The loop driver assumes that if the passed in fd is opened with
O_DIRECT, the caller wants to use direct I/O on the loop device.
However, if the underlying filesystem has a different block size than
the loop block queue, direct I/O can't be enabled. Instead of requiring
userspace to manually change the blocksize and re-enable direct I/O,
just change the queue block size to match.
Signed-off-by: Martijn Coenen <[email protected]>
---
drivers/block/loop.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index ab7ca5989097a..1138162ff6c4d 100644
--- a/drivers/block/loop.c
+++ b/drivers/block/loop.c
@@ -994,6 +994,12 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
if (!(lo_flags & LO_FLAGS_READ_ONLY) && file->f_op->fsync)
blk_queue_write_cache(lo->lo_queue, true, false);
+ if (io_is_direct(lo->lo_backing_file) && inode->i_sb->s_bdev) {
+ /* In case of direct I/O, match underlying block size */
+ blk_queue_logical_block_size(lo->lo_queue,
+ bdev_logical_block_size(inode->i_sb->s_bdev));
+ }
+
loop_update_rotational(lo);
loop_update_dio(lo);
set_capacity(lo->lo_disk, size);
--
2.23.0.187.g17f5b7556c-goog
On Wed, Aug 28, 2019 at 12:32:29PM +0200, Martijn Coenen wrote:
> The loop driver assumes that if the passed in fd is opened with
> O_DIRECT, the caller wants to use direct I/O on the loop device.
> However, if the underlying filesystem has a different block size than
> the loop block queue, direct I/O can't be enabled. Instead of requiring
> userspace to manually change the blocksize and re-enable direct I/O,
> just change the queue block size to match.
Why can't we enable the block device in that case? All the usual
block filesystems support 512 byte aligned direct I/O with a 4k
file system block size (as long as the underlying block device
sector size is also 512 bytes).
On Fri, Aug 30, 2019 at 5:50 PM Christoph Hellwig <[email protected]> wrote:
>
> On Wed, Aug 28, 2019 at 12:32:29PM +0200, Martijn Coenen wrote:
> > The loop driver assumes that if the passed in fd is opened with
> > O_DIRECT, the caller wants to use direct I/O on the loop device.
> > However, if the underlying filesystem has a different block size than
> > the loop block queue, direct I/O can't be enabled. Instead of requiring
> > userspace to manually change the blocksize and re-enable direct I/O,
> > just change the queue block size to match.
>
> Why can't we enable the block device in that case? All the usual
> block filesystems support 512 byte aligned direct I/O with a 4k
> file system block size (as long as the underlying block device
> sector size is also 512 bytes).
Sorry, I didn't word that correctly: it's not the logical block size
of the filesystem, but the logical block size of the underlying block
device that loop's queue must match (or exceed). With the current loop
code, if the backing file is opened with O_DIRECT and resides on a
block device with a 512 bytes logical block size, the loop device will
correctly use direct I/O. If instead the backing file happened to
reside on a block device with a 4k logical block size, the loop device
would silently fall back to cached mode. I think there's a benefit in
the behavior being consistent independent of the block size of the
backing device, and I don't see a good reason for not automatically
switching loop's logical/physical queue sizes to match the backing
device in this specific case.
Will send a v2.
Thanks,
Martijn
The loop driver assumes that if the passed in fd is opened with
O_DIRECT, the caller wants to use direct I/O on the loop device.
However, if the underlying block device has a different block size than
the loop block queue, direct I/O can't be enabled. Instead of requiring
userspace to manually change the blocksize and re-enable direct I/O,
just change the queue block sizes to match, as well as the io_min size.
Signed-off-by: Martijn Coenen <[email protected]>
---
v2 changes:
- Fixed commit message to say the block size must match the underlying
block device, not the underlying filesystem.
- Also set physical blocksize and minimal io size correspondingly.
drivers/block/loop.c | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index ab7ca5989097a..b547182037af2 100644
--- a/drivers/block/loop.c
+++ b/drivers/block/loop.c
@@ -994,6 +994,16 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
if (!(lo_flags & LO_FLAGS_READ_ONLY) && file->f_op->fsync)
blk_queue_write_cache(lo->lo_queue, true, false);
+ if (io_is_direct(lo->lo_backing_file) && inode->i_sb->s_bdev) {
+ /* In case of direct I/O, match underlying block size */
+ unsigned short bsize = bdev_logical_block_size(
+ inode->i_sb->s_bdev);
+
+ blk_queue_logical_block_size(lo->lo_queue, bsize);
+ blk_queue_physical_block_size(lo->lo_queue, bsize);
+ blk_queue_io_min(lo->lo_queue, bsize);
+ }
+
loop_update_rotational(lo);
loop_update_dio(lo);
set_capacity(lo->lo_disk, size);
--
2.23.0.187.g17f5b7556c-goog
Hi Jens, Ming,
Do you have any thoughts about this patch?
Thanks,
Martijn
On Wed, Sep 4, 2019 at 9:49 PM Martijn Coenen <[email protected]> wrote:
>
> The loop driver assumes that if the passed in fd is opened with
> O_DIRECT, the caller wants to use direct I/O on the loop device.
> However, if the underlying block device has a different block size than
> the loop block queue, direct I/O can't be enabled. Instead of requiring
> userspace to manually change the blocksize and re-enable direct I/O,
> just change the queue block sizes to match, as well as the io_min size.
>
> Signed-off-by: Martijn Coenen <[email protected]>
> ---
> v2 changes:
> - Fixed commit message to say the block size must match the underlying
> block device, not the underlying filesystem.
> - Also set physical blocksize and minimal io size correspondingly.
>
> drivers/block/loop.c | 10 ++++++++++
> 1 file changed, 10 insertions(+)
>
> diff --git a/drivers/block/loop.c b/drivers/block/loop.c
> index ab7ca5989097a..b547182037af2 100644
> --- a/drivers/block/loop.c
> +++ b/drivers/block/loop.c
> @@ -994,6 +994,16 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
> if (!(lo_flags & LO_FLAGS_READ_ONLY) && file->f_op->fsync)
> blk_queue_write_cache(lo->lo_queue, true, false);
>
> + if (io_is_direct(lo->lo_backing_file) && inode->i_sb->s_bdev) {
> + /* In case of direct I/O, match underlying block size */
> + unsigned short bsize = bdev_logical_block_size(
> + inode->i_sb->s_bdev);
> +
> + blk_queue_logical_block_size(lo->lo_queue, bsize);
> + blk_queue_physical_block_size(lo->lo_queue, bsize);
> + blk_queue_io_min(lo->lo_queue, bsize);
> + }
> +
> loop_update_rotational(lo);
> loop_update_dio(lo);
> set_capacity(lo->lo_disk, size);
> --
> 2.23.0.187.g17f5b7556c-goog
>
Looks fine:
Reviewed-by: Christoph Hellwig <[email protected]>
On Mon, Sep 30, 2019 at 10:24 AM Christoph Hellwig <[email protected]> wrote:
>
> Looks fine:
>
> Reviewed-by: Christoph Hellwig <[email protected]>
Thanks Christoph! Jens, Ming, are you also ok with taking this?
On 9/4/19 1:49 PM, Martijn Coenen wrote:
> The loop driver assumes that if the passed in fd is opened with
> O_DIRECT, the caller wants to use direct I/O on the loop device.
> However, if the underlying block device has a different block size than
> the loop block queue, direct I/O can't be enabled. Instead of requiring
> userspace to manually change the blocksize and re-enable direct I/O,
> just change the queue block sizes to match, as well as the io_min size.
Applied, thanks.
--
Jens Axboe