| Message ID | 20190410061006.GA18813@kadam (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | block: make blk_queue_zone_sectors() return sector_t type | expand |
On 2019/04/10 15:10, Dan Carpenter wrote: > My static checker complains about this line from dmz_get_zoned_device() > > aligned_capacity = dev->capacity & ~(blk_queue_zone_sectors(q) - 1); > > The problem is that "aligned_capacity" and "dev->capacity" are sector_t > type (which is a u64) but blk_queue_zone_sectors(q) returns a u32 so the > higher 32 bits in aligned_capacity are always cleared to zero. > > Declaring blk_queue_zone_sectors() as a sector_t addresses this warning > and it feels intuitive based on the function name. I updated > bdev_zone_sectors() as well just to be consistent. > > Fixes: 114e025968b5 ("dm zoned: ignore last smaller runt zone") > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> > --- > Please review this one extra carefully. I'm not positive it's correct. > > include/linux/blkdev.h | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h > index 4b85dc066264..1785a7f506be 100644 > --- a/include/linux/blkdev.h > +++ b/include/linux/blkdev.h > @@ -670,7 +670,7 @@ static inline bool blk_queue_is_zoned(struct request_queue *q) > } > } > > -static inline unsigned int blk_queue_zone_sectors(struct request_queue *q) > +static inline sector_t blk_queue_zone_sectors(struct request_queue *q) > { > return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0; > } > @@ -1419,7 +1419,7 @@ static inline bool bdev_is_zoned(struct block_device *bdev) > return false; > } > > -static inline unsigned int bdev_zone_sectors(struct block_device *bdev) > +static inline sector_t bdev_zone_sectors(struct block_device *bdev) > { > struct request_queue *q = bdev_get_queue(bdev); Indeed, using sector_t instead of unsigned int is more intuitive and changing to sector_t matches the 64bits zone length in struct blk_zone too, so no problem on this front. However, q->limits.chunk_sectors is an unsigned int, which is why I used that type, to reflect the fact that even though the API allows values larger than 4G sectors, the handling through the queue limits truncates that to at most 4G-1 sectors. Furthermore, the unsigned int type is used wherever blk_queue_zone_sectors() and bdev_zone_sectors() are called (block/ioctl.c, block/blk-zoned.c, f2fs, etc). That needs in depth checking. I would rather keep the unsigned int since the queue limits will give only 32 bit values anyway. What about a cast for the aligned capacity line ? Something like: aligned_capacity = dev->capacity & ~((u64)blk_queue_zone_sectors(q) - 1); Would that make your checker happy ?
On Wed, Apr 10, 2019 at 07:20:27AM +0000, Damien Le Moal wrote: > On 2019/04/10 15:10, Dan Carpenter wrote: > > My static checker complains about this line from dmz_get_zoned_device() > > > > aligned_capacity = dev->capacity & ~(blk_queue_zone_sectors(q) - 1); > > > > The problem is that "aligned_capacity" and "dev->capacity" are sector_t > > type (which is a u64) but blk_queue_zone_sectors(q) returns a u32 so the > > higher 32 bits in aligned_capacity are always cleared to zero. > > > > Declaring blk_queue_zone_sectors() as a sector_t addresses this warning > > and it feels intuitive based on the function name. I updated > > bdev_zone_sectors() as well just to be consistent. > > > > Fixes: 114e025968b5 ("dm zoned: ignore last smaller runt zone") > > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> > > --- > > Please review this one extra carefully. I'm not positive it's correct. > > > > include/linux/blkdev.h | 4 ++-- > > 1 file changed, 2 insertions(+), 2 deletions(-) > > > > diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h > > index 4b85dc066264..1785a7f506be 100644 > > --- a/include/linux/blkdev.h > > +++ b/include/linux/blkdev.h > > @@ -670,7 +670,7 @@ static inline bool blk_queue_is_zoned(struct request_queue *q) > > } > > } > > > > -static inline unsigned int blk_queue_zone_sectors(struct request_queue *q) > > +static inline sector_t blk_queue_zone_sectors(struct request_queue *q) > > { > > return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0; > > } > > @@ -1419,7 +1419,7 @@ static inline bool bdev_is_zoned(struct block_device *bdev) > > return false; > > } > > > > -static inline unsigned int bdev_zone_sectors(struct block_device *bdev) > > +static inline sector_t bdev_zone_sectors(struct block_device *bdev) > > { > > struct request_queue *q = bdev_get_queue(bdev); > > Indeed, using sector_t instead of unsigned int is more intuitive and changing to > sector_t matches the 64bits zone length in struct blk_zone too, so no problem on > this front. > > However, q->limits.chunk_sectors is an unsigned int, which is why I used that > type, to reflect the fact that even though the API allows values larger than 4G > sectors, the handling through the queue limits truncates that to at most 4G-1 > sectors. Furthermore, the unsigned int type is used wherever > blk_queue_zone_sectors() and bdev_zone_sectors() are called (block/ioctl.c, > block/blk-zoned.c, f2fs, etc). That needs in depth checking. > > I would rather keep the unsigned int since the queue limits will give only 32 > bit values anyway. What about a cast for the aligned capacity line ? Something like: > > aligned_capacity = dev->capacity & ~((u64)blk_queue_zone_sectors(q) - 1); > > Would that make your checker happy ? > Yeah. Or a sector_t cast. aligned_capacity = dev->capacity & ~((sector_t)blk_queue_zone_sectors(q) - 1); regards, dan carpenter
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 4b85dc066264..1785a7f506be 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -670,7 +670,7 @@ static inline bool blk_queue_is_zoned(struct request_queue *q) } } -static inline unsigned int blk_queue_zone_sectors(struct request_queue *q) +static inline sector_t blk_queue_zone_sectors(struct request_queue *q) { return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0; } @@ -1419,7 +1419,7 @@ static inline bool bdev_is_zoned(struct block_device *bdev) return false; } -static inline unsigned int bdev_zone_sectors(struct block_device *bdev) +static inline sector_t bdev_zone_sectors(struct block_device *bdev) { struct request_queue *q = bdev_get_queue(bdev);
My static checker complains about this line from dmz_get_zoned_device() aligned_capacity = dev->capacity & ~(blk_queue_zone_sectors(q) - 1); The problem is that "aligned_capacity" and "dev->capacity" are sector_t type (which is a u64) but blk_queue_zone_sectors(q) returns a u32 so the higher 32 bits in aligned_capacity are always cleared to zero. Declaring blk_queue_zone_sectors() as a sector_t addresses this warning and it feels intuitive based on the function name. I updated bdev_zone_sectors() as well just to be consistent. Fixes: 114e025968b5 ("dm zoned: ignore last smaller runt zone") Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> --- Please review this one extra carefully. I'm not positive it's correct. include/linux/blkdev.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)