xfs
[Top] [All Lists]

Re: [PATCH, RFC] xfs: batched discard support

To: Mark Lord <liml@xxxxxx>
Subject: Re: [PATCH, RFC] xfs: batched discard support
From: James Bottomley <James.Bottomley@xxxxxxx>
Date: Thu, 20 Aug 2009 08:42:36 -0600
Cc: Ric Wheeler <rwheeler@xxxxxxxxxx>, Ingo Molnar <mingo@xxxxxxx>, Christoph Hellwig <hch@xxxxxxxxxxxxx>, Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>, Paul Mackerras <paulus@xxxxxxxxx>, Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>, xfs@xxxxxxxxxxx, linux-fsdevel@xxxxxxxxxxxxxxx, linux-scsi@xxxxxxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx, jens.axboe@xxxxxxxxxx, IDE/ATA development list <linux-ide@xxxxxxxxxxxxxxx>, Neil Brown <neilb@xxxxxxx>
In-reply-to: <4A8D5FDB.7080505@xxxxxx>
References: <20090816004705.GA7347@xxxxxxxxxxxxx> <20090819203916.GA25296@xxxxxxx> <4A8CA956.2060406@xxxxxx> <4A8D5442.1000302@xxxxxxxxxx> <4A8D5FDB.7080505@xxxxxx>
On Thu, 2009-08-20 at 10:38 -0400, Mark Lord wrote:
> Would it still be okay to do the TRIMs when the entire parity stripe
> (across all members) is being discarded?  (As opposed to just partial
> data there being dropped)

Not really.  The problem is that array verification is done at the block
level not the fs level (although, I suppose, we could change that).  So
a fully discarded stripe still has to verify OK (as in what's read for
the parity must match what's read for the data).  All of this is the
reason for the TPRZ bit for SCSI UNMAP ... and why WRITE_SAME is also
under consideration for discards in T10.

James


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