[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <744be7c4-8e00-7876-5819-a1d07d3d423f@suse.com>
Date: Thu, 25 Aug 2022 13:58:59 +0200
From: Jan Beulich <jbeulich@...e.com>
To: Juergen Gross <jgross@...e.com>
Cc: Stefano Stabellini <sstabellini@...nel.org>,
Oleksandr Tyshchenko <oleksandr_tyshchenko@...m.com>,
stable@...r.kernel.org,
Rustam Subkhankulov <subkhankulov@...ras.ru>,
xen-devel@...ts.xenproject.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3] xen/privcmd: fix error exit of privcmd_ioctl_dm_op()
On 25.08.2022 13:40, Juergen Gross wrote:
> --- a/drivers/xen/privcmd.c
> +++ b/drivers/xen/privcmd.c
> @@ -581,7 +581,7 @@ static int lock_pages(
> struct privcmd_dm_op_buf kbufs[], unsigned int num,
> struct page *pages[], unsigned int nr_pages, unsigned int *pinned)
> {
> - unsigned int i;
> + unsigned int i, off = 0;
>
> for (i = 0; i < num; i++) {
> unsigned int requested;
> @@ -589,19 +589,23 @@ static int lock_pages(
>
> requested = DIV_ROUND_UP(
> offset_in_page(kbufs[i].uptr) + kbufs[i].size,
> - PAGE_SIZE);
> + PAGE_SIZE) - off;
> if (requested > nr_pages)
> return -ENOSPC;
>
> page_count = pin_user_pages_fast(
> - (unsigned long) kbufs[i].uptr,
> + (unsigned long)kbufs[i].uptr + off * PAGE_SIZE,
> requested, FOLL_WRITE, pages);
> - if (page_count < 0)
> - return page_count;
> + if (page_count <= 0)
> + return page_count ? : -EFAULT;
>
> *pinned += page_count;
> nr_pages -= page_count;
> pages += page_count;
> +
> + off = requested - page_count;
> + if (off)
> + i--;
> }
Initially I thought this would go wrong only on the 3rd iteration, but
meanwhile I think it's wrong already on the 2nd. What I think you need
is
if (page_count < requested)
i--;
off += page_count;
or with the i++ from the loop header absorbed here
if (page_count == requested)
i++;
off += page_count;
Plus of course off needs resetting to zero whenever i advances. I.e.
if (page_count == requested) {
i++;
off = 0;
} else {
off += page_count;
}
Jan
Powered by blists - more mailing lists