xref: /openbmc/linux/drivers/net/ethernet/ibm/ibmvnic.c (revision 4da722ca)
1 /**************************************************************************/
2 /*                                                                        */
3 /*  IBM System i and System p Virtual NIC Device Driver                   */
4 /*  Copyright (C) 2014 IBM Corp.                                          */
5 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
6 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
7 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
8 /*                                                                        */
9 /*  This program is free software; you can redistribute it and/or modify  */
10 /*  it under the terms of the GNU General Public License as published by  */
11 /*  the Free Software Foundation; either version 2 of the License, or     */
12 /*  (at your option) any later version.                                   */
13 /*                                                                        */
14 /*  This program is distributed in the hope that it will be useful,       */
15 /*  but WITHOUT ANY WARRANTY; without even the implied warranty of        */
16 /*  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         */
17 /*  GNU General Public License for more details.                          */
18 /*                                                                        */
19 /*  You should have received a copy of the GNU General Public License     */
20 /*  along with this program.                                              */
21 /*                                                                        */
22 /* This module contains the implementation of a virtual ethernet device   */
23 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
24 /* option of the RS/6000 Platform Architecture to interface with virtual  */
25 /* ethernet NICs that are presented to the partition by the hypervisor.   */
26 /*									   */
27 /* Messages are passed between the VNIC driver and the VNIC server using  */
28 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
29 /* issue and receive commands that initiate communication with the server */
30 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
31 /* are used by the driver to notify the server that a packet is           */
32 /* ready for transmission or that a buffer has been added to receive a    */
33 /* packet. Subsequently, sCRQs are used by the server to notify the       */
34 /* driver that a packet transmission has been completed or that a packet  */
35 /* has been received and placed in a waiting buffer.                      */
36 /*                                                                        */
37 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
38 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
39 /* or receive has been completed, the VNIC driver is required to use      */
40 /* "long term mapping". This entails that large, continuous DMA mapped    */
41 /* buffers are allocated on driver initialization and these buffers are   */
42 /* then continuously reused to pass skbs to and from the VNIC server.     */
43 /*                                                                        */
44 /**************************************************************************/
45 
46 #include <linux/module.h>
47 #include <linux/moduleparam.h>
48 #include <linux/types.h>
49 #include <linux/errno.h>
50 #include <linux/completion.h>
51 #include <linux/ioport.h>
52 #include <linux/dma-mapping.h>
53 #include <linux/kernel.h>
54 #include <linux/netdevice.h>
55 #include <linux/etherdevice.h>
56 #include <linux/skbuff.h>
57 #include <linux/init.h>
58 #include <linux/delay.h>
59 #include <linux/mm.h>
60 #include <linux/ethtool.h>
61 #include <linux/proc_fs.h>
62 #include <linux/in.h>
63 #include <linux/ip.h>
64 #include <linux/ipv6.h>
65 #include <linux/irq.h>
66 #include <linux/kthread.h>
67 #include <linux/seq_file.h>
68 #include <linux/interrupt.h>
69 #include <net/net_namespace.h>
70 #include <asm/hvcall.h>
71 #include <linux/atomic.h>
72 #include <asm/vio.h>
73 #include <asm/iommu.h>
74 #include <linux/uaccess.h>
75 #include <asm/firmware.h>
76 #include <linux/workqueue.h>
77 #include <linux/if_vlan.h>
78 
79 #include "ibmvnic.h"
80 
81 static const char ibmvnic_driver_name[] = "ibmvnic";
82 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
83 
84 MODULE_AUTHOR("Santiago Leon");
85 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
86 MODULE_LICENSE("GPL");
87 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
88 
89 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
90 static int ibmvnic_remove(struct vio_dev *);
91 static void release_sub_crqs(struct ibmvnic_adapter *);
92 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
93 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
94 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
95 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
96 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
97 		       union sub_crq *sub_crq);
98 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
99 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
100 static int enable_scrq_irq(struct ibmvnic_adapter *,
101 			   struct ibmvnic_sub_crq_queue *);
102 static int disable_scrq_irq(struct ibmvnic_adapter *,
103 			    struct ibmvnic_sub_crq_queue *);
104 static int pending_scrq(struct ibmvnic_adapter *,
105 			struct ibmvnic_sub_crq_queue *);
106 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
107 					struct ibmvnic_sub_crq_queue *);
108 static int ibmvnic_poll(struct napi_struct *napi, int data);
109 static void send_map_query(struct ibmvnic_adapter *adapter);
110 static void send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
111 static void send_request_unmap(struct ibmvnic_adapter *, u8);
112 static void send_login(struct ibmvnic_adapter *adapter);
113 static void send_cap_queries(struct ibmvnic_adapter *adapter);
114 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
115 static int ibmvnic_init(struct ibmvnic_adapter *);
116 static void release_crq_queue(struct ibmvnic_adapter *);
117 
118 struct ibmvnic_stat {
119 	char name[ETH_GSTRING_LEN];
120 	int offset;
121 };
122 
123 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
124 			     offsetof(struct ibmvnic_statistics, stat))
125 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
126 
127 static const struct ibmvnic_stat ibmvnic_stats[] = {
128 	{"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
129 	{"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
130 	{"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
131 	{"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
132 	{"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
133 	{"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
134 	{"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
135 	{"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
136 	{"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
137 	{"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
138 	{"align_errors", IBMVNIC_STAT_OFF(align_errors)},
139 	{"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
140 	{"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
141 	{"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
142 	{"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
143 	{"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
144 	{"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
145 	{"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
146 	{"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
147 	{"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
148 	{"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
149 	{"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
150 };
151 
152 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
153 			  unsigned long length, unsigned long *number,
154 			  unsigned long *irq)
155 {
156 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
157 	long rc;
158 
159 	rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
160 	*number = retbuf[0];
161 	*irq = retbuf[1];
162 
163 	return rc;
164 }
165 
166 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
167 				struct ibmvnic_long_term_buff *ltb, int size)
168 {
169 	struct device *dev = &adapter->vdev->dev;
170 
171 	ltb->size = size;
172 	ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
173 				       GFP_KERNEL);
174 
175 	if (!ltb->buff) {
176 		dev_err(dev, "Couldn't alloc long term buffer\n");
177 		return -ENOMEM;
178 	}
179 	ltb->map_id = adapter->map_id;
180 	adapter->map_id++;
181 
182 	init_completion(&adapter->fw_done);
183 	send_request_map(adapter, ltb->addr,
184 			 ltb->size, ltb->map_id);
185 	wait_for_completion(&adapter->fw_done);
186 
187 	if (adapter->fw_done_rc) {
188 		dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
189 			adapter->fw_done_rc);
190 		return -1;
191 	}
192 	return 0;
193 }
194 
195 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
196 				struct ibmvnic_long_term_buff *ltb)
197 {
198 	struct device *dev = &adapter->vdev->dev;
199 
200 	if (!ltb->buff)
201 		return;
202 
203 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
204 	    adapter->reset_reason != VNIC_RESET_MOBILITY)
205 		send_request_unmap(adapter, ltb->map_id);
206 	dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
207 }
208 
209 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
210 				struct ibmvnic_long_term_buff *ltb)
211 {
212 	memset(ltb->buff, 0, ltb->size);
213 
214 	init_completion(&adapter->fw_done);
215 	send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
216 	wait_for_completion(&adapter->fw_done);
217 
218 	if (adapter->fw_done_rc) {
219 		dev_info(&adapter->vdev->dev,
220 			 "Reset failed, attempting to free and reallocate buffer\n");
221 		free_long_term_buff(adapter, ltb);
222 		return alloc_long_term_buff(adapter, ltb, ltb->size);
223 	}
224 	return 0;
225 }
226 
227 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
228 {
229 	int i;
230 
231 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
232 	     i++)
233 		adapter->rx_pool[i].active = 0;
234 }
235 
236 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
237 			      struct ibmvnic_rx_pool *pool)
238 {
239 	int count = pool->size - atomic_read(&pool->available);
240 	struct device *dev = &adapter->vdev->dev;
241 	int buffers_added = 0;
242 	unsigned long lpar_rc;
243 	union sub_crq sub_crq;
244 	struct sk_buff *skb;
245 	unsigned int offset;
246 	dma_addr_t dma_addr;
247 	unsigned char *dst;
248 	u64 *handle_array;
249 	int shift = 0;
250 	int index;
251 	int i;
252 
253 	if (!pool->active)
254 		return;
255 
256 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
257 				      be32_to_cpu(adapter->login_rsp_buf->
258 				      off_rxadd_subcrqs));
259 
260 	for (i = 0; i < count; ++i) {
261 		skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
262 		if (!skb) {
263 			dev_err(dev, "Couldn't replenish rx buff\n");
264 			adapter->replenish_no_mem++;
265 			break;
266 		}
267 
268 		index = pool->free_map[pool->next_free];
269 
270 		if (pool->rx_buff[index].skb)
271 			dev_err(dev, "Inconsistent free_map!\n");
272 
273 		/* Copy the skb to the long term mapped DMA buffer */
274 		offset = index * pool->buff_size;
275 		dst = pool->long_term_buff.buff + offset;
276 		memset(dst, 0, pool->buff_size);
277 		dma_addr = pool->long_term_buff.addr + offset;
278 		pool->rx_buff[index].data = dst;
279 
280 		pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
281 		pool->rx_buff[index].dma = dma_addr;
282 		pool->rx_buff[index].skb = skb;
283 		pool->rx_buff[index].pool_index = pool->index;
284 		pool->rx_buff[index].size = pool->buff_size;
285 
286 		memset(&sub_crq, 0, sizeof(sub_crq));
287 		sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
288 		sub_crq.rx_add.correlator =
289 		    cpu_to_be64((u64)&pool->rx_buff[index]);
290 		sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
291 		sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
292 
293 		/* The length field of the sCRQ is defined to be 24 bits so the
294 		 * buffer size needs to be left shifted by a byte before it is
295 		 * converted to big endian to prevent the last byte from being
296 		 * truncated.
297 		 */
298 #ifdef __LITTLE_ENDIAN__
299 		shift = 8;
300 #endif
301 		sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
302 
303 		lpar_rc = send_subcrq(adapter, handle_array[pool->index],
304 				      &sub_crq);
305 		if (lpar_rc != H_SUCCESS)
306 			goto failure;
307 
308 		buffers_added++;
309 		adapter->replenish_add_buff_success++;
310 		pool->next_free = (pool->next_free + 1) % pool->size;
311 	}
312 	atomic_add(buffers_added, &pool->available);
313 	return;
314 
315 failure:
316 	dev_info(dev, "replenish pools failure\n");
317 	pool->free_map[pool->next_free] = index;
318 	pool->rx_buff[index].skb = NULL;
319 	if (!dma_mapping_error(dev, dma_addr))
320 		dma_unmap_single(dev, dma_addr, pool->buff_size,
321 				 DMA_FROM_DEVICE);
322 
323 	dev_kfree_skb_any(skb);
324 	adapter->replenish_add_buff_failure++;
325 	atomic_add(buffers_added, &pool->available);
326 
327 	if (lpar_rc == H_CLOSED) {
328 		/* Disable buffer pool replenishment and report carrier off if
329 		 * queue is closed. Firmware guarantees that a signal will
330 		 * be sent to the driver, triggering a reset.
331 		 */
332 		deactivate_rx_pools(adapter);
333 		netif_carrier_off(adapter->netdev);
334 	}
335 }
336 
337 static void replenish_pools(struct ibmvnic_adapter *adapter)
338 {
339 	int i;
340 
341 	adapter->replenish_task_cycles++;
342 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
343 	     i++) {
344 		if (adapter->rx_pool[i].active)
345 			replenish_rx_pool(adapter, &adapter->rx_pool[i]);
346 	}
347 }
348 
349 static void release_stats_token(struct ibmvnic_adapter *adapter)
350 {
351 	struct device *dev = &adapter->vdev->dev;
352 
353 	if (!adapter->stats_token)
354 		return;
355 
356 	dma_unmap_single(dev, adapter->stats_token,
357 			 sizeof(struct ibmvnic_statistics),
358 			 DMA_FROM_DEVICE);
359 	adapter->stats_token = 0;
360 }
361 
362 static int init_stats_token(struct ibmvnic_adapter *adapter)
363 {
364 	struct device *dev = &adapter->vdev->dev;
365 	dma_addr_t stok;
366 
367 	stok = dma_map_single(dev, &adapter->stats,
368 			      sizeof(struct ibmvnic_statistics),
369 			      DMA_FROM_DEVICE);
370 	if (dma_mapping_error(dev, stok)) {
371 		dev_err(dev, "Couldn't map stats buffer\n");
372 		return -1;
373 	}
374 
375 	adapter->stats_token = stok;
376 	return 0;
377 }
378 
379 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
380 {
381 	struct ibmvnic_rx_pool *rx_pool;
382 	int rx_scrqs;
383 	int i, j, rc;
384 
385 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
386 	for (i = 0; i < rx_scrqs; i++) {
387 		rx_pool = &adapter->rx_pool[i];
388 
389 		rc = reset_long_term_buff(adapter, &rx_pool->long_term_buff);
390 		if (rc)
391 			return rc;
392 
393 		for (j = 0; j < rx_pool->size; j++)
394 			rx_pool->free_map[j] = j;
395 
396 		memset(rx_pool->rx_buff, 0,
397 		       rx_pool->size * sizeof(struct ibmvnic_rx_buff));
398 
399 		atomic_set(&rx_pool->available, 0);
400 		rx_pool->next_alloc = 0;
401 		rx_pool->next_free = 0;
402 		rx_pool->active = 1;
403 	}
404 
405 	return 0;
406 }
407 
408 static void release_rx_pools(struct ibmvnic_adapter *adapter)
409 {
410 	struct ibmvnic_rx_pool *rx_pool;
411 	int rx_scrqs;
412 	int i, j;
413 
414 	if (!adapter->rx_pool)
415 		return;
416 
417 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
418 	for (i = 0; i < rx_scrqs; i++) {
419 		rx_pool = &adapter->rx_pool[i];
420 
421 		kfree(rx_pool->free_map);
422 		free_long_term_buff(adapter, &rx_pool->long_term_buff);
423 
424 		if (!rx_pool->rx_buff)
425 			continue;
426 
427 		for (j = 0; j < rx_pool->size; j++) {
428 			if (rx_pool->rx_buff[j].skb) {
429 				dev_kfree_skb_any(rx_pool->rx_buff[i].skb);
430 				rx_pool->rx_buff[i].skb = NULL;
431 			}
432 		}
433 
434 		kfree(rx_pool->rx_buff);
435 	}
436 
437 	kfree(adapter->rx_pool);
438 	adapter->rx_pool = NULL;
439 }
440 
441 static int init_rx_pools(struct net_device *netdev)
442 {
443 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
444 	struct device *dev = &adapter->vdev->dev;
445 	struct ibmvnic_rx_pool *rx_pool;
446 	int rxadd_subcrqs;
447 	u64 *size_array;
448 	int i, j;
449 
450 	rxadd_subcrqs =
451 		be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
452 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
453 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
454 
455 	adapter->rx_pool = kcalloc(rxadd_subcrqs,
456 				   sizeof(struct ibmvnic_rx_pool),
457 				   GFP_KERNEL);
458 	if (!adapter->rx_pool) {
459 		dev_err(dev, "Failed to allocate rx pools\n");
460 		return -1;
461 	}
462 
463 	for (i = 0; i < rxadd_subcrqs; i++) {
464 		rx_pool = &adapter->rx_pool[i];
465 
466 		netdev_dbg(adapter->netdev,
467 			   "Initializing rx_pool %d, %lld buffs, %lld bytes each\n",
468 			   i, adapter->req_rx_add_entries_per_subcrq,
469 			   be64_to_cpu(size_array[i]));
470 
471 		rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
472 		rx_pool->index = i;
473 		rx_pool->buff_size = be64_to_cpu(size_array[i]);
474 		rx_pool->active = 1;
475 
476 		rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
477 					    GFP_KERNEL);
478 		if (!rx_pool->free_map) {
479 			release_rx_pools(adapter);
480 			return -1;
481 		}
482 
483 		rx_pool->rx_buff = kcalloc(rx_pool->size,
484 					   sizeof(struct ibmvnic_rx_buff),
485 					   GFP_KERNEL);
486 		if (!rx_pool->rx_buff) {
487 			dev_err(dev, "Couldn't alloc rx buffers\n");
488 			release_rx_pools(adapter);
489 			return -1;
490 		}
491 
492 		if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
493 					 rx_pool->size * rx_pool->buff_size)) {
494 			release_rx_pools(adapter);
495 			return -1;
496 		}
497 
498 		for (j = 0; j < rx_pool->size; ++j)
499 			rx_pool->free_map[j] = j;
500 
501 		atomic_set(&rx_pool->available, 0);
502 		rx_pool->next_alloc = 0;
503 		rx_pool->next_free = 0;
504 	}
505 
506 	return 0;
507 }
508 
509 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
510 {
511 	struct ibmvnic_tx_pool *tx_pool;
512 	int tx_scrqs;
513 	int i, j, rc;
514 
515 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
516 	for (i = 0; i < tx_scrqs; i++) {
517 		tx_pool = &adapter->tx_pool[i];
518 
519 		rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
520 		if (rc)
521 			return rc;
522 
523 		memset(tx_pool->tx_buff, 0,
524 		       adapter->req_tx_entries_per_subcrq *
525 		       sizeof(struct ibmvnic_tx_buff));
526 
527 		for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
528 			tx_pool->free_map[j] = j;
529 
530 		tx_pool->consumer_index = 0;
531 		tx_pool->producer_index = 0;
532 	}
533 
534 	return 0;
535 }
536 
537 static void release_tx_pools(struct ibmvnic_adapter *adapter)
538 {
539 	struct ibmvnic_tx_pool *tx_pool;
540 	int i, tx_scrqs;
541 
542 	if (!adapter->tx_pool)
543 		return;
544 
545 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
546 	for (i = 0; i < tx_scrqs; i++) {
547 		tx_pool = &adapter->tx_pool[i];
548 		kfree(tx_pool->tx_buff);
549 		free_long_term_buff(adapter, &tx_pool->long_term_buff);
550 		kfree(tx_pool->free_map);
551 	}
552 
553 	kfree(adapter->tx_pool);
554 	adapter->tx_pool = NULL;
555 }
556 
557 static int init_tx_pools(struct net_device *netdev)
558 {
559 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
560 	struct device *dev = &adapter->vdev->dev;
561 	struct ibmvnic_tx_pool *tx_pool;
562 	int tx_subcrqs;
563 	int i, j;
564 
565 	tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
566 	adapter->tx_pool = kcalloc(tx_subcrqs,
567 				   sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
568 	if (!adapter->tx_pool)
569 		return -1;
570 
571 	for (i = 0; i < tx_subcrqs; i++) {
572 		tx_pool = &adapter->tx_pool[i];
573 		tx_pool->tx_buff = kcalloc(adapter->req_tx_entries_per_subcrq,
574 					   sizeof(struct ibmvnic_tx_buff),
575 					   GFP_KERNEL);
576 		if (!tx_pool->tx_buff) {
577 			dev_err(dev, "tx pool buffer allocation failed\n");
578 			release_tx_pools(adapter);
579 			return -1;
580 		}
581 
582 		if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
583 					 adapter->req_tx_entries_per_subcrq *
584 					 adapter->req_mtu)) {
585 			release_tx_pools(adapter);
586 			return -1;
587 		}
588 
589 		tx_pool->free_map = kcalloc(adapter->req_tx_entries_per_subcrq,
590 					    sizeof(int), GFP_KERNEL);
591 		if (!tx_pool->free_map) {
592 			release_tx_pools(adapter);
593 			return -1;
594 		}
595 
596 		for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
597 			tx_pool->free_map[j] = j;
598 
599 		tx_pool->consumer_index = 0;
600 		tx_pool->producer_index = 0;
601 	}
602 
603 	return 0;
604 }
605 
606 static void release_error_buffers(struct ibmvnic_adapter *adapter)
607 {
608 	struct device *dev = &adapter->vdev->dev;
609 	struct ibmvnic_error_buff *error_buff, *tmp;
610 	unsigned long flags;
611 
612 	spin_lock_irqsave(&adapter->error_list_lock, flags);
613 	list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list) {
614 		list_del(&error_buff->list);
615 		dma_unmap_single(dev, error_buff->dma, error_buff->len,
616 				 DMA_FROM_DEVICE);
617 		kfree(error_buff->buff);
618 		kfree(error_buff);
619 	}
620 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
621 }
622 
623 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
624 {
625 	int i;
626 
627 	if (adapter->napi_enabled)
628 		return;
629 
630 	for (i = 0; i < adapter->req_rx_queues; i++)
631 		napi_enable(&adapter->napi[i]);
632 
633 	adapter->napi_enabled = true;
634 }
635 
636 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
637 {
638 	int i;
639 
640 	if (!adapter->napi_enabled)
641 		return;
642 
643 	for (i = 0; i < adapter->req_rx_queues; i++)
644 		napi_disable(&adapter->napi[i]);
645 
646 	adapter->napi_enabled = false;
647 }
648 
649 static int ibmvnic_login(struct net_device *netdev)
650 {
651 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
652 	unsigned long timeout = msecs_to_jiffies(30000);
653 	struct device *dev = &adapter->vdev->dev;
654 
655 	do {
656 		if (adapter->renegotiate) {
657 			adapter->renegotiate = false;
658 			release_sub_crqs(adapter);
659 
660 			reinit_completion(&adapter->init_done);
661 			send_cap_queries(adapter);
662 			if (!wait_for_completion_timeout(&adapter->init_done,
663 							 timeout)) {
664 				dev_err(dev, "Capabilities query timeout\n");
665 				return -1;
666 			}
667 		}
668 
669 		reinit_completion(&adapter->init_done);
670 		send_login(adapter);
671 		if (!wait_for_completion_timeout(&adapter->init_done,
672 						 timeout)) {
673 			dev_err(dev, "Login timeout\n");
674 			return -1;
675 		}
676 	} while (adapter->renegotiate);
677 
678 	return 0;
679 }
680 
681 static void release_resources(struct ibmvnic_adapter *adapter)
682 {
683 	int i;
684 
685 	release_tx_pools(adapter);
686 	release_rx_pools(adapter);
687 
688 	release_stats_token(adapter);
689 	release_error_buffers(adapter);
690 
691 	if (adapter->napi) {
692 		for (i = 0; i < adapter->req_rx_queues; i++) {
693 			if (&adapter->napi[i])
694 				netif_napi_del(&adapter->napi[i]);
695 		}
696 	}
697 }
698 
699 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
700 {
701 	struct net_device *netdev = adapter->netdev;
702 	unsigned long timeout = msecs_to_jiffies(30000);
703 	union ibmvnic_crq crq;
704 	bool resend;
705 	int rc;
706 
707 	netdev_err(netdev, "setting link state %d\n", link_state);
708 	memset(&crq, 0, sizeof(crq));
709 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
710 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
711 	crq.logical_link_state.link_state = link_state;
712 
713 	do {
714 		resend = false;
715 
716 		reinit_completion(&adapter->init_done);
717 		rc = ibmvnic_send_crq(adapter, &crq);
718 		if (rc) {
719 			netdev_err(netdev, "Failed to set link state\n");
720 			return rc;
721 		}
722 
723 		if (!wait_for_completion_timeout(&adapter->init_done,
724 						 timeout)) {
725 			netdev_err(netdev, "timeout setting link state\n");
726 			return -1;
727 		}
728 
729 		if (adapter->init_done_rc == 1) {
730 			/* Partuial success, delay and re-send */
731 			mdelay(1000);
732 			resend = true;
733 		}
734 	} while (resend);
735 
736 	return 0;
737 }
738 
739 static int set_real_num_queues(struct net_device *netdev)
740 {
741 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
742 	int rc;
743 
744 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
745 	if (rc) {
746 		netdev_err(netdev, "failed to set the number of tx queues\n");
747 		return rc;
748 	}
749 
750 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
751 	if (rc)
752 		netdev_err(netdev, "failed to set the number of rx queues\n");
753 
754 	return rc;
755 }
756 
757 static int init_resources(struct ibmvnic_adapter *adapter)
758 {
759 	struct net_device *netdev = adapter->netdev;
760 	int i, rc;
761 
762 	rc = set_real_num_queues(netdev);
763 	if (rc)
764 		return rc;
765 
766 	rc = init_stats_token(adapter);
767 	if (rc)
768 		return rc;
769 
770 	adapter->map_id = 1;
771 	adapter->napi = kcalloc(adapter->req_rx_queues,
772 				sizeof(struct napi_struct), GFP_KERNEL);
773 	if (!adapter->napi)
774 		return -ENOMEM;
775 
776 	for (i = 0; i < adapter->req_rx_queues; i++) {
777 		netif_napi_add(netdev, &adapter->napi[i], ibmvnic_poll,
778 			       NAPI_POLL_WEIGHT);
779 	}
780 
781 	send_map_query(adapter);
782 
783 	rc = init_rx_pools(netdev);
784 	if (rc)
785 		return rc;
786 
787 	rc = init_tx_pools(netdev);
788 	return rc;
789 }
790 
791 static int __ibmvnic_open(struct net_device *netdev)
792 {
793 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
794 	enum vnic_state prev_state = adapter->state;
795 	int i, rc;
796 
797 	adapter->state = VNIC_OPENING;
798 	replenish_pools(adapter);
799 	ibmvnic_napi_enable(adapter);
800 
801 	/* We're ready to receive frames, enable the sub-crq interrupts and
802 	 * set the logical link state to up
803 	 */
804 	for (i = 0; i < adapter->req_rx_queues; i++) {
805 		if (prev_state == VNIC_CLOSED)
806 			enable_irq(adapter->rx_scrq[i]->irq);
807 		else
808 			enable_scrq_irq(adapter, adapter->rx_scrq[i]);
809 	}
810 
811 	for (i = 0; i < adapter->req_tx_queues; i++) {
812 		if (prev_state == VNIC_CLOSED)
813 			enable_irq(adapter->tx_scrq[i]->irq);
814 		else
815 			enable_scrq_irq(adapter, adapter->tx_scrq[i]);
816 	}
817 
818 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
819 	if (rc) {
820 		for (i = 0; i < adapter->req_rx_queues; i++)
821 			napi_disable(&adapter->napi[i]);
822 		release_resources(adapter);
823 		return rc;
824 	}
825 
826 	netif_tx_start_all_queues(netdev);
827 
828 	if (prev_state == VNIC_CLOSED) {
829 		for (i = 0; i < adapter->req_rx_queues; i++)
830 			napi_schedule(&adapter->napi[i]);
831 	}
832 
833 	adapter->state = VNIC_OPEN;
834 	return rc;
835 }
836 
837 static int ibmvnic_open(struct net_device *netdev)
838 {
839 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
840 	int rc;
841 
842 	mutex_lock(&adapter->reset_lock);
843 
844 	if (adapter->state != VNIC_CLOSED) {
845 		rc = ibmvnic_login(netdev);
846 		if (rc) {
847 			mutex_unlock(&adapter->reset_lock);
848 			return rc;
849 		}
850 
851 		rc = init_resources(adapter);
852 		if (rc) {
853 			netdev_err(netdev, "failed to initialize resources\n");
854 			release_resources(adapter);
855 			mutex_unlock(&adapter->reset_lock);
856 			return rc;
857 		}
858 	}
859 
860 	rc = __ibmvnic_open(netdev);
861 	mutex_unlock(&adapter->reset_lock);
862 
863 	return rc;
864 }
865 
866 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
867 {
868 	struct ibmvnic_tx_pool *tx_pool;
869 	u64 tx_entries;
870 	int tx_scrqs;
871 	int i, j;
872 
873 	if (!adapter->tx_pool)
874 		return;
875 
876 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
877 	tx_entries = adapter->req_tx_entries_per_subcrq;
878 
879 	/* Free any remaining skbs in the tx buffer pools */
880 	for (i = 0; i < tx_scrqs; i++) {
881 		tx_pool = &adapter->tx_pool[i];
882 		if (!tx_pool)
883 			continue;
884 
885 		for (j = 0; j < tx_entries; j++) {
886 			if (tx_pool->tx_buff[j].skb) {
887 				dev_kfree_skb_any(tx_pool->tx_buff[j].skb);
888 				tx_pool->tx_buff[j].skb = NULL;
889 			}
890 		}
891 	}
892 }
893 
894 static int __ibmvnic_close(struct net_device *netdev)
895 {
896 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
897 	int rc = 0;
898 	int i;
899 
900 	adapter->state = VNIC_CLOSING;
901 
902 	/* ensure that transmissions are stopped if called by do_reset */
903 	if (adapter->resetting)
904 		netif_tx_disable(netdev);
905 	else
906 		netif_tx_stop_all_queues(netdev);
907 
908 	ibmvnic_napi_disable(adapter);
909 
910 	if (adapter->tx_scrq) {
911 		for (i = 0; i < adapter->req_tx_queues; i++)
912 			if (adapter->tx_scrq[i]->irq)
913 				disable_irq(adapter->tx_scrq[i]->irq);
914 	}
915 
916 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
917 	if (rc)
918 		return rc;
919 
920 	if (adapter->rx_scrq) {
921 		for (i = 0; i < adapter->req_rx_queues; i++) {
922 			int retries = 10;
923 
924 			while (pending_scrq(adapter, adapter->rx_scrq[i])) {
925 				retries--;
926 				mdelay(100);
927 
928 				if (retries == 0)
929 					break;
930 			}
931 
932 			if (adapter->rx_scrq[i]->irq)
933 				disable_irq(adapter->rx_scrq[i]->irq);
934 		}
935 	}
936 
937 	clean_tx_pools(adapter);
938 	adapter->state = VNIC_CLOSED;
939 	return rc;
940 }
941 
942 static int ibmvnic_close(struct net_device *netdev)
943 {
944 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
945 	int rc;
946 
947 	mutex_lock(&adapter->reset_lock);
948 	rc = __ibmvnic_close(netdev);
949 	mutex_unlock(&adapter->reset_lock);
950 
951 	return rc;
952 }
953 
954 /**
955  * build_hdr_data - creates L2/L3/L4 header data buffer
956  * @hdr_field - bitfield determining needed headers
957  * @skb - socket buffer
958  * @hdr_len - array of header lengths
959  * @tot_len - total length of data
960  *
961  * Reads hdr_field to determine which headers are needed by firmware.
962  * Builds a buffer containing these headers.  Saves individual header
963  * lengths and total buffer length to be used to build descriptors.
964  */
965 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
966 			  int *hdr_len, u8 *hdr_data)
967 {
968 	int len = 0;
969 	u8 *hdr;
970 
971 	hdr_len[0] = sizeof(struct ethhdr);
972 
973 	if (skb->protocol == htons(ETH_P_IP)) {
974 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
975 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
976 			hdr_len[2] = tcp_hdrlen(skb);
977 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
978 			hdr_len[2] = sizeof(struct udphdr);
979 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
980 		hdr_len[1] = sizeof(struct ipv6hdr);
981 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
982 			hdr_len[2] = tcp_hdrlen(skb);
983 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
984 			hdr_len[2] = sizeof(struct udphdr);
985 	}
986 
987 	memset(hdr_data, 0, 120);
988 	if ((hdr_field >> 6) & 1) {
989 		hdr = skb_mac_header(skb);
990 		memcpy(hdr_data, hdr, hdr_len[0]);
991 		len += hdr_len[0];
992 	}
993 
994 	if ((hdr_field >> 5) & 1) {
995 		hdr = skb_network_header(skb);
996 		memcpy(hdr_data + len, hdr, hdr_len[1]);
997 		len += hdr_len[1];
998 	}
999 
1000 	if ((hdr_field >> 4) & 1) {
1001 		hdr = skb_transport_header(skb);
1002 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1003 		len += hdr_len[2];
1004 	}
1005 	return len;
1006 }
1007 
1008 /**
1009  * create_hdr_descs - create header and header extension descriptors
1010  * @hdr_field - bitfield determining needed headers
1011  * @data - buffer containing header data
1012  * @len - length of data buffer
1013  * @hdr_len - array of individual header lengths
1014  * @scrq_arr - descriptor array
1015  *
1016  * Creates header and, if needed, header extension descriptors and
1017  * places them in a descriptor array, scrq_arr
1018  */
1019 
1020 static void create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1021 			     union sub_crq *scrq_arr)
1022 {
1023 	union sub_crq hdr_desc;
1024 	int tmp_len = len;
1025 	u8 *data, *cur;
1026 	int tmp;
1027 
1028 	while (tmp_len > 0) {
1029 		cur = hdr_data + len - tmp_len;
1030 
1031 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1032 		if (cur != hdr_data) {
1033 			data = hdr_desc.hdr_ext.data;
1034 			tmp = tmp_len > 29 ? 29 : tmp_len;
1035 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1036 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1037 			hdr_desc.hdr_ext.len = tmp;
1038 		} else {
1039 			data = hdr_desc.hdr.data;
1040 			tmp = tmp_len > 24 ? 24 : tmp_len;
1041 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1042 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1043 			hdr_desc.hdr.len = tmp;
1044 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1045 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1046 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1047 			hdr_desc.hdr.flag = hdr_field << 1;
1048 		}
1049 		memcpy(data, cur, tmp);
1050 		tmp_len -= tmp;
1051 		*scrq_arr = hdr_desc;
1052 		scrq_arr++;
1053 	}
1054 }
1055 
1056 /**
1057  * build_hdr_descs_arr - build a header descriptor array
1058  * @skb - socket buffer
1059  * @num_entries - number of descriptors to be sent
1060  * @subcrq - first TX descriptor
1061  * @hdr_field - bit field determining which headers will be sent
1062  *
1063  * This function will build a TX descriptor array with applicable
1064  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1065  */
1066 
1067 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1068 				int *num_entries, u8 hdr_field)
1069 {
1070 	int hdr_len[3] = {0, 0, 0};
1071 	int tot_len, len;
1072 	u8 *hdr_data = txbuff->hdr_data;
1073 
1074 	tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1075 				 txbuff->hdr_data);
1076 	len = tot_len;
1077 	len -= 24;
1078 	if (len > 0)
1079 		num_entries += len % 29 ? len / 29 + 1 : len / 29;
1080 	create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1081 			 txbuff->indir_arr + 1);
1082 }
1083 
1084 static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1085 {
1086 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1087 	int queue_num = skb_get_queue_mapping(skb);
1088 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1089 	struct device *dev = &adapter->vdev->dev;
1090 	struct ibmvnic_tx_buff *tx_buff = NULL;
1091 	struct ibmvnic_sub_crq_queue *tx_scrq;
1092 	struct ibmvnic_tx_pool *tx_pool;
1093 	unsigned int tx_send_failed = 0;
1094 	unsigned int tx_map_failed = 0;
1095 	unsigned int tx_dropped = 0;
1096 	unsigned int tx_packets = 0;
1097 	unsigned int tx_bytes = 0;
1098 	dma_addr_t data_dma_addr;
1099 	struct netdev_queue *txq;
1100 	unsigned long lpar_rc;
1101 	union sub_crq tx_crq;
1102 	unsigned int offset;
1103 	int num_entries = 1;
1104 	unsigned char *dst;
1105 	u64 *handle_array;
1106 	int index = 0;
1107 	int ret = 0;
1108 
1109 	if (adapter->resetting) {
1110 		if (!netif_subqueue_stopped(netdev, skb))
1111 			netif_stop_subqueue(netdev, queue_num);
1112 		dev_kfree_skb_any(skb);
1113 
1114 		tx_send_failed++;
1115 		tx_dropped++;
1116 		ret = NETDEV_TX_OK;
1117 		goto out;
1118 	}
1119 
1120 	tx_pool = &adapter->tx_pool[queue_num];
1121 	tx_scrq = adapter->tx_scrq[queue_num];
1122 	txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1123 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1124 		be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1125 
1126 	index = tx_pool->free_map[tx_pool->consumer_index];
1127 	offset = index * adapter->req_mtu;
1128 	dst = tx_pool->long_term_buff.buff + offset;
1129 	memset(dst, 0, adapter->req_mtu);
1130 	skb_copy_from_linear_data(skb, dst, skb->len);
1131 	data_dma_addr = tx_pool->long_term_buff.addr + offset;
1132 
1133 	tx_pool->consumer_index =
1134 	    (tx_pool->consumer_index + 1) %
1135 		adapter->req_tx_entries_per_subcrq;
1136 
1137 	tx_buff = &tx_pool->tx_buff[index];
1138 	tx_buff->skb = skb;
1139 	tx_buff->data_dma[0] = data_dma_addr;
1140 	tx_buff->data_len[0] = skb->len;
1141 	tx_buff->index = index;
1142 	tx_buff->pool_index = queue_num;
1143 	tx_buff->last_frag = true;
1144 
1145 	memset(&tx_crq, 0, sizeof(tx_crq));
1146 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1147 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1148 	tx_crq.v1.n_crq_elem = 1;
1149 	tx_crq.v1.n_sge = 1;
1150 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1151 	tx_crq.v1.correlator = cpu_to_be32(index);
1152 	tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1153 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1154 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1155 
1156 	if (adapter->vlan_header_insertion) {
1157 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1158 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1159 	}
1160 
1161 	if (skb->protocol == htons(ETH_P_IP)) {
1162 		if (ip_hdr(skb)->version == 4)
1163 			tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1164 		else if (ip_hdr(skb)->version == 6)
1165 			tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1166 
1167 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1168 			tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1169 		else if (ip_hdr(skb)->protocol != IPPROTO_TCP)
1170 			tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1171 	}
1172 
1173 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1174 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1175 		hdrs += 2;
1176 	}
1177 	/* determine if l2/3/4 headers are sent to firmware */
1178 	if ((*hdrs >> 7) & 1 &&
1179 	    (skb->protocol == htons(ETH_P_IP) ||
1180 	     skb->protocol == htons(ETH_P_IPV6))) {
1181 		build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1182 		tx_crq.v1.n_crq_elem = num_entries;
1183 		tx_buff->indir_arr[0] = tx_crq;
1184 		tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1185 						    sizeof(tx_buff->indir_arr),
1186 						    DMA_TO_DEVICE);
1187 		if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1188 			dev_kfree_skb_any(skb);
1189 			tx_buff->skb = NULL;
1190 			if (!firmware_has_feature(FW_FEATURE_CMO))
1191 				dev_err(dev, "tx: unable to map descriptor array\n");
1192 			tx_map_failed++;
1193 			tx_dropped++;
1194 			ret = NETDEV_TX_OK;
1195 			goto out;
1196 		}
1197 		lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1198 					       (u64)tx_buff->indir_dma,
1199 					       (u64)num_entries);
1200 	} else {
1201 		lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1202 				      &tx_crq);
1203 	}
1204 	if (lpar_rc != H_SUCCESS) {
1205 		dev_err(dev, "tx failed with code %ld\n", lpar_rc);
1206 
1207 		if (tx_pool->consumer_index == 0)
1208 			tx_pool->consumer_index =
1209 				adapter->req_tx_entries_per_subcrq - 1;
1210 		else
1211 			tx_pool->consumer_index--;
1212 
1213 		dev_kfree_skb_any(skb);
1214 		tx_buff->skb = NULL;
1215 
1216 		if (lpar_rc == H_CLOSED) {
1217 			/* Disable TX and report carrier off if queue is closed.
1218 			 * Firmware guarantees that a signal will be sent to the
1219 			 * driver, triggering a reset or some other action.
1220 			 */
1221 			netif_tx_stop_all_queues(netdev);
1222 			netif_carrier_off(netdev);
1223 		}
1224 
1225 		tx_send_failed++;
1226 		tx_dropped++;
1227 		ret = NETDEV_TX_OK;
1228 		goto out;
1229 	}
1230 
1231 	if (atomic_inc_return(&tx_scrq->used)
1232 					>= adapter->req_tx_entries_per_subcrq) {
1233 		netdev_info(netdev, "Stopping queue %d\n", queue_num);
1234 		netif_stop_subqueue(netdev, queue_num);
1235 	}
1236 
1237 	tx_packets++;
1238 	tx_bytes += skb->len;
1239 	txq->trans_start = jiffies;
1240 	ret = NETDEV_TX_OK;
1241 
1242 out:
1243 	netdev->stats.tx_dropped += tx_dropped;
1244 	netdev->stats.tx_bytes += tx_bytes;
1245 	netdev->stats.tx_packets += tx_packets;
1246 	adapter->tx_send_failed += tx_send_failed;
1247 	adapter->tx_map_failed += tx_map_failed;
1248 
1249 	return ret;
1250 }
1251 
1252 static void ibmvnic_set_multi(struct net_device *netdev)
1253 {
1254 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1255 	struct netdev_hw_addr *ha;
1256 	union ibmvnic_crq crq;
1257 
1258 	memset(&crq, 0, sizeof(crq));
1259 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1260 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1261 
1262 	if (netdev->flags & IFF_PROMISC) {
1263 		if (!adapter->promisc_supported)
1264 			return;
1265 	} else {
1266 		if (netdev->flags & IFF_ALLMULTI) {
1267 			/* Accept all multicast */
1268 			memset(&crq, 0, sizeof(crq));
1269 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1270 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1271 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1272 			ibmvnic_send_crq(adapter, &crq);
1273 		} else if (netdev_mc_empty(netdev)) {
1274 			/* Reject all multicast */
1275 			memset(&crq, 0, sizeof(crq));
1276 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1277 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1278 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1279 			ibmvnic_send_crq(adapter, &crq);
1280 		} else {
1281 			/* Accept one or more multicast(s) */
1282 			netdev_for_each_mc_addr(ha, netdev) {
1283 				memset(&crq, 0, sizeof(crq));
1284 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1285 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1286 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1287 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1288 						ha->addr);
1289 				ibmvnic_send_crq(adapter, &crq);
1290 			}
1291 		}
1292 	}
1293 }
1294 
1295 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1296 {
1297 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1298 	struct sockaddr *addr = p;
1299 	union ibmvnic_crq crq;
1300 
1301 	if (!is_valid_ether_addr(addr->sa_data))
1302 		return -EADDRNOTAVAIL;
1303 
1304 	memset(&crq, 0, sizeof(crq));
1305 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1306 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1307 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1308 	ibmvnic_send_crq(adapter, &crq);
1309 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1310 	return 0;
1311 }
1312 
1313 /**
1314  * do_reset returns zero if we are able to keep processing reset events, or
1315  * non-zero if we hit a fatal error and must halt.
1316  */
1317 static int do_reset(struct ibmvnic_adapter *adapter,
1318 		    struct ibmvnic_rwi *rwi, u32 reset_state)
1319 {
1320 	struct net_device *netdev = adapter->netdev;
1321 	int i, rc;
1322 
1323 	netif_carrier_off(netdev);
1324 	adapter->reset_reason = rwi->reset_reason;
1325 
1326 	if (rwi->reset_reason == VNIC_RESET_MOBILITY) {
1327 		rc = ibmvnic_reenable_crq_queue(adapter);
1328 		if (rc)
1329 			return 0;
1330 	}
1331 
1332 	rc = __ibmvnic_close(netdev);
1333 	if (rc)
1334 		return rc;
1335 
1336 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1337 		/* remove the closed state so when we call open it appears
1338 		 * we are coming from the probed state.
1339 		 */
1340 		adapter->state = VNIC_PROBED;
1341 
1342 		rc = ibmvnic_init(adapter);
1343 		if (rc)
1344 			return 0;
1345 
1346 		/* If the adapter was in PROBE state prior to the reset,
1347 		 * exit here.
1348 		 */
1349 		if (reset_state == VNIC_PROBED)
1350 			return 0;
1351 
1352 		rc = ibmvnic_login(netdev);
1353 		if (rc) {
1354 			adapter->state = VNIC_PROBED;
1355 			return 0;
1356 		}
1357 
1358 		rc = reset_tx_pools(adapter);
1359 		if (rc)
1360 			return rc;
1361 
1362 		rc = reset_rx_pools(adapter);
1363 		if (rc)
1364 			return rc;
1365 
1366 		if (reset_state == VNIC_CLOSED)
1367 			return 0;
1368 	}
1369 
1370 	rc = __ibmvnic_open(netdev);
1371 	if (rc) {
1372 		if (list_empty(&adapter->rwi_list))
1373 			adapter->state = VNIC_CLOSED;
1374 		else
1375 			adapter->state = reset_state;
1376 
1377 		return 0;
1378 	}
1379 
1380 	netif_carrier_on(netdev);
1381 
1382 	/* kick napi */
1383 	for (i = 0; i < adapter->req_rx_queues; i++)
1384 		napi_schedule(&adapter->napi[i]);
1385 
1386 	if (adapter->reset_reason != VNIC_RESET_FAILOVER)
1387 		netdev_notify_peers(netdev);
1388 
1389 	return 0;
1390 }
1391 
1392 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1393 {
1394 	struct ibmvnic_rwi *rwi;
1395 
1396 	mutex_lock(&adapter->rwi_lock);
1397 
1398 	if (!list_empty(&adapter->rwi_list)) {
1399 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1400 				       list);
1401 		list_del(&rwi->list);
1402 	} else {
1403 		rwi = NULL;
1404 	}
1405 
1406 	mutex_unlock(&adapter->rwi_lock);
1407 	return rwi;
1408 }
1409 
1410 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1411 {
1412 	struct ibmvnic_rwi *rwi;
1413 
1414 	rwi = get_next_rwi(adapter);
1415 	while (rwi) {
1416 		kfree(rwi);
1417 		rwi = get_next_rwi(adapter);
1418 	}
1419 }
1420 
1421 static void __ibmvnic_reset(struct work_struct *work)
1422 {
1423 	struct ibmvnic_rwi *rwi;
1424 	struct ibmvnic_adapter *adapter;
1425 	struct net_device *netdev;
1426 	u32 reset_state;
1427 	int rc;
1428 
1429 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1430 	netdev = adapter->netdev;
1431 
1432 	mutex_lock(&adapter->reset_lock);
1433 	adapter->resetting = true;
1434 	reset_state = adapter->state;
1435 
1436 	rwi = get_next_rwi(adapter);
1437 	while (rwi) {
1438 		rc = do_reset(adapter, rwi, reset_state);
1439 		kfree(rwi);
1440 		if (rc)
1441 			break;
1442 
1443 		rwi = get_next_rwi(adapter);
1444 	}
1445 
1446 	if (rc) {
1447 		free_all_rwi(adapter);
1448 		mutex_unlock(&adapter->reset_lock);
1449 		return;
1450 	}
1451 
1452 	adapter->resetting = false;
1453 	mutex_unlock(&adapter->reset_lock);
1454 }
1455 
1456 static void ibmvnic_reset(struct ibmvnic_adapter *adapter,
1457 			  enum ibmvnic_reset_reason reason)
1458 {
1459 	struct ibmvnic_rwi *rwi, *tmp;
1460 	struct net_device *netdev = adapter->netdev;
1461 	struct list_head *entry;
1462 
1463 	if (adapter->state == VNIC_REMOVING ||
1464 	    adapter->state == VNIC_REMOVED) {
1465 		netdev_dbg(netdev, "Adapter removing, skipping reset\n");
1466 		return;
1467 	}
1468 
1469 	if (adapter->state == VNIC_PROBING) {
1470 		netdev_warn(netdev, "Adapter reset during probe\n");
1471 		adapter->init_done_rc = EAGAIN;
1472 		return;
1473 	}
1474 
1475 	mutex_lock(&adapter->rwi_lock);
1476 
1477 	list_for_each(entry, &adapter->rwi_list) {
1478 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
1479 		if (tmp->reset_reason == reason) {
1480 			netdev_err(netdev, "Matching reset found, skipping\n");
1481 			mutex_unlock(&adapter->rwi_lock);
1482 			return;
1483 		}
1484 	}
1485 
1486 	rwi = kzalloc(sizeof(*rwi), GFP_KERNEL);
1487 	if (!rwi) {
1488 		mutex_unlock(&adapter->rwi_lock);
1489 		ibmvnic_close(netdev);
1490 		return;
1491 	}
1492 
1493 	rwi->reset_reason = reason;
1494 	list_add_tail(&rwi->list, &adapter->rwi_list);
1495 	mutex_unlock(&adapter->rwi_lock);
1496 	schedule_work(&adapter->ibmvnic_reset);
1497 }
1498 
1499 static void ibmvnic_tx_timeout(struct net_device *dev)
1500 {
1501 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1502 
1503 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
1504 }
1505 
1506 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
1507 				  struct ibmvnic_rx_buff *rx_buff)
1508 {
1509 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
1510 
1511 	rx_buff->skb = NULL;
1512 
1513 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
1514 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
1515 
1516 	atomic_dec(&pool->available);
1517 }
1518 
1519 static int ibmvnic_poll(struct napi_struct *napi, int budget)
1520 {
1521 	struct net_device *netdev = napi->dev;
1522 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1523 	int scrq_num = (int)(napi - adapter->napi);
1524 	int frames_processed = 0;
1525 
1526 restart_poll:
1527 	while (frames_processed < budget) {
1528 		struct sk_buff *skb;
1529 		struct ibmvnic_rx_buff *rx_buff;
1530 		union sub_crq *next;
1531 		u32 length;
1532 		u16 offset;
1533 		u8 flags = 0;
1534 
1535 		if (unlikely(adapter->resetting)) {
1536 			enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1537 			napi_complete_done(napi, frames_processed);
1538 			return frames_processed;
1539 		}
1540 
1541 		if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
1542 			break;
1543 		next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
1544 		rx_buff =
1545 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
1546 							  rx_comp.correlator);
1547 		/* do error checking */
1548 		if (next->rx_comp.rc) {
1549 			netdev_err(netdev, "rx error %x\n", next->rx_comp.rc);
1550 			/* free the entry */
1551 			next->rx_comp.first = 0;
1552 			remove_buff_from_pool(adapter, rx_buff);
1553 			continue;
1554 		}
1555 
1556 		length = be32_to_cpu(next->rx_comp.len);
1557 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
1558 		flags = next->rx_comp.flags;
1559 		skb = rx_buff->skb;
1560 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
1561 					length);
1562 
1563 		/* VLAN Header has been stripped by the system firmware and
1564 		 * needs to be inserted by the driver
1565 		 */
1566 		if (adapter->rx_vlan_header_insertion &&
1567 		    (flags & IBMVNIC_VLAN_STRIPPED))
1568 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
1569 					       ntohs(next->rx_comp.vlan_tci));
1570 
1571 		/* free the entry */
1572 		next->rx_comp.first = 0;
1573 		remove_buff_from_pool(adapter, rx_buff);
1574 
1575 		skb_put(skb, length);
1576 		skb->protocol = eth_type_trans(skb, netdev);
1577 		skb_record_rx_queue(skb, scrq_num);
1578 
1579 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
1580 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
1581 			skb->ip_summed = CHECKSUM_UNNECESSARY;
1582 		}
1583 
1584 		length = skb->len;
1585 		napi_gro_receive(napi, skb); /* send it up */
1586 		netdev->stats.rx_packets++;
1587 		netdev->stats.rx_bytes += length;
1588 		frames_processed++;
1589 	}
1590 
1591 	if (adapter->state != VNIC_CLOSING)
1592 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
1593 
1594 	if (frames_processed < budget) {
1595 		enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1596 		napi_complete_done(napi, frames_processed);
1597 		if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
1598 		    napi_reschedule(napi)) {
1599 			disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1600 			goto restart_poll;
1601 		}
1602 	}
1603 	return frames_processed;
1604 }
1605 
1606 #ifdef CONFIG_NET_POLL_CONTROLLER
1607 static void ibmvnic_netpoll_controller(struct net_device *dev)
1608 {
1609 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1610 	int i;
1611 
1612 	replenish_pools(netdev_priv(dev));
1613 	for (i = 0; i < adapter->req_rx_queues; i++)
1614 		ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
1615 				     adapter->rx_scrq[i]);
1616 }
1617 #endif
1618 
1619 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
1620 {
1621 	return -EOPNOTSUPP;
1622 }
1623 
1624 static const struct net_device_ops ibmvnic_netdev_ops = {
1625 	.ndo_open		= ibmvnic_open,
1626 	.ndo_stop		= ibmvnic_close,
1627 	.ndo_start_xmit		= ibmvnic_xmit,
1628 	.ndo_set_rx_mode	= ibmvnic_set_multi,
1629 	.ndo_set_mac_address	= ibmvnic_set_mac,
1630 	.ndo_validate_addr	= eth_validate_addr,
1631 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
1632 #ifdef CONFIG_NET_POLL_CONTROLLER
1633 	.ndo_poll_controller	= ibmvnic_netpoll_controller,
1634 #endif
1635 	.ndo_change_mtu		= ibmvnic_change_mtu,
1636 };
1637 
1638 /* ethtool functions */
1639 
1640 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
1641 				      struct ethtool_link_ksettings *cmd)
1642 {
1643 	u32 supported, advertising;
1644 
1645 	supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
1646 			  SUPPORTED_FIBRE);
1647 	advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
1648 			    ADVERTISED_FIBRE);
1649 	cmd->base.speed = SPEED_1000;
1650 	cmd->base.duplex = DUPLEX_FULL;
1651 	cmd->base.port = PORT_FIBRE;
1652 	cmd->base.phy_address = 0;
1653 	cmd->base.autoneg = AUTONEG_ENABLE;
1654 
1655 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
1656 						supported);
1657 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
1658 						advertising);
1659 
1660 	return 0;
1661 }
1662 
1663 static void ibmvnic_get_drvinfo(struct net_device *dev,
1664 				struct ethtool_drvinfo *info)
1665 {
1666 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
1667 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
1668 }
1669 
1670 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
1671 {
1672 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1673 
1674 	return adapter->msg_enable;
1675 }
1676 
1677 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
1678 {
1679 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1680 
1681 	adapter->msg_enable = data;
1682 }
1683 
1684 static u32 ibmvnic_get_link(struct net_device *netdev)
1685 {
1686 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1687 
1688 	/* Don't need to send a query because we request a logical link up at
1689 	 * init and then we wait for link state indications
1690 	 */
1691 	return adapter->logical_link_state;
1692 }
1693 
1694 static void ibmvnic_get_ringparam(struct net_device *netdev,
1695 				  struct ethtool_ringparam *ring)
1696 {
1697 	ring->rx_max_pending = 0;
1698 	ring->tx_max_pending = 0;
1699 	ring->rx_mini_max_pending = 0;
1700 	ring->rx_jumbo_max_pending = 0;
1701 	ring->rx_pending = 0;
1702 	ring->tx_pending = 0;
1703 	ring->rx_mini_pending = 0;
1704 	ring->rx_jumbo_pending = 0;
1705 }
1706 
1707 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
1708 {
1709 	int i;
1710 
1711 	if (stringset != ETH_SS_STATS)
1712 		return;
1713 
1714 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
1715 		memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
1716 }
1717 
1718 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
1719 {
1720 	switch (sset) {
1721 	case ETH_SS_STATS:
1722 		return ARRAY_SIZE(ibmvnic_stats);
1723 	default:
1724 		return -EOPNOTSUPP;
1725 	}
1726 }
1727 
1728 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
1729 				      struct ethtool_stats *stats, u64 *data)
1730 {
1731 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1732 	union ibmvnic_crq crq;
1733 	int i;
1734 
1735 	memset(&crq, 0, sizeof(crq));
1736 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
1737 	crq.request_statistics.cmd = REQUEST_STATISTICS;
1738 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
1739 	crq.request_statistics.len =
1740 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
1741 
1742 	/* Wait for data to be written */
1743 	init_completion(&adapter->stats_done);
1744 	ibmvnic_send_crq(adapter, &crq);
1745 	wait_for_completion(&adapter->stats_done);
1746 
1747 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
1748 		data[i] = IBMVNIC_GET_STAT(adapter, ibmvnic_stats[i].offset);
1749 }
1750 
1751 static const struct ethtool_ops ibmvnic_ethtool_ops = {
1752 	.get_drvinfo		= ibmvnic_get_drvinfo,
1753 	.get_msglevel		= ibmvnic_get_msglevel,
1754 	.set_msglevel		= ibmvnic_set_msglevel,
1755 	.get_link		= ibmvnic_get_link,
1756 	.get_ringparam		= ibmvnic_get_ringparam,
1757 	.get_strings            = ibmvnic_get_strings,
1758 	.get_sset_count         = ibmvnic_get_sset_count,
1759 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
1760 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
1761 };
1762 
1763 /* Routines for managing CRQs/sCRQs  */
1764 
1765 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
1766 				   struct ibmvnic_sub_crq_queue *scrq)
1767 {
1768 	int rc;
1769 
1770 	if (scrq->irq) {
1771 		free_irq(scrq->irq, scrq);
1772 		irq_dispose_mapping(scrq->irq);
1773 		scrq->irq = 0;
1774 	}
1775 
1776 	memset(scrq->msgs, 0, 4 * PAGE_SIZE);
1777 	scrq->cur = 0;
1778 
1779 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
1780 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
1781 	return rc;
1782 }
1783 
1784 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
1785 {
1786 	int i, rc;
1787 
1788 	for (i = 0; i < adapter->req_tx_queues; i++) {
1789 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
1790 		if (rc)
1791 			return rc;
1792 	}
1793 
1794 	for (i = 0; i < adapter->req_rx_queues; i++) {
1795 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
1796 		if (rc)
1797 			return rc;
1798 	}
1799 
1800 	return rc;
1801 }
1802 
1803 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
1804 				  struct ibmvnic_sub_crq_queue *scrq)
1805 {
1806 	struct device *dev = &adapter->vdev->dev;
1807 	long rc;
1808 
1809 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
1810 
1811 	/* Close the sub-crqs */
1812 	do {
1813 		rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
1814 					adapter->vdev->unit_address,
1815 					scrq->crq_num);
1816 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
1817 
1818 	if (rc) {
1819 		netdev_err(adapter->netdev,
1820 			   "Failed to release sub-CRQ %16lx, rc = %ld\n",
1821 			   scrq->crq_num, rc);
1822 	}
1823 
1824 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
1825 			 DMA_BIDIRECTIONAL);
1826 	free_pages((unsigned long)scrq->msgs, 2);
1827 	kfree(scrq);
1828 }
1829 
1830 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
1831 							*adapter)
1832 {
1833 	struct device *dev = &adapter->vdev->dev;
1834 	struct ibmvnic_sub_crq_queue *scrq;
1835 	int rc;
1836 
1837 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
1838 	if (!scrq)
1839 		return NULL;
1840 
1841 	scrq->msgs =
1842 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
1843 	if (!scrq->msgs) {
1844 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
1845 		goto zero_page_failed;
1846 	}
1847 
1848 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
1849 					 DMA_BIDIRECTIONAL);
1850 	if (dma_mapping_error(dev, scrq->msg_token)) {
1851 		dev_warn(dev, "Couldn't map crq queue messages page\n");
1852 		goto map_failed;
1853 	}
1854 
1855 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
1856 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
1857 
1858 	if (rc == H_RESOURCE)
1859 		rc = ibmvnic_reset_crq(adapter);
1860 
1861 	if (rc == H_CLOSED) {
1862 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
1863 	} else if (rc) {
1864 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
1865 		goto reg_failed;
1866 	}
1867 
1868 	scrq->adapter = adapter;
1869 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
1870 	spin_lock_init(&scrq->lock);
1871 
1872 	netdev_dbg(adapter->netdev,
1873 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
1874 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
1875 
1876 	return scrq;
1877 
1878 reg_failed:
1879 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
1880 			 DMA_BIDIRECTIONAL);
1881 map_failed:
1882 	free_pages((unsigned long)scrq->msgs, 2);
1883 zero_page_failed:
1884 	kfree(scrq);
1885 
1886 	return NULL;
1887 }
1888 
1889 static void release_sub_crqs(struct ibmvnic_adapter *adapter)
1890 {
1891 	int i;
1892 
1893 	if (adapter->tx_scrq) {
1894 		for (i = 0; i < adapter->req_tx_queues; i++) {
1895 			if (!adapter->tx_scrq[i])
1896 				continue;
1897 
1898 			if (adapter->tx_scrq[i]->irq) {
1899 				free_irq(adapter->tx_scrq[i]->irq,
1900 					 adapter->tx_scrq[i]);
1901 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
1902 				adapter->tx_scrq[i]->irq = 0;
1903 			}
1904 
1905 			release_sub_crq_queue(adapter, adapter->tx_scrq[i]);
1906 		}
1907 
1908 		kfree(adapter->tx_scrq);
1909 		adapter->tx_scrq = NULL;
1910 	}
1911 
1912 	if (adapter->rx_scrq) {
1913 		for (i = 0; i < adapter->req_rx_queues; i++) {
1914 			if (!adapter->rx_scrq[i])
1915 				continue;
1916 
1917 			if (adapter->rx_scrq[i]->irq) {
1918 				free_irq(adapter->rx_scrq[i]->irq,
1919 					 adapter->rx_scrq[i]);
1920 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
1921 				adapter->rx_scrq[i]->irq = 0;
1922 			}
1923 
1924 			release_sub_crq_queue(adapter, adapter->rx_scrq[i]);
1925 		}
1926 
1927 		kfree(adapter->rx_scrq);
1928 		adapter->rx_scrq = NULL;
1929 	}
1930 }
1931 
1932 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
1933 			    struct ibmvnic_sub_crq_queue *scrq)
1934 {
1935 	struct device *dev = &adapter->vdev->dev;
1936 	unsigned long rc;
1937 
1938 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
1939 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
1940 	if (rc)
1941 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
1942 			scrq->hw_irq, rc);
1943 	return rc;
1944 }
1945 
1946 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
1947 			   struct ibmvnic_sub_crq_queue *scrq)
1948 {
1949 	struct device *dev = &adapter->vdev->dev;
1950 	unsigned long rc;
1951 
1952 	if (scrq->hw_irq > 0x100000000ULL) {
1953 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
1954 		return 1;
1955 	}
1956 
1957 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
1958 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
1959 	if (rc)
1960 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
1961 			scrq->hw_irq, rc);
1962 	return rc;
1963 }
1964 
1965 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
1966 			       struct ibmvnic_sub_crq_queue *scrq)
1967 {
1968 	struct device *dev = &adapter->vdev->dev;
1969 	struct ibmvnic_tx_buff *txbuff;
1970 	union sub_crq *next;
1971 	int index;
1972 	int i, j;
1973 	u8 first;
1974 
1975 restart_loop:
1976 	while (pending_scrq(adapter, scrq)) {
1977 		unsigned int pool = scrq->pool_index;
1978 
1979 		next = ibmvnic_next_scrq(adapter, scrq);
1980 		for (i = 0; i < next->tx_comp.num_comps; i++) {
1981 			if (next->tx_comp.rcs[i]) {
1982 				dev_err(dev, "tx error %x\n",
1983 					next->tx_comp.rcs[i]);
1984 				continue;
1985 			}
1986 			index = be32_to_cpu(next->tx_comp.correlators[i]);
1987 			txbuff = &adapter->tx_pool[pool].tx_buff[index];
1988 
1989 			for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
1990 				if (!txbuff->data_dma[j])
1991 					continue;
1992 
1993 				txbuff->data_dma[j] = 0;
1994 			}
1995 			/* if sub_crq was sent indirectly */
1996 			first = txbuff->indir_arr[0].generic.first;
1997 			if (first == IBMVNIC_CRQ_CMD) {
1998 				dma_unmap_single(dev, txbuff->indir_dma,
1999 						 sizeof(txbuff->indir_arr),
2000 						 DMA_TO_DEVICE);
2001 			}
2002 
2003 			if (txbuff->last_frag) {
2004 				dev_kfree_skb_any(txbuff->skb);
2005 				txbuff->skb = NULL;
2006 			}
2007 
2008 			adapter->tx_pool[pool].free_map[adapter->tx_pool[pool].
2009 						     producer_index] = index;
2010 			adapter->tx_pool[pool].producer_index =
2011 			    (adapter->tx_pool[pool].producer_index + 1) %
2012 			    adapter->req_tx_entries_per_subcrq;
2013 		}
2014 		/* remove tx_comp scrq*/
2015 		next->tx_comp.first = 0;
2016 
2017 		if (atomic_sub_return(next->tx_comp.num_comps, &scrq->used) <=
2018 		    (adapter->req_tx_entries_per_subcrq / 2) &&
2019 		    __netif_subqueue_stopped(adapter->netdev,
2020 					     scrq->pool_index)) {
2021 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2022 			netdev_info(adapter->netdev, "Started queue %d\n",
2023 				    scrq->pool_index);
2024 		}
2025 	}
2026 
2027 	enable_scrq_irq(adapter, scrq);
2028 
2029 	if (pending_scrq(adapter, scrq)) {
2030 		disable_scrq_irq(adapter, scrq);
2031 		goto restart_loop;
2032 	}
2033 
2034 	return 0;
2035 }
2036 
2037 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2038 {
2039 	struct ibmvnic_sub_crq_queue *scrq = instance;
2040 	struct ibmvnic_adapter *adapter = scrq->adapter;
2041 
2042 	disable_scrq_irq(adapter, scrq);
2043 	ibmvnic_complete_tx(adapter, scrq);
2044 
2045 	return IRQ_HANDLED;
2046 }
2047 
2048 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2049 {
2050 	struct ibmvnic_sub_crq_queue *scrq = instance;
2051 	struct ibmvnic_adapter *adapter = scrq->adapter;
2052 
2053 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2054 		disable_scrq_irq(adapter, scrq);
2055 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
2056 	}
2057 
2058 	return IRQ_HANDLED;
2059 }
2060 
2061 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2062 {
2063 	struct device *dev = &adapter->vdev->dev;
2064 	struct ibmvnic_sub_crq_queue *scrq;
2065 	int i = 0, j = 0;
2066 	int rc = 0;
2067 
2068 	for (i = 0; i < adapter->req_tx_queues; i++) {
2069 		scrq = adapter->tx_scrq[i];
2070 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2071 
2072 		if (!scrq->irq) {
2073 			rc = -EINVAL;
2074 			dev_err(dev, "Error mapping irq\n");
2075 			goto req_tx_irq_failed;
2076 		}
2077 
2078 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2079 				 0, "ibmvnic_tx", scrq);
2080 
2081 		if (rc) {
2082 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2083 				scrq->irq, rc);
2084 			irq_dispose_mapping(scrq->irq);
2085 			goto req_rx_irq_failed;
2086 		}
2087 	}
2088 
2089 	for (i = 0; i < adapter->req_rx_queues; i++) {
2090 		scrq = adapter->rx_scrq[i];
2091 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2092 		if (!scrq->irq) {
2093 			rc = -EINVAL;
2094 			dev_err(dev, "Error mapping irq\n");
2095 			goto req_rx_irq_failed;
2096 		}
2097 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2098 				 0, "ibmvnic_rx", scrq);
2099 		if (rc) {
2100 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2101 				scrq->irq, rc);
2102 			irq_dispose_mapping(scrq->irq);
2103 			goto req_rx_irq_failed;
2104 		}
2105 	}
2106 	return rc;
2107 
2108 req_rx_irq_failed:
2109 	for (j = 0; j < i; j++) {
2110 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2111 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2112 	}
2113 	i = adapter->req_tx_queues;
2114 req_tx_irq_failed:
2115 	for (j = 0; j < i; j++) {
2116 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2117 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2118 	}
2119 	release_sub_crqs(adapter);
2120 	return rc;
2121 }
2122 
2123 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2124 {
2125 	struct device *dev = &adapter->vdev->dev;
2126 	struct ibmvnic_sub_crq_queue **allqueues;
2127 	int registered_queues = 0;
2128 	int total_queues;
2129 	int more = 0;
2130 	int i;
2131 
2132 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2133 
2134 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2135 	if (!allqueues)
2136 		return -1;
2137 
2138 	for (i = 0; i < total_queues; i++) {
2139 		allqueues[i] = init_sub_crq_queue(adapter);
2140 		if (!allqueues[i]) {
2141 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2142 			break;
2143 		}
2144 		registered_queues++;
2145 	}
2146 
2147 	/* Make sure we were able to register the minimum number of queues */
2148 	if (registered_queues <
2149 	    adapter->min_tx_queues + adapter->min_rx_queues) {
2150 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2151 		goto tx_failed;
2152 	}
2153 
2154 	/* Distribute the failed allocated queues*/
2155 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
2156 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2157 		switch (i % 3) {
2158 		case 0:
2159 			if (adapter->req_rx_queues > adapter->min_rx_queues)
2160 				adapter->req_rx_queues--;
2161 			else
2162 				more++;
2163 			break;
2164 		case 1:
2165 			if (adapter->req_tx_queues > adapter->min_tx_queues)
2166 				adapter->req_tx_queues--;
2167 			else
2168 				more++;
2169 			break;
2170 		}
2171 	}
2172 
2173 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
2174 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
2175 	if (!adapter->tx_scrq)
2176 		goto tx_failed;
2177 
2178 	for (i = 0; i < adapter->req_tx_queues; i++) {
2179 		adapter->tx_scrq[i] = allqueues[i];
2180 		adapter->tx_scrq[i]->pool_index = i;
2181 	}
2182 
2183 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
2184 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
2185 	if (!adapter->rx_scrq)
2186 		goto rx_failed;
2187 
2188 	for (i = 0; i < adapter->req_rx_queues; i++) {
2189 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
2190 		adapter->rx_scrq[i]->scrq_num = i;
2191 	}
2192 
2193 	kfree(allqueues);
2194 	return 0;
2195 
2196 rx_failed:
2197 	kfree(adapter->tx_scrq);
2198 	adapter->tx_scrq = NULL;
2199 tx_failed:
2200 	for (i = 0; i < registered_queues; i++)
2201 		release_sub_crq_queue(adapter, allqueues[i]);
2202 	kfree(allqueues);
2203 	return -1;
2204 }
2205 
2206 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
2207 {
2208 	struct device *dev = &adapter->vdev->dev;
2209 	union ibmvnic_crq crq;
2210 
2211 	if (!retry) {
2212 		/* Sub-CRQ entries are 32 byte long */
2213 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
2214 
2215 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
2216 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
2217 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
2218 			return;
2219 		}
2220 
2221 		/* Get the minimum between the queried max and the entries
2222 		 * that fit in our PAGE_SIZE
2223 		 */
2224 		adapter->req_tx_entries_per_subcrq =
2225 		    adapter->max_tx_entries_per_subcrq > entries_page ?
2226 		    entries_page : adapter->max_tx_entries_per_subcrq;
2227 		adapter->req_rx_add_entries_per_subcrq =
2228 		    adapter->max_rx_add_entries_per_subcrq > entries_page ?
2229 		    entries_page : adapter->max_rx_add_entries_per_subcrq;
2230 
2231 		adapter->req_tx_queues = adapter->opt_tx_comp_sub_queues;
2232 		adapter->req_rx_queues = adapter->opt_rx_comp_queues;
2233 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
2234 
2235 		adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
2236 	}
2237 
2238 	memset(&crq, 0, sizeof(crq));
2239 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
2240 	crq.request_capability.cmd = REQUEST_CAPABILITY;
2241 
2242 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
2243 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
2244 	atomic_inc(&adapter->running_cap_crqs);
2245 	ibmvnic_send_crq(adapter, &crq);
2246 
2247 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
2248 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
2249 	atomic_inc(&adapter->running_cap_crqs);
2250 	ibmvnic_send_crq(adapter, &crq);
2251 
2252 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
2253 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
2254 	atomic_inc(&adapter->running_cap_crqs);
2255 	ibmvnic_send_crq(adapter, &crq);
2256 
2257 	crq.request_capability.capability =
2258 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
2259 	crq.request_capability.number =
2260 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
2261 	atomic_inc(&adapter->running_cap_crqs);
2262 	ibmvnic_send_crq(adapter, &crq);
2263 
2264 	crq.request_capability.capability =
2265 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
2266 	crq.request_capability.number =
2267 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
2268 	atomic_inc(&adapter->running_cap_crqs);
2269 	ibmvnic_send_crq(adapter, &crq);
2270 
2271 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
2272 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
2273 	atomic_inc(&adapter->running_cap_crqs);
2274 	ibmvnic_send_crq(adapter, &crq);
2275 
2276 	if (adapter->netdev->flags & IFF_PROMISC) {
2277 		if (adapter->promisc_supported) {
2278 			crq.request_capability.capability =
2279 			    cpu_to_be16(PROMISC_REQUESTED);
2280 			crq.request_capability.number = cpu_to_be64(1);
2281 			atomic_inc(&adapter->running_cap_crqs);
2282 			ibmvnic_send_crq(adapter, &crq);
2283 		}
2284 	} else {
2285 		crq.request_capability.capability =
2286 		    cpu_to_be16(PROMISC_REQUESTED);
2287 		crq.request_capability.number = cpu_to_be64(0);
2288 		atomic_inc(&adapter->running_cap_crqs);
2289 		ibmvnic_send_crq(adapter, &crq);
2290 	}
2291 }
2292 
2293 static int pending_scrq(struct ibmvnic_adapter *adapter,
2294 			struct ibmvnic_sub_crq_queue *scrq)
2295 {
2296 	union sub_crq *entry = &scrq->msgs[scrq->cur];
2297 
2298 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
2299 		return 1;
2300 	else
2301 		return 0;
2302 }
2303 
2304 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
2305 					struct ibmvnic_sub_crq_queue *scrq)
2306 {
2307 	union sub_crq *entry;
2308 	unsigned long flags;
2309 
2310 	spin_lock_irqsave(&scrq->lock, flags);
2311 	entry = &scrq->msgs[scrq->cur];
2312 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2313 		if (++scrq->cur == scrq->size)
2314 			scrq->cur = 0;
2315 	} else {
2316 		entry = NULL;
2317 	}
2318 	spin_unlock_irqrestore(&scrq->lock, flags);
2319 
2320 	return entry;
2321 }
2322 
2323 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
2324 {
2325 	struct ibmvnic_crq_queue *queue = &adapter->crq;
2326 	union ibmvnic_crq *crq;
2327 
2328 	crq = &queue->msgs[queue->cur];
2329 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2330 		if (++queue->cur == queue->size)
2331 			queue->cur = 0;
2332 	} else {
2333 		crq = NULL;
2334 	}
2335 
2336 	return crq;
2337 }
2338 
2339 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
2340 		       union sub_crq *sub_crq)
2341 {
2342 	unsigned int ua = adapter->vdev->unit_address;
2343 	struct device *dev = &adapter->vdev->dev;
2344 	u64 *u64_crq = (u64 *)sub_crq;
2345 	int rc;
2346 
2347 	netdev_dbg(adapter->netdev,
2348 		   "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
2349 		   (unsigned long int)cpu_to_be64(remote_handle),
2350 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
2351 		   (unsigned long int)cpu_to_be64(u64_crq[1]),
2352 		   (unsigned long int)cpu_to_be64(u64_crq[2]),
2353 		   (unsigned long int)cpu_to_be64(u64_crq[3]));
2354 
2355 	/* Make sure the hypervisor sees the complete request */
2356 	mb();
2357 
2358 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
2359 				cpu_to_be64(remote_handle),
2360 				cpu_to_be64(u64_crq[0]),
2361 				cpu_to_be64(u64_crq[1]),
2362 				cpu_to_be64(u64_crq[2]),
2363 				cpu_to_be64(u64_crq[3]));
2364 
2365 	if (rc) {
2366 		if (rc == H_CLOSED)
2367 			dev_warn(dev, "CRQ Queue closed\n");
2368 		dev_err(dev, "Send error (rc=%d)\n", rc);
2369 	}
2370 
2371 	return rc;
2372 }
2373 
2374 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
2375 				u64 remote_handle, u64 ioba, u64 num_entries)
2376 {
2377 	unsigned int ua = adapter->vdev->unit_address;
2378 	struct device *dev = &adapter->vdev->dev;
2379 	int rc;
2380 
2381 	/* Make sure the hypervisor sees the complete request */
2382 	mb();
2383 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
2384 				cpu_to_be64(remote_handle),
2385 				ioba, num_entries);
2386 
2387 	if (rc) {
2388 		if (rc == H_CLOSED)
2389 			dev_warn(dev, "CRQ Queue closed\n");
2390 		dev_err(dev, "Send (indirect) error (rc=%d)\n", rc);
2391 	}
2392 
2393 	return rc;
2394 }
2395 
2396 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
2397 			    union ibmvnic_crq *crq)
2398 {
2399 	unsigned int ua = adapter->vdev->unit_address;
2400 	struct device *dev = &adapter->vdev->dev;
2401 	u64 *u64_crq = (u64 *)crq;
2402 	int rc;
2403 
2404 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
2405 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
2406 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
2407 
2408 	/* Make sure the hypervisor sees the complete request */
2409 	mb();
2410 
2411 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
2412 				cpu_to_be64(u64_crq[0]),
2413 				cpu_to_be64(u64_crq[1]));
2414 
2415 	if (rc) {
2416 		if (rc == H_CLOSED)
2417 			dev_warn(dev, "CRQ Queue closed\n");
2418 		dev_warn(dev, "Send error (rc=%d)\n", rc);
2419 	}
2420 
2421 	return rc;
2422 }
2423 
2424 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
2425 {
2426 	union ibmvnic_crq crq;
2427 
2428 	memset(&crq, 0, sizeof(crq));
2429 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
2430 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
2431 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
2432 
2433 	return ibmvnic_send_crq(adapter, &crq);
2434 }
2435 
2436 static int send_version_xchg(struct ibmvnic_adapter *adapter)
2437 {
2438 	union ibmvnic_crq crq;
2439 
2440 	memset(&crq, 0, sizeof(crq));
2441 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
2442 	crq.version_exchange.cmd = VERSION_EXCHANGE;
2443 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
2444 
2445 	return ibmvnic_send_crq(adapter, &crq);
2446 }
2447 
2448 static void send_login(struct ibmvnic_adapter *adapter)
2449 {
2450 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
2451 	struct ibmvnic_login_buffer *login_buffer;
2452 	struct device *dev = &adapter->vdev->dev;
2453 	dma_addr_t rsp_buffer_token;
2454 	dma_addr_t buffer_token;
2455 	size_t rsp_buffer_size;
2456 	union ibmvnic_crq crq;
2457 	size_t buffer_size;
2458 	__be64 *tx_list_p;
2459 	__be64 *rx_list_p;
2460 	int i;
2461 
2462 	buffer_size =
2463 	    sizeof(struct ibmvnic_login_buffer) +
2464 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues);
2465 
2466 	login_buffer = kmalloc(buffer_size, GFP_ATOMIC);
2467 	if (!login_buffer)
2468 		goto buf_alloc_failed;
2469 
2470 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
2471 				      DMA_TO_DEVICE);
2472 	if (dma_mapping_error(dev, buffer_token)) {
2473 		dev_err(dev, "Couldn't map login buffer\n");
2474 		goto buf_map_failed;
2475 	}
2476 
2477 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
2478 			  sizeof(u64) * adapter->req_tx_queues +
2479 			  sizeof(u64) * adapter->req_rx_queues +
2480 			  sizeof(u64) * adapter->req_rx_queues +
2481 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
2482 
2483 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
2484 	if (!login_rsp_buffer)
2485 		goto buf_rsp_alloc_failed;
2486 
2487 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
2488 					  rsp_buffer_size, DMA_FROM_DEVICE);
2489 	if (dma_mapping_error(dev, rsp_buffer_token)) {
2490 		dev_err(dev, "Couldn't map login rsp buffer\n");
2491 		goto buf_rsp_map_failed;
2492 	}
2493 
2494 	adapter->login_buf = login_buffer;
2495 	adapter->login_buf_token = buffer_token;
2496 	adapter->login_buf_sz = buffer_size;
2497 	adapter->login_rsp_buf = login_rsp_buffer;
2498 	adapter->login_rsp_buf_token = rsp_buffer_token;
2499 	adapter->login_rsp_buf_sz = rsp_buffer_size;
2500 
2501 	login_buffer->len = cpu_to_be32(buffer_size);
2502 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
2503 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
2504 	login_buffer->off_txcomp_subcrqs =
2505 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
2506 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
2507 	login_buffer->off_rxcomp_subcrqs =
2508 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
2509 			sizeof(u64) * adapter->req_tx_queues);
2510 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
2511 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
2512 
2513 	tx_list_p = (__be64 *)((char *)login_buffer +
2514 				      sizeof(struct ibmvnic_login_buffer));
2515 	rx_list_p = (__be64 *)((char *)login_buffer +
2516 				      sizeof(struct ibmvnic_login_buffer) +
2517 				      sizeof(u64) * adapter->req_tx_queues);
2518 
2519 	for (i = 0; i < adapter->req_tx_queues; i++) {
2520 		if (adapter->tx_scrq[i]) {
2521 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
2522 						   crq_num);
2523 		}
2524 	}
2525 
2526 	for (i = 0; i < adapter->req_rx_queues; i++) {
2527 		if (adapter->rx_scrq[i]) {
2528 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
2529 						   crq_num);
2530 		}
2531 	}
2532 
2533 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
2534 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
2535 		netdev_dbg(adapter->netdev, "%016lx\n",
2536 			   ((unsigned long int *)(adapter->login_buf))[i]);
2537 	}
2538 
2539 	memset(&crq, 0, sizeof(crq));
2540 	crq.login.first = IBMVNIC_CRQ_CMD;
2541 	crq.login.cmd = LOGIN;
2542 	crq.login.ioba = cpu_to_be32(buffer_token);
2543 	crq.login.len = cpu_to_be32(buffer_size);
2544 	ibmvnic_send_crq(adapter, &crq);
2545 
2546 	return;
2547 
2548 buf_rsp_map_failed:
2549 	kfree(login_rsp_buffer);
2550 buf_rsp_alloc_failed:
2551 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
2552 buf_map_failed:
2553 	kfree(login_buffer);
2554 buf_alloc_failed:
2555 	return;
2556 }
2557 
2558 static void send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
2559 			     u32 len, u8 map_id)
2560 {
2561 	union ibmvnic_crq crq;
2562 
2563 	memset(&crq, 0, sizeof(crq));
2564 	crq.request_map.first = IBMVNIC_CRQ_CMD;
2565 	crq.request_map.cmd = REQUEST_MAP;
2566 	crq.request_map.map_id = map_id;
2567 	crq.request_map.ioba = cpu_to_be32(addr);
2568 	crq.request_map.len = cpu_to_be32(len);
2569 	ibmvnic_send_crq(adapter, &crq);
2570 }
2571 
2572 static void send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
2573 {
2574 	union ibmvnic_crq crq;
2575 
2576 	memset(&crq, 0, sizeof(crq));
2577 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
2578 	crq.request_unmap.cmd = REQUEST_UNMAP;
2579 	crq.request_unmap.map_id = map_id;
2580 	ibmvnic_send_crq(adapter, &crq);
2581 }
2582 
2583 static void send_map_query(struct ibmvnic_adapter *adapter)
2584 {
2585 	union ibmvnic_crq crq;
2586 
2587 	memset(&crq, 0, sizeof(crq));
2588 	crq.query_map.first = IBMVNIC_CRQ_CMD;
2589 	crq.query_map.cmd = QUERY_MAP;
2590 	ibmvnic_send_crq(adapter, &crq);
2591 }
2592 
2593 /* Send a series of CRQs requesting various capabilities of the VNIC server */
2594 static void send_cap_queries(struct ibmvnic_adapter *adapter)
2595 {
2596 	union ibmvnic_crq crq;
2597 
2598 	atomic_set(&adapter->running_cap_crqs, 0);
2599 	memset(&crq, 0, sizeof(crq));
2600 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
2601 	crq.query_capability.cmd = QUERY_CAPABILITY;
2602 
2603 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
2604 	atomic_inc(&adapter->running_cap_crqs);
2605 	ibmvnic_send_crq(adapter, &crq);
2606 
2607 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
2608 	atomic_inc(&adapter->running_cap_crqs);
2609 	ibmvnic_send_crq(adapter, &crq);
2610 
2611 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
2612 	atomic_inc(&adapter->running_cap_crqs);
2613 	ibmvnic_send_crq(adapter, &crq);
2614 
2615 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
2616 	atomic_inc(&adapter->running_cap_crqs);
2617 	ibmvnic_send_crq(adapter, &crq);
2618 
2619 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
2620 	atomic_inc(&adapter->running_cap_crqs);
2621 	ibmvnic_send_crq(adapter, &crq);
2622 
2623 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
2624 	atomic_inc(&adapter->running_cap_crqs);
2625 	ibmvnic_send_crq(adapter, &crq);
2626 
2627 	crq.query_capability.capability =
2628 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
2629 	atomic_inc(&adapter->running_cap_crqs);
2630 	ibmvnic_send_crq(adapter, &crq);
2631 
2632 	crq.query_capability.capability =
2633 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
2634 	atomic_inc(&adapter->running_cap_crqs);
2635 	ibmvnic_send_crq(adapter, &crq);
2636 
2637 	crq.query_capability.capability =
2638 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
2639 	atomic_inc(&adapter->running_cap_crqs);
2640 	ibmvnic_send_crq(adapter, &crq);
2641 
2642 	crq.query_capability.capability =
2643 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
2644 	atomic_inc(&adapter->running_cap_crqs);
2645 	ibmvnic_send_crq(adapter, &crq);
2646 
2647 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
2648 	atomic_inc(&adapter->running_cap_crqs);
2649 	ibmvnic_send_crq(adapter, &crq);
2650 
2651 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
2652 	atomic_inc(&adapter->running_cap_crqs);
2653 	ibmvnic_send_crq(adapter, &crq);
2654 
2655 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
2656 	atomic_inc(&adapter->running_cap_crqs);
2657 	ibmvnic_send_crq(adapter, &crq);
2658 
2659 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
2660 	atomic_inc(&adapter->running_cap_crqs);
2661 	ibmvnic_send_crq(adapter, &crq);
2662 
2663 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
2664 	atomic_inc(&adapter->running_cap_crqs);
2665 	ibmvnic_send_crq(adapter, &crq);
2666 
2667 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
2668 	atomic_inc(&adapter->running_cap_crqs);
2669 	ibmvnic_send_crq(adapter, &crq);
2670 
2671 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
2672 	atomic_inc(&adapter->running_cap_crqs);
2673 	ibmvnic_send_crq(adapter, &crq);
2674 
2675 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
2676 	atomic_inc(&adapter->running_cap_crqs);
2677 	ibmvnic_send_crq(adapter, &crq);
2678 
2679 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
2680 	atomic_inc(&adapter->running_cap_crqs);
2681 	ibmvnic_send_crq(adapter, &crq);
2682 
2683 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
2684 	atomic_inc(&adapter->running_cap_crqs);
2685 	ibmvnic_send_crq(adapter, &crq);
2686 
2687 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
2688 	atomic_inc(&adapter->running_cap_crqs);
2689 	ibmvnic_send_crq(adapter, &crq);
2690 
2691 	crq.query_capability.capability =
2692 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
2693 	atomic_inc(&adapter->running_cap_crqs);
2694 	ibmvnic_send_crq(adapter, &crq);
2695 
2696 	crq.query_capability.capability =
2697 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
2698 	atomic_inc(&adapter->running_cap_crqs);
2699 	ibmvnic_send_crq(adapter, &crq);
2700 
2701 	crq.query_capability.capability =
2702 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
2703 	atomic_inc(&adapter->running_cap_crqs);
2704 	ibmvnic_send_crq(adapter, &crq);
2705 
2706 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
2707 	atomic_inc(&adapter->running_cap_crqs);
2708 	ibmvnic_send_crq(adapter, &crq);
2709 }
2710 
2711 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
2712 {
2713 	struct device *dev = &adapter->vdev->dev;
2714 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
2715 	union ibmvnic_crq crq;
2716 	int i;
2717 
2718 	dma_unmap_single(dev, adapter->ip_offload_tok,
2719 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
2720 
2721 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
2722 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
2723 		netdev_dbg(adapter->netdev, "%016lx\n",
2724 			   ((unsigned long int *)(buf))[i]);
2725 
2726 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
2727 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
2728 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
2729 		   buf->tcp_ipv4_chksum);
2730 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
2731 		   buf->tcp_ipv6_chksum);
2732 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
2733 		   buf->udp_ipv4_chksum);
2734 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
2735 		   buf->udp_ipv6_chksum);
2736 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
2737 		   buf->large_tx_ipv4);
2738 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
2739 		   buf->large_tx_ipv6);
2740 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
2741 		   buf->large_rx_ipv4);
2742 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
2743 		   buf->large_rx_ipv6);
2744 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
2745 		   buf->max_ipv4_header_size);
2746 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
2747 		   buf->max_ipv6_header_size);
2748 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
2749 		   buf->max_tcp_header_size);
2750 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
2751 		   buf->max_udp_header_size);
2752 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
2753 		   buf->max_large_tx_size);
2754 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
2755 		   buf->max_large_rx_size);
2756 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
2757 		   buf->ipv6_extension_header);
2758 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
2759 		   buf->tcp_pseudosum_req);
2760 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
2761 		   buf->num_ipv6_ext_headers);
2762 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
2763 		   buf->off_ipv6_ext_headers);
2764 
2765 	adapter->ip_offload_ctrl_tok =
2766 	    dma_map_single(dev, &adapter->ip_offload_ctrl,
2767 			   sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
2768 
2769 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
2770 		dev_err(dev, "Couldn't map ip offload control buffer\n");
2771 		return;
2772 	}
2773 
2774 	adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
2775 	adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
2776 	adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
2777 	adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
2778 	adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
2779 
2780 	/* large_tx/rx disabled for now, additional features needed */
2781 	adapter->ip_offload_ctrl.large_tx_ipv4 = 0;
2782 	adapter->ip_offload_ctrl.large_tx_ipv6 = 0;
2783 	adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
2784 	adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
2785 
2786 	adapter->netdev->features = NETIF_F_GSO;
2787 
2788 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
2789 		adapter->netdev->features |= NETIF_F_IP_CSUM;
2790 
2791 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
2792 		adapter->netdev->features |= NETIF_F_IPV6_CSUM;
2793 
2794 	if ((adapter->netdev->features &
2795 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
2796 		adapter->netdev->features |= NETIF_F_RXCSUM;
2797 
2798 	memset(&crq, 0, sizeof(crq));
2799 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
2800 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
2801 	crq.control_ip_offload.len =
2802 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
2803 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
2804 	ibmvnic_send_crq(adapter, &crq);
2805 }
2806 
2807 static void handle_error_info_rsp(union ibmvnic_crq *crq,
2808 				  struct ibmvnic_adapter *adapter)
2809 {
2810 	struct device *dev = &adapter->vdev->dev;
2811 	struct ibmvnic_error_buff *error_buff, *tmp;
2812 	unsigned long flags;
2813 	bool found = false;
2814 	int i;
2815 
2816 	if (!crq->request_error_rsp.rc.code) {
2817 		dev_info(dev, "Request Error Rsp returned with rc=%x\n",
2818 			 crq->request_error_rsp.rc.code);
2819 		return;
2820 	}
2821 
2822 	spin_lock_irqsave(&adapter->error_list_lock, flags);
2823 	list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list)
2824 		if (error_buff->error_id == crq->request_error_rsp.error_id) {
2825 			found = true;
2826 			list_del(&error_buff->list);
2827 			break;
2828 		}
2829 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
2830 
2831 	if (!found) {
2832 		dev_err(dev, "Couldn't find error id %x\n",
2833 			be32_to_cpu(crq->request_error_rsp.error_id));
2834 		return;
2835 	}
2836 
2837 	dev_err(dev, "Detailed info for error id %x:",
2838 		be32_to_cpu(crq->request_error_rsp.error_id));
2839 
2840 	for (i = 0; i < error_buff->len; i++) {
2841 		pr_cont("%02x", (int)error_buff->buff[i]);
2842 		if (i % 8 == 7)
2843 			pr_cont(" ");
2844 	}
2845 	pr_cont("\n");
2846 
2847 	dma_unmap_single(dev, error_buff->dma, error_buff->len,
2848 			 DMA_FROM_DEVICE);
2849 	kfree(error_buff->buff);
2850 	kfree(error_buff);
2851 }
2852 
2853 static void request_error_information(struct ibmvnic_adapter *adapter,
2854 				      union ibmvnic_crq *err_crq)
2855 {
2856 	struct device *dev = &adapter->vdev->dev;
2857 	struct net_device *netdev = adapter->netdev;
2858 	struct ibmvnic_error_buff *error_buff;
2859 	unsigned long timeout = msecs_to_jiffies(30000);
2860 	union ibmvnic_crq crq;
2861 	unsigned long flags;
2862 	int rc, detail_len;
2863 
2864 	error_buff = kmalloc(sizeof(*error_buff), GFP_ATOMIC);
2865 	if (!error_buff)
2866 		return;
2867 
2868 	detail_len = be32_to_cpu(err_crq->error_indication.detail_error_sz);
2869 	error_buff->buff = kmalloc(detail_len, GFP_ATOMIC);
2870 	if (!error_buff->buff) {
2871 		kfree(error_buff);
2872 		return;
2873 	}
2874 
2875 	error_buff->dma = dma_map_single(dev, error_buff->buff, detail_len,
2876 					 DMA_FROM_DEVICE);
2877 	if (dma_mapping_error(dev, error_buff->dma)) {
2878 		netdev_err(netdev, "Couldn't map error buffer\n");
2879 		kfree(error_buff->buff);
2880 		kfree(error_buff);
2881 		return;
2882 	}
2883 
2884 	error_buff->len = detail_len;
2885 	error_buff->error_id = err_crq->error_indication.error_id;
2886 
2887 	spin_lock_irqsave(&adapter->error_list_lock, flags);
2888 	list_add_tail(&error_buff->list, &adapter->errors);
2889 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
2890 
2891 	memset(&crq, 0, sizeof(crq));
2892 	crq.request_error_info.first = IBMVNIC_CRQ_CMD;
2893 	crq.request_error_info.cmd = REQUEST_ERROR_INFO;
2894 	crq.request_error_info.ioba = cpu_to_be32(error_buff->dma);
2895 	crq.request_error_info.len = cpu_to_be32(detail_len);
2896 	crq.request_error_info.error_id = err_crq->error_indication.error_id;
2897 
2898 	rc = ibmvnic_send_crq(adapter, &crq);
2899 	if (rc) {
2900 		netdev_err(netdev, "failed to request error information\n");
2901 		goto err_info_fail;
2902 	}
2903 
2904 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
2905 		netdev_err(netdev, "timeout waiting for error information\n");
2906 		goto err_info_fail;
2907 	}
2908 
2909 	return;
2910 
2911 err_info_fail:
2912 	spin_lock_irqsave(&adapter->error_list_lock, flags);
2913 	list_del(&error_buff->list);
2914 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
2915 
2916 	kfree(error_buff->buff);
2917 	kfree(error_buff);
2918 }
2919 
2920 static void handle_error_indication(union ibmvnic_crq *crq,
2921 				    struct ibmvnic_adapter *adapter)
2922 {
2923 	struct device *dev = &adapter->vdev->dev;
2924 
2925 	dev_err(dev, "Firmware reports %serror id %x, cause %d\n",
2926 		crq->error_indication.flags
2927 			& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
2928 		be32_to_cpu(crq->error_indication.error_id),
2929 		be16_to_cpu(crq->error_indication.error_cause));
2930 
2931 	if (be32_to_cpu(crq->error_indication.error_id))
2932 		request_error_information(adapter, crq);
2933 
2934 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
2935 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
2936 	else
2937 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
2938 }
2939 
2940 static void handle_change_mac_rsp(union ibmvnic_crq *crq,
2941 				  struct ibmvnic_adapter *adapter)
2942 {
2943 	struct net_device *netdev = adapter->netdev;
2944 	struct device *dev = &adapter->vdev->dev;
2945 	long rc;
2946 
2947 	rc = crq->change_mac_addr_rsp.rc.code;
2948 	if (rc) {
2949 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
2950 		return;
2951 	}
2952 	memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
2953 	       ETH_ALEN);
2954 }
2955 
2956 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
2957 				   struct ibmvnic_adapter *adapter)
2958 {
2959 	struct device *dev = &adapter->vdev->dev;
2960 	u64 *req_value;
2961 	char *name;
2962 
2963 	atomic_dec(&adapter->running_cap_crqs);
2964 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
2965 	case REQ_TX_QUEUES:
2966 		req_value = &adapter->req_tx_queues;
2967 		name = "tx";
2968 		break;
2969 	case REQ_RX_QUEUES:
2970 		req_value = &adapter->req_rx_queues;
2971 		name = "rx";
2972 		break;
2973 	case REQ_RX_ADD_QUEUES:
2974 		req_value = &adapter->req_rx_add_queues;
2975 		name = "rx_add";
2976 		break;
2977 	case REQ_TX_ENTRIES_PER_SUBCRQ:
2978 		req_value = &adapter->req_tx_entries_per_subcrq;
2979 		name = "tx_entries_per_subcrq";
2980 		break;
2981 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
2982 		req_value = &adapter->req_rx_add_entries_per_subcrq;
2983 		name = "rx_add_entries_per_subcrq";
2984 		break;
2985 	case REQ_MTU:
2986 		req_value = &adapter->req_mtu;
2987 		name = "mtu";
2988 		break;
2989 	case PROMISC_REQUESTED:
2990 		req_value = &adapter->promisc;
2991 		name = "promisc";
2992 		break;
2993 	default:
2994 		dev_err(dev, "Got invalid cap request rsp %d\n",
2995 			crq->request_capability.capability);
2996 		return;
2997 	}
2998 
2999 	switch (crq->request_capability_rsp.rc.code) {
3000 	case SUCCESS:
3001 		break;
3002 	case PARTIALSUCCESS:
3003 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3004 			 *req_value,
3005 			 (long int)be64_to_cpu(crq->request_capability_rsp.
3006 					       number), name);
3007 		release_sub_crqs(adapter);
3008 		*req_value = be64_to_cpu(crq->request_capability_rsp.number);
3009 		ibmvnic_send_req_caps(adapter, 1);
3010 		return;
3011 	default:
3012 		dev_err(dev, "Error %d in request cap rsp\n",
3013 			crq->request_capability_rsp.rc.code);
3014 		return;
3015 	}
3016 
3017 	/* Done receiving requested capabilities, query IP offload support */
3018 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
3019 		union ibmvnic_crq newcrq;
3020 		int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3021 		struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
3022 		    &adapter->ip_offload_buf;
3023 
3024 		adapter->wait_capability = false;
3025 		adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
3026 							 buf_sz,
3027 							 DMA_FROM_DEVICE);
3028 
3029 		if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3030 			if (!firmware_has_feature(FW_FEATURE_CMO))
3031 				dev_err(dev, "Couldn't map offload buffer\n");
3032 			return;
3033 		}
3034 
3035 		memset(&newcrq, 0, sizeof(newcrq));
3036 		newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3037 		newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3038 		newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
3039 		newcrq.query_ip_offload.ioba =
3040 		    cpu_to_be32(adapter->ip_offload_tok);
3041 
3042 		ibmvnic_send_crq(adapter, &newcrq);
3043 	}
3044 }
3045 
3046 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
3047 			    struct ibmvnic_adapter *adapter)
3048 {
3049 	struct device *dev = &adapter->vdev->dev;
3050 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
3051 	struct ibmvnic_login_buffer *login = adapter->login_buf;
3052 	int i;
3053 
3054 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
3055 			 DMA_BIDIRECTIONAL);
3056 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
3057 			 adapter->login_rsp_buf_sz, DMA_BIDIRECTIONAL);
3058 
3059 	/* If the number of queues requested can't be allocated by the
3060 	 * server, the login response will return with code 1. We will need
3061 	 * to resend the login buffer with fewer queues requested.
3062 	 */
3063 	if (login_rsp_crq->generic.rc.code) {
3064 		adapter->renegotiate = true;
3065 		complete(&adapter->init_done);
3066 		return 0;
3067 	}
3068 
3069 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
3070 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
3071 		netdev_dbg(adapter->netdev, "%016lx\n",
3072 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
3073 	}
3074 
3075 	/* Sanity checks */
3076 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
3077 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
3078 	     adapter->req_rx_add_queues !=
3079 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
3080 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
3081 		ibmvnic_remove(adapter->vdev);
3082 		return -EIO;
3083 	}
3084 	complete(&adapter->init_done);
3085 
3086 	return 0;
3087 }
3088 
3089 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
3090 				     struct ibmvnic_adapter *adapter)
3091 {
3092 	struct device *dev = &adapter->vdev->dev;
3093 	long rc;
3094 
3095 	rc = crq->request_unmap_rsp.rc.code;
3096 	if (rc)
3097 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
3098 }
3099 
3100 static void handle_query_map_rsp(union ibmvnic_crq *crq,
3101 				 struct ibmvnic_adapter *adapter)
3102 {
3103 	struct net_device *netdev = adapter->netdev;
3104 	struct device *dev = &adapter->vdev->dev;
3105 	long rc;
3106 
3107 	rc = crq->query_map_rsp.rc.code;
3108 	if (rc) {
3109 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
3110 		return;
3111 	}
3112 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
3113 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
3114 		   crq->query_map_rsp.free_pages);
3115 }
3116 
3117 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
3118 				 struct ibmvnic_adapter *adapter)
3119 {
3120 	struct net_device *netdev = adapter->netdev;
3121 	struct device *dev = &adapter->vdev->dev;
3122 	long rc;
3123 
3124 	atomic_dec(&adapter->running_cap_crqs);
3125 	netdev_dbg(netdev, "Outstanding queries: %d\n",
3126 		   atomic_read(&adapter->running_cap_crqs));
3127 	rc = crq->query_capability.rc.code;
3128 	if (rc) {
3129 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
3130 		goto out;
3131 	}
3132 
3133 	switch (be16_to_cpu(crq->query_capability.capability)) {
3134 	case MIN_TX_QUEUES:
3135 		adapter->min_tx_queues =
3136 		    be64_to_cpu(crq->query_capability.number);
3137 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
3138 			   adapter->min_tx_queues);
3139 		break;
3140 	case MIN_RX_QUEUES:
3141 		adapter->min_rx_queues =
3142 		    be64_to_cpu(crq->query_capability.number);
3143 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
3144 			   adapter->min_rx_queues);
3145 		break;
3146 	case MIN_RX_ADD_QUEUES:
3147 		adapter->min_rx_add_queues =
3148 		    be64_to_cpu(crq->query_capability.number);
3149 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
3150 			   adapter->min_rx_add_queues);
3151 		break;
3152 	case MAX_TX_QUEUES:
3153 		adapter->max_tx_queues =
3154 		    be64_to_cpu(crq->query_capability.number);
3155 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
3156 			   adapter->max_tx_queues);
3157 		break;
3158 	case MAX_RX_QUEUES:
3159 		adapter->max_rx_queues =
3160 		    be64_to_cpu(crq->query_capability.number);
3161 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
3162 			   adapter->max_rx_queues);
3163 		break;
3164 	case MAX_RX_ADD_QUEUES:
3165 		adapter->max_rx_add_queues =
3166 		    be64_to_cpu(crq->query_capability.number);
3167 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
3168 			   adapter->max_rx_add_queues);
3169 		break;
3170 	case MIN_TX_ENTRIES_PER_SUBCRQ:
3171 		adapter->min_tx_entries_per_subcrq =
3172 		    be64_to_cpu(crq->query_capability.number);
3173 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
3174 			   adapter->min_tx_entries_per_subcrq);
3175 		break;
3176 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
3177 		adapter->min_rx_add_entries_per_subcrq =
3178 		    be64_to_cpu(crq->query_capability.number);
3179 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
3180 			   adapter->min_rx_add_entries_per_subcrq);
3181 		break;
3182 	case MAX_TX_ENTRIES_PER_SUBCRQ:
3183 		adapter->max_tx_entries_per_subcrq =
3184 		    be64_to_cpu(crq->query_capability.number);
3185 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
3186 			   adapter->max_tx_entries_per_subcrq);
3187 		break;
3188 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
3189 		adapter->max_rx_add_entries_per_subcrq =
3190 		    be64_to_cpu(crq->query_capability.number);
3191 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
3192 			   adapter->max_rx_add_entries_per_subcrq);
3193 		break;
3194 	case TCP_IP_OFFLOAD:
3195 		adapter->tcp_ip_offload =
3196 		    be64_to_cpu(crq->query_capability.number);
3197 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
3198 			   adapter->tcp_ip_offload);
3199 		break;
3200 	case PROMISC_SUPPORTED:
3201 		adapter->promisc_supported =
3202 		    be64_to_cpu(crq->query_capability.number);
3203 		netdev_dbg(netdev, "promisc_supported = %lld\n",
3204 			   adapter->promisc_supported);
3205 		break;
3206 	case MIN_MTU:
3207 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
3208 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
3209 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
3210 		break;
3211 	case MAX_MTU:
3212 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
3213 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
3214 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
3215 		break;
3216 	case MAX_MULTICAST_FILTERS:
3217 		adapter->max_multicast_filters =
3218 		    be64_to_cpu(crq->query_capability.number);
3219 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
3220 			   adapter->max_multicast_filters);
3221 		break;
3222 	case VLAN_HEADER_INSERTION:
3223 		adapter->vlan_header_insertion =
3224 		    be64_to_cpu(crq->query_capability.number);
3225 		if (adapter->vlan_header_insertion)
3226 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
3227 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
3228 			   adapter->vlan_header_insertion);
3229 		break;
3230 	case RX_VLAN_HEADER_INSERTION:
3231 		adapter->rx_vlan_header_insertion =
3232 		    be64_to_cpu(crq->query_capability.number);
3233 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
3234 			   adapter->rx_vlan_header_insertion);
3235 		break;
3236 	case MAX_TX_SG_ENTRIES:
3237 		adapter->max_tx_sg_entries =
3238 		    be64_to_cpu(crq->query_capability.number);
3239 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
3240 			   adapter->max_tx_sg_entries);
3241 		break;
3242 	case RX_SG_SUPPORTED:
3243 		adapter->rx_sg_supported =
3244 		    be64_to_cpu(crq->query_capability.number);
3245 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
3246 			   adapter->rx_sg_supported);
3247 		break;
3248 	case OPT_TX_COMP_SUB_QUEUES:
3249 		adapter->opt_tx_comp_sub_queues =
3250 		    be64_to_cpu(crq->query_capability.number);
3251 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
3252 			   adapter->opt_tx_comp_sub_queues);
3253 		break;
3254 	case OPT_RX_COMP_QUEUES:
3255 		adapter->opt_rx_comp_queues =
3256 		    be64_to_cpu(crq->query_capability.number);
3257 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
3258 			   adapter->opt_rx_comp_queues);
3259 		break;
3260 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
3261 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
3262 		    be64_to_cpu(crq->query_capability.number);
3263 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
3264 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
3265 		break;
3266 	case OPT_TX_ENTRIES_PER_SUBCRQ:
3267 		adapter->opt_tx_entries_per_subcrq =
3268 		    be64_to_cpu(crq->query_capability.number);
3269 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
3270 			   adapter->opt_tx_entries_per_subcrq);
3271 		break;
3272 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
3273 		adapter->opt_rxba_entries_per_subcrq =
3274 		    be64_to_cpu(crq->query_capability.number);
3275 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
3276 			   adapter->opt_rxba_entries_per_subcrq);
3277 		break;
3278 	case TX_RX_DESC_REQ:
3279 		adapter->tx_rx_desc_req = crq->query_capability.number;
3280 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
3281 			   adapter->tx_rx_desc_req);
3282 		break;
3283 
3284 	default:
3285 		netdev_err(netdev, "Got invalid cap rsp %d\n",
3286 			   crq->query_capability.capability);
3287 	}
3288 
3289 out:
3290 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
3291 		adapter->wait_capability = false;
3292 		ibmvnic_send_req_caps(adapter, 0);
3293 	}
3294 }
3295 
3296 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
3297 			       struct ibmvnic_adapter *adapter)
3298 {
3299 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
3300 	struct net_device *netdev = adapter->netdev;
3301 	struct device *dev = &adapter->vdev->dev;
3302 	u64 *u64_crq = (u64 *)crq;
3303 	long rc;
3304 
3305 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
3306 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3307 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
3308 	switch (gen_crq->first) {
3309 	case IBMVNIC_CRQ_INIT_RSP:
3310 		switch (gen_crq->cmd) {
3311 		case IBMVNIC_CRQ_INIT:
3312 			dev_info(dev, "Partner initialized\n");
3313 			adapter->from_passive_init = true;
3314 			complete(&adapter->init_done);
3315 			break;
3316 		case IBMVNIC_CRQ_INIT_COMPLETE:
3317 			dev_info(dev, "Partner initialization complete\n");
3318 			send_version_xchg(adapter);
3319 			break;
3320 		default:
3321 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
3322 		}
3323 		return;
3324 	case IBMVNIC_CRQ_XPORT_EVENT:
3325 		netif_carrier_off(netdev);
3326 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
3327 			dev_info(dev, "Migrated, re-enabling adapter\n");
3328 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
3329 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
3330 			dev_info(dev, "Backing device failover detected\n");
3331 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
3332 		} else {
3333 			/* The adapter lost the connection */
3334 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
3335 				gen_crq->cmd);
3336 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3337 		}
3338 		return;
3339 	case IBMVNIC_CRQ_CMD_RSP:
3340 		break;
3341 	default:
3342 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
3343 			gen_crq->first);
3344 		return;
3345 	}
3346 
3347 	switch (gen_crq->cmd) {
3348 	case VERSION_EXCHANGE_RSP:
3349 		rc = crq->version_exchange_rsp.rc.code;
3350 		if (rc) {
3351 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
3352 			break;
3353 		}
3354 		dev_info(dev, "Partner protocol version is %d\n",
3355 			 crq->version_exchange_rsp.version);
3356 		if (be16_to_cpu(crq->version_exchange_rsp.version) <
3357 		    ibmvnic_version)
3358 			ibmvnic_version =
3359 			    be16_to_cpu(crq->version_exchange_rsp.version);
3360 		send_cap_queries(adapter);
3361 		break;
3362 	case QUERY_CAPABILITY_RSP:
3363 		handle_query_cap_rsp(crq, adapter);
3364 		break;
3365 	case QUERY_MAP_RSP:
3366 		handle_query_map_rsp(crq, adapter);
3367 		break;
3368 	case REQUEST_MAP_RSP:
3369 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
3370 		complete(&adapter->fw_done);
3371 		break;
3372 	case REQUEST_UNMAP_RSP:
3373 		handle_request_unmap_rsp(crq, adapter);
3374 		break;
3375 	case REQUEST_CAPABILITY_RSP:
3376 		handle_request_cap_rsp(crq, adapter);
3377 		break;
3378 	case LOGIN_RSP:
3379 		netdev_dbg(netdev, "Got Login Response\n");
3380 		handle_login_rsp(crq, adapter);
3381 		break;
3382 	case LOGICAL_LINK_STATE_RSP:
3383 		netdev_dbg(netdev,
3384 			   "Got Logical Link State Response, state: %d rc: %d\n",
3385 			   crq->logical_link_state_rsp.link_state,
3386 			   crq->logical_link_state_rsp.rc.code);
3387 		adapter->logical_link_state =
3388 		    crq->logical_link_state_rsp.link_state;
3389 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
3390 		complete(&adapter->init_done);
3391 		break;
3392 	case LINK_STATE_INDICATION:
3393 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
3394 		adapter->phys_link_state =
3395 		    crq->link_state_indication.phys_link_state;
3396 		adapter->logical_link_state =
3397 		    crq->link_state_indication.logical_link_state;
3398 		break;
3399 	case CHANGE_MAC_ADDR_RSP:
3400 		netdev_dbg(netdev, "Got MAC address change Response\n");
3401 		handle_change_mac_rsp(crq, adapter);
3402 		break;
3403 	case ERROR_INDICATION:
3404 		netdev_dbg(netdev, "Got Error Indication\n");
3405 		handle_error_indication(crq, adapter);
3406 		break;
3407 	case REQUEST_ERROR_RSP:
3408 		netdev_dbg(netdev, "Got Error Detail Response\n");
3409 		handle_error_info_rsp(crq, adapter);
3410 		break;
3411 	case REQUEST_STATISTICS_RSP:
3412 		netdev_dbg(netdev, "Got Statistics Response\n");
3413 		complete(&adapter->stats_done);
3414 		break;
3415 	case QUERY_IP_OFFLOAD_RSP:
3416 		netdev_dbg(netdev, "Got Query IP offload Response\n");
3417 		handle_query_ip_offload_rsp(adapter);
3418 		break;
3419 	case MULTICAST_CTRL_RSP:
3420 		netdev_dbg(netdev, "Got multicast control Response\n");
3421 		break;
3422 	case CONTROL_IP_OFFLOAD_RSP:
3423 		netdev_dbg(netdev, "Got Control IP offload Response\n");
3424 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
3425 				 sizeof(adapter->ip_offload_ctrl),
3426 				 DMA_TO_DEVICE);
3427 		complete(&adapter->init_done);
3428 		break;
3429 	case COLLECT_FW_TRACE_RSP:
3430 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
3431 		complete(&adapter->fw_done);
3432 		break;
3433 	default:
3434 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
3435 			   gen_crq->cmd);
3436 	}
3437 }
3438 
3439 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
3440 {
3441 	struct ibmvnic_adapter *adapter = instance;
3442 
3443 	tasklet_schedule(&adapter->tasklet);
3444 	return IRQ_HANDLED;
3445 }
3446 
3447 static void ibmvnic_tasklet(void *data)
3448 {
3449 	struct ibmvnic_adapter *adapter = data;
3450 	struct ibmvnic_crq_queue *queue = &adapter->crq;
3451 	union ibmvnic_crq *crq;
3452 	unsigned long flags;
3453 	bool done = false;
3454 
3455 	spin_lock_irqsave(&queue->lock, flags);
3456 	while (!done) {
3457 		/* Pull all the valid messages off the CRQ */
3458 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
3459 			ibmvnic_handle_crq(crq, adapter);
3460 			crq->generic.first = 0;
3461 		}
3462 
3463 		/* remain in tasklet until all
3464 		 * capabilities responses are received
3465 		 */
3466 		if (!adapter->wait_capability)
3467 			done = true;
3468 	}
3469 	/* if capabilities CRQ's were sent in this tasklet, the following
3470 	 * tasklet must wait until all responses are received
3471 	 */
3472 	if (atomic_read(&adapter->running_cap_crqs) != 0)
3473 		adapter->wait_capability = true;
3474 	spin_unlock_irqrestore(&queue->lock, flags);
3475 }
3476 
3477 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
3478 {
3479 	struct vio_dev *vdev = adapter->vdev;
3480 	int rc;
3481 
3482 	do {
3483 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
3484 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
3485 
3486 	if (rc)
3487 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
3488 
3489 	return rc;
3490 }
3491 
3492 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
3493 {
3494 	struct ibmvnic_crq_queue *crq = &adapter->crq;
3495 	struct device *dev = &adapter->vdev->dev;
3496 	struct vio_dev *vdev = adapter->vdev;
3497 	int rc;
3498 
3499 	/* Close the CRQ */
3500 	do {
3501 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3502 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3503 
3504 	/* Clean out the queue */
3505 	memset(crq->msgs, 0, PAGE_SIZE);
3506 	crq->cur = 0;
3507 
3508 	/* And re-open it again */
3509 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
3510 				crq->msg_token, PAGE_SIZE);
3511 
3512 	if (rc == H_CLOSED)
3513 		/* Adapter is good, but other end is not ready */
3514 		dev_warn(dev, "Partner adapter not ready\n");
3515 	else if (rc != 0)
3516 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
3517 
3518 	return rc;
3519 }
3520 
3521 static void release_crq_queue(struct ibmvnic_adapter *adapter)
3522 {
3523 	struct ibmvnic_crq_queue *crq = &adapter->crq;
3524 	struct vio_dev *vdev = adapter->vdev;
3525 	long rc;
3526 
3527 	if (!crq->msgs)
3528 		return;
3529 
3530 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
3531 	free_irq(vdev->irq, adapter);
3532 	tasklet_kill(&adapter->tasklet);
3533 	do {
3534 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3535 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3536 
3537 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
3538 			 DMA_BIDIRECTIONAL);
3539 	free_page((unsigned long)crq->msgs);
3540 	crq->msgs = NULL;
3541 }
3542 
3543 static int init_crq_queue(struct ibmvnic_adapter *adapter)
3544 {
3545 	struct ibmvnic_crq_queue *crq = &adapter->crq;
3546 	struct device *dev = &adapter->vdev->dev;
3547 	struct vio_dev *vdev = adapter->vdev;
3548 	int rc, retrc = -ENOMEM;
3549 
3550 	if (crq->msgs)
3551 		return 0;
3552 
3553 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
3554 	/* Should we allocate more than one page? */
3555 
3556 	if (!crq->msgs)
3557 		return -ENOMEM;
3558 
3559 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
3560 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
3561 					DMA_BIDIRECTIONAL);
3562 	if (dma_mapping_error(dev, crq->msg_token))
3563 		goto map_failed;
3564 
3565 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
3566 				crq->msg_token, PAGE_SIZE);
3567 
3568 	if (rc == H_RESOURCE)
3569 		/* maybe kexecing and resource is busy. try a reset */
3570 		rc = ibmvnic_reset_crq(adapter);
3571 	retrc = rc;
3572 
3573 	if (rc == H_CLOSED) {
3574 		dev_warn(dev, "Partner adapter not ready\n");
3575 	} else if (rc) {
3576 		dev_warn(dev, "Error %d opening adapter\n", rc);
3577 		goto reg_crq_failed;
3578 	}
3579 
3580 	retrc = 0;
3581 
3582 	tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
3583 		     (unsigned long)adapter);
3584 
3585 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
3586 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
3587 			 adapter);
3588 	if (rc) {
3589 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
3590 			vdev->irq, rc);
3591 		goto req_irq_failed;
3592 	}
3593 
3594 	rc = vio_enable_interrupts(vdev);
3595 	if (rc) {
3596 		dev_err(dev, "Error %d enabling interrupts\n", rc);
3597 		goto req_irq_failed;
3598 	}
3599 
3600 	crq->cur = 0;
3601 	spin_lock_init(&crq->lock);
3602 
3603 	return retrc;
3604 
3605 req_irq_failed:
3606 	tasklet_kill(&adapter->tasklet);
3607 	do {
3608 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3609 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3610 reg_crq_failed:
3611 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
3612 map_failed:
3613 	free_page((unsigned long)crq->msgs);
3614 	crq->msgs = NULL;
3615 	return retrc;
3616 }
3617 
3618 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
3619 {
3620 	struct device *dev = &adapter->vdev->dev;
3621 	unsigned long timeout = msecs_to_jiffies(30000);
3622 	int rc;
3623 
3624 	if (adapter->resetting) {
3625 		rc = ibmvnic_reset_crq(adapter);
3626 		if (!rc)
3627 			rc = vio_enable_interrupts(adapter->vdev);
3628 	} else {
3629 		rc = init_crq_queue(adapter);
3630 	}
3631 
3632 	if (rc) {
3633 		dev_err(dev, "Couldn't initialize crq. rc=%d\n", rc);
3634 		return rc;
3635 	}
3636 
3637 	adapter->from_passive_init = false;
3638 
3639 	init_completion(&adapter->init_done);
3640 	adapter->init_done_rc = 0;
3641 	ibmvnic_send_crq_init(adapter);
3642 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
3643 		dev_err(dev, "Initialization sequence timed out\n");
3644 		return -1;
3645 	}
3646 
3647 	if (adapter->init_done_rc) {
3648 		release_crq_queue(adapter);
3649 		return adapter->init_done_rc;
3650 	}
3651 
3652 	if (adapter->from_passive_init) {
3653 		adapter->state = VNIC_OPEN;
3654 		adapter->from_passive_init = false;
3655 		return -1;
3656 	}
3657 
3658 	if (adapter->resetting)
3659 		rc = reset_sub_crq_queues(adapter);
3660 	else
3661 		rc = init_sub_crqs(adapter);
3662 	if (rc) {
3663 		dev_err(dev, "Initialization of sub crqs failed\n");
3664 		release_crq_queue(adapter);
3665 		return rc;
3666 	}
3667 
3668 	rc = init_sub_crq_irqs(adapter);
3669 	if (rc) {
3670 		dev_err(dev, "Failed to initialize sub crq irqs\n");
3671 		release_crq_queue(adapter);
3672 	}
3673 
3674 	return rc;
3675 }
3676 
3677 static struct device_attribute dev_attr_failover;
3678 
3679 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
3680 {
3681 	struct ibmvnic_adapter *adapter;
3682 	struct net_device *netdev;
3683 	unsigned char *mac_addr_p;
3684 	int rc;
3685 
3686 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
3687 		dev->unit_address);
3688 
3689 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
3690 							VETH_MAC_ADDR, NULL);
3691 	if (!mac_addr_p) {
3692 		dev_err(&dev->dev,
3693 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
3694 			__FILE__, __LINE__);
3695 		return 0;
3696 	}
3697 
3698 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
3699 				   IBMVNIC_MAX_TX_QUEUES);
3700 	if (!netdev)
3701 		return -ENOMEM;
3702 
3703 	adapter = netdev_priv(netdev);
3704 	adapter->state = VNIC_PROBING;
3705 	dev_set_drvdata(&dev->dev, netdev);
3706 	adapter->vdev = dev;
3707 	adapter->netdev = netdev;
3708 
3709 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
3710 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
3711 	netdev->irq = dev->irq;
3712 	netdev->netdev_ops = &ibmvnic_netdev_ops;
3713 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
3714 	SET_NETDEV_DEV(netdev, &dev->dev);
3715 
3716 	spin_lock_init(&adapter->stats_lock);
3717 
3718 	INIT_LIST_HEAD(&adapter->errors);
3719 	spin_lock_init(&adapter->error_list_lock);
3720 
3721 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
3722 	INIT_LIST_HEAD(&adapter->rwi_list);
3723 	mutex_init(&adapter->reset_lock);
3724 	mutex_init(&adapter->rwi_lock);
3725 	adapter->resetting = false;
3726 
3727 	do {
3728 		rc = ibmvnic_init(adapter);
3729 		if (rc && rc != EAGAIN) {
3730 			free_netdev(netdev);
3731 			return rc;
3732 		}
3733 	} while (rc == EAGAIN);
3734 
3735 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
3736 
3737 	rc = device_create_file(&dev->dev, &dev_attr_failover);
3738 	if (rc) {
3739 		free_netdev(netdev);
3740 		return rc;
3741 	}
3742 
3743 	rc = register_netdev(netdev);
3744 	if (rc) {
3745 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
3746 		device_remove_file(&dev->dev, &dev_attr_failover);
3747 		free_netdev(netdev);
3748 		return rc;
3749 	}
3750 	dev_info(&dev->dev, "ibmvnic registered\n");
3751 
3752 	adapter->state = VNIC_PROBED;
3753 	return 0;
3754 }
3755 
3756 static int ibmvnic_remove(struct vio_dev *dev)
3757 {
3758 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
3759 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
3760 
3761 	adapter->state = VNIC_REMOVING;
3762 	unregister_netdev(netdev);
3763 	mutex_lock(&adapter->reset_lock);
3764 
3765 	release_resources(adapter);
3766 	release_sub_crqs(adapter);
3767 	release_crq_queue(adapter);
3768 
3769 	adapter->state = VNIC_REMOVED;
3770 
3771 	mutex_unlock(&adapter->reset_lock);
3772 	device_remove_file(&dev->dev, &dev_attr_failover);
3773 	free_netdev(netdev);
3774 	dev_set_drvdata(&dev->dev, NULL);
3775 
3776 	return 0;
3777 }
3778 
3779 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
3780 			      const char *buf, size_t count)
3781 {
3782 	struct net_device *netdev = dev_get_drvdata(dev);
3783 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
3784 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
3785 	__be64 session_token;
3786 	long rc;
3787 
3788 	if (!sysfs_streq(buf, "1"))
3789 		return -EINVAL;
3790 
3791 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
3792 			 H_GET_SESSION_TOKEN, 0, 0, 0);
3793 	if (rc) {
3794 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
3795 			   rc);
3796 		return -EINVAL;
3797 	}
3798 
3799 	session_token = (__be64)retbuf[0];
3800 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
3801 		   be64_to_cpu(session_token));
3802 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3803 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
3804 	if (rc) {
3805 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
3806 			   rc);
3807 		return -EINVAL;
3808 	}
3809 
3810 	return count;
3811 }
3812 
3813 static DEVICE_ATTR(failover, 0200, NULL, failover_store);
3814 
3815 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
3816 {
3817 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
3818 	struct ibmvnic_adapter *adapter;
3819 	struct iommu_table *tbl;
3820 	unsigned long ret = 0;
3821 	int i;
3822 
3823 	tbl = get_iommu_table_base(&vdev->dev);
3824 
3825 	/* netdev inits at probe time along with the structures we need below*/
3826 	if (!netdev)
3827 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
3828 
3829 	adapter = netdev_priv(netdev);
3830 
3831 	ret += PAGE_SIZE; /* the crq message queue */
3832 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
3833 
3834 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
3835 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
3836 
3837 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
3838 	     i++)
3839 		ret += adapter->rx_pool[i].size *
3840 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
3841 
3842 	return ret;
3843 }
3844 
3845 static int ibmvnic_resume(struct device *dev)
3846 {
3847 	struct net_device *netdev = dev_get_drvdata(dev);
3848 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
3849 	int i;
3850 
3851 	if (adapter->state != VNIC_OPEN)
3852 		return 0;
3853 
3854 	/* kick the interrupt handlers just in case we lost an interrupt */
3855 	for (i = 0; i < adapter->req_rx_queues; i++)
3856 		ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
3857 				     adapter->rx_scrq[i]);
3858 
3859 	return 0;
3860 }
3861 
3862 static struct vio_device_id ibmvnic_device_table[] = {
3863 	{"network", "IBM,vnic"},
3864 	{"", "" }
3865 };
3866 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
3867 
3868 static const struct dev_pm_ops ibmvnic_pm_ops = {
3869 	.resume = ibmvnic_resume
3870 };
3871 
3872 static struct vio_driver ibmvnic_driver = {
3873 	.id_table       = ibmvnic_device_table,
3874 	.probe          = ibmvnic_probe,
3875 	.remove         = ibmvnic_remove,
3876 	.get_desired_dma = ibmvnic_get_desired_dma,
3877 	.name		= ibmvnic_driver_name,
3878 	.pm		= &ibmvnic_pm_ops,
3879 };
3880 
3881 /* module functions */
3882 static int __init ibmvnic_module_init(void)
3883 {
3884 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
3885 		IBMVNIC_DRIVER_VERSION);
3886 
3887 	return vio_register_driver(&ibmvnic_driver);
3888 }
3889 
3890 static void __exit ibmvnic_module_exit(void)
3891 {
3892 	vio_unregister_driver(&ibmvnic_driver);
3893 }
3894 
3895 module_init(ibmvnic_module_init);
3896 module_exit(ibmvnic_module_exit);
3897