2 * IBM Power Virtual Ethernet Device Driver
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License as published by
6 * the Free Software Foundation; either version 2 of the License, or
7 * (at your option) any later version.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, see <http://www.gnu.org/licenses/>.
17 * Copyright (C) IBM Corporation, 2003, 2010
19 * Authors: Dave Larson <larson1@us.ibm.com>
20 * Santiago Leon <santil@linux.vnet.ibm.com>
21 * Brian King <brking@linux.vnet.ibm.com>
22 * Robert Jennings <rcj@linux.vnet.ibm.com>
23 * Anton Blanchard <anton@au.ibm.com>
26 #include <linux/module.h>
27 #include <linux/moduleparam.h>
28 #include <linux/types.h>
29 #include <linux/errno.h>
30 #include <linux/dma-mapping.h>
31 #include <linux/kernel.h>
32 #include <linux/netdevice.h>
33 #include <linux/etherdevice.h>
34 #include <linux/skbuff.h>
35 #include <linux/init.h>
36 #include <linux/interrupt.h>
39 #include <linux/ethtool.h>
42 #include <linux/ipv6.h>
43 #include <linux/slab.h>
44 #include <asm/hvcall.h>
45 #include <linux/atomic.h>
47 #include <asm/iommu.h>
48 #include <asm/firmware.h>
52 static irqreturn_t
ibmveth_interrupt(int irq
, void *dev_instance
);
53 static void ibmveth_rxq_harvest_buffer(struct ibmveth_adapter
*adapter
);
54 static unsigned long ibmveth_get_desired_dma(struct vio_dev
*vdev
);
56 static struct kobj_type ktype_veth_pool
;
59 static const char ibmveth_driver_name
[] = "ibmveth";
60 static const char ibmveth_driver_string
[] = "IBM Power Virtual Ethernet Driver";
61 #define ibmveth_driver_version "1.06"
63 MODULE_AUTHOR("Santiago Leon <santil@linux.vnet.ibm.com>");
64 MODULE_DESCRIPTION("IBM Power Virtual Ethernet Driver");
65 MODULE_LICENSE("GPL");
66 MODULE_VERSION(ibmveth_driver_version
);
68 static unsigned int tx_copybreak __read_mostly
= 128;
69 module_param(tx_copybreak
, uint
, 0644);
70 MODULE_PARM_DESC(tx_copybreak
,
71 "Maximum size of packet that is copied to a new buffer on transmit");
73 static unsigned int rx_copybreak __read_mostly
= 128;
74 module_param(rx_copybreak
, uint
, 0644);
75 MODULE_PARM_DESC(rx_copybreak
,
76 "Maximum size of packet that is copied to a new buffer on receive");
78 static unsigned int rx_flush __read_mostly
= 0;
79 module_param(rx_flush
, uint
, 0644);
80 MODULE_PARM_DESC(rx_flush
, "Flush receive buffers before use");
82 static bool old_large_send __read_mostly
;
83 module_param(old_large_send
, bool, S_IRUGO
);
84 MODULE_PARM_DESC(old_large_send
,
85 "Use old large send method on firmware that supports the new method");
88 char name
[ETH_GSTRING_LEN
];
92 #define IBMVETH_STAT_OFF(stat) offsetof(struct ibmveth_adapter, stat)
93 #define IBMVETH_GET_STAT(a, off) *((u64 *)(((unsigned long)(a)) + off))
95 struct ibmveth_stat ibmveth_stats
[] = {
96 { "replenish_task_cycles", IBMVETH_STAT_OFF(replenish_task_cycles
) },
97 { "replenish_no_mem", IBMVETH_STAT_OFF(replenish_no_mem
) },
98 { "replenish_add_buff_failure",
99 IBMVETH_STAT_OFF(replenish_add_buff_failure
) },
100 { "replenish_add_buff_success",
101 IBMVETH_STAT_OFF(replenish_add_buff_success
) },
102 { "rx_invalid_buffer", IBMVETH_STAT_OFF(rx_invalid_buffer
) },
103 { "rx_no_buffer", IBMVETH_STAT_OFF(rx_no_buffer
) },
104 { "tx_map_failed", IBMVETH_STAT_OFF(tx_map_failed
) },
105 { "tx_send_failed", IBMVETH_STAT_OFF(tx_send_failed
) },
106 { "fw_enabled_ipv4_csum", IBMVETH_STAT_OFF(fw_ipv4_csum_support
) },
107 { "fw_enabled_ipv6_csum", IBMVETH_STAT_OFF(fw_ipv6_csum_support
) },
108 { "tx_large_packets", IBMVETH_STAT_OFF(tx_large_packets
) },
109 { "rx_large_packets", IBMVETH_STAT_OFF(rx_large_packets
) },
110 { "fw_enabled_large_send", IBMVETH_STAT_OFF(fw_large_send_support
) }
113 /* simple methods of getting data from the current rxq entry */
114 static inline u32
ibmveth_rxq_flags(struct ibmveth_adapter
*adapter
)
116 return be32_to_cpu(adapter
->rx_queue
.queue_addr
[adapter
->rx_queue
.index
].flags_off
);
119 static inline int ibmveth_rxq_toggle(struct ibmveth_adapter
*adapter
)
121 return (ibmveth_rxq_flags(adapter
) & IBMVETH_RXQ_TOGGLE
) >>
122 IBMVETH_RXQ_TOGGLE_SHIFT
;
125 static inline int ibmveth_rxq_pending_buffer(struct ibmveth_adapter
*adapter
)
127 return ibmveth_rxq_toggle(adapter
) == adapter
->rx_queue
.toggle
;
130 static inline int ibmveth_rxq_buffer_valid(struct ibmveth_adapter
*adapter
)
132 return ibmveth_rxq_flags(adapter
) & IBMVETH_RXQ_VALID
;
135 static inline int ibmveth_rxq_frame_offset(struct ibmveth_adapter
*adapter
)
137 return ibmveth_rxq_flags(adapter
) & IBMVETH_RXQ_OFF_MASK
;
140 static inline int ibmveth_rxq_large_packet(struct ibmveth_adapter
*adapter
)
142 return ibmveth_rxq_flags(adapter
) & IBMVETH_RXQ_LRG_PKT
;
145 static inline int ibmveth_rxq_frame_length(struct ibmveth_adapter
*adapter
)
147 return be32_to_cpu(adapter
->rx_queue
.queue_addr
[adapter
->rx_queue
.index
].length
);
150 static inline int ibmveth_rxq_csum_good(struct ibmveth_adapter
*adapter
)
152 return ibmveth_rxq_flags(adapter
) & IBMVETH_RXQ_CSUM_GOOD
;
155 /* setup the initial settings for a buffer pool */
156 static void ibmveth_init_buffer_pool(struct ibmveth_buff_pool
*pool
,
157 u32 pool_index
, u32 pool_size
,
158 u32 buff_size
, u32 pool_active
)
160 pool
->size
= pool_size
;
161 pool
->index
= pool_index
;
162 pool
->buff_size
= buff_size
;
163 pool
->threshold
= pool_size
* 7 / 8;
164 pool
->active
= pool_active
;
167 /* allocate and setup an buffer pool - called during open */
168 static int ibmveth_alloc_buffer_pool(struct ibmveth_buff_pool
*pool
)
172 pool
->free_map
= kmalloc(sizeof(u16
) * pool
->size
, GFP_KERNEL
);
177 pool
->dma_addr
= kcalloc(pool
->size
, sizeof(dma_addr_t
), GFP_KERNEL
);
178 if (!pool
->dma_addr
) {
179 kfree(pool
->free_map
);
180 pool
->free_map
= NULL
;
184 pool
->skbuff
= kcalloc(pool
->size
, sizeof(void *), GFP_KERNEL
);
187 kfree(pool
->dma_addr
);
188 pool
->dma_addr
= NULL
;
190 kfree(pool
->free_map
);
191 pool
->free_map
= NULL
;
195 for (i
= 0; i
< pool
->size
; ++i
)
196 pool
->free_map
[i
] = i
;
198 atomic_set(&pool
->available
, 0);
199 pool
->producer_index
= 0;
200 pool
->consumer_index
= 0;
205 static inline void ibmveth_flush_buffer(void *addr
, unsigned long length
)
207 unsigned long offset
;
209 for (offset
= 0; offset
< length
; offset
+= SMP_CACHE_BYTES
)
210 asm("dcbfl %0,%1" :: "b" (addr
), "r" (offset
));
213 /* replenish the buffers for a pool. note that we don't need to
214 * skb_reserve these since they are used for incoming...
216 static void ibmveth_replenish_buffer_pool(struct ibmveth_adapter
*adapter
,
217 struct ibmveth_buff_pool
*pool
)
220 u32 count
= pool
->size
- atomic_read(&pool
->available
);
221 u32 buffers_added
= 0;
223 unsigned int free_index
, index
;
225 unsigned long lpar_rc
;
230 for (i
= 0; i
< count
; ++i
) {
231 union ibmveth_buf_desc desc
;
233 skb
= netdev_alloc_skb(adapter
->netdev
, pool
->buff_size
);
236 netdev_dbg(adapter
->netdev
,
237 "replenish: unable to allocate skb\n");
238 adapter
->replenish_no_mem
++;
242 free_index
= pool
->consumer_index
;
243 pool
->consumer_index
++;
244 if (pool
->consumer_index
>= pool
->size
)
245 pool
->consumer_index
= 0;
246 index
= pool
->free_map
[free_index
];
248 BUG_ON(index
== IBM_VETH_INVALID_MAP
);
249 BUG_ON(pool
->skbuff
[index
] != NULL
);
251 dma_addr
= dma_map_single(&adapter
->vdev
->dev
, skb
->data
,
252 pool
->buff_size
, DMA_FROM_DEVICE
);
254 if (dma_mapping_error(&adapter
->vdev
->dev
, dma_addr
))
257 pool
->free_map
[free_index
] = IBM_VETH_INVALID_MAP
;
258 pool
->dma_addr
[index
] = dma_addr
;
259 pool
->skbuff
[index
] = skb
;
261 correlator
= ((u64
)pool
->index
<< 32) | index
;
262 *(u64
*)skb
->data
= correlator
;
264 desc
.fields
.flags_len
= IBMVETH_BUF_VALID
| pool
->buff_size
;
265 desc
.fields
.address
= dma_addr
;
268 unsigned int len
= min(pool
->buff_size
,
269 adapter
->netdev
->mtu
+
271 ibmveth_flush_buffer(skb
->data
, len
);
273 lpar_rc
= h_add_logical_lan_buffer(adapter
->vdev
->unit_address
,
276 if (lpar_rc
!= H_SUCCESS
) {
280 adapter
->replenish_add_buff_success
++;
285 atomic_add(buffers_added
, &(pool
->available
));
289 pool
->free_map
[free_index
] = index
;
290 pool
->skbuff
[index
] = NULL
;
291 if (pool
->consumer_index
== 0)
292 pool
->consumer_index
= pool
->size
- 1;
294 pool
->consumer_index
--;
295 if (!dma_mapping_error(&adapter
->vdev
->dev
, dma_addr
))
296 dma_unmap_single(&adapter
->vdev
->dev
,
297 pool
->dma_addr
[index
], pool
->buff_size
,
299 dev_kfree_skb_any(skb
);
300 adapter
->replenish_add_buff_failure
++;
303 atomic_add(buffers_added
, &(pool
->available
));
307 * The final 8 bytes of the buffer list is a counter of frames dropped
308 * because there was not a buffer in the buffer list capable of holding
311 static void ibmveth_update_rx_no_buffer(struct ibmveth_adapter
*adapter
)
313 __be64
*p
= adapter
->buffer_list_addr
+ 4096 - 8;
315 adapter
->rx_no_buffer
= be64_to_cpup(p
);
318 /* replenish routine */
319 static void ibmveth_replenish_task(struct ibmveth_adapter
*adapter
)
323 adapter
->replenish_task_cycles
++;
325 for (i
= (IBMVETH_NUM_BUFF_POOLS
- 1); i
>= 0; i
--) {
326 struct ibmveth_buff_pool
*pool
= &adapter
->rx_buff_pool
[i
];
329 (atomic_read(&pool
->available
) < pool
->threshold
))
330 ibmveth_replenish_buffer_pool(adapter
, pool
);
333 ibmveth_update_rx_no_buffer(adapter
);
336 /* empty and free ana buffer pool - also used to do cleanup in error paths */
337 static void ibmveth_free_buffer_pool(struct ibmveth_adapter
*adapter
,
338 struct ibmveth_buff_pool
*pool
)
342 kfree(pool
->free_map
);
343 pool
->free_map
= NULL
;
345 if (pool
->skbuff
&& pool
->dma_addr
) {
346 for (i
= 0; i
< pool
->size
; ++i
) {
347 struct sk_buff
*skb
= pool
->skbuff
[i
];
349 dma_unmap_single(&adapter
->vdev
->dev
,
353 dev_kfree_skb_any(skb
);
354 pool
->skbuff
[i
] = NULL
;
359 if (pool
->dma_addr
) {
360 kfree(pool
->dma_addr
);
361 pool
->dma_addr
= NULL
;
370 /* remove a buffer from a pool */
371 static void ibmveth_remove_buffer_from_pool(struct ibmveth_adapter
*adapter
,
374 unsigned int pool
= correlator
>> 32;
375 unsigned int index
= correlator
& 0xffffffffUL
;
376 unsigned int free_index
;
379 BUG_ON(pool
>= IBMVETH_NUM_BUFF_POOLS
);
380 BUG_ON(index
>= adapter
->rx_buff_pool
[pool
].size
);
382 skb
= adapter
->rx_buff_pool
[pool
].skbuff
[index
];
386 adapter
->rx_buff_pool
[pool
].skbuff
[index
] = NULL
;
388 dma_unmap_single(&adapter
->vdev
->dev
,
389 adapter
->rx_buff_pool
[pool
].dma_addr
[index
],
390 adapter
->rx_buff_pool
[pool
].buff_size
,
393 free_index
= adapter
->rx_buff_pool
[pool
].producer_index
;
394 adapter
->rx_buff_pool
[pool
].producer_index
++;
395 if (adapter
->rx_buff_pool
[pool
].producer_index
>=
396 adapter
->rx_buff_pool
[pool
].size
)
397 adapter
->rx_buff_pool
[pool
].producer_index
= 0;
398 adapter
->rx_buff_pool
[pool
].free_map
[free_index
] = index
;
402 atomic_dec(&(adapter
->rx_buff_pool
[pool
].available
));
405 /* get the current buffer on the rx queue */
406 static inline struct sk_buff
*ibmveth_rxq_get_buffer(struct ibmveth_adapter
*adapter
)
408 u64 correlator
= adapter
->rx_queue
.queue_addr
[adapter
->rx_queue
.index
].correlator
;
409 unsigned int pool
= correlator
>> 32;
410 unsigned int index
= correlator
& 0xffffffffUL
;
412 BUG_ON(pool
>= IBMVETH_NUM_BUFF_POOLS
);
413 BUG_ON(index
>= adapter
->rx_buff_pool
[pool
].size
);
415 return adapter
->rx_buff_pool
[pool
].skbuff
[index
];
418 /* recycle the current buffer on the rx queue */
419 static int ibmveth_rxq_recycle_buffer(struct ibmveth_adapter
*adapter
)
421 u32 q_index
= adapter
->rx_queue
.index
;
422 u64 correlator
= adapter
->rx_queue
.queue_addr
[q_index
].correlator
;
423 unsigned int pool
= correlator
>> 32;
424 unsigned int index
= correlator
& 0xffffffffUL
;
425 union ibmveth_buf_desc desc
;
426 unsigned long lpar_rc
;
429 BUG_ON(pool
>= IBMVETH_NUM_BUFF_POOLS
);
430 BUG_ON(index
>= adapter
->rx_buff_pool
[pool
].size
);
432 if (!adapter
->rx_buff_pool
[pool
].active
) {
433 ibmveth_rxq_harvest_buffer(adapter
);
434 ibmveth_free_buffer_pool(adapter
, &adapter
->rx_buff_pool
[pool
]);
438 desc
.fields
.flags_len
= IBMVETH_BUF_VALID
|
439 adapter
->rx_buff_pool
[pool
].buff_size
;
440 desc
.fields
.address
= adapter
->rx_buff_pool
[pool
].dma_addr
[index
];
442 lpar_rc
= h_add_logical_lan_buffer(adapter
->vdev
->unit_address
, desc
.desc
);
444 if (lpar_rc
!= H_SUCCESS
) {
445 netdev_dbg(adapter
->netdev
, "h_add_logical_lan_buffer failed "
446 "during recycle rc=%ld", lpar_rc
);
447 ibmveth_remove_buffer_from_pool(adapter
, adapter
->rx_queue
.queue_addr
[adapter
->rx_queue
.index
].correlator
);
451 if (++adapter
->rx_queue
.index
== adapter
->rx_queue
.num_slots
) {
452 adapter
->rx_queue
.index
= 0;
453 adapter
->rx_queue
.toggle
= !adapter
->rx_queue
.toggle
;
460 static void ibmveth_rxq_harvest_buffer(struct ibmveth_adapter
*adapter
)
462 ibmveth_remove_buffer_from_pool(adapter
, adapter
->rx_queue
.queue_addr
[adapter
->rx_queue
.index
].correlator
);
464 if (++adapter
->rx_queue
.index
== adapter
->rx_queue
.num_slots
) {
465 adapter
->rx_queue
.index
= 0;
466 adapter
->rx_queue
.toggle
= !adapter
->rx_queue
.toggle
;
470 static void ibmveth_cleanup(struct ibmveth_adapter
*adapter
)
473 struct device
*dev
= &adapter
->vdev
->dev
;
475 if (adapter
->buffer_list_addr
!= NULL
) {
476 if (!dma_mapping_error(dev
, adapter
->buffer_list_dma
)) {
477 dma_unmap_single(dev
, adapter
->buffer_list_dma
, 4096,
479 adapter
->buffer_list_dma
= DMA_ERROR_CODE
;
481 free_page((unsigned long)adapter
->buffer_list_addr
);
482 adapter
->buffer_list_addr
= NULL
;
485 if (adapter
->filter_list_addr
!= NULL
) {
486 if (!dma_mapping_error(dev
, adapter
->filter_list_dma
)) {
487 dma_unmap_single(dev
, adapter
->filter_list_dma
, 4096,
489 adapter
->filter_list_dma
= DMA_ERROR_CODE
;
491 free_page((unsigned long)adapter
->filter_list_addr
);
492 adapter
->filter_list_addr
= NULL
;
495 if (adapter
->rx_queue
.queue_addr
!= NULL
) {
496 dma_free_coherent(dev
, adapter
->rx_queue
.queue_len
,
497 adapter
->rx_queue
.queue_addr
,
498 adapter
->rx_queue
.queue_dma
);
499 adapter
->rx_queue
.queue_addr
= NULL
;
502 for (i
= 0; i
< IBMVETH_NUM_BUFF_POOLS
; i
++)
503 if (adapter
->rx_buff_pool
[i
].active
)
504 ibmveth_free_buffer_pool(adapter
,
505 &adapter
->rx_buff_pool
[i
]);
507 if (adapter
->bounce_buffer
!= NULL
) {
508 if (!dma_mapping_error(dev
, adapter
->bounce_buffer_dma
)) {
509 dma_unmap_single(&adapter
->vdev
->dev
,
510 adapter
->bounce_buffer_dma
,
511 adapter
->netdev
->mtu
+ IBMVETH_BUFF_OH
,
513 adapter
->bounce_buffer_dma
= DMA_ERROR_CODE
;
515 kfree(adapter
->bounce_buffer
);
516 adapter
->bounce_buffer
= NULL
;
520 static int ibmveth_register_logical_lan(struct ibmveth_adapter
*adapter
,
521 union ibmveth_buf_desc rxq_desc
, u64 mac_address
)
523 int rc
, try_again
= 1;
526 * After a kexec the adapter will still be open, so our attempt to
527 * open it will fail. So if we get a failure we free the adapter and
528 * try again, but only once.
531 rc
= h_register_logical_lan(adapter
->vdev
->unit_address
,
532 adapter
->buffer_list_dma
, rxq_desc
.desc
,
533 adapter
->filter_list_dma
, mac_address
);
535 if (rc
!= H_SUCCESS
&& try_again
) {
537 rc
= h_free_logical_lan(adapter
->vdev
->unit_address
);
538 } while (H_IS_LONG_BUSY(rc
) || (rc
== H_BUSY
));
547 static u64
ibmveth_encode_mac_addr(u8
*mac
)
552 for (i
= 0; i
< ETH_ALEN
; i
++)
553 encoded
= (encoded
<< 8) | mac
[i
];
558 static int ibmveth_open(struct net_device
*netdev
)
560 struct ibmveth_adapter
*adapter
= netdev_priv(netdev
);
563 unsigned long lpar_rc
;
565 union ibmveth_buf_desc rxq_desc
;
569 netdev_dbg(netdev
, "open starting\n");
571 napi_enable(&adapter
->napi
);
573 for(i
= 0; i
< IBMVETH_NUM_BUFF_POOLS
; i
++)
574 rxq_entries
+= adapter
->rx_buff_pool
[i
].size
;
576 adapter
->buffer_list_addr
= (void*) get_zeroed_page(GFP_KERNEL
);
577 adapter
->filter_list_addr
= (void*) get_zeroed_page(GFP_KERNEL
);
579 if (!adapter
->buffer_list_addr
|| !adapter
->filter_list_addr
) {
580 netdev_err(netdev
, "unable to allocate filter or buffer list "
586 dev
= &adapter
->vdev
->dev
;
588 adapter
->rx_queue
.queue_len
= sizeof(struct ibmveth_rx_q_entry
) *
590 adapter
->rx_queue
.queue_addr
=
591 dma_alloc_coherent(dev
, adapter
->rx_queue
.queue_len
,
592 &adapter
->rx_queue
.queue_dma
, GFP_KERNEL
);
593 if (!adapter
->rx_queue
.queue_addr
) {
598 adapter
->buffer_list_dma
= dma_map_single(dev
,
599 adapter
->buffer_list_addr
, 4096, DMA_BIDIRECTIONAL
);
600 adapter
->filter_list_dma
= dma_map_single(dev
,
601 adapter
->filter_list_addr
, 4096, DMA_BIDIRECTIONAL
);
603 if ((dma_mapping_error(dev
, adapter
->buffer_list_dma
)) ||
604 (dma_mapping_error(dev
, adapter
->filter_list_dma
))) {
605 netdev_err(netdev
, "unable to map filter or buffer list "
611 adapter
->rx_queue
.index
= 0;
612 adapter
->rx_queue
.num_slots
= rxq_entries
;
613 adapter
->rx_queue
.toggle
= 1;
615 mac_address
= ibmveth_encode_mac_addr(netdev
->dev_addr
);
617 rxq_desc
.fields
.flags_len
= IBMVETH_BUF_VALID
|
618 adapter
->rx_queue
.queue_len
;
619 rxq_desc
.fields
.address
= adapter
->rx_queue
.queue_dma
;
621 netdev_dbg(netdev
, "buffer list @ 0x%p\n", adapter
->buffer_list_addr
);
622 netdev_dbg(netdev
, "filter list @ 0x%p\n", adapter
->filter_list_addr
);
623 netdev_dbg(netdev
, "receive q @ 0x%p\n", adapter
->rx_queue
.queue_addr
);
625 h_vio_signal(adapter
->vdev
->unit_address
, VIO_IRQ_DISABLE
);
627 lpar_rc
= ibmveth_register_logical_lan(adapter
, rxq_desc
, mac_address
);
629 if (lpar_rc
!= H_SUCCESS
) {
630 netdev_err(netdev
, "h_register_logical_lan failed with %ld\n",
632 netdev_err(netdev
, "buffer TCE:0x%llx filter TCE:0x%llx rxq "
633 "desc:0x%llx MAC:0x%llx\n",
634 adapter
->buffer_list_dma
,
635 adapter
->filter_list_dma
,
642 for (i
= 0; i
< IBMVETH_NUM_BUFF_POOLS
; i
++) {
643 if (!adapter
->rx_buff_pool
[i
].active
)
645 if (ibmveth_alloc_buffer_pool(&adapter
->rx_buff_pool
[i
])) {
646 netdev_err(netdev
, "unable to alloc pool\n");
647 adapter
->rx_buff_pool
[i
].active
= 0;
653 netdev_dbg(netdev
, "registering irq 0x%x\n", netdev
->irq
);
654 rc
= request_irq(netdev
->irq
, ibmveth_interrupt
, 0, netdev
->name
,
657 netdev_err(netdev
, "unable to request irq 0x%x, rc %d\n",
660 lpar_rc
= h_free_logical_lan(adapter
->vdev
->unit_address
);
661 } while (H_IS_LONG_BUSY(lpar_rc
) || (lpar_rc
== H_BUSY
));
666 adapter
->bounce_buffer
=
667 kmalloc(netdev
->mtu
+ IBMVETH_BUFF_OH
, GFP_KERNEL
);
668 if (!adapter
->bounce_buffer
) {
670 goto err_out_free_irq
;
672 adapter
->bounce_buffer_dma
=
673 dma_map_single(&adapter
->vdev
->dev
, adapter
->bounce_buffer
,
674 netdev
->mtu
+ IBMVETH_BUFF_OH
, DMA_BIDIRECTIONAL
);
675 if (dma_mapping_error(dev
, adapter
->bounce_buffer_dma
)) {
676 netdev_err(netdev
, "unable to map bounce buffer\n");
678 goto err_out_free_irq
;
681 netdev_dbg(netdev
, "initial replenish cycle\n");
682 ibmveth_interrupt(netdev
->irq
, netdev
);
684 netif_start_queue(netdev
);
686 netdev_dbg(netdev
, "open complete\n");
691 free_irq(netdev
->irq
, netdev
);
693 ibmveth_cleanup(adapter
);
694 napi_disable(&adapter
->napi
);
698 static int ibmveth_close(struct net_device
*netdev
)
700 struct ibmveth_adapter
*adapter
= netdev_priv(netdev
);
703 netdev_dbg(netdev
, "close starting\n");
705 napi_disable(&adapter
->napi
);
707 if (!adapter
->pool_config
)
708 netif_stop_queue(netdev
);
710 h_vio_signal(adapter
->vdev
->unit_address
, VIO_IRQ_DISABLE
);
713 lpar_rc
= h_free_logical_lan(adapter
->vdev
->unit_address
);
714 } while (H_IS_LONG_BUSY(lpar_rc
) || (lpar_rc
== H_BUSY
));
716 if (lpar_rc
!= H_SUCCESS
) {
717 netdev_err(netdev
, "h_free_logical_lan failed with %lx, "
718 "continuing with close\n", lpar_rc
);
721 free_irq(netdev
->irq
, netdev
);
723 ibmveth_update_rx_no_buffer(adapter
);
725 ibmveth_cleanup(adapter
);
727 netdev_dbg(netdev
, "close complete\n");
732 static int netdev_get_settings(struct net_device
*dev
, struct ethtool_cmd
*cmd
)
734 cmd
->supported
= (SUPPORTED_1000baseT_Full
| SUPPORTED_Autoneg
|
736 cmd
->advertising
= (ADVERTISED_1000baseT_Full
| ADVERTISED_Autoneg
|
738 ethtool_cmd_speed_set(cmd
, SPEED_1000
);
739 cmd
->duplex
= DUPLEX_FULL
;
740 cmd
->port
= PORT_FIBRE
;
741 cmd
->phy_address
= 0;
742 cmd
->transceiver
= XCVR_INTERNAL
;
743 cmd
->autoneg
= AUTONEG_ENABLE
;
749 static void netdev_get_drvinfo(struct net_device
*dev
,
750 struct ethtool_drvinfo
*info
)
752 strlcpy(info
->driver
, ibmveth_driver_name
, sizeof(info
->driver
));
753 strlcpy(info
->version
, ibmveth_driver_version
, sizeof(info
->version
));
756 static netdev_features_t
ibmveth_fix_features(struct net_device
*dev
,
757 netdev_features_t features
)
760 * Since the ibmveth firmware interface does not have the
761 * concept of separate tx/rx checksum offload enable, if rx
762 * checksum is disabled we also have to disable tx checksum
763 * offload. Once we disable rx checksum offload, we are no
764 * longer allowed to send tx buffers that are not properly
768 if (!(features
& NETIF_F_RXCSUM
))
769 features
&= ~NETIF_F_CSUM_MASK
;
774 static int ibmveth_set_csum_offload(struct net_device
*dev
, u32 data
)
776 struct ibmveth_adapter
*adapter
= netdev_priv(dev
);
777 unsigned long set_attr
, clr_attr
, ret_attr
;
778 unsigned long set_attr6
, clr_attr6
;
779 long ret
, ret4
, ret6
;
780 int rc1
= 0, rc2
= 0;
783 if (netif_running(dev
)) {
785 adapter
->pool_config
= 1;
787 adapter
->pool_config
= 0;
796 set_attr
= IBMVETH_ILLAN_IPV4_TCP_CSUM
;
797 set_attr6
= IBMVETH_ILLAN_IPV6_TCP_CSUM
;
799 clr_attr
= IBMVETH_ILLAN_IPV4_TCP_CSUM
;
800 clr_attr6
= IBMVETH_ILLAN_IPV6_TCP_CSUM
;
803 ret
= h_illan_attributes(adapter
->vdev
->unit_address
, 0, 0, &ret_attr
);
805 if (ret
== H_SUCCESS
&& !(ret_attr
& IBMVETH_ILLAN_ACTIVE_TRUNK
) &&
806 !(ret_attr
& IBMVETH_ILLAN_TRUNK_PRI_MASK
) &&
807 (ret_attr
& IBMVETH_ILLAN_PADDED_PKT_CSUM
)) {
808 ret4
= h_illan_attributes(adapter
->vdev
->unit_address
, clr_attr
,
809 set_attr
, &ret_attr
);
811 if (ret4
!= H_SUCCESS
) {
812 netdev_err(dev
, "unable to change IPv4 checksum "
813 "offload settings. %d rc=%ld\n",
816 h_illan_attributes(adapter
->vdev
->unit_address
,
817 set_attr
, clr_attr
, &ret_attr
);
820 dev
->features
&= ~NETIF_F_IP_CSUM
;
823 adapter
->fw_ipv4_csum_support
= data
;
826 ret6
= h_illan_attributes(adapter
->vdev
->unit_address
,
827 clr_attr6
, set_attr6
, &ret_attr
);
829 if (ret6
!= H_SUCCESS
) {
830 netdev_err(dev
, "unable to change IPv6 checksum "
831 "offload settings. %d rc=%ld\n",
834 h_illan_attributes(adapter
->vdev
->unit_address
,
835 set_attr6
, clr_attr6
, &ret_attr
);
838 dev
->features
&= ~NETIF_F_IPV6_CSUM
;
841 adapter
->fw_ipv6_csum_support
= data
;
843 if (ret4
== H_SUCCESS
|| ret6
== H_SUCCESS
)
844 adapter
->rx_csum
= data
;
849 netdev_err(dev
, "unable to change checksum offload settings."
850 " %d rc=%ld ret_attr=%lx\n", data
, ret
,
855 rc2
= ibmveth_open(dev
);
857 return rc1
? rc1
: rc2
;
860 static int ibmveth_set_tso(struct net_device
*dev
, u32 data
)
862 struct ibmveth_adapter
*adapter
= netdev_priv(dev
);
863 unsigned long set_attr
, clr_attr
, ret_attr
;
865 int rc1
= 0, rc2
= 0;
868 if (netif_running(dev
)) {
870 adapter
->pool_config
= 1;
872 adapter
->pool_config
= 0;
879 set_attr
= IBMVETH_ILLAN_LRG_SR_ENABLED
;
881 clr_attr
= IBMVETH_ILLAN_LRG_SR_ENABLED
;
883 ret1
= h_illan_attributes(adapter
->vdev
->unit_address
, 0, 0, &ret_attr
);
885 if (ret1
== H_SUCCESS
&& (ret_attr
& IBMVETH_ILLAN_LRG_SND_SUPPORT
) &&
887 ret2
= h_illan_attributes(adapter
->vdev
->unit_address
, clr_attr
,
888 set_attr
, &ret_attr
);
890 if (ret2
!= H_SUCCESS
) {
891 netdev_err(dev
, "unable to change tso settings. %d rc=%ld\n",
894 h_illan_attributes(adapter
->vdev
->unit_address
,
895 set_attr
, clr_attr
, &ret_attr
);
898 dev
->features
&= ~(NETIF_F_TSO
| NETIF_F_TSO6
);
902 adapter
->fw_large_send_support
= data
;
903 adapter
->large_send
= data
;
906 /* Older firmware version of large send offload does not
910 dev
->features
&= ~NETIF_F_TSO6
;
911 netdev_info(dev
, "TSO feature requires all partitions to have updated driver");
913 adapter
->large_send
= data
;
917 rc2
= ibmveth_open(dev
);
919 return rc1
? rc1
: rc2
;
922 static int ibmveth_set_features(struct net_device
*dev
,
923 netdev_features_t features
)
925 struct ibmveth_adapter
*adapter
= netdev_priv(dev
);
926 int rx_csum
= !!(features
& NETIF_F_RXCSUM
);
927 int large_send
= !!(features
& (NETIF_F_TSO
| NETIF_F_TSO6
));
928 int rc1
= 0, rc2
= 0;
930 if (rx_csum
!= adapter
->rx_csum
) {
931 rc1
= ibmveth_set_csum_offload(dev
, rx_csum
);
932 if (rc1
&& !adapter
->rx_csum
)
934 features
& ~(NETIF_F_CSUM_MASK
|
938 if (large_send
!= adapter
->large_send
) {
939 rc2
= ibmveth_set_tso(dev
, large_send
);
940 if (rc2
&& !adapter
->large_send
)
942 features
& ~(NETIF_F_TSO
| NETIF_F_TSO6
);
945 return rc1
? rc1
: rc2
;
948 static void ibmveth_get_strings(struct net_device
*dev
, u32 stringset
, u8
*data
)
952 if (stringset
!= ETH_SS_STATS
)
955 for (i
= 0; i
< ARRAY_SIZE(ibmveth_stats
); i
++, data
+= ETH_GSTRING_LEN
)
956 memcpy(data
, ibmveth_stats
[i
].name
, ETH_GSTRING_LEN
);
959 static int ibmveth_get_sset_count(struct net_device
*dev
, int sset
)
963 return ARRAY_SIZE(ibmveth_stats
);
969 static void ibmveth_get_ethtool_stats(struct net_device
*dev
,
970 struct ethtool_stats
*stats
, u64
*data
)
973 struct ibmveth_adapter
*adapter
= netdev_priv(dev
);
975 for (i
= 0; i
< ARRAY_SIZE(ibmveth_stats
); i
++)
976 data
[i
] = IBMVETH_GET_STAT(adapter
, ibmveth_stats
[i
].offset
);
979 static const struct ethtool_ops netdev_ethtool_ops
= {
980 .get_drvinfo
= netdev_get_drvinfo
,
981 .get_settings
= netdev_get_settings
,
982 .get_link
= ethtool_op_get_link
,
983 .get_strings
= ibmveth_get_strings
,
984 .get_sset_count
= ibmveth_get_sset_count
,
985 .get_ethtool_stats
= ibmveth_get_ethtool_stats
,
988 static int ibmveth_ioctl(struct net_device
*dev
, struct ifreq
*ifr
, int cmd
)
993 #define page_offset(v) ((unsigned long)(v) & ((1 << 12) - 1))
995 static int ibmveth_send(struct ibmveth_adapter
*adapter
,
996 union ibmveth_buf_desc
*descs
, unsigned long mss
)
998 unsigned long correlator
;
999 unsigned int retry_count
;
1003 * The retry count sets a maximum for the number of broadcast and
1004 * multicast destinations within the system.
1009 ret
= h_send_logical_lan(adapter
->vdev
->unit_address
,
1010 descs
[0].desc
, descs
[1].desc
,
1011 descs
[2].desc
, descs
[3].desc
,
1012 descs
[4].desc
, descs
[5].desc
,
1013 correlator
, &correlator
, mss
,
1014 adapter
->fw_large_send_support
);
1015 } while ((ret
== H_BUSY
) && (retry_count
--));
1017 if (ret
!= H_SUCCESS
&& ret
!= H_DROPPED
) {
1018 netdev_err(adapter
->netdev
, "tx: h_send_logical_lan failed "
1019 "with rc=%ld\n", ret
);
1026 static netdev_tx_t
ibmveth_start_xmit(struct sk_buff
*skb
,
1027 struct net_device
*netdev
)
1029 struct ibmveth_adapter
*adapter
= netdev_priv(netdev
);
1030 unsigned int desc_flags
;
1031 union ibmveth_buf_desc descs
[6];
1033 int force_bounce
= 0;
1034 dma_addr_t dma_addr
;
1035 unsigned long mss
= 0;
1038 * veth handles a maximum of 6 segments including the header, so
1039 * we have to linearize the skb if there are more than this.
1041 if (skb_shinfo(skb
)->nr_frags
> 5 && __skb_linearize(skb
)) {
1042 netdev
->stats
.tx_dropped
++;
1046 /* veth can't checksum offload UDP */
1047 if (skb
->ip_summed
== CHECKSUM_PARTIAL
&&
1048 ((skb
->protocol
== htons(ETH_P_IP
) &&
1049 ip_hdr(skb
)->protocol
!= IPPROTO_TCP
) ||
1050 (skb
->protocol
== htons(ETH_P_IPV6
) &&
1051 ipv6_hdr(skb
)->nexthdr
!= IPPROTO_TCP
)) &&
1052 skb_checksum_help(skb
)) {
1054 netdev_err(netdev
, "tx: failed to checksum packet\n");
1055 netdev
->stats
.tx_dropped
++;
1059 desc_flags
= IBMVETH_BUF_VALID
;
1061 if (skb_is_gso(skb
) && adapter
->fw_large_send_support
)
1062 desc_flags
|= IBMVETH_BUF_LRG_SND
;
1064 if (skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1065 unsigned char *buf
= skb_transport_header(skb
) +
1068 desc_flags
|= (IBMVETH_BUF_NO_CSUM
| IBMVETH_BUF_CSUM_GOOD
);
1070 /* Need to zero out the checksum */
1076 memset(descs
, 0, sizeof(descs
));
1079 * If a linear packet is below the rx threshold then
1080 * copy it into the static bounce buffer. This avoids the
1081 * cost of a TCE insert and remove.
1083 if (force_bounce
|| (!skb_is_nonlinear(skb
) &&
1084 (skb
->len
< tx_copybreak
))) {
1085 skb_copy_from_linear_data(skb
, adapter
->bounce_buffer
,
1088 descs
[0].fields
.flags_len
= desc_flags
| skb
->len
;
1089 descs
[0].fields
.address
= adapter
->bounce_buffer_dma
;
1091 if (ibmveth_send(adapter
, descs
, 0)) {
1092 adapter
->tx_send_failed
++;
1093 netdev
->stats
.tx_dropped
++;
1095 netdev
->stats
.tx_packets
++;
1096 netdev
->stats
.tx_bytes
+= skb
->len
;
1102 /* Map the header */
1103 dma_addr
= dma_map_single(&adapter
->vdev
->dev
, skb
->data
,
1104 skb_headlen(skb
), DMA_TO_DEVICE
);
1105 if (dma_mapping_error(&adapter
->vdev
->dev
, dma_addr
))
1108 descs
[0].fields
.flags_len
= desc_flags
| skb_headlen(skb
);
1109 descs
[0].fields
.address
= dma_addr
;
1112 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1113 const skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1115 dma_addr
= skb_frag_dma_map(&adapter
->vdev
->dev
, frag
, 0,
1116 skb_frag_size(frag
), DMA_TO_DEVICE
);
1118 if (dma_mapping_error(&adapter
->vdev
->dev
, dma_addr
))
1119 goto map_failed_frags
;
1121 descs
[i
+1].fields
.flags_len
= desc_flags
| skb_frag_size(frag
);
1122 descs
[i
+1].fields
.address
= dma_addr
;
1125 if (skb_is_gso(skb
)) {
1126 if (adapter
->fw_large_send_support
) {
1127 mss
= (unsigned long)skb_shinfo(skb
)->gso_size
;
1128 adapter
->tx_large_packets
++;
1129 } else if (!skb_is_gso_v6(skb
)) {
1130 /* Put -1 in the IP checksum to tell phyp it
1131 * is a largesend packet. Put the mss in
1134 ip_hdr(skb
)->check
= 0xffff;
1135 tcp_hdr(skb
)->check
=
1136 cpu_to_be16(skb_shinfo(skb
)->gso_size
);
1137 adapter
->tx_large_packets
++;
1141 if (ibmveth_send(adapter
, descs
, mss
)) {
1142 adapter
->tx_send_failed
++;
1143 netdev
->stats
.tx_dropped
++;
1145 netdev
->stats
.tx_packets
++;
1146 netdev
->stats
.tx_bytes
+= skb
->len
;
1149 dma_unmap_single(&adapter
->vdev
->dev
,
1150 descs
[0].fields
.address
,
1151 descs
[0].fields
.flags_len
& IBMVETH_BUF_LEN_MASK
,
1154 for (i
= 1; i
< skb_shinfo(skb
)->nr_frags
+ 1; i
++)
1155 dma_unmap_page(&adapter
->vdev
->dev
, descs
[i
].fields
.address
,
1156 descs
[i
].fields
.flags_len
& IBMVETH_BUF_LEN_MASK
,
1160 dev_consume_skb_any(skb
);
1161 return NETDEV_TX_OK
;
1165 for (i
= 0; i
< last
; i
++)
1166 dma_unmap_page(&adapter
->vdev
->dev
, descs
[i
].fields
.address
,
1167 descs
[i
].fields
.flags_len
& IBMVETH_BUF_LEN_MASK
,
1171 if (!firmware_has_feature(FW_FEATURE_CMO
))
1172 netdev_err(netdev
, "tx: unable to map xmit buffer\n");
1173 adapter
->tx_map_failed
++;
1174 if (skb_linearize(skb
)) {
1175 netdev
->stats
.tx_dropped
++;
1182 static void ibmveth_rx_mss_helper(struct sk_buff
*skb
, u16 mss
, int lrg_pkt
)
1184 struct tcphdr
*tcph
;
1188 /* only TCP packets will be aggregated */
1189 if (skb
->protocol
== htons(ETH_P_IP
)) {
1190 struct iphdr
*iph
= (struct iphdr
*)skb
->data
;
1192 if (iph
->protocol
== IPPROTO_TCP
) {
1193 offset
= iph
->ihl
* 4;
1194 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV4
;
1198 } else if (skb
->protocol
== htons(ETH_P_IPV6
)) {
1199 struct ipv6hdr
*iph6
= (struct ipv6hdr
*)skb
->data
;
1201 if (iph6
->nexthdr
== IPPROTO_TCP
) {
1202 offset
= sizeof(struct ipv6hdr
);
1203 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV6
;
1210 /* if mss is not set through Large Packet bit/mss in rx buffer,
1211 * expect that the mss will be written to the tcp header checksum.
1213 tcph
= (struct tcphdr
*)(skb
->data
+ offset
);
1215 skb_shinfo(skb
)->gso_size
= mss
;
1216 } else if (offset
) {
1217 skb_shinfo(skb
)->gso_size
= ntohs(tcph
->check
);
1221 if (skb_shinfo(skb
)->gso_size
) {
1222 hdr_len
= offset
+ tcph
->doff
* 4;
1223 skb_shinfo(skb
)->gso_segs
=
1224 DIV_ROUND_UP(skb
->len
- hdr_len
,
1225 skb_shinfo(skb
)->gso_size
);
1229 static int ibmveth_poll(struct napi_struct
*napi
, int budget
)
1231 struct ibmveth_adapter
*adapter
=
1232 container_of(napi
, struct ibmveth_adapter
, napi
);
1233 struct net_device
*netdev
= adapter
->netdev
;
1234 int frames_processed
= 0;
1235 unsigned long lpar_rc
;
1240 while (frames_processed
< budget
) {
1241 if (!ibmveth_rxq_pending_buffer(adapter
))
1245 if (!ibmveth_rxq_buffer_valid(adapter
)) {
1246 wmb(); /* suggested by larson1 */
1247 adapter
->rx_invalid_buffer
++;
1248 netdev_dbg(netdev
, "recycling invalid buffer\n");
1249 ibmveth_rxq_recycle_buffer(adapter
);
1251 struct sk_buff
*skb
, *new_skb
;
1252 int length
= ibmveth_rxq_frame_length(adapter
);
1253 int offset
= ibmveth_rxq_frame_offset(adapter
);
1254 int csum_good
= ibmveth_rxq_csum_good(adapter
);
1255 int lrg_pkt
= ibmveth_rxq_large_packet(adapter
);
1257 skb
= ibmveth_rxq_get_buffer(adapter
);
1259 /* if the large packet bit is set in the rx queue
1260 * descriptor, the mss will be written by PHYP eight
1261 * bytes from the start of the rx buffer, which is
1262 * skb->data at this stage
1265 __be64
*rxmss
= (__be64
*)(skb
->data
+ 8);
1267 mss
= (u16
)be64_to_cpu(*rxmss
);
1271 if (length
< rx_copybreak
)
1272 new_skb
= netdev_alloc_skb(netdev
, length
);
1275 skb_copy_to_linear_data(new_skb
,
1279 ibmveth_flush_buffer(skb
->data
,
1281 if (!ibmveth_rxq_recycle_buffer(adapter
))
1285 ibmveth_rxq_harvest_buffer(adapter
);
1286 skb_reserve(skb
, offset
);
1289 skb_put(skb
, length
);
1290 skb
->protocol
= eth_type_trans(skb
, netdev
);
1293 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1294 if (be16_to_cpu(skb
->protocol
) == ETH_P_IP
) {
1295 iph
= (struct iphdr
*)skb
->data
;
1297 /* If the IP checksum is not offloaded and if the packet
1298 * is large send, the checksum must be rebuilt.
1300 if (iph
->check
== 0xffff) {
1302 iph
->check
= ip_fast_csum((unsigned char *)iph
, iph
->ihl
);
1307 if (length
> netdev
->mtu
+ ETH_HLEN
) {
1308 ibmveth_rx_mss_helper(skb
, mss
, lrg_pkt
);
1309 adapter
->rx_large_packets
++;
1312 napi_gro_receive(napi
, skb
); /* send it up */
1314 netdev
->stats
.rx_packets
++;
1315 netdev
->stats
.rx_bytes
+= length
;
1320 ibmveth_replenish_task(adapter
);
1322 if (frames_processed
< budget
) {
1323 napi_complete(napi
);
1325 /* We think we are done - reenable interrupts,
1326 * then check once more to make sure we are done.
1328 lpar_rc
= h_vio_signal(adapter
->vdev
->unit_address
,
1331 BUG_ON(lpar_rc
!= H_SUCCESS
);
1333 if (ibmveth_rxq_pending_buffer(adapter
) &&
1334 napi_reschedule(napi
)) {
1335 lpar_rc
= h_vio_signal(adapter
->vdev
->unit_address
,
1341 return frames_processed
;
1344 static irqreturn_t
ibmveth_interrupt(int irq
, void *dev_instance
)
1346 struct net_device
*netdev
= dev_instance
;
1347 struct ibmveth_adapter
*adapter
= netdev_priv(netdev
);
1348 unsigned long lpar_rc
;
1350 if (napi_schedule_prep(&adapter
->napi
)) {
1351 lpar_rc
= h_vio_signal(adapter
->vdev
->unit_address
,
1353 BUG_ON(lpar_rc
!= H_SUCCESS
);
1354 __napi_schedule(&adapter
->napi
);
1359 static void ibmveth_set_multicast_list(struct net_device
*netdev
)
1361 struct ibmveth_adapter
*adapter
= netdev_priv(netdev
);
1362 unsigned long lpar_rc
;
1364 if ((netdev
->flags
& IFF_PROMISC
) ||
1365 (netdev_mc_count(netdev
) > adapter
->mcastFilterSize
)) {
1366 lpar_rc
= h_multicast_ctrl(adapter
->vdev
->unit_address
,
1367 IbmVethMcastEnableRecv
|
1368 IbmVethMcastDisableFiltering
,
1370 if (lpar_rc
!= H_SUCCESS
) {
1371 netdev_err(netdev
, "h_multicast_ctrl rc=%ld when "
1372 "entering promisc mode\n", lpar_rc
);
1375 struct netdev_hw_addr
*ha
;
1376 /* clear the filter table & disable filtering */
1377 lpar_rc
= h_multicast_ctrl(adapter
->vdev
->unit_address
,
1378 IbmVethMcastEnableRecv
|
1379 IbmVethMcastDisableFiltering
|
1380 IbmVethMcastClearFilterTable
,
1382 if (lpar_rc
!= H_SUCCESS
) {
1383 netdev_err(netdev
, "h_multicast_ctrl rc=%ld when "
1384 "attempting to clear filter table\n",
1387 /* add the addresses to the filter table */
1388 netdev_for_each_mc_addr(ha
, netdev
) {
1389 /* add the multicast address to the filter table */
1391 mcast_addr
= ibmveth_encode_mac_addr(ha
->addr
);
1392 lpar_rc
= h_multicast_ctrl(adapter
->vdev
->unit_address
,
1393 IbmVethMcastAddFilter
,
1395 if (lpar_rc
!= H_SUCCESS
) {
1396 netdev_err(netdev
, "h_multicast_ctrl rc=%ld "
1397 "when adding an entry to the filter "
1398 "table\n", lpar_rc
);
1402 /* re-enable filtering */
1403 lpar_rc
= h_multicast_ctrl(adapter
->vdev
->unit_address
,
1404 IbmVethMcastEnableFiltering
,
1406 if (lpar_rc
!= H_SUCCESS
) {
1407 netdev_err(netdev
, "h_multicast_ctrl rc=%ld when "
1408 "enabling filtering\n", lpar_rc
);
1413 static int ibmveth_change_mtu(struct net_device
*dev
, int new_mtu
)
1415 struct ibmveth_adapter
*adapter
= netdev_priv(dev
);
1416 struct vio_dev
*viodev
= adapter
->vdev
;
1417 int new_mtu_oh
= new_mtu
+ IBMVETH_BUFF_OH
;
1419 int need_restart
= 0;
1421 if (new_mtu
< IBMVETH_MIN_MTU
)
1424 for (i
= 0; i
< IBMVETH_NUM_BUFF_POOLS
; i
++)
1425 if (new_mtu_oh
<= adapter
->rx_buff_pool
[i
].buff_size
)
1428 if (i
== IBMVETH_NUM_BUFF_POOLS
)
1431 /* Deactivate all the buffer pools so that the next loop can activate
1432 only the buffer pools necessary to hold the new MTU */
1433 if (netif_running(adapter
->netdev
)) {
1435 adapter
->pool_config
= 1;
1436 ibmveth_close(adapter
->netdev
);
1437 adapter
->pool_config
= 0;
1440 /* Look for an active buffer pool that can hold the new MTU */
1441 for (i
= 0; i
< IBMVETH_NUM_BUFF_POOLS
; i
++) {
1442 adapter
->rx_buff_pool
[i
].active
= 1;
1444 if (new_mtu_oh
<= adapter
->rx_buff_pool
[i
].buff_size
) {
1446 vio_cmo_set_dev_desired(viodev
,
1447 ibmveth_get_desired_dma
1450 return ibmveth_open(adapter
->netdev
);
1456 if (need_restart
&& (rc
= ibmveth_open(adapter
->netdev
)))
1462 #ifdef CONFIG_NET_POLL_CONTROLLER
1463 static void ibmveth_poll_controller(struct net_device
*dev
)
1465 ibmveth_replenish_task(netdev_priv(dev
));
1466 ibmveth_interrupt(dev
->irq
, dev
);
1471 * ibmveth_get_desired_dma - Calculate IO memory desired by the driver
1473 * @vdev: struct vio_dev for the device whose desired IO mem is to be returned
1476 * Number of bytes of IO data the driver will need to perform well.
1478 static unsigned long ibmveth_get_desired_dma(struct vio_dev
*vdev
)
1480 struct net_device
*netdev
= dev_get_drvdata(&vdev
->dev
);
1481 struct ibmveth_adapter
*adapter
;
1482 struct iommu_table
*tbl
;
1487 tbl
= get_iommu_table_base(&vdev
->dev
);
1489 /* netdev inits at probe time along with the structures we need below*/
1491 return IOMMU_PAGE_ALIGN(IBMVETH_IO_ENTITLEMENT_DEFAULT
, tbl
);
1493 adapter
= netdev_priv(netdev
);
1495 ret
= IBMVETH_BUFF_LIST_SIZE
+ IBMVETH_FILT_LIST_SIZE
;
1496 ret
+= IOMMU_PAGE_ALIGN(netdev
->mtu
, tbl
);
1498 for (i
= 0; i
< IBMVETH_NUM_BUFF_POOLS
; i
++) {
1499 /* add the size of the active receive buffers */
1500 if (adapter
->rx_buff_pool
[i
].active
)
1502 adapter
->rx_buff_pool
[i
].size
*
1503 IOMMU_PAGE_ALIGN(adapter
->rx_buff_pool
[i
].
1505 rxqentries
+= adapter
->rx_buff_pool
[i
].size
;
1507 /* add the size of the receive queue entries */
1508 ret
+= IOMMU_PAGE_ALIGN(
1509 rxqentries
* sizeof(struct ibmveth_rx_q_entry
), tbl
);
1514 static int ibmveth_set_mac_addr(struct net_device
*dev
, void *p
)
1516 struct ibmveth_adapter
*adapter
= netdev_priv(dev
);
1517 struct sockaddr
*addr
= p
;
1521 if (!is_valid_ether_addr(addr
->sa_data
))
1522 return -EADDRNOTAVAIL
;
1524 mac_address
= ibmveth_encode_mac_addr(addr
->sa_data
);
1525 rc
= h_change_logical_lan_mac(adapter
->vdev
->unit_address
, mac_address
);
1527 netdev_err(adapter
->netdev
, "h_change_logical_lan_mac failed with rc=%d\n", rc
);
1531 ether_addr_copy(dev
->dev_addr
, addr
->sa_data
);
1536 static const struct net_device_ops ibmveth_netdev_ops
= {
1537 .ndo_open
= ibmveth_open
,
1538 .ndo_stop
= ibmveth_close
,
1539 .ndo_start_xmit
= ibmveth_start_xmit
,
1540 .ndo_set_rx_mode
= ibmveth_set_multicast_list
,
1541 .ndo_do_ioctl
= ibmveth_ioctl
,
1542 .ndo_change_mtu
= ibmveth_change_mtu
,
1543 .ndo_fix_features
= ibmveth_fix_features
,
1544 .ndo_set_features
= ibmveth_set_features
,
1545 .ndo_validate_addr
= eth_validate_addr
,
1546 .ndo_set_mac_address
= ibmveth_set_mac_addr
,
1547 #ifdef CONFIG_NET_POLL_CONTROLLER
1548 .ndo_poll_controller
= ibmveth_poll_controller
,
1552 static int ibmveth_probe(struct vio_dev
*dev
, const struct vio_device_id
*id
)
1555 struct net_device
*netdev
;
1556 struct ibmveth_adapter
*adapter
;
1557 unsigned char *mac_addr_p
;
1558 unsigned int *mcastFilterSize_p
;
1560 unsigned long ret_attr
;
1562 dev_dbg(&dev
->dev
, "entering ibmveth_probe for UA 0x%x\n",
1565 mac_addr_p
= (unsigned char *)vio_get_attribute(dev
, VETH_MAC_ADDR
,
1568 dev_err(&dev
->dev
, "Can't find VETH_MAC_ADDR attribute\n");
1571 /* Workaround for old/broken pHyp */
1574 else if (mac_len
!= 6) {
1575 dev_err(&dev
->dev
, "VETH_MAC_ADDR attribute wrong len %d\n",
1580 mcastFilterSize_p
= (unsigned int *)vio_get_attribute(dev
,
1581 VETH_MCAST_FILTER_SIZE
, NULL
);
1582 if (!mcastFilterSize_p
) {
1583 dev_err(&dev
->dev
, "Can't find VETH_MCAST_FILTER_SIZE "
1588 netdev
= alloc_etherdev(sizeof(struct ibmveth_adapter
));
1593 adapter
= netdev_priv(netdev
);
1594 dev_set_drvdata(&dev
->dev
, netdev
);
1596 adapter
->vdev
= dev
;
1597 adapter
->netdev
= netdev
;
1598 adapter
->mcastFilterSize
= *mcastFilterSize_p
;
1599 adapter
->pool_config
= 0;
1601 netif_napi_add(netdev
, &adapter
->napi
, ibmveth_poll
, 16);
1603 netdev
->irq
= dev
->irq
;
1604 netdev
->netdev_ops
= &ibmveth_netdev_ops
;
1605 netdev
->ethtool_ops
= &netdev_ethtool_ops
;
1606 SET_NETDEV_DEV(netdev
, &dev
->dev
);
1607 netdev
->hw_features
= NETIF_F_SG
| NETIF_F_RXCSUM
|
1608 NETIF_F_IP_CSUM
| NETIF_F_IPV6_CSUM
;
1610 netdev
->features
|= netdev
->hw_features
;
1612 ret
= h_illan_attributes(adapter
->vdev
->unit_address
, 0, 0, &ret_attr
);
1614 /* If running older firmware, TSO should not be enabled by default */
1615 if (ret
== H_SUCCESS
&& (ret_attr
& IBMVETH_ILLAN_LRG_SND_SUPPORT
) &&
1617 netdev
->hw_features
|= NETIF_F_TSO
| NETIF_F_TSO6
;
1618 netdev
->features
|= netdev
->hw_features
;
1620 netdev
->hw_features
|= NETIF_F_TSO
;
1623 memcpy(netdev
->dev_addr
, mac_addr_p
, ETH_ALEN
);
1625 if (firmware_has_feature(FW_FEATURE_CMO
))
1626 memcpy(pool_count
, pool_count_cmo
, sizeof(pool_count
));
1628 for (i
= 0; i
< IBMVETH_NUM_BUFF_POOLS
; i
++) {
1629 struct kobject
*kobj
= &adapter
->rx_buff_pool
[i
].kobj
;
1632 ibmveth_init_buffer_pool(&adapter
->rx_buff_pool
[i
], i
,
1633 pool_count
[i
], pool_size
[i
],
1635 error
= kobject_init_and_add(kobj
, &ktype_veth_pool
,
1636 &dev
->dev
.kobj
, "pool%d", i
);
1638 kobject_uevent(kobj
, KOBJ_ADD
);
1641 netdev_dbg(netdev
, "adapter @ 0x%p\n", adapter
);
1643 adapter
->buffer_list_dma
= DMA_ERROR_CODE
;
1644 adapter
->filter_list_dma
= DMA_ERROR_CODE
;
1645 adapter
->rx_queue
.queue_dma
= DMA_ERROR_CODE
;
1647 netdev_dbg(netdev
, "registering netdev...\n");
1649 ibmveth_set_features(netdev
, netdev
->features
);
1651 rc
= register_netdev(netdev
);
1654 netdev_dbg(netdev
, "failed to register netdev rc=%d\n", rc
);
1655 free_netdev(netdev
);
1659 netdev_dbg(netdev
, "registered\n");
1664 static int ibmveth_remove(struct vio_dev
*dev
)
1666 struct net_device
*netdev
= dev_get_drvdata(&dev
->dev
);
1667 struct ibmveth_adapter
*adapter
= netdev_priv(netdev
);
1670 for (i
= 0; i
< IBMVETH_NUM_BUFF_POOLS
; i
++)
1671 kobject_put(&adapter
->rx_buff_pool
[i
].kobj
);
1673 unregister_netdev(netdev
);
1675 free_netdev(netdev
);
1676 dev_set_drvdata(&dev
->dev
, NULL
);
1681 static struct attribute veth_active_attr
;
1682 static struct attribute veth_num_attr
;
1683 static struct attribute veth_size_attr
;
1685 static ssize_t
veth_pool_show(struct kobject
*kobj
,
1686 struct attribute
*attr
, char *buf
)
1688 struct ibmveth_buff_pool
*pool
= container_of(kobj
,
1689 struct ibmveth_buff_pool
,
1692 if (attr
== &veth_active_attr
)
1693 return sprintf(buf
, "%d\n", pool
->active
);
1694 else if (attr
== &veth_num_attr
)
1695 return sprintf(buf
, "%d\n", pool
->size
);
1696 else if (attr
== &veth_size_attr
)
1697 return sprintf(buf
, "%d\n", pool
->buff_size
);
1701 static ssize_t
veth_pool_store(struct kobject
*kobj
, struct attribute
*attr
,
1702 const char *buf
, size_t count
)
1704 struct ibmveth_buff_pool
*pool
= container_of(kobj
,
1705 struct ibmveth_buff_pool
,
1707 struct net_device
*netdev
= dev_get_drvdata(
1708 container_of(kobj
->parent
, struct device
, kobj
));
1709 struct ibmveth_adapter
*adapter
= netdev_priv(netdev
);
1710 long value
= simple_strtol(buf
, NULL
, 10);
1713 if (attr
== &veth_active_attr
) {
1714 if (value
&& !pool
->active
) {
1715 if (netif_running(netdev
)) {
1716 if (ibmveth_alloc_buffer_pool(pool
)) {
1718 "unable to alloc pool\n");
1722 adapter
->pool_config
= 1;
1723 ibmveth_close(netdev
);
1724 adapter
->pool_config
= 0;
1725 if ((rc
= ibmveth_open(netdev
)))
1730 } else if (!value
&& pool
->active
) {
1731 int mtu
= netdev
->mtu
+ IBMVETH_BUFF_OH
;
1733 /* Make sure there is a buffer pool with buffers that
1734 can hold a packet of the size of the MTU */
1735 for (i
= 0; i
< IBMVETH_NUM_BUFF_POOLS
; i
++) {
1736 if (pool
== &adapter
->rx_buff_pool
[i
])
1738 if (!adapter
->rx_buff_pool
[i
].active
)
1740 if (mtu
<= adapter
->rx_buff_pool
[i
].buff_size
)
1744 if (i
== IBMVETH_NUM_BUFF_POOLS
) {
1745 netdev_err(netdev
, "no active pool >= MTU\n");
1749 if (netif_running(netdev
)) {
1750 adapter
->pool_config
= 1;
1751 ibmveth_close(netdev
);
1753 adapter
->pool_config
= 0;
1754 if ((rc
= ibmveth_open(netdev
)))
1759 } else if (attr
== &veth_num_attr
) {
1760 if (value
<= 0 || value
> IBMVETH_MAX_POOL_COUNT
) {
1763 if (netif_running(netdev
)) {
1764 adapter
->pool_config
= 1;
1765 ibmveth_close(netdev
);
1766 adapter
->pool_config
= 0;
1768 if ((rc
= ibmveth_open(netdev
)))
1774 } else if (attr
== &veth_size_attr
) {
1775 if (value
<= IBMVETH_BUFF_OH
|| value
> IBMVETH_MAX_BUF_SIZE
) {
1778 if (netif_running(netdev
)) {
1779 adapter
->pool_config
= 1;
1780 ibmveth_close(netdev
);
1781 adapter
->pool_config
= 0;
1782 pool
->buff_size
= value
;
1783 if ((rc
= ibmveth_open(netdev
)))
1786 pool
->buff_size
= value
;
1791 /* kick the interrupt handler to allocate/deallocate pools */
1792 ibmveth_interrupt(netdev
->irq
, netdev
);
1797 #define ATTR(_name, _mode) \
1798 struct attribute veth_##_name##_attr = { \
1799 .name = __stringify(_name), .mode = _mode, \
1802 static ATTR(active
, 0644);
1803 static ATTR(num
, 0644);
1804 static ATTR(size
, 0644);
1806 static struct attribute
*veth_pool_attrs
[] = {
1813 static const struct sysfs_ops veth_pool_ops
= {
1814 .show
= veth_pool_show
,
1815 .store
= veth_pool_store
,
1818 static struct kobj_type ktype_veth_pool
= {
1820 .sysfs_ops
= &veth_pool_ops
,
1821 .default_attrs
= veth_pool_attrs
,
1824 static int ibmveth_resume(struct device
*dev
)
1826 struct net_device
*netdev
= dev_get_drvdata(dev
);
1827 ibmveth_interrupt(netdev
->irq
, netdev
);
1831 static struct vio_device_id ibmveth_device_table
[] = {
1832 { "network", "IBM,l-lan"},
1835 MODULE_DEVICE_TABLE(vio
, ibmveth_device_table
);
1837 static struct dev_pm_ops ibmveth_pm_ops
= {
1838 .resume
= ibmveth_resume
1841 static struct vio_driver ibmveth_driver
= {
1842 .id_table
= ibmveth_device_table
,
1843 .probe
= ibmveth_probe
,
1844 .remove
= ibmveth_remove
,
1845 .get_desired_dma
= ibmveth_get_desired_dma
,
1846 .name
= ibmveth_driver_name
,
1847 .pm
= &ibmveth_pm_ops
,
1850 static int __init
ibmveth_module_init(void)
1852 printk(KERN_DEBUG
"%s: %s %s\n", ibmveth_driver_name
,
1853 ibmveth_driver_string
, ibmveth_driver_version
);
1855 return vio_register_driver(&ibmveth_driver
);
1858 static void __exit
ibmveth_module_exit(void)
1860 vio_unregister_driver(&ibmveth_driver
);
1863 module_init(ibmveth_module_init
);
1864 module_exit(ibmveth_module_exit
);