2022-10-26 16:58:43

by Keith Busch

Subject: [PATCH] iomap: directly use logical block size

From: Keith Busch <[email protected]>

Don't transform the logical block size to a bit shift only to shift it
back to the original block size. Just use the size.

Signed-off-by: Keith Busch <[email protected]>
---
fs/iomap/direct-io.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/fs/iomap/direct-io.c b/fs/iomap/direct-io.c
index 4eb559a16c9e..503b97e5a115 100644
--- a/fs/iomap/direct-io.c
+++ b/fs/iomap/direct-io.c
@@ -240,7 +240,7 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
{
const struct iomap *iomap = &iter->iomap;
struct inode *inode = iter->inode;
- unsigned int blkbits = blksize_bits(bdev_logical_block_size(iomap->bdev));
+ unsigned int blksz = bdev_logical_block_size(iomap->bdev);
unsigned int fs_block_size = i_blocksize(inode), pad;
loff_t length = iomap_length(iter);
loff_t pos = iter->pos;
@@ -252,7 +252,7 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
size_t copied = 0;
size_t orig_count;

- if ((pos | length) & ((1 << blkbits) - 1) ||
+ if ((pos | length) & (blksz - 1) ||
!bdev_iter_is_aligned(iomap->bdev, dio->submit.iter))
return -EINVAL;

--
2.30.2

2022-10-26 18:19:29

by Chaitanya Kulkarni

Subject: Re: [PATCH] iomap: directly use logical block size

On 10/26/2022 9:51 AM, Keith Busch wrote:
> From: Keith Busch <[email protected]>
>
> Don't transform the logical block size to a bit shift only to shift it
> back to the original block size. Just use the size.
>
> Signed-off-by: Keith Busch <[email protected]>
> ---

we should also change any other callers like this.

Reviewed-by: Chaitanya Kulkarni <[email protected]>

-ck

2022-10-26 18:52:30

by Darrick J. Wong

Subject: Re: [PATCH] iomap: directly use logical block size

On Wed, Oct 26, 2022 at 09:51:33AM -0700, Keith Busch wrote:
> From: Keith Busch <[email protected]>
>
> Don't transform the logical block size to a bit shift only to shift it
> back to the original block size. Just use the size.
>
> Signed-off-by: Keith Busch <[email protected]>
> ---
> fs/iomap/direct-io.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/fs/iomap/direct-io.c b/fs/iomap/direct-io.c
> index 4eb559a16c9e..503b97e5a115 100644
> --- a/fs/iomap/direct-io.c
> +++ b/fs/iomap/direct-io.c
> @@ -240,7 +240,7 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
> {
> const struct iomap *iomap = &iter->iomap;
> struct inode *inode = iter->inode;
> - unsigned int blkbits = blksize_bits(bdev_logical_block_size(iomap->bdev));
> + unsigned int blksz = bdev_logical_block_size(iomap->bdev);

/me looks at what blksize_bits does (assumes block size > 256) and rolls
his eyes.

Regardless, this looks correct to me, so

Reviewed-by: Darrick J. Wong <[email protected]>

--D


> unsigned int fs_block_size = i_blocksize(inode), pad;
> loff_t length = iomap_length(iter);
> loff_t pos = iter->pos;
> @@ -252,7 +252,7 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
> size_t copied = 0;
> size_t orig_count;
>
> - if ((pos | length) & ((1 << blkbits) - 1) ||
> + if ((pos | length) & (blksz - 1) ||
> !bdev_iter_is_aligned(iomap->bdev, dio->submit.iter))
> return -EINVAL;
>
> --
> 2.30.2
>

2022-10-28 21:46:39

by Bart Van Assche

Subject: Re: [PATCH] iomap: directly use logical block size

On 10/26/22 09:51, Keith Busch wrote:
> From: Keith Busch <[email protected]>
>
> Don't transform the logical block size to a bit shift only to shift it
> back to the original block size. Just use the size.
>
> Signed-off-by: Keith Busch <[email protected]>
> ---
> fs/iomap/direct-io.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/fs/iomap/direct-io.c b/fs/iomap/direct-io.c
> index 4eb559a16c9e..503b97e5a115 100644
> --- a/fs/iomap/direct-io.c
> +++ b/fs/iomap/direct-io.c
> @@ -240,7 +240,7 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
> {
> const struct iomap *iomap = &iter->iomap;
> struct inode *inode = iter->inode;
> - unsigned int blkbits = blksize_bits(bdev_logical_block_size(iomap->bdev));
> + unsigned int blksz = bdev_logical_block_size(iomap->bdev);
> unsigned int fs_block_size = i_blocksize(inode), pad;
> loff_t length = iomap_length(iter);
> loff_t pos = iter->pos;
> @@ -252,7 +252,7 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
> size_t copied = 0;
> size_t orig_count;
>
> - if ((pos | length) & ((1 << blkbits) - 1) ||
> + if ((pos | length) & (blksz - 1) ||
> !bdev_iter_is_aligned(iomap->bdev, dio->submit.iter))
> return -EINVAL;

Reviewed-by: Bart Van Assche <[email protected]>

2022-10-30 08:55:07

by Christoph Hellwig

Subject: Re: [PATCH] iomap: directly use logical block size

On Wed, Oct 26, 2022 at 09:51:33AM -0700, Keith Busch wrote:
> From: Keith Busch <[email protected]>
>
> Don't transform the logical block size to a bit shift only to shift it
> back to the original block size. Just use the size.

This looks reasonable. But given that the blksz variable is only used
once, why not open code the bdev_logical_block_size in the conditional?