]> www.infradead.org Git - users/dwmw2/linux.git/commit
btrfs: add io_uring command for encoded reads (ENCODED_READ ioctl)
authorMark Harmstone <maharmstone@fb.com>
Tue, 22 Oct 2024 14:50:20 +0000 (15:50 +0100)
committerDavid Sterba <dsterba@suse.com>
Mon, 11 Nov 2024 13:34:21 +0000 (14:34 +0100)
commit34310c442e175f286b4c06ab5caa4e0b267ea31c
tree184883ad5bc19c8ff81676e75c1b4dbaf5f66e90
parent68d3b27e05c7ca5545e88465f5e2be6eda0e11df
btrfs: add io_uring command for encoded reads (ENCODED_READ ioctl)

Add an io_uring command for encoded reads, using the same interface as
the existing BTRFS_IOC_ENCODED_READ ioctl.

btrfs_uring_encoded_read() is an io_uring version of
btrfs_ioctl_encoded_read(), which validates the user input and calls
btrfs_encoded_read() to read the appropriate metadata. If we determine
that we need to read an extent from disk, we call
btrfs_encoded_read_regular_fill_pages() through
btrfs_uring_read_extent() to prepare the bio.

The existing btrfs_encoded_read_regular_fill_pages() is changed so that
if it is passed a valid uring_ctx, rather than waking up any waiting
threads it calls btrfs_uring_read_extent_endio(). This in turn copies
the read data back to userspace, and calls io_uring_cmd_done() to
complete the io_uring command.

Because we're potentially doing a non-blocking read,
btrfs_uring_read_extent() doesn't clean up after itself if it returns
-EIOCBQUEUED. Instead, it allocates a priv struct, populates the fields
there that we will need to unlock the inode and free our allocations,
and defers this to the btrfs_uring_read_finished() that gets called when
the bio completes.

Signed-off-by: Mark Harmstone <maharmstone@fb.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/btrfs_inode.h
fs/btrfs/file.c
fs/btrfs/inode.c
fs/btrfs/ioctl.c
fs/btrfs/ioctl.h
fs/btrfs/send.c