2 * IBM eServer eHCA Infiniband device driver for Linux on POWER
4 * Functions for EQs, NEQs and interrupts
6 * Authors: Heiko J Schick <schickhj@de.ibm.com>
7 * Khadija Souissi <souissi@de.ibm.com>
8 * Hoang-Nam Nguyen <hnguyen@de.ibm.com>
9 * Joachim Fenkes <fenkes@de.ibm.com>
11 * Copyright (c) 2005 IBM Corporation
13 * All rights reserved.
15 * This source code is distributed under a dual license of GPL v2.0 and OpenIB
20 * Redistribution and use in source and binary forms, with or without
21 * modification, are permitted provided that the following conditions are met:
23 * Redistributions of source code must retain the above copyright notice, this
24 * list of conditions and the following disclaimer.
26 * Redistributions in binary form must reproduce the above copyright notice,
27 * this list of conditions and the following disclaimer in the documentation
28 * and/or other materials
29 * provided with the distribution.
31 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
32 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
33 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
34 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
35 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
36 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
37 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
38 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER
39 * IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
40 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
41 * POSSIBILITY OF SUCH DAMAGE.
44 #include "ehca_classes.h"
46 #include "ehca_iverbs.h"
47 #include "ehca_tools.h"
50 #include "ipz_pt_fn.h"
52 #define EQE_COMPLETION_EVENT EHCA_BMASK_IBM( 1, 1)
53 #define EQE_CQ_QP_NUMBER EHCA_BMASK_IBM( 8, 31)
54 #define EQE_EE_IDENTIFIER EHCA_BMASK_IBM( 2, 7)
55 #define EQE_CQ_NUMBER EHCA_BMASK_IBM( 8, 31)
56 #define EQE_QP_NUMBER EHCA_BMASK_IBM( 8, 31)
57 #define EQE_QP_TOKEN EHCA_BMASK_IBM(32, 63)
58 #define EQE_CQ_TOKEN EHCA_BMASK_IBM(32, 63)
60 #define NEQE_COMPLETION_EVENT EHCA_BMASK_IBM( 1, 1)
61 #define NEQE_EVENT_CODE EHCA_BMASK_IBM( 2, 7)
62 #define NEQE_PORT_NUMBER EHCA_BMASK_IBM( 8, 15)
63 #define NEQE_PORT_AVAILABILITY EHCA_BMASK_IBM(16, 16)
64 #define NEQE_DISRUPTIVE EHCA_BMASK_IBM(16, 16)
65 #define NEQE_SPECIFIC_EVENT EHCA_BMASK_IBM(16, 23)
67 #define ERROR_DATA_LENGTH EHCA_BMASK_IBM(52, 63)
68 #define ERROR_DATA_TYPE EHCA_BMASK_IBM( 0, 7)
70 static void queue_comp_task(struct ehca_cq
*__cq
);
72 static struct ehca_comp_pool
*pool
;
74 static inline void comp_event_callback(struct ehca_cq
*cq
)
76 if (!cq
->ib_cq
.comp_handler
)
79 spin_lock(&cq
->cb_lock
);
80 cq
->ib_cq
.comp_handler(&cq
->ib_cq
, cq
->ib_cq
.cq_context
);
81 spin_unlock(&cq
->cb_lock
);
86 static void print_error_data(struct ehca_shca
*shca
, void *data
,
87 u64
*rblock
, int length
)
89 u64 type
= EHCA_BMASK_GET(ERROR_DATA_TYPE
, rblock
[2]);
90 u64 resource
= rblock
[1];
93 case 0x1: /* Queue Pair */
95 struct ehca_qp
*qp
= (struct ehca_qp
*)data
;
97 /* only print error data if AER is set */
101 ehca_err(&shca
->ib_device
,
102 "QP 0x%x (resource=%lx) has errors.",
103 qp
->ib_qp
.qp_num
, resource
);
106 case 0x4: /* Completion Queue */
108 struct ehca_cq
*cq
= (struct ehca_cq
*)data
;
110 ehca_err(&shca
->ib_device
,
111 "CQ 0x%x (resource=%lx) has errors.",
112 cq
->cq_number
, resource
);
116 ehca_err(&shca
->ib_device
,
117 "Unknown error type: %lx on %s.",
118 type
, shca
->ib_device
.name
);
122 ehca_err(&shca
->ib_device
, "Error data is available: %lx.", resource
);
123 ehca_err(&shca
->ib_device
, "EHCA ----- error data begin "
124 "---------------------------------------------------");
125 ehca_dmp(rblock
, length
, "resource=%lx", resource
);
126 ehca_err(&shca
->ib_device
, "EHCA ----- error data end "
127 "----------------------------------------------------");
132 int ehca_error_data(struct ehca_shca
*shca
, void *data
,
138 unsigned long block_count
;
140 rblock
= ehca_alloc_fw_ctrlblock(GFP_ATOMIC
);
142 ehca_err(&shca
->ib_device
, "Cannot allocate rblock memory.");
147 /* rblock must be 4K aligned and should be 4K large */
148 ret
= hipz_h_error_data(shca
->ipz_hca_handle
,
153 if (ret
== H_R_STATE
)
154 ehca_err(&shca
->ib_device
,
155 "No error data is available: %lx.", resource
);
156 else if (ret
== H_SUCCESS
) {
159 length
= EHCA_BMASK_GET(ERROR_DATA_LENGTH
, rblock
[0]);
161 if (length
> EHCA_PAGESIZE
)
162 length
= EHCA_PAGESIZE
;
164 print_error_data(shca
, data
, rblock
, length
);
166 ehca_err(&shca
->ib_device
,
167 "Error data could not be fetched: %lx", resource
);
169 ehca_free_fw_ctrlblock(rblock
);
176 static void dispatch_qp_event(struct ehca_shca
*shca
, struct ehca_qp
*qp
,
177 enum ib_event_type event_type
)
179 struct ib_event event
;
181 event
.device
= &shca
->ib_device
;
182 event
.event
= event_type
;
184 if (qp
->ext_type
== EQPT_SRQ
) {
185 if (!qp
->ib_srq
.event_handler
)
188 event
.element
.srq
= &qp
->ib_srq
;
189 qp
->ib_srq
.event_handler(&event
, qp
->ib_srq
.srq_context
);
191 if (!qp
->ib_qp
.event_handler
)
194 event
.element
.qp
= &qp
->ib_qp
;
195 qp
->ib_qp
.event_handler(&event
, qp
->ib_qp
.qp_context
);
199 static void qp_event_callback(struct ehca_shca
*shca
, u64 eqe
,
200 enum ib_event_type event_type
, int fatal
)
203 u32 token
= EHCA_BMASK_GET(EQE_QP_TOKEN
, eqe
);
205 read_lock(&ehca_qp_idr_lock
);
206 qp
= idr_find(&ehca_qp_idr
, token
);
207 read_unlock(&ehca_qp_idr_lock
);
213 ehca_error_data(shca
, qp
, qp
->ipz_qp_handle
.handle
);
215 dispatch_qp_event(shca
, qp
, fatal
&& qp
->ext_type
== EQPT_SRQ
?
216 IB_EVENT_SRQ_ERR
: event_type
);
219 * eHCA only processes one WQE at a time for SRQ base QPs,
220 * so the last WQE has been processed as soon as the QP enters
223 if (fatal
&& qp
->ext_type
== EQPT_SRQBASE
)
224 dispatch_qp_event(shca
, qp
, IB_EVENT_QP_LAST_WQE_REACHED
);
229 static void cq_event_callback(struct ehca_shca
*shca
,
233 u32 token
= EHCA_BMASK_GET(EQE_CQ_TOKEN
, eqe
);
235 read_lock(&ehca_cq_idr_lock
);
236 cq
= idr_find(&ehca_cq_idr
, token
);
238 atomic_inc(&cq
->nr_events
);
239 read_unlock(&ehca_cq_idr_lock
);
244 ehca_error_data(shca
, cq
, cq
->ipz_cq_handle
.handle
);
246 if (atomic_dec_and_test(&cq
->nr_events
))
247 wake_up(&cq
->wait_completion
);
252 static void parse_identifier(struct ehca_shca
*shca
, u64 eqe
)
254 u8 identifier
= EHCA_BMASK_GET(EQE_EE_IDENTIFIER
, eqe
);
256 switch (identifier
) {
257 case 0x02: /* path migrated */
258 qp_event_callback(shca
, eqe
, IB_EVENT_PATH_MIG
, 0);
260 case 0x03: /* communication established */
261 qp_event_callback(shca
, eqe
, IB_EVENT_COMM_EST
, 0);
263 case 0x04: /* send queue drained */
264 qp_event_callback(shca
, eqe
, IB_EVENT_SQ_DRAINED
, 0);
266 case 0x05: /* QP error */
267 case 0x06: /* QP error */
268 qp_event_callback(shca
, eqe
, IB_EVENT_QP_FATAL
, 1);
270 case 0x07: /* CQ error */
271 case 0x08: /* CQ error */
272 cq_event_callback(shca
, eqe
);
274 case 0x09: /* MRMWPTE error */
275 ehca_err(&shca
->ib_device
, "MRMWPTE error.");
277 case 0x0A: /* port event */
278 ehca_err(&shca
->ib_device
, "Port event.");
280 case 0x0B: /* MR access error */
281 ehca_err(&shca
->ib_device
, "MR access error.");
283 case 0x0C: /* EQ error */
284 ehca_err(&shca
->ib_device
, "EQ error.");
286 case 0x0D: /* P/Q_Key mismatch */
287 ehca_err(&shca
->ib_device
, "P/Q_Key mismatch.");
289 case 0x10: /* sampling complete */
290 ehca_err(&shca
->ib_device
, "Sampling complete.");
292 case 0x11: /* unaffiliated access error */
293 ehca_err(&shca
->ib_device
, "Unaffiliated access error.");
295 case 0x12: /* path migrating */
296 ehca_err(&shca
->ib_device
, "Path migrating.");
298 case 0x13: /* interface trace stopped */
299 ehca_err(&shca
->ib_device
, "Interface trace stopped.");
301 case 0x14: /* first error capture info available */
302 ehca_info(&shca
->ib_device
, "First error capture available");
304 case 0x15: /* SRQ limit reached */
305 qp_event_callback(shca
, eqe
, IB_EVENT_SRQ_LIMIT_REACHED
, 0);
308 ehca_err(&shca
->ib_device
, "Unknown identifier: %x on %s.",
309 identifier
, shca
->ib_device
.name
);
316 static void dispatch_port_event(struct ehca_shca
*shca
, int port_num
,
317 enum ib_event_type type
, const char *msg
)
319 struct ib_event event
;
321 ehca_info(&shca
->ib_device
, "port %d %s.", port_num
, msg
);
322 event
.device
= &shca
->ib_device
;
324 event
.element
.port_num
= port_num
;
325 ib_dispatch_event(&event
);
328 static void notify_port_conf_change(struct ehca_shca
*shca
, int port_num
)
330 struct ehca_sma_attr new_attr
;
331 struct ehca_sma_attr
*old_attr
= &shca
->sport
[port_num
- 1].saved_attr
;
333 ehca_query_sma_attr(shca
, port_num
, &new_attr
);
335 if (new_attr
.sm_sl
!= old_attr
->sm_sl
||
336 new_attr
.sm_lid
!= old_attr
->sm_lid
)
337 dispatch_port_event(shca
, port_num
, IB_EVENT_SM_CHANGE
,
340 if (new_attr
.lid
!= old_attr
->lid
||
341 new_attr
.lmc
!= old_attr
->lmc
)
342 dispatch_port_event(shca
, port_num
, IB_EVENT_LID_CHANGE
,
345 if (new_attr
.pkey_tbl_len
!= old_attr
->pkey_tbl_len
||
346 memcmp(new_attr
.pkeys
, old_attr
->pkeys
,
347 sizeof(u16
) * new_attr
.pkey_tbl_len
))
348 dispatch_port_event(shca
, port_num
, IB_EVENT_PKEY_CHANGE
,
351 *old_attr
= new_attr
;
354 static void parse_ec(struct ehca_shca
*shca
, u64 eqe
)
356 u8 ec
= EHCA_BMASK_GET(NEQE_EVENT_CODE
, eqe
);
357 u8 port
= EHCA_BMASK_GET(NEQE_PORT_NUMBER
, eqe
);
359 struct ehca_sport
*sport
= &shca
->sport
[port
- 1];
363 case 0x30: /* port availability change */
364 if (EHCA_BMASK_GET(NEQE_PORT_AVAILABILITY
, eqe
)) {
366 /* replay modify_qp for sqps */
367 spin_lock_irqsave(&sport
->mod_sqp_lock
, flags
);
368 suppress_event
= !sport
->ibqp_sqp
[IB_QPT_GSI
];
369 if (sport
->ibqp_sqp
[IB_QPT_SMI
])
370 ehca_recover_sqp(sport
->ibqp_sqp
[IB_QPT_SMI
]);
372 ehca_recover_sqp(sport
->ibqp_sqp
[IB_QPT_GSI
]);
373 spin_unlock_irqrestore(&sport
->mod_sqp_lock
, flags
);
375 /* AQP1 was destroyed, ignore this event */
379 sport
->port_state
= IB_PORT_ACTIVE
;
380 dispatch_port_event(shca
, port
, IB_EVENT_PORT_ACTIVE
,
382 ehca_query_sma_attr(shca
, port
,
385 sport
->port_state
= IB_PORT_DOWN
;
386 dispatch_port_event(shca
, port
, IB_EVENT_PORT_ERR
,
391 /* port configuration change
392 * disruptive change is caused by
393 * LID, PKEY or SM change
395 if (EHCA_BMASK_GET(NEQE_DISRUPTIVE
, eqe
)) {
396 ehca_warn(&shca
->ib_device
, "disruptive port "
397 "%d configuration change", port
);
399 sport
->port_state
= IB_PORT_DOWN
;
400 dispatch_port_event(shca
, port
, IB_EVENT_PORT_ERR
,
403 sport
->port_state
= IB_PORT_ACTIVE
;
404 dispatch_port_event(shca
, port
, IB_EVENT_PORT_ACTIVE
,
406 ehca_query_sma_attr(shca
, port
,
409 notify_port_conf_change(shca
, port
);
411 case 0x32: /* adapter malfunction */
412 ehca_err(&shca
->ib_device
, "Adapter malfunction.");
414 case 0x33: /* trace stopped */
415 ehca_err(&shca
->ib_device
, "Traced stopped.");
417 case 0x34: /* util async event */
418 spec_event
= EHCA_BMASK_GET(NEQE_SPECIFIC_EVENT
, eqe
);
419 if (spec_event
== 0x80) /* client reregister required */
420 dispatch_port_event(shca
, port
,
421 IB_EVENT_CLIENT_REREGISTER
,
422 "client reregister req.");
424 ehca_warn(&shca
->ib_device
, "Unknown util async "
425 "event %x on port %x", spec_event
, port
);
428 ehca_err(&shca
->ib_device
, "Unknown event code: %x on %s.",
429 ec
, shca
->ib_device
.name
);
436 static inline void reset_eq_pending(struct ehca_cq
*cq
)
439 struct h_galpa gal
= cq
->galpas
.kernel
;
441 hipz_galpa_store_cq(gal
, cqx_ep
, 0x0);
442 CQx_EP
= hipz_galpa_load(gal
, CQTEMM_OFFSET(cqx_ep
));
447 irqreturn_t
ehca_interrupt_neq(int irq
, void *dev_id
)
449 struct ehca_shca
*shca
= (struct ehca_shca
*)dev_id
;
451 tasklet_hi_schedule(&shca
->neq
.interrupt_task
);
456 void ehca_tasklet_neq(unsigned long data
)
458 struct ehca_shca
*shca
= (struct ehca_shca
*)data
;
459 struct ehca_eqe
*eqe
;
462 eqe
= (struct ehca_eqe
*)ehca_poll_eq(shca
, &shca
->neq
);
465 if (!EHCA_BMASK_GET(NEQE_COMPLETION_EVENT
, eqe
->entry
))
466 parse_ec(shca
, eqe
->entry
);
468 eqe
= (struct ehca_eqe
*)ehca_poll_eq(shca
, &shca
->neq
);
471 ret
= hipz_h_reset_event(shca
->ipz_hca_handle
,
472 shca
->neq
.ipz_eq_handle
, 0xFFFFFFFFFFFFFFFFL
);
474 if (ret
!= H_SUCCESS
)
475 ehca_err(&shca
->ib_device
, "Can't clear notification events.");
480 irqreturn_t
ehca_interrupt_eq(int irq
, void *dev_id
)
482 struct ehca_shca
*shca
= (struct ehca_shca
*)dev_id
;
484 tasklet_hi_schedule(&shca
->eq
.interrupt_task
);
490 static inline void process_eqe(struct ehca_shca
*shca
, struct ehca_eqe
*eqe
)
496 eqe_value
= eqe
->entry
;
497 ehca_dbg(&shca
->ib_device
, "eqe_value=%lx", eqe_value
);
498 if (EHCA_BMASK_GET(EQE_COMPLETION_EVENT
, eqe_value
)) {
499 ehca_dbg(&shca
->ib_device
, "Got completion event");
500 token
= EHCA_BMASK_GET(EQE_CQ_TOKEN
, eqe_value
);
501 read_lock(&ehca_cq_idr_lock
);
502 cq
= idr_find(&ehca_cq_idr
, token
);
504 atomic_inc(&cq
->nr_events
);
505 read_unlock(&ehca_cq_idr_lock
);
507 ehca_err(&shca
->ib_device
,
508 "Invalid eqe for non-existing cq token=%x",
512 reset_eq_pending(cq
);
513 if (ehca_scaling_code
)
516 comp_event_callback(cq
);
517 if (atomic_dec_and_test(&cq
->nr_events
))
518 wake_up(&cq
->wait_completion
);
521 ehca_dbg(&shca
->ib_device
, "Got non completion event");
522 parse_identifier(shca
, eqe_value
);
526 void ehca_process_eq(struct ehca_shca
*shca
, int is_irq
)
528 struct ehca_eq
*eq
= &shca
->eq
;
529 struct ehca_eqe_cache_entry
*eqe_cache
= eq
->eqe_cache
;
535 spin_lock_irqsave(&eq
->irq_spinlock
, flags
);
537 const int max_query_cnt
= 100;
541 int_state
= hipz_h_query_int_state(
542 shca
->ipz_hca_handle
, eq
->ist
);
545 } while (int_state
&& query_cnt
< max_query_cnt
);
546 if (unlikely((query_cnt
== max_query_cnt
)))
547 ehca_dbg(&shca
->ib_device
, "int_state=%x query_cnt=%x",
548 int_state
, query_cnt
);
551 /* read out all eqes */
555 eqe_cache
[eqe_cnt
].eqe
=
556 (struct ehca_eqe
*)ehca_poll_eq(shca
, eq
);
557 if (!eqe_cache
[eqe_cnt
].eqe
)
559 eqe_value
= eqe_cache
[eqe_cnt
].eqe
->entry
;
560 if (EHCA_BMASK_GET(EQE_COMPLETION_EVENT
, eqe_value
)) {
561 token
= EHCA_BMASK_GET(EQE_CQ_TOKEN
, eqe_value
);
562 read_lock(&ehca_cq_idr_lock
);
563 eqe_cache
[eqe_cnt
].cq
= idr_find(&ehca_cq_idr
, token
);
564 if (eqe_cache
[eqe_cnt
].cq
)
565 atomic_inc(&eqe_cache
[eqe_cnt
].cq
->nr_events
);
566 read_unlock(&ehca_cq_idr_lock
);
567 if (!eqe_cache
[eqe_cnt
].cq
) {
568 ehca_err(&shca
->ib_device
,
569 "Invalid eqe for non-existing cq "
574 eqe_cache
[eqe_cnt
].cq
= NULL
;
576 } while (eqe_cnt
< EHCA_EQE_CACHE_SIZE
);
579 ehca_dbg(&shca
->ib_device
,
580 "No eqe found for irq event");
581 goto unlock_irq_spinlock
;
583 ehca_dbg(&shca
->ib_device
, "deadman found %x eqe", eqe_cnt
);
584 if (unlikely(eqe_cnt
== EHCA_EQE_CACHE_SIZE
))
585 ehca_dbg(&shca
->ib_device
, "too many eqes for one irq event");
586 /* enable irq for new packets */
587 for (i
= 0; i
< eqe_cnt
; i
++) {
588 if (eq
->eqe_cache
[i
].cq
)
589 reset_eq_pending(eq
->eqe_cache
[i
].cq
);
592 spin_lock(&eq
->spinlock
);
593 eq_empty
= (!ipz_eqit_eq_peek_valid(&shca
->eq
.ipz_queue
));
594 spin_unlock(&eq
->spinlock
);
595 /* call completion handler for cached eqes */
596 for (i
= 0; i
< eqe_cnt
; i
++)
597 if (eq
->eqe_cache
[i
].cq
) {
598 if (ehca_scaling_code
)
599 queue_comp_task(eq
->eqe_cache
[i
].cq
);
601 struct ehca_cq
*cq
= eq
->eqe_cache
[i
].cq
;
602 comp_event_callback(cq
);
603 if (atomic_dec_and_test(&cq
->nr_events
))
604 wake_up(&cq
->wait_completion
);
607 ehca_dbg(&shca
->ib_device
, "Got non completion event");
608 parse_identifier(shca
, eq
->eqe_cache
[i
].eqe
->entry
);
610 /* poll eq if not empty */
612 goto unlock_irq_spinlock
;
614 struct ehca_eqe
*eqe
;
615 eqe
= (struct ehca_eqe
*)ehca_poll_eq(shca
, &shca
->eq
);
618 process_eqe(shca
, eqe
);
622 spin_unlock_irqrestore(&eq
->irq_spinlock
, flags
);
625 void ehca_tasklet_eq(unsigned long data
)
627 ehca_process_eq((struct ehca_shca
*)data
, 1);
630 static inline int find_next_online_cpu(struct ehca_comp_pool
*pool
)
635 WARN_ON_ONCE(!in_interrupt());
636 if (ehca_debug_level
)
637 ehca_dmp(&cpu_online_map
, sizeof(cpumask_t
), "");
639 spin_lock_irqsave(&pool
->last_cpu_lock
, flags
);
640 cpu
= next_cpu(pool
->last_cpu
, cpu_online_map
);
642 cpu
= first_cpu(cpu_online_map
);
643 pool
->last_cpu
= cpu
;
644 spin_unlock_irqrestore(&pool
->last_cpu_lock
, flags
);
649 static void __queue_comp_task(struct ehca_cq
*__cq
,
650 struct ehca_cpu_comp_task
*cct
)
654 spin_lock_irqsave(&cct
->task_lock
, flags
);
655 spin_lock(&__cq
->task_lock
);
657 if (__cq
->nr_callbacks
== 0) {
658 __cq
->nr_callbacks
++;
659 list_add_tail(&__cq
->entry
, &cct
->cq_list
);
661 wake_up(&cct
->wait_queue
);
663 __cq
->nr_callbacks
++;
665 spin_unlock(&__cq
->task_lock
);
666 spin_unlock_irqrestore(&cct
->task_lock
, flags
);
669 static void queue_comp_task(struct ehca_cq
*__cq
)
672 struct ehca_cpu_comp_task
*cct
;
676 cpu_id
= find_next_online_cpu(pool
);
677 BUG_ON(!cpu_online(cpu_id
));
679 cct
= per_cpu_ptr(pool
->cpu_comp_tasks
, cpu_id
);
682 spin_lock_irqsave(&cct
->task_lock
, flags
);
683 cq_jobs
= cct
->cq_jobs
;
684 spin_unlock_irqrestore(&cct
->task_lock
, flags
);
686 cpu_id
= find_next_online_cpu(pool
);
687 cct
= per_cpu_ptr(pool
->cpu_comp_tasks
, cpu_id
);
691 __queue_comp_task(__cq
, cct
);
694 static void run_comp_task(struct ehca_cpu_comp_task
*cct
)
699 spin_lock_irqsave(&cct
->task_lock
, flags
);
701 while (!list_empty(&cct
->cq_list
)) {
702 cq
= list_entry(cct
->cq_list
.next
, struct ehca_cq
, entry
);
703 spin_unlock_irqrestore(&cct
->task_lock
, flags
);
705 comp_event_callback(cq
);
706 if (atomic_dec_and_test(&cq
->nr_events
))
707 wake_up(&cq
->wait_completion
);
709 spin_lock_irqsave(&cct
->task_lock
, flags
);
710 spin_lock(&cq
->task_lock
);
712 if (!cq
->nr_callbacks
) {
713 list_del_init(cct
->cq_list
.next
);
716 spin_unlock(&cq
->task_lock
);
719 spin_unlock_irqrestore(&cct
->task_lock
, flags
);
722 static int comp_task(void *__cct
)
724 struct ehca_cpu_comp_task
*cct
= __cct
;
726 DECLARE_WAITQUEUE(wait
, current
);
728 set_current_state(TASK_INTERRUPTIBLE
);
729 while (!kthread_should_stop()) {
730 add_wait_queue(&cct
->wait_queue
, &wait
);
732 spin_lock_irq(&cct
->task_lock
);
733 cql_empty
= list_empty(&cct
->cq_list
);
734 spin_unlock_irq(&cct
->task_lock
);
738 __set_current_state(TASK_RUNNING
);
740 remove_wait_queue(&cct
->wait_queue
, &wait
);
742 spin_lock_irq(&cct
->task_lock
);
743 cql_empty
= list_empty(&cct
->cq_list
);
744 spin_unlock_irq(&cct
->task_lock
);
746 run_comp_task(__cct
);
748 set_current_state(TASK_INTERRUPTIBLE
);
750 __set_current_state(TASK_RUNNING
);
755 static struct task_struct
*create_comp_task(struct ehca_comp_pool
*pool
,
758 struct ehca_cpu_comp_task
*cct
;
760 cct
= per_cpu_ptr(pool
->cpu_comp_tasks
, cpu
);
761 spin_lock_init(&cct
->task_lock
);
762 INIT_LIST_HEAD(&cct
->cq_list
);
763 init_waitqueue_head(&cct
->wait_queue
);
764 cct
->task
= kthread_create(comp_task
, cct
, "ehca_comp/%d", cpu
);
769 static void destroy_comp_task(struct ehca_comp_pool
*pool
,
772 struct ehca_cpu_comp_task
*cct
;
773 struct task_struct
*task
;
774 unsigned long flags_cct
;
776 cct
= per_cpu_ptr(pool
->cpu_comp_tasks
, cpu
);
778 spin_lock_irqsave(&cct
->task_lock
, flags_cct
);
784 spin_unlock_irqrestore(&cct
->task_lock
, flags_cct
);
790 static void __cpuinit
take_over_work(struct ehca_comp_pool
*pool
, int cpu
)
792 struct ehca_cpu_comp_task
*cct
= per_cpu_ptr(pool
->cpu_comp_tasks
, cpu
);
795 unsigned long flags_cct
;
797 spin_lock_irqsave(&cct
->task_lock
, flags_cct
);
799 list_splice_init(&cct
->cq_list
, &list
);
801 while (!list_empty(&list
)) {
802 cq
= list_entry(cct
->cq_list
.next
, struct ehca_cq
, entry
);
804 list_del(&cq
->entry
);
805 __queue_comp_task(cq
, per_cpu_ptr(pool
->cpu_comp_tasks
,
806 smp_processor_id()));
809 spin_unlock_irqrestore(&cct
->task_lock
, flags_cct
);
813 static int __cpuinit
comp_pool_callback(struct notifier_block
*nfb
,
814 unsigned long action
,
817 unsigned int cpu
= (unsigned long)hcpu
;
818 struct ehca_cpu_comp_task
*cct
;
822 case CPU_UP_PREPARE_FROZEN
:
823 ehca_gen_dbg("CPU: %x (CPU_PREPARE)", cpu
);
824 if (!create_comp_task(pool
, cpu
)) {
825 ehca_gen_err("Can't create comp_task for cpu: %x", cpu
);
829 case CPU_UP_CANCELED
:
830 case CPU_UP_CANCELED_FROZEN
:
831 ehca_gen_dbg("CPU: %x (CPU_CANCELED)", cpu
);
832 cct
= per_cpu_ptr(pool
->cpu_comp_tasks
, cpu
);
833 kthread_bind(cct
->task
, any_online_cpu(cpu_online_map
));
834 destroy_comp_task(pool
, cpu
);
837 case CPU_ONLINE_FROZEN
:
838 ehca_gen_dbg("CPU: %x (CPU_ONLINE)", cpu
);
839 cct
= per_cpu_ptr(pool
->cpu_comp_tasks
, cpu
);
840 kthread_bind(cct
->task
, cpu
);
841 wake_up_process(cct
->task
);
843 case CPU_DOWN_PREPARE
:
844 case CPU_DOWN_PREPARE_FROZEN
:
845 ehca_gen_dbg("CPU: %x (CPU_DOWN_PREPARE)", cpu
);
847 case CPU_DOWN_FAILED
:
848 case CPU_DOWN_FAILED_FROZEN
:
849 ehca_gen_dbg("CPU: %x (CPU_DOWN_FAILED)", cpu
);
852 case CPU_DEAD_FROZEN
:
853 ehca_gen_dbg("CPU: %x (CPU_DEAD)", cpu
);
854 destroy_comp_task(pool
, cpu
);
855 take_over_work(pool
, cpu
);
862 static struct notifier_block comp_pool_callback_nb __cpuinitdata
= {
863 .notifier_call
= comp_pool_callback
,
867 int ehca_create_comp_pool(void)
870 struct task_struct
*task
;
872 if (!ehca_scaling_code
)
875 pool
= kzalloc(sizeof(struct ehca_comp_pool
), GFP_KERNEL
);
879 spin_lock_init(&pool
->last_cpu_lock
);
880 pool
->last_cpu
= any_online_cpu(cpu_online_map
);
882 pool
->cpu_comp_tasks
= alloc_percpu(struct ehca_cpu_comp_task
);
883 if (pool
->cpu_comp_tasks
== NULL
) {
888 for_each_online_cpu(cpu
) {
889 task
= create_comp_task(pool
, cpu
);
891 kthread_bind(task
, cpu
);
892 wake_up_process(task
);
896 register_hotcpu_notifier(&comp_pool_callback_nb
);
898 printk(KERN_INFO
"eHCA scaling code enabled\n");
903 void ehca_destroy_comp_pool(void)
907 if (!ehca_scaling_code
)
910 unregister_hotcpu_notifier(&comp_pool_callback_nb
);
912 for (i
= 0; i
< NR_CPUS
; i
++) {
914 destroy_comp_task(pool
, i
);
916 free_percpu(pool
->cpu_comp_tasks
);