[PATCH 1/1]: Thaws refrigerated bdi flusher threads before invoking kthread_stop on them

[PATCH 1/1]: Thaws refrigerated bdi flusher threads before invoking kthread_stop on them

[Romit Dasgupta]

Kicks out frozen bdi flusher task out of the refrigerator when the flusher task
needs to exit.
Signed-off-by: Romit Dasgupta <romit@ti.com>
---
diff --git a/mm/backing-dev.c b/mm/backing-dev.c
index 5a37e20..c757b05 100644
--- a/mm/backing-dev.c
+++ b/mm/backing-dev.c
@@ -606,8 +606,11 @@ static void bdi_wb_shutdown(struct backing_dev_info *bdi)
 	 * Finally, kill the kernel threads. We don't need to be RCU
 	 * safe anymore, since the bdi is gone from visibility.
 	 */
-	list_for_each_entry(wb, &bdi->wb_list, list)
+	list_for_each_entry(wb, &bdi->wb_list, list) {
+		if (unlikely(frozen(wb->task)))
+			wb->task->flags &= ~PF_FROZEN;
 		kthread_stop(wb->task);
+	}
 }
 
 void bdi_unregister(struct backing_dev_info *bdi)

Re: [PATCH 1/1]: Thaws refrigerated bdi flusher threads before invoking kthread_stop on them

[Pavel Machek]

On Wed 2009-11-11 14:00:16, Romit Dasgupta wrote:
> Kicks out frozen bdi flusher task out of the refrigerator when the flusher task
> needs to exit.


> Signed-off-by: Romit Dasgupta <romit@ti.com>

Ok, its slightly "interesting", but better than modifying common
code. Looks ok to me.

ACK.
								Pavel

> ---
> diff --git a/mm/backing-dev.c b/mm/backing-dev.c
> index 5a37e20..c757b05 100644
> --- a/mm/backing-dev.c
> +++ b/mm/backing-dev.c
> @@ -606,8 +606,11 @@ static void bdi_wb_shutdown(struct backing_dev_info *bdi)
>  	 * Finally, kill the kernel threads. We don't need to be RCU
>  	 * safe anymore, since the bdi is gone from visibility.
>  	 */
> -	list_for_each_entry(wb, &bdi->wb_list, list)
> +	list_for_each_entry(wb, &bdi->wb_list, list) {
> +		if (unlikely(frozen(wb->task)))
> +			wb->task->flags &= ~PF_FROZEN;
>  		kthread_stop(wb->task);
> +	}
>  }
>  
>  void bdi_unregister(struct backing_dev_info *bdi)
> 

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [PATCH 1/1]: Thaws refrigerated bdi flusher threads before invoking kthread_stop on them

[Rafael J. Wysocki]

On Wednesday 11 November 2009, Pavel Machek wrote:
> On Wed 2009-11-11 14:00:16, Romit Dasgupta wrote:
> > Kicks out frozen bdi flusher task out of the refrigerator when the flusher task
> > needs to exit.
> 
> 
> > Signed-off-by: Romit Dasgupta <romit@ti.com>
> 
> Ok, its slightly "interesting", but better than modifying common
> code. Looks ok to me.
> 
> ACK.

Agreed.

Jens, any objections?

> > ---
> > diff --git a/mm/backing-dev.c b/mm/backing-dev.c
> > index 5a37e20..c757b05 100644
> > --- a/mm/backing-dev.c
> > +++ b/mm/backing-dev.c
> > @@ -606,8 +606,11 @@ static void bdi_wb_shutdown(struct backing_dev_info *bdi)
> >  	 * Finally, kill the kernel threads. We don't need to be RCU
> >  	 * safe anymore, since the bdi is gone from visibility.
> >  	 */
> > -	list_for_each_entry(wb, &bdi->wb_list, list)
> > +	list_for_each_entry(wb, &bdi->wb_list, list) {
> > +		if (unlikely(frozen(wb->task)))
> > +			wb->task->flags &= ~PF_FROZEN;
> >  		kthread_stop(wb->task);
> > +	}
> >  }
> >  
> >  void bdi_unregister(struct backing_dev_info *bdi)
> > 

For completness, below is the information from the Romit's introductory message
(Romit, I really think that should go into the chagelog):

"Few days back I started facing problems during system suspend with
MMC, SD card installed. I will restate how to reproduce the problem:

1) Mount a file system from MMC/SD card.
2) Unmount the file system. This creates a flusher task.
3) Attempt suspend to RAM. System is unresponsive.

This is because the bdi flusher thread is already in the refrigerator
and will remain so until it is thawed. The MMC driver suspend routine
ultimately will issue a 'kthread_stop' on the bdi flusher thread and
will block until the flusher thread is exited. Since the bdi flusher
thread is in the refrigerator it never cleans up until thawed.

Enabling khungtaskd gave the following dump: (the dump wraps beyond 80
cols).

INFO: task sh:387 blocked for more than 120 seconds.
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
sh            D c027e86c     0   387      1 0x00000000
[<c027e86c>] (schedule+0x2e0/0x36c) from [<c027ee78>] (schedule_timeout+0x18/0x1ec)
[<c027ee78>] (schedule_timeout+0x18/0x1ec) from [<c027ed10>] (wait_for_common+0xe0/0x198)
[<c027ed10>] (wait_for_common+0xe0/0x198) from [<c0063dd4>] (kthread_stop+0x44/0x78)
[<c0063dd4>] (kthread_stop+0x44/0x78) from [<c0090938>] (bdi_unregister+0x64/0xa4)
[<c0090938>] (bdi_unregister+0x64/0xa4) from [<c0173958>] (unlink_gendisk+0x20/0x3c)
[<c0173958>] (unlink_gendisk+0x20/0x3c) from [<c00f2338>] (del_gendisk+0x84/0xb4)
[<c00f2338>] (del_gendisk+0x84/0xb4) from [<c01e6840>] (mmc_blk_remove+0x24/0x44)
[<c01e6840>] (mmc_blk_remove+0x24/0x44) from [<c01e14f0>] (mmc_bus_remove+0x18/0x20)
[<c01e14f0>] (mmc_bus_remove+0x18/0x20) from [<c01af6ac>] (__device_release_driver+0x64/0xa4)
[<c01af6ac>] (__device_release_driver+0x64/0xa4) from [<c01af7e4>] (device_release_driver+0x1c/0x28)
[<c01af7e4>] (device_release_driver+0x1c/0x28) from [<c01aed5c>] (bus_remove_device+0x7c/0x90)
[<c01aed5c>] (bus_remove_device+0x7c/0x90) from [<c01ad538>] (device_del+0x110/0x160)
[<c01ad538>] (device_del+0x110/0x160) from [<c01e15a8>] (mmc_remove_card+0x50/0x64)
[<c01e15a8>] (mmc_remove_card+0x50/0x64) from [<c01e2ea4>] (mmc_sd_remove+0x24/0x30)
[<c01e2ea4>] (mmc_sd_remove+0x24/0x30) from [<c01e0dcc>] (mmc_suspend_host+0x110/0x1a8)
[<c01e0dcc>] (mmc_suspend_host+0x110/0x1a8) from [<c01e7d04>] (omap_hsmmc_suspend+0x74/0x104)
[<c01e7d04>] (omap_hsmmc_suspend+0x74/0x104) from [<c01b08e8>] (platform_pm_suspend+0x50/0x5c)
[<c01b08e8>] (platform_pm_suspend+0x50/0x5c) from [<c01b27f0>] (pm_op+0x30/0x74)
[<c01b27f0>] (pm_op+0x30/0x74) from [<c01b2ec8>] (dpm_suspend_start+0x3b4/0x518)
[<c01b2ec8>] (dpm_suspend_start+0x3b4/0x518) from [<c0078b20>] (suspend_devices_and_enter+0x3c/0x1c4)
[<c0078b20>] (suspend_devices_and_enter+0x3c/0x1c4) from [<c0078d88>] (enter_state+0xe0/0x138)
[<c0078d88>] (enter_state+0xe0/0x138) from [<c0078444>] (state_store+0x94/0xbc)
[<c0078444>] (state_store+0x94/0xbc) from [<c017e124>] (kobj_attr_store+0x18/0x1c)
[<c017e124>] (kobj_attr_store+0x18/0x1c) from [<c00f3a08>] (sysfs_write_file+0x108/0x13c)
[<c00f3a08>] (sysfs_write_file+0x108/0x13c) from [<c00a76b8>] (vfs_write+0xac/0x154)
[<c00a76b8>] (vfs_write+0xac/0x154) from [<c00a780c>] (sys_write+0x3c/0x68)
[<c00a780c>] (sys_write+0x3c/0x68) from [<c0025e60>] (ret_fast_syscall+0x0/0x2c)

Earlier I had sent a patch to thaw any refrigerated kernel thread when
some active thread has invoked 'kthread_stop' on it. This was done with
the assumption that all such kernel threads should invoke
'kthread_should_stop' after 'try_to_freeze' and exit if necessary. It
looks there are some kernel threads which do not follow this. With that
in mind I am sending a different patch to fix the above issue (in the
next mail)."

Rafael

Re: [PATCH 1/1]: Thaws refrigerated bdi flusher threads before invoking kthread_stop on them

[Romit Dasgupta]

Hello Rafael, 
              As suggested I have added the relevant information in the
changelog. The patch is below:
> For completness, below is the information from the Romit's introductory message
> (Romit, I really think that should go into the chagelog):

Kicks out frozen bdi flusher task out of the refrigerator when the said task
needs to exit.
Steps to reproduce this.
1) Mount a file system from MMC/SD card.
2) Unmount the file system. This creates a flusher task.
3) Attempt suspend to RAM. System is unresponsive.

This is because the bdi flusher thread is already in the refrigerator and will
remain so until it is thawed. The MMC driver suspend routine ultimately will
issue a 'kthread_stop' on the bdi flusher thread and will block until the
flusher thread is exited. Since the bdi flusher thread is in the refrigerator
it never cleans up until thawed.

Signed-off-by: Romit Dasgupta <romit@ti.com>
---
diff --git a/mm/backing-dev.c b/mm/backing-dev.c
index 5a37e20..c757b05 100644
--- a/mm/backing-dev.c
+++ b/mm/backing-dev.c
@@ -606,8 +606,11 @@ static void bdi_wb_shutdown(struct backing_dev_info *bdi)
 	 * Finally, kill the kernel threads. We don't need to be RCU
 	 * safe anymore, since the bdi is gone from visibility.
 	 */
-	list_for_each_entry(wb, &bdi->wb_list, list)
+	list_for_each_entry(wb, &bdi->wb_list, list) {
+		if (unlikely(frozen(wb->task)))
+			wb->task->flags &= ~PF_FROZEN;
 		kthread_stop(wb->task);
+	}
 }
 
 void bdi_unregister(struct backing_dev_info *bdi)

Re: [PATCH 1/1]: Thaws refrigerated bdi flusher threads before invoking kthread_stop on them

[Rafael J. Wysocki]

On Wednesday 11 November 2009, Romit Dasgupta wrote:
> Hello Rafael, 
>               As suggested I have added the relevant information in the
> changelog. The patch is below:
> > For completness, below is the information from the Romit's introductory message
> > (Romit, I really think that should go into the chagelog):
> 
> Kicks out frozen bdi flusher task out of the refrigerator when the said task
> needs to exit.
> Steps to reproduce this.
> 1) Mount a file system from MMC/SD card.
> 2) Unmount the file system. This creates a flusher task.
> 3) Attempt suspend to RAM. System is unresponsive.
> 
> This is because the bdi flusher thread is already in the refrigerator and will
> remain so until it is thawed. The MMC driver suspend routine ultimately will
> issue a 'kthread_stop' on the bdi flusher thread and will block until the
> flusher thread is exited. Since the bdi flusher thread is in the refrigerator
> it never cleans up until thawed.
> 
> Signed-off-by: Romit Dasgupta <romit@ti.com>

Thanks!  I'd still like to know that this change is fine with Jens, though.

Rafael


> ---
> diff --git a/mm/backing-dev.c b/mm/backing-dev.c
> index 5a37e20..c757b05 100644
> --- a/mm/backing-dev.c
> +++ b/mm/backing-dev.c
> @@ -606,8 +606,11 @@ static void bdi_wb_shutdown(struct backing_dev_info *bdi)
>  	 * Finally, kill the kernel threads. We don't need to be RCU
>  	 * safe anymore, since the bdi is gone from visibility.
>  	 */
> -	list_for_each_entry(wb, &bdi->wb_list, list)
> +	list_for_each_entry(wb, &bdi->wb_list, list) {
> +		if (unlikely(frozen(wb->task)))
> +			wb->task->flags &= ~PF_FROZEN;
>  		kthread_stop(wb->task);
> +	}
>  }
>  
>  void bdi_unregister(struct backing_dev_info *bdi)

Re: [PATCH 1/1]: Thaws refrigerated bdi flusher threads before invoking kthread_stop on them

[Jens Axboe]

On Wed, Nov 11 2009, Rafael J. Wysocki wrote:
> On Wednesday 11 November 2009, Pavel Machek wrote:
> > On Wed 2009-11-11 14:00:16, Romit Dasgupta wrote:
> > > Kicks out frozen bdi flusher task out of the refrigerator when the flusher task
> > > needs to exit.
> > 
> > 
> > > Signed-off-by: Romit Dasgupta <romit@ti.com>
> > 
> > Ok, its slightly "interesting", but better than modifying common
> > code. Looks ok to me.
> > 
> > ACK.
> 
> Agreed.
> 
> Jens, any objections?

Nope, looks fine to me. Though I'd probably prefer just doing an
unconditional PF_FROZEN clear.

/*
 * Force unfreeze of the bdi threads before stopping it, since otherwise
 * it would never exit if it is stuck in the refrigerator.
 */
list_for_each_entry(wb, &bdi->wb_list, list) {
	wb->task->flags &= ~PF_FROZEN;
	kthread_stop(wb->task);
}

And the comment too, it's not enough to stuff this into the commit.

-- 
Jens Axboe

Re: [PATCH 1/1]: Thaws refrigerated bdi flusher threads before invoking kthread_stop on them

[Rafael J. Wysocki]

On Wednesday 11 November 2009, Jens Axboe wrote:
> On Wed, Nov 11 2009, Rafael J. Wysocki wrote:
> > On Wednesday 11 November 2009, Pavel Machek wrote:
> > > On Wed 2009-11-11 14:00:16, Romit Dasgupta wrote:
> > > > Kicks out frozen bdi flusher task out of the refrigerator when the flusher task
> > > > needs to exit.
> > > 
> > > 
> > > > Signed-off-by: Romit Dasgupta <romit@ti.com>
> > > 
> > > Ok, its slightly "interesting", but better than modifying common
> > > code. Looks ok to me.
> > > 
> > > ACK.
> > 
> > Agreed.
> > 
> > Jens, any objections?
> 
> Nope, looks fine to me. Though I'd probably prefer just doing an
> unconditional PF_FROZEN clear.
> 
> /*
>  * Force unfreeze of the bdi threads before stopping it, since otherwise
>  * it would never exit if it is stuck in the refrigerator.
>  */
> list_for_each_entry(wb, &bdi->wb_list, list) {
> 	wb->task->flags &= ~PF_FROZEN;
> 	kthread_stop(wb->task);
> }
> 
> And the comment too, it's not enough to stuff this into the commit.

The last version had the changelog fixed.

Romit, please rework as suggested by Jens and resubmit.

Thanks,
Rafael

Re: [PATCH 1/1]: Thaws refrigerated bdi flusher threads before invoking kthread_stop on them

[Romit Dasgupta]

Resubmitting after incorporationg the suggestions:
> 
> Romit, please rework as suggested by Jens and resubmit.

Unfreezes the bdi flusher task when the said task needs to exit.

Steps to reproduce this.
1) Mount a file system from MMC/SD card.
2) Unmount the file system. This creates a flusher task.
3) Attempt suspend to RAM. System is unresponsive.

This is because the bdi flusher thread is already in the refrigerator and will
remain so until it is thawed. The MMC driver suspend routine call stack will
ultimately issue a 'kthread_stop' on the bdi flusher thread and will block
until the flusher thread is exited. Since the bdi flusher thread is in the
refrigerator it never cleans up until thawed.

Signed-off-by: Romit Dasgupta <romit@ti.com>
---
diff --git a/mm/backing-dev.c b/mm/backing-dev.c
index 5a37e20..5a9ab6f 100644
--- a/mm/backing-dev.c
+++ b/mm/backing-dev.c
@@ -603,11 +603,15 @@ static void bdi_wb_shutdown(struct backing_dev_info *bdi)
 	bdi_remove_from_list(bdi);
 
 	/*
+	 * Force unfreeze of bdi threads before stopping it, otherwise
+	 * it would never exit if it is stuck in the refrigerator.
 	 * Finally, kill the kernel threads. We don't need to be RCU
 	 * safe anymore, since the bdi is gone from visibility.
 	 */
-	list_for_each_entry(wb, &bdi->wb_list, list)
+	list_for_each_entry(wb, &bdi->wb_list, list) {
+		wb->task->flags &= ~PF_FROZEN;
 		kthread_stop(wb->task);
+	}
 }
 
 void bdi_unregister(struct backing_dev_info *bdi)

Re: [PATCH 1/1]: Thaws refrigerated bdi flusher threads before invoking kthread_stop on them

[Jens Axboe]

On Thu, Nov 12 2009, Romit Dasgupta wrote:
> Resubmitting after incorporationg the suggestions:
> > 
> > Romit, please rework as suggested by Jens and resubmit.
> 
> Unfreezes the bdi flusher task when the said task needs to exit.

Thanks, this looks more palatable. I will merge it and make sure it goes
upstream soon.

-- 
Jens Axboe