xref: /openbmc/linux/drivers/net/ethernet/ibm/ibmvnic.c (revision 8cb5d748)
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_crqs(struct ibmvnic_adapter *);
115 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
116 static int ibmvnic_init(struct ibmvnic_adapter *);
117 static void release_crq_queue(struct ibmvnic_adapter *);
118 
119 struct ibmvnic_stat {
120 	char name[ETH_GSTRING_LEN];
121 	int offset;
122 };
123 
124 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
125 			     offsetof(struct ibmvnic_statistics, stat))
126 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
127 
128 static const struct ibmvnic_stat ibmvnic_stats[] = {
129 	{"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
130 	{"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
131 	{"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
132 	{"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
133 	{"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
134 	{"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
135 	{"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
136 	{"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
137 	{"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
138 	{"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
139 	{"align_errors", IBMVNIC_STAT_OFF(align_errors)},
140 	{"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
141 	{"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
142 	{"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
143 	{"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
144 	{"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
145 	{"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
146 	{"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
147 	{"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
148 	{"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
149 	{"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
150 	{"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
151 };
152 
153 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
154 			  unsigned long length, unsigned long *number,
155 			  unsigned long *irq)
156 {
157 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
158 	long rc;
159 
160 	rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
161 	*number = retbuf[0];
162 	*irq = retbuf[1];
163 
164 	return rc;
165 }
166 
167 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
168 				struct ibmvnic_long_term_buff *ltb, int size)
169 {
170 	struct device *dev = &adapter->vdev->dev;
171 
172 	ltb->size = size;
173 	ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
174 				       GFP_KERNEL);
175 
176 	if (!ltb->buff) {
177 		dev_err(dev, "Couldn't alloc long term buffer\n");
178 		return -ENOMEM;
179 	}
180 	ltb->map_id = adapter->map_id;
181 	adapter->map_id++;
182 
183 	init_completion(&adapter->fw_done);
184 	send_request_map(adapter, ltb->addr,
185 			 ltb->size, ltb->map_id);
186 	wait_for_completion(&adapter->fw_done);
187 
188 	if (adapter->fw_done_rc) {
189 		dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
190 			adapter->fw_done_rc);
191 		return -1;
192 	}
193 	return 0;
194 }
195 
196 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
197 				struct ibmvnic_long_term_buff *ltb)
198 {
199 	struct device *dev = &adapter->vdev->dev;
200 
201 	if (!ltb->buff)
202 		return;
203 
204 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
205 	    adapter->reset_reason != VNIC_RESET_MOBILITY)
206 		send_request_unmap(adapter, ltb->map_id);
207 	dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
208 }
209 
210 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
211 				struct ibmvnic_long_term_buff *ltb)
212 {
213 	memset(ltb->buff, 0, ltb->size);
214 
215 	init_completion(&adapter->fw_done);
216 	send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
217 	wait_for_completion(&adapter->fw_done);
218 
219 	if (adapter->fw_done_rc) {
220 		dev_info(&adapter->vdev->dev,
221 			 "Reset failed, attempting to free and reallocate buffer\n");
222 		free_long_term_buff(adapter, ltb);
223 		return alloc_long_term_buff(adapter, ltb, ltb->size);
224 	}
225 	return 0;
226 }
227 
228 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
229 {
230 	int i;
231 
232 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
233 	     i++)
234 		adapter->rx_pool[i].active = 0;
235 }
236 
237 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
238 			      struct ibmvnic_rx_pool *pool)
239 {
240 	int count = pool->size - atomic_read(&pool->available);
241 	struct device *dev = &adapter->vdev->dev;
242 	int buffers_added = 0;
243 	unsigned long lpar_rc;
244 	union sub_crq sub_crq;
245 	struct sk_buff *skb;
246 	unsigned int offset;
247 	dma_addr_t dma_addr;
248 	unsigned char *dst;
249 	u64 *handle_array;
250 	int shift = 0;
251 	int index;
252 	int i;
253 
254 	if (!pool->active)
255 		return;
256 
257 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
258 				      be32_to_cpu(adapter->login_rsp_buf->
259 				      off_rxadd_subcrqs));
260 
261 	for (i = 0; i < count; ++i) {
262 		skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
263 		if (!skb) {
264 			dev_err(dev, "Couldn't replenish rx buff\n");
265 			adapter->replenish_no_mem++;
266 			break;
267 		}
268 
269 		index = pool->free_map[pool->next_free];
270 
271 		if (pool->rx_buff[index].skb)
272 			dev_err(dev, "Inconsistent free_map!\n");
273 
274 		/* Copy the skb to the long term mapped DMA buffer */
275 		offset = index * pool->buff_size;
276 		dst = pool->long_term_buff.buff + offset;
277 		memset(dst, 0, pool->buff_size);
278 		dma_addr = pool->long_term_buff.addr + offset;
279 		pool->rx_buff[index].data = dst;
280 
281 		pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
282 		pool->rx_buff[index].dma = dma_addr;
283 		pool->rx_buff[index].skb = skb;
284 		pool->rx_buff[index].pool_index = pool->index;
285 		pool->rx_buff[index].size = pool->buff_size;
286 
287 		memset(&sub_crq, 0, sizeof(sub_crq));
288 		sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
289 		sub_crq.rx_add.correlator =
290 		    cpu_to_be64((u64)&pool->rx_buff[index]);
291 		sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
292 		sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
293 
294 		/* The length field of the sCRQ is defined to be 24 bits so the
295 		 * buffer size needs to be left shifted by a byte before it is
296 		 * converted to big endian to prevent the last byte from being
297 		 * truncated.
298 		 */
299 #ifdef __LITTLE_ENDIAN__
300 		shift = 8;
301 #endif
302 		sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
303 
304 		lpar_rc = send_subcrq(adapter, handle_array[pool->index],
305 				      &sub_crq);
306 		if (lpar_rc != H_SUCCESS)
307 			goto failure;
308 
309 		buffers_added++;
310 		adapter->replenish_add_buff_success++;
311 		pool->next_free = (pool->next_free + 1) % pool->size;
312 	}
313 	atomic_add(buffers_added, &pool->available);
314 	return;
315 
316 failure:
317 	dev_info(dev, "replenish pools failure\n");
318 	pool->free_map[pool->next_free] = index;
319 	pool->rx_buff[index].skb = NULL;
320 	if (!dma_mapping_error(dev, dma_addr))
321 		dma_unmap_single(dev, dma_addr, pool->buff_size,
322 				 DMA_FROM_DEVICE);
323 
324 	dev_kfree_skb_any(skb);
325 	adapter->replenish_add_buff_failure++;
326 	atomic_add(buffers_added, &pool->available);
327 
328 	if (lpar_rc == H_CLOSED) {
329 		/* Disable buffer pool replenishment and report carrier off if
330 		 * queue is closed. Firmware guarantees that a signal will
331 		 * be sent to the driver, triggering a reset.
332 		 */
333 		deactivate_rx_pools(adapter);
334 		netif_carrier_off(adapter->netdev);
335 	}
336 }
337 
338 static void replenish_pools(struct ibmvnic_adapter *adapter)
339 {
340 	int i;
341 
342 	adapter->replenish_task_cycles++;
343 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
344 	     i++) {
345 		if (adapter->rx_pool[i].active)
346 			replenish_rx_pool(adapter, &adapter->rx_pool[i]);
347 	}
348 }
349 
350 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
351 {
352 	kfree(adapter->tx_stats_buffers);
353 	kfree(adapter->rx_stats_buffers);
354 }
355 
356 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
357 {
358 	adapter->tx_stats_buffers =
359 				kcalloc(adapter->req_tx_queues,
360 					sizeof(struct ibmvnic_tx_queue_stats),
361 					GFP_KERNEL);
362 	if (!adapter->tx_stats_buffers)
363 		return -ENOMEM;
364 
365 	adapter->rx_stats_buffers =
366 				kcalloc(adapter->req_rx_queues,
367 					sizeof(struct ibmvnic_rx_queue_stats),
368 					GFP_KERNEL);
369 	if (!adapter->rx_stats_buffers)
370 		return -ENOMEM;
371 
372 	return 0;
373 }
374 
375 static void release_stats_token(struct ibmvnic_adapter *adapter)
376 {
377 	struct device *dev = &adapter->vdev->dev;
378 
379 	if (!adapter->stats_token)
380 		return;
381 
382 	dma_unmap_single(dev, adapter->stats_token,
383 			 sizeof(struct ibmvnic_statistics),
384 			 DMA_FROM_DEVICE);
385 	adapter->stats_token = 0;
386 }
387 
388 static int init_stats_token(struct ibmvnic_adapter *adapter)
389 {
390 	struct device *dev = &adapter->vdev->dev;
391 	dma_addr_t stok;
392 
393 	stok = dma_map_single(dev, &adapter->stats,
394 			      sizeof(struct ibmvnic_statistics),
395 			      DMA_FROM_DEVICE);
396 	if (dma_mapping_error(dev, stok)) {
397 		dev_err(dev, "Couldn't map stats buffer\n");
398 		return -1;
399 	}
400 
401 	adapter->stats_token = stok;
402 	netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
403 	return 0;
404 }
405 
406 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
407 {
408 	struct ibmvnic_rx_pool *rx_pool;
409 	int rx_scrqs;
410 	int i, j, rc;
411 
412 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
413 	for (i = 0; i < rx_scrqs; i++) {
414 		rx_pool = &adapter->rx_pool[i];
415 
416 		netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
417 
418 		rc = reset_long_term_buff(adapter, &rx_pool->long_term_buff);
419 		if (rc)
420 			return rc;
421 
422 		for (j = 0; j < rx_pool->size; j++)
423 			rx_pool->free_map[j] = j;
424 
425 		memset(rx_pool->rx_buff, 0,
426 		       rx_pool->size * sizeof(struct ibmvnic_rx_buff));
427 
428 		atomic_set(&rx_pool->available, 0);
429 		rx_pool->next_alloc = 0;
430 		rx_pool->next_free = 0;
431 		rx_pool->active = 1;
432 	}
433 
434 	return 0;
435 }
436 
437 static void release_rx_pools(struct ibmvnic_adapter *adapter)
438 {
439 	struct ibmvnic_rx_pool *rx_pool;
440 	int rx_scrqs;
441 	int i, j;
442 
443 	if (!adapter->rx_pool)
444 		return;
445 
446 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
447 	for (i = 0; i < rx_scrqs; i++) {
448 		rx_pool = &adapter->rx_pool[i];
449 
450 		netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
451 
452 		kfree(rx_pool->free_map);
453 		free_long_term_buff(adapter, &rx_pool->long_term_buff);
454 
455 		if (!rx_pool->rx_buff)
456 			continue;
457 
458 		for (j = 0; j < rx_pool->size; j++) {
459 			if (rx_pool->rx_buff[j].skb) {
460 				dev_kfree_skb_any(rx_pool->rx_buff[i].skb);
461 				rx_pool->rx_buff[i].skb = NULL;
462 			}
463 		}
464 
465 		kfree(rx_pool->rx_buff);
466 	}
467 
468 	kfree(adapter->rx_pool);
469 	adapter->rx_pool = NULL;
470 }
471 
472 static int init_rx_pools(struct net_device *netdev)
473 {
474 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
475 	struct device *dev = &adapter->vdev->dev;
476 	struct ibmvnic_rx_pool *rx_pool;
477 	int rxadd_subcrqs;
478 	u64 *size_array;
479 	int i, j;
480 
481 	rxadd_subcrqs =
482 		be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
483 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
484 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
485 
486 	adapter->rx_pool = kcalloc(rxadd_subcrqs,
487 				   sizeof(struct ibmvnic_rx_pool),
488 				   GFP_KERNEL);
489 	if (!adapter->rx_pool) {
490 		dev_err(dev, "Failed to allocate rx pools\n");
491 		return -1;
492 	}
493 
494 	for (i = 0; i < rxadd_subcrqs; i++) {
495 		rx_pool = &adapter->rx_pool[i];
496 
497 		netdev_dbg(adapter->netdev,
498 			   "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
499 			   i, adapter->req_rx_add_entries_per_subcrq,
500 			   be64_to_cpu(size_array[i]));
501 
502 		rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
503 		rx_pool->index = i;
504 		rx_pool->buff_size = be64_to_cpu(size_array[i]);
505 		rx_pool->active = 1;
506 
507 		rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
508 					    GFP_KERNEL);
509 		if (!rx_pool->free_map) {
510 			release_rx_pools(adapter);
511 			return -1;
512 		}
513 
514 		rx_pool->rx_buff = kcalloc(rx_pool->size,
515 					   sizeof(struct ibmvnic_rx_buff),
516 					   GFP_KERNEL);
517 		if (!rx_pool->rx_buff) {
518 			dev_err(dev, "Couldn't alloc rx buffers\n");
519 			release_rx_pools(adapter);
520 			return -1;
521 		}
522 
523 		if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
524 					 rx_pool->size * rx_pool->buff_size)) {
525 			release_rx_pools(adapter);
526 			return -1;
527 		}
528 
529 		for (j = 0; j < rx_pool->size; ++j)
530 			rx_pool->free_map[j] = j;
531 
532 		atomic_set(&rx_pool->available, 0);
533 		rx_pool->next_alloc = 0;
534 		rx_pool->next_free = 0;
535 	}
536 
537 	return 0;
538 }
539 
540 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
541 {
542 	struct ibmvnic_tx_pool *tx_pool;
543 	int tx_scrqs;
544 	int i, j, rc;
545 
546 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
547 	for (i = 0; i < tx_scrqs; i++) {
548 		netdev_dbg(adapter->netdev, "Re-setting tx_pool[%d]\n", i);
549 
550 		tx_pool = &adapter->tx_pool[i];
551 
552 		rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
553 		if (rc)
554 			return rc;
555 
556 		memset(tx_pool->tx_buff, 0,
557 		       adapter->req_tx_entries_per_subcrq *
558 		       sizeof(struct ibmvnic_tx_buff));
559 
560 		for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
561 			tx_pool->free_map[j] = j;
562 
563 		tx_pool->consumer_index = 0;
564 		tx_pool->producer_index = 0;
565 	}
566 
567 	return 0;
568 }
569 
570 static void release_tx_pools(struct ibmvnic_adapter *adapter)
571 {
572 	struct ibmvnic_tx_pool *tx_pool;
573 	int i, tx_scrqs;
574 
575 	if (!adapter->tx_pool)
576 		return;
577 
578 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
579 	for (i = 0; i < tx_scrqs; i++) {
580 		netdev_dbg(adapter->netdev, "Releasing tx_pool[%d]\n", i);
581 		tx_pool = &adapter->tx_pool[i];
582 		kfree(tx_pool->tx_buff);
583 		free_long_term_buff(adapter, &tx_pool->long_term_buff);
584 		kfree(tx_pool->free_map);
585 	}
586 
587 	kfree(adapter->tx_pool);
588 	adapter->tx_pool = NULL;
589 }
590 
591 static int init_tx_pools(struct net_device *netdev)
592 {
593 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
594 	struct device *dev = &adapter->vdev->dev;
595 	struct ibmvnic_tx_pool *tx_pool;
596 	int tx_subcrqs;
597 	int i, j;
598 
599 	tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
600 	adapter->tx_pool = kcalloc(tx_subcrqs,
601 				   sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
602 	if (!adapter->tx_pool)
603 		return -1;
604 
605 	for (i = 0; i < tx_subcrqs; i++) {
606 		tx_pool = &adapter->tx_pool[i];
607 
608 		netdev_dbg(adapter->netdev,
609 			   "Initializing tx_pool[%d], %lld buffs\n",
610 			   i, adapter->req_tx_entries_per_subcrq);
611 
612 		tx_pool->tx_buff = kcalloc(adapter->req_tx_entries_per_subcrq,
613 					   sizeof(struct ibmvnic_tx_buff),
614 					   GFP_KERNEL);
615 		if (!tx_pool->tx_buff) {
616 			dev_err(dev, "tx pool buffer allocation failed\n");
617 			release_tx_pools(adapter);
618 			return -1;
619 		}
620 
621 		if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
622 					 adapter->req_tx_entries_per_subcrq *
623 					 adapter->req_mtu)) {
624 			release_tx_pools(adapter);
625 			return -1;
626 		}
627 
628 		tx_pool->free_map = kcalloc(adapter->req_tx_entries_per_subcrq,
629 					    sizeof(int), GFP_KERNEL);
630 		if (!tx_pool->free_map) {
631 			release_tx_pools(adapter);
632 			return -1;
633 		}
634 
635 		for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
636 			tx_pool->free_map[j] = j;
637 
638 		tx_pool->consumer_index = 0;
639 		tx_pool->producer_index = 0;
640 	}
641 
642 	return 0;
643 }
644 
645 static void release_error_buffers(struct ibmvnic_adapter *adapter)
646 {
647 	struct device *dev = &adapter->vdev->dev;
648 	struct ibmvnic_error_buff *error_buff, *tmp;
649 	unsigned long flags;
650 
651 	spin_lock_irqsave(&adapter->error_list_lock, flags);
652 	list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list) {
653 		list_del(&error_buff->list);
654 		dma_unmap_single(dev, error_buff->dma, error_buff->len,
655 				 DMA_FROM_DEVICE);
656 		kfree(error_buff->buff);
657 		kfree(error_buff);
658 	}
659 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
660 }
661 
662 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
663 {
664 	int i;
665 
666 	if (adapter->napi_enabled)
667 		return;
668 
669 	for (i = 0; i < adapter->req_rx_queues; i++)
670 		napi_enable(&adapter->napi[i]);
671 
672 	adapter->napi_enabled = true;
673 }
674 
675 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
676 {
677 	int i;
678 
679 	if (!adapter->napi_enabled)
680 		return;
681 
682 	for (i = 0; i < adapter->req_rx_queues; i++) {
683 		netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
684 		napi_disable(&adapter->napi[i]);
685 	}
686 
687 	adapter->napi_enabled = false;
688 }
689 
690 static int ibmvnic_login(struct net_device *netdev)
691 {
692 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
693 	unsigned long timeout = msecs_to_jiffies(30000);
694 	struct device *dev = &adapter->vdev->dev;
695 	int rc;
696 
697 	do {
698 		if (adapter->renegotiate) {
699 			adapter->renegotiate = false;
700 			release_sub_crqs(adapter);
701 
702 			reinit_completion(&adapter->init_done);
703 			send_cap_queries(adapter);
704 			if (!wait_for_completion_timeout(&adapter->init_done,
705 							 timeout)) {
706 				dev_err(dev, "Capabilities query timeout\n");
707 				return -1;
708 			}
709 			rc = init_sub_crqs(adapter);
710 			if (rc) {
711 				dev_err(dev,
712 					"Initialization of SCRQ's failed\n");
713 				return -1;
714 			}
715 			rc = init_sub_crq_irqs(adapter);
716 			if (rc) {
717 				dev_err(dev,
718 					"Initialization of SCRQ's irqs failed\n");
719 				return -1;
720 			}
721 		}
722 
723 		reinit_completion(&adapter->init_done);
724 		send_login(adapter);
725 		if (!wait_for_completion_timeout(&adapter->init_done,
726 						 timeout)) {
727 			dev_err(dev, "Login timeout\n");
728 			return -1;
729 		}
730 	} while (adapter->renegotiate);
731 
732 	return 0;
733 }
734 
735 static void release_resources(struct ibmvnic_adapter *adapter)
736 {
737 	int i;
738 
739 	release_tx_pools(adapter);
740 	release_rx_pools(adapter);
741 
742 	release_stats_token(adapter);
743 	release_stats_buffers(adapter);
744 	release_error_buffers(adapter);
745 
746 	if (adapter->napi) {
747 		for (i = 0; i < adapter->req_rx_queues; i++) {
748 			if (&adapter->napi[i]) {
749 				netdev_dbg(adapter->netdev,
750 					   "Releasing napi[%d]\n", i);
751 				netif_napi_del(&adapter->napi[i]);
752 			}
753 		}
754 	}
755 }
756 
757 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
758 {
759 	struct net_device *netdev = adapter->netdev;
760 	unsigned long timeout = msecs_to_jiffies(30000);
761 	union ibmvnic_crq crq;
762 	bool resend;
763 	int rc;
764 
765 	netdev_dbg(netdev, "setting link state %d\n", link_state);
766 
767 	memset(&crq, 0, sizeof(crq));
768 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
769 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
770 	crq.logical_link_state.link_state = link_state;
771 
772 	do {
773 		resend = false;
774 
775 		reinit_completion(&adapter->init_done);
776 		rc = ibmvnic_send_crq(adapter, &crq);
777 		if (rc) {
778 			netdev_err(netdev, "Failed to set link state\n");
779 			return rc;
780 		}
781 
782 		if (!wait_for_completion_timeout(&adapter->init_done,
783 						 timeout)) {
784 			netdev_err(netdev, "timeout setting link state\n");
785 			return -1;
786 		}
787 
788 		if (adapter->init_done_rc == 1) {
789 			/* Partuial success, delay and re-send */
790 			mdelay(1000);
791 			resend = true;
792 		}
793 	} while (resend);
794 
795 	return 0;
796 }
797 
798 static int set_real_num_queues(struct net_device *netdev)
799 {
800 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
801 	int rc;
802 
803 	netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
804 		   adapter->req_tx_queues, adapter->req_rx_queues);
805 
806 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
807 	if (rc) {
808 		netdev_err(netdev, "failed to set the number of tx queues\n");
809 		return rc;
810 	}
811 
812 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
813 	if (rc)
814 		netdev_err(netdev, "failed to set the number of rx queues\n");
815 
816 	return rc;
817 }
818 
819 static int init_resources(struct ibmvnic_adapter *adapter)
820 {
821 	struct net_device *netdev = adapter->netdev;
822 	int i, rc;
823 
824 	rc = set_real_num_queues(netdev);
825 	if (rc)
826 		return rc;
827 
828 	rc = init_stats_buffers(adapter);
829 	if (rc)
830 		return rc;
831 
832 	rc = init_stats_token(adapter);
833 	if (rc)
834 		return rc;
835 
836 	adapter->map_id = 1;
837 	adapter->napi = kcalloc(adapter->req_rx_queues,
838 				sizeof(struct napi_struct), GFP_KERNEL);
839 	if (!adapter->napi)
840 		return -ENOMEM;
841 
842 	for (i = 0; i < adapter->req_rx_queues; i++) {
843 		netdev_dbg(netdev, "Adding napi[%d]\n", i);
844 		netif_napi_add(netdev, &adapter->napi[i], ibmvnic_poll,
845 			       NAPI_POLL_WEIGHT);
846 	}
847 
848 	send_map_query(adapter);
849 
850 	rc = init_rx_pools(netdev);
851 	if (rc)
852 		return rc;
853 
854 	rc = init_tx_pools(netdev);
855 	return rc;
856 }
857 
858 static int __ibmvnic_open(struct net_device *netdev)
859 {
860 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
861 	enum vnic_state prev_state = adapter->state;
862 	int i, rc;
863 
864 	adapter->state = VNIC_OPENING;
865 	replenish_pools(adapter);
866 	ibmvnic_napi_enable(adapter);
867 
868 	/* We're ready to receive frames, enable the sub-crq interrupts and
869 	 * set the logical link state to up
870 	 */
871 	for (i = 0; i < adapter->req_rx_queues; i++) {
872 		netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
873 		if (prev_state == VNIC_CLOSED)
874 			enable_irq(adapter->rx_scrq[i]->irq);
875 		else
876 			enable_scrq_irq(adapter, adapter->rx_scrq[i]);
877 	}
878 
879 	for (i = 0; i < adapter->req_tx_queues; i++) {
880 		netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
881 		if (prev_state == VNIC_CLOSED)
882 			enable_irq(adapter->tx_scrq[i]->irq);
883 		else
884 			enable_scrq_irq(adapter, adapter->tx_scrq[i]);
885 	}
886 
887 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
888 	if (rc) {
889 		for (i = 0; i < adapter->req_rx_queues; i++)
890 			napi_disable(&adapter->napi[i]);
891 		release_resources(adapter);
892 		return rc;
893 	}
894 
895 	netif_tx_start_all_queues(netdev);
896 
897 	if (prev_state == VNIC_CLOSED) {
898 		for (i = 0; i < adapter->req_rx_queues; i++)
899 			napi_schedule(&adapter->napi[i]);
900 	}
901 
902 	adapter->state = VNIC_OPEN;
903 	return rc;
904 }
905 
906 static int ibmvnic_open(struct net_device *netdev)
907 {
908 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
909 	int rc;
910 
911 	mutex_lock(&adapter->reset_lock);
912 
913 	if (adapter->state != VNIC_CLOSED) {
914 		rc = ibmvnic_login(netdev);
915 		if (rc) {
916 			mutex_unlock(&adapter->reset_lock);
917 			return rc;
918 		}
919 
920 		rc = init_resources(adapter);
921 		if (rc) {
922 			netdev_err(netdev, "failed to initialize resources\n");
923 			release_resources(adapter);
924 			mutex_unlock(&adapter->reset_lock);
925 			return rc;
926 		}
927 	}
928 
929 	rc = __ibmvnic_open(netdev);
930 	mutex_unlock(&adapter->reset_lock);
931 
932 	return rc;
933 }
934 
935 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
936 {
937 	struct ibmvnic_tx_pool *tx_pool;
938 	u64 tx_entries;
939 	int tx_scrqs;
940 	int i, j;
941 
942 	if (!adapter->tx_pool)
943 		return;
944 
945 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
946 	tx_entries = adapter->req_tx_entries_per_subcrq;
947 
948 	/* Free any remaining skbs in the tx buffer pools */
949 	for (i = 0; i < tx_scrqs; i++) {
950 		tx_pool = &adapter->tx_pool[i];
951 		if (!tx_pool)
952 			continue;
953 
954 		netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
955 		for (j = 0; j < tx_entries; j++) {
956 			if (tx_pool->tx_buff[j].skb) {
957 				dev_kfree_skb_any(tx_pool->tx_buff[j].skb);
958 				tx_pool->tx_buff[j].skb = NULL;
959 			}
960 		}
961 	}
962 }
963 
964 static int __ibmvnic_close(struct net_device *netdev)
965 {
966 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
967 	int rc = 0;
968 	int i;
969 
970 	adapter->state = VNIC_CLOSING;
971 
972 	/* ensure that transmissions are stopped if called by do_reset */
973 	if (adapter->resetting)
974 		netif_tx_disable(netdev);
975 	else
976 		netif_tx_stop_all_queues(netdev);
977 
978 	ibmvnic_napi_disable(adapter);
979 
980 	if (adapter->tx_scrq) {
981 		for (i = 0; i < adapter->req_tx_queues; i++)
982 			if (adapter->tx_scrq[i]->irq) {
983 				netdev_dbg(adapter->netdev,
984 					   "Disabling tx_scrq[%d] irq\n", i);
985 				disable_irq(adapter->tx_scrq[i]->irq);
986 			}
987 	}
988 
989 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
990 	if (rc)
991 		return rc;
992 
993 	if (adapter->rx_scrq) {
994 		for (i = 0; i < adapter->req_rx_queues; i++) {
995 			int retries = 10;
996 
997 			while (pending_scrq(adapter, adapter->rx_scrq[i])) {
998 				retries--;
999 				mdelay(100);
1000 
1001 				if (retries == 0)
1002 					break;
1003 			}
1004 
1005 			if (adapter->rx_scrq[i]->irq) {
1006 				netdev_dbg(adapter->netdev,
1007 					   "Disabling rx_scrq[%d] irq\n", i);
1008 				disable_irq(adapter->rx_scrq[i]->irq);
1009 			}
1010 		}
1011 	}
1012 
1013 	clean_tx_pools(adapter);
1014 	adapter->state = VNIC_CLOSED;
1015 	return rc;
1016 }
1017 
1018 static int ibmvnic_close(struct net_device *netdev)
1019 {
1020 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1021 	int rc;
1022 
1023 	mutex_lock(&adapter->reset_lock);
1024 	rc = __ibmvnic_close(netdev);
1025 	mutex_unlock(&adapter->reset_lock);
1026 
1027 	return rc;
1028 }
1029 
1030 /**
1031  * build_hdr_data - creates L2/L3/L4 header data buffer
1032  * @hdr_field - bitfield determining needed headers
1033  * @skb - socket buffer
1034  * @hdr_len - array of header lengths
1035  * @tot_len - total length of data
1036  *
1037  * Reads hdr_field to determine which headers are needed by firmware.
1038  * Builds a buffer containing these headers.  Saves individual header
1039  * lengths and total buffer length to be used to build descriptors.
1040  */
1041 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1042 			  int *hdr_len, u8 *hdr_data)
1043 {
1044 	int len = 0;
1045 	u8 *hdr;
1046 
1047 	hdr_len[0] = sizeof(struct ethhdr);
1048 
1049 	if (skb->protocol == htons(ETH_P_IP)) {
1050 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
1051 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1052 			hdr_len[2] = tcp_hdrlen(skb);
1053 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1054 			hdr_len[2] = sizeof(struct udphdr);
1055 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1056 		hdr_len[1] = sizeof(struct ipv6hdr);
1057 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1058 			hdr_len[2] = tcp_hdrlen(skb);
1059 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1060 			hdr_len[2] = sizeof(struct udphdr);
1061 	}
1062 
1063 	memset(hdr_data, 0, 120);
1064 	if ((hdr_field >> 6) & 1) {
1065 		hdr = skb_mac_header(skb);
1066 		memcpy(hdr_data, hdr, hdr_len[0]);
1067 		len += hdr_len[0];
1068 	}
1069 
1070 	if ((hdr_field >> 5) & 1) {
1071 		hdr = skb_network_header(skb);
1072 		memcpy(hdr_data + len, hdr, hdr_len[1]);
1073 		len += hdr_len[1];
1074 	}
1075 
1076 	if ((hdr_field >> 4) & 1) {
1077 		hdr = skb_transport_header(skb);
1078 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1079 		len += hdr_len[2];
1080 	}
1081 	return len;
1082 }
1083 
1084 /**
1085  * create_hdr_descs - create header and header extension descriptors
1086  * @hdr_field - bitfield determining needed headers
1087  * @data - buffer containing header data
1088  * @len - length of data buffer
1089  * @hdr_len - array of individual header lengths
1090  * @scrq_arr - descriptor array
1091  *
1092  * Creates header and, if needed, header extension descriptors and
1093  * places them in a descriptor array, scrq_arr
1094  */
1095 
1096 static void create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1097 			     union sub_crq *scrq_arr)
1098 {
1099 	union sub_crq hdr_desc;
1100 	int tmp_len = len;
1101 	u8 *data, *cur;
1102 	int tmp;
1103 
1104 	while (tmp_len > 0) {
1105 		cur = hdr_data + len - tmp_len;
1106 
1107 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1108 		if (cur != hdr_data) {
1109 			data = hdr_desc.hdr_ext.data;
1110 			tmp = tmp_len > 29 ? 29 : tmp_len;
1111 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1112 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1113 			hdr_desc.hdr_ext.len = tmp;
1114 		} else {
1115 			data = hdr_desc.hdr.data;
1116 			tmp = tmp_len > 24 ? 24 : tmp_len;
1117 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1118 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1119 			hdr_desc.hdr.len = tmp;
1120 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1121 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1122 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1123 			hdr_desc.hdr.flag = hdr_field << 1;
1124 		}
1125 		memcpy(data, cur, tmp);
1126 		tmp_len -= tmp;
1127 		*scrq_arr = hdr_desc;
1128 		scrq_arr++;
1129 	}
1130 }
1131 
1132 /**
1133  * build_hdr_descs_arr - build a header descriptor array
1134  * @skb - socket buffer
1135  * @num_entries - number of descriptors to be sent
1136  * @subcrq - first TX descriptor
1137  * @hdr_field - bit field determining which headers will be sent
1138  *
1139  * This function will build a TX descriptor array with applicable
1140  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1141  */
1142 
1143 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1144 				int *num_entries, u8 hdr_field)
1145 {
1146 	int hdr_len[3] = {0, 0, 0};
1147 	int tot_len, len;
1148 	u8 *hdr_data = txbuff->hdr_data;
1149 
1150 	tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1151 				 txbuff->hdr_data);
1152 	len = tot_len;
1153 	len -= 24;
1154 	if (len > 0)
1155 		num_entries += len % 29 ? len / 29 + 1 : len / 29;
1156 	create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1157 			 txbuff->indir_arr + 1);
1158 }
1159 
1160 static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1161 {
1162 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1163 	int queue_num = skb_get_queue_mapping(skb);
1164 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1165 	struct device *dev = &adapter->vdev->dev;
1166 	struct ibmvnic_tx_buff *tx_buff = NULL;
1167 	struct ibmvnic_sub_crq_queue *tx_scrq;
1168 	struct ibmvnic_tx_pool *tx_pool;
1169 	unsigned int tx_send_failed = 0;
1170 	unsigned int tx_map_failed = 0;
1171 	unsigned int tx_dropped = 0;
1172 	unsigned int tx_packets = 0;
1173 	unsigned int tx_bytes = 0;
1174 	dma_addr_t data_dma_addr;
1175 	struct netdev_queue *txq;
1176 	unsigned long lpar_rc;
1177 	union sub_crq tx_crq;
1178 	unsigned int offset;
1179 	int num_entries = 1;
1180 	unsigned char *dst;
1181 	u64 *handle_array;
1182 	int index = 0;
1183 	int ret = 0;
1184 
1185 	if (adapter->resetting) {
1186 		if (!netif_subqueue_stopped(netdev, skb))
1187 			netif_stop_subqueue(netdev, queue_num);
1188 		dev_kfree_skb_any(skb);
1189 
1190 		tx_send_failed++;
1191 		tx_dropped++;
1192 		ret = NETDEV_TX_OK;
1193 		goto out;
1194 	}
1195 
1196 	tx_pool = &adapter->tx_pool[queue_num];
1197 	tx_scrq = adapter->tx_scrq[queue_num];
1198 	txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1199 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1200 		be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1201 
1202 	index = tx_pool->free_map[tx_pool->consumer_index];
1203 	offset = index * adapter->req_mtu;
1204 	dst = tx_pool->long_term_buff.buff + offset;
1205 	memset(dst, 0, adapter->req_mtu);
1206 	skb_copy_from_linear_data(skb, dst, skb->len);
1207 	data_dma_addr = tx_pool->long_term_buff.addr + offset;
1208 
1209 	tx_pool->consumer_index =
1210 	    (tx_pool->consumer_index + 1) %
1211 		adapter->req_tx_entries_per_subcrq;
1212 
1213 	tx_buff = &tx_pool->tx_buff[index];
1214 	tx_buff->skb = skb;
1215 	tx_buff->data_dma[0] = data_dma_addr;
1216 	tx_buff->data_len[0] = skb->len;
1217 	tx_buff->index = index;
1218 	tx_buff->pool_index = queue_num;
1219 	tx_buff->last_frag = true;
1220 
1221 	memset(&tx_crq, 0, sizeof(tx_crq));
1222 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1223 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1224 	tx_crq.v1.n_crq_elem = 1;
1225 	tx_crq.v1.n_sge = 1;
1226 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1227 	tx_crq.v1.correlator = cpu_to_be32(index);
1228 	tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1229 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1230 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1231 
1232 	if (adapter->vlan_header_insertion) {
1233 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1234 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1235 	}
1236 
1237 	if (skb->protocol == htons(ETH_P_IP)) {
1238 		if (ip_hdr(skb)->version == 4)
1239 			tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1240 		else if (ip_hdr(skb)->version == 6)
1241 			tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1242 
1243 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1244 			tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1245 		else if (ip_hdr(skb)->protocol != IPPROTO_TCP)
1246 			tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1247 	}
1248 
1249 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1250 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1251 		hdrs += 2;
1252 	}
1253 	/* determine if l2/3/4 headers are sent to firmware */
1254 	if ((*hdrs >> 7) & 1 &&
1255 	    (skb->protocol == htons(ETH_P_IP) ||
1256 	     skb->protocol == htons(ETH_P_IPV6))) {
1257 		build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1258 		tx_crq.v1.n_crq_elem = num_entries;
1259 		tx_buff->indir_arr[0] = tx_crq;
1260 		tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1261 						    sizeof(tx_buff->indir_arr),
1262 						    DMA_TO_DEVICE);
1263 		if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1264 			dev_kfree_skb_any(skb);
1265 			tx_buff->skb = NULL;
1266 			if (!firmware_has_feature(FW_FEATURE_CMO))
1267 				dev_err(dev, "tx: unable to map descriptor array\n");
1268 			tx_map_failed++;
1269 			tx_dropped++;
1270 			ret = NETDEV_TX_OK;
1271 			goto out;
1272 		}
1273 		lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1274 					       (u64)tx_buff->indir_dma,
1275 					       (u64)num_entries);
1276 	} else {
1277 		lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1278 				      &tx_crq);
1279 	}
1280 	if (lpar_rc != H_SUCCESS) {
1281 		dev_err(dev, "tx failed with code %ld\n", lpar_rc);
1282 
1283 		if (tx_pool->consumer_index == 0)
1284 			tx_pool->consumer_index =
1285 				adapter->req_tx_entries_per_subcrq - 1;
1286 		else
1287 			tx_pool->consumer_index--;
1288 
1289 		dev_kfree_skb_any(skb);
1290 		tx_buff->skb = NULL;
1291 
1292 		if (lpar_rc == H_CLOSED) {
1293 			/* Disable TX and report carrier off if queue is closed.
1294 			 * Firmware guarantees that a signal will be sent to the
1295 			 * driver, triggering a reset or some other action.
1296 			 */
1297 			netif_tx_stop_all_queues(netdev);
1298 			netif_carrier_off(netdev);
1299 		}
1300 
1301 		tx_send_failed++;
1302 		tx_dropped++;
1303 		ret = NETDEV_TX_OK;
1304 		goto out;
1305 	}
1306 
1307 	if (atomic_inc_return(&tx_scrq->used)
1308 					>= adapter->req_tx_entries_per_subcrq) {
1309 		netdev_info(netdev, "Stopping queue %d\n", queue_num);
1310 		netif_stop_subqueue(netdev, queue_num);
1311 	}
1312 
1313 	tx_packets++;
1314 	tx_bytes += skb->len;
1315 	txq->trans_start = jiffies;
1316 	ret = NETDEV_TX_OK;
1317 
1318 out:
1319 	netdev->stats.tx_dropped += tx_dropped;
1320 	netdev->stats.tx_bytes += tx_bytes;
1321 	netdev->stats.tx_packets += tx_packets;
1322 	adapter->tx_send_failed += tx_send_failed;
1323 	adapter->tx_map_failed += tx_map_failed;
1324 	adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1325 	adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1326 	adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1327 
1328 	return ret;
1329 }
1330 
1331 static void ibmvnic_set_multi(struct net_device *netdev)
1332 {
1333 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1334 	struct netdev_hw_addr *ha;
1335 	union ibmvnic_crq crq;
1336 
1337 	memset(&crq, 0, sizeof(crq));
1338 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1339 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1340 
1341 	if (netdev->flags & IFF_PROMISC) {
1342 		if (!adapter->promisc_supported)
1343 			return;
1344 	} else {
1345 		if (netdev->flags & IFF_ALLMULTI) {
1346 			/* Accept all multicast */
1347 			memset(&crq, 0, sizeof(crq));
1348 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1349 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1350 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1351 			ibmvnic_send_crq(adapter, &crq);
1352 		} else if (netdev_mc_empty(netdev)) {
1353 			/* Reject all multicast */
1354 			memset(&crq, 0, sizeof(crq));
1355 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1356 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1357 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1358 			ibmvnic_send_crq(adapter, &crq);
1359 		} else {
1360 			/* Accept one or more multicast(s) */
1361 			netdev_for_each_mc_addr(ha, netdev) {
1362 				memset(&crq, 0, sizeof(crq));
1363 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1364 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1365 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1366 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1367 						ha->addr);
1368 				ibmvnic_send_crq(adapter, &crq);
1369 			}
1370 		}
1371 	}
1372 }
1373 
1374 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1375 {
1376 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1377 	struct sockaddr *addr = p;
1378 	union ibmvnic_crq crq;
1379 
1380 	if (!is_valid_ether_addr(addr->sa_data))
1381 		return -EADDRNOTAVAIL;
1382 
1383 	memset(&crq, 0, sizeof(crq));
1384 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1385 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1386 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1387 	ibmvnic_send_crq(adapter, &crq);
1388 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1389 	return 0;
1390 }
1391 
1392 /**
1393  * do_reset returns zero if we are able to keep processing reset events, or
1394  * non-zero if we hit a fatal error and must halt.
1395  */
1396 static int do_reset(struct ibmvnic_adapter *adapter,
1397 		    struct ibmvnic_rwi *rwi, u32 reset_state)
1398 {
1399 	struct net_device *netdev = adapter->netdev;
1400 	int i, rc;
1401 
1402 	netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1403 		   rwi->reset_reason);
1404 
1405 	netif_carrier_off(netdev);
1406 	adapter->reset_reason = rwi->reset_reason;
1407 
1408 	if (rwi->reset_reason == VNIC_RESET_MOBILITY) {
1409 		rc = ibmvnic_reenable_crq_queue(adapter);
1410 		if (rc)
1411 			return 0;
1412 	}
1413 
1414 	rc = __ibmvnic_close(netdev);
1415 	if (rc)
1416 		return rc;
1417 
1418 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1419 		/* remove the closed state so when we call open it appears
1420 		 * we are coming from the probed state.
1421 		 */
1422 		adapter->state = VNIC_PROBED;
1423 
1424 		rc = ibmvnic_init(adapter);
1425 		if (rc)
1426 			return 0;
1427 
1428 		/* If the adapter was in PROBE state prior to the reset,
1429 		 * exit here.
1430 		 */
1431 		if (reset_state == VNIC_PROBED)
1432 			return 0;
1433 
1434 		rc = ibmvnic_login(netdev);
1435 		if (rc) {
1436 			adapter->state = VNIC_PROBED;
1437 			return 0;
1438 		}
1439 
1440 		rc = reset_tx_pools(adapter);
1441 		if (rc)
1442 			return rc;
1443 
1444 		rc = reset_rx_pools(adapter);
1445 		if (rc)
1446 			return rc;
1447 
1448 		if (reset_state == VNIC_CLOSED)
1449 			return 0;
1450 	}
1451 
1452 	rc = __ibmvnic_open(netdev);
1453 	if (rc) {
1454 		if (list_empty(&adapter->rwi_list))
1455 			adapter->state = VNIC_CLOSED;
1456 		else
1457 			adapter->state = reset_state;
1458 
1459 		return 0;
1460 	}
1461 
1462 	netif_carrier_on(netdev);
1463 
1464 	/* kick napi */
1465 	for (i = 0; i < adapter->req_rx_queues; i++)
1466 		napi_schedule(&adapter->napi[i]);
1467 
1468 	if (adapter->reset_reason != VNIC_RESET_FAILOVER)
1469 		netdev_notify_peers(netdev);
1470 
1471 	return 0;
1472 }
1473 
1474 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1475 {
1476 	struct ibmvnic_rwi *rwi;
1477 
1478 	mutex_lock(&adapter->rwi_lock);
1479 
1480 	if (!list_empty(&adapter->rwi_list)) {
1481 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1482 				       list);
1483 		list_del(&rwi->list);
1484 	} else {
1485 		rwi = NULL;
1486 	}
1487 
1488 	mutex_unlock(&adapter->rwi_lock);
1489 	return rwi;
1490 }
1491 
1492 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1493 {
1494 	struct ibmvnic_rwi *rwi;
1495 
1496 	rwi = get_next_rwi(adapter);
1497 	while (rwi) {
1498 		kfree(rwi);
1499 		rwi = get_next_rwi(adapter);
1500 	}
1501 }
1502 
1503 static void __ibmvnic_reset(struct work_struct *work)
1504 {
1505 	struct ibmvnic_rwi *rwi;
1506 	struct ibmvnic_adapter *adapter;
1507 	struct net_device *netdev;
1508 	u32 reset_state;
1509 	int rc;
1510 
1511 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1512 	netdev = adapter->netdev;
1513 
1514 	mutex_lock(&adapter->reset_lock);
1515 	adapter->resetting = true;
1516 	reset_state = adapter->state;
1517 
1518 	rwi = get_next_rwi(adapter);
1519 	while (rwi) {
1520 		rc = do_reset(adapter, rwi, reset_state);
1521 		kfree(rwi);
1522 		if (rc)
1523 			break;
1524 
1525 		rwi = get_next_rwi(adapter);
1526 	}
1527 
1528 	if (rc) {
1529 		netdev_dbg(adapter->netdev, "Reset failed\n");
1530 		free_all_rwi(adapter);
1531 		mutex_unlock(&adapter->reset_lock);
1532 		return;
1533 	}
1534 
1535 	adapter->resetting = false;
1536 	mutex_unlock(&adapter->reset_lock);
1537 }
1538 
1539 static void ibmvnic_reset(struct ibmvnic_adapter *adapter,
1540 			  enum ibmvnic_reset_reason reason)
1541 {
1542 	struct ibmvnic_rwi *rwi, *tmp;
1543 	struct net_device *netdev = adapter->netdev;
1544 	struct list_head *entry;
1545 
1546 	if (adapter->state == VNIC_REMOVING ||
1547 	    adapter->state == VNIC_REMOVED) {
1548 		netdev_dbg(netdev, "Adapter removing, skipping reset\n");
1549 		return;
1550 	}
1551 
1552 	if (adapter->state == VNIC_PROBING) {
1553 		netdev_warn(netdev, "Adapter reset during probe\n");
1554 		adapter->init_done_rc = EAGAIN;
1555 		return;
1556 	}
1557 
1558 	mutex_lock(&adapter->rwi_lock);
1559 
1560 	list_for_each(entry, &adapter->rwi_list) {
1561 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
1562 		if (tmp->reset_reason == reason) {
1563 			netdev_dbg(netdev, "Skipping matching reset\n");
1564 			mutex_unlock(&adapter->rwi_lock);
1565 			return;
1566 		}
1567 	}
1568 
1569 	rwi = kzalloc(sizeof(*rwi), GFP_KERNEL);
1570 	if (!rwi) {
1571 		mutex_unlock(&adapter->rwi_lock);
1572 		ibmvnic_close(netdev);
1573 		return;
1574 	}
1575 
1576 	rwi->reset_reason = reason;
1577 	list_add_tail(&rwi->list, &adapter->rwi_list);
1578 	mutex_unlock(&adapter->rwi_lock);
1579 
1580 	netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
1581 	schedule_work(&adapter->ibmvnic_reset);
1582 }
1583 
1584 static void ibmvnic_tx_timeout(struct net_device *dev)
1585 {
1586 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1587 
1588 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
1589 }
1590 
1591 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
1592 				  struct ibmvnic_rx_buff *rx_buff)
1593 {
1594 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
1595 
1596 	rx_buff->skb = NULL;
1597 
1598 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
1599 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
1600 
1601 	atomic_dec(&pool->available);
1602 }
1603 
1604 static int ibmvnic_poll(struct napi_struct *napi, int budget)
1605 {
1606 	struct net_device *netdev = napi->dev;
1607 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1608 	int scrq_num = (int)(napi - adapter->napi);
1609 	int frames_processed = 0;
1610 
1611 restart_poll:
1612 	while (frames_processed < budget) {
1613 		struct sk_buff *skb;
1614 		struct ibmvnic_rx_buff *rx_buff;
1615 		union sub_crq *next;
1616 		u32 length;
1617 		u16 offset;
1618 		u8 flags = 0;
1619 
1620 		if (unlikely(adapter->resetting)) {
1621 			enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1622 			napi_complete_done(napi, frames_processed);
1623 			return frames_processed;
1624 		}
1625 
1626 		if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
1627 			break;
1628 		next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
1629 		rx_buff =
1630 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
1631 							  rx_comp.correlator);
1632 		/* do error checking */
1633 		if (next->rx_comp.rc) {
1634 			netdev_dbg(netdev, "rx buffer returned with rc %x\n",
1635 				   be16_to_cpu(next->rx_comp.rc));
1636 			/* free the entry */
1637 			next->rx_comp.first = 0;
1638 			remove_buff_from_pool(adapter, rx_buff);
1639 			continue;
1640 		}
1641 
1642 		length = be32_to_cpu(next->rx_comp.len);
1643 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
1644 		flags = next->rx_comp.flags;
1645 		skb = rx_buff->skb;
1646 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
1647 					length);
1648 
1649 		/* VLAN Header has been stripped by the system firmware and
1650 		 * needs to be inserted by the driver
1651 		 */
1652 		if (adapter->rx_vlan_header_insertion &&
1653 		    (flags & IBMVNIC_VLAN_STRIPPED))
1654 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
1655 					       ntohs(next->rx_comp.vlan_tci));
1656 
1657 		/* free the entry */
1658 		next->rx_comp.first = 0;
1659 		remove_buff_from_pool(adapter, rx_buff);
1660 
1661 		skb_put(skb, length);
1662 		skb->protocol = eth_type_trans(skb, netdev);
1663 		skb_record_rx_queue(skb, scrq_num);
1664 
1665 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
1666 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
1667 			skb->ip_summed = CHECKSUM_UNNECESSARY;
1668 		}
1669 
1670 		length = skb->len;
1671 		napi_gro_receive(napi, skb); /* send it up */
1672 		netdev->stats.rx_packets++;
1673 		netdev->stats.rx_bytes += length;
1674 		adapter->rx_stats_buffers[scrq_num].packets++;
1675 		adapter->rx_stats_buffers[scrq_num].bytes += length;
1676 		frames_processed++;
1677 	}
1678 
1679 	if (adapter->state != VNIC_CLOSING)
1680 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
1681 
1682 	if (frames_processed < budget) {
1683 		enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1684 		napi_complete_done(napi, frames_processed);
1685 		if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
1686 		    napi_reschedule(napi)) {
1687 			disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1688 			goto restart_poll;
1689 		}
1690 	}
1691 	return frames_processed;
1692 }
1693 
1694 #ifdef CONFIG_NET_POLL_CONTROLLER
1695 static void ibmvnic_netpoll_controller(struct net_device *dev)
1696 {
1697 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1698 	int i;
1699 
1700 	replenish_pools(netdev_priv(dev));
1701 	for (i = 0; i < adapter->req_rx_queues; i++)
1702 		ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
1703 				     adapter->rx_scrq[i]);
1704 }
1705 #endif
1706 
1707 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
1708 {
1709 	return -EOPNOTSUPP;
1710 }
1711 
1712 static const struct net_device_ops ibmvnic_netdev_ops = {
1713 	.ndo_open		= ibmvnic_open,
1714 	.ndo_stop		= ibmvnic_close,
1715 	.ndo_start_xmit		= ibmvnic_xmit,
1716 	.ndo_set_rx_mode	= ibmvnic_set_multi,
1717 	.ndo_set_mac_address	= ibmvnic_set_mac,
1718 	.ndo_validate_addr	= eth_validate_addr,
1719 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
1720 #ifdef CONFIG_NET_POLL_CONTROLLER
1721 	.ndo_poll_controller	= ibmvnic_netpoll_controller,
1722 #endif
1723 	.ndo_change_mtu		= ibmvnic_change_mtu,
1724 };
1725 
1726 /* ethtool functions */
1727 
1728 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
1729 				      struct ethtool_link_ksettings *cmd)
1730 {
1731 	u32 supported, advertising;
1732 
1733 	supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
1734 			  SUPPORTED_FIBRE);
1735 	advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
1736 			    ADVERTISED_FIBRE);
1737 	cmd->base.speed = SPEED_1000;
1738 	cmd->base.duplex = DUPLEX_FULL;
1739 	cmd->base.port = PORT_FIBRE;
1740 	cmd->base.phy_address = 0;
1741 	cmd->base.autoneg = AUTONEG_ENABLE;
1742 
1743 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
1744 						supported);
1745 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
1746 						advertising);
1747 
1748 	return 0;
1749 }
1750 
1751 static void ibmvnic_get_drvinfo(struct net_device *dev,
1752 				struct ethtool_drvinfo *info)
1753 {
1754 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
1755 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
1756 }
1757 
1758 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
1759 {
1760 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1761 
1762 	return adapter->msg_enable;
1763 }
1764 
1765 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
1766 {
1767 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1768 
1769 	adapter->msg_enable = data;
1770 }
1771 
1772 static u32 ibmvnic_get_link(struct net_device *netdev)
1773 {
1774 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1775 
1776 	/* Don't need to send a query because we request a logical link up at
1777 	 * init and then we wait for link state indications
1778 	 */
1779 	return adapter->logical_link_state;
1780 }
1781 
1782 static void ibmvnic_get_ringparam(struct net_device *netdev,
1783 				  struct ethtool_ringparam *ring)
1784 {
1785 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1786 
1787 	ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
1788 	ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
1789 	ring->rx_mini_max_pending = 0;
1790 	ring->rx_jumbo_max_pending = 0;
1791 	ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
1792 	ring->tx_pending = adapter->req_tx_entries_per_subcrq;
1793 	ring->rx_mini_pending = 0;
1794 	ring->rx_jumbo_pending = 0;
1795 }
1796 
1797 static void ibmvnic_get_channels(struct net_device *netdev,
1798 				 struct ethtool_channels *channels)
1799 {
1800 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1801 
1802 	channels->max_rx = adapter->max_rx_queues;
1803 	channels->max_tx = adapter->max_tx_queues;
1804 	channels->max_other = 0;
1805 	channels->max_combined = 0;
1806 	channels->rx_count = adapter->req_rx_queues;
1807 	channels->tx_count = adapter->req_tx_queues;
1808 	channels->other_count = 0;
1809 	channels->combined_count = 0;
1810 }
1811 
1812 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
1813 {
1814 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1815 	int i;
1816 
1817 	if (stringset != ETH_SS_STATS)
1818 		return;
1819 
1820 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
1821 		memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
1822 
1823 	for (i = 0; i < adapter->req_tx_queues; i++) {
1824 		snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
1825 		data += ETH_GSTRING_LEN;
1826 
1827 		snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
1828 		data += ETH_GSTRING_LEN;
1829 
1830 		snprintf(data, ETH_GSTRING_LEN, "tx%d_dropped_packets", i);
1831 		data += ETH_GSTRING_LEN;
1832 	}
1833 
1834 	for (i = 0; i < adapter->req_rx_queues; i++) {
1835 		snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
1836 		data += ETH_GSTRING_LEN;
1837 
1838 		snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
1839 		data += ETH_GSTRING_LEN;
1840 
1841 		snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
1842 		data += ETH_GSTRING_LEN;
1843 	}
1844 }
1845 
1846 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
1847 {
1848 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1849 
1850 	switch (sset) {
1851 	case ETH_SS_STATS:
1852 		return ARRAY_SIZE(ibmvnic_stats) +
1853 		       adapter->req_tx_queues * NUM_TX_STATS +
1854 		       adapter->req_rx_queues * NUM_RX_STATS;
1855 	default:
1856 		return -EOPNOTSUPP;
1857 	}
1858 }
1859 
1860 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
1861 				      struct ethtool_stats *stats, u64 *data)
1862 {
1863 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1864 	union ibmvnic_crq crq;
1865 	int i, j;
1866 
1867 	memset(&crq, 0, sizeof(crq));
1868 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
1869 	crq.request_statistics.cmd = REQUEST_STATISTICS;
1870 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
1871 	crq.request_statistics.len =
1872 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
1873 
1874 	/* Wait for data to be written */
1875 	init_completion(&adapter->stats_done);
1876 	ibmvnic_send_crq(adapter, &crq);
1877 	wait_for_completion(&adapter->stats_done);
1878 
1879 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
1880 		data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
1881 						ibmvnic_stats[i].offset));
1882 
1883 	for (j = 0; j < adapter->req_tx_queues; j++) {
1884 		data[i] = adapter->tx_stats_buffers[j].packets;
1885 		i++;
1886 		data[i] = adapter->tx_stats_buffers[j].bytes;
1887 		i++;
1888 		data[i] = adapter->tx_stats_buffers[j].dropped_packets;
1889 		i++;
1890 	}
1891 
1892 	for (j = 0; j < adapter->req_rx_queues; j++) {
1893 		data[i] = adapter->rx_stats_buffers[j].packets;
1894 		i++;
1895 		data[i] = adapter->rx_stats_buffers[j].bytes;
1896 		i++;
1897 		data[i] = adapter->rx_stats_buffers[j].interrupts;
1898 		i++;
1899 	}
1900 }
1901 
1902 static const struct ethtool_ops ibmvnic_ethtool_ops = {
1903 	.get_drvinfo		= ibmvnic_get_drvinfo,
1904 	.get_msglevel		= ibmvnic_get_msglevel,
1905 	.set_msglevel		= ibmvnic_set_msglevel,
1906 	.get_link		= ibmvnic_get_link,
1907 	.get_ringparam		= ibmvnic_get_ringparam,
1908 	.get_channels		= ibmvnic_get_channels,
1909 	.get_strings            = ibmvnic_get_strings,
1910 	.get_sset_count         = ibmvnic_get_sset_count,
1911 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
1912 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
1913 };
1914 
1915 /* Routines for managing CRQs/sCRQs  */
1916 
1917 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
1918 				   struct ibmvnic_sub_crq_queue *scrq)
1919 {
1920 	int rc;
1921 
1922 	if (scrq->irq) {
1923 		free_irq(scrq->irq, scrq);
1924 		irq_dispose_mapping(scrq->irq);
1925 		scrq->irq = 0;
1926 	}
1927 
1928 	memset(scrq->msgs, 0, 4 * PAGE_SIZE);
1929 	scrq->cur = 0;
1930 
1931 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
1932 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
1933 	return rc;
1934 }
1935 
1936 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
1937 {
1938 	int i, rc;
1939 
1940 	for (i = 0; i < adapter->req_tx_queues; i++) {
1941 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
1942 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
1943 		if (rc)
1944 			return rc;
1945 	}
1946 
1947 	for (i = 0; i < adapter->req_rx_queues; i++) {
1948 		netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
1949 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
1950 		if (rc)
1951 			return rc;
1952 	}
1953 
1954 	return rc;
1955 }
1956 
1957 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
1958 				  struct ibmvnic_sub_crq_queue *scrq)
1959 {
1960 	struct device *dev = &adapter->vdev->dev;
1961 	long rc;
1962 
1963 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
1964 
1965 	/* Close the sub-crqs */
1966 	do {
1967 		rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
1968 					adapter->vdev->unit_address,
1969 					scrq->crq_num);
1970 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
1971 
1972 	if (rc) {
1973 		netdev_err(adapter->netdev,
1974 			   "Failed to release sub-CRQ %16lx, rc = %ld\n",
1975 			   scrq->crq_num, rc);
1976 	}
1977 
1978 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
1979 			 DMA_BIDIRECTIONAL);
1980 	free_pages((unsigned long)scrq->msgs, 2);
1981 	kfree(scrq);
1982 }
1983 
1984 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
1985 							*adapter)
1986 {
1987 	struct device *dev = &adapter->vdev->dev;
1988 	struct ibmvnic_sub_crq_queue *scrq;
1989 	int rc;
1990 
1991 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
1992 	if (!scrq)
1993 		return NULL;
1994 
1995 	scrq->msgs =
1996 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
1997 	if (!scrq->msgs) {
1998 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
1999 		goto zero_page_failed;
2000 	}
2001 
2002 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2003 					 DMA_BIDIRECTIONAL);
2004 	if (dma_mapping_error(dev, scrq->msg_token)) {
2005 		dev_warn(dev, "Couldn't map crq queue messages page\n");
2006 		goto map_failed;
2007 	}
2008 
2009 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2010 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2011 
2012 	if (rc == H_RESOURCE)
2013 		rc = ibmvnic_reset_crq(adapter);
2014 
2015 	if (rc == H_CLOSED) {
2016 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
2017 	} else if (rc) {
2018 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
2019 		goto reg_failed;
2020 	}
2021 
2022 	scrq->adapter = adapter;
2023 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2024 	spin_lock_init(&scrq->lock);
2025 
2026 	netdev_dbg(adapter->netdev,
2027 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2028 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
2029 
2030 	return scrq;
2031 
2032 reg_failed:
2033 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2034 			 DMA_BIDIRECTIONAL);
2035 map_failed:
2036 	free_pages((unsigned long)scrq->msgs, 2);
2037 zero_page_failed:
2038 	kfree(scrq);
2039 
2040 	return NULL;
2041 }
2042 
2043 static void release_sub_crqs(struct ibmvnic_adapter *adapter)
2044 {
2045 	int i;
2046 
2047 	if (adapter->tx_scrq) {
2048 		for (i = 0; i < adapter->req_tx_queues; i++) {
2049 			if (!adapter->tx_scrq[i])
2050 				continue;
2051 
2052 			netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2053 				   i);
2054 			if (adapter->tx_scrq[i]->irq) {
2055 				free_irq(adapter->tx_scrq[i]->irq,
2056 					 adapter->tx_scrq[i]);
2057 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2058 				adapter->tx_scrq[i]->irq = 0;
2059 			}
2060 
2061 			release_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2062 		}
2063 
2064 		kfree(adapter->tx_scrq);
2065 		adapter->tx_scrq = NULL;
2066 	}
2067 
2068 	if (adapter->rx_scrq) {
2069 		for (i = 0; i < adapter->req_rx_queues; i++) {
2070 			if (!adapter->rx_scrq[i])
2071 				continue;
2072 
2073 			netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2074 				   i);
2075 			if (adapter->rx_scrq[i]->irq) {
2076 				free_irq(adapter->rx_scrq[i]->irq,
2077 					 adapter->rx_scrq[i]);
2078 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2079 				adapter->rx_scrq[i]->irq = 0;
2080 			}
2081 
2082 			release_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2083 		}
2084 
2085 		kfree(adapter->rx_scrq);
2086 		adapter->rx_scrq = NULL;
2087 	}
2088 }
2089 
2090 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2091 			    struct ibmvnic_sub_crq_queue *scrq)
2092 {
2093 	struct device *dev = &adapter->vdev->dev;
2094 	unsigned long rc;
2095 
2096 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2097 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2098 	if (rc)
2099 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2100 			scrq->hw_irq, rc);
2101 	return rc;
2102 }
2103 
2104 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2105 			   struct ibmvnic_sub_crq_queue *scrq)
2106 {
2107 	struct device *dev = &adapter->vdev->dev;
2108 	unsigned long rc;
2109 
2110 	if (scrq->hw_irq > 0x100000000ULL) {
2111 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2112 		return 1;
2113 	}
2114 
2115 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2116 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2117 	if (rc)
2118 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2119 			scrq->hw_irq, rc);
2120 	return rc;
2121 }
2122 
2123 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2124 			       struct ibmvnic_sub_crq_queue *scrq)
2125 {
2126 	struct device *dev = &adapter->vdev->dev;
2127 	struct ibmvnic_tx_buff *txbuff;
2128 	union sub_crq *next;
2129 	int index;
2130 	int i, j;
2131 	u8 first;
2132 
2133 restart_loop:
2134 	while (pending_scrq(adapter, scrq)) {
2135 		unsigned int pool = scrq->pool_index;
2136 
2137 		next = ibmvnic_next_scrq(adapter, scrq);
2138 		for (i = 0; i < next->tx_comp.num_comps; i++) {
2139 			if (next->tx_comp.rcs[i]) {
2140 				dev_err(dev, "tx error %x\n",
2141 					next->tx_comp.rcs[i]);
2142 				continue;
2143 			}
2144 			index = be32_to_cpu(next->tx_comp.correlators[i]);
2145 			txbuff = &adapter->tx_pool[pool].tx_buff[index];
2146 
2147 			for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2148 				if (!txbuff->data_dma[j])
2149 					continue;
2150 
2151 				txbuff->data_dma[j] = 0;
2152 			}
2153 			/* if sub_crq was sent indirectly */
2154 			first = txbuff->indir_arr[0].generic.first;
2155 			if (first == IBMVNIC_CRQ_CMD) {
2156 				dma_unmap_single(dev, txbuff->indir_dma,
2157 						 sizeof(txbuff->indir_arr),
2158 						 DMA_TO_DEVICE);
2159 			}
2160 
2161 			if (txbuff->last_frag) {
2162 				dev_kfree_skb_any(txbuff->skb);
2163 				txbuff->skb = NULL;
2164 			}
2165 
2166 			adapter->tx_pool[pool].free_map[adapter->tx_pool[pool].
2167 						     producer_index] = index;
2168 			adapter->tx_pool[pool].producer_index =
2169 			    (adapter->tx_pool[pool].producer_index + 1) %
2170 			    adapter->req_tx_entries_per_subcrq;
2171 		}
2172 		/* remove tx_comp scrq*/
2173 		next->tx_comp.first = 0;
2174 
2175 		if (atomic_sub_return(next->tx_comp.num_comps, &scrq->used) <=
2176 		    (adapter->req_tx_entries_per_subcrq / 2) &&
2177 		    __netif_subqueue_stopped(adapter->netdev,
2178 					     scrq->pool_index)) {
2179 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2180 			netdev_info(adapter->netdev, "Started queue %d\n",
2181 				    scrq->pool_index);
2182 		}
2183 	}
2184 
2185 	enable_scrq_irq(adapter, scrq);
2186 
2187 	if (pending_scrq(adapter, scrq)) {
2188 		disable_scrq_irq(adapter, scrq);
2189 		goto restart_loop;
2190 	}
2191 
2192 	return 0;
2193 }
2194 
2195 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2196 {
2197 	struct ibmvnic_sub_crq_queue *scrq = instance;
2198 	struct ibmvnic_adapter *adapter = scrq->adapter;
2199 
2200 	disable_scrq_irq(adapter, scrq);
2201 	ibmvnic_complete_tx(adapter, scrq);
2202 
2203 	return IRQ_HANDLED;
2204 }
2205 
2206 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2207 {
2208 	struct ibmvnic_sub_crq_queue *scrq = instance;
2209 	struct ibmvnic_adapter *adapter = scrq->adapter;
2210 
2211 	adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2212 
2213 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2214 		disable_scrq_irq(adapter, scrq);
2215 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
2216 	}
2217 
2218 	return IRQ_HANDLED;
2219 }
2220 
2221 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2222 {
2223 	struct device *dev = &adapter->vdev->dev;
2224 	struct ibmvnic_sub_crq_queue *scrq;
2225 	int i = 0, j = 0;
2226 	int rc = 0;
2227 
2228 	for (i = 0; i < adapter->req_tx_queues; i++) {
2229 		netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2230 			   i);
2231 		scrq = adapter->tx_scrq[i];
2232 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2233 
2234 		if (!scrq->irq) {
2235 			rc = -EINVAL;
2236 			dev_err(dev, "Error mapping irq\n");
2237 			goto req_tx_irq_failed;
2238 		}
2239 
2240 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2241 				 0, "ibmvnic_tx", scrq);
2242 
2243 		if (rc) {
2244 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2245 				scrq->irq, rc);
2246 			irq_dispose_mapping(scrq->irq);
2247 			goto req_rx_irq_failed;
2248 		}
2249 	}
2250 
2251 	for (i = 0; i < adapter->req_rx_queues; i++) {
2252 		netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2253 			   i);
2254 		scrq = adapter->rx_scrq[i];
2255 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2256 		if (!scrq->irq) {
2257 			rc = -EINVAL;
2258 			dev_err(dev, "Error mapping irq\n");
2259 			goto req_rx_irq_failed;
2260 		}
2261 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2262 				 0, "ibmvnic_rx", scrq);
2263 		if (rc) {
2264 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2265 				scrq->irq, rc);
2266 			irq_dispose_mapping(scrq->irq);
2267 			goto req_rx_irq_failed;
2268 		}
2269 	}
2270 	return rc;
2271 
2272 req_rx_irq_failed:
2273 	for (j = 0; j < i; j++) {
2274 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2275 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2276 	}
2277 	i = adapter->req_tx_queues;
2278 req_tx_irq_failed:
2279 	for (j = 0; j < i; j++) {
2280 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2281 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2282 	}
2283 	release_sub_crqs(adapter);
2284 	return rc;
2285 }
2286 
2287 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2288 {
2289 	struct device *dev = &adapter->vdev->dev;
2290 	struct ibmvnic_sub_crq_queue **allqueues;
2291 	int registered_queues = 0;
2292 	int total_queues;
2293 	int more = 0;
2294 	int i;
2295 
2296 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2297 
2298 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2299 	if (!allqueues)
2300 		return -1;
2301 
2302 	for (i = 0; i < total_queues; i++) {
2303 		allqueues[i] = init_sub_crq_queue(adapter);
2304 		if (!allqueues[i]) {
2305 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2306 			break;
2307 		}
2308 		registered_queues++;
2309 	}
2310 
2311 	/* Make sure we were able to register the minimum number of queues */
2312 	if (registered_queues <
2313 	    adapter->min_tx_queues + adapter->min_rx_queues) {
2314 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2315 		goto tx_failed;
2316 	}
2317 
2318 	/* Distribute the failed allocated queues*/
2319 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
2320 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2321 		switch (i % 3) {
2322 		case 0:
2323 			if (adapter->req_rx_queues > adapter->min_rx_queues)
2324 				adapter->req_rx_queues--;
2325 			else
2326 				more++;
2327 			break;
2328 		case 1:
2329 			if (adapter->req_tx_queues > adapter->min_tx_queues)
2330 				adapter->req_tx_queues--;
2331 			else
2332 				more++;
2333 			break;
2334 		}
2335 	}
2336 
2337 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
2338 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
2339 	if (!adapter->tx_scrq)
2340 		goto tx_failed;
2341 
2342 	for (i = 0; i < adapter->req_tx_queues; i++) {
2343 		adapter->tx_scrq[i] = allqueues[i];
2344 		adapter->tx_scrq[i]->pool_index = i;
2345 	}
2346 
2347 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
2348 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
2349 	if (!adapter->rx_scrq)
2350 		goto rx_failed;
2351 
2352 	for (i = 0; i < adapter->req_rx_queues; i++) {
2353 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
2354 		adapter->rx_scrq[i]->scrq_num = i;
2355 	}
2356 
2357 	kfree(allqueues);
2358 	return 0;
2359 
2360 rx_failed:
2361 	kfree(adapter->tx_scrq);
2362 	adapter->tx_scrq = NULL;
2363 tx_failed:
2364 	for (i = 0; i < registered_queues; i++)
2365 		release_sub_crq_queue(adapter, allqueues[i]);
2366 	kfree(allqueues);
2367 	return -1;
2368 }
2369 
2370 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
2371 {
2372 	struct device *dev = &adapter->vdev->dev;
2373 	union ibmvnic_crq crq;
2374 
2375 	if (!retry) {
2376 		/* Sub-CRQ entries are 32 byte long */
2377 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
2378 
2379 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
2380 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
2381 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
2382 			return;
2383 		}
2384 
2385 		/* Get the minimum between the queried max and the entries
2386 		 * that fit in our PAGE_SIZE
2387 		 */
2388 		adapter->req_tx_entries_per_subcrq =
2389 		    adapter->max_tx_entries_per_subcrq > entries_page ?
2390 		    entries_page : adapter->max_tx_entries_per_subcrq;
2391 		adapter->req_rx_add_entries_per_subcrq =
2392 		    adapter->max_rx_add_entries_per_subcrq > entries_page ?
2393 		    entries_page : adapter->max_rx_add_entries_per_subcrq;
2394 
2395 		adapter->req_tx_queues = adapter->opt_tx_comp_sub_queues;
2396 		adapter->req_rx_queues = adapter->opt_rx_comp_queues;
2397 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
2398 
2399 		adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
2400 	}
2401 
2402 	memset(&crq, 0, sizeof(crq));
2403 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
2404 	crq.request_capability.cmd = REQUEST_CAPABILITY;
2405 
2406 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
2407 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
2408 	atomic_inc(&adapter->running_cap_crqs);
2409 	ibmvnic_send_crq(adapter, &crq);
2410 
2411 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
2412 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
2413 	atomic_inc(&adapter->running_cap_crqs);
2414 	ibmvnic_send_crq(adapter, &crq);
2415 
2416 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
2417 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
2418 	atomic_inc(&adapter->running_cap_crqs);
2419 	ibmvnic_send_crq(adapter, &crq);
2420 
2421 	crq.request_capability.capability =
2422 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
2423 	crq.request_capability.number =
2424 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
2425 	atomic_inc(&adapter->running_cap_crqs);
2426 	ibmvnic_send_crq(adapter, &crq);
2427 
2428 	crq.request_capability.capability =
2429 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
2430 	crq.request_capability.number =
2431 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
2432 	atomic_inc(&adapter->running_cap_crqs);
2433 	ibmvnic_send_crq(adapter, &crq);
2434 
2435 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
2436 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
2437 	atomic_inc(&adapter->running_cap_crqs);
2438 	ibmvnic_send_crq(adapter, &crq);
2439 
2440 	if (adapter->netdev->flags & IFF_PROMISC) {
2441 		if (adapter->promisc_supported) {
2442 			crq.request_capability.capability =
2443 			    cpu_to_be16(PROMISC_REQUESTED);
2444 			crq.request_capability.number = cpu_to_be64(1);
2445 			atomic_inc(&adapter->running_cap_crqs);
2446 			ibmvnic_send_crq(adapter, &crq);
2447 		}
2448 	} else {
2449 		crq.request_capability.capability =
2450 		    cpu_to_be16(PROMISC_REQUESTED);
2451 		crq.request_capability.number = cpu_to_be64(0);
2452 		atomic_inc(&adapter->running_cap_crqs);
2453 		ibmvnic_send_crq(adapter, &crq);
2454 	}
2455 }
2456 
2457 static int pending_scrq(struct ibmvnic_adapter *adapter,
2458 			struct ibmvnic_sub_crq_queue *scrq)
2459 {
2460 	union sub_crq *entry = &scrq->msgs[scrq->cur];
2461 
2462 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
2463 		return 1;
2464 	else
2465 		return 0;
2466 }
2467 
2468 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
2469 					struct ibmvnic_sub_crq_queue *scrq)
2470 {
2471 	union sub_crq *entry;
2472 	unsigned long flags;
2473 
2474 	spin_lock_irqsave(&scrq->lock, flags);
2475 	entry = &scrq->msgs[scrq->cur];
2476 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2477 		if (++scrq->cur == scrq->size)
2478 			scrq->cur = 0;
2479 	} else {
2480 		entry = NULL;
2481 	}
2482 	spin_unlock_irqrestore(&scrq->lock, flags);
2483 
2484 	return entry;
2485 }
2486 
2487 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
2488 {
2489 	struct ibmvnic_crq_queue *queue = &adapter->crq;
2490 	union ibmvnic_crq *crq;
2491 
2492 	crq = &queue->msgs[queue->cur];
2493 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2494 		if (++queue->cur == queue->size)
2495 			queue->cur = 0;
2496 	} else {
2497 		crq = NULL;
2498 	}
2499 
2500 	return crq;
2501 }
2502 
2503 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
2504 		       union sub_crq *sub_crq)
2505 {
2506 	unsigned int ua = adapter->vdev->unit_address;
2507 	struct device *dev = &adapter->vdev->dev;
2508 	u64 *u64_crq = (u64 *)sub_crq;
2509 	int rc;
2510 
2511 	netdev_dbg(adapter->netdev,
2512 		   "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
2513 		   (unsigned long int)cpu_to_be64(remote_handle),
2514 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
2515 		   (unsigned long int)cpu_to_be64(u64_crq[1]),
2516 		   (unsigned long int)cpu_to_be64(u64_crq[2]),
2517 		   (unsigned long int)cpu_to_be64(u64_crq[3]));
2518 
2519 	/* Make sure the hypervisor sees the complete request */
2520 	mb();
2521 
2522 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
2523 				cpu_to_be64(remote_handle),
2524 				cpu_to_be64(u64_crq[0]),
2525 				cpu_to_be64(u64_crq[1]),
2526 				cpu_to_be64(u64_crq[2]),
2527 				cpu_to_be64(u64_crq[3]));
2528 
2529 	if (rc) {
2530 		if (rc == H_CLOSED)
2531 			dev_warn(dev, "CRQ Queue closed\n");
2532 		dev_err(dev, "Send error (rc=%d)\n", rc);
2533 	}
2534 
2535 	return rc;
2536 }
2537 
2538 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
2539 				u64 remote_handle, u64 ioba, u64 num_entries)
2540 {
2541 	unsigned int ua = adapter->vdev->unit_address;
2542 	struct device *dev = &adapter->vdev->dev;
2543 	int rc;
2544 
2545 	/* Make sure the hypervisor sees the complete request */
2546 	mb();
2547 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
2548 				cpu_to_be64(remote_handle),
2549 				ioba, num_entries);
2550 
2551 	if (rc) {
2552 		if (rc == H_CLOSED)
2553 			dev_warn(dev, "CRQ Queue closed\n");
2554 		dev_err(dev, "Send (indirect) error (rc=%d)\n", rc);
2555 	}
2556 
2557 	return rc;
2558 }
2559 
2560 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
2561 			    union ibmvnic_crq *crq)
2562 {
2563 	unsigned int ua = adapter->vdev->unit_address;
2564 	struct device *dev = &adapter->vdev->dev;
2565 	u64 *u64_crq = (u64 *)crq;
2566 	int rc;
2567 
2568 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
2569 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
2570 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
2571 
2572 	/* Make sure the hypervisor sees the complete request */
2573 	mb();
2574 
2575 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
2576 				cpu_to_be64(u64_crq[0]),
2577 				cpu_to_be64(u64_crq[1]));
2578 
2579 	if (rc) {
2580 		if (rc == H_CLOSED)
2581 			dev_warn(dev, "CRQ Queue closed\n");
2582 		dev_warn(dev, "Send error (rc=%d)\n", rc);
2583 	}
2584 
2585 	return rc;
2586 }
2587 
2588 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
2589 {
2590 	union ibmvnic_crq crq;
2591 
2592 	memset(&crq, 0, sizeof(crq));
2593 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
2594 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
2595 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
2596 
2597 	return ibmvnic_send_crq(adapter, &crq);
2598 }
2599 
2600 static int send_version_xchg(struct ibmvnic_adapter *adapter)
2601 {
2602 	union ibmvnic_crq crq;
2603 
2604 	memset(&crq, 0, sizeof(crq));
2605 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
2606 	crq.version_exchange.cmd = VERSION_EXCHANGE;
2607 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
2608 
2609 	return ibmvnic_send_crq(adapter, &crq);
2610 }
2611 
2612 static void send_login(struct ibmvnic_adapter *adapter)
2613 {
2614 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
2615 	struct ibmvnic_login_buffer *login_buffer;
2616 	struct device *dev = &adapter->vdev->dev;
2617 	dma_addr_t rsp_buffer_token;
2618 	dma_addr_t buffer_token;
2619 	size_t rsp_buffer_size;
2620 	union ibmvnic_crq crq;
2621 	size_t buffer_size;
2622 	__be64 *tx_list_p;
2623 	__be64 *rx_list_p;
2624 	int i;
2625 
2626 	buffer_size =
2627 	    sizeof(struct ibmvnic_login_buffer) +
2628 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues);
2629 
2630 	login_buffer = kmalloc(buffer_size, GFP_ATOMIC);
2631 	if (!login_buffer)
2632 		goto buf_alloc_failed;
2633 
2634 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
2635 				      DMA_TO_DEVICE);
2636 	if (dma_mapping_error(dev, buffer_token)) {
2637 		dev_err(dev, "Couldn't map login buffer\n");
2638 		goto buf_map_failed;
2639 	}
2640 
2641 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
2642 			  sizeof(u64) * adapter->req_tx_queues +
2643 			  sizeof(u64) * adapter->req_rx_queues +
2644 			  sizeof(u64) * adapter->req_rx_queues +
2645 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
2646 
2647 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
2648 	if (!login_rsp_buffer)
2649 		goto buf_rsp_alloc_failed;
2650 
2651 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
2652 					  rsp_buffer_size, DMA_FROM_DEVICE);
2653 	if (dma_mapping_error(dev, rsp_buffer_token)) {
2654 		dev_err(dev, "Couldn't map login rsp buffer\n");
2655 		goto buf_rsp_map_failed;
2656 	}
2657 
2658 	adapter->login_buf = login_buffer;
2659 	adapter->login_buf_token = buffer_token;
2660 	adapter->login_buf_sz = buffer_size;
2661 	adapter->login_rsp_buf = login_rsp_buffer;
2662 	adapter->login_rsp_buf_token = rsp_buffer_token;
2663 	adapter->login_rsp_buf_sz = rsp_buffer_size;
2664 
2665 	login_buffer->len = cpu_to_be32(buffer_size);
2666 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
2667 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
2668 	login_buffer->off_txcomp_subcrqs =
2669 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
2670 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
2671 	login_buffer->off_rxcomp_subcrqs =
2672 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
2673 			sizeof(u64) * adapter->req_tx_queues);
2674 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
2675 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
2676 
2677 	tx_list_p = (__be64 *)((char *)login_buffer +
2678 				      sizeof(struct ibmvnic_login_buffer));
2679 	rx_list_p = (__be64 *)((char *)login_buffer +
2680 				      sizeof(struct ibmvnic_login_buffer) +
2681 				      sizeof(u64) * adapter->req_tx_queues);
2682 
2683 	for (i = 0; i < adapter->req_tx_queues; i++) {
2684 		if (adapter->tx_scrq[i]) {
2685 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
2686 						   crq_num);
2687 		}
2688 	}
2689 
2690 	for (i = 0; i < adapter->req_rx_queues; i++) {
2691 		if (adapter->rx_scrq[i]) {
2692 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
2693 						   crq_num);
2694 		}
2695 	}
2696 
2697 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
2698 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
2699 		netdev_dbg(adapter->netdev, "%016lx\n",
2700 			   ((unsigned long int *)(adapter->login_buf))[i]);
2701 	}
2702 
2703 	memset(&crq, 0, sizeof(crq));
2704 	crq.login.first = IBMVNIC_CRQ_CMD;
2705 	crq.login.cmd = LOGIN;
2706 	crq.login.ioba = cpu_to_be32(buffer_token);
2707 	crq.login.len = cpu_to_be32(buffer_size);
2708 	ibmvnic_send_crq(adapter, &crq);
2709 
2710 	return;
2711 
2712 buf_rsp_map_failed:
2713 	kfree(login_rsp_buffer);
2714 buf_rsp_alloc_failed:
2715 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
2716 buf_map_failed:
2717 	kfree(login_buffer);
2718 buf_alloc_failed:
2719 	return;
2720 }
2721 
2722 static void send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
2723 			     u32 len, u8 map_id)
2724 {
2725 	union ibmvnic_crq crq;
2726 
2727 	memset(&crq, 0, sizeof(crq));
2728 	crq.request_map.first = IBMVNIC_CRQ_CMD;
2729 	crq.request_map.cmd = REQUEST_MAP;
2730 	crq.request_map.map_id = map_id;
2731 	crq.request_map.ioba = cpu_to_be32(addr);
2732 	crq.request_map.len = cpu_to_be32(len);
2733 	ibmvnic_send_crq(adapter, &crq);
2734 }
2735 
2736 static void send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
2737 {
2738 	union ibmvnic_crq crq;
2739 
2740 	memset(&crq, 0, sizeof(crq));
2741 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
2742 	crq.request_unmap.cmd = REQUEST_UNMAP;
2743 	crq.request_unmap.map_id = map_id;
2744 	ibmvnic_send_crq(adapter, &crq);
2745 }
2746 
2747 static void send_map_query(struct ibmvnic_adapter *adapter)
2748 {
2749 	union ibmvnic_crq crq;
2750 
2751 	memset(&crq, 0, sizeof(crq));
2752 	crq.query_map.first = IBMVNIC_CRQ_CMD;
2753 	crq.query_map.cmd = QUERY_MAP;
2754 	ibmvnic_send_crq(adapter, &crq);
2755 }
2756 
2757 /* Send a series of CRQs requesting various capabilities of the VNIC server */
2758 static void send_cap_queries(struct ibmvnic_adapter *adapter)
2759 {
2760 	union ibmvnic_crq crq;
2761 
2762 	atomic_set(&adapter->running_cap_crqs, 0);
2763 	memset(&crq, 0, sizeof(crq));
2764 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
2765 	crq.query_capability.cmd = QUERY_CAPABILITY;
2766 
2767 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
2768 	atomic_inc(&adapter->running_cap_crqs);
2769 	ibmvnic_send_crq(adapter, &crq);
2770 
2771 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
2772 	atomic_inc(&adapter->running_cap_crqs);
2773 	ibmvnic_send_crq(adapter, &crq);
2774 
2775 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
2776 	atomic_inc(&adapter->running_cap_crqs);
2777 	ibmvnic_send_crq(adapter, &crq);
2778 
2779 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
2780 	atomic_inc(&adapter->running_cap_crqs);
2781 	ibmvnic_send_crq(adapter, &crq);
2782 
2783 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
2784 	atomic_inc(&adapter->running_cap_crqs);
2785 	ibmvnic_send_crq(adapter, &crq);
2786 
2787 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
2788 	atomic_inc(&adapter->running_cap_crqs);
2789 	ibmvnic_send_crq(adapter, &crq);
2790 
2791 	crq.query_capability.capability =
2792 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
2793 	atomic_inc(&adapter->running_cap_crqs);
2794 	ibmvnic_send_crq(adapter, &crq);
2795 
2796 	crq.query_capability.capability =
2797 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
2798 	atomic_inc(&adapter->running_cap_crqs);
2799 	ibmvnic_send_crq(adapter, &crq);
2800 
2801 	crq.query_capability.capability =
2802 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
2803 	atomic_inc(&adapter->running_cap_crqs);
2804 	ibmvnic_send_crq(adapter, &crq);
2805 
2806 	crq.query_capability.capability =
2807 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
2808 	atomic_inc(&adapter->running_cap_crqs);
2809 	ibmvnic_send_crq(adapter, &crq);
2810 
2811 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
2812 	atomic_inc(&adapter->running_cap_crqs);
2813 	ibmvnic_send_crq(adapter, &crq);
2814 
2815 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
2816 	atomic_inc(&adapter->running_cap_crqs);
2817 	ibmvnic_send_crq(adapter, &crq);
2818 
2819 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
2820 	atomic_inc(&adapter->running_cap_crqs);
2821 	ibmvnic_send_crq(adapter, &crq);
2822 
2823 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
2824 	atomic_inc(&adapter->running_cap_crqs);
2825 	ibmvnic_send_crq(adapter, &crq);
2826 
2827 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
2828 	atomic_inc(&adapter->running_cap_crqs);
2829 	ibmvnic_send_crq(adapter, &crq);
2830 
2831 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
2832 	atomic_inc(&adapter->running_cap_crqs);
2833 	ibmvnic_send_crq(adapter, &crq);
2834 
2835 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
2836 	atomic_inc(&adapter->running_cap_crqs);
2837 	ibmvnic_send_crq(adapter, &crq);
2838 
2839 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
2840 	atomic_inc(&adapter->running_cap_crqs);
2841 	ibmvnic_send_crq(adapter, &crq);
2842 
2843 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
2844 	atomic_inc(&adapter->running_cap_crqs);
2845 	ibmvnic_send_crq(adapter, &crq);
2846 
2847 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
2848 	atomic_inc(&adapter->running_cap_crqs);
2849 	ibmvnic_send_crq(adapter, &crq);
2850 
2851 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
2852 	atomic_inc(&adapter->running_cap_crqs);
2853 	ibmvnic_send_crq(adapter, &crq);
2854 
2855 	crq.query_capability.capability =
2856 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
2857 	atomic_inc(&adapter->running_cap_crqs);
2858 	ibmvnic_send_crq(adapter, &crq);
2859 
2860 	crq.query_capability.capability =
2861 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
2862 	atomic_inc(&adapter->running_cap_crqs);
2863 	ibmvnic_send_crq(adapter, &crq);
2864 
2865 	crq.query_capability.capability =
2866 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
2867 	atomic_inc(&adapter->running_cap_crqs);
2868 	ibmvnic_send_crq(adapter, &crq);
2869 
2870 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
2871 	atomic_inc(&adapter->running_cap_crqs);
2872 	ibmvnic_send_crq(adapter, &crq);
2873 }
2874 
2875 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
2876 {
2877 	struct device *dev = &adapter->vdev->dev;
2878 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
2879 	union ibmvnic_crq crq;
2880 	int i;
2881 
2882 	dma_unmap_single(dev, adapter->ip_offload_tok,
2883 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
2884 
2885 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
2886 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
2887 		netdev_dbg(adapter->netdev, "%016lx\n",
2888 			   ((unsigned long int *)(buf))[i]);
2889 
2890 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
2891 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
2892 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
2893 		   buf->tcp_ipv4_chksum);
2894 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
2895 		   buf->tcp_ipv6_chksum);
2896 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
2897 		   buf->udp_ipv4_chksum);
2898 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
2899 		   buf->udp_ipv6_chksum);
2900 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
2901 		   buf->large_tx_ipv4);
2902 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
2903 		   buf->large_tx_ipv6);
2904 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
2905 		   buf->large_rx_ipv4);
2906 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
2907 		   buf->large_rx_ipv6);
2908 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
2909 		   buf->max_ipv4_header_size);
2910 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
2911 		   buf->max_ipv6_header_size);
2912 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
2913 		   buf->max_tcp_header_size);
2914 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
2915 		   buf->max_udp_header_size);
2916 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
2917 		   buf->max_large_tx_size);
2918 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
2919 		   buf->max_large_rx_size);
2920 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
2921 		   buf->ipv6_extension_header);
2922 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
2923 		   buf->tcp_pseudosum_req);
2924 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
2925 		   buf->num_ipv6_ext_headers);
2926 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
2927 		   buf->off_ipv6_ext_headers);
2928 
2929 	adapter->ip_offload_ctrl_tok =
2930 	    dma_map_single(dev, &adapter->ip_offload_ctrl,
2931 			   sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
2932 
2933 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
2934 		dev_err(dev, "Couldn't map ip offload control buffer\n");
2935 		return;
2936 	}
2937 
2938 	adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
2939 	adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
2940 	adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
2941 	adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
2942 	adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
2943 
2944 	/* large_tx/rx disabled for now, additional features needed */
2945 	adapter->ip_offload_ctrl.large_tx_ipv4 = 0;
2946 	adapter->ip_offload_ctrl.large_tx_ipv6 = 0;
2947 	adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
2948 	adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
2949 
2950 	adapter->netdev->features = NETIF_F_GSO;
2951 
2952 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
2953 		adapter->netdev->features |= NETIF_F_IP_CSUM;
2954 
2955 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
2956 		adapter->netdev->features |= NETIF_F_IPV6_CSUM;
2957 
2958 	if ((adapter->netdev->features &
2959 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
2960 		adapter->netdev->features |= NETIF_F_RXCSUM;
2961 
2962 	memset(&crq, 0, sizeof(crq));
2963 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
2964 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
2965 	crq.control_ip_offload.len =
2966 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
2967 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
2968 	ibmvnic_send_crq(adapter, &crq);
2969 }
2970 
2971 static void handle_error_info_rsp(union ibmvnic_crq *crq,
2972 				  struct ibmvnic_adapter *adapter)
2973 {
2974 	struct device *dev = &adapter->vdev->dev;
2975 	struct ibmvnic_error_buff *error_buff, *tmp;
2976 	unsigned long flags;
2977 	bool found = false;
2978 	int i;
2979 
2980 	if (!crq->request_error_rsp.rc.code) {
2981 		dev_info(dev, "Request Error Rsp returned with rc=%x\n",
2982 			 crq->request_error_rsp.rc.code);
2983 		return;
2984 	}
2985 
2986 	spin_lock_irqsave(&adapter->error_list_lock, flags);
2987 	list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list)
2988 		if (error_buff->error_id == crq->request_error_rsp.error_id) {
2989 			found = true;
2990 			list_del(&error_buff->list);
2991 			break;
2992 		}
2993 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
2994 
2995 	if (!found) {
2996 		dev_err(dev, "Couldn't find error id %x\n",
2997 			be32_to_cpu(crq->request_error_rsp.error_id));
2998 		return;
2999 	}
3000 
3001 	dev_err(dev, "Detailed info for error id %x:",
3002 		be32_to_cpu(crq->request_error_rsp.error_id));
3003 
3004 	for (i = 0; i < error_buff->len; i++) {
3005 		pr_cont("%02x", (int)error_buff->buff[i]);
3006 		if (i % 8 == 7)
3007 			pr_cont(" ");
3008 	}
3009 	pr_cont("\n");
3010 
3011 	dma_unmap_single(dev, error_buff->dma, error_buff->len,
3012 			 DMA_FROM_DEVICE);
3013 	kfree(error_buff->buff);
3014 	kfree(error_buff);
3015 }
3016 
3017 static void request_error_information(struct ibmvnic_adapter *adapter,
3018 				      union ibmvnic_crq *err_crq)
3019 {
3020 	struct device *dev = &adapter->vdev->dev;
3021 	struct net_device *netdev = adapter->netdev;
3022 	struct ibmvnic_error_buff *error_buff;
3023 	unsigned long timeout = msecs_to_jiffies(30000);
3024 	union ibmvnic_crq crq;
3025 	unsigned long flags;
3026 	int rc, detail_len;
3027 
3028 	error_buff = kmalloc(sizeof(*error_buff), GFP_ATOMIC);
3029 	if (!error_buff)
3030 		return;
3031 
3032 	detail_len = be32_to_cpu(err_crq->error_indication.detail_error_sz);
3033 	error_buff->buff = kmalloc(detail_len, GFP_ATOMIC);
3034 	if (!error_buff->buff) {
3035 		kfree(error_buff);
3036 		return;
3037 	}
3038 
3039 	error_buff->dma = dma_map_single(dev, error_buff->buff, detail_len,
3040 					 DMA_FROM_DEVICE);
3041 	if (dma_mapping_error(dev, error_buff->dma)) {
3042 		netdev_err(netdev, "Couldn't map error buffer\n");
3043 		kfree(error_buff->buff);
3044 		kfree(error_buff);
3045 		return;
3046 	}
3047 
3048 	error_buff->len = detail_len;
3049 	error_buff->error_id = err_crq->error_indication.error_id;
3050 
3051 	spin_lock_irqsave(&adapter->error_list_lock, flags);
3052 	list_add_tail(&error_buff->list, &adapter->errors);
3053 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3054 
3055 	memset(&crq, 0, sizeof(crq));
3056 	crq.request_error_info.first = IBMVNIC_CRQ_CMD;
3057 	crq.request_error_info.cmd = REQUEST_ERROR_INFO;
3058 	crq.request_error_info.ioba = cpu_to_be32(error_buff->dma);
3059 	crq.request_error_info.len = cpu_to_be32(detail_len);
3060 	crq.request_error_info.error_id = err_crq->error_indication.error_id;
3061 
3062 	rc = ibmvnic_send_crq(adapter, &crq);
3063 	if (rc) {
3064 		netdev_err(netdev, "failed to request error information\n");
3065 		goto err_info_fail;
3066 	}
3067 
3068 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
3069 		netdev_err(netdev, "timeout waiting for error information\n");
3070 		goto err_info_fail;
3071 	}
3072 
3073 	return;
3074 
3075 err_info_fail:
3076 	spin_lock_irqsave(&adapter->error_list_lock, flags);
3077 	list_del(&error_buff->list);
3078 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3079 
3080 	kfree(error_buff->buff);
3081 	kfree(error_buff);
3082 }
3083 
3084 static void handle_error_indication(union ibmvnic_crq *crq,
3085 				    struct ibmvnic_adapter *adapter)
3086 {
3087 	struct device *dev = &adapter->vdev->dev;
3088 
3089 	dev_err(dev, "Firmware reports %serror id %x, cause %d\n",
3090 		crq->error_indication.flags
3091 			& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3092 		be32_to_cpu(crq->error_indication.error_id),
3093 		be16_to_cpu(crq->error_indication.error_cause));
3094 
3095 	if (be32_to_cpu(crq->error_indication.error_id))
3096 		request_error_information(adapter, crq);
3097 
3098 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3099 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3100 	else
3101 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3102 }
3103 
3104 static void handle_change_mac_rsp(union ibmvnic_crq *crq,
3105 				  struct ibmvnic_adapter *adapter)
3106 {
3107 	struct net_device *netdev = adapter->netdev;
3108 	struct device *dev = &adapter->vdev->dev;
3109 	long rc;
3110 
3111 	rc = crq->change_mac_addr_rsp.rc.code;
3112 	if (rc) {
3113 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3114 		return;
3115 	}
3116 	memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
3117 	       ETH_ALEN);
3118 }
3119 
3120 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3121 				   struct ibmvnic_adapter *adapter)
3122 {
3123 	struct device *dev = &adapter->vdev->dev;
3124 	u64 *req_value;
3125 	char *name;
3126 
3127 	atomic_dec(&adapter->running_cap_crqs);
3128 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3129 	case REQ_TX_QUEUES:
3130 		req_value = &adapter->req_tx_queues;
3131 		name = "tx";
3132 		break;
3133 	case REQ_RX_QUEUES:
3134 		req_value = &adapter->req_rx_queues;
3135 		name = "rx";
3136 		break;
3137 	case REQ_RX_ADD_QUEUES:
3138 		req_value = &adapter->req_rx_add_queues;
3139 		name = "rx_add";
3140 		break;
3141 	case REQ_TX_ENTRIES_PER_SUBCRQ:
3142 		req_value = &adapter->req_tx_entries_per_subcrq;
3143 		name = "tx_entries_per_subcrq";
3144 		break;
3145 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3146 		req_value = &adapter->req_rx_add_entries_per_subcrq;
3147 		name = "rx_add_entries_per_subcrq";
3148 		break;
3149 	case REQ_MTU:
3150 		req_value = &adapter->req_mtu;
3151 		name = "mtu";
3152 		break;
3153 	case PROMISC_REQUESTED:
3154 		req_value = &adapter->promisc;
3155 		name = "promisc";
3156 		break;
3157 	default:
3158 		dev_err(dev, "Got invalid cap request rsp %d\n",
3159 			crq->request_capability.capability);
3160 		return;
3161 	}
3162 
3163 	switch (crq->request_capability_rsp.rc.code) {
3164 	case SUCCESS:
3165 		break;
3166 	case PARTIALSUCCESS:
3167 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3168 			 *req_value,
3169 			 (long int)be64_to_cpu(crq->request_capability_rsp.
3170 					       number), name);
3171 		*req_value = be64_to_cpu(crq->request_capability_rsp.number);
3172 		ibmvnic_send_req_caps(adapter, 1);
3173 		return;
3174 	default:
3175 		dev_err(dev, "Error %d in request cap rsp\n",
3176 			crq->request_capability_rsp.rc.code);
3177 		return;
3178 	}
3179 
3180 	/* Done receiving requested capabilities, query IP offload support */
3181 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
3182 		union ibmvnic_crq newcrq;
3183 		int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3184 		struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
3185 		    &adapter->ip_offload_buf;
3186 
3187 		adapter->wait_capability = false;
3188 		adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
3189 							 buf_sz,
3190 							 DMA_FROM_DEVICE);
3191 
3192 		if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3193 			if (!firmware_has_feature(FW_FEATURE_CMO))
3194 				dev_err(dev, "Couldn't map offload buffer\n");
3195 			return;
3196 		}
3197 
3198 		memset(&newcrq, 0, sizeof(newcrq));
3199 		newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3200 		newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3201 		newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
3202 		newcrq.query_ip_offload.ioba =
3203 		    cpu_to_be32(adapter->ip_offload_tok);
3204 
3205 		ibmvnic_send_crq(adapter, &newcrq);
3206 	}
3207 }
3208 
3209 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
3210 			    struct ibmvnic_adapter *adapter)
3211 {
3212 	struct device *dev = &adapter->vdev->dev;
3213 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
3214 	struct ibmvnic_login_buffer *login = adapter->login_buf;
3215 	int i;
3216 
3217 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
3218 			 DMA_BIDIRECTIONAL);
3219 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
3220 			 adapter->login_rsp_buf_sz, DMA_BIDIRECTIONAL);
3221 
3222 	/* If the number of queues requested can't be allocated by the
3223 	 * server, the login response will return with code 1. We will need
3224 	 * to resend the login buffer with fewer queues requested.
3225 	 */
3226 	if (login_rsp_crq->generic.rc.code) {
3227 		adapter->renegotiate = true;
3228 		complete(&adapter->init_done);
3229 		return 0;
3230 	}
3231 
3232 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
3233 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
3234 		netdev_dbg(adapter->netdev, "%016lx\n",
3235 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
3236 	}
3237 
3238 	/* Sanity checks */
3239 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
3240 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
3241 	     adapter->req_rx_add_queues !=
3242 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
3243 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
3244 		ibmvnic_remove(adapter->vdev);
3245 		return -EIO;
3246 	}
3247 	complete(&adapter->init_done);
3248 
3249 	return 0;
3250 }
3251 
3252 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
3253 				     struct ibmvnic_adapter *adapter)
3254 {
3255 	struct device *dev = &adapter->vdev->dev;
3256 	long rc;
3257 
3258 	rc = crq->request_unmap_rsp.rc.code;
3259 	if (rc)
3260 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
3261 }
3262 
3263 static void handle_query_map_rsp(union ibmvnic_crq *crq,
3264 				 struct ibmvnic_adapter *adapter)
3265 {
3266 	struct net_device *netdev = adapter->netdev;
3267 	struct device *dev = &adapter->vdev->dev;
3268 	long rc;
3269 
3270 	rc = crq->query_map_rsp.rc.code;
3271 	if (rc) {
3272 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
3273 		return;
3274 	}
3275 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
3276 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
3277 		   crq->query_map_rsp.free_pages);
3278 }
3279 
3280 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
3281 				 struct ibmvnic_adapter *adapter)
3282 {
3283 	struct net_device *netdev = adapter->netdev;
3284 	struct device *dev = &adapter->vdev->dev;
3285 	long rc;
3286 
3287 	atomic_dec(&adapter->running_cap_crqs);
3288 	netdev_dbg(netdev, "Outstanding queries: %d\n",
3289 		   atomic_read(&adapter->running_cap_crqs));
3290 	rc = crq->query_capability.rc.code;
3291 	if (rc) {
3292 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
3293 		goto out;
3294 	}
3295 
3296 	switch (be16_to_cpu(crq->query_capability.capability)) {
3297 	case MIN_TX_QUEUES:
3298 		adapter->min_tx_queues =
3299 		    be64_to_cpu(crq->query_capability.number);
3300 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
3301 			   adapter->min_tx_queues);
3302 		break;
3303 	case MIN_RX_QUEUES:
3304 		adapter->min_rx_queues =
3305 		    be64_to_cpu(crq->query_capability.number);
3306 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
3307 			   adapter->min_rx_queues);
3308 		break;
3309 	case MIN_RX_ADD_QUEUES:
3310 		adapter->min_rx_add_queues =
3311 		    be64_to_cpu(crq->query_capability.number);
3312 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
3313 			   adapter->min_rx_add_queues);
3314 		break;
3315 	case MAX_TX_QUEUES:
3316 		adapter->max_tx_queues =
3317 		    be64_to_cpu(crq->query_capability.number);
3318 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
3319 			   adapter->max_tx_queues);
3320 		break;
3321 	case MAX_RX_QUEUES:
3322 		adapter->max_rx_queues =
3323 		    be64_to_cpu(crq->query_capability.number);
3324 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
3325 			   adapter->max_rx_queues);
3326 		break;
3327 	case MAX_RX_ADD_QUEUES:
3328 		adapter->max_rx_add_queues =
3329 		    be64_to_cpu(crq->query_capability.number);
3330 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
3331 			   adapter->max_rx_add_queues);
3332 		break;
3333 	case MIN_TX_ENTRIES_PER_SUBCRQ:
3334 		adapter->min_tx_entries_per_subcrq =
3335 		    be64_to_cpu(crq->query_capability.number);
3336 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
3337 			   adapter->min_tx_entries_per_subcrq);
3338 		break;
3339 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
3340 		adapter->min_rx_add_entries_per_subcrq =
3341 		    be64_to_cpu(crq->query_capability.number);
3342 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
3343 			   adapter->min_rx_add_entries_per_subcrq);
3344 		break;
3345 	case MAX_TX_ENTRIES_PER_SUBCRQ:
3346 		adapter->max_tx_entries_per_subcrq =
3347 		    be64_to_cpu(crq->query_capability.number);
3348 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
3349 			   adapter->max_tx_entries_per_subcrq);
3350 		break;
3351 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
3352 		adapter->max_rx_add_entries_per_subcrq =
3353 		    be64_to_cpu(crq->query_capability.number);
3354 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
3355 			   adapter->max_rx_add_entries_per_subcrq);
3356 		break;
3357 	case TCP_IP_OFFLOAD:
3358 		adapter->tcp_ip_offload =
3359 		    be64_to_cpu(crq->query_capability.number);
3360 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
3361 			   adapter->tcp_ip_offload);
3362 		break;
3363 	case PROMISC_SUPPORTED:
3364 		adapter->promisc_supported =
3365 		    be64_to_cpu(crq->query_capability.number);
3366 		netdev_dbg(netdev, "promisc_supported = %lld\n",
3367 			   adapter->promisc_supported);
3368 		break;
3369 	case MIN_MTU:
3370 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
3371 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
3372 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
3373 		break;
3374 	case MAX_MTU:
3375 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
3376 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
3377 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
3378 		break;
3379 	case MAX_MULTICAST_FILTERS:
3380 		adapter->max_multicast_filters =
3381 		    be64_to_cpu(crq->query_capability.number);
3382 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
3383 			   adapter->max_multicast_filters);
3384 		break;
3385 	case VLAN_HEADER_INSERTION:
3386 		adapter->vlan_header_insertion =
3387 		    be64_to_cpu(crq->query_capability.number);
3388 		if (adapter->vlan_header_insertion)
3389 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
3390 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
3391 			   adapter->vlan_header_insertion);
3392 		break;
3393 	case RX_VLAN_HEADER_INSERTION:
3394 		adapter->rx_vlan_header_insertion =
3395 		    be64_to_cpu(crq->query_capability.number);
3396 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
3397 			   adapter->rx_vlan_header_insertion);
3398 		break;
3399 	case MAX_TX_SG_ENTRIES:
3400 		adapter->max_tx_sg_entries =
3401 		    be64_to_cpu(crq->query_capability.number);
3402 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
3403 			   adapter->max_tx_sg_entries);
3404 		break;
3405 	case RX_SG_SUPPORTED:
3406 		adapter->rx_sg_supported =
3407 		    be64_to_cpu(crq->query_capability.number);
3408 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
3409 			   adapter->rx_sg_supported);
3410 		break;
3411 	case OPT_TX_COMP_SUB_QUEUES:
3412 		adapter->opt_tx_comp_sub_queues =
3413 		    be64_to_cpu(crq->query_capability.number);
3414 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
3415 			   adapter->opt_tx_comp_sub_queues);
3416 		break;
3417 	case OPT_RX_COMP_QUEUES:
3418 		adapter->opt_rx_comp_queues =
3419 		    be64_to_cpu(crq->query_capability.number);
3420 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
3421 			   adapter->opt_rx_comp_queues);
3422 		break;
3423 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
3424 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
3425 		    be64_to_cpu(crq->query_capability.number);
3426 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
3427 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
3428 		break;
3429 	case OPT_TX_ENTRIES_PER_SUBCRQ:
3430 		adapter->opt_tx_entries_per_subcrq =
3431 		    be64_to_cpu(crq->query_capability.number);
3432 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
3433 			   adapter->opt_tx_entries_per_subcrq);
3434 		break;
3435 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
3436 		adapter->opt_rxba_entries_per_subcrq =
3437 		    be64_to_cpu(crq->query_capability.number);
3438 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
3439 			   adapter->opt_rxba_entries_per_subcrq);
3440 		break;
3441 	case TX_RX_DESC_REQ:
3442 		adapter->tx_rx_desc_req = crq->query_capability.number;
3443 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
3444 			   adapter->tx_rx_desc_req);
3445 		break;
3446 
3447 	default:
3448 		netdev_err(netdev, "Got invalid cap rsp %d\n",
3449 			   crq->query_capability.capability);
3450 	}
3451 
3452 out:
3453 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
3454 		adapter->wait_capability = false;
3455 		ibmvnic_send_req_caps(adapter, 0);
3456 	}
3457 }
3458 
3459 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
3460 			       struct ibmvnic_adapter *adapter)
3461 {
3462 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
3463 	struct net_device *netdev = adapter->netdev;
3464 	struct device *dev = &adapter->vdev->dev;
3465 	u64 *u64_crq = (u64 *)crq;
3466 	long rc;
3467 
3468 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
3469 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3470 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
3471 	switch (gen_crq->first) {
3472 	case IBMVNIC_CRQ_INIT_RSP:
3473 		switch (gen_crq->cmd) {
3474 		case IBMVNIC_CRQ_INIT:
3475 			dev_info(dev, "Partner initialized\n");
3476 			adapter->from_passive_init = true;
3477 			complete(&adapter->init_done);
3478 			break;
3479 		case IBMVNIC_CRQ_INIT_COMPLETE:
3480 			dev_info(dev, "Partner initialization complete\n");
3481 			send_version_xchg(adapter);
3482 			break;
3483 		default:
3484 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
3485 		}
3486 		return;
3487 	case IBMVNIC_CRQ_XPORT_EVENT:
3488 		netif_carrier_off(netdev);
3489 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
3490 			dev_info(dev, "Migrated, re-enabling adapter\n");
3491 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
3492 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
3493 			dev_info(dev, "Backing device failover detected\n");
3494 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
3495 		} else {
3496 			/* The adapter lost the connection */
3497 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
3498 				gen_crq->cmd);
3499 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3500 		}
3501 		return;
3502 	case IBMVNIC_CRQ_CMD_RSP:
3503 		break;
3504 	default:
3505 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
3506 			gen_crq->first);
3507 		return;
3508 	}
3509 
3510 	switch (gen_crq->cmd) {
3511 	case VERSION_EXCHANGE_RSP:
3512 		rc = crq->version_exchange_rsp.rc.code;
3513 		if (rc) {
3514 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
3515 			break;
3516 		}
3517 		dev_info(dev, "Partner protocol version is %d\n",
3518 			 crq->version_exchange_rsp.version);
3519 		if (be16_to_cpu(crq->version_exchange_rsp.version) <
3520 		    ibmvnic_version)
3521 			ibmvnic_version =
3522 			    be16_to_cpu(crq->version_exchange_rsp.version);
3523 		send_cap_queries(adapter);
3524 		break;
3525 	case QUERY_CAPABILITY_RSP:
3526 		handle_query_cap_rsp(crq, adapter);
3527 		break;
3528 	case QUERY_MAP_RSP:
3529 		handle_query_map_rsp(crq, adapter);
3530 		break;
3531 	case REQUEST_MAP_RSP:
3532 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
3533 		complete(&adapter->fw_done);
3534 		break;
3535 	case REQUEST_UNMAP_RSP:
3536 		handle_request_unmap_rsp(crq, adapter);
3537 		break;
3538 	case REQUEST_CAPABILITY_RSP:
3539 		handle_request_cap_rsp(crq, adapter);
3540 		break;
3541 	case LOGIN_RSP:
3542 		netdev_dbg(netdev, "Got Login Response\n");
3543 		handle_login_rsp(crq, adapter);
3544 		break;
3545 	case LOGICAL_LINK_STATE_RSP:
3546 		netdev_dbg(netdev,
3547 			   "Got Logical Link State Response, state: %d rc: %d\n",
3548 			   crq->logical_link_state_rsp.link_state,
3549 			   crq->logical_link_state_rsp.rc.code);
3550 		adapter->logical_link_state =
3551 		    crq->logical_link_state_rsp.link_state;
3552 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
3553 		complete(&adapter->init_done);
3554 		break;
3555 	case LINK_STATE_INDICATION:
3556 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
3557 		adapter->phys_link_state =
3558 		    crq->link_state_indication.phys_link_state;
3559 		adapter->logical_link_state =
3560 		    crq->link_state_indication.logical_link_state;
3561 		break;
3562 	case CHANGE_MAC_ADDR_RSP:
3563 		netdev_dbg(netdev, "Got MAC address change Response\n");
3564 		handle_change_mac_rsp(crq, adapter);
3565 		break;
3566 	case ERROR_INDICATION:
3567 		netdev_dbg(netdev, "Got Error Indication\n");
3568 		handle_error_indication(crq, adapter);
3569 		break;
3570 	case REQUEST_ERROR_RSP:
3571 		netdev_dbg(netdev, "Got Error Detail Response\n");
3572 		handle_error_info_rsp(crq, adapter);
3573 		break;
3574 	case REQUEST_STATISTICS_RSP:
3575 		netdev_dbg(netdev, "Got Statistics Response\n");
3576 		complete(&adapter->stats_done);
3577 		break;
3578 	case QUERY_IP_OFFLOAD_RSP:
3579 		netdev_dbg(netdev, "Got Query IP offload Response\n");
3580 		handle_query_ip_offload_rsp(adapter);
3581 		break;
3582 	case MULTICAST_CTRL_RSP:
3583 		netdev_dbg(netdev, "Got multicast control Response\n");
3584 		break;
3585 	case CONTROL_IP_OFFLOAD_RSP:
3586 		netdev_dbg(netdev, "Got Control IP offload Response\n");
3587 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
3588 				 sizeof(adapter->ip_offload_ctrl),
3589 				 DMA_TO_DEVICE);
3590 		complete(&adapter->init_done);
3591 		break;
3592 	case COLLECT_FW_TRACE_RSP:
3593 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
3594 		complete(&adapter->fw_done);
3595 		break;
3596 	default:
3597 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
3598 			   gen_crq->cmd);
3599 	}
3600 }
3601 
3602 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
3603 {
3604 	struct ibmvnic_adapter *adapter = instance;
3605 
3606 	tasklet_schedule(&adapter->tasklet);
3607 	return IRQ_HANDLED;
3608 }
3609 
3610 static void ibmvnic_tasklet(void *data)
3611 {
3612 	struct ibmvnic_adapter *adapter = data;
3613 	struct ibmvnic_crq_queue *queue = &adapter->crq;
3614 	union ibmvnic_crq *crq;
3615 	unsigned long flags;
3616 	bool done = false;
3617 
3618 	spin_lock_irqsave(&queue->lock, flags);
3619 	while (!done) {
3620 		/* Pull all the valid messages off the CRQ */
3621 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
3622 			ibmvnic_handle_crq(crq, adapter);
3623 			crq->generic.first = 0;
3624 		}
3625 
3626 		/* remain in tasklet until all
3627 		 * capabilities responses are received
3628 		 */
3629 		if (!adapter->wait_capability)
3630 			done = true;
3631 	}
3632 	/* if capabilities CRQ's were sent in this tasklet, the following
3633 	 * tasklet must wait until all responses are received
3634 	 */
3635 	if (atomic_read(&adapter->running_cap_crqs) != 0)
3636 		adapter->wait_capability = true;
3637 	spin_unlock_irqrestore(&queue->lock, flags);
3638 }
3639 
3640 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
3641 {
3642 	struct vio_dev *vdev = adapter->vdev;
3643 	int rc;
3644 
3645 	do {
3646 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
3647 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
3648 
3649 	if (rc)
3650 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
3651 
3652 	return rc;
3653 }
3654 
3655 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
3656 {
3657 	struct ibmvnic_crq_queue *crq = &adapter->crq;
3658 	struct device *dev = &adapter->vdev->dev;
3659 	struct vio_dev *vdev = adapter->vdev;
3660 	int rc;
3661 
3662 	/* Close the CRQ */
3663 	do {
3664 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3665 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3666 
3667 	/* Clean out the queue */
3668 	memset(crq->msgs, 0, PAGE_SIZE);
3669 	crq->cur = 0;
3670 
3671 	/* And re-open it again */
3672 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
3673 				crq->msg_token, PAGE_SIZE);
3674 
3675 	if (rc == H_CLOSED)
3676 		/* Adapter is good, but other end is not ready */
3677 		dev_warn(dev, "Partner adapter not ready\n");
3678 	else if (rc != 0)
3679 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
3680 
3681 	return rc;
3682 }
3683 
3684 static void release_crq_queue(struct ibmvnic_adapter *adapter)
3685 {
3686 	struct ibmvnic_crq_queue *crq = &adapter->crq;
3687 	struct vio_dev *vdev = adapter->vdev;
3688 	long rc;
3689 
3690 	if (!crq->msgs)
3691 		return;
3692 
3693 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
3694 	free_irq(vdev->irq, adapter);
3695 	tasklet_kill(&adapter->tasklet);
3696 	do {
3697 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3698 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3699 
3700 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
3701 			 DMA_BIDIRECTIONAL);
3702 	free_page((unsigned long)crq->msgs);
3703 	crq->msgs = NULL;
3704 }
3705 
3706 static int init_crq_queue(struct ibmvnic_adapter *adapter)
3707 {
3708 	struct ibmvnic_crq_queue *crq = &adapter->crq;
3709 	struct device *dev = &adapter->vdev->dev;
3710 	struct vio_dev *vdev = adapter->vdev;
3711 	int rc, retrc = -ENOMEM;
3712 
3713 	if (crq->msgs)
3714 		return 0;
3715 
3716 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
3717 	/* Should we allocate more than one page? */
3718 
3719 	if (!crq->msgs)
3720 		return -ENOMEM;
3721 
3722 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
3723 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
3724 					DMA_BIDIRECTIONAL);
3725 	if (dma_mapping_error(dev, crq->msg_token))
3726 		goto map_failed;
3727 
3728 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
3729 				crq->msg_token, PAGE_SIZE);
3730 
3731 	if (rc == H_RESOURCE)
3732 		/* maybe kexecing and resource is busy. try a reset */
3733 		rc = ibmvnic_reset_crq(adapter);
3734 	retrc = rc;
3735 
3736 	if (rc == H_CLOSED) {
3737 		dev_warn(dev, "Partner adapter not ready\n");
3738 	} else if (rc) {
3739 		dev_warn(dev, "Error %d opening adapter\n", rc);
3740 		goto reg_crq_failed;
3741 	}
3742 
3743 	retrc = 0;
3744 
3745 	tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
3746 		     (unsigned long)adapter);
3747 
3748 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
3749 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
3750 			 adapter);
3751 	if (rc) {
3752 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
3753 			vdev->irq, rc);
3754 		goto req_irq_failed;
3755 	}
3756 
3757 	rc = vio_enable_interrupts(vdev);
3758 	if (rc) {
3759 		dev_err(dev, "Error %d enabling interrupts\n", rc);
3760 		goto req_irq_failed;
3761 	}
3762 
3763 	crq->cur = 0;
3764 	spin_lock_init(&crq->lock);
3765 
3766 	return retrc;
3767 
3768 req_irq_failed:
3769 	tasklet_kill(&adapter->tasklet);
3770 	do {
3771 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3772 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3773 reg_crq_failed:
3774 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
3775 map_failed:
3776 	free_page((unsigned long)crq->msgs);
3777 	crq->msgs = NULL;
3778 	return retrc;
3779 }
3780 
3781 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
3782 {
3783 	struct device *dev = &adapter->vdev->dev;
3784 	unsigned long timeout = msecs_to_jiffies(30000);
3785 	int rc;
3786 
3787 	if (adapter->resetting) {
3788 		rc = ibmvnic_reset_crq(adapter);
3789 		if (!rc)
3790 			rc = vio_enable_interrupts(adapter->vdev);
3791 	} else {
3792 		rc = init_crq_queue(adapter);
3793 	}
3794 
3795 	if (rc) {
3796 		dev_err(dev, "Couldn't initialize crq. rc=%d\n", rc);
3797 		return rc;
3798 	}
3799 
3800 	adapter->from_passive_init = false;
3801 
3802 	init_completion(&adapter->init_done);
3803 	adapter->init_done_rc = 0;
3804 	ibmvnic_send_crq_init(adapter);
3805 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
3806 		dev_err(dev, "Initialization sequence timed out\n");
3807 		return -1;
3808 	}
3809 
3810 	if (adapter->init_done_rc) {
3811 		release_crq_queue(adapter);
3812 		return adapter->init_done_rc;
3813 	}
3814 
3815 	if (adapter->from_passive_init) {
3816 		adapter->state = VNIC_OPEN;
3817 		adapter->from_passive_init = false;
3818 		return -1;
3819 	}
3820 
3821 	if (adapter->resetting)
3822 		rc = reset_sub_crq_queues(adapter);
3823 	else
3824 		rc = init_sub_crqs(adapter);
3825 	if (rc) {
3826 		dev_err(dev, "Initialization of sub crqs failed\n");
3827 		release_crq_queue(adapter);
3828 		return rc;
3829 	}
3830 
3831 	rc = init_sub_crq_irqs(adapter);
3832 	if (rc) {
3833 		dev_err(dev, "Failed to initialize sub crq irqs\n");
3834 		release_crq_queue(adapter);
3835 	}
3836 
3837 	return rc;
3838 }
3839 
3840 static struct device_attribute dev_attr_failover;
3841 
3842 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
3843 {
3844 	struct ibmvnic_adapter *adapter;
3845 	struct net_device *netdev;
3846 	unsigned char *mac_addr_p;
3847 	int rc;
3848 
3849 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
3850 		dev->unit_address);
3851 
3852 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
3853 							VETH_MAC_ADDR, NULL);
3854 	if (!mac_addr_p) {
3855 		dev_err(&dev->dev,
3856 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
3857 			__FILE__, __LINE__);
3858 		return 0;
3859 	}
3860 
3861 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
3862 				   IBMVNIC_MAX_TX_QUEUES);
3863 	if (!netdev)
3864 		return -ENOMEM;
3865 
3866 	adapter = netdev_priv(netdev);
3867 	adapter->state = VNIC_PROBING;
3868 	dev_set_drvdata(&dev->dev, netdev);
3869 	adapter->vdev = dev;
3870 	adapter->netdev = netdev;
3871 
3872 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
3873 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
3874 	netdev->irq = dev->irq;
3875 	netdev->netdev_ops = &ibmvnic_netdev_ops;
3876 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
3877 	SET_NETDEV_DEV(netdev, &dev->dev);
3878 
3879 	spin_lock_init(&adapter->stats_lock);
3880 
3881 	INIT_LIST_HEAD(&adapter->errors);
3882 	spin_lock_init(&adapter->error_list_lock);
3883 
3884 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
3885 	INIT_LIST_HEAD(&adapter->rwi_list);
3886 	mutex_init(&adapter->reset_lock);
3887 	mutex_init(&adapter->rwi_lock);
3888 	adapter->resetting = false;
3889 
3890 	do {
3891 		rc = ibmvnic_init(adapter);
3892 		if (rc && rc != EAGAIN)
3893 			goto ibmvnic_init_fail;
3894 	} while (rc == EAGAIN);
3895 
3896 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
3897 
3898 	rc = device_create_file(&dev->dev, &dev_attr_failover);
3899 	if (rc)
3900 		goto ibmvnic_init_fail;
3901 
3902 	rc = register_netdev(netdev);
3903 	if (rc) {
3904 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
3905 		goto ibmvnic_register_fail;
3906 	}
3907 	dev_info(&dev->dev, "ibmvnic registered\n");
3908 
3909 	adapter->state = VNIC_PROBED;
3910 	return 0;
3911 
3912 ibmvnic_register_fail:
3913 	device_remove_file(&dev->dev, &dev_attr_failover);
3914 
3915 ibmvnic_init_fail:
3916 	release_sub_crqs(adapter);
3917 	release_crq_queue(adapter);
3918 	free_netdev(netdev);
3919 
3920 	return rc;
3921 }
3922 
3923 static int ibmvnic_remove(struct vio_dev *dev)
3924 {
3925 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
3926 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
3927 
3928 	adapter->state = VNIC_REMOVING;
3929 	unregister_netdev(netdev);
3930 	mutex_lock(&adapter->reset_lock);
3931 
3932 	release_resources(adapter);
3933 	release_sub_crqs(adapter);
3934 	release_crq_queue(adapter);
3935 
3936 	adapter->state = VNIC_REMOVED;
3937 
3938 	mutex_unlock(&adapter->reset_lock);
3939 	device_remove_file(&dev->dev, &dev_attr_failover);
3940 	free_netdev(netdev);
3941 	dev_set_drvdata(&dev->dev, NULL);
3942 
3943 	return 0;
3944 }
3945 
3946 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
3947 			      const char *buf, size_t count)
3948 {
3949 	struct net_device *netdev = dev_get_drvdata(dev);
3950 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
3951 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
3952 	__be64 session_token;
3953 	long rc;
3954 
3955 	if (!sysfs_streq(buf, "1"))
3956 		return -EINVAL;
3957 
3958 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
3959 			 H_GET_SESSION_TOKEN, 0, 0, 0);
3960 	if (rc) {
3961 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
3962 			   rc);
3963 		return -EINVAL;
3964 	}
3965 
3966 	session_token = (__be64)retbuf[0];
3967 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
3968 		   be64_to_cpu(session_token));
3969 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3970 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
3971 	if (rc) {
3972 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
3973 			   rc);
3974 		return -EINVAL;
3975 	}
3976 
3977 	return count;
3978 }
3979 
3980 static DEVICE_ATTR(failover, 0200, NULL, failover_store);
3981 
3982 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
3983 {
3984 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
3985 	struct ibmvnic_adapter *adapter;
3986 	struct iommu_table *tbl;
3987 	unsigned long ret = 0;
3988 	int i;
3989 
3990 	tbl = get_iommu_table_base(&vdev->dev);
3991 
3992 	/* netdev inits at probe time along with the structures we need below*/
3993 	if (!netdev)
3994 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
3995 
3996 	adapter = netdev_priv(netdev);
3997 
3998 	ret += PAGE_SIZE; /* the crq message queue */
3999 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4000 
4001 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4002 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
4003 
4004 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4005 	     i++)
4006 		ret += adapter->rx_pool[i].size *
4007 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4008 
4009 	return ret;
4010 }
4011 
4012 static int ibmvnic_resume(struct device *dev)
4013 {
4014 	struct net_device *netdev = dev_get_drvdata(dev);
4015 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4016 
4017 	if (adapter->state != VNIC_OPEN)
4018 		return 0;
4019 
4020 	tasklet_schedule(&adapter->tasklet);
4021 
4022 	return 0;
4023 }
4024 
4025 static const struct vio_device_id ibmvnic_device_table[] = {
4026 	{"network", "IBM,vnic"},
4027 	{"", "" }
4028 };
4029 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
4030 
4031 static const struct dev_pm_ops ibmvnic_pm_ops = {
4032 	.resume = ibmvnic_resume
4033 };
4034 
4035 static struct vio_driver ibmvnic_driver = {
4036 	.id_table       = ibmvnic_device_table,
4037 	.probe          = ibmvnic_probe,
4038 	.remove         = ibmvnic_remove,
4039 	.get_desired_dma = ibmvnic_get_desired_dma,
4040 	.name		= ibmvnic_driver_name,
4041 	.pm		= &ibmvnic_pm_ops,
4042 };
4043 
4044 /* module functions */
4045 static int __init ibmvnic_module_init(void)
4046 {
4047 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
4048 		IBMVNIC_DRIVER_VERSION);
4049 
4050 	return vio_register_driver(&ibmvnic_driver);
4051 }
4052 
4053 static void __exit ibmvnic_module_exit(void)
4054 {
4055 	vio_unregister_driver(&ibmvnic_driver);
4056 }
4057 
4058 module_init(ibmvnic_module_init);
4059 module_exit(ibmvnic_module_exit);
4060