[Top] [All Lists]

Re: [PATCH V2] xfs: prevent xfs_bmap_extsize_align() to exceed maximum e

To: Dave Chinner <david@xxxxxxxxxxxxx>
Subject: Re: [PATCH V2] xfs: prevent xfs_bmap_extsize_align() to exceed maximum extent size.
From: Alain Renaud <arenaud@xxxxxxx>
Date: Fri, 20 Jul 2012 10:05:01 +0200
Cc: xfs@xxxxxxxxxxx
In-reply-to: <20120720014454.GI23387@dastard>
References: <20120712154554.377970666@xxxxxxx> <4FFFF3D8.7060001@xxxxxxx> <50059437.6070705@xxxxxxx> <20120719033027.GD23387@dastard> <5007C33B.5010603@xxxxxxx> <20120720014454.GI23387@dastard>
User-agent: Mozilla/5.0 (X11; Linux i686; rv:13.0) Gecko/20120615 Thunderbird/13.0.1
On 12-07-20 03:44 AM, Dave Chinner wrote:

If you look in the function where I put this code is it AFTER the
head/len alignment so the HEAD is already align so I do not need to
account for it.
No need to shout. I did look at the code, and IMO you do need to
account for both head and tail extension and you definitely can't do
it there.....
Sorry for the cap I did not intend to shout.

However the problem does exist. the len limitation in xfs_bmapi_write is not sufficient
if you set the extsize.

Here is what I see.

xfs_bmapi_write() <= lenght limited to MAXEXTLEN;
                      xfs_bmap_extsize_align()  <== Align the offest / len

Since the alignment is outward like you explain bellow we return len > MAXEXTLEN
causing failure to report ENOSPC.

You can reproduce the problem the following way.

# uname -r

# mkfs.xfs -f -b size=512 -d agcount=1,size=4294967296 /dev/sdb5

meta-data=/dev/sdb5              isize=256    agcount=1, agsize=8388608 blks
         =                       sectsz=512   attr=2, projid32bit=0
data     =                       bsize=512    blocks=8388608, imaxpct=25
         =                       sunit=0      swidth=0 blks
naming   =version 2              bsize=4096   ascii-ci=0
log      =internal log           bsize=512    blocks=20480, version=2
         =                       sectsz=512   sunit=0 blks, lazy-count=1
realtime =none                   extsz=4096   blocks=0, rtextents=0

# mount /dev/sdb5 /a

# xfs_io -f -c "extsize 2m" -c "resvsp 0 1g" /a/data1
XFS_IOC_RESVSP64: No space left on device

So Maybe the fix is to limit the bma.length in xfs_bmapi_write to MAXEXTLEN - (2 * extsz - 1) like you do in xfs_bmapi_reserve_delalloc().

Hope this is making sense.

Alain Renaud     - Cluster File System Engineer
arenaud@xxxxxxx  - SGI

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