diff mbox series

[v3,16/16] xfs: add pre-content fsnotify hook for write faults

Message ID 89eb3a19d19c9b4bc19b6edbc708a8a33a911516.1723228772.git.josef@toxicpanda.com (mailing list archive)
State Superseded, archived
Headers show
Series fanotify: add pre-content hooks | expand

Commit Message

Josef Bacik Aug. 9, 2024, 6:44 p.m. UTC
xfs has it's own handling for write faults, so we need to add the
pre-content fsnotify hook for this case.  Reads go through filemap_fault
so they're handled properly there.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
---
 fs/xfs/xfs_file.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

Comments

Dave Chinner Aug. 9, 2024, 11:06 p.m. UTC | #1
On Fri, Aug 09, 2024 at 02:44:24PM -0400, Josef Bacik wrote:
> xfs has it's own handling for write faults, so we need to add the
> pre-content fsnotify hook for this case.  Reads go through filemap_fault
> so they're handled properly there.
> 
> Signed-off-by: Josef Bacik <josef@toxicpanda.com>
> ---
>  fs/xfs/xfs_file.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
> index 4cdc54dc9686..a00436dd29d1 100644
> --- a/fs/xfs/xfs_file.c
> +++ b/fs/xfs/xfs_file.c
> @@ -1328,8 +1328,13 @@ __xfs_filemap_fault(
>  
>  	trace_xfs_filemap_fault(XFS_I(inode), order, write_fault);
>  
> -	if (write_fault)
> -		return xfs_write_fault(vmf, order);
> +	if (write_fault) {
> +		vm_fault_t ret = filemap_maybe_emit_fsnotify_event(vmf);
> +		if (unlikely(ret))
> +			return ret;
> +		xfs_write_fault(vmf, order);
> +	}
> +
>  	if (IS_DAX(inode))
>  		return xfs_dax_read_fault(vmf, order);
>  	return filemap_fault(vmf);

Looks good now.

Acked-by: Dave Chinner <dchinner@redhat.com>
Christoph Hellwig Aug. 11, 2024, 8:40 a.m. UTC | #2
On Fri, Aug 09, 2024 at 02:44:24PM -0400, Josef Bacik wrote:
> -	if (write_fault)
> -		return xfs_write_fault(vmf, order);
> +	if (write_fault) {
> +		vm_fault_t ret = filemap_maybe_emit_fsnotify_event(vmf);
> +		if (unlikely(ret))
> +			return ret;
> +		xfs_write_fault(vmf, order);

This now loses the return value from xfs_write_fault and falls through
to actually do the read fault as well.  I somehow doubt this can work
at all as-is.

Please also move the filemap_maybe_emit_fsnotify_event call into
xfs_write_fault to keep the write fault handling isolated there,
which is the reason that it exists.
diff mbox series

Patch

diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index 4cdc54dc9686..a00436dd29d1 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -1328,8 +1328,13 @@  __xfs_filemap_fault(
 
 	trace_xfs_filemap_fault(XFS_I(inode), order, write_fault);
 
-	if (write_fault)
-		return xfs_write_fault(vmf, order);
+	if (write_fault) {
+		vm_fault_t ret = filemap_maybe_emit_fsnotify_event(vmf);
+		if (unlikely(ret))
+			return ret;
+		xfs_write_fault(vmf, order);
+	}
+
 	if (IS_DAX(inode))
 		return xfs_dax_read_fault(vmf, order);
 	return filemap_fault(vmf);