On Wed, Oct 29, 2014 at 05:35:19PM +0100, Andrea Arcangeli wrote:
> This allows those get_user_pages calls to pass FAULT_FLAG_ALLOW_RETRY
> to the page fault in order to release the mmap_sem during the I/O.
>
> Signed-off-by: Andrea Arcangeli <aarcange@xxxxxxxxxx>
Reviewed-by: Peter Feiner <pfeiner@xxxxxxxxxx>
> diff --git a/net/ceph/pagevec.c b/net/ceph/pagevec.c
> index 5550130..096d914 100644
> --- a/net/ceph/pagevec.c
> +++ b/net/ceph/pagevec.c
> @@ -23,17 +23,15 @@ struct page **ceph_get_direct_page_vector(const void __user *data,
> if (!pages)
> return ERR_PTR(-ENOMEM);
>
> - down_read(¤t->mm->mmap_sem);
> while (got < num_pages) {
> - rc = get_user_pages(current, current->mm,
> + rc = get_user_pages_unlocked(current, current->mm,
> (unsigned long)data + ((unsigned long)got * PAGE_SIZE),
> - num_pages - got, write_page, 0, pages + got, NULL);
> + num_pages - got, write_page, 0, pages + got);
> if (rc < 0)
> break;
> BUG_ON(rc == 0);
> got += rc;
> }
> - up_read(¤t->mm->mmap_sem);
> if (rc < 0)
> goto fail;
> return pages;
I spent a while looking at this to make sure that BUG_ON(rc == 0) won't
trigger. AFAICT, __get_user_pages_locked can't return 0 since __get_user_pages
only returns 0 when nonblocking is not NULL; when __get_user_pages_locked
calls __get_user_pages with nonblocking != NULL (i.e., the first call in the
body of the for(;;) loop) and __get_user_pages returns 0, then
__get_user_pages_locked will call __get_user_pages again with nonblocking ==
NULL.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>