xref: /linux/drivers/net/ethernet/ibm/ibmvnic.c (revision 7f356166aebb0d956d367dfe55e19d7783277d09)
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(20000);
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 			return rc;
873 
874 		if (!wait_for_completion_timeout(&adapter->init_done,
875 						 timeout)) {
876 			netdev_warn(netdev, "Login timed out, retrying...\n");
877 			retry = true;
878 			adapter->init_done_rc = 0;
879 			retry_count++;
880 			continue;
881 		}
882 
883 		if (adapter->init_done_rc == ABORTED) {
884 			netdev_warn(netdev, "Login aborted, retrying...\n");
885 			retry = true;
886 			adapter->init_done_rc = 0;
887 			retry_count++;
888 			/* FW or device may be busy, so
889 			 * wait a bit before retrying login
890 			 */
891 			msleep(500);
892 		} else if (adapter->init_done_rc == PARTIALSUCCESS) {
893 			retry_count++;
894 			release_sub_crqs(adapter, 1);
895 
896 			retry = true;
897 			netdev_dbg(netdev,
898 				   "Received partial success, retrying...\n");
899 			adapter->init_done_rc = 0;
900 			reinit_completion(&adapter->init_done);
901 			send_query_cap(adapter);
902 			if (!wait_for_completion_timeout(&adapter->init_done,
903 							 timeout)) {
904 				netdev_warn(netdev,
905 					    "Capabilities query timed out\n");
906 				return -1;
907 			}
908 
909 			rc = init_sub_crqs(adapter);
910 			if (rc) {
911 				netdev_warn(netdev,
912 					    "SCRQ initialization failed\n");
913 				return -1;
914 			}
915 
916 			rc = init_sub_crq_irqs(adapter);
917 			if (rc) {
918 				netdev_warn(netdev,
919 					    "SCRQ irq initialization failed\n");
920 				return -1;
921 			}
922 		} else if (adapter->init_done_rc) {
923 			netdev_warn(netdev, "Adapter login failed\n");
924 			return -1;
925 		}
926 	} while (retry);
927 
928 	__ibmvnic_set_mac(netdev, adapter->mac_addr);
929 
930 	return 0;
931 }
932 
933 static void release_login_buffer(struct ibmvnic_adapter *adapter)
934 {
935 	kfree(adapter->login_buf);
936 	adapter->login_buf = NULL;
937 }
938 
939 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
940 {
941 	kfree(adapter->login_rsp_buf);
942 	adapter->login_rsp_buf = NULL;
943 }
944 
945 static void release_resources(struct ibmvnic_adapter *adapter)
946 {
947 	release_vpd_data(adapter);
948 
949 	release_tx_pools(adapter);
950 	release_rx_pools(adapter);
951 
952 	release_napi(adapter);
953 	release_login_rsp_buffer(adapter);
954 }
955 
956 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
957 {
958 	struct net_device *netdev = adapter->netdev;
959 	unsigned long timeout = msecs_to_jiffies(20000);
960 	union ibmvnic_crq crq;
961 	bool resend;
962 	int rc;
963 
964 	netdev_dbg(netdev, "setting link state %d\n", link_state);
965 
966 	memset(&crq, 0, sizeof(crq));
967 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
968 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
969 	crq.logical_link_state.link_state = link_state;
970 
971 	do {
972 		resend = false;
973 
974 		reinit_completion(&adapter->init_done);
975 		rc = ibmvnic_send_crq(adapter, &crq);
976 		if (rc) {
977 			netdev_err(netdev, "Failed to set link state\n");
978 			return rc;
979 		}
980 
981 		if (!wait_for_completion_timeout(&adapter->init_done,
982 						 timeout)) {
983 			netdev_err(netdev, "timeout setting link state\n");
984 			return -1;
985 		}
986 
987 		if (adapter->init_done_rc == PARTIALSUCCESS) {
988 			/* Partuial success, delay and re-send */
989 			mdelay(1000);
990 			resend = true;
991 		} else if (adapter->init_done_rc) {
992 			netdev_warn(netdev, "Unable to set link state, rc=%d\n",
993 				    adapter->init_done_rc);
994 			return adapter->init_done_rc;
995 		}
996 	} while (resend);
997 
998 	return 0;
999 }
1000 
1001 static int set_real_num_queues(struct net_device *netdev)
1002 {
1003 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1004 	int rc;
1005 
1006 	netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
1007 		   adapter->req_tx_queues, adapter->req_rx_queues);
1008 
1009 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
1010 	if (rc) {
1011 		netdev_err(netdev, "failed to set the number of tx queues\n");
1012 		return rc;
1013 	}
1014 
1015 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
1016 	if (rc)
1017 		netdev_err(netdev, "failed to set the number of rx queues\n");
1018 
1019 	return rc;
1020 }
1021 
1022 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
1023 {
1024 	struct device *dev = &adapter->vdev->dev;
1025 	union ibmvnic_crq crq;
1026 	int len = 0;
1027 	int rc;
1028 
1029 	if (adapter->vpd->buff)
1030 		len = adapter->vpd->len;
1031 
1032 	mutex_lock(&adapter->fw_lock);
1033 	adapter->fw_done_rc = 0;
1034 	reinit_completion(&adapter->fw_done);
1035 
1036 	crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
1037 	crq.get_vpd_size.cmd = GET_VPD_SIZE;
1038 	rc = ibmvnic_send_crq(adapter, &crq);
1039 	if (rc) {
1040 		mutex_unlock(&adapter->fw_lock);
1041 		return rc;
1042 	}
1043 
1044 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1045 	if (rc) {
1046 		dev_err(dev, "Could not retrieve VPD size, rc = %d\n", rc);
1047 		mutex_unlock(&adapter->fw_lock);
1048 		return rc;
1049 	}
1050 	mutex_unlock(&adapter->fw_lock);
1051 
1052 	if (!adapter->vpd->len)
1053 		return -ENODATA;
1054 
1055 	if (!adapter->vpd->buff)
1056 		adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
1057 	else if (adapter->vpd->len != len)
1058 		adapter->vpd->buff =
1059 			krealloc(adapter->vpd->buff,
1060 				 adapter->vpd->len, GFP_KERNEL);
1061 
1062 	if (!adapter->vpd->buff) {
1063 		dev_err(dev, "Could allocate VPD buffer\n");
1064 		return -ENOMEM;
1065 	}
1066 
1067 	adapter->vpd->dma_addr =
1068 		dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
1069 			       DMA_FROM_DEVICE);
1070 	if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
1071 		dev_err(dev, "Could not map VPD buffer\n");
1072 		kfree(adapter->vpd->buff);
1073 		adapter->vpd->buff = NULL;
1074 		return -ENOMEM;
1075 	}
1076 
1077 	mutex_lock(&adapter->fw_lock);
1078 	adapter->fw_done_rc = 0;
1079 	reinit_completion(&adapter->fw_done);
1080 
1081 	crq.get_vpd.first = IBMVNIC_CRQ_CMD;
1082 	crq.get_vpd.cmd = GET_VPD;
1083 	crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
1084 	crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1085 	rc = ibmvnic_send_crq(adapter, &crq);
1086 	if (rc) {
1087 		kfree(adapter->vpd->buff);
1088 		adapter->vpd->buff = NULL;
1089 		mutex_unlock(&adapter->fw_lock);
1090 		return rc;
1091 	}
1092 
1093 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1094 	if (rc) {
1095 		dev_err(dev, "Unable to retrieve VPD, rc = %d\n", rc);
1096 		kfree(adapter->vpd->buff);
1097 		adapter->vpd->buff = NULL;
1098 		mutex_unlock(&adapter->fw_lock);
1099 		return rc;
1100 	}
1101 
1102 	mutex_unlock(&adapter->fw_lock);
1103 	return 0;
1104 }
1105 
1106 static int init_resources(struct ibmvnic_adapter *adapter)
1107 {
1108 	struct net_device *netdev = adapter->netdev;
1109 	int rc;
1110 
1111 	rc = set_real_num_queues(netdev);
1112 	if (rc)
1113 		return rc;
1114 
1115 	adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1116 	if (!adapter->vpd)
1117 		return -ENOMEM;
1118 
1119 	/* Vital Product Data (VPD) */
1120 	rc = ibmvnic_get_vpd(adapter);
1121 	if (rc) {
1122 		netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1123 		return rc;
1124 	}
1125 
1126 	adapter->map_id = 1;
1127 
1128 	rc = init_napi(adapter);
1129 	if (rc)
1130 		return rc;
1131 
1132 	send_query_map(adapter);
1133 
1134 	rc = init_rx_pools(netdev);
1135 	if (rc)
1136 		return rc;
1137 
1138 	rc = init_tx_pools(netdev);
1139 	return rc;
1140 }
1141 
1142 static int __ibmvnic_open(struct net_device *netdev)
1143 {
1144 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1145 	enum vnic_state prev_state = adapter->state;
1146 	int i, rc;
1147 
1148 	adapter->state = VNIC_OPENING;
1149 	replenish_pools(adapter);
1150 	ibmvnic_napi_enable(adapter);
1151 
1152 	/* We're ready to receive frames, enable the sub-crq interrupts and
1153 	 * set the logical link state to up
1154 	 */
1155 	for (i = 0; i < adapter->req_rx_queues; i++) {
1156 		netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1157 		if (prev_state == VNIC_CLOSED)
1158 			enable_irq(adapter->rx_scrq[i]->irq);
1159 		enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1160 	}
1161 
1162 	for (i = 0; i < adapter->req_tx_queues; i++) {
1163 		netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1164 		if (prev_state == VNIC_CLOSED)
1165 			enable_irq(adapter->tx_scrq[i]->irq);
1166 		enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1167 		netdev_tx_reset_queue(netdev_get_tx_queue(netdev, i));
1168 	}
1169 
1170 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1171 	if (rc) {
1172 		for (i = 0; i < adapter->req_rx_queues; i++)
1173 			napi_disable(&adapter->napi[i]);
1174 		release_resources(adapter);
1175 		return rc;
1176 	}
1177 
1178 	netif_tx_start_all_queues(netdev);
1179 
1180 	if (prev_state == VNIC_CLOSED) {
1181 		for (i = 0; i < adapter->req_rx_queues; i++)
1182 			napi_schedule(&adapter->napi[i]);
1183 	}
1184 
1185 	adapter->state = VNIC_OPEN;
1186 	return rc;
1187 }
1188 
1189 static int ibmvnic_open(struct net_device *netdev)
1190 {
1191 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1192 	int rc;
1193 
1194 	/* If device failover is pending, just set device state and return.
1195 	 * Device operation will be handled by reset routine.
1196 	 */
1197 	if (adapter->failover_pending) {
1198 		adapter->state = VNIC_OPEN;
1199 		return 0;
1200 	}
1201 
1202 	if (adapter->state != VNIC_CLOSED) {
1203 		rc = ibmvnic_login(netdev);
1204 		if (rc)
1205 			goto out;
1206 
1207 		rc = init_resources(adapter);
1208 		if (rc) {
1209 			netdev_err(netdev, "failed to initialize resources\n");
1210 			release_resources(adapter);
1211 			goto out;
1212 		}
1213 	}
1214 
1215 	rc = __ibmvnic_open(netdev);
1216 
1217 out:
1218 	/*
1219 	 * If open fails due to a pending failover, set device state and
1220 	 * return. Device operation will be handled by reset routine.
1221 	 */
1222 	if (rc && adapter->failover_pending) {
1223 		adapter->state = VNIC_OPEN;
1224 		rc = 0;
1225 	}
1226 	return rc;
1227 }
1228 
1229 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1230 {
1231 	struct ibmvnic_rx_pool *rx_pool;
1232 	struct ibmvnic_rx_buff *rx_buff;
1233 	u64 rx_entries;
1234 	int rx_scrqs;
1235 	int i, j;
1236 
1237 	if (!adapter->rx_pool)
1238 		return;
1239 
1240 	rx_scrqs = adapter->num_active_rx_pools;
1241 	rx_entries = adapter->req_rx_add_entries_per_subcrq;
1242 
1243 	/* Free any remaining skbs in the rx buffer pools */
1244 	for (i = 0; i < rx_scrqs; i++) {
1245 		rx_pool = &adapter->rx_pool[i];
1246 		if (!rx_pool || !rx_pool->rx_buff)
1247 			continue;
1248 
1249 		netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1250 		for (j = 0; j < rx_entries; j++) {
1251 			rx_buff = &rx_pool->rx_buff[j];
1252 			if (rx_buff && rx_buff->skb) {
1253 				dev_kfree_skb_any(rx_buff->skb);
1254 				rx_buff->skb = NULL;
1255 			}
1256 		}
1257 	}
1258 }
1259 
1260 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1261 			      struct ibmvnic_tx_pool *tx_pool)
1262 {
1263 	struct ibmvnic_tx_buff *tx_buff;
1264 	u64 tx_entries;
1265 	int i;
1266 
1267 	if (!tx_pool || !tx_pool->tx_buff)
1268 		return;
1269 
1270 	tx_entries = tx_pool->num_buffers;
1271 
1272 	for (i = 0; i < tx_entries; i++) {
1273 		tx_buff = &tx_pool->tx_buff[i];
1274 		if (tx_buff && tx_buff->skb) {
1275 			dev_kfree_skb_any(tx_buff->skb);
1276 			tx_buff->skb = NULL;
1277 		}
1278 	}
1279 }
1280 
1281 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1282 {
1283 	int tx_scrqs;
1284 	int i;
1285 
1286 	if (!adapter->tx_pool || !adapter->tso_pool)
1287 		return;
1288 
1289 	tx_scrqs = adapter->num_active_tx_pools;
1290 
1291 	/* Free any remaining skbs in the tx buffer pools */
1292 	for (i = 0; i < tx_scrqs; i++) {
1293 		netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1294 		clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1295 		clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1296 	}
1297 }
1298 
1299 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1300 {
1301 	struct net_device *netdev = adapter->netdev;
1302 	int i;
1303 
1304 	if (adapter->tx_scrq) {
1305 		for (i = 0; i < adapter->req_tx_queues; i++)
1306 			if (adapter->tx_scrq[i]->irq) {
1307 				netdev_dbg(netdev,
1308 					   "Disabling tx_scrq[%d] irq\n", i);
1309 				disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1310 				disable_irq(adapter->tx_scrq[i]->irq);
1311 			}
1312 	}
1313 
1314 	if (adapter->rx_scrq) {
1315 		for (i = 0; i < adapter->req_rx_queues; i++) {
1316 			if (adapter->rx_scrq[i]->irq) {
1317 				netdev_dbg(netdev,
1318 					   "Disabling rx_scrq[%d] irq\n", i);
1319 				disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1320 				disable_irq(adapter->rx_scrq[i]->irq);
1321 			}
1322 		}
1323 	}
1324 }
1325 
1326 static void ibmvnic_cleanup(struct net_device *netdev)
1327 {
1328 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1329 
1330 	/* ensure that transmissions are stopped if called by do_reset */
1331 	if (test_bit(0, &adapter->resetting))
1332 		netif_tx_disable(netdev);
1333 	else
1334 		netif_tx_stop_all_queues(netdev);
1335 
1336 	ibmvnic_napi_disable(adapter);
1337 	ibmvnic_disable_irqs(adapter);
1338 
1339 	clean_rx_pools(adapter);
1340 	clean_tx_pools(adapter);
1341 }
1342 
1343 static int __ibmvnic_close(struct net_device *netdev)
1344 {
1345 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1346 	int rc = 0;
1347 
1348 	adapter->state = VNIC_CLOSING;
1349 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1350 	if (rc)
1351 		return rc;
1352 	adapter->state = VNIC_CLOSED;
1353 	return 0;
1354 }
1355 
1356 static int ibmvnic_close(struct net_device *netdev)
1357 {
1358 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1359 	int rc;
1360 
1361 	/* If device failover is pending, just set device state and return.
1362 	 * Device operation will be handled by reset routine.
1363 	 */
1364 	if (adapter->failover_pending) {
1365 		adapter->state = VNIC_CLOSED;
1366 		return 0;
1367 	}
1368 
1369 	rc = __ibmvnic_close(netdev);
1370 	ibmvnic_cleanup(netdev);
1371 
1372 	return rc;
1373 }
1374 
1375 /**
1376  * build_hdr_data - creates L2/L3/L4 header data buffer
1377  * @hdr_field - bitfield determining needed headers
1378  * @skb - socket buffer
1379  * @hdr_len - array of header lengths
1380  * @tot_len - total length of data
1381  *
1382  * Reads hdr_field to determine which headers are needed by firmware.
1383  * Builds a buffer containing these headers.  Saves individual header
1384  * lengths and total buffer length to be used to build descriptors.
1385  */
1386 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1387 			  int *hdr_len, u8 *hdr_data)
1388 {
1389 	int len = 0;
1390 	u8 *hdr;
1391 
1392 	if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1393 		hdr_len[0] = sizeof(struct vlan_ethhdr);
1394 	else
1395 		hdr_len[0] = sizeof(struct ethhdr);
1396 
1397 	if (skb->protocol == htons(ETH_P_IP)) {
1398 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
1399 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1400 			hdr_len[2] = tcp_hdrlen(skb);
1401 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1402 			hdr_len[2] = sizeof(struct udphdr);
1403 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1404 		hdr_len[1] = sizeof(struct ipv6hdr);
1405 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1406 			hdr_len[2] = tcp_hdrlen(skb);
1407 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1408 			hdr_len[2] = sizeof(struct udphdr);
1409 	} else if (skb->protocol == htons(ETH_P_ARP)) {
1410 		hdr_len[1] = arp_hdr_len(skb->dev);
1411 		hdr_len[2] = 0;
1412 	}
1413 
1414 	memset(hdr_data, 0, 120);
1415 	if ((hdr_field >> 6) & 1) {
1416 		hdr = skb_mac_header(skb);
1417 		memcpy(hdr_data, hdr, hdr_len[0]);
1418 		len += hdr_len[0];
1419 	}
1420 
1421 	if ((hdr_field >> 5) & 1) {
1422 		hdr = skb_network_header(skb);
1423 		memcpy(hdr_data + len, hdr, hdr_len[1]);
1424 		len += hdr_len[1];
1425 	}
1426 
1427 	if ((hdr_field >> 4) & 1) {
1428 		hdr = skb_transport_header(skb);
1429 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1430 		len += hdr_len[2];
1431 	}
1432 	return len;
1433 }
1434 
1435 /**
1436  * create_hdr_descs - create header and header extension descriptors
1437  * @hdr_field - bitfield determining needed headers
1438  * @data - buffer containing header data
1439  * @len - length of data buffer
1440  * @hdr_len - array of individual header lengths
1441  * @scrq_arr - descriptor array
1442  *
1443  * Creates header and, if needed, header extension descriptors and
1444  * places them in a descriptor array, scrq_arr
1445  */
1446 
1447 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1448 			    union sub_crq *scrq_arr)
1449 {
1450 	union sub_crq hdr_desc;
1451 	int tmp_len = len;
1452 	int num_descs = 0;
1453 	u8 *data, *cur;
1454 	int tmp;
1455 
1456 	while (tmp_len > 0) {
1457 		cur = hdr_data + len - tmp_len;
1458 
1459 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1460 		if (cur != hdr_data) {
1461 			data = hdr_desc.hdr_ext.data;
1462 			tmp = tmp_len > 29 ? 29 : tmp_len;
1463 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1464 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1465 			hdr_desc.hdr_ext.len = tmp;
1466 		} else {
1467 			data = hdr_desc.hdr.data;
1468 			tmp = tmp_len > 24 ? 24 : tmp_len;
1469 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1470 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1471 			hdr_desc.hdr.len = tmp;
1472 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1473 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1474 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1475 			hdr_desc.hdr.flag = hdr_field << 1;
1476 		}
1477 		memcpy(data, cur, tmp);
1478 		tmp_len -= tmp;
1479 		*scrq_arr = hdr_desc;
1480 		scrq_arr++;
1481 		num_descs++;
1482 	}
1483 
1484 	return num_descs;
1485 }
1486 
1487 /**
1488  * build_hdr_descs_arr - build a header descriptor array
1489  * @skb - socket buffer
1490  * @num_entries - number of descriptors to be sent
1491  * @subcrq - first TX descriptor
1492  * @hdr_field - bit field determining which headers will be sent
1493  *
1494  * This function will build a TX descriptor array with applicable
1495  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1496  */
1497 
1498 static void build_hdr_descs_arr(struct sk_buff *skb,
1499 				union sub_crq *indir_arr,
1500 				int *num_entries, u8 hdr_field)
1501 {
1502 	int hdr_len[3] = {0, 0, 0};
1503 	u8 hdr_data[140] = {0};
1504 	int tot_len;
1505 
1506 	tot_len = build_hdr_data(hdr_field, skb, hdr_len,
1507 				 hdr_data);
1508 	*num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1509 					 indir_arr + 1);
1510 }
1511 
1512 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1513 				    struct net_device *netdev)
1514 {
1515 	/* For some backing devices, mishandling of small packets
1516 	 * can result in a loss of connection or TX stall. Device
1517 	 * architects recommend that no packet should be smaller
1518 	 * than the minimum MTU value provided to the driver, so
1519 	 * pad any packets to that length
1520 	 */
1521 	if (skb->len < netdev->min_mtu)
1522 		return skb_put_padto(skb, netdev->min_mtu);
1523 
1524 	return 0;
1525 }
1526 
1527 static void ibmvnic_tx_scrq_clean_buffer(struct ibmvnic_adapter *adapter,
1528 					 struct ibmvnic_sub_crq_queue *tx_scrq)
1529 {
1530 	struct ibmvnic_ind_xmit_queue *ind_bufp;
1531 	struct ibmvnic_tx_buff *tx_buff;
1532 	struct ibmvnic_tx_pool *tx_pool;
1533 	union sub_crq tx_scrq_entry;
1534 	int queue_num;
1535 	int entries;
1536 	int index;
1537 	int i;
1538 
1539 	ind_bufp = &tx_scrq->ind_buf;
1540 	entries = (u64)ind_bufp->index;
1541 	queue_num = tx_scrq->pool_index;
1542 
1543 	for (i = entries - 1; i >= 0; --i) {
1544 		tx_scrq_entry = ind_bufp->indir_arr[i];
1545 		if (tx_scrq_entry.v1.type != IBMVNIC_TX_DESC)
1546 			continue;
1547 		index = be32_to_cpu(tx_scrq_entry.v1.correlator);
1548 		if (index & IBMVNIC_TSO_POOL_MASK) {
1549 			tx_pool = &adapter->tso_pool[queue_num];
1550 			index &= ~IBMVNIC_TSO_POOL_MASK;
1551 		} else {
1552 			tx_pool = &adapter->tx_pool[queue_num];
1553 		}
1554 		tx_pool->free_map[tx_pool->consumer_index] = index;
1555 		tx_pool->consumer_index = tx_pool->consumer_index == 0 ?
1556 					  tx_pool->num_buffers - 1 :
1557 					  tx_pool->consumer_index - 1;
1558 		tx_buff = &tx_pool->tx_buff[index];
1559 		adapter->netdev->stats.tx_packets--;
1560 		adapter->netdev->stats.tx_bytes -= tx_buff->skb->len;
1561 		adapter->tx_stats_buffers[queue_num].packets--;
1562 		adapter->tx_stats_buffers[queue_num].bytes -=
1563 						tx_buff->skb->len;
1564 		dev_kfree_skb_any(tx_buff->skb);
1565 		tx_buff->skb = NULL;
1566 		adapter->netdev->stats.tx_dropped++;
1567 	}
1568 	ind_bufp->index = 0;
1569 	if (atomic_sub_return(entries, &tx_scrq->used) <=
1570 	    (adapter->req_tx_entries_per_subcrq / 2) &&
1571 	    __netif_subqueue_stopped(adapter->netdev, queue_num)) {
1572 		netif_wake_subqueue(adapter->netdev, queue_num);
1573 		netdev_dbg(adapter->netdev, "Started queue %d\n",
1574 			   queue_num);
1575 	}
1576 }
1577 
1578 static int ibmvnic_tx_scrq_flush(struct ibmvnic_adapter *adapter,
1579 				 struct ibmvnic_sub_crq_queue *tx_scrq)
1580 {
1581 	struct ibmvnic_ind_xmit_queue *ind_bufp;
1582 	u64 dma_addr;
1583 	u64 entries;
1584 	u64 handle;
1585 	int rc;
1586 
1587 	ind_bufp = &tx_scrq->ind_buf;
1588 	dma_addr = (u64)ind_bufp->indir_dma;
1589 	entries = (u64)ind_bufp->index;
1590 	handle = tx_scrq->handle;
1591 
1592 	if (!entries)
1593 		return 0;
1594 	rc = send_subcrq_indirect(adapter, handle, dma_addr, entries);
1595 	if (rc)
1596 		ibmvnic_tx_scrq_clean_buffer(adapter, tx_scrq);
1597 	else
1598 		ind_bufp->index = 0;
1599 	return 0;
1600 }
1601 
1602 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1603 {
1604 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1605 	int queue_num = skb_get_queue_mapping(skb);
1606 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1607 	struct device *dev = &adapter->vdev->dev;
1608 	struct ibmvnic_ind_xmit_queue *ind_bufp;
1609 	struct ibmvnic_tx_buff *tx_buff = NULL;
1610 	struct ibmvnic_sub_crq_queue *tx_scrq;
1611 	struct ibmvnic_tx_pool *tx_pool;
1612 	unsigned int tx_send_failed = 0;
1613 	netdev_tx_t ret = NETDEV_TX_OK;
1614 	unsigned int tx_map_failed = 0;
1615 	union sub_crq indir_arr[16];
1616 	unsigned int tx_dropped = 0;
1617 	unsigned int tx_packets = 0;
1618 	unsigned int tx_bytes = 0;
1619 	dma_addr_t data_dma_addr;
1620 	struct netdev_queue *txq;
1621 	unsigned long lpar_rc;
1622 	union sub_crq tx_crq;
1623 	unsigned int offset;
1624 	int num_entries = 1;
1625 	unsigned char *dst;
1626 	int index = 0;
1627 	u8 proto = 0;
1628 
1629 	tx_scrq = adapter->tx_scrq[queue_num];
1630 	txq = netdev_get_tx_queue(netdev, queue_num);
1631 	ind_bufp = &tx_scrq->ind_buf;
1632 
1633 	if (test_bit(0, &adapter->resetting)) {
1634 		if (!netif_subqueue_stopped(netdev, skb))
1635 			netif_stop_subqueue(netdev, queue_num);
1636 		dev_kfree_skb_any(skb);
1637 
1638 		tx_send_failed++;
1639 		tx_dropped++;
1640 		ret = NETDEV_TX_OK;
1641 		ibmvnic_tx_scrq_flush(adapter, tx_scrq);
1642 		goto out;
1643 	}
1644 
1645 	if (ibmvnic_xmit_workarounds(skb, netdev)) {
1646 		tx_dropped++;
1647 		tx_send_failed++;
1648 		ret = NETDEV_TX_OK;
1649 		ibmvnic_tx_scrq_flush(adapter, tx_scrq);
1650 		goto out;
1651 	}
1652 	if (skb_is_gso(skb))
1653 		tx_pool = &adapter->tso_pool[queue_num];
1654 	else
1655 		tx_pool = &adapter->tx_pool[queue_num];
1656 
1657 	index = tx_pool->free_map[tx_pool->consumer_index];
1658 
1659 	if (index == IBMVNIC_INVALID_MAP) {
1660 		dev_kfree_skb_any(skb);
1661 		tx_send_failed++;
1662 		tx_dropped++;
1663 		ret = NETDEV_TX_OK;
1664 		ibmvnic_tx_scrq_flush(adapter, tx_scrq);
1665 		goto out;
1666 	}
1667 
1668 	tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1669 
1670 	offset = index * tx_pool->buf_size;
1671 	dst = tx_pool->long_term_buff.buff + offset;
1672 	memset(dst, 0, tx_pool->buf_size);
1673 	data_dma_addr = tx_pool->long_term_buff.addr + offset;
1674 
1675 	if (skb_shinfo(skb)->nr_frags) {
1676 		int cur, i;
1677 
1678 		/* Copy the head */
1679 		skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1680 		cur = skb_headlen(skb);
1681 
1682 		/* Copy the frags */
1683 		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1684 			const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1685 
1686 			memcpy(dst + cur,
1687 			       page_address(skb_frag_page(frag)) +
1688 			       skb_frag_off(frag), skb_frag_size(frag));
1689 			cur += skb_frag_size(frag);
1690 		}
1691 	} else {
1692 		skb_copy_from_linear_data(skb, dst, skb->len);
1693 	}
1694 
1695 	tx_pool->consumer_index =
1696 	    (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1697 
1698 	tx_buff = &tx_pool->tx_buff[index];
1699 	tx_buff->skb = skb;
1700 	tx_buff->index = index;
1701 	tx_buff->pool_index = queue_num;
1702 
1703 	memset(&tx_crq, 0, sizeof(tx_crq));
1704 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1705 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1706 	tx_crq.v1.n_crq_elem = 1;
1707 	tx_crq.v1.n_sge = 1;
1708 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1709 
1710 	if (skb_is_gso(skb))
1711 		tx_crq.v1.correlator =
1712 			cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1713 	else
1714 		tx_crq.v1.correlator = cpu_to_be32(index);
1715 	tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1716 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1717 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1718 
1719 	if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1720 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1721 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1722 	}
1723 
1724 	if (skb->protocol == htons(ETH_P_IP)) {
1725 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1726 		proto = ip_hdr(skb)->protocol;
1727 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1728 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1729 		proto = ipv6_hdr(skb)->nexthdr;
1730 	}
1731 
1732 	if (proto == IPPROTO_TCP)
1733 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1734 	else if (proto == IPPROTO_UDP)
1735 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1736 
1737 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1738 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1739 		hdrs += 2;
1740 	}
1741 	if (skb_is_gso(skb)) {
1742 		tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1743 		tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1744 		hdrs += 2;
1745 	}
1746 
1747 	if ((*hdrs >> 7) & 1)
1748 		build_hdr_descs_arr(skb, indir_arr, &num_entries, *hdrs);
1749 
1750 	tx_crq.v1.n_crq_elem = num_entries;
1751 	tx_buff->num_entries = num_entries;
1752 	/* flush buffer if current entry can not fit */
1753 	if (num_entries + ind_bufp->index > IBMVNIC_MAX_IND_DESCS) {
1754 		lpar_rc = ibmvnic_tx_scrq_flush(adapter, tx_scrq);
1755 		if (lpar_rc != H_SUCCESS)
1756 			goto tx_flush_err;
1757 	}
1758 
1759 	indir_arr[0] = tx_crq;
1760 	memcpy(&ind_bufp->indir_arr[ind_bufp->index], &indir_arr[0],
1761 	       num_entries * sizeof(struct ibmvnic_generic_scrq));
1762 	ind_bufp->index += num_entries;
1763 	if (__netdev_tx_sent_queue(txq, skb->len,
1764 				   netdev_xmit_more() &&
1765 				   ind_bufp->index < IBMVNIC_MAX_IND_DESCS)) {
1766 		lpar_rc = ibmvnic_tx_scrq_flush(adapter, tx_scrq);
1767 		if (lpar_rc != H_SUCCESS)
1768 			goto tx_err;
1769 	}
1770 
1771 	if (atomic_add_return(num_entries, &tx_scrq->used)
1772 					>= adapter->req_tx_entries_per_subcrq) {
1773 		netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1774 		netif_stop_subqueue(netdev, queue_num);
1775 	}
1776 
1777 	tx_packets++;
1778 	tx_bytes += skb->len;
1779 	txq->trans_start = jiffies;
1780 	ret = NETDEV_TX_OK;
1781 	goto out;
1782 
1783 tx_flush_err:
1784 	dev_kfree_skb_any(skb);
1785 	tx_buff->skb = NULL;
1786 	tx_pool->consumer_index = tx_pool->consumer_index == 0 ?
1787 				  tx_pool->num_buffers - 1 :
1788 				  tx_pool->consumer_index - 1;
1789 	tx_dropped++;
1790 tx_err:
1791 	if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1792 		dev_err_ratelimited(dev, "tx: send failed\n");
1793 
1794 	if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1795 		/* Disable TX and report carrier off if queue is closed
1796 		 * or pending failover.
1797 		 * Firmware guarantees that a signal will be sent to the
1798 		 * driver, triggering a reset or some other action.
1799 		 */
1800 		netif_tx_stop_all_queues(netdev);
1801 		netif_carrier_off(netdev);
1802 	}
1803 out:
1804 	netdev->stats.tx_dropped += tx_dropped;
1805 	netdev->stats.tx_bytes += tx_bytes;
1806 	netdev->stats.tx_packets += tx_packets;
1807 	adapter->tx_send_failed += tx_send_failed;
1808 	adapter->tx_map_failed += tx_map_failed;
1809 	adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1810 	adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1811 	adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1812 
1813 	return ret;
1814 }
1815 
1816 static void ibmvnic_set_multi(struct net_device *netdev)
1817 {
1818 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1819 	struct netdev_hw_addr *ha;
1820 	union ibmvnic_crq crq;
1821 
1822 	memset(&crq, 0, sizeof(crq));
1823 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1824 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1825 
1826 	if (netdev->flags & IFF_PROMISC) {
1827 		if (!adapter->promisc_supported)
1828 			return;
1829 	} else {
1830 		if (netdev->flags & IFF_ALLMULTI) {
1831 			/* Accept all multicast */
1832 			memset(&crq, 0, sizeof(crq));
1833 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1834 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1835 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1836 			ibmvnic_send_crq(adapter, &crq);
1837 		} else if (netdev_mc_empty(netdev)) {
1838 			/* Reject all multicast */
1839 			memset(&crq, 0, sizeof(crq));
1840 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1841 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1842 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1843 			ibmvnic_send_crq(adapter, &crq);
1844 		} else {
1845 			/* Accept one or more multicast(s) */
1846 			netdev_for_each_mc_addr(ha, netdev) {
1847 				memset(&crq, 0, sizeof(crq));
1848 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1849 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1850 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1851 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1852 						ha->addr);
1853 				ibmvnic_send_crq(adapter, &crq);
1854 			}
1855 		}
1856 	}
1857 }
1858 
1859 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1860 {
1861 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1862 	union ibmvnic_crq crq;
1863 	int rc;
1864 
1865 	if (!is_valid_ether_addr(dev_addr)) {
1866 		rc = -EADDRNOTAVAIL;
1867 		goto err;
1868 	}
1869 
1870 	memset(&crq, 0, sizeof(crq));
1871 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1872 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1873 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1874 
1875 	mutex_lock(&adapter->fw_lock);
1876 	adapter->fw_done_rc = 0;
1877 	reinit_completion(&adapter->fw_done);
1878 
1879 	rc = ibmvnic_send_crq(adapter, &crq);
1880 	if (rc) {
1881 		rc = -EIO;
1882 		mutex_unlock(&adapter->fw_lock);
1883 		goto err;
1884 	}
1885 
1886 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1887 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1888 	if (rc || adapter->fw_done_rc) {
1889 		rc = -EIO;
1890 		mutex_unlock(&adapter->fw_lock);
1891 		goto err;
1892 	}
1893 	mutex_unlock(&adapter->fw_lock);
1894 	return 0;
1895 err:
1896 	ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1897 	return rc;
1898 }
1899 
1900 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1901 {
1902 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1903 	struct sockaddr *addr = p;
1904 	int rc;
1905 
1906 	rc = 0;
1907 	if (!is_valid_ether_addr(addr->sa_data))
1908 		return -EADDRNOTAVAIL;
1909 
1910 	if (adapter->state != VNIC_PROBED) {
1911 		ether_addr_copy(adapter->mac_addr, addr->sa_data);
1912 		rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1913 	}
1914 
1915 	return rc;
1916 }
1917 
1918 /**
1919  * do_change_param_reset returns zero if we are able to keep processing reset
1920  * events, or non-zero if we hit a fatal error and must halt.
1921  */
1922 static int do_change_param_reset(struct ibmvnic_adapter *adapter,
1923 				 struct ibmvnic_rwi *rwi,
1924 				 u32 reset_state)
1925 {
1926 	struct net_device *netdev = adapter->netdev;
1927 	int i, rc;
1928 
1929 	netdev_dbg(adapter->netdev, "Change param resetting driver (%d)\n",
1930 		   rwi->reset_reason);
1931 
1932 	netif_carrier_off(netdev);
1933 	adapter->reset_reason = rwi->reset_reason;
1934 
1935 	ibmvnic_cleanup(netdev);
1936 
1937 	if (reset_state == VNIC_OPEN) {
1938 		rc = __ibmvnic_close(netdev);
1939 		if (rc)
1940 			goto out;
1941 	}
1942 
1943 	release_resources(adapter);
1944 	release_sub_crqs(adapter, 1);
1945 	release_crq_queue(adapter);
1946 
1947 	adapter->state = VNIC_PROBED;
1948 
1949 	rc = init_crq_queue(adapter);
1950 
1951 	if (rc) {
1952 		netdev_err(adapter->netdev,
1953 			   "Couldn't initialize crq. rc=%d\n", rc);
1954 		return rc;
1955 	}
1956 
1957 	rc = ibmvnic_reset_init(adapter, true);
1958 	if (rc) {
1959 		rc = IBMVNIC_INIT_FAILED;
1960 		goto out;
1961 	}
1962 
1963 	/* If the adapter was in PROBE state prior to the reset,
1964 	 * exit here.
1965 	 */
1966 	if (reset_state == VNIC_PROBED)
1967 		goto out;
1968 
1969 	rc = ibmvnic_login(netdev);
1970 	if (rc) {
1971 		goto out;
1972 	}
1973 
1974 	rc = init_resources(adapter);
1975 	if (rc)
1976 		goto out;
1977 
1978 	ibmvnic_disable_irqs(adapter);
1979 
1980 	adapter->state = VNIC_CLOSED;
1981 
1982 	if (reset_state == VNIC_CLOSED)
1983 		return 0;
1984 
1985 	rc = __ibmvnic_open(netdev);
1986 	if (rc) {
1987 		rc = IBMVNIC_OPEN_FAILED;
1988 		goto out;
1989 	}
1990 
1991 	/* refresh device's multicast list */
1992 	ibmvnic_set_multi(netdev);
1993 
1994 	/* kick napi */
1995 	for (i = 0; i < adapter->req_rx_queues; i++)
1996 		napi_schedule(&adapter->napi[i]);
1997 
1998 out:
1999 	if (rc)
2000 		adapter->state = reset_state;
2001 	return rc;
2002 }
2003 
2004 /**
2005  * do_reset returns zero if we are able to keep processing reset events, or
2006  * non-zero if we hit a fatal error and must halt.
2007  */
2008 static int do_reset(struct ibmvnic_adapter *adapter,
2009 		    struct ibmvnic_rwi *rwi, u32 reset_state)
2010 {
2011 	u64 old_num_rx_queues, old_num_tx_queues;
2012 	u64 old_num_rx_slots, old_num_tx_slots;
2013 	struct net_device *netdev = adapter->netdev;
2014 	int i, rc;
2015 
2016 	netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
2017 		   rwi->reset_reason);
2018 
2019 	rtnl_lock();
2020 	/*
2021 	 * Now that we have the rtnl lock, clear any pending failover.
2022 	 * This will ensure ibmvnic_open() has either completed or will
2023 	 * block until failover is complete.
2024 	 */
2025 	if (rwi->reset_reason == VNIC_RESET_FAILOVER)
2026 		adapter->failover_pending = false;
2027 
2028 	netif_carrier_off(netdev);
2029 	adapter->reset_reason = rwi->reset_reason;
2030 
2031 	old_num_rx_queues = adapter->req_rx_queues;
2032 	old_num_tx_queues = adapter->req_tx_queues;
2033 	old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
2034 	old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
2035 
2036 	ibmvnic_cleanup(netdev);
2037 
2038 	if (reset_state == VNIC_OPEN &&
2039 	    adapter->reset_reason != VNIC_RESET_MOBILITY &&
2040 	    adapter->reset_reason != VNIC_RESET_FAILOVER) {
2041 		adapter->state = VNIC_CLOSING;
2042 
2043 		/* Release the RTNL lock before link state change and
2044 		 * re-acquire after the link state change to allow
2045 		 * linkwatch_event to grab the RTNL lock and run during
2046 		 * a reset.
2047 		 */
2048 		rtnl_unlock();
2049 		rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
2050 		rtnl_lock();
2051 		if (rc)
2052 			goto out;
2053 
2054 		if (adapter->state != VNIC_CLOSING) {
2055 			rc = -1;
2056 			goto out;
2057 		}
2058 
2059 		adapter->state = VNIC_CLOSED;
2060 	}
2061 
2062 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
2063 		/* remove the closed state so when we call open it appears
2064 		 * we are coming from the probed state.
2065 		 */
2066 		adapter->state = VNIC_PROBED;
2067 
2068 		if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
2069 			rc = ibmvnic_reenable_crq_queue(adapter);
2070 			release_sub_crqs(adapter, 1);
2071 		} else {
2072 			rc = ibmvnic_reset_crq(adapter);
2073 			if (rc == H_CLOSED || rc == H_SUCCESS) {
2074 				rc = vio_enable_interrupts(adapter->vdev);
2075 				if (rc)
2076 					netdev_err(adapter->netdev,
2077 						   "Reset failed to enable interrupts. rc=%d\n",
2078 						   rc);
2079 			}
2080 		}
2081 
2082 		if (rc) {
2083 			netdev_err(adapter->netdev,
2084 				   "Reset couldn't initialize crq. rc=%d\n", rc);
2085 			goto out;
2086 		}
2087 
2088 		rc = ibmvnic_reset_init(adapter, true);
2089 		if (rc) {
2090 			rc = IBMVNIC_INIT_FAILED;
2091 			goto out;
2092 		}
2093 
2094 		/* If the adapter was in PROBE state prior to the reset,
2095 		 * exit here.
2096 		 */
2097 		if (reset_state == VNIC_PROBED) {
2098 			rc = 0;
2099 			goto out;
2100 		}
2101 
2102 		rc = ibmvnic_login(netdev);
2103 		if (rc) {
2104 			goto out;
2105 		}
2106 
2107 		if (adapter->req_rx_queues != old_num_rx_queues ||
2108 		    adapter->req_tx_queues != old_num_tx_queues ||
2109 		    adapter->req_rx_add_entries_per_subcrq !=
2110 		    old_num_rx_slots ||
2111 		    adapter->req_tx_entries_per_subcrq !=
2112 		    old_num_tx_slots ||
2113 		    !adapter->rx_pool ||
2114 		    !adapter->tso_pool ||
2115 		    !adapter->tx_pool) {
2116 			release_rx_pools(adapter);
2117 			release_tx_pools(adapter);
2118 			release_napi(adapter);
2119 			release_vpd_data(adapter);
2120 
2121 			rc = init_resources(adapter);
2122 			if (rc)
2123 				goto out;
2124 
2125 		} else {
2126 			rc = reset_tx_pools(adapter);
2127 			if (rc) {
2128 				netdev_dbg(adapter->netdev, "reset tx pools failed (%d)\n",
2129 						rc);
2130 				goto out;
2131 			}
2132 
2133 			rc = reset_rx_pools(adapter);
2134 			if (rc) {
2135 				netdev_dbg(adapter->netdev, "reset rx pools failed (%d)\n",
2136 						rc);
2137 				goto out;
2138 			}
2139 		}
2140 		ibmvnic_disable_irqs(adapter);
2141 	}
2142 	adapter->state = VNIC_CLOSED;
2143 
2144 	if (reset_state == VNIC_CLOSED) {
2145 		rc = 0;
2146 		goto out;
2147 	}
2148 
2149 	rc = __ibmvnic_open(netdev);
2150 	if (rc) {
2151 		rc = IBMVNIC_OPEN_FAILED;
2152 		goto out;
2153 	}
2154 
2155 	/* refresh device's multicast list */
2156 	ibmvnic_set_multi(netdev);
2157 
2158 	/* kick napi */
2159 	for (i = 0; i < adapter->req_rx_queues; i++)
2160 		napi_schedule(&adapter->napi[i]);
2161 
2162 	if (adapter->reset_reason == VNIC_RESET_FAILOVER ||
2163 	    adapter->reset_reason == VNIC_RESET_MOBILITY) {
2164 		call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
2165 		call_netdevice_notifiers(NETDEV_RESEND_IGMP, netdev);
2166 	}
2167 
2168 	rc = 0;
2169 
2170 out:
2171 	/* restore the adapter state if reset failed */
2172 	if (rc)
2173 		adapter->state = reset_state;
2174 	rtnl_unlock();
2175 
2176 	return rc;
2177 }
2178 
2179 static int do_hard_reset(struct ibmvnic_adapter *adapter,
2180 			 struct ibmvnic_rwi *rwi, u32 reset_state)
2181 {
2182 	struct net_device *netdev = adapter->netdev;
2183 	int rc;
2184 
2185 	netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
2186 		   rwi->reset_reason);
2187 
2188 	netif_carrier_off(netdev);
2189 	adapter->reset_reason = rwi->reset_reason;
2190 
2191 	ibmvnic_cleanup(netdev);
2192 	release_resources(adapter);
2193 	release_sub_crqs(adapter, 0);
2194 	release_crq_queue(adapter);
2195 
2196 	/* remove the closed state so when we call open it appears
2197 	 * we are coming from the probed state.
2198 	 */
2199 	adapter->state = VNIC_PROBED;
2200 
2201 	reinit_completion(&adapter->init_done);
2202 	rc = init_crq_queue(adapter);
2203 	if (rc) {
2204 		netdev_err(adapter->netdev,
2205 			   "Couldn't initialize crq. rc=%d\n", rc);
2206 		goto out;
2207 	}
2208 
2209 	rc = ibmvnic_reset_init(adapter, false);
2210 	if (rc)
2211 		goto out;
2212 
2213 	/* If the adapter was in PROBE state prior to the reset,
2214 	 * exit here.
2215 	 */
2216 	if (reset_state == VNIC_PROBED)
2217 		goto out;
2218 
2219 	rc = ibmvnic_login(netdev);
2220 	if (rc)
2221 		goto out;
2222 
2223 	rc = init_resources(adapter);
2224 	if (rc)
2225 		goto out;
2226 
2227 	ibmvnic_disable_irqs(adapter);
2228 	adapter->state = VNIC_CLOSED;
2229 
2230 	if (reset_state == VNIC_CLOSED)
2231 		goto out;
2232 
2233 	rc = __ibmvnic_open(netdev);
2234 	if (rc) {
2235 		rc = IBMVNIC_OPEN_FAILED;
2236 		goto out;
2237 	}
2238 
2239 	call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
2240 	call_netdevice_notifiers(NETDEV_RESEND_IGMP, netdev);
2241 out:
2242 	/* restore adapter state if reset failed */
2243 	if (rc)
2244 		adapter->state = reset_state;
2245 	return rc;
2246 }
2247 
2248 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
2249 {
2250 	struct ibmvnic_rwi *rwi;
2251 	unsigned long flags;
2252 
2253 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2254 
2255 	if (!list_empty(&adapter->rwi_list)) {
2256 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
2257 				       list);
2258 		list_del(&rwi->list);
2259 	} else {
2260 		rwi = NULL;
2261 	}
2262 
2263 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2264 	return rwi;
2265 }
2266 
2267 static void __ibmvnic_reset(struct work_struct *work)
2268 {
2269 	struct ibmvnic_rwi *rwi;
2270 	struct ibmvnic_adapter *adapter;
2271 	bool saved_state = false;
2272 	unsigned long flags;
2273 	u32 reset_state;
2274 	int rc = 0;
2275 
2276 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
2277 
2278 	if (test_and_set_bit_lock(0, &adapter->resetting)) {
2279 		schedule_delayed_work(&adapter->ibmvnic_delayed_reset,
2280 				      IBMVNIC_RESET_DELAY);
2281 		return;
2282 	}
2283 
2284 	rwi = get_next_rwi(adapter);
2285 	while (rwi) {
2286 		spin_lock_irqsave(&adapter->state_lock, flags);
2287 
2288 		if (adapter->state == VNIC_REMOVING ||
2289 		    adapter->state == VNIC_REMOVED) {
2290 			spin_unlock_irqrestore(&adapter->state_lock, flags);
2291 			kfree(rwi);
2292 			rc = EBUSY;
2293 			break;
2294 		}
2295 
2296 		if (!saved_state) {
2297 			reset_state = adapter->state;
2298 			saved_state = true;
2299 		}
2300 		spin_unlock_irqrestore(&adapter->state_lock, flags);
2301 
2302 		if (rwi->reset_reason == VNIC_RESET_CHANGE_PARAM) {
2303 			/* CHANGE_PARAM requestor holds rtnl_lock */
2304 			rc = do_change_param_reset(adapter, rwi, reset_state);
2305 		} else if (adapter->force_reset_recovery) {
2306 			/*
2307 			 * Since we are doing a hard reset now, clear the
2308 			 * failover_pending flag so we don't ignore any
2309 			 * future MOBILITY or other resets.
2310 			 */
2311 			adapter->failover_pending = false;
2312 
2313 			/* Transport event occurred during previous reset */
2314 			if (adapter->wait_for_reset) {
2315 				/* Previous was CHANGE_PARAM; caller locked */
2316 				adapter->force_reset_recovery = false;
2317 				rc = do_hard_reset(adapter, rwi, reset_state);
2318 			} else {
2319 				rtnl_lock();
2320 				adapter->force_reset_recovery = false;
2321 				rc = do_hard_reset(adapter, rwi, reset_state);
2322 				rtnl_unlock();
2323 			}
2324 			if (rc) {
2325 				/* give backing device time to settle down */
2326 				netdev_dbg(adapter->netdev,
2327 					   "[S:%d] Hard reset failed, waiting 60 secs\n",
2328 					   adapter->state);
2329 				set_current_state(TASK_UNINTERRUPTIBLE);
2330 				schedule_timeout(60 * HZ);
2331 			}
2332 		} else if (!(rwi->reset_reason == VNIC_RESET_FATAL &&
2333 				adapter->from_passive_init)) {
2334 			rc = do_reset(adapter, rwi, reset_state);
2335 		}
2336 		kfree(rwi);
2337 		adapter->last_reset_time = jiffies;
2338 
2339 		if (rc)
2340 			netdev_dbg(adapter->netdev, "Reset failed, rc=%d\n", rc);
2341 
2342 		rwi = get_next_rwi(adapter);
2343 
2344 		if (rwi && (rwi->reset_reason == VNIC_RESET_FAILOVER ||
2345 			    rwi->reset_reason == VNIC_RESET_MOBILITY))
2346 			adapter->force_reset_recovery = true;
2347 	}
2348 
2349 	if (adapter->wait_for_reset) {
2350 		adapter->reset_done_rc = rc;
2351 		complete(&adapter->reset_done);
2352 	}
2353 
2354 	clear_bit_unlock(0, &adapter->resetting);
2355 }
2356 
2357 static void __ibmvnic_delayed_reset(struct work_struct *work)
2358 {
2359 	struct ibmvnic_adapter *adapter;
2360 
2361 	adapter = container_of(work, struct ibmvnic_adapter,
2362 			       ibmvnic_delayed_reset.work);
2363 	__ibmvnic_reset(&adapter->ibmvnic_reset);
2364 }
2365 
2366 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2367 			 enum ibmvnic_reset_reason reason)
2368 {
2369 	struct list_head *entry, *tmp_entry;
2370 	struct ibmvnic_rwi *rwi, *tmp;
2371 	struct net_device *netdev = adapter->netdev;
2372 	unsigned long flags;
2373 	int ret;
2374 
2375 	/*
2376 	 * If failover is pending don't schedule any other reset.
2377 	 * Instead let the failover complete. If there is already a
2378 	 * a failover reset scheduled, we will detect and drop the
2379 	 * duplicate reset when walking the ->rwi_list below.
2380 	 */
2381 	if (adapter->state == VNIC_REMOVING ||
2382 	    adapter->state == VNIC_REMOVED ||
2383 	    (adapter->failover_pending && reason != VNIC_RESET_FAILOVER)) {
2384 		ret = EBUSY;
2385 		netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2386 		goto err;
2387 	}
2388 
2389 	if (adapter->state == VNIC_PROBING) {
2390 		netdev_warn(netdev, "Adapter reset during probe\n");
2391 		ret = adapter->init_done_rc = EAGAIN;
2392 		goto err;
2393 	}
2394 
2395 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2396 
2397 	list_for_each(entry, &adapter->rwi_list) {
2398 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
2399 		if (tmp->reset_reason == reason) {
2400 			netdev_dbg(netdev, "Skipping matching reset\n");
2401 			spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2402 			ret = EBUSY;
2403 			goto err;
2404 		}
2405 	}
2406 
2407 	rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2408 	if (!rwi) {
2409 		spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2410 		ibmvnic_close(netdev);
2411 		ret = ENOMEM;
2412 		goto err;
2413 	}
2414 	/* if we just received a transport event,
2415 	 * flush reset queue and process this reset
2416 	 */
2417 	if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2418 		list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2419 			list_del(entry);
2420 	}
2421 	rwi->reset_reason = reason;
2422 	list_add_tail(&rwi->list, &adapter->rwi_list);
2423 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2424 	netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2425 	schedule_work(&adapter->ibmvnic_reset);
2426 
2427 	return 0;
2428 err:
2429 	return -ret;
2430 }
2431 
2432 static void ibmvnic_tx_timeout(struct net_device *dev, unsigned int txqueue)
2433 {
2434 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2435 
2436 	if (test_bit(0, &adapter->resetting)) {
2437 		netdev_err(adapter->netdev,
2438 			   "Adapter is resetting, skip timeout reset\n");
2439 		return;
2440 	}
2441 	/* No queuing up reset until at least 5 seconds (default watchdog val)
2442 	 * after last reset
2443 	 */
2444 	if (time_before(jiffies, (adapter->last_reset_time + dev->watchdog_timeo))) {
2445 		netdev_dbg(dev, "Not yet time to tx timeout.\n");
2446 		return;
2447 	}
2448 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2449 }
2450 
2451 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2452 				  struct ibmvnic_rx_buff *rx_buff)
2453 {
2454 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2455 
2456 	rx_buff->skb = NULL;
2457 
2458 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2459 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2460 
2461 	atomic_dec(&pool->available);
2462 }
2463 
2464 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2465 {
2466 	struct ibmvnic_sub_crq_queue *rx_scrq;
2467 	struct ibmvnic_adapter *adapter;
2468 	struct net_device *netdev;
2469 	int frames_processed;
2470 	int scrq_num;
2471 
2472 	netdev = napi->dev;
2473 	adapter = netdev_priv(netdev);
2474 	scrq_num = (int)(napi - adapter->napi);
2475 	frames_processed = 0;
2476 	rx_scrq = adapter->rx_scrq[scrq_num];
2477 
2478 restart_poll:
2479 	while (frames_processed < budget) {
2480 		struct sk_buff *skb;
2481 		struct ibmvnic_rx_buff *rx_buff;
2482 		union sub_crq *next;
2483 		u32 length;
2484 		u16 offset;
2485 		u8 flags = 0;
2486 
2487 		if (unlikely(test_bit(0, &adapter->resetting) &&
2488 			     adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2489 			enable_scrq_irq(adapter, rx_scrq);
2490 			napi_complete_done(napi, frames_processed);
2491 			return frames_processed;
2492 		}
2493 
2494 		if (!pending_scrq(adapter, rx_scrq))
2495 			break;
2496 		/* The queue entry at the current index is peeked at above
2497 		 * to determine that there is a valid descriptor awaiting
2498 		 * processing. We want to be sure that the current slot
2499 		 * holds a valid descriptor before reading its contents.
2500 		 */
2501 		dma_rmb();
2502 		next = ibmvnic_next_scrq(adapter, rx_scrq);
2503 		rx_buff =
2504 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2505 							  rx_comp.correlator);
2506 		/* do error checking */
2507 		if (next->rx_comp.rc) {
2508 			netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2509 				   be16_to_cpu(next->rx_comp.rc));
2510 			/* free the entry */
2511 			next->rx_comp.first = 0;
2512 			dev_kfree_skb_any(rx_buff->skb);
2513 			remove_buff_from_pool(adapter, rx_buff);
2514 			continue;
2515 		} else if (!rx_buff->skb) {
2516 			/* free the entry */
2517 			next->rx_comp.first = 0;
2518 			remove_buff_from_pool(adapter, rx_buff);
2519 			continue;
2520 		}
2521 
2522 		length = be32_to_cpu(next->rx_comp.len);
2523 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
2524 		flags = next->rx_comp.flags;
2525 		skb = rx_buff->skb;
2526 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
2527 					length);
2528 
2529 		/* VLAN Header has been stripped by the system firmware and
2530 		 * needs to be inserted by the driver
2531 		 */
2532 		if (adapter->rx_vlan_header_insertion &&
2533 		    (flags & IBMVNIC_VLAN_STRIPPED))
2534 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2535 					       ntohs(next->rx_comp.vlan_tci));
2536 
2537 		/* free the entry */
2538 		next->rx_comp.first = 0;
2539 		remove_buff_from_pool(adapter, rx_buff);
2540 
2541 		skb_put(skb, length);
2542 		skb->protocol = eth_type_trans(skb, netdev);
2543 		skb_record_rx_queue(skb, scrq_num);
2544 
2545 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2546 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2547 			skb->ip_summed = CHECKSUM_UNNECESSARY;
2548 		}
2549 
2550 		length = skb->len;
2551 		napi_gro_receive(napi, skb); /* send it up */
2552 		netdev->stats.rx_packets++;
2553 		netdev->stats.rx_bytes += length;
2554 		adapter->rx_stats_buffers[scrq_num].packets++;
2555 		adapter->rx_stats_buffers[scrq_num].bytes += length;
2556 		frames_processed++;
2557 	}
2558 
2559 	if (adapter->state != VNIC_CLOSING &&
2560 	    ((atomic_read(&adapter->rx_pool[scrq_num].available) <
2561 	      adapter->req_rx_add_entries_per_subcrq / 2) ||
2562 	      frames_processed < budget))
2563 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2564 	if (frames_processed < budget) {
2565 		if (napi_complete_done(napi, frames_processed)) {
2566 			enable_scrq_irq(adapter, rx_scrq);
2567 			if (pending_scrq(adapter, rx_scrq)) {
2568 				rmb();
2569 				if (napi_reschedule(napi)) {
2570 					disable_scrq_irq(adapter, rx_scrq);
2571 					goto restart_poll;
2572 				}
2573 			}
2574 		}
2575 	}
2576 	return frames_processed;
2577 }
2578 
2579 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2580 {
2581 	int rc, ret;
2582 
2583 	adapter->fallback.mtu = adapter->req_mtu;
2584 	adapter->fallback.rx_queues = adapter->req_rx_queues;
2585 	adapter->fallback.tx_queues = adapter->req_tx_queues;
2586 	adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2587 	adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2588 
2589 	reinit_completion(&adapter->reset_done);
2590 	adapter->wait_for_reset = true;
2591 	rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2592 
2593 	if (rc) {
2594 		ret = rc;
2595 		goto out;
2596 	}
2597 	rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done, 60000);
2598 	if (rc) {
2599 		ret = -ENODEV;
2600 		goto out;
2601 	}
2602 
2603 	ret = 0;
2604 	if (adapter->reset_done_rc) {
2605 		ret = -EIO;
2606 		adapter->desired.mtu = adapter->fallback.mtu;
2607 		adapter->desired.rx_queues = adapter->fallback.rx_queues;
2608 		adapter->desired.tx_queues = adapter->fallback.tx_queues;
2609 		adapter->desired.rx_entries = adapter->fallback.rx_entries;
2610 		adapter->desired.tx_entries = adapter->fallback.tx_entries;
2611 
2612 		reinit_completion(&adapter->reset_done);
2613 		adapter->wait_for_reset = true;
2614 		rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2615 		if (rc) {
2616 			ret = rc;
2617 			goto out;
2618 		}
2619 		rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done,
2620 						 60000);
2621 		if (rc) {
2622 			ret = -ENODEV;
2623 			goto out;
2624 		}
2625 	}
2626 out:
2627 	adapter->wait_for_reset = false;
2628 
2629 	return ret;
2630 }
2631 
2632 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2633 {
2634 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2635 
2636 	adapter->desired.mtu = new_mtu + ETH_HLEN;
2637 
2638 	return wait_for_reset(adapter);
2639 }
2640 
2641 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2642 						struct net_device *dev,
2643 						netdev_features_t features)
2644 {
2645 	/* Some backing hardware adapters can not
2646 	 * handle packets with a MSS less than 224
2647 	 * or with only one segment.
2648 	 */
2649 	if (skb_is_gso(skb)) {
2650 		if (skb_shinfo(skb)->gso_size < 224 ||
2651 		    skb_shinfo(skb)->gso_segs == 1)
2652 			features &= ~NETIF_F_GSO_MASK;
2653 	}
2654 
2655 	return features;
2656 }
2657 
2658 static const struct net_device_ops ibmvnic_netdev_ops = {
2659 	.ndo_open		= ibmvnic_open,
2660 	.ndo_stop		= ibmvnic_close,
2661 	.ndo_start_xmit		= ibmvnic_xmit,
2662 	.ndo_set_rx_mode	= ibmvnic_set_multi,
2663 	.ndo_set_mac_address	= ibmvnic_set_mac,
2664 	.ndo_validate_addr	= eth_validate_addr,
2665 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
2666 	.ndo_change_mtu		= ibmvnic_change_mtu,
2667 	.ndo_features_check     = ibmvnic_features_check,
2668 };
2669 
2670 /* ethtool functions */
2671 
2672 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2673 				      struct ethtool_link_ksettings *cmd)
2674 {
2675 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2676 	int rc;
2677 
2678 	rc = send_query_phys_parms(adapter);
2679 	if (rc) {
2680 		adapter->speed = SPEED_UNKNOWN;
2681 		adapter->duplex = DUPLEX_UNKNOWN;
2682 	}
2683 	cmd->base.speed = adapter->speed;
2684 	cmd->base.duplex = adapter->duplex;
2685 	cmd->base.port = PORT_FIBRE;
2686 	cmd->base.phy_address = 0;
2687 	cmd->base.autoneg = AUTONEG_ENABLE;
2688 
2689 	return 0;
2690 }
2691 
2692 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2693 				struct ethtool_drvinfo *info)
2694 {
2695 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2696 
2697 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2698 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2699 	strlcpy(info->fw_version, adapter->fw_version,
2700 		sizeof(info->fw_version));
2701 }
2702 
2703 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2704 {
2705 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2706 
2707 	return adapter->msg_enable;
2708 }
2709 
2710 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2711 {
2712 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2713 
2714 	adapter->msg_enable = data;
2715 }
2716 
2717 static u32 ibmvnic_get_link(struct net_device *netdev)
2718 {
2719 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2720 
2721 	/* Don't need to send a query because we request a logical link up at
2722 	 * init and then we wait for link state indications
2723 	 */
2724 	return adapter->logical_link_state;
2725 }
2726 
2727 static void ibmvnic_get_ringparam(struct net_device *netdev,
2728 				  struct ethtool_ringparam *ring)
2729 {
2730 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2731 
2732 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2733 		ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2734 		ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2735 	} else {
2736 		ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2737 		ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2738 	}
2739 	ring->rx_mini_max_pending = 0;
2740 	ring->rx_jumbo_max_pending = 0;
2741 	ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2742 	ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2743 	ring->rx_mini_pending = 0;
2744 	ring->rx_jumbo_pending = 0;
2745 }
2746 
2747 static int ibmvnic_set_ringparam(struct net_device *netdev,
2748 				 struct ethtool_ringparam *ring)
2749 {
2750 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2751 	int ret;
2752 
2753 	ret = 0;
2754 	adapter->desired.rx_entries = ring->rx_pending;
2755 	adapter->desired.tx_entries = ring->tx_pending;
2756 
2757 	ret = wait_for_reset(adapter);
2758 
2759 	if (!ret &&
2760 	    (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2761 	     adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2762 		netdev_info(netdev,
2763 			    "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2764 			    ring->rx_pending, ring->tx_pending,
2765 			    adapter->req_rx_add_entries_per_subcrq,
2766 			    adapter->req_tx_entries_per_subcrq);
2767 	return ret;
2768 }
2769 
2770 static void ibmvnic_get_channels(struct net_device *netdev,
2771 				 struct ethtool_channels *channels)
2772 {
2773 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2774 
2775 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2776 		channels->max_rx = adapter->max_rx_queues;
2777 		channels->max_tx = adapter->max_tx_queues;
2778 	} else {
2779 		channels->max_rx = IBMVNIC_MAX_QUEUES;
2780 		channels->max_tx = IBMVNIC_MAX_QUEUES;
2781 	}
2782 
2783 	channels->max_other = 0;
2784 	channels->max_combined = 0;
2785 	channels->rx_count = adapter->req_rx_queues;
2786 	channels->tx_count = adapter->req_tx_queues;
2787 	channels->other_count = 0;
2788 	channels->combined_count = 0;
2789 }
2790 
2791 static int ibmvnic_set_channels(struct net_device *netdev,
2792 				struct ethtool_channels *channels)
2793 {
2794 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2795 	int ret;
2796 
2797 	ret = 0;
2798 	adapter->desired.rx_queues = channels->rx_count;
2799 	adapter->desired.tx_queues = channels->tx_count;
2800 
2801 	ret = wait_for_reset(adapter);
2802 
2803 	if (!ret &&
2804 	    (adapter->req_rx_queues != channels->rx_count ||
2805 	     adapter->req_tx_queues != channels->tx_count))
2806 		netdev_info(netdev,
2807 			    "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2808 			    channels->rx_count, channels->tx_count,
2809 			    adapter->req_rx_queues, adapter->req_tx_queues);
2810 	return ret;
2811 
2812 }
2813 
2814 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2815 {
2816 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2817 	int i;
2818 
2819 	switch (stringset) {
2820 	case ETH_SS_STATS:
2821 		for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2822 				i++, data += ETH_GSTRING_LEN)
2823 			memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2824 
2825 		for (i = 0; i < adapter->req_tx_queues; i++) {
2826 			snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2827 			data += ETH_GSTRING_LEN;
2828 
2829 			snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2830 			data += ETH_GSTRING_LEN;
2831 
2832 			snprintf(data, ETH_GSTRING_LEN,
2833 				 "tx%d_dropped_packets", i);
2834 			data += ETH_GSTRING_LEN;
2835 		}
2836 
2837 		for (i = 0; i < adapter->req_rx_queues; i++) {
2838 			snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2839 			data += ETH_GSTRING_LEN;
2840 
2841 			snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2842 			data += ETH_GSTRING_LEN;
2843 
2844 			snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2845 			data += ETH_GSTRING_LEN;
2846 		}
2847 		break;
2848 
2849 	case ETH_SS_PRIV_FLAGS:
2850 		for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2851 			strcpy(data + i * ETH_GSTRING_LEN,
2852 			       ibmvnic_priv_flags[i]);
2853 		break;
2854 	default:
2855 		return;
2856 	}
2857 }
2858 
2859 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2860 {
2861 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2862 
2863 	switch (sset) {
2864 	case ETH_SS_STATS:
2865 		return ARRAY_SIZE(ibmvnic_stats) +
2866 		       adapter->req_tx_queues * NUM_TX_STATS +
2867 		       adapter->req_rx_queues * NUM_RX_STATS;
2868 	case ETH_SS_PRIV_FLAGS:
2869 		return ARRAY_SIZE(ibmvnic_priv_flags);
2870 	default:
2871 		return -EOPNOTSUPP;
2872 	}
2873 }
2874 
2875 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2876 				      struct ethtool_stats *stats, u64 *data)
2877 {
2878 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2879 	union ibmvnic_crq crq;
2880 	int i, j;
2881 	int rc;
2882 
2883 	memset(&crq, 0, sizeof(crq));
2884 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2885 	crq.request_statistics.cmd = REQUEST_STATISTICS;
2886 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2887 	crq.request_statistics.len =
2888 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
2889 
2890 	/* Wait for data to be written */
2891 	reinit_completion(&adapter->stats_done);
2892 	rc = ibmvnic_send_crq(adapter, &crq);
2893 	if (rc)
2894 		return;
2895 	rc = ibmvnic_wait_for_completion(adapter, &adapter->stats_done, 10000);
2896 	if (rc)
2897 		return;
2898 
2899 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2900 		data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2901 						ibmvnic_stats[i].offset));
2902 
2903 	for (j = 0; j < adapter->req_tx_queues; j++) {
2904 		data[i] = adapter->tx_stats_buffers[j].packets;
2905 		i++;
2906 		data[i] = adapter->tx_stats_buffers[j].bytes;
2907 		i++;
2908 		data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2909 		i++;
2910 	}
2911 
2912 	for (j = 0; j < adapter->req_rx_queues; j++) {
2913 		data[i] = adapter->rx_stats_buffers[j].packets;
2914 		i++;
2915 		data[i] = adapter->rx_stats_buffers[j].bytes;
2916 		i++;
2917 		data[i] = adapter->rx_stats_buffers[j].interrupts;
2918 		i++;
2919 	}
2920 }
2921 
2922 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2923 {
2924 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2925 
2926 	return adapter->priv_flags;
2927 }
2928 
2929 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2930 {
2931 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2932 	bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2933 
2934 	if (which_maxes)
2935 		adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2936 	else
2937 		adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2938 
2939 	return 0;
2940 }
2941 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2942 	.get_drvinfo		= ibmvnic_get_drvinfo,
2943 	.get_msglevel		= ibmvnic_get_msglevel,
2944 	.set_msglevel		= ibmvnic_set_msglevel,
2945 	.get_link		= ibmvnic_get_link,
2946 	.get_ringparam		= ibmvnic_get_ringparam,
2947 	.set_ringparam		= ibmvnic_set_ringparam,
2948 	.get_channels		= ibmvnic_get_channels,
2949 	.set_channels		= ibmvnic_set_channels,
2950 	.get_strings            = ibmvnic_get_strings,
2951 	.get_sset_count         = ibmvnic_get_sset_count,
2952 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
2953 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
2954 	.get_priv_flags		= ibmvnic_get_priv_flags,
2955 	.set_priv_flags		= ibmvnic_set_priv_flags,
2956 };
2957 
2958 /* Routines for managing CRQs/sCRQs  */
2959 
2960 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2961 				   struct ibmvnic_sub_crq_queue *scrq)
2962 {
2963 	int rc;
2964 
2965 	if (!scrq) {
2966 		netdev_dbg(adapter->netdev,
2967 			   "Invalid scrq reset. irq (%d) or msgs (%p).\n",
2968 			   scrq->irq, scrq->msgs);
2969 		return -EINVAL;
2970 	}
2971 
2972 	if (scrq->irq) {
2973 		free_irq(scrq->irq, scrq);
2974 		irq_dispose_mapping(scrq->irq);
2975 		scrq->irq = 0;
2976 	}
2977 
2978 	if (scrq->msgs) {
2979 		memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2980 		atomic_set(&scrq->used, 0);
2981 		scrq->cur = 0;
2982 		scrq->ind_buf.index = 0;
2983 	} else {
2984 		netdev_dbg(adapter->netdev, "Invalid scrq reset\n");
2985 		return -EINVAL;
2986 	}
2987 
2988 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2989 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2990 	return rc;
2991 }
2992 
2993 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2994 {
2995 	int i, rc;
2996 
2997 	if (!adapter->tx_scrq || !adapter->rx_scrq)
2998 		return -EINVAL;
2999 
3000 	for (i = 0; i < adapter->req_tx_queues; i++) {
3001 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
3002 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
3003 		if (rc)
3004 			return rc;
3005 	}
3006 
3007 	for (i = 0; i < adapter->req_rx_queues; i++) {
3008 		netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
3009 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
3010 		if (rc)
3011 			return rc;
3012 	}
3013 
3014 	return rc;
3015 }
3016 
3017 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
3018 				  struct ibmvnic_sub_crq_queue *scrq,
3019 				  bool do_h_free)
3020 {
3021 	struct device *dev = &adapter->vdev->dev;
3022 	long rc;
3023 
3024 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
3025 
3026 	if (do_h_free) {
3027 		/* Close the sub-crqs */
3028 		do {
3029 			rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
3030 						adapter->vdev->unit_address,
3031 						scrq->crq_num);
3032 		} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3033 
3034 		if (rc) {
3035 			netdev_err(adapter->netdev,
3036 				   "Failed to release sub-CRQ %16lx, rc = %ld\n",
3037 				   scrq->crq_num, rc);
3038 		}
3039 	}
3040 
3041 	dma_free_coherent(dev,
3042 			  IBMVNIC_IND_ARR_SZ,
3043 			  scrq->ind_buf.indir_arr,
3044 			  scrq->ind_buf.indir_dma);
3045 
3046 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
3047 			 DMA_BIDIRECTIONAL);
3048 	free_pages((unsigned long)scrq->msgs, 2);
3049 	kfree(scrq);
3050 }
3051 
3052 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
3053 							*adapter)
3054 {
3055 	struct device *dev = &adapter->vdev->dev;
3056 	struct ibmvnic_sub_crq_queue *scrq;
3057 	int rc;
3058 
3059 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
3060 	if (!scrq)
3061 		return NULL;
3062 
3063 	scrq->msgs =
3064 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
3065 	if (!scrq->msgs) {
3066 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
3067 		goto zero_page_failed;
3068 	}
3069 
3070 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
3071 					 DMA_BIDIRECTIONAL);
3072 	if (dma_mapping_error(dev, scrq->msg_token)) {
3073 		dev_warn(dev, "Couldn't map crq queue messages page\n");
3074 		goto map_failed;
3075 	}
3076 
3077 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
3078 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
3079 
3080 	if (rc == H_RESOURCE)
3081 		rc = ibmvnic_reset_crq(adapter);
3082 
3083 	if (rc == H_CLOSED) {
3084 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
3085 	} else if (rc) {
3086 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
3087 		goto reg_failed;
3088 	}
3089 
3090 	scrq->adapter = adapter;
3091 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
3092 	scrq->ind_buf.index = 0;
3093 
3094 	scrq->ind_buf.indir_arr =
3095 		dma_alloc_coherent(dev,
3096 				   IBMVNIC_IND_ARR_SZ,
3097 				   &scrq->ind_buf.indir_dma,
3098 				   GFP_KERNEL);
3099 
3100 	if (!scrq->ind_buf.indir_arr)
3101 		goto indir_failed;
3102 
3103 	spin_lock_init(&scrq->lock);
3104 
3105 	netdev_dbg(adapter->netdev,
3106 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
3107 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
3108 
3109 	return scrq;
3110 
3111 indir_failed:
3112 	do {
3113 		rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
3114 					adapter->vdev->unit_address,
3115 					scrq->crq_num);
3116 	} while (rc == H_BUSY || rc == H_IS_LONG_BUSY(rc));
3117 reg_failed:
3118 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
3119 			 DMA_BIDIRECTIONAL);
3120 map_failed:
3121 	free_pages((unsigned long)scrq->msgs, 2);
3122 zero_page_failed:
3123 	kfree(scrq);
3124 
3125 	return NULL;
3126 }
3127 
3128 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
3129 {
3130 	int i;
3131 
3132 	if (adapter->tx_scrq) {
3133 		for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
3134 			if (!adapter->tx_scrq[i])
3135 				continue;
3136 
3137 			netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
3138 				   i);
3139 			if (adapter->tx_scrq[i]->irq) {
3140 				free_irq(adapter->tx_scrq[i]->irq,
3141 					 adapter->tx_scrq[i]);
3142 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
3143 				adapter->tx_scrq[i]->irq = 0;
3144 			}
3145 
3146 			release_sub_crq_queue(adapter, adapter->tx_scrq[i],
3147 					      do_h_free);
3148 		}
3149 
3150 		kfree(adapter->tx_scrq);
3151 		adapter->tx_scrq = NULL;
3152 		adapter->num_active_tx_scrqs = 0;
3153 	}
3154 
3155 	if (adapter->rx_scrq) {
3156 		for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
3157 			if (!adapter->rx_scrq[i])
3158 				continue;
3159 
3160 			netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
3161 				   i);
3162 			if (adapter->rx_scrq[i]->irq) {
3163 				free_irq(adapter->rx_scrq[i]->irq,
3164 					 adapter->rx_scrq[i]);
3165 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
3166 				adapter->rx_scrq[i]->irq = 0;
3167 			}
3168 
3169 			release_sub_crq_queue(adapter, adapter->rx_scrq[i],
3170 					      do_h_free);
3171 		}
3172 
3173 		kfree(adapter->rx_scrq);
3174 		adapter->rx_scrq = NULL;
3175 		adapter->num_active_rx_scrqs = 0;
3176 	}
3177 }
3178 
3179 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
3180 			    struct ibmvnic_sub_crq_queue *scrq)
3181 {
3182 	struct device *dev = &adapter->vdev->dev;
3183 	unsigned long rc;
3184 
3185 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3186 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3187 	if (rc)
3188 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
3189 			scrq->hw_irq, rc);
3190 	return rc;
3191 }
3192 
3193 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
3194 			   struct ibmvnic_sub_crq_queue *scrq)
3195 {
3196 	struct device *dev = &adapter->vdev->dev;
3197 	unsigned long rc;
3198 
3199 	if (scrq->hw_irq > 0x100000000ULL) {
3200 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
3201 		return 1;
3202 	}
3203 
3204 	if (test_bit(0, &adapter->resetting) &&
3205 	    adapter->reset_reason == VNIC_RESET_MOBILITY) {
3206 		u64 val = (0xff000000) | scrq->hw_irq;
3207 
3208 		rc = plpar_hcall_norets(H_EOI, val);
3209 		/* H_EOI would fail with rc = H_FUNCTION when running
3210 		 * in XIVE mode which is expected, but not an error.
3211 		 */
3212 		if (rc && (rc != H_FUNCTION))
3213 			dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
3214 				val, rc);
3215 	}
3216 
3217 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3218 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3219 	if (rc)
3220 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
3221 			scrq->hw_irq, rc);
3222 	return rc;
3223 }
3224 
3225 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
3226 			       struct ibmvnic_sub_crq_queue *scrq)
3227 {
3228 	struct device *dev = &adapter->vdev->dev;
3229 	struct ibmvnic_tx_pool *tx_pool;
3230 	struct ibmvnic_tx_buff *txbuff;
3231 	struct netdev_queue *txq;
3232 	union sub_crq *next;
3233 	int index;
3234 	int i;
3235 
3236 restart_loop:
3237 	while (pending_scrq(adapter, scrq)) {
3238 		unsigned int pool = scrq->pool_index;
3239 		int num_entries = 0;
3240 		int total_bytes = 0;
3241 		int num_packets = 0;
3242 
3243 		/* The queue entry at the current index is peeked at above
3244 		 * to determine that there is a valid descriptor awaiting
3245 		 * processing. We want to be sure that the current slot
3246 		 * holds a valid descriptor before reading its contents.
3247 		 */
3248 		dma_rmb();
3249 
3250 		next = ibmvnic_next_scrq(adapter, scrq);
3251 		for (i = 0; i < next->tx_comp.num_comps; i++) {
3252 			if (next->tx_comp.rcs[i])
3253 				dev_err(dev, "tx error %x\n",
3254 					next->tx_comp.rcs[i]);
3255 			index = be32_to_cpu(next->tx_comp.correlators[i]);
3256 			if (index & IBMVNIC_TSO_POOL_MASK) {
3257 				tx_pool = &adapter->tso_pool[pool];
3258 				index &= ~IBMVNIC_TSO_POOL_MASK;
3259 			} else {
3260 				tx_pool = &adapter->tx_pool[pool];
3261 			}
3262 
3263 			txbuff = &tx_pool->tx_buff[index];
3264 			num_packets++;
3265 			num_entries += txbuff->num_entries;
3266 			if (txbuff->skb) {
3267 				total_bytes += txbuff->skb->len;
3268 				dev_consume_skb_irq(txbuff->skb);
3269 				txbuff->skb = NULL;
3270 			} else {
3271 				netdev_warn(adapter->netdev,
3272 					    "TX completion received with NULL socket buffer\n");
3273 			}
3274 			tx_pool->free_map[tx_pool->producer_index] = index;
3275 			tx_pool->producer_index =
3276 				(tx_pool->producer_index + 1) %
3277 					tx_pool->num_buffers;
3278 		}
3279 		/* remove tx_comp scrq*/
3280 		next->tx_comp.first = 0;
3281 
3282 		txq = netdev_get_tx_queue(adapter->netdev, scrq->pool_index);
3283 		netdev_tx_completed_queue(txq, num_packets, total_bytes);
3284 
3285 		if (atomic_sub_return(num_entries, &scrq->used) <=
3286 		    (adapter->req_tx_entries_per_subcrq / 2) &&
3287 		    __netif_subqueue_stopped(adapter->netdev,
3288 					     scrq->pool_index)) {
3289 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
3290 			netdev_dbg(adapter->netdev, "Started queue %d\n",
3291 				   scrq->pool_index);
3292 		}
3293 	}
3294 
3295 	enable_scrq_irq(adapter, scrq);
3296 
3297 	if (pending_scrq(adapter, scrq)) {
3298 		disable_scrq_irq(adapter, scrq);
3299 		goto restart_loop;
3300 	}
3301 
3302 	return 0;
3303 }
3304 
3305 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
3306 {
3307 	struct ibmvnic_sub_crq_queue *scrq = instance;
3308 	struct ibmvnic_adapter *adapter = scrq->adapter;
3309 
3310 	disable_scrq_irq(adapter, scrq);
3311 	ibmvnic_complete_tx(adapter, scrq);
3312 
3313 	return IRQ_HANDLED;
3314 }
3315 
3316 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
3317 {
3318 	struct ibmvnic_sub_crq_queue *scrq = instance;
3319 	struct ibmvnic_adapter *adapter = scrq->adapter;
3320 
3321 	/* When booting a kdump kernel we can hit pending interrupts
3322 	 * prior to completing driver initialization.
3323 	 */
3324 	if (unlikely(adapter->state != VNIC_OPEN))
3325 		return IRQ_NONE;
3326 
3327 	adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
3328 
3329 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
3330 		disable_scrq_irq(adapter, scrq);
3331 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
3332 	}
3333 
3334 	return IRQ_HANDLED;
3335 }
3336 
3337 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
3338 {
3339 	struct device *dev = &adapter->vdev->dev;
3340 	struct ibmvnic_sub_crq_queue *scrq;
3341 	int i = 0, j = 0;
3342 	int rc = 0;
3343 
3344 	for (i = 0; i < adapter->req_tx_queues; i++) {
3345 		netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
3346 			   i);
3347 		scrq = adapter->tx_scrq[i];
3348 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3349 
3350 		if (!scrq->irq) {
3351 			rc = -EINVAL;
3352 			dev_err(dev, "Error mapping irq\n");
3353 			goto req_tx_irq_failed;
3354 		}
3355 
3356 		snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
3357 			 adapter->vdev->unit_address, i);
3358 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
3359 				 0, scrq->name, scrq);
3360 
3361 		if (rc) {
3362 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
3363 				scrq->irq, rc);
3364 			irq_dispose_mapping(scrq->irq);
3365 			goto req_tx_irq_failed;
3366 		}
3367 	}
3368 
3369 	for (i = 0; i < adapter->req_rx_queues; i++) {
3370 		netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
3371 			   i);
3372 		scrq = adapter->rx_scrq[i];
3373 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3374 		if (!scrq->irq) {
3375 			rc = -EINVAL;
3376 			dev_err(dev, "Error mapping irq\n");
3377 			goto req_rx_irq_failed;
3378 		}
3379 		snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
3380 			 adapter->vdev->unit_address, i);
3381 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
3382 				 0, scrq->name, scrq);
3383 		if (rc) {
3384 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
3385 				scrq->irq, rc);
3386 			irq_dispose_mapping(scrq->irq);
3387 			goto req_rx_irq_failed;
3388 		}
3389 	}
3390 	return rc;
3391 
3392 req_rx_irq_failed:
3393 	for (j = 0; j < i; j++) {
3394 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
3395 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
3396 	}
3397 	i = adapter->req_tx_queues;
3398 req_tx_irq_failed:
3399 	for (j = 0; j < i; j++) {
3400 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
3401 		irq_dispose_mapping(adapter->tx_scrq[j]->irq);
3402 	}
3403 	release_sub_crqs(adapter, 1);
3404 	return rc;
3405 }
3406 
3407 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
3408 {
3409 	struct device *dev = &adapter->vdev->dev;
3410 	struct ibmvnic_sub_crq_queue **allqueues;
3411 	int registered_queues = 0;
3412 	int total_queues;
3413 	int more = 0;
3414 	int i;
3415 
3416 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
3417 
3418 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
3419 	if (!allqueues)
3420 		return -1;
3421 
3422 	for (i = 0; i < total_queues; i++) {
3423 		allqueues[i] = init_sub_crq_queue(adapter);
3424 		if (!allqueues[i]) {
3425 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
3426 			break;
3427 		}
3428 		registered_queues++;
3429 	}
3430 
3431 	/* Make sure we were able to register the minimum number of queues */
3432 	if (registered_queues <
3433 	    adapter->min_tx_queues + adapter->min_rx_queues) {
3434 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
3435 		goto tx_failed;
3436 	}
3437 
3438 	/* Distribute the failed allocated queues*/
3439 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
3440 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3441 		switch (i % 3) {
3442 		case 0:
3443 			if (adapter->req_rx_queues > adapter->min_rx_queues)
3444 				adapter->req_rx_queues--;
3445 			else
3446 				more++;
3447 			break;
3448 		case 1:
3449 			if (adapter->req_tx_queues > adapter->min_tx_queues)
3450 				adapter->req_tx_queues--;
3451 			else
3452 				more++;
3453 			break;
3454 		}
3455 	}
3456 
3457 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3458 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
3459 	if (!adapter->tx_scrq)
3460 		goto tx_failed;
3461 
3462 	for (i = 0; i < adapter->req_tx_queues; i++) {
3463 		adapter->tx_scrq[i] = allqueues[i];
3464 		adapter->tx_scrq[i]->pool_index = i;
3465 		adapter->num_active_tx_scrqs++;
3466 	}
3467 
3468 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3469 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
3470 	if (!adapter->rx_scrq)
3471 		goto rx_failed;
3472 
3473 	for (i = 0; i < adapter->req_rx_queues; i++) {
3474 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3475 		adapter->rx_scrq[i]->scrq_num = i;
3476 		adapter->num_active_rx_scrqs++;
3477 	}
3478 
3479 	kfree(allqueues);
3480 	return 0;
3481 
3482 rx_failed:
3483 	kfree(adapter->tx_scrq);
3484 	adapter->tx_scrq = NULL;
3485 tx_failed:
3486 	for (i = 0; i < registered_queues; i++)
3487 		release_sub_crq_queue(adapter, allqueues[i], 1);
3488 	kfree(allqueues);
3489 	return -1;
3490 }
3491 
3492 static void send_request_cap(struct ibmvnic_adapter *adapter, int retry)
3493 {
3494 	struct device *dev = &adapter->vdev->dev;
3495 	union ibmvnic_crq crq;
3496 	int max_entries;
3497 
3498 	if (!retry) {
3499 		/* Sub-CRQ entries are 32 byte long */
3500 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3501 
3502 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
3503 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
3504 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3505 			return;
3506 		}
3507 
3508 		if (adapter->desired.mtu)
3509 			adapter->req_mtu = adapter->desired.mtu;
3510 		else
3511 			adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3512 
3513 		if (!adapter->desired.tx_entries)
3514 			adapter->desired.tx_entries =
3515 					adapter->max_tx_entries_per_subcrq;
3516 		if (!adapter->desired.rx_entries)
3517 			adapter->desired.rx_entries =
3518 					adapter->max_rx_add_entries_per_subcrq;
3519 
3520 		max_entries = IBMVNIC_MAX_LTB_SIZE /
3521 			      (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3522 
3523 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3524 			adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3525 			adapter->desired.tx_entries = max_entries;
3526 		}
3527 
3528 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3529 			adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3530 			adapter->desired.rx_entries = max_entries;
3531 		}
3532 
3533 		if (adapter->desired.tx_entries)
3534 			adapter->req_tx_entries_per_subcrq =
3535 					adapter->desired.tx_entries;
3536 		else
3537 			adapter->req_tx_entries_per_subcrq =
3538 					adapter->max_tx_entries_per_subcrq;
3539 
3540 		if (adapter->desired.rx_entries)
3541 			adapter->req_rx_add_entries_per_subcrq =
3542 					adapter->desired.rx_entries;
3543 		else
3544 			adapter->req_rx_add_entries_per_subcrq =
3545 					adapter->max_rx_add_entries_per_subcrq;
3546 
3547 		if (adapter->desired.tx_queues)
3548 			adapter->req_tx_queues =
3549 					adapter->desired.tx_queues;
3550 		else
3551 			adapter->req_tx_queues =
3552 					adapter->opt_tx_comp_sub_queues;
3553 
3554 		if (adapter->desired.rx_queues)
3555 			adapter->req_rx_queues =
3556 					adapter->desired.rx_queues;
3557 		else
3558 			adapter->req_rx_queues =
3559 					adapter->opt_rx_comp_queues;
3560 
3561 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3562 	}
3563 
3564 	memset(&crq, 0, sizeof(crq));
3565 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
3566 	crq.request_capability.cmd = REQUEST_CAPABILITY;
3567 
3568 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3569 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3570 	atomic_inc(&adapter->running_cap_crqs);
3571 	ibmvnic_send_crq(adapter, &crq);
3572 
3573 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3574 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3575 	atomic_inc(&adapter->running_cap_crqs);
3576 	ibmvnic_send_crq(adapter, &crq);
3577 
3578 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3579 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3580 	atomic_inc(&adapter->running_cap_crqs);
3581 	ibmvnic_send_crq(adapter, &crq);
3582 
3583 	crq.request_capability.capability =
3584 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3585 	crq.request_capability.number =
3586 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3587 	atomic_inc(&adapter->running_cap_crqs);
3588 	ibmvnic_send_crq(adapter, &crq);
3589 
3590 	crq.request_capability.capability =
3591 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3592 	crq.request_capability.number =
3593 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3594 	atomic_inc(&adapter->running_cap_crqs);
3595 	ibmvnic_send_crq(adapter, &crq);
3596 
3597 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3598 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3599 	atomic_inc(&adapter->running_cap_crqs);
3600 	ibmvnic_send_crq(adapter, &crq);
3601 
3602 	if (adapter->netdev->flags & IFF_PROMISC) {
3603 		if (adapter->promisc_supported) {
3604 			crq.request_capability.capability =
3605 			    cpu_to_be16(PROMISC_REQUESTED);
3606 			crq.request_capability.number = cpu_to_be64(1);
3607 			atomic_inc(&adapter->running_cap_crqs);
3608 			ibmvnic_send_crq(adapter, &crq);
3609 		}
3610 	} else {
3611 		crq.request_capability.capability =
3612 		    cpu_to_be16(PROMISC_REQUESTED);
3613 		crq.request_capability.number = cpu_to_be64(0);
3614 		atomic_inc(&adapter->running_cap_crqs);
3615 		ibmvnic_send_crq(adapter, &crq);
3616 	}
3617 }
3618 
3619 static int pending_scrq(struct ibmvnic_adapter *adapter,
3620 			struct ibmvnic_sub_crq_queue *scrq)
3621 {
3622 	union sub_crq *entry = &scrq->msgs[scrq->cur];
3623 
3624 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3625 		return 1;
3626 	else
3627 		return 0;
3628 }
3629 
3630 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3631 					struct ibmvnic_sub_crq_queue *scrq)
3632 {
3633 	union sub_crq *entry;
3634 	unsigned long flags;
3635 
3636 	spin_lock_irqsave(&scrq->lock, flags);
3637 	entry = &scrq->msgs[scrq->cur];
3638 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3639 		if (++scrq->cur == scrq->size)
3640 			scrq->cur = 0;
3641 	} else {
3642 		entry = NULL;
3643 	}
3644 	spin_unlock_irqrestore(&scrq->lock, flags);
3645 
3646 	/* Ensure that the entire buffer descriptor has been
3647 	 * loaded before reading its contents
3648 	 */
3649 	dma_rmb();
3650 
3651 	return entry;
3652 }
3653 
3654 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3655 {
3656 	struct ibmvnic_crq_queue *queue = &adapter->crq;
3657 	union ibmvnic_crq *crq;
3658 
3659 	crq = &queue->msgs[queue->cur];
3660 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3661 		if (++queue->cur == queue->size)
3662 			queue->cur = 0;
3663 	} else {
3664 		crq = NULL;
3665 	}
3666 
3667 	return crq;
3668 }
3669 
3670 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3671 {
3672 	switch (rc) {
3673 	case H_PARAMETER:
3674 		dev_warn_ratelimited(dev,
3675 				     "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3676 				     func, rc);
3677 		break;
3678 	case H_CLOSED:
3679 		dev_warn_ratelimited(dev,
3680 				     "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3681 				     func, rc);
3682 		break;
3683 	default:
3684 		dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3685 		break;
3686 	}
3687 }
3688 
3689 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3690 				u64 remote_handle, u64 ioba, u64 num_entries)
3691 {
3692 	unsigned int ua = adapter->vdev->unit_address;
3693 	struct device *dev = &adapter->vdev->dev;
3694 	int rc;
3695 
3696 	/* Make sure the hypervisor sees the complete request */
3697 	mb();
3698 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3699 				cpu_to_be64(remote_handle),
3700 				ioba, num_entries);
3701 
3702 	if (rc)
3703 		print_subcrq_error(dev, rc, __func__);
3704 
3705 	return rc;
3706 }
3707 
3708 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3709 			    union ibmvnic_crq *crq)
3710 {
3711 	unsigned int ua = adapter->vdev->unit_address;
3712 	struct device *dev = &adapter->vdev->dev;
3713 	u64 *u64_crq = (u64 *)crq;
3714 	int rc;
3715 
3716 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3717 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3718 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
3719 
3720 	if (!adapter->crq.active &&
3721 	    crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3722 		dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3723 		return -EINVAL;
3724 	}
3725 
3726 	/* Make sure the hypervisor sees the complete request */
3727 	mb();
3728 
3729 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3730 				cpu_to_be64(u64_crq[0]),
3731 				cpu_to_be64(u64_crq[1]));
3732 
3733 	if (rc) {
3734 		if (rc == H_CLOSED) {
3735 			dev_warn(dev, "CRQ Queue closed\n");
3736 			/* do not reset, report the fail, wait for passive init from server */
3737 		}
3738 
3739 		dev_warn(dev, "Send error (rc=%d)\n", rc);
3740 	}
3741 
3742 	return rc;
3743 }
3744 
3745 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3746 {
3747 	struct device *dev = &adapter->vdev->dev;
3748 	union ibmvnic_crq crq;
3749 	int retries = 100;
3750 	int rc;
3751 
3752 	memset(&crq, 0, sizeof(crq));
3753 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3754 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
3755 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3756 
3757 	do {
3758 		rc = ibmvnic_send_crq(adapter, &crq);
3759 		if (rc != H_CLOSED)
3760 			break;
3761 		retries--;
3762 		msleep(50);
3763 
3764 	} while (retries > 0);
3765 
3766 	if (rc) {
3767 		dev_err(dev, "Failed to send init request, rc = %d\n", rc);
3768 		return rc;
3769 	}
3770 
3771 	return 0;
3772 }
3773 
3774 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3775 {
3776 	union ibmvnic_crq crq;
3777 
3778 	memset(&crq, 0, sizeof(crq));
3779 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3780 	crq.version_exchange.cmd = VERSION_EXCHANGE;
3781 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3782 
3783 	return ibmvnic_send_crq(adapter, &crq);
3784 }
3785 
3786 struct vnic_login_client_data {
3787 	u8	type;
3788 	__be16	len;
3789 	char	name[];
3790 } __packed;
3791 
3792 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3793 {
3794 	int len;
3795 
3796 	/* Calculate the amount of buffer space needed for the
3797 	 * vnic client data in the login buffer. There are four entries,
3798 	 * OS name, LPAR name, device name, and a null last entry.
3799 	 */
3800 	len = 4 * sizeof(struct vnic_login_client_data);
3801 	len += 6; /* "Linux" plus NULL */
3802 	len += strlen(utsname()->nodename) + 1;
3803 	len += strlen(adapter->netdev->name) + 1;
3804 
3805 	return len;
3806 }
3807 
3808 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3809 				 struct vnic_login_client_data *vlcd)
3810 {
3811 	const char *os_name = "Linux";
3812 	int len;
3813 
3814 	/* Type 1 - LPAR OS */
3815 	vlcd->type = 1;
3816 	len = strlen(os_name) + 1;
3817 	vlcd->len = cpu_to_be16(len);
3818 	strncpy(vlcd->name, os_name, len);
3819 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3820 
3821 	/* Type 2 - LPAR name */
3822 	vlcd->type = 2;
3823 	len = strlen(utsname()->nodename) + 1;
3824 	vlcd->len = cpu_to_be16(len);
3825 	strncpy(vlcd->name, utsname()->nodename, len);
3826 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3827 
3828 	/* Type 3 - device name */
3829 	vlcd->type = 3;
3830 	len = strlen(adapter->netdev->name) + 1;
3831 	vlcd->len = cpu_to_be16(len);
3832 	strncpy(vlcd->name, adapter->netdev->name, len);
3833 }
3834 
3835 static int send_login(struct ibmvnic_adapter *adapter)
3836 {
3837 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3838 	struct ibmvnic_login_buffer *login_buffer;
3839 	struct device *dev = &adapter->vdev->dev;
3840 	struct vnic_login_client_data *vlcd;
3841 	dma_addr_t rsp_buffer_token;
3842 	dma_addr_t buffer_token;
3843 	size_t rsp_buffer_size;
3844 	union ibmvnic_crq crq;
3845 	int client_data_len;
3846 	size_t buffer_size;
3847 	__be64 *tx_list_p;
3848 	__be64 *rx_list_p;
3849 	int rc;
3850 	int i;
3851 
3852 	if (!adapter->tx_scrq || !adapter->rx_scrq) {
3853 		netdev_err(adapter->netdev,
3854 			   "RX or TX queues are not allocated, device login failed\n");
3855 		return -1;
3856 	}
3857 
3858 	release_login_rsp_buffer(adapter);
3859 	client_data_len = vnic_client_data_len(adapter);
3860 
3861 	buffer_size =
3862 	    sizeof(struct ibmvnic_login_buffer) +
3863 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3864 	    client_data_len;
3865 
3866 	login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3867 	if (!login_buffer)
3868 		goto buf_alloc_failed;
3869 
3870 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3871 				      DMA_TO_DEVICE);
3872 	if (dma_mapping_error(dev, buffer_token)) {
3873 		dev_err(dev, "Couldn't map login buffer\n");
3874 		goto buf_map_failed;
3875 	}
3876 
3877 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3878 			  sizeof(u64) * adapter->req_tx_queues +
3879 			  sizeof(u64) * adapter->req_rx_queues +
3880 			  sizeof(u64) * adapter->req_rx_queues +
3881 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3882 
3883 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3884 	if (!login_rsp_buffer)
3885 		goto buf_rsp_alloc_failed;
3886 
3887 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3888 					  rsp_buffer_size, DMA_FROM_DEVICE);
3889 	if (dma_mapping_error(dev, rsp_buffer_token)) {
3890 		dev_err(dev, "Couldn't map login rsp buffer\n");
3891 		goto buf_rsp_map_failed;
3892 	}
3893 
3894 	adapter->login_buf = login_buffer;
3895 	adapter->login_buf_token = buffer_token;
3896 	adapter->login_buf_sz = buffer_size;
3897 	adapter->login_rsp_buf = login_rsp_buffer;
3898 	adapter->login_rsp_buf_token = rsp_buffer_token;
3899 	adapter->login_rsp_buf_sz = rsp_buffer_size;
3900 
3901 	login_buffer->len = cpu_to_be32(buffer_size);
3902 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3903 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3904 	login_buffer->off_txcomp_subcrqs =
3905 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3906 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3907 	login_buffer->off_rxcomp_subcrqs =
3908 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3909 			sizeof(u64) * adapter->req_tx_queues);
3910 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3911 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3912 
3913 	tx_list_p = (__be64 *)((char *)login_buffer +
3914 				      sizeof(struct ibmvnic_login_buffer));
3915 	rx_list_p = (__be64 *)((char *)login_buffer +
3916 				      sizeof(struct ibmvnic_login_buffer) +
3917 				      sizeof(u64) * adapter->req_tx_queues);
3918 
3919 	for (i = 0; i < adapter->req_tx_queues; i++) {
3920 		if (adapter->tx_scrq[i]) {
3921 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3922 						   crq_num);
3923 		}
3924 	}
3925 
3926 	for (i = 0; i < adapter->req_rx_queues; i++) {
3927 		if (adapter->rx_scrq[i]) {
3928 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3929 						   crq_num);
3930 		}
3931 	}
3932 
3933 	/* Insert vNIC login client data */
3934 	vlcd = (struct vnic_login_client_data *)
3935 		((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3936 	login_buffer->client_data_offset =
3937 			cpu_to_be32((char *)vlcd - (char *)login_buffer);
3938 	login_buffer->client_data_len = cpu_to_be32(client_data_len);
3939 
3940 	vnic_add_client_data(adapter, vlcd);
3941 
3942 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
3943 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3944 		netdev_dbg(adapter->netdev, "%016lx\n",
3945 			   ((unsigned long int *)(adapter->login_buf))[i]);
3946 	}
3947 
3948 	memset(&crq, 0, sizeof(crq));
3949 	crq.login.first = IBMVNIC_CRQ_CMD;
3950 	crq.login.cmd = LOGIN;
3951 	crq.login.ioba = cpu_to_be32(buffer_token);
3952 	crq.login.len = cpu_to_be32(buffer_size);
3953 
3954 	adapter->login_pending = true;
3955 	rc = ibmvnic_send_crq(adapter, &crq);
3956 	if (rc) {
3957 		adapter->login_pending = false;
3958 		netdev_err(adapter->netdev, "Failed to send login, rc=%d\n", rc);
3959 		goto buf_rsp_map_failed;
3960 	}
3961 
3962 	return 0;
3963 
3964 buf_rsp_map_failed:
3965 	kfree(login_rsp_buffer);
3966 	adapter->login_rsp_buf = NULL;
3967 buf_rsp_alloc_failed:
3968 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3969 buf_map_failed:
3970 	kfree(login_buffer);
3971 	adapter->login_buf = NULL;
3972 buf_alloc_failed:
3973 	return -1;
3974 }
3975 
3976 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3977 			    u32 len, u8 map_id)
3978 {
3979 	union ibmvnic_crq crq;
3980 
3981 	memset(&crq, 0, sizeof(crq));
3982 	crq.request_map.first = IBMVNIC_CRQ_CMD;
3983 	crq.request_map.cmd = REQUEST_MAP;
3984 	crq.request_map.map_id = map_id;
3985 	crq.request_map.ioba = cpu_to_be32(addr);
3986 	crq.request_map.len = cpu_to_be32(len);
3987 	return ibmvnic_send_crq(adapter, &crq);
3988 }
3989 
3990 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3991 {
3992 	union ibmvnic_crq crq;
3993 
3994 	memset(&crq, 0, sizeof(crq));
3995 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3996 	crq.request_unmap.cmd = REQUEST_UNMAP;
3997 	crq.request_unmap.map_id = map_id;
3998 	return ibmvnic_send_crq(adapter, &crq);
3999 }
4000 
4001 static void send_query_map(struct ibmvnic_adapter *adapter)
4002 {
4003 	union ibmvnic_crq crq;
4004 
4005 	memset(&crq, 0, sizeof(crq));
4006 	crq.query_map.first = IBMVNIC_CRQ_CMD;
4007 	crq.query_map.cmd = QUERY_MAP;
4008 	ibmvnic_send_crq(adapter, &crq);
4009 }
4010 
4011 /* Send a series of CRQs requesting various capabilities of the VNIC server */
4012 static void send_query_cap(struct ibmvnic_adapter *adapter)
4013 {
4014 	union ibmvnic_crq crq;
4015 
4016 	atomic_set(&adapter->running_cap_crqs, 0);
4017 	memset(&crq, 0, sizeof(crq));
4018 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
4019 	crq.query_capability.cmd = QUERY_CAPABILITY;
4020 
4021 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
4022 	atomic_inc(&adapter->running_cap_crqs);
4023 	ibmvnic_send_crq(adapter, &crq);
4024 
4025 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
4026 	atomic_inc(&adapter->running_cap_crqs);
4027 	ibmvnic_send_crq(adapter, &crq);
4028 
4029 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
4030 	atomic_inc(&adapter->running_cap_crqs);
4031 	ibmvnic_send_crq(adapter, &crq);
4032 
4033 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
4034 	atomic_inc(&adapter->running_cap_crqs);
4035 	ibmvnic_send_crq(adapter, &crq);
4036 
4037 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
4038 	atomic_inc(&adapter->running_cap_crqs);
4039 	ibmvnic_send_crq(adapter, &crq);
4040 
4041 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
4042 	atomic_inc(&adapter->running_cap_crqs);
4043 	ibmvnic_send_crq(adapter, &crq);
4044 
4045 	crq.query_capability.capability =
4046 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
4047 	atomic_inc(&adapter->running_cap_crqs);
4048 	ibmvnic_send_crq(adapter, &crq);
4049 
4050 	crq.query_capability.capability =
4051 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
4052 	atomic_inc(&adapter->running_cap_crqs);
4053 	ibmvnic_send_crq(adapter, &crq);
4054 
4055 	crq.query_capability.capability =
4056 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
4057 	atomic_inc(&adapter->running_cap_crqs);
4058 	ibmvnic_send_crq(adapter, &crq);
4059 
4060 	crq.query_capability.capability =
4061 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
4062 	atomic_inc(&adapter->running_cap_crqs);
4063 	ibmvnic_send_crq(adapter, &crq);
4064 
4065 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
4066 	atomic_inc(&adapter->running_cap_crqs);
4067 	ibmvnic_send_crq(adapter, &crq);
4068 
4069 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
4070 	atomic_inc(&adapter->running_cap_crqs);
4071 	ibmvnic_send_crq(adapter, &crq);
4072 
4073 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
4074 	atomic_inc(&adapter->running_cap_crqs);
4075 	ibmvnic_send_crq(adapter, &crq);
4076 
4077 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
4078 	atomic_inc(&adapter->running_cap_crqs);
4079 	ibmvnic_send_crq(adapter, &crq);
4080 
4081 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
4082 	atomic_inc(&adapter->running_cap_crqs);
4083 	ibmvnic_send_crq(adapter, &crq);
4084 
4085 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
4086 	atomic_inc(&adapter->running_cap_crqs);
4087 	ibmvnic_send_crq(adapter, &crq);
4088 
4089 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
4090 	atomic_inc(&adapter->running_cap_crqs);
4091 	ibmvnic_send_crq(adapter, &crq);
4092 
4093 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
4094 	atomic_inc(&adapter->running_cap_crqs);
4095 	ibmvnic_send_crq(adapter, &crq);
4096 
4097 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
4098 	atomic_inc(&adapter->running_cap_crqs);
4099 	ibmvnic_send_crq(adapter, &crq);
4100 
4101 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
4102 	atomic_inc(&adapter->running_cap_crqs);
4103 	ibmvnic_send_crq(adapter, &crq);
4104 
4105 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
4106 	atomic_inc(&adapter->running_cap_crqs);
4107 	ibmvnic_send_crq(adapter, &crq);
4108 
4109 	crq.query_capability.capability =
4110 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
4111 	atomic_inc(&adapter->running_cap_crqs);
4112 	ibmvnic_send_crq(adapter, &crq);
4113 
4114 	crq.query_capability.capability =
4115 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
4116 	atomic_inc(&adapter->running_cap_crqs);
4117 	ibmvnic_send_crq(adapter, &crq);
4118 
4119 	crq.query_capability.capability =
4120 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
4121 	atomic_inc(&adapter->running_cap_crqs);
4122 	ibmvnic_send_crq(adapter, &crq);
4123 
4124 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
4125 	atomic_inc(&adapter->running_cap_crqs);
4126 	ibmvnic_send_crq(adapter, &crq);
4127 }
4128 
4129 static void send_query_ip_offload(struct ibmvnic_adapter *adapter)
4130 {
4131 	int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4132 	struct device *dev = &adapter->vdev->dev;
4133 	union ibmvnic_crq crq;
4134 
4135 	adapter->ip_offload_tok =
4136 		dma_map_single(dev,
4137 			       &adapter->ip_offload_buf,
4138 			       buf_sz,
4139 			       DMA_FROM_DEVICE);
4140 
4141 	if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4142 		if (!firmware_has_feature(FW_FEATURE_CMO))
4143 			dev_err(dev, "Couldn't map offload buffer\n");
4144 		return;
4145 	}
4146 
4147 	memset(&crq, 0, sizeof(crq));
4148 	crq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4149 	crq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4150 	crq.query_ip_offload.len = cpu_to_be32(buf_sz);
4151 	crq.query_ip_offload.ioba =
4152 	    cpu_to_be32(adapter->ip_offload_tok);
4153 
4154 	ibmvnic_send_crq(adapter, &crq);
4155 }
4156 
4157 static void send_control_ip_offload(struct ibmvnic_adapter *adapter)
4158 {
4159 	struct ibmvnic_control_ip_offload_buffer *ctrl_buf = &adapter->ip_offload_ctrl;
4160 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
4161 	struct device *dev = &adapter->vdev->dev;
4162 	netdev_features_t old_hw_features = 0;
4163 	union ibmvnic_crq crq;
4164 
4165 	adapter->ip_offload_ctrl_tok =
4166 		dma_map_single(dev,
4167 			       ctrl_buf,
4168 			       sizeof(adapter->ip_offload_ctrl),
4169 			       DMA_TO_DEVICE);
4170 
4171 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
4172 		dev_err(dev, "Couldn't map ip offload control buffer\n");
4173 		return;
4174 	}
4175 
4176 	ctrl_buf->len = cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4177 	ctrl_buf->version = cpu_to_be32(INITIAL_VERSION_IOB);
4178 	ctrl_buf->ipv4_chksum = buf->ipv4_chksum;
4179 	ctrl_buf->ipv6_chksum = buf->ipv6_chksum;
4180 	ctrl_buf->tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
4181 	ctrl_buf->udp_ipv4_chksum = buf->udp_ipv4_chksum;
4182 	ctrl_buf->tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
4183 	ctrl_buf->udp_ipv6_chksum = buf->udp_ipv6_chksum;
4184 	ctrl_buf->large_tx_ipv4 = buf->large_tx_ipv4;
4185 	ctrl_buf->large_tx_ipv6 = buf->large_tx_ipv6;
4186 
4187 	/* large_rx disabled for now, additional features needed */
4188 	ctrl_buf->large_rx_ipv4 = 0;
4189 	ctrl_buf->large_rx_ipv6 = 0;
4190 
4191 	if (adapter->state != VNIC_PROBING) {
4192 		old_hw_features = adapter->netdev->hw_features;
4193 		adapter->netdev->hw_features = 0;
4194 	}
4195 
4196 	adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
4197 
4198 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
4199 		adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
4200 
4201 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
4202 		adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
4203 
4204 	if ((adapter->netdev->features &
4205 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
4206 		adapter->netdev->hw_features |= NETIF_F_RXCSUM;
4207 
4208 	if (buf->large_tx_ipv4)
4209 		adapter->netdev->hw_features |= NETIF_F_TSO;
4210 	if (buf->large_tx_ipv6)
4211 		adapter->netdev->hw_features |= NETIF_F_TSO6;
4212 
4213 	if (adapter->state == VNIC_PROBING) {
4214 		adapter->netdev->features |= adapter->netdev->hw_features;
4215 	} else if (old_hw_features != adapter->netdev->hw_features) {
4216 		netdev_features_t tmp = 0;
4217 
4218 		/* disable features no longer supported */
4219 		adapter->netdev->features &= adapter->netdev->hw_features;
4220 		/* turn on features now supported if previously enabled */
4221 		tmp = (old_hw_features ^ adapter->netdev->hw_features) &
4222 			adapter->netdev->hw_features;
4223 		adapter->netdev->features |=
4224 				tmp & adapter->netdev->wanted_features;
4225 	}
4226 
4227 	memset(&crq, 0, sizeof(crq));
4228 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
4229 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
4230 	crq.control_ip_offload.len =
4231 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4232 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
4233 	ibmvnic_send_crq(adapter, &crq);
4234 }
4235 
4236 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
4237 				struct ibmvnic_adapter *adapter)
4238 {
4239 	struct device *dev = &adapter->vdev->dev;
4240 
4241 	if (crq->get_vpd_size_rsp.rc.code) {
4242 		dev_err(dev, "Error retrieving VPD size, rc=%x\n",
4243 			crq->get_vpd_size_rsp.rc.code);
4244 		complete(&adapter->fw_done);
4245 		return;
4246 	}
4247 
4248 	adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
4249 	complete(&adapter->fw_done);
4250 }
4251 
4252 static void handle_vpd_rsp(union ibmvnic_crq *crq,
4253 			   struct ibmvnic_adapter *adapter)
4254 {
4255 	struct device *dev = &adapter->vdev->dev;
4256 	unsigned char *substr = NULL;
4257 	u8 fw_level_len = 0;
4258 
4259 	memset(adapter->fw_version, 0, 32);
4260 
4261 	dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
4262 			 DMA_FROM_DEVICE);
4263 
4264 	if (crq->get_vpd_rsp.rc.code) {
4265 		dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
4266 			crq->get_vpd_rsp.rc.code);
4267 		goto complete;
4268 	}
4269 
4270 	/* get the position of the firmware version info
4271 	 * located after the ASCII 'RM' substring in the buffer
4272 	 */
4273 	substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
4274 	if (!substr) {
4275 		dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
4276 		goto complete;
4277 	}
4278 
4279 	/* get length of firmware level ASCII substring */
4280 	if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
4281 		fw_level_len = *(substr + 2);
4282 	} else {
4283 		dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
4284 		goto complete;
4285 	}
4286 
4287 	/* copy firmware version string from vpd into adapter */
4288 	if ((substr + 3 + fw_level_len) <
4289 	    (adapter->vpd->buff + adapter->vpd->len)) {
4290 		strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
4291 	} else {
4292 		dev_info(dev, "FW substr extrapolated VPD buff\n");
4293 	}
4294 
4295 complete:
4296 	if (adapter->fw_version[0] == '\0')
4297 		strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
4298 	complete(&adapter->fw_done);
4299 }
4300 
4301 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
4302 {
4303 	struct device *dev = &adapter->vdev->dev;
4304 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
4305 	int i;
4306 
4307 	dma_unmap_single(dev, adapter->ip_offload_tok,
4308 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
4309 
4310 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
4311 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
4312 		netdev_dbg(adapter->netdev, "%016lx\n",
4313 			   ((unsigned long int *)(buf))[i]);
4314 
4315 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
4316 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
4317 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
4318 		   buf->tcp_ipv4_chksum);
4319 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
4320 		   buf->tcp_ipv6_chksum);
4321 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
4322 		   buf->udp_ipv4_chksum);
4323 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
4324 		   buf->udp_ipv6_chksum);
4325 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
4326 		   buf->large_tx_ipv4);
4327 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
4328 		   buf->large_tx_ipv6);
4329 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
4330 		   buf->large_rx_ipv4);
4331 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
4332 		   buf->large_rx_ipv6);
4333 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
4334 		   buf->max_ipv4_header_size);
4335 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
4336 		   buf->max_ipv6_header_size);
4337 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
4338 		   buf->max_tcp_header_size);
4339 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
4340 		   buf->max_udp_header_size);
4341 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
4342 		   buf->max_large_tx_size);
4343 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
4344 		   buf->max_large_rx_size);
4345 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
4346 		   buf->ipv6_extension_header);
4347 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
4348 		   buf->tcp_pseudosum_req);
4349 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
4350 		   buf->num_ipv6_ext_headers);
4351 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
4352 		   buf->off_ipv6_ext_headers);
4353 
4354 	send_control_ip_offload(adapter);
4355 }
4356 
4357 static const char *ibmvnic_fw_err_cause(u16 cause)
4358 {
4359 	switch (cause) {
4360 	case ADAPTER_PROBLEM:
4361 		return "adapter problem";
4362 	case BUS_PROBLEM:
4363 		return "bus problem";
4364 	case FW_PROBLEM:
4365 		return "firmware problem";
4366 	case DD_PROBLEM:
4367 		return "device driver problem";
4368 	case EEH_RECOVERY:
4369 		return "EEH recovery";
4370 	case FW_UPDATED:
4371 		return "firmware updated";
4372 	case LOW_MEMORY:
4373 		return "low Memory";
4374 	default:
4375 		return "unknown";
4376 	}
4377 }
4378 
4379 static void handle_error_indication(union ibmvnic_crq *crq,
4380 				    struct ibmvnic_adapter *adapter)
4381 {
4382 	struct device *dev = &adapter->vdev->dev;
4383 	u16 cause;
4384 
4385 	cause = be16_to_cpu(crq->error_indication.error_cause);
4386 
4387 	dev_warn_ratelimited(dev,
4388 			     "Firmware reports %serror, cause: %s. Starting recovery...\n",
4389 			     crq->error_indication.flags
4390 				& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
4391 			     ibmvnic_fw_err_cause(cause));
4392 
4393 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
4394 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4395 	else
4396 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
4397 }
4398 
4399 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
4400 				 struct ibmvnic_adapter *adapter)
4401 {
4402 	struct net_device *netdev = adapter->netdev;
4403 	struct device *dev = &adapter->vdev->dev;
4404 	long rc;
4405 
4406 	rc = crq->change_mac_addr_rsp.rc.code;
4407 	if (rc) {
4408 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
4409 		goto out;
4410 	}
4411 	/* crq->change_mac_addr.mac_addr is the requested one
4412 	 * crq->change_mac_addr_rsp.mac_addr is the returned valid one.
4413 	 */
4414 	ether_addr_copy(netdev->dev_addr,
4415 			&crq->change_mac_addr_rsp.mac_addr[0]);
4416 	ether_addr_copy(adapter->mac_addr,
4417 			&crq->change_mac_addr_rsp.mac_addr[0]);
4418 out:
4419 	complete(&adapter->fw_done);
4420 	return rc;
4421 }
4422 
4423 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
4424 				   struct ibmvnic_adapter *adapter)
4425 {
4426 	struct device *dev = &adapter->vdev->dev;
4427 	u64 *req_value;
4428 	char *name;
4429 
4430 	atomic_dec(&adapter->running_cap_crqs);
4431 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
4432 	case REQ_TX_QUEUES:
4433 		req_value = &adapter->req_tx_queues;
4434 		name = "tx";
4435 		break;
4436 	case REQ_RX_QUEUES:
4437 		req_value = &adapter->req_rx_queues;
4438 		name = "rx";
4439 		break;
4440 	case REQ_RX_ADD_QUEUES:
4441 		req_value = &adapter->req_rx_add_queues;
4442 		name = "rx_add";
4443 		break;
4444 	case REQ_TX_ENTRIES_PER_SUBCRQ:
4445 		req_value = &adapter->req_tx_entries_per_subcrq;
4446 		name = "tx_entries_per_subcrq";
4447 		break;
4448 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
4449 		req_value = &adapter->req_rx_add_entries_per_subcrq;
4450 		name = "rx_add_entries_per_subcrq";
4451 		break;
4452 	case REQ_MTU:
4453 		req_value = &adapter->req_mtu;
4454 		name = "mtu";
4455 		break;
4456 	case PROMISC_REQUESTED:
4457 		req_value = &adapter->promisc;
4458 		name = "promisc";
4459 		break;
4460 	default:
4461 		dev_err(dev, "Got invalid cap request rsp %d\n",
4462 			crq->request_capability.capability);
4463 		return;
4464 	}
4465 
4466 	switch (crq->request_capability_rsp.rc.code) {
4467 	case SUCCESS:
4468 		break;
4469 	case PARTIALSUCCESS:
4470 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
4471 			 *req_value,
4472 			 (long int)be64_to_cpu(crq->request_capability_rsp.
4473 					       number), name);
4474 
4475 		if (be16_to_cpu(crq->request_capability_rsp.capability) ==
4476 		    REQ_MTU) {
4477 			pr_err("mtu of %llu is not supported. Reverting.\n",
4478 			       *req_value);
4479 			*req_value = adapter->fallback.mtu;
4480 		} else {
4481 			*req_value =
4482 				be64_to_cpu(crq->request_capability_rsp.number);
4483 		}
4484 
4485 		send_request_cap(adapter, 1);
4486 		return;
4487 	default:
4488 		dev_err(dev, "Error %d in request cap rsp\n",
4489 			crq->request_capability_rsp.rc.code);
4490 		return;
4491 	}
4492 
4493 	/* Done receiving requested capabilities, query IP offload support */
4494 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4495 		adapter->wait_capability = false;
4496 		send_query_ip_offload(adapter);
4497 	}
4498 }
4499 
4500 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4501 			    struct ibmvnic_adapter *adapter)
4502 {
4503 	struct device *dev = &adapter->vdev->dev;
4504 	struct net_device *netdev = adapter->netdev;
4505 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4506 	struct ibmvnic_login_buffer *login = adapter->login_buf;
4507 	u64 *tx_handle_array;
4508 	u64 *rx_handle_array;
4509 	int num_tx_pools;
4510 	int num_rx_pools;
4511 	u64 *size_array;
4512 	int i;
4513 
4514 	/* CHECK: Test/set of login_pending does not need to be atomic
4515 	 * because only ibmvnic_tasklet tests/clears this.
4516 	 */
4517 	if (!adapter->login_pending) {
4518 		netdev_warn(netdev, "Ignoring unexpected login response\n");
4519 		return 0;
4520 	}
4521 	adapter->login_pending = false;
4522 
4523 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4524 			 DMA_TO_DEVICE);
4525 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
4526 			 adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4527 
4528 	/* If the number of queues requested can't be allocated by the
4529 	 * server, the login response will return with code 1. We will need
4530 	 * to resend the login buffer with fewer queues requested.
4531 	 */
4532 	if (login_rsp_crq->generic.rc.code) {
4533 		adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4534 		complete(&adapter->init_done);
4535 		return 0;
4536 	}
4537 
4538 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4539 
4540 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4541 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4542 		netdev_dbg(adapter->netdev, "%016lx\n",
4543 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4544 	}
4545 
4546 	/* Sanity checks */
4547 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4548 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
4549 	     adapter->req_rx_add_queues !=
4550 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4551 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4552 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4553 		return -EIO;
4554 	}
4555 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4556 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
4557 	/* variable buffer sizes are not supported, so just read the
4558 	 * first entry.
4559 	 */
4560 	adapter->cur_rx_buf_sz = be64_to_cpu(size_array[0]);
4561 
4562 	num_tx_pools = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
4563 	num_rx_pools = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4564 
4565 	tx_handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4566 				  be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
4567 	rx_handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4568 				  be32_to_cpu(adapter->login_rsp_buf->off_rxadd_subcrqs));
4569 
4570 	for (i = 0; i < num_tx_pools; i++)
4571 		adapter->tx_scrq[i]->handle = tx_handle_array[i];
4572 
4573 	for (i = 0; i < num_rx_pools; i++)
4574 		adapter->rx_scrq[i]->handle = rx_handle_array[i];
4575 
4576 	adapter->num_active_tx_scrqs = num_tx_pools;
4577 	adapter->num_active_rx_scrqs = num_rx_pools;
4578 	release_login_rsp_buffer(adapter);
4579 	release_login_buffer(adapter);
4580 	complete(&adapter->init_done);
4581 
4582 	return 0;
4583 }
4584 
4585 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4586 				     struct ibmvnic_adapter *adapter)
4587 {
4588 	struct device *dev = &adapter->vdev->dev;
4589 	long rc;
4590 
4591 	rc = crq->request_unmap_rsp.rc.code;
4592 	if (rc)
4593 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4594 }
4595 
4596 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4597 				 struct ibmvnic_adapter *adapter)
4598 {
4599 	struct net_device *netdev = adapter->netdev;
4600 	struct device *dev = &adapter->vdev->dev;
4601 	long rc;
4602 
4603 	rc = crq->query_map_rsp.rc.code;
4604 	if (rc) {
4605 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4606 		return;
4607 	}
4608 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4609 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4610 		   crq->query_map_rsp.free_pages);
4611 }
4612 
4613 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4614 				 struct ibmvnic_adapter *adapter)
4615 {
4616 	struct net_device *netdev = adapter->netdev;
4617 	struct device *dev = &adapter->vdev->dev;
4618 	long rc;
4619 
4620 	atomic_dec(&adapter->running_cap_crqs);
4621 	netdev_dbg(netdev, "Outstanding queries: %d\n",
4622 		   atomic_read(&adapter->running_cap_crqs));
4623 	rc = crq->query_capability.rc.code;
4624 	if (rc) {
4625 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4626 		goto out;
4627 	}
4628 
4629 	switch (be16_to_cpu(crq->query_capability.capability)) {
4630 	case MIN_TX_QUEUES:
4631 		adapter->min_tx_queues =
4632 		    be64_to_cpu(crq->query_capability.number);
4633 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
4634 			   adapter->min_tx_queues);
4635 		break;
4636 	case MIN_RX_QUEUES:
4637 		adapter->min_rx_queues =
4638 		    be64_to_cpu(crq->query_capability.number);
4639 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
4640 			   adapter->min_rx_queues);
4641 		break;
4642 	case MIN_RX_ADD_QUEUES:
4643 		adapter->min_rx_add_queues =
4644 		    be64_to_cpu(crq->query_capability.number);
4645 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4646 			   adapter->min_rx_add_queues);
4647 		break;
4648 	case MAX_TX_QUEUES:
4649 		adapter->max_tx_queues =
4650 		    be64_to_cpu(crq->query_capability.number);
4651 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
4652 			   adapter->max_tx_queues);
4653 		break;
4654 	case MAX_RX_QUEUES:
4655 		adapter->max_rx_queues =
4656 		    be64_to_cpu(crq->query_capability.number);
4657 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
4658 			   adapter->max_rx_queues);
4659 		break;
4660 	case MAX_RX_ADD_QUEUES:
4661 		adapter->max_rx_add_queues =
4662 		    be64_to_cpu(crq->query_capability.number);
4663 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4664 			   adapter->max_rx_add_queues);
4665 		break;
4666 	case MIN_TX_ENTRIES_PER_SUBCRQ:
4667 		adapter->min_tx_entries_per_subcrq =
4668 		    be64_to_cpu(crq->query_capability.number);
4669 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4670 			   adapter->min_tx_entries_per_subcrq);
4671 		break;
4672 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4673 		adapter->min_rx_add_entries_per_subcrq =
4674 		    be64_to_cpu(crq->query_capability.number);
4675 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4676 			   adapter->min_rx_add_entries_per_subcrq);
4677 		break;
4678 	case MAX_TX_ENTRIES_PER_SUBCRQ:
4679 		adapter->max_tx_entries_per_subcrq =
4680 		    be64_to_cpu(crq->query_capability.number);
4681 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4682 			   adapter->max_tx_entries_per_subcrq);
4683 		break;
4684 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4685 		adapter->max_rx_add_entries_per_subcrq =
4686 		    be64_to_cpu(crq->query_capability.number);
4687 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4688 			   adapter->max_rx_add_entries_per_subcrq);
4689 		break;
4690 	case TCP_IP_OFFLOAD:
4691 		adapter->tcp_ip_offload =
4692 		    be64_to_cpu(crq->query_capability.number);
4693 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4694 			   adapter->tcp_ip_offload);
4695 		break;
4696 	case PROMISC_SUPPORTED:
4697 		adapter->promisc_supported =
4698 		    be64_to_cpu(crq->query_capability.number);
4699 		netdev_dbg(netdev, "promisc_supported = %lld\n",
4700 			   adapter->promisc_supported);
4701 		break;
4702 	case MIN_MTU:
4703 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4704 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4705 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4706 		break;
4707 	case MAX_MTU:
4708 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4709 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4710 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4711 		break;
4712 	case MAX_MULTICAST_FILTERS:
4713 		adapter->max_multicast_filters =
4714 		    be64_to_cpu(crq->query_capability.number);
4715 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4716 			   adapter->max_multicast_filters);
4717 		break;
4718 	case VLAN_HEADER_INSERTION:
4719 		adapter->vlan_header_insertion =
4720 		    be64_to_cpu(crq->query_capability.number);
4721 		if (adapter->vlan_header_insertion)
4722 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4723 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4724 			   adapter->vlan_header_insertion);
4725 		break;
4726 	case RX_VLAN_HEADER_INSERTION:
4727 		adapter->rx_vlan_header_insertion =
4728 		    be64_to_cpu(crq->query_capability.number);
4729 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4730 			   adapter->rx_vlan_header_insertion);
4731 		break;
4732 	case MAX_TX_SG_ENTRIES:
4733 		adapter->max_tx_sg_entries =
4734 		    be64_to_cpu(crq->query_capability.number);
4735 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4736 			   adapter->max_tx_sg_entries);
4737 		break;
4738 	case RX_SG_SUPPORTED:
4739 		adapter->rx_sg_supported =
4740 		    be64_to_cpu(crq->query_capability.number);
4741 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4742 			   adapter->rx_sg_supported);
4743 		break;
4744 	case OPT_TX_COMP_SUB_QUEUES:
4745 		adapter->opt_tx_comp_sub_queues =
4746 		    be64_to_cpu(crq->query_capability.number);
4747 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4748 			   adapter->opt_tx_comp_sub_queues);
4749 		break;
4750 	case OPT_RX_COMP_QUEUES:
4751 		adapter->opt_rx_comp_queues =
4752 		    be64_to_cpu(crq->query_capability.number);
4753 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4754 			   adapter->opt_rx_comp_queues);
4755 		break;
4756 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4757 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
4758 		    be64_to_cpu(crq->query_capability.number);
4759 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4760 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
4761 		break;
4762 	case OPT_TX_ENTRIES_PER_SUBCRQ:
4763 		adapter->opt_tx_entries_per_subcrq =
4764 		    be64_to_cpu(crq->query_capability.number);
4765 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4766 			   adapter->opt_tx_entries_per_subcrq);
4767 		break;
4768 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4769 		adapter->opt_rxba_entries_per_subcrq =
4770 		    be64_to_cpu(crq->query_capability.number);
4771 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4772 			   adapter->opt_rxba_entries_per_subcrq);
4773 		break;
4774 	case TX_RX_DESC_REQ:
4775 		adapter->tx_rx_desc_req = crq->query_capability.number;
4776 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4777 			   adapter->tx_rx_desc_req);
4778 		break;
4779 
4780 	default:
4781 		netdev_err(netdev, "Got invalid cap rsp %d\n",
4782 			   crq->query_capability.capability);
4783 	}
4784 
4785 out:
4786 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4787 		adapter->wait_capability = false;
4788 		send_request_cap(adapter, 0);
4789 	}
4790 }
4791 
4792 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4793 {
4794 	union ibmvnic_crq crq;
4795 	int rc;
4796 
4797 	memset(&crq, 0, sizeof(crq));
4798 	crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4799 	crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4800 
4801 	mutex_lock(&adapter->fw_lock);
4802 	adapter->fw_done_rc = 0;
4803 	reinit_completion(&adapter->fw_done);
4804 
4805 	rc = ibmvnic_send_crq(adapter, &crq);
4806 	if (rc) {
4807 		mutex_unlock(&adapter->fw_lock);
4808 		return rc;
4809 	}
4810 
4811 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
4812 	if (rc) {
4813 		mutex_unlock(&adapter->fw_lock);
4814 		return rc;
4815 	}
4816 
4817 	mutex_unlock(&adapter->fw_lock);
4818 	return adapter->fw_done_rc ? -EIO : 0;
4819 }
4820 
4821 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4822 				       struct ibmvnic_adapter *adapter)
4823 {
4824 	struct net_device *netdev = adapter->netdev;
4825 	int rc;
4826 	__be32 rspeed = cpu_to_be32(crq->query_phys_parms_rsp.speed);
4827 
4828 	rc = crq->query_phys_parms_rsp.rc.code;
4829 	if (rc) {
4830 		netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4831 		return rc;
4832 	}
4833 	switch (rspeed) {
4834 	case IBMVNIC_10MBPS:
4835 		adapter->speed = SPEED_10;
4836 		break;
4837 	case IBMVNIC_100MBPS:
4838 		adapter->speed = SPEED_100;
4839 		break;
4840 	case IBMVNIC_1GBPS:
4841 		adapter->speed = SPEED_1000;
4842 		break;
4843 	case IBMVNIC_10GBPS:
4844 		adapter->speed = SPEED_10000;
4845 		break;
4846 	case IBMVNIC_25GBPS:
4847 		adapter->speed = SPEED_25000;
4848 		break;
4849 	case IBMVNIC_40GBPS:
4850 		adapter->speed = SPEED_40000;
4851 		break;
4852 	case IBMVNIC_50GBPS:
4853 		adapter->speed = SPEED_50000;
4854 		break;
4855 	case IBMVNIC_100GBPS:
4856 		adapter->speed = SPEED_100000;
4857 		break;
4858 	case IBMVNIC_200GBPS:
4859 		adapter->speed = SPEED_200000;
4860 		break;
4861 	default:
4862 		if (netif_carrier_ok(netdev))
4863 			netdev_warn(netdev, "Unknown speed 0x%08x\n", rspeed);
4864 		adapter->speed = SPEED_UNKNOWN;
4865 	}
4866 	if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4867 		adapter->duplex = DUPLEX_FULL;
4868 	else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4869 		adapter->duplex = DUPLEX_HALF;
4870 	else
4871 		adapter->duplex = DUPLEX_UNKNOWN;
4872 
4873 	return rc;
4874 }
4875 
4876 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4877 			       struct ibmvnic_adapter *adapter)
4878 {
4879 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4880 	struct net_device *netdev = adapter->netdev;
4881 	struct device *dev = &adapter->vdev->dev;
4882 	u64 *u64_crq = (u64 *)crq;
4883 	long rc;
4884 
4885 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4886 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
4887 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
4888 	switch (gen_crq->first) {
4889 	case IBMVNIC_CRQ_INIT_RSP:
4890 		switch (gen_crq->cmd) {
4891 		case IBMVNIC_CRQ_INIT:
4892 			dev_info(dev, "Partner initialized\n");
4893 			adapter->from_passive_init = true;
4894 			/* Discard any stale login responses from prev reset.
4895 			 * CHECK: should we clear even on INIT_COMPLETE?
4896 			 */
4897 			adapter->login_pending = false;
4898 
4899 			if (!completion_done(&adapter->init_done)) {
4900 				complete(&adapter->init_done);
4901 				adapter->init_done_rc = -EIO;
4902 			}
4903 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4904 			break;
4905 		case IBMVNIC_CRQ_INIT_COMPLETE:
4906 			dev_info(dev, "Partner initialization complete\n");
4907 			adapter->crq.active = true;
4908 			send_version_xchg(adapter);
4909 			break;
4910 		default:
4911 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4912 		}
4913 		return;
4914 	case IBMVNIC_CRQ_XPORT_EVENT:
4915 		netif_carrier_off(netdev);
4916 		adapter->crq.active = false;
4917 		/* terminate any thread waiting for a response
4918 		 * from the device
4919 		 */
4920 		if (!completion_done(&adapter->fw_done)) {
4921 			adapter->fw_done_rc = -EIO;
4922 			complete(&adapter->fw_done);
4923 		}
4924 		if (!completion_done(&adapter->stats_done))
4925 			complete(&adapter->stats_done);
4926 		if (test_bit(0, &adapter->resetting))
4927 			adapter->force_reset_recovery = true;
4928 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4929 			dev_info(dev, "Migrated, re-enabling adapter\n");
4930 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4931 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4932 			dev_info(dev, "Backing device failover detected\n");
4933 			adapter->failover_pending = true;
4934 		} else {
4935 			/* The adapter lost the connection */
4936 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4937 				gen_crq->cmd);
4938 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4939 		}
4940 		return;
4941 	case IBMVNIC_CRQ_CMD_RSP:
4942 		break;
4943 	default:
4944 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
4945 			gen_crq->first);
4946 		return;
4947 	}
4948 
4949 	switch (gen_crq->cmd) {
4950 	case VERSION_EXCHANGE_RSP:
4951 		rc = crq->version_exchange_rsp.rc.code;
4952 		if (rc) {
4953 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4954 			break;
4955 		}
4956 		ibmvnic_version =
4957 			    be16_to_cpu(crq->version_exchange_rsp.version);
4958 		dev_info(dev, "Partner protocol version is %d\n",
4959 			 ibmvnic_version);
4960 		send_query_cap(adapter);
4961 		break;
4962 	case QUERY_CAPABILITY_RSP:
4963 		handle_query_cap_rsp(crq, adapter);
4964 		break;
4965 	case QUERY_MAP_RSP:
4966 		handle_query_map_rsp(crq, adapter);
4967 		break;
4968 	case REQUEST_MAP_RSP:
4969 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4970 		complete(&adapter->fw_done);
4971 		break;
4972 	case REQUEST_UNMAP_RSP:
4973 		handle_request_unmap_rsp(crq, adapter);
4974 		break;
4975 	case REQUEST_CAPABILITY_RSP:
4976 		handle_request_cap_rsp(crq, adapter);
4977 		break;
4978 	case LOGIN_RSP:
4979 		netdev_dbg(netdev, "Got Login Response\n");
4980 		handle_login_rsp(crq, adapter);
4981 		break;
4982 	case LOGICAL_LINK_STATE_RSP:
4983 		netdev_dbg(netdev,
4984 			   "Got Logical Link State Response, state: %d rc: %d\n",
4985 			   crq->logical_link_state_rsp.link_state,
4986 			   crq->logical_link_state_rsp.rc.code);
4987 		adapter->logical_link_state =
4988 		    crq->logical_link_state_rsp.link_state;
4989 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4990 		complete(&adapter->init_done);
4991 		break;
4992 	case LINK_STATE_INDICATION:
4993 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
4994 		adapter->phys_link_state =
4995 		    crq->link_state_indication.phys_link_state;
4996 		adapter->logical_link_state =
4997 		    crq->link_state_indication.logical_link_state;
4998 		if (adapter->phys_link_state && adapter->logical_link_state)
4999 			netif_carrier_on(netdev);
5000 		else
5001 			netif_carrier_off(netdev);
5002 		break;
5003 	case CHANGE_MAC_ADDR_RSP:
5004 		netdev_dbg(netdev, "Got MAC address change Response\n");
5005 		adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
5006 		break;
5007 	case ERROR_INDICATION:
5008 		netdev_dbg(netdev, "Got Error Indication\n");
5009 		handle_error_indication(crq, adapter);
5010 		break;
5011 	case REQUEST_STATISTICS_RSP:
5012 		netdev_dbg(netdev, "Got Statistics Response\n");
5013 		complete(&adapter->stats_done);
5014 		break;
5015 	case QUERY_IP_OFFLOAD_RSP:
5016 		netdev_dbg(netdev, "Got Query IP offload Response\n");
5017 		handle_query_ip_offload_rsp(adapter);
5018 		break;
5019 	case MULTICAST_CTRL_RSP:
5020 		netdev_dbg(netdev, "Got multicast control Response\n");
5021 		break;
5022 	case CONTROL_IP_OFFLOAD_RSP:
5023 		netdev_dbg(netdev, "Got Control IP offload Response\n");
5024 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
5025 				 sizeof(adapter->ip_offload_ctrl),
5026 				 DMA_TO_DEVICE);
5027 		complete(&adapter->init_done);
5028 		break;
5029 	case COLLECT_FW_TRACE_RSP:
5030 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
5031 		complete(&adapter->fw_done);
5032 		break;
5033 	case GET_VPD_SIZE_RSP:
5034 		handle_vpd_size_rsp(crq, adapter);
5035 		break;
5036 	case GET_VPD_RSP:
5037 		handle_vpd_rsp(crq, adapter);
5038 		break;
5039 	case QUERY_PHYS_PARMS_RSP:
5040 		adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
5041 		complete(&adapter->fw_done);
5042 		break;
5043 	default:
5044 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
5045 			   gen_crq->cmd);
5046 	}
5047 }
5048 
5049 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
5050 {
5051 	struct ibmvnic_adapter *adapter = instance;
5052 
5053 	tasklet_schedule(&adapter->tasklet);
5054 	return IRQ_HANDLED;
5055 }
5056 
5057 static void ibmvnic_tasklet(struct tasklet_struct *t)
5058 {
5059 	struct ibmvnic_adapter *adapter = from_tasklet(adapter, t, tasklet);
5060 	struct ibmvnic_crq_queue *queue = &adapter->crq;
5061 	union ibmvnic_crq *crq;
5062 	unsigned long flags;
5063 	bool done = false;
5064 
5065 	spin_lock_irqsave(&queue->lock, flags);
5066 	while (!done) {
5067 		/* Pull all the valid messages off the CRQ */
5068 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
5069 			ibmvnic_handle_crq(crq, adapter);
5070 			crq->generic.first = 0;
5071 		}
5072 
5073 		/* remain in tasklet until all
5074 		 * capabilities responses are received
5075 		 */
5076 		if (!adapter->wait_capability)
5077 			done = true;
5078 	}
5079 	/* if capabilities CRQ's were sent in this tasklet, the following
5080 	 * tasklet must wait until all responses are received
5081 	 */
5082 	if (atomic_read(&adapter->running_cap_crqs) != 0)
5083 		adapter->wait_capability = true;
5084 	spin_unlock_irqrestore(&queue->lock, flags);
5085 }
5086 
5087 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
5088 {
5089 	struct vio_dev *vdev = adapter->vdev;
5090 	int rc;
5091 
5092 	do {
5093 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
5094 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
5095 
5096 	if (rc)
5097 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
5098 
5099 	return rc;
5100 }
5101 
5102 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
5103 {
5104 	struct ibmvnic_crq_queue *crq = &adapter->crq;
5105 	struct device *dev = &adapter->vdev->dev;
5106 	struct vio_dev *vdev = adapter->vdev;
5107 	int rc;
5108 
5109 	/* Close the CRQ */
5110 	do {
5111 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
5112 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
5113 
5114 	/* Clean out the queue */
5115 	if (!crq->msgs)
5116 		return -EINVAL;
5117 
5118 	memset(crq->msgs, 0, PAGE_SIZE);
5119 	crq->cur = 0;
5120 	crq->active = false;
5121 
5122 	/* And re-open it again */
5123 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
5124 				crq->msg_token, PAGE_SIZE);
5125 
5126 	if (rc == H_CLOSED)
5127 		/* Adapter is good, but other end is not ready */
5128 		dev_warn(dev, "Partner adapter not ready\n");
5129 	else if (rc != 0)
5130 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
5131 
5132 	return rc;
5133 }
5134 
5135 static void release_crq_queue(struct ibmvnic_adapter *adapter)
5136 {
5137 	struct ibmvnic_crq_queue *crq = &adapter->crq;
5138 	struct vio_dev *vdev = adapter->vdev;
5139 	long rc;
5140 
5141 	if (!crq->msgs)
5142 		return;
5143 
5144 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
5145 	free_irq(vdev->irq, adapter);
5146 	tasklet_kill(&adapter->tasklet);
5147 	do {
5148 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
5149 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
5150 
5151 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
5152 			 DMA_BIDIRECTIONAL);
5153 	free_page((unsigned long)crq->msgs);
5154 	crq->msgs = NULL;
5155 	crq->active = false;
5156 }
5157 
5158 static int init_crq_queue(struct ibmvnic_adapter *adapter)
5159 {
5160 	struct ibmvnic_crq_queue *crq = &adapter->crq;
5161 	struct device *dev = &adapter->vdev->dev;
5162 	struct vio_dev *vdev = adapter->vdev;
5163 	int rc, retrc = -ENOMEM;
5164 
5165 	if (crq->msgs)
5166 		return 0;
5167 
5168 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
5169 	/* Should we allocate more than one page? */
5170 
5171 	if (!crq->msgs)
5172 		return -ENOMEM;
5173 
5174 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
5175 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
5176 					DMA_BIDIRECTIONAL);
5177 	if (dma_mapping_error(dev, crq->msg_token))
5178 		goto map_failed;
5179 
5180 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
5181 				crq->msg_token, PAGE_SIZE);
5182 
5183 	if (rc == H_RESOURCE)
5184 		/* maybe kexecing and resource is busy. try a reset */
5185 		rc = ibmvnic_reset_crq(adapter);
5186 	retrc = rc;
5187 
5188 	if (rc == H_CLOSED) {
5189 		dev_warn(dev, "Partner adapter not ready\n");
5190 	} else if (rc) {
5191 		dev_warn(dev, "Error %d opening adapter\n", rc);
5192 		goto reg_crq_failed;
5193 	}
5194 
5195 	retrc = 0;
5196 
5197 	tasklet_setup(&adapter->tasklet, (void *)ibmvnic_tasklet);
5198 
5199 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
5200 	snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
5201 		 adapter->vdev->unit_address);
5202 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
5203 	if (rc) {
5204 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
5205 			vdev->irq, rc);
5206 		goto req_irq_failed;
5207 	}
5208 
5209 	rc = vio_enable_interrupts(vdev);
5210 	if (rc) {
5211 		dev_err(dev, "Error %d enabling interrupts\n", rc);
5212 		goto req_irq_failed;
5213 	}
5214 
5215 	crq->cur = 0;
5216 	spin_lock_init(&crq->lock);
5217 
5218 	return retrc;
5219 
5220 req_irq_failed:
5221 	tasklet_kill(&adapter->tasklet);
5222 	do {
5223 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
5224 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
5225 reg_crq_failed:
5226 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
5227 map_failed:
5228 	free_page((unsigned long)crq->msgs);
5229 	crq->msgs = NULL;
5230 	return retrc;
5231 }
5232 
5233 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter, bool reset)
5234 {
5235 	struct device *dev = &adapter->vdev->dev;
5236 	unsigned long timeout = msecs_to_jiffies(20000);
5237 	u64 old_num_rx_queues, old_num_tx_queues;
5238 	int rc;
5239 
5240 	adapter->from_passive_init = false;
5241 
5242 	if (reset) {
5243 		old_num_rx_queues = adapter->req_rx_queues;
5244 		old_num_tx_queues = adapter->req_tx_queues;
5245 		reinit_completion(&adapter->init_done);
5246 	}
5247 
5248 	adapter->init_done_rc = 0;
5249 	rc = ibmvnic_send_crq_init(adapter);
5250 	if (rc) {
5251 		dev_err(dev, "Send crq init failed with error %d\n", rc);
5252 		return rc;
5253 	}
5254 
5255 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
5256 		dev_err(dev, "Initialization sequence timed out\n");
5257 		return -1;
5258 	}
5259 
5260 	if (adapter->init_done_rc) {
5261 		release_crq_queue(adapter);
5262 		return adapter->init_done_rc;
5263 	}
5264 
5265 	if (adapter->from_passive_init) {
5266 		adapter->state = VNIC_OPEN;
5267 		adapter->from_passive_init = false;
5268 		return -1;
5269 	}
5270 
5271 	if (reset &&
5272 	    test_bit(0, &adapter->resetting) && !adapter->wait_for_reset &&
5273 	    adapter->reset_reason != VNIC_RESET_MOBILITY) {
5274 		if (adapter->req_rx_queues != old_num_rx_queues ||
5275 		    adapter->req_tx_queues != old_num_tx_queues) {
5276 			release_sub_crqs(adapter, 0);
5277 			rc = init_sub_crqs(adapter);
5278 		} else {
5279 			rc = reset_sub_crq_queues(adapter);
5280 		}
5281 	} else {
5282 		rc = init_sub_crqs(adapter);
5283 	}
5284 
5285 	if (rc) {
5286 		dev_err(dev, "Initialization of sub crqs failed\n");
5287 		release_crq_queue(adapter);
5288 		return rc;
5289 	}
5290 
5291 	rc = init_sub_crq_irqs(adapter);
5292 	if (rc) {
5293 		dev_err(dev, "Failed to initialize sub crq irqs\n");
5294 		release_crq_queue(adapter);
5295 	}
5296 
5297 	return rc;
5298 }
5299 
5300 static struct device_attribute dev_attr_failover;
5301 
5302 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
5303 {
5304 	struct ibmvnic_adapter *adapter;
5305 	struct net_device *netdev;
5306 	unsigned char *mac_addr_p;
5307 	int rc;
5308 
5309 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
5310 		dev->unit_address);
5311 
5312 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
5313 							VETH_MAC_ADDR, NULL);
5314 	if (!mac_addr_p) {
5315 		dev_err(&dev->dev,
5316 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
5317 			__FILE__, __LINE__);
5318 		return 0;
5319 	}
5320 
5321 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
5322 				   IBMVNIC_MAX_QUEUES);
5323 	if (!netdev)
5324 		return -ENOMEM;
5325 
5326 	adapter = netdev_priv(netdev);
5327 	adapter->state = VNIC_PROBING;
5328 	dev_set_drvdata(&dev->dev, netdev);
5329 	adapter->vdev = dev;
5330 	adapter->netdev = netdev;
5331 	adapter->login_pending = false;
5332 
5333 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
5334 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
5335 	netdev->irq = dev->irq;
5336 	netdev->netdev_ops = &ibmvnic_netdev_ops;
5337 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
5338 	SET_NETDEV_DEV(netdev, &dev->dev);
5339 
5340 	spin_lock_init(&adapter->stats_lock);
5341 
5342 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
5343 	INIT_DELAYED_WORK(&adapter->ibmvnic_delayed_reset,
5344 			  __ibmvnic_delayed_reset);
5345 	INIT_LIST_HEAD(&adapter->rwi_list);
5346 	spin_lock_init(&adapter->rwi_lock);
5347 	spin_lock_init(&adapter->state_lock);
5348 	mutex_init(&adapter->fw_lock);
5349 	init_completion(&adapter->init_done);
5350 	init_completion(&adapter->fw_done);
5351 	init_completion(&adapter->reset_done);
5352 	init_completion(&adapter->stats_done);
5353 	clear_bit(0, &adapter->resetting);
5354 
5355 	do {
5356 		rc = init_crq_queue(adapter);
5357 		if (rc) {
5358 			dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
5359 				rc);
5360 			goto ibmvnic_init_fail;
5361 		}
5362 
5363 		rc = ibmvnic_reset_init(adapter, false);
5364 		if (rc && rc != EAGAIN)
5365 			goto ibmvnic_init_fail;
5366 	} while (rc == EAGAIN);
5367 
5368 	rc = init_stats_buffers(adapter);
5369 	if (rc)
5370 		goto ibmvnic_init_fail;
5371 
5372 	rc = init_stats_token(adapter);
5373 	if (rc)
5374 		goto ibmvnic_stats_fail;
5375 
5376 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
5377 	netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
5378 	netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
5379 
5380 	rc = device_create_file(&dev->dev, &dev_attr_failover);
5381 	if (rc)
5382 		goto ibmvnic_dev_file_err;
5383 
5384 	netif_carrier_off(netdev);
5385 	rc = register_netdev(netdev);
5386 	if (rc) {
5387 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
5388 		goto ibmvnic_register_fail;
5389 	}
5390 	dev_info(&dev->dev, "ibmvnic registered\n");
5391 
5392 	adapter->state = VNIC_PROBED;
5393 
5394 	adapter->wait_for_reset = false;
5395 	adapter->last_reset_time = jiffies;
5396 	return 0;
5397 
5398 ibmvnic_register_fail:
5399 	device_remove_file(&dev->dev, &dev_attr_failover);
5400 
5401 ibmvnic_dev_file_err:
5402 	release_stats_token(adapter);
5403 
5404 ibmvnic_stats_fail:
5405 	release_stats_buffers(adapter);
5406 
5407 ibmvnic_init_fail:
5408 	release_sub_crqs(adapter, 1);
5409 	release_crq_queue(adapter);
5410 	mutex_destroy(&adapter->fw_lock);
5411 	free_netdev(netdev);
5412 
5413 	return rc;
5414 }
5415 
5416 static int ibmvnic_remove(struct vio_dev *dev)
5417 {
5418 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
5419 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5420 	unsigned long flags;
5421 
5422 	spin_lock_irqsave(&adapter->state_lock, flags);
5423 	if (test_bit(0, &adapter->resetting)) {
5424 		spin_unlock_irqrestore(&adapter->state_lock, flags);
5425 		return -EBUSY;
5426 	}
5427 
5428 	adapter->state = VNIC_REMOVING;
5429 	spin_unlock_irqrestore(&adapter->state_lock, flags);
5430 
5431 	flush_work(&adapter->ibmvnic_reset);
5432 	flush_delayed_work(&adapter->ibmvnic_delayed_reset);
5433 
5434 	rtnl_lock();
5435 	unregister_netdevice(netdev);
5436 
5437 	release_resources(adapter);
5438 	release_sub_crqs(adapter, 1);
5439 	release_crq_queue(adapter);
5440 
5441 	release_stats_token(adapter);
5442 	release_stats_buffers(adapter);
5443 
5444 	adapter->state = VNIC_REMOVED;
5445 
5446 	rtnl_unlock();
5447 	mutex_destroy(&adapter->fw_lock);
5448 	device_remove_file(&dev->dev, &dev_attr_failover);
5449 	free_netdev(netdev);
5450 	dev_set_drvdata(&dev->dev, NULL);
5451 
5452 	return 0;
5453 }
5454 
5455 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
5456 			      const char *buf, size_t count)
5457 {
5458 	struct net_device *netdev = dev_get_drvdata(dev);
5459 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5460 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
5461 	__be64 session_token;
5462 	long rc;
5463 
5464 	if (!sysfs_streq(buf, "1"))
5465 		return -EINVAL;
5466 
5467 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
5468 			 H_GET_SESSION_TOKEN, 0, 0, 0);
5469 	if (rc) {
5470 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
5471 			   rc);
5472 		return -EINVAL;
5473 	}
5474 
5475 	session_token = (__be64)retbuf[0];
5476 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
5477 		   be64_to_cpu(session_token));
5478 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
5479 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
5480 	if (rc) {
5481 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
5482 			   rc);
5483 		return -EINVAL;
5484 	}
5485 
5486 	return count;
5487 }
5488 
5489 static DEVICE_ATTR_WO(failover);
5490 
5491 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
5492 {
5493 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
5494 	struct ibmvnic_adapter *adapter;
5495 	struct iommu_table *tbl;
5496 	unsigned long ret = 0;
5497 	int i;
5498 
5499 	tbl = get_iommu_table_base(&vdev->dev);
5500 
5501 	/* netdev inits at probe time along with the structures we need below*/
5502 	if (!netdev)
5503 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
5504 
5505 	adapter = netdev_priv(netdev);
5506 
5507 	ret += PAGE_SIZE; /* the crq message queue */
5508 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
5509 
5510 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
5511 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
5512 
5513 	for (i = 0; i < adapter->num_active_rx_pools; i++)
5514 		ret += adapter->rx_pool[i].size *
5515 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
5516 
5517 	return ret;
5518 }
5519 
5520 static int ibmvnic_resume(struct device *dev)
5521 {
5522 	struct net_device *netdev = dev_get_drvdata(dev);
5523 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5524 
5525 	if (adapter->state != VNIC_OPEN)
5526 		return 0;
5527 
5528 	tasklet_schedule(&adapter->tasklet);
5529 
5530 	return 0;
5531 }
5532 
5533 static const struct vio_device_id ibmvnic_device_table[] = {
5534 	{"network", "IBM,vnic"},
5535 	{"", "" }
5536 };
5537 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5538 
5539 static const struct dev_pm_ops ibmvnic_pm_ops = {
5540 	.resume = ibmvnic_resume
5541 };
5542 
5543 static struct vio_driver ibmvnic_driver = {
5544 	.id_table       = ibmvnic_device_table,
5545 	.probe          = ibmvnic_probe,
5546 	.remove         = ibmvnic_remove,
5547 	.get_desired_dma = ibmvnic_get_desired_dma,
5548 	.name		= ibmvnic_driver_name,
5549 	.pm		= &ibmvnic_pm_ops,
5550 };
5551 
5552 /* module functions */
5553 static int __init ibmvnic_module_init(void)
5554 {
5555 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5556 		IBMVNIC_DRIVER_VERSION);
5557 
5558 	return vio_register_driver(&ibmvnic_driver);
5559 }
5560 
5561 static void __exit ibmvnic_module_exit(void)
5562 {
5563 	vio_unregister_driver(&ibmvnic_driver);
5564 }
5565 
5566 module_init(ibmvnic_module_init);
5567 module_exit(ibmvnic_module_exit);
5568