[Top] [All Lists]

Re: Still seeing hangs in xlog_grant_log_space

To: xfs@xxxxxxxxxxx
Subject: Re: Still seeing hangs in xlog_grant_log_space
From: Chris J Arges <chris.j.arges@xxxxxxxxxxxxx>
Date: Tue, 5 Jun 2012 08:21:34 -0700 (PDT)
In-reply-to: <CAH4wwdGWHSZoveLJMxu5pjr22NEEeW7oG8TS+snoM8RY=ZeRmg@xxxxxxxxxxxxxx>
References: <CADLDEKsP4DsXf_G07ub+a-ODbrJbsiprRJUX1fJdaQ41TB7+Xg@xxxxxxxxxxxxxx> <CAH4wwdGWHSZoveLJMxu5pjr22NEEeW7oG8TS+snoM8RY=ZeRmg@xxxxxxxxxxxxxx>

Peter Watkins-3 wrote:
> Does your kernel have the effect of
> 0bf6a5bd4b55b466964ead6fa566d8f346a828ee xfs: convert the xfsaild
> thread to a workqueue

This patch is present in the Ubuntu Precise 3.2 kernel, and I have been able
to reproduce the failure with this patch applied.

Peter Watkins-3 wrote:
> c7eead1e118fb7e34ee8f5063c3c090c054c3820 xfs: revert to using a
> kthread for AIL pushing

I see this patch is 0030807c66f058230bcb20d2573bcaf28852e804 upstream, and
this patch is also present in the 3.2 Precise kernel that was tested and
exhibited the failure.

Peter Watkins-3 wrote:
> In particular, is this code in xfs_trans_ail_push:
>        smp_wmb();
>        xfs_trans_ail_copy_lsn(ailp, &ailp->xa_target, &threshold_lsn);
>        smp_wmb();

In fs/xfs/xfs_trans_ail.c:xfs_ail_push():

         * Ensure that the new target is noticed in push code before it
         * the XFS_AIL_PUSHING_BIT.
        xfs_trans_ail_copy_lsn(ailp, &ailp->xa_target, &threshold_lsn);

xfs mailing list

View this message in context: 
Sent from the Xfs - General mailing list archive at Nabble.com.

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