From: Dave Chinner <david@fromorbit.com>
To: "Darrick J. Wong" <darrick.wong@oracle.com>
Cc: linux-xfs@vger.kernel.org
Subject: Re: [PATCH 04/10] xfs: check directory bestfree information in the verifier
Date: Mon, 4 Jun 2018 10:10:19 +1000 [thread overview]
Message-ID: <20180604001019.GC10363@dastard> (raw)
In-Reply-To: <152806816970.18187.13658571880127164744.stgit@magnolia>
On Sun, Jun 03, 2018 at 04:22:49PM -0700, Darrick J. Wong wrote:
> From: Darrick J. Wong <darrick.wong@oracle.com>
>
> Create a variant of xfs_dir2_data_freefind that is suitable for use in a
> verifier. Because _freefind is called by the verifier, we simply
> duplicate the _freefind function, convert the ASSERTs to return
> __this_address, and modify the verifier to call our new function. Once
> we've made it impossible for directory blocks with bad bestfree data to
> make it into the filesystem we can remove the DEBUG code from the
> regular _freefind function.
>
> Underlying argument: corruption of on-disk metadata should return
> -EFSCORRUPTED instead of blowing ASSERTs.
>
> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> ---
> fs/xfs/libxfs/xfs_dir2_data.c | 121 +++++++++++++++++++++++++++++------------
> 1 file changed, 86 insertions(+), 35 deletions(-)
>
>
> diff --git a/fs/xfs/libxfs/xfs_dir2_data.c b/fs/xfs/libxfs/xfs_dir2_data.c
> index cb67ec730b9b..bc5c0ba46ec6 100644
> --- a/fs/xfs/libxfs/xfs_dir2_data.c
> +++ b/fs/xfs/libxfs/xfs_dir2_data.c
> @@ -33,6 +33,11 @@
> #include "xfs_cksum.h"
> #include "xfs_log.h"
>
> +static xfs_failaddr_t xfs_dir2_data_freefind_verify(
> + struct xfs_dir2_data_hdr *hdr, struct xfs_dir2_data_free *bf,
> + struct xfs_dir2_data_unused *dup,
> + struct xfs_dir2_data_free **bf_ent);
> +
> /*
> * Check the consistency of the data block.
> * The input can also be a block-format directory.
> @@ -52,6 +57,7 @@ __xfs_dir3_data_check(
> xfs_dir2_data_free_t *dfp; /* bestfree entry */
> xfs_dir2_data_unused_t *dup; /* unused entry */
> char *endp; /* end of useful data */
> + xfs_failaddr_t fa;
> int freeseen; /* mask of bestfrees seen */
> xfs_dahash_t hash; /* hash of current name */
> int i; /* leaf index */
This could be placed inside the loop scope, right?
> - ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC) ||
> - hdr->magic == cpu_to_be32(XFS_DIR3_DATA_MAGIC) ||
> - hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC) ||
> - hdr->magic == cpu_to_be32(XFS_DIR3_BLOCK_MAGIC));
> - for (dfp = &bf[0], seenzero = matched = 0;
> - dfp < &bf[XFS_DIR2_DATA_FD_COUNT];
> - dfp++) {
> + for (dfp = &bf[0]; dfp < &bf[XFS_DIR2_DATA_FD_COUNT]; dfp++) {
> if (!dfp->offset) {
> - ASSERT(!dfp->length);
> - seenzero = 1;
> + if (dfp->length)
> + return __this_address;
> + seenzero = true;
> continue;
> }
> - ASSERT(seenzero == 0);
> + if (seenzero)
> + return __this_address;
> if (be16_to_cpu(dfp->offset) == off) {
> - matched = 1;
> - ASSERT(dfp->length == dup->length);
> - } else if (off < be16_to_cpu(dfp->offset))
> - ASSERT(off + be16_to_cpu(dup->length) <= be16_to_cpu(dfp->offset));
> - else
> - ASSERT(be16_to_cpu(dfp->offset) + be16_to_cpu(dfp->length) <= off);
> - ASSERT(matched || be16_to_cpu(dfp->length) >= be16_to_cpu(dup->length));
> - if (dfp > &bf[0])
> - ASSERT(be16_to_cpu(dfp[-1].length) >= be16_to_cpu(dfp[0].length));
> + matched = true;
> + if (dfp->length != dup->length)
> + return __this_address;
> + } else if (be16_to_cpu(dfp->offset) > off) {
> + if (off + be16_to_cpu(dup->length) >
> + be16_to_cpu(dfp->offset))
can you indent the second line further to indicate it is a
continuation of the logic statement on the previous line rather than
a new logic condition? i.e.
if (off + be16_to_cpu(dup->length) >
be16_to_cpu(dfp->offset))
> + return __this_address;
> + } else {
> + if (be16_to_cpu(dfp->offset) +
> + be16_to_cpu(dfp->length) > off)
> + return __this_address;
Same here?
> + }
> + if (!matched &&
> + be16_to_cpu(dfp->length) < be16_to_cpu(dup->length))
> + return __this_address;
> + if (dfp > &bf[0] &&
> + be16_to_cpu(dfp[-1].length) < be16_to_cpu(dfp[0].length))
> + return __this_address;
> }
> -#endif
> +
> + /*
> + * If this is smaller than the smallest bestfree entry,
> + * it can't be there since they're sorted.
> + */
> + if (be16_to_cpu(dup->length) <
> + be16_to_cpu(bf[XFS_DIR2_DATA_FD_COUNT - 1].length))
> + return NULL;
> + /*
> + * Look at the three bestfree entries for our guy.
> + */
> + for (dfp = &bf[0]; dfp < &bf[XFS_DIR2_DATA_FD_COUNT]; dfp++) {
> + if (!dfp->offset)
> + return NULL;
> + if (be16_to_cpu(dfp->offset) == off) {
> + *bf_ent = dfp;
> + return NULL;
> + }
> + }
> + /*
> + * Didn't find it. This only happens if there are duplicate lengths.
> + */
> + return NULL;
And this tail is basically a duplicate of what now remains in
xfs_dir2_data_freefind(). Can you call that function rather than
duplicating the search code?
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
next prev parent reply other threads:[~2018-06-04 0:10 UTC|newest]
Thread overview: 26+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-06-03 23:22 [PATCH 00/10] xfs: fix various checking problems Darrick J. Wong
2018-06-03 23:22 ` [PATCH 01/10] xfs: don't forbid setting dax flag on directories if device doesn't dax Darrick J. Wong
2018-06-03 23:41 ` Dave Chinner
2018-06-04 4:25 ` Darrick J. Wong
2018-06-03 23:22 ` [PATCH 02/10] xfs: strengthen btree pointer checks before use Darrick J. Wong
2018-06-03 23:45 ` Dave Chinner
2018-06-03 23:22 ` [PATCH 03/10] xfs: don't assert when on-disk btree pointers are garbage Darrick J. Wong
2018-06-03 23:49 ` Dave Chinner
2018-06-03 23:55 ` Dave Chinner
2018-06-03 23:22 ` [PATCH 04/10] xfs: check directory bestfree information in the verifier Darrick J. Wong
2018-06-04 0:10 ` Dave Chinner [this message]
2018-06-04 4:23 ` Darrick J. Wong
2018-06-03 23:22 ` [PATCH 05/10] xfs: don't assert when reporting on-disk corruption while loading btree Darrick J. Wong
2018-06-03 23:51 ` Dave Chinner
2018-06-03 23:23 ` [PATCH 06/10] xfs: remove redundant ASSERT on insufficient bestfree length in _leaf_addname Darrick J. Wong
2018-06-04 0:11 ` Dave Chinner
2018-06-03 23:23 ` [PATCH 07/10] xfs: xfs_alloc_get_rec should return EFSCORRUPTED for obvious bnobt corruption Darrick J. Wong
2018-06-04 0:14 ` Dave Chinner
2018-06-04 4:32 ` Darrick J. Wong
2018-06-04 23:22 ` Dave Chinner
2018-06-03 23:23 ` [PATCH 08/10] xfs: btree lookup shouldn't ASSERT on empty btree nodes Darrick J. Wong
2018-06-04 0:14 ` Dave Chinner
2018-06-03 23:23 ` [PATCH 09/10] xfs: don't ASSERT on short form btree root pointer of zero Darrick J. Wong
2018-06-04 0:15 ` Dave Chinner
2018-06-03 23:23 ` [PATCH 10/10] xfs: don't return garbage buffers in xfs_da3_node_read Darrick J. Wong
2018-06-04 0:18 ` Dave Chinner
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20180604001019.GC10363@dastard \
--to=david@fromorbit.com \
--cc=darrick.wong@oracle.com \
--cc=linux-xfs@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox