1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32#include <linux/module.h>
33#include <linux/moduleparam.h>
34#include <linux/debugfs.h>
35#include <linux/vmalloc.h>
36#include <linux/math64.h>
37
38#include <rdma/ib_verbs.h>
39
40#include "iw_cxgb4.h"
41
42#define DRV_VERSION "0.1"
43
44MODULE_AUTHOR("Steve Wise");
45MODULE_DESCRIPTION("Chelsio T4/T5 RDMA Driver");
46MODULE_LICENSE("Dual BSD/GPL");
47
48static int allow_db_fc_on_t5;
49module_param(allow_db_fc_on_t5, int, 0644);
50MODULE_PARM_DESC(allow_db_fc_on_t5,
51 "Allow DB Flow Control on T5 (default = 0)");
52
53static int allow_db_coalescing_on_t5;
54module_param(allow_db_coalescing_on_t5, int, 0644);
55MODULE_PARM_DESC(allow_db_coalescing_on_t5,
56 "Allow DB Coalescing on T5 (default = 0)");
57
58int c4iw_wr_log = 0;
59module_param(c4iw_wr_log, int, 0444);
60MODULE_PARM_DESC(c4iw_wr_log, "Enables logging of work request timing data.");
61
62static int c4iw_wr_log_size_order = 12;
63module_param(c4iw_wr_log_size_order, int, 0444);
64MODULE_PARM_DESC(c4iw_wr_log_size_order,
65 "Number of entries (log2) in the work request timing log.");
66
67static LIST_HEAD(uld_ctx_list);
68static DEFINE_MUTEX(dev_mutex);
69static struct workqueue_struct *reg_workq;
70
71#define DB_FC_RESUME_SIZE 64
72#define DB_FC_RESUME_DELAY 1
73#define DB_FC_DRAIN_THRESH 0
74
75static struct dentry *c4iw_debugfs_root;
76
77struct c4iw_debugfs_data {
78 struct c4iw_dev *devp;
79 char *buf;
80 int bufsize;
81 int pos;
82};
83
84static int count_idrs(int id, void *p, void *data)
85{
86 int *countp = data;
87
88 *countp = *countp + 1;
89 return 0;
90}
91
92static ssize_t debugfs_read(struct file *file, char __user *buf, size_t count,
93 loff_t *ppos)
94{
95 struct c4iw_debugfs_data *d = file->private_data;
96
97 return simple_read_from_buffer(buf, count, ppos, d->buf, d->pos);
98}
99
100void c4iw_log_wr_stats(struct t4_wq *wq, struct t4_cqe *cqe)
101{
102 struct wr_log_entry le;
103 int idx;
104
105 if (!wq->rdev->wr_log)
106 return;
107
108 idx = (atomic_inc_return(&wq->rdev->wr_log_idx) - 1) &
109 (wq->rdev->wr_log_size - 1);
110 le.poll_sge_ts = cxgb4_read_sge_timestamp(wq->rdev->lldi.ports[0]);
111 le.poll_host_time = ktime_get();
112 le.valid = 1;
113 le.cqe_sge_ts = CQE_TS(cqe);
114 if (SQ_TYPE(cqe)) {
115 le.qid = wq->sq.qid;
116 le.opcode = CQE_OPCODE(cqe);
117 le.post_host_time = wq->sq.sw_sq[wq->sq.cidx].host_time;
118 le.post_sge_ts = wq->sq.sw_sq[wq->sq.cidx].sge_ts;
119 le.wr_id = CQE_WRID_SQ_IDX(cqe);
120 } else {
121 le.qid = wq->rq.qid;
122 le.opcode = FW_RI_RECEIVE;
123 le.post_host_time = wq->rq.sw_rq[wq->rq.cidx].host_time;
124 le.post_sge_ts = wq->rq.sw_rq[wq->rq.cidx].sge_ts;
125 le.wr_id = CQE_WRID_MSN(cqe);
126 }
127 wq->rdev->wr_log[idx] = le;
128}
129
130static int wr_log_show(struct seq_file *seq, void *v)
131{
132 struct c4iw_dev *dev = seq->private;
133 ktime_t prev_time;
134 struct wr_log_entry *lep;
135 int prev_time_set = 0;
136 int idx, end;
137
138#define ts2ns(ts) div64_u64((ts) * dev->rdev.lldi.cclk_ps, 1000)
139
140 idx = atomic_read(&dev->rdev.wr_log_idx) &
141 (dev->rdev.wr_log_size - 1);
142 end = idx - 1;
143 if (end < 0)
144 end = dev->rdev.wr_log_size - 1;
145 lep = &dev->rdev.wr_log[idx];
146 while (idx != end) {
147 if (lep->valid) {
148 if (!prev_time_set) {
149 prev_time_set = 1;
150 prev_time = lep->poll_host_time;
151 }
152 seq_printf(seq, "%04u: nsec %llu qid %u opcode "
153 "%u %s 0x%x host_wr_delta nsec %llu "
154 "post_sge_ts 0x%llx cqe_sge_ts 0x%llx "
155 "poll_sge_ts 0x%llx post_poll_delta_ns %llu "
156 "cqe_poll_delta_ns %llu\n",
157 idx,
158 ktime_to_ns(ktime_sub(lep->poll_host_time,
159 prev_time)),
160 lep->qid, lep->opcode,
161 lep->opcode == FW_RI_RECEIVE ?
162 "msn" : "wrid",
163 lep->wr_id,
164 ktime_to_ns(ktime_sub(lep->poll_host_time,
165 lep->post_host_time)),
166 lep->post_sge_ts, lep->cqe_sge_ts,
167 lep->poll_sge_ts,
168 ts2ns(lep->poll_sge_ts - lep->post_sge_ts),
169 ts2ns(lep->poll_sge_ts - lep->cqe_sge_ts));
170 prev_time = lep->poll_host_time;
171 }
172 idx++;
173 if (idx > (dev->rdev.wr_log_size - 1))
174 idx = 0;
175 lep = &dev->rdev.wr_log[idx];
176 }
177#undef ts2ns
178 return 0;
179}
180
181static int wr_log_open(struct inode *inode, struct file *file)
182{
183 return single_open(file, wr_log_show, inode->i_private);
184}
185
186static ssize_t wr_log_clear(struct file *file, const char __user *buf,
187 size_t count, loff_t *pos)
188{
189 struct c4iw_dev *dev = ((struct seq_file *)file->private_data)->private;
190 int i;
191
192 if (dev->rdev.wr_log)
193 for (i = 0; i < dev->rdev.wr_log_size; i++)
194 dev->rdev.wr_log[i].valid = 0;
195 return count;
196}
197
198static const struct file_operations wr_log_debugfs_fops = {
199 .owner = THIS_MODULE,
200 .open = wr_log_open,
201 .release = single_release,
202 .read = seq_read,
203 .llseek = seq_lseek,
204 .write = wr_log_clear,
205};
206
207static struct sockaddr_in zero_sin = {
208 .sin_family = AF_INET,
209};
210
211static struct sockaddr_in6 zero_sin6 = {
212 .sin6_family = AF_INET6,
213};
214
215static void set_ep_sin_addrs(struct c4iw_ep *ep,
216 struct sockaddr_in **lsin,
217 struct sockaddr_in **rsin,
218 struct sockaddr_in **m_lsin,
219 struct sockaddr_in **m_rsin)
220{
221 struct iw_cm_id *id = ep->com.cm_id;
222
223 *lsin = (struct sockaddr_in *)&ep->com.local_addr;
224 *rsin = (struct sockaddr_in *)&ep->com.remote_addr;
225 if (id) {
226 *m_lsin = (struct sockaddr_in *)&id->m_local_addr;
227 *m_rsin = (struct sockaddr_in *)&id->m_remote_addr;
228 } else {
229 *m_lsin = &zero_sin;
230 *m_rsin = &zero_sin;
231 }
232}
233
234static void set_ep_sin6_addrs(struct c4iw_ep *ep,
235 struct sockaddr_in6 **lsin6,
236 struct sockaddr_in6 **rsin6,
237 struct sockaddr_in6 **m_lsin6,
238 struct sockaddr_in6 **m_rsin6)
239{
240 struct iw_cm_id *id = ep->com.cm_id;
241
242 *lsin6 = (struct sockaddr_in6 *)&ep->com.local_addr;
243 *rsin6 = (struct sockaddr_in6 *)&ep->com.remote_addr;
244 if (id) {
245 *m_lsin6 = (struct sockaddr_in6 *)&id->m_local_addr;
246 *m_rsin6 = (struct sockaddr_in6 *)&id->m_remote_addr;
247 } else {
248 *m_lsin6 = &zero_sin6;
249 *m_rsin6 = &zero_sin6;
250 }
251}
252
253static int dump_qp(int id, void *p, void *data)
254{
255 struct c4iw_qp *qp = p;
256 struct c4iw_debugfs_data *qpd = data;
257 int space;
258 int cc;
259
260 if (id != qp->wq.sq.qid)
261 return 0;
262
263 space = qpd->bufsize - qpd->pos - 1;
264 if (space == 0)
265 return 1;
266
267 if (qp->ep) {
268 struct c4iw_ep *ep = qp->ep;
269
270 if (ep->com.local_addr.ss_family == AF_INET) {
271 struct sockaddr_in *lsin;
272 struct sockaddr_in *rsin;
273 struct sockaddr_in *m_lsin;
274 struct sockaddr_in *m_rsin;
275
276 set_ep_sin_addrs(ep, &lsin, &rsin, &m_lsin, &m_rsin);
277 cc = snprintf(qpd->buf + qpd->pos, space,
278 "rc qp sq id %u rq id %u state %u "
279 "onchip %u ep tid %u state %u "
280 "%pI4:%u/%u->%pI4:%u/%u\n",
281 qp->wq.sq.qid, qp->wq.rq.qid,
282 (int)qp->attr.state,
283 qp->wq.sq.flags & T4_SQ_ONCHIP,
284 ep->hwtid, (int)ep->com.state,
285 &lsin->sin_addr, ntohs(lsin->sin_port),
286 ntohs(m_lsin->sin_port),
287 &rsin->sin_addr, ntohs(rsin->sin_port),
288 ntohs(m_rsin->sin_port));
289 } else {
290 struct sockaddr_in6 *lsin6;
291 struct sockaddr_in6 *rsin6;
292 struct sockaddr_in6 *m_lsin6;
293 struct sockaddr_in6 *m_rsin6;
294
295 set_ep_sin6_addrs(ep, &lsin6, &rsin6, &m_lsin6,
296 &m_rsin6);
297 cc = snprintf(qpd->buf + qpd->pos, space,
298 "rc qp sq id %u rq id %u state %u "
299 "onchip %u ep tid %u state %u "
300 "%pI6:%u/%u->%pI6:%u/%u\n",
301 qp->wq.sq.qid, qp->wq.rq.qid,
302 (int)qp->attr.state,
303 qp->wq.sq.flags & T4_SQ_ONCHIP,
304 ep->hwtid, (int)ep->com.state,
305 &lsin6->sin6_addr,
306 ntohs(lsin6->sin6_port),
307 ntohs(m_lsin6->sin6_port),
308 &rsin6->sin6_addr,
309 ntohs(rsin6->sin6_port),
310 ntohs(m_rsin6->sin6_port));
311 }
312 } else
313 cc = snprintf(qpd->buf + qpd->pos, space,
314 "qp sq id %u rq id %u state %u onchip %u\n",
315 qp->wq.sq.qid, qp->wq.rq.qid,
316 (int)qp->attr.state,
317 qp->wq.sq.flags & T4_SQ_ONCHIP);
318 if (cc < space)
319 qpd->pos += cc;
320 return 0;
321}
322
323static int qp_release(struct inode *inode, struct file *file)
324{
325 struct c4iw_debugfs_data *qpd = file->private_data;
326 if (!qpd) {
327 pr_info("%s null qpd?\n", __func__);
328 return 0;
329 }
330 vfree(qpd->buf);
331 kfree(qpd);
332 return 0;
333}
334
335static int qp_open(struct inode *inode, struct file *file)
336{
337 struct c4iw_debugfs_data *qpd;
338 int count = 1;
339
340 qpd = kmalloc(sizeof *qpd, GFP_KERNEL);
341 if (!qpd)
342 return -ENOMEM;
343
344 qpd->devp = inode->i_private;
345 qpd->pos = 0;
346
347 spin_lock_irq(&qpd->devp->lock);
348 idr_for_each(&qpd->devp->qpidr, count_idrs, &count);
349 spin_unlock_irq(&qpd->devp->lock);
350
351 qpd->bufsize = count * 180;
352 qpd->buf = vmalloc(qpd->bufsize);
353 if (!qpd->buf) {
354 kfree(qpd);
355 return -ENOMEM;
356 }
357
358 spin_lock_irq(&qpd->devp->lock);
359 idr_for_each(&qpd->devp->qpidr, dump_qp, qpd);
360 spin_unlock_irq(&qpd->devp->lock);
361
362 qpd->buf[qpd->pos++] = 0;
363 file->private_data = qpd;
364 return 0;
365}
366
367static const struct file_operations qp_debugfs_fops = {
368 .owner = THIS_MODULE,
369 .open = qp_open,
370 .release = qp_release,
371 .read = debugfs_read,
372 .llseek = default_llseek,
373};
374
375static int dump_stag(int id, void *p, void *data)
376{
377 struct c4iw_debugfs_data *stagd = data;
378 int space;
379 int cc;
380 struct fw_ri_tpte tpte;
381 int ret;
382
383 space = stagd->bufsize - stagd->pos - 1;
384 if (space == 0)
385 return 1;
386
387 ret = cxgb4_read_tpte(stagd->devp->rdev.lldi.ports[0], (u32)id<<8,
388 (__be32 *)&tpte);
389 if (ret) {
390 dev_err(&stagd->devp->rdev.lldi.pdev->dev,
391 "%s cxgb4_read_tpte err %d\n", __func__, ret);
392 return ret;
393 }
394 cc = snprintf(stagd->buf + stagd->pos, space,
395 "stag: idx 0x%x valid %d key 0x%x state %d pdid %d "
396 "perm 0x%x ps %d len 0x%llx va 0x%llx\n",
397 (u32)id<<8,
398 FW_RI_TPTE_VALID_G(ntohl(tpte.valid_to_pdid)),
399 FW_RI_TPTE_STAGKEY_G(ntohl(tpte.valid_to_pdid)),
400 FW_RI_TPTE_STAGSTATE_G(ntohl(tpte.valid_to_pdid)),
401 FW_RI_TPTE_PDID_G(ntohl(tpte.valid_to_pdid)),
402 FW_RI_TPTE_PERM_G(ntohl(tpte.locread_to_qpid)),
403 FW_RI_TPTE_PS_G(ntohl(tpte.locread_to_qpid)),
404 ((u64)ntohl(tpte.len_hi) << 32) | ntohl(tpte.len_lo),
405 ((u64)ntohl(tpte.va_hi) << 32) | ntohl(tpte.va_lo_fbo));
406 if (cc < space)
407 stagd->pos += cc;
408 return 0;
409}
410
411static int stag_release(struct inode *inode, struct file *file)
412{
413 struct c4iw_debugfs_data *stagd = file->private_data;
414 if (!stagd) {
415 pr_info("%s null stagd?\n", __func__);
416 return 0;
417 }
418 vfree(stagd->buf);
419 kfree(stagd);
420 return 0;
421}
422
423static int stag_open(struct inode *inode, struct file *file)
424{
425 struct c4iw_debugfs_data *stagd;
426 int ret = 0;
427 int count = 1;
428
429 stagd = kmalloc(sizeof *stagd, GFP_KERNEL);
430 if (!stagd) {
431 ret = -ENOMEM;
432 goto out;
433 }
434 stagd->devp = inode->i_private;
435 stagd->pos = 0;
436
437 spin_lock_irq(&stagd->devp->lock);
438 idr_for_each(&stagd->devp->mmidr, count_idrs, &count);
439 spin_unlock_irq(&stagd->devp->lock);
440
441 stagd->bufsize = count * 256;
442 stagd->buf = vmalloc(stagd->bufsize);
443 if (!stagd->buf) {
444 ret = -ENOMEM;
445 goto err1;
446 }
447
448 spin_lock_irq(&stagd->devp->lock);
449 idr_for_each(&stagd->devp->mmidr, dump_stag, stagd);
450 spin_unlock_irq(&stagd->devp->lock);
451
452 stagd->buf[stagd->pos++] = 0;
453 file->private_data = stagd;
454 goto out;
455err1:
456 kfree(stagd);
457out:
458 return ret;
459}
460
461static const struct file_operations stag_debugfs_fops = {
462 .owner = THIS_MODULE,
463 .open = stag_open,
464 .release = stag_release,
465 .read = debugfs_read,
466 .llseek = default_llseek,
467};
468
469static char *db_state_str[] = {"NORMAL", "FLOW_CONTROL", "RECOVERY", "STOPPED"};
470
471static int stats_show(struct seq_file *seq, void *v)
472{
473 struct c4iw_dev *dev = seq->private;
474
475 seq_printf(seq, " Object: %10s %10s %10s %10s\n", "Total", "Current",
476 "Max", "Fail");
477 seq_printf(seq, " PDID: %10llu %10llu %10llu %10llu\n",
478 dev->rdev.stats.pd.total, dev->rdev.stats.pd.cur,
479 dev->rdev.stats.pd.max, dev->rdev.stats.pd.fail);
480 seq_printf(seq, " QID: %10llu %10llu %10llu %10llu\n",
481 dev->rdev.stats.qid.total, dev->rdev.stats.qid.cur,
482 dev->rdev.stats.qid.max, dev->rdev.stats.qid.fail);
483 seq_printf(seq, " TPTMEM: %10llu %10llu %10llu %10llu\n",
484 dev->rdev.stats.stag.total, dev->rdev.stats.stag.cur,
485 dev->rdev.stats.stag.max, dev->rdev.stats.stag.fail);
486 seq_printf(seq, " PBLMEM: %10llu %10llu %10llu %10llu\n",
487 dev->rdev.stats.pbl.total, dev->rdev.stats.pbl.cur,
488 dev->rdev.stats.pbl.max, dev->rdev.stats.pbl.fail);
489 seq_printf(seq, " RQTMEM: %10llu %10llu %10llu %10llu\n",
490 dev->rdev.stats.rqt.total, dev->rdev.stats.rqt.cur,
491 dev->rdev.stats.rqt.max, dev->rdev.stats.rqt.fail);
492 seq_printf(seq, " OCQPMEM: %10llu %10llu %10llu %10llu\n",
493 dev->rdev.stats.ocqp.total, dev->rdev.stats.ocqp.cur,
494 dev->rdev.stats.ocqp.max, dev->rdev.stats.ocqp.fail);
495 seq_printf(seq, " DB FULL: %10llu\n", dev->rdev.stats.db_full);
496 seq_printf(seq, " DB EMPTY: %10llu\n", dev->rdev.stats.db_empty);
497 seq_printf(seq, " DB DROP: %10llu\n", dev->rdev.stats.db_drop);
498 seq_printf(seq, " DB State: %s Transitions %llu FC Interruptions %llu\n",
499 db_state_str[dev->db_state],
500 dev->rdev.stats.db_state_transitions,
501 dev->rdev.stats.db_fc_interruptions);
502 seq_printf(seq, "TCAM_FULL: %10llu\n", dev->rdev.stats.tcam_full);
503 seq_printf(seq, "ACT_OFLD_CONN_FAILS: %10llu\n",
504 dev->rdev.stats.act_ofld_conn_fails);
505 seq_printf(seq, "PAS_OFLD_CONN_FAILS: %10llu\n",
506 dev->rdev.stats.pas_ofld_conn_fails);
507 seq_printf(seq, "NEG_ADV_RCVD: %10llu\n", dev->rdev.stats.neg_adv);
508 seq_printf(seq, "AVAILABLE IRD: %10u\n", dev->avail_ird);
509 return 0;
510}
511
512static int stats_open(struct inode *inode, struct file *file)
513{
514 return single_open(file, stats_show, inode->i_private);
515}
516
517static ssize_t stats_clear(struct file *file, const char __user *buf,
518 size_t count, loff_t *pos)
519{
520 struct c4iw_dev *dev = ((struct seq_file *)file->private_data)->private;
521
522 mutex_lock(&dev->rdev.stats.lock);
523 dev->rdev.stats.pd.max = 0;
524 dev->rdev.stats.pd.fail = 0;
525 dev->rdev.stats.qid.max = 0;
526 dev->rdev.stats.qid.fail = 0;
527 dev->rdev.stats.stag.max = 0;
528 dev->rdev.stats.stag.fail = 0;
529 dev->rdev.stats.pbl.max = 0;
530 dev->rdev.stats.pbl.fail = 0;
531 dev->rdev.stats.rqt.max = 0;
532 dev->rdev.stats.rqt.fail = 0;
533 dev->rdev.stats.ocqp.max = 0;
534 dev->rdev.stats.ocqp.fail = 0;
535 dev->rdev.stats.db_full = 0;
536 dev->rdev.stats.db_empty = 0;
537 dev->rdev.stats.db_drop = 0;
538 dev->rdev.stats.db_state_transitions = 0;
539 dev->rdev.stats.tcam_full = 0;
540 dev->rdev.stats.act_ofld_conn_fails = 0;
541 dev->rdev.stats.pas_ofld_conn_fails = 0;
542 mutex_unlock(&dev->rdev.stats.lock);
543 return count;
544}
545
546static const struct file_operations stats_debugfs_fops = {
547 .owner = THIS_MODULE,
548 .open = stats_open,
549 .release = single_release,
550 .read = seq_read,
551 .llseek = seq_lseek,
552 .write = stats_clear,
553};
554
555static int dump_ep(int id, void *p, void *data)
556{
557 struct c4iw_ep *ep = p;
558 struct c4iw_debugfs_data *epd = data;
559 int space;
560 int cc;
561
562 space = epd->bufsize - epd->pos - 1;
563 if (space == 0)
564 return 1;
565
566 if (ep->com.local_addr.ss_family == AF_INET) {
567 struct sockaddr_in *lsin;
568 struct sockaddr_in *rsin;
569 struct sockaddr_in *m_lsin;
570 struct sockaddr_in *m_rsin;
571
572 set_ep_sin_addrs(ep, &lsin, &rsin, &m_lsin, &m_rsin);
573 cc = snprintf(epd->buf + epd->pos, space,
574 "ep %p cm_id %p qp %p state %d flags 0x%lx "
575 "history 0x%lx hwtid %d atid %d "
576 "conn_na %u abort_na %u "
577 "%pI4:%d/%d <-> %pI4:%d/%d\n",
578 ep, ep->com.cm_id, ep->com.qp,
579 (int)ep->com.state, ep->com.flags,
580 ep->com.history, ep->hwtid, ep->atid,
581 ep->stats.connect_neg_adv,
582 ep->stats.abort_neg_adv,
583 &lsin->sin_addr, ntohs(lsin->sin_port),
584 ntohs(m_lsin->sin_port),
585 &rsin->sin_addr, ntohs(rsin->sin_port),
586 ntohs(m_rsin->sin_port));
587 } else {
588 struct sockaddr_in6 *lsin6;
589 struct sockaddr_in6 *rsin6;
590 struct sockaddr_in6 *m_lsin6;
591 struct sockaddr_in6 *m_rsin6;
592
593 set_ep_sin6_addrs(ep, &lsin6, &rsin6, &m_lsin6, &m_rsin6);
594 cc = snprintf(epd->buf + epd->pos, space,
595 "ep %p cm_id %p qp %p state %d flags 0x%lx "
596 "history 0x%lx hwtid %d atid %d "
597 "conn_na %u abort_na %u "
598 "%pI6:%d/%d <-> %pI6:%d/%d\n",
599 ep, ep->com.cm_id, ep->com.qp,
600 (int)ep->com.state, ep->com.flags,
601 ep->com.history, ep->hwtid, ep->atid,
602 ep->stats.connect_neg_adv,
603 ep->stats.abort_neg_adv,
604 &lsin6->sin6_addr, ntohs(lsin6->sin6_port),
605 ntohs(m_lsin6->sin6_port),
606 &rsin6->sin6_addr, ntohs(rsin6->sin6_port),
607 ntohs(m_rsin6->sin6_port));
608 }
609 if (cc < space)
610 epd->pos += cc;
611 return 0;
612}
613
614static int dump_listen_ep(int id, void *p, void *data)
615{
616 struct c4iw_listen_ep *ep = p;
617 struct c4iw_debugfs_data *epd = data;
618 int space;
619 int cc;
620
621 space = epd->bufsize - epd->pos - 1;
622 if (space == 0)
623 return 1;
624
625 if (ep->com.local_addr.ss_family == AF_INET) {
626 struct sockaddr_in *lsin = (struct sockaddr_in *)
627 &ep->com.cm_id->local_addr;
628 struct sockaddr_in *m_lsin = (struct sockaddr_in *)
629 &ep->com.cm_id->m_local_addr;
630
631 cc = snprintf(epd->buf + epd->pos, space,
632 "ep %p cm_id %p state %d flags 0x%lx stid %d "
633 "backlog %d %pI4:%d/%d\n",
634 ep, ep->com.cm_id, (int)ep->com.state,
635 ep->com.flags, ep->stid, ep->backlog,
636 &lsin->sin_addr, ntohs(lsin->sin_port),
637 ntohs(m_lsin->sin_port));
638 } else {
639 struct sockaddr_in6 *lsin6 = (struct sockaddr_in6 *)
640 &ep->com.cm_id->local_addr;
641 struct sockaddr_in6 *m_lsin6 = (struct sockaddr_in6 *)
642 &ep->com.cm_id->m_local_addr;
643
644 cc = snprintf(epd->buf + epd->pos, space,
645 "ep %p cm_id %p state %d flags 0x%lx stid %d "
646 "backlog %d %pI6:%d/%d\n",
647 ep, ep->com.cm_id, (int)ep->com.state,
648 ep->com.flags, ep->stid, ep->backlog,
649 &lsin6->sin6_addr, ntohs(lsin6->sin6_port),
650 ntohs(m_lsin6->sin6_port));
651 }
652 if (cc < space)
653 epd->pos += cc;
654 return 0;
655}
656
657static int ep_release(struct inode *inode, struct file *file)
658{
659 struct c4iw_debugfs_data *epd = file->private_data;
660 if (!epd) {
661 pr_info("%s null qpd?\n", __func__);
662 return 0;
663 }
664 vfree(epd->buf);
665 kfree(epd);
666 return 0;
667}
668
669static int ep_open(struct inode *inode, struct file *file)
670{
671 struct c4iw_debugfs_data *epd;
672 int ret = 0;
673 int count = 1;
674
675 epd = kmalloc(sizeof(*epd), GFP_KERNEL);
676 if (!epd) {
677 ret = -ENOMEM;
678 goto out;
679 }
680 epd->devp = inode->i_private;
681 epd->pos = 0;
682
683 spin_lock_irq(&epd->devp->lock);
684 idr_for_each(&epd->devp->hwtid_idr, count_idrs, &count);
685 idr_for_each(&epd->devp->atid_idr, count_idrs, &count);
686 idr_for_each(&epd->devp->stid_idr, count_idrs, &count);
687 spin_unlock_irq(&epd->devp->lock);
688
689 epd->bufsize = count * 240;
690 epd->buf = vmalloc(epd->bufsize);
691 if (!epd->buf) {
692 ret = -ENOMEM;
693 goto err1;
694 }
695
696 spin_lock_irq(&epd->devp->lock);
697 idr_for_each(&epd->devp->hwtid_idr, dump_ep, epd);
698 idr_for_each(&epd->devp->atid_idr, dump_ep, epd);
699 idr_for_each(&epd->devp->stid_idr, dump_listen_ep, epd);
700 spin_unlock_irq(&epd->devp->lock);
701
702 file->private_data = epd;
703 goto out;
704err1:
705 kfree(epd);
706out:
707 return ret;
708}
709
710static const struct file_operations ep_debugfs_fops = {
711 .owner = THIS_MODULE,
712 .open = ep_open,
713 .release = ep_release,
714 .read = debugfs_read,
715};
716
717static int setup_debugfs(struct c4iw_dev *devp)
718{
719 if (!devp->debugfs_root)
720 return -1;
721
722 debugfs_create_file_size("qps", S_IWUSR, devp->debugfs_root,
723 (void *)devp, &qp_debugfs_fops, 4096);
724
725 debugfs_create_file_size("stags", S_IWUSR, devp->debugfs_root,
726 (void *)devp, &stag_debugfs_fops, 4096);
727
728 debugfs_create_file_size("stats", S_IWUSR, devp->debugfs_root,
729 (void *)devp, &stats_debugfs_fops, 4096);
730
731 debugfs_create_file_size("eps", S_IWUSR, devp->debugfs_root,
732 (void *)devp, &ep_debugfs_fops, 4096);
733
734 if (c4iw_wr_log)
735 debugfs_create_file_size("wr_log", S_IWUSR, devp->debugfs_root,
736 (void *)devp, &wr_log_debugfs_fops, 4096);
737 return 0;
738}
739
740void c4iw_release_dev_ucontext(struct c4iw_rdev *rdev,
741 struct c4iw_dev_ucontext *uctx)
742{
743 struct list_head *pos, *nxt;
744 struct c4iw_qid_list *entry;
745
746 mutex_lock(&uctx->lock);
747 list_for_each_safe(pos, nxt, &uctx->qpids) {
748 entry = list_entry(pos, struct c4iw_qid_list, entry);
749 list_del_init(&entry->entry);
750 if (!(entry->qid & rdev->qpmask)) {
751 c4iw_put_resource(&rdev->resource.qid_table,
752 entry->qid);
753 mutex_lock(&rdev->stats.lock);
754 rdev->stats.qid.cur -= rdev->qpmask + 1;
755 mutex_unlock(&rdev->stats.lock);
756 }
757 kfree(entry);
758 }
759
760 list_for_each_safe(pos, nxt, &uctx->cqids) {
761 entry = list_entry(pos, struct c4iw_qid_list, entry);
762 list_del_init(&entry->entry);
763 kfree(entry);
764 }
765 mutex_unlock(&uctx->lock);
766}
767
768void c4iw_init_dev_ucontext(struct c4iw_rdev *rdev,
769 struct c4iw_dev_ucontext *uctx)
770{
771 INIT_LIST_HEAD(&uctx->qpids);
772 INIT_LIST_HEAD(&uctx->cqids);
773 mutex_init(&uctx->lock);
774}
775
776
777static int c4iw_rdev_open(struct c4iw_rdev *rdev)
778{
779 int err;
780
781 c4iw_init_dev_ucontext(rdev, &rdev->uctx);
782
783
784
785
786
787
788 if (rdev->lldi.udb_density != rdev->lldi.ucq_density) {
789 pr_err("%s: unsupported udb/ucq densities %u/%u\n",
790 pci_name(rdev->lldi.pdev), rdev->lldi.udb_density,
791 rdev->lldi.ucq_density);
792 return -EINVAL;
793 }
794 if (rdev->lldi.vr->qp.start != rdev->lldi.vr->cq.start ||
795 rdev->lldi.vr->qp.size != rdev->lldi.vr->cq.size) {
796 pr_err("%s: unsupported qp and cq id ranges qp start %u size %u cq start %u size %u\n",
797 pci_name(rdev->lldi.pdev), rdev->lldi.vr->qp.start,
798 rdev->lldi.vr->qp.size, rdev->lldi.vr->cq.size,
799 rdev->lldi.vr->cq.size);
800 return -EINVAL;
801 }
802
803 rdev->qpmask = rdev->lldi.udb_density - 1;
804 rdev->cqmask = rdev->lldi.ucq_density - 1;
805 pr_debug("dev %s stag start 0x%0x size 0x%0x num stags %d pbl start 0x%0x size 0x%0x rq start 0x%0x size 0x%0x qp qid start %u size %u cq qid start %u size %u\n",
806 pci_name(rdev->lldi.pdev), rdev->lldi.vr->stag.start,
807 rdev->lldi.vr->stag.size, c4iw_num_stags(rdev),
808 rdev->lldi.vr->pbl.start,
809 rdev->lldi.vr->pbl.size, rdev->lldi.vr->rq.start,
810 rdev->lldi.vr->rq.size,
811 rdev->lldi.vr->qp.start,
812 rdev->lldi.vr->qp.size,
813 rdev->lldi.vr->cq.start,
814 rdev->lldi.vr->cq.size);
815 pr_debug("udb %pR db_reg %p gts_reg %p qpmask 0x%x cqmask 0x%x\n",
816 &rdev->lldi.pdev->resource[2],
817 rdev->lldi.db_reg, rdev->lldi.gts_reg,
818 rdev->qpmask, rdev->cqmask);
819
820 if (c4iw_num_stags(rdev) == 0)
821 return -EINVAL;
822
823 rdev->stats.pd.total = T4_MAX_NUM_PD;
824 rdev->stats.stag.total = rdev->lldi.vr->stag.size;
825 rdev->stats.pbl.total = rdev->lldi.vr->pbl.size;
826 rdev->stats.rqt.total = rdev->lldi.vr->rq.size;
827 rdev->stats.ocqp.total = rdev->lldi.vr->ocq.size;
828 rdev->stats.qid.total = rdev->lldi.vr->qp.size;
829
830 err = c4iw_init_resource(rdev, c4iw_num_stags(rdev), T4_MAX_NUM_PD);
831 if (err) {
832 pr_err("error %d initializing resources\n", err);
833 return err;
834 }
835 err = c4iw_pblpool_create(rdev);
836 if (err) {
837 pr_err("error %d initializing pbl pool\n", err);
838 goto destroy_resource;
839 }
840 err = c4iw_rqtpool_create(rdev);
841 if (err) {
842 pr_err("error %d initializing rqt pool\n", err);
843 goto destroy_pblpool;
844 }
845 err = c4iw_ocqp_pool_create(rdev);
846 if (err) {
847 pr_err("error %d initializing ocqp pool\n", err);
848 goto destroy_rqtpool;
849 }
850 rdev->status_page = (struct t4_dev_status_page *)
851 __get_free_page(GFP_KERNEL);
852 if (!rdev->status_page) {
853 err = -ENOMEM;
854 goto destroy_ocqp_pool;
855 }
856 rdev->status_page->qp_start = rdev->lldi.vr->qp.start;
857 rdev->status_page->qp_size = rdev->lldi.vr->qp.size;
858 rdev->status_page->cq_start = rdev->lldi.vr->cq.start;
859 rdev->status_page->cq_size = rdev->lldi.vr->cq.size;
860
861 if (c4iw_wr_log) {
862 rdev->wr_log = kzalloc((1 << c4iw_wr_log_size_order) *
863 sizeof(*rdev->wr_log), GFP_KERNEL);
864 if (rdev->wr_log) {
865 rdev->wr_log_size = 1 << c4iw_wr_log_size_order;
866 atomic_set(&rdev->wr_log_idx, 0);
867 }
868 }
869
870 rdev->free_workq = create_singlethread_workqueue("iw_cxgb4_free");
871 if (!rdev->free_workq) {
872 err = -ENOMEM;
873 goto err_free_status_page_and_wr_log;
874 }
875
876 rdev->status_page->db_off = 0;
877
878 return 0;
879err_free_status_page_and_wr_log:
880 if (c4iw_wr_log && rdev->wr_log)
881 kfree(rdev->wr_log);
882 free_page((unsigned long)rdev->status_page);
883destroy_ocqp_pool:
884 c4iw_ocqp_pool_destroy(rdev);
885destroy_rqtpool:
886 c4iw_rqtpool_destroy(rdev);
887destroy_pblpool:
888 c4iw_pblpool_destroy(rdev);
889destroy_resource:
890 c4iw_destroy_resource(&rdev->resource);
891 return err;
892}
893
894static void c4iw_rdev_close(struct c4iw_rdev *rdev)
895{
896 destroy_workqueue(rdev->free_workq);
897 kfree(rdev->wr_log);
898 c4iw_release_dev_ucontext(rdev, &rdev->uctx);
899 free_page((unsigned long)rdev->status_page);
900 c4iw_pblpool_destroy(rdev);
901 c4iw_rqtpool_destroy(rdev);
902 c4iw_ocqp_pool_destroy(rdev);
903 c4iw_destroy_resource(&rdev->resource);
904}
905
906void c4iw_dealloc(struct uld_ctx *ctx)
907{
908 c4iw_rdev_close(&ctx->dev->rdev);
909 WARN_ON_ONCE(!idr_is_empty(&ctx->dev->cqidr));
910 idr_destroy(&ctx->dev->cqidr);
911 WARN_ON_ONCE(!idr_is_empty(&ctx->dev->qpidr));
912 idr_destroy(&ctx->dev->qpidr);
913 WARN_ON_ONCE(!idr_is_empty(&ctx->dev->mmidr));
914 idr_destroy(&ctx->dev->mmidr);
915 wait_event(ctx->dev->wait, idr_is_empty(&ctx->dev->hwtid_idr));
916 idr_destroy(&ctx->dev->hwtid_idr);
917 idr_destroy(&ctx->dev->stid_idr);
918 idr_destroy(&ctx->dev->atid_idr);
919 if (ctx->dev->rdev.bar2_kva)
920 iounmap(ctx->dev->rdev.bar2_kva);
921 if (ctx->dev->rdev.oc_mw_kva)
922 iounmap(ctx->dev->rdev.oc_mw_kva);
923 ib_dealloc_device(&ctx->dev->ibdev);
924 ctx->dev = NULL;
925}
926
927static void c4iw_remove(struct uld_ctx *ctx)
928{
929 pr_debug("c4iw_dev %p\n", ctx->dev);
930 c4iw_unregister_device(ctx->dev);
931 c4iw_dealloc(ctx);
932}
933
934static int rdma_supported(const struct cxgb4_lld_info *infop)
935{
936 return infop->vr->stag.size > 0 && infop->vr->pbl.size > 0 &&
937 infop->vr->rq.size > 0 && infop->vr->qp.size > 0 &&
938 infop->vr->cq.size > 0;
939}
940
941static struct c4iw_dev *c4iw_alloc(const struct cxgb4_lld_info *infop)
942{
943 struct c4iw_dev *devp;
944 int ret;
945
946 if (!rdma_supported(infop)) {
947 pr_info("%s: RDMA not supported on this device\n",
948 pci_name(infop->pdev));
949 return ERR_PTR(-ENOSYS);
950 }
951 if (!ocqp_supported(infop))
952 pr_info("%s: On-Chip Queues not supported on this device\n",
953 pci_name(infop->pdev));
954
955 devp = (struct c4iw_dev *)ib_alloc_device(sizeof(*devp));
956 if (!devp) {
957 pr_err("Cannot allocate ib device\n");
958 return ERR_PTR(-ENOMEM);
959 }
960 devp->rdev.lldi = *infop;
961
962
963 pr_debug("Ing. padding boundary is %d, egrsstatuspagesize = %d\n",
964 devp->rdev.lldi.sge_ingpadboundary,
965 devp->rdev.lldi.sge_egrstatuspagesize);
966
967 devp->rdev.hw_queue.t4_eq_status_entries =
968 devp->rdev.lldi.sge_egrstatuspagesize / 64;
969 devp->rdev.hw_queue.t4_max_eq_size = 65520;
970 devp->rdev.hw_queue.t4_max_iq_size = 65520;
971 devp->rdev.hw_queue.t4_max_rq_size = 8192 -
972 devp->rdev.hw_queue.t4_eq_status_entries - 1;
973 devp->rdev.hw_queue.t4_max_sq_size =
974 devp->rdev.hw_queue.t4_max_eq_size -
975 devp->rdev.hw_queue.t4_eq_status_entries - 1;
976 devp->rdev.hw_queue.t4_max_qp_depth =
977 devp->rdev.hw_queue.t4_max_rq_size;
978 devp->rdev.hw_queue.t4_max_cq_depth =
979 devp->rdev.hw_queue.t4_max_iq_size - 2;
980 devp->rdev.hw_queue.t4_stat_len =
981 devp->rdev.lldi.sge_egrstatuspagesize;
982
983
984
985
986
987
988 devp->rdev.bar2_pa = pci_resource_start(devp->rdev.lldi.pdev, 2);
989 if (!is_t4(devp->rdev.lldi.adapter_type)) {
990 devp->rdev.bar2_kva = ioremap_wc(devp->rdev.bar2_pa,
991 pci_resource_len(devp->rdev.lldi.pdev, 2));
992 if (!devp->rdev.bar2_kva) {
993 pr_err("Unable to ioremap BAR2\n");
994 ib_dealloc_device(&devp->ibdev);
995 return ERR_PTR(-EINVAL);
996 }
997 } else if (ocqp_supported(infop)) {
998 devp->rdev.oc_mw_pa =
999 pci_resource_start(devp->rdev.lldi.pdev, 2) +
1000 pci_resource_len(devp->rdev.lldi.pdev, 2) -
1001 roundup_pow_of_two(devp->rdev.lldi.vr->ocq.size);
1002 devp->rdev.oc_mw_kva = ioremap_wc(devp->rdev.oc_mw_pa,
1003 devp->rdev.lldi.vr->ocq.size);
1004 if (!devp->rdev.oc_mw_kva) {
1005 pr_err("Unable to ioremap onchip mem\n");
1006 ib_dealloc_device(&devp->ibdev);
1007 return ERR_PTR(-EINVAL);
1008 }
1009 }
1010
1011 pr_debug("ocq memory: hw_start 0x%x size %u mw_pa 0x%lx mw_kva %p\n",
1012 devp->rdev.lldi.vr->ocq.start, devp->rdev.lldi.vr->ocq.size,
1013 devp->rdev.oc_mw_pa, devp->rdev.oc_mw_kva);
1014
1015 ret = c4iw_rdev_open(&devp->rdev);
1016 if (ret) {
1017 pr_err("Unable to open CXIO rdev err %d\n", ret);
1018 ib_dealloc_device(&devp->ibdev);
1019 return ERR_PTR(ret);
1020 }
1021
1022 idr_init(&devp->cqidr);
1023 idr_init(&devp->qpidr);
1024 idr_init(&devp->mmidr);
1025 idr_init(&devp->hwtid_idr);
1026 idr_init(&devp->stid_idr);
1027 idr_init(&devp->atid_idr);
1028 spin_lock_init(&devp->lock);
1029 mutex_init(&devp->rdev.stats.lock);
1030 mutex_init(&devp->db_mutex);
1031 INIT_LIST_HEAD(&devp->db_fc_list);
1032 init_waitqueue_head(&devp->wait);
1033 devp->avail_ird = devp->rdev.lldi.max_ird_adapter;
1034
1035 if (c4iw_debugfs_root) {
1036 devp->debugfs_root = debugfs_create_dir(
1037 pci_name(devp->rdev.lldi.pdev),
1038 c4iw_debugfs_root);
1039 setup_debugfs(devp);
1040 }
1041
1042
1043 return devp;
1044}
1045
1046static void *c4iw_uld_add(const struct cxgb4_lld_info *infop)
1047{
1048 struct uld_ctx *ctx;
1049 static int vers_printed;
1050 int i;
1051
1052 if (!vers_printed++)
1053 pr_info("Chelsio T4/T5 RDMA Driver - version %s\n",
1054 DRV_VERSION);
1055
1056 ctx = kzalloc(sizeof *ctx, GFP_KERNEL);
1057 if (!ctx) {
1058 ctx = ERR_PTR(-ENOMEM);
1059 goto out;
1060 }
1061 ctx->lldi = *infop;
1062
1063 pr_debug("found device %s nchan %u nrxq %u ntxq %u nports %u\n",
1064 pci_name(ctx->lldi.pdev),
1065 ctx->lldi.nchan, ctx->lldi.nrxq,
1066 ctx->lldi.ntxq, ctx->lldi.nports);
1067
1068 mutex_lock(&dev_mutex);
1069 list_add_tail(&ctx->entry, &uld_ctx_list);
1070 mutex_unlock(&dev_mutex);
1071
1072 for (i = 0; i < ctx->lldi.nrxq; i++)
1073 pr_debug("rxqid[%u] %u\n", i, ctx->lldi.rxq_ids[i]);
1074out:
1075 return ctx;
1076}
1077
1078static inline struct sk_buff *copy_gl_to_skb_pkt(const struct pkt_gl *gl,
1079 const __be64 *rsp,
1080 u32 pktshift)
1081{
1082 struct sk_buff *skb;
1083
1084
1085
1086
1087
1088
1089
1090
1091 skb = alloc_skb(gl->tot_len + sizeof(struct cpl_pass_accept_req) +
1092 sizeof(struct rss_header) - pktshift, GFP_ATOMIC);
1093 if (unlikely(!skb))
1094 return NULL;
1095
1096 __skb_put(skb, gl->tot_len + sizeof(struct cpl_pass_accept_req) +
1097 sizeof(struct rss_header) - pktshift);
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107 skb_copy_to_linear_data(skb, rsp, sizeof(struct cpl_pass_accept_req) +
1108 sizeof(struct rss_header));
1109 skb_copy_to_linear_data_offset(skb, sizeof(struct rss_header) +
1110 sizeof(struct cpl_pass_accept_req),
1111 gl->va + pktshift,
1112 gl->tot_len - pktshift);
1113 return skb;
1114}
1115
1116static inline int recv_rx_pkt(struct c4iw_dev *dev, const struct pkt_gl *gl,
1117 const __be64 *rsp)
1118{
1119 unsigned int opcode = *(u8 *)rsp;
1120 struct sk_buff *skb;
1121
1122 if (opcode != CPL_RX_PKT)
1123 goto out;
1124
1125 skb = copy_gl_to_skb_pkt(gl , rsp, dev->rdev.lldi.sge_pktshift);
1126 if (skb == NULL)
1127 goto out;
1128
1129 if (c4iw_handlers[opcode] == NULL) {
1130 pr_info("%s no handler opcode 0x%x...\n", __func__, opcode);
1131 kfree_skb(skb);
1132 goto out;
1133 }
1134 c4iw_handlers[opcode](dev, skb);
1135 return 1;
1136out:
1137 return 0;
1138}
1139
1140static int c4iw_uld_rx_handler(void *handle, const __be64 *rsp,
1141 const struct pkt_gl *gl)
1142{
1143 struct uld_ctx *ctx = handle;
1144 struct c4iw_dev *dev = ctx->dev;
1145 struct sk_buff *skb;
1146 u8 opcode;
1147
1148 if (gl == NULL) {
1149
1150 unsigned int len = 64 - sizeof(struct rsp_ctrl) - 8;
1151
1152 skb = alloc_skb(256, GFP_ATOMIC);
1153 if (!skb)
1154 goto nomem;
1155 __skb_put(skb, len);
1156 skb_copy_to_linear_data(skb, &rsp[1], len);
1157 } else if (gl == CXGB4_MSG_AN) {
1158 const struct rsp_ctrl *rc = (void *)rsp;
1159
1160 u32 qid = be32_to_cpu(rc->pldbuflen_qid);
1161 c4iw_ev_handler(dev, qid);
1162 return 0;
1163 } else if (unlikely(*(u8 *)rsp != *(u8 *)gl->va)) {
1164 if (recv_rx_pkt(dev, gl, rsp))
1165 return 0;
1166
1167 pr_info("%s: unexpected FL contents at %p, RSS %#llx, FL %#llx, len %u\n",
1168 pci_name(ctx->lldi.pdev), gl->va,
1169 be64_to_cpu(*rsp),
1170 be64_to_cpu(*(__force __be64 *)gl->va),
1171 gl->tot_len);
1172
1173 return 0;
1174 } else {
1175 skb = cxgb4_pktgl_to_skb(gl, 128, 128);
1176 if (unlikely(!skb))
1177 goto nomem;
1178 }
1179
1180 opcode = *(u8 *)rsp;
1181 if (c4iw_handlers[opcode]) {
1182 c4iw_handlers[opcode](dev, skb);
1183 } else {
1184 pr_info("%s no handler opcode 0x%x...\n", __func__, opcode);
1185 kfree_skb(skb);
1186 }
1187
1188 return 0;
1189nomem:
1190 return -1;
1191}
1192
1193static int c4iw_uld_state_change(void *handle, enum cxgb4_state new_state)
1194{
1195 struct uld_ctx *ctx = handle;
1196
1197 pr_debug("new_state %u\n", new_state);
1198 switch (new_state) {
1199 case CXGB4_STATE_UP:
1200 pr_info("%s: Up\n", pci_name(ctx->lldi.pdev));
1201 if (!ctx->dev) {
1202 ctx->dev = c4iw_alloc(&ctx->lldi);
1203 if (IS_ERR(ctx->dev)) {
1204 pr_err("%s: initialization failed: %ld\n",
1205 pci_name(ctx->lldi.pdev),
1206 PTR_ERR(ctx->dev));
1207 ctx->dev = NULL;
1208 break;
1209 }
1210
1211 INIT_WORK(&ctx->reg_work, c4iw_register_device);
1212 queue_work(reg_workq, &ctx->reg_work);
1213 }
1214 break;
1215 case CXGB4_STATE_DOWN:
1216 pr_info("%s: Down\n", pci_name(ctx->lldi.pdev));
1217 if (ctx->dev)
1218 c4iw_remove(ctx);
1219 break;
1220 case CXGB4_STATE_START_RECOVERY:
1221 pr_info("%s: Fatal Error\n", pci_name(ctx->lldi.pdev));
1222 if (ctx->dev) {
1223 struct ib_event event;
1224
1225 ctx->dev->rdev.flags |= T4_FATAL_ERROR;
1226 memset(&event, 0, sizeof event);
1227 event.event = IB_EVENT_DEVICE_FATAL;
1228 event.device = &ctx->dev->ibdev;
1229 ib_dispatch_event(&event);
1230 c4iw_remove(ctx);
1231 }
1232 break;
1233 case CXGB4_STATE_DETACH:
1234 pr_info("%s: Detach\n", pci_name(ctx->lldi.pdev));
1235 if (ctx->dev)
1236 c4iw_remove(ctx);
1237 break;
1238 }
1239 return 0;
1240}
1241
1242static int disable_qp_db(int id, void *p, void *data)
1243{
1244 struct c4iw_qp *qp = p;
1245
1246 t4_disable_wq_db(&qp->wq);
1247 return 0;
1248}
1249
1250static void stop_queues(struct uld_ctx *ctx)
1251{
1252 unsigned long flags;
1253
1254 spin_lock_irqsave(&ctx->dev->lock, flags);
1255 ctx->dev->rdev.stats.db_state_transitions++;
1256 ctx->dev->db_state = STOPPED;
1257 if (ctx->dev->rdev.flags & T4_STATUS_PAGE_DISABLED)
1258 idr_for_each(&ctx->dev->qpidr, disable_qp_db, NULL);
1259 else
1260 ctx->dev->rdev.status_page->db_off = 1;
1261 spin_unlock_irqrestore(&ctx->dev->lock, flags);
1262}
1263
1264static int enable_qp_db(int id, void *p, void *data)
1265{
1266 struct c4iw_qp *qp = p;
1267
1268 t4_enable_wq_db(&qp->wq);
1269 return 0;
1270}
1271
1272static void resume_rc_qp(struct c4iw_qp *qp)
1273{
1274 spin_lock(&qp->lock);
1275 t4_ring_sq_db(&qp->wq, qp->wq.sq.wq_pidx_inc, NULL);
1276 qp->wq.sq.wq_pidx_inc = 0;
1277 t4_ring_rq_db(&qp->wq, qp->wq.rq.wq_pidx_inc, NULL);
1278 qp->wq.rq.wq_pidx_inc = 0;
1279 spin_unlock(&qp->lock);
1280}
1281
1282static void resume_a_chunk(struct uld_ctx *ctx)
1283{
1284 int i;
1285 struct c4iw_qp *qp;
1286
1287 for (i = 0; i < DB_FC_RESUME_SIZE; i++) {
1288 qp = list_first_entry(&ctx->dev->db_fc_list, struct c4iw_qp,
1289 db_fc_entry);
1290 list_del_init(&qp->db_fc_entry);
1291 resume_rc_qp(qp);
1292 if (list_empty(&ctx->dev->db_fc_list))
1293 break;
1294 }
1295}
1296
1297static void resume_queues(struct uld_ctx *ctx)
1298{
1299 spin_lock_irq(&ctx->dev->lock);
1300 if (ctx->dev->db_state != STOPPED)
1301 goto out;
1302 ctx->dev->db_state = FLOW_CONTROL;
1303 while (1) {
1304 if (list_empty(&ctx->dev->db_fc_list)) {
1305 WARN_ON(ctx->dev->db_state != FLOW_CONTROL);
1306 ctx->dev->db_state = NORMAL;
1307 ctx->dev->rdev.stats.db_state_transitions++;
1308 if (ctx->dev->rdev.flags & T4_STATUS_PAGE_DISABLED) {
1309 idr_for_each(&ctx->dev->qpidr, enable_qp_db,
1310 NULL);
1311 } else {
1312 ctx->dev->rdev.status_page->db_off = 0;
1313 }
1314 break;
1315 } else {
1316 if (cxgb4_dbfifo_count(ctx->dev->rdev.lldi.ports[0], 1)
1317 < (ctx->dev->rdev.lldi.dbfifo_int_thresh <<
1318 DB_FC_DRAIN_THRESH)) {
1319 resume_a_chunk(ctx);
1320 }
1321 if (!list_empty(&ctx->dev->db_fc_list)) {
1322 spin_unlock_irq(&ctx->dev->lock);
1323 if (DB_FC_RESUME_DELAY) {
1324 set_current_state(TASK_UNINTERRUPTIBLE);
1325 schedule_timeout(DB_FC_RESUME_DELAY);
1326 }
1327 spin_lock_irq(&ctx->dev->lock);
1328 if (ctx->dev->db_state != FLOW_CONTROL)
1329 break;
1330 }
1331 }
1332 }
1333out:
1334 if (ctx->dev->db_state != NORMAL)
1335 ctx->dev->rdev.stats.db_fc_interruptions++;
1336 spin_unlock_irq(&ctx->dev->lock);
1337}
1338
1339struct qp_list {
1340 unsigned idx;
1341 struct c4iw_qp **qps;
1342};
1343
1344static int add_and_ref_qp(int id, void *p, void *data)
1345{
1346 struct qp_list *qp_listp = data;
1347 struct c4iw_qp *qp = p;
1348
1349 c4iw_qp_add_ref(&qp->ibqp);
1350 qp_listp->qps[qp_listp->idx++] = qp;
1351 return 0;
1352}
1353
1354static int count_qps(int id, void *p, void *data)
1355{
1356 unsigned *countp = data;
1357 (*countp)++;
1358 return 0;
1359}
1360
1361static void deref_qps(struct qp_list *qp_list)
1362{
1363 int idx;
1364
1365 for (idx = 0; idx < qp_list->idx; idx++)
1366 c4iw_qp_rem_ref(&qp_list->qps[idx]->ibqp);
1367}
1368
1369static void recover_lost_dbs(struct uld_ctx *ctx, struct qp_list *qp_list)
1370{
1371 int idx;
1372 int ret;
1373
1374 for (idx = 0; idx < qp_list->idx; idx++) {
1375 struct c4iw_qp *qp = qp_list->qps[idx];
1376
1377 spin_lock_irq(&qp->rhp->lock);
1378 spin_lock(&qp->lock);
1379 ret = cxgb4_sync_txq_pidx(qp->rhp->rdev.lldi.ports[0],
1380 qp->wq.sq.qid,
1381 t4_sq_host_wq_pidx(&qp->wq),
1382 t4_sq_wq_size(&qp->wq));
1383 if (ret) {
1384 pr_err("%s: Fatal error - DB overflow recovery failed - error syncing SQ qid %u\n",
1385 pci_name(ctx->lldi.pdev), qp->wq.sq.qid);
1386 spin_unlock(&qp->lock);
1387 spin_unlock_irq(&qp->rhp->lock);
1388 return;
1389 }
1390 qp->wq.sq.wq_pidx_inc = 0;
1391
1392 ret = cxgb4_sync_txq_pidx(qp->rhp->rdev.lldi.ports[0],
1393 qp->wq.rq.qid,
1394 t4_rq_host_wq_pidx(&qp->wq),
1395 t4_rq_wq_size(&qp->wq));
1396
1397 if (ret) {
1398 pr_err("%s: Fatal error - DB overflow recovery failed - error syncing RQ qid %u\n",
1399 pci_name(ctx->lldi.pdev), qp->wq.rq.qid);
1400 spin_unlock(&qp->lock);
1401 spin_unlock_irq(&qp->rhp->lock);
1402 return;
1403 }
1404 qp->wq.rq.wq_pidx_inc = 0;
1405 spin_unlock(&qp->lock);
1406 spin_unlock_irq(&qp->rhp->lock);
1407
1408
1409 while (cxgb4_dbfifo_count(qp->rhp->rdev.lldi.ports[0], 1) > 0) {
1410 set_current_state(TASK_UNINTERRUPTIBLE);
1411 schedule_timeout(usecs_to_jiffies(10));
1412 }
1413 }
1414}
1415
1416static void recover_queues(struct uld_ctx *ctx)
1417{
1418 int count = 0;
1419 struct qp_list qp_list;
1420 int ret;
1421
1422
1423 set_current_state(TASK_UNINTERRUPTIBLE);
1424 schedule_timeout(usecs_to_jiffies(1000));
1425
1426
1427 ret = cxgb4_flush_eq_cache(ctx->dev->rdev.lldi.ports[0]);
1428 if (ret) {
1429 pr_err("%s: Fatal error - DB overflow recovery failed\n",
1430 pci_name(ctx->lldi.pdev));
1431 return;
1432 }
1433
1434
1435 spin_lock_irq(&ctx->dev->lock);
1436 WARN_ON(ctx->dev->db_state != STOPPED);
1437 ctx->dev->db_state = RECOVERY;
1438 idr_for_each(&ctx->dev->qpidr, count_qps, &count);
1439
1440 qp_list.qps = kzalloc(count * sizeof *qp_list.qps, GFP_ATOMIC);
1441 if (!qp_list.qps) {
1442 spin_unlock_irq(&ctx->dev->lock);
1443 return;
1444 }
1445 qp_list.idx = 0;
1446
1447
1448 idr_for_each(&ctx->dev->qpidr, add_and_ref_qp, &qp_list);
1449
1450 spin_unlock_irq(&ctx->dev->lock);
1451
1452
1453 recover_lost_dbs(ctx, &qp_list);
1454
1455
1456 deref_qps(&qp_list);
1457 kfree(qp_list.qps);
1458
1459 spin_lock_irq(&ctx->dev->lock);
1460 WARN_ON(ctx->dev->db_state != RECOVERY);
1461 ctx->dev->db_state = STOPPED;
1462 spin_unlock_irq(&ctx->dev->lock);
1463}
1464
1465static int c4iw_uld_control(void *handle, enum cxgb4_control control, ...)
1466{
1467 struct uld_ctx *ctx = handle;
1468
1469 switch (control) {
1470 case CXGB4_CONTROL_DB_FULL:
1471 stop_queues(ctx);
1472 ctx->dev->rdev.stats.db_full++;
1473 break;
1474 case CXGB4_CONTROL_DB_EMPTY:
1475 resume_queues(ctx);
1476 mutex_lock(&ctx->dev->rdev.stats.lock);
1477 ctx->dev->rdev.stats.db_empty++;
1478 mutex_unlock(&ctx->dev->rdev.stats.lock);
1479 break;
1480 case CXGB4_CONTROL_DB_DROP:
1481 recover_queues(ctx);
1482 mutex_lock(&ctx->dev->rdev.stats.lock);
1483 ctx->dev->rdev.stats.db_drop++;
1484 mutex_unlock(&ctx->dev->rdev.stats.lock);
1485 break;
1486 default:
1487 pr_warn("%s: unknown control cmd %u\n",
1488 pci_name(ctx->lldi.pdev), control);
1489 break;
1490 }
1491 return 0;
1492}
1493
1494static struct cxgb4_uld_info c4iw_uld_info = {
1495 .name = DRV_NAME,
1496 .nrxq = MAX_ULD_QSETS,
1497 .ntxq = MAX_ULD_QSETS,
1498 .rxq_size = 511,
1499 .ciq = true,
1500 .lro = false,
1501 .add = c4iw_uld_add,
1502 .rx_handler = c4iw_uld_rx_handler,
1503 .state_change = c4iw_uld_state_change,
1504 .control = c4iw_uld_control,
1505};
1506
1507void _c4iw_free_wr_wait(struct kref *kref)
1508{
1509 struct c4iw_wr_wait *wr_waitp;
1510
1511 wr_waitp = container_of(kref, struct c4iw_wr_wait, kref);
1512 pr_debug("Free wr_wait %p\n", wr_waitp);
1513 kfree(wr_waitp);
1514}
1515
1516struct c4iw_wr_wait *c4iw_alloc_wr_wait(gfp_t gfp)
1517{
1518 struct c4iw_wr_wait *wr_waitp;
1519
1520 wr_waitp = kzalloc(sizeof(*wr_waitp), gfp);
1521 if (wr_waitp) {
1522 kref_init(&wr_waitp->kref);
1523 pr_debug("wr_wait %p\n", wr_waitp);
1524 }
1525 return wr_waitp;
1526}
1527
1528static int __init c4iw_init_module(void)
1529{
1530 int err;
1531
1532 err = c4iw_cm_init();
1533 if (err)
1534 return err;
1535
1536 c4iw_debugfs_root = debugfs_create_dir(DRV_NAME, NULL);
1537 if (!c4iw_debugfs_root)
1538 pr_warn("could not create debugfs entry, continuing\n");
1539
1540 reg_workq = create_singlethread_workqueue("Register_iWARP_device");
1541 if (!reg_workq) {
1542 pr_err("Failed creating workqueue to register iwarp device\n");
1543 return -ENOMEM;
1544 }
1545
1546 cxgb4_register_uld(CXGB4_ULD_RDMA, &c4iw_uld_info);
1547
1548 return 0;
1549}
1550
1551static void __exit c4iw_exit_module(void)
1552{
1553 struct uld_ctx *ctx, *tmp;
1554
1555 mutex_lock(&dev_mutex);
1556 list_for_each_entry_safe(ctx, tmp, &uld_ctx_list, entry) {
1557 if (ctx->dev)
1558 c4iw_remove(ctx);
1559 kfree(ctx);
1560 }
1561 mutex_unlock(&dev_mutex);
1562 flush_workqueue(reg_workq);
1563 destroy_workqueue(reg_workq);
1564 cxgb4_unregister_uld(CXGB4_ULD_RDMA);
1565 c4iw_cm_term();
1566 debugfs_remove_recursive(c4iw_debugfs_root);
1567}
1568
1569module_init(c4iw_init_module);
1570module_exit(c4iw_exit_module);
1571