I/O errors in -test7-mm1 tree on ia64

DISCUSSION Jesse Barnes 3 participants 6 messages Oct 16, 2003 – Oct 17, 2003

Share & tools

Jesse Barnes

author

2003-10-16 18:55 UTC #0 lore raw

I just compiled and booted -test7-mm1 on an Altix (I needed to create an empty include/asm-ia64/kgdb.h to get tg3.c to compile), but I'm getting the following errors after starting init:

NET: Registered protocol family 17
XFS mounting filesystem sdb6
VFS: Mounted root (xfs filesystem) readonly.
Freeing unused kernel memory: 432kB freed
INIT: version 2.78 booting
Welcome to SGI ProPack(TM) for Linux(R)
Press 'I' to enter interactive startup.

Mounting proc filesystem:  [  OK  ]
Configuring kernel parameters:  [  OK  ]
Setting clock  (utc): Thu Oct 16 11:22:35 PDT 2003 [  OK  ]

Setting hostname margin.engr.sgi.com: [ OK ]
Checking root filesystem
[/sbin/fsck.xfs (1) -- /] fsck.xfs -a /dev/sdb6
[ OK ]
Remounting root filesystem in read-write mode: [ OK ]
I/O error in filesystem ("sdb6") meta-data dev sdb6 block 0x7d090e ("xlog_iodone") error 5 buf count 32768
Filesystem "sdb6": Log I/O Error Detected. Shutting down filesystem: sdb6 Please umount the filesystem, and rectify the problem(s)
can't link lock file /etc/mtab~: Input/output error (use -n flag to override)
INIT: can't open(/etc/ioctl.save, O_WRONLY): Input/output error
INIT: can't open(/etc/ioctl.save, O_WRONLY): Input/output error
INIT: Entering runlevel: 3
INIT: cannot execute "/etc/rc.d/rc"
INIT: cannot execute "/sbin/agetty"
INIT: cannot execute "/sbin/agetty"

I don't see this when using Linus' BK tree as of a few minutes ago, and the only changes I've made are adding the kgdb.h for ia64 and adding in the Altix console driver. Any ideas? I'll try reverting some patches and looking around a bit more.

Thanks,
Jesse

William Lee Irwin III

2003-10-16 19:46 UTC #1 lore raw

Quoted 4 lines · Jesse Barnes

I don't see this when using Linus' BK tree as of a few minutes ago, and the only changes I've made are adding the kgdb.h for ia64 and adding in the Altix console driver. Any ideas? I'll try reverting some patches and looking around a bit more.

Well, the first thing to try is backing out invalidate_inodes-speedup.patch

-- wli

Jesse Barnes

author

2003-10-16 20:47 UTC #2 lore raw

Quoted 7 lines · William Lee Irwin III

Quoted 4 lines · Jesse Barnes
I don't see this when using Linus' BK tree as of a few minutes ago, and the only changes I've made are adding the kgdb.h for ia64 and adding in the Altix console driver. Any ideas? I'll try reverting some patches and looking around a bit more.
Well, the first thing to try is backing out invalidate_inodes-speedup.patch

That didn't seem to help. Got the same errors.

Jesse

Jens Axboe

2003-10-17 10:12 UTC #3 lore raw

Quoted 10 lines · Jesse Barnes

Quoted 7 lines · William Lee Irwin III
Quoted 4 lines · Jesse Barnes
I don't see this when using Linus' BK tree as of a few minutes ago, and the only changes I've made are adding the kgdb.h for ia64 and adding in the Altix console driver. Any ideas? I'll try reverting some patches and looking around a bit more.
Well, the first thing to try is backing out invalidate_inodes-speedup.patch
That didn't seem to help. Got the same errors.

xfs/pagebuf/page_buf.c~ renamed from xfs/pagebuf/page_buf.c +0 -2

@@ -1406,10 +1406,8 @@
 		int cmd = WRITE;
 		if (pb->pb_flags & PBF_READ)
 			cmd = READ;
-#if 0
 		else if (pb->pb_flags & PBF_FLUSH)
 			cmd = WRITESYNC;
-#endif
 		submit_bio(cmd, bio);
 		if (size)
 			goto next_chunk;

it was a mistake to enable barriers unconditionally on XFS when it has no fallback logic. If you apply the above on test7-mm1, it should work fine for you.

Signature · 1 line

Jens Axboe

Jens Axboe

2003-10-17 10:14 UTC #4 lore raw

Quoted 26 lines · Jens Axboe

Quoted 10 lines · Jesse Barnes
Quoted 7 lines · William Lee Irwin III
Quoted 4 lines · Jesse Barnes
I don't see this when using Linus' BK tree as of a few minutes ago, and the only changes I've made are adding the kgdb.h for ia64 and adding in the Altix console driver. Any ideas? I'll try reverting some patches and looking around a bit more.
Well, the first thing to try is backing out invalidate_inodes-speedup.patch
That didn't seem to help. Got the same errors.
xfs/pagebuf/page_buf.c~ renamed from xfs/pagebuf/page_buf.c +0 -2
@@ -1406,10 +1406,8 @@
 		int cmd = WRITE;
 		if (pb->pb_flags & PBF_READ)
 			cmd = READ;
-#if 0
 		else if (pb->pb_flags & PBF_FLUSH)
 			cmd = WRITESYNC;
-#endif
 		submit_bio(cmd, bio);
 		if (size)
 			goto next_chunk;

Quoted 3 lines

it was a mistake to enable barriers unconditionally on XFS when it has no fallback logic. If you apply the above on test7-mm1, it should work fine for you.

Patch is reversed, you need to use -R. Funky.

Signature · 1 line

Jens Axboe

Jesse Barnes

author

2003-10-17 22:16 UTC #5 lore raw

Quoted 14 lines · Jens Axboe

Quoted 14 lines

xfs/pagebuf/page_buf.c~ renamed from xfs/pagebuf/page_buf.c +0 -2

@@ -1406,10 +1406,8 @@
 		int cmd = WRITE;
 		if (pb->pb_flags & PBF_READ)
 			cmd = READ;
-#if 0
 		else if (pb->pb_flags & PBF_FLUSH)
 			cmd = WRITESYNC;
-#endif
 		submit_bio(cmd, bio);
 		if (size)
 			goto next_chunk;

Quoted 5 lines

Quoted 3 lines
it was a mistake to enable barriers unconditionally on XFS when it has no fallback logic. If you apply the above on test7-mm1, it should work fine for you.
Patch is reversed, you need to use -R. Funky.

That did the trick. Thanks.

Jesse