From: "Darrick J. Wong" <djwong@kernel.org>
To: John Garry <john.g.garry@oracle.com>
Cc: brauner@kernel.org, hch@lst.de, viro@zeniv.linux.org.uk,
jack@suse.cz, cem@kernel.org, linux-fsdevel@vger.kernel.org,
dchinner@redhat.com, linux-xfs@vger.kernel.org,
linux-kernel@vger.kernel.org, ojaswin@linux.ibm.com,
ritesh.list@gmail.com, martin.petersen@oracle.com,
linux-ext4@vger.kernel.org, linux-block@vger.kernel.org,
catherine.hoang@oracle.com, linux-api@vger.kernel.org
Subject: [PATCH v9.1 05/15] xfs: ignore HW which cannot atomic write a single block
Date: Tue, 29 Apr 2025 22:18:22 -0700 [thread overview]
Message-ID: <20250430051822.GY25675@frogsfrogsfrogs> (raw)
In-Reply-To: <20250425164504.3263637-6-john.g.garry@oracle.com>
From: Darrick J. Wong <djwong@kernel.org>
Currently only HW which can write at least 1x block is supported.
For supporting atomic writes > 1x block, a CoW-based method will also be
used and this will not be resticted to using HW which can write >= 1x
block.
However for deciding if HW-based atomic writes can be used, we need to
start adding checks for write length < HW min, which complicates the
code. Indeed, a statx field similar to unit_max_opt should also be
added for this minimum, which is undesirable.
HW which can only write > 1x blocks would be uncommon and quite weird,
so let's just not support it.
Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
---
v9.1: move the atomic hw geometry calls to xfs_setup_devices
---
fs/xfs/xfs_buf.h | 3 ++-
fs/xfs/xfs_inode.h | 14 ++------------
fs/xfs/xfs_buf.c | 41 ++++++++++++++++++++++++++++++++++-------
fs/xfs/xfs_super.c | 6 +++++-
4 files changed, 43 insertions(+), 21 deletions(-)
diff --git a/fs/xfs/xfs_buf.h b/fs/xfs/xfs_buf.h
index d0b065a9a9f0d2..6f691779887f77 100644
--- a/fs/xfs/xfs_buf.h
+++ b/fs/xfs/xfs_buf.h
@@ -112,7 +112,7 @@ struct xfs_buftarg {
struct percpu_counter bt_readahead_count;
struct ratelimit_state bt_ioerror_rl;
- /* Atomic write unit values */
+ /* Atomic write unit values, bytes */
unsigned int bt_bdev_awu_min;
unsigned int bt_bdev_awu_max;
@@ -375,6 +375,7 @@ extern void xfs_free_buftarg(struct xfs_buftarg *);
extern void xfs_buftarg_wait(struct xfs_buftarg *);
extern void xfs_buftarg_drain(struct xfs_buftarg *);
extern int xfs_setsize_buftarg(struct xfs_buftarg *, unsigned int);
+void xfs_buftarg_config_atomic_writes(struct xfs_buftarg *btp);
#define xfs_getsize_buftarg(buftarg) block_size((buftarg)->bt_bdev)
#define xfs_readonly_buftarg(buftarg) bdev_read_only((buftarg)->bt_bdev)
diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h
index bdbbff0d8d9920..d7e2b902ef5c97 100644
--- a/fs/xfs/xfs_inode.h
+++ b/fs/xfs/xfs_inode.h
@@ -356,19 +356,9 @@ static inline bool xfs_inode_has_bigrtalloc(const struct xfs_inode *ip)
(XFS_IS_REALTIME_INODE(ip) ? \
(ip)->i_mount->m_rtdev_targp : (ip)->i_mount->m_ddev_targp)
-static inline bool
-xfs_inode_can_hw_atomic_write(
- struct xfs_inode *ip)
+static inline bool xfs_inode_can_hw_atomic_write(const struct xfs_inode *ip)
{
- struct xfs_mount *mp = ip->i_mount;
- struct xfs_buftarg *target = xfs_inode_buftarg(ip);
-
- if (mp->m_sb.sb_blocksize < target->bt_bdev_awu_min)
- return false;
- if (mp->m_sb.sb_blocksize > target->bt_bdev_awu_max)
- return false;
-
- return true;
+ return xfs_inode_buftarg(ip)->bt_bdev_awu_max > 0;
}
/*
diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
index 5ae77ffdc947b1..c1bd5654c3afa8 100644
--- a/fs/xfs/xfs_buf.c
+++ b/fs/xfs/xfs_buf.c
@@ -1779,6 +1779,40 @@ xfs_init_buftarg(
return -ENOMEM;
}
+/*
+ * Configure this buffer target for hardware-assisted atomic writes if the
+ * underlying block device supports is congruent with the filesystem geometry.
+ */
+void
+xfs_buftarg_config_atomic_writes(
+ struct xfs_buftarg *btp)
+{
+ struct xfs_mount *mp = btp->bt_mount;
+ unsigned int min_bytes, max_bytes;
+
+ ASSERT(btp->bt_bdev != NULL);
+
+ if (!bdev_can_atomic_write(btp->bt_bdev))
+ return;
+
+ min_bytes = bdev_atomic_write_unit_min_bytes(btp->bt_bdev);
+ max_bytes = bdev_atomic_write_unit_max_bytes(btp->bt_bdev);
+
+ /*
+ * Ignore atomic write geometry that is nonsense or doesn't even cover
+ * a single fsblock.
+ */
+ if (min_bytes > max_bytes ||
+ min_bytes > mp->m_sb.sb_blocksize ||
+ max_bytes < mp->m_sb.sb_blocksize) {
+ min_bytes = 0;
+ max_bytes = 0;
+ }
+
+ btp->bt_bdev_awu_min = min_bytes;
+ btp->bt_bdev_awu_max = max_bytes;
+}
+
struct xfs_buftarg *
xfs_alloc_buftarg(
struct xfs_mount *mp,
@@ -1799,13 +1833,6 @@ xfs_alloc_buftarg(
btp->bt_daxdev = fs_dax_get_by_bdev(btp->bt_bdev, &btp->bt_dax_part_off,
mp, ops);
- if (bdev_can_atomic_write(btp->bt_bdev)) {
- btp->bt_bdev_awu_min = bdev_atomic_write_unit_min_bytes(
- btp->bt_bdev);
- btp->bt_bdev_awu_max = bdev_atomic_write_unit_max_bytes(
- btp->bt_bdev);
- }
-
/*
* When allocating the buftargs we have not yet read the super block and
* thus don't know the file system sector size yet.
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index b2dd0c0bf50979..af4c541251d859 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -520,7 +520,8 @@ xfs_open_devices(
}
/*
- * Setup xfs_mount buffer target pointers based on superblock
+ * Setup xfs_mount buffer target pointers based on superblock, and configure
+ * the atomic write capabilities now that we've validated the blocksize.
*/
STATIC int
xfs_setup_devices(
@@ -531,6 +532,7 @@ xfs_setup_devices(
error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_sectsize);
if (error)
return error;
+ xfs_buftarg_config_atomic_writes(mp->m_ddev_targp);
if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
unsigned int log_sector_size = BBSIZE;
@@ -541,6 +543,7 @@ xfs_setup_devices(
log_sector_size);
if (error)
return error;
+ xfs_buftarg_config_atomic_writes(mp->m_logdev_targp);
}
if (mp->m_sb.sb_rtstart) {
@@ -555,6 +558,7 @@ xfs_setup_devices(
mp->m_sb.sb_sectsize);
if (error)
return error;
+ xfs_buftarg_config_atomic_writes(mp->m_rtdev_targp);
}
return 0;
next prev parent reply other threads:[~2025-04-30 5:18 UTC|newest]
Thread overview: 32+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-04-25 16:44 [PATCH v9 00/15] large atomic writes for xfs John Garry
2025-04-25 16:44 ` [PATCH v9 01/15] fs: add atomic write unit max opt to statx John Garry
2025-04-25 16:44 ` [PATCH v9 02/15] xfs: add helpers to compute log item overhead John Garry
2025-04-25 16:44 ` [PATCH v9 03/15] xfs: add helpers to compute transaction reservation for finishing intent items John Garry
2025-04-25 16:44 ` [PATCH v9 04/15] xfs: rename xfs_inode_can_atomicwrite() -> xfs_inode_can_hw_atomic_write() John Garry
2025-04-25 16:44 ` [PATCH v9 05/15] xfs: ignore HW which cannot atomic write a single block John Garry
2025-04-29 12:21 ` Christoph Hellwig
2025-04-29 14:44 ` Darrick J. Wong
2025-04-30 12:59 ` Christoph Hellwig
2025-05-01 16:22 ` Darrick J. Wong
2025-05-01 19:53 ` Darrick J. Wong
2025-04-30 5:18 ` Darrick J. Wong [this message]
2025-04-25 16:44 ` [PATCH v9 06/15] xfs: allow block allocator to take an alignment hint John Garry
2025-04-25 16:44 ` [PATCH v9 07/15] xfs: refactor xfs_reflink_end_cow_extent() John Garry
2025-04-25 16:44 ` [PATCH v9 08/15] xfs: refine atomic write size check in xfs_file_write_iter() John Garry
2025-04-25 16:44 ` [PATCH v9 09/15] xfs: add xfs_atomic_write_cow_iomap_begin() John Garry
2025-04-25 16:44 ` [PATCH v9 10/15] xfs: add large atomic writes checks in xfs_direct_write_iomap_begin() John Garry
2025-04-25 16:45 ` [PATCH v9 11/15] xfs: commit CoW-based atomic writes atomically John Garry
2025-04-25 16:45 ` [PATCH v9 12/15] xfs: add xfs_file_dio_write_atomic() John Garry
2025-04-25 16:45 ` [PATCH v9 13/15] xfs: add xfs_compute_atomic_write_unit_max() John Garry
2025-04-30 7:52 ` John Garry
2025-05-01 4:30 ` Darrick J. Wong
2025-05-01 5:00 ` John Garry
2025-05-01 16:23 ` Darrick J. Wong
2025-04-25 16:45 ` [PATCH v9 14/15] xfs: update atomic write limits John Garry
2025-04-25 16:45 ` [PATCH v9 15/15] xfs: allow sysadmins to specify a maximum atomic write limit at mount time John Garry
2025-04-29 12:22 ` Christoph Hellwig
2025-04-29 14:38 ` Darrick J. Wong
2025-04-30 14:14 ` [PATCH v9 00/15] large atomic writes for xfs John Garry
2025-05-01 4:31 ` Darrick J. Wong
2025-05-01 5:04 ` John Garry
2025-05-01 13:44 ` 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=20250430051822.GY25675@frogsfrogsfrogs \
--to=djwong@kernel.org \
--cc=brauner@kernel.org \
--cc=catherine.hoang@oracle.com \
--cc=cem@kernel.org \
--cc=dchinner@redhat.com \
--cc=hch@lst.de \
--cc=jack@suse.cz \
--cc=john.g.garry@oracle.com \
--cc=linux-api@vger.kernel.org \
--cc=linux-block@vger.kernel.org \
--cc=linux-ext4@vger.kernel.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 \
--cc=viro@zeniv.linux.org.uk \
/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).