10 files changed, 194 insertions, 82 deletions

diff --git a/block/bio-integrity.c b/block/bio-integrity.c
index bc423f7b02da..f14b4abbebd8 100644
--- a/block/bio-integrity.c
+++ b/block/bio-integrity.c
@@ -520,7 +520,7 @@ void bio_integrity_endio(struct bio *bio, int error)
          */
         if (error) {
                 bio->bi_end_io = bip->bip_end_io;
-                bio_endio(bio, error);
+                bio_endio_nodec(bio, error);
 
                 return;
         }
diff --git a/block/blk-core.c b/block/blk-core.c
index c359d72e9d76..bf930f481d43 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -1252,7 +1252,6 @@ void blk_rq_set_block_pc(struct request *rq)
         rq->__sector = (sector_t) -1;
         rq->bio = rq->biotail = NULL;
         memset(rq->__cmd, 0, sizeof(rq->__cmd));
-        rq->cmd = rq->__cmd;
 }
 EXPORT_SYMBOL(blk_rq_set_block_pc);
 
diff --git a/block/blk-exec.c b/block/blk-exec.c
index f4d27b12c90b..9924725fa50d 100644
--- a/block/blk-exec.c
+++ b/block/blk-exec.c
@@ -56,6 +56,7 @@ void blk_execute_rq_nowait(struct request_queue *q, struct gendisk *bd_disk,
         bool is_pm_resume;
 
         WARN_ON(irqs_disabled());
+        WARN_ON(rq->cmd_type == REQ_TYPE_FS);
 
         rq->rq_disk = bd_disk;
         rq->end_io = done;
diff --git a/block/blk-merge.c b/block/blk-merge.c
index 54535831f1e1..77881798f793 100644
--- a/block/blk-merge.c
+++ b/block/blk-merge.c
@@ -10,10 +10,11 @@
 #include "blk.h"
 
 static unsigned int __blk_recalc_rq_segments(struct request_queue *q,
-                                             struct bio *bio)
+                                             struct bio *bio,
+                                             bool no_sg_merge)
 {
         struct bio_vec bv, bvprv = { NULL };
-        int cluster, high, highprv = 1, no_sg_merge;
+        int cluster, high, highprv = 1;
         unsigned int seg_size, nr_phys_segs;
         struct bio *fbio, *bbio;
         struct bvec_iter iter;
@@ -35,7 +36,6 @@ static unsigned int __blk_recalc_rq_segments(struct request_queue *q,
         cluster = blk_queue_cluster(q);
         seg_size = 0;
         nr_phys_segs = 0;
-        no_sg_merge = test_bit(QUEUE_FLAG_NO_SG_MERGE, &q->queue_flags);
         high = 0;
         for_each_bio(bio) {
                 bio_for_each_segment(bv, bio, iter) {
@@ -88,18 +88,23 @@ new_segment:
 
 void blk_recalc_rq_segments(struct request *rq)
 {
-        rq->nr_phys_segments = __blk_recalc_rq_segments(rq->q, rq->bio);
+        bool no_sg_merge = !!test_bit(QUEUE_FLAG_NO_SG_MERGE,
+                        &rq->q->queue_flags);
+
+        rq->nr_phys_segments = __blk_recalc_rq_segments(rq->q, rq->bio,
+                        no_sg_merge);
 }
 
 void blk_recount_segments(struct request_queue *q, struct bio *bio)
 {
-        if (test_bit(QUEUE_FLAG_NO_SG_MERGE, &q->queue_flags))
+        if (test_bit(QUEUE_FLAG_NO_SG_MERGE, &q->queue_flags) &&
+                        bio->bi_vcnt < queue_max_segments(q))
                 bio->bi_phys_segments = bio->bi_vcnt;
         else {
                 struct bio *nxt = bio->bi_next;
 
                 bio->bi_next = NULL;
-                bio->bi_phys_segments = __blk_recalc_rq_segments(q, bio);
+                bio->bi_phys_segments = __blk_recalc_rq_segments(q, bio, false);
                 bio->bi_next = nxt;
         }
 
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 5189cb1e478a..df8e1e09dd17 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -112,18 +112,31 @@ static void blk_mq_usage_counter_release(struct percpu_ref *ref)
  */
 void blk_mq_freeze_queue(struct request_queue *q)
 {
+        bool freeze;
+
         spin_lock_irq(q->queue_lock);
-        q->mq_freeze_depth++;
+        freeze = !q->mq_freeze_depth++;
         spin_unlock_irq(q->queue_lock);
 
-        percpu_ref_kill(&q->mq_usage_counter);
-        blk_mq_run_queues(q, false);
+        if (freeze) {
+                /*
+                 * XXX: Temporary kludge to work around SCSI blk-mq stall.
+                 * SCSI synchronously creates and destroys many queues
+                 * back-to-back during probe leading to lengthy stalls.
+                 * This will be fixed by keeping ->mq_usage_counter in
+                 * atomic mode until genhd registration, but, for now,
+                 * let's work around using expedited synchronization.
+                 */
+                __percpu_ref_kill_expedited(&q->mq_usage_counter);
+
+                blk_mq_run_queues(q, false);
+        }
         wait_event(q->mq_freeze_wq, percpu_ref_is_zero(&q->mq_usage_counter));
 }
 
 static void blk_mq_unfreeze_queue(struct request_queue *q)
 {
-        bool wake = false;
+        bool wake;
 
         spin_lock_irq(q->queue_lock);
         wake = !--q->mq_freeze_depth;
@@ -172,6 +185,8 @@ static void blk_mq_rq_ctx_init(struct request_queue *q, struct blk_mq_ctx *ctx,
         /* tag was already set */
         rq->errors = 0;
 
+        rq->cmd = rq->__cmd;
+
         rq->extra_len = 0;
         rq->sense_len = 0;
         rq->resid_len = 0;
@@ -197,7 +212,6 @@ __blk_mq_alloc_request(struct blk_mq_alloc_data *data, int rw)
         if (tag != BLK_MQ_TAG_FAIL) {
                 rq = data->hctx->tags->rqs[tag];
 
-                rq->cmd_flags = 0;
                 if (blk_mq_tag_busy(data->hctx)) {
                         rq->cmd_flags = REQ_MQ_INFLIGHT;
                         atomic_inc(&data->hctx->nr_active);
@@ -252,6 +266,7 @@ static void __blk_mq_free_request(struct blk_mq_hw_ctx *hctx,
 
         if (rq->cmd_flags & REQ_MQ_INFLIGHT)
                 atomic_dec(&hctx->nr_active);
+        rq->cmd_flags = 0;
 
         clear_bit(REQ_ATOM_STARTED, &rq->atomic_flags);
         blk_mq_put_tag(hctx, tag, &ctx->last_tag);
@@ -387,6 +402,12 @@ static void blk_mq_start_request(struct request *rq, bool last)
         blk_add_timer(rq);
 
         /*
+         * Ensure that ->deadline is visible before set the started
+         * flag and clear the completed flag.
+         */
+        smp_mb__before_atomic();
+
+        /*
          * Mark us as started and clear complete. Complete might have been
          * set if requeue raced with timeout, which then marked it as
          * complete. So be sure to clear complete again when we start
@@ -467,7 +488,11 @@ static void blk_mq_requeue_work(struct work_struct *work)
                 blk_mq_insert_request(rq, false, false, false);
         }
 
-        blk_mq_run_queues(q, false);
+        /*
+         * Use the start variant of queue running here, so that running
+         * the requeue work will kick stopped queues.
+         */
+        blk_mq_start_hw_queues(q);
 }
 
 void blk_mq_add_to_requeue_list(struct request *rq, bool at_head)
@@ -951,14 +976,9 @@ void blk_mq_insert_request(struct request *rq, bool at_head, bool run_queue,
 
         hctx = q->mq_ops->map_queue(q, ctx->cpu);
 
-        if (rq->cmd_flags & (REQ_FLUSH | REQ_FUA) &&
-            !(rq->cmd_flags & (REQ_FLUSH_SEQ))) {
-                blk_insert_flush(rq);
-        } else {
-                spin_lock(&ctx->lock);
-                __blk_mq_insert_request(hctx, rq, at_head);
-                spin_unlock(&ctx->lock);
-        }
+        spin_lock(&ctx->lock);
+        __blk_mq_insert_request(hctx, rq, at_head);
+        spin_unlock(&ctx->lock);
 
         if (run_queue)
                 blk_mq_run_hw_queue(hctx, async);
@@ -1068,13 +1088,17 @@ static void blk_mq_bio_to_request(struct request *rq, struct bio *bio)
                 blk_account_io_start(rq, 1);
 }
 
+static inline bool hctx_allow_merges(struct blk_mq_hw_ctx *hctx)
+{
+        return (hctx->flags & BLK_MQ_F_SHOULD_MERGE) &&
+                !blk_queue_nomerges(hctx->queue);
+}
+
 static inline bool blk_mq_merge_queue_io(struct blk_mq_hw_ctx *hctx,
                                          struct blk_mq_ctx *ctx,
                                          struct request *rq, struct bio *bio)
 {
-        struct request_queue *q = hctx->queue;
-
-        if (!(hctx->flags & BLK_MQ_F_SHOULD_MERGE)) {
+        if (!hctx_allow_merges(hctx)) {
                 blk_mq_bio_to_request(rq, bio);
                 spin_lock(&ctx->lock);
 insert_rq:
@@ -1082,6 +1106,8 @@ insert_rq:
                 spin_unlock(&ctx->lock);
                 return false;
         } else {
+                struct request_queue *q = hctx->queue;
+
                 spin_lock(&ctx->lock);
                 if (!blk_mq_attempt_merge(q, ctx, bio)) {
                         blk_mq_bio_to_request(rq, bio);
@@ -1309,6 +1335,7 @@ static void blk_mq_free_rq_map(struct blk_mq_tag_set *set,
                                 continue;
                         set->ops->exit_request(set->driver_data, tags->rqs[i],
                                                 hctx_idx, i);
+                        tags->rqs[i] = NULL;
                 }
         }
 
@@ -1342,8 +1369,9 @@ static struct blk_mq_tags *blk_mq_init_rq_map(struct blk_mq_tag_set *set,
 
         INIT_LIST_HEAD(&tags->page_list);
 
-        tags->rqs = kmalloc_node(set->queue_depth * sizeof(struct request *),
-                                        GFP_KERNEL, set->numa_node);
+        tags->rqs = kzalloc_node(set->queue_depth * sizeof(struct request *),
+                                 GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY,
+                                 set->numa_node);
         if (!tags->rqs) {
                 blk_mq_free_tags(tags);
                 return NULL;
@@ -1367,8 +1395,9 @@ static struct blk_mq_tags *blk_mq_init_rq_map(struct blk_mq_tag_set *set,
                         this_order--;
 
                 do {
-                        page = alloc_pages_node(set->numa_node, GFP_KERNEL,
-                                                this_order);
+                        page = alloc_pages_node(set->numa_node,
+                                GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY,
+                                this_order);
                         if (page)
                                 break;
                         if (!this_order--)
@@ -1389,11 +1418,15 @@ static struct blk_mq_tags *blk_mq_init_rq_map(struct blk_mq_tag_set *set,
                 left -= to_do * rq_size;
                 for (j = 0; j < to_do; j++) {
                         tags->rqs[i] = p;
+                        tags->rqs[i]->atomic_flags = 0;
+                        tags->rqs[i]->cmd_flags = 0;
                         if (set->ops->init_request) {
                                 if (set->ops->init_request(set->driver_data,
                                                 tags->rqs[i], hctx_idx, i,
-                                                set->numa_node))
+                                                set->numa_node)) {
+                                        tags->rqs[i] = NULL;
                                         goto fail;
+                                }
                         }
 
                         p += rq_size;
@@ -1404,7 +1437,6 @@ static struct blk_mq_tags *blk_mq_init_rq_map(struct blk_mq_tag_set *set,
         return tags;
 
 fail:
-        pr_warn("%s: failed to allocate requests\n", __func__);
         blk_mq_free_rq_map(set, tags, hctx_idx);
         return NULL;
 }
@@ -1574,7 +1606,7 @@ static int blk_mq_init_hw_queues(struct request_queue *q,
                 hctx->tags = set->tags[i];
 
                 /*
-                 * Allocate space for all possible cpus to avoid allocation in
+                 * Allocate space for all possible cpus to avoid allocation at
                  * runtime
                  */
                 hctx->ctxs = kmalloc_node(nr_cpu_ids * sizeof(void *),
@@ -1662,8 +1694,8 @@ static void blk_mq_map_swqueue(struct request_queue *q)
 
         queue_for_each_hw_ctx(q, hctx, i) {
                 /*
-                 * If not software queues are mapped to this hardware queue,
-                 * disable it and free the request entries
+                 * If no software queues are mapped to this hardware queue,
+                 * disable it and free the request entries.
                  */
                 if (!hctx->nr_ctx) {
                         struct blk_mq_tag_set *set = q->tag_set;
@@ -1713,14 +1745,10 @@ static void blk_mq_del_queue_tag_set(struct request_queue *q)
 {
         struct blk_mq_tag_set *set = q->tag_set;
 
-        blk_mq_freeze_queue(q);
-
         mutex_lock(&set->tag_list_lock);
         list_del_init(&q->tag_set_list);
         blk_mq_update_tag_set_depth(set);
         mutex_unlock(&set->tag_list_lock);
-
-        blk_mq_unfreeze_queue(q);
 }
 
 static void blk_mq_add_queue_tag_set(struct blk_mq_tag_set *set,
@@ -1928,6 +1956,60 @@ static int blk_mq_queue_reinit_notify(struct notifier_block *nb,
         return NOTIFY_OK;
 }
 
+static int __blk_mq_alloc_rq_maps(struct blk_mq_tag_set *set)
+{
+        int i;
+
+        for (i = 0; i < set->nr_hw_queues; i++) {
+                set->tags[i] = blk_mq_init_rq_map(set, i);
+                if (!set->tags[i])
+                        goto out_unwind;
+        }
+
+        return 0;
+
+out_unwind:
+        while (--i >= 0)
+                blk_mq_free_rq_map(set, set->tags[i], i);
+
+        return -ENOMEM;
+}
+
+/*
+ * Allocate the request maps associated with this tag_set. Note that this
+ * may reduce the depth asked for, if memory is tight. set->queue_depth
+ * will be updated to reflect the allocated depth.
+ */
+static int blk_mq_alloc_rq_maps(struct blk_mq_tag_set *set)
+{
+        unsigned int depth;
+        int err;
+
+        depth = set->queue_depth;
+        do {
+                err = __blk_mq_alloc_rq_maps(set);
+                if (!err)
+                        break;
+
+                set->queue_depth >>= 1;
+                if (set->queue_depth < set->reserved_tags + BLK_MQ_TAG_MIN) {
+                        err = -ENOMEM;
+                        break;
+                }
+        } while (set->queue_depth);
+
+        if (!set->queue_depth || err) {
+                pr_err("blk-mq: failed to allocate request map\n");
+                return -ENOMEM;
+        }
+
+        if (depth != set->queue_depth)
+                pr_info("blk-mq: reduced tag depth (%u -> %u)\n",
+                                                depth, set->queue_depth);
+
+        return 0;
+}
+
 /*
  * Alloc a tag set to be associated with one or more request queues.
  * May fail with EINVAL for various error conditions. May adjust the
@@ -1936,8 +2018,6 @@ static int blk_mq_queue_reinit_notify(struct notifier_block *nb,
  */
 int blk_mq_alloc_tag_set(struct blk_mq_tag_set *set)
 {
-        int i;
-
         if (!set->nr_hw_queues)
                 return -EINVAL;
         if (!set->queue_depth)
@@ -1958,23 +2038,18 @@ int blk_mq_alloc_tag_set(struct blk_mq_tag_set *set)
                                  sizeof(struct blk_mq_tags *),
                                  GFP_KERNEL, set->numa_node);
         if (!set->tags)
-                goto out;
+                return -ENOMEM;
 
-        for (i = 0; i < set->nr_hw_queues; i++) {
-                set->tags[i] = blk_mq_init_rq_map(set, i);
-                if (!set->tags[i])
-                        goto out_unwind;
-        }
+        if (blk_mq_alloc_rq_maps(set))
+                goto enomem;
 
         mutex_init(&set->tag_list_lock);
         INIT_LIST_HEAD(&set->tag_list);
 
         return 0;
-
-out_unwind:
-        while (--i >= 0)
-                blk_mq_free_rq_map(set, set->tags[i], i);
-out:
+enomem:
+        kfree(set->tags);
+        set->tags = NULL;
         return -ENOMEM;
 }
 EXPORT_SYMBOL(blk_mq_alloc_tag_set);
@@ -1989,6 +2064,7 @@ void blk_mq_free_tag_set(struct blk_mq_tag_set *set)
         }
 
         kfree(set->tags);
+        set->tags = NULL;
 }
 EXPORT_SYMBOL(blk_mq_free_tag_set);
 
diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
index 4db5abf96b9e..17f5c84ce7bf 100644
--- a/block/blk-sysfs.c
+++ b/block/blk-sysfs.c
@@ -554,8 +554,10 @@ int blk_register_queue(struct gendisk *disk)
          * Initialization must be complete by now.  Finish the initial
          * bypass from queue allocation.
          */
-        queue_flag_set_unlocked(QUEUE_FLAG_INIT_DONE, q);
-        blk_queue_bypass_end(q);
+        if (!blk_queue_init_done(q)) {
+                queue_flag_set_unlocked(QUEUE_FLAG_INIT_DONE, q);
+                blk_queue_bypass_end(q);
+        }
 
         ret = blk_trace_init_sysfs(dev);
         if (ret)
diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index cadc37841744..3f31cf9508e6 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -1272,15 +1272,22 @@ __cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
         rb_insert_color(&cfqg->rb_node, &st->rb);
 }
 
+/*
+ * This has to be called only on activation of cfqg
+ */
 static void
 cfq_update_group_weight(struct cfq_group *cfqg)
 {
-        BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node));
-
         if (cfqg->new_weight) {
                 cfqg->weight = cfqg->new_weight;
                 cfqg->new_weight = 0;
         }
+}
+
+static void
+cfq_update_group_leaf_weight(struct cfq_group *cfqg)
+{
+        BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node));
 
         if (cfqg->new_leaf_weight) {
                 cfqg->leaf_weight = cfqg->new_leaf_weight;
@@ -1299,7 +1306,12 @@ cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
         /* add to the service tree */
         BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node));
 
-        cfq_update_group_weight(cfqg);
+        /*
+         * Update leaf_weight.  We cannot update weight at this point
+         * because cfqg might already have been activated and is
+         * contributing its current weight to the parent's child_weight.
+         */
+        cfq_update_group_leaf_weight(cfqg);
         __cfq_group_service_tree_add(st, cfqg);
 
         /*
@@ -1323,6 +1335,7 @@ cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
          */
         while ((parent = cfqg_parent(pos))) {
                 if (propagate) {
+                        cfq_update_group_weight(pos);
                         propagate = !parent->nr_active++;
                         parent->children_weight += pos->weight;
                 }
diff --git a/block/genhd.c b/block/genhd.c
index 791f41943132..e6723bd4d7a1 100644
--- a/block/genhd.c
+++ b/block/genhd.c
@@ -28,10 +28,10 @@ struct kobject *block_depr;
 /* for extended dynamic devt allocation, currently only one major is used */
 #define NR_EXT_DEVT             (1 << MINORBITS)
 
-/* For extended devt allocation.  ext_devt_mutex prevents look up
+/* For extended devt allocation.  ext_devt_lock prevents look up
  * results from going away underneath its user.
  */
-static DEFINE_MUTEX(ext_devt_mutex);
+static DEFINE_SPINLOCK(ext_devt_lock);
 static DEFINE_IDR(ext_devt_idr);
 
 static struct device_type disk_type;
@@ -420,9 +420,13 @@ int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
         }
 
         /* allocate ext devt */
-        mutex_lock(&ext_devt_mutex);
-        idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_KERNEL);
-        mutex_unlock(&ext_devt_mutex);
+        idr_preload(GFP_KERNEL);
+
+        spin_lock(&ext_devt_lock);
+        idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT);
+        spin_unlock(&ext_devt_lock);
+
+        idr_preload_end();
         if (idx < 0)
                 return idx == -ENOSPC ? -EBUSY : idx;
 
@@ -441,15 +445,13 @@ int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
  */
 void blk_free_devt(dev_t devt)
 {
-        might_sleep();
-
         if (devt == MKDEV(0, 0))
                 return;
 
         if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
-                mutex_lock(&ext_devt_mutex);
+                spin_lock(&ext_devt_lock);
                 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
-                mutex_unlock(&ext_devt_mutex);
+                spin_unlock(&ext_devt_lock);
         }
 }
 
@@ -665,7 +667,6 @@ void del_gendisk(struct gendisk *disk)
                 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
         pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
         device_del(disk_to_dev(disk));
-        blk_free_devt(disk_to_dev(disk)->devt);
 }
 EXPORT_SYMBOL(del_gendisk);
 
@@ -690,13 +691,13 @@ struct gendisk *get_gendisk(dev_t devt, int *partno)
         } else {
                 struct hd_struct *part;
 
-                mutex_lock(&ext_devt_mutex);
+                spin_lock(&ext_devt_lock);
                 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
                 if (part && get_disk(part_to_disk(part))) {
                         *partno = part->partno;
                         disk = part_to_disk(part);
                 }
-                mutex_unlock(&ext_devt_mutex);
+                spin_unlock(&ext_devt_lock);
         }
 
         return disk;
@@ -1098,6 +1099,7 @@ static void disk_release(struct device *dev)
 {
         struct gendisk *disk = dev_to_disk(dev);
 
+        blk_free_devt(dev->devt);
         disk_release_events(disk);
         kfree(disk->random);
         disk_replace_part_tbl(disk, NULL);
diff --git a/block/partition-generic.c b/block/partition-generic.c
index 789cdea05893..0d9e5f97f0a8 100644
--- a/block/partition-generic.c
+++ b/block/partition-generic.c
@@ -211,6 +211,7 @@ static const struct attribute_group *part_attr_groups[] = {
 static void part_release(struct device *dev)
 {
         struct hd_struct *p = dev_to_part(dev);
+        blk_free_devt(dev->devt);
         free_part_stats(p);
         free_part_info(p);
         kfree(p);
@@ -253,7 +254,6 @@ void delete_partition(struct gendisk *disk, int partno)
         rcu_assign_pointer(ptbl->last_lookup, NULL);
         kobject_put(part->holder_dir);
         device_del(part_to_dev(part));
-        blk_free_devt(part_devt(part));
 
         hd_struct_put(part);
 }
diff --git a/block/scsi_ioctl.c b/block/scsi_ioctl.c
index 51bf5155ee75..9b8eaeca6a79 100644
--- a/block/scsi_ioctl.c
+++ b/block/scsi_ioctl.c
@@ -279,7 +279,6 @@ static int blk_complete_sghdr_rq(struct request *rq, struct sg_io_hdr *hdr,
         r = blk_rq_unmap_user(bio);
         if (!ret)
                 ret = r;
-        blk_put_request(rq);
 
         return ret;
 }
@@ -297,8 +296,6 @@ static int sg_io(struct request_queue *q, struct gendisk *bd_disk,
 
         if (hdr->interface_id != 'S')
                 return -EINVAL;
-        if (hdr->cmd_len > BLK_MAX_CDB)
-                return -EINVAL;
 
         if (hdr->dxfer_len > (queue_max_hw_sectors(q) << 9))
                 return -EIO;
@@ -317,16 +314,23 @@ static int sg_io(struct request_queue *q, struct gendisk *bd_disk,
         if (hdr->flags & SG_FLAG_Q_AT_HEAD)
                 at_head = 1;
 
+        ret = -ENOMEM;
         rq = blk_get_request(q, writing ? WRITE : READ, GFP_KERNEL);
         if (!rq)
-                return -ENOMEM;
+                goto out;
         blk_rq_set_block_pc(rq);
 
-        if (blk_fill_sghdr_rq(q, rq, hdr, mode)) {
-                blk_put_request(rq);
-                return -EFAULT;
+        if (hdr->cmd_len > BLK_MAX_CDB) {
+                rq->cmd = kzalloc(hdr->cmd_len, GFP_KERNEL);
+                if (!rq->cmd)
+                        goto out_put_request;
         }
 
+        ret = -EFAULT;
+        if (blk_fill_sghdr_rq(q, rq, hdr, mode))
+                goto out_free_cdb;
+
+        ret = 0;
         if (hdr->iovec_count) {
                 size_t iov_data_len;
                 struct iovec *iov = NULL;
@@ -335,7 +339,7 @@ static int sg_io(struct request_queue *q, struct gendisk *bd_disk,
                                             0, NULL, &iov);
                 if (ret < 0) {
                         kfree(iov);
-                        goto out;
+                        goto out_free_cdb;
                 }
 
                 iov_data_len = ret;
@@ -358,7 +362,7 @@ static int sg_io(struct request_queue *q, struct gendisk *bd_disk,
                                       GFP_KERNEL);
 
         if (ret)
-                goto out;
+                goto out_free_cdb;
 
         bio = rq->bio;
         memset(sense, 0, sizeof(sense));
@@ -376,9 +380,14 @@ static int sg_io(struct request_queue *q, struct gendisk *bd_disk,
 
         hdr->duration = jiffies_to_msecs(jiffies - start_time);
 
-        return blk_complete_sghdr_rq(rq, hdr, bio);
-out:
+        ret = blk_complete_sghdr_rq(rq, hdr, bio);
+
+out_free_cdb:
+        if (rq->cmd != rq->__cmd)
+                kfree(rq->cmd);
+out_put_request:
         blk_put_request(rq);
+out:
         return ret;
 }
 
@@ -448,6 +457,11 @@ int sg_scsi_ioctl(struct request_queue *q, struct gendisk *disk, fmode_t mode,
         }
 
         rq = blk_get_request(q, in_len ? WRITE : READ, __GFP_WAIT);
+        if (!rq) {
+                err = -ENOMEM;
+                goto error;
+        }
+        blk_rq_set_block_pc(rq);
 
         cmdlen = COMMAND_SIZE(opcode);
 
@@ -501,7 +515,6 @@ int sg_scsi_ioctl(struct request_queue *q, struct gendisk *disk, fmode_t mode,
         memset(sense, 0, sizeof(sense));
         rq->sense = sense;
         rq->sense_len = 0;
-        blk_rq_set_block_pc(rq);
 
         blk_execute_rq(q, disk, rq, 0);
 
@@ -521,7 +534,8 @@ out:
         
 error:
         kfree(buffer);
-        blk_put_request(rq);
+        if (rq)
+                blk_put_request(rq);
         return err;
 }
 EXPORT_SYMBOL_GPL(sg_scsi_ioctl);