From: John Garry <john.g.garry@oracle.com>
To: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>,
brauner@kernel.org, djwong@kernel.org, cem@kernel.org,
linux-xfs@vger.kernel.org, linux-fsdevel@vger.kernel.org,
linux-kernel@vger.kernel.org, ojaswin@linux.ibm.com,
ritesh.list@gmail.com, martin.petersen@oracle.com
Subject: Re: [PATCH RFC v5 10/10] iomap: Rename ATOMIC flags again
Date: Thu, 13 Mar 2025 07:41:11 +0000 [thread overview]
Message-ID: <3aeb1d0e-6c74-4bfe-914d-22ba4152bc7f@oracle.com> (raw)
In-Reply-To: <Z9KC7UHOutY61C5K@infradead.org>
On 13/03/2025 07:02, Christoph Hellwig wrote:
> On Thu, Mar 13, 2025 at 06:28:03AM +0000, John Garry wrote:
>>>> I'd rather have a
>>>>
>>>> blk_opf_t extra_flags;
>>>>
>>>> in the caller that gets REQ_FUA and REQ_ATOMIC assigned as needed,
>>>> and then just clear
>>> Yep, that is cleaner..
>>
>> This suggestion is not clear to me.
>>
>> Is it that iomap_dio_bio_iter() [the only caller of iomap_dio_bio_opflags()]
>> sets REQ_FUA and REQ_ATOMIC in extra_flags, and then we extra_flags |
>> bio_opf?
>
> Yes.
>
>> Note that iomap_dio_bio_opflags() does still use use_fua for clearing
>> IOMAP_DIO_WRITE_THROUGH.
>
> You can check for REQ_FUA in extra_flags (or the entire op).
> >> And to me it seems nicer to set all the REQ_ flags in one place.
>
> Passing multiple bool arguments just loses way too much context. But
> if you really want everything in one place you could probably build
> the entire blk_opf_t in iomap_dio_bio_iter, and avoid having to
> recalculate it for every bio.
>
Yeah, when we start taking use_fua and atomic_bio args from
iomap_dio_bio_opflags(), then iomap_dio_bio_opflags() becomes a shell of
the function.
So how about this (I would re-add the write through comment):
--- a/fs/iomap/direct-io.c
+++ b/fs/iomap/direct-io.c
@@ -311,30 +311,6 @@ static int iomap_dio_zero(const struct iomap_iter
*iter, struct iomap_dio *dio,
return 0;
}
-/*
- * Figure out the bio's operation flags from the dio request, the
- * mapping, and whether or not we want FUA. Note that we can end up
- * clearing the WRITE_THROUGH flag in the dio request.
- */
-static inline blk_opf_t iomap_dio_bio_opflags(struct iomap_dio *dio,
- const struct iomap *iomap, bool use_fua, bool atomic_hw)
-{
- blk_opf_t opflags = REQ_SYNC | REQ_IDLE;
-
- if (!(dio->flags & IOMAP_DIO_WRITE))
- return REQ_OP_READ;
-
- opflags |= REQ_OP_WRITE;
- if (use_fua)
- opflags |= REQ_FUA;
- else
- dio->flags &= ~IOMAP_DIO_WRITE_THROUGH;
- if (atomic_hw)
- opflags |= REQ_ATOMIC;
-
- return opflags;
-}
-
static int iomap_dio_bio_iter(struct iomap_iter *iter, struct
iomap_dio *dio)
{
const struct iomap *iomap = &iter->iomap;
@@ -346,13 +322,20 @@ static int iomap_dio_bio_iter(struct iomap_iter
*iter, struct iomap_dio *dio)
blk_opf_t bio_opf;
struct bio *bio;
bool need_zeroout = false;
- bool use_fua = false;
int nr_pages, ret = 0;
u64 copied = 0;
size_t orig_count;
- if (atomic_hw && length != iter->len)
- return -EINVAL;
+ if (dio->flags & IOMAP_DIO_WRITE) {
+ bio_opf = REQ_OP_WRITE;
+ if (atomic_hw) {
+ if (length != iter->len)
+ return -EINVAL;
+ bio_opf |= REQ_ATOMIC;
+ }
+ } else {
+ bio_opf = REQ_OP_READ;
+ }
if ((pos | length) & (bdev_logical_block_size(iomap->bdev) - 1) ||
!bdev_iter_is_aligned(iomap->bdev, dio->submit.iter))
@@ -382,10 +365,12 @@ static int iomap_dio_bio_iter(struct iomap_iter
*iter, struct iomap_dio *dio)
*/
if (!(iomap->flags & (IOMAP_F_SHARED|IOMAP_F_DIRTY)) &&
(dio->flags & IOMAP_DIO_WRITE_THROUGH) &&
- (bdev_fua(iomap->bdev) || !bdev_write_cache(iomap->bdev)))
- use_fua = true;
- else if (dio->flags & IOMAP_DIO_NEED_SYNC)
+ (bdev_fua(iomap->bdev) || !bdev_write_cache(iomap->bdev))) {
+ bio_opf |= REQ_FUA; //reads as well?
+ dio->flags &= ~IOMAP_DIO_WRITE_THROUGH;
+ } else if (dio->flags & IOMAP_DIO_NEED_SYNC) {
dio->flags &= ~IOMAP_DIO_CALLER_COMP;
+ }
}
/*
@@ -407,7 +392,7 @@ static int iomap_dio_bio_iter(struct iomap_iter
*iter, struct iomap_dio *dio)
* during completion processing.
*/
if (need_zeroout ||
- ((dio->flags & IOMAP_DIO_NEED_SYNC) && !use_fua) ||
+ ((dio->flags & IOMAP_DIO_NEED_SYNC) && !(bio_opf & REQ_FUA)) ||
((dio->flags & IOMAP_DIO_WRITE) && pos >= i_size_read(inode)))
dio->flags &= ~IOMAP_DIO_CALLER_COMP;
@@ -428,8 +413,6 @@ static int iomap_dio_bio_iter(struct iomap_iter
*iter, struct iomap_dio *dio)
goto out;
}
- bio_opf = iomap_dio_bio_opflags(dio, iomap, use_fua, atomic_hw);
-
nr_pages = bio_iov_vecs_to_alloc(dio->submit.iter, BIO_MAX_VECS);
do {
size_t n;
--
next prev parent reply other threads:[~2025-03-13 7:41 UTC|newest]
Thread overview: 63+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-03-10 18:39 [PATCH v5 00/10] large atomic writes for xfs with CoW John Garry
2025-03-10 18:39 ` [PATCH v5 01/10] xfs: Pass flags to xfs_reflink_allocate_cow() John Garry
2025-03-12 7:15 ` Christoph Hellwig
2025-03-12 8:19 ` John Garry
2025-03-10 18:39 ` [PATCH v5 02/10] xfs: Switch atomic write size check in xfs_file_write_iter() John Garry
2025-03-12 7:17 ` Christoph Hellwig
2025-03-12 8:21 ` John Garry
2025-03-10 18:39 ` [PATCH v5 03/10] xfs: Refactor xfs_reflink_end_cow_extent() John Garry
2025-03-12 7:24 ` Christoph Hellwig
2025-03-12 8:27 ` John Garry
2025-03-12 8:35 ` Christoph Hellwig
2025-03-12 15:46 ` Darrick J. Wong
2025-03-12 22:06 ` John Garry
2025-03-12 23:22 ` Darrick J. Wong
2025-03-13 1:25 ` Dave Chinner
2025-03-13 4:51 ` Darrick J. Wong
2025-03-13 6:11 ` John Garry
2025-03-18 0:43 ` Dave Chinner
2025-03-13 7:21 ` Dave Chinner
2025-03-22 5:19 ` Darrick J. Wong
2025-03-10 18:39 ` [PATCH v5 04/10] xfs: Reflink CoW-based atomic write support John Garry
2025-03-12 7:27 ` Christoph Hellwig
2025-03-12 9:13 ` John Garry
2025-03-12 13:45 ` Christoph Hellwig
2025-03-12 14:48 ` John Garry
2025-03-10 18:39 ` [PATCH v5 05/10] xfs: Iomap SW-based " John Garry
2025-03-12 7:37 ` Christoph Hellwig
2025-03-12 9:00 ` John Garry
2025-03-12 13:52 ` Christoph Hellwig
2025-03-12 14:57 ` John Garry
2025-03-12 15:55 ` Christoph Hellwig
2025-03-12 16:11 ` John Garry
2025-03-10 18:39 ` [PATCH v5 06/10] xfs: Add xfs_file_dio_write_atomic() John Garry
2025-03-10 18:39 ` [PATCH v5 07/10] xfs: Commit CoW-based atomic writes atomically John Garry
2025-03-12 7:39 ` Christoph Hellwig
2025-03-12 9:04 ` John Garry
2025-03-12 13:54 ` Christoph Hellwig
2025-03-12 15:01 ` John Garry
2025-03-10 18:39 ` [PATCH v5 08/10] xfs: Update atomic write max size John Garry
2025-03-11 14:40 ` Carlos Maiolino
2025-03-12 7:41 ` Christoph Hellwig
2025-03-12 8:09 ` John Garry
2025-03-12 8:13 ` Christoph Hellwig
2025-03-12 8:14 ` John Garry
2025-03-10 18:39 ` [PATCH v5 09/10] xfs: Allow block allocator to take an alignment hint John Garry
2025-03-12 7:42 ` Christoph Hellwig
2025-03-12 8:05 ` John Garry
2025-03-12 13:45 ` Christoph Hellwig
2025-03-12 14:47 ` John Garry
2025-03-12 16:00 ` Darrick J. Wong
2025-03-12 16:28 ` John Garry
2025-03-10 18:39 ` [PATCH RFC v5 10/10] iomap: Rename ATOMIC flags again John Garry
2025-03-12 7:13 ` Christoph Hellwig
2025-03-12 23:59 ` Dave Chinner
2025-03-13 6:28 ` John Garry
2025-03-13 7:02 ` Christoph Hellwig
2025-03-13 7:41 ` John Garry [this message]
2025-03-13 7:49 ` Christoph Hellwig
2025-03-13 7:53 ` John Garry
2025-03-13 8:09 ` Christoph Hellwig
2025-03-13 8:18 ` Christoph Hellwig
2025-03-13 8:24 ` John Garry
2025-03-13 8:28 ` 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=3aeb1d0e-6c74-4bfe-914d-22ba4152bc7f@oracle.com \
--to=john.g.garry@oracle.com \
--cc=brauner@kernel.org \
--cc=cem@kernel.org \
--cc=david@fromorbit.com \
--cc=djwong@kernel.org \
--cc=hch@infradead.org \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-xfs@vger.kernel.org \
--cc=martin.petersen@oracle.com \
--cc=ojaswin@linux.ibm.com \
--cc=ritesh.list@gmail.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).