drivers/net/ethernet/ibm/ibmvnic.c

   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/if_arp.h>
  63 #include <linux/in.h>
  64 #include <linux/ip.h>
  65 #include <linux/ipv6.h>
  66 #include <linux/irq.h>
  67 #include <linux/kthread.h>
  68 #include <linux/seq_file.h>
  69 #include <linux/interrupt.h>
  70 #include <net/net_namespace.h>
  71 #include <asm/hvcall.h>
  72 #include <linux/atomic.h>
  73 #include <asm/vio.h>
  74 #include <asm/iommu.h>
  75 #include <linux/uaccess.h>
  76 #include <asm/firmware.h>
  77 #include <linux/workqueue.h>
  78 #include <linux/if_vlan.h>
  79 #include <linux/utsname.h>
  80
  81 #include "ibmvnic.h"
  82
  83 static const char ibmvnic_driver_name[] = "ibmvnic";
  84 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
  85
  86 MODULE_AUTHOR("Santiago Leon");
  87 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
  88 MODULE_LICENSE("GPL");
  89 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
  90
  91 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
  92 static int ibmvnic_remove(struct vio_dev *);
  93 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
  94 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
  95 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
  96 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
  97 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
  98 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
  99                        union sub_crq *sub_crq);
 100 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
 101 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
 102 static int enable_scrq_irq(struct ibmvnic_adapter *,
 103                            struct ibmvnic_sub_crq_queue *);
 104 static int disable_scrq_irq(struct ibmvnic_adapter *,
 105                             struct ibmvnic_sub_crq_queue *);
 106 static int pending_scrq(struct ibmvnic_adapter *,
 107                         struct ibmvnic_sub_crq_queue *);
 108 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
 109                                         struct ibmvnic_sub_crq_queue *);
 110 static int ibmvnic_poll(struct napi_struct *napi, int data);
 111 static void send_map_query(struct ibmvnic_adapter *adapter);
 112 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
 113 static int send_request_unmap(struct ibmvnic_adapter *, u8);
 114 static int send_login(struct ibmvnic_adapter *adapter);
 115 static void send_cap_queries(struct ibmvnic_adapter *adapter);
 116 static int init_sub_crqs(struct ibmvnic_adapter *);
 117 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
 118 static int ibmvnic_init(struct ibmvnic_adapter *);
 119 static int ibmvnic_reset_init(struct ibmvnic_adapter *);
 120 static void release_crq_queue(struct ibmvnic_adapter *);
 121 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p);
 122 static int init_crq_queue(struct ibmvnic_adapter *adapter);
 123
 124 struct ibmvnic_stat {
 125         char name[ETH_GSTRING_LEN];
 126         int offset;
 127 };
 128
 129 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
 130                              offsetof(struct ibmvnic_statistics, stat))
 131 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
 132
 133 static const struct ibmvnic_stat ibmvnic_stats[] = {
 134         {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
 135         {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
 136         {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
 137         {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
 138         {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
 139         {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
 140         {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
 141         {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
 142         {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
 143         {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
 144         {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
 145         {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
 146         {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
 147         {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
 148         {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
 149         {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
 150         {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
 151         {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
 152         {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
 153         {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
 154         {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
 155         {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
 156 };
 157
 158 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
 159                           unsigned long length, unsigned long *number,
 160                           unsigned long *irq)
 161 {
 162         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
 163         long rc;
 164
 165         rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
 166         *number = retbuf[0];
 167         *irq = retbuf[1];
 168
 169         return rc;
 170 }
 171
 172 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
 173                                 struct ibmvnic_long_term_buff *ltb, int size)
 174 {
 175         struct device *dev = &adapter->vdev->dev;
 176         int rc;
 177
 178         ltb->size = size;
 179         ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
 180                                        GFP_KERNEL);
 181
 182         if (!ltb->buff) {
 183                 dev_err(dev, "Couldn't alloc long term buffer\n");
 184                 return -ENOMEM;
 185         }
 186         ltb->map_id = adapter->map_id;
 187         adapter->map_id++;
 188
 189         init_completion(&adapter->fw_done);
 190         rc = send_request_map(adapter, ltb->addr,
 191                               ltb->size, ltb->map_id);
 192         if (rc) {
 193                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 194                 return rc;
 195         }
 196         wait_for_completion(&adapter->fw_done);
 197
 198         if (adapter->fw_done_rc) {
 199                 dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
 200                         adapter->fw_done_rc);
 201                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 202                 return -1;
 203         }
 204         return 0;
 205 }
 206
 207 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
 208                                 struct ibmvnic_long_term_buff *ltb)
 209 {
 210         struct device *dev = &adapter->vdev->dev;
 211
 212         if (!ltb->buff)
 213                 return;
 214
 215         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
 216             adapter->reset_reason != VNIC_RESET_MOBILITY)
 217                 send_request_unmap(adapter, ltb->map_id);
 218         dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 219 }
 220
 221 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
 222                                 struct ibmvnic_long_term_buff *ltb)
 223 {
 224         int rc;
 225
 226         memset(ltb->buff, 0, ltb->size);
 227
 228         init_completion(&adapter->fw_done);
 229         rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
 230         if (rc)
 231                 return rc;
 232         wait_for_completion(&adapter->fw_done);
 233
 234         if (adapter->fw_done_rc) {
 235                 dev_info(&adapter->vdev->dev,
 236                          "Reset failed, attempting to free and reallocate buffer\n");
 237                 free_long_term_buff(adapter, ltb);
 238                 return alloc_long_term_buff(adapter, ltb, ltb->size);
 239         }
 240         return 0;
 241 }
 242
 243 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
 244 {
 245         int i;
 246
 247         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 248              i++)
 249                 adapter->rx_pool[i].active = 0;
 250 }
 251
 252 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
 253                               struct ibmvnic_rx_pool *pool)
 254 {
 255         int count = pool->size - atomic_read(&pool->available);
 256         struct device *dev = &adapter->vdev->dev;
 257         int buffers_added = 0;
 258         unsigned long lpar_rc;
 259         union sub_crq sub_crq;
 260         struct sk_buff *skb;
 261         unsigned int offset;
 262         dma_addr_t dma_addr;
 263         unsigned char *dst;
 264         u64 *handle_array;
 265         int shift = 0;
 266         int index;
 267         int i;
 268
 269         if (!pool->active)
 270                 return;
 271
 272         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 273                                       be32_to_cpu(adapter->login_rsp_buf->
 274                                       off_rxadd_subcrqs));
 275
 276         for (i = 0; i < count; ++i) {
 277                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 278                 if (!skb) {
 279                         dev_err(dev, "Couldn't replenish rx buff\n");
 280                         adapter->replenish_no_mem++;
 281                         break;
 282                 }
 283
 284                 index = pool->free_map[pool->next_free];
 285
 286                 if (pool->rx_buff[index].skb)
 287                         dev_err(dev, "Inconsistent free_map!\n");
 288
 289                 /* Copy the skb to the long term mapped DMA buffer */
 290                 offset = index * pool->buff_size;
 291                 dst = pool->long_term_buff.buff + offset;
 292                 memset(dst, 0, pool->buff_size);
 293                 dma_addr = pool->long_term_buff.addr + offset;
 294                 pool->rx_buff[index].data = dst;
 295
 296                 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
 297                 pool->rx_buff[index].dma = dma_addr;
 298                 pool->rx_buff[index].skb = skb;
 299                 pool->rx_buff[index].pool_index = pool->index;
 300                 pool->rx_buff[index].size = pool->buff_size;
 301
 302                 memset(&sub_crq, 0, sizeof(sub_crq));
 303                 sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
 304                 sub_crq.rx_add.correlator =
 305                     cpu_to_be64((u64)&pool->rx_buff[index]);
 306                 sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
 307                 sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
 308
 309                 /* The length field of the sCRQ is defined to be 24 bits so the
 310                  * buffer size needs to be left shifted by a byte before it is
 311                  * converted to big endian to prevent the last byte from being
 312                  * truncated.
 313                  */
 314 #ifdef __LITTLE_ENDIAN__
 315                 shift = 8;
 316 #endif
 317                 sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
 318
 319                 lpar_rc = send_subcrq(adapter, handle_array[pool->index],
 320                                       &sub_crq);
 321                 if (lpar_rc != H_SUCCESS)
 322                         goto failure;
 323
 324                 buffers_added++;
 325                 adapter->replenish_add_buff_success++;
 326                 pool->next_free = (pool->next_free + 1) % pool->size;
 327         }
 328         atomic_add(buffers_added, &pool->available);
 329         return;
 330
 331 failure:
 332         if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
 333                 dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
 334         pool->free_map[pool->next_free] = index;
 335         pool->rx_buff[index].skb = NULL;
 336
 337         dev_kfree_skb_any(skb);
 338         adapter->replenish_add_buff_failure++;
 339         atomic_add(buffers_added, &pool->available);
 340
 341         if (lpar_rc == H_CLOSED || adapter->failover_pending) {
 342                 /* Disable buffer pool replenishment and report carrier off if
 343                  * queue is closed or pending failover.
 344                  * Firmware guarantees that a signal will be sent to the
 345                  * driver, triggering a reset.
 346                  */
 347                 deactivate_rx_pools(adapter);
 348                 netif_carrier_off(adapter->netdev);
 349         }
 350 }
 351
 352 static void replenish_pools(struct ibmvnic_adapter *adapter)
 353 {
 354         int i;
 355
 356         adapter->replenish_task_cycles++;
 357         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 358              i++) {
 359                 if (adapter->rx_pool[i].active)
 360                         replenish_rx_pool(adapter, &adapter->rx_pool[i]);
 361         }
 362 }
 363
 364 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
 365 {
 366         kfree(adapter->tx_stats_buffers);
 367         kfree(adapter->rx_stats_buffers);
 368         adapter->tx_stats_buffers = NULL;
 369         adapter->rx_stats_buffers = NULL;
 370 }
 371
 372 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
 373 {
 374         adapter->tx_stats_buffers =
 375                                 kcalloc(IBMVNIC_MAX_QUEUES,
 376                                         sizeof(struct ibmvnic_tx_queue_stats),
 377                                         GFP_KERNEL);
 378         if (!adapter->tx_stats_buffers)
 379                 return -ENOMEM;
 380
 381         adapter->rx_stats_buffers =
 382                                 kcalloc(IBMVNIC_MAX_QUEUES,
 383                                         sizeof(struct ibmvnic_rx_queue_stats),
 384                                         GFP_KERNEL);
 385         if (!adapter->rx_stats_buffers)
 386                 return -ENOMEM;
 387
 388         return 0;
 389 }
 390
 391 static void release_stats_token(struct ibmvnic_adapter *adapter)
 392 {
 393         struct device *dev = &adapter->vdev->dev;
 394
 395         if (!adapter->stats_token)
 396                 return;
 397
 398         dma_unmap_single(dev, adapter->stats_token,
 399                          sizeof(struct ibmvnic_statistics),
 400                          DMA_FROM_DEVICE);
 401         adapter->stats_token = 0;
 402 }
 403
 404 static int init_stats_token(struct ibmvnic_adapter *adapter)
 405 {
 406         struct device *dev = &adapter->vdev->dev;
 407         dma_addr_t stok;
 408
 409         stok = dma_map_single(dev, &adapter->stats,
 410                               sizeof(struct ibmvnic_statistics),
 411                               DMA_FROM_DEVICE);
 412         if (dma_mapping_error(dev, stok)) {
 413                 dev_err(dev, "Couldn't map stats buffer\n");
 414                 return -1;
 415         }
 416
 417         adapter->stats_token = stok;
 418         netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
 419         return 0;
 420 }
 421
 422 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
 423 {
 424         struct ibmvnic_rx_pool *rx_pool;
 425         int rx_scrqs;
 426         int i, j, rc;
 427         u64 *size_array;
 428
 429         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 430                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 431
 432         rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 433         for (i = 0; i < rx_scrqs; i++) {
 434                 rx_pool = &adapter->rx_pool[i];
 435
 436                 netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
 437
 438                 if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
 439                         free_long_term_buff(adapter, &rx_pool->long_term_buff);
 440                         rx_pool->buff_size = be64_to_cpu(size_array[i]);
 441                         rc = alloc_long_term_buff(adapter,
 442                                                   &rx_pool->long_term_buff,
 443                                                   rx_pool->size *
 444                                                   rx_pool->buff_size);
 445                 } else {
 446                         rc = reset_long_term_buff(adapter,
 447                                                   &rx_pool->long_term_buff);
 448                 }
 449
 450                 if (rc)
 451                         return rc;
 452
 453                 for (j = 0; j < rx_pool->size; j++)
 454                         rx_pool->free_map[j] = j;
 455
 456                 memset(rx_pool->rx_buff, 0,
 457                        rx_pool->size * sizeof(struct ibmvnic_rx_buff));
 458
 459                 atomic_set(&rx_pool->available, 0);
 460                 rx_pool->next_alloc = 0;
 461                 rx_pool->next_free = 0;
 462                 rx_pool->active = 1;
 463         }
 464
 465         return 0;
 466 }
 467
 468 static void release_rx_pools(struct ibmvnic_adapter *adapter)
 469 {
 470         struct ibmvnic_rx_pool *rx_pool;
 471         int i, j;
 472
 473         if (!adapter->rx_pool)
 474                 return;
 475
 476         for (i = 0; i < adapter->num_active_rx_pools; i++) {
 477                 rx_pool = &adapter->rx_pool[i];
 478
 479                 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
 480
 481                 kfree(rx_pool->free_map);
 482                 free_long_term_buff(adapter, &rx_pool->long_term_buff);
 483
 484                 if (!rx_pool->rx_buff)
 485                         continue;
 486
 487                 for (j = 0; j < rx_pool->size; j++) {
 488                         if (rx_pool->rx_buff[j].skb) {
 489                                 dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
 490                                 rx_pool->rx_buff[j].skb = NULL;
 491                         }
 492                 }
 493
 494                 kfree(rx_pool->rx_buff);
 495         }
 496
 497         kfree(adapter->rx_pool);
 498         adapter->rx_pool = NULL;
 499         adapter->num_active_rx_pools = 0;
 500 }
 501
 502 static int init_rx_pools(struct net_device *netdev)
 503 {
 504         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 505         struct device *dev = &adapter->vdev->dev;
 506         struct ibmvnic_rx_pool *rx_pool;
 507         int rxadd_subcrqs;
 508         u64 *size_array;
 509         int i, j;
 510
 511         rxadd_subcrqs =
 512                 be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 513         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 514                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 515
 516         adapter->rx_pool = kcalloc(rxadd_subcrqs,
 517                                    sizeof(struct ibmvnic_rx_pool),
 518                                    GFP_KERNEL);
 519         if (!adapter->rx_pool) {
 520                 dev_err(dev, "Failed to allocate rx pools\n");
 521                 return -1;
 522         }
 523
 524         adapter->num_active_rx_pools = rxadd_subcrqs;
 525
 526         for (i = 0; i < rxadd_subcrqs; i++) {
 527                 rx_pool = &adapter->rx_pool[i];
 528
 529                 netdev_dbg(adapter->netdev,
 530                            "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
 531                            i, adapter->req_rx_add_entries_per_subcrq,
 532                            be64_to_cpu(size_array[i]));
 533
 534                 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
 535                 rx_pool->index = i;
 536                 rx_pool->buff_size = be64_to_cpu(size_array[i]);
 537                 rx_pool->active = 1;
 538
 539                 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
 540                                             GFP_KERNEL);
 541                 if (!rx_pool->free_map) {
 542                         release_rx_pools(adapter);
 543                         return -1;
 544                 }
 545
 546                 rx_pool->rx_buff = kcalloc(rx_pool->size,
 547                                            sizeof(struct ibmvnic_rx_buff),
 548                                            GFP_KERNEL);
 549                 if (!rx_pool->rx_buff) {
 550                         dev_err(dev, "Couldn't alloc rx buffers\n");
 551                         release_rx_pools(adapter);
 552                         return -1;
 553                 }
 554
 555                 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 556                                          rx_pool->size * rx_pool->buff_size)) {
 557                         release_rx_pools(adapter);
 558                         return -1;
 559                 }
 560
 561                 for (j = 0; j < rx_pool->size; ++j)
 562                         rx_pool->free_map[j] = j;
 563
 564                 atomic_set(&rx_pool->available, 0);
 565                 rx_pool->next_alloc = 0;
 566                 rx_pool->next_free = 0;
 567         }
 568
 569         return 0;
 570 }
 571
 572 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
 573                              struct ibmvnic_tx_pool *tx_pool)
 574 {
 575         int rc, i;
 576
 577         rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
 578         if (rc)
 579                 return rc;
 580
 581         memset(tx_pool->tx_buff, 0,
 582                tx_pool->num_buffers *
 583                sizeof(struct ibmvnic_tx_buff));
 584
 585         for (i = 0; i < tx_pool->num_buffers; i++)
 586                 tx_pool->free_map[i] = i;
 587
 588         tx_pool->consumer_index = 0;
 589         tx_pool->producer_index = 0;
 590
 591         return 0;
 592 }
 593
 594 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
 595 {
 596         int tx_scrqs;
 597         int i, rc;
 598
 599         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 600         for (i = 0; i < tx_scrqs; i++) {
 601                 rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
 602                 if (rc)
 603                         return rc;
 604                 rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
 605                 if (rc)
 606                         return rc;
 607         }
 608
 609         return 0;
 610 }
 611
 612 static void release_vpd_data(struct ibmvnic_adapter *adapter)
 613 {
 614         if (!adapter->vpd)
 615                 return;
 616
 617         kfree(adapter->vpd->buff);
 618         kfree(adapter->vpd);
 619
 620         adapter->vpd = NULL;
 621 }
 622
 623 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
 624                                 struct ibmvnic_tx_pool *tx_pool)
 625 {
 626         kfree(tx_pool->tx_buff);
 627         kfree(tx_pool->free_map);
 628         free_long_term_buff(adapter, &tx_pool->long_term_buff);
 629 }
 630
 631 static void release_tx_pools(struct ibmvnic_adapter *adapter)
 632 {
 633         int i;
 634
 635         if (!adapter->tx_pool)
 636                 return;
 637
 638         for (i = 0; i < adapter->num_active_tx_pools; i++) {
 639                 release_one_tx_pool(adapter, &adapter->tx_pool[i]);
 640                 release_one_tx_pool(adapter, &adapter->tso_pool[i]);
 641         }
 642
 643         kfree(adapter->tx_pool);
 644         adapter->tx_pool = NULL;
 645         kfree(adapter->tso_pool);
 646         adapter->tso_pool = NULL;
 647         adapter->num_active_tx_pools = 0;
 648 }
 649
 650 static int init_one_tx_pool(struct net_device *netdev,
 651                             struct ibmvnic_tx_pool *tx_pool,
 652                             int num_entries, int buf_size)
 653 {
 654         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 655         int i;
 656
 657         tx_pool->tx_buff = kcalloc(num_entries,
 658                                    sizeof(struct ibmvnic_tx_buff),
 659                                    GFP_KERNEL);
 660         if (!tx_pool->tx_buff)
 661                 return -1;
 662
 663         if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
 664                                  num_entries * buf_size))
 665                 return -1;
 666
 667         tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
 668         if (!tx_pool->free_map)
 669                 return -1;
 670
 671         for (i = 0; i < num_entries; i++)
 672                 tx_pool->free_map[i] = i;
 673
 674         tx_pool->consumer_index = 0;
 675         tx_pool->producer_index = 0;
 676         tx_pool->num_buffers = num_entries;
 677         tx_pool->buf_size = buf_size;
 678
 679         return 0;
 680 }
 681
 682 static int init_tx_pools(struct net_device *netdev)
 683 {
 684         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 685         int tx_subcrqs;
 686         int i, rc;
 687
 688         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 689         adapter->tx_pool = kcalloc(tx_subcrqs,
 690                                    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 691         if (!adapter->tx_pool)
 692                 return -1;
 693
 694         adapter->tso_pool = kcalloc(tx_subcrqs,
 695                                     sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 696         if (!adapter->tso_pool)
 697                 return -1;
 698
 699         adapter->num_active_tx_pools = tx_subcrqs;
 700
 701         for (i = 0; i < tx_subcrqs; i++) {
 702                 rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
 703                                       adapter->req_tx_entries_per_subcrq,
 704                                       adapter->req_mtu + VLAN_HLEN);
 705                 if (rc) {
 706                         release_tx_pools(adapter);
 707                         return rc;
 708                 }
 709
 710                 rc = init_one_tx_pool(netdev, &adapter->tso_pool[i],
 711                                       IBMVNIC_TSO_BUFS,
 712                                       IBMVNIC_TSO_BUF_SZ);
 713                 if (rc) {
 714                         release_tx_pools(adapter);
 715                         return rc;
 716                 }
 717         }
 718
 719         return 0;
 720 }
 721
 722 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
 723 {
 724         int i;
 725
 726         if (adapter->napi_enabled)
 727                 return;
 728
 729         for (i = 0; i < adapter->req_rx_queues; i++)
 730                 napi_enable(&adapter->napi[i]);
 731
 732         adapter->napi_enabled = true;
 733 }
 734
 735 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
 736 {
 737         int i;
 738
 739         if (!adapter->napi_enabled)
 740                 return;
 741
 742         for (i = 0; i < adapter->req_rx_queues; i++) {
 743                 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
 744                 napi_disable(&adapter->napi[i]);
 745         }
 746
 747         adapter->napi_enabled = false;
 748 }
 749
 750 static int init_napi(struct ibmvnic_adapter *adapter)
 751 {
 752         int i;
 753
 754         adapter->napi = kcalloc(adapter->req_rx_queues,
 755                                 sizeof(struct napi_struct), GFP_KERNEL);
 756         if (!adapter->napi)
 757                 return -ENOMEM;
 758
 759         for (i = 0; i < adapter->req_rx_queues; i++) {
 760                 netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
 761                 netif_napi_add(adapter->netdev, &adapter->napi[i],
 762                                ibmvnic_poll, NAPI_POLL_WEIGHT);
 763         }
 764
 765         adapter->num_active_rx_napi = adapter->req_rx_queues;
 766         return 0;
 767 }
 768
 769 static void release_napi(struct ibmvnic_adapter *adapter)
 770 {
 771         int i;
 772
 773         if (!adapter->napi)
 774                 return;
 775
 776         for (i = 0; i < adapter->num_active_rx_napi; i++) {
 777                 if (&adapter->napi[i]) {
 778                         netdev_dbg(adapter->netdev,
 779                                    "Releasing napi[%d]\n", i);
 780                         netif_napi_del(&adapter->napi[i]);
 781                 }
 782         }
 783
 784         kfree(adapter->napi);
 785         adapter->napi = NULL;
 786         adapter->num_active_rx_napi = 0;
 787         adapter->napi_enabled = false;
 788 }
 789
 790 static int ibmvnic_login(struct net_device *netdev)
 791 {
 792         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 793         unsigned long timeout = msecs_to_jiffies(30000);
 794         int retry_count = 0;
 795         bool retry;
 796         int rc;
 797
 798         do {
 799                 retry = false;
 800                 if (retry_count > IBMVNIC_MAX_QUEUES) {
 801                         netdev_warn(netdev, "Login attempts exceeded\n");
 802                         return -1;
 803                 }
 804
 805                 adapter->init_done_rc = 0;
 806                 reinit_completion(&adapter->init_done);
 807                 rc = send_login(adapter);
 808                 if (rc) {
 809                         netdev_warn(netdev, "Unable to login\n");
 810                         return rc;
 811                 }
 812
 813                 if (!wait_for_completion_timeout(&adapter->init_done,
 814                                                  timeout)) {
 815                         netdev_warn(netdev, "Login timed out\n");
 816                         return -1;
 817                 }
 818
 819                 if (adapter->init_done_rc == PARTIALSUCCESS) {
 820                         retry_count++;
 821                         release_sub_crqs(adapter, 1);
 822
 823                         retry = true;
 824                         netdev_dbg(netdev,
 825                                    "Received partial success, retrying...\n");
 826                         adapter->init_done_rc = 0;
 827                         reinit_completion(&adapter->init_done);
 828                         send_cap_queries(adapter);
 829                         if (!wait_for_completion_timeout(&adapter->init_done,
 830                                                          timeout)) {
 831                                 netdev_warn(netdev,
 832                                             "Capabilities query timed out\n");
 833                                 return -1;
 834                         }
 835
 836                         rc = init_sub_crqs(adapter);
 837                         if (rc) {
 838                                 netdev_warn(netdev,
 839                                             "SCRQ initialization failed\n");
 840                                 return -1;
 841                         }
 842
 843                         rc = init_sub_crq_irqs(adapter);
 844                         if (rc) {
 845                                 netdev_warn(netdev,
 846                                             "SCRQ irq initialization failed\n");
 847                                 return -1;
 848                         }
 849                 } else if (adapter->init_done_rc) {
 850                         netdev_warn(netdev, "Adapter login failed\n");
 851                         return -1;
 852                 }
 853         } while (retry);
 854
 855         /* handle pending MAC address changes after successful login */
 856         if (adapter->mac_change_pending) {
 857                 __ibmvnic_set_mac(netdev, &adapter->desired.mac);
 858                 adapter->mac_change_pending = false;
 859         }
 860
 861         return 0;
 862 }
 863
 864 static void release_login_buffer(struct ibmvnic_adapter *adapter)
 865 {
 866         kfree(adapter->login_buf);
 867         adapter->login_buf = NULL;
 868 }
 869
 870 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
 871 {
 872         kfree(adapter->login_rsp_buf);
 873         adapter->login_rsp_buf = NULL;
 874 }
 875
 876 static void release_resources(struct ibmvnic_adapter *adapter)
 877 {
 878         release_vpd_data(adapter);
 879
 880         release_tx_pools(adapter);
 881         release_rx_pools(adapter);
 882
 883         release_napi(adapter);
 884         release_login_rsp_buffer(adapter);
 885 }
 886
 887 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
 888 {
 889         struct net_device *netdev = adapter->netdev;
 890         unsigned long timeout = msecs_to_jiffies(30000);
 891         union ibmvnic_crq crq;
 892         bool resend;
 893         int rc;
 894
 895         netdev_dbg(netdev, "setting link state %d\n", link_state);
 896
 897         memset(&crq, 0, sizeof(crq));
 898         crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
 899         crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
 900         crq.logical_link_state.link_state = link_state;
 901
 902         do {
 903                 resend = false;
 904
 905                 reinit_completion(&adapter->init_done);
 906                 rc = ibmvnic_send_crq(adapter, &crq);
 907                 if (rc) {
 908                         netdev_err(netdev, "Failed to set link state\n");
 909                         return rc;
 910                 }
 911
 912                 if (!wait_for_completion_timeout(&adapter->init_done,
 913                                                  timeout)) {
 914                         netdev_err(netdev, "timeout setting link state\n");
 915                         return -1;
 916                 }
 917
 918                 if (adapter->init_done_rc == 1) {
 919                         /* Partuial success, delay and re-send */
 920                         mdelay(1000);
 921                         resend = true;
 922                 } else if (adapter->init_done_rc) {
 923                         netdev_warn(netdev, "Unable to set link state, rc=%d\n",
 924                                     adapter->init_done_rc);
 925                         return adapter->init_done_rc;
 926                 }
 927         } while (resend);
 928
 929         return 0;
 930 }
 931
 932 static int set_real_num_queues(struct net_device *netdev)
 933 {
 934         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 935         int rc;
 936
 937         netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
 938                    adapter->req_tx_queues, adapter->req_rx_queues);
 939
 940         rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
 941         if (rc) {
 942                 netdev_err(netdev, "failed to set the number of tx queues\n");
 943                 return rc;
 944         }
 945
 946         rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
 947         if (rc)
 948                 netdev_err(netdev, "failed to set the number of rx queues\n");
 949
 950         return rc;
 951 }
 952
 953 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
 954 {
 955         struct device *dev = &adapter->vdev->dev;
 956         union ibmvnic_crq crq;
 957         int len = 0;
 958         int rc;
 959
 960         if (adapter->vpd->buff)
 961                 len = adapter->vpd->len;
 962
 963         init_completion(&adapter->fw_done);
 964         crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
 965         crq.get_vpd_size.cmd = GET_VPD_SIZE;
 966         rc = ibmvnic_send_crq(adapter, &crq);
 967         if (rc)
 968                 return rc;
 969         wait_for_completion(&adapter->fw_done);
 970
 971         if (!adapter->vpd->len)
 972                 return -ENODATA;
 973
 974         if (!adapter->vpd->buff)
 975                 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
 976         else if (adapter->vpd->len != len)
 977                 adapter->vpd->buff =
 978                         krealloc(adapter->vpd->buff,
 979                                  adapter->vpd->len, GFP_KERNEL);
 980
 981         if (!adapter->vpd->buff) {
 982                 dev_err(dev, "Could allocate VPD buffer\n");
 983                 return -ENOMEM;
 984         }
 985
 986         adapter->vpd->dma_addr =
 987                 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
 988                                DMA_FROM_DEVICE);
 989         if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
 990                 dev_err(dev, "Could not map VPD buffer\n");
 991                 kfree(adapter->vpd->buff);
 992                 adapter->vpd->buff = NULL;
 993                 return -ENOMEM;
 994         }
 995
 996         reinit_completion(&adapter->fw_done);
 997         crq.get_vpd.first = IBMVNIC_CRQ_CMD;
 998         crq.get_vpd.cmd = GET_VPD;
 999         crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
1000         crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1001         rc = ibmvnic_send_crq(adapter, &crq);
1002         if (rc) {
1003                 kfree(adapter->vpd->buff);
1004                 adapter->vpd->buff = NULL;
1005                 return rc;
1006         }
1007         wait_for_completion(&adapter->fw_done);
1008
1009         return 0;
1010 }
1011
1012 static int init_resources(struct ibmvnic_adapter *adapter)
1013 {
1014         struct net_device *netdev = adapter->netdev;
1015         int rc;
1016
1017         rc = set_real_num_queues(netdev);
1018         if (rc)
1019                 return rc;
1020
1021         adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1022         if (!adapter->vpd)
1023                 return -ENOMEM;
1024
1025         /* Vital Product Data (VPD) */
1026         rc = ibmvnic_get_vpd(adapter);
1027         if (rc) {
1028                 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1029                 return rc;
1030         }
1031
1032         adapter->map_id = 1;
1033
1034         rc = init_napi(adapter);
1035         if (rc)
1036                 return rc;
1037
1038         send_map_query(adapter);
1039
1040         rc = init_rx_pools(netdev);
1041         if (rc)
1042                 return rc;
1043
1044         rc = init_tx_pools(netdev);
1045         return rc;
1046 }
1047
1048 static int __ibmvnic_open(struct net_device *netdev)
1049 {
1050         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1051         enum vnic_state prev_state = adapter->state;
1052         int i, rc;
1053
1054         adapter->state = VNIC_OPENING;
1055         replenish_pools(adapter);
1056         ibmvnic_napi_enable(adapter);
1057
1058         /* We're ready to receive frames, enable the sub-crq interrupts and
1059          * set the logical link state to up
1060          */
1061         for (i = 0; i < adapter->req_rx_queues; i++) {
1062                 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1063                 if (prev_state == VNIC_CLOSED)
1064                         enable_irq(adapter->rx_scrq[i]->irq);
1065                 enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1066         }
1067
1068         for (i = 0; i < adapter->req_tx_queues; i++) {
1069                 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1070                 if (prev_state == VNIC_CLOSED)
1071                         enable_irq(adapter->tx_scrq[i]->irq);
1072                 enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1073         }
1074
1075         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1076         if (rc) {
1077                 for (i = 0; i < adapter->req_rx_queues; i++)
1078                         napi_disable(&adapter->napi[i]);
1079                 release_resources(adapter);
1080                 return rc;
1081         }
1082
1083         netif_tx_start_all_queues(netdev);
1084
1085         if (prev_state == VNIC_CLOSED) {
1086                 for (i = 0; i < adapter->req_rx_queues; i++)
1087                         napi_schedule(&adapter->napi[i]);
1088         }
1089
1090         adapter->state = VNIC_OPEN;
1091         return rc;
1092 }
1093
1094 static int ibmvnic_open(struct net_device *netdev)
1095 {
1096         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1097         int rc;
1098
1099         /* If device failover is pending, just set device state and return.
1100          * Device operation will be handled by reset routine.
1101          */
1102         if (adapter->failover_pending) {
1103                 adapter->state = VNIC_OPEN;
1104                 return 0;
1105         }
1106
1107         if (adapter->state != VNIC_CLOSED) {
1108                 rc = ibmvnic_login(netdev);
1109                 if (rc)
1110                         return rc;
1111
1112                 rc = init_resources(adapter);
1113                 if (rc) {
1114                         netdev_err(netdev, "failed to initialize resources\n");
1115                         release_resources(adapter);
1116                         return rc;
1117                 }
1118         }
1119
1120         rc = __ibmvnic_open(netdev);
1121         netif_carrier_on(netdev);
1122
1123         return rc;
1124 }
1125
1126 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1127 {
1128         struct ibmvnic_rx_pool *rx_pool;
1129         struct ibmvnic_rx_buff *rx_buff;
1130         u64 rx_entries;
1131         int rx_scrqs;
1132         int i, j;
1133
1134         if (!adapter->rx_pool)
1135                 return;
1136
1137         rx_scrqs = adapter->num_active_rx_pools;
1138         rx_entries = adapter->req_rx_add_entries_per_subcrq;
1139
1140         /* Free any remaining skbs in the rx buffer pools */
1141         for (i = 0; i < rx_scrqs; i++) {
1142                 rx_pool = &adapter->rx_pool[i];
1143                 if (!rx_pool || !rx_pool->rx_buff)
1144                         continue;
1145
1146                 netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1147                 for (j = 0; j < rx_entries; j++) {
1148                         rx_buff = &rx_pool->rx_buff[j];
1149                         if (rx_buff && rx_buff->skb) {
1150                                 dev_kfree_skb_any(rx_buff->skb);
1151                                 rx_buff->skb = NULL;
1152                         }
1153                 }
1154         }
1155 }
1156
1157 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1158                               struct ibmvnic_tx_pool *tx_pool)
1159 {
1160         struct ibmvnic_tx_buff *tx_buff;
1161         u64 tx_entries;
1162         int i;
1163
1164         if (!tx_pool || !tx_pool->tx_buff)
1165                 return;
1166
1167         tx_entries = tx_pool->num_buffers;
1168
1169         for (i = 0; i < tx_entries; i++) {
1170                 tx_buff = &tx_pool->tx_buff[i];
1171                 if (tx_buff && tx_buff->skb) {
1172                         dev_kfree_skb_any(tx_buff->skb);
1173                         tx_buff->skb = NULL;
1174                 }
1175         }
1176 }
1177
1178 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1179 {
1180         int tx_scrqs;
1181         int i;
1182
1183         if (!adapter->tx_pool || !adapter->tso_pool)
1184                 return;
1185
1186         tx_scrqs = adapter->num_active_tx_pools;
1187
1188         /* Free any remaining skbs in the tx buffer pools */
1189         for (i = 0; i < tx_scrqs; i++) {
1190                 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1191                 clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1192                 clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1193         }
1194 }
1195
1196 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1197 {
1198         struct net_device *netdev = adapter->netdev;
1199         int i;
1200
1201         if (adapter->tx_scrq) {
1202                 for (i = 0; i < adapter->req_tx_queues; i++)
1203                         if (adapter->tx_scrq[i]->irq) {
1204                                 netdev_dbg(netdev,
1205                                            "Disabling tx_scrq[%d] irq\n", i);
1206                                 disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1207                                 disable_irq(adapter->tx_scrq[i]->irq);
1208                         }
1209         }
1210
1211         if (adapter->rx_scrq) {
1212                 for (i = 0; i < adapter->req_rx_queues; i++) {
1213                         if (adapter->rx_scrq[i]->irq) {
1214                                 netdev_dbg(netdev,
1215                                            "Disabling rx_scrq[%d] irq\n", i);
1216                                 disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1217                                 disable_irq(adapter->rx_scrq[i]->irq);
1218                         }
1219                 }
1220         }
1221 }
1222
1223 static void ibmvnic_cleanup(struct net_device *netdev)
1224 {
1225         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1226
1227         /* ensure that transmissions are stopped if called by do_reset */
1228         if (adapter->resetting)
1229                 netif_tx_disable(netdev);
1230         else
1231                 netif_tx_stop_all_queues(netdev);
1232
1233         ibmvnic_napi_disable(adapter);
1234         ibmvnic_disable_irqs(adapter);
1235
1236         clean_rx_pools(adapter);
1237         clean_tx_pools(adapter);
1238 }
1239
1240 static int __ibmvnic_close(struct net_device *netdev)
1241 {
1242         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1243         int rc = 0;
1244
1245         adapter->state = VNIC_CLOSING;
1246         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1247         if (rc)
1248                 return rc;
1249         adapter->state = VNIC_CLOSED;
1250         return 0;
1251 }
1252
1253 static int ibmvnic_close(struct net_device *netdev)
1254 {
1255         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1256         int rc;
1257
1258         /* If device failover is pending, just set device state and return.
1259          * Device operation will be handled by reset routine.
1260          */
1261         if (adapter->failover_pending) {
1262                 adapter->state = VNIC_CLOSED;
1263                 return 0;
1264         }
1265
1266         rc = __ibmvnic_close(netdev);
1267         ibmvnic_cleanup(netdev);
1268
1269         return rc;
1270 }
1271
1272 /**
1273  * build_hdr_data - creates L2/L3/L4 header data buffer
1274  * @hdr_field - bitfield determining needed headers
1275  * @skb - socket buffer
1276  * @hdr_len - array of header lengths
1277  * @tot_len - total length of data
1278  *
1279  * Reads hdr_field to determine which headers are needed by firmware.
1280  * Builds a buffer containing these headers.  Saves individual header
1281  * lengths and total buffer length to be used to build descriptors.
1282  */
1283 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1284                           int *hdr_len, u8 *hdr_data)
1285 {
1286         int len = 0;
1287         u8 *hdr;
1288
1289         if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1290                 hdr_len[0] = sizeof(struct vlan_ethhdr);
1291         else
1292                 hdr_len[0] = sizeof(struct ethhdr);
1293
1294         if (skb->protocol == htons(ETH_P_IP)) {
1295                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1296                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1297                         hdr_len[2] = tcp_hdrlen(skb);
1298                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1299                         hdr_len[2] = sizeof(struct udphdr);
1300         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1301                 hdr_len[1] = sizeof(struct ipv6hdr);
1302                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1303                         hdr_len[2] = tcp_hdrlen(skb);
1304                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1305                         hdr_len[2] = sizeof(struct udphdr);
1306         } else if (skb->protocol == htons(ETH_P_ARP)) {
1307                 hdr_len[1] = arp_hdr_len(skb->dev);
1308                 hdr_len[2] = 0;
1309         }
1310
1311         memset(hdr_data, 0, 120);
1312         if ((hdr_field >> 6) & 1) {
1313                 hdr = skb_mac_header(skb);
1314                 memcpy(hdr_data, hdr, hdr_len[0]);
1315                 len += hdr_len[0];
1316         }
1317
1318         if ((hdr_field >> 5) & 1) {
1319                 hdr = skb_network_header(skb);
1320                 memcpy(hdr_data + len, hdr, hdr_len[1]);
1321                 len += hdr_len[1];
1322         }
1323
1324         if ((hdr_field >> 4) & 1) {
1325                 hdr = skb_transport_header(skb);
1326                 memcpy(hdr_data + len, hdr, hdr_len[2]);
1327                 len += hdr_len[2];
1328         }
1329         return len;
1330 }
1331
1332 /**
1333  * create_hdr_descs - create header and header extension descriptors
1334  * @hdr_field - bitfield determining needed headers
1335  * @data - buffer containing header data
1336  * @len - length of data buffer
1337  * @hdr_len - array of individual header lengths
1338  * @scrq_arr - descriptor array
1339  *
1340  * Creates header and, if needed, header extension descriptors and
1341  * places them in a descriptor array, scrq_arr
1342  */
1343
1344 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1345                             union sub_crq *scrq_arr)
1346 {
1347         union sub_crq hdr_desc;
1348         int tmp_len = len;
1349         int num_descs = 0;
1350         u8 *data, *cur;
1351         int tmp;
1352
1353         while (tmp_len > 0) {
1354                 cur = hdr_data + len - tmp_len;
1355
1356                 memset(&hdr_desc, 0, sizeof(hdr_desc));
1357                 if (cur != hdr_data) {
1358                         data = hdr_desc.hdr_ext.data;
1359                         tmp = tmp_len > 29 ? 29 : tmp_len;
1360                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1361                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1362                         hdr_desc.hdr_ext.len = tmp;
1363                 } else {
1364                         data = hdr_desc.hdr.data;
1365                         tmp = tmp_len > 24 ? 24 : tmp_len;
1366                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1367                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1368                         hdr_desc.hdr.len = tmp;
1369                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1370                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1371                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1372                         hdr_desc.hdr.flag = hdr_field << 1;
1373                 }
1374                 memcpy(data, cur, tmp);
1375                 tmp_len -= tmp;
1376                 *scrq_arr = hdr_desc;
1377                 scrq_arr++;
1378                 num_descs++;
1379         }
1380
1381         return num_descs;
1382 }
1383
1384 /**
1385  * build_hdr_descs_arr - build a header descriptor array
1386  * @skb - socket buffer
1387  * @num_entries - number of descriptors to be sent
1388  * @subcrq - first TX descriptor
1389  * @hdr_field - bit field determining which headers will be sent
1390  *
1391  * This function will build a TX descriptor array with applicable
1392  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1393  */
1394
1395 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1396                                 int *num_entries, u8 hdr_field)
1397 {
1398         int hdr_len[3] = {0, 0, 0};
1399         int tot_len;
1400         u8 *hdr_data = txbuff->hdr_data;
1401
1402         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1403                                  txbuff->hdr_data);
1404         *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1405                          txbuff->indir_arr + 1);
1406 }
1407
1408 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1409                                     struct net_device *netdev)
1410 {
1411         /* For some backing devices, mishandling of small packets
1412          * can result in a loss of connection or TX stall. Device
1413          * architects recommend that no packet should be smaller
1414          * than the minimum MTU value provided to the driver, so
1415          * pad any packets to that length
1416          */
1417         if (skb->len < netdev->min_mtu)
1418                 return skb_put_padto(skb, netdev->min_mtu);
1419
1420         return 0;
1421 }
1422
1423 static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1424 {
1425         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1426         int queue_num = skb_get_queue_mapping(skb);
1427         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1428         struct device *dev = &adapter->vdev->dev;
1429         struct ibmvnic_tx_buff *tx_buff = NULL;
1430         struct ibmvnic_sub_crq_queue *tx_scrq;
1431         struct ibmvnic_tx_pool *tx_pool;
1432         unsigned int tx_send_failed = 0;
1433         unsigned int tx_map_failed = 0;
1434         unsigned int tx_dropped = 0;
1435         unsigned int tx_packets = 0;
1436         unsigned int tx_bytes = 0;
1437         dma_addr_t data_dma_addr;
1438         struct netdev_queue *txq;
1439         unsigned long lpar_rc;
1440         union sub_crq tx_crq;
1441         unsigned int offset;
1442         int num_entries = 1;
1443         unsigned char *dst;
1444         u64 *handle_array;
1445         int index = 0;
1446         u8 proto = 0;
1447         int ret = 0;
1448
1449         if (adapter->resetting) {
1450                 if (!netif_subqueue_stopped(netdev, skb))
1451                         netif_stop_subqueue(netdev, queue_num);
1452                 dev_kfree_skb_any(skb);
1453
1454                 tx_send_failed++;
1455                 tx_dropped++;
1456                 ret = NETDEV_TX_OK;
1457                 goto out;
1458         }
1459
1460         if (ibmvnic_xmit_workarounds(skb, netdev)) {
1461                 tx_dropped++;
1462                 tx_send_failed++;
1463                 ret = NETDEV_TX_OK;
1464                 goto out;
1465         }
1466         if (skb_is_gso(skb))
1467                 tx_pool = &adapter->tso_pool[queue_num];
1468         else
1469                 tx_pool = &adapter->tx_pool[queue_num];
1470
1471         tx_scrq = adapter->tx_scrq[queue_num];
1472         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1473         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1474                 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1475
1476         index = tx_pool->free_map[tx_pool->consumer_index];
1477
1478         if (index == IBMVNIC_INVALID_MAP) {
1479                 dev_kfree_skb_any(skb);
1480                 tx_send_failed++;
1481                 tx_dropped++;
1482                 ret = NETDEV_TX_OK;
1483                 goto out;
1484         }
1485
1486         tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1487
1488         offset = index * tx_pool->buf_size;
1489         dst = tx_pool->long_term_buff.buff + offset;
1490         memset(dst, 0, tx_pool->buf_size);
1491         data_dma_addr = tx_pool->long_term_buff.addr + offset;
1492
1493         if (skb_shinfo(skb)->nr_frags) {
1494                 int cur, i;
1495
1496                 /* Copy the head */
1497                 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1498                 cur = skb_headlen(skb);
1499
1500                 /* Copy the frags */
1501                 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1502                         const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1503
1504                         memcpy(dst + cur,
1505                                page_address(skb_frag_page(frag)) +
1506                                frag->page_offset, skb_frag_size(frag));
1507                         cur += skb_frag_size(frag);
1508                 }
1509         } else {
1510                 skb_copy_from_linear_data(skb, dst, skb->len);
1511         }
1512
1513         tx_pool->consumer_index =
1514             (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1515
1516         tx_buff = &tx_pool->tx_buff[index];
1517         tx_buff->skb = skb;
1518         tx_buff->data_dma[0] = data_dma_addr;
1519         tx_buff->data_len[0] = skb->len;
1520         tx_buff->index = index;
1521         tx_buff->pool_index = queue_num;
1522         tx_buff->last_frag = true;
1523
1524         memset(&tx_crq, 0, sizeof(tx_crq));
1525         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1526         tx_crq.v1.type = IBMVNIC_TX_DESC;
1527         tx_crq.v1.n_crq_elem = 1;
1528         tx_crq.v1.n_sge = 1;
1529         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1530
1531         if (skb_is_gso(skb))
1532                 tx_crq.v1.correlator =
1533                         cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1534         else
1535                 tx_crq.v1.correlator = cpu_to_be32(index);
1536         tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1537         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1538         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1539
1540         if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1541                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1542                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1543         }
1544
1545         if (skb->protocol == htons(ETH_P_IP)) {
1546                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1547                 proto = ip_hdr(skb)->protocol;
1548         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1549                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1550                 proto = ipv6_hdr(skb)->nexthdr;
1551         }
1552
1553         if (proto == IPPROTO_TCP)
1554                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1555         else if (proto == IPPROTO_UDP)
1556                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1557
1558         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1559                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1560                 hdrs += 2;
1561         }
1562         if (skb_is_gso(skb)) {
1563                 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1564                 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1565                 hdrs += 2;
1566         }
1567         /* determine if l2/3/4 headers are sent to firmware */
1568         if ((*hdrs >> 7) & 1) {
1569                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1570                 tx_crq.v1.n_crq_elem = num_entries;
1571                 tx_buff->num_entries = num_entries;
1572                 tx_buff->indir_arr[0] = tx_crq;
1573                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1574                                                     sizeof(tx_buff->indir_arr),
1575                                                     DMA_TO_DEVICE);
1576                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1577                         dev_kfree_skb_any(skb);
1578                         tx_buff->skb = NULL;
1579                         if (!firmware_has_feature(FW_FEATURE_CMO))
1580                                 dev_err(dev, "tx: unable to map descriptor array\n");
1581                         tx_map_failed++;
1582                         tx_dropped++;
1583                         ret = NETDEV_TX_OK;
1584                         goto tx_err_out;
1585                 }
1586                 lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1587                                                (u64)tx_buff->indir_dma,
1588                                                (u64)num_entries);
1589                 dma_unmap_single(dev, tx_buff->indir_dma,
1590                                  sizeof(tx_buff->indir_arr), DMA_TO_DEVICE);
1591         } else {
1592                 tx_buff->num_entries = num_entries;
1593                 lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1594                                       &tx_crq);
1595         }
1596         if (lpar_rc != H_SUCCESS) {
1597                 if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1598                         dev_err_ratelimited(dev, "tx: send failed\n");
1599                 dev_kfree_skb_any(skb);
1600                 tx_buff->skb = NULL;
1601
1602                 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1603                         /* Disable TX and report carrier off if queue is closed
1604                          * or pending failover.
1605                          * Firmware guarantees that a signal will be sent to the
1606                          * driver, triggering a reset or some other action.
1607                          */
1608                         netif_tx_stop_all_queues(netdev);
1609                         netif_carrier_off(netdev);
1610                 }
1611
1612                 tx_send_failed++;
1613                 tx_dropped++;
1614                 ret = NETDEV_TX_OK;
1615                 goto tx_err_out;
1616         }
1617
1618         if (atomic_add_return(num_entries, &tx_scrq->used)
1619                                         >= adapter->req_tx_entries_per_subcrq) {
1620                 netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1621                 netif_stop_subqueue(netdev, queue_num);
1622         }
1623
1624         tx_packets++;
1625         tx_bytes += skb->len;
1626         txq->trans_start = jiffies;
1627         ret = NETDEV_TX_OK;
1628         goto out;
1629
1630 tx_err_out:
1631         /* roll back consumer index and map array*/
1632         if (tx_pool->consumer_index == 0)
1633                 tx_pool->consumer_index =
1634                         tx_pool->num_buffers - 1;
1635         else
1636                 tx_pool->consumer_index--;
1637         tx_pool->free_map[tx_pool->consumer_index] = index;
1638 out:
1639         netdev->stats.tx_dropped += tx_dropped;
1640         netdev->stats.tx_bytes += tx_bytes;
1641         netdev->stats.tx_packets += tx_packets;
1642         adapter->tx_send_failed += tx_send_failed;
1643         adapter->tx_map_failed += tx_map_failed;
1644         adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1645         adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1646         adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1647
1648         return ret;
1649 }
1650
1651 static void ibmvnic_set_multi(struct net_device *netdev)
1652 {
1653         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1654         struct netdev_hw_addr *ha;
1655         union ibmvnic_crq crq;
1656
1657         memset(&crq, 0, sizeof(crq));
1658         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1659         crq.request_capability.cmd = REQUEST_CAPABILITY;
1660
1661         if (netdev->flags & IFF_PROMISC) {
1662                 if (!adapter->promisc_supported)
1663                         return;
1664         } else {
1665                 if (netdev->flags & IFF_ALLMULTI) {
1666                         /* Accept all multicast */
1667                         memset(&crq, 0, sizeof(crq));
1668                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1669                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1670                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1671                         ibmvnic_send_crq(adapter, &crq);
1672                 } else if (netdev_mc_empty(netdev)) {
1673                         /* Reject all multicast */
1674                         memset(&crq, 0, sizeof(crq));
1675                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1676                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1677                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1678                         ibmvnic_send_crq(adapter, &crq);
1679                 } else {
1680                         /* Accept one or more multicast(s) */
1681                         netdev_for_each_mc_addr(ha, netdev) {
1682                                 memset(&crq, 0, sizeof(crq));
1683                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1684                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1685                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1686                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1687                                                 ha->addr);
1688                                 ibmvnic_send_crq(adapter, &crq);
1689                         }
1690                 }
1691         }
1692 }
1693
1694 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p)
1695 {
1696         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1697         struct sockaddr *addr = p;
1698         union ibmvnic_crq crq;
1699         int rc;
1700
1701         if (!is_valid_ether_addr(addr->sa_data))
1702                 return -EADDRNOTAVAIL;
1703
1704         memset(&crq, 0, sizeof(crq));
1705         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1706         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1707         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1708
1709         init_completion(&adapter->fw_done);
1710         rc = ibmvnic_send_crq(adapter, &crq);
1711         if (rc)
1712                 return rc;
1713         wait_for_completion(&adapter->fw_done);
1714         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1715         return adapter->fw_done_rc ? -EIO : 0;
1716 }
1717
1718 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1719 {
1720         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1721         struct sockaddr *addr = p;
1722         int rc;
1723
1724         if (adapter->state == VNIC_PROBED) {
1725                 memcpy(&adapter->desired.mac, addr, sizeof(struct sockaddr));
1726                 adapter->mac_change_pending = true;
1727                 return 0;
1728         }
1729
1730         rc = __ibmvnic_set_mac(netdev, addr);
1731
1732         return rc;
1733 }
1734
1735 /**
1736  * do_reset returns zero if we are able to keep processing reset events, or
1737  * non-zero if we hit a fatal error and must halt.
1738  */
1739 static int do_reset(struct ibmvnic_adapter *adapter,
1740                     struct ibmvnic_rwi *rwi, u32 reset_state)
1741 {
1742         u64 old_num_rx_queues, old_num_tx_queues;
1743         u64 old_num_rx_slots, old_num_tx_slots;
1744         struct net_device *netdev = adapter->netdev;
1745         int i, rc;
1746
1747         netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1748                    rwi->reset_reason);
1749
1750         netif_carrier_off(netdev);
1751         adapter->reset_reason = rwi->reset_reason;
1752
1753         old_num_rx_queues = adapter->req_rx_queues;
1754         old_num_tx_queues = adapter->req_tx_queues;
1755         old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1756         old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1757
1758         ibmvnic_cleanup(netdev);
1759
1760         if (reset_state == VNIC_OPEN &&
1761             adapter->reset_reason != VNIC_RESET_MOBILITY &&
1762             adapter->reset_reason != VNIC_RESET_FAILOVER) {
1763                 rc = __ibmvnic_close(netdev);
1764                 if (rc)
1765                         return rc;
1766         }
1767
1768         if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1769             adapter->wait_for_reset) {
1770                 release_resources(adapter);
1771                 release_sub_crqs(adapter, 1);
1772                 release_crq_queue(adapter);
1773         }
1774
1775         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1776                 /* remove the closed state so when we call open it appears
1777                  * we are coming from the probed state.
1778                  */
1779                 adapter->state = VNIC_PROBED;
1780
1781                 if (adapter->wait_for_reset) {
1782                         rc = init_crq_queue(adapter);
1783                 } else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1784                         rc = ibmvnic_reenable_crq_queue(adapter);
1785                         release_sub_crqs(adapter, 1);
1786                 } else {
1787                         rc = ibmvnic_reset_crq(adapter);
1788                         if (!rc)
1789                                 rc = vio_enable_interrupts(adapter->vdev);
1790                 }
1791
1792                 if (rc) {
1793                         netdev_err(adapter->netdev,
1794                                    "Couldn't initialize crq. rc=%d\n", rc);
1795                         return rc;
1796                 }
1797
1798                 rc = ibmvnic_reset_init(adapter);
1799                 if (rc)
1800                         return IBMVNIC_INIT_FAILED;
1801
1802                 /* If the adapter was in PROBE state prior to the reset,
1803                  * exit here.
1804                  */
1805                 if (reset_state == VNIC_PROBED)
1806                         return 0;
1807
1808                 rc = ibmvnic_login(netdev);
1809                 if (rc) {
1810                         adapter->state = reset_state;
1811                         return rc;
1812                 }
1813
1814                 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1815                     adapter->wait_for_reset) {
1816                         rc = init_resources(adapter);
1817                         if (rc)
1818                                 return rc;
1819                 } else if (adapter->req_rx_queues != old_num_rx_queues ||
1820                            adapter->req_tx_queues != old_num_tx_queues ||
1821                            adapter->req_rx_add_entries_per_subcrq !=
1822                                                         old_num_rx_slots ||
1823                            adapter->req_tx_entries_per_subcrq !=
1824                                                         old_num_tx_slots) {
1825                         release_rx_pools(adapter);
1826                         release_tx_pools(adapter);
1827                         release_napi(adapter);
1828                         release_vpd_data(adapter);
1829
1830                         rc = init_resources(adapter);
1831                         if (rc)
1832                                 return rc;
1833
1834                 } else {
1835                         rc = reset_tx_pools(adapter);
1836                         if (rc)
1837                                 return rc;
1838
1839                         rc = reset_rx_pools(adapter);
1840                         if (rc)
1841                                 return rc;
1842                 }
1843                 ibmvnic_disable_irqs(adapter);
1844         }
1845         adapter->state = VNIC_CLOSED;
1846
1847         if (reset_state == VNIC_CLOSED)
1848                 return 0;
1849
1850         rc = __ibmvnic_open(netdev);
1851         if (rc) {
1852                 if (list_empty(&adapter->rwi_list))
1853                         adapter->state = VNIC_CLOSED;
1854                 else
1855                         adapter->state = reset_state;
1856
1857                 return 0;
1858         }
1859
1860         /* refresh device's multicast list */
1861         ibmvnic_set_multi(netdev);
1862
1863         /* kick napi */
1864         for (i = 0; i < adapter->req_rx_queues; i++)
1865                 napi_schedule(&adapter->napi[i]);
1866
1867         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
1868             adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
1869                 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
1870
1871         netif_carrier_on(netdev);
1872
1873         return 0;
1874 }
1875
1876 static int do_hard_reset(struct ibmvnic_adapter *adapter,
1877                          struct ibmvnic_rwi *rwi, u32 reset_state)
1878 {
1879         struct net_device *netdev = adapter->netdev;
1880         int rc;
1881
1882         netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
1883                    rwi->reset_reason);
1884
1885         netif_carrier_off(netdev);
1886         adapter->reset_reason = rwi->reset_reason;
1887
1888         ibmvnic_cleanup(netdev);
1889         release_resources(adapter);
1890         release_sub_crqs(adapter, 0);
1891         release_crq_queue(adapter);
1892
1893         /* remove the closed state so when we call open it appears
1894          * we are coming from the probed state.
1895          */
1896         adapter->state = VNIC_PROBED;
1897
1898         reinit_completion(&adapter->init_done);
1899         rc = init_crq_queue(adapter);
1900         if (rc) {
1901                 netdev_err(adapter->netdev,
1902                            "Couldn't initialize crq. rc=%d\n", rc);
1903                 return rc;
1904         }
1905
1906         rc = ibmvnic_init(adapter);
1907         if (rc)
1908                 return rc;
1909
1910         /* If the adapter was in PROBE state prior to the reset,
1911          * exit here.
1912          */
1913         if (reset_state == VNIC_PROBED)
1914                 return 0;
1915
1916         rc = ibmvnic_login(netdev);
1917         if (rc) {
1918                 adapter->state = VNIC_PROBED;
1919                 return 0;
1920         }
1921
1922         rc = init_resources(adapter);
1923         if (rc)
1924                 return rc;
1925
1926         ibmvnic_disable_irqs(adapter);
1927         adapter->state = VNIC_CLOSED;
1928
1929         if (reset_state == VNIC_CLOSED)
1930                 return 0;
1931
1932         rc = __ibmvnic_open(netdev);
1933         if (rc) {
1934                 if (list_empty(&adapter->rwi_list))
1935                         adapter->state = VNIC_CLOSED;
1936                 else
1937                         adapter->state = reset_state;
1938
1939                 return 0;
1940         }
1941
1942         netif_carrier_on(netdev);
1943
1944         return 0;
1945 }
1946
1947 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1948 {
1949         struct ibmvnic_rwi *rwi;
1950         unsigned long flags;
1951
1952         spin_lock_irqsave(&adapter->rwi_lock, flags);
1953
1954         if (!list_empty(&adapter->rwi_list)) {
1955                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1956                                        list);
1957                 list_del(&rwi->list);
1958         } else {
1959                 rwi = NULL;
1960         }
1961
1962         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
1963         return rwi;
1964 }
1965
1966 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1967 {
1968         struct ibmvnic_rwi *rwi;
1969
1970         rwi = get_next_rwi(adapter);
1971         while (rwi) {
1972                 kfree(rwi);
1973                 rwi = get_next_rwi(adapter);
1974         }
1975 }
1976
1977 static void __ibmvnic_reset(struct work_struct *work)
1978 {
1979         struct ibmvnic_rwi *rwi;
1980         struct ibmvnic_adapter *adapter;
1981         struct net_device *netdev;
1982         bool we_lock_rtnl = false;
1983         u32 reset_state;
1984         int rc = 0;
1985
1986         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1987         netdev = adapter->netdev;
1988
1989         /* netif_set_real_num_xx_queues needs to take rtnl lock here
1990          * unless wait_for_reset is set, in which case the rtnl lock
1991          * has already been taken before initializing the reset
1992          */
1993         if (!adapter->wait_for_reset) {
1994                 rtnl_lock();
1995                 we_lock_rtnl = true;
1996         }
1997         reset_state = adapter->state;
1998
1999         rwi = get_next_rwi(adapter);
2000         while (rwi) {
2001                 if (adapter->state == VNIC_REMOVING ||
2002                     adapter->state == VNIC_REMOVED) {
2003                         kfree(rwi);
2004                         rc = EBUSY;
2005                         break;
2006                 }
2007
2008                 if (adapter->force_reset_recovery) {
2009                         adapter->force_reset_recovery = false;
2010                         rc = do_hard_reset(adapter, rwi, reset_state);
2011                 } else {
2012                         rc = do_reset(adapter, rwi, reset_state);
2013                 }
2014                 kfree(rwi);
2015                 if (rc && rc != IBMVNIC_INIT_FAILED &&
2016                     !adapter->force_reset_recovery)
2017                         break;
2018
2019                 rwi = get_next_rwi(adapter);
2020         }
2021
2022         if (adapter->wait_for_reset) {
2023                 adapter->wait_for_reset = false;
2024                 adapter->reset_done_rc = rc;
2025                 complete(&adapter->reset_done);
2026         }
2027
2028         if (rc) {
2029                 netdev_dbg(adapter->netdev, "Reset failed\n");
2030                 free_all_rwi(adapter);
2031         }
2032
2033         adapter->resetting = false;
2034         if (we_lock_rtnl)
2035                 rtnl_unlock();
2036 }
2037
2038 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2039                          enum ibmvnic_reset_reason reason)
2040 {
2041         struct list_head *entry, *tmp_entry;
2042         struct ibmvnic_rwi *rwi, *tmp;
2043         struct net_device *netdev = adapter->netdev;
2044         unsigned long flags;
2045         int ret;
2046
2047         if (adapter->state == VNIC_REMOVING ||
2048             adapter->state == VNIC_REMOVED ||
2049             adapter->failover_pending) {
2050                 ret = EBUSY;
2051                 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2052                 goto err;
2053         }
2054
2055         if (adapter->state == VNIC_PROBING) {
2056                 netdev_warn(netdev, "Adapter reset during probe\n");
2057                 ret = adapter->init_done_rc = EAGAIN;
2058                 goto err;
2059         }
2060
2061         spin_lock_irqsave(&adapter->rwi_lock, flags);
2062
2063         list_for_each(entry, &adapter->rwi_list) {
2064                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
2065                 if (tmp->reset_reason == reason) {
2066                         netdev_dbg(netdev, "Skipping matching reset\n");
2067                         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2068                         ret = EBUSY;
2069                         goto err;
2070                 }
2071         }
2072
2073         rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2074         if (!rwi) {
2075                 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2076                 ibmvnic_close(netdev);
2077                 ret = ENOMEM;
2078                 goto err;
2079         }
2080         /* if we just received a transport event,
2081          * flush reset queue and process this reset
2082          */
2083         if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2084                 list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2085                         list_del(entry);
2086         }
2087         rwi->reset_reason = reason;
2088         list_add_tail(&rwi->list, &adapter->rwi_list);
2089         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2090         adapter->resetting = true;
2091         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2092         schedule_work(&adapter->ibmvnic_reset);
2093
2094         return 0;
2095 err:
2096         if (adapter->wait_for_reset)
2097                 adapter->wait_for_reset = false;
2098         return -ret;
2099 }
2100
2101 static void ibmvnic_tx_timeout(struct net_device *dev)
2102 {
2103         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2104
2105         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2106 }
2107
2108 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2109                                   struct ibmvnic_rx_buff *rx_buff)
2110 {
2111         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2112
2113         rx_buff->skb = NULL;
2114
2115         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2116         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2117
2118         atomic_dec(&pool->available);
2119 }
2120
2121 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2122 {
2123         struct net_device *netdev = napi->dev;
2124         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2125         int scrq_num = (int)(napi - adapter->napi);
2126         int frames_processed = 0;
2127
2128 restart_poll:
2129         while (frames_processed < budget) {
2130                 struct sk_buff *skb;
2131                 struct ibmvnic_rx_buff *rx_buff;
2132                 union sub_crq *next;
2133                 u32 length;
2134                 u16 offset;
2135                 u8 flags = 0;
2136
2137                 if (unlikely(adapter->resetting &&
2138                              adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2139                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2140                         napi_complete_done(napi, frames_processed);
2141                         return frames_processed;
2142                 }
2143
2144                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2145                         break;
2146                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2147                 rx_buff =
2148                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2149                                                           rx_comp.correlator);
2150                 /* do error checking */
2151                 if (next->rx_comp.rc) {
2152                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2153                                    be16_to_cpu(next->rx_comp.rc));
2154                         /* free the entry */
2155                         next->rx_comp.first = 0;
2156                         dev_kfree_skb_any(rx_buff->skb);
2157                         remove_buff_from_pool(adapter, rx_buff);
2158                         continue;
2159                 } else if (!rx_buff->skb) {
2160                         /* free the entry */
2161                         next->rx_comp.first = 0;
2162                         remove_buff_from_pool(adapter, rx_buff);
2163                         continue;
2164                 }
2165
2166                 length = be32_to_cpu(next->rx_comp.len);
2167                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2168                 flags = next->rx_comp.flags;
2169                 skb = rx_buff->skb;
2170                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2171                                         length);
2172
2173                 /* VLAN Header has been stripped by the system firmware and
2174                  * needs to be inserted by the driver
2175                  */
2176                 if (adapter->rx_vlan_header_insertion &&
2177                     (flags & IBMVNIC_VLAN_STRIPPED))
2178                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2179                                                ntohs(next->rx_comp.vlan_tci));
2180
2181                 /* free the entry */
2182                 next->rx_comp.first = 0;
2183                 remove_buff_from_pool(adapter, rx_buff);
2184
2185                 skb_put(skb, length);
2186                 skb->protocol = eth_type_trans(skb, netdev);
2187                 skb_record_rx_queue(skb, scrq_num);
2188
2189                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2190                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2191                         skb->ip_summed = CHECKSUM_UNNECESSARY;
2192                 }
2193
2194                 length = skb->len;
2195                 napi_gro_receive(napi, skb); /* send it up */
2196                 netdev->stats.rx_packets++;
2197                 netdev->stats.rx_bytes += length;
2198                 adapter->rx_stats_buffers[scrq_num].packets++;
2199                 adapter->rx_stats_buffers[scrq_num].bytes += length;
2200                 frames_processed++;
2201         }
2202
2203         if (adapter->state != VNIC_CLOSING)
2204                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2205
2206         if (frames_processed < budget) {
2207                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2208                 napi_complete_done(napi, frames_processed);
2209                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2210                     napi_reschedule(napi)) {
2211                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2212                         goto restart_poll;
2213                 }
2214         }
2215         return frames_processed;
2216 }
2217
2218 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2219 {
2220         int rc, ret;
2221
2222         adapter->fallback.mtu = adapter->req_mtu;
2223         adapter->fallback.rx_queues = adapter->req_rx_queues;
2224         adapter->fallback.tx_queues = adapter->req_tx_queues;
2225         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2226         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2227
2228         init_completion(&adapter->reset_done);
2229         adapter->wait_for_reset = true;
2230         rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2231         if (rc)
2232                 return rc;
2233         wait_for_completion(&adapter->reset_done);
2234
2235         ret = 0;
2236         if (adapter->reset_done_rc) {
2237                 ret = -EIO;
2238                 adapter->desired.mtu = adapter->fallback.mtu;
2239                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2240                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2241                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2242                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2243
2244                 init_completion(&adapter->reset_done);
2245                 adapter->wait_for_reset = true;
2246                 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2247                 if (rc)
2248                         return ret;
2249                 wait_for_completion(&adapter->reset_done);
2250         }
2251         adapter->wait_for_reset = false;
2252
2253         return ret;
2254 }
2255
2256 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2257 {
2258         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2259
2260         adapter->desired.mtu = new_mtu + ETH_HLEN;
2261
2262         return wait_for_reset(adapter);
2263 }
2264
2265 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2266                                                 struct net_device *dev,
2267                                                 netdev_features_t features)
2268 {
2269         /* Some backing hardware adapters can not
2270          * handle packets with a MSS less than 224
2271          * or with only one segment.
2272          */
2273         if (skb_is_gso(skb)) {
2274                 if (skb_shinfo(skb)->gso_size < 224 ||
2275                     skb_shinfo(skb)->gso_segs == 1)
2276                         features &= ~NETIF_F_GSO_MASK;
2277         }
2278
2279         return features;
2280 }
2281
2282 static const struct net_device_ops ibmvnic_netdev_ops = {
2283         .ndo_open               = ibmvnic_open,
2284         .ndo_stop               = ibmvnic_close,
2285         .ndo_start_xmit         = ibmvnic_xmit,
2286         .ndo_set_rx_mode        = ibmvnic_set_multi,
2287         .ndo_set_mac_address    = ibmvnic_set_mac,
2288         .ndo_validate_addr      = eth_validate_addr,
2289         .ndo_tx_timeout         = ibmvnic_tx_timeout,
2290         .ndo_change_mtu         = ibmvnic_change_mtu,
2291         .ndo_features_check     = ibmvnic_features_check,
2292 };
2293
2294 /* ethtool functions */
2295
2296 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2297                                       struct ethtool_link_ksettings *cmd)
2298 {
2299         u32 supported, advertising;
2300
2301         supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
2302                           SUPPORTED_FIBRE);
2303         advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
2304                             ADVERTISED_FIBRE);
2305         cmd->base.speed = SPEED_1000;
2306         cmd->base.duplex = DUPLEX_FULL;
2307         cmd->base.port = PORT_FIBRE;
2308         cmd->base.phy_address = 0;
2309         cmd->base.autoneg = AUTONEG_ENABLE;
2310
2311         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
2312                                                 supported);
2313         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
2314                                                 advertising);
2315
2316         return 0;
2317 }
2318
2319 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2320                                 struct ethtool_drvinfo *info)
2321 {
2322         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2323
2324         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2325         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2326         strlcpy(info->fw_version, adapter->fw_version,
2327                 sizeof(info->fw_version));
2328 }
2329
2330 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2331 {
2332         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2333
2334         return adapter->msg_enable;
2335 }
2336
2337 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2338 {
2339         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2340
2341         adapter->msg_enable = data;
2342 }
2343
2344 static u32 ibmvnic_get_link(struct net_device *netdev)
2345 {
2346         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2347
2348         /* Don't need to send a query because we request a logical link up at
2349          * init and then we wait for link state indications
2350          */
2351         return adapter->logical_link_state;
2352 }
2353
2354 static void ibmvnic_get_ringparam(struct net_device *netdev,
2355                                   struct ethtool_ringparam *ring)
2356 {
2357         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2358
2359         ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2360         ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2361         ring->rx_mini_max_pending = 0;
2362         ring->rx_jumbo_max_pending = 0;
2363         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2364         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2365         ring->rx_mini_pending = 0;
2366         ring->rx_jumbo_pending = 0;
2367 }
2368
2369 static int ibmvnic_set_ringparam(struct net_device *netdev,
2370                                  struct ethtool_ringparam *ring)
2371 {
2372         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2373
2374         if (ring->rx_pending > adapter->max_rx_add_entries_per_subcrq  ||
2375             ring->tx_pending > adapter->max_tx_entries_per_subcrq) {
2376                 netdev_err(netdev, "Invalid request.\n");
2377                 netdev_err(netdev, "Max tx buffers = %llu\n",
2378                            adapter->max_rx_add_entries_per_subcrq);
2379                 netdev_err(netdev, "Max rx buffers = %llu\n",
2380                            adapter->max_tx_entries_per_subcrq);
2381                 return -EINVAL;
2382         }
2383
2384         adapter->desired.rx_entries = ring->rx_pending;
2385         adapter->desired.tx_entries = ring->tx_pending;
2386
2387         return wait_for_reset(adapter);
2388 }
2389
2390 static void ibmvnic_get_channels(struct net_device *netdev,
2391                                  struct ethtool_channels *channels)
2392 {
2393         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2394
2395         channels->max_rx = adapter->max_rx_queues;
2396         channels->max_tx = adapter->max_tx_queues;
2397         channels->max_other = 0;
2398         channels->max_combined = 0;
2399         channels->rx_count = adapter->req_rx_queues;
2400         channels->tx_count = adapter->req_tx_queues;
2401         channels->other_count = 0;
2402         channels->combined_count = 0;
2403 }
2404
2405 static int ibmvnic_set_channels(struct net_device *netdev,
2406                                 struct ethtool_channels *channels)
2407 {
2408         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2409
2410         adapter->desired.rx_queues = channels->rx_count;
2411         adapter->desired.tx_queues = channels->tx_count;
2412
2413         return wait_for_reset(adapter);
2414 }
2415
2416 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2417 {
2418         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2419         int i;
2420
2421         if (stringset != ETH_SS_STATS)
2422                 return;
2423
2424         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
2425                 memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2426
2427         for (i = 0; i < adapter->req_tx_queues; i++) {
2428                 snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2429                 data += ETH_GSTRING_LEN;
2430
2431                 snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2432                 data += ETH_GSTRING_LEN;
2433
2434                 snprintf(data, ETH_GSTRING_LEN, "tx%d_dropped_packets", i);
2435                 data += ETH_GSTRING_LEN;
2436         }
2437
2438         for (i = 0; i < adapter->req_rx_queues; i++) {
2439                 snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2440                 data += ETH_GSTRING_LEN;
2441
2442                 snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2443                 data += ETH_GSTRING_LEN;
2444
2445                 snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2446                 data += ETH_GSTRING_LEN;
2447         }
2448 }
2449
2450 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2451 {
2452         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2453
2454         switch (sset) {
2455         case ETH_SS_STATS:
2456                 return ARRAY_SIZE(ibmvnic_stats) +
2457                        adapter->req_tx_queues * NUM_TX_STATS +
2458                        adapter->req_rx_queues * NUM_RX_STATS;
2459         default:
2460                 return -EOPNOTSUPP;
2461         }
2462 }
2463
2464 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2465                                       struct ethtool_stats *stats, u64 *data)
2466 {
2467         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2468         union ibmvnic_crq crq;
2469         int i, j;
2470         int rc;
2471
2472         memset(&crq, 0, sizeof(crq));
2473         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2474         crq.request_statistics.cmd = REQUEST_STATISTICS;
2475         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2476         crq.request_statistics.len =
2477             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2478
2479         /* Wait for data to be written */
2480         init_completion(&adapter->stats_done);
2481         rc = ibmvnic_send_crq(adapter, &crq);
2482         if (rc)
2483                 return;
2484         wait_for_completion(&adapter->stats_done);
2485
2486         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2487                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2488                                                 ibmvnic_stats[i].offset));
2489
2490         for (j = 0; j < adapter->req_tx_queues; j++) {
2491                 data[i] = adapter->tx_stats_buffers[j].packets;
2492                 i++;
2493                 data[i] = adapter->tx_stats_buffers[j].bytes;
2494                 i++;
2495                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2496                 i++;
2497         }
2498
2499         for (j = 0; j < adapter->req_rx_queues; j++) {
2500                 data[i] = adapter->rx_stats_buffers[j].packets;
2501                 i++;
2502                 data[i] = adapter->rx_stats_buffers[j].bytes;
2503                 i++;
2504                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2505                 i++;
2506         }
2507 }
2508
2509 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2510         .get_drvinfo            = ibmvnic_get_drvinfo,
2511         .get_msglevel           = ibmvnic_get_msglevel,
2512         .set_msglevel           = ibmvnic_set_msglevel,
2513         .get_link               = ibmvnic_get_link,
2514         .get_ringparam          = ibmvnic_get_ringparam,
2515         .set_ringparam          = ibmvnic_set_ringparam,
2516         .get_channels           = ibmvnic_get_channels,
2517         .set_channels           = ibmvnic_set_channels,
2518         .get_strings            = ibmvnic_get_strings,
2519         .get_sset_count         = ibmvnic_get_sset_count,
2520         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2521         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2522 };
2523
2524 /* Routines for managing CRQs/sCRQs  */
2525
2526 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2527                                    struct ibmvnic_sub_crq_queue *scrq)
2528 {
2529         int rc;
2530
2531         if (scrq->irq) {
2532                 free_irq(scrq->irq, scrq);
2533                 irq_dispose_mapping(scrq->irq);
2534                 scrq->irq = 0;
2535         }
2536
2537         memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2538         atomic_set(&scrq->used, 0);
2539         scrq->cur = 0;
2540
2541         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2542                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2543         return rc;
2544 }
2545
2546 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2547 {
2548         int i, rc;
2549
2550         for (i = 0; i < adapter->req_tx_queues; i++) {
2551                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2552                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2553                 if (rc)
2554                         return rc;
2555         }
2556
2557         for (i = 0; i < adapter->req_rx_queues; i++) {
2558                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2559                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2560                 if (rc)
2561                         return rc;
2562         }
2563
2564         return rc;
2565 }
2566
2567 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2568                                   struct ibmvnic_sub_crq_queue *scrq,
2569                                   bool do_h_free)
2570 {
2571         struct device *dev = &adapter->vdev->dev;
2572         long rc;
2573
2574         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2575
2576         if (do_h_free) {
2577                 /* Close the sub-crqs */
2578                 do {
2579                         rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2580                                                 adapter->vdev->unit_address,
2581                                                 scrq->crq_num);
2582                 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2583
2584                 if (rc) {
2585                         netdev_err(adapter->netdev,
2586                                    "Failed to release sub-CRQ %16lx, rc = %ld\n",
2587                                    scrq->crq_num, rc);
2588                 }
2589         }
2590
2591         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2592                          DMA_BIDIRECTIONAL);
2593         free_pages((unsigned long)scrq->msgs, 2);
2594         kfree(scrq);
2595 }
2596
2597 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2598                                                         *adapter)
2599 {
2600         struct device *dev = &adapter->vdev->dev;
2601         struct ibmvnic_sub_crq_queue *scrq;
2602         int rc;
2603
2604         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2605         if (!scrq)
2606                 return NULL;
2607
2608         scrq->msgs =
2609                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2610         if (!scrq->msgs) {
2611                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2612                 goto zero_page_failed;
2613         }
2614
2615         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2616                                          DMA_BIDIRECTIONAL);
2617         if (dma_mapping_error(dev, scrq->msg_token)) {
2618                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2619                 goto map_failed;
2620         }
2621
2622         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2623                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2624
2625         if (rc == H_RESOURCE)
2626                 rc = ibmvnic_reset_crq(adapter);
2627
2628         if (rc == H_CLOSED) {
2629                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2630         } else if (rc) {
2631                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2632                 goto reg_failed;
2633         }
2634
2635         scrq->adapter = adapter;
2636         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2637         spin_lock_init(&scrq->lock);
2638
2639         netdev_dbg(adapter->netdev,
2640                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2641                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2642
2643         return scrq;
2644
2645 reg_failed:
2646         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2647                          DMA_BIDIRECTIONAL);
2648 map_failed:
2649         free_pages((unsigned long)scrq->msgs, 2);
2650 zero_page_failed:
2651         kfree(scrq);
2652
2653         return NULL;
2654 }
2655
2656 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2657 {
2658         int i;
2659
2660         if (adapter->tx_scrq) {
2661                 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2662                         if (!adapter->tx_scrq[i])
2663                                 continue;
2664
2665                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2666                                    i);
2667                         if (adapter->tx_scrq[i]->irq) {
2668                                 free_irq(adapter->tx_scrq[i]->irq,
2669                                          adapter->tx_scrq[i]);
2670                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2671                                 adapter->tx_scrq[i]->irq = 0;
2672                         }
2673
2674                         release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2675                                               do_h_free);
2676                 }
2677
2678                 kfree(adapter->tx_scrq);
2679                 adapter->tx_scrq = NULL;
2680                 adapter->num_active_tx_scrqs = 0;
2681         }
2682
2683         if (adapter->rx_scrq) {
2684                 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2685                         if (!adapter->rx_scrq[i])
2686                                 continue;
2687
2688                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2689                                    i);
2690                         if (adapter->rx_scrq[i]->irq) {
2691                                 free_irq(adapter->rx_scrq[i]->irq,
2692                                          adapter->rx_scrq[i]);
2693                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2694                                 adapter->rx_scrq[i]->irq = 0;
2695                         }
2696
2697                         release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2698                                               do_h_free);
2699                 }
2700
2701                 kfree(adapter->rx_scrq);
2702                 adapter->rx_scrq = NULL;
2703                 adapter->num_active_rx_scrqs = 0;
2704         }
2705 }
2706
2707 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2708                             struct ibmvnic_sub_crq_queue *scrq)
2709 {
2710         struct device *dev = &adapter->vdev->dev;
2711         unsigned long rc;
2712
2713         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2714                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2715         if (rc)
2716                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2717                         scrq->hw_irq, rc);
2718         return rc;
2719 }
2720
2721 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2722                            struct ibmvnic_sub_crq_queue *scrq)
2723 {
2724         struct device *dev = &adapter->vdev->dev;
2725         unsigned long rc;
2726
2727         if (scrq->hw_irq > 0x100000000ULL) {
2728                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2729                 return 1;
2730         }
2731
2732         if (adapter->resetting &&
2733             adapter->reset_reason == VNIC_RESET_MOBILITY) {
2734                 u64 val = (0xff000000) | scrq->hw_irq;
2735
2736                 rc = plpar_hcall_norets(H_EOI, val);
2737                 if (rc)
2738                         dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2739                                 val, rc);
2740         }
2741
2742         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2743                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2744         if (rc)
2745                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2746                         scrq->hw_irq, rc);
2747         return rc;
2748 }
2749
2750 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2751                                struct ibmvnic_sub_crq_queue *scrq)
2752 {
2753         struct device *dev = &adapter->vdev->dev;
2754         struct ibmvnic_tx_pool *tx_pool;
2755         struct ibmvnic_tx_buff *txbuff;
2756         union sub_crq *next;
2757         int index;
2758         int i, j;
2759
2760 restart_loop:
2761         while (pending_scrq(adapter, scrq)) {
2762                 unsigned int pool = scrq->pool_index;
2763                 int num_entries = 0;
2764
2765                 next = ibmvnic_next_scrq(adapter, scrq);
2766                 for (i = 0; i < next->tx_comp.num_comps; i++) {
2767                         if (next->tx_comp.rcs[i]) {
2768                                 dev_err(dev, "tx error %x\n",
2769                                         next->tx_comp.rcs[i]);
2770                                 continue;
2771                         }
2772                         index = be32_to_cpu(next->tx_comp.correlators[i]);
2773                         if (index & IBMVNIC_TSO_POOL_MASK) {
2774                                 tx_pool = &adapter->tso_pool[pool];
2775                                 index &= ~IBMVNIC_TSO_POOL_MASK;
2776                         } else {
2777                                 tx_pool = &adapter->tx_pool[pool];
2778                         }
2779
2780                         txbuff = &tx_pool->tx_buff[index];
2781
2782                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2783                                 if (!txbuff->data_dma[j])
2784                                         continue;
2785
2786                                 txbuff->data_dma[j] = 0;
2787                         }
2788
2789                         if (txbuff->last_frag) {
2790                                 dev_kfree_skb_any(txbuff->skb);
2791                                 txbuff->skb = NULL;
2792                         }
2793
2794                         num_entries += txbuff->num_entries;
2795
2796                         tx_pool->free_map[tx_pool->producer_index] = index;
2797                         tx_pool->producer_index =
2798                                 (tx_pool->producer_index + 1) %
2799                                         tx_pool->num_buffers;
2800                 }
2801                 /* remove tx_comp scrq*/
2802                 next->tx_comp.first = 0;
2803
2804                 if (atomic_sub_return(num_entries, &scrq->used) <=
2805                     (adapter->req_tx_entries_per_subcrq / 2) &&
2806                     __netif_subqueue_stopped(adapter->netdev,
2807                                              scrq->pool_index)) {
2808                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2809                         netdev_dbg(adapter->netdev, "Started queue %d\n",
2810                                    scrq->pool_index);
2811                 }
2812         }
2813
2814         enable_scrq_irq(adapter, scrq);
2815
2816         if (pending_scrq(adapter, scrq)) {
2817                 disable_scrq_irq(adapter, scrq);
2818                 goto restart_loop;
2819         }
2820
2821         return 0;
2822 }
2823
2824 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2825 {
2826         struct ibmvnic_sub_crq_queue *scrq = instance;
2827         struct ibmvnic_adapter *adapter = scrq->adapter;
2828
2829         disable_scrq_irq(adapter, scrq);
2830         ibmvnic_complete_tx(adapter, scrq);
2831
2832         return IRQ_HANDLED;
2833 }
2834
2835 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2836 {
2837         struct ibmvnic_sub_crq_queue *scrq = instance;
2838         struct ibmvnic_adapter *adapter = scrq->adapter;
2839
2840         /* When booting a kdump kernel we can hit pending interrupts
2841          * prior to completing driver initialization.
2842          */
2843         if (unlikely(adapter->state != VNIC_OPEN))
2844                 return IRQ_NONE;
2845
2846         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2847
2848         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2849                 disable_scrq_irq(adapter, scrq);
2850                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
2851         }
2852
2853         return IRQ_HANDLED;
2854 }
2855
2856 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2857 {
2858         struct device *dev = &adapter->vdev->dev;
2859         struct ibmvnic_sub_crq_queue *scrq;
2860         int i = 0, j = 0;
2861         int rc = 0;
2862
2863         for (i = 0; i < adapter->req_tx_queues; i++) {
2864                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2865                            i);
2866                 scrq = adapter->tx_scrq[i];
2867                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2868
2869                 if (!scrq->irq) {
2870                         rc = -EINVAL;
2871                         dev_err(dev, "Error mapping irq\n");
2872                         goto req_tx_irq_failed;
2873                 }
2874
2875                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2876                                  0, "ibmvnic_tx", scrq);
2877
2878                 if (rc) {
2879                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2880                                 scrq->irq, rc);
2881                         irq_dispose_mapping(scrq->irq);
2882                         goto req_tx_irq_failed;
2883                 }
2884         }
2885
2886         for (i = 0; i < adapter->req_rx_queues; i++) {
2887                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2888                            i);
2889                 scrq = adapter->rx_scrq[i];
2890                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2891                 if (!scrq->irq) {
2892                         rc = -EINVAL;
2893                         dev_err(dev, "Error mapping irq\n");
2894                         goto req_rx_irq_failed;
2895                 }
2896                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2897                                  0, "ibmvnic_rx", scrq);
2898                 if (rc) {
2899                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2900                                 scrq->irq, rc);
2901                         irq_dispose_mapping(scrq->irq);
2902                         goto req_rx_irq_failed;
2903                 }
2904         }
2905         return rc;
2906
2907 req_rx_irq_failed:
2908         for (j = 0; j < i; j++) {
2909                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2910                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2911         }
2912         i = adapter->req_tx_queues;
2913 req_tx_irq_failed:
2914         for (j = 0; j < i; j++) {
2915                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2916                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2917         }
2918         release_sub_crqs(adapter, 1);
2919         return rc;
2920 }
2921
2922 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2923 {
2924         struct device *dev = &adapter->vdev->dev;
2925         struct ibmvnic_sub_crq_queue **allqueues;
2926         int registered_queues = 0;
2927         int total_queues;
2928         int more = 0;
2929         int i;
2930
2931         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2932
2933         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2934         if (!allqueues)
2935                 return -1;
2936
2937         for (i = 0; i < total_queues; i++) {
2938                 allqueues[i] = init_sub_crq_queue(adapter);
2939                 if (!allqueues[i]) {
2940                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2941                         break;
2942                 }
2943                 registered_queues++;
2944         }
2945
2946         /* Make sure we were able to register the minimum number of queues */
2947         if (registered_queues <
2948             adapter->min_tx_queues + adapter->min_rx_queues) {
2949                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2950                 goto tx_failed;
2951         }
2952
2953         /* Distribute the failed allocated queues*/
2954         for (i = 0; i < total_queues - registered_queues + more ; i++) {
2955                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2956                 switch (i % 3) {
2957                 case 0:
2958                         if (adapter->req_rx_queues > adapter->min_rx_queues)
2959                                 adapter->req_rx_queues--;
2960                         else
2961                                 more++;
2962                         break;
2963                 case 1:
2964                         if (adapter->req_tx_queues > adapter->min_tx_queues)
2965                                 adapter->req_tx_queues--;
2966                         else
2967                                 more++;
2968                         break;
2969                 }
2970         }
2971
2972         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
2973                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
2974         if (!adapter->tx_scrq)
2975                 goto tx_failed;
2976
2977         for (i = 0; i < adapter->req_tx_queues; i++) {
2978                 adapter->tx_scrq[i] = allqueues[i];
2979                 adapter->tx_scrq[i]->pool_index = i;
2980                 adapter->num_active_tx_scrqs++;
2981         }
2982
2983         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
2984                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
2985         if (!adapter->rx_scrq)
2986                 goto rx_failed;
2987
2988         for (i = 0; i < adapter->req_rx_queues; i++) {
2989                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
2990                 adapter->rx_scrq[i]->scrq_num = i;
2991                 adapter->num_active_rx_scrqs++;
2992         }
2993
2994         kfree(allqueues);
2995         return 0;
2996
2997 rx_failed:
2998         kfree(adapter->tx_scrq);
2999         adapter->tx_scrq = NULL;
3000 tx_failed:
3001         for (i = 0; i < registered_queues; i++)
3002                 release_sub_crq_queue(adapter, allqueues[i], 1);
3003         kfree(allqueues);
3004         return -1;
3005 }
3006
3007 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3008 {
3009         struct device *dev = &adapter->vdev->dev;
3010         union ibmvnic_crq crq;
3011         int max_entries;
3012
3013         if (!retry) {
3014                 /* Sub-CRQ entries are 32 byte long */
3015                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3016
3017                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
3018                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
3019                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3020                         return;
3021                 }
3022
3023                 if (adapter->desired.mtu)
3024                         adapter->req_mtu = adapter->desired.mtu;
3025                 else
3026                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3027
3028                 if (!adapter->desired.tx_entries)
3029                         adapter->desired.tx_entries =
3030                                         adapter->max_tx_entries_per_subcrq;
3031                 if (!adapter->desired.rx_entries)
3032                         adapter->desired.rx_entries =
3033                                         adapter->max_rx_add_entries_per_subcrq;
3034
3035                 max_entries = IBMVNIC_MAX_LTB_SIZE /
3036                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3037
3038                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3039                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3040                         adapter->desired.tx_entries = max_entries;
3041                 }
3042
3043                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3044                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3045                         adapter->desired.rx_entries = max_entries;
3046                 }
3047
3048                 if (adapter->desired.tx_entries)
3049                         adapter->req_tx_entries_per_subcrq =
3050                                         adapter->desired.tx_entries;
3051                 else
3052                         adapter->req_tx_entries_per_subcrq =
3053                                         adapter->max_tx_entries_per_subcrq;
3054
3055                 if (adapter->desired.rx_entries)
3056                         adapter->req_rx_add_entries_per_subcrq =
3057                                         adapter->desired.rx_entries;
3058                 else
3059                         adapter->req_rx_add_entries_per_subcrq =
3060                                         adapter->max_rx_add_entries_per_subcrq;
3061
3062                 if (adapter->desired.tx_queues)
3063                         adapter->req_tx_queues =
3064                                         adapter->desired.tx_queues;
3065                 else
3066                         adapter->req_tx_queues =
3067                                         adapter->opt_tx_comp_sub_queues;
3068
3069                 if (adapter->desired.rx_queues)
3070                         adapter->req_rx_queues =
3071                                         adapter->desired.rx_queues;
3072                 else
3073                         adapter->req_rx_queues =
3074                                         adapter->opt_rx_comp_queues;
3075
3076                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3077         }
3078
3079         memset(&crq, 0, sizeof(crq));
3080         crq.request_capability.first = IBMVNIC_CRQ_CMD;
3081         crq.request_capability.cmd = REQUEST_CAPABILITY;
3082
3083         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3084         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3085         atomic_inc(&adapter->running_cap_crqs);
3086         ibmvnic_send_crq(adapter, &crq);
3087
3088         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3089         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3090         atomic_inc(&adapter->running_cap_crqs);
3091         ibmvnic_send_crq(adapter, &crq);
3092
3093         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3094         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3095         atomic_inc(&adapter->running_cap_crqs);
3096         ibmvnic_send_crq(adapter, &crq);
3097
3098         crq.request_capability.capability =
3099             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3100         crq.request_capability.number =
3101             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3102         atomic_inc(&adapter->running_cap_crqs);
3103         ibmvnic_send_crq(adapter, &crq);
3104
3105         crq.request_capability.capability =
3106             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3107         crq.request_capability.number =
3108             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3109         atomic_inc(&adapter->running_cap_crqs);
3110         ibmvnic_send_crq(adapter, &crq);
3111
3112         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3113         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3114         atomic_inc(&adapter->running_cap_crqs);
3115         ibmvnic_send_crq(adapter, &crq);
3116
3117         if (adapter->netdev->flags & IFF_PROMISC) {
3118                 if (adapter->promisc_supported) {
3119                         crq.request_capability.capability =
3120                             cpu_to_be16(PROMISC_REQUESTED);
3121                         crq.request_capability.number = cpu_to_be64(1);
3122                         atomic_inc(&adapter->running_cap_crqs);
3123                         ibmvnic_send_crq(adapter, &crq);
3124                 }
3125         } else {
3126                 crq.request_capability.capability =
3127                     cpu_to_be16(PROMISC_REQUESTED);
3128                 crq.request_capability.number = cpu_to_be64(0);
3129                 atomic_inc(&adapter->running_cap_crqs);
3130                 ibmvnic_send_crq(adapter, &crq);
3131         }
3132 }
3133
3134 static int pending_scrq(struct ibmvnic_adapter *adapter,
3135                         struct ibmvnic_sub_crq_queue *scrq)
3136 {
3137         union sub_crq *entry = &scrq->msgs[scrq->cur];
3138
3139         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3140                 return 1;
3141         else
3142                 return 0;
3143 }
3144
3145 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3146                                         struct ibmvnic_sub_crq_queue *scrq)
3147 {
3148         union sub_crq *entry;
3149         unsigned long flags;
3150
3151         spin_lock_irqsave(&scrq->lock, flags);
3152         entry = &scrq->msgs[scrq->cur];
3153         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3154                 if (++scrq->cur == scrq->size)
3155                         scrq->cur = 0;
3156         } else {
3157                 entry = NULL;
3158         }
3159         spin_unlock_irqrestore(&scrq->lock, flags);
3160
3161         return entry;
3162 }
3163
3164 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3165 {
3166         struct ibmvnic_crq_queue *queue = &adapter->crq;
3167         union ibmvnic_crq *crq;
3168
3169         crq = &queue->msgs[queue->cur];
3170         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3171                 if (++queue->cur == queue->size)
3172                         queue->cur = 0;
3173         } else {
3174                 crq = NULL;
3175         }
3176
3177         return crq;
3178 }
3179
3180 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3181 {
3182         switch (rc) {
3183         case H_PARAMETER:
3184                 dev_warn_ratelimited(dev,
3185                                      "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3186                                      func, rc);
3187                 break;
3188         case H_CLOSED:
3189                 dev_warn_ratelimited(dev,
3190                                      "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3191                                      func, rc);
3192                 break;
3193         default:
3194                 dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3195                 break;
3196         }
3197 }
3198
3199 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3200                        union sub_crq *sub_crq)
3201 {
3202         unsigned int ua = adapter->vdev->unit_address;
3203         struct device *dev = &adapter->vdev->dev;
3204         u64 *u64_crq = (u64 *)sub_crq;
3205         int rc;
3206
3207         netdev_dbg(adapter->netdev,
3208                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3209                    (unsigned long int)cpu_to_be64(remote_handle),
3210                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3211                    (unsigned long int)cpu_to_be64(u64_crq[1]),
3212                    (unsigned long int)cpu_to_be64(u64_crq[2]),
3213                    (unsigned long int)cpu_to_be64(u64_crq[3]));
3214
3215         /* Make sure the hypervisor sees the complete request */
3216         mb();
3217
3218         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3219                                 cpu_to_be64(remote_handle),
3220                                 cpu_to_be64(u64_crq[0]),
3221                                 cpu_to_be64(u64_crq[1]),
3222                                 cpu_to_be64(u64_crq[2]),
3223                                 cpu_to_be64(u64_crq[3]));
3224
3225         if (rc)
3226                 print_subcrq_error(dev, rc, __func__);
3227
3228         return rc;
3229 }
3230
3231 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3232                                 u64 remote_handle, u64 ioba, u64 num_entries)
3233 {
3234         unsigned int ua = adapter->vdev->unit_address;
3235         struct device *dev = &adapter->vdev->dev;
3236         int rc;
3237
3238         /* Make sure the hypervisor sees the complete request */
3239         mb();
3240         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3241                                 cpu_to_be64(remote_handle),
3242                                 ioba, num_entries);
3243
3244         if (rc)
3245                 print_subcrq_error(dev, rc, __func__);
3246
3247         return rc;
3248 }
3249
3250 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3251                             union ibmvnic_crq *crq)
3252 {
3253         unsigned int ua = adapter->vdev->unit_address;
3254         struct device *dev = &adapter->vdev->dev;
3255         u64 *u64_crq = (u64 *)crq;
3256         int rc;
3257
3258         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3259                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3260                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3261
3262         if (!adapter->crq.active &&
3263             crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3264                 dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3265                 return -EINVAL;
3266         }
3267
3268         /* Make sure the hypervisor sees the complete request */
3269         mb();
3270
3271         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3272                                 cpu_to_be64(u64_crq[0]),
3273                                 cpu_to_be64(u64_crq[1]));
3274
3275         if (rc) {
3276                 if (rc == H_CLOSED) {
3277                         dev_warn(dev, "CRQ Queue closed\n");
3278                         if (adapter->resetting)
3279                                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3280                 }
3281
3282                 dev_warn(dev, "Send error (rc=%d)\n", rc);
3283         }
3284
3285         return rc;
3286 }
3287
3288 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3289 {
3290         union ibmvnic_crq crq;
3291
3292         memset(&crq, 0, sizeof(crq));
3293         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3294         crq.generic.cmd = IBMVNIC_CRQ_INIT;
3295         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3296
3297         return ibmvnic_send_crq(adapter, &crq);
3298 }
3299
3300 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3301 {
3302         union ibmvnic_crq crq;
3303
3304         memset(&crq, 0, sizeof(crq));
3305         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3306         crq.version_exchange.cmd = VERSION_EXCHANGE;
3307         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3308
3309         return ibmvnic_send_crq(adapter, &crq);
3310 }
3311
3312 struct vnic_login_client_data {
3313         u8      type;
3314         __be16  len;
3315         char    name[];
3316 } __packed;
3317
3318 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3319 {
3320         int len;
3321
3322         /* Calculate the amount of buffer space needed for the
3323          * vnic client data in the login buffer. There are four entries,
3324          * OS name, LPAR name, device name, and a null last entry.
3325          */
3326         len = 4 * sizeof(struct vnic_login_client_data);
3327         len += 6; /* "Linux" plus NULL */
3328         len += strlen(utsname()->nodename) + 1;
3329         len += strlen(adapter->netdev->name) + 1;
3330
3331         return len;
3332 }
3333
3334 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3335                                  struct vnic_login_client_data *vlcd)
3336 {
3337         const char *os_name = "Linux";
3338         int len;
3339
3340         /* Type 1 - LPAR OS */
3341         vlcd->type = 1;
3342         len = strlen(os_name) + 1;
3343         vlcd->len = cpu_to_be16(len);
3344         strncpy(vlcd->name, os_name, len);
3345         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3346
3347         /* Type 2 - LPAR name */
3348         vlcd->type = 2;
3349         len = strlen(utsname()->nodename) + 1;
3350         vlcd->len = cpu_to_be16(len);
3351         strncpy(vlcd->name, utsname()->nodename, len);
3352         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3353
3354         /* Type 3 - device name */
3355         vlcd->type = 3;
3356         len = strlen(adapter->netdev->name) + 1;
3357         vlcd->len = cpu_to_be16(len);
3358         strncpy(vlcd->name, adapter->netdev->name, len);
3359 }
3360
3361 static int send_login(struct ibmvnic_adapter *adapter)
3362 {
3363         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3364         struct ibmvnic_login_buffer *login_buffer;
3365         struct device *dev = &adapter->vdev->dev;
3366         dma_addr_t rsp_buffer_token;
3367         dma_addr_t buffer_token;
3368         size_t rsp_buffer_size;
3369         union ibmvnic_crq crq;
3370         size_t buffer_size;
3371         __be64 *tx_list_p;
3372         __be64 *rx_list_p;
3373         int client_data_len;
3374         struct vnic_login_client_data *vlcd;
3375         int i;
3376
3377         if (!adapter->tx_scrq || !adapter->rx_scrq) {
3378                 netdev_err(adapter->netdev,
3379                            "RX or TX queues are not allocated, device login failed\n");
3380                 return -1;
3381         }
3382
3383         release_login_rsp_buffer(adapter);
3384         client_data_len = vnic_client_data_len(adapter);
3385
3386         buffer_size =
3387             sizeof(struct ibmvnic_login_buffer) +
3388             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3389             client_data_len;
3390
3391         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3392         if (!login_buffer)
3393                 goto buf_alloc_failed;
3394
3395         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3396                                       DMA_TO_DEVICE);
3397         if (dma_mapping_error(dev, buffer_token)) {
3398                 dev_err(dev, "Couldn't map login buffer\n");
3399                 goto buf_map_failed;
3400         }
3401
3402         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3403                           sizeof(u64) * adapter->req_tx_queues +
3404                           sizeof(u64) * adapter->req_rx_queues +
3405                           sizeof(u64) * adapter->req_rx_queues +
3406                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3407
3408         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3409         if (!login_rsp_buffer)
3410                 goto buf_rsp_alloc_failed;
3411
3412         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3413                                           rsp_buffer_size, DMA_FROM_DEVICE);
3414         if (dma_mapping_error(dev, rsp_buffer_token)) {
3415                 dev_err(dev, "Couldn't map login rsp buffer\n");
3416                 goto buf_rsp_map_failed;
3417         }
3418
3419         adapter->login_buf = login_buffer;
3420         adapter->login_buf_token = buffer_token;
3421         adapter->login_buf_sz = buffer_size;
3422         adapter->login_rsp_buf = login_rsp_buffer;
3423         adapter->login_rsp_buf_token = rsp_buffer_token;
3424         adapter->login_rsp_buf_sz = rsp_buffer_size;
3425
3426         login_buffer->len = cpu_to_be32(buffer_size);
3427         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3428         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3429         login_buffer->off_txcomp_subcrqs =
3430             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3431         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3432         login_buffer->off_rxcomp_subcrqs =
3433             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3434                         sizeof(u64) * adapter->req_tx_queues);
3435         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3436         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3437
3438         tx_list_p = (__be64 *)((char *)login_buffer +
3439                                       sizeof(struct ibmvnic_login_buffer));
3440         rx_list_p = (__be64 *)((char *)login_buffer +
3441                                       sizeof(struct ibmvnic_login_buffer) +
3442                                       sizeof(u64) * adapter->req_tx_queues);
3443
3444         for (i = 0; i < adapter->req_tx_queues; i++) {
3445                 if (adapter->tx_scrq[i]) {
3446                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3447                                                    crq_num);
3448                 }
3449         }
3450
3451         for (i = 0; i < adapter->req_rx_queues; i++) {
3452                 if (adapter->rx_scrq[i]) {
3453                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3454                                                    crq_num);
3455                 }
3456         }
3457
3458         /* Insert vNIC login client data */
3459         vlcd = (struct vnic_login_client_data *)
3460                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3461         login_buffer->client_data_offset =
3462                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3463         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3464
3465         vnic_add_client_data(adapter, vlcd);
3466
3467         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3468         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3469                 netdev_dbg(adapter->netdev, "%016lx\n",
3470                            ((unsigned long int *)(adapter->login_buf))[i]);
3471         }
3472
3473         memset(&crq, 0, sizeof(crq));
3474         crq.login.first = IBMVNIC_CRQ_CMD;
3475         crq.login.cmd = LOGIN;
3476         crq.login.ioba = cpu_to_be32(buffer_token);
3477         crq.login.len = cpu_to_be32(buffer_size);
3478         ibmvnic_send_crq(adapter, &crq);
3479
3480         return 0;
3481
3482 buf_rsp_map_failed:
3483         kfree(login_rsp_buffer);
3484 buf_rsp_alloc_failed:
3485         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3486 buf_map_failed:
3487         kfree(login_buffer);
3488 buf_alloc_failed:
3489         return -1;
3490 }
3491
3492 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3493                             u32 len, u8 map_id)
3494 {
3495         union ibmvnic_crq crq;
3496
3497         memset(&crq, 0, sizeof(crq));
3498         crq.request_map.first = IBMVNIC_CRQ_CMD;
3499         crq.request_map.cmd = REQUEST_MAP;
3500         crq.request_map.map_id = map_id;
3501         crq.request_map.ioba = cpu_to_be32(addr);
3502         crq.request_map.len = cpu_to_be32(len);
3503         return ibmvnic_send_crq(adapter, &crq);
3504 }
3505
3506 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3507 {
3508         union ibmvnic_crq crq;
3509
3510         memset(&crq, 0, sizeof(crq));
3511         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3512         crq.request_unmap.cmd = REQUEST_UNMAP;
3513         crq.request_unmap.map_id = map_id;
3514         return ibmvnic_send_crq(adapter, &crq);
3515 }
3516
3517 static void send_map_query(struct ibmvnic_adapter *adapter)
3518 {
3519         union ibmvnic_crq crq;
3520
3521         memset(&crq, 0, sizeof(crq));
3522         crq.query_map.first = IBMVNIC_CRQ_CMD;
3523         crq.query_map.cmd = QUERY_MAP;
3524         ibmvnic_send_crq(adapter, &crq);
3525 }
3526
3527 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3528 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3529 {
3530         union ibmvnic_crq crq;
3531
3532         atomic_set(&adapter->running_cap_crqs, 0);
3533         memset(&crq, 0, sizeof(crq));
3534         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3535         crq.query_capability.cmd = QUERY_CAPABILITY;
3536
3537         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3538         atomic_inc(&adapter->running_cap_crqs);
3539         ibmvnic_send_crq(adapter, &crq);
3540
3541         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3542         atomic_inc(&adapter->running_cap_crqs);
3543         ibmvnic_send_crq(adapter, &crq);
3544
3545         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3546         atomic_inc(&adapter->running_cap_crqs);
3547         ibmvnic_send_crq(adapter, &crq);
3548
3549         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3550         atomic_inc(&adapter->running_cap_crqs);
3551         ibmvnic_send_crq(adapter, &crq);
3552
3553         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3554         atomic_inc(&adapter->running_cap_crqs);
3555         ibmvnic_send_crq(adapter, &crq);
3556
3557         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3558         atomic_inc(&adapter->running_cap_crqs);
3559         ibmvnic_send_crq(adapter, &crq);
3560
3561         crq.query_capability.capability =
3562             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3563         atomic_inc(&adapter->running_cap_crqs);
3564         ibmvnic_send_crq(adapter, &crq);
3565
3566         crq.query_capability.capability =
3567             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3568         atomic_inc(&adapter->running_cap_crqs);
3569         ibmvnic_send_crq(adapter, &crq);
3570
3571         crq.query_capability.capability =
3572             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3573         atomic_inc(&adapter->running_cap_crqs);
3574         ibmvnic_send_crq(adapter, &crq);
3575
3576         crq.query_capability.capability =
3577             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3578         atomic_inc(&adapter->running_cap_crqs);
3579         ibmvnic_send_crq(adapter, &crq);
3580
3581         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3582         atomic_inc(&adapter->running_cap_crqs);
3583         ibmvnic_send_crq(adapter, &crq);
3584
3585         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3586         atomic_inc(&adapter->running_cap_crqs);
3587         ibmvnic_send_crq(adapter, &crq);
3588
3589         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3590         atomic_inc(&adapter->running_cap_crqs);
3591         ibmvnic_send_crq(adapter, &crq);
3592
3593         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3594         atomic_inc(&adapter->running_cap_crqs);
3595         ibmvnic_send_crq(adapter, &crq);
3596
3597         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3598         atomic_inc(&adapter->running_cap_crqs);
3599         ibmvnic_send_crq(adapter, &crq);
3600
3601         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3602         atomic_inc(&adapter->running_cap_crqs);
3603         ibmvnic_send_crq(adapter, &crq);
3604
3605         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3606         atomic_inc(&adapter->running_cap_crqs);
3607         ibmvnic_send_crq(adapter, &crq);
3608
3609         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3610         atomic_inc(&adapter->running_cap_crqs);
3611         ibmvnic_send_crq(adapter, &crq);
3612
3613         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3614         atomic_inc(&adapter->running_cap_crqs);
3615         ibmvnic_send_crq(adapter, &crq);
3616
3617         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3618         atomic_inc(&adapter->running_cap_crqs);
3619         ibmvnic_send_crq(adapter, &crq);
3620
3621         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3622         atomic_inc(&adapter->running_cap_crqs);
3623         ibmvnic_send_crq(adapter, &crq);
3624
3625         crq.query_capability.capability =
3626                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3627         atomic_inc(&adapter->running_cap_crqs);
3628         ibmvnic_send_crq(adapter, &crq);
3629
3630         crq.query_capability.capability =
3631                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3632         atomic_inc(&adapter->running_cap_crqs);
3633         ibmvnic_send_crq(adapter, &crq);
3634
3635         crq.query_capability.capability =
3636                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3637         atomic_inc(&adapter->running_cap_crqs);
3638         ibmvnic_send_crq(adapter, &crq);
3639
3640         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3641         atomic_inc(&adapter->running_cap_crqs);
3642         ibmvnic_send_crq(adapter, &crq);
3643 }
3644
3645 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3646                                 struct ibmvnic_adapter *adapter)
3647 {
3648         struct device *dev = &adapter->vdev->dev;
3649
3650         if (crq->get_vpd_size_rsp.rc.code) {
3651                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3652                         crq->get_vpd_size_rsp.rc.code);
3653                 complete(&adapter->fw_done);
3654                 return;
3655         }
3656
3657         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3658         complete(&adapter->fw_done);
3659 }
3660
3661 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3662                            struct ibmvnic_adapter *adapter)
3663 {
3664         struct device *dev = &adapter->vdev->dev;
3665         unsigned char *substr = NULL;
3666         u8 fw_level_len = 0;
3667
3668         memset(adapter->fw_version, 0, 32);
3669
3670         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3671                          DMA_FROM_DEVICE);
3672
3673         if (crq->get_vpd_rsp.rc.code) {
3674                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3675                         crq->get_vpd_rsp.rc.code);
3676                 goto complete;
3677         }
3678
3679         /* get the position of the firmware version info
3680          * located after the ASCII 'RM' substring in the buffer
3681          */
3682         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3683         if (!substr) {
3684                 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3685                 goto complete;
3686         }
3687
3688         /* get length of firmware level ASCII substring */
3689         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3690                 fw_level_len = *(substr + 2);
3691         } else {
3692                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3693                 goto complete;
3694         }
3695
3696         /* copy firmware version string from vpd into adapter */
3697         if ((substr + 3 + fw_level_len) <
3698             (adapter->vpd->buff + adapter->vpd->len)) {
3699                 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3700         } else {
3701                 dev_info(dev, "FW substr extrapolated VPD buff\n");
3702         }
3703
3704 complete:
3705         if (adapter->fw_version[0] == '\0')
3706                 strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3707         complete(&adapter->fw_done);
3708 }
3709
3710 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3711 {
3712         struct device *dev = &adapter->vdev->dev;
3713         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3714         union ibmvnic_crq crq;
3715         int i;
3716
3717         dma_unmap_single(dev, adapter->ip_offload_tok,
3718                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3719
3720         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3721         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3722                 netdev_dbg(adapter->netdev, "%016lx\n",
3723                            ((unsigned long int *)(buf))[i]);
3724
3725         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3726         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3727         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3728                    buf->tcp_ipv4_chksum);
3729         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3730                    buf->tcp_ipv6_chksum);
3731         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3732                    buf->udp_ipv4_chksum);
3733         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3734                    buf->udp_ipv6_chksum);
3735         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3736                    buf->large_tx_ipv4);
3737         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3738                    buf->large_tx_ipv6);
3739         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3740                    buf->large_rx_ipv4);
3741         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3742                    buf->large_rx_ipv6);
3743         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3744                    buf->max_ipv4_header_size);
3745         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3746                    buf->max_ipv6_header_size);
3747         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3748                    buf->max_tcp_header_size);
3749         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3750                    buf->max_udp_header_size);
3751         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3752                    buf->max_large_tx_size);
3753         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3754                    buf->max_large_rx_size);
3755         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3756                    buf->ipv6_extension_header);
3757         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3758                    buf->tcp_pseudosum_req);
3759         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3760                    buf->num_ipv6_ext_headers);
3761         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3762                    buf->off_ipv6_ext_headers);
3763
3764         adapter->ip_offload_ctrl_tok =
3765             dma_map_single(dev, &adapter->ip_offload_ctrl,
3766                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3767
3768         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3769                 dev_err(dev, "Couldn't map ip offload control buffer\n");
3770                 return;
3771         }
3772
3773         adapter->ip_offload_ctrl.len =
3774             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3775         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3776         adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3777         adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3778         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3779         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3780         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3781         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3782         adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3783         adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3784
3785         /* large_rx disabled for now, additional features needed */
3786         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3787         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3788
3789         adapter->netdev->features = NETIF_F_SG | NETIF_F_GSO;
3790
3791         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3792                 adapter->netdev->features |= NETIF_F_IP_CSUM;
3793
3794         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3795                 adapter->netdev->features |= NETIF_F_IPV6_CSUM;
3796
3797         if ((adapter->netdev->features &
3798             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3799                 adapter->netdev->features |= NETIF_F_RXCSUM;
3800
3801         if (buf->large_tx_ipv4)
3802                 adapter->netdev->features |= NETIF_F_TSO;
3803         if (buf->large_tx_ipv6)
3804                 adapter->netdev->features |= NETIF_F_TSO6;
3805
3806         adapter->netdev->hw_features |= adapter->netdev->features;
3807
3808         memset(&crq, 0, sizeof(crq));
3809         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3810         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3811         crq.control_ip_offload.len =
3812             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3813         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3814         ibmvnic_send_crq(adapter, &crq);
3815 }
3816
3817 static const char *ibmvnic_fw_err_cause(u16 cause)
3818 {
3819         switch (cause) {
3820         case ADAPTER_PROBLEM:
3821                 return "adapter problem";
3822         case BUS_PROBLEM:
3823                 return "bus problem";
3824         case FW_PROBLEM:
3825                 return "firmware problem";
3826         case DD_PROBLEM:
3827                 return "device driver problem";
3828         case EEH_RECOVERY:
3829                 return "EEH recovery";
3830         case FW_UPDATED:
3831                 return "firmware updated";
3832         case LOW_MEMORY:
3833                 return "low Memory";
3834         default:
3835                 return "unknown";
3836         }
3837 }
3838
3839 static void handle_error_indication(union ibmvnic_crq *crq,
3840                                     struct ibmvnic_adapter *adapter)
3841 {
3842         struct device *dev = &adapter->vdev->dev;
3843         u16 cause;
3844
3845         cause = be16_to_cpu(crq->error_indication.error_cause);
3846
3847         dev_warn_ratelimited(dev,
3848                              "Firmware reports %serror, cause: %s. Starting recovery...\n",
3849                              crq->error_indication.flags
3850                                 & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3851                              ibmvnic_fw_err_cause(cause));
3852
3853         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3854                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3855         else
3856                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3857 }
3858
3859 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3860                                  struct ibmvnic_adapter *adapter)
3861 {
3862         struct net_device *netdev = adapter->netdev;
3863         struct device *dev = &adapter->vdev->dev;
3864         long rc;
3865
3866         rc = crq->change_mac_addr_rsp.rc.code;
3867         if (rc) {
3868                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3869                 goto out;
3870         }
3871         memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
3872                ETH_ALEN);
3873 out:
3874         complete(&adapter->fw_done);
3875         return rc;
3876 }
3877
3878 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3879                                    struct ibmvnic_adapter *adapter)
3880 {
3881         struct device *dev = &adapter->vdev->dev;
3882         u64 *req_value;
3883         char *name;
3884
3885         atomic_dec(&adapter->running_cap_crqs);
3886         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3887         case REQ_TX_QUEUES:
3888                 req_value = &adapter->req_tx_queues;
3889                 name = "tx";
3890                 break;
3891         case REQ_RX_QUEUES:
3892                 req_value = &adapter->req_rx_queues;
3893                 name = "rx";
3894                 break;
3895         case REQ_RX_ADD_QUEUES:
3896                 req_value = &adapter->req_rx_add_queues;
3897                 name = "rx_add";
3898                 break;
3899         case REQ_TX_ENTRIES_PER_SUBCRQ:
3900                 req_value = &adapter->req_tx_entries_per_subcrq;
3901                 name = "tx_entries_per_subcrq";
3902                 break;
3903         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3904                 req_value = &adapter->req_rx_add_entries_per_subcrq;
3905                 name = "rx_add_entries_per_subcrq";
3906                 break;
3907         case REQ_MTU:
3908                 req_value = &adapter->req_mtu;
3909                 name = "mtu";
3910                 break;
3911         case PROMISC_REQUESTED:
3912                 req_value = &adapter->promisc;
3913                 name = "promisc";
3914                 break;
3915         default:
3916                 dev_err(dev, "Got invalid cap request rsp %d\n",
3917                         crq->request_capability.capability);
3918                 return;
3919         }
3920
3921         switch (crq->request_capability_rsp.rc.code) {
3922         case SUCCESS:
3923                 break;
3924         case PARTIALSUCCESS:
3925                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3926                          *req_value,
3927                          (long int)be64_to_cpu(crq->request_capability_rsp.
3928                                                number), name);
3929
3930                 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3931                     REQ_MTU) {
3932                         pr_err("mtu of %llu is not supported. Reverting.\n",
3933                                *req_value);
3934                         *req_value = adapter->fallback.mtu;
3935                 } else {
3936                         *req_value =
3937                                 be64_to_cpu(crq->request_capability_rsp.number);
3938                 }
3939
3940                 ibmvnic_send_req_caps(adapter, 1);
3941                 return;
3942         default:
3943                 dev_err(dev, "Error %d in request cap rsp\n",
3944                         crq->request_capability_rsp.rc.code);
3945                 return;
3946         }
3947
3948         /* Done receiving requested capabilities, query IP offload support */
3949         if (atomic_read(&adapter->running_cap_crqs) == 0) {
3950                 union ibmvnic_crq newcrq;
3951                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3952                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
3953                     &adapter->ip_offload_buf;
3954
3955                 adapter->wait_capability = false;
3956                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
3957                                                          buf_sz,
3958                                                          DMA_FROM_DEVICE);
3959
3960                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3961                         if (!firmware_has_feature(FW_FEATURE_CMO))
3962                                 dev_err(dev, "Couldn't map offload buffer\n");
3963                         return;
3964                 }
3965
3966                 memset(&newcrq, 0, sizeof(newcrq));
3967                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3968                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3969                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
3970                 newcrq.query_ip_offload.ioba =
3971                     cpu_to_be32(adapter->ip_offload_tok);
3972
3973                 ibmvnic_send_crq(adapter, &newcrq);
3974         }
3975 }
3976
3977 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
3978                             struct ibmvnic_adapter *adapter)
3979 {
3980         struct device *dev = &adapter->vdev->dev;
3981         struct net_device *netdev = adapter->netdev;
3982         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
3983         struct ibmvnic_login_buffer *login = adapter->login_buf;
3984         int i;
3985
3986         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
3987                          DMA_TO_DEVICE);
3988         dma_unmap_single(dev, adapter->login_rsp_buf_token,
3989                          adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
3990
3991         /* If the number of queues requested can't be allocated by the
3992          * server, the login response will return with code 1. We will need
3993          * to resend the login buffer with fewer queues requested.
3994          */
3995         if (login_rsp_crq->generic.rc.code) {
3996                 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
3997                 complete(&adapter->init_done);
3998                 return 0;
3999         }
4000
4001         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4002
4003         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4004         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4005                 netdev_dbg(adapter->netdev, "%016lx\n",
4006                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4007         }
4008
4009         /* Sanity checks */
4010         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4011             (be32_to_cpu(login->num_rxcomp_subcrqs) *
4012              adapter->req_rx_add_queues !=
4013              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4014                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4015                 ibmvnic_remove(adapter->vdev);
4016                 return -EIO;
4017         }
4018         release_login_buffer(adapter);
4019         complete(&adapter->init_done);
4020
4021         return 0;
4022 }
4023
4024 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4025                                      struct ibmvnic_adapter *adapter)
4026 {
4027         struct device *dev = &adapter->vdev->dev;
4028         long rc;
4029
4030         rc = crq->request_unmap_rsp.rc.code;
4031         if (rc)
4032                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4033 }
4034
4035 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4036                                  struct ibmvnic_adapter *adapter)
4037 {
4038         struct net_device *netdev = adapter->netdev;
4039         struct device *dev = &adapter->vdev->dev;
4040         long rc;
4041
4042         rc = crq->query_map_rsp.rc.code;
4043         if (rc) {
4044                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4045                 return;
4046         }
4047         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4048                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4049                    crq->query_map_rsp.free_pages);
4050 }
4051
4052 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4053                                  struct ibmvnic_adapter *adapter)
4054 {
4055         struct net_device *netdev = adapter->netdev;
4056         struct device *dev = &adapter->vdev->dev;
4057         long rc;
4058
4059         atomic_dec(&adapter->running_cap_crqs);
4060         netdev_dbg(netdev, "Outstanding queries: %d\n",
4061                    atomic_read(&adapter->running_cap_crqs));
4062         rc = crq->query_capability.rc.code;
4063         if (rc) {
4064                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4065                 goto out;
4066         }
4067
4068         switch (be16_to_cpu(crq->query_capability.capability)) {
4069         case MIN_TX_QUEUES:
4070                 adapter->min_tx_queues =
4071                     be64_to_cpu(crq->query_capability.number);
4072                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4073                            adapter->min_tx_queues);
4074                 break;
4075         case MIN_RX_QUEUES:
4076                 adapter->min_rx_queues =
4077                     be64_to_cpu(crq->query_capability.number);
4078                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4079                            adapter->min_rx_queues);
4080                 break;
4081         case MIN_RX_ADD_QUEUES:
4082                 adapter->min_rx_add_queues =
4083                     be64_to_cpu(crq->query_capability.number);
4084                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4085                            adapter->min_rx_add_queues);
4086                 break;
4087         case MAX_TX_QUEUES:
4088                 adapter->max_tx_queues =
4089                     be64_to_cpu(crq->query_capability.number);
4090                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4091                            adapter->max_tx_queues);
4092                 break;
4093         case MAX_RX_QUEUES:
4094                 adapter->max_rx_queues =
4095                     be64_to_cpu(crq->query_capability.number);
4096                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4097                            adapter->max_rx_queues);
4098                 break;
4099         case MAX_RX_ADD_QUEUES:
4100                 adapter->max_rx_add_queues =
4101                     be64_to_cpu(crq->query_capability.number);
4102                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4103                            adapter->max_rx_add_queues);
4104                 break;
4105         case MIN_TX_ENTRIES_PER_SUBCRQ:
4106                 adapter->min_tx_entries_per_subcrq =
4107                     be64_to_cpu(crq->query_capability.number);
4108                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4109                            adapter->min_tx_entries_per_subcrq);
4110                 break;
4111         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4112                 adapter->min_rx_add_entries_per_subcrq =
4113                     be64_to_cpu(crq->query_capability.number);
4114                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4115                            adapter->min_rx_add_entries_per_subcrq);
4116                 break;
4117         case MAX_TX_ENTRIES_PER_SUBCRQ:
4118                 adapter->max_tx_entries_per_subcrq =
4119                     be64_to_cpu(crq->query_capability.number);
4120                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4121                            adapter->max_tx_entries_per_subcrq);
4122                 break;
4123         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4124                 adapter->max_rx_add_entries_per_subcrq =
4125                     be64_to_cpu(crq->query_capability.number);
4126                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4127                            adapter->max_rx_add_entries_per_subcrq);
4128                 break;
4129         case TCP_IP_OFFLOAD:
4130                 adapter->tcp_ip_offload =
4131                     be64_to_cpu(crq->query_capability.number);
4132                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4133                            adapter->tcp_ip_offload);
4134                 break;
4135         case PROMISC_SUPPORTED:
4136                 adapter->promisc_supported =
4137                     be64_to_cpu(crq->query_capability.number);
4138                 netdev_dbg(netdev, "promisc_supported = %lld\n",
4139                            adapter->promisc_supported);
4140                 break;
4141         case MIN_MTU:
4142                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4143                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4144                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4145                 break;
4146         case MAX_MTU:
4147                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4148                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4149                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4150                 break;
4151         case MAX_MULTICAST_FILTERS:
4152                 adapter->max_multicast_filters =
4153                     be64_to_cpu(crq->query_capability.number);
4154                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4155                            adapter->max_multicast_filters);
4156                 break;
4157         case VLAN_HEADER_INSERTION:
4158                 adapter->vlan_header_insertion =
4159                     be64_to_cpu(crq->query_capability.number);
4160                 if (adapter->vlan_header_insertion)
4161                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4162                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4163                            adapter->vlan_header_insertion);
4164                 break;
4165         case RX_VLAN_HEADER_INSERTION:
4166                 adapter->rx_vlan_header_insertion =
4167                     be64_to_cpu(crq->query_capability.number);
4168                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4169                            adapter->rx_vlan_header_insertion);
4170                 break;
4171         case MAX_TX_SG_ENTRIES:
4172                 adapter->max_tx_sg_entries =
4173                     be64_to_cpu(crq->query_capability.number);
4174                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4175                            adapter->max_tx_sg_entries);
4176                 break;
4177         case RX_SG_SUPPORTED:
4178                 adapter->rx_sg_supported =
4179                     be64_to_cpu(crq->query_capability.number);
4180                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4181                            adapter->rx_sg_supported);
4182                 break;
4183         case OPT_TX_COMP_SUB_QUEUES:
4184                 adapter->opt_tx_comp_sub_queues =
4185                     be64_to_cpu(crq->query_capability.number);
4186                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4187                            adapter->opt_tx_comp_sub_queues);
4188                 break;
4189         case OPT_RX_COMP_QUEUES:
4190                 adapter->opt_rx_comp_queues =
4191                     be64_to_cpu(crq->query_capability.number);
4192                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4193                            adapter->opt_rx_comp_queues);
4194                 break;
4195         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4196                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4197                     be64_to_cpu(crq->query_capability.number);
4198                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4199                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
4200                 break;
4201         case OPT_TX_ENTRIES_PER_SUBCRQ:
4202                 adapter->opt_tx_entries_per_subcrq =
4203                     be64_to_cpu(crq->query_capability.number);
4204                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4205                            adapter->opt_tx_entries_per_subcrq);
4206                 break;
4207         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4208                 adapter->opt_rxba_entries_per_subcrq =
4209                     be64_to_cpu(crq->query_capability.number);
4210                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4211                            adapter->opt_rxba_entries_per_subcrq);
4212                 break;
4213         case TX_RX_DESC_REQ:
4214                 adapter->tx_rx_desc_req = crq->query_capability.number;
4215                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4216                            adapter->tx_rx_desc_req);
4217                 break;
4218
4219         default:
4220                 netdev_err(netdev, "Got invalid cap rsp %d\n",
4221                            crq->query_capability.capability);
4222         }
4223
4224 out:
4225         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4226                 adapter->wait_capability = false;
4227                 ibmvnic_send_req_caps(adapter, 0);
4228         }
4229 }
4230
4231 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4232                                struct ibmvnic_adapter *adapter)
4233 {
4234         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4235         struct net_device *netdev = adapter->netdev;
4236         struct device *dev = &adapter->vdev->dev;
4237         u64 *u64_crq = (u64 *)crq;
4238         long rc;
4239
4240         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4241                    (unsigned long int)cpu_to_be64(u64_crq[0]),
4242                    (unsigned long int)cpu_to_be64(u64_crq[1]));
4243         switch (gen_crq->first) {
4244         case IBMVNIC_CRQ_INIT_RSP:
4245                 switch (gen_crq->cmd) {
4246                 case IBMVNIC_CRQ_INIT:
4247                         dev_info(dev, "Partner initialized\n");
4248                         adapter->from_passive_init = true;
4249                         adapter->failover_pending = false;
4250                         if (!completion_done(&adapter->init_done)) {
4251                                 complete(&adapter->init_done);
4252                                 adapter->init_done_rc = -EIO;
4253                         }
4254                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4255                         break;
4256                 case IBMVNIC_CRQ_INIT_COMPLETE:
4257                         dev_info(dev, "Partner initialization complete\n");
4258                         adapter->crq.active = true;
4259                         send_version_xchg(adapter);
4260                         break;
4261                 default:
4262                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4263                 }
4264                 return;
4265         case IBMVNIC_CRQ_XPORT_EVENT:
4266                 netif_carrier_off(netdev);
4267                 adapter->crq.active = false;
4268                 if (adapter->resetting)
4269                         adapter->force_reset_recovery = true;
4270                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4271                         dev_info(dev, "Migrated, re-enabling adapter\n");
4272                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4273                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4274                         dev_info(dev, "Backing device failover detected\n");
4275                         adapter->failover_pending = true;
4276                 } else {
4277                         /* The adapter lost the connection */
4278                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4279                                 gen_crq->cmd);
4280                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4281                 }
4282                 return;
4283         case IBMVNIC_CRQ_CMD_RSP:
4284                 break;
4285         default:
4286                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
4287                         gen_crq->first);
4288                 return;
4289         }
4290
4291         switch (gen_crq->cmd) {
4292         case VERSION_EXCHANGE_RSP:
4293                 rc = crq->version_exchange_rsp.rc.code;
4294                 if (rc) {
4295                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4296                         break;
4297                 }
4298                 dev_info(dev, "Partner protocol version is %d\n",
4299                          crq->version_exchange_rsp.version);
4300                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
4301                     ibmvnic_version)
4302                         ibmvnic_version =
4303                             be16_to_cpu(crq->version_exchange_rsp.version);
4304                 send_cap_queries(adapter);
4305                 break;
4306         case QUERY_CAPABILITY_RSP:
4307                 handle_query_cap_rsp(crq, adapter);
4308                 break;
4309         case QUERY_MAP_RSP:
4310                 handle_query_map_rsp(crq, adapter);
4311                 break;
4312         case REQUEST_MAP_RSP:
4313                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4314                 complete(&adapter->fw_done);
4315                 break;
4316         case REQUEST_UNMAP_RSP:
4317                 handle_request_unmap_rsp(crq, adapter);
4318                 break;
4319         case REQUEST_CAPABILITY_RSP:
4320                 handle_request_cap_rsp(crq, adapter);
4321                 break;
4322         case LOGIN_RSP:
4323                 netdev_dbg(netdev, "Got Login Response\n");
4324                 handle_login_rsp(crq, adapter);
4325                 break;
4326         case LOGICAL_LINK_STATE_RSP:
4327                 netdev_dbg(netdev,
4328                            "Got Logical Link State Response, state: %d rc: %d\n",
4329                            crq->logical_link_state_rsp.link_state,
4330                            crq->logical_link_state_rsp.rc.code);
4331                 adapter->logical_link_state =
4332                     crq->logical_link_state_rsp.link_state;
4333                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4334                 complete(&adapter->init_done);
4335                 break;
4336         case LINK_STATE_INDICATION:
4337                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4338                 adapter->phys_link_state =
4339                     crq->link_state_indication.phys_link_state;
4340                 adapter->logical_link_state =
4341                     crq->link_state_indication.logical_link_state;
4342                 break;
4343         case CHANGE_MAC_ADDR_RSP:
4344                 netdev_dbg(netdev, "Got MAC address change Response\n");
4345                 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4346                 break;
4347         case ERROR_INDICATION:
4348                 netdev_dbg(netdev, "Got Error Indication\n");
4349                 handle_error_indication(crq, adapter);
4350                 break;
4351         case REQUEST_STATISTICS_RSP:
4352                 netdev_dbg(netdev, "Got Statistics Response\n");
4353                 complete(&adapter->stats_done);
4354                 break;
4355         case QUERY_IP_OFFLOAD_RSP:
4356                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4357                 handle_query_ip_offload_rsp(adapter);
4358                 break;
4359         case MULTICAST_CTRL_RSP:
4360                 netdev_dbg(netdev, "Got multicast control Response\n");
4361                 break;
4362         case CONTROL_IP_OFFLOAD_RSP:
4363                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4364                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4365                                  sizeof(adapter->ip_offload_ctrl),
4366                                  DMA_TO_DEVICE);
4367                 complete(&adapter->init_done);
4368                 break;
4369         case COLLECT_FW_TRACE_RSP:
4370                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4371                 complete(&adapter->fw_done);
4372                 break;
4373         case GET_VPD_SIZE_RSP:
4374                 handle_vpd_size_rsp(crq, adapter);
4375                 break;
4376         case GET_VPD_RSP:
4377                 handle_vpd_rsp(crq, adapter);
4378                 break;
4379         default:
4380                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4381                            gen_crq->cmd);
4382         }
4383 }
4384
4385 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4386 {
4387         struct ibmvnic_adapter *adapter = instance;
4388
4389         tasklet_schedule(&adapter->tasklet);
4390         return IRQ_HANDLED;
4391 }
4392
4393 static void ibmvnic_tasklet(void *data)
4394 {
4395         struct ibmvnic_adapter *adapter = data;
4396         struct ibmvnic_crq_queue *queue = &adapter->crq;
4397         union ibmvnic_crq *crq;
4398         unsigned long flags;
4399         bool done = false;
4400
4401         spin_lock_irqsave(&queue->lock, flags);
4402         while (!done) {
4403                 /* Pull all the valid messages off the CRQ */
4404                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4405                         ibmvnic_handle_crq(crq, adapter);
4406                         crq->generic.first = 0;
4407                 }
4408
4409                 /* remain in tasklet until all
4410                  * capabilities responses are received
4411                  */
4412                 if (!adapter->wait_capability)
4413                         done = true;
4414         }
4415         /* if capabilities CRQ's were sent in this tasklet, the following
4416          * tasklet must wait until all responses are received
4417          */
4418         if (atomic_read(&adapter->running_cap_crqs) != 0)
4419                 adapter->wait_capability = true;
4420         spin_unlock_irqrestore(&queue->lock, flags);
4421 }
4422
4423 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4424 {
4425         struct vio_dev *vdev = adapter->vdev;
4426         int rc;
4427
4428         do {
4429                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4430         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4431
4432         if (rc)
4433                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4434
4435         return rc;
4436 }
4437
4438 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4439 {
4440         struct ibmvnic_crq_queue *crq = &adapter->crq;
4441         struct device *dev = &adapter->vdev->dev;
4442         struct vio_dev *vdev = adapter->vdev;
4443         int rc;
4444
4445         /* Close the CRQ */
4446         do {
4447                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4448         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4449
4450         /* Clean out the queue */
4451         memset(crq->msgs, 0, PAGE_SIZE);
4452         crq->cur = 0;
4453         crq->active = false;
4454
4455         /* And re-open it again */
4456         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4457                                 crq->msg_token, PAGE_SIZE);
4458
4459         if (rc == H_CLOSED)
4460                 /* Adapter is good, but other end is not ready */
4461                 dev_warn(dev, "Partner adapter not ready\n");
4462         else if (rc != 0)
4463                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4464
4465         return rc;
4466 }
4467
4468 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4469 {
4470         struct ibmvnic_crq_queue *crq = &adapter->crq;
4471         struct vio_dev *vdev = adapter->vdev;
4472         long rc;
4473
4474         if (!crq->msgs)
4475                 return;
4476
4477         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4478         free_irq(vdev->irq, adapter);
4479         tasklet_kill(&adapter->tasklet);
4480         do {
4481                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4482         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4483
4484         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4485                          DMA_BIDIRECTIONAL);
4486         free_page((unsigned long)crq->msgs);
4487         crq->msgs = NULL;
4488         crq->active = false;
4489 }
4490
4491 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4492 {
4493         struct ibmvnic_crq_queue *crq = &adapter->crq;
4494         struct device *dev = &adapter->vdev->dev;
4495         struct vio_dev *vdev = adapter->vdev;
4496         int rc, retrc = -ENOMEM;
4497
4498         if (crq->msgs)
4499                 return 0;
4500
4501         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4502         /* Should we allocate more than one page? */
4503
4504         if (!crq->msgs)
4505                 return -ENOMEM;
4506
4507         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4508         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4509                                         DMA_BIDIRECTIONAL);
4510         if (dma_mapping_error(dev, crq->msg_token))
4511                 goto map_failed;
4512
4513         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4514                                 crq->msg_token, PAGE_SIZE);
4515
4516         if (rc == H_RESOURCE)
4517                 /* maybe kexecing and resource is busy. try a reset */
4518                 rc = ibmvnic_reset_crq(adapter);
4519         retrc = rc;
4520
4521         if (rc == H_CLOSED) {
4522                 dev_warn(dev, "Partner adapter not ready\n");
4523         } else if (rc) {
4524                 dev_warn(dev, "Error %d opening adapter\n", rc);
4525                 goto reg_crq_failed;
4526         }
4527
4528         retrc = 0;
4529
4530         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4531                      (unsigned long)adapter);
4532
4533         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4534         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
4535                          adapter);
4536         if (rc) {
4537                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4538                         vdev->irq, rc);
4539                 goto req_irq_failed;
4540         }
4541
4542         rc = vio_enable_interrupts(vdev);
4543         if (rc) {
4544                 dev_err(dev, "Error %d enabling interrupts\n", rc);
4545                 goto req_irq_failed;
4546         }
4547
4548         crq->cur = 0;
4549         spin_lock_init(&crq->lock);
4550
4551         return retrc;
4552
4553 req_irq_failed:
4554         tasklet_kill(&adapter->tasklet);
4555         do {
4556                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4557         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4558 reg_crq_failed:
4559         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4560 map_failed:
4561         free_page((unsigned long)crq->msgs);
4562         crq->msgs = NULL;
4563         return retrc;
4564 }
4565
4566 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4567 {
4568         struct device *dev = &adapter->vdev->dev;
4569         unsigned long timeout = msecs_to_jiffies(30000);
4570         u64 old_num_rx_queues, old_num_tx_queues;
4571         int rc;
4572
4573         adapter->from_passive_init = false;
4574
4575         old_num_rx_queues = adapter->req_rx_queues;
4576         old_num_tx_queues = adapter->req_tx_queues;
4577
4578         reinit_completion(&adapter->init_done);
4579         adapter->init_done_rc = 0;
4580         ibmvnic_send_crq_init(adapter);
4581         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4582                 dev_err(dev, "Initialization sequence timed out\n");
4583                 return -1;
4584         }
4585
4586         if (adapter->init_done_rc) {
4587                 release_crq_queue(adapter);
4588                 return adapter->init_done_rc;
4589         }
4590
4591         if (adapter->from_passive_init) {
4592                 adapter->state = VNIC_OPEN;
4593                 adapter->from_passive_init = false;
4594                 return -1;
4595         }
4596
4597         if (adapter->resetting && !adapter->wait_for_reset &&
4598             adapter->reset_reason != VNIC_RESET_MOBILITY) {
4599                 if (adapter->req_rx_queues != old_num_rx_queues ||
4600                     adapter->req_tx_queues != old_num_tx_queues) {
4601                         release_sub_crqs(adapter, 0);
4602                         rc = init_sub_crqs(adapter);
4603                 } else {
4604                         rc = reset_sub_crq_queues(adapter);
4605                 }
4606         } else {
4607                 rc = init_sub_crqs(adapter);
4608         }
4609
4610         if (rc) {
4611                 dev_err(dev, "Initialization of sub crqs failed\n");
4612                 release_crq_queue(adapter);
4613                 return rc;
4614         }
4615
4616         rc = init_sub_crq_irqs(adapter);
4617         if (rc) {
4618                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4619                 release_crq_queue(adapter);
4620         }
4621
4622         return rc;
4623 }
4624
4625 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4626 {
4627         struct device *dev = &adapter->vdev->dev;
4628         unsigned long timeout = msecs_to_jiffies(30000);
4629         int rc;
4630
4631         adapter->from_passive_init = false;
4632
4633         adapter->init_done_rc = 0;
4634         ibmvnic_send_crq_init(adapter);
4635         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4636                 dev_err(dev, "Initialization sequence timed out\n");
4637                 return -1;
4638         }
4639
4640         if (adapter->init_done_rc) {
4641                 release_crq_queue(adapter);
4642                 return adapter->init_done_rc;
4643         }
4644
4645         if (adapter->from_passive_init) {
4646                 adapter->state = VNIC_OPEN;
4647                 adapter->from_passive_init = false;
4648                 return -1;
4649         }
4650
4651         rc = init_sub_crqs(adapter);
4652         if (rc) {
4653                 dev_err(dev, "Initialization of sub crqs failed\n");
4654                 release_crq_queue(adapter);
4655                 return rc;
4656         }
4657
4658         rc = init_sub_crq_irqs(adapter);
4659         if (rc) {
4660                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4661                 release_crq_queue(adapter);
4662         }
4663
4664         return rc;
4665 }
4666
4667 static struct device_attribute dev_attr_failover;
4668
4669 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4670 {
4671         struct ibmvnic_adapter *adapter;
4672         struct net_device *netdev;
4673         unsigned char *mac_addr_p;
4674         int rc;
4675
4676         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4677                 dev->unit_address);
4678
4679         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4680                                                         VETH_MAC_ADDR, NULL);
4681         if (!mac_addr_p) {
4682                 dev_err(&dev->dev,
4683                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4684                         __FILE__, __LINE__);
4685                 return 0;
4686         }
4687
4688         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4689                                    IBMVNIC_MAX_QUEUES);
4690         if (!netdev)
4691                 return -ENOMEM;
4692
4693         adapter = netdev_priv(netdev);
4694         adapter->state = VNIC_PROBING;
4695         dev_set_drvdata(&dev->dev, netdev);
4696         adapter->vdev = dev;
4697         adapter->netdev = netdev;
4698
4699         ether_addr_copy(adapter->mac_addr, mac_addr_p);
4700         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4701         netdev->irq = dev->irq;
4702         netdev->netdev_ops = &ibmvnic_netdev_ops;
4703         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4704         SET_NETDEV_DEV(netdev, &dev->dev);
4705
4706         spin_lock_init(&adapter->stats_lock);
4707
4708         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4709         INIT_LIST_HEAD(&adapter->rwi_list);
4710         spin_lock_init(&adapter->rwi_lock);
4711         init_completion(&adapter->init_done);
4712         adapter->resetting = false;
4713
4714         adapter->mac_change_pending = false;
4715
4716         do {
4717                 rc = init_crq_queue(adapter);
4718                 if (rc) {
4719                         dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
4720                                 rc);
4721                         goto ibmvnic_init_fail;
4722                 }
4723
4724                 rc = ibmvnic_init(adapter);
4725                 if (rc && rc != EAGAIN)
4726                         goto ibmvnic_init_fail;
4727         } while (rc == EAGAIN);
4728
4729         rc = init_stats_buffers(adapter);
4730         if (rc)
4731                 goto ibmvnic_init_fail;
4732
4733         rc = init_stats_token(adapter);
4734         if (rc)
4735                 goto ibmvnic_stats_fail;
4736
4737         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4738         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4739         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4740
4741         rc = device_create_file(&dev->dev, &dev_attr_failover);
4742         if (rc)
4743                 goto ibmvnic_dev_file_err;
4744
4745         netif_carrier_off(netdev);
4746         rc = register_netdev(netdev);
4747         if (rc) {
4748                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4749                 goto ibmvnic_register_fail;
4750         }
4751         dev_info(&dev->dev, "ibmvnic registered\n");
4752
4753         adapter->state = VNIC_PROBED;
4754
4755         adapter->wait_for_reset = false;
4756
4757         return 0;
4758
4759 ibmvnic_register_fail:
4760         device_remove_file(&dev->dev, &dev_attr_failover);
4761
4762 ibmvnic_dev_file_err:
4763         release_stats_token(adapter);
4764
4765 ibmvnic_stats_fail:
4766         release_stats_buffers(adapter);
4767
4768 ibmvnic_init_fail:
4769         release_sub_crqs(adapter, 1);
4770         release_crq_queue(adapter);
4771         free_netdev(netdev);
4772
4773         return rc;
4774 }
4775
4776 static int ibmvnic_remove(struct vio_dev *dev)
4777 {
4778         struct net_device *netdev = dev_get_drvdata(&dev->dev);
4779         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4780
4781         adapter->state = VNIC_REMOVING;
4782         rtnl_lock();
4783         unregister_netdevice(netdev);
4784
4785         release_resources(adapter);
4786         release_sub_crqs(adapter, 1);
4787         release_crq_queue(adapter);
4788
4789         release_stats_token(adapter);
4790         release_stats_buffers(adapter);
4791
4792         adapter->state = VNIC_REMOVED;
4793
4794         rtnl_unlock();
4795         device_remove_file(&dev->dev, &dev_attr_failover);
4796         free_netdev(netdev);
4797         dev_set_drvdata(&dev->dev, NULL);
4798
4799         return 0;
4800 }
4801
4802 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4803                               const char *buf, size_t count)
4804 {
4805         struct net_device *netdev = dev_get_drvdata(dev);
4806         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4807         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4808         __be64 session_token;
4809         long rc;
4810
4811         if (!sysfs_streq(buf, "1"))
4812                 return -EINVAL;
4813
4814         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4815                          H_GET_SESSION_TOKEN, 0, 0, 0);
4816         if (rc) {
4817                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4818                            rc);
4819                 return -EINVAL;
4820         }
4821
4822         session_token = (__be64)retbuf[0];
4823         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4824                    be64_to_cpu(session_token));
4825         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4826                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
4827         if (rc) {
4828                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4829                            rc);
4830                 return -EINVAL;
4831         }
4832
4833         return count;
4834 }
4835
4836 static DEVICE_ATTR_WO(failover);
4837
4838 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4839 {
4840         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4841         struct ibmvnic_adapter *adapter;
4842         struct iommu_table *tbl;
4843         unsigned long ret = 0;
4844         int i;
4845
4846         tbl = get_iommu_table_base(&vdev->dev);
4847
4848         /* netdev inits at probe time along with the structures we need below*/
4849         if (!netdev)
4850                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4851
4852         adapter = netdev_priv(netdev);
4853
4854         ret += PAGE_SIZE; /* the crq message queue */
4855         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4856
4857         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4858                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
4859
4860         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4861              i++)
4862                 ret += adapter->rx_pool[i].size *
4863                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4864
4865         return ret;
4866 }
4867
4868 static int ibmvnic_resume(struct device *dev)
4869 {
4870         struct net_device *netdev = dev_get_drvdata(dev);
4871         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4872
4873         if (adapter->state != VNIC_OPEN)
4874                 return 0;
4875
4876         tasklet_schedule(&adapter->tasklet);
4877
4878         return 0;
4879 }
4880
4881 static const struct vio_device_id ibmvnic_device_table[] = {
4882         {"network", "IBM,vnic"},
4883         {"", "" }
4884 };
4885 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
4886
4887 static const struct dev_pm_ops ibmvnic_pm_ops = {
4888         .resume = ibmvnic_resume
4889 };
4890
4891 static struct vio_driver ibmvnic_driver = {
4892         .id_table       = ibmvnic_device_table,
4893         .probe          = ibmvnic_probe,
4894         .remove         = ibmvnic_remove,
4895         .get_desired_dma = ibmvnic_get_desired_dma,
4896         .name           = ibmvnic_driver_name,
4897         .pm             = &ibmvnic_pm_ops,
4898 };
4899
4900 /* module functions */
4901 static int __init ibmvnic_module_init(void)
4902 {
4903         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
4904                 IBMVNIC_DRIVER_VERSION);
4905
4906         return vio_register_driver(&ibmvnic_driver);
4907 }
4908
4909 static void __exit ibmvnic_module_exit(void)
4910 {
4911         vio_unregister_driver(&ibmvnic_driver);
4912 }
4913
4914 module_init(ibmvnic_module_init);
4915 module_exit(ibmvnic_module_exit);