[PATCH] Freezer: Fix JFFS2 garbage collector freezing issue (was: Re: JFFS2 garbage collection threads not freezing?)

[PATCH] Freezer: Fix JFFS2 garbage collector freezing issue (was: Re: JFFS2 garbage collection threads not freezing?)

[Rafael J. Wysocki]

On Friday, 30 of November 2007, Pete MacKay wrote:
> We've ported from 2.6.18 to 2.6.24-rc3 on a pxa270-based machine and can
> no longer enter 'standby' or 'mem' states.  I've stripped down the kernel
> and talked with the freezer maintainer and we believe the jffs2 garbage
> collection threads may not be calling try_to_freeze() or marking
> themselves properly.  (We have two mounted jffs2 file systems - mtd2 and
> mtd3).  The following is some cleaned-up output:
> 
> Freezing remaining freezable tasks ...
> Freezing of tasks failed after 20.00 seconds (2 tasks refusing to freeze):
>   task                PC stack   pid father
> init          D c0244b90     0     1      0
> kthreadd      S c0244b90     0     2      0
> ksoftirqd/0   S c0244b90     0     3      2
> events/0      S c0244b90     0     4      2
> khelper       S c0244b90     0     5      2
> kblockd/0     S c0244b90     0    51      2
> ksuspend_usbd D c0244b90     0    59      2
> khubd         D c0244b90     0    64      2
> kseriod       D c0244b90     0    67      2
> pdflush       D c0244b90     0    90      2
> pdflush       D c0244b90     0    91      2
> kswapd0       D c0244b90     0    92      2
> aio/0         S c0244b90     0    93      2
> mtdblockd     S c0244b90     0   192      2
> udevd         D c0244b90     0   619      1
> jffs2_gcd_mtd R running      0  1533      2
> jffs2_gcd_mtd R running      0  1598      2
> syslogd       D c0244b90     0  2248      1
> klogd         D c0244b90     0  2250      1
> hcid          D c0244b90     0  2273      1
> sdpd          D c0244b90     0  2277      1
> hidd          D c0244b90     0  2281      1
> sh            D c0244b90     0  2292      1
> getty         D c0244b90     0  2293      1
> echo          R running      0  2303   2292
>  jffs2_gcd_mtd2
>  jffs2_gcd_mtd3
> 
> Restarting tasks ... done.

I think that the appended patch is necessary to fix this problem.

Greetings,
Rafael

---
Subject: Freezer: Fix JFFS2 garbage collector freezing issue
From: Rafael J. Wysocki <rjw@sisk.pl>

Fix breakage caused by commit d5d8c5976d6adeddb8208c240460411e2198b393
"freezer: do not send signals to kernel threads" in
jffs2_garbage_collect_thread() that assumed it would be sent signals
by the freezer.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Pete MacKay <armlinux@architechnical.net>
Cc: Andrew Morton <akpm@linux-foundation.org>
---
 fs/jffs2/background.c |    8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

Index: linux-2.6/fs/jffs2/background.c
===================================================================
--- linux-2.6.orig/fs/jffs2/background.c
+++ linux-2.6/fs/jffs2/background.c
@@ -103,15 +103,17 @@ static int jffs2_garbage_collect_thread(
 		   get there first. */
 		yield();
 
+		/* If system suspend is in progress, go to the refrigerator and
+		   start again when the suspend is done */
+		if (try_to_freeze())
+			goto again;
+
 		/* Put_super will send a SIGKILL and then wait on the sem.
 		 */
 		while (signal_pending(current)) {
 			siginfo_t info;
 			unsigned long signr;
 
-			if (try_to_freeze())
-				goto again;
-
 			signr = dequeue_signal_lock(current, &current->blocked, &info);
 
 			switch(signr) {

Re: [PATCH] Freezer: Fix JFFS2 garbage collector freezing issue (was: Re: JFFS2 garbage collection threads not freezing?)

[Rafael J. Wysocki]

On Saturday, 1 of December 2007, Rafael J. Wysocki wrote:
> On Friday, 30 of November 2007, Pete MacKay wrote:
[--snip--]
> ---
> Subject: Freezer: Fix JFFS2 garbage collector freezing issue
> From: Rafael J. Wysocki <rjw@sisk.pl>
> 
> Fix breakage caused by commit d5d8c5976d6adeddb8208c240460411e2198b393
> "freezer: do not send signals to kernel threads" in
> jffs2_garbage_collect_thread() that assumed it would be sent signals
> by the freezer.
> 
> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
> Cc: Pete MacKay <armlinux@architechnical.net>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> ---
>  fs/jffs2/background.c |    8 +++++---
>  1 file changed, 5 insertions(+), 3 deletions(-)
> 
> Index: linux-2.6/fs/jffs2/background.c
> ===================================================================
> --- linux-2.6.orig/fs/jffs2/background.c
> +++ linux-2.6/fs/jffs2/background.c
> @@ -103,15 +103,17 @@ static int jffs2_garbage_collect_thread(
>  		   get there first. */
>  		yield();
>  
> +		/* If system suspend is in progress, go to the refrigerator and
> +		   start again when the suspend is done */
> +		if (try_to_freeze())
> +			goto again;
> +

This still has the problem that, if the freeze request comes exactly here, the
loop below will not allow us to freeze.

>  		/* Put_super will send a SIGKILL and then wait on the sem.
>  		 */
>  		while (signal_pending(current)) {
>  			siginfo_t info;
>  			unsigned long signr;
>  
> -			if (try_to_freeze())
> -				goto again;
> -
>  			signr = dequeue_signal_lock(current, &current->blocked, &info);
>  
>  			switch(signr) {
> --

I'll send another patch for this in a while.

Greetings,
Rafael