer_token;
3797 	dma_addr_t buffer_token;
3798 	size_t rsp_buffer_size;
3799 	union ibmvnic_crq crq;
3800 	size_t buffer_size;
3801 	__be64 *tx_list_p;
3802 	__be64 *rx_list_p;
3803 	int client_data_len;
3804 	struct vnic_login_client_data *vlcd;
3805 	int i;
3806 
3807 	if (!adapter->tx_scrq || !adapter->rx_scrq) {
3808 		netdev_err(adapter->netdev,
3809 			   "RX or TX queues are not allocated, device login failed\n");
3810 		return -1;
3811 	}
3812 
3813 	release_login_rsp_buffer(adapter);
3814 	client_data_len = vnic_client_data_len(adapter);
3815 
3816 	buffer_size =
3817 	    sizeof(struct ibmvnic_login_buffer) +
3818 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3819 	    client_data_len;
3820 
3821 	login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3822 	if (!login_buffer)
3823 		goto buf_alloc_failed;
3824 
3825 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3826 				      DMA_TO_DEVICE);
3827 	if (dma_mapping_error(dev, buffer_token)) {
3828 		dev_err(dev, "Couldn't map login buffer\n");
3829 		goto buf_map_failed;
3830 	}
3831 
3832 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3833 			  sizeof(u64) * adapter->req_tx_queues +
3834 			  sizeof(u64) * adapter->req_rx_queues +
3835 			  sizeof(u64) * adapter->req_rx_queues +
3836 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3837 
3838 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3839 	if (!login_rsp_buffer)
3840 		goto buf_rsp_alloc_failed;
3841 
3842 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3843 					  rsp_buffer_size, DMA_FROM_DEVICE);
3844 	if (dma_mapping_error(dev, rsp_buffer_token)) {
3845 		dev_err(dev, "Couldn't map login rsp buffer\n");
3846 		goto buf_rsp_map_failed;
3847 	}
3848 
3849 	adapter->login_buf = login_buffer;
3850 	adapter->login_buf_token = buffer_token;
3851 	adapter->login_buf_sz = buffer_size;
3852 	adapter->login_rsp_buf = login_rsp_buffer;
3853 	adapter->login_rsp_buf_token = rsp_buffer_token;
3854 	adapter->login_rsp_buf_sz = rsp_buffer_size;
3855 
3856 	login_buffer->len = cpu_to_be32(buffer_size);
3857 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3858 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3859 	login_buffer->off_txcomp_subcrqs =
3860 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3861 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3862 	login_buffer->off_rxcomp_subcrqs =
3863 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3864 			sizeof(u64) * adapter->req_tx_queues);
3865 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3866 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3867 
3868 	tx_list_p = (__be64 *)((char *)login_buffer +
3869 				      sizeof(struct ibmvnic_login_buffer));
3870 	rx_list_p = (__be64 *)((char *)login_buffer +
3871 				      sizeof(struct ibmvnic_login_buffer) +
3872 				      sizeof(u64) * adapter->req_tx_queues);
3873 
3874 	for (i = 0; i < adapter->req_tx_queues; i++) {
3875 		if (adapter->tx_scrq[i]) {
3876 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3877 						   crq_num);
3878 		}
3879 	}
3880 
3881 	for (i = 0; i < adapter->req_rx_queues; i++) {
3882 		if (adapter->rx_scrq[i]) {
3883 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3884 						   crq_num);
3885 		}
3886 	}
3887 
3888 	/* Insert vNIC login client data */
3889 	vlcd = (struct vnic_login_client_data *)
3890 		((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3891 	login_buffer->client_data_offset =
3892 			cpu_to_be32((char *)vlcd - (char *)login_buffer);
3893 	login_buffer->client_data_len = cpu_to_be32(client_data_len);
3894 
3895 	vnic_add_client_data(adapter, vlcd);
3896 
3897 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
3898 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3899 		netdev_dbg(adapter->netdev, "%016lx\n",
3900 			   ((unsigned long int *)(adapter->login_buf))[i]);
3901 	}
3902 
3903 	memset(&crq, 0, sizeof(crq));
3904 	crq.login.first = IBMVNIC_CRQ_CMD;
3905 	crq.login.cmd = LOGIN;
3906 	crq.login.ioba = cpu_to_be32(buffer_token);
3907 	crq.login.len = cpu_to_be32(buffer_size);
3908 	ibmvnic_send_crq(adapter, &crq);
3909 
3910 	return 0;
3911 
3912 buf_rsp_map_failed:
3913 	kfree(login_rsp_buffer);
3914 buf_rsp_alloc_failed:
3915 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3916 buf_map_failed:
3917 	kfree(login_buffer);
3918 buf_alloc_failed:
3919 	return -1;
3920 }
3921 
3922 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3923 			    u32 len, u8 map_id)
3924 {
3925 	union ibmvnic_crq crq;
3926 
3927 	memset(&crq, 0, sizeof(crq));
3928 	crq.request_map.first = IBMVNIC_CRQ_CMD;
3929 	crq.request_map.cmd = REQUEST_MAP;
3930 	crq.request_map.map_id = map_id;
3931 	crq.request_map.ioba = cpu_to_be32(addr);
3932 	crq.request_map.len = cpu_to_be32(len);
3933 	return ibmvnic_send_crq(adapter, &crq);
3934 }
3935 
3936 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3937 {
3938 	union ibmvnic_crq crq;
3939 
3940 	memset(&crq, 0, sizeof(crq));
3941 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3942 	crq.request_unmap.cmd = REQUEST_UNMAP;
3943 	crq.request_unmap.map_id = map_id;
3944 	return ibmvnic_send_crq(adapter, &crq);
3945 }
3946 
3947 static void send_query_map(struct ibmvnic_adapter *adapter)
3948 {
3949 	union ibmvnic_crq crq;
3950 
3951 	memset(&crq, 0, sizeof(crq));
3952 	crq.query_map.first = IBMVNIC_CRQ_CMD;
3953 	crq.query_map.cmd = QUERY_MAP;
3954 	ibmvnic_send_crq(adapter, &crq);
3955 }
3956 
3957 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3958 static void send_query_cap(struct ibmvnic_adapter *adapter)
3959 {
3960 	union ibmvnic_crq crq;
3961 
3962 	atomic_set(&adapter->running_cap_crqs, 0);
3963 	memset(&crq, 0, sizeof(crq));
3964 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
3965 	crq.query_capability.cmd = QUERY_CAPABILITY;
3966 
3967 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3968 	atomic_inc(&adapter->running_cap_crqs);
3969 	ibmvnic_send_crq(adapter, &crq);
3970 
3971 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3972 	atomic_inc(&adapter->running_cap_crqs);
3973 	ibmvnic_send_crq(adapter, &crq);
3974 
3975 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3976 	atomic_inc(&adapter->running_cap_crqs);
3977 	ibmvnic_send_crq(adapter, &crq);
3978 
3979 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3980 	atomic_inc(&adapter->running_cap_crqs);
3981 	ibmvnic_send_crq(adapter, &crq);
3982 
3983 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3984 	atomic_inc(&adapter->running_cap_crqs);
3985 	ibmvnic_send_crq(adapter, &crq);
3986 
3987 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3988 	atomic_inc(&adapter->running_cap_crqs);
3989 	ibmvnic_send_crq(adapter, &crq);
3990 
3991 	crq.query_capability.capability =
3992 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3993 	atomic_inc(&adapter->running_cap_crqs);
3994 	ibmvnic_send_crq(adapter, &crq);
3995 
3996 	crq.query_capability.capability =
3997 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3998 	atomic_inc(&adapter->running_cap_crqs);
3999 	ibmvnic_send_crq(adapter, &crq);
4000 
4001 	crq.query_capability.capability =
4002 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
4003 	atomic_inc(&adapter->running_cap_crqs);
4004 	ibmvnic_send_crq(adapter, &crq);
4005 
4006 	crq.query_capability.capability =
4007 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
4008 	atomic_inc(&adapter->running_cap_crqs);
4009 	ibmvnic_send_crq(adapter, &crq);
4010 
4011 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
4012 	atomic_inc(&adapter->running_cap_crqs);
4013 	ibmvnic_send_crq(adapter, &crq);
4014 
4015 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
4016 	atomic_inc(&adapter->running_cap_crqs);
4017 	ibmvnic_send_crq(adapter, &crq);
4018 
4019 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
4020 	atomic_inc(&adapter->running_cap_crqs);
4021 	ibmvnic_send_crq(adapter, &crq);
4022 
4023 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
4024 	atomic_inc(&adapter->running_cap_crqs);
4025 	ibmvnic_send_crq(adapter, &crq);
4026 
4027 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
4028 	atomic_inc(&adapter->running_cap_crqs);
4029 	ibmvnic_send_crq(adapter, &crq);
4030 
4031 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
4032 	atomic_inc(&adapter->running_cap_crqs);
4033 	ibmvnic_send_crq(adapter, &crq);
4034 
4035 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
4036 	atomic_inc(&adapter->running_cap_crqs);
4037 	ibmvnic_send_crq(adapter, &crq);
4038 
4039 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
4040 	atomic_inc(&adapter->running_cap_crqs);
4041 	ibmvnic_send_crq(adapter, &crq);
4042 
4043 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
4044 	atomic_inc(&adapter->running_cap_crqs);
4045 	ibmvnic_send_crq(adapter, &crq);
4046 
4047 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
4048 	atomic_inc(&adapter->running_cap_crqs);
4049 	ibmvnic_send_crq(adapter, &crq);
4050 
4051 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
4052 	atomic_inc(&adapter->running_cap_crqs);
4053 	ibmvnic_send_crq(adapter, &crq);
4054 
4055 	crq.query_capability.capability =
4056 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
4057 	atomic_inc(&adapter->running_cap_crqs);
4058 	ibmvnic_send_crq(adapter, &crq);
4059 
4060 	crq.query_capability.capability =
4061 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
4062 	atomic_inc(&adapter->running_cap_crqs);
4063 	ibmvnic_send_crq(adapter, &crq);
4064 
4065 	crq.query_capability.capability =
4066 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
4067 	atomic_inc(&adapter->running_cap_crqs);
4068 	ibmvnic_send_crq(adapter, &crq);
4069 
4070 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
4071 	atomic_inc(&adapter->running_cap_crqs);
4072 	ibmvnic_send_crq(adapter, &crq);
4073 }
4074 
4075 static void send_query_ip_offload(struct ibmvnic_adapter *adapter)
4076 {
4077 	int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4078 	struct device *dev = &adapter->vdev->dev;
4079 	union ibmvnic_crq crq;
4080 
4081 	adapter->ip_offload_tok =
4082 		dma_map_single(dev,
4083 			       &adapter->ip_offload_buf,
4084 			       buf_sz,
4085 			       DMA_FROM_DEVICE);
4086 
4087 	if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4088 		if (!firmware_has_feature(FW_FEATURE_CMO))
4089 			dev_err(dev, "Couldn't map offload buffer\n");
4090 		return;
4091 	}
4092 
4093 	memset(&crq, 0, sizeof(crq));
4094 	crq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4095 	crq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4096 	crq.query_ip_offload.len = cpu_to_be32(buf_sz);
4097 	crq.query_ip_offload.ioba =
4098 	    cpu_to_be32(adapter->ip_offload_tok);
4099 
4100 	ibmvnic_send_crq(adapter, &crq);
4101 }
4102 
4103 static void send_control_ip_offload(struct ibmvnic_adapter *adapter)
4104 {
4105 	struct ibmvnic_control_ip_offload_buffer *ctrl_buf = &adapter->ip_offload_ctrl;
4106 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
4107 	struct device *dev = &adapter->vdev->dev;
4108 	netdev_features_t old_hw_features = 0;
4109 	union ibmvnic_crq crq;
4110 
4111 	adapter->ip_offload_ctrl_tok =
4112 		dma_map_single(dev,
4113 			       ctrl_buf,
4114 			       sizeof(adapter->ip_offload_ctrl),
4115 			       DMA_TO_DEVICE);
4116 
4117 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
4118 		dev_err(dev, "Couldn't map ip offload control buffer\n");
4119 		return;
4120 	}
4121 
4122 	ctrl_buf->len = cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4123 	ctrl_buf->version = cpu_to_be32(INITIAL_VERSION_IOB);
4124 	ctrl_buf->ipv4_chksum = buf->ipv4_chksum;
4125 	ctrl_buf->ipv6_chksum = buf->ipv6_chksum;
4126 	ctrl_buf->tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
4127 	ctrl_buf->udp_ipv4_chksum = buf->udp_ipv4_chksum;
4128 	ctrl_buf->tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
4129 	ctrl_buf->udp_ipv6_chksum = buf->udp_ipv6_chksum;
4130 	ctrl_buf->large_tx_ipv4 = buf->large_tx_ipv4;
4131 	ctrl_buf->large_tx_ipv6 = buf->large_tx_ipv6;
4132 
4133 	/* large_rx disabled for now, additional features needed */
4134 	ctrl_buf->large_rx_ipv4 = 0;
4135 	ctrl_buf->large_rx_ipv6 = 0;
4136 
4137 	if (adapter->state != VNIC_PROBING) {
4138 		old_hw_features = adapter->netdev->hw_features;
4139 		adapter->netdev->hw_features = 0;
4140 	}
4141 
4142 	adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
4143 
4144 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
4145 		adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
4146 
4147 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
4148 		adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
4149 
4150 	if ((adapter->netdev->features &
4151 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
4152 		adapter->netdev->hw_features |= NETIF_F_RXCSUM;
4153 
4154 	if (buf->large_tx_ipv4)
4155 		adapter->netdev->hw_features |= NETIF_F_TSO;
4156 	if (buf->large_tx_ipv6)
4157 		adapter->netdev->hw_features |= NETIF_F_TSO6;
4158 
4159 	if (adapter->state == VNIC_PROBING) {
4160 		adapter->netdev->features |= adapter->netdev->hw_features;
4161 	} else if (old_hw_features != adapter->netdev->hw_features) {
4162 		netdev_features_t tmp = 0;
4163 
4164 		/* disable features no longer supported */
4165 		adapter->netdev->features &= adapter->netdev->hw_features;
4166 		/* turn on features now supported if previously enabled */
4167 		tmp = (old_hw_features ^ adapter->netdev->hw_features) &
4168 			adapter->netdev->hw_features;
4169 		adapter->netdev->features |=
4170 				tmp & adapter->netdev->wanted_features;
4171 	}
4172 
4173 	memset(&crq, 0, sizeof(crq));
4174 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
4175 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
4176 	crq.control_ip_offload.len =
4177 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4178 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
4179 	ibmvnic_send_crq(adapter, &crq);
4180 }
4181 
4182 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
4183 				struct ibmvnic_adapter *adapter)
4184 {
4185 	struct device *dev = &adapter->vdev->dev;
4186 
4187 	if (crq->get_vpd_size_rsp.rc.code) {
4188 		dev_err(dev, "Error retrieving VPD size, rc=%x\n",
4189 			crq->get_vpd_size_rsp.rc.code);
4190 		complete(&adapter->fw_done);
4191 		return;
4192 	}
4193 
4194 	adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
4195 	complete(&adapter->fw_done);
4196 }
4197 
4198 static void handle_vpd_rsp(union ibmvnic_crq *crq,
4199 			   struct ibmvnic_adapter *adapter)
4200 {
4201 	struct device *dev = &adapter->vdev->dev;
4202 	unsigned char *substr = NULL;
4203 	u8 fw_level_len = 0;
4204 
4205 	memset(adapter->fw_version, 0, 32);
4206 
4207 	dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
4208 			 DMA_FROM_DEVICE);
4209 
4210 	if (crq->get_vpd_rsp.rc.code) {
4211 		dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
4212 			crq->get_vpd_rsp.rc.code);
4213 		goto complete;
4214 	}
4215 
4216 	/* get the position of the firmware version info
4217 	 * located after the ASCII 'RM' substring in the buffer
4218 	 */
4219 	substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
4220 	if (!substr) {
4221 		dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
4222 		goto complete;
4223 	}
4224 
4225 	/* get length of firmware level ASCII substring */
4226 	if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
4227 		fw_level_len = *(substr + 2);
4228 	} else {
4229 		dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
4230 		goto complete;
4231 	}
4232 
4233 	/* copy firmware version string from vpd into adapter */
4234 	if ((substr + 3 + fw_level_len) <
4235 	    (adapter->vpd->buff + adapter->vpd->len)) {
4236 		strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
4237 	} else {
4238 		dev_info(dev, "FW substr extrapolated VPD buff\n");
4239 	}
4240 
4241 complete:
4242 	if (adapter->fw_version[0] == '\0')
4243 		strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
4244 	complete(&adapter->fw_done);
4245 }
4246 
4247 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
4248 {
4249 	struct device *dev = &adapter->vdev->dev;
4250 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
4251 	int i;
4252 
4253 	dma_unmap_single(dev, adapter->ip_offload_tok,
4254 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
4255 
4256 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
4257 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
4258 		netdev_dbg(adapter->netdev, "%016lx\n",
4259 			   ((unsigned long int *)(buf))[i]);
4260 
4261 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
4262 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
4263 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
4264 		   buf->tcp_ipv4_chksum);
4265 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
4266 		   buf->tcp_ipv6_chksum);
4267 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
4268 		   buf->udp_ipv4_chksum);
4269 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
4270 		   buf->udp_ipv6_chksum);
4271 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
4272 		   buf->large_tx_ipv4);
4273 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
4274 		   buf->large_tx_ipv6);
4275 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
4276 		   buf->large_rx_ipv4);
4277 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
4278 		   buf->large_rx_ipv6);
4279 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
4280 		   buf->max_ipv4_header_size);
4281 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
4282 		   buf->max_ipv6_header_size);
4283 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
4284 		   buf->max_tcp_header_size);
4285 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
4286 		   buf->max_udp_header_size);
4287 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
4288 		   buf->max_large_tx_size);
4289 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
4290 		   buf->max_large_rx_size);
4291 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
4292 		   buf->ipv6_extension_header);
4293 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
4294 		   buf->tcp_pseudosum_req);
4295 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
4296 		   buf->num_ipv6_ext_headers);
4297 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
4298 		   buf->off_ipv6_ext_headers);
4299 
4300 	send_control_ip_offload(adapter);
4301 }
4302 
4303 static const char *ibmvnic_fw_err_cause(u16 cause)
4304 {
4305 	switch (cause) {
4306 	case ADAPTER_PROBLEM:
4307 		return "adapter problem";
4308 	case BUS_PROBLEM:
4309 		return "bus problem";
4310 	case FW_PROBLEM:
4311 		return "firmware problem";
4312 	case DD_PROBLEM:
4313 		return "device driver problem";
4314 	case EEH_RECOVERY:
4315 		return "EEH recovery";
4316 	case FW_UPDATED:
4317 		return "firmware updated";
4318 	case LOW_MEMORY:
4319 		return "low Memory";
4320 	default:
4321 		return "unknown";
4322 	}
4323 }
4324 
4325 static void handle_error_indication(union ibmvnic_crq *crq,
4326 				    struct ibmvnic_adapter *adapter)
4327 {
4328 	struct device *dev = &adapter->vdev->dev;
4329 	u16 cause;
4330 
4331 	cause = be16_to_cpu(crq->error_indication.error_cause);
4332 
4333 	dev_warn_ratelimited(dev,
4334 			     "Firmware reports %serror, cause: %s. Starting recovery...\n",
4335 			     crq->error_indication.flags
4336 				& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
4337 			     ibmvnic_fw_err_cause(cause));
4338 
4339 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
4340 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4341 	else
4342 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
4343 }
4344 
4345 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
4346 				 struct ibmvnic_adapter *adapter)
4347 {
4348 	struct net_device *netdev = adapter->netdev;
4349 	struct device *dev = &adapter->vdev->dev;
4350 	long rc;
4351 
4352 	rc = crq->change_mac_addr_rsp.rc.code;
4353 	if (rc) {
4354 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
4355 		goto out;
4356 	}
4357 	/* crq->change_mac_addr.mac_addr is the requested one
4358 	 * crq->change_mac_addr_rsp.mac_addr is the returned valid one.
4359 	 */
4360 	ether_addr_copy(netdev->dev_addr,
4361 			&crq->change_mac_addr_rsp.mac_addr[0]);
4362 	ether_addr_copy(adapter->mac_addr,
4363 			&crq->change_mac_addr_rsp.mac_addr[0]);
4364 out:
4365 	complete(&adapter->fw_done);
4366 	return rc;
4367 }
4368 
4369 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
4370 				   struct ibmvnic_adapter *adapter)
4371 {
4372 	struct device *dev = &adapter->vdev->dev;
4373 	u64 *req_value;
4374 	char *name;
4375 
4376 	atomic_dec(&adapter->running_cap_crqs);
4377 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
4378 	case REQ_TX_QUEUES:
4379 		req_value = &adapter->req_tx_queues;
4380 		name = "tx";
4381 		break;
4382 	case REQ_RX_QUEUES:
4383 		req_value = &adapter->req_rx_queues;
4384 		name = "rx";
4385 		break;
4386 	case REQ_RX_ADD_QUEUES:
4387 		req_value = &adapter->req_rx_add_queues;
4388 		name = "rx_add";
4389 		break;
4390 	case REQ_TX_ENTRIES_PER_SUBCRQ:
4391 		req_value = &adapter->req_tx_entries_per_subcrq;
4392 		name = "tx_entries_per_subcrq";
4393 		break;
4394 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
4395 		req_value = &adapter->req_rx_add_entries_per_subcrq;
4396 		name = "rx_add_entries_per_subcrq";
4397 		break;
4398 	case REQ_MTU:
4399 		req_value = &adapter->req_mtu;
4400 		name = "mtu";
4401 		break;
4402 	case PROMISC_REQUESTED:
4403 		req_value = &adapter->promisc;
4404 		name = "promisc";
4405 		break;
4406 	default:
4407 		dev_err(dev, "Got invalid cap request rsp %d\n",
4408 			crq->request_capability.capability);
4409 		return;
4410 	}
4411 
4412 	switch (crq->request_capability_rsp.rc.code) {
4413 	case SUCCESS:
4414 		break;
4415 	case PARTIALSUCCESS:
4416 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
4417 			 *req_value,
4418 			 (long int)be64_to_cpu(crq->request_capability_rsp.
4419 					       number), name);
4420 
4421 		if (be16_to_cpu(crq->request_capability_rsp.capability) ==
4422 		    REQ_MTU) {
4423 			pr_err("mtu of %llu is not supported. Reverting.\n",
4424 			       *req_value);
4425 			*req_value = adapter->fallback.mtu;
4426 		} else {
4427 			*req_value =
4428 				be64_to_cpu(crq->request_capability_rsp.number);
4429 		}
4430 
4431 		send_request_cap(adapter, 1);
4432 		return;
4433 	default:
4434 		dev_err(dev, "Error %d in request cap rsp\n",
4435 			crq->request_capability_rsp.rc.code);
4436 		return;
4437 	}
4438 
4439 	/* Done receiving requested capabilities, query IP offload support */
4440 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4441 		adapter->wait_capability = false;
4442 		send_query_ip_offload(adapter);
4443 	}
4444 }
4445 
4446 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4447 			    struct ibmvnic_adapter *adapter)
4448 {
4449 	struct device *dev = &adapter->vdev->dev;
4450 	struct net_device *netdev = adapter->netdev;
4451 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4452 	struct ibmvnic_login_buffer *login = adapter->login_buf;
4453 	u64 *tx_handle_array;
4454 	u64 *rx_handle_array;
4455 	int num_tx_pools;
4456 	int num_rx_pools;
4457 	u64 *size_array;
4458 	int i;
4459 
4460 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4461 			 DMA_TO_DEVICE);
4462 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
4463 			 adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4464 
4465 	/* If the number of queues requested can't be allocated by the
4466 	 * server, the login response will return with code 1. We will need
4467 	 * to resend the login buffer with fewer queues requested.
4468 	 */
4469 	if (login_rsp_crq->generic.rc.code) {
4470 		adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4471 		complete(&adapter->init_done);
4472 		return 0;
4473 	}
4474 
4475 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4476 
4477 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4478 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4479 		netdev_dbg(adapter->netdev, "%016lx\n",
4480 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4481 	}
4482 
4483 	/* Sanity checks */
4484 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4485 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
4486 	     adapter->req_rx_add_queues !=
4487 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4488 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4489 		ibmvnic_remove(adapter->vdev);
4490 		return -EIO;
4491 	}
4492 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4493 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
4494 	/* variable buffer sizes are not supported, so just read the
4495 	 * first entry.
4496 	 */
4497 	adapter->cur_rx_buf_sz = be64_to_cpu(size_array[0]);
4498 
4499 	num_tx_pools = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
4500 	num_rx_pools = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4501 
4502 	tx_handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4503 				  be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
4504 	rx_handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4505 				  be32_to_cpu(adapter->login_rsp_buf->off_rxadd_subcrqs));
4506 
4507 	for (i = 0; i < num_tx_pools; i++)
4508 		adapter->tx_scrq[i]->handle = tx_handle_array[i];
4509 
4510 	for (i = 0; i < num_rx_pools; i++)
4511 		adapter->rx_scrq[i]->handle = rx_handle_array[i];
4512 
4513 	adapter->num_active_tx_scrqs = num_tx_pools;
4514 	adapter->num_active_rx_scrqs = num_rx_pools;
4515 	release_login_rsp_buffer(adapter);
4516 	release_login_buffer(adapter);
4517 	complete(&adapter->init_done);
4518 
4519 	return 0;
4520 }
4521 
4522 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4523 				     struct ibmvnic_adapter *adapter)
4524 {
4525 	struct device *dev = &adapter->vdev->dev;
4526 	long rc;
4527 
4528 	rc = crq->request_unmap_rsp.rc.code;
4529 	if (rc)
4530 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4531 }
4532 
4533 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4534 				 struct ibmvnic_adapter *adapter)
4535 {
4536 	struct net_device *netdev = adapter->netdev;
4537 	struct device *dev = &adapter->vdev->dev;
4538 	long rc;
4539 
4540 	rc = crq->query_map_rsp.rc.code;
4541 	if (rc) {
4542 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4543 		return;
4544 	}
4545 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4546 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4547 		   crq->query_map_rsp.free_pages);
4548 }
4549 
4550 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4551 				 struct ibmvnic_adapter *adapter)
4552 {
4553 	struct net_device *netdev = adapter->netdev;
4554 	struct device *dev = &adapter->vdev->dev;
4555 	long rc;
4556 
4557 	atomic_dec(&adapter->running_cap_crqs);
4558 	netdev_dbg(netdev, "Outstanding queries: %d\n",
4559 		   atomic_read(&adapter->running_cap_crqs));
4560 	rc = crq->query_capability.rc.code;
4561 	if (rc) {
4562 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4563 		goto out;
4564 	}
4565 
4566 	switch (be16_to_cpu(crq->query_capability.capability)) {
4567 	case MIN_TX_QUEUES:
4568 		adapter->min_tx_queues =
4569 		    be64_to_cpu(crq->query_capability.number);
4570 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
4571 			   adapter->min_tx_queues);
4572 		break;
4573 	case MIN_RX_QUEUES:
4574 		adapter->min_rx_queues =
4575 		    be64_to_cpu(crq->query_capability.number);
4576 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
4577 			   adapter->min_rx_queues);
4578 		break;
4579 	case MIN_RX_ADD_QUEUES:
4580 		adapter->min_rx_add_queues =
4581 		    be64_to_cpu(crq->query_capability.number);
4582 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4583 			   adapter->min_rx_add_queues);
4584 		break;
4585 	case MAX_TX_QUEUES:
4586 		adapter->max_tx_queues =
4587 		    be64_to_cpu(crq->query_capability.number);
4588 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
4589 			   adapter->max_tx_queues);
4590 		break;
4591 	case MAX_RX_QUEUES:
4592 		adapter->max_rx_queues =
4593 		    be64_to_cpu(crq->query_capability.number);
4594 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
4595 			   adapter->max_rx_queues);
4596 		break;
4597 	case MAX_RX_ADD_QUEUES:
4598 		adapter->max_rx_add_queues =
4599 		    be64_to_cpu(crq->query_capability.number);
4600 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4601 			   adapter->max_rx_add_queues);
4602 		break;
4603 	case MIN_TX_ENTRIES_PER_SUBCRQ:
4604 		adapter->min_tx_entries_per_subcrq =
4605 		    be64_to_cpu(crq->query_capability.number);
4606 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4607 			   adapter->min_tx_entries_per_subcrq);
4608 		break;
4609 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4610 		adapter->min_rx_add_entries_per_subcrq =
4611 		    be64_to_cpu(crq->query_capability.number);
4612 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4613 			   adapter->min_rx_add_entries_per_subcrq);
4614 		break;
4615 	case MAX_TX_ENTRIES_PER_SUBCRQ:
4616 		adapter->max_tx_entries_per_subcrq =
4617 		    be64_to_cpu(crq->query_capability.number);
4618 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4619 			   adapter->max_tx_entries_per_subcrq);
4620 		break;
4621 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4622 		adapter->max_rx_add_entries_per_subcrq =
4623 		    be64_to_cpu(crq->query_capability.number);
4624 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4625 			   adapter->max_rx_add_entries_per_subcrq);
4626 		break;
4627 	case TCP_IP_OFFLOAD:
4628 		adapter->tcp_ip_offload =
4629 		    be64_to_cpu(crq->query_capability.number);
4630 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4631 			   adapter->tcp_ip_offload);
4632 		break;
4633 	case PROMISC_SUPPORTED:
4634 		adapter->promisc_supported =
4635 		    be64_to_cpu(crq->query_capability.number);
4636 		netdev_dbg(netdev, "promisc_supported = %lld\n",
4637 			   adapter->promisc_supported);
4638 		break;
4639 	case MIN_MTU:
4640 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4641 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4642 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4643 		break;
4644 	case MAX_MTU:
4645 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4646 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4647 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4648 		break;
4649 	case MAX_MULTICAST_FILTERS:
4650 		adapter->max_multicast_filters =
4651 		    be64_to_cpu(crq->query_capability.number);
4652 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4653 			   adapter->max_multicast_filters);
4654 		break;
4655 	case VLAN_HEADER_INSERTION:
4656 		adapter->vlan_header_insertion =
4657 		    be64_to_cpu(crq->query_capability.number);
4658 		if (adapter->vlan_header_insertion)
4659 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4660 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4661 			   adapter->vlan_header_insertion);
4662 		break;
4663 	case RX_VLAN_HEADER_INSERTION:
4664 		adapter->rx_vlan_header_insertion =
4665 		    be64_to_cpu(crq->query_capability.number);
4666 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4667 			   adapter->rx_vlan_header_insertion);
4668 		break;
4669 	case MAX_TX_SG_ENTRIES:
4670 		adapter->max_tx_sg_entries =
4671 		    be64_to_cpu(crq->query_capability.number);
4672 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4673 			   adapter->max_tx_sg_entries);
4674 		break;
4675 	case RX_SG_SUPPORTED:
4676 		adapter->rx_sg_supported =
4677 		    be64_to_cpu(crq->query_capability.number);
4678 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4679 			   adapter->rx_sg_supported);
4680 		break;
4681 	case OPT_TX_COMP_SUB_QUEUES:
4682 		adapter->opt_tx_comp_sub_queues =
4683 		    be64_to_cpu(crq->query_capability.number);
4684 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4685 			   adapter->opt_tx_comp_sub_queues);
4686 		break;
4687 	case OPT_RX_COMP_QUEUES:
4688 		adapter->opt_rx_comp_queues =
4689 		    be64_to_cpu(crq->query_capability.number);
4690 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4691 			   adapter->opt_rx_comp_queues);
4692 		break;
4693 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4694 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
4695 		    be64_to_cpu(crq->query_capability.number);
4696 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4697 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
4698 		break;
4699 	case OPT_TX_ENTRIES_PER_SUBCRQ:
4700 		adapter->opt_tx_entries_per_subcrq =
4701 		    be64_to_cpu(crq->query_capability.number);
4702 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4703 			   adapter->opt_tx_entries_per_subcrq);
4704 		break;
4705 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4706 		adapter->opt_rxba_entries_per_subcrq =
4707 		    be64_to_cpu(crq->query_capability.number);
4708 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4709 			   adapter->opt_rxba_entries_per_subcrq);
4710 		break;
4711 	case TX_RX_DESC_REQ:
4712 		adapter->tx_rx_desc_req = crq->query_capability.number;
4713 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4714 			   adapter->tx_rx_desc_req);
4715 		break;
4716 
4717 	default:
4718 		netdev_err(netdev, "Got invalid cap rsp %d\n",
4719 			   crq->query_capability.capability);
4720 	}
4721 
4722 out:
4723 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4724 		adapter->wait_capability = false;
4725 		send_request_cap(adapter, 0);
4726 	}
4727 }
4728 
4729 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4730 {
4731 	union ibmvnic_crq crq;
4732 	int rc;
4733 
4734 	memset(&crq, 0, sizeof(crq));
4735 	crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4736 	crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4737 
4738 	mutex_lock(&adapter->fw_lock);
4739 	adapter->fw_done_rc = 0;
4740 	reinit_completion(&adapter->fw_done);
4741 
4742 	rc = ibmvnic_send_crq(adapter, &crq);
4743 	if (rc) {
4744 		mutex_unlock(&adapter->fw_lock);
4745 		return rc;
4746 	}
4747 
4748 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
4749 	if (rc) {
4750 		mutex_unlock(&adapter->fw_lock);
4751 		return rc;
4752 	}
4753 
4754 	mutex_unlock(&adapter->fw_lock);
4755 	return adapter->fw_done_rc ? -EIO : 0;
4756 }
4757 
4758 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4759 				       struct ibmvnic_adapter *adapter)
4760 {
4761 	struct net_device *netdev = adapter->netdev;
4762 	int rc;
4763 	__be32 rspeed = cpu_to_be32(crq->query_phys_parms_rsp.speed);
4764 
4765 	rc = crq->query_phys_parms_rsp.rc.code;
4766 	if (rc) {
4767 		netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4768 		return rc;
4769 	}
4770 	switch (rspeed) {
4771 	case IBMVNIC_10MBPS:
4772 		adapter->speed = SPEED_10;
4773 		break;
4774 	case IBMVNIC_100MBPS:
4775 		adapter->speed = SPEED_100;
4776 		break;
4777 	case IBMVNIC_1GBPS:
4778 		adapter->speed = SPEED_1000;
4779 		break;
4780 	case IBMVNIC_10GBPS:
4781 		adapter->speed = SPEED_10000;
4782 		break;
4783 	case IBMVNIC_25GBPS:
4784 		adapter->speed = SPEED_25000;
4785 		break;
4786 	case IBMVNIC_40GBPS:
4787 		adapter->speed = SPEED_40000;
4788 		break;
4789 	case IBMVNIC_50GBPS:
4790 		adapter->speed = SPEED_50000;
4791 		break;
4792 	case IBMVNIC_100GBPS:
4793 		adapter->speed = SPEED_100000;
4794 		break;
4795 	case IBMVNIC_200GBPS:
4796 		adapter->speed = SPEED_200000;
4797 		break;
4798 	default:
4799 		if (netif_carrier_ok(netdev))
4800 			netdev_warn(netdev, "Unknown speed 0x%08x\n", rspeed);
4801 		adapter->speed = SPEED_UNKNOWN;
4802 	}
4803 	if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4804 		adapter->duplex = DUPLEX_FULL;
4805 	else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4806 		adapter->duplex = DUPLEX_HALF;
4807 	else
4808 		adapter->duplex = DUPLEX_UNKNOWN;
4809 
4810 	return rc;
4811 }
4812 
4813 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4814 			       struct ibmvnic_adapter *adapter)
4815 {
4816 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4817 	struct net_device *netdev = adapter->netdev;
4818 	struct device *dev = &adapter->vdev->dev;
4819 	u64 *u64_crq = (u64 *)crq;
4820 	long rc;
4821 
4822 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4823 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
4824 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
4825 	switch (gen_crq->first) {
4826 	case IBMVNIC_CRQ_INIT_RSP:
4827 		switch (gen_crq->cmd) {
4828 		case IBMVNIC_CRQ_INIT:
4829 			dev_info(dev, "Partner initialized\n");
4830 			adapter->from_passive_init = true;
4831 			if (!completion_done(&adapter->init_done)) {
4832 				complete(&adapter->init_done);
4833 				adapter->init_done_rc = -EIO;
4834 			}
4835 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4836 			break;
4837 		case IBMVNIC_CRQ_INIT_COMPLETE:
4838 			dev_info(dev, "Partner initialization complete\n");
4839 			adapter->crq.active = true;
4840 			send_version_xchg(adapter);
4841 			break;
4842 		default:
4843 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4844 		}
4845 		return;
4846 	case IBMVNIC_CRQ_XPORT_EVENT:
4847 		netif_carrier_off(netdev);
4848 		adapter->crq.active = false;
4849 		/* terminate any thread waiting for a response
4850 		 * from the device
4851 		 */
4852 		if (!completion_done(&adapter->fw_done)) {
4853 			adapter->fw_done_rc = -EIO;
4854 			complete(&adapter->fw_done);
4855 		}
4856 		if (!completion_done(&adapter->stats_done))
4857 			complete(&adapter->stats_done);
4858 		if (test_bit(0, &adapter->resetting))
4859 			adapter->force_reset_recovery = true;
4860 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4861 			dev_info(dev, "Migrated, re-enabling adapter\n");
4862 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4863 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4864 			dev_info(dev, "Backing device failover detected\n");
4865 			adapter->failover_pending = true;
4866 		} else {
4867 			/* The adapter lost the connection */
4868 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4869 				gen_crq->cmd);
4870 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4871 		}
4872 		return;
4873 	case IBMVNIC_CRQ_CMD_RSP:
4874 		break;
4875 	default:
4876 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
4877 			gen_crq->first);
4878 		return;
4879 	}
4880 
4881 	switch (gen_crq->cmd) {
4882 	case VERSION_EXCHANGE_RSP:
4883 		rc = crq->version_exchange_rsp.rc.code;
4884 		if (rc) {
4885 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4886 			break;
4887 		}
4888 		ibmvnic_version =
4889 			    be16_to_cpu(crq->version_exchange_rsp.version);
4890 		dev_info(dev, "Partner protocol version is %d\n",
4891 			 ibmvnic_version);
4892 		send_query_cap(adapter);
4893 		break;
4894 	case QUERY_CAPABILITY_RSP:
4895 		handle_query_cap_rsp(crq, adapter);
4896 		break;
4897 	case QUERY_MAP_RSP:
4898 		handle_query_map_rsp(crq, adapter);
4899 		break;
4900 	case REQUEST_MAP_RSP:
4901 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4902 		complete(&adapter->fw_done);
4903 		break;
4904 	case REQUEST_UNMAP_RSP:
4905 		handle_request_unmap_rsp(crq, adapter);
4906 		break;
4907 	case REQUEST_CAPABILITY_RSP:
4908 		handle_request_cap_rsp(crq, adapter);
4909 		break;
4910 	case LOGIN_RSP:
4911 		netdev_dbg(netdev, "Got Login Response\n");
4912 		handle_login_rsp(crq, adapter);
4913 		break;
4914 	case LOGICAL_LINK_STATE_RSP:
4915 		netdev_dbg(netdev,
4916 			   "Got Logical Link State Response, state: %d rc: %d\n",
4917 			   crq->logical_link_state_rsp.link_state,
4918 			   crq->logical_link_state_rsp.rc.code);
4919 		adapter->logical_link_state =
4920 		    crq->logical_link_state_rsp.link_state;
4921 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4922 		complete(&adapter->init_done);
4923 		break;
4924 	case LINK_STATE_INDICATION:
4925 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
4926 		adapter->phys_link_state =
4927 		    crq->link_state_indication.phys_link_state;
4928 		adapter->logical_link_state =
4929 		    crq->link_state_indication.logical_link_state;
4930 		if (adapter->phys_link_state && adapter->logical_link_state)
4931 			netif_carrier_on(netdev);
4932 		else
4933 			netif_carrier_off(netdev);
4934 		break;
4935 	case CHANGE_MAC_ADDR_RSP:
4936 		netdev_dbg(netdev, "Got MAC address change Response\n");
4937 		adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4938 		break;
4939 	case ERROR_INDICATION:
4940 		netdev_dbg(netdev, "Got Error Indication\n");
4941 		handle_error_indication(crq, adapter);
4942 		break;
4943 	case REQUEST_STATISTICS_RSP:
4944 		netdev_dbg(netdev, "Got Statistics Response\n");
4945 		complete(&adapter->stats_done);
4946 		break;
4947 	case QUERY_IP_OFFLOAD_RSP:
4948 		netdev_dbg(netdev, "Got Query IP offload Response\n");
4949 		handle_query_ip_offload_rsp(adapter);
4950 		break;
4951 	case MULTICAST_CTRL_RSP:
4952 		netdev_dbg(netdev, "Got multicast control Response\n");
4953 		break;
4954 	case CONTROL_IP_OFFLOAD_RSP:
4955 		netdev_dbg(netdev, "Got Control IP offload Response\n");
4956 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4957 				 sizeof(adapter->ip_offload_ctrl),
4958 				 DMA_TO_DEVICE);
4959 		complete(&adapter->init_done);
4960 		break;
4961 	case COLLECT_FW_TRACE_RSP:
4962 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4963 		complete(&adapter->fw_done);
4964 		break;
4965 	case GET_VPD_SIZE_RSP:
4966 		handle_vpd_size_rsp(crq, adapter);
4967 		break;
4968 	case GET_VPD_RSP:
4969 		handle_vpd_rsp(crq, adapter);
4970 		break;
4971 	case QUERY_PHYS_PARMS_RSP:
4972 		adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4973 		complete(&adapter->fw_done);
4974 		break;
4975 	default:
4976 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4977 			   gen_crq->cmd);
4978 	}
4979 }
4980 
4981 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4982 {
4983 	struct ibmvnic_adapter *adapter = instance;
4984 
4985 	tasklet_schedule(&adapter->tasklet);
4986 	return IRQ_HANDLED;
4987 }
4988 
4989 static void ibmvnic_tasklet(struct tasklet_struct *t)
4990 {
4991 	struct ibmvnic_adapter *adapter = from_tasklet(adapter, t, tasklet);
4992 	struct ibmvnic_crq_queue *queue = &adapter->crq;
4993 	union ibmvnic_crq *crq;
4994 	unsigned long flags;
4995 	bool done = false;
4996 
4997 	spin_lock_irqsave(&queue->lock, flags);
4998 	while (!done) {
4999 		/* Pull all the valid messages off the CRQ */
5000 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
5001 			ibmvnic_handle_crq(crq, adapter);
5002 			crq->generic.first = 0;
5003 		}
5004 
5005 		/* remain in tasklet until all
5006 		 * capabilities responses are received
5007 		 */
5008 		if (!adapter->wait_capability)
5009 			done = true;
5010 	}
5011 	/* if capabilities CRQ's were sent in this tasklet, the following
5012 	 * tasklet must wait until all responses are received
5013 	 */
5014 	if (atomic_read(&adapter->running_cap_crqs) != 0)
5015 		adapter->wait_capability = true;
5016 	spin_unlock_irqrestore(&queue->lock, flags);
5017 }
5018 
5019 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
5020 {
5021 	struct vio_dev *vdev = adapter->vdev;
5022 	int rc;
5023 
5024 	do {
5025 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
5026 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
5027 
5028 	if (rc)
5029 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
5030 
5031 	return rc;
5032 }
5033 
5034 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
5035 {
5036 	struct ibmvnic_crq_queue *crq = &adapter->crq;
5037 	struct device *dev = &adapter->vdev->dev;
5038 	struct vio_dev *vdev = adapter->vdev;
5039 	int rc;
5040 
5041 	/* Close the CRQ */
5042 	do {
5043 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
5044 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
5045 
5046 	/* Clean out the queue */
5047 	memset(crq->msgs, 0, PAGE_SIZE);
5048 	crq->cur = 0;
5049 	crq->active = false;
5050 
5051 	/* And re-open it again */
5052 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
5053 				crq->msg_token, PAGE_SIZE);
5054 
5055 	if (rc == H_CLOSED)
5056 		/* Adapter is good, but other end is not ready */
5057 		dev_warn(dev, "Partner adapter not ready\n");
5058 	else if (rc != 0)
5059 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
5060 
5061 	return rc;
5062 }
5063 
5064 static void release_crq_queue(struct ibmvnic_adapter *adapter)
5065 {
5066 	struct ibmvnic_crq_queue *crq = &adapter->crq;
5067 	struct vio_dev *vdev = adapter->vdev;
5068 	long rc;
5069 
5070 	if (!crq->msgs)
5071 		return;
5072 
5073 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
5074 	free_irq(vdev->irq, adapter);
5075 	tasklet_kill(&adapter->tasklet);
5076 	do {
5077 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
5078 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
5079 
5080 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
5081 			 DMA_BIDIRECTIONAL);
5082 	free_page((unsigned long)crq->msgs);
5083 	crq->msgs = NULL;
5084 	crq->active = false;
5085 }
5086 
5087 static int init_crq_queue(struct ibmvnic_adapter *adapter)
5088 {
5089 	struct ibmvnic_crq_queue *crq = &adapter->crq;
5090 	struct device *dev = &adapter->vdev->dev;
5091 	struct vio_dev *vdev = adapter->vdev;
5092 	int rc, retrc = -ENOMEM;
5093 
5094 	if (crq->msgs)
5095 		return 0;
5096 
5097 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
5098 	/* Should we allocate more than one page? */
5099 
5100 	if (!crq->msgs)
5101 		return -ENOMEM;
5102 
5103 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
5104 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
5105 					DMA_BIDIRECTIONAL);
5106 	if (dma_mapping_error(dev, crq->msg_token))
5107 		goto map_failed;
5108 
5109 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
5110 				crq->msg_token, PAGE_SIZE);
5111 
5112 	if (rc == H_RESOURCE)
5113 		/* maybe kexecing and resource is busy. try a reset */
5114 		rc = ibmvnic_reset_crq(adapter);
5115 	retrc = rc;
5116 
5117 	if (rc == H_CLOSED) {
5118 		dev_warn(dev, "Partner adapter not ready\n");
5119 	} else if (rc) {
5120 		dev_warn(dev, "Error %d opening adapter\n", rc);
5121 		goto reg_crq_failed;
5122 	}
5123 
5124 	retrc = 0;
5125 
5126 	tasklet_setup(&adapter->tasklet, (void *)ibmvnic_tasklet);
5127 
5128 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
5129 	snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
5130 		 adapter->vdev->unit_address);
5131 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
5132 	if (rc) {
5133 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
5134 			vdev->irq, rc);
5135 		goto req_irq_failed;
5136 	}
5137 
5138 	rc = vio_enable_interrupts(vdev);
5139 	if (rc) {
5140 		dev_err(dev, "Error %d enabling interrupts\n", rc);
5141 		goto req_irq_failed;
5142 	}
5143 
5144 	crq->cur = 0;
5145 	spin_lock_init(&crq->lock);
5146 
5147 	return retrc;
5148 
5149 req_irq_failed:
5150 	tasklet_kill(&adapter->tasklet);
5151 	do {
5152 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
5153 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
5154 reg_crq_failed:
5155 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
5156 map_failed:
5157 	free_page((unsigned long)crq->msgs);
5158 	crq->msgs = NULL;
5159 	return retrc;
5160 }
5161 
5162 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter, bool reset)
5163 {
5164 	struct device *dev = &adapter->vdev->dev;
5165 	unsigned long timeout = msecs_to_jiffies(30000);
5166 	u64 old_num_rx_queues, old_num_tx_queues;
5167 	int rc;
5168 
5169 	adapter->from_passive_init = false;
5170 
5171 	if (reset) {
5172 		old_num_rx_queues = adapter->req_rx_queues;
5173 		old_num_tx_queues = adapter->req_tx_queues;
5174 		reinit_completion(&adapter->init_done);
5175 	}
5176 
5177 	adapter->init_done_rc = 0;
5178 	rc = ibmvnic_send_crq_init(adapter);
5179 	if (rc) {
5180 		dev_err(dev, "Send crq init failed with error %d\n", rc);
5181 		return rc;
5182 	}
5183 
5184 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
5185 		dev_err(dev, "Initialization sequence timed out\n");
5186 		return -1;
5187 	}
5188 
5189 	if (adapter->init_done_rc) {
5190 		release_crq_queue(adapter);
5191 		return adapter->init_done_rc;
5192 	}
5193 
5194 	if (adapter->from_passive_init) {
5195 		adapter->state = VNIC_OPEN;
5196 		adapter->from_passive_init = false;
5197 		return -1;
5198 	}
5199 
5200 	if (reset &&
5201 	    test_bit(0, &adapter->resetting) && !adapter->wait_for_reset &&
5202 	    adapter->reset_reason != VNIC_RESET_MOBILITY) {
5203 		if (adapter->req_rx_queues != old_num_rx_queues ||
5204 		    adapter->req_tx_queues != old_num_tx_queues) {
5205 			release_sub_crqs(adapter, 0);
5206 			rc = init_sub_crqs(adapter);
5207 		} else {
5208 			rc = reset_sub_crq_queues(adapter);
5209 		}
5210 	} else {
5211 		rc = init_sub_crqs(adapter);
5212 	}
5213 
5214 	if (rc) {
5215 		dev_err(dev, "Initialization of sub crqs failed\n");
5216 		release_crq_queue(adapter);
5217 		return rc;
5218 	}
5219 
5220 	rc = init_sub_crq_irqs(adapter);
5221 	if (rc) {
5222 		dev_err(dev, "Failed to initialize sub crq irqs\n");
5223 		release_crq_queue(adapter);
5224 	}
5225 
5226 	return rc;
5227 }
5228 
5229 static struct device_attribute dev_attr_failover;
5230 
5231 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
5232 {
5233 	struct ibmvnic_adapter *adapter;
5234 	struct net_device *netdev;
5235 	unsigned char *mac_addr_p;
5236 	int rc;
5237 
5238 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
5239 		dev->unit_address);
5240 
5241 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
5242 							VETH_MAC_ADDR, NULL);
5243 	if (!mac_addr_p) {
5244 		dev_err(&dev->dev,
5245 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
5246 			__FILE__, __LINE__);
5247 		return 0;
5248 	}
5249 
5250 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
5251 				   IBMVNIC_MAX_QUEUES);
5252 	if (!netdev)
5253 		return -ENOMEM;
5254 
5255 	adapter = netdev_priv(netdev);
5256 	adapter->state = VNIC_PROBING;
5257 	dev_set_drvdata(&dev->dev, netdev);
5258 	adapter->vdev = dev;
5259 	adapter->netdev = netdev;
5260 
5261 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
5262 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
5263 	netdev->irq = dev->irq;
5264 	netdev->netdev_ops = &ibmvnic_netdev_ops;
5265 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
5266 	SET_NETDEV_DEV(netdev, &dev->dev);
5267 
5268 	spin_lock_init(&adapter->stats_lock);
5269 
5270 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
5271 	INIT_DELAYED_WORK(&adapter->ibmvnic_delayed_reset,
5272 			  __ibmvnic_delayed_reset);
5273 	INIT_LIST_HEAD(&adapter->rwi_list);
5274 	spin_lock_init(&adapter->rwi_lock);
5275 	spin_lock_init(&adapter->state_lock);
5276 	mutex_init(&adapter->fw_lock);
5277 	init_completion(&adapter->init_done);
5278 	init_completion(&adapter->fw_done);
5279 	init_completion(&adapter->reset_done);
5280 	init_completion(&adapter->stats_done);
5281 	clear_bit(0, &adapter->resetting);
5282 
5283 	do {
5284 		rc = init_crq_queue(adapter);
5285 		if (rc) {
5286 			dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
5287 				rc);
5288 			goto ibmvnic_init_fail;
5289 		}
5290 
5291 		rc = ibmvnic_reset_init(adapter, false);
5292 		if (rc && rc != EAGAIN)
5293 			goto ibmvnic_init_fail;
5294 	} while (rc == EAGAIN);
5295 
5296 	rc = init_stats_buffers(adapter);
5297 	if (rc)
5298 		goto ibmvnic_init_fail;
5299 
5300 	rc = init_stats_token(adapter);
5301 	if (rc)
5302 		goto ibmvnic_stats_fail;
5303 
5304 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
5305 	netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
5306 	netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
5307 
5308 	rc = device_create_file(&dev->dev, &dev_attr_failover);
5309 	if (rc)
5310 		goto ibmvnic_dev_file_err;
5311 
5312 	netif_carrier_off(netdev);
5313 	rc = register_netdev(netdev);
5314 	if (rc) {
5315 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
5316 		goto ibmvnic_register_fail;
5317 	}
5318 	dev_info(&dev->dev, "ibmvnic registered\n");
5319 
5320 	adapter->state = VNIC_PROBED;
5321 
5322 	adapter->wait_for_reset = false;
5323 
5324 	return 0;
5325 
5326 ibmvnic_register_fail:
5327 	device_remove_file(&dev->dev, &dev_attr_failover);
5328 
5329 ibmvnic_dev_file_err:
5330 	release_stats_token(adapter);
5331 
5332 ibmvnic_stats_fail:
5333 	release_stats_buffers(adapter);
5334 
5335 ibmvnic_init_fail:
5336 	release_sub_crqs(adapter, 1);
5337 	release_crq_queue(adapter);
5338 	mutex_destroy(&adapter->fw_lock);
5339 	free_netdev(netdev);
5340 
5341 	return rc;
5342 }
5343 
5344 static int ibmvnic_remove(struct vio_dev *dev)
5345 {
5346 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
5347 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5348 	unsigned long flags;
5349 
5350 	spin_lock_irqsave(&adapter->state_lock, flags);
5351 	if (adapter->state == VNIC_RESETTING) {
5352 		spin_unlock_irqrestore(&adapter->state_lock, flags);
5353 		return -EBUSY;
5354 	}
5355 
5356 	adapter->state = VNIC_REMOVING;
5357 	spin_unlock_irqrestore(&adapter->state_lock, flags);
5358 
5359 	flush_work(&adapter->ibmvnic_reset);
5360 	flush_delayed_work(&adapter->ibmvnic_delayed_reset);
5361 
5362 	rtnl_lock();
5363 	unregister_netdevice(netdev);
5364 
5365 	release_resources(adapter);
5366 	release_sub_crqs(adapter, 1);
5367 	release_crq_queue(adapter);
5368 
5369 	release_stats_token(adapter);
5370 	release_stats_buffers(adapter);
5371 
5372 	adapter->state = VNIC_REMOVED;
5373 
5374 	rtnl_unlock();
5375 	mutex_destroy(&adapter->fw_lock);
5376 	device_remove_file(&dev->dev, &dev_attr_failover);
5377 	free_netdev(netdev);
5378 	dev_set_drvdata(&dev->dev, NULL);
5379 
5380 	return 0;
5381 }
5382 
5383 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
5384 			      const char *buf, size_t count)
5385 {
5386 	struct net_device *netdev = dev_get_drvdata(dev);
5387 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5388 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
5389 	__be64 session_token;
5390 	long rc;
5391 
5392 	if (!sysfs_streq(buf, "1"))
5393 		return -EINVAL;
5394 
5395 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
5396 			 H_GET_SESSION_TOKEN, 0, 0, 0);
5397 	if (rc) {
5398 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
5399 			   rc);
5400 		return -EINVAL;
5401 	}
5402 
5403 	session_token = (__be64)retbuf[0];
5404 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
5405 		   be64_to_cpu(session_token));
5406 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
5407 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
5408 	if (rc) {
5409 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
5410 			   rc);
5411 		return -EINVAL;
5412 	}
5413 
5414 	return count;
5415 }
5416 
5417 static DEVICE_ATTR_WO(failover);
5418 
5419 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
5420 {
5421 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
5422 	struct ibmvnic_adapter *adapter;
5423 	struct iommu_table *tbl;
5424 	unsigned long ret = 0;
5425 	int i;
5426 
5427 	tbl = get_iommu_table_base(&vdev->dev);
5428 
5429 	/* netdev inits at probe time along with the structures we need below*/
5430 	if (!netdev)
5431 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
5432 
5433 	adapter = netdev_priv(netdev);
5434 
5435 	ret += PAGE_SIZE; /* the crq message queue */
5436 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
5437 
5438 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
5439 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
5440 
5441 	for (i = 0; i < adapter->num_active_rx_pools; i++)
5442 		ret += adapter->rx_pool[i].size *
5443 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
5444 
5445 	return ret;
5446 }
5447 
5448 static int ibmvnic_resume(struct device *dev)
5449 {
5450 	struct net_device *netdev = dev_get_drvdata(dev);
5451 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5452 
5453 	if (adapter->state != VNIC_OPEN)
5454 		return 0;
5455 
5456 	tasklet_schedule(&adapter->tasklet);
5457 
5458 	return 0;
5459 }
5460 
5461 static const struct vio_device_id ibmvnic_device_table[] = {
5462 	{"network", "IBM,vnic"},
5463 	{"", "" }
5464 };
5465 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5466 
5467 static const struct dev_pm_ops ibmvnic_pm_ops = {
5468 	.resume = ibmvnic_resume
5469 };
5470 
5471 static struct vio_driver ibmvnic_driver = {
5472 	.id_table       = ibmvnic_device_table,
5473 	.probe          = ibmvnic_probe,
5474 	.remove         = ibmvnic_remove,
5475 	.get_desired_dma = ibmvnic_get_desired_dma,
5476 	.name		= ibmvnic_driver_name,
5477 	.pm		= &ibmvnic_pm_ops,
5478 };
5479 
5480 /* module functions */
5481 static int __init ibmvnic_module_init(void)
5482 {
5483 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5484 		IBMVNIC_DRIVER_VERSION);
5485 
5486 	return vio_register_driver(&ibmvnic_driver);
5487 }
5488 
5489 static void __exit ibmvnic_module_exit(void)
5490 {
5491 	vio_unregister_driver(&ibmvnic_driver);
5492 }
5493 
5494 module_init(ibmvnic_module_init);
5495 module_exit(ibmvnic_module_exit);
5496