From: "Darrick J. Wong" <darrick.wong@oracle.com>
To: Christoph Hellwig <hch@lst.de>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.com>,
linux-xfs@vger.kernel.org, linux-fsdevel@vger.kernel.org
Subject: Re: [PATCH 08/20] iomap: use a srcmap for a read-modify-write I/O
Date: Wed, 9 Oct 2019 10:16:57 -0700 [thread overview]
Message-ID: <20191009171657.GG13108@magnolia> (raw)
In-Reply-To: <20191009062824.GA29833@lst.de>
On Wed, Oct 09, 2019 at 08:28:24AM +0200, Christoph Hellwig wrote:
> On Tue, Oct 08, 2019 at 08:00:44AM -0700, Darrick J. Wong wrote:
> > > unsigned long vaddr = vmf->address;
> > > loff_t pos = (loff_t)vmf->pgoff << PAGE_SHIFT;
> > > struct iomap iomap = { 0 };
> >
> > Does this definition ^^^^^ need to be converted too? You convert the
> > one in iomap_apply()...
>
> Doesn't strictly need to, but it sure would look nicer and fit the theme.
>
> > /*
> > * The @iomap and @srcmap parameters should be set to a hole
> > * prior to calling ->iomap_begin.
> > */
> > #define IOMAP_EMPTY_RECORD { .type = IOMAP_HOLE }
> >
> > ...and later...
> >
> > struct iomap srcmap = IOMAP_EMPTY_RECORD;
> >
> > ..but meh, I'm not sure that adds much.
>
> I don't really see the point.
Yeah. Agreed.
> > > unsigned flags = IOMAP_FAULT;
> > > int error, major = 0;
> > > bool write = vmf->flags & FAULT_FLAG_WRITE;
> > > @@ -1292,7 +1293,7 @@ static vm_fault_t dax_iomap_pte_fault(struct vm_fault *vmf, pfn_t *pfnp,
> > > * the file system block size to be equal the page size, which means
> > > * that we never have to deal with more than a single extent here.
> > > */
> > > - error = ops->iomap_begin(inode, pos, PAGE_SIZE, flags, &iomap);
> > > + error = ops->iomap_begin(inode, pos, PAGE_SIZE, flags, &iomap, &srcmap);
> >
> > ->iomap_begin callers are never supposed to touch srcmap, right?
> > Maybe we ought to check that srcmap.io_type == HOLE, at least until
> > someone fixes this code to dax-copy the data from srcmap to iomap?
>
> What do you mean with touch? ->iomap_begin fills it out and then the
> caller looks at it, at least for places that can deal with
> read-modify-write operations (DAX currently can't).
Yes, I grok that the DAX code should never get fed a shared mapping, but
maybe we ought to have a WARN_ON_ONCE just in case some filesystem AI
programmer decides to backport a fs patch that results in sending a
non-hole srcmap back to the dax iomap callers. /We/ know that you
should never do this, but does the AI know? <grumble>
(Yeah, pure paranoia on my part :P)
--D
next prev parent reply other threads:[~2019-10-09 17:17 UTC|newest]
Thread overview: 32+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-08 7:15 iomap and xfs COW cleanups v2 Christoph Hellwig
2019-10-08 7:15 ` [PATCH 01/20] iomap: better document the IOMAP_F_* flags Christoph Hellwig
2019-10-08 7:15 ` [PATCH 02/20] iomap: remove the unused iomap argument to __iomap_write_end Christoph Hellwig
2019-10-08 7:15 ` [PATCH 03/20] iomap: always use AOP_FLAG_NOFS in iomap_write_begin Christoph Hellwig
2019-10-08 7:15 ` [PATCH 04/20] iomap: ignore non-shared or non-data blocks in xfs_file_dirty Christoph Hellwig
2019-10-08 7:15 ` [PATCH 05/20] iomap: move the zeroing case out of iomap_read_page_sync Christoph Hellwig
2019-10-08 7:15 ` [PATCH 06/20] iomap: use write_begin to read pages to unshare Christoph Hellwig
2019-10-08 15:12 ` Darrick J. Wong
2019-10-08 7:15 ` [PATCH 07/20] iomap: renumber IOMAP_HOLE to 0 Christoph Hellwig
2019-10-08 15:01 ` Darrick J. Wong
2019-10-08 7:15 ` [PATCH 08/20] iomap: use a srcmap for a read-modify-write I/O Christoph Hellwig
2019-10-08 15:00 ` Darrick J. Wong
2019-10-09 6:28 ` Christoph Hellwig
2019-10-09 17:16 ` Darrick J. Wong [this message]
2019-10-14 23:27 ` Darrick J. Wong
2019-10-15 13:00 ` Goldwyn Rodrigues
2019-10-08 7:15 ` [PATCH 09/20] xfs: also call xfs_file_iomap_end_delalloc for zeroing operations Christoph Hellwig
2019-10-08 7:15 ` [PATCH 10/20] xfs: remove xfs_reflink_dirty_extents Christoph Hellwig
2019-10-08 15:20 ` Darrick J. Wong
2019-10-08 7:15 ` [PATCH 11/20] xfs: pass two imaps to xfs_reflink_allocate_cow Christoph Hellwig
2019-10-08 7:15 ` [PATCH 12/20] xfs: refactor xfs_file_iomap_begin_delay Christoph Hellwig
2019-10-08 7:15 ` [PATCH 13/20] xfs: fill out the srcmap in iomap_begin Christoph Hellwig
2019-10-08 15:26 ` Darrick J. Wong
2019-10-08 7:15 ` [PATCH 14/20] xfs: factor out a helper to calculate the end_fsb Christoph Hellwig
2019-10-08 7:15 ` [PATCH 15/20] xfs: split out a new set of read-only iomap ops Christoph Hellwig
2019-10-08 15:27 ` Darrick J. Wong
2019-10-08 7:15 ` [PATCH 16/20] xfs: move xfs_file_iomap_begin_delay around Christoph Hellwig
2019-10-08 7:15 ` [PATCH 17/20] xfs: split the iomap ops for buffered vs direct writes Christoph Hellwig
2019-10-08 15:28 ` Darrick J. Wong
2019-10-08 7:15 ` [PATCH 18/20] xfs: rename the whichfork variable in xfs_buffered_write_iomap_begin Christoph Hellwig
2019-10-08 7:15 ` [PATCH 19/20] xfs: cleanup xfs_iomap_write_unwritten Christoph Hellwig
2019-10-08 7:15 ` [PATCH 20/20] xfs: improve the IOMAP_NOWAIT check for COW inodes Christoph Hellwig
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=20191009171657.GG13108@magnolia \
--to=darrick.wong@oracle.com \
--cc=hch@lst.de \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-xfs@vger.kernel.org \
--cc=rgoldwyn@suse.com \
/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;
as well as URLs for NNTP newsgroup(s).