On 6/7/21 12:29 AM, Al Viro wrote:
> On Sun, Jun 06, 2021 at 03:05:49PM -0700, Linus Torvalds wrote:
[...]
>
> BTW, speaking of initializers... Pavel, could you check if
> the following breaks anything? Unless I'm misreading __io_import_fixed(),
> looks like that's what it's trying to do...
It's a version of iov_iter_advance() that assumes all bvecs are
single-paged and all possibly besides first/last are page
aligned/sized.
Looks and works well, will try the full set later.
btw, as that assumption is not true in general, I'd suggest to add
a comment. Don't like the idea of it being misused...
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index f46acbbeed57..9bd2da9a4c3d 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -2773,57 +2773,14 @@ static int __io_import_fixed(struct io_kiocb *req, int rw, struct iov_iter *iter
> {
> size_t len = req->rw.len;
> u64 buf_end, buf_addr = req->rw.addr;
> - size_t offset;
>
> if (unlikely(check_add_overflow(buf_addr, (u64)len, &buf_end)))
> return -EFAULT;
> /* not inside the mapped region */
> if (unlikely(buf_addr < imu->ubuf || buf_end > imu->ubuf_end))
> return -EFAULT;
> -
> - /*
> - * May not be a start of buffer, set size appropriately
> - * and advance us to the beginning.
> - */
> - offset = buf_addr - imu->ubuf;
> - iov_iter_bvec(iter, rw, imu->bvec, imu->nr_bvecs, offset + len);
> -
> - if (offset) {
> - /*
> - * Don't use iov_iter_advance() here, as it's really slow for
> - * using the latter parts of a big fixed buffer - it iterates
> - * over each segment manually. We can cheat a bit here, because
> - * we know that:
> - *
> - * 1) it's a BVEC iter, we set it up
> - * 2) all bvecs are PAGE_SIZE in size, except potentially the
> - * first and last bvec
> - *
> - * So just find our index, and adjust the iterator afterwards.
> - * If the offset is within the first bvec (or the whole first
> - * bvec, just use iov_iter_advance(). This makes it easier
> - * since we can just skip the first segment, which may not
> - * be PAGE_SIZE aligned.
> - */
> - const struct bio_vec *bvec = imu->bvec;
> -
> - if (offset <= bvec->bv_len) {
> - iov_iter_advance(iter, offset);
> - } else {
> - unsigned long seg_skip;
> -
> - /* skip first vec */
> - offset -= bvec->bv_len;
> - seg_skip = 1 + (offset >> PAGE_SHIFT);
> -
> - iter->bvec = bvec + seg_skip;
> - iter->nr_segs -= seg_skip;
> - iter->count -= bvec->bv_len + offset;
> - iter->iov_offset = offset & ~PAGE_MASK;
> - }
> - }
> -
> - return 0;
> + return import_pagevec(rw, buf_addr, len, imu->ubuf,
> + imu->nr_bvecs, imu->bvec, iter);
> }
>
> static int io_import_fixed(struct io_kiocb *req, int rw, struct iov_iter *iter)
> diff --git a/include/linux/uio.h b/include/linux/uio.h
> index fd88d9911dad..f6291e981d07 100644
> --- a/include/linux/uio.h
> +++ b/include/linux/uio.h
> @@ -299,5 +299,8 @@ ssize_t __import_iovec(int type, const struct iovec __user *uvec,
> struct iov_iter *i, bool compat);
> int import_single_range(int type, void __user *buf, size_t len,
> struct iovec *iov, struct iov_iter *i);
> +int import_pagevec(int rw, unsigned long from, size_t len,
> + unsigned long base, unsigned nr_pages,
> + struct bio_vec *bvec, struct iov_iter *i);
>
> #endif
> diff --git a/lib/iov_iter.c b/lib/iov_iter.c
> index 11b39bd1d1ab..4a771fcb529b 100644
> --- a/lib/iov_iter.c
> +++ b/lib/iov_iter.c
> @@ -1982,3 +1982,21 @@ int import_single_range(int rw, void __user *buf, size_t len,
> return 0;
> }
> EXPORT_SYMBOL(import_single_range);
> +
> +int import_pagevec(int rw, unsigned long from, size_t len,
> + unsigned long base, unsigned nr_pages,
> + struct bio_vec *bvec, struct iov_iter *i)
> +
> +{
> + unsigned long skip_pages = (from >> PAGE_SHIFT) - (base >> PAGE_SHIFT);
> +
> + *i = (struct iov_iter){
> + .iter_type = ITER_BVEC,
> + .data_source = rw,
> + .bvec = bvec + skip_pages,
> + .nr_segs = nr_pages - skip_pages,
> + .iov_offset = skip_pages ? from & ~PAGE_MASK : from - base,
> + .count = len
> + };
> + return 0;
> +}
>
--
Pavel Begunkov
