Hi all,
After merging the scsi-mkp tree, today's linux-next build (sparc64
defconfig) failed like this:
drivers/mtd/nand/raw/intel-nand-controller.c:17:10: fatal error: linux/mtd/nand_ecc.h: No such file or directory
17 | #include <linux/mtd/nand_ecc.h>
| ^~~~~~~~~~~~~~~~~~~~~~
Caused by commit
81a395cdc176 ("scsi: block: Do not accept any requests while suspended")
# CONFIG_PM is not set
I have applied the following patch:
From: Stephen Rothwell <[email protected]>
Date: Tue, 8 Dec 2020 20:12:33 +1100
Subject: [PATCH] scsi: block: fix for "scsi: block: Do not accept any requests while suspended"
Fixes: 81a395cdc176 ("scsi: block: Do not accept any requests while suspended")
Signed-off-by: Stephen Rothwell <[email protected]>
---
block/blk-core.c | 18 ++++++++++++++----
1 file changed, 14 insertions(+), 4 deletions(-)
diff --git a/block/blk-core.c b/block/blk-core.c
index a71a5c9429d6..9c9aec1382be 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -421,6 +421,18 @@ void blk_cleanup_queue(struct request_queue *q)
}
EXPORT_SYMBOL(blk_cleanup_queue);
+#ifdef CONFIG_PM
+static bool rq_suspended(struct request_queue *q)
+{
+ return q->rpm_status == RPM_SUSPENDED;
+}
+#else
+static bool rq_suspended(struct request_queue *q)
+{
+ return false;
+}
+#endif
+
/**
* blk_queue_enter() - try to increase q->q_usage_counter
* @q: request queue pointer
@@ -440,12 +452,10 @@ int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags)
* responsible for ensuring that that counter is
* globally visible before the queue is unfrozen.
*/
- if ((pm && q->rpm_status != RPM_SUSPENDED) ||
- !blk_queue_pm_only(q)) {
+ if ((pm && !rq_suspended(q)) || !blk_queue_pm_only(q))
success = true;
- } else {
+ else
percpu_ref_put(&q->q_usage_counter);
- }
}
rcu_read_unlock();
--
2.29.2
--
Cheers,
Stephen Rothwell
On Tue, Dec 08, 2020 at 08:38:59PM +1100, Stephen Rothwell wrote:
> Hi all,
>
> On Tue, 8 Dec 2020 20:28:53 +1100 Stephen Rothwell <[email protected]> wrote:
> >
> > Hi all,
> >
> > After merging the scsi-mkp tree, today's linux-next build (sparc64
> > defconfig) failed like this:
> >
> > drivers/mtd/nand/raw/intel-nand-controller.c:17:10: fatal error: linux/mtd/nand_ecc.h: No such file or directory
> > 17 | #include <linux/mtd/nand_ecc.h>
> > | ^~~~~~~~~~~~~~~~~~~~~~
>
> Clearly, it did not fail like that :-)
>
> block/blk-core.c: In function 'blk_queue_enter':
> block/blk-core.c:443:18: error: 'struct request_queue' has no member named 'rpm_status'; did you mean 'stats'?
> if ((pm && q->rpm_status != RPM_SUSPENDED) ||
> ^~~~~~~~~~
> stats
>
> > Caused by commit
> >
> > 81a395cdc176 ("scsi: block: Do not accept any requests while suspended")
> >
> > # CONFIG_PM is not set
> >
> > I have applied the following patch:
> >
> > From: Stephen Rothwell <[email protected]>
> > Date: Tue, 8 Dec 2020 20:12:33 +1100
> > Subject: [PATCH] scsi: block: fix for "scsi: block: Do not accept any requests while suspended"
> >
> > Fixes: 81a395cdc176 ("scsi: block: Do not accept any requests while suspended")
> > Signed-off-by: Stephen Rothwell <[email protected]>
> > ---
> > block/blk-core.c | 18 ++++++++++++++----
> > 1 file changed, 14 insertions(+), 4 deletions(-)
> >
> > diff --git a/block/blk-core.c b/block/blk-core.c
> > index a71a5c9429d6..9c9aec1382be 100644
> > --- a/block/blk-core.c
> > +++ b/block/blk-core.c
> > @@ -421,6 +421,18 @@ void blk_cleanup_queue(struct request_queue *q)
> > }
> > EXPORT_SYMBOL(blk_cleanup_queue);
> >
> > +#ifdef CONFIG_PM
> > +static bool rq_suspended(struct request_queue *q)
> > +{
> > + return q->rpm_status == RPM_SUSPENDED;
> > +}
> > +#else
> > +static bool rq_suspended(struct request_queue *q)
> > +{
> > + return false;
> > +}
> > +#endif
> > +
> > /**
> > * blk_queue_enter() - try to increase q->q_usage_counter
> > * @q: request queue pointer
> > @@ -440,12 +452,10 @@ int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags)
> > * responsible for ensuring that that counter is
> > * globally visible before the queue is unfrozen.
> > */
> > - if ((pm && q->rpm_status != RPM_SUSPENDED) ||
> > - !blk_queue_pm_only(q)) {
> > + if ((pm && !rq_suspended(q)) || !blk_queue_pm_only(q))
> > success = true;
> > - } else {
> > + else
> > percpu_ref_put(&q->q_usage_counter);
> > - }
> > }
> > rcu_read_unlock();
Yes, that certainly is the proper fix. It's all to easy to miss these
issues that depend on your kernel configuration.
Bart, can you fold it into a new version of the patch?
Alan Stern
Hi all,
On Tue, 8 Dec 2020 20:28:53 +1100 Stephen Rothwell <[email protected]> wrote:
>
> Hi all,
>
> After merging the scsi-mkp tree, today's linux-next build (sparc64
> defconfig) failed like this:
>
> drivers/mtd/nand/raw/intel-nand-controller.c:17:10: fatal error: linux/mtd/nand_ecc.h: No such file or directory
> 17 | #include <linux/mtd/nand_ecc.h>
> | ^~~~~~~~~~~~~~~~~~~~~~
Clearly, it did not fail like that :-)
block/blk-core.c: In function 'blk_queue_enter':
block/blk-core.c:443:18: error: 'struct request_queue' has no member named 'rpm_status'; did you mean 'stats'?
if ((pm && q->rpm_status != RPM_SUSPENDED) ||
^~~~~~~~~~
stats
> Caused by commit
>
> 81a395cdc176 ("scsi: block: Do not accept any requests while suspended")
>
> # CONFIG_PM is not set
>
> I have applied the following patch:
>
> From: Stephen Rothwell <[email protected]>
> Date: Tue, 8 Dec 2020 20:12:33 +1100
> Subject: [PATCH] scsi: block: fix for "scsi: block: Do not accept any requests while suspended"
>
> Fixes: 81a395cdc176 ("scsi: block: Do not accept any requests while suspended")
> Signed-off-by: Stephen Rothwell <[email protected]>
> ---
> block/blk-core.c | 18 ++++++++++++++----
> 1 file changed, 14 insertions(+), 4 deletions(-)
>
> diff --git a/block/blk-core.c b/block/blk-core.c
> index a71a5c9429d6..9c9aec1382be 100644
> --- a/block/blk-core.c
> +++ b/block/blk-core.c
> @@ -421,6 +421,18 @@ void blk_cleanup_queue(struct request_queue *q)
> }
> EXPORT_SYMBOL(blk_cleanup_queue);
>
> +#ifdef CONFIG_PM
> +static bool rq_suspended(struct request_queue *q)
> +{
> + return q->rpm_status == RPM_SUSPENDED;
> +}
> +#else
> +static bool rq_suspended(struct request_queue *q)
> +{
> + return false;
> +}
> +#endif
> +
> /**
> * blk_queue_enter() - try to increase q->q_usage_counter
> * @q: request queue pointer
> @@ -440,12 +452,10 @@ int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags)
> * responsible for ensuring that that counter is
> * globally visible before the queue is unfrozen.
> */
> - if ((pm && q->rpm_status != RPM_SUSPENDED) ||
> - !blk_queue_pm_only(q)) {
> + if ((pm && !rq_suspended(q)) || !blk_queue_pm_only(q))
> success = true;
> - } else {
> + else
> percpu_ref_put(&q->q_usage_counter);
> - }
> }
> rcu_read_unlock();
>
> --
> 2.29.2
--
Cheers,
Stephen Rothwell
On 12/8/20 9:55 AM, Alan Stern wrote:
> Yes, that certainly is the proper fix. It's all to easy to miss these
> issues that depend on your kernel configuration.
>
> Bart, can you fold it into a new version of the patch?
Sure, I will do that.
Thanks,
Bart.