Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60

Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60

[Christian Kujau]

When writing to a newly created btrfs (vanilla 2.6.33-rc6, sparc64) the 
following messages are printed:

[28617.650231] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
[28617.745783] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
[28654.589492] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
[28654.685036] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
[28689.595394] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]

...and so on. I've seen these "unaligned access at TPC" on sparc64 
before[0], but they were not btrfs related. Btw, *reading* does not cause 
these messages.

At one point and when lots of I/O was going on, the following appeared once, inbetween
the other messages:

[31419.847052] Kernel unaligned access at TPC[102a3f18] btrfs_csum_final+0x38/0x60 [btrfs]
[31447.243224] Kernel unaligned access at TPC[102a3f18] btrfs_csum_final+0x38/0x60 [btrfs]
[31447.338727] Kernel unaligned access at TPC[102a3f18] btrfs_csum_final+0x38/0x60 [btrfs]
[33888.613296] btrfs: loop0 checksum verify failed on 39084032 wanted EAFC29EA found BBFFFFFF level 0
[34421.508271] Kernel unaligned access at TPC[102a3f18] btrfs_csum_final+0x38/0x60 [btrfs]
[34421.603872] Kernel unaligned access at TPC[102a3f18] btrfs_csum_final+0x38/0x60 [btrfs]
[34452.965546] Kernel unaligned access at TPC[102a3f18] btrfs_csum_final+0x38/0x60 [btrfs]

No messages from the backing device have been reported though, but I'm 
not so sure about the system's RAM, so it might be a hardware issue after 
all. I was not able to reproduce the "checksum verify failed" yet.

Full dmesg and .config: http://nerdbynature.de/bits/2.6.33-rc6/btrfs/

Christian.


[0] http://linux.derkeiler.com/Mailing-Lists/Kernel/2009-02/msg01327.html
-- 
BOFH excuse #369:

Virus transmitted from computer to sysadmins.

Re: Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60

[David Miller]

From: Christian Kujau <lists@nerdbynature.de>
Date: Fri, 5 Feb 2010 10:36:52 -0800 (PST)

> When writing to a newly created btrfs (vanilla 2.6.33-rc6, sparc64) the 
> following messages are printed:
> 
> [28617.650231] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
> [28617.745783] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
> [28654.589492] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
> [28654.685036] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
> [28689.595394] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]

Can you rerun your test with the following patch applied?

It will obtain more information for the btrfs developers.

Thanks!

diff --git a/arch/sparc/kernel/unaligned_64.c b/arch/sparc/kernel/unaligned_64.c
index 378ca82..cbde2ea 100644
--- a/arch/sparc/kernel/unaligned_64.c
+++ b/arch/sparc/kernel/unaligned_64.c
@@ -283,6 +283,9 @@ static void log_unaligned(struct pt_regs *regs)
 		count++;
 		printk("Kernel unaligned access at TPC[%lx] %pS\n",
 		       regs->tpc, (void *) regs->tpc);
+		printk("Caller [%lx:%pS]\n",
+		       regs->u_regs[UREG_RETPC],
+		       (void *) regs->u_regs[UREG_RETPC]);
 	}
 }
 

Re: Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60

[Christian Kujau]

On Fri, 5 Feb 2010 at 12:01, David Miller wrote:
> Can you rerun your test with the following patch applied?
> It will obtain more information for the btrfs developers.

Thanks, David! Here it is:

[ 1861.965178] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
[ 1862.060546] Caller [100a6044:crc32c+0x44/0x80 [libcrc32c]]
[ 1862.126652] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
[ 1862.221928] Caller [100a6044:crc32c+0x44/0x80 [libcrc32c]]

It's always libcrc32c and the numbers stay the same too. Full dmesg here:
http://nerdbynature.de/bits/2.6.33-rc6/btrfs/

HTH,
Christian.
-- 
BOFH excuse #37:

heavy gravity fluctuation, move computer to floor rapidly

Re: Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60

[David Miller]

From: Christian Kujau <lists@nerdbynature.de>
Date: Fri, 5 Feb 2010 21:13:00 -0800 (PST)

> On Fri, 5 Feb 2010 at 12:01, David Miller wrote:
>> Can you rerun your test with the following patch applied?
>> It will obtain more information for the btrfs developers.
> 
> Thanks, David! Here it is:
> 
> [ 1861.965178] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
> [ 1862.060546] Caller [100a6044:crc32c+0x44/0x80 [libcrc32c]]
> [ 1862.126652] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
> [ 1862.221928] Caller [100a6044:crc32c+0x44/0x80 [libcrc32c]]
> 
> It's always libcrc32c and the numbers stay the same too. Full dmesg here:
> http://nerdbynature.de/bits/2.6.33-rc6/btrfs/

My debugging patch didn't work correctly.

Can you try using this one instead?

Thanks!

diff --git a/arch/sparc/kernel/unaligned_64.c b/arch/sparc/kernel/unaligned_64.c
index 378ca82..cbde2ea 100644
--- a/arch/sparc/kernel/unaligned_64.c
+++ b/arch/sparc/kernel/unaligned_64.c
@@ -283,6 +283,9 @@ static void log_unaligned(struct pt_regs *regs)
 		count++;
 		printk("Kernel unaligned access at TPC[%lx] %pS\n",
 		       regs->tpc, (void *) regs->tpc);
+		printk("Caller [%lx:%pS]\n",
+		       regs->tnpc,
+		       (void *) regs->tnpc);
 	}
 }
 

Re: Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60

[Christian Kujau]

On Fri, 5 Feb 2010 at 21:32, David Miller wrote:
> My debugging patch didn't work correctly.
> Can you try using this one instead?

Hm, now it looks like this, but I don't know how it'd reveal more 
information:

[  210.707051] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
[  210.802236] Caller [10101f1c:btrfs_csum_final+0x3c/0x60 [btrfs]]
[  210.874620] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
[  210.969899] Caller [10101f1c:btrfs_csum_final+0x3c/0x60 [btrfs]]
[  228.724982] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
[  228.820220] Caller [10101f1c:btrfs_csum_final+0x3c/0x60 [btrfs]]
[  228.892286] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
[  228.987851] Caller [10101f1c:btrfs_csum_final+0x3c/0x60 [btrfs]]

I'm open for more patches :-)

Christian.
-- 
BOFH excuse #277:

Your Flux Capacitor has gone bad.

Re: Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60

[David Miller]

From: Christian Kujau <lists@nerdbynature.de>
Date: Fri, 5 Feb 2010 22:28:47 -0800 (PST)

> Hm, now it looks like this, but I don't know how it'd reveal more 
> information:
> 
> [  210.707051] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
> [  210.802236] Caller [10101f1c:btrfs_csum_final+0x3c/0x60 [btrfs]]
> [  210.874620] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
> [  210.969899] Caller [10101f1c:btrfs_csum_final+0x3c/0x60 [btrfs]]
> [  228.724982] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
> [  228.820220] Caller [10101f1c:btrfs_csum_final+0x3c/0x60 [btrfs]]
> [  228.892286] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
> [  228.987851] Caller [10101f1c:btrfs_csum_final+0x3c/0x60 [btrfs]]
> 
> I'm open for more patches :-)

Confusing for me too.  I'll try to think about this some more.

Re: Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60

[Chris Mason]

On Fri, Feb 05, 2010 at 10:43:10PM -0800, David Miller wrote:
> From: Christian Kujau <lists@nerdbynature.de>
> Date: Fri, 5 Feb 2010 22:28:47 -0800 (PST)
> 
> > Hm, now it looks like this, but I don't know how it'd reveal more 
> > information:
> > 
> > [  210.707051] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
> > [  210.802236] Caller [10101f1c:btrfs_csum_final+0x3c/0x60 [btrfs]]
> > [  210.874620] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
> > [  210.969899] Caller [10101f1c:btrfs_csum_final+0x3c/0x60 [btrfs]]
> > [  228.724982] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
> > [  228.820220] Caller [10101f1c:btrfs_csum_final+0x3c/0x60 [btrfs]]
> > [  228.892286] Kernel unaligned access at TPC[10101f18] btrfs_csum_final+0x38/0x60 [btrfs]
> > [  228.987851] Caller [10101f1c:btrfs_csum_final+0x3c/0x60 [btrfs]]
> > 
> > I'm open for more patches :-)
> 
> Confusing for me too.  I'll try to think about this some more.

It is probably the directory code.  I switched that to crc32c since
you tried things.

-chris