[Top] [All Lists]

Re: kernel dump (may be related to previous fragmentation problem)

To: Ying-Hung Chen <ying@xxxxxxxxxxxxxx>
Subject: Re: kernel dump (may be related to previous fragmentation problem)
From: Eric Sandeen <sandeen@xxxxxxx>
Date: Mon, 26 Sep 2005 10:26:15 -0500
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <4337ADB7.3060700@xxxxxxxxxxxxxx>
References: <4337ADB7.3060700@xxxxxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Mozilla Thunderbird 1.0 (X11/20041206)
Ying-Hung Chen wrote:
Hi there,

We are seeing two (very similar?) kernel dumps (order:5, mode 0xd0 /
order:4 mode 0x50) on our systems,

the kernel didn't dump, this is just a backtrace. It is unable to allocate enough contiguous memory to accommodate your extent list, so yes, it is probably related to your fragmentation problem.

If you use the method I suggested previously for preallocating space, this problem should go away.

I have search the mailing list and it seem to somehow related to the
fragmentation problem? (ours is at > 99%) though someone maintained this
is not Fatal, but our system usually becomes unresponsive when the dump
starts to appear (sometimes its ok after a while, but most of time it
just dumps forever and system becomes unusable and we have to reboot)

it's in a loop trying to allocate memory, and it can't.

are there fixes / workaround we can try?

Use preallocation, as I suggested.

Are these two message dump the same thing?

basically, yes.


<Prev in Thread] Current Thread [Next in Thread>