aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2009-04-24 10:48:24 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2009-04-24 10:48:24 -0400
commit97c68d00dbb2decda4b3dce79ce55df04246a834 (patch)
treef7c2f3185e4dfb073f6d268b4aa0a4d9a0015ece
parent596a5c4e43b08e90c0cd788861f859bf7ed6eddc (diff)
parentf2d1f0ae7851be5ebd9613a80dac139270938809 (diff)
Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block
* 'for-linus' of git://git.kernel.dk/linux-2.6-block: cfq-iosched: cache prio_tree root in cfqq->p_root cfq-iosched: fix bug with aliased request and cooperation detection cfq-iosched: clear ->prio_trees[] on cfqd alloc block: fix intermittent dm timeout based oops umem: fix request_queue lock warning block: simplify I/O stat accounting pktcdvd.h should include mempool.h cfq-iosched: use the default seek distance when there aren't enough seek samples cfq-iosched: make seek_mean converge more quickly block: make blk_abort_queue() ignore non-request based devices block: include empty disks in /proc/diskstats bio: use bio_kmalloc() in copy/map functions bio: fix bio_kmalloc() block: fix queue bounce limit setting block: fix SG_IO vector request data length handling scatterlist: make sure sg_miter_next() doesn't return 0 sized mappings
-rw-r--r--block/blk-core.c6
-rw-r--r--block/blk-merge.c5
-rw-r--r--block/blk-settings.c20
-rw-r--r--block/blk-sysfs.c4
-rw-r--r--block/blk-timeout.c13
-rw-r--r--block/blk.h7
-rw-r--r--block/cfq-iosched.c66
-rw-r--r--block/genhd.c12
-rw-r--r--block/scsi_ioctl.c13
-rw-r--r--drivers/block/umem.c1
-rw-r--r--fs/bio.c124
-rw-r--r--include/linux/bio.h1
-rw-r--r--include/linux/blkdev.h3
-rw-r--r--include/linux/genhd.h1
-rw-r--r--include/linux/pktcdvd.h1
-rw-r--r--lib/scatterlist.c9
16 files changed, 166 insertions, 120 deletions
diff --git a/block/blk-core.c b/block/blk-core.c
index 07ab75403e1a..2998fe3a2377 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -643,7 +643,7 @@ static inline void blk_free_request(struct request_queue *q, struct request *rq)
643} 643}
644 644
645static struct request * 645static struct request *
646blk_alloc_request(struct request_queue *q, int rw, int priv, gfp_t gfp_mask) 646blk_alloc_request(struct request_queue *q, int flags, int priv, gfp_t gfp_mask)
647{ 647{
648 struct request *rq = mempool_alloc(q->rq.rq_pool, gfp_mask); 648 struct request *rq = mempool_alloc(q->rq.rq_pool, gfp_mask);
649 649
@@ -652,7 +652,7 @@ blk_alloc_request(struct request_queue *q, int rw, int priv, gfp_t gfp_mask)
652 652
653 blk_rq_init(q, rq); 653 blk_rq_init(q, rq);
654 654
655 rq->cmd_flags = rw | REQ_ALLOCED; 655 rq->cmd_flags = flags | REQ_ALLOCED;
656 656
657 if (priv) { 657 if (priv) {
658 if (unlikely(elv_set_request(q, rq, gfp_mask))) { 658 if (unlikely(elv_set_request(q, rq, gfp_mask))) {
@@ -792,6 +792,8 @@ static struct request *get_request(struct request_queue *q, int rw_flags,
792 if (priv) 792 if (priv)
793 rl->elvpriv++; 793 rl->elvpriv++;
794 794
795 if (blk_queue_io_stat(q))
796 rw_flags |= REQ_IO_STAT;
795 spin_unlock_irq(q->queue_lock); 797 spin_unlock_irq(q->queue_lock);
796 798
797 rq = blk_alloc_request(q, rw_flags, priv, gfp_mask); 799 rq = blk_alloc_request(q, rw_flags, priv, gfp_mask);
diff --git a/block/blk-merge.c b/block/blk-merge.c
index 63760ca3da0f..23d2a6fe34a3 100644
--- a/block/blk-merge.c
+++ b/block/blk-merge.c
@@ -402,7 +402,10 @@ static int attempt_merge(struct request_queue *q, struct request *req,
402 402
403 elv_merge_requests(q, req, next); 403 elv_merge_requests(q, req, next);
404 404
405 blk_account_io_merge(req); 405 /*
406 * 'next' is going away, so update stats accordingly
407 */
408 blk_account_io_merge(next);
406 409
407 req->ioprio = ioprio_best(req->ioprio, next->ioprio); 410 req->ioprio = ioprio_best(req->ioprio, next->ioprio);
408 if (blk_rq_cpu_valid(next)) 411 if (blk_rq_cpu_valid(next))
diff --git a/block/blk-settings.c b/block/blk-settings.c
index 69c42adde52b..57af728d94bb 100644
--- a/block/blk-settings.c
+++ b/block/blk-settings.c
@@ -156,26 +156,28 @@ EXPORT_SYMBOL(blk_queue_make_request);
156 156
157/** 157/**
158 * blk_queue_bounce_limit - set bounce buffer limit for queue 158 * blk_queue_bounce_limit - set bounce buffer limit for queue
159 * @q: the request queue for the device 159 * @q: the request queue for the device
160 * @dma_addr: bus address limit 160 * @dma_mask: the maximum address the device can handle
161 * 161 *
162 * Description: 162 * Description:
163 * Different hardware can have different requirements as to what pages 163 * Different hardware can have different requirements as to what pages
164 * it can do I/O directly to. A low level driver can call 164 * it can do I/O directly to. A low level driver can call
165 * blk_queue_bounce_limit to have lower memory pages allocated as bounce 165 * blk_queue_bounce_limit to have lower memory pages allocated as bounce
166 * buffers for doing I/O to pages residing above @dma_addr. 166 * buffers for doing I/O to pages residing above @dma_mask.
167 **/ 167 **/
168void blk_queue_bounce_limit(struct request_queue *q, u64 dma_addr) 168void blk_queue_bounce_limit(struct request_queue *q, u64 dma_mask)
169{ 169{
170 unsigned long b_pfn = dma_addr >> PAGE_SHIFT; 170 unsigned long b_pfn = dma_mask >> PAGE_SHIFT;
171 int dma = 0; 171 int dma = 0;
172 172
173 q->bounce_gfp = GFP_NOIO; 173 q->bounce_gfp = GFP_NOIO;
174#if BITS_PER_LONG == 64 174#if BITS_PER_LONG == 64
175 /* Assume anything <= 4GB can be handled by IOMMU. 175 /*
176 Actually some IOMMUs can handle everything, but I don't 176 * Assume anything <= 4GB can be handled by IOMMU. Actually
177 know of a way to test this here. */ 177 * some IOMMUs can handle everything, but I don't know of a
178 if (b_pfn < (min_t(u64, 0x100000000UL, BLK_BOUNCE_HIGH) >> PAGE_SHIFT)) 178 * way to test this here.
179 */
180 if (b_pfn < (min_t(u64, 0xffffffffUL, BLK_BOUNCE_HIGH) >> PAGE_SHIFT))
179 dma = 1; 181 dma = 1;
180 q->bounce_pfn = max_low_pfn; 182 q->bounce_pfn = max_low_pfn;
181#else 183#else
diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
index cac4e9febe6a..3ff9bba3379a 100644
--- a/block/blk-sysfs.c
+++ b/block/blk-sysfs.c
@@ -209,14 +209,10 @@ static ssize_t queue_iostats_store(struct request_queue *q, const char *page,
209 ssize_t ret = queue_var_store(&stats, page, count); 209 ssize_t ret = queue_var_store(&stats, page, count);
210 210
211 spin_lock_irq(q->queue_lock); 211 spin_lock_irq(q->queue_lock);
212 elv_quiesce_start(q);
213
214 if (stats) 212 if (stats)
215 queue_flag_set(QUEUE_FLAG_IO_STAT, q); 213 queue_flag_set(QUEUE_FLAG_IO_STAT, q);
216 else 214 else
217 queue_flag_clear(QUEUE_FLAG_IO_STAT, q); 215 queue_flag_clear(QUEUE_FLAG_IO_STAT, q);
218
219 elv_quiesce_end(q);
220 spin_unlock_irq(q->queue_lock); 216 spin_unlock_irq(q->queue_lock);
221 217
222 return ret; 218 return ret;
diff --git a/block/blk-timeout.c b/block/blk-timeout.c
index bbbdc4b8ccf2..1ec0d503cacd 100644
--- a/block/blk-timeout.c
+++ b/block/blk-timeout.c
@@ -211,6 +211,12 @@ void blk_abort_queue(struct request_queue *q)
211 struct request *rq, *tmp; 211 struct request *rq, *tmp;
212 LIST_HEAD(list); 212 LIST_HEAD(list);
213 213
214 /*
215 * Not a request based block device, nothing to abort
216 */
217 if (!q->request_fn)
218 return;
219
214 spin_lock_irqsave(q->queue_lock, flags); 220 spin_lock_irqsave(q->queue_lock, flags);
215 221
216 elv_abort_queue(q); 222 elv_abort_queue(q);
@@ -224,6 +230,13 @@ void blk_abort_queue(struct request_queue *q)
224 list_for_each_entry_safe(rq, tmp, &list, timeout_list) 230 list_for_each_entry_safe(rq, tmp, &list, timeout_list)
225 blk_abort_request(rq); 231 blk_abort_request(rq);
226 232
233 /*
234 * Occasionally, blk_abort_request() will return without
235 * deleting the element from the list. Make sure we add those back
236 * instead of leaving them on the local stack list.
237 */
238 list_splice(&list, &q->timeout_list);
239
227 spin_unlock_irqrestore(q->queue_lock, flags); 240 spin_unlock_irqrestore(q->queue_lock, flags);
228 241
229} 242}
diff --git a/block/blk.h b/block/blk.h
index 5dfc41267a08..79c85f7c9ff5 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -114,12 +114,7 @@ static inline int blk_cpu_to_group(int cpu)
114 114
115static inline int blk_do_io_stat(struct request *rq) 115static inline int blk_do_io_stat(struct request *rq)
116{ 116{
117 struct gendisk *disk = rq->rq_disk; 117 return rq->rq_disk && blk_rq_io_stat(rq);
118
119 if (!disk || !disk->queue)
120 return 0;
121
122 return blk_queue_io_stat(disk->queue) && (rq->cmd_flags & REQ_ELVPRIV);
123} 118}
124 119
125#endif 120#endif
diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index 0d3b70de3d80..a55a9bd75bd1 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -154,6 +154,8 @@ struct cfq_queue {
154 unsigned long rb_key; 154 unsigned long rb_key;
155 /* prio tree member */ 155 /* prio tree member */
156 struct rb_node p_node; 156 struct rb_node p_node;
157 /* prio tree root we belong to, if any */
158 struct rb_root *p_root;
157 /* sorted list of pending requests */ 159 /* sorted list of pending requests */
158 struct rb_root sort_list; 160 struct rb_root sort_list;
159 /* if fifo isn't expired, next request to serve */ 161 /* if fifo isn't expired, next request to serve */
@@ -558,10 +560,10 @@ static void cfq_service_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq,
558} 560}
559 561
560static struct cfq_queue * 562static struct cfq_queue *
561cfq_prio_tree_lookup(struct cfq_data *cfqd, int ioprio, sector_t sector, 563cfq_prio_tree_lookup(struct cfq_data *cfqd, struct rb_root *root,
562 struct rb_node **ret_parent, struct rb_node ***rb_link) 564 sector_t sector, struct rb_node **ret_parent,
565 struct rb_node ***rb_link)
563{ 566{
564 struct rb_root *root = &cfqd->prio_trees[ioprio];
565 struct rb_node **p, *parent; 567 struct rb_node **p, *parent;
566 struct cfq_queue *cfqq = NULL; 568 struct cfq_queue *cfqq = NULL;
567 569
@@ -584,34 +586,38 @@ cfq_prio_tree_lookup(struct cfq_data *cfqd, int ioprio, sector_t sector,
584 else 586 else
585 break; 587 break;
586 p = n; 588 p = n;
589 cfqq = NULL;
587 } 590 }
588 591
589 *ret_parent = parent; 592 *ret_parent = parent;
590 if (rb_link) 593 if (rb_link)
591 *rb_link = p; 594 *rb_link = p;
592 return NULL; 595 return cfqq;
593} 596}
594 597
595static void cfq_prio_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq) 598static void cfq_prio_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq)
596{ 599{
597 struct rb_root *root = &cfqd->prio_trees[cfqq->ioprio];
598 struct rb_node **p, *parent; 600 struct rb_node **p, *parent;
599 struct cfq_queue *__cfqq; 601 struct cfq_queue *__cfqq;
600 602
601 if (!RB_EMPTY_NODE(&cfqq->p_node)) 603 if (cfqq->p_root) {
602 rb_erase_init(&cfqq->p_node, root); 604 rb_erase(&cfqq->p_node, cfqq->p_root);
605 cfqq->p_root = NULL;
606 }
603 607
604 if (cfq_class_idle(cfqq)) 608 if (cfq_class_idle(cfqq))
605 return; 609 return;
606 if (!cfqq->next_rq) 610 if (!cfqq->next_rq)
607 return; 611 return;
608 612
609 __cfqq = cfq_prio_tree_lookup(cfqd, cfqq->ioprio, cfqq->next_rq->sector, 613 cfqq->p_root = &cfqd->prio_trees[cfqq->org_ioprio];
614 __cfqq = cfq_prio_tree_lookup(cfqd, cfqq->p_root, cfqq->next_rq->sector,
610 &parent, &p); 615 &parent, &p);
611 BUG_ON(__cfqq); 616 if (!__cfqq) {
612 617 rb_link_node(&cfqq->p_node, parent, p);
613 rb_link_node(&cfqq->p_node, parent, p); 618 rb_insert_color(&cfqq->p_node, cfqq->p_root);
614 rb_insert_color(&cfqq->p_node, root); 619 } else
620 cfqq->p_root = NULL;
615} 621}
616 622
617/* 623/*
@@ -656,8 +662,10 @@ static void cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
656 662
657 if (!RB_EMPTY_NODE(&cfqq->rb_node)) 663 if (!RB_EMPTY_NODE(&cfqq->rb_node))
658 cfq_rb_erase(&cfqq->rb_node, &cfqd->service_tree); 664 cfq_rb_erase(&cfqq->rb_node, &cfqd->service_tree);
659 if (!RB_EMPTY_NODE(&cfqq->p_node)) 665 if (cfqq->p_root) {
660 rb_erase_init(&cfqq->p_node, &cfqd->prio_trees[cfqq->ioprio]); 666 rb_erase(&cfqq->p_node, cfqq->p_root);
667 cfqq->p_root = NULL;
668 }
661 669
662 BUG_ON(!cfqd->busy_queues); 670 BUG_ON(!cfqd->busy_queues);
663 cfqd->busy_queues--; 671 cfqd->busy_queues--;
@@ -947,20 +955,24 @@ static inline sector_t cfq_dist_from_last(struct cfq_data *cfqd,
947 return cfqd->last_position - rq->sector; 955 return cfqd->last_position - rq->sector;
948} 956}
949 957
958#define CIC_SEEK_THR 8 * 1024
959#define CIC_SEEKY(cic) ((cic)->seek_mean > CIC_SEEK_THR)
960
950static inline int cfq_rq_close(struct cfq_data *cfqd, struct request *rq) 961static inline int cfq_rq_close(struct cfq_data *cfqd, struct request *rq)
951{ 962{
952 struct cfq_io_context *cic = cfqd->active_cic; 963 struct cfq_io_context *cic = cfqd->active_cic;
964 sector_t sdist = cic->seek_mean;
953 965
954 if (!sample_valid(cic->seek_samples)) 966 if (!sample_valid(cic->seek_samples))
955 return 0; 967 sdist = CIC_SEEK_THR;
956 968
957 return cfq_dist_from_last(cfqd, rq) <= cic->seek_mean; 969 return cfq_dist_from_last(cfqd, rq) <= sdist;
958} 970}
959 971
960static struct cfq_queue *cfqq_close(struct cfq_data *cfqd, 972static struct cfq_queue *cfqq_close(struct cfq_data *cfqd,
961 struct cfq_queue *cur_cfqq) 973 struct cfq_queue *cur_cfqq)
962{ 974{
963 struct rb_root *root = &cfqd->prio_trees[cur_cfqq->ioprio]; 975 struct rb_root *root = &cfqd->prio_trees[cur_cfqq->org_ioprio];
964 struct rb_node *parent, *node; 976 struct rb_node *parent, *node;
965 struct cfq_queue *__cfqq; 977 struct cfq_queue *__cfqq;
966 sector_t sector = cfqd->last_position; 978 sector_t sector = cfqd->last_position;
@@ -972,8 +984,7 @@ static struct cfq_queue *cfqq_close(struct cfq_data *cfqd,
972 * First, if we find a request starting at the end of the last 984 * First, if we find a request starting at the end of the last
973 * request, choose it. 985 * request, choose it.
974 */ 986 */
975 __cfqq = cfq_prio_tree_lookup(cfqd, cur_cfqq->ioprio, 987 __cfqq = cfq_prio_tree_lookup(cfqd, root, sector, &parent, NULL);
976 sector, &parent, NULL);
977 if (__cfqq) 988 if (__cfqq)
978 return __cfqq; 989 return __cfqq;
979 990
@@ -1039,9 +1050,6 @@ static struct cfq_queue *cfq_close_cooperator(struct cfq_data *cfqd,
1039 return cfqq; 1050 return cfqq;
1040} 1051}
1041 1052
1042
1043#define CIC_SEEKY(cic) ((cic)->seek_mean > (8 * 1024))
1044
1045static void cfq_arm_slice_timer(struct cfq_data *cfqd) 1053static void cfq_arm_slice_timer(struct cfq_data *cfqd)
1046{ 1054{
1047 struct cfq_queue *cfqq = cfqd->active_queue; 1055 struct cfq_queue *cfqq = cfqd->active_queue;
@@ -1908,7 +1916,9 @@ cfq_update_io_seektime(struct cfq_data *cfqd, struct cfq_io_context *cic,
1908 sector_t sdist; 1916 sector_t sdist;
1909 u64 total; 1917 u64 total;
1910 1918
1911 if (cic->last_request_pos < rq->sector) 1919 if (!cic->last_request_pos)
1920 sdist = 0;
1921 else if (cic->last_request_pos < rq->sector)
1912 sdist = rq->sector - cic->last_request_pos; 1922 sdist = rq->sector - cic->last_request_pos;
1913 else 1923 else
1914 sdist = cic->last_request_pos - rq->sector; 1924 sdist = cic->last_request_pos - rq->sector;
@@ -2443,12 +2453,22 @@ static void cfq_exit_queue(struct elevator_queue *e)
2443static void *cfq_init_queue(struct request_queue *q) 2453static void *cfq_init_queue(struct request_queue *q)
2444{ 2454{
2445 struct cfq_data *cfqd; 2455 struct cfq_data *cfqd;
2456 int i;
2446 2457
2447 cfqd = kmalloc_node(sizeof(*cfqd), GFP_KERNEL | __GFP_ZERO, q->node); 2458 cfqd = kmalloc_node(sizeof(*cfqd), GFP_KERNEL | __GFP_ZERO, q->node);
2448 if (!cfqd) 2459 if (!cfqd)
2449 return NULL; 2460 return NULL;
2450 2461
2451 cfqd->service_tree = CFQ_RB_ROOT; 2462 cfqd->service_tree = CFQ_RB_ROOT;
2463
2464 /*
2465 * Not strictly needed (since RB_ROOT just clears the node and we
2466 * zeroed cfqd on alloc), but better be safe in case someone decides
2467 * to add magic to the rb code
2468 */
2469 for (i = 0; i < CFQ_PRIO_LISTS; i++)
2470 cfqd->prio_trees[i] = RB_ROOT;
2471
2452 INIT_LIST_HEAD(&cfqd->cic_list); 2472 INIT_LIST_HEAD(&cfqd->cic_list);
2453 2473
2454 cfqd->queue = q; 2474 cfqd->queue = q;
diff --git a/block/genhd.c b/block/genhd.c
index a9ec910974c1..1a4916e01732 100644
--- a/block/genhd.c
+++ b/block/genhd.c
@@ -98,7 +98,7 @@ void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
98 98
99 if (flags & DISK_PITER_REVERSE) 99 if (flags & DISK_PITER_REVERSE)
100 piter->idx = ptbl->len - 1; 100 piter->idx = ptbl->len - 1;
101 else if (flags & DISK_PITER_INCL_PART0) 101 else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
102 piter->idx = 0; 102 piter->idx = 0;
103 else 103 else
104 piter->idx = 1; 104 piter->idx = 1;
@@ -134,7 +134,8 @@ struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
134 /* determine iteration parameters */ 134 /* determine iteration parameters */
135 if (piter->flags & DISK_PITER_REVERSE) { 135 if (piter->flags & DISK_PITER_REVERSE) {
136 inc = -1; 136 inc = -1;
137 if (piter->flags & DISK_PITER_INCL_PART0) 137 if (piter->flags & (DISK_PITER_INCL_PART0 |
138 DISK_PITER_INCL_EMPTY_PART0))
138 end = -1; 139 end = -1;
139 else 140 else
140 end = 0; 141 end = 0;
@@ -150,7 +151,10 @@ struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
150 part = rcu_dereference(ptbl->part[piter->idx]); 151 part = rcu_dereference(ptbl->part[piter->idx]);
151 if (!part) 152 if (!part)
152 continue; 153 continue;
153 if (!(piter->flags & DISK_PITER_INCL_EMPTY) && !part->nr_sects) 154 if (!part->nr_sects &&
155 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
156 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
157 piter->idx == 0))
154 continue; 158 continue;
155 159
156 get_device(part_to_dev(part)); 160 get_device(part_to_dev(part));
@@ -1011,7 +1015,7 @@ static int diskstats_show(struct seq_file *seqf, void *v)
1011 "\n\n"); 1015 "\n\n");
1012 */ 1016 */
1013 1017
1014 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_PART0); 1018 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
1015 while ((hd = disk_part_iter_next(&piter))) { 1019 while ((hd = disk_part_iter_next(&piter))) {
1016 cpu = part_stat_lock(); 1020 cpu = part_stat_lock();
1017 part_round_stats(cpu, hd); 1021 part_round_stats(cpu, hd);
diff --git a/block/scsi_ioctl.c b/block/scsi_ioctl.c
index 84b7f8709f41..82a0ca2f6729 100644
--- a/block/scsi_ioctl.c
+++ b/block/scsi_ioctl.c
@@ -290,6 +290,7 @@ static int sg_io(struct request_queue *q, struct gendisk *bd_disk,
290 290
291 if (hdr->iovec_count) { 291 if (hdr->iovec_count) {
292 const int size = sizeof(struct sg_iovec) * hdr->iovec_count; 292 const int size = sizeof(struct sg_iovec) * hdr->iovec_count;
293 size_t iov_data_len;
293 struct sg_iovec *iov; 294 struct sg_iovec *iov;
294 295
295 iov = kmalloc(size, GFP_KERNEL); 296 iov = kmalloc(size, GFP_KERNEL);
@@ -304,8 +305,18 @@ static int sg_io(struct request_queue *q, struct gendisk *bd_disk,
304 goto out; 305 goto out;
305 } 306 }
306 307
308 /* SG_IO howto says that the shorter of the two wins */
309 iov_data_len = iov_length((struct iovec *)iov,
310 hdr->iovec_count);
311 if (hdr->dxfer_len < iov_data_len) {
312 hdr->iovec_count = iov_shorten((struct iovec *)iov,
313 hdr->iovec_count,
314 hdr->dxfer_len);
315 iov_data_len = hdr->dxfer_len;
316 }
317
307 ret = blk_rq_map_user_iov(q, rq, NULL, iov, hdr->iovec_count, 318 ret = blk_rq_map_user_iov(q, rq, NULL, iov, hdr->iovec_count,
308 hdr->dxfer_len, GFP_KERNEL); 319 iov_data_len, GFP_KERNEL);
309 kfree(iov); 320 kfree(iov);
310 } else if (hdr->dxfer_len) 321 } else if (hdr->dxfer_len)
311 ret = blk_rq_map_user(q, rq, NULL, hdr->dxferp, hdr->dxfer_len, 322 ret = blk_rq_map_user(q, rq, NULL, hdr->dxferp, hdr->dxfer_len,
diff --git a/drivers/block/umem.c b/drivers/block/umem.c
index 9744d59a69f2..858c34dd032d 100644
--- a/drivers/block/umem.c
+++ b/drivers/block/umem.c
@@ -906,6 +906,7 @@ static int __devinit mm_pci_probe(struct pci_dev *dev,
906 goto failed_alloc; 906 goto failed_alloc;
907 907
908 blk_queue_make_request(card->queue, mm_make_request); 908 blk_queue_make_request(card->queue, mm_make_request);
909 card->queue->queue_lock = &card->lock;
909 card->queue->queuedata = card; 910 card->queue->queuedata = card;
910 card->queue->unplug_fn = mm_unplug_device; 911 card->queue->unplug_fn = mm_unplug_device;
911 912
diff --git a/fs/bio.c b/fs/bio.c
index cd42bb882f30..7bbc98f0eda1 100644
--- a/fs/bio.c
+++ b/fs/bio.c
@@ -175,14 +175,6 @@ struct bio_vec *bvec_alloc_bs(gfp_t gfp_mask, int nr, unsigned long *idx,
175 struct bio_vec *bvl; 175 struct bio_vec *bvl;
176 176
177 /* 177 /*
178 * If 'bs' is given, lookup the pool and do the mempool alloc.
179 * If not, this is a bio_kmalloc() allocation and just do a
180 * kzalloc() for the exact number of vecs right away.
181 */
182 if (!bs)
183 bvl = kmalloc(nr * sizeof(struct bio_vec), gfp_mask);
184
185 /*
186 * see comment near bvec_array define! 178 * see comment near bvec_array define!
187 */ 179 */
188 switch (nr) { 180 switch (nr) {
@@ -260,21 +252,6 @@ void bio_free(struct bio *bio, struct bio_set *bs)
260 mempool_free(p, bs->bio_pool); 252 mempool_free(p, bs->bio_pool);
261} 253}
262 254
263/*
264 * default destructor for a bio allocated with bio_alloc_bioset()
265 */
266static void bio_fs_destructor(struct bio *bio)
267{
268 bio_free(bio, fs_bio_set);
269}
270
271static void bio_kmalloc_destructor(struct bio *bio)
272{
273 if (bio_has_allocated_vec(bio))
274 kfree(bio->bi_io_vec);
275 kfree(bio);
276}
277
278void bio_init(struct bio *bio) 255void bio_init(struct bio *bio)
279{ 256{
280 memset(bio, 0, sizeof(*bio)); 257 memset(bio, 0, sizeof(*bio));
@@ -301,21 +278,15 @@ void bio_init(struct bio *bio)
301 **/ 278 **/
302struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs) 279struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs)
303{ 280{
281 unsigned long idx = BIO_POOL_NONE;
304 struct bio_vec *bvl = NULL; 282 struct bio_vec *bvl = NULL;
305 struct bio *bio = NULL; 283 struct bio *bio;
306 unsigned long idx = 0; 284 void *p;
307 void *p = NULL; 285
308 286 p = mempool_alloc(bs->bio_pool, gfp_mask);
309 if (bs) { 287 if (unlikely(!p))
310 p = mempool_alloc(bs->bio_pool, gfp_mask); 288 return NULL;
311 if (!p) 289 bio = p + bs->front_pad;
312 goto err;
313 bio = p + bs->front_pad;
314 } else {
315 bio = kmalloc(sizeof(*bio), gfp_mask);
316 if (!bio)
317 goto err;
318 }
319 290
320 bio_init(bio); 291 bio_init(bio);
321 292
@@ -332,22 +303,50 @@ struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs)
332 303
333 nr_iovecs = bvec_nr_vecs(idx); 304 nr_iovecs = bvec_nr_vecs(idx);
334 } 305 }
306out_set:
335 bio->bi_flags |= idx << BIO_POOL_OFFSET; 307 bio->bi_flags |= idx << BIO_POOL_OFFSET;
336 bio->bi_max_vecs = nr_iovecs; 308 bio->bi_max_vecs = nr_iovecs;
337out_set:
338 bio->bi_io_vec = bvl; 309 bio->bi_io_vec = bvl;
339
340 return bio; 310 return bio;
341 311
342err_free: 312err_free:
343 if (bs) 313 mempool_free(p, bs->bio_pool);
344 mempool_free(p, bs->bio_pool);
345 else
346 kfree(bio);
347err:
348 return NULL; 314 return NULL;
349} 315}
350 316
317static void bio_fs_destructor(struct bio *bio)
318{
319 bio_free(bio, fs_bio_set);
320}
321
322/**
323 * bio_alloc - allocate a new bio, memory pool backed
324 * @gfp_mask: allocation mask to use
325 * @nr_iovecs: number of iovecs
326 *
327 * Allocate a new bio with @nr_iovecs bvecs. If @gfp_mask
328 * contains __GFP_WAIT, the allocation is guaranteed to succeed.
329 *
330 * RETURNS:
331 * Pointer to new bio on success, NULL on failure.
332 */
333struct bio *bio_alloc(gfp_t gfp_mask, int nr_iovecs)
334{
335 struct bio *bio = bio_alloc_bioset(gfp_mask, nr_iovecs, fs_bio_set);
336
337 if (bio)
338 bio->bi_destructor = bio_fs_destructor;
339
340 return bio;
341}
342
343static void bio_kmalloc_destructor(struct bio *bio)
344{
345 if (bio_integrity(bio))
346 bio_integrity_free(bio);
347 kfree(bio);
348}
349
351/** 350/**
352 * bio_alloc - allocate a bio for I/O 351 * bio_alloc - allocate a bio for I/O
353 * @gfp_mask: the GFP_ mask given to the slab allocator 352 * @gfp_mask: the GFP_ mask given to the slab allocator
@@ -366,29 +365,20 @@ err:
366 * do so can cause livelocks under memory pressure. 365 * do so can cause livelocks under memory pressure.
367 * 366 *
368 **/ 367 **/
369struct bio *bio_alloc(gfp_t gfp_mask, int nr_iovecs)
370{
371 struct bio *bio = bio_alloc_bioset(gfp_mask, nr_iovecs, fs_bio_set);
372
373 if (bio)
374 bio->bi_destructor = bio_fs_destructor;
375
376 return bio;
377}
378
379/*
380 * Like bio_alloc(), but doesn't use a mempool backing. This means that
381 * it CAN fail, but while bio_alloc() can only be used for allocations
382 * that have a short (finite) life span, bio_kmalloc() should be used
383 * for more permanent bio allocations (like allocating some bio's for
384 * initalization or setup purposes).
385 */
386struct bio *bio_kmalloc(gfp_t gfp_mask, int nr_iovecs) 368struct bio *bio_kmalloc(gfp_t gfp_mask, int nr_iovecs)
387{ 369{
388 struct bio *bio = bio_alloc_bioset(gfp_mask, nr_iovecs, NULL); 370 struct bio *bio;
389 371
390 if (bio) 372 bio = kmalloc(sizeof(struct bio) + nr_iovecs * sizeof(struct bio_vec),
391 bio->bi_destructor = bio_kmalloc_destructor; 373 gfp_mask);
374 if (unlikely(!bio))
375 return NULL;
376
377 bio_init(bio);
378 bio->bi_flags |= BIO_POOL_NONE << BIO_POOL_OFFSET;
379 bio->bi_max_vecs = nr_iovecs;
380 bio->bi_io_vec = bio->bi_inline_vecs;
381 bio->bi_destructor = bio_kmalloc_destructor;
392 382
393 return bio; 383 return bio;
394} 384}
@@ -832,7 +822,7 @@ struct bio *bio_copy_user_iov(struct request_queue *q,
832 return ERR_PTR(-ENOMEM); 822 return ERR_PTR(-ENOMEM);
833 823
834 ret = -ENOMEM; 824 ret = -ENOMEM;
835 bio = bio_alloc(gfp_mask, nr_pages); 825 bio = bio_kmalloc(gfp_mask, nr_pages);
836 if (!bio) 826 if (!bio)
837 goto out_bmd; 827 goto out_bmd;
838 828
@@ -956,7 +946,7 @@ static struct bio *__bio_map_user_iov(struct request_queue *q,
956 if (!nr_pages) 946 if (!nr_pages)
957 return ERR_PTR(-EINVAL); 947 return ERR_PTR(-EINVAL);
958 948
959 bio = bio_alloc(gfp_mask, nr_pages); 949 bio = bio_kmalloc(gfp_mask, nr_pages);
960 if (!bio) 950 if (!bio)
961 return ERR_PTR(-ENOMEM); 951 return ERR_PTR(-ENOMEM);
962 952
@@ -1140,7 +1130,7 @@ static struct bio *__bio_map_kern(struct request_queue *q, void *data,
1140 int offset, i; 1130 int offset, i;
1141 struct bio *bio; 1131 struct bio *bio;
1142 1132
1143 bio = bio_alloc(gfp_mask, nr_pages); 1133 bio = bio_kmalloc(gfp_mask, nr_pages);
1144 if (!bio) 1134 if (!bio)
1145 return ERR_PTR(-ENOMEM); 1135 return ERR_PTR(-ENOMEM);
1146 1136
diff --git a/include/linux/bio.h b/include/linux/bio.h
index b89cf2d82898..7b214fd672a2 100644
--- a/include/linux/bio.h
+++ b/include/linux/bio.h
@@ -132,6 +132,7 @@ struct bio {
132 * top 4 bits of bio flags indicate the pool this bio came from 132 * top 4 bits of bio flags indicate the pool this bio came from
133 */ 133 */
134#define BIO_POOL_BITS (4) 134#define BIO_POOL_BITS (4)
135#define BIO_POOL_NONE ((1UL << BIO_POOL_BITS) - 1)
135#define BIO_POOL_OFFSET (BITS_PER_LONG - BIO_POOL_BITS) 136#define BIO_POOL_OFFSET (BITS_PER_LONG - BIO_POOL_BITS)
136#define BIO_POOL_MASK (1UL << BIO_POOL_OFFSET) 137#define BIO_POOL_MASK (1UL << BIO_POOL_OFFSET)
137#define BIO_POOL_IDX(bio) ((bio)->bi_flags >> BIO_POOL_OFFSET) 138#define BIO_POOL_IDX(bio) ((bio)->bi_flags >> BIO_POOL_OFFSET)
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index ba54c834a590..2755d5c6da22 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -118,6 +118,7 @@ enum rq_flag_bits {
118 __REQ_COPY_USER, /* contains copies of user pages */ 118 __REQ_COPY_USER, /* contains copies of user pages */
119 __REQ_INTEGRITY, /* integrity metadata has been remapped */ 119 __REQ_INTEGRITY, /* integrity metadata has been remapped */
120 __REQ_NOIDLE, /* Don't anticipate more IO after this one */ 120 __REQ_NOIDLE, /* Don't anticipate more IO after this one */
121 __REQ_IO_STAT, /* account I/O stat */
121 __REQ_NR_BITS, /* stops here */ 122 __REQ_NR_BITS, /* stops here */
122}; 123};
123 124
@@ -145,6 +146,7 @@ enum rq_flag_bits {
145#define REQ_COPY_USER (1 << __REQ_COPY_USER) 146#define REQ_COPY_USER (1 << __REQ_COPY_USER)
146#define REQ_INTEGRITY (1 << __REQ_INTEGRITY) 147#define REQ_INTEGRITY (1 << __REQ_INTEGRITY)
147#define REQ_NOIDLE (1 << __REQ_NOIDLE) 148#define REQ_NOIDLE (1 << __REQ_NOIDLE)
149#define REQ_IO_STAT (1 << __REQ_IO_STAT)
148 150
149#define BLK_MAX_CDB 16 151#define BLK_MAX_CDB 16
150 152
@@ -598,6 +600,7 @@ enum {
598 blk_failfast_transport(rq) || \ 600 blk_failfast_transport(rq) || \
599 blk_failfast_driver(rq)) 601 blk_failfast_driver(rq))
600#define blk_rq_started(rq) ((rq)->cmd_flags & REQ_STARTED) 602#define blk_rq_started(rq) ((rq)->cmd_flags & REQ_STARTED)
603#define blk_rq_io_stat(rq) ((rq)->cmd_flags & REQ_IO_STAT)
601 604
602#define blk_account_rq(rq) (blk_rq_started(rq) && (blk_fs_request(rq) || blk_discard_rq(rq))) 605#define blk_account_rq(rq) (blk_rq_started(rq) && (blk_fs_request(rq) || blk_discard_rq(rq)))
603 606
diff --git a/include/linux/genhd.h b/include/linux/genhd.h
index 634c53028fb8..a1a28caed23d 100644
--- a/include/linux/genhd.h
+++ b/include/linux/genhd.h
@@ -214,6 +214,7 @@ static inline void disk_put_part(struct hd_struct *part)
214#define DISK_PITER_REVERSE (1 << 0) /* iterate in the reverse direction */ 214#define DISK_PITER_REVERSE (1 << 0) /* iterate in the reverse direction */
215#define DISK_PITER_INCL_EMPTY (1 << 1) /* include 0-sized parts */ 215#define DISK_PITER_INCL_EMPTY (1 << 1) /* include 0-sized parts */
216#define DISK_PITER_INCL_PART0 (1 << 2) /* include partition 0 */ 216#define DISK_PITER_INCL_PART0 (1 << 2) /* include partition 0 */
217#define DISK_PITER_INCL_EMPTY_PART0 (1 << 3) /* include empty partition 0 */
217 218
218struct disk_part_iter { 219struct disk_part_iter {
219 struct gendisk *disk; 220 struct gendisk *disk;
diff --git a/include/linux/pktcdvd.h b/include/linux/pktcdvd.h
index 04b4d7330e6d..d745f5b6c7b0 100644
--- a/include/linux/pktcdvd.h
+++ b/include/linux/pktcdvd.h
@@ -113,6 +113,7 @@ struct pkt_ctrl_command {
113#include <linux/cdrom.h> 113#include <linux/cdrom.h>
114#include <linux/kobject.h> 114#include <linux/kobject.h>
115#include <linux/sysfs.h> 115#include <linux/sysfs.h>
116#include <linux/mempool.h>
116 117
117/* default bio write queue congestion marks */ 118/* default bio write queue congestion marks */
118#define PKT_WRITE_CONGESTION_ON 10000 119#define PKT_WRITE_CONGESTION_ON 10000
diff --git a/lib/scatterlist.c b/lib/scatterlist.c
index b7b449dafbe5..a295e404e908 100644
--- a/lib/scatterlist.c
+++ b/lib/scatterlist.c
@@ -347,9 +347,12 @@ bool sg_miter_next(struct sg_mapping_iter *miter)
347 sg_miter_stop(miter); 347 sg_miter_stop(miter);
348 348
349 /* get to the next sg if necessary. __offset is adjusted by stop */ 349 /* get to the next sg if necessary. __offset is adjusted by stop */
350 if (miter->__offset == miter->__sg->length && --miter->__nents) { 350 while (miter->__offset == miter->__sg->length) {
351 miter->__sg = sg_next(miter->__sg); 351 if (--miter->__nents) {
352 miter->__offset = 0; 352 miter->__sg = sg_next(miter->__sg);
353 miter->__offset = 0;
354 } else
355 return false;
353 } 356 }
354 357
355 /* map the next page */ 358 /* map the next page */