[PATCH 0/5] percpu_ida+Co: Make percpu_ida_alloc accept task state bitmask

[PATCH 0/5] percpu_ida+Co: Make percpu_ida_alloc accept task state bitmask

[Nicholas A. Bellinger]

From: Nicholas Bellinger <nab@linux-iscsi.org>

Hello Linus,

Here is the series for converting percpu_ida_alloc() + consumer
usage to accept the task state bitmask parameter, w/o the extra
legacy gfp_t wrapper.

Also, given that the first three patches needs to be backported
to different kernel versions, the series has been broken up into
seperate commits for blk-mq (v3.13.y) and target/* (v3.12.y) with
the minimal set of changes marked for stable.

The last two patches propigate task state bitmask usage further
up the blk-mq + target/iscsi stack, and as improvements are not
marked for stable code.

CC'ing Ingo + Peter for TASK_RUNNING + prepare_to_wait() bit.

Thank you,

--nab

Kent Overstreet (1):
  percpu_ida: Make percpu_ida_alloc accept task state bitmask

Nicholas Bellinger (4):
  blk-mq: Update percpu_ida_alloc to use task state bitmask
  target: Update percpu_ida_alloc to use task state bitmask
  blk-mq: Convert gfp_t parameters to task state bitmask
  target/iscsi: Convert gfp_t parameter to task state bitmask

 block/blk-core.c                         |    4 +++-
 block/blk-flush.c                        |    2 +-
 block/blk-mq-tag.c                       |   14 +++++++-------
 block/blk-mq-tag.h                       |    2 +-
 block/blk-mq.c                           |   28 ++++++++++++++--------------
 drivers/infiniband/ulp/isert/ib_isert.c  |   14 +++++++-------
 drivers/target/iscsi/iscsi_target.c      |   14 +++++++-------
 drivers/target/iscsi/iscsi_target_util.c |    9 ++++++---
 drivers/target/iscsi/iscsi_target_util.h |    2 +-
 drivers/vhost/scsi.c                     |    2 +-
 include/linux/blk-mq.h                   |    4 ++--
 include/linux/percpu_ida.h               |    3 ++-
 include/target/iscsi/iscsi_transport.h   |    2 +-
 lib/percpu_ida.c                         |   17 +++++++++++------
 14 files changed, 64 insertions(+), 53 deletions(-)

-- 
1.7.10.4

[PATCH 1/5] percpu_ida: Make percpu_ida_alloc accept task state bitmask

[Nicholas A. Bellinger]

From: Kent Overstreet <kmo@daterainc.com>

This patch changes percpu_ida_alloc() to accept task state bitmask
for prepare_to_wait() to support interruptible sleep for callers
that require it.

Previously, this code was assuming uninterruptible sleep for all cases
thus preventing signals from being delivered to a scheduled process
context in the tag stealing slow path.

It now accepts TASK_INTERRUPTIBLE or TASK_UNINTERRUPTIBLE when the
caller is able to sleep waiting for a new tag, or TASK_RUNNING when
the caller cannot sleep, and is forced to return a negative tag.

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: <stable@vger.kernel.org> #3.12+
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
---
 include/linux/percpu_ida.h |    3 ++-
 lib/percpu_ida.c           |   17 +++++++++++------
 2 files changed, 13 insertions(+), 7 deletions(-)

diff --git a/include/linux/percpu_ida.h b/include/linux/percpu_ida.h
index 1900bd0..f5cfdd6 100644
--- a/include/linux/percpu_ida.h
+++ b/include/linux/percpu_ida.h
@@ -4,6 +4,7 @@
 #include <linux/types.h>
 #include <linux/bitops.h>
 #include <linux/init.h>
+#include <linux/sched.h>
 #include <linux/spinlock_types.h>
 #include <linux/wait.h>
 #include <linux/cpumask.h>
@@ -61,7 +62,7 @@ struct percpu_ida {
 /* Max size of percpu freelist, */
 #define IDA_DEFAULT_PCPU_SIZE	((IDA_DEFAULT_PCPU_BATCH_MOVE * 3) / 2)
 
-int percpu_ida_alloc(struct percpu_ida *pool, gfp_t gfp);
+int percpu_ida_alloc(struct percpu_ida *pool, int state);
 void percpu_ida_free(struct percpu_ida *pool, unsigned tag);
 
 void percpu_ida_destroy(struct percpu_ida *pool);
diff --git a/lib/percpu_ida.c b/lib/percpu_ida.c
index 9d054bf..4579749 100644
--- a/lib/percpu_ida.c
+++ b/lib/percpu_ida.c
@@ -132,22 +132,22 @@ static inline unsigned alloc_local_tag(struct percpu_ida_cpu *tags)
 /**
  * percpu_ida_alloc - allocate a tag
  * @pool: pool to allocate from
- * @gfp: gfp flags
+ * @state: task state for prepare_to_wait
  *
  * Returns a tag - an integer in the range [0..nr_tags) (passed to
  * tag_pool_init()), or otherwise -ENOSPC on allocation failure.
  *
  * Safe to be called from interrupt context (assuming it isn't passed
- * __GFP_WAIT, of course).
+ * TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE, of course).
  *
  * @gfp indicates whether or not to wait until a free id is available (it's not
  * used for internal memory allocations); thus if passed __GFP_WAIT we may sleep
  * however long it takes until another thread frees an id (same semantics as a
  * mempool).
  *
- * Will not fail if passed __GFP_WAIT.
+ * Will not fail if passed TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE.
  */
-int percpu_ida_alloc(struct percpu_ida *pool, gfp_t gfp)
+int percpu_ida_alloc(struct percpu_ida *pool, int state)
 {
 	DEFINE_WAIT(wait);
 	struct percpu_ida_cpu *tags;
@@ -174,7 +174,7 @@ int percpu_ida_alloc(struct percpu_ida *pool, gfp_t gfp)
 		 *
 		 * global lock held and irqs disabled, don't need percpu lock
 		 */
-		prepare_to_wait(&pool->wait, &wait, TASK_UNINTERRUPTIBLE);
+		prepare_to_wait(&pool->wait, &wait, state);
 
 		if (!tags->nr_free)
 			alloc_global_tags(pool, tags);
@@ -191,9 +191,14 @@ int percpu_ida_alloc(struct percpu_ida *pool, gfp_t gfp)
 		spin_unlock(&pool->lock);
 		local_irq_restore(flags);
 
-		if (tag >= 0 || !(gfp & __GFP_WAIT))
+		if (tag >= 0 || state == TASK_RUNNING)
 			break;
 
+		if (signal_pending_state(state, current)) {
+			tag = -ERESTARTSYS;
+			break;
+		}
+
 		schedule();
 
 		local_irq_save(flags);
-- 
1.7.10.4

Re: [PATCH 1/5] percpu_ida: Make percpu_ida_alloc accept task state bitmask

[Linus Torvalds]

On Sun, Jan 19, 2014 at 2:16 AM, Nicholas A. Bellinger
<nab@linux-iscsi.org> wrote:
>
> This patch changes percpu_ida_alloc() to accept task state bitmask
> for prepare_to_wait() to support interruptible sleep for callers
> that require it.

This patch-series is not bisectable. Afaik, the first patch will break
the build (or at least cause the end result to not actually work).

This kind of "split up one large patch into many small patches THAT
DON'T ACTUALLY WORK INDIVIDUALLY" model is pure and utter garbage.

So a big NAK on this series as being completely broken.

To fix it, I would suggest:

 - make the first patch change all *existing* users (that only have
the atomic vs uninterruptible semantics) pass in either
TASK_UNINTERRUPTIBLE or TASK_RUNNING depending on whether they had
__GFP_WAIT or not.

   So the first patch would not change *any* semantics or behavior, it
would only change the calling convention.

 - do the cleanup patches to block/blk-mq-tag.c to not have those
"gfp" calling convention, and instead passing in the state natively

 - add the TASK_INTERRUPTIBLE case last (which includes the new
"signal_pending_state()" logic in percpu_ida_alloc())

that way, all patches compile cleanly and should each work
individually, and they all do clearly just one thing. And the biggest
patch in the series (the first one) doesn't actually make any semantic
changes.

                Linus

Re: [PATCH 1/5] percpu_ida: Make percpu_ida_alloc accept task state bitmask

[Nicholas A. Bellinger]

On Sun, 2014-01-19 at 18:38 -0800, Linus Torvalds wrote:
> On Sun, Jan 19, 2014 at 2:16 AM, Nicholas A. Bellinger
> <nab@linux-iscsi.org> wrote:
> >
> > This patch changes percpu_ida_alloc() to accept task state bitmask
> > for prepare_to_wait() to support interruptible sleep for callers
> > that require it.
> 
> This patch-series is not bisectable. Afaik, the first patch will break
> the build (or at least cause the end result to not actually work).
> 
> This kind of "split up one large patch into many small patches THAT
> DON'T ACTUALLY WORK INDIVIDUALLY" model is pure and utter garbage.
> 
> So a big NAK on this series as being completely broken.
> 

So the late night reasoning was to allow the patches to apply cleanly to
stable.  That was, indeed, a bad decision.

> To fix it, I would suggest:
> 
>  - make the first patch change all *existing* users (that only have
> the atomic vs uninterruptible semantics) pass in either
> TASK_UNINTERRUPTIBLE or TASK_RUNNING depending on whether they had
> __GFP_WAIT or not.
> 
>    So the first patch would not change *any* semantics or behavior, it
> would only change the calling convention.
> 
>  - do the cleanup patches to block/blk-mq-tag.c to not have those
> "gfp" calling convention, and instead passing in the state natively
> 
>  - add the TASK_INTERRUPTIBLE case last (which includes the new
> "signal_pending_state()" logic in percpu_ida_alloc())
> 
> that way, all patches compile cleanly and should each work
> individually, and they all do clearly just one thing. And the biggest
> patch in the series (the first one) doesn't actually make any semantic
> changes.

-v2 sent out.  Please review.

Jens, please review the blk-mq specific changes in patch #1, and let me
know if you'd like to pick-up #2 via the block tree, or have it included
in target-pending/for-next.

Thanks,

--nab

[PATCH 2/5] blk-mq: Update percpu_ida_alloc to use task state bitmask

[Nicholas A. Bellinger]

From: Nicholas Bellinger <nab@linux-iscsi.org>

This patch updates percpu_ida_alloc() usage within blk-mq tag
allocation code to use TASK_UNINTERRUPTIBLE (may sleep) or
ASK_RUNNING (cannot sleep), following the changes to percpu_ida
to allow for task state bitmask to be passed from the caller.

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: <stable@vger.kernel.org> #3.13+
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
---
 block/blk-mq-tag.c |    6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
index d64a02f..5d70edc 100644
--- a/block/blk-mq-tag.c
+++ b/block/blk-mq-tag.c
@@ -36,7 +36,8 @@ static unsigned int __blk_mq_get_tag(struct blk_mq_tags *tags, gfp_t gfp)
 {
 	int tag;
 
-	tag = percpu_ida_alloc(&tags->free_tags, gfp);
+	tag = percpu_ida_alloc(&tags->free_tags, (gfp & __GFP_WAIT) ?
+			       TASK_UNINTERRUPTIBLE : TASK_RUNNING);
 	if (tag < 0)
 		return BLK_MQ_TAG_FAIL;
 	return tag + tags->nr_reserved_tags;
@@ -52,7 +53,8 @@ static unsigned int __blk_mq_get_reserved_tag(struct blk_mq_tags *tags,
 		return BLK_MQ_TAG_FAIL;
 	}
 
-	tag = percpu_ida_alloc(&tags->reserved_tags, gfp);
+	tag = percpu_ida_alloc(&tags->reserved_tags, (gfp & __GFP_WAIT) ?
+			       TASK_UNINTERRUPTIBLE : TASK_RUNNING);
 	if (tag < 0)
 		return BLK_MQ_TAG_FAIL;
 	return tag;
-- 
1.7.10.4

[PATCH 3/5] target: Update percpu_ida_alloc to use task state bitmask

[Nicholas A. Bellinger]

From: Nicholas Bellinger <nab@linux-iscsi.org>

This patch updates percpu_ida_alloc() for target/vhost to TASK_RUNNING
(cannot sleep) instead of GFP_ATOMIC, and updates iscsi-target to use
TASK_INTERRUPTIBLE (may sleep) or TASK_RUNNING (cannot sleep).

Both are following the changes to percpu_ida to allow for task state
bitmask to be passed from the caller.

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: <stable@vger.kernel.org> #3.12+
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
---
 drivers/target/iscsi/iscsi_target_util.c |    8 ++++++--
 drivers/vhost/scsi.c                     |    2 +-
 2 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/drivers/target/iscsi/iscsi_target_util.c b/drivers/target/iscsi/iscsi_target_util.c
index 0819e68..5477eca 100644
--- a/drivers/target/iscsi/iscsi_target_util.c
+++ b/drivers/target/iscsi/iscsi_target_util.c
@@ -156,9 +156,13 @@ struct iscsi_cmd *iscsit_allocate_cmd(struct iscsi_conn *conn, gfp_t gfp_mask)
 {
 	struct iscsi_cmd *cmd;
 	struct se_session *se_sess = conn->sess->se_sess;
-	int size, tag;
+	int size, tag, state = (gfp_mask & __GFP_WAIT) ? TASK_INTERRUPTIBLE :
+				TASK_RUNNING;
+
+	tag = percpu_ida_alloc(&se_sess->sess_tag_pool, state);
+	if (tag < 0)
+		return NULL;
 
-	tag = percpu_ida_alloc(&se_sess->sess_tag_pool, gfp_mask);
 	size = sizeof(struct iscsi_cmd) + conn->conn_transport->priv_size;
 	cmd = (struct iscsi_cmd *)(se_sess->sess_cmd_map + (tag * size));
 	memset(cmd, 0, size);
diff --git a/drivers/vhost/scsi.c b/drivers/vhost/scsi.c
index 84488a8..2d084fb 100644
--- a/drivers/vhost/scsi.c
+++ b/drivers/vhost/scsi.c
@@ -728,7 +728,7 @@ vhost_scsi_get_tag(struct vhost_virtqueue *vq,
 	}
 	se_sess = tv_nexus->tvn_se_sess;
 
-	tag = percpu_ida_alloc(&se_sess->sess_tag_pool, GFP_ATOMIC);
+	tag = percpu_ida_alloc(&se_sess->sess_tag_pool, TASK_RUNNING);
 	if (tag < 0) {
 		pr_err("Unable to obtain tag for tcm_vhost_cmd\n");
 		return ERR_PTR(-ENOMEM);
-- 
1.7.10.4

[PATCH 4/5] blk-mq: Convert gfp_t parameters to task state bitmask

[Nicholas A. Bellinger]

From: Nicholas Bellinger <nab@linux-iscsi.org>

This patch propigates the use of task state bitmask for
percpu_ida_alloc() up the blk-mq callchain, to the point in
blk_get_request() where the blk-mq vs. blk-old split occurs.

Along with the obvious parameters changes, there are two cases
in mq_flush_work() + blk_mq_make_request() where the original
code was using __GFP_WAIT|GFP_ATOMIC that always expect a tag
which have been converted to TASK_UNINTERRUPTIBLE.

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
---
 block/blk-core.c       |    4 +++-
 block/blk-flush.c      |    2 +-
 block/blk-mq-tag.c     |   16 +++++++---------
 block/blk-mq-tag.h     |    2 +-
 block/blk-mq.c         |   28 ++++++++++++++--------------
 include/linux/blk-mq.h |    4 ++--
 6 files changed, 28 insertions(+), 28 deletions(-)

diff --git a/block/blk-core.c b/block/blk-core.c
index 8bdd012..ab0dc9a 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -1112,7 +1112,9 @@ static struct request *blk_old_get_request(struct request_queue *q, int rw,
 struct request *blk_get_request(struct request_queue *q, int rw, gfp_t gfp_mask)
 {
 	if (q->mq_ops)
-		return blk_mq_alloc_request(q, rw, gfp_mask, false);
+		return blk_mq_alloc_request(q, rw, (gfp_mask & __GFP_WAIT) ?
+					    TASK_UNINTERRUPTIBLE : TASK_RUNNING,
+					    false);
 	else
 		return blk_old_get_request(q, rw, gfp_mask);
 }
diff --git a/block/blk-flush.c b/block/blk-flush.c
index fb6f3c0..8dd6ff8 100644
--- a/block/blk-flush.c
+++ b/block/blk-flush.c
@@ -286,7 +286,7 @@ static void mq_flush_work(struct work_struct *work)
 
 	/* We don't need set REQ_FLUSH_SEQ, it's for consistency */
 	rq = blk_mq_alloc_request(q, WRITE_FLUSH|REQ_FLUSH_SEQ,
-		__GFP_WAIT|GFP_ATOMIC, true);
+				  TASK_UNINTERRUPTIBLE, true);
 	rq->cmd_type = REQ_TYPE_FS;
 	rq->end_io = flush_end_io;
 
diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
index 5d70edc..20777bd 100644
--- a/block/blk-mq-tag.c
+++ b/block/blk-mq-tag.c
@@ -32,19 +32,18 @@ bool blk_mq_has_free_tags(struct blk_mq_tags *tags)
 		percpu_ida_free_tags(&tags->free_tags, nr_cpu_ids) != 0;
 }
 
-static unsigned int __blk_mq_get_tag(struct blk_mq_tags *tags, gfp_t gfp)
+static unsigned int __blk_mq_get_tag(struct blk_mq_tags *tags, int state)
 {
 	int tag;
 
-	tag = percpu_ida_alloc(&tags->free_tags, (gfp & __GFP_WAIT) ?
-			       TASK_UNINTERRUPTIBLE : TASK_RUNNING);
+	tag = percpu_ida_alloc(&tags->free_tags, state);
 	if (tag < 0)
 		return BLK_MQ_TAG_FAIL;
 	return tag + tags->nr_reserved_tags;
 }
 
 static unsigned int __blk_mq_get_reserved_tag(struct blk_mq_tags *tags,
-					      gfp_t gfp)
+					      int state)
 {
 	int tag;
 
@@ -53,19 +52,18 @@ static unsigned int __blk_mq_get_reserved_tag(struct blk_mq_tags *tags,
 		return BLK_MQ_TAG_FAIL;
 	}
 
-	tag = percpu_ida_alloc(&tags->reserved_tags, (gfp & __GFP_WAIT) ?
-			       TASK_UNINTERRUPTIBLE : TASK_RUNNING);
+	tag = percpu_ida_alloc(&tags->reserved_tags, state);
 	if (tag < 0)
 		return BLK_MQ_TAG_FAIL;
 	return tag;
 }
 
-unsigned int blk_mq_get_tag(struct blk_mq_tags *tags, gfp_t gfp, bool reserved)
+unsigned int blk_mq_get_tag(struct blk_mq_tags *tags, int state, bool reserved)
 {
 	if (!reserved)
-		return __blk_mq_get_tag(tags, gfp);
+		return __blk_mq_get_tag(tags, state);
 
-	return __blk_mq_get_reserved_tag(tags, gfp);
+	return __blk_mq_get_reserved_tag(tags, state);
 }
 
 static void __blk_mq_put_tag(struct blk_mq_tags *tags, unsigned int tag)
diff --git a/block/blk-mq-tag.h b/block/blk-mq-tag.h
index 947ba2c..b3c1487 100644
--- a/block/blk-mq-tag.h
+++ b/block/blk-mq-tag.h
@@ -6,7 +6,7 @@ struct blk_mq_tags;
 extern struct blk_mq_tags *blk_mq_init_tags(unsigned int nr_tags, unsigned int reserved_tags, int node);
 extern void blk_mq_free_tags(struct blk_mq_tags *tags);
 
-extern unsigned int blk_mq_get_tag(struct blk_mq_tags *tags, gfp_t gfp, bool reserved);
+extern unsigned int blk_mq_get_tag(struct blk_mq_tags *tags, int state, bool reserved);
 extern void blk_mq_wait_for_tags(struct blk_mq_tags *tags);
 extern void blk_mq_put_tag(struct blk_mq_tags *tags, unsigned int tag);
 extern void blk_mq_tag_busy_iter(struct blk_mq_tags *tags, void (*fn)(void *data, unsigned long *), void *data);
diff --git a/block/blk-mq.c b/block/blk-mq.c
index c79126e..80bbfbd 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -75,13 +75,13 @@ static void blk_mq_hctx_mark_pending(struct blk_mq_hw_ctx *hctx,
 		set_bit(ctx->index_hw, hctx->ctx_map);
 }
 
-static struct request *blk_mq_alloc_rq(struct blk_mq_hw_ctx *hctx, gfp_t gfp,
+static struct request *blk_mq_alloc_rq(struct blk_mq_hw_ctx *hctx, int state,
 				       bool reserved)
 {
 	struct request *rq;
 	unsigned int tag;
 
-	tag = blk_mq_get_tag(hctx->tags, gfp, reserved);
+	tag = blk_mq_get_tag(hctx->tags, state, reserved);
 	if (tag != BLK_MQ_TAG_FAIL) {
 		rq = hctx->rqs[tag];
 		rq->tag = tag;
@@ -183,13 +183,13 @@ static void blk_mq_rq_ctx_init(struct request_queue *q, struct blk_mq_ctx *ctx,
 }
 
 static struct request *__blk_mq_alloc_request(struct blk_mq_hw_ctx *hctx,
-					      gfp_t gfp, bool reserved)
+					      int state, bool reserved)
 {
-	return blk_mq_alloc_rq(hctx, gfp, reserved);
+	return blk_mq_alloc_rq(hctx, state, reserved);
 }
 
 static struct request *blk_mq_alloc_request_pinned(struct request_queue *q,
-						   int rw, gfp_t gfp,
+						   int rw, int state,
 						   bool reserved)
 {
 	struct request *rq;
@@ -198,14 +198,14 @@ static struct request *blk_mq_alloc_request_pinned(struct request_queue *q,
 		struct blk_mq_ctx *ctx = blk_mq_get_ctx(q);
 		struct blk_mq_hw_ctx *hctx = q->mq_ops->map_queue(q, ctx->cpu);
 
-		rq = __blk_mq_alloc_request(hctx, gfp & ~__GFP_WAIT, reserved);
+		rq = __blk_mq_alloc_request(hctx, state, reserved);
 		if (rq) {
 			blk_mq_rq_ctx_init(q, ctx, rq, rw);
 			break;
 		}
 
 		blk_mq_put_ctx(ctx);
-		if (!(gfp & __GFP_WAIT))
+		if (state == TASK_RUNNING)
 			break;
 
 		__blk_mq_run_hw_queue(hctx);
@@ -216,28 +216,28 @@ static struct request *blk_mq_alloc_request_pinned(struct request_queue *q,
 }
 
 struct request *blk_mq_alloc_request(struct request_queue *q, int rw,
-		gfp_t gfp, bool reserved)
+		int state, bool reserved)
 {
 	struct request *rq;
 
 	if (blk_mq_queue_enter(q))
 		return NULL;
 
-	rq = blk_mq_alloc_request_pinned(q, rw, gfp, reserved);
+	rq = blk_mq_alloc_request_pinned(q, rw, state, reserved);
 	if (rq)
 		blk_mq_put_ctx(rq->mq_ctx);
 	return rq;
 }
 
 struct request *blk_mq_alloc_reserved_request(struct request_queue *q, int rw,
-					      gfp_t gfp)
+					      int state)
 {
 	struct request *rq;
 
 	if (blk_mq_queue_enter(q))
 		return NULL;
 
-	rq = blk_mq_alloc_request_pinned(q, rw, gfp, true);
+	rq = blk_mq_alloc_request_pinned(q, rw, state, true);
 	if (rq)
 		blk_mq_put_ctx(rq->mq_ctx);
 	return rq;
@@ -928,14 +928,14 @@ static void blk_mq_make_request(struct request_queue *q, struct bio *bio)
 	hctx = q->mq_ops->map_queue(q, ctx->cpu);
 
 	trace_block_getrq(q, bio, rw);
-	rq = __blk_mq_alloc_request(hctx, GFP_ATOMIC, false);
+	rq = __blk_mq_alloc_request(hctx, TASK_RUNNING, false);
 	if (likely(rq))
 		blk_mq_rq_ctx_init(q, ctx, rq, rw);
 	else {
 		blk_mq_put_ctx(ctx);
 		trace_block_sleeprq(q, bio, rw);
-		rq = blk_mq_alloc_request_pinned(q, rw, __GFP_WAIT|GFP_ATOMIC,
-							false);
+		rq = blk_mq_alloc_request_pinned(q, rw, TASK_UNINTERRUPTIBLE,
+						 false);
 		ctx = rq->mq_ctx;
 		hctx = q->mq_ops->map_queue(q, ctx->cpu);
 	}
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index ab0e9b2..91ee75a 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -124,8 +124,8 @@ void blk_mq_insert_request(struct request_queue *, struct request *, bool);
 void blk_mq_run_queues(struct request_queue *q, bool async);
 void blk_mq_free_request(struct request *rq);
 bool blk_mq_can_queue(struct blk_mq_hw_ctx *);
-struct request *blk_mq_alloc_request(struct request_queue *q, int rw, gfp_t gfp, bool reserved);
-struct request *blk_mq_alloc_reserved_request(struct request_queue *q, int rw, gfp_t gfp);
+struct request *blk_mq_alloc_request(struct request_queue *q, int rw, int state, bool reserved);
+struct request *blk_mq_alloc_reserved_request(struct request_queue *q, int rw, int state);
 struct request *blk_mq_rq_from_tag(struct request_queue *q, unsigned int tag);
 
 struct blk_mq_hw_ctx *blk_mq_map_queue(struct request_queue *, const int ctx_index);
-- 
1.7.10.4

[PATCH 5/5] target/iscsi: Convert gfp_t parameter to task state bitmask

[Nicholas A. Bellinger]

From: Nicholas Bellinger <nab@linux-iscsi.org>

This patch propigates the use of task state bitmask now used by
percpu_ida_alloc() up the iscsi-target callchain, replacing the
use of GFP_ATOMIC + GFP_KERNEL.

Also, drop the unnecessary gfp_t parameter to isert_allocate_cmd(),
and just pass TASK_INTERRUPTIBLE into iscsit_allocate_cmd().

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
---
 drivers/infiniband/ulp/isert/ib_isert.c  |   14 +++++++-------
 drivers/target/iscsi/iscsi_target.c      |   14 +++++++-------
 drivers/target/iscsi/iscsi_target_util.c |    7 +++----
 drivers/target/iscsi/iscsi_target_util.h |    2 +-
 include/target/iscsi/iscsi_transport.h   |    2 +-
 5 files changed, 19 insertions(+), 20 deletions(-)

diff --git a/drivers/infiniband/ulp/isert/ib_isert.c b/drivers/infiniband/ulp/isert/ib_isert.c
index 421182b..63bcf69 100644
--- a/drivers/infiniband/ulp/isert/ib_isert.c
+++ b/drivers/infiniband/ulp/isert/ib_isert.c
@@ -1028,13 +1028,13 @@ isert_rx_login_req(struct iser_rx_desc *rx_desc, int rx_buflen,
 }
 
 static struct iscsi_cmd
-*isert_allocate_cmd(struct iscsi_conn *conn, gfp_t gfp)
+*isert_allocate_cmd(struct iscsi_conn *conn)
 {
 	struct isert_conn *isert_conn = (struct isert_conn *)conn->context;
 	struct isert_cmd *isert_cmd;
 	struct iscsi_cmd *cmd;
 
-	cmd = iscsit_allocate_cmd(conn, gfp);
+	cmd = iscsit_allocate_cmd(conn, TASK_INTERRUPTIBLE);
 	if (!cmd) {
 		pr_err("Unable to allocate iscsi_cmd + isert_cmd\n");
 		return NULL;
@@ -1223,7 +1223,7 @@ isert_rx_opcode(struct isert_conn *isert_conn, struct iser_rx_desc *rx_desc,
 
 	switch (opcode) {
 	case ISCSI_OP_SCSI_CMD:
-		cmd = isert_allocate_cmd(conn, GFP_KERNEL);
+		cmd = isert_allocate_cmd(conn);
 		if (!cmd)
 			break;
 
@@ -1237,7 +1237,7 @@ isert_rx_opcode(struct isert_conn *isert_conn, struct iser_rx_desc *rx_desc,
 					rx_desc, (unsigned char *)hdr);
 		break;
 	case ISCSI_OP_NOOP_OUT:
-		cmd = isert_allocate_cmd(conn, GFP_KERNEL);
+		cmd = isert_allocate_cmd(conn);
 		if (!cmd)
 			break;
 
@@ -1250,7 +1250,7 @@ isert_rx_opcode(struct isert_conn *isert_conn, struct iser_rx_desc *rx_desc,
 						(unsigned char *)hdr);
 		break;
 	case ISCSI_OP_SCSI_TMFUNC:
-		cmd = isert_allocate_cmd(conn, GFP_KERNEL);
+		cmd = isert_allocate_cmd(conn);
 		if (!cmd)
 			break;
 
@@ -1258,7 +1258,7 @@ isert_rx_opcode(struct isert_conn *isert_conn, struct iser_rx_desc *rx_desc,
 						(unsigned char *)hdr);
 		break;
 	case ISCSI_OP_LOGOUT:
-		cmd = isert_allocate_cmd(conn, GFP_KERNEL);
+		cmd = isert_allocate_cmd(conn);
 		if (!cmd)
 			break;
 
@@ -1269,7 +1269,7 @@ isert_rx_opcode(struct isert_conn *isert_conn, struct iser_rx_desc *rx_desc,
 						    HZ);
 		break;
 	case ISCSI_OP_TEXT:
-		cmd = isert_allocate_cmd(conn, GFP_KERNEL);
+		cmd = isert_allocate_cmd(conn);
 		if (!cmd)
 			break;
 
diff --git a/drivers/target/iscsi/iscsi_target.c b/drivers/target/iscsi/iscsi_target.c
index d70e911..2a52752 100644
--- a/drivers/target/iscsi/iscsi_target.c
+++ b/drivers/target/iscsi/iscsi_target.c
@@ -621,7 +621,7 @@ static int iscsit_add_reject(
 {
 	struct iscsi_cmd *cmd;
 
-	cmd = iscsit_allocate_cmd(conn, GFP_KERNEL);
+	cmd = iscsit_allocate_cmd(conn, TASK_INTERRUPTIBLE);
 	if (!cmd)
 		return -1;
 
@@ -2476,7 +2476,7 @@ static void iscsit_build_conn_drop_async_message(struct iscsi_conn *conn)
 	if (!conn_p)
 		return;
 
-	cmd = iscsit_allocate_cmd(conn_p, GFP_ATOMIC);
+	cmd = iscsit_allocate_cmd(conn_p, TASK_RUNNING);
 	if (!cmd) {
 		iscsit_dec_conn_usage_count(conn_p);
 		return;
@@ -3952,7 +3952,7 @@ static int iscsi_target_rx_opcode(struct iscsi_conn *conn, unsigned char *buf)
 
 	switch (hdr->opcode & ISCSI_OPCODE_MASK) {
 	case ISCSI_OP_SCSI_CMD:
-		cmd = iscsit_allocate_cmd(conn, GFP_KERNEL);
+		cmd = iscsit_allocate_cmd(conn, TASK_INTERRUPTIBLE);
 		if (!cmd)
 			goto reject;
 
@@ -3964,28 +3964,28 @@ static int iscsi_target_rx_opcode(struct iscsi_conn *conn, unsigned char *buf)
 	case ISCSI_OP_NOOP_OUT:
 		cmd = NULL;
 		if (hdr->ttt == cpu_to_be32(0xFFFFFFFF)) {
-			cmd = iscsit_allocate_cmd(conn, GFP_KERNEL);
+			cmd = iscsit_allocate_cmd(conn, TASK_INTERRUPTIBLE);
 			if (!cmd)
 				goto reject;
 		}
 		ret = iscsit_handle_nop_out(conn, cmd, buf);
 		break;
 	case ISCSI_OP_SCSI_TMFUNC:
-		cmd = iscsit_allocate_cmd(conn, GFP_KERNEL);
+		cmd = iscsit_allocate_cmd(conn, TASK_INTERRUPTIBLE);
 		if (!cmd)
 			goto reject;
 
 		ret = iscsit_handle_task_mgt_cmd(conn, cmd, buf);
 		break;
 	case ISCSI_OP_TEXT:
-		cmd = iscsit_allocate_cmd(conn, GFP_KERNEL);
+		cmd = iscsit_allocate_cmd(conn, TASK_INTERRUPTIBLE);
 		if (!cmd)
 			goto reject;
 
 		ret = iscsit_handle_text_cmd(conn, cmd, buf);
 		break;
 	case ISCSI_OP_LOGOUT:
-		cmd = iscsit_allocate_cmd(conn, GFP_KERNEL);
+		cmd = iscsit_allocate_cmd(conn, TASK_INTERRUPTIBLE);
 		if (!cmd)
 			goto reject;
 
diff --git a/drivers/target/iscsi/iscsi_target_util.c b/drivers/target/iscsi/iscsi_target_util.c
index 5477eca..e655b04 100644
--- a/drivers/target/iscsi/iscsi_target_util.c
+++ b/drivers/target/iscsi/iscsi_target_util.c
@@ -152,12 +152,11 @@ void iscsit_free_r2ts_from_list(struct iscsi_cmd *cmd)
  * May be called from software interrupt (timer) context for allocating
  * iSCSI NopINs.
  */
-struct iscsi_cmd *iscsit_allocate_cmd(struct iscsi_conn *conn, gfp_t gfp_mask)
+struct iscsi_cmd *iscsit_allocate_cmd(struct iscsi_conn *conn, int state)
 {
 	struct iscsi_cmd *cmd;
 	struct se_session *se_sess = conn->sess->se_sess;
-	int size, tag, state = (gfp_mask & __GFP_WAIT) ? TASK_INTERRUPTIBLE :
-				TASK_RUNNING;
+	int size, tag;
 
 	tag = percpu_ida_alloc(&se_sess->sess_tag_pool, state);
 	if (tag < 0)
@@ -930,7 +929,7 @@ static int iscsit_add_nopin(struct iscsi_conn *conn, int want_response)
 	u8 state;
 	struct iscsi_cmd *cmd;
 
-	cmd = iscsit_allocate_cmd(conn, GFP_ATOMIC);
+	cmd = iscsit_allocate_cmd(conn, TASK_RUNNING);
 	if (!cmd)
 		return -1;
 
diff --git a/drivers/target/iscsi/iscsi_target_util.h b/drivers/target/iscsi/iscsi_target_util.h
index e4fc34a..561a424 100644
--- a/drivers/target/iscsi/iscsi_target_util.h
+++ b/drivers/target/iscsi/iscsi_target_util.h
@@ -9,7 +9,7 @@ extern struct iscsi_r2t *iscsit_get_r2t_from_list(struct iscsi_cmd *);
 extern void iscsit_free_r2t(struct iscsi_r2t *, struct iscsi_cmd *);
 extern void iscsit_free_r2ts_from_list(struct iscsi_cmd *);
 extern struct iscsi_cmd *iscsit_alloc_cmd(struct iscsi_conn *, gfp_t);
-extern struct iscsi_cmd *iscsit_allocate_cmd(struct iscsi_conn *, gfp_t);
+extern struct iscsi_cmd *iscsit_allocate_cmd(struct iscsi_conn *, int);
 extern struct iscsi_seq *iscsit_get_seq_holder_for_datain(struct iscsi_cmd *, u32);
 extern struct iscsi_seq *iscsit_get_seq_holder_for_r2t(struct iscsi_cmd *);
 extern struct iscsi_r2t *iscsit_get_holder_for_r2tsn(struct iscsi_cmd *, u32);
diff --git a/include/target/iscsi/iscsi_transport.h b/include/target/iscsi/iscsi_transport.h
index a12589c..ae5a171 100644
--- a/include/target/iscsi/iscsi_transport.h
+++ b/include/target/iscsi/iscsi_transport.h
@@ -94,7 +94,7 @@ extern int iscsit_tmr_post_handler(struct iscsi_cmd *, struct iscsi_conn *);
 /*
  * From iscsi_target_util.c
  */
-extern struct iscsi_cmd *iscsit_allocate_cmd(struct iscsi_conn *, gfp_t);
+extern struct iscsi_cmd *iscsit_allocate_cmd(struct iscsi_conn *, int);
 extern int iscsit_sequence_cmd(struct iscsi_conn *, struct iscsi_cmd *,
 			       unsigned char *, __be32);
 extern void iscsit_release_cmd(struct iscsi_cmd *);
-- 
1.7.10.4