drivers/net/ibmveth.c

   1 /**************************************************************************/
   2 /*                                                                        */
   3 /* IBM eServer i/pSeries Virtual Ethernet Device Driver                   */
   4 /* Copyright (C) 2003 IBM Corp.                                           */
   5 /*  Originally written by Dave Larson (larson1@us.ibm.com)                */
   6 /*  Maintained by Santiago Leon (santil@us.ibm.com)                       */
   7 /*                                                                        */
   8 /*  This program is free software; you can redistribute it and/or modify  */
   9 /*  it under the terms of the GNU General Public License as published by  */
  10 /*  the Free Software Foundation; either version 2 of the License, or     */
  11 /*  (at your option) any later version.                                   */
  12 /*                                                                        */
  13 /*  This program is distributed in the hope that it will be useful,       */
  14 /*  but WITHOUT ANY WARRANTY; without even the implied warranty of        */
  15 /*  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         */
  16 /*  GNU General Public License for more details.                          */
  17 /*                                                                        */
  18 /*  You should have received a copy of the GNU General Public License     */
  19 /*  along with this program; if not, write to the Free Software           */
  20 /*  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  */
  21 /*                                                                   USA  */
  22 /*                                                                        */
  23 /* This module contains the implementation of a virtual ethernet device   */
  24 /* for use with IBM i/pSeries LPAR Linux.  It utilizes the logical LAN    */
  25 /* option of the RS/6000 Platform Architechture to interface with virtual */
  26 /* ethernet NICs that are presented to the partition by the hypervisor.   */
  27 /*                                                                        */
  28 /**************************************************************************/
  29 /*
  30   TODO:
  31   - add support for sysfs
  32   - possibly remove procfs support
  33 */
  34
  35 #include <linux/module.h>
  36 #include <linux/types.h>
  37 #include <linux/errno.h>
  38 #include <linux/ioport.h>
  39 #include <linux/dma-mapping.h>
  40 #include <linux/kernel.h>
  41 #include <linux/netdevice.h>
  42 #include <linux/etherdevice.h>
  43 #include <linux/skbuff.h>
  44 #include <linux/init.h>
  45 #include <linux/delay.h>
  46 #include <linux/mm.h>
  47 #include <linux/ethtool.h>
  48 #include <linux/proc_fs.h>
  49 #include <linux/in.h>
  50 #include <linux/ip.h>
  51 #include <net/net_namespace.h>
  52 #include <asm/semaphore.h>
  53 #include <asm/hvcall.h>
  54 #include <asm/atomic.h>
  55 #include <asm/vio.h>
  56 #include <asm/uaccess.h>
  57 #include <linux/seq_file.h>
  58
  59 #include "ibmveth.h"
  60
  61 #undef DEBUG
  62
  63 #define ibmveth_printk(fmt, args...) \
  64   printk(KERN_DEBUG "%s: " fmt, __FILE__, ## args)
  65
  66 #define ibmveth_error_printk(fmt, args...) \
  67   printk(KERN_ERR "(%s:%3.3d ua:%x) ERROR: " fmt, __FILE__, __LINE__ , adapter->vdev->unit_address, ## args)
  68
  69 #ifdef DEBUG
  70 #define ibmveth_debug_printk_no_adapter(fmt, args...) \
  71   printk(KERN_DEBUG "(%s:%3.3d): " fmt, __FILE__, __LINE__ , ## args)
  72 #define ibmveth_debug_printk(fmt, args...) \
  73   printk(KERN_DEBUG "(%s:%3.3d ua:%x): " fmt, __FILE__, __LINE__ , adapter->vdev->unit_address, ## args)
  74 #define ibmveth_assert(expr) \
  75   if(!(expr)) {                                   \
  76     printk(KERN_DEBUG "assertion failed (%s:%3.3d ua:%x): %s\n", __FILE__, __LINE__, adapter->vdev->unit_address, #expr); \
  77     BUG(); \
  78   }
  79 #else
  80 #define ibmveth_debug_printk_no_adapter(fmt, args...)
  81 #define ibmveth_debug_printk(fmt, args...)
  82 #define ibmveth_assert(expr)
  83 #endif
  84
  85 static int ibmveth_open(struct net_device *dev);
  86 static int ibmveth_close(struct net_device *dev);
  87 static int ibmveth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd);
  88 static int ibmveth_poll(struct napi_struct *napi, int budget);
  89 static int ibmveth_start_xmit(struct sk_buff *skb, struct net_device *dev);
  90 static struct net_device_stats *ibmveth_get_stats(struct net_device *dev);
  91 static void ibmveth_set_multicast_list(struct net_device *dev);
  92 static int ibmveth_change_mtu(struct net_device *dev, int new_mtu);
  93 static void ibmveth_proc_register_driver(void);
  94 static void ibmveth_proc_unregister_driver(void);
  95 static void ibmveth_proc_register_adapter(struct ibmveth_adapter *adapter);
  96 static void ibmveth_proc_unregister_adapter(struct ibmveth_adapter *adapter);
  97 static irqreturn_t ibmveth_interrupt(int irq, void *dev_instance);
  98 static void ibmveth_rxq_harvest_buffer(struct ibmveth_adapter *adapter);
  99 static struct kobj_type ktype_veth_pool;
 100
 101 #ifdef CONFIG_PROC_FS
 102 #define IBMVETH_PROC_DIR "ibmveth"
 103 static struct proc_dir_entry *ibmveth_proc_dir;
 104 #endif
 105
 106 static const char ibmveth_driver_name[] = "ibmveth";
 107 static const char ibmveth_driver_string[] = "IBM i/pSeries Virtual Ethernet Driver";
 108 #define ibmveth_driver_version "1.03"
 109
 110 MODULE_AUTHOR("Santiago Leon <santil@us.ibm.com>");
 111 MODULE_DESCRIPTION("IBM i/pSeries Virtual Ethernet Driver");
 112 MODULE_LICENSE("GPL");
 113 MODULE_VERSION(ibmveth_driver_version);
 114
 115 struct ibmveth_stat {
 116         char name[ETH_GSTRING_LEN];
 117         int offset;
 118 };
 119
 120 #define IBMVETH_STAT_OFF(stat) offsetof(struct ibmveth_adapter, stat)
 121 #define IBMVETH_GET_STAT(a, off) *((u64 *)(((unsigned long)(a)) + off))
 122
 123 struct ibmveth_stat ibmveth_stats[] = {
 124         { "replenish_task_cycles", IBMVETH_STAT_OFF(replenish_task_cycles) },
 125         { "replenish_no_mem", IBMVETH_STAT_OFF(replenish_no_mem) },
 126         { "replenish_add_buff_failure", IBMVETH_STAT_OFF(replenish_add_buff_failure) },
 127         { "replenish_add_buff_success", IBMVETH_STAT_OFF(replenish_add_buff_success) },
 128         { "rx_invalid_buffer", IBMVETH_STAT_OFF(rx_invalid_buffer) },
 129         { "rx_no_buffer", IBMVETH_STAT_OFF(rx_no_buffer) },
 130         { "tx_map_failed", IBMVETH_STAT_OFF(tx_map_failed) },
 131         { "tx_send_failed", IBMVETH_STAT_OFF(tx_send_failed) },
 132 };
 133
 134 /* simple methods of getting data from the current rxq entry */
 135 static inline u32 ibmveth_rxq_flags(struct ibmveth_adapter *adapter)
 136 {
 137         return adapter->rx_queue.queue_addr[adapter->rx_queue.index].flags_off;
 138 }
 139
 140 static inline int ibmveth_rxq_toggle(struct ibmveth_adapter *adapter)
 141 {
 142         return (ibmveth_rxq_flags(adapter) & IBMVETH_RXQ_TOGGLE) >> IBMVETH_RXQ_TOGGLE_SHIFT;
 143 }
 144
 145 static inline int ibmveth_rxq_pending_buffer(struct ibmveth_adapter *adapter)
 146 {
 147         return (ibmveth_rxq_toggle(adapter) == adapter->rx_queue.toggle);
 148 }
 149
 150 static inline int ibmveth_rxq_buffer_valid(struct ibmveth_adapter *adapter)
 151 {
 152         return (ibmveth_rxq_flags(adapter) & IBMVETH_RXQ_VALID);
 153 }
 154
 155 static inline int ibmveth_rxq_frame_offset(struct ibmveth_adapter *adapter)
 156 {
 157         return (ibmveth_rxq_flags(adapter) & IBMVETH_RXQ_OFF_MASK);
 158 }
 159
 160 static inline int ibmveth_rxq_frame_length(struct ibmveth_adapter *adapter)
 161 {
 162         return (adapter->rx_queue.queue_addr[adapter->rx_queue.index].length);
 163 }
 164
 165 static inline int ibmveth_rxq_csum_good(struct ibmveth_adapter *adapter)
 166 {
 167         return (ibmveth_rxq_flags(adapter) & IBMVETH_RXQ_CSUM_GOOD);
 168 }
 169
 170 /* setup the initial settings for a buffer pool */
 171 static void ibmveth_init_buffer_pool(struct ibmveth_buff_pool *pool, u32 pool_index, u32 pool_size, u32 buff_size, u32 pool_active)
 172 {
 173         pool->size = pool_size;
 174         pool->index = pool_index;
 175         pool->buff_size = buff_size;
 176         pool->threshold = pool_size / 2;
 177         pool->active = pool_active;
 178 }
 179
 180 /* allocate and setup an buffer pool - called during open */
 181 static int ibmveth_alloc_buffer_pool(struct ibmveth_buff_pool *pool)
 182 {
 183         int i;
 184
 185         pool->free_map = kmalloc(sizeof(u16) * pool->size, GFP_KERNEL);
 186
 187         if(!pool->free_map) {
 188                 return -1;
 189         }
 190
 191         pool->dma_addr = kmalloc(sizeof(dma_addr_t) * pool->size, GFP_KERNEL);
 192         if(!pool->dma_addr) {
 193                 kfree(pool->free_map);
 194                 pool->free_map = NULL;
 195                 return -1;
 196         }
 197
 198         pool->skbuff = kmalloc(sizeof(void*) * pool->size, GFP_KERNEL);
 199
 200         if(!pool->skbuff) {
 201                 kfree(pool->dma_addr);
 202                 pool->dma_addr = NULL;
 203
 204                 kfree(pool->free_map);
 205                 pool->free_map = NULL;
 206                 return -1;
 207         }
 208
 209         memset(pool->skbuff, 0, sizeof(void*) * pool->size);
 210         memset(pool->dma_addr, 0, sizeof(dma_addr_t) * pool->size);
 211
 212         for(i = 0; i < pool->size; ++i) {
 213                 pool->free_map[i] = i;
 214         }
 215
 216         atomic_set(&pool->available, 0);
 217         pool->producer_index = 0;
 218         pool->consumer_index = 0;
 219
 220         return 0;
 221 }
 222
 223 /* replenish the buffers for a pool.  note that we don't need to
 224  * skb_reserve these since they are used for incoming...
 225  */
 226 static void ibmveth_replenish_buffer_pool(struct ibmveth_adapter *adapter, struct ibmveth_buff_pool *pool)
 227 {
 228         u32 i;
 229         u32 count = pool->size - atomic_read(&pool->available);
 230         u32 buffers_added = 0;
 231
 232         mb();
 233
 234         for(i = 0; i < count; ++i) {
 235                 struct sk_buff *skb;
 236                 unsigned int free_index, index;
 237                 u64 correlator;
 238                 union ibmveth_buf_desc desc;
 239                 unsigned long lpar_rc;
 240                 dma_addr_t dma_addr;
 241
 242                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 243
 244                 if(!skb) {
 245                         ibmveth_debug_printk("replenish: unable to allocate skb\n");
 246                         adapter->replenish_no_mem++;
 247                         break;
 248                 }
 249
 250                 free_index = pool->consumer_index;
 251                 pool->consumer_index = (pool->consumer_index + 1) % pool->size;
 252                 index = pool->free_map[free_index];
 253
 254                 ibmveth_assert(index != IBM_VETH_INVALID_MAP);
 255                 ibmveth_assert(pool->skbuff[index] == NULL);
 256
 257                 dma_addr = dma_map_single(&adapter->vdev->dev, skb->data,
 258                                 pool->buff_size, DMA_FROM_DEVICE);
 259
 260                 pool->free_map[free_index] = IBM_VETH_INVALID_MAP;
 261                 pool->dma_addr[index] = dma_addr;
 262                 pool->skbuff[index] = skb;
 263
 264                 correlator = ((u64)pool->index << 32) | index;
 265                 *(u64*)skb->data = correlator;
 266
 267                 desc.fields.flags_len = IBMVETH_BUF_VALID | pool->buff_size;
 268                 desc.fields.address = dma_addr;
 269
 270                 lpar_rc = h_add_logical_lan_buffer(adapter->vdev->unit_address, desc.desc);
 271
 272                 if(lpar_rc != H_SUCCESS) {
 273                         pool->free_map[free_index] = index;
 274                         pool->skbuff[index] = NULL;
 275                         if (pool->consumer_index == 0)
 276                                 pool->consumer_index = pool->size - 1;
 277                         else
 278                                 pool->consumer_index--;
 279                         dma_unmap_single(&adapter->vdev->dev,
 280                                         pool->dma_addr[index], pool->buff_size,
 281                                         DMA_FROM_DEVICE);
 282                         dev_kfree_skb_any(skb);
 283                         adapter->replenish_add_buff_failure++;
 284                         break;
 285                 } else {
 286                         buffers_added++;
 287                         adapter->replenish_add_buff_success++;
 288                 }
 289         }
 290
 291         mb();
 292         atomic_add(buffers_added, &(pool->available));
 293 }
 294
 295 /* replenish routine */
 296 static void ibmveth_replenish_task(struct ibmveth_adapter *adapter)
 297 {
 298         int i;
 299
 300         adapter->replenish_task_cycles++;
 301
 302         for(i = 0; i < IbmVethNumBufferPools; i++)
 303                 if(adapter->rx_buff_pool[i].active)
 304                         ibmveth_replenish_buffer_pool(adapter,
 305                                                      &adapter->rx_buff_pool[i]);
 306
 307         adapter->rx_no_buffer = *(u64*)(((char*)adapter->buffer_list_addr) + 4096 - 8);
 308 }
 309
 310 /* empty and free ana buffer pool - also used to do cleanup in error paths */
 311 static void ibmveth_free_buffer_pool(struct ibmveth_adapter *adapter, struct ibmveth_buff_pool *pool)
 312 {
 313         int i;
 314
 315         kfree(pool->free_map);
 316         pool->free_map = NULL;
 317
 318         if(pool->skbuff && pool->dma_addr) {
 319                 for(i = 0; i < pool->size; ++i) {
 320                         struct sk_buff *skb = pool->skbuff[i];
 321                         if(skb) {
 322                                 dma_unmap_single(&adapter->vdev->dev,
 323                                                  pool->dma_addr[i],
 324                                                  pool->buff_size,
 325                                                  DMA_FROM_DEVICE);
 326                                 dev_kfree_skb_any(skb);
 327                                 pool->skbuff[i] = NULL;
 328                         }
 329                 }
 330         }
 331
 332         if(pool->dma_addr) {
 333                 kfree(pool->dma_addr);
 334                 pool->dma_addr = NULL;
 335         }
 336
 337         if(pool->skbuff) {
 338                 kfree(pool->skbuff);
 339                 pool->skbuff = NULL;
 340         }
 341 }
 342
 343 /* remove a buffer from a pool */
 344 static void ibmveth_remove_buffer_from_pool(struct ibmveth_adapter *adapter, u64 correlator)
 345 {
 346         unsigned int pool  = correlator >> 32;
 347         unsigned int index = correlator & 0xffffffffUL;
 348         unsigned int free_index;
 349         struct sk_buff *skb;
 350
 351         ibmveth_assert(pool < IbmVethNumBufferPools);
 352         ibmveth_assert(index < adapter->rx_buff_pool[pool].size);
 353
 354         skb = adapter->rx_buff_pool[pool].skbuff[index];
 355
 356         ibmveth_assert(skb != NULL);
 357
 358         adapter->rx_buff_pool[pool].skbuff[index] = NULL;
 359
 360         dma_unmap_single(&adapter->vdev->dev,
 361                          adapter->rx_buff_pool[pool].dma_addr[index],
 362                          adapter->rx_buff_pool[pool].buff_size,
 363                          DMA_FROM_DEVICE);
 364
 365         free_index = adapter->rx_buff_pool[pool].producer_index;
 366         adapter->rx_buff_pool[pool].producer_index
 367                 = (adapter->rx_buff_pool[pool].producer_index + 1)
 368                 % adapter->rx_buff_pool[pool].size;
 369         adapter->rx_buff_pool[pool].free_map[free_index] = index;
 370
 371         mb();
 372
 373         atomic_dec(&(adapter->rx_buff_pool[pool].available));
 374 }
 375
 376 /* get the current buffer on the rx queue */
 377 static inline struct sk_buff *ibmveth_rxq_get_buffer(struct ibmveth_adapter *adapter)
 378 {
 379         u64 correlator = adapter->rx_queue.queue_addr[adapter->rx_queue.index].correlator;
 380         unsigned int pool = correlator >> 32;
 381         unsigned int index = correlator & 0xffffffffUL;
 382
 383         ibmveth_assert(pool < IbmVethNumBufferPools);
 384         ibmveth_assert(index < adapter->rx_buff_pool[pool].size);
 385
 386         return adapter->rx_buff_pool[pool].skbuff[index];
 387 }
 388
 389 /* recycle the current buffer on the rx queue */
 390 static void ibmveth_rxq_recycle_buffer(struct ibmveth_adapter *adapter)
 391 {
 392         u32 q_index = adapter->rx_queue.index;
 393         u64 correlator = adapter->rx_queue.queue_addr[q_index].correlator;
 394         unsigned int pool = correlator >> 32;
 395         unsigned int index = correlator & 0xffffffffUL;
 396         union ibmveth_buf_desc desc;
 397         unsigned long lpar_rc;
 398
 399         ibmveth_assert(pool < IbmVethNumBufferPools);
 400         ibmveth_assert(index < adapter->rx_buff_pool[pool].size);
 401
 402         if(!adapter->rx_buff_pool[pool].active) {
 403                 ibmveth_rxq_harvest_buffer(adapter);
 404                 ibmveth_free_buffer_pool(adapter, &adapter->rx_buff_pool[pool]);
 405                 return;
 406         }
 407
 408         desc.fields.flags_len = IBMVETH_BUF_VALID |
 409                 adapter->rx_buff_pool[pool].buff_size;
 410         desc.fields.address = adapter->rx_buff_pool[pool].dma_addr[index];
 411
 412         lpar_rc = h_add_logical_lan_buffer(adapter->vdev->unit_address, desc.desc);
 413
 414         if(lpar_rc != H_SUCCESS) {
 415                 ibmveth_debug_printk("h_add_logical_lan_buffer failed during recycle rc=%ld", lpar_rc);
 416                 ibmveth_remove_buffer_from_pool(adapter, adapter->rx_queue.queue_addr[adapter->rx_queue.index].correlator);
 417         }
 418
 419         if(++adapter->rx_queue.index == adapter->rx_queue.num_slots) {
 420                 adapter->rx_queue.index = 0;
 421                 adapter->rx_queue.toggle = !adapter->rx_queue.toggle;
 422         }
 423 }
 424
 425 static void ibmveth_rxq_harvest_buffer(struct ibmveth_adapter *adapter)
 426 {
 427         ibmveth_remove_buffer_from_pool(adapter, adapter->rx_queue.queue_addr[adapter->rx_queue.index].correlator);
 428
 429         if(++adapter->rx_queue.index == adapter->rx_queue.num_slots) {
 430                 adapter->rx_queue.index = 0;
 431                 adapter->rx_queue.toggle = !adapter->rx_queue.toggle;
 432         }
 433 }
 434
 435 static void ibmveth_cleanup(struct ibmveth_adapter *adapter)
 436 {
 437         int i;
 438
 439         if(adapter->buffer_list_addr != NULL) {
 440                 if(!dma_mapping_error(adapter->buffer_list_dma)) {
 441                         dma_unmap_single(&adapter->vdev->dev,
 442                                         adapter->buffer_list_dma, 4096,
 443                                         DMA_BIDIRECTIONAL);
 444                         adapter->buffer_list_dma = DMA_ERROR_CODE;
 445                 }
 446                 free_page((unsigned long)adapter->buffer_list_addr);
 447                 adapter->buffer_list_addr = NULL;
 448         }
 449
 450         if(adapter->filter_list_addr != NULL) {
 451                 if(!dma_mapping_error(adapter->filter_list_dma)) {
 452                         dma_unmap_single(&adapter->vdev->dev,
 453                                         adapter->filter_list_dma, 4096,
 454                                         DMA_BIDIRECTIONAL);
 455                         adapter->filter_list_dma = DMA_ERROR_CODE;
 456                 }
 457                 free_page((unsigned long)adapter->filter_list_addr);
 458                 adapter->filter_list_addr = NULL;
 459         }
 460
 461         if(adapter->rx_queue.queue_addr != NULL) {
 462                 if(!dma_mapping_error(adapter->rx_queue.queue_dma)) {
 463                         dma_unmap_single(&adapter->vdev->dev,
 464                                         adapter->rx_queue.queue_dma,
 465                                         adapter->rx_queue.queue_len,
 466                                         DMA_BIDIRECTIONAL);
 467                         adapter->rx_queue.queue_dma = DMA_ERROR_CODE;
 468                 }
 469                 kfree(adapter->rx_queue.queue_addr);
 470                 adapter->rx_queue.queue_addr = NULL;
 471         }
 472
 473         for(i = 0; i<IbmVethNumBufferPools; i++)
 474                 if (adapter->rx_buff_pool[i].active)
 475                         ibmveth_free_buffer_pool(adapter,
 476                                                  &adapter->rx_buff_pool[i]);
 477 }
 478
 479 static int ibmveth_register_logical_lan(struct ibmveth_adapter *adapter,
 480         union ibmveth_buf_desc rxq_desc, u64 mac_address)
 481 {
 482         int rc, try_again = 1;
 483
 484         /* After a kexec the adapter will still be open, so our attempt to
 485         * open it will fail. So if we get a failure we free the adapter and
 486         * try again, but only once. */
 487 retry:
 488         rc = h_register_logical_lan(adapter->vdev->unit_address,
 489                                     adapter->buffer_list_dma, rxq_desc.desc,
 490                                     adapter->filter_list_dma, mac_address);
 491
 492         if (rc != H_SUCCESS && try_again) {
 493                 do {
 494                         rc = h_free_logical_lan(adapter->vdev->unit_address);
 495                 } while (H_IS_LONG_BUSY(rc) || (rc == H_BUSY));
 496
 497                 try_again = 0;
 498                 goto retry;
 499         }
 500
 501         return rc;
 502 }
 503
 504 static int ibmveth_open(struct net_device *netdev)
 505 {
 506         struct ibmveth_adapter *adapter = netdev->priv;
 507         u64 mac_address = 0;
 508         int rxq_entries = 1;
 509         unsigned long lpar_rc;
 510         int rc;
 511         union ibmveth_buf_desc rxq_desc;
 512         int i;
 513
 514         ibmveth_debug_printk("open starting\n");
 515
 516         napi_enable(&adapter->napi);
 517
 518         for(i = 0; i<IbmVethNumBufferPools; i++)
 519                 rxq_entries += adapter->rx_buff_pool[i].size;
 520
 521         adapter->buffer_list_addr = (void*) get_zeroed_page(GFP_KERNEL);
 522         adapter->filter_list_addr = (void*) get_zeroed_page(GFP_KERNEL);
 523
 524         if(!adapter->buffer_list_addr || !adapter->filter_list_addr) {
 525                 ibmveth_error_printk("unable to allocate filter or buffer list pages\n");
 526                 ibmveth_cleanup(adapter);
 527                 napi_disable(&adapter->napi);
 528                 return -ENOMEM;
 529         }
 530
 531         adapter->rx_queue.queue_len = sizeof(struct ibmveth_rx_q_entry) * rxq_entries;
 532         adapter->rx_queue.queue_addr = kmalloc(adapter->rx_queue.queue_len, GFP_KERNEL);
 533
 534         if(!adapter->rx_queue.queue_addr) {
 535                 ibmveth_error_printk("unable to allocate rx queue pages\n");
 536                 ibmveth_cleanup(adapter);
 537                 napi_disable(&adapter->napi);
 538                 return -ENOMEM;
 539         }
 540
 541         adapter->buffer_list_dma = dma_map_single(&adapter->vdev->dev,
 542                         adapter->buffer_list_addr, 4096, DMA_BIDIRECTIONAL);
 543         adapter->filter_list_dma = dma_map_single(&adapter->vdev->dev,
 544                         adapter->filter_list_addr, 4096, DMA_BIDIRECTIONAL);
 545         adapter->rx_queue.queue_dma = dma_map_single(&adapter->vdev->dev,
 546                         adapter->rx_queue.queue_addr,
 547                         adapter->rx_queue.queue_len, DMA_BIDIRECTIONAL);
 548
 549         if((dma_mapping_error(adapter->buffer_list_dma) ) ||
 550            (dma_mapping_error(adapter->filter_list_dma)) ||
 551            (dma_mapping_error(adapter->rx_queue.queue_dma))) {
 552                 ibmveth_error_printk("unable to map filter or buffer list pages\n");
 553                 ibmveth_cleanup(adapter);
 554                 napi_disable(&adapter->napi);
 555                 return -ENOMEM;
 556         }
 557
 558         adapter->rx_queue.index = 0;
 559         adapter->rx_queue.num_slots = rxq_entries;
 560         adapter->rx_queue.toggle = 1;
 561
 562         memcpy(&mac_address, netdev->dev_addr, netdev->addr_len);
 563         mac_address = mac_address >> 16;
 564
 565         rxq_desc.fields.flags_len = IBMVETH_BUF_VALID | adapter->rx_queue.queue_len;
 566         rxq_desc.fields.address = adapter->rx_queue.queue_dma;
 567
 568         ibmveth_debug_printk("buffer list @ 0x%p\n", adapter->buffer_list_addr);
 569         ibmveth_debug_printk("filter list @ 0x%p\n", adapter->filter_list_addr);
 570         ibmveth_debug_printk("receive q   @ 0x%p\n", adapter->rx_queue.queue_addr);
 571
 572         h_vio_signal(adapter->vdev->unit_address, VIO_IRQ_DISABLE);
 573
 574         lpar_rc = ibmveth_register_logical_lan(adapter, rxq_desc, mac_address);
 575
 576         if(lpar_rc != H_SUCCESS) {
 577                 ibmveth_error_printk("h_register_logical_lan failed with %ld\n", lpar_rc);
 578                 ibmveth_error_printk("buffer TCE:0x%lx filter TCE:0x%lx rxq desc:0x%lx MAC:0x%lx\n",
 579                                      adapter->buffer_list_dma,
 580                                      adapter->filter_list_dma,
 581                                      rxq_desc.desc,
 582                                      mac_address);
 583                 ibmveth_cleanup(adapter);
 584                 napi_disable(&adapter->napi);
 585                 return -ENONET;
 586         }
 587
 588         for(i = 0; i<IbmVethNumBufferPools; i++) {
 589                 if(!adapter->rx_buff_pool[i].active)
 590                         continue;
 591                 if (ibmveth_alloc_buffer_pool(&adapter->rx_buff_pool[i])) {
 592                         ibmveth_error_printk("unable to alloc pool\n");
 593                         adapter->rx_buff_pool[i].active = 0;
 594                         ibmveth_cleanup(adapter);
 595                         napi_disable(&adapter->napi);
 596                         return -ENOMEM ;
 597                 }
 598         }
 599
 600         ibmveth_debug_printk("registering irq 0x%x\n", netdev->irq);
 601         if((rc = request_irq(netdev->irq, &ibmveth_interrupt, 0, netdev->name, netdev)) != 0) {
 602                 ibmveth_error_printk("unable to request irq 0x%x, rc %d\n", netdev->irq, rc);
 603                 do {
 604                         rc = h_free_logical_lan(adapter->vdev->unit_address);
 605                 } while (H_IS_LONG_BUSY(rc) || (rc == H_BUSY));
 606
 607                 ibmveth_cleanup(adapter);
 608                 napi_disable(&adapter->napi);
 609                 return rc;
 610         }
 611
 612         ibmveth_debug_printk("initial replenish cycle\n");
 613         ibmveth_interrupt(netdev->irq, netdev);
 614
 615         netif_start_queue(netdev);
 616
 617         ibmveth_debug_printk("open complete\n");
 618
 619         return 0;
 620 }
 621
 622 static int ibmveth_close(struct net_device *netdev)
 623 {
 624         struct ibmveth_adapter *adapter = netdev->priv;
 625         long lpar_rc;
 626
 627         ibmveth_debug_printk("close starting\n");
 628
 629         napi_disable(&adapter->napi);
 630
 631         if (!adapter->pool_config)
 632                 netif_stop_queue(netdev);
 633
 634         free_irq(netdev->irq, netdev);
 635
 636         do {
 637                 lpar_rc = h_free_logical_lan(adapter->vdev->unit_address);
 638         } while (H_IS_LONG_BUSY(lpar_rc) || (lpar_rc == H_BUSY));
 639
 640         if(lpar_rc != H_SUCCESS)
 641         {
 642                 ibmveth_error_printk("h_free_logical_lan failed with %lx, continuing with close\n",
 643                                      lpar_rc);
 644         }
 645
 646         adapter->rx_no_buffer = *(u64*)(((char*)adapter->buffer_list_addr) + 4096 - 8);
 647
 648         ibmveth_cleanup(adapter);
 649
 650         ibmveth_debug_printk("close complete\n");
 651
 652         return 0;
 653 }
 654
 655 static int netdev_get_settings(struct net_device *dev, struct ethtool_cmd *cmd) {
 656         cmd->supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg | SUPPORTED_FIBRE);
 657         cmd->advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg | ADVERTISED_FIBRE);
 658         cmd->speed = SPEED_1000;
 659         cmd->duplex = DUPLEX_FULL;
 660         cmd->port = PORT_FIBRE;
 661         cmd->phy_address = 0;
 662         cmd->transceiver = XCVR_INTERNAL;
 663         cmd->autoneg = AUTONEG_ENABLE;
 664         cmd->maxtxpkt = 0;
 665         cmd->maxrxpkt = 1;
 666         return 0;
 667 }
 668
 669 static void netdev_get_drvinfo (struct net_device *dev, struct ethtool_drvinfo *info) {
 670         strncpy(info->driver, ibmveth_driver_name, sizeof(info->driver) - 1);
 671         strncpy(info->version, ibmveth_driver_version, sizeof(info->version) - 1);
 672 }
 673
 674 static u32 netdev_get_link(struct net_device *dev) {
 675         return 1;
 676 }
 677
 678 static void ibmveth_set_rx_csum_flags(struct net_device *dev, u32 data)
 679 {
 680         struct ibmveth_adapter *adapter = dev->priv;
 681
 682         if (data)
 683                 adapter->rx_csum = 1;
 684         else {
 685                 /*
 686                  * Since the ibmveth firmware interface does not have the concept of
 687                  * separate tx/rx checksum offload enable, if rx checksum is disabled
 688                  * we also have to disable tx checksum offload. Once we disable rx
 689                  * checksum offload, we are no longer allowed to send tx buffers that
 690                  * are not properly checksummed.
 691                  */
 692                 adapter->rx_csum = 0;
 693                 dev->features &= ~NETIF_F_IP_CSUM;
 694         }
 695 }
 696
 697 static void ibmveth_set_tx_csum_flags(struct net_device *dev, u32 data)
 698 {
 699         struct ibmveth_adapter *adapter = dev->priv;
 700
 701         if (data) {
 702                 dev->features |= NETIF_F_IP_CSUM;
 703                 adapter->rx_csum = 1;
 704         } else
 705                 dev->features &= ~NETIF_F_IP_CSUM;
 706 }
 707
 708 static int ibmveth_set_csum_offload(struct net_device *dev, u32 data,
 709                                     void (*done) (struct net_device *, u32))
 710 {
 711         struct ibmveth_adapter *adapter = dev->priv;
 712         u64 set_attr, clr_attr, ret_attr;
 713         long ret;
 714         int rc1 = 0, rc2 = 0;
 715         int restart = 0;
 716
 717         if (netif_running(dev)) {
 718                 restart = 1;
 719                 adapter->pool_config = 1;
 720                 ibmveth_close(dev);
 721                 adapter->pool_config = 0;
 722         }
 723
 724         set_attr = 0;
 725         clr_attr = 0;
 726
 727         if (data)
 728                 set_attr = IBMVETH_ILLAN_IPV4_TCP_CSUM;
 729         else
 730                 clr_attr = IBMVETH_ILLAN_IPV4_TCP_CSUM;
 731
 732         ret = h_illan_attributes(adapter->vdev->unit_address, 0, 0, &ret_attr);
 733
 734         if (ret == H_SUCCESS && !(ret_attr & IBMVETH_ILLAN_ACTIVE_TRUNK) &&
 735             !(ret_attr & IBMVETH_ILLAN_TRUNK_PRI_MASK) &&
 736             (ret_attr & IBMVETH_ILLAN_PADDED_PKT_CSUM)) {
 737                 ret = h_illan_attributes(adapter->vdev->unit_address, clr_attr,
 738                                          set_attr, &ret_attr);
 739
 740                 if (ret != H_SUCCESS) {
 741                         rc1 = -EIO;
 742                         ibmveth_error_printk("unable to change checksum offload settings."
 743                                              " %d rc=%ld\n", data, ret);
 744
 745                         ret = h_illan_attributes(adapter->vdev->unit_address,
 746                                                  set_attr, clr_attr, &ret_attr);
 747                 } else
 748                         done(dev, data);
 749         } else {
 750                 rc1 = -EIO;
 751                 ibmveth_error_printk("unable to change checksum offload settings."
 752                                      " %d rc=%ld ret_attr=%lx\n", data, ret, ret_attr);
 753         }
 754
 755         if (restart)
 756                 rc2 = ibmveth_open(dev);
 757
 758         return rc1 ? rc1 : rc2;
 759 }
 760
 761 static int ibmveth_set_rx_csum(struct net_device *dev, u32 data)
 762 {
 763         struct ibmveth_adapter *adapter = dev->priv;
 764
 765         if ((data && adapter->rx_csum) || (!data && !adapter->rx_csum))
 766                 return 0;
 767
 768         return ibmveth_set_csum_offload(dev, data, ibmveth_set_rx_csum_flags);
 769 }
 770
 771 static int ibmveth_set_tx_csum(struct net_device *dev, u32 data)
 772 {
 773         struct ibmveth_adapter *adapter = dev->priv;
 774         int rc = 0;
 775
 776         if (data && (dev->features & NETIF_F_IP_CSUM))
 777                 return 0;
 778         if (!data && !(dev->features & NETIF_F_IP_CSUM))
 779                 return 0;
 780
 781         if (data && !adapter->rx_csum)
 782                 rc = ibmveth_set_csum_offload(dev, data, ibmveth_set_tx_csum_flags);
 783         else
 784                 ibmveth_set_tx_csum_flags(dev, data);
 785
 786         return rc;
 787 }
 788
 789 static u32 ibmveth_get_rx_csum(struct net_device *dev)
 790 {
 791         struct ibmveth_adapter *adapter = dev->priv;
 792         return adapter->rx_csum;
 793 }
 794
 795 static void ibmveth_get_strings(struct net_device *dev, u32 stringset, u8 *data)
 796 {
 797         int i;
 798
 799         if (stringset != ETH_SS_STATS)
 800                 return;
 801
 802         for (i = 0; i < ARRAY_SIZE(ibmveth_stats); i++, data += ETH_GSTRING_LEN)
 803                 memcpy(data, ibmveth_stats[i].name, ETH_GSTRING_LEN);
 804 }
 805
 806 static int ibmveth_get_stats_count(struct net_device *dev)
 807 {
 808         return ARRAY_SIZE(ibmveth_stats);
 809 }
 810
 811 static void ibmveth_get_ethtool_stats(struct net_device *dev,
 812                                       struct ethtool_stats *stats, u64 *data)
 813 {
 814         int i;
 815         struct ibmveth_adapter *adapter = dev->priv;
 816
 817         for (i = 0; i < ARRAY_SIZE(ibmveth_stats); i++)
 818                 data[i] = IBMVETH_GET_STAT(adapter, ibmveth_stats[i].offset);
 819 }
 820
 821 static const struct ethtool_ops netdev_ethtool_ops = {
 822         .get_drvinfo            = netdev_get_drvinfo,
 823         .get_settings           = netdev_get_settings,
 824         .get_link               = netdev_get_link,
 825         .get_sg                 = ethtool_op_get_sg,
 826         .get_tx_csum            = ethtool_op_get_tx_csum,
 827         .set_tx_csum            = ibmveth_set_tx_csum,
 828         .get_rx_csum            = ibmveth_get_rx_csum,
 829         .set_rx_csum            = ibmveth_set_rx_csum,
 830         .get_tso                        = ethtool_op_get_tso,
 831         .get_ufo                        = ethtool_op_get_ufo,
 832         .get_strings            = ibmveth_get_strings,
 833         .get_stats_count                = ibmveth_get_stats_count,
 834         .get_ethtool_stats      = ibmveth_get_ethtool_stats,
 835 };
 836
 837 static int ibmveth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
 838 {
 839         return -EOPNOTSUPP;
 840 }
 841
 842 #define page_offset(v) ((unsigned long)(v) & ((1 << 12) - 1))
 843
 844 static int ibmveth_start_xmit(struct sk_buff *skb, struct net_device *netdev)
 845 {
 846         struct ibmveth_adapter *adapter = netdev->priv;
 847         union ibmveth_buf_desc desc;
 848         unsigned long lpar_rc;
 849         unsigned long correlator;
 850         unsigned long flags;
 851         unsigned int retry_count;
 852         unsigned int tx_dropped = 0;
 853         unsigned int tx_bytes = 0;
 854         unsigned int tx_packets = 0;
 855         unsigned int tx_send_failed = 0;
 856         unsigned int tx_map_failed = 0;
 857
 858         desc.fields.flags_len = IBMVETH_BUF_VALID | skb->len;
 859         desc.fields.address = dma_map_single(&adapter->vdev->dev, skb->data,
 860                                              skb->len, DMA_TO_DEVICE);
 861
 862         if (skb->ip_summed == CHECKSUM_PARTIAL &&
 863             ip_hdr(skb)->protocol != IPPROTO_TCP && skb_checksum_help(skb)) {
 864                 ibmveth_error_printk("tx: failed to checksum packet\n");
 865                 tx_dropped++;
 866                 goto out;
 867         }
 868
 869         if (skb->ip_summed == CHECKSUM_PARTIAL) {
 870                 unsigned char *buf = skb_transport_header(skb) + skb->csum_offset;
 871
 872                 desc.fields.flags_len |= (IBMVETH_BUF_NO_CSUM | IBMVETH_BUF_CSUM_GOOD);
 873
 874                 /* Need to zero out the checksum */
 875                 buf[0] = 0;
 876                 buf[1] = 0;
 877         }
 878
 879         if (dma_mapping_error(desc.fields.address)) {
 880                 ibmveth_error_printk("tx: unable to map xmit buffer\n");
 881                 tx_map_failed++;
 882                 tx_dropped++;
 883                 goto out;
 884         }
 885
 886         /* send the frame. Arbitrarily set retrycount to 1024 */
 887         correlator = 0;
 888         retry_count = 1024;
 889         do {
 890                 lpar_rc = h_send_logical_lan(adapter->vdev->unit_address,
 891                                              desc.desc, 0, 0, 0, 0, 0,
 892                                              correlator, &correlator);
 893         } while ((lpar_rc == H_BUSY) && (retry_count--));
 894
 895         if(lpar_rc != H_SUCCESS && lpar_rc != H_DROPPED) {
 896                 ibmveth_error_printk("tx: h_send_logical_lan failed with rc=%ld\n", lpar_rc);
 897                 ibmveth_error_printk("tx: valid=%d, len=%d, address=0x%08x\n",
 898                                      (desc.fields.flags_len & IBMVETH_BUF_VALID) ? 1 : 0,
 899                                      skb->len, desc.fields.address);
 900                 tx_send_failed++;
 901                 tx_dropped++;
 902         } else {
 903                 tx_packets++;
 904                 tx_bytes += skb->len;
 905                 netdev->trans_start = jiffies;
 906         }
 907
 908         dma_unmap_single(&adapter->vdev->dev, desc.fields.address,
 909                          skb->len, DMA_TO_DEVICE);
 910
 911 out:    spin_lock_irqsave(&adapter->stats_lock, flags);
 912         adapter->stats.tx_dropped += tx_dropped;
 913         adapter->stats.tx_bytes += tx_bytes;
 914         adapter->stats.tx_packets += tx_packets;
 915         adapter->tx_send_failed += tx_send_failed;
 916         adapter->tx_map_failed += tx_map_failed;
 917         spin_unlock_irqrestore(&adapter->stats_lock, flags);
 918
 919         dev_kfree_skb(skb);
 920         return 0;
 921 }
 922
 923 static int ibmveth_poll(struct napi_struct *napi, int budget)
 924 {
 925         struct ibmveth_adapter *adapter = container_of(napi, struct ibmveth_adapter, napi);
 926         struct net_device *netdev = adapter->netdev;
 927         int frames_processed = 0;
 928         unsigned long lpar_rc;
 929
 930  restart_poll:
 931         do {
 932                 struct sk_buff *skb;
 933
 934                 if (!ibmveth_rxq_pending_buffer(adapter))
 935                         break;
 936
 937                 rmb();
 938                 if (!ibmveth_rxq_buffer_valid(adapter)) {
 939                         wmb(); /* suggested by larson1 */
 940                         adapter->rx_invalid_buffer++;
 941                         ibmveth_debug_printk("recycling invalid buffer\n");
 942                         ibmveth_rxq_recycle_buffer(adapter);
 943                 } else {
 944                         int length = ibmveth_rxq_frame_length(adapter);
 945                         int offset = ibmveth_rxq_frame_offset(adapter);
 946                         int csum_good = ibmveth_rxq_csum_good(adapter);
 947
 948                         skb = ibmveth_rxq_get_buffer(adapter);
 949                         if (csum_good)
 950                                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 951
 952                         ibmveth_rxq_harvest_buffer(adapter);
 953
 954                         skb_reserve(skb, offset);
 955                         skb_put(skb, length);
 956                         skb->protocol = eth_type_trans(skb, netdev);
 957
 958                         netif_receive_skb(skb); /* send it up */
 959
 960                         adapter->stats.rx_packets++;
 961                         adapter->stats.rx_bytes += length;
 962                         frames_processed++;
 963                         netdev->last_rx = jiffies;
 964                 }
 965         } while (frames_processed < budget);
 966
 967         ibmveth_replenish_task(adapter);
 968
 969         if (frames_processed < budget) {
 970                 /* We think we are done - reenable interrupts,
 971                  * then check once more to make sure we are done.
 972                  */
 973                 lpar_rc = h_vio_signal(adapter->vdev->unit_address,
 974                                        VIO_IRQ_ENABLE);
 975
 976                 ibmveth_assert(lpar_rc == H_SUCCESS);
 977
 978                 netif_rx_complete(netdev, napi);
 979
 980                 if (ibmveth_rxq_pending_buffer(adapter) &&
 981                     netif_rx_reschedule(netdev, napi)) {
 982                         lpar_rc = h_vio_signal(adapter->vdev->unit_address,
 983                                                VIO_IRQ_DISABLE);
 984                         goto restart_poll;
 985                 }
 986         }
 987
 988         return frames_processed;
 989 }
 990
 991 static irqreturn_t ibmveth_interrupt(int irq, void *dev_instance)
 992 {
 993         struct net_device *netdev = dev_instance;
 994         struct ibmveth_adapter *adapter = netdev->priv;
 995         unsigned long lpar_rc;
 996
 997         if (netif_rx_schedule_prep(netdev, &adapter->napi)) {
 998                 lpar_rc = h_vio_signal(adapter->vdev->unit_address,
 999                                        VIO_IRQ_DISABLE);
1000                 ibmveth_assert(lpar_rc == H_SUCCESS);
1001                 __netif_rx_schedule(netdev, &adapter->napi);
1002         }
1003         return IRQ_HANDLED;
1004 }
1005
1006 static struct net_device_stats *ibmveth_get_stats(struct net_device *dev)
1007 {
1008         struct ibmveth_adapter *adapter = dev->priv;
1009         return &adapter->stats;
1010 }
1011
1012 static void ibmveth_set_multicast_list(struct net_device *netdev)
1013 {
1014         struct ibmveth_adapter *adapter = netdev->priv;
1015         unsigned long lpar_rc;
1016
1017         if((netdev->flags & IFF_PROMISC) || (netdev->mc_count > adapter->mcastFilterSize)) {
1018                 lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
1019                                            IbmVethMcastEnableRecv |
1020                                            IbmVethMcastDisableFiltering,
1021                                            0);
1022                 if(lpar_rc != H_SUCCESS) {
1023                         ibmveth_error_printk("h_multicast_ctrl rc=%ld when entering promisc mode\n", lpar_rc);
1024                 }
1025         } else {
1026                 struct dev_mc_list *mclist = netdev->mc_list;
1027                 int i;
1028                 /* clear the filter table & disable filtering */
1029                 lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
1030                                            IbmVethMcastEnableRecv |
1031                                            IbmVethMcastDisableFiltering |
1032                                            IbmVethMcastClearFilterTable,
1033                                            0);
1034                 if(lpar_rc != H_SUCCESS) {
1035                         ibmveth_error_printk("h_multicast_ctrl rc=%ld when attempting to clear filter table\n", lpar_rc);
1036                 }
1037                 /* add the addresses to the filter table */
1038                 for(i = 0; i < netdev->mc_count; ++i, mclist = mclist->next) {
1039                         // add the multicast address to the filter table
1040                         unsigned long mcast_addr = 0;
1041                         memcpy(((char *)&mcast_addr)+2, mclist->dmi_addr, 6);
1042                         lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
1043                                                    IbmVethMcastAddFilter,
1044                                                    mcast_addr);
1045                         if(lpar_rc != H_SUCCESS) {
1046                                 ibmveth_error_printk("h_multicast_ctrl rc=%ld when adding an entry to the filter table\n", lpar_rc);
1047                         }
1048                 }
1049
1050                 /* re-enable filtering */
1051                 lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
1052                                            IbmVethMcastEnableFiltering,
1053                                            0);
1054                 if(lpar_rc != H_SUCCESS) {
1055                         ibmveth_error_printk("h_multicast_ctrl rc=%ld when enabling filtering\n", lpar_rc);
1056                 }
1057         }
1058 }
1059
1060 static int ibmveth_change_mtu(struct net_device *dev, int new_mtu)
1061 {
1062         struct ibmveth_adapter *adapter = dev->priv;
1063         int new_mtu_oh = new_mtu + IBMVETH_BUFF_OH;
1064         int reinit = 0;
1065         int i, rc;
1066
1067         if (new_mtu < IBMVETH_MAX_MTU)
1068                 return -EINVAL;
1069
1070         for (i = 0; i < IbmVethNumBufferPools; i++)
1071                 if (new_mtu_oh < adapter->rx_buff_pool[i].buff_size)
1072                         break;
1073
1074         if (i == IbmVethNumBufferPools)
1075                 return -EINVAL;
1076
1077         /* Look for an active buffer pool that can hold the new MTU */
1078         for(i = 0; i<IbmVethNumBufferPools; i++) {
1079                 if (!adapter->rx_buff_pool[i].active) {
1080                         adapter->rx_buff_pool[i].active = 1;
1081                         reinit = 1;
1082                 }
1083
1084                 if (new_mtu_oh < adapter->rx_buff_pool[i].buff_size) {
1085                         if (reinit && netif_running(adapter->netdev)) {
1086                                 adapter->pool_config = 1;
1087                                 ibmveth_close(adapter->netdev);
1088                                 adapter->pool_config = 0;
1089                                 dev->mtu = new_mtu;
1090                                 if ((rc = ibmveth_open(adapter->netdev)))
1091                                         return rc;
1092                         } else
1093                                 dev->mtu = new_mtu;
1094                         return 0;
1095                 }
1096         }
1097         return -EINVAL;
1098 }
1099
1100 #ifdef CONFIG_NET_POLL_CONTROLLER
1101 static void ibmveth_poll_controller(struct net_device *dev)
1102 {
1103         ibmveth_replenish_task(dev->priv);
1104         ibmveth_interrupt(dev->irq, dev);
1105 }
1106 #endif
1107
1108 static int __devinit ibmveth_probe(struct vio_dev *dev, const struct vio_device_id *id)
1109 {
1110         int rc, i;
1111         long ret;
1112         struct net_device *netdev;
1113         struct ibmveth_adapter *adapter;
1114         u64 set_attr, ret_attr;
1115
1116         unsigned char *mac_addr_p;
1117         unsigned int *mcastFilterSize_p;
1118
1119
1120         ibmveth_debug_printk_no_adapter("entering ibmveth_probe for UA 0x%x\n",
1121                                         dev->unit_address);
1122
1123         mac_addr_p = (unsigned char *) vio_get_attribute(dev,
1124                                                 VETH_MAC_ADDR, NULL);
1125         if(!mac_addr_p) {
1126                 printk(KERN_ERR "(%s:%3.3d) ERROR: Can't find VETH_MAC_ADDR "
1127                                 "attribute\n", __FILE__, __LINE__);
1128                 return 0;
1129         }
1130
1131         mcastFilterSize_p = (unsigned int *) vio_get_attribute(dev,
1132                                                 VETH_MCAST_FILTER_SIZE, NULL);
1133         if(!mcastFilterSize_p) {
1134                 printk(KERN_ERR "(%s:%3.3d) ERROR: Can't find "
1135                                 "VETH_MCAST_FILTER_SIZE attribute\n",
1136                                 __FILE__, __LINE__);
1137                 return 0;
1138         }
1139
1140         netdev = alloc_etherdev(sizeof(struct ibmveth_adapter));
1141
1142         if(!netdev)
1143                 return -ENOMEM;
1144
1145         adapter = netdev->priv;
1146         dev->dev.driver_data = netdev;
1147
1148         adapter->vdev = dev;
1149         adapter->netdev = netdev;
1150         adapter->mcastFilterSize= *mcastFilterSize_p;
1151         adapter->pool_config = 0;
1152
1153         netif_napi_add(netdev, &adapter->napi, ibmveth_poll, 16);
1154
1155         /*      Some older boxes running PHYP non-natively have an OF that
1156                 returns a 8-byte local-mac-address field (and the first
1157                 2 bytes have to be ignored) while newer boxes' OF return
1158                 a 6-byte field. Note that IEEE 1275 specifies that
1159                 local-mac-address must be a 6-byte field.
1160                 The RPA doc specifies that the first byte must be 10b, so
1161                 we'll just look for it to solve this 8 vs. 6 byte field issue */
1162
1163         if ((*mac_addr_p & 0x3) != 0x02)
1164                 mac_addr_p += 2;
1165
1166         adapter->mac_addr = 0;
1167         memcpy(&adapter->mac_addr, mac_addr_p, 6);
1168
1169         netdev->irq = dev->irq;
1170         netdev->open               = ibmveth_open;
1171         netdev->stop               = ibmveth_close;
1172         netdev->hard_start_xmit    = ibmveth_start_xmit;
1173         netdev->get_stats          = ibmveth_get_stats;
1174         netdev->set_multicast_list = ibmveth_set_multicast_list;
1175         netdev->do_ioctl           = ibmveth_ioctl;
1176         netdev->ethtool_ops           = &netdev_ethtool_ops;
1177         netdev->change_mtu         = ibmveth_change_mtu;
1178         SET_NETDEV_DEV(netdev, &dev->dev);
1179 #ifdef CONFIG_NET_POLL_CONTROLLER
1180         netdev->poll_controller = ibmveth_poll_controller;
1181 #endif
1182         netdev->features |= NETIF_F_LLTX;
1183         spin_lock_init(&adapter->stats_lock);
1184
1185         memcpy(&netdev->dev_addr, &adapter->mac_addr, netdev->addr_len);
1186
1187         for(i = 0; i<IbmVethNumBufferPools; i++) {
1188                 struct kobject *kobj = &adapter->rx_buff_pool[i].kobj;
1189                 ibmveth_init_buffer_pool(&adapter->rx_buff_pool[i], i,
1190                                          pool_count[i], pool_size[i],
1191                                          pool_active[i]);
1192                 kobj->parent = &dev->dev.kobj;
1193                 sprintf(kobj->name, "pool%d", i);
1194                 kobj->ktype = &ktype_veth_pool;
1195                 kobject_register(kobj);
1196         }
1197
1198         ibmveth_debug_printk("adapter @ 0x%p\n", adapter);
1199
1200         adapter->buffer_list_dma = DMA_ERROR_CODE;
1201         adapter->filter_list_dma = DMA_ERROR_CODE;
1202         adapter->rx_queue.queue_dma = DMA_ERROR_CODE;
1203
1204         ibmveth_debug_printk("registering netdev...\n");
1205
1206         ret = h_illan_attributes(dev->unit_address, 0, 0, &ret_attr);
1207
1208         if (ret == H_SUCCESS && !(ret_attr & IBMVETH_ILLAN_ACTIVE_TRUNK) &&
1209             !(ret_attr & IBMVETH_ILLAN_TRUNK_PRI_MASK) &&
1210             (ret_attr & IBMVETH_ILLAN_PADDED_PKT_CSUM)) {
1211                 set_attr = IBMVETH_ILLAN_IPV4_TCP_CSUM;
1212
1213                 ret = h_illan_attributes(dev->unit_address, 0, set_attr, &ret_attr);
1214
1215                 if (ret == H_SUCCESS) {
1216                         adapter->rx_csum = 1;
1217                         netdev->features |= NETIF_F_IP_CSUM;
1218                 } else
1219                         ret = h_illan_attributes(dev->unit_address, set_attr, 0, &ret_attr);
1220         }
1221
1222         rc = register_netdev(netdev);
1223
1224         if(rc) {
1225                 ibmveth_debug_printk("failed to register netdev rc=%d\n", rc);
1226                 free_netdev(netdev);
1227                 return rc;
1228         }
1229
1230         ibmveth_debug_printk("registered\n");
1231
1232         ibmveth_proc_register_adapter(adapter);
1233
1234         return 0;
1235 }
1236
1237 static int __devexit ibmveth_remove(struct vio_dev *dev)
1238 {
1239         struct net_device *netdev = dev->dev.driver_data;
1240         struct ibmveth_adapter *adapter = netdev->priv;
1241         int i;
1242
1243         for(i = 0; i<IbmVethNumBufferPools; i++)
1244                 kobject_unregister(&adapter->rx_buff_pool[i].kobj);
1245
1246         unregister_netdev(netdev);
1247
1248         ibmveth_proc_unregister_adapter(adapter);
1249
1250         free_netdev(netdev);
1251         return 0;
1252 }
1253
1254 #ifdef CONFIG_PROC_FS
1255 static void ibmveth_proc_register_driver(void)
1256 {
1257         ibmveth_proc_dir = proc_mkdir(IBMVETH_PROC_DIR, init_net.proc_net);
1258         if (ibmveth_proc_dir) {
1259         }
1260 }
1261
1262 static void ibmveth_proc_unregister_driver(void)
1263 {
1264         remove_proc_entry(IBMVETH_PROC_DIR, init_net.proc_net);
1265 }
1266
1267 static void *ibmveth_seq_start(struct seq_file *seq, loff_t *pos)
1268 {
1269         if (*pos == 0) {
1270                 return (void *)1;
1271         } else {
1272                 return NULL;
1273         }
1274 }
1275
1276 static void *ibmveth_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1277 {
1278         ++*pos;
1279         return NULL;
1280 }
1281
1282 static void ibmveth_seq_stop(struct seq_file *seq, void *v)
1283 {
1284 }
1285
1286 static int ibmveth_seq_show(struct seq_file *seq, void *v)
1287 {
1288         struct ibmveth_adapter *adapter = seq->private;
1289         char *current_mac = ((char*) &adapter->netdev->dev_addr);
1290         char *firmware_mac = ((char*) &adapter->mac_addr) ;
1291
1292         seq_printf(seq, "%s %s\n\n", ibmveth_driver_string, ibmveth_driver_version);
1293
1294         seq_printf(seq, "Unit Address:    0x%x\n", adapter->vdev->unit_address);
1295         seq_printf(seq, "Current MAC:     %02X:%02X:%02X:%02X:%02X:%02X\n",
1296                    current_mac[0], current_mac[1], current_mac[2],
1297                    current_mac[3], current_mac[4], current_mac[5]);
1298         seq_printf(seq, "Firmware MAC:    %02X:%02X:%02X:%02X:%02X:%02X\n",
1299                    firmware_mac[0], firmware_mac[1], firmware_mac[2],
1300                    firmware_mac[3], firmware_mac[4], firmware_mac[5]);
1301
1302         seq_printf(seq, "\nAdapter Statistics:\n");
1303         seq_printf(seq, "  TX:  vio_map_single failres:      %ld\n", adapter->tx_map_failed);
1304         seq_printf(seq, "       send failures:               %ld\n", adapter->tx_send_failed);
1305         seq_printf(seq, "  RX:  replenish task cycles:       %ld\n", adapter->replenish_task_cycles);
1306         seq_printf(seq, "       alloc_skb_failures:          %ld\n", adapter->replenish_no_mem);
1307         seq_printf(seq, "       add buffer failures:         %ld\n", adapter->replenish_add_buff_failure);
1308         seq_printf(seq, "       invalid buffers:             %ld\n", adapter->rx_invalid_buffer);
1309         seq_printf(seq, "       no buffers:                  %ld\n", adapter->rx_no_buffer);
1310
1311         return 0;
1312 }
1313 static struct seq_operations ibmveth_seq_ops = {
1314         .start = ibmveth_seq_start,
1315         .next  = ibmveth_seq_next,
1316         .stop  = ibmveth_seq_stop,
1317         .show  = ibmveth_seq_show,
1318 };
1319
1320 static int ibmveth_proc_open(struct inode *inode, struct file *file)
1321 {
1322         struct seq_file *seq;
1323         struct proc_dir_entry *proc;
1324         int rc;
1325
1326         rc = seq_open(file, &ibmveth_seq_ops);
1327         if (!rc) {
1328                 /* recover the pointer buried in proc_dir_entry data */
1329                 seq = file->private_data;
1330                 proc = PDE(inode);
1331                 seq->private = proc->data;
1332         }
1333         return rc;
1334 }
1335
1336 static const struct file_operations ibmveth_proc_fops = {
1337         .owner   = THIS_MODULE,
1338         .open    = ibmveth_proc_open,
1339         .read    = seq_read,
1340         .llseek  = seq_lseek,
1341         .release = seq_release,
1342 };
1343
1344 static void ibmveth_proc_register_adapter(struct ibmveth_adapter *adapter)
1345 {
1346         struct proc_dir_entry *entry;
1347         if (ibmveth_proc_dir) {
1348                 char u_addr[10];
1349                 sprintf(u_addr, "%x", adapter->vdev->unit_address);
1350                 entry = create_proc_entry(u_addr, S_IFREG, ibmveth_proc_dir);
1351                 if (!entry) {
1352                         ibmveth_error_printk("Cannot create adapter proc entry");
1353                 } else {
1354                         entry->data = (void *) adapter;
1355                         entry->proc_fops = &ibmveth_proc_fops;
1356                 }
1357         }
1358         return;
1359 }
1360
1361 static void ibmveth_proc_unregister_adapter(struct ibmveth_adapter *adapter)
1362 {
1363         if (ibmveth_proc_dir) {
1364                 char u_addr[10];
1365                 sprintf(u_addr, "%x", adapter->vdev->unit_address);
1366                 remove_proc_entry(u_addr, ibmveth_proc_dir);
1367         }
1368 }
1369
1370 #else /* CONFIG_PROC_FS */
1371 static void ibmveth_proc_register_adapter(struct ibmveth_adapter *adapter)
1372 {
1373 }
1374
1375 static void ibmveth_proc_unregister_adapter(struct ibmveth_adapter *adapter)
1376 {
1377 }
1378 static void ibmveth_proc_register_driver(void)
1379 {
1380 }
1381
1382 static void ibmveth_proc_unregister_driver(void)
1383 {
1384 }
1385 #endif /* CONFIG_PROC_FS */
1386
1387 static struct attribute veth_active_attr;
1388 static struct attribute veth_num_attr;
1389 static struct attribute veth_size_attr;
1390
1391 static ssize_t veth_pool_show(struct kobject * kobj,
1392                               struct attribute * attr, char * buf)
1393 {
1394         struct ibmveth_buff_pool *pool = container_of(kobj,
1395                                                       struct ibmveth_buff_pool,
1396                                                       kobj);
1397
1398         if (attr == &veth_active_attr)
1399                 return sprintf(buf, "%d\n", pool->active);
1400         else if (attr == &veth_num_attr)
1401                 return sprintf(buf, "%d\n", pool->size);
1402         else if (attr == &veth_size_attr)
1403                 return sprintf(buf, "%d\n", pool->buff_size);
1404         return 0;
1405 }
1406
1407 static ssize_t veth_pool_store(struct kobject * kobj, struct attribute * attr,
1408 const char * buf, size_t count)
1409 {
1410         struct ibmveth_buff_pool *pool = container_of(kobj,
1411                                                       struct ibmveth_buff_pool,
1412                                                       kobj);
1413         struct net_device *netdev =
1414             container_of(kobj->parent, struct device, kobj)->driver_data;
1415         struct ibmveth_adapter *adapter = netdev->priv;
1416         long value = simple_strtol(buf, NULL, 10);
1417         long rc;
1418
1419         if (attr == &veth_active_attr) {
1420                 if (value && !pool->active) {
1421                         if (netif_running(netdev)) {
1422                                 if(ibmveth_alloc_buffer_pool(pool)) {
1423                                         ibmveth_error_printk("unable to alloc pool\n");
1424                                         return -ENOMEM;
1425                                 }
1426                                 pool->active = 1;
1427                                 adapter->pool_config = 1;
1428                                 ibmveth_close(netdev);
1429                                 adapter->pool_config = 0;
1430                                 if ((rc = ibmveth_open(netdev)))
1431                                         return rc;
1432                         } else
1433                                 pool->active = 1;
1434                 } else if (!value && pool->active) {
1435                         int mtu = netdev->mtu + IBMVETH_BUFF_OH;
1436                         int i;
1437                         /* Make sure there is a buffer pool with buffers that
1438                            can hold a packet of the size of the MTU */
1439                         for (i = 0; i < IbmVethNumBufferPools; i++) {
1440                                 if (pool == &adapter->rx_buff_pool[i])
1441                                         continue;
1442                                 if (!adapter->rx_buff_pool[i].active)
1443                                         continue;
1444                                 if (mtu <= adapter->rx_buff_pool[i].buff_size)
1445                                         break;
1446                         }
1447
1448                         if (i == IbmVethNumBufferPools) {
1449                                 ibmveth_error_printk("no active pool >= MTU\n");
1450                                 return -EPERM;
1451                         }
1452
1453                         pool->active = 0;
1454                         if (netif_running(netdev)) {
1455                                 adapter->pool_config = 1;
1456                                 ibmveth_close(netdev);
1457                                 adapter->pool_config = 0;
1458                                 if ((rc = ibmveth_open(netdev)))
1459                                         return rc;
1460                         }
1461                 }
1462         } else if (attr == &veth_num_attr) {
1463                 if (value <= 0 || value > IBMVETH_MAX_POOL_COUNT)
1464                         return -EINVAL;
1465                 else {
1466                         if (netif_running(netdev)) {
1467                                 adapter->pool_config = 1;
1468                                 ibmveth_close(netdev);
1469                                 adapter->pool_config = 0;
1470                                 pool->size = value;
1471                                 if ((rc = ibmveth_open(netdev)))
1472                                         return rc;
1473                         } else
1474                                 pool->size = value;
1475                 }
1476         } else if (attr == &veth_size_attr) {
1477                 if (value <= IBMVETH_BUFF_OH || value > IBMVETH_MAX_BUF_SIZE)
1478                         return -EINVAL;
1479                 else {
1480                         if (netif_running(netdev)) {
1481                                 adapter->pool_config = 1;
1482                                 ibmveth_close(netdev);
1483                                 adapter->pool_config = 0;
1484                                 pool->buff_size = value;
1485                                 if ((rc = ibmveth_open(netdev)))
1486                                         return rc;
1487                         } else
1488                                 pool->buff_size = value;
1489                 }
1490         }
1491
1492         /* kick the interrupt handler to allocate/deallocate pools */
1493         ibmveth_interrupt(netdev->irq, netdev);
1494         return count;
1495 }
1496
1497
1498 #define ATTR(_name, _mode)      \
1499         struct attribute veth_##_name##_attr = {               \
1500         .name = __stringify(_name), .mode = _mode, \
1501         };
1502
1503 static ATTR(active, 0644);
1504 static ATTR(num, 0644);
1505 static ATTR(size, 0644);
1506
1507 static struct attribute * veth_pool_attrs[] = {
1508         &veth_active_attr,
1509         &veth_num_attr,
1510         &veth_size_attr,
1511         NULL,
1512 };
1513
1514 static struct sysfs_ops veth_pool_ops = {
1515         .show   = veth_pool_show,
1516         .store  = veth_pool_store,
1517 };
1518
1519 static struct kobj_type ktype_veth_pool = {
1520         .release        = NULL,
1521         .sysfs_ops      = &veth_pool_ops,
1522         .default_attrs  = veth_pool_attrs,
1523 };
1524
1525
1526 static struct vio_device_id ibmveth_device_table[] __devinitdata= {
1527         { "network", "IBM,l-lan"},
1528         { "", "" }
1529 };
1530 MODULE_DEVICE_TABLE(vio, ibmveth_device_table);
1531
1532 static struct vio_driver ibmveth_driver = {
1533         .id_table       = ibmveth_device_table,
1534         .probe          = ibmveth_probe,
1535         .remove         = ibmveth_remove,
1536         .driver         = {
1537                 .name   = ibmveth_driver_name,
1538                 .owner  = THIS_MODULE,
1539         }
1540 };
1541
1542 static int __init ibmveth_module_init(void)
1543 {
1544         ibmveth_printk("%s: %s %s\n", ibmveth_driver_name, ibmveth_driver_string, ibmveth_driver_version);
1545
1546         ibmveth_proc_register_driver();
1547
1548         return vio_register_driver(&ibmveth_driver);
1549 }
1550
1551 static void __exit ibmveth_module_exit(void)
1552 {
1553         vio_unregister_driver(&ibmveth_driver);
1554         ibmveth_proc_unregister_driver();
1555 }
1556
1557 module_init(ibmveth_module_init);
1558 module_exit(ibmveth_module_exit);