On Fri, 13 Aug 2021, David Howells wrote:
> Trying to use a swapfile on NFS results in every DIO write failing with
> ETXTBSY because generic_write_checks(), as called by nfs_direct_write()
> from nfs_direct_IO(), forbids writes to swapfiles.
Could we just remove this check from generic_write_checks(), and instead
call deny_write_access() in swap_on?
Then user-space wouldn't be able to open a swap-file for write, so there
would be no need to check on every single write.
NeilBrown
>
> Fix this by introducing a new kiocb flag, IOCB_SWAP, that's set by the swap
> code to indicate that the swapper is doing this operation and so overrule
> the check in generic_write_checks().
>
> Without this patch, the following is seen:
>
> Write error on dio swapfile (3800334336)
>
> Altering __swap_writepage() to show the error shows:
>
> Write error (-26) on dio swapfile (3800334336)
>
> Tested by swapping off all swap partitions and then swapping on a prepared
> NFS file (CONFIG_NFS_SWAP=y is also needed). Enough copies of the
> following program then need to be run to force swapping to occur (at least
> one per gigabyte of RAM):
>
> #include <stdbool.h>
> #include <stdio.h>
> #include <stdlib.h>
> #include <unistd.h>
> #include <sys/mman.h>
> int main()
> {
> unsigned int pid = getpid(), iterations = 0;
> size_t i, j, size = 1024 * 1024 * 1024;
> char *p;
> bool mismatch;
> p = malloc(size);
> if (!p) {
> perror("malloc");
> exit(1);
> }
> srand(pid);
> for (i = 0; i < size; i += 4)
> *(unsigned int *)(p + i) = rand();
> do {
> for (j = 0; j < 16; j++) {
> for (i = 0; i < size; i += 4096)
> *(unsigned int *)(p + i) += 1;
> iterations++;
> }
> mismatch = false;
> srand(pid);
> for (i = 0; i < size; i += 4) {
> unsigned int r = rand();
> unsigned int v = *(unsigned int *)(p + i);
> if (i % 4096 == 0)
> v -= iterations;
> if (v != r) {
> fprintf(stderr, "mismatch %zx: %x != %x (diff %x)\n",
> i, v, r, v - r);
> mismatch = true;
> }
> }
> } while (!mismatch);
> exit(1);
> }
>
>
> Fixes: dc617f29dbe5 ("vfs: don't allow writes to swap files")
> Signed-off-by: David Howells <dhowells@xxxxxxxxxx>
> cc: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
> cc: Christoph Hellwig <hch@xxxxxx>
> cc: Trond Myklebust <trond.myklebust@xxxxxxxxxxxxxxx>
> cc: linux-nfs@xxxxxxxxxxxxxxx
> ---
>
> fs/read_write.c | 2 +-
> include/linux/fs.h | 1 +
> mm/page_io.c | 7 ++++---
> 3 files changed, 6 insertions(+), 4 deletions(-)
>
> diff --git a/fs/read_write.c b/fs/read_write.c
> index 9db7adf160d2..daef721ca67e 100644
> --- a/fs/read_write.c
> +++ b/fs/read_write.c
> @@ -1646,7 +1646,7 @@ ssize_t generic_write_checks(struct kiocb *iocb, struct iov_iter *from)
> loff_t count;
> int ret;
>
> - if (IS_SWAPFILE(inode))
> + if (IS_SWAPFILE(inode) && !(iocb->ki_flags & IOCB_SWAP))
> return -ETXTBSY;
>
> if (!iov_iter_count(from))
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index 640574294216..b3e6a20f28ef 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -319,6 +319,7 @@ enum rw_hint {
> /* iocb->ki_waitq is valid */
> #define IOCB_WAITQ (1 << 19)
> #define IOCB_NOIO (1 << 20)
> +#define IOCB_SWAP (1 << 21) /* This is a swap request */
>
> struct kiocb {
> struct file *ki_filp;
> diff --git a/mm/page_io.c b/mm/page_io.c
> index d597bc6e6e45..edb72bf624d2 100644
> --- a/mm/page_io.c
> +++ b/mm/page_io.c
> @@ -303,7 +303,8 @@ int __swap_writepage(struct page *page, struct writeback_control *wbc,
>
> iov_iter_bvec(&from, WRITE, &bv, 1, PAGE_SIZE);
> init_sync_kiocb(&kiocb, swap_file);
> - kiocb.ki_pos = page_file_offset(page);
> + kiocb.ki_pos = page_file_offset(page);
> + kiocb.ki_flags = IOCB_DIRECT | IOCB_WRITE | IOCB_SWAP;
>
> set_page_writeback(page);
> unlock_page(page);
> @@ -324,8 +325,8 @@ int __swap_writepage(struct page *page, struct writeback_control *wbc,
> */
> set_page_dirty(page);
> ClearPageReclaim(page);
> - pr_err_ratelimited("Write error on dio swapfile (%llu)\n",
> - page_file_offset(page));
> + pr_err_ratelimited("Write error (%d) on dio swapfile (%llu)\n",
> + ret, page_file_offset(page));
> }
> end_page_writeback(page);
> return ret;
>
>
>
