On Fri, Nov 08, 2024 at 04:12:50PM -0800, Joanne Koong wrote:
> Add support for folios larger than one page size for non-writeback
> writes.
>
> Signed-off-by: Joanne Koong <joannelkoong@xxxxxxxxx>
> ---
> fs/fuse/file.c | 29 ++++++++++++++++++-----------
> 1 file changed, 18 insertions(+), 11 deletions(-)
>
> diff --git a/fs/fuse/file.c b/fs/fuse/file.c
> index a89fdc55a40b..6ee23ab9b7f2 100644
> --- a/fs/fuse/file.c
> +++ b/fs/fuse/file.c
> @@ -1146,19 +1146,15 @@ static ssize_t fuse_fill_write_pages(struct fuse_io_args *ia,
> num = min(num, max_pages << PAGE_SHIFT);
>
> ap->args.in_pages = true;
> - ap->descs[0].offset = offset;
>
> while (num) {
> size_t tmp;
> struct folio *folio;
> pgoff_t index = pos >> PAGE_SHIFT;
> - unsigned int bytes = min(PAGE_SIZE - offset, num);
> -
> - again:
> - err = -EFAULT;
> - if (fault_in_iov_iter_readable(ii, bytes))
> - break;
> + unsigned int bytes;
> + unsigned int folio_offset;
>
> + again:
> folio = __filemap_get_folio(mapping, index, FGP_WRITEBEGIN,
> mapping_gfp_mask(mapping));
> if (IS_ERR(folio)) {
> @@ -1166,10 +1162,20 @@ static ssize_t fuse_fill_write_pages(struct fuse_io_args *ia,
> break;
> }
>
> + folio_offset = ((index - folio->index) << PAGE_SHIFT) + offset;
> + bytes = min(folio_size(folio) - folio_offset, num);
> +
> + err = -EFAULT;
> + if (fault_in_iov_iter_readable(ii, bytes)) {
> + folio_unlock(folio);
> + folio_put(folio);
> + break;
> + }
This is a deadlock potentially. mmap the file, then buffered write to it with
the mmaped region and you will deadlock because fault_in_iov_iter_readable()
will try to lock the folio to submit the read, and you're already holding the
lock from the __filemap_get_folio() call.
If you look at iomap_write_iter it does a the fault_in_iov_iter_readable() call
before getting the folio, and it just uses the mapping_max_folio_size() value to
fault in larger chunks for whatever folio we manage to grab. Thanks,
Josef
