[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <AFE1174E-74AD-4EA3-ADFC-E7512BEC60AE@dilger.ca>
Date: Wed, 14 Dec 2022 14:15:30 -0700
From: Andreas Dilger <adilger@...ger.ca>
To: "Ritesh Harjani (IBM)" <ritesh.list@...il.com>
Cc: Theodore Ts'o <tytso@....edu>, linux-ext4@...r.kernel.org,
Harshad Shirwadkar <harshadshirwadkar@...il.com>,
Li Xi <lixi@....com>, Wang Shilong <wangshilong1991@...il.com>
Subject: Re: [RFCv1 15/72] tst_bitmaps_pthread: Add merge bitmaps test using
pthreads
On Nov 7, 2022, at 5:21 AM, Ritesh Harjani (IBM) <ritesh.list@...il.com> wrote:
>
> This patch adds a test to verify the core bitmaps merge APIs
> for rbtree bitmap type.
>
> Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@...il.com>
Reviewed-by: Andreas Dilger <adilger@...ger.ca>
> +/*
> + * Note we use EXT2FS_BMAP64_BITARRAY always for used_bitmap, this is because
> + * EXT2FS_BMAP64_RBTREE does not support parallel scan due to rcursor
> + * optimization.
> + */
Is this going to be a problem in the future? I think for pass1/pass5 there
are no rbtree bitmaps that are *source* bitmaps for multi-threaded operation,
but I suspect that once we try to parallelize pass2 this could break.
Is there any check in the code that prevents multi-thread access to an rbtree
bitmap? Making a copy for each thread, or ideally copying the rcursor for
each thread, to allow at least multi-threaded reads. I don't think there is
a high need for multi-threaded write
Cheers, Andreas
Download attachment "signature.asc" of type "application/pgp-signature" (874 bytes)
Powered by blists - more mailing lists