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/in.h>
  63 #include <linux/ip.h>
  64 #include <linux/ipv6.h>
  65 #include <linux/irq.h>
  66 #include <linux/kthread.h>
  67 #include <linux/seq_file.h>
  68 #include <linux/interrupt.h>
  69 #include <net/net_namespace.h>
  70 #include <asm/hvcall.h>
  71 #include <linux/atomic.h>
  72 #include <asm/vio.h>
  73 #include <asm/iommu.h>
  74 #include <linux/uaccess.h>
  75 #include <asm/firmware.h>
  76 #include <linux/workqueue.h>
  77 #include <linux/if_vlan.h>
  78 #include <linux/utsname.h>
  79
  80 #include "ibmvnic.h"
  81
  82 static const char ibmvnic_driver_name[] = "ibmvnic";
  83 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
  84
  85 MODULE_AUTHOR("Santiago Leon");
  86 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
  87 MODULE_LICENSE("GPL");
  88 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
  89
  90 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
  91 static int ibmvnic_remove(struct vio_dev *);
  92 static void release_sub_crqs(struct ibmvnic_adapter *);
  93 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
  94 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
  95 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
  96 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
  97 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
  98                        union sub_crq *sub_crq);
  99 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
 100 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
 101 static int enable_scrq_irq(struct ibmvnic_adapter *,
 102                            struct ibmvnic_sub_crq_queue *);
 103 static int disable_scrq_irq(struct ibmvnic_adapter *,
 104                             struct ibmvnic_sub_crq_queue *);
 105 static int pending_scrq(struct ibmvnic_adapter *,
 106                         struct ibmvnic_sub_crq_queue *);
 107 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
 108                                         struct ibmvnic_sub_crq_queue *);
 109 static int ibmvnic_poll(struct napi_struct *napi, int data);
 110 static void send_map_query(struct ibmvnic_adapter *adapter);
 111 static void send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
 112 static void send_request_unmap(struct ibmvnic_adapter *, u8);
 113 static void send_login(struct ibmvnic_adapter *adapter);
 114 static void send_cap_queries(struct ibmvnic_adapter *adapter);
 115 static int init_sub_crqs(struct ibmvnic_adapter *);
 116 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
 117 static int ibmvnic_init(struct ibmvnic_adapter *);
 118 static void release_crq_queue(struct ibmvnic_adapter *);
 119 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p);
 120
 121 struct ibmvnic_stat {
 122         char name[ETH_GSTRING_LEN];
 123         int offset;
 124 };
 125
 126 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
 127                              offsetof(struct ibmvnic_statistics, stat))
 128 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
 129
 130 static const struct ibmvnic_stat ibmvnic_stats[] = {
 131         {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
 132         {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
 133         {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
 134         {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
 135         {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
 136         {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
 137         {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
 138         {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
 139         {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
 140         {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
 141         {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
 142         {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
 143         {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
 144         {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
 145         {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
 146         {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
 147         {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
 148         {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
 149         {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
 150         {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
 151         {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
 152         {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
 153 };
 154
 155 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
 156                           unsigned long length, unsigned long *number,
 157                           unsigned long *irq)
 158 {
 159         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
 160         long rc;
 161
 162         rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
 163         *number = retbuf[0];
 164         *irq = retbuf[1];
 165
 166         return rc;
 167 }
 168
 169 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
 170                                 struct ibmvnic_long_term_buff *ltb, int size)
 171 {
 172         struct device *dev = &adapter->vdev->dev;
 173
 174         ltb->size = size;
 175         ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
 176                                        GFP_KERNEL);
 177
 178         if (!ltb->buff) {
 179                 dev_err(dev, "Couldn't alloc long term buffer\n");
 180                 return -ENOMEM;
 181         }
 182         ltb->map_id = adapter->map_id;
 183         adapter->map_id++;
 184
 185         init_completion(&adapter->fw_done);
 186         send_request_map(adapter, ltb->addr,
 187                          ltb->size, ltb->map_id);
 188         wait_for_completion(&adapter->fw_done);
 189
 190         if (adapter->fw_done_rc) {
 191                 dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
 192                         adapter->fw_done_rc);
 193                 return -1;
 194         }
 195         return 0;
 196 }
 197
 198 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
 199                                 struct ibmvnic_long_term_buff *ltb)
 200 {
 201         struct device *dev = &adapter->vdev->dev;
 202
 203         if (!ltb->buff)
 204                 return;
 205
 206         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
 207             adapter->reset_reason != VNIC_RESET_MOBILITY)
 208                 send_request_unmap(adapter, ltb->map_id);
 209         dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 210 }
 211
 212 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
 213                                 struct ibmvnic_long_term_buff *ltb)
 214 {
 215         memset(ltb->buff, 0, ltb->size);
 216
 217         init_completion(&adapter->fw_done);
 218         send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
 219         wait_for_completion(&adapter->fw_done);
 220
 221         if (adapter->fw_done_rc) {
 222                 dev_info(&adapter->vdev->dev,
 223                          "Reset failed, attempting to free and reallocate buffer\n");
 224                 free_long_term_buff(adapter, ltb);
 225                 return alloc_long_term_buff(adapter, ltb, ltb->size);
 226         }
 227         return 0;
 228 }
 229
 230 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
 231 {
 232         int i;
 233
 234         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 235              i++)
 236                 adapter->rx_pool[i].active = 0;
 237 }
 238
 239 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
 240                               struct ibmvnic_rx_pool *pool)
 241 {
 242         int count = pool->size - atomic_read(&pool->available);
 243         struct device *dev = &adapter->vdev->dev;
 244         int buffers_added = 0;
 245         unsigned long lpar_rc;
 246         union sub_crq sub_crq;
 247         struct sk_buff *skb;
 248         unsigned int offset;
 249         dma_addr_t dma_addr;
 250         unsigned char *dst;
 251         u64 *handle_array;
 252         int shift = 0;
 253         int index;
 254         int i;
 255
 256         if (!pool->active)
 257                 return;
 258
 259         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 260                                       be32_to_cpu(adapter->login_rsp_buf->
 261                                       off_rxadd_subcrqs));
 262
 263         for (i = 0; i < count; ++i) {
 264                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 265                 if (!skb) {
 266                         dev_err(dev, "Couldn't replenish rx buff\n");
 267                         adapter->replenish_no_mem++;
 268                         break;
 269                 }
 270
 271                 index = pool->free_map[pool->next_free];
 272
 273                 if (pool->rx_buff[index].skb)
 274                         dev_err(dev, "Inconsistent free_map!\n");
 275
 276                 /* Copy the skb to the long term mapped DMA buffer */
 277                 offset = index * pool->buff_size;
 278                 dst = pool->long_term_buff.buff + offset;
 279                 memset(dst, 0, pool->buff_size);
 280                 dma_addr = pool->long_term_buff.addr + offset;
 281                 pool->rx_buff[index].data = dst;
 282
 283                 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
 284                 pool->rx_buff[index].dma = dma_addr;
 285                 pool->rx_buff[index].skb = skb;
 286                 pool->rx_buff[index].pool_index = pool->index;
 287                 pool->rx_buff[index].size = pool->buff_size;
 288
 289                 memset(&sub_crq, 0, sizeof(sub_crq));
 290                 sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
 291                 sub_crq.rx_add.correlator =
 292                     cpu_to_be64((u64)&pool->rx_buff[index]);
 293                 sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
 294                 sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
 295
 296                 /* The length field of the sCRQ is defined to be 24 bits so the
 297                  * buffer size needs to be left shifted by a byte before it is
 298                  * converted to big endian to prevent the last byte from being
 299                  * truncated.
 300                  */
 301 #ifdef __LITTLE_ENDIAN__
 302                 shift = 8;
 303 #endif
 304                 sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
 305
 306                 lpar_rc = send_subcrq(adapter, handle_array[pool->index],
 307                                       &sub_crq);
 308                 if (lpar_rc != H_SUCCESS)
 309                         goto failure;
 310
 311                 buffers_added++;
 312                 adapter->replenish_add_buff_success++;
 313                 pool->next_free = (pool->next_free + 1) % pool->size;
 314         }
 315         atomic_add(buffers_added, &pool->available);
 316         return;
 317
 318 failure:
 319         dev_info(dev, "replenish pools failure\n");
 320         pool->free_map[pool->next_free] = index;
 321         pool->rx_buff[index].skb = NULL;
 322         if (!dma_mapping_error(dev, dma_addr))
 323                 dma_unmap_single(dev, dma_addr, pool->buff_size,
 324                                  DMA_FROM_DEVICE);
 325
 326         dev_kfree_skb_any(skb);
 327         adapter->replenish_add_buff_failure++;
 328         atomic_add(buffers_added, &pool->available);
 329
 330         if (lpar_rc == H_CLOSED) {
 331                 /* Disable buffer pool replenishment and report carrier off if
 332                  * queue is closed. Firmware guarantees that a signal will
 333                  * be sent to the driver, triggering a reset.
 334                  */
 335                 deactivate_rx_pools(adapter);
 336                 netif_carrier_off(adapter->netdev);
 337         }
 338 }
 339
 340 static void replenish_pools(struct ibmvnic_adapter *adapter)
 341 {
 342         int i;
 343
 344         adapter->replenish_task_cycles++;
 345         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 346              i++) {
 347                 if (adapter->rx_pool[i].active)
 348                         replenish_rx_pool(adapter, &adapter->rx_pool[i]);
 349         }
 350 }
 351
 352 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
 353 {
 354         kfree(adapter->tx_stats_buffers);
 355         kfree(adapter->rx_stats_buffers);
 356 }
 357
 358 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
 359 {
 360         adapter->tx_stats_buffers =
 361                                 kcalloc(adapter->req_tx_queues,
 362                                         sizeof(struct ibmvnic_tx_queue_stats),
 363                                         GFP_KERNEL);
 364         if (!adapter->tx_stats_buffers)
 365                 return -ENOMEM;
 366
 367         adapter->rx_stats_buffers =
 368                                 kcalloc(adapter->req_rx_queues,
 369                                         sizeof(struct ibmvnic_rx_queue_stats),
 370                                         GFP_KERNEL);
 371         if (!adapter->rx_stats_buffers)
 372                 return -ENOMEM;
 373
 374         return 0;
 375 }
 376
 377 static void release_stats_token(struct ibmvnic_adapter *adapter)
 378 {
 379         struct device *dev = &adapter->vdev->dev;
 380
 381         if (!adapter->stats_token)
 382                 return;
 383
 384         dma_unmap_single(dev, adapter->stats_token,
 385                          sizeof(struct ibmvnic_statistics),
 386                          DMA_FROM_DEVICE);
 387         adapter->stats_token = 0;
 388 }
 389
 390 static int init_stats_token(struct ibmvnic_adapter *adapter)
 391 {
 392         struct device *dev = &adapter->vdev->dev;
 393         dma_addr_t stok;
 394
 395         stok = dma_map_single(dev, &adapter->stats,
 396                               sizeof(struct ibmvnic_statistics),
 397                               DMA_FROM_DEVICE);
 398         if (dma_mapping_error(dev, stok)) {
 399                 dev_err(dev, "Couldn't map stats buffer\n");
 400                 return -1;
 401         }
 402
 403         adapter->stats_token = stok;
 404         netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
 405         return 0;
 406 }
 407
 408 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
 409 {
 410         struct ibmvnic_rx_pool *rx_pool;
 411         int rx_scrqs;
 412         int i, j, rc;
 413
 414         rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 415         for (i = 0; i < rx_scrqs; i++) {
 416                 rx_pool = &adapter->rx_pool[i];
 417
 418                 netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
 419
 420                 rc = reset_long_term_buff(adapter, &rx_pool->long_term_buff);
 421                 if (rc)
 422                         return rc;
 423
 424                 for (j = 0; j < rx_pool->size; j++)
 425                         rx_pool->free_map[j] = j;
 426
 427                 memset(rx_pool->rx_buff, 0,
 428                        rx_pool->size * sizeof(struct ibmvnic_rx_buff));
 429
 430                 atomic_set(&rx_pool->available, 0);
 431                 rx_pool->next_alloc = 0;
 432                 rx_pool->next_free = 0;
 433                 rx_pool->active = 1;
 434         }
 435
 436         return 0;
 437 }
 438
 439 static void release_rx_pools(struct ibmvnic_adapter *adapter)
 440 {
 441         struct ibmvnic_rx_pool *rx_pool;
 442         int rx_scrqs;
 443         int i, j;
 444
 445         if (!adapter->rx_pool)
 446                 return;
 447
 448         rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 449         for (i = 0; i < rx_scrqs; i++) {
 450                 rx_pool = &adapter->rx_pool[i];
 451
 452                 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
 453
 454                 kfree(rx_pool->free_map);
 455                 free_long_term_buff(adapter, &rx_pool->long_term_buff);
 456
 457                 if (!rx_pool->rx_buff)
 458                         continue;
 459
 460                 for (j = 0; j < rx_pool->size; j++) {
 461                         if (rx_pool->rx_buff[j].skb) {
 462                                 dev_kfree_skb_any(rx_pool->rx_buff[i].skb);
 463                                 rx_pool->rx_buff[i].skb = NULL;
 464                         }
 465                 }
 466
 467                 kfree(rx_pool->rx_buff);
 468         }
 469
 470         kfree(adapter->rx_pool);
 471         adapter->rx_pool = NULL;
 472 }
 473
 474 static int init_rx_pools(struct net_device *netdev)
 475 {
 476         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 477         struct device *dev = &adapter->vdev->dev;
 478         struct ibmvnic_rx_pool *rx_pool;
 479         int rxadd_subcrqs;
 480         u64 *size_array;
 481         int i, j;
 482
 483         rxadd_subcrqs =
 484                 be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 485         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 486                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 487
 488         adapter->rx_pool = kcalloc(rxadd_subcrqs,
 489                                    sizeof(struct ibmvnic_rx_pool),
 490                                    GFP_KERNEL);
 491         if (!adapter->rx_pool) {
 492                 dev_err(dev, "Failed to allocate rx pools\n");
 493                 return -1;
 494         }
 495
 496         for (i = 0; i < rxadd_subcrqs; i++) {
 497                 rx_pool = &adapter->rx_pool[i];
 498
 499                 netdev_dbg(adapter->netdev,
 500                            "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
 501                            i, adapter->req_rx_add_entries_per_subcrq,
 502                            be64_to_cpu(size_array[i]));
 503
 504                 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
 505                 rx_pool->index = i;
 506                 rx_pool->buff_size = be64_to_cpu(size_array[i]);
 507                 rx_pool->active = 1;
 508
 509                 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
 510                                             GFP_KERNEL);
 511                 if (!rx_pool->free_map) {
 512                         release_rx_pools(adapter);
 513                         return -1;
 514                 }
 515
 516                 rx_pool->rx_buff = kcalloc(rx_pool->size,
 517                                            sizeof(struct ibmvnic_rx_buff),
 518                                            GFP_KERNEL);
 519                 if (!rx_pool->rx_buff) {
 520                         dev_err(dev, "Couldn't alloc rx buffers\n");
 521                         release_rx_pools(adapter);
 522                         return -1;
 523                 }
 524
 525                 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 526                                          rx_pool->size * rx_pool->buff_size)) {
 527                         release_rx_pools(adapter);
 528                         return -1;
 529                 }
 530
 531                 for (j = 0; j < rx_pool->size; ++j)
 532                         rx_pool->free_map[j] = j;
 533
 534                 atomic_set(&rx_pool->available, 0);
 535                 rx_pool->next_alloc = 0;
 536                 rx_pool->next_free = 0;
 537         }
 538
 539         return 0;
 540 }
 541
 542 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
 543 {
 544         struct ibmvnic_tx_pool *tx_pool;
 545         int tx_scrqs;
 546         int i, j, rc;
 547
 548         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 549         for (i = 0; i < tx_scrqs; i++) {
 550                 netdev_dbg(adapter->netdev, "Re-setting tx_pool[%d]\n", i);
 551
 552                 tx_pool = &adapter->tx_pool[i];
 553
 554                 rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
 555                 if (rc)
 556                         return rc;
 557
 558                 rc = reset_long_term_buff(adapter, &tx_pool->tso_ltb);
 559                 if (rc)
 560                         return rc;
 561
 562                 memset(tx_pool->tx_buff, 0,
 563                        adapter->req_tx_entries_per_subcrq *
 564                        sizeof(struct ibmvnic_tx_buff));
 565
 566                 for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
 567                         tx_pool->free_map[j] = j;
 568
 569                 tx_pool->consumer_index = 0;
 570                 tx_pool->producer_index = 0;
 571                 tx_pool->tso_index = 0;
 572         }
 573
 574         return 0;
 575 }
 576
 577 static void release_vpd_data(struct ibmvnic_adapter *adapter)
 578 {
 579         if (!adapter->vpd)
 580                 return;
 581
 582         kfree(adapter->vpd->buff);
 583         kfree(adapter->vpd);
 584 }
 585
 586 static void release_tx_pools(struct ibmvnic_adapter *adapter)
 587 {
 588         struct ibmvnic_tx_pool *tx_pool;
 589         int i, tx_scrqs;
 590
 591         if (!adapter->tx_pool)
 592                 return;
 593
 594         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 595         for (i = 0; i < tx_scrqs; i++) {
 596                 netdev_dbg(adapter->netdev, "Releasing tx_pool[%d]\n", i);
 597                 tx_pool = &adapter->tx_pool[i];
 598                 kfree(tx_pool->tx_buff);
 599                 free_long_term_buff(adapter, &tx_pool->long_term_buff);
 600                 free_long_term_buff(adapter, &tx_pool->tso_ltb);
 601                 kfree(tx_pool->free_map);
 602         }
 603
 604         kfree(adapter->tx_pool);
 605         adapter->tx_pool = NULL;
 606 }
 607
 608 static int init_tx_pools(struct net_device *netdev)
 609 {
 610         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 611         struct device *dev = &adapter->vdev->dev;
 612         struct ibmvnic_tx_pool *tx_pool;
 613         int tx_subcrqs;
 614         int i, j;
 615
 616         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 617         adapter->tx_pool = kcalloc(tx_subcrqs,
 618                                    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 619         if (!adapter->tx_pool)
 620                 return -1;
 621
 622         for (i = 0; i < tx_subcrqs; i++) {
 623                 tx_pool = &adapter->tx_pool[i];
 624
 625                 netdev_dbg(adapter->netdev,
 626                            "Initializing tx_pool[%d], %lld buffs\n",
 627                            i, adapter->req_tx_entries_per_subcrq);
 628
 629                 tx_pool->tx_buff = kcalloc(adapter->req_tx_entries_per_subcrq,
 630                                            sizeof(struct ibmvnic_tx_buff),
 631                                            GFP_KERNEL);
 632                 if (!tx_pool->tx_buff) {
 633                         dev_err(dev, "tx pool buffer allocation failed\n");
 634                         release_tx_pools(adapter);
 635                         return -1;
 636                 }
 637
 638                 if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
 639                                          adapter->req_tx_entries_per_subcrq *
 640                                          adapter->req_mtu)) {
 641                         release_tx_pools(adapter);
 642                         return -1;
 643                 }
 644
 645                 /* alloc TSO ltb */
 646                 if (alloc_long_term_buff(adapter, &tx_pool->tso_ltb,
 647                                          IBMVNIC_TSO_BUFS *
 648                                          IBMVNIC_TSO_BUF_SZ)) {
 649                         release_tx_pools(adapter);
 650                         return -1;
 651                 }
 652
 653                 tx_pool->tso_index = 0;
 654
 655                 tx_pool->free_map = kcalloc(adapter->req_tx_entries_per_subcrq,
 656                                             sizeof(int), GFP_KERNEL);
 657                 if (!tx_pool->free_map) {
 658                         release_tx_pools(adapter);
 659                         return -1;
 660                 }
 661
 662                 for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
 663                         tx_pool->free_map[j] = j;
 664
 665                 tx_pool->consumer_index = 0;
 666                 tx_pool->producer_index = 0;
 667         }
 668
 669         return 0;
 670 }
 671
 672 static void release_error_buffers(struct ibmvnic_adapter *adapter)
 673 {
 674         struct device *dev = &adapter->vdev->dev;
 675         struct ibmvnic_error_buff *error_buff, *tmp;
 676         unsigned long flags;
 677
 678         spin_lock_irqsave(&adapter->error_list_lock, flags);
 679         list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list) {
 680                 list_del(&error_buff->list);
 681                 dma_unmap_single(dev, error_buff->dma, error_buff->len,
 682                                  DMA_FROM_DEVICE);
 683                 kfree(error_buff->buff);
 684                 kfree(error_buff);
 685         }
 686         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
 687 }
 688
 689 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
 690 {
 691         int i;
 692
 693         if (adapter->napi_enabled)
 694                 return;
 695
 696         for (i = 0; i < adapter->req_rx_queues; i++)
 697                 napi_enable(&adapter->napi[i]);
 698
 699         adapter->napi_enabled = true;
 700 }
 701
 702 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
 703 {
 704         int i;
 705
 706         if (!adapter->napi_enabled)
 707                 return;
 708
 709         for (i = 0; i < adapter->req_rx_queues; i++) {
 710                 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
 711                 napi_disable(&adapter->napi[i]);
 712         }
 713
 714         adapter->napi_enabled = false;
 715 }
 716
 717 static int ibmvnic_login(struct net_device *netdev)
 718 {
 719         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 720         unsigned long timeout = msecs_to_jiffies(30000);
 721         struct device *dev = &adapter->vdev->dev;
 722         int rc;
 723
 724         do {
 725                 if (adapter->renegotiate) {
 726                         adapter->renegotiate = false;
 727                         release_sub_crqs(adapter);
 728
 729                         reinit_completion(&adapter->init_done);
 730                         send_cap_queries(adapter);
 731                         if (!wait_for_completion_timeout(&adapter->init_done,
 732                                                          timeout)) {
 733                                 dev_err(dev, "Capabilities query timeout\n");
 734                                 return -1;
 735                         }
 736                         rc = init_sub_crqs(adapter);
 737                         if (rc) {
 738                                 dev_err(dev,
 739                                         "Initialization of SCRQ's failed\n");
 740                                 return -1;
 741                         }
 742                         rc = init_sub_crq_irqs(adapter);
 743                         if (rc) {
 744                                 dev_err(dev,
 745                                         "Initialization of SCRQ's irqs failed\n");
 746                                 return -1;
 747                         }
 748                 }
 749
 750                 reinit_completion(&adapter->init_done);
 751                 send_login(adapter);
 752                 if (!wait_for_completion_timeout(&adapter->init_done,
 753                                                  timeout)) {
 754                         dev_err(dev, "Login timeout\n");
 755                         return -1;
 756                 }
 757         } while (adapter->renegotiate);
 758
 759         /* handle pending MAC address changes after successful login */
 760         if (adapter->mac_change_pending) {
 761                 __ibmvnic_set_mac(netdev, &adapter->desired.mac);
 762                 adapter->mac_change_pending = false;
 763         }
 764
 765         return 0;
 766 }
 767
 768 static void release_resources(struct ibmvnic_adapter *adapter)
 769 {
 770         int i;
 771
 772         release_vpd_data(adapter);
 773
 774         release_tx_pools(adapter);
 775         release_rx_pools(adapter);
 776
 777         release_stats_token(adapter);
 778         release_stats_buffers(adapter);
 779         release_error_buffers(adapter);
 780
 781         if (adapter->napi) {
 782                 for (i = 0; i < adapter->req_rx_queues; i++) {
 783                         if (&adapter->napi[i]) {
 784                                 netdev_dbg(adapter->netdev,
 785                                            "Releasing napi[%d]\n", i);
 786                                 netif_napi_del(&adapter->napi[i]);
 787                         }
 788                 }
 789         }
 790 }
 791
 792 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
 793 {
 794         struct net_device *netdev = adapter->netdev;
 795         unsigned long timeout = msecs_to_jiffies(30000);
 796         union ibmvnic_crq crq;
 797         bool resend;
 798         int rc;
 799
 800         netdev_dbg(netdev, "setting link state %d\n", link_state);
 801
 802         memset(&crq, 0, sizeof(crq));
 803         crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
 804         crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
 805         crq.logical_link_state.link_state = link_state;
 806
 807         do {
 808                 resend = false;
 809
 810                 reinit_completion(&adapter->init_done);
 811                 rc = ibmvnic_send_crq(adapter, &crq);
 812                 if (rc) {
 813                         netdev_err(netdev, "Failed to set link state\n");
 814                         return rc;
 815                 }
 816
 817                 if (!wait_for_completion_timeout(&adapter->init_done,
 818                                                  timeout)) {
 819                         netdev_err(netdev, "timeout setting link state\n");
 820                         return -1;
 821                 }
 822
 823                 if (adapter->init_done_rc == 1) {
 824                         /* Partuial success, delay and re-send */
 825                         mdelay(1000);
 826                         resend = true;
 827                 }
 828         } while (resend);
 829
 830         return 0;
 831 }
 832
 833 static int set_real_num_queues(struct net_device *netdev)
 834 {
 835         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 836         int rc;
 837
 838         netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
 839                    adapter->req_tx_queues, adapter->req_rx_queues);
 840
 841         rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
 842         if (rc) {
 843                 netdev_err(netdev, "failed to set the number of tx queues\n");
 844                 return rc;
 845         }
 846
 847         rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
 848         if (rc)
 849                 netdev_err(netdev, "failed to set the number of rx queues\n");
 850
 851         return rc;
 852 }
 853
 854 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
 855 {
 856         struct device *dev = &adapter->vdev->dev;
 857         union ibmvnic_crq crq;
 858         int len = 0;
 859
 860         if (adapter->vpd->buff)
 861                 len = adapter->vpd->len;
 862
 863         reinit_completion(&adapter->fw_done);
 864         crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
 865         crq.get_vpd_size.cmd = GET_VPD_SIZE;
 866         ibmvnic_send_crq(adapter, &crq);
 867         wait_for_completion(&adapter->fw_done);
 868
 869         if (!adapter->vpd->len)
 870                 return -ENODATA;
 871
 872         if (!adapter->vpd->buff)
 873                 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
 874         else if (adapter->vpd->len != len)
 875                 adapter->vpd->buff =
 876                         krealloc(adapter->vpd->buff,
 877                                  adapter->vpd->len, GFP_KERNEL);
 878
 879         if (!adapter->vpd->buff) {
 880                 dev_err(dev, "Could allocate VPD buffer\n");
 881                 return -ENOMEM;
 882         }
 883
 884         adapter->vpd->dma_addr =
 885                 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
 886                                DMA_FROM_DEVICE);
 887         if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
 888                 dev_err(dev, "Could not map VPD buffer\n");
 889                 kfree(adapter->vpd->buff);
 890                 return -ENOMEM;
 891         }
 892
 893         reinit_completion(&adapter->fw_done);
 894         crq.get_vpd.first = IBMVNIC_CRQ_CMD;
 895         crq.get_vpd.cmd = GET_VPD;
 896         crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
 897         crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
 898         ibmvnic_send_crq(adapter, &crq);
 899         wait_for_completion(&adapter->fw_done);
 900
 901         return 0;
 902 }
 903
 904 static int init_resources(struct ibmvnic_adapter *adapter)
 905 {
 906         struct net_device *netdev = adapter->netdev;
 907         int i, rc;
 908
 909         rc = set_real_num_queues(netdev);
 910         if (rc)
 911                 return rc;
 912
 913         rc = init_stats_buffers(adapter);
 914         if (rc)
 915                 return rc;
 916
 917         rc = init_stats_token(adapter);
 918         if (rc)
 919                 return rc;
 920
 921         adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
 922         if (!adapter->vpd)
 923                 return -ENOMEM;
 924
 925         adapter->map_id = 1;
 926         adapter->napi = kcalloc(adapter->req_rx_queues,
 927                                 sizeof(struct napi_struct), GFP_KERNEL);
 928         if (!adapter->napi)
 929                 return -ENOMEM;
 930
 931         for (i = 0; i < adapter->req_rx_queues; i++) {
 932                 netdev_dbg(netdev, "Adding napi[%d]\n", i);
 933                 netif_napi_add(netdev, &adapter->napi[i], ibmvnic_poll,
 934                                NAPI_POLL_WEIGHT);
 935         }
 936
 937         send_map_query(adapter);
 938
 939         rc = init_rx_pools(netdev);
 940         if (rc)
 941                 return rc;
 942
 943         rc = init_tx_pools(netdev);
 944         return rc;
 945 }
 946
 947 static int __ibmvnic_open(struct net_device *netdev)
 948 {
 949         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 950         enum vnic_state prev_state = adapter->state;
 951         int i, rc;
 952
 953         adapter->state = VNIC_OPENING;
 954         replenish_pools(adapter);
 955         ibmvnic_napi_enable(adapter);
 956
 957         /* We're ready to receive frames, enable the sub-crq interrupts and
 958          * set the logical link state to up
 959          */
 960         for (i = 0; i < adapter->req_rx_queues; i++) {
 961                 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
 962                 if (prev_state == VNIC_CLOSED)
 963                         enable_irq(adapter->rx_scrq[i]->irq);
 964                 else
 965                         enable_scrq_irq(adapter, adapter->rx_scrq[i]);
 966         }
 967
 968         for (i = 0; i < adapter->req_tx_queues; i++) {
 969                 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
 970                 if (prev_state == VNIC_CLOSED)
 971                         enable_irq(adapter->tx_scrq[i]->irq);
 972                 else
 973                         enable_scrq_irq(adapter, adapter->tx_scrq[i]);
 974         }
 975
 976         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
 977         if (rc) {
 978                 for (i = 0; i < adapter->req_rx_queues; i++)
 979                         napi_disable(&adapter->napi[i]);
 980                 release_resources(adapter);
 981                 return rc;
 982         }
 983
 984         netif_tx_start_all_queues(netdev);
 985
 986         if (prev_state == VNIC_CLOSED) {
 987                 for (i = 0; i < adapter->req_rx_queues; i++)
 988                         napi_schedule(&adapter->napi[i]);
 989         }
 990
 991         adapter->state = VNIC_OPEN;
 992         return rc;
 993 }
 994
 995 static int ibmvnic_open(struct net_device *netdev)
 996 {
 997         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 998         int rc, vpd;
 999
1000         mutex_lock(&adapter->reset_lock);
1001
1002         if (adapter->state != VNIC_CLOSED) {
1003                 rc = ibmvnic_login(netdev);
1004                 if (rc) {
1005                         mutex_unlock(&adapter->reset_lock);
1006                         return rc;
1007                 }
1008
1009                 rc = init_resources(adapter);
1010                 if (rc) {
1011                         netdev_err(netdev, "failed to initialize resources\n");
1012                         release_resources(adapter);
1013                         mutex_unlock(&adapter->reset_lock);
1014                         return rc;
1015                 }
1016         }
1017
1018         rc = __ibmvnic_open(netdev);
1019         netif_carrier_on(netdev);
1020
1021         /* Vital Product Data (VPD) */
1022         vpd = ibmvnic_get_vpd(adapter);
1023         if (vpd)
1024                 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1025
1026         mutex_unlock(&adapter->reset_lock);
1027
1028         return rc;
1029 }
1030
1031 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1032 {
1033         struct ibmvnic_tx_pool *tx_pool;
1034         u64 tx_entries;
1035         int tx_scrqs;
1036         int i, j;
1037
1038         if (!adapter->tx_pool)
1039                 return;
1040
1041         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
1042         tx_entries = adapter->req_tx_entries_per_subcrq;
1043
1044         /* Free any remaining skbs in the tx buffer pools */
1045         for (i = 0; i < tx_scrqs; i++) {
1046                 tx_pool = &adapter->tx_pool[i];
1047                 if (!tx_pool)
1048                         continue;
1049
1050                 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1051                 for (j = 0; j < tx_entries; j++) {
1052                         if (tx_pool->tx_buff[j].skb) {
1053                                 dev_kfree_skb_any(tx_pool->tx_buff[j].skb);
1054                                 tx_pool->tx_buff[j].skb = NULL;
1055                         }
1056                 }
1057         }
1058 }
1059
1060 static int __ibmvnic_close(struct net_device *netdev)
1061 {
1062         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1063         int rc = 0;
1064         int i;
1065
1066         adapter->state = VNIC_CLOSING;
1067
1068         /* ensure that transmissions are stopped if called by do_reset */
1069         if (adapter->resetting)
1070                 netif_tx_disable(netdev);
1071         else
1072                 netif_tx_stop_all_queues(netdev);
1073
1074         ibmvnic_napi_disable(adapter);
1075
1076         if (adapter->tx_scrq) {
1077                 for (i = 0; i < adapter->req_tx_queues; i++)
1078                         if (adapter->tx_scrq[i]->irq) {
1079                                 netdev_dbg(adapter->netdev,
1080                                            "Disabling tx_scrq[%d] irq\n", i);
1081                                 disable_irq(adapter->tx_scrq[i]->irq);
1082                         }
1083         }
1084
1085         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1086         if (rc)
1087                 return rc;
1088
1089         if (adapter->rx_scrq) {
1090                 for (i = 0; i < adapter->req_rx_queues; i++) {
1091                         int retries = 10;
1092
1093                         while (pending_scrq(adapter, adapter->rx_scrq[i])) {
1094                                 retries--;
1095                                 mdelay(100);
1096
1097                                 if (retries == 0)
1098                                         break;
1099                         }
1100
1101                         if (adapter->rx_scrq[i]->irq) {
1102                                 netdev_dbg(adapter->netdev,
1103                                            "Disabling rx_scrq[%d] irq\n", i);
1104                                 disable_irq(adapter->rx_scrq[i]->irq);
1105                         }
1106                 }
1107         }
1108
1109         clean_tx_pools(adapter);
1110         adapter->state = VNIC_CLOSED;
1111         return rc;
1112 }
1113
1114 static int ibmvnic_close(struct net_device *netdev)
1115 {
1116         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1117         int rc;
1118
1119         mutex_lock(&adapter->reset_lock);
1120         rc = __ibmvnic_close(netdev);
1121         mutex_unlock(&adapter->reset_lock);
1122
1123         return rc;
1124 }
1125
1126 /**
1127  * build_hdr_data - creates L2/L3/L4 header data buffer
1128  * @hdr_field - bitfield determining needed headers
1129  * @skb - socket buffer
1130  * @hdr_len - array of header lengths
1131  * @tot_len - total length of data
1132  *
1133  * Reads hdr_field to determine which headers are needed by firmware.
1134  * Builds a buffer containing these headers.  Saves individual header
1135  * lengths and total buffer length to be used to build descriptors.
1136  */
1137 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1138                           int *hdr_len, u8 *hdr_data)
1139 {
1140         int len = 0;
1141         u8 *hdr;
1142
1143         hdr_len[0] = sizeof(struct ethhdr);
1144
1145         if (skb->protocol == htons(ETH_P_IP)) {
1146                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1147                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1148                         hdr_len[2] = tcp_hdrlen(skb);
1149                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1150                         hdr_len[2] = sizeof(struct udphdr);
1151         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1152                 hdr_len[1] = sizeof(struct ipv6hdr);
1153                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1154                         hdr_len[2] = tcp_hdrlen(skb);
1155                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1156                         hdr_len[2] = sizeof(struct udphdr);
1157         }
1158
1159         memset(hdr_data, 0, 120);
1160         if ((hdr_field >> 6) & 1) {
1161                 hdr = skb_mac_header(skb);
1162                 memcpy(hdr_data, hdr, hdr_len[0]);
1163                 len += hdr_len[0];
1164         }
1165
1166         if ((hdr_field >> 5) & 1) {
1167                 hdr = skb_network_header(skb);
1168                 memcpy(hdr_data + len, hdr, hdr_len[1]);
1169                 len += hdr_len[1];
1170         }
1171
1172         if ((hdr_field >> 4) & 1) {
1173                 hdr = skb_transport_header(skb);
1174                 memcpy(hdr_data + len, hdr, hdr_len[2]);
1175                 len += hdr_len[2];
1176         }
1177         return len;
1178 }
1179
1180 /**
1181  * create_hdr_descs - create header and header extension descriptors
1182  * @hdr_field - bitfield determining needed headers
1183  * @data - buffer containing header data
1184  * @len - length of data buffer
1185  * @hdr_len - array of individual header lengths
1186  * @scrq_arr - descriptor array
1187  *
1188  * Creates header and, if needed, header extension descriptors and
1189  * places them in a descriptor array, scrq_arr
1190  */
1191
1192 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1193                             union sub_crq *scrq_arr)
1194 {
1195         union sub_crq hdr_desc;
1196         int tmp_len = len;
1197         int num_descs = 0;
1198         u8 *data, *cur;
1199         int tmp;
1200
1201         while (tmp_len > 0) {
1202                 cur = hdr_data + len - tmp_len;
1203
1204                 memset(&hdr_desc, 0, sizeof(hdr_desc));
1205                 if (cur != hdr_data) {
1206                         data = hdr_desc.hdr_ext.data;
1207                         tmp = tmp_len > 29 ? 29 : tmp_len;
1208                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1209                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1210                         hdr_desc.hdr_ext.len = tmp;
1211                 } else {
1212                         data = hdr_desc.hdr.data;
1213                         tmp = tmp_len > 24 ? 24 : tmp_len;
1214                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1215                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1216                         hdr_desc.hdr.len = tmp;
1217                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1218                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1219                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1220                         hdr_desc.hdr.flag = hdr_field << 1;
1221                 }
1222                 memcpy(data, cur, tmp);
1223                 tmp_len -= tmp;
1224                 *scrq_arr = hdr_desc;
1225                 scrq_arr++;
1226                 num_descs++;
1227         }
1228
1229         return num_descs;
1230 }
1231
1232 /**
1233  * build_hdr_descs_arr - build a header descriptor array
1234  * @skb - socket buffer
1235  * @num_entries - number of descriptors to be sent
1236  * @subcrq - first TX descriptor
1237  * @hdr_field - bit field determining which headers will be sent
1238  *
1239  * This function will build a TX descriptor array with applicable
1240  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1241  */
1242
1243 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1244                                 int *num_entries, u8 hdr_field)
1245 {
1246         int hdr_len[3] = {0, 0, 0};
1247         int tot_len;
1248         u8 *hdr_data = txbuff->hdr_data;
1249
1250         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1251                                  txbuff->hdr_data);
1252         *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1253                          txbuff->indir_arr + 1);
1254 }
1255
1256 static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1257 {
1258         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1259         int queue_num = skb_get_queue_mapping(skb);
1260         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1261         struct device *dev = &adapter->vdev->dev;
1262         struct ibmvnic_tx_buff *tx_buff = NULL;
1263         struct ibmvnic_sub_crq_queue *tx_scrq;
1264         struct ibmvnic_tx_pool *tx_pool;
1265         unsigned int tx_send_failed = 0;
1266         unsigned int tx_map_failed = 0;
1267         unsigned int tx_dropped = 0;
1268         unsigned int tx_packets = 0;
1269         unsigned int tx_bytes = 0;
1270         dma_addr_t data_dma_addr;
1271         struct netdev_queue *txq;
1272         unsigned long lpar_rc;
1273         union sub_crq tx_crq;
1274         unsigned int offset;
1275         int num_entries = 1;
1276         unsigned char *dst;
1277         u64 *handle_array;
1278         int index = 0;
1279         int ret = 0;
1280
1281         if (adapter->resetting) {
1282                 if (!netif_subqueue_stopped(netdev, skb))
1283                         netif_stop_subqueue(netdev, queue_num);
1284                 dev_kfree_skb_any(skb);
1285
1286                 tx_send_failed++;
1287                 tx_dropped++;
1288                 ret = NETDEV_TX_OK;
1289                 goto out;
1290         }
1291
1292         tx_pool = &adapter->tx_pool[queue_num];
1293         tx_scrq = adapter->tx_scrq[queue_num];
1294         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1295         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1296                 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1297
1298         index = tx_pool->free_map[tx_pool->consumer_index];
1299
1300         if (skb_is_gso(skb)) {
1301                 offset = tx_pool->tso_index * IBMVNIC_TSO_BUF_SZ;
1302                 dst = tx_pool->tso_ltb.buff + offset;
1303                 memset(dst, 0, IBMVNIC_TSO_BUF_SZ);
1304                 data_dma_addr = tx_pool->tso_ltb.addr + offset;
1305                 tx_pool->tso_index++;
1306                 if (tx_pool->tso_index == IBMVNIC_TSO_BUFS)
1307                         tx_pool->tso_index = 0;
1308         } else {
1309                 offset = index * adapter->req_mtu;
1310                 dst = tx_pool->long_term_buff.buff + offset;
1311                 memset(dst, 0, adapter->req_mtu);
1312                 data_dma_addr = tx_pool->long_term_buff.addr + offset;
1313         }
1314
1315         if (skb_shinfo(skb)->nr_frags) {
1316                 int cur, i;
1317
1318                 /* Copy the head */
1319                 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1320                 cur = skb_headlen(skb);
1321
1322                 /* Copy the frags */
1323                 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1324                         const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1325
1326                         memcpy(dst + cur,
1327                                page_address(skb_frag_page(frag)) +
1328                                frag->page_offset, skb_frag_size(frag));
1329                         cur += skb_frag_size(frag);
1330                 }
1331         } else {
1332                 skb_copy_from_linear_data(skb, dst, skb->len);
1333         }
1334
1335         tx_pool->consumer_index =
1336             (tx_pool->consumer_index + 1) %
1337                 adapter->req_tx_entries_per_subcrq;
1338
1339         tx_buff = &tx_pool->tx_buff[index];
1340         tx_buff->skb = skb;
1341         tx_buff->data_dma[0] = data_dma_addr;
1342         tx_buff->data_len[0] = skb->len;
1343         tx_buff->index = index;
1344         tx_buff->pool_index = queue_num;
1345         tx_buff->last_frag = true;
1346
1347         memset(&tx_crq, 0, sizeof(tx_crq));
1348         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1349         tx_crq.v1.type = IBMVNIC_TX_DESC;
1350         tx_crq.v1.n_crq_elem = 1;
1351         tx_crq.v1.n_sge = 1;
1352         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1353         tx_crq.v1.correlator = cpu_to_be32(index);
1354         if (skb_is_gso(skb))
1355                 tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->tso_ltb.map_id);
1356         else
1357                 tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1358         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1359         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1360
1361         if (adapter->vlan_header_insertion) {
1362                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1363                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1364         }
1365
1366         if (skb->protocol == htons(ETH_P_IP)) {
1367                 if (ip_hdr(skb)->version == 4)
1368                         tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1369                 else if (ip_hdr(skb)->version == 6)
1370                         tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1371
1372                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1373                         tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1374                 else if (ip_hdr(skb)->protocol != IPPROTO_TCP)
1375                         tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1376         }
1377
1378         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1379                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1380                 hdrs += 2;
1381         }
1382         if (skb_is_gso(skb)) {
1383                 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1384                 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1385                 hdrs += 2;
1386         }
1387         /* determine if l2/3/4 headers are sent to firmware */
1388         if ((*hdrs >> 7) & 1 &&
1389             (skb->protocol == htons(ETH_P_IP) ||
1390              skb->protocol == htons(ETH_P_IPV6))) {
1391                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1392                 tx_crq.v1.n_crq_elem = num_entries;
1393                 tx_buff->indir_arr[0] = tx_crq;
1394                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1395                                                     sizeof(tx_buff->indir_arr),
1396                                                     DMA_TO_DEVICE);
1397                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1398                         dev_kfree_skb_any(skb);
1399                         tx_buff->skb = NULL;
1400                         if (!firmware_has_feature(FW_FEATURE_CMO))
1401                                 dev_err(dev, "tx: unable to map descriptor array\n");
1402                         tx_map_failed++;
1403                         tx_dropped++;
1404                         ret = NETDEV_TX_OK;
1405                         goto out;
1406                 }
1407                 lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1408                                                (u64)tx_buff->indir_dma,
1409                                                (u64)num_entries);
1410         } else {
1411                 lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1412                                       &tx_crq);
1413         }
1414         if (lpar_rc != H_SUCCESS) {
1415                 dev_err(dev, "tx failed with code %ld\n", lpar_rc);
1416
1417                 if (tx_pool->consumer_index == 0)
1418                         tx_pool->consumer_index =
1419                                 adapter->req_tx_entries_per_subcrq - 1;
1420                 else
1421                         tx_pool->consumer_index--;
1422
1423                 dev_kfree_skb_any(skb);
1424                 tx_buff->skb = NULL;
1425
1426                 if (lpar_rc == H_CLOSED) {
1427                         /* Disable TX and report carrier off if queue is closed.
1428                          * Firmware guarantees that a signal will be sent to the
1429                          * driver, triggering a reset or some other action.
1430                          */
1431                         netif_tx_stop_all_queues(netdev);
1432                         netif_carrier_off(netdev);
1433                 }
1434
1435                 tx_send_failed++;
1436                 tx_dropped++;
1437                 ret = NETDEV_TX_OK;
1438                 goto out;
1439         }
1440
1441         if (atomic_inc_return(&tx_scrq->used)
1442                                         >= adapter->req_tx_entries_per_subcrq) {
1443                 netdev_info(netdev, "Stopping queue %d\n", queue_num);
1444                 netif_stop_subqueue(netdev, queue_num);
1445         }
1446
1447         tx_packets++;
1448         tx_bytes += skb->len;
1449         txq->trans_start = jiffies;
1450         ret = NETDEV_TX_OK;
1451
1452 out:
1453         netdev->stats.tx_dropped += tx_dropped;
1454         netdev->stats.tx_bytes += tx_bytes;
1455         netdev->stats.tx_packets += tx_packets;
1456         adapter->tx_send_failed += tx_send_failed;
1457         adapter->tx_map_failed += tx_map_failed;
1458         adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1459         adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1460         adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1461
1462         return ret;
1463 }
1464
1465 static void ibmvnic_set_multi(struct net_device *netdev)
1466 {
1467         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1468         struct netdev_hw_addr *ha;
1469         union ibmvnic_crq crq;
1470
1471         memset(&crq, 0, sizeof(crq));
1472         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1473         crq.request_capability.cmd = REQUEST_CAPABILITY;
1474
1475         if (netdev->flags & IFF_PROMISC) {
1476                 if (!adapter->promisc_supported)
1477                         return;
1478         } else {
1479                 if (netdev->flags & IFF_ALLMULTI) {
1480                         /* Accept all multicast */
1481                         memset(&crq, 0, sizeof(crq));
1482                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1483                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1484                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1485                         ibmvnic_send_crq(adapter, &crq);
1486                 } else if (netdev_mc_empty(netdev)) {
1487                         /* Reject all multicast */
1488                         memset(&crq, 0, sizeof(crq));
1489                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1490                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1491                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1492                         ibmvnic_send_crq(adapter, &crq);
1493                 } else {
1494                         /* Accept one or more multicast(s) */
1495                         netdev_for_each_mc_addr(ha, netdev) {
1496                                 memset(&crq, 0, sizeof(crq));
1497                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1498                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1499                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1500                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1501                                                 ha->addr);
1502                                 ibmvnic_send_crq(adapter, &crq);
1503                         }
1504                 }
1505         }
1506 }
1507
1508 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p)
1509 {
1510         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1511         struct sockaddr *addr = p;
1512         union ibmvnic_crq crq;
1513
1514         if (!is_valid_ether_addr(addr->sa_data))
1515                 return -EADDRNOTAVAIL;
1516
1517         memset(&crq, 0, sizeof(crq));
1518         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1519         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1520         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1521         ibmvnic_send_crq(adapter, &crq);
1522         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1523         return 0;
1524 }
1525
1526 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1527 {
1528         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1529         struct sockaddr *addr = p;
1530
1531         if (adapter->state == VNIC_PROBED) {
1532                 memcpy(&adapter->desired.mac, addr, sizeof(struct sockaddr));
1533                 adapter->mac_change_pending = true;
1534                 return 0;
1535         }
1536
1537         __ibmvnic_set_mac(netdev, addr);
1538
1539         return 0;
1540 }
1541
1542 /**
1543  * do_reset returns zero if we are able to keep processing reset events, or
1544  * non-zero if we hit a fatal error and must halt.
1545  */
1546 static int do_reset(struct ibmvnic_adapter *adapter,
1547                     struct ibmvnic_rwi *rwi, u32 reset_state)
1548 {
1549         struct net_device *netdev = adapter->netdev;
1550         int i, rc;
1551
1552         netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1553                    rwi->reset_reason);
1554
1555         netif_carrier_off(netdev);
1556         adapter->reset_reason = rwi->reset_reason;
1557
1558         if (rwi->reset_reason == VNIC_RESET_MOBILITY) {
1559                 rc = ibmvnic_reenable_crq_queue(adapter);
1560                 if (rc)
1561                         return 0;
1562         }
1563
1564         rc = __ibmvnic_close(netdev);
1565         if (rc)
1566                 return rc;
1567
1568         if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1569             adapter->wait_for_reset) {
1570                 release_resources(adapter);
1571                 release_sub_crqs(adapter);
1572                 release_crq_queue(adapter);
1573         }
1574
1575         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1576                 /* remove the closed state so when we call open it appears
1577                  * we are coming from the probed state.
1578                  */
1579                 adapter->state = VNIC_PROBED;
1580
1581                 rc = ibmvnic_init(adapter);
1582                 if (rc)
1583                         return IBMVNIC_INIT_FAILED;
1584
1585                 /* If the adapter was in PROBE state prior to the reset,
1586                  * exit here.
1587                  */
1588                 if (reset_state == VNIC_PROBED)
1589                         return 0;
1590
1591                 rc = ibmvnic_login(netdev);
1592                 if (rc) {
1593                         adapter->state = VNIC_PROBED;
1594                         return 0;
1595                 }
1596
1597                 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1598                     adapter->wait_for_reset) {
1599                         rc = init_resources(adapter);
1600                         if (rc)
1601                                 return rc;
1602                 } else {
1603                         rc = reset_tx_pools(adapter);
1604                         if (rc)
1605                                 return rc;
1606
1607                         rc = reset_rx_pools(adapter);
1608                         if (rc)
1609                                 return rc;
1610
1611                         if (reset_state == VNIC_CLOSED)
1612                                 return 0;
1613                 }
1614         }
1615
1616         rc = __ibmvnic_open(netdev);
1617         if (rc) {
1618                 if (list_empty(&adapter->rwi_list))
1619                         adapter->state = VNIC_CLOSED;
1620                 else
1621                         adapter->state = reset_state;
1622
1623                 return 0;
1624         }
1625
1626         netif_carrier_on(netdev);
1627
1628         /* kick napi */
1629         for (i = 0; i < adapter->req_rx_queues; i++)
1630                 napi_schedule(&adapter->napi[i]);
1631
1632         if (adapter->reset_reason != VNIC_RESET_FAILOVER)
1633                 netdev_notify_peers(netdev);
1634
1635         return 0;
1636 }
1637
1638 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1639 {
1640         struct ibmvnic_rwi *rwi;
1641
1642         mutex_lock(&adapter->rwi_lock);
1643
1644         if (!list_empty(&adapter->rwi_list)) {
1645                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1646                                        list);
1647                 list_del(&rwi->list);
1648         } else {
1649                 rwi = NULL;
1650         }
1651
1652         mutex_unlock(&adapter->rwi_lock);
1653         return rwi;
1654 }
1655
1656 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1657 {
1658         struct ibmvnic_rwi *rwi;
1659
1660         rwi = get_next_rwi(adapter);
1661         while (rwi) {
1662                 kfree(rwi);
1663                 rwi = get_next_rwi(adapter);
1664         }
1665 }
1666
1667 static void __ibmvnic_reset(struct work_struct *work)
1668 {
1669         struct ibmvnic_rwi *rwi;
1670         struct ibmvnic_adapter *adapter;
1671         struct net_device *netdev;
1672         u32 reset_state;
1673         int rc = 0;
1674
1675         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1676         netdev = adapter->netdev;
1677
1678         mutex_lock(&adapter->reset_lock);
1679         adapter->resetting = true;
1680         reset_state = adapter->state;
1681
1682         rwi = get_next_rwi(adapter);
1683         while (rwi) {
1684                 rc = do_reset(adapter, rwi, reset_state);
1685                 kfree(rwi);
1686                 if (rc && rc != IBMVNIC_INIT_FAILED)
1687                         break;
1688
1689                 rwi = get_next_rwi(adapter);
1690         }
1691
1692         if (adapter->wait_for_reset) {
1693                 adapter->wait_for_reset = false;
1694                 adapter->reset_done_rc = rc;
1695                 complete(&adapter->reset_done);
1696         }
1697
1698         if (rc) {
1699                 netdev_dbg(adapter->netdev, "Reset failed\n");
1700                 free_all_rwi(adapter);
1701                 mutex_unlock(&adapter->reset_lock);
1702                 return;
1703         }
1704
1705         adapter->resetting = false;
1706         mutex_unlock(&adapter->reset_lock);
1707 }
1708
1709 static void ibmvnic_reset(struct ibmvnic_adapter *adapter,
1710                           enum ibmvnic_reset_reason reason)
1711 {
1712         struct ibmvnic_rwi *rwi, *tmp;
1713         struct net_device *netdev = adapter->netdev;
1714         struct list_head *entry;
1715
1716         if (adapter->state == VNIC_REMOVING ||
1717             adapter->state == VNIC_REMOVED) {
1718                 netdev_dbg(netdev, "Adapter removing, skipping reset\n");
1719                 return;
1720         }
1721
1722         if (adapter->state == VNIC_PROBING) {
1723                 netdev_warn(netdev, "Adapter reset during probe\n");
1724                 adapter->init_done_rc = EAGAIN;
1725                 return;
1726         }
1727
1728         mutex_lock(&adapter->rwi_lock);
1729
1730         list_for_each(entry, &adapter->rwi_list) {
1731                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
1732                 if (tmp->reset_reason == reason) {
1733                         netdev_dbg(netdev, "Skipping matching reset\n");
1734                         mutex_unlock(&adapter->rwi_lock);
1735                         return;
1736                 }
1737         }
1738
1739         rwi = kzalloc(sizeof(*rwi), GFP_KERNEL);
1740         if (!rwi) {
1741                 mutex_unlock(&adapter->rwi_lock);
1742                 ibmvnic_close(netdev);
1743                 return;
1744         }
1745
1746         rwi->reset_reason = reason;
1747         list_add_tail(&rwi->list, &adapter->rwi_list);
1748         mutex_unlock(&adapter->rwi_lock);
1749
1750         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
1751         schedule_work(&adapter->ibmvnic_reset);
1752 }
1753
1754 static void ibmvnic_tx_timeout(struct net_device *dev)
1755 {
1756         struct ibmvnic_adapter *adapter = netdev_priv(dev);
1757
1758         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
1759 }
1760
1761 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
1762                                   struct ibmvnic_rx_buff *rx_buff)
1763 {
1764         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
1765
1766         rx_buff->skb = NULL;
1767
1768         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
1769         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
1770
1771         atomic_dec(&pool->available);
1772 }
1773
1774 static int ibmvnic_poll(struct napi_struct *napi, int budget)
1775 {
1776         struct net_device *netdev = napi->dev;
1777         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1778         int scrq_num = (int)(napi - adapter->napi);
1779         int frames_processed = 0;
1780
1781 restart_poll:
1782         while (frames_processed < budget) {
1783                 struct sk_buff *skb;
1784                 struct ibmvnic_rx_buff *rx_buff;
1785                 union sub_crq *next;
1786                 u32 length;
1787                 u16 offset;
1788                 u8 flags = 0;
1789
1790                 if (unlikely(adapter->resetting)) {
1791                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1792                         napi_complete_done(napi, frames_processed);
1793                         return frames_processed;
1794                 }
1795
1796                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
1797                         break;
1798                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
1799                 rx_buff =
1800                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
1801                                                           rx_comp.correlator);
1802                 /* do error checking */
1803                 if (next->rx_comp.rc) {
1804                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
1805                                    be16_to_cpu(next->rx_comp.rc));
1806                         /* free the entry */
1807                         next->rx_comp.first = 0;
1808                         remove_buff_from_pool(adapter, rx_buff);
1809                         continue;
1810                 }
1811
1812                 length = be32_to_cpu(next->rx_comp.len);
1813                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
1814                 flags = next->rx_comp.flags;
1815                 skb = rx_buff->skb;
1816                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
1817                                         length);
1818
1819                 /* VLAN Header has been stripped by the system firmware and
1820                  * needs to be inserted by the driver
1821                  */
1822                 if (adapter->rx_vlan_header_insertion &&
1823                     (flags & IBMVNIC_VLAN_STRIPPED))
1824                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
1825                                                ntohs(next->rx_comp.vlan_tci));
1826
1827                 /* free the entry */
1828                 next->rx_comp.first = 0;
1829                 remove_buff_from_pool(adapter, rx_buff);
1830
1831                 skb_put(skb, length);
1832                 skb->protocol = eth_type_trans(skb, netdev);
1833                 skb_record_rx_queue(skb, scrq_num);
1834
1835                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
1836                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
1837                         skb->ip_summed = CHECKSUM_UNNECESSARY;
1838                 }
1839
1840                 length = skb->len;
1841                 napi_gro_receive(napi, skb); /* send it up */
1842                 netdev->stats.rx_packets++;
1843                 netdev->stats.rx_bytes += length;
1844                 adapter->rx_stats_buffers[scrq_num].packets++;
1845                 adapter->rx_stats_buffers[scrq_num].bytes += length;
1846                 frames_processed++;
1847         }
1848
1849         if (adapter->state != VNIC_CLOSING)
1850                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
1851
1852         if (frames_processed < budget) {
1853                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1854                 napi_complete_done(napi, frames_processed);
1855                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
1856                     napi_reschedule(napi)) {
1857                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1858                         goto restart_poll;
1859                 }
1860         }
1861         return frames_processed;
1862 }
1863
1864 #ifdef CONFIG_NET_POLL_CONTROLLER
1865 static void ibmvnic_netpoll_controller(struct net_device *dev)
1866 {
1867         struct ibmvnic_adapter *adapter = netdev_priv(dev);
1868         int i;
1869
1870         replenish_pools(netdev_priv(dev));
1871         for (i = 0; i < adapter->req_rx_queues; i++)
1872                 ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
1873                                      adapter->rx_scrq[i]);
1874 }
1875 #endif
1876
1877 static int wait_for_reset(struct ibmvnic_adapter *adapter)
1878 {
1879         adapter->fallback.mtu = adapter->req_mtu;
1880         adapter->fallback.rx_queues = adapter->req_rx_queues;
1881         adapter->fallback.tx_queues = adapter->req_tx_queues;
1882         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
1883         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
1884
1885         init_completion(&adapter->reset_done);
1886         ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
1887         adapter->wait_for_reset = true;
1888         wait_for_completion(&adapter->reset_done);
1889
1890         if (adapter->reset_done_rc) {
1891                 adapter->desired.mtu = adapter->fallback.mtu;
1892                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
1893                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
1894                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
1895                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
1896
1897                 init_completion(&adapter->reset_done);
1898                 ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
1899                 wait_for_completion(&adapter->reset_done);
1900         }
1901         adapter->wait_for_reset = false;
1902
1903         return adapter->reset_done_rc;
1904 }
1905
1906 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
1907 {
1908         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1909
1910         adapter->desired.mtu = new_mtu + ETH_HLEN;
1911
1912         return wait_for_reset(adapter);
1913 }
1914
1915 static const struct net_device_ops ibmvnic_netdev_ops = {
1916         .ndo_open               = ibmvnic_open,
1917         .ndo_stop               = ibmvnic_close,
1918         .ndo_start_xmit         = ibmvnic_xmit,
1919         .ndo_set_rx_mode        = ibmvnic_set_multi,
1920         .ndo_set_mac_address    = ibmvnic_set_mac,
1921         .ndo_validate_addr      = eth_validate_addr,
1922         .ndo_tx_timeout         = ibmvnic_tx_timeout,
1923 #ifdef CONFIG_NET_POLL_CONTROLLER
1924         .ndo_poll_controller    = ibmvnic_netpoll_controller,
1925 #endif
1926         .ndo_change_mtu         = ibmvnic_change_mtu,
1927 };
1928
1929 /* ethtool functions */
1930
1931 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
1932                                       struct ethtool_link_ksettings *cmd)
1933 {
1934         u32 supported, advertising;
1935
1936         supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
1937                           SUPPORTED_FIBRE);
1938         advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
1939                             ADVERTISED_FIBRE);
1940         cmd->base.speed = SPEED_1000;
1941         cmd->base.duplex = DUPLEX_FULL;
1942         cmd->base.port = PORT_FIBRE;
1943         cmd->base.phy_address = 0;
1944         cmd->base.autoneg = AUTONEG_ENABLE;
1945
1946         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
1947                                                 supported);
1948         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
1949                                                 advertising);
1950
1951         return 0;
1952 }
1953
1954 static void ibmvnic_get_drvinfo(struct net_device *netdev,
1955                                 struct ethtool_drvinfo *info)
1956 {
1957         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1958
1959         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
1960         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
1961         strlcpy(info->fw_version, adapter->fw_version,
1962                 sizeof(info->fw_version));
1963 }
1964
1965 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
1966 {
1967         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1968
1969         return adapter->msg_enable;
1970 }
1971
1972 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
1973 {
1974         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1975
1976         adapter->msg_enable = data;
1977 }
1978
1979 static u32 ibmvnic_get_link(struct net_device *netdev)
1980 {
1981         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1982
1983         /* Don't need to send a query because we request a logical link up at
1984          * init and then we wait for link state indications
1985          */
1986         return adapter->logical_link_state;
1987 }
1988
1989 static void ibmvnic_get_ringparam(struct net_device *netdev,
1990                                   struct ethtool_ringparam *ring)
1991 {
1992         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1993
1994         ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
1995         ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
1996         ring->rx_mini_max_pending = 0;
1997         ring->rx_jumbo_max_pending = 0;
1998         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
1999         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2000         ring->rx_mini_pending = 0;
2001         ring->rx_jumbo_pending = 0;
2002 }
2003
2004 static int ibmvnic_set_ringparam(struct net_device *netdev,
2005                                  struct ethtool_ringparam *ring)
2006 {
2007         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2008
2009         if (ring->rx_pending > adapter->max_rx_add_entries_per_subcrq  ||
2010             ring->tx_pending > adapter->max_tx_entries_per_subcrq) {
2011                 netdev_err(netdev, "Invalid request.\n");
2012                 netdev_err(netdev, "Max tx buffers = %llu\n",
2013                            adapter->max_rx_add_entries_per_subcrq);
2014                 netdev_err(netdev, "Max rx buffers = %llu\n",
2015                            adapter->max_tx_entries_per_subcrq);
2016                 return -EINVAL;
2017         }
2018
2019         adapter->desired.rx_entries = ring->rx_pending;
2020         adapter->desired.tx_entries = ring->tx_pending;
2021
2022         return wait_for_reset(adapter);
2023 }
2024
2025 static void ibmvnic_get_channels(struct net_device *netdev,
2026                                  struct ethtool_channels *channels)
2027 {
2028         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2029
2030         channels->max_rx = adapter->max_rx_queues;
2031         channels->max_tx = adapter->max_tx_queues;
2032         channels->max_other = 0;
2033         channels->max_combined = 0;
2034         channels->rx_count = adapter->req_rx_queues;
2035         channels->tx_count = adapter->req_tx_queues;
2036         channels->other_count = 0;
2037         channels->combined_count = 0;
2038 }
2039
2040 static int ibmvnic_set_channels(struct net_device *netdev,
2041                                 struct ethtool_channels *channels)
2042 {
2043         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2044
2045         adapter->desired.rx_queues = channels->rx_count;
2046         adapter->desired.tx_queues = channels->tx_count;
2047
2048         return wait_for_reset(adapter);
2049 }
2050
2051 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2052 {
2053         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2054         int i;
2055
2056         if (stringset != ETH_SS_STATS)
2057                 return;
2058
2059         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
2060                 memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2061
2062         for (i = 0; i < adapter->req_tx_queues; i++) {
2063                 snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2064                 data += ETH_GSTRING_LEN;
2065
2066                 snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2067                 data += ETH_GSTRING_LEN;
2068
2069                 snprintf(data, ETH_GSTRING_LEN, "tx%d_dropped_packets", i);
2070                 data += ETH_GSTRING_LEN;
2071         }
2072
2073         for (i = 0; i < adapter->req_rx_queues; i++) {
2074                 snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2075                 data += ETH_GSTRING_LEN;
2076
2077                 snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2078                 data += ETH_GSTRING_LEN;
2079
2080                 snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2081                 data += ETH_GSTRING_LEN;
2082         }
2083 }
2084
2085 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2086 {
2087         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2088
2089         switch (sset) {
2090         case ETH_SS_STATS:
2091                 return ARRAY_SIZE(ibmvnic_stats) +
2092                        adapter->req_tx_queues * NUM_TX_STATS +
2093                        adapter->req_rx_queues * NUM_RX_STATS;
2094         default:
2095                 return -EOPNOTSUPP;
2096         }
2097 }
2098
2099 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2100                                       struct ethtool_stats *stats, u64 *data)
2101 {
2102         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2103         union ibmvnic_crq crq;
2104         int i, j;
2105
2106         memset(&crq, 0, sizeof(crq));
2107         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2108         crq.request_statistics.cmd = REQUEST_STATISTICS;
2109         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2110         crq.request_statistics.len =
2111             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2112
2113         /* Wait for data to be written */
2114         init_completion(&adapter->stats_done);
2115         ibmvnic_send_crq(adapter, &crq);
2116         wait_for_completion(&adapter->stats_done);
2117
2118         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2119                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2120                                                 ibmvnic_stats[i].offset));
2121
2122         for (j = 0; j < adapter->req_tx_queues; j++) {
2123                 data[i] = adapter->tx_stats_buffers[j].packets;
2124                 i++;
2125                 data[i] = adapter->tx_stats_buffers[j].bytes;
2126                 i++;
2127                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2128                 i++;
2129         }
2130
2131         for (j = 0; j < adapter->req_rx_queues; j++) {
2132                 data[i] = adapter->rx_stats_buffers[j].packets;
2133                 i++;
2134                 data[i] = adapter->rx_stats_buffers[j].bytes;
2135                 i++;
2136                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2137                 i++;
2138         }
2139 }
2140
2141 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2142         .get_drvinfo            = ibmvnic_get_drvinfo,
2143         .get_msglevel           = ibmvnic_get_msglevel,
2144         .set_msglevel           = ibmvnic_set_msglevel,
2145         .get_link               = ibmvnic_get_link,
2146         .get_ringparam          = ibmvnic_get_ringparam,
2147         .set_ringparam          = ibmvnic_set_ringparam,
2148         .get_channels           = ibmvnic_get_channels,
2149         .set_channels           = ibmvnic_set_channels,
2150         .get_strings            = ibmvnic_get_strings,
2151         .get_sset_count         = ibmvnic_get_sset_count,
2152         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2153         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2154 };
2155
2156 /* Routines for managing CRQs/sCRQs  */
2157
2158 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2159                                    struct ibmvnic_sub_crq_queue *scrq)
2160 {
2161         int rc;
2162
2163         if (scrq->irq) {
2164                 free_irq(scrq->irq, scrq);
2165                 irq_dispose_mapping(scrq->irq);
2166                 scrq->irq = 0;
2167         }
2168
2169         memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2170         scrq->cur = 0;
2171
2172         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2173                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2174         return rc;
2175 }
2176
2177 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2178 {
2179         int i, rc;
2180
2181         for (i = 0; i < adapter->req_tx_queues; i++) {
2182                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2183                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2184                 if (rc)
2185                         return rc;
2186         }
2187
2188         for (i = 0; i < adapter->req_rx_queues; i++) {
2189                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2190                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2191                 if (rc)
2192                         return rc;
2193         }
2194
2195         return rc;
2196 }
2197
2198 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2199                                   struct ibmvnic_sub_crq_queue *scrq)
2200 {
2201         struct device *dev = &adapter->vdev->dev;
2202         long rc;
2203
2204         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2205
2206         /* Close the sub-crqs */
2207         do {
2208                 rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2209                                         adapter->vdev->unit_address,
2210                                         scrq->crq_num);
2211         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2212
2213         if (rc) {
2214                 netdev_err(adapter->netdev,
2215                            "Failed to release sub-CRQ %16lx, rc = %ld\n",
2216                            scrq->crq_num, rc);
2217         }
2218
2219         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2220                          DMA_BIDIRECTIONAL);
2221         free_pages((unsigned long)scrq->msgs, 2);
2222         kfree(scrq);
2223 }
2224
2225 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2226                                                         *adapter)
2227 {
2228         struct device *dev = &adapter->vdev->dev;
2229         struct ibmvnic_sub_crq_queue *scrq;
2230         int rc;
2231
2232         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2233         if (!scrq)
2234                 return NULL;
2235
2236         scrq->msgs =
2237                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2238         if (!scrq->msgs) {
2239                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2240                 goto zero_page_failed;
2241         }
2242
2243         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2244                                          DMA_BIDIRECTIONAL);
2245         if (dma_mapping_error(dev, scrq->msg_token)) {
2246                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2247                 goto map_failed;
2248         }
2249
2250         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2251                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2252
2253         if (rc == H_RESOURCE)
2254                 rc = ibmvnic_reset_crq(adapter);
2255
2256         if (rc == H_CLOSED) {
2257                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2258         } else if (rc) {
2259                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2260                 goto reg_failed;
2261         }
2262
2263         scrq->adapter = adapter;
2264         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2265         spin_lock_init(&scrq->lock);
2266
2267         netdev_dbg(adapter->netdev,
2268                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2269                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2270
2271         return scrq;
2272
2273 reg_failed:
2274         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2275                          DMA_BIDIRECTIONAL);
2276 map_failed:
2277         free_pages((unsigned long)scrq->msgs, 2);
2278 zero_page_failed:
2279         kfree(scrq);
2280
2281         return NULL;
2282 }
2283
2284 static void release_sub_crqs(struct ibmvnic_adapter *adapter)
2285 {
2286         int i;
2287
2288         if (adapter->tx_scrq) {
2289                 for (i = 0; i < adapter->req_tx_queues; i++) {
2290                         if (!adapter->tx_scrq[i])
2291                                 continue;
2292
2293                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2294                                    i);
2295                         if (adapter->tx_scrq[i]->irq) {
2296                                 free_irq(adapter->tx_scrq[i]->irq,
2297                                          adapter->tx_scrq[i]);
2298                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2299                                 adapter->tx_scrq[i]->irq = 0;
2300                         }
2301
2302                         release_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2303                 }
2304
2305                 kfree(adapter->tx_scrq);
2306                 adapter->tx_scrq = NULL;
2307         }
2308
2309         if (adapter->rx_scrq) {
2310                 for (i = 0; i < adapter->req_rx_queues; i++) {
2311                         if (!adapter->rx_scrq[i])
2312                                 continue;
2313
2314                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2315                                    i);
2316                         if (adapter->rx_scrq[i]->irq) {
2317                                 free_irq(adapter->rx_scrq[i]->irq,
2318                                          adapter->rx_scrq[i]);
2319                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2320                                 adapter->rx_scrq[i]->irq = 0;
2321                         }
2322
2323                         release_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2324                 }
2325
2326                 kfree(adapter->rx_scrq);
2327                 adapter->rx_scrq = NULL;
2328         }
2329 }
2330
2331 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2332                             struct ibmvnic_sub_crq_queue *scrq)
2333 {
2334         struct device *dev = &adapter->vdev->dev;
2335         unsigned long rc;
2336
2337         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2338                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2339         if (rc)
2340                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2341                         scrq->hw_irq, rc);
2342         return rc;
2343 }
2344
2345 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2346                            struct ibmvnic_sub_crq_queue *scrq)
2347 {
2348         struct device *dev = &adapter->vdev->dev;
2349         unsigned long rc;
2350
2351         if (scrq->hw_irq > 0x100000000ULL) {
2352                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2353                 return 1;
2354         }
2355
2356         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2357                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2358         if (rc)
2359                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2360                         scrq->hw_irq, rc);
2361         return rc;
2362 }
2363
2364 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2365                                struct ibmvnic_sub_crq_queue *scrq)
2366 {
2367         struct device *dev = &adapter->vdev->dev;
2368         struct ibmvnic_tx_buff *txbuff;
2369         union sub_crq *next;
2370         int index;
2371         int i, j;
2372         u8 first;
2373
2374 restart_loop:
2375         while (pending_scrq(adapter, scrq)) {
2376                 unsigned int pool = scrq->pool_index;
2377
2378                 next = ibmvnic_next_scrq(adapter, scrq);
2379                 for (i = 0; i < next->tx_comp.num_comps; i++) {
2380                         if (next->tx_comp.rcs[i]) {
2381                                 dev_err(dev, "tx error %x\n",
2382                                         next->tx_comp.rcs[i]);
2383                                 continue;
2384                         }
2385                         index = be32_to_cpu(next->tx_comp.correlators[i]);
2386                         txbuff = &adapter->tx_pool[pool].tx_buff[index];
2387
2388                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2389                                 if (!txbuff->data_dma[j])
2390                                         continue;
2391
2392                                 txbuff->data_dma[j] = 0;
2393                         }
2394                         /* if sub_crq was sent indirectly */
2395                         first = txbuff->indir_arr[0].generic.first;
2396                         if (first == IBMVNIC_CRQ_CMD) {
2397                                 dma_unmap_single(dev, txbuff->indir_dma,
2398                                                  sizeof(txbuff->indir_arr),
2399                                                  DMA_TO_DEVICE);
2400                         }
2401
2402                         if (txbuff->last_frag) {
2403                                 dev_kfree_skb_any(txbuff->skb);
2404                                 txbuff->skb = NULL;
2405                         }
2406
2407                         adapter->tx_pool[pool].free_map[adapter->tx_pool[pool].
2408                                                      producer_index] = index;
2409                         adapter->tx_pool[pool].producer_index =
2410                             (adapter->tx_pool[pool].producer_index + 1) %
2411                             adapter->req_tx_entries_per_subcrq;
2412                 }
2413                 /* remove tx_comp scrq*/
2414                 next->tx_comp.first = 0;
2415
2416                 if (atomic_sub_return(next->tx_comp.num_comps, &scrq->used) <=
2417                     (adapter->req_tx_entries_per_subcrq / 2) &&
2418                     __netif_subqueue_stopped(adapter->netdev,
2419                                              scrq->pool_index)) {
2420                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2421                         netdev_info(adapter->netdev, "Started queue %d\n",
2422                                     scrq->pool_index);
2423                 }
2424         }
2425
2426         enable_scrq_irq(adapter, scrq);
2427
2428         if (pending_scrq(adapter, scrq)) {
2429                 disable_scrq_irq(adapter, scrq);
2430                 goto restart_loop;
2431         }
2432
2433         return 0;
2434 }
2435
2436 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2437 {
2438         struct ibmvnic_sub_crq_queue *scrq = instance;
2439         struct ibmvnic_adapter *adapter = scrq->adapter;
2440
2441         disable_scrq_irq(adapter, scrq);
2442         ibmvnic_complete_tx(adapter, scrq);
2443
2444         return IRQ_HANDLED;
2445 }
2446
2447 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2448 {
2449         struct ibmvnic_sub_crq_queue *scrq = instance;
2450         struct ibmvnic_adapter *adapter = scrq->adapter;
2451
2452         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2453
2454         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2455                 disable_scrq_irq(adapter, scrq);
2456                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
2457         }
2458
2459         return IRQ_HANDLED;
2460 }
2461
2462 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2463 {
2464         struct device *dev = &adapter->vdev->dev;
2465         struct ibmvnic_sub_crq_queue *scrq;
2466         int i = 0, j = 0;
2467         int rc = 0;
2468
2469         for (i = 0; i < adapter->req_tx_queues; i++) {
2470                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2471                            i);
2472                 scrq = adapter->tx_scrq[i];
2473                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2474
2475                 if (!scrq->irq) {
2476                         rc = -EINVAL;
2477                         dev_err(dev, "Error mapping irq\n");
2478                         goto req_tx_irq_failed;
2479                 }
2480
2481                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2482                                  0, "ibmvnic_tx", scrq);
2483
2484                 if (rc) {
2485                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2486                                 scrq->irq, rc);
2487                         irq_dispose_mapping(scrq->irq);
2488                         goto req_rx_irq_failed;
2489                 }
2490         }
2491
2492         for (i = 0; i < adapter->req_rx_queues; i++) {
2493                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2494                            i);
2495                 scrq = adapter->rx_scrq[i];
2496                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2497                 if (!scrq->irq) {
2498                         rc = -EINVAL;
2499                         dev_err(dev, "Error mapping irq\n");
2500                         goto req_rx_irq_failed;
2501                 }
2502                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2503                                  0, "ibmvnic_rx", scrq);
2504                 if (rc) {
2505                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2506                                 scrq->irq, rc);
2507                         irq_dispose_mapping(scrq->irq);
2508                         goto req_rx_irq_failed;
2509                 }
2510         }
2511         return rc;
2512
2513 req_rx_irq_failed:
2514         for (j = 0; j < i; j++) {
2515                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2516                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2517         }
2518         i = adapter->req_tx_queues;
2519 req_tx_irq_failed:
2520         for (j = 0; j < i; j++) {
2521                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2522                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2523         }
2524         release_sub_crqs(adapter);
2525         return rc;
2526 }
2527
2528 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2529 {
2530         struct device *dev = &adapter->vdev->dev;
2531         struct ibmvnic_sub_crq_queue **allqueues;
2532         int registered_queues = 0;
2533         int total_queues;
2534         int more = 0;
2535         int i;
2536
2537         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2538
2539         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2540         if (!allqueues)
2541                 return -1;
2542
2543         for (i = 0; i < total_queues; i++) {
2544                 allqueues[i] = init_sub_crq_queue(adapter);
2545                 if (!allqueues[i]) {
2546                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2547                         break;
2548                 }
2549                 registered_queues++;
2550         }
2551
2552         /* Make sure we were able to register the minimum number of queues */
2553         if (registered_queues <
2554             adapter->min_tx_queues + adapter->min_rx_queues) {
2555                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2556                 goto tx_failed;
2557         }
2558
2559         /* Distribute the failed allocated queues*/
2560         for (i = 0; i < total_queues - registered_queues + more ; i++) {
2561                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2562                 switch (i % 3) {
2563                 case 0:
2564                         if (adapter->req_rx_queues > adapter->min_rx_queues)
2565                                 adapter->req_rx_queues--;
2566                         else
2567                                 more++;
2568                         break;
2569                 case 1:
2570                         if (adapter->req_tx_queues > adapter->min_tx_queues)
2571                                 adapter->req_tx_queues--;
2572                         else
2573                                 more++;
2574                         break;
2575                 }
2576         }
2577
2578         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
2579                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
2580         if (!adapter->tx_scrq)
2581                 goto tx_failed;
2582
2583         for (i = 0; i < adapter->req_tx_queues; i++) {
2584                 adapter->tx_scrq[i] = allqueues[i];
2585                 adapter->tx_scrq[i]->pool_index = i;
2586         }
2587
2588         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
2589                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
2590         if (!adapter->rx_scrq)
2591                 goto rx_failed;
2592
2593         for (i = 0; i < adapter->req_rx_queues; i++) {
2594                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
2595                 adapter->rx_scrq[i]->scrq_num = i;
2596         }
2597
2598         kfree(allqueues);
2599         return 0;
2600
2601 rx_failed:
2602         kfree(adapter->tx_scrq);
2603         adapter->tx_scrq = NULL;
2604 tx_failed:
2605         for (i = 0; i < registered_queues; i++)
2606                 release_sub_crq_queue(adapter, allqueues[i]);
2607         kfree(allqueues);
2608         return -1;
2609 }
2610
2611 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
2612 {
2613         struct device *dev = &adapter->vdev->dev;
2614         union ibmvnic_crq crq;
2615         int max_entries;
2616
2617         if (!retry) {
2618                 /* Sub-CRQ entries are 32 byte long */
2619                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
2620
2621                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
2622                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
2623                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
2624                         return;
2625                 }
2626
2627                 if (adapter->desired.mtu)
2628                         adapter->req_mtu = adapter->desired.mtu;
2629                 else
2630                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
2631
2632                 if (!adapter->desired.tx_entries)
2633                         adapter->desired.tx_entries =
2634                                         adapter->max_tx_entries_per_subcrq;
2635                 if (!adapter->desired.rx_entries)
2636                         adapter->desired.rx_entries =
2637                                         adapter->max_rx_add_entries_per_subcrq;
2638
2639                 max_entries = IBMVNIC_MAX_LTB_SIZE /
2640                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
2641
2642                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
2643                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
2644                         adapter->desired.tx_entries = max_entries;
2645                 }
2646
2647                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
2648                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
2649                         adapter->desired.rx_entries = max_entries;
2650                 }
2651
2652                 if (adapter->desired.tx_entries)
2653                         adapter->req_tx_entries_per_subcrq =
2654                                         adapter->desired.tx_entries;
2655                 else
2656                         adapter->req_tx_entries_per_subcrq =
2657                                         adapter->max_tx_entries_per_subcrq;
2658
2659                 if (adapter->desired.rx_entries)
2660                         adapter->req_rx_add_entries_per_subcrq =
2661                                         adapter->desired.rx_entries;
2662                 else
2663                         adapter->req_rx_add_entries_per_subcrq =
2664                                         adapter->max_rx_add_entries_per_subcrq;
2665
2666                 if (adapter->desired.tx_queues)
2667                         adapter->req_tx_queues =
2668                                         adapter->desired.tx_queues;
2669                 else
2670                         adapter->req_tx_queues =
2671                                         adapter->opt_tx_comp_sub_queues;
2672
2673                 if (adapter->desired.rx_queues)
2674                         adapter->req_rx_queues =
2675                                         adapter->desired.rx_queues;
2676                 else
2677                         adapter->req_rx_queues =
2678                                         adapter->opt_rx_comp_queues;
2679
2680                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
2681         }
2682
2683         memset(&crq, 0, sizeof(crq));
2684         crq.request_capability.first = IBMVNIC_CRQ_CMD;
2685         crq.request_capability.cmd = REQUEST_CAPABILITY;
2686
2687         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
2688         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
2689         atomic_inc(&adapter->running_cap_crqs);
2690         ibmvnic_send_crq(adapter, &crq);
2691
2692         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
2693         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
2694         atomic_inc(&adapter->running_cap_crqs);
2695         ibmvnic_send_crq(adapter, &crq);
2696
2697         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
2698         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
2699         atomic_inc(&adapter->running_cap_crqs);
2700         ibmvnic_send_crq(adapter, &crq);
2701
2702         crq.request_capability.capability =
2703             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
2704         crq.request_capability.number =
2705             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
2706         atomic_inc(&adapter->running_cap_crqs);
2707         ibmvnic_send_crq(adapter, &crq);
2708
2709         crq.request_capability.capability =
2710             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
2711         crq.request_capability.number =
2712             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
2713         atomic_inc(&adapter->running_cap_crqs);
2714         ibmvnic_send_crq(adapter, &crq);
2715
2716         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
2717         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
2718         atomic_inc(&adapter->running_cap_crqs);
2719         ibmvnic_send_crq(adapter, &crq);
2720
2721         if (adapter->netdev->flags & IFF_PROMISC) {
2722                 if (adapter->promisc_supported) {
2723                         crq.request_capability.capability =
2724                             cpu_to_be16(PROMISC_REQUESTED);
2725                         crq.request_capability.number = cpu_to_be64(1);
2726                         atomic_inc(&adapter->running_cap_crqs);
2727                         ibmvnic_send_crq(adapter, &crq);
2728                 }
2729         } else {
2730                 crq.request_capability.capability =
2731                     cpu_to_be16(PROMISC_REQUESTED);
2732                 crq.request_capability.number = cpu_to_be64(0);
2733                 atomic_inc(&adapter->running_cap_crqs);
2734                 ibmvnic_send_crq(adapter, &crq);
2735         }
2736 }
2737
2738 static int pending_scrq(struct ibmvnic_adapter *adapter,
2739                         struct ibmvnic_sub_crq_queue *scrq)
2740 {
2741         union sub_crq *entry = &scrq->msgs[scrq->cur];
2742
2743         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
2744                 return 1;
2745         else
2746                 return 0;
2747 }
2748
2749 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
2750                                         struct ibmvnic_sub_crq_queue *scrq)
2751 {
2752         union sub_crq *entry;
2753         unsigned long flags;
2754
2755         spin_lock_irqsave(&scrq->lock, flags);
2756         entry = &scrq->msgs[scrq->cur];
2757         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2758                 if (++scrq->cur == scrq->size)
2759                         scrq->cur = 0;
2760         } else {
2761                 entry = NULL;
2762         }
2763         spin_unlock_irqrestore(&scrq->lock, flags);
2764
2765         return entry;
2766 }
2767
2768 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
2769 {
2770         struct ibmvnic_crq_queue *queue = &adapter->crq;
2771         union ibmvnic_crq *crq;
2772
2773         crq = &queue->msgs[queue->cur];
2774         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2775                 if (++queue->cur == queue->size)
2776                         queue->cur = 0;
2777         } else {
2778                 crq = NULL;
2779         }
2780
2781         return crq;
2782 }
2783
2784 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
2785                        union sub_crq *sub_crq)
2786 {
2787         unsigned int ua = adapter->vdev->unit_address;
2788         struct device *dev = &adapter->vdev->dev;
2789         u64 *u64_crq = (u64 *)sub_crq;
2790         int rc;
2791
2792         netdev_dbg(adapter->netdev,
2793                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
2794                    (unsigned long int)cpu_to_be64(remote_handle),
2795                    (unsigned long int)cpu_to_be64(u64_crq[0]),
2796                    (unsigned long int)cpu_to_be64(u64_crq[1]),
2797                    (unsigned long int)cpu_to_be64(u64_crq[2]),
2798                    (unsigned long int)cpu_to_be64(u64_crq[3]));
2799
2800         /* Make sure the hypervisor sees the complete request */
2801         mb();
2802
2803         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
2804                                 cpu_to_be64(remote_handle),
2805                                 cpu_to_be64(u64_crq[0]),
2806                                 cpu_to_be64(u64_crq[1]),
2807                                 cpu_to_be64(u64_crq[2]),
2808                                 cpu_to_be64(u64_crq[3]));
2809
2810         if (rc) {
2811                 if (rc == H_CLOSED)
2812                         dev_warn(dev, "CRQ Queue closed\n");
2813                 dev_err(dev, "Send error (rc=%d)\n", rc);
2814         }
2815
2816         return rc;
2817 }
2818
2819 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
2820                                 u64 remote_handle, u64 ioba, u64 num_entries)
2821 {
2822         unsigned int ua = adapter->vdev->unit_address;
2823         struct device *dev = &adapter->vdev->dev;
2824         int rc;
2825
2826         /* Make sure the hypervisor sees the complete request */
2827         mb();
2828         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
2829                                 cpu_to_be64(remote_handle),
2830                                 ioba, num_entries);
2831
2832         if (rc) {
2833                 if (rc == H_CLOSED)
2834                         dev_warn(dev, "CRQ Queue closed\n");
2835                 dev_err(dev, "Send (indirect) error (rc=%d)\n", rc);
2836         }
2837
2838         return rc;
2839 }
2840
2841 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
2842                             union ibmvnic_crq *crq)
2843 {
2844         unsigned int ua = adapter->vdev->unit_address;
2845         struct device *dev = &adapter->vdev->dev;
2846         u64 *u64_crq = (u64 *)crq;
2847         int rc;
2848
2849         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
2850                    (unsigned long int)cpu_to_be64(u64_crq[0]),
2851                    (unsigned long int)cpu_to_be64(u64_crq[1]));
2852
2853         /* Make sure the hypervisor sees the complete request */
2854         mb();
2855
2856         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
2857                                 cpu_to_be64(u64_crq[0]),
2858                                 cpu_to_be64(u64_crq[1]));
2859
2860         if (rc) {
2861                 if (rc == H_CLOSED)
2862                         dev_warn(dev, "CRQ Queue closed\n");
2863                 dev_warn(dev, "Send error (rc=%d)\n", rc);
2864         }
2865
2866         return rc;
2867 }
2868
2869 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
2870 {
2871         union ibmvnic_crq crq;
2872
2873         memset(&crq, 0, sizeof(crq));
2874         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
2875         crq.generic.cmd = IBMVNIC_CRQ_INIT;
2876         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
2877
2878         return ibmvnic_send_crq(adapter, &crq);
2879 }
2880
2881 static int send_version_xchg(struct ibmvnic_adapter *adapter)
2882 {
2883         union ibmvnic_crq crq;
2884
2885         memset(&crq, 0, sizeof(crq));
2886         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
2887         crq.version_exchange.cmd = VERSION_EXCHANGE;
2888         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
2889
2890         return ibmvnic_send_crq(adapter, &crq);
2891 }
2892
2893 struct vnic_login_client_data {
2894         u8      type;
2895         __be16  len;
2896         char    name;
2897 } __packed;
2898
2899 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
2900 {
2901         int len;
2902
2903         /* Calculate the amount of buffer space needed for the
2904          * vnic client data in the login buffer. There are four entries,
2905          * OS name, LPAR name, device name, and a null last entry.
2906          */
2907         len = 4 * sizeof(struct vnic_login_client_data);
2908         len += 6; /* "Linux" plus NULL */
2909         len += strlen(utsname()->nodename) + 1;
2910         len += strlen(adapter->netdev->name) + 1;
2911
2912         return len;
2913 }
2914
2915 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
2916                                  struct vnic_login_client_data *vlcd)
2917 {
2918         const char *os_name = "Linux";
2919         int len;
2920
2921         /* Type 1 - LPAR OS */
2922         vlcd->type = 1;
2923         len = strlen(os_name) + 1;
2924         vlcd->len = cpu_to_be16(len);
2925         strncpy(&vlcd->name, os_name, len);
2926         vlcd = (struct vnic_login_client_data *)((char *)&vlcd->name + len);
2927
2928         /* Type 2 - LPAR name */
2929         vlcd->type = 2;
2930         len = strlen(utsname()->nodename) + 1;
2931         vlcd->len = cpu_to_be16(len);
2932         strncpy(&vlcd->name, utsname()->nodename, len);
2933         vlcd = (struct vnic_login_client_data *)((char *)&vlcd->name + len);
2934
2935         /* Type 3 - device name */
2936         vlcd->type = 3;
2937         len = strlen(adapter->netdev->name) + 1;
2938         vlcd->len = cpu_to_be16(len);
2939         strncpy(&vlcd->name, adapter->netdev->name, len);
2940 }
2941
2942 static void send_login(struct ibmvnic_adapter *adapter)
2943 {
2944         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
2945         struct ibmvnic_login_buffer *login_buffer;
2946         struct device *dev = &adapter->vdev->dev;
2947         dma_addr_t rsp_buffer_token;
2948         dma_addr_t buffer_token;
2949         size_t rsp_buffer_size;
2950         union ibmvnic_crq crq;
2951         size_t buffer_size;
2952         __be64 *tx_list_p;
2953         __be64 *rx_list_p;
2954         int client_data_len;
2955         struct vnic_login_client_data *vlcd;
2956         int i;
2957
2958         client_data_len = vnic_client_data_len(adapter);
2959
2960         buffer_size =
2961             sizeof(struct ibmvnic_login_buffer) +
2962             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
2963             client_data_len;
2964
2965         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
2966         if (!login_buffer)
2967                 goto buf_alloc_failed;
2968
2969         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
2970                                       DMA_TO_DEVICE);
2971         if (dma_mapping_error(dev, buffer_token)) {
2972                 dev_err(dev, "Couldn't map login buffer\n");
2973                 goto buf_map_failed;
2974         }
2975
2976         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
2977                           sizeof(u64) * adapter->req_tx_queues +
2978                           sizeof(u64) * adapter->req_rx_queues +
2979                           sizeof(u64) * adapter->req_rx_queues +
2980                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
2981
2982         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
2983         if (!login_rsp_buffer)
2984                 goto buf_rsp_alloc_failed;
2985
2986         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
2987                                           rsp_buffer_size, DMA_FROM_DEVICE);
2988         if (dma_mapping_error(dev, rsp_buffer_token)) {
2989                 dev_err(dev, "Couldn't map login rsp buffer\n");
2990                 goto buf_rsp_map_failed;
2991         }
2992
2993         adapter->login_buf = login_buffer;
2994         adapter->login_buf_token = buffer_token;
2995         adapter->login_buf_sz = buffer_size;
2996         adapter->login_rsp_buf = login_rsp_buffer;
2997         adapter->login_rsp_buf_token = rsp_buffer_token;
2998         adapter->login_rsp_buf_sz = rsp_buffer_size;
2999
3000         login_buffer->len = cpu_to_be32(buffer_size);
3001         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3002         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3003         login_buffer->off_txcomp_subcrqs =
3004             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3005         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3006         login_buffer->off_rxcomp_subcrqs =
3007             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3008                         sizeof(u64) * adapter->req_tx_queues);
3009         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3010         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3011
3012         tx_list_p = (__be64 *)((char *)login_buffer +
3013                                       sizeof(struct ibmvnic_login_buffer));
3014         rx_list_p = (__be64 *)((char *)login_buffer +
3015                                       sizeof(struct ibmvnic_login_buffer) +
3016                                       sizeof(u64) * adapter->req_tx_queues);
3017
3018         for (i = 0; i < adapter->req_tx_queues; i++) {
3019                 if (adapter->tx_scrq[i]) {
3020                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3021                                                    crq_num);
3022                 }
3023         }
3024
3025         for (i = 0; i < adapter->req_rx_queues; i++) {
3026                 if (adapter->rx_scrq[i]) {
3027                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3028                                                    crq_num);
3029                 }
3030         }
3031
3032         /* Insert vNIC login client data */
3033         vlcd = (struct vnic_login_client_data *)
3034                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3035         login_buffer->client_data_offset =
3036                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3037         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3038
3039         vnic_add_client_data(adapter, vlcd);
3040
3041         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3042         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3043                 netdev_dbg(adapter->netdev, "%016lx\n",
3044                            ((unsigned long int *)(adapter->login_buf))[i]);
3045         }
3046
3047         memset(&crq, 0, sizeof(crq));
3048         crq.login.first = IBMVNIC_CRQ_CMD;
3049         crq.login.cmd = LOGIN;
3050         crq.login.ioba = cpu_to_be32(buffer_token);
3051         crq.login.len = cpu_to_be32(buffer_size);
3052         ibmvnic_send_crq(adapter, &crq);
3053
3054         return;
3055
3056 buf_rsp_map_failed:
3057         kfree(login_rsp_buffer);
3058 buf_rsp_alloc_failed:
3059         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3060 buf_map_failed:
3061         kfree(login_buffer);
3062 buf_alloc_failed:
3063         return;
3064 }
3065
3066 static void send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3067                              u32 len, u8 map_id)
3068 {
3069         union ibmvnic_crq crq;
3070
3071         memset(&crq, 0, sizeof(crq));
3072         crq.request_map.first = IBMVNIC_CRQ_CMD;
3073         crq.request_map.cmd = REQUEST_MAP;
3074         crq.request_map.map_id = map_id;
3075         crq.request_map.ioba = cpu_to_be32(addr);
3076         crq.request_map.len = cpu_to_be32(len);
3077         ibmvnic_send_crq(adapter, &crq);
3078 }
3079
3080 static void send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3081 {
3082         union ibmvnic_crq crq;
3083
3084         memset(&crq, 0, sizeof(crq));
3085         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3086         crq.request_unmap.cmd = REQUEST_UNMAP;
3087         crq.request_unmap.map_id = map_id;
3088         ibmvnic_send_crq(adapter, &crq);
3089 }
3090
3091 static void send_map_query(struct ibmvnic_adapter *adapter)
3092 {
3093         union ibmvnic_crq crq;
3094
3095         memset(&crq, 0, sizeof(crq));
3096         crq.query_map.first = IBMVNIC_CRQ_CMD;
3097         crq.query_map.cmd = QUERY_MAP;
3098         ibmvnic_send_crq(adapter, &crq);
3099 }
3100
3101 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3102 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3103 {
3104         union ibmvnic_crq crq;
3105
3106         atomic_set(&adapter->running_cap_crqs, 0);
3107         memset(&crq, 0, sizeof(crq));
3108         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3109         crq.query_capability.cmd = QUERY_CAPABILITY;
3110
3111         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3112         atomic_inc(&adapter->running_cap_crqs);
3113         ibmvnic_send_crq(adapter, &crq);
3114
3115         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3116         atomic_inc(&adapter->running_cap_crqs);
3117         ibmvnic_send_crq(adapter, &crq);
3118
3119         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3120         atomic_inc(&adapter->running_cap_crqs);
3121         ibmvnic_send_crq(adapter, &crq);
3122
3123         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3124         atomic_inc(&adapter->running_cap_crqs);
3125         ibmvnic_send_crq(adapter, &crq);
3126
3127         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3128         atomic_inc(&adapter->running_cap_crqs);
3129         ibmvnic_send_crq(adapter, &crq);
3130
3131         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3132         atomic_inc(&adapter->running_cap_crqs);
3133         ibmvnic_send_crq(adapter, &crq);
3134
3135         crq.query_capability.capability =
3136             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3137         atomic_inc(&adapter->running_cap_crqs);
3138         ibmvnic_send_crq(adapter, &crq);
3139
3140         crq.query_capability.capability =
3141             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3142         atomic_inc(&adapter->running_cap_crqs);
3143         ibmvnic_send_crq(adapter, &crq);
3144
3145         crq.query_capability.capability =
3146             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3147         atomic_inc(&adapter->running_cap_crqs);
3148         ibmvnic_send_crq(adapter, &crq);
3149
3150         crq.query_capability.capability =
3151             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3152         atomic_inc(&adapter->running_cap_crqs);
3153         ibmvnic_send_crq(adapter, &crq);
3154
3155         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3156         atomic_inc(&adapter->running_cap_crqs);
3157         ibmvnic_send_crq(adapter, &crq);
3158
3159         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3160         atomic_inc(&adapter->running_cap_crqs);
3161         ibmvnic_send_crq(adapter, &crq);
3162
3163         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3164         atomic_inc(&adapter->running_cap_crqs);
3165         ibmvnic_send_crq(adapter, &crq);
3166
3167         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3168         atomic_inc(&adapter->running_cap_crqs);
3169         ibmvnic_send_crq(adapter, &crq);
3170
3171         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3172         atomic_inc(&adapter->running_cap_crqs);
3173         ibmvnic_send_crq(adapter, &crq);
3174
3175         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3176         atomic_inc(&adapter->running_cap_crqs);
3177         ibmvnic_send_crq(adapter, &crq);
3178
3179         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3180         atomic_inc(&adapter->running_cap_crqs);
3181         ibmvnic_send_crq(adapter, &crq);
3182
3183         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3184         atomic_inc(&adapter->running_cap_crqs);
3185         ibmvnic_send_crq(adapter, &crq);
3186
3187         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3188         atomic_inc(&adapter->running_cap_crqs);
3189         ibmvnic_send_crq(adapter, &crq);
3190
3191         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3192         atomic_inc(&adapter->running_cap_crqs);
3193         ibmvnic_send_crq(adapter, &crq);
3194
3195         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3196         atomic_inc(&adapter->running_cap_crqs);
3197         ibmvnic_send_crq(adapter, &crq);
3198
3199         crq.query_capability.capability =
3200                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3201         atomic_inc(&adapter->running_cap_crqs);
3202         ibmvnic_send_crq(adapter, &crq);
3203
3204         crq.query_capability.capability =
3205                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3206         atomic_inc(&adapter->running_cap_crqs);
3207         ibmvnic_send_crq(adapter, &crq);
3208
3209         crq.query_capability.capability =
3210                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3211         atomic_inc(&adapter->running_cap_crqs);
3212         ibmvnic_send_crq(adapter, &crq);
3213
3214         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3215         atomic_inc(&adapter->running_cap_crqs);
3216         ibmvnic_send_crq(adapter, &crq);
3217 }
3218
3219 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3220                                 struct ibmvnic_adapter *adapter)
3221 {
3222         struct device *dev = &adapter->vdev->dev;
3223
3224         if (crq->get_vpd_size_rsp.rc.code) {
3225                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3226                         crq->get_vpd_size_rsp.rc.code);
3227                 complete(&adapter->fw_done);
3228                 return;
3229         }
3230
3231         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3232         complete(&adapter->fw_done);
3233 }
3234
3235 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3236                            struct ibmvnic_adapter *adapter)
3237 {
3238         struct device *dev = &adapter->vdev->dev;
3239         unsigned char *substr = NULL, *ptr = NULL;
3240         u8 fw_level_len = 0;
3241
3242         memset(adapter->fw_version, 0, 32);
3243
3244         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3245                          DMA_FROM_DEVICE);
3246
3247         if (crq->get_vpd_rsp.rc.code) {
3248                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3249                         crq->get_vpd_rsp.rc.code);
3250                 goto complete;
3251         }
3252
3253         /* get the position of the firmware version info
3254          * located after the ASCII 'RM' substring in the buffer
3255          */
3256         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3257         if (!substr) {
3258                 dev_info(dev, "No FW level provided by VPD\n");
3259                 goto complete;
3260         }
3261
3262         /* get length of firmware level ASCII substring */
3263         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3264                 fw_level_len = *(substr + 2);
3265         } else {
3266                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3267                 goto complete;
3268         }
3269
3270         /* copy firmware version string from vpd into adapter */
3271         if ((substr + 3 + fw_level_len) <
3272             (adapter->vpd->buff + adapter->vpd->len)) {
3273                 ptr = strncpy((char *)adapter->fw_version,
3274                               substr + 3, fw_level_len);
3275
3276                 if (!ptr)
3277                         dev_err(dev, "Failed to isolate FW level string\n");
3278         } else {
3279                 dev_info(dev, "FW substr extrapolated VPD buff\n");
3280         }
3281
3282 complete:
3283         complete(&adapter->fw_done);
3284 }
3285
3286 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3287 {
3288         struct device *dev = &adapter->vdev->dev;
3289         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3290         union ibmvnic_crq crq;
3291         int i;
3292
3293         dma_unmap_single(dev, adapter->ip_offload_tok,
3294                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3295
3296         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3297         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3298                 netdev_dbg(adapter->netdev, "%016lx\n",
3299                            ((unsigned long int *)(buf))[i]);
3300
3301         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3302         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3303         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3304                    buf->tcp_ipv4_chksum);
3305         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3306                    buf->tcp_ipv6_chksum);
3307         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3308                    buf->udp_ipv4_chksum);
3309         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3310                    buf->udp_ipv6_chksum);
3311         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3312                    buf->large_tx_ipv4);
3313         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3314                    buf->large_tx_ipv6);
3315         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3316                    buf->large_rx_ipv4);
3317         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3318                    buf->large_rx_ipv6);
3319         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3320                    buf->max_ipv4_header_size);
3321         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3322                    buf->max_ipv6_header_size);
3323         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3324                    buf->max_tcp_header_size);
3325         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3326                    buf->max_udp_header_size);
3327         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3328                    buf->max_large_tx_size);
3329         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3330                    buf->max_large_rx_size);
3331         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3332                    buf->ipv6_extension_header);
3333         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3334                    buf->tcp_pseudosum_req);
3335         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3336                    buf->num_ipv6_ext_headers);
3337         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3338                    buf->off_ipv6_ext_headers);
3339
3340         adapter->ip_offload_ctrl_tok =
3341             dma_map_single(dev, &adapter->ip_offload_ctrl,
3342                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3343
3344         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3345                 dev_err(dev, "Couldn't map ip offload control buffer\n");
3346                 return;
3347         }
3348
3349         adapter->ip_offload_ctrl.len =
3350             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3351         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3352         adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3353         adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3354         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3355         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3356         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3357         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3358         adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3359         adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3360
3361         /* large_rx disabled for now, additional features needed */
3362         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3363         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3364
3365         adapter->netdev->features = NETIF_F_SG | NETIF_F_GSO;
3366
3367         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3368                 adapter->netdev->features |= NETIF_F_IP_CSUM;
3369
3370         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3371                 adapter->netdev->features |= NETIF_F_IPV6_CSUM;
3372
3373         if ((adapter->netdev->features &
3374             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3375                 adapter->netdev->features |= NETIF_F_RXCSUM;
3376
3377         if (buf->large_tx_ipv4)
3378                 adapter->netdev->features |= NETIF_F_TSO;
3379         if (buf->large_tx_ipv6)
3380                 adapter->netdev->features |= NETIF_F_TSO6;
3381
3382         adapter->netdev->hw_features |= adapter->netdev->features;
3383
3384         memset(&crq, 0, sizeof(crq));
3385         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3386         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3387         crq.control_ip_offload.len =
3388             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3389         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3390         ibmvnic_send_crq(adapter, &crq);
3391 }
3392
3393 static void handle_error_info_rsp(union ibmvnic_crq *crq,
3394                                   struct ibmvnic_adapter *adapter)
3395 {
3396         struct device *dev = &adapter->vdev->dev;
3397         struct ibmvnic_error_buff *error_buff, *tmp;
3398         unsigned long flags;
3399         bool found = false;
3400         int i;
3401
3402         if (!crq->request_error_rsp.rc.code) {
3403                 dev_info(dev, "Request Error Rsp returned with rc=%x\n",
3404                          crq->request_error_rsp.rc.code);
3405                 return;
3406         }
3407
3408         spin_lock_irqsave(&adapter->error_list_lock, flags);
3409         list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list)
3410                 if (error_buff->error_id == crq->request_error_rsp.error_id) {
3411                         found = true;
3412                         list_del(&error_buff->list);
3413                         break;
3414                 }
3415         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3416
3417         if (!found) {
3418                 dev_err(dev, "Couldn't find error id %x\n",
3419                         be32_to_cpu(crq->request_error_rsp.error_id));
3420                 return;
3421         }
3422
3423         dev_err(dev, "Detailed info for error id %x:",
3424                 be32_to_cpu(crq->request_error_rsp.error_id));
3425
3426         for (i = 0; i < error_buff->len; i++) {
3427                 pr_cont("%02x", (int)error_buff->buff[i]);
3428                 if (i % 8 == 7)
3429                         pr_cont(" ");
3430         }
3431         pr_cont("\n");
3432
3433         dma_unmap_single(dev, error_buff->dma, error_buff->len,
3434                          DMA_FROM_DEVICE);
3435         kfree(error_buff->buff);
3436         kfree(error_buff);
3437 }
3438
3439 static void request_error_information(struct ibmvnic_adapter *adapter,
3440                                       union ibmvnic_crq *err_crq)
3441 {
3442         struct device *dev = &adapter->vdev->dev;
3443         struct net_device *netdev = adapter->netdev;
3444         struct ibmvnic_error_buff *error_buff;
3445         unsigned long timeout = msecs_to_jiffies(30000);
3446         union ibmvnic_crq crq;
3447         unsigned long flags;
3448         int rc, detail_len;
3449
3450         error_buff = kmalloc(sizeof(*error_buff), GFP_ATOMIC);
3451         if (!error_buff)
3452                 return;
3453
3454         detail_len = be32_to_cpu(err_crq->error_indication.detail_error_sz);
3455         error_buff->buff = kmalloc(detail_len, GFP_ATOMIC);
3456         if (!error_buff->buff) {
3457                 kfree(error_buff);
3458                 return;
3459         }
3460
3461         error_buff->dma = dma_map_single(dev, error_buff->buff, detail_len,
3462                                          DMA_FROM_DEVICE);
3463         if (dma_mapping_error(dev, error_buff->dma)) {
3464                 netdev_err(netdev, "Couldn't map error buffer\n");
3465                 kfree(error_buff->buff);
3466                 kfree(error_buff);
3467                 return;
3468         }
3469
3470         error_buff->len = detail_len;
3471         error_buff->error_id = err_crq->error_indication.error_id;
3472
3473         spin_lock_irqsave(&adapter->error_list_lock, flags);
3474         list_add_tail(&error_buff->list, &adapter->errors);
3475         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3476
3477         memset(&crq, 0, sizeof(crq));
3478         crq.request_error_info.first = IBMVNIC_CRQ_CMD;
3479         crq.request_error_info.cmd = REQUEST_ERROR_INFO;
3480         crq.request_error_info.ioba = cpu_to_be32(error_buff->dma);
3481         crq.request_error_info.len = cpu_to_be32(detail_len);
3482         crq.request_error_info.error_id = err_crq->error_indication.error_id;
3483
3484         rc = ibmvnic_send_crq(adapter, &crq);
3485         if (rc) {
3486                 netdev_err(netdev, "failed to request error information\n");
3487                 goto err_info_fail;
3488         }
3489
3490         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
3491                 netdev_err(netdev, "timeout waiting for error information\n");
3492                 goto err_info_fail;
3493         }
3494
3495         return;
3496
3497 err_info_fail:
3498         spin_lock_irqsave(&adapter->error_list_lock, flags);
3499         list_del(&error_buff->list);
3500         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3501
3502         kfree(error_buff->buff);
3503         kfree(error_buff);
3504 }
3505
3506 static void handle_error_indication(union ibmvnic_crq *crq,
3507                                     struct ibmvnic_adapter *adapter)
3508 {
3509         struct device *dev = &adapter->vdev->dev;
3510
3511         dev_err(dev, "Firmware reports %serror id %x, cause %d\n",
3512                 crq->error_indication.flags
3513                         & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3514                 be32_to_cpu(crq->error_indication.error_id),
3515                 be16_to_cpu(crq->error_indication.error_cause));
3516
3517         if (be32_to_cpu(crq->error_indication.error_id))
3518                 request_error_information(adapter, crq);
3519
3520         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3521                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3522         else
3523                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3524 }
3525
3526 static void handle_change_mac_rsp(union ibmvnic_crq *crq,
3527                                   struct ibmvnic_adapter *adapter)
3528 {
3529         struct net_device *netdev = adapter->netdev;
3530         struct device *dev = &adapter->vdev->dev;
3531         long rc;
3532
3533         rc = crq->change_mac_addr_rsp.rc.code;
3534         if (rc) {
3535                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3536                 return;
3537         }
3538         memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
3539                ETH_ALEN);
3540 }
3541
3542 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3543                                    struct ibmvnic_adapter *adapter)
3544 {
3545         struct device *dev = &adapter->vdev->dev;
3546         u64 *req_value;
3547         char *name;
3548
3549         atomic_dec(&adapter->running_cap_crqs);
3550         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3551         case REQ_TX_QUEUES:
3552                 req_value = &adapter->req_tx_queues;
3553                 name = "tx";
3554                 break;
3555         case REQ_RX_QUEUES:
3556                 req_value = &adapter->req_rx_queues;
3557                 name = "rx";
3558                 break;
3559         case REQ_RX_ADD_QUEUES:
3560                 req_value = &adapter->req_rx_add_queues;
3561                 name = "rx_add";
3562                 break;
3563         case REQ_TX_ENTRIES_PER_SUBCRQ:
3564                 req_value = &adapter->req_tx_entries_per_subcrq;
3565                 name = "tx_entries_per_subcrq";
3566                 break;
3567         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3568                 req_value = &adapter->req_rx_add_entries_per_subcrq;
3569                 name = "rx_add_entries_per_subcrq";
3570                 break;
3571         case REQ_MTU:
3572                 req_value = &adapter->req_mtu;
3573                 name = "mtu";
3574                 break;
3575         case PROMISC_REQUESTED:
3576                 req_value = &adapter->promisc;
3577                 name = "promisc";
3578                 break;
3579         default:
3580                 dev_err(dev, "Got invalid cap request rsp %d\n",
3581                         crq->request_capability.capability);
3582                 return;
3583         }
3584
3585         switch (crq->request_capability_rsp.rc.code) {
3586         case SUCCESS:
3587                 break;
3588         case PARTIALSUCCESS:
3589                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3590                          *req_value,
3591                          (long int)be64_to_cpu(crq->request_capability_rsp.
3592                                                number), name);
3593                 *req_value = be64_to_cpu(crq->request_capability_rsp.number);
3594                 ibmvnic_send_req_caps(adapter, 1);
3595                 return;
3596         default:
3597                 dev_err(dev, "Error %d in request cap rsp\n",
3598                         crq->request_capability_rsp.rc.code);
3599                 return;
3600         }
3601
3602         /* Done receiving requested capabilities, query IP offload support */
3603         if (atomic_read(&adapter->running_cap_crqs) == 0) {
3604                 union ibmvnic_crq newcrq;
3605                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3606                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
3607                     &adapter->ip_offload_buf;
3608
3609                 adapter->wait_capability = false;
3610                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
3611                                                          buf_sz,
3612                                                          DMA_FROM_DEVICE);
3613
3614                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3615                         if (!firmware_has_feature(FW_FEATURE_CMO))
3616                                 dev_err(dev, "Couldn't map offload buffer\n");
3617                         return;
3618                 }
3619
3620                 memset(&newcrq, 0, sizeof(newcrq));
3621                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3622                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3623                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
3624                 newcrq.query_ip_offload.ioba =
3625                     cpu_to_be32(adapter->ip_offload_tok);
3626
3627                 ibmvnic_send_crq(adapter, &newcrq);
3628         }
3629 }
3630
3631 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
3632                             struct ibmvnic_adapter *adapter)
3633 {
3634         struct device *dev = &adapter->vdev->dev;
3635         struct net_device *netdev = adapter->netdev;
3636         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
3637         struct ibmvnic_login_buffer *login = adapter->login_buf;
3638         int i;
3639
3640         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
3641                          DMA_BIDIRECTIONAL);
3642         dma_unmap_single(dev, adapter->login_rsp_buf_token,
3643                          adapter->login_rsp_buf_sz, DMA_BIDIRECTIONAL);
3644
3645         /* If the number of queues requested can't be allocated by the
3646          * server, the login response will return with code 1. We will need
3647          * to resend the login buffer with fewer queues requested.
3648          */
3649         if (login_rsp_crq->generic.rc.code) {
3650                 adapter->renegotiate = true;
3651                 complete(&adapter->init_done);
3652                 return 0;
3653         }
3654
3655         netdev->mtu = adapter->req_mtu - ETH_HLEN;
3656
3657         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
3658         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
3659                 netdev_dbg(adapter->netdev, "%016lx\n",
3660                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
3661         }
3662
3663         /* Sanity checks */
3664         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
3665             (be32_to_cpu(login->num_rxcomp_subcrqs) *
3666              adapter->req_rx_add_queues !=
3667              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
3668                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
3669                 ibmvnic_remove(adapter->vdev);
3670                 return -EIO;
3671         }
3672         complete(&adapter->init_done);
3673
3674         return 0;
3675 }
3676
3677 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
3678                                      struct ibmvnic_adapter *adapter)
3679 {
3680         struct device *dev = &adapter->vdev->dev;
3681         long rc;
3682
3683         rc = crq->request_unmap_rsp.rc.code;
3684         if (rc)
3685                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
3686 }
3687
3688 static void handle_query_map_rsp(union ibmvnic_crq *crq,
3689                                  struct ibmvnic_adapter *adapter)
3690 {
3691         struct net_device *netdev = adapter->netdev;
3692         struct device *dev = &adapter->vdev->dev;
3693         long rc;
3694
3695         rc = crq->query_map_rsp.rc.code;
3696         if (rc) {
3697                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
3698                 return;
3699         }
3700         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
3701                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
3702                    crq->query_map_rsp.free_pages);
3703 }
3704
3705 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
3706                                  struct ibmvnic_adapter *adapter)
3707 {
3708         struct net_device *netdev = adapter->netdev;
3709         struct device *dev = &adapter->vdev->dev;
3710         long rc;
3711
3712         atomic_dec(&adapter->running_cap_crqs);
3713         netdev_dbg(netdev, "Outstanding queries: %d\n",
3714                    atomic_read(&adapter->running_cap_crqs));
3715         rc = crq->query_capability.rc.code;
3716         if (rc) {
3717                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
3718                 goto out;
3719         }
3720
3721         switch (be16_to_cpu(crq->query_capability.capability)) {
3722         case MIN_TX_QUEUES:
3723                 adapter->min_tx_queues =
3724                     be64_to_cpu(crq->query_capability.number);
3725                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
3726                            adapter->min_tx_queues);
3727                 break;
3728         case MIN_RX_QUEUES:
3729                 adapter->min_rx_queues =
3730                     be64_to_cpu(crq->query_capability.number);
3731                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
3732                            adapter->min_rx_queues);
3733                 break;
3734         case MIN_RX_ADD_QUEUES:
3735                 adapter->min_rx_add_queues =
3736                     be64_to_cpu(crq->query_capability.number);
3737                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
3738                            adapter->min_rx_add_queues);
3739                 break;
3740         case MAX_TX_QUEUES:
3741                 adapter->max_tx_queues =
3742                     be64_to_cpu(crq->query_capability.number);
3743                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
3744                            adapter->max_tx_queues);
3745                 break;
3746         case MAX_RX_QUEUES:
3747                 adapter->max_rx_queues =
3748                     be64_to_cpu(crq->query_capability.number);
3749                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
3750                            adapter->max_rx_queues);
3751                 break;
3752         case MAX_RX_ADD_QUEUES:
3753                 adapter->max_rx_add_queues =
3754                     be64_to_cpu(crq->query_capability.number);
3755                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
3756                            adapter->max_rx_add_queues);
3757                 break;
3758         case MIN_TX_ENTRIES_PER_SUBCRQ:
3759                 adapter->min_tx_entries_per_subcrq =
3760                     be64_to_cpu(crq->query_capability.number);
3761                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
3762                            adapter->min_tx_entries_per_subcrq);
3763                 break;
3764         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
3765                 adapter->min_rx_add_entries_per_subcrq =
3766                     be64_to_cpu(crq->query_capability.number);
3767                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
3768                            adapter->min_rx_add_entries_per_subcrq);
3769                 break;
3770         case MAX_TX_ENTRIES_PER_SUBCRQ:
3771                 adapter->max_tx_entries_per_subcrq =
3772                     be64_to_cpu(crq->query_capability.number);
3773                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
3774                            adapter->max_tx_entries_per_subcrq);
3775                 break;
3776         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
3777                 adapter->max_rx_add_entries_per_subcrq =
3778                     be64_to_cpu(crq->query_capability.number);
3779                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
3780                            adapter->max_rx_add_entries_per_subcrq);
3781                 break;
3782         case TCP_IP_OFFLOAD:
3783                 adapter->tcp_ip_offload =
3784                     be64_to_cpu(crq->query_capability.number);
3785                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
3786                            adapter->tcp_ip_offload);
3787                 break;
3788         case PROMISC_SUPPORTED:
3789                 adapter->promisc_supported =
3790                     be64_to_cpu(crq->query_capability.number);
3791                 netdev_dbg(netdev, "promisc_supported = %lld\n",
3792                            adapter->promisc_supported);
3793                 break;
3794         case MIN_MTU:
3795                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
3796                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
3797                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
3798                 break;
3799         case MAX_MTU:
3800                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
3801                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
3802                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
3803                 break;
3804         case MAX_MULTICAST_FILTERS:
3805                 adapter->max_multicast_filters =
3806                     be64_to_cpu(crq->query_capability.number);
3807                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
3808                            adapter->max_multicast_filters);
3809                 break;
3810         case VLAN_HEADER_INSERTION:
3811                 adapter->vlan_header_insertion =
3812                     be64_to_cpu(crq->query_capability.number);
3813                 if (adapter->vlan_header_insertion)
3814                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
3815                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
3816                            adapter->vlan_header_insertion);
3817                 break;
3818         case RX_VLAN_HEADER_INSERTION:
3819                 adapter->rx_vlan_header_insertion =
3820                     be64_to_cpu(crq->query_capability.number);
3821                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
3822                            adapter->rx_vlan_header_insertion);
3823                 break;
3824         case MAX_TX_SG_ENTRIES:
3825                 adapter->max_tx_sg_entries =
3826                     be64_to_cpu(crq->query_capability.number);
3827                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
3828                            adapter->max_tx_sg_entries);
3829                 break;
3830         case RX_SG_SUPPORTED:
3831                 adapter->rx_sg_supported =
3832                     be64_to_cpu(crq->query_capability.number);
3833                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
3834                            adapter->rx_sg_supported);
3835                 break;
3836         case OPT_TX_COMP_SUB_QUEUES:
3837                 adapter->opt_tx_comp_sub_queues =
3838                     be64_to_cpu(crq->query_capability.number);
3839                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
3840                            adapter->opt_tx_comp_sub_queues);
3841                 break;
3842         case OPT_RX_COMP_QUEUES:
3843                 adapter->opt_rx_comp_queues =
3844                     be64_to_cpu(crq->query_capability.number);
3845                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
3846                            adapter->opt_rx_comp_queues);
3847                 break;
3848         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
3849                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
3850                     be64_to_cpu(crq->query_capability.number);
3851                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
3852                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
3853                 break;
3854         case OPT_TX_ENTRIES_PER_SUBCRQ:
3855                 adapter->opt_tx_entries_per_subcrq =
3856                     be64_to_cpu(crq->query_capability.number);
3857                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
3858                            adapter->opt_tx_entries_per_subcrq);
3859                 break;
3860         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
3861                 adapter->opt_rxba_entries_per_subcrq =
3862                     be64_to_cpu(crq->query_capability.number);
3863                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
3864                            adapter->opt_rxba_entries_per_subcrq);
3865                 break;
3866         case TX_RX_DESC_REQ:
3867                 adapter->tx_rx_desc_req = crq->query_capability.number;
3868                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
3869                            adapter->tx_rx_desc_req);
3870                 break;
3871
3872         default:
3873                 netdev_err(netdev, "Got invalid cap rsp %d\n",
3874                            crq->query_capability.capability);
3875         }
3876
3877 out:
3878         if (atomic_read(&adapter->running_cap_crqs) == 0) {
3879                 adapter->wait_capability = false;
3880                 ibmvnic_send_req_caps(adapter, 0);
3881         }
3882 }
3883
3884 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
3885                                struct ibmvnic_adapter *adapter)
3886 {
3887         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
3888         struct net_device *netdev = adapter->netdev;
3889         struct device *dev = &adapter->vdev->dev;
3890         u64 *u64_crq = (u64 *)crq;
3891         long rc;
3892
3893         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
3894                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3895                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3896         switch (gen_crq->first) {
3897         case IBMVNIC_CRQ_INIT_RSP:
3898                 switch (gen_crq->cmd) {
3899                 case IBMVNIC_CRQ_INIT:
3900                         dev_info(dev, "Partner initialized\n");
3901                         adapter->from_passive_init = true;
3902                         complete(&adapter->init_done);
3903                         break;
3904                 case IBMVNIC_CRQ_INIT_COMPLETE:
3905                         dev_info(dev, "Partner initialization complete\n");
3906                         send_version_xchg(adapter);
3907                         break;
3908                 default:
3909                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
3910                 }
3911                 return;
3912         case IBMVNIC_CRQ_XPORT_EVENT:
3913                 netif_carrier_off(netdev);
3914                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
3915                         dev_info(dev, "Migrated, re-enabling adapter\n");
3916                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
3917                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
3918                         dev_info(dev, "Backing device failover detected\n");
3919                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
3920                 } else {
3921                         /* The adapter lost the connection */
3922                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
3923                                 gen_crq->cmd);
3924                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3925                 }
3926                 return;
3927         case IBMVNIC_CRQ_CMD_RSP:
3928                 break;
3929         default:
3930                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
3931                         gen_crq->first);
3932                 return;
3933         }
3934
3935         switch (gen_crq->cmd) {
3936         case VERSION_EXCHANGE_RSP:
3937                 rc = crq->version_exchange_rsp.rc.code;
3938                 if (rc) {
3939                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
3940                         break;
3941                 }
3942                 dev_info(dev, "Partner protocol version is %d\n",
3943                          crq->version_exchange_rsp.version);
3944                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
3945                     ibmvnic_version)
3946                         ibmvnic_version =
3947                             be16_to_cpu(crq->version_exchange_rsp.version);
3948                 send_cap_queries(adapter);
3949                 break;
3950         case QUERY_CAPABILITY_RSP:
3951                 handle_query_cap_rsp(crq, adapter);
3952                 break;
3953         case QUERY_MAP_RSP:
3954                 handle_query_map_rsp(crq, adapter);
3955                 break;
3956         case REQUEST_MAP_RSP:
3957                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
3958                 complete(&adapter->fw_done);
3959                 break;
3960         case REQUEST_UNMAP_RSP:
3961                 handle_request_unmap_rsp(crq, adapter);
3962                 break;
3963         case REQUEST_CAPABILITY_RSP:
3964                 handle_request_cap_rsp(crq, adapter);
3965                 break;
3966         case LOGIN_RSP:
3967                 netdev_dbg(netdev, "Got Login Response\n");
3968                 handle_login_rsp(crq, adapter);
3969                 break;
3970         case LOGICAL_LINK_STATE_RSP:
3971                 netdev_dbg(netdev,
3972                            "Got Logical Link State Response, state: %d rc: %d\n",
3973                            crq->logical_link_state_rsp.link_state,
3974                            crq->logical_link_state_rsp.rc.code);
3975                 adapter->logical_link_state =
3976                     crq->logical_link_state_rsp.link_state;
3977                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
3978                 complete(&adapter->init_done);
3979                 break;
3980         case LINK_STATE_INDICATION:
3981                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
3982                 adapter->phys_link_state =
3983                     crq->link_state_indication.phys_link_state;
3984                 adapter->logical_link_state =
3985                     crq->link_state_indication.logical_link_state;
3986                 break;
3987         case CHANGE_MAC_ADDR_RSP:
3988                 netdev_dbg(netdev, "Got MAC address change Response\n");
3989                 handle_change_mac_rsp(crq, adapter);
3990                 break;
3991         case ERROR_INDICATION:
3992                 netdev_dbg(netdev, "Got Error Indication\n");
3993                 handle_error_indication(crq, adapter);
3994                 break;
3995         case REQUEST_ERROR_RSP:
3996                 netdev_dbg(netdev, "Got Error Detail Response\n");
3997                 handle_error_info_rsp(crq, adapter);
3998                 break;
3999         case REQUEST_STATISTICS_RSP:
4000                 netdev_dbg(netdev, "Got Statistics Response\n");
4001                 complete(&adapter->stats_done);
4002                 break;
4003         case QUERY_IP_OFFLOAD_RSP:
4004                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4005                 handle_query_ip_offload_rsp(adapter);
4006                 break;
4007         case MULTICAST_CTRL_RSP:
4008                 netdev_dbg(netdev, "Got multicast control Response\n");
4009                 break;
4010         case CONTROL_IP_OFFLOAD_RSP:
4011                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4012                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4013                                  sizeof(adapter->ip_offload_ctrl),
4014                                  DMA_TO_DEVICE);
4015                 complete(&adapter->init_done);
4016                 break;
4017         case COLLECT_FW_TRACE_RSP:
4018                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4019                 complete(&adapter->fw_done);
4020                 break;
4021         case GET_VPD_SIZE_RSP:
4022                 handle_vpd_size_rsp(crq, adapter);
4023                 break;
4024         case GET_VPD_RSP:
4025                 handle_vpd_rsp(crq, adapter);
4026                 break;
4027         default:
4028                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4029                            gen_crq->cmd);
4030         }
4031 }
4032
4033 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4034 {
4035         struct ibmvnic_adapter *adapter = instance;
4036
4037         tasklet_schedule(&adapter->tasklet);
4038         return IRQ_HANDLED;
4039 }
4040
4041 static void ibmvnic_tasklet(void *data)
4042 {
4043         struct ibmvnic_adapter *adapter = data;
4044         struct ibmvnic_crq_queue *queue = &adapter->crq;
4045         union ibmvnic_crq *crq;
4046         unsigned long flags;
4047         bool done = false;
4048
4049         spin_lock_irqsave(&queue->lock, flags);
4050         while (!done) {
4051                 /* Pull all the valid messages off the CRQ */
4052                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4053                         ibmvnic_handle_crq(crq, adapter);
4054                         crq->generic.first = 0;
4055                 }
4056
4057                 /* remain in tasklet until all
4058                  * capabilities responses are received
4059                  */
4060                 if (!adapter->wait_capability)
4061                         done = true;
4062         }
4063         /* if capabilities CRQ's were sent in this tasklet, the following
4064          * tasklet must wait until all responses are received
4065          */
4066         if (atomic_read(&adapter->running_cap_crqs) != 0)
4067                 adapter->wait_capability = true;
4068         spin_unlock_irqrestore(&queue->lock, flags);
4069 }
4070
4071 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4072 {
4073         struct vio_dev *vdev = adapter->vdev;
4074         int rc;
4075
4076         do {
4077                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4078         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4079
4080         if (rc)
4081                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4082
4083         return rc;
4084 }
4085
4086 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4087 {
4088         struct ibmvnic_crq_queue *crq = &adapter->crq;
4089         struct device *dev = &adapter->vdev->dev;
4090         struct vio_dev *vdev = adapter->vdev;
4091         int rc;
4092
4093         /* Close the CRQ */
4094         do {
4095                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4096         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4097
4098         /* Clean out the queue */
4099         memset(crq->msgs, 0, PAGE_SIZE);
4100         crq->cur = 0;
4101
4102         /* And re-open it again */
4103         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4104                                 crq->msg_token, PAGE_SIZE);
4105
4106         if (rc == H_CLOSED)
4107                 /* Adapter is good, but other end is not ready */
4108                 dev_warn(dev, "Partner adapter not ready\n");
4109         else if (rc != 0)
4110                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4111
4112         return rc;
4113 }
4114
4115 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4116 {
4117         struct ibmvnic_crq_queue *crq = &adapter->crq;
4118         struct vio_dev *vdev = adapter->vdev;
4119         long rc;
4120
4121         if (!crq->msgs)
4122                 return;
4123
4124         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4125         free_irq(vdev->irq, adapter);
4126         tasklet_kill(&adapter->tasklet);
4127         do {
4128                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4129         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4130
4131         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4132                          DMA_BIDIRECTIONAL);
4133         free_page((unsigned long)crq->msgs);
4134         crq->msgs = NULL;
4135 }
4136
4137 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4138 {
4139         struct ibmvnic_crq_queue *crq = &adapter->crq;
4140         struct device *dev = &adapter->vdev->dev;
4141         struct vio_dev *vdev = adapter->vdev;
4142         int rc, retrc = -ENOMEM;
4143
4144         if (crq->msgs)
4145                 return 0;
4146
4147         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4148         /* Should we allocate more than one page? */
4149
4150         if (!crq->msgs)
4151                 return -ENOMEM;
4152
4153         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4154         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4155                                         DMA_BIDIRECTIONAL);
4156         if (dma_mapping_error(dev, crq->msg_token))
4157                 goto map_failed;
4158
4159         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4160                                 crq->msg_token, PAGE_SIZE);
4161
4162         if (rc == H_RESOURCE)
4163                 /* maybe kexecing and resource is busy. try a reset */
4164                 rc = ibmvnic_reset_crq(adapter);
4165         retrc = rc;
4166
4167         if (rc == H_CLOSED) {
4168                 dev_warn(dev, "Partner adapter not ready\n");
4169         } else if (rc) {
4170                 dev_warn(dev, "Error %d opening adapter\n", rc);
4171                 goto reg_crq_failed;
4172         }
4173
4174         retrc = 0;
4175
4176         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4177                      (unsigned long)adapter);
4178
4179         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4180         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
4181                          adapter);
4182         if (rc) {
4183                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4184                         vdev->irq, rc);
4185                 goto req_irq_failed;
4186         }
4187
4188         rc = vio_enable_interrupts(vdev);
4189         if (rc) {
4190                 dev_err(dev, "Error %d enabling interrupts\n", rc);
4191                 goto req_irq_failed;
4192         }
4193
4194         crq->cur = 0;
4195         spin_lock_init(&crq->lock);
4196
4197         return retrc;
4198
4199 req_irq_failed:
4200         tasklet_kill(&adapter->tasklet);
4201         do {
4202                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4203         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4204 reg_crq_failed:
4205         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4206 map_failed:
4207         free_page((unsigned long)crq->msgs);
4208         crq->msgs = NULL;
4209         return retrc;
4210 }
4211
4212 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4213 {
4214         struct device *dev = &adapter->vdev->dev;
4215         unsigned long timeout = msecs_to_jiffies(30000);
4216         int rc;
4217
4218         if (adapter->resetting && !adapter->wait_for_reset) {
4219                 rc = ibmvnic_reset_crq(adapter);
4220                 if (!rc)
4221                         rc = vio_enable_interrupts(adapter->vdev);
4222         } else {
4223                 rc = init_crq_queue(adapter);
4224         }
4225
4226         if (rc) {
4227                 dev_err(dev, "Couldn't initialize crq. rc=%d\n", rc);
4228                 return rc;
4229         }
4230
4231         adapter->from_passive_init = false;
4232
4233         init_completion(&adapter->init_done);
4234         adapter->init_done_rc = 0;
4235         ibmvnic_send_crq_init(adapter);
4236         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4237                 dev_err(dev, "Initialization sequence timed out\n");
4238                 return -1;
4239         }
4240
4241         if (adapter->init_done_rc) {
4242                 release_crq_queue(adapter);
4243                 return adapter->init_done_rc;
4244         }
4245
4246         if (adapter->from_passive_init) {
4247                 adapter->state = VNIC_OPEN;
4248                 adapter->from_passive_init = false;
4249                 return -1;
4250         }
4251
4252         if (adapter->resetting && !adapter->wait_for_reset)
4253                 rc = reset_sub_crq_queues(adapter);
4254         else
4255                 rc = init_sub_crqs(adapter);
4256         if (rc) {
4257                 dev_err(dev, "Initialization of sub crqs failed\n");
4258                 release_crq_queue(adapter);
4259                 return rc;
4260         }
4261
4262         rc = init_sub_crq_irqs(adapter);
4263         if (rc) {
4264                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4265                 release_crq_queue(adapter);
4266         }
4267
4268         return rc;
4269 }
4270
4271 static struct device_attribute dev_attr_failover;
4272
4273 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4274 {
4275         struct ibmvnic_adapter *adapter;
4276         struct net_device *netdev;
4277         unsigned char *mac_addr_p;
4278         int rc;
4279
4280         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4281                 dev->unit_address);
4282
4283         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4284                                                         VETH_MAC_ADDR, NULL);
4285         if (!mac_addr_p) {
4286                 dev_err(&dev->dev,
4287                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4288                         __FILE__, __LINE__);
4289                 return 0;
4290         }
4291
4292         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4293                                    IBMVNIC_MAX_TX_QUEUES);
4294         if (!netdev)
4295                 return -ENOMEM;
4296
4297         adapter = netdev_priv(netdev);
4298         adapter->state = VNIC_PROBING;
4299         dev_set_drvdata(&dev->dev, netdev);
4300         adapter->vdev = dev;
4301         adapter->netdev = netdev;
4302
4303         ether_addr_copy(adapter->mac_addr, mac_addr_p);
4304         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4305         netdev->irq = dev->irq;
4306         netdev->netdev_ops = &ibmvnic_netdev_ops;
4307         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4308         SET_NETDEV_DEV(netdev, &dev->dev);
4309
4310         spin_lock_init(&adapter->stats_lock);
4311
4312         INIT_LIST_HEAD(&adapter->errors);
4313         spin_lock_init(&adapter->error_list_lock);
4314
4315         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4316         INIT_LIST_HEAD(&adapter->rwi_list);
4317         mutex_init(&adapter->reset_lock);
4318         mutex_init(&adapter->rwi_lock);
4319         adapter->resetting = false;
4320
4321         adapter->mac_change_pending = false;
4322
4323         do {
4324                 rc = ibmvnic_init(adapter);
4325                 if (rc && rc != EAGAIN)
4326                         goto ibmvnic_init_fail;
4327         } while (rc == EAGAIN);
4328
4329         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4330         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4331         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4332
4333         rc = device_create_file(&dev->dev, &dev_attr_failover);
4334         if (rc)
4335                 goto ibmvnic_init_fail;
4336
4337         netif_carrier_off(netdev);
4338         rc = register_netdev(netdev);
4339         if (rc) {
4340                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4341                 goto ibmvnic_register_fail;
4342         }
4343         dev_info(&dev->dev, "ibmvnic registered\n");
4344
4345         adapter->state = VNIC_PROBED;
4346
4347         adapter->wait_for_reset = false;
4348
4349         return 0;
4350
4351 ibmvnic_register_fail:
4352         device_remove_file(&dev->dev, &dev_attr_failover);
4353
4354 ibmvnic_init_fail:
4355         release_sub_crqs(adapter);
4356         release_crq_queue(adapter);
4357         free_netdev(netdev);
4358
4359         return rc;
4360 }
4361
4362 static int ibmvnic_remove(struct vio_dev *dev)
4363 {
4364         struct net_device *netdev = dev_get_drvdata(&dev->dev);
4365         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4366
4367         adapter->state = VNIC_REMOVING;
4368         unregister_netdev(netdev);
4369         mutex_lock(&adapter->reset_lock);
4370
4371         release_resources(adapter);
4372         release_sub_crqs(adapter);
4373         release_crq_queue(adapter);
4374
4375         adapter->state = VNIC_REMOVED;
4376
4377         mutex_unlock(&adapter->reset_lock);
4378         device_remove_file(&dev->dev, &dev_attr_failover);
4379         free_netdev(netdev);
4380         dev_set_drvdata(&dev->dev, NULL);
4381
4382         return 0;
4383 }
4384
4385 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4386                               const char *buf, size_t count)
4387 {
4388         struct net_device *netdev = dev_get_drvdata(dev);
4389         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4390         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4391         __be64 session_token;
4392         long rc;
4393
4394         if (!sysfs_streq(buf, "1"))
4395                 return -EINVAL;
4396
4397         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4398                          H_GET_SESSION_TOKEN, 0, 0, 0);
4399         if (rc) {
4400                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4401                            rc);
4402                 return -EINVAL;
4403         }
4404
4405         session_token = (__be64)retbuf[0];
4406         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4407                    be64_to_cpu(session_token));
4408         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4409                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
4410         if (rc) {
4411                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4412                            rc);
4413                 return -EINVAL;
4414         }
4415
4416         return count;
4417 }
4418
4419 static DEVICE_ATTR(failover, 0200, NULL, failover_store);
4420
4421 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4422 {
4423         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4424         struct ibmvnic_adapter *adapter;
4425         struct iommu_table *tbl;
4426         unsigned long ret = 0;
4427         int i;
4428
4429         tbl = get_iommu_table_base(&vdev->dev);
4430
4431         /* netdev inits at probe time along with the structures we need below*/
4432         if (!netdev)
4433                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4434
4435         adapter = netdev_priv(netdev);
4436
4437         ret += PAGE_SIZE; /* the crq message queue */
4438         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4439
4440         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4441                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
4442
4443         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4444              i++)
4445                 ret += adapter->rx_pool[i].size *
4446                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4447
4448         return ret;
4449 }
4450
4451 static int ibmvnic_resume(struct device *dev)
4452 {
4453         struct net_device *netdev = dev_get_drvdata(dev);
4454         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4455
4456         if (adapter->state != VNIC_OPEN)
4457                 return 0;
4458
4459         tasklet_schedule(&adapter->tasklet);
4460
4461         return 0;
4462 }
4463
4464 static const struct vio_device_id ibmvnic_device_table[] = {
4465         {"network", "IBM,vnic"},
4466         {"", "" }
4467 };
4468 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
4469
4470 static const struct dev_pm_ops ibmvnic_pm_ops = {
4471         .resume = ibmvnic_resume
4472 };
4473
4474 static struct vio_driver ibmvnic_driver = {
4475         .id_table       = ibmvnic_device_table,
4476         .probe          = ibmvnic_probe,
4477         .remove         = ibmvnic_remove,
4478         .get_desired_dma = ibmvnic_get_desired_dma,
4479         .name           = ibmvnic_driver_name,
4480         .pm             = &ibmvnic_pm_ops,
4481 };
4482
4483 /* module functions */
4484 static int __init ibmvnic_module_init(void)
4485 {
4486         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
4487                 IBMVNIC_DRIVER_VERSION);
4488
4489         return vio_register_driver(&ibmvnic_driver);
4490 }
4491
4492 static void __exit ibmvnic_module_exit(void)
4493 {
4494         vio_unregister_driver(&ibmvnic_driver);
4495 }
4496
4497 module_init(ibmvnic_module_init);
4498 module_exit(ibmvnic_module_exit);