Great! I can confirm this indeed fixes my reproducer, and the iov_iter
now looks sensible in gdb as far as I can judge.
Tested-by: Arnout Engelen <arnout@xxxxxxxx>
On Mon, Aug 11, 2025, at 09:39, Dominique Martinet via B4 Relay wrote:
> From: Dominique Martinet <asmadeus@xxxxxxxxxxxxx>
>
> After advancing into a folioq it makes more sense to point to the next
> slot than at the end of the current slot.
> This should not be needed for correctness, but this also happens to
> "fix" the 9p bug with iterate_folioq() not copying properly.
>
> Signed-off-by: Dominique Martinet <asmadeus@xxxxxxxxxxxxx>
> ---
> lib/iov_iter.c | 6 +++---
> 1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/lib/iov_iter.c b/lib/iov_iter.c
> index f9193f952f49945297479483755d68a34c6d4ffe..65c05134ab934e1e0bf5d010fff22983bfe9c680 100644
> --- a/lib/iov_iter.c
> +++ b/lib/iov_iter.c
> @@ -1032,9 +1032,6 @@ static ssize_t iter_folioq_get_pages(struct iov_iter *iter,
> maxpages--;
> }
>
> - if (maxpages == 0 || extracted >= maxsize)
> - break;
> -
> if (iov_offset >= fsize) {
> iov_offset = 0;
> slot++;
> @@ -1043,6 +1040,9 @@ static ssize_t iter_folioq_get_pages(struct iov_iter *iter,
> slot = 0;
> }
> }
> +
> + if (maxpages == 0 || extracted >= maxsize)
> + break;
> }
>
> iter->count = count;
>
> --
> 2.50.1
>
>
>
--
Arnout Engelen
Engelen Open Source
https://engelen.eu
