2 * Copyright (c) 2007 Mellanox Technologies. All rights reserved.
4 * This software is available to you under a choice of one of two
5 * licenses. You may choose to be licensed under the terms of the GNU
6 * General Public License (GPL) Version 2, available from the file
7 * COPYING in the main directory of this source tree, or the
8 * OpenIB.org BSD license below:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
18 * - Redistributions in binary form must reproduce the above
19 * copyright notice, this list of conditions and the following
20 * disclaimer in the documentation and/or other materials
21 * provided with the distribution.
23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
37 #include <linux/compiler.h>
38 #include <linux/list.h>
39 #include <linux/mutex.h>
40 #include <linux/netdevice.h>
41 #include <linux/inet_lro.h>
43 #include <linux/mlx4/device.h>
44 #include <linux/mlx4/qp.h>
45 #include <linux/mlx4/cq.h>
46 #include <linux/mlx4/srq.h>
47 #include <linux/mlx4/doorbell.h>
51 #define DRV_NAME "mlx4_en"
52 #define DRV_VERSION "1.4.1.1"
53 #define DRV_RELDATE "June 2009"
56 #define MLX4_EN_MSG_LEVEL (NETIF_MSG_LINK | NETIF_MSG_IFDOWN)
58 #define en_print(level, priv, format, arg...) \
60 if ((priv)->registered) \
61 printk(level "%s: %s: " format, DRV_NAME, \
62 (priv->dev)->name, ## arg); \
64 printk(level "%s: %s: Port %d: " format, \
65 DRV_NAME, dev_name(&priv->mdev->pdev->dev), \
66 (priv)->port, ## arg); \
69 #define en_dbg(mlevel, priv, format, arg...) \
71 if (NETIF_MSG_##mlevel & priv->msg_enable) \
72 en_print(KERN_DEBUG, priv, format, ## arg) \
74 #define en_warn(priv, format, arg...) \
75 en_print(KERN_WARNING, priv, format, ## arg)
76 #define en_err(priv, format, arg...) \
77 en_print(KERN_ERR, priv, format, ## arg)
79 #define mlx4_err(mdev, format, arg...) \
80 printk(KERN_ERR "%s %s: " format , DRV_NAME ,\
81 dev_name(&mdev->pdev->dev) , ## arg)
82 #define mlx4_info(mdev, format, arg...) \
83 printk(KERN_INFO "%s %s: " format , DRV_NAME ,\
84 dev_name(&mdev->pdev->dev) , ## arg)
85 #define mlx4_warn(mdev, format, arg...) \
86 printk(KERN_WARNING "%s %s: " format , DRV_NAME ,\
87 dev_name(&mdev->pdev->dev) , ## arg)
94 #define MLX4_EN_PAGE_SHIFT 12
95 #define MLX4_EN_PAGE_SIZE (1 << MLX4_EN_PAGE_SHIFT)
96 #define MAX_TX_RINGS 16
97 #define MAX_RX_RINGS 16
98 #define MAX_RSS_MAP_SIZE 64
101 #define HEADROOM (2048 / TXBB_SIZE + 1)
102 #define STAMP_STRIDE 64
103 #define STAMP_DWORDS (STAMP_STRIDE / 4)
104 #define STAMP_SHIFT 31
105 #define STAMP_VAL 0x7fffffff
106 #define STATS_DELAY (HZ / 4)
108 /* Typical TSO descriptor with 16 gather entries is 352 bytes... */
109 #define MAX_DESC_SIZE 512
110 #define MAX_DESC_TXBBS (MAX_DESC_SIZE / TXBB_SIZE)
113 * OS related constants and tunables
116 #define MLX4_EN_WATCHDOG_TIMEOUT (15 * HZ)
118 #define MLX4_EN_ALLOC_ORDER 2
119 #define MLX4_EN_ALLOC_SIZE (PAGE_SIZE << MLX4_EN_ALLOC_ORDER)
121 #define MLX4_EN_MAX_LRO_DESCRIPTORS 32
123 /* Receive fragment sizes; we use at most 4 fragments (for 9600 byte MTU
124 * and 4K allocations) */
126 FRAG_SZ0
= 512 - NET_IP_ALIGN
,
129 FRAG_SZ3
= MLX4_EN_ALLOC_SIZE
131 #define MLX4_EN_MAX_RX_FRAGS 4
133 /* Maximum ring sizes */
134 #define MLX4_EN_MAX_TX_SIZE 8192
135 #define MLX4_EN_MAX_RX_SIZE 8192
137 /* Minimum ring size for our page-allocation sceme to work */
138 #define MLX4_EN_MIN_RX_SIZE (MLX4_EN_ALLOC_SIZE / SMP_CACHE_BYTES)
139 #define MLX4_EN_MIN_TX_SIZE (4096 / TXBB_SIZE)
141 #define MLX4_EN_SMALL_PKT_SIZE 64
142 #define MLX4_EN_NUM_TX_RINGS 8
143 #define MLX4_EN_NUM_PPP_RINGS 8
144 #define MLX4_EN_DEF_TX_RING_SIZE 512
145 #define MLX4_EN_DEF_RX_RING_SIZE 1024
147 /* Target number of packets to coalesce with interrupt moderation */
148 #define MLX4_EN_RX_COAL_TARGET 44
149 #define MLX4_EN_RX_COAL_TIME 0x10
151 #define MLX4_EN_TX_COAL_PKTS 5
152 #define MLX4_EN_TX_COAL_TIME 0x80
154 #define MLX4_EN_RX_RATE_LOW 400000
155 #define MLX4_EN_RX_COAL_TIME_LOW 0
156 #define MLX4_EN_RX_RATE_HIGH 450000
157 #define MLX4_EN_RX_COAL_TIME_HIGH 128
158 #define MLX4_EN_RX_SIZE_THRESH 1024
159 #define MLX4_EN_RX_RATE_THRESH (1000000 / MLX4_EN_RX_COAL_TIME_HIGH)
160 #define MLX4_EN_SAMPLE_INTERVAL 0
162 #define MLX4_EN_AUTO_CONF 0xffff
164 #define MLX4_EN_DEF_RX_PAUSE 1
165 #define MLX4_EN_DEF_TX_PAUSE 1
167 /* Interval between sucessive polls in the Tx routine when polling is used
168 instead of interrupts (in per-core Tx rings) - should be power of 2 */
169 #define MLX4_EN_TX_POLL_MODER 16
170 #define MLX4_EN_TX_POLL_TIMEOUT (HZ / 4)
172 #define ETH_LLC_SNAP_SIZE 8
174 #define SMALL_PACKET_SIZE (256 - NET_IP_ALIGN)
175 #define HEADER_COPY_SIZE (128 - NET_IP_ALIGN)
177 #define MLX4_EN_MIN_MTU 46
178 #define ETH_BCAST 0xffffffffffffULL
180 #ifdef MLX4_EN_PERF_STAT
181 /* Number of samples to 'average' */
183 #define AVG_FACTOR 1024
184 #define NUM_PERF_STATS NUM_PERF_COUNTERS
186 #define INC_PERF_COUNTER(cnt) (++(cnt))
187 #define ADD_PERF_COUNTER(cnt, add) ((cnt) += (add))
188 #define AVG_PERF_COUNTER(cnt, sample) \
189 ((cnt) = ((cnt) * (AVG_SIZE - 1) + (sample) * AVG_FACTOR) / AVG_SIZE)
190 #define GET_PERF_COUNTER(cnt) (cnt)
191 #define GET_AVG_PERF_COUNTER(cnt) ((cnt) / AVG_FACTOR)
195 #define NUM_PERF_STATS 0
196 #define INC_PERF_COUNTER(cnt) do {} while (0)
197 #define ADD_PERF_COUNTER(cnt, add) do {} while (0)
198 #define AVG_PERF_COUNTER(cnt, sample) do {} while (0)
199 #define GET_PERF_COUNTER(cnt) (0)
200 #define GET_AVG_PERF_COUNTER(cnt) (0)
201 #endif /* MLX4_EN_PERF_STAT */
216 #define ROUNDUP_LOG2(x) ilog2(roundup_pow_of_two(x))
217 #define XNOR(x, y) (!(x) == !(y))
218 #define ILLEGAL_MAC(addr) (addr == 0xffffffffffffULL || addr == 0x0)
221 struct mlx4_en_tx_info
{
230 #define MLX4_EN_BIT_DESC_OWN 0x80000000
231 #define CTRL_SIZE sizeof(struct mlx4_wqe_ctrl_seg)
232 #define MLX4_EN_MEMTYPE_PAD 0x100
233 #define DS_SIZE sizeof(struct mlx4_wqe_data_seg)
236 struct mlx4_en_tx_desc
{
237 struct mlx4_wqe_ctrl_seg ctrl
;
239 struct mlx4_wqe_data_seg data
; /* at least one data segment */
240 struct mlx4_wqe_lso_seg lso
;
241 struct mlx4_wqe_inline_seg inl
;
245 #define MLX4_EN_USE_SRQ 0x01000000
247 struct mlx4_en_rx_alloc
{
252 struct mlx4_en_tx_ring
{
253 struct mlx4_hwq_resources wqres
;
254 u32 size
; /* number of TXBBs */
257 u16 cqn
; /* index of port CQ associated with this ring */
265 struct mlx4_en_tx_info
*tx_info
;
269 struct mlx4_qp_context context
;
271 enum mlx4_qp_state qp_state
;
272 struct mlx4_srq dummy
;
274 unsigned long packets
;
275 spinlock_t comp_lock
;
278 struct mlx4_en_rx_desc
{
279 struct mlx4_wqe_srq_next_seg next
;
280 /* actual number of entries depends on rx ring stride */
281 struct mlx4_wqe_data_seg data
[0];
284 struct mlx4_en_rx_ring
{
286 struct mlx4_hwq_resources wqres
;
287 struct mlx4_en_rx_alloc page_alloc
[MLX4_EN_MAX_RX_FRAGS
];
288 struct net_lro_mgr lro
;
289 u32 size
; /* number of Rx descs*/
294 u16 cqn
; /* index of port CQ associated with this ring */
301 unsigned long packets
;
305 static inline int mlx4_en_can_lro(__be16 status
)
307 return (status
& cpu_to_be16(MLX4_CQE_STATUS_IPV4
|
308 MLX4_CQE_STATUS_IPV4F
|
309 MLX4_CQE_STATUS_IPV6
|
310 MLX4_CQE_STATUS_IPV4OPT
|
311 MLX4_CQE_STATUS_TCP
|
312 MLX4_CQE_STATUS_UDP
|
313 MLX4_CQE_STATUS_IPOK
)) ==
314 cpu_to_be16(MLX4_CQE_STATUS_IPV4
|
315 MLX4_CQE_STATUS_IPOK
|
316 MLX4_CQE_STATUS_TCP
);
321 struct mlx4_hwq_resources wqres
;
324 struct net_device
*dev
;
325 struct napi_struct napi
;
326 /* Per-core Tx cq processing support */
327 struct timer_list timer
;
334 struct mlx4_cqe
*buf
;
335 #define MLX4_EN_OPCODE_ERROR 0x1e
338 struct mlx4_en_port_profile
{
350 struct mlx4_en_profile
{
357 struct mlx4_en_port_profile prof
[MLX4_MAX_PORTS
+ 1];
361 struct mlx4_dev
*dev
;
362 struct pci_dev
*pdev
;
363 struct mutex state_lock
;
364 struct net_device
*pndev
[MLX4_MAX_PORTS
+ 1];
367 struct mlx4_en_profile profile
;
369 struct workqueue_struct
*workqueue
;
370 struct device
*dma_device
;
371 void __iomem
*uar_map
;
372 struct mlx4_uar priv_uar
;
379 struct mlx4_en_rss_map
{
382 u16 map
[MAX_RSS_MAP_SIZE
];
383 struct mlx4_qp qps
[MAX_RSS_MAP_SIZE
];
384 enum mlx4_qp_state state
[MAX_RSS_MAP_SIZE
];
385 struct mlx4_qp indir_qp
;
386 enum mlx4_qp_state indir_state
;
389 struct mlx4_en_rss_context
{
398 struct mlx4_en_pkt_stats
{
399 unsigned long broadcast
;
400 unsigned long rx_prio
[8];
401 unsigned long tx_prio
[8];
402 #define NUM_PKT_STATS 17
405 struct mlx4_en_port_stats
{
406 unsigned long lro_aggregated
;
407 unsigned long lro_flushed
;
408 unsigned long lro_no_desc
;
409 unsigned long tso_packets
;
410 unsigned long queue_stopped
;
411 unsigned long wake_queue
;
412 unsigned long tx_timeout
;
413 unsigned long rx_alloc_failed
;
414 unsigned long rx_chksum_good
;
415 unsigned long rx_chksum_none
;
416 unsigned long tx_chksum_offload
;
417 #define NUM_PORT_STATS 11
420 struct mlx4_en_perf_stats
{
427 #define NUM_PERF_COUNTERS 6
430 struct mlx4_en_frag_info
{
432 u16 frag_prefix_size
;
439 struct mlx4_en_priv
{
440 struct mlx4_en_dev
*mdev
;
441 struct mlx4_en_port_profile
*prof
;
442 struct net_device
*dev
;
443 struct vlan_group
*vlgrp
;
444 struct net_device_stats stats
;
445 struct net_device_stats ret_stats
;
446 spinlock_t stats_lock
;
448 unsigned long last_moder_packets
;
449 unsigned long last_moder_tx_packets
;
450 unsigned long last_moder_bytes
;
451 unsigned long last_moder_jiffies
;
462 u16 adaptive_rx_coal
;
465 struct mlx4_hwq_resources res
;
479 struct mlx4_en_rss_map rss_map
;
481 #define MLX4_EN_FLAG_PROMISC 0x1
485 struct mlx4_en_frag_info frag_info
[MLX4_EN_MAX_RX_FRAGS
];
489 struct mlx4_en_tx_ring tx_ring
[MAX_TX_RINGS
];
490 struct mlx4_en_rx_ring rx_ring
[MAX_RX_RINGS
];
491 struct mlx4_en_cq tx_cq
[MAX_TX_RINGS
];
492 struct mlx4_en_cq rx_cq
[MAX_RX_RINGS
];
493 struct work_struct mcast_task
;
494 struct work_struct mac_task
;
495 struct work_struct watchdog_task
;
496 struct work_struct linkstate_task
;
497 struct delayed_work stats_task
;
498 struct mlx4_en_perf_stats pstats
;
499 struct mlx4_en_pkt_stats pkstats
;
500 struct mlx4_en_port_stats port_stats
;
501 struct dev_mc_list
*mc_list
;
502 struct mlx4_en_stat_out_mbox hw_stats
;
506 void mlx4_en_destroy_netdev(struct net_device
*dev
);
507 int mlx4_en_init_netdev(struct mlx4_en_dev
*mdev
, int port
,
508 struct mlx4_en_port_profile
*prof
);
510 int mlx4_en_start_port(struct net_device
*dev
);
511 void mlx4_en_stop_port(struct net_device
*dev
);
513 void mlx4_en_free_resources(struct mlx4_en_priv
*priv
);
514 int mlx4_en_alloc_resources(struct mlx4_en_priv
*priv
);
516 int mlx4_en_create_cq(struct mlx4_en_priv
*priv
, struct mlx4_en_cq
*cq
,
517 int entries
, int ring
, enum cq_type mode
);
518 void mlx4_en_destroy_cq(struct mlx4_en_priv
*priv
, struct mlx4_en_cq
*cq
);
519 int mlx4_en_activate_cq(struct mlx4_en_priv
*priv
, struct mlx4_en_cq
*cq
);
520 void mlx4_en_deactivate_cq(struct mlx4_en_priv
*priv
, struct mlx4_en_cq
*cq
);
521 int mlx4_en_set_cq_moder(struct mlx4_en_priv
*priv
, struct mlx4_en_cq
*cq
);
522 int mlx4_en_arm_cq(struct mlx4_en_priv
*priv
, struct mlx4_en_cq
*cq
);
524 void mlx4_en_poll_tx_cq(unsigned long data
);
525 void mlx4_en_tx_irq(struct mlx4_cq
*mcq
);
526 u16
mlx4_en_select_queue(struct net_device
*dev
, struct sk_buff
*skb
);
527 int mlx4_en_xmit(struct sk_buff
*skb
, struct net_device
*dev
);
529 int mlx4_en_create_tx_ring(struct mlx4_en_priv
*priv
, struct mlx4_en_tx_ring
*ring
,
530 u32 size
, u16 stride
);
531 void mlx4_en_destroy_tx_ring(struct mlx4_en_priv
*priv
, struct mlx4_en_tx_ring
*ring
);
532 int mlx4_en_activate_tx_ring(struct mlx4_en_priv
*priv
,
533 struct mlx4_en_tx_ring
*ring
,
535 void mlx4_en_deactivate_tx_ring(struct mlx4_en_priv
*priv
,
536 struct mlx4_en_tx_ring
*ring
);
538 int mlx4_en_create_rx_ring(struct mlx4_en_priv
*priv
,
539 struct mlx4_en_rx_ring
*ring
,
540 u32 size
, u16 stride
);
541 void mlx4_en_destroy_rx_ring(struct mlx4_en_priv
*priv
,
542 struct mlx4_en_rx_ring
*ring
);
543 int mlx4_en_activate_rx_rings(struct mlx4_en_priv
*priv
);
544 void mlx4_en_deactivate_rx_ring(struct mlx4_en_priv
*priv
,
545 struct mlx4_en_rx_ring
*ring
);
546 int mlx4_en_process_rx_cq(struct net_device
*dev
,
547 struct mlx4_en_cq
*cq
,
549 int mlx4_en_poll_rx_cq(struct napi_struct
*napi
, int budget
);
550 void mlx4_en_fill_qp_context(struct mlx4_en_priv
*priv
, int size
, int stride
,
551 int is_tx
, int rss
, int qpn
, int cqn
, int srqn
,
552 struct mlx4_qp_context
*context
);
553 void mlx4_en_sqp_event(struct mlx4_qp
*qp
, enum mlx4_event event
);
554 int mlx4_en_map_buffer(struct mlx4_buf
*buf
);
555 void mlx4_en_unmap_buffer(struct mlx4_buf
*buf
);
557 void mlx4_en_calc_rx_buf(struct net_device
*dev
);
558 void mlx4_en_set_default_rss_map(struct mlx4_en_priv
*priv
,
559 struct mlx4_en_rss_map
*rss_map
,
560 int num_entries
, int num_rings
);
561 int mlx4_en_config_rss_steer(struct mlx4_en_priv
*priv
);
562 void mlx4_en_release_rss_steer(struct mlx4_en_priv
*priv
);
563 int mlx4_en_free_tx_buf(struct net_device
*dev
, struct mlx4_en_tx_ring
*ring
);
564 void mlx4_en_rx_irq(struct mlx4_cq
*mcq
);
566 int mlx4_SET_MCAST_FLTR(struct mlx4_dev
*dev
, u8 port
, u64 mac
, u64 clear
, u8 mode
);
567 int mlx4_SET_VLAN_FLTR(struct mlx4_dev
*dev
, u8 port
, struct vlan_group
*grp
);
568 int mlx4_SET_PORT_general(struct mlx4_dev
*dev
, u8 port
, int mtu
,
569 u8 pptx
, u8 pfctx
, u8 pprx
, u8 pfcrx
);
570 int mlx4_SET_PORT_qpn_calc(struct mlx4_dev
*dev
, u8 port
, u32 base_qpn
,
573 int mlx4_en_DUMP_ETH_STATS(struct mlx4_en_dev
*mdev
, u8 port
, u8 reset
);
578 extern const struct ethtool_ops mlx4_en_ethtool_ops
;