Re: sysfs Kernel BUG when RAID bitmap file has IO errors

[Andrew Morton]

On Wed, 12 Mar 2008 10:51:38 +0100
Tomasz Chmielewski <mangoo@xxxxxxxx> wrote:

> Tomasz Chmielewski schrieb:
> 
> (...)
> 
> > Let's access "/sys/block/md0/md/dev-sdd1/super":
> > 
> > # cat /sys/block/md0/md/dev-sdd1/super
> > 
> > # dmesg -c
> > ------------[ cut here ]------------
> > Kernel BUG at 78178626 [verbose debug info unavailable]
> 
> It turns out a broken RAID bitmap file has nothing to do with it - the 
> same happens on a different machine without a bitmap file:
> 
> ------------[ cut here ]------------
> Kernel BUG at 7817736a [verbose debug info unavailable]

argh.  Please do enable CONFIG_DEBUG_BUGVERBOSE.

> invalid opcode: 0000 [#1]
> Modules linked in: as_iosched nfs lockd nfs_acl sunrpc bonding dm_mirror 
> dm_snapshot e1000 sata_mv
> 
> Pid: 2494, comm: cat Not tainted (2.6.24.3-1 #1)
> EIP: 0060:[<7817736a>] EFLAGS: 00010212 CPU: 0
> EIP is at sysfs_read_file+0x88/0xd4
> EAX: 00000001 EBX: 961b5880 ECX: 00000000 EDX: 964ef360
> ESI: 00001000 EDI: 964ef3c0 EBP: 9705bd04 ESP: 971f1f54
>   DS: 007b ES: 007b FS: 0000 GS: 0033 SS: 0068
> Process cat (pid: 2494, ti=971f0000 task=970ad9a0 task.ti=971f0000)
> Stack: 96443080 0804b4d8 00001000 0804e000 961b5894 7835f6f0 96193400 
> 0804e000
>         781772e2 00001000 78149bd5 971f1fa0 00001000 96193400 fffffff7 
> 0804e000
>         971f0000 78149f03 971f1fa0 00000000 00000000 00000000 00000003 
> 0804e000
> Call Trace:
>   [<781772e2>] sysfs_read_file+0x0/0xd4
>   [<78149bd5>] vfs_read+0x88/0x10a
>   [<78149f03>] sys_read+0x41/0x67
>   [<78103bba>] syscall_call+0x7/0xb
>   =======================
> Code: c0 74 61 8b 47 18 8b 4b 0c 8b 40 04 89 43 24 89 e8 8b 74 24 14 8b 
> 57 14 ff 16 89 c6 89 f8 e8 18 0b 00 00 81 fe ff 0f 00 00 7e 04 <0f> 0b 
> eb fe 85 f6 78 31 c7 43 20 00 00 00 00 89 33 eb 07 be f4
> EIP: [<7817736a>] sysfs_read_file+0x88/0xd4 SS:ESP 0068:971f1f54

I assume this is the BUG_ON(count >= (ssize_t)PAGE_SIZE) in
fill_read_buffer().

This was reported recently and we prepared a debug patch but the
reporter was unable to trigger the bug again.

Please add the below and retest?


From: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>

Try to find the culprit who caused
http://bugzilla.kernel.org/show_bug.cgi?id=10150

Cc: <balajirrao@xxxxxxxxx>
Cc: Greg KH <greg@xxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---

 drivers/base/core.c |    5 +++++
 fs/sysfs/file.c     |    8 +++++++-
 2 files changed, 12 insertions(+), 1 deletion(-)

diff -puN fs/Kconfig~driver-core-debug-for-bad-dev_attr_show-return-value fs/Kconfig
diff -puN fs/sysfs/file.c~driver-core-debug-for-bad-dev_attr_show-return-value fs/sysfs/file.c
--- a/fs/sysfs/file.c~driver-core-debug-for-bad-dev_attr_show-return-value
+++ a/fs/sysfs/file.c
@@ -12,6 +12,7 @@
 
 #include <linux/module.h>
 #include <linux/kobject.h>
+#include <linux/kallsyms.h>
 #include <linux/namei.h>
 #include <linux/poll.h>
 #include <linux/list.h>
@@ -94,7 +95,12 @@ static int fill_read_buffer(struct dentr
 	 * The code works fine with PAGE_SIZE return but it's likely to
 	 * indicate truncated result or overflow in normal use cases.
 	 */
-	BUG_ON(count >= (ssize_t)PAGE_SIZE);
+	if (count >= (ssize_t)PAGE_SIZE) {
+		print_symbol("fill_read_buffer: %s returned bad count\n",
+			(unsigned long)ops->show);
+		/* Try to struggle along */
+		count = PAGE_SIZE - 1;
+	}
 	if (count >= 0) {
 		buffer->needs_read_fill = 0;
 		buffer->count = count;
diff -puN drivers/base/core.c~driver-core-debug-for-bad-dev_attr_show-return-value drivers/base/core.c
--- a/drivers/base/core.c~driver-core-debug-for-bad-dev_attr_show-return-value
+++ a/drivers/base/core.c
@@ -19,6 +19,7 @@
 #include <linux/kdev_t.h>
 #include <linux/notifier.h>
 #include <linux/genhd.h>
+#include <linux/kallsyms.h>
 #include <asm/semaphore.h>
 
 #include "base.h"
@@ -68,6 +69,10 @@ static ssize_t dev_attr_show(struct kobj
 
 	if (dev_attr->show)
 		ret = dev_attr->show(dev, dev_attr, buf);
+	if (ret >= (ssize_t)PAGE_SIZE) {
+		print_symbol("dev_attr_show: %s returned bad count\n",
+				(unsigned long)dev_attr->show);
+	}
 	return ret;
 }
 
_

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/