| Message ID | 20210510201201.1558972-1-willy@infradead.org (mailing list archive) |
|---|---|
| State | Accepted |
| Headers | show |
| Series | mm/filemap: Fix readahead return types | expand |
On Mon, May 10, 2021 at 09:12:01PM +0100, Matthew Wilcox (Oracle) wrote: > A readahead request will not allocate more memory than can be represented > by a size_t, even on systems that have HIGHMEM available. Change the > length functions from returning an loff_t to a size_t. > > Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> Looks reasonable to me; is this a 5.13 bugfix or just something that doesn't look right (i.e. save it for 5.14)? Reviewed-by: Darrick J. Wong <djwong@kernel.org> --D > --- > fs/iomap/buffered-io.c | 4 ++-- > include/linux/pagemap.h | 6 +++--- > 2 files changed, 5 insertions(+), 5 deletions(-) > > diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c > index f2cd2034a87b..9023717c5188 100644 > --- a/fs/iomap/buffered-io.c > +++ b/fs/iomap/buffered-io.c > @@ -394,7 +394,7 @@ void iomap_readahead(struct readahead_control *rac, const struct iomap_ops *ops) > { > struct inode *inode = rac->mapping->host; > loff_t pos = readahead_pos(rac); > - loff_t length = readahead_length(rac); > + size_t length = readahead_length(rac); > struct iomap_readpage_ctx ctx = { > .rac = rac, > }; > @@ -402,7 +402,7 @@ void iomap_readahead(struct readahead_control *rac, const struct iomap_ops *ops) > trace_iomap_readahead(inode, readahead_count(rac)); > > while (length > 0) { > - loff_t ret = iomap_apply(inode, pos, length, 0, ops, > + ssize_t ret = iomap_apply(inode, pos, length, 0, ops, > &ctx, iomap_readahead_actor); > if (ret <= 0) { > WARN_ON_ONCE(ret == 0); > diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h > index a4bd41128bf3..e89df447fae3 100644 > --- a/include/linux/pagemap.h > +++ b/include/linux/pagemap.h > @@ -997,9 +997,9 @@ static inline loff_t readahead_pos(struct readahead_control *rac) > * readahead_length - The number of bytes in this readahead request. > * @rac: The readahead request. > */ > -static inline loff_t readahead_length(struct readahead_control *rac) > +static inline size_t readahead_length(struct readahead_control *rac) > { > - return (loff_t)rac->_nr_pages * PAGE_SIZE; > + return rac->_nr_pages * PAGE_SIZE; > } > > /** > @@ -1024,7 +1024,7 @@ static inline unsigned int readahead_count(struct readahead_control *rac) > * readahead_batch_length - The number of bytes in the current batch. > * @rac: The readahead request. > */ > -static inline loff_t readahead_batch_length(struct readahead_control *rac) > +static inline size_t readahead_batch_length(struct readahead_control *rac) > { > return rac->_batch_count * PAGE_SIZE; > } > -- > 2.30.2 >
On Mon, May 10, 2021 at 03:27:56PM -0700, Darrick J. Wong wrote: > On Mon, May 10, 2021 at 09:12:01PM +0100, Matthew Wilcox (Oracle) wrote: > > A readahead request will not allocate more memory than can be represented > > by a size_t, even on systems that have HIGHMEM available. Change the > > length functions from returning an loff_t to a size_t. > > > > Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> > > Looks reasonable to me; is this a 5.13 bugfix or just something that > doesn't look right (i.e. save it for 5.14)? > > Reviewed-by: Darrick J. Wong <djwong@kernel.org> Thanks! Linus was unhappy about it, and I promised to fix it. I leave it up to Andrew whether he wants to see this in 5.13 or 5.14. https://lore.kernel.org/linux-btrfs/CAHk-=wj1KRvb=hie1VUTGo1D_ckD+Suo0-M2Nh5Kek1Wu=2Ppw@mail.gmail.com/ (I went a little beyond what he was directly unhappy with and reviewed all the readahead.*length users for unnecessary loff_t usage)
diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c index f2cd2034a87b..9023717c5188 100644 --- a/fs/iomap/buffered-io.c +++ b/fs/iomap/buffered-io.c @@ -394,7 +394,7 @@ void iomap_readahead(struct readahead_control *rac, const struct iomap_ops *ops) { struct inode *inode = rac->mapping->host; loff_t pos = readahead_pos(rac); - loff_t length = readahead_length(rac); + size_t length = readahead_length(rac); struct iomap_readpage_ctx ctx = { .rac = rac, }; @@ -402,7 +402,7 @@ void iomap_readahead(struct readahead_control *rac, const struct iomap_ops *ops) trace_iomap_readahead(inode, readahead_count(rac)); while (length > 0) { - loff_t ret = iomap_apply(inode, pos, length, 0, ops, + ssize_t ret = iomap_apply(inode, pos, length, 0, ops, &ctx, iomap_readahead_actor); if (ret <= 0) { WARN_ON_ONCE(ret == 0); diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h index a4bd41128bf3..e89df447fae3 100644 --- a/include/linux/pagemap.h +++ b/include/linux/pagemap.h @@ -997,9 +997,9 @@ static inline loff_t readahead_pos(struct readahead_control *rac) * readahead_length - The number of bytes in this readahead request. * @rac: The readahead request. */ -static inline loff_t readahead_length(struct readahead_control *rac) +static inline size_t readahead_length(struct readahead_control *rac) { - return (loff_t)rac->_nr_pages * PAGE_SIZE; + return rac->_nr_pages * PAGE_SIZE; } /** @@ -1024,7 +1024,7 @@ static inline unsigned int readahead_count(struct readahead_control *rac) * readahead_batch_length - The number of bytes in the current batch. * @rac: The readahead request. */ -static inline loff_t readahead_batch_length(struct readahead_control *rac) +static inline size_t readahead_batch_length(struct readahead_control *rac) { return rac->_batch_count * PAGE_SIZE; }
A readahead request will not allocate more memory than can be represented by a size_t, even on systems that have HIGHMEM available. Change the length functions from returning an loff_t to a size_t. Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> --- fs/iomap/buffered-io.c | 4 ++-- include/linux/pagemap.h | 6 +++--- 2 files changed, 5 insertions(+), 5 deletions(-)