Bug 76411

Summary:	Kernel panic: Watchdog detected hard LOCKUP on cpu 0 - freezes the whole machine
Product:	File System	Reporter:	Tormen (tormen)
Component:	btrfs	Assignee:	Josef Bacik (josef)
Status:	RESOLVED OBSOLETE
Severity:	normal	CC:	dsterba, szg00000
Priority:	P1
Hardware:	All
OS:	Linux
Kernel Version:	3.14.4	Subsystem:
Regression:	No	Bisected commit-id:
Attachments:	Screenshot of the kernel panic

Description Tormen 2014-05-18 15:50:44 UTC

Hi, 

I tried with debian linux-image 3.13, then with latest debian 3.14.4, same problem:

I always get a kernel panic " Watchdog detected hard LOCKUP on cpu 0 " after some while when using btrfs and rsyncing (rsync -ax) 2TB (from an xfs partition onto a 3TB btrfs parition (created with kernel 3.14.4, mkfs.btrfs v0.19) mounted with compress=zlib.

When using xfs, everything works as it should (same machine, same source, same target, same kernel, same rsync command.

The partition contains backup data so MANY smaller files.

The kernel panic locks up the machine completely (even NumLock disabled), so the machine is really DEAD ;)
Traces of the kernel panic are ONLY printed to screen, they did not even get through /dev/kmsg via nc into a logfile on another machine of /dev/kmsg
I took a photo if your interested.

Not sure how to further debug this.


Tormen

Comment 1 Tormen 2014-05-18 15:52:02 UTC

The xfs partition is plain vanilla (mkfs.xfs /dev/xxx, no special mount options what so ever).

Comment 2 Tormen 2014-05-18 16:43:28 UTC

Created attachment 136631 [details]
Screenshot of the kernel panic

Comment 3 Tormen 2014-05-18 16:47:39 UTC

I know that in the screenshot it does not say anything about btrfs.
But I tried it 3 times: 2x with 3.13 1x with 3.14.4 on an otherwise IDLE server and xfs -> xfs worked right away (and is still running ;), so this  +  the fact that I NEVER saw this kernel panic before the moment I tried btrfs clearly points to btrfs as "guilty" variable in the equation.

My guess: A combination of the amount of data + SOOO many little files + compress=zlib.

But you tell me ;)

Thanks again in advance.

Let me know if you'll need anything. As it's only my backup data, I can still do tests for you on a spare drive.

Comment 4 Tormen 2014-05-18 16:48:18 UTC

(and the 3 times it always ended in the screenshot I attached)

Comment 5 David Sterba 2022-10-03 15:01:18 UTC

This is a semi-automated bugzilla cleanup, report is against an old kernel version. If the problem still happens, please open a new bug. Thanks.