[Top] [All Lists]

Re: XFS hung on kernel

To: Dave Chinner <david@xxxxxxxxxxxxx>
Subject: Re: XFS hung on kernel
From: Ilia Mirkin <imirkin@xxxxxxxxxxxx>
Date: Thu, 5 Aug 2010 02:42:38 -0400
Cc: xfs@xxxxxxxxxxx
Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:sender:received :in-reply-to:references:date:x-google-sender-auth:message-id:subject :from:to:cc:content-type:content-transfer-encoding; bh=wsu0V4E2yCBoN1qrtmYTZ7Jkqn1zWWEVQbSmIPyYMmM=; b=CM2yX+A4sQzm07DPzDyW1B17fe3zbiFQ3opqUOiGU96CNCHyAHEULrAFMGpEcQqCy0 3LGs1AcgONlcIniKYxosBS9m8k4sMQ7ewEIkvKWn6+8x180kaOryBhCeqDEu+FZQEXtR 8fxQYqrpGROaW6pUrfCGmbDxZwvNQd9P/h6VI=
Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:sender:in-reply-to:references:date :x-google-sender-auth:message-id:subject:from:to:cc:content-type :content-transfer-encoding; b=lSFHCmSjCj8JIOVoKiMMszDf/3w+jNWYZ55i2TDBvwrq3tGXMO1Dr1srkPQWE8uH7S hGlRhdGrXb9kYPVdQHL2J6vAX/WKrre1Sjzo23lWSvj+45rXCOdeNdkZog9mvEmPBatn rJpP+Nj2Ryo1hR0v1LsNLH53c89rul37ieEjg=
In-reply-to: <20100805062541.GC26402@dastard>
References: <AANLkTilX3l8TbUztLStj_u9OqOZnBrsNQxmeV4DuBmYJ@xxxxxxxxxxxxxx> <20100718012033.GA18888@dastard> <AANLkTikEv75KRyRTs4awmG894NSKMnBkJNJPYsypMdWf@xxxxxxxxxxxxxx> <20100718235036.GC32635@dastard> <AANLkTi=EzePfBRdEFQo2BRtKy=464Ci6zPG=UyYJiwNp@xxxxxxxxxxxxxx> <20100805062541.GC26402@dastard>
Sender: ibmirkin@xxxxxxxxx
On Thu, Aug 5, 2010 at 2:25 AM, Dave Chinner <david@xxxxxxxxxxxxx> wrote:
> a couple of things to see if the hang can be caused faster? Maybe
> trying these different initial conditions:

Together or independently?

>        - use a smaller test file

OK. My theory [based on no evidence] was that it had to be bigger than
RAM size. I'll try less though.

>        - allocate the file with worst case fragmentation by
>          writing it backwards in synchronous 4k blocks before
>          running the test

I created the file with

dd if=/dev/zero of=... bs=1M count=...

How would I "write backwards"? Like truncate and then seek to the end,
write, seek back, write, etc? Is there a utility that might do it

>        - preallocating the test file with fallocate()


> Also, it woul dbe interesting to find out if a newer kernel also
> hangs in a similar manner, say 2.6.35?

Will do.

Ilia Mirkin

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