From: Wengang Wang <wen.gang.wang@oracle.com>
To: "linux-xfs@vger.kernel.org" <linux-xfs@vger.kernel.org>
Subject: Re: [PATCH] xfs: make src file readable during reflink
Date: Thu, 23 Jun 2022 16:12:36 +0000 [thread overview]
Message-ID: <44420D34-8DF7-428B-8881-0825A28232B2@oracle.com> (raw)
In-Reply-To: <20220618011631.61826-1-wen.gang.wang@oracle.com>
Hi,
Could anyone please review this patch?
thanks,
wengang
> On Jun 17, 2022, at 6:16 PM, Wengang Wang <wen.gang.wang@oracle.com> wrote:
>
> inode io lock and mmap lock are exclusively locked durning reflink copy,
> read operations are blocked during that time. In case the reflink copy
> needs a long time to finish, read operations could be blocked for that long
> too.
>
> The real case is that reflinks take serveral minutes or even longer with
> huge source files. Those source files are hundreds of GB long and badly
> fragmented, so the reflink copy needs to process more than one million
> extents.
>
> As the source of copy, besides some neccessary modification happens (say
> dirty page flushing), it plays readonly role. Locking source file
> exclusively through out the reflink copy is unnecessary.
>
> This patch downgrade exclusive locks on source file to shared modes after
> page cache flushing and before cloing the extents.
>
> Signed-off-by: Wengang Wang <wen.gang.wang@oracle.com>
> ---
> fs/xfs/xfs_file.c | 8 +++++++-
> fs/xfs/xfs_inode.c | 36 ++++++++++++++++++++++++++++++++++++
> fs/xfs/xfs_inode.h | 4 ++++
> 3 files changed, 47 insertions(+), 1 deletion(-)
>
> diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
> index 5a171c0b244b..99bbb188deb4 100644
> --- a/fs/xfs/xfs_file.c
> +++ b/fs/xfs/xfs_file.c
> @@ -1125,6 +1125,12 @@ xfs_file_remap_range(
> if (ret || len == 0)
> return ret;
>
> + /*
> + * From now on, we read only from src, so downgrade locks on src
> + * to allow read operations in parallel.
> + */
> + xfs_ilock_io_mmap_downgrade_src(src, dest);
> +
> trace_xfs_reflink_remap_range(src, pos_in, len, dest, pos_out);
>
> ret = xfs_reflink_remap_blocks(src, pos_in, dest, pos_out, len,
> @@ -1152,7 +1158,7 @@ xfs_file_remap_range(
> if (xfs_file_sync_writes(file_in) || xfs_file_sync_writes(file_out))
> xfs_log_force_inode(dest);
> out_unlock:
> - xfs_iunlock2_io_mmap(src, dest);
> + xfs_iunlock2_io_mmap_src_shared(src, dest);
> if (ret)
> trace_xfs_reflink_remap_range_error(dest, ret, _RET_IP_);
> return remapped > 0 ? remapped : ret;
> diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
> index 52d6f2c7d58b..721abefbb1fa 100644
> --- a/fs/xfs/xfs_inode.c
> +++ b/fs/xfs/xfs_inode.c
> @@ -3786,6 +3786,22 @@ xfs_ilock2_io_mmap(
> return 0;
> }
>
> +/*
> + * Downgrade the locks on src file if src and dest are not the same one.
> + */
> +void
> +xfs_ilock_io_mmap_downgrade_src(
> + struct xfs_inode *src,
> + struct xfs_inode *dest)
> +{
> + if (src != dest) {
> + struct inode *inode = VFS_I(src);
> +
> + downgrade_write(&inode->i_mapping->invalidate_lock);
> + downgrade_write(&inode->i_rwsem);
> + }
> +}
> +
> /* Unlock both inodes to allow IO and mmap activity. */
> void
> xfs_iunlock2_io_mmap(
> @@ -3798,3 +3814,23 @@ xfs_iunlock2_io_mmap(
> if (ip1 != ip2)
> inode_unlock(VFS_I(ip1));
> }
> +
> +/* Unlock the exclusive locks on dest file.
> + * Also unlock the shared locks on src if src and dest are not the same one
> + */
> +void
> +xfs_iunlock2_io_mmap_src_shared(
> + struct xfs_inode *src,
> + struct xfs_inode *dest)
> +{
> + struct inode *src_inode = VFS_I(src);
> + struct inode *dest_inode = VFS_I(dest);
> +
> + inode_unlock(dest_inode);
> + up_write(&dest_inode->i_mapping->invalidate_lock);
> + if (src == dest)
> + return;
> +
> + inode_unlock_shared(src_inode);
> + up_read(&src_inode->i_mapping->invalidate_lock);
> +}
> diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h
> index 7be6f8e705ab..02b44e1a7e4e 100644
> --- a/fs/xfs/xfs_inode.h
> +++ b/fs/xfs/xfs_inode.h
> @@ -512,5 +512,9 @@ void xfs_end_io(struct work_struct *work);
>
> int xfs_ilock2_io_mmap(struct xfs_inode *ip1, struct xfs_inode *ip2);
> void xfs_iunlock2_io_mmap(struct xfs_inode *ip1, struct xfs_inode *ip2);
> +void xfs_ilock_io_mmap_downgrade_src(struct xfs_inode *src,
> + struct xfs_inode *dest);
> +void xfs_iunlock2_io_mmap_src_shared(struct xfs_inode *src,
> + struct xfs_inode *dest);
>
> #endif /* __XFS_INODE_H__ */
> --
> 2.21.0 (Apple Git-122.2)
>
next parent reply other threads:[~2022-06-23 16:12 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <20220618011631.61826-1-wen.gang.wang@oracle.com>
2022-06-23 16:12 ` Wengang Wang [this message]
2022-06-23 17:37 ` [PATCH] xfs: make src file readable during reflink Darrick J. Wong
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=44420D34-8DF7-428B-8881-0825A28232B2@oracle.com \
--to=wen.gang.wang@oracle.com \
--cc=linux-xfs@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox