From: Eryu Guan <guaneryu@gmail.com>
To: Amir Goldstein <amir73il@gmail.com>
Cc: Miklos Szeredi <miklos@szeredi.hu>,
Eddie Horng <eddiehorng.tw@gmail.com>,
linux-unionfs@vger.kernel.org, fstests@vger.kernel.org
Subject: Re: [PATCH v2 1/2] fstests: Add more sanity to seek_sanity_test
Date: Wed, 6 Mar 2019 20:28:26 +0800 [thread overview]
Message-ID: <20190306122826.GH2824@desktop> (raw)
In-Reply-To: <20190303150940.12734-1-amir73il@gmail.com>
On Sun, Mar 03, 2019 at 05:09:39PM +0200, Amir Goldstein wrote:
> seek_sanity_test checks for one of several SEEK_DATA/HOLE
> behaviors and allows for the default behavior of filesystems,
> where SEEK_HOLE always returns EOF.
>
> This means that if filesystem has a regression in finding
> holes, the sanity test won't catch it. And indeed this regression
> happened in overlayfs on kernel v4.19 and went unnoticed.
>
> To improve test coverage, add a flag -f to seek_sanity_test to
> indicate that the default behavior is not acceptable.
> Whitelist all filesystem types that are expected to detect holes
> and use wrapper when invoking seek_sanity_test to add the -f flag
> to those filesystems.
>
> Overlayfs inherits expected behavior from base fs type.
>
> Signed-off-by: Amir Goldstein <amir73il@gmail.com>
> ---
>
> Eryu,
>
> After this change, the generic/seek group tests will start
> failing with overlayfs over base fs with true SEEK_HOLE support
> on upstream kernel.
>
> We had missing coverage of SEEK_HOLE, so we missed a regression
> in kernel v4.19 when overlayfs SEEK_HOLE stopped finding holes
> in sparse files.
>
> Thanks,
> Amir.
>
> Changes since v1:
> - Use whitelist of filesystems types that support true SEEK_HOLE
> instead of inferring SEEK_HOLE support from PUNCH_HOLE support
> - Split from patch of new PUNCH_HOLE+SEEK_HOLE test
>
> common/rc | 46 ++++++++++++++++++++++++++++++++++++++++++
> src/seek_sanity_test.c | 14 +++++++++++--
> tests/generic/285 | 2 +-
> tests/generic/436 | 2 +-
> tests/generic/445 | 2 +-
> tests/generic/448 | 2 +-
> tests/generic/490 | 2 +-
> 7 files changed, 63 insertions(+), 7 deletions(-)
>
> diff --git a/common/rc b/common/rc
> index c88c304c..d2482c20 100644
> --- a/common/rc
> +++ b/common/rc
> @@ -2298,6 +2298,52 @@ _ext4_disable_extent_zeroout()
> echo 0 >/sys/fs/ext4/$sdev/extent_max_zeroout_kb
> }
>
> +# Is this filesystem type expected to support seeking and finding holes?
> +# filesystems without SEEK_HOLE support, accept the SEEK_HOLE argument,
> +# but always return EOF offset for SEEK_HOLE.
> +# We use this whitelist FSTYP, to set expectation and avoid silent
> +# regression of filesystem seek hole behavior.
I don't think this description is that accurate, we're not checking if
the fs supports SEEK_HOLE or not, but checking if it supports
non-default behavior of SEEK_HOLE, where SEEK_HOLE always returns EOF.
> +#
> +# Return 0 for true
> +_fstyp_support_seek_data_hole()
So I think the function should be renamed too, but I couldn't come up
with a better name.. _fstyp_has_non_default_seek_data_hole ?
> +{
> + if [ -z $1 ]; then
> + local fstyp=$FSTYP
> + else
> + local fstyp=$1
> + fi
> +
> + case "$fstyp" in
> + btrfs|ext4|xfs|ceph|cifs|f2fs|gfs2|nfs*|ocfs2|tmpfs)
> + return 0
> + ;;
> + overlay)
> + if [ ! -z $OVL_BASE_FSTYP -a $OVL_BASE_FSTYP != "overlay" ]; then
> + _fstyp_support_seek_data_hole $OVL_BASE_FSTYP
> + return $?
> + else
> + # Do not assume that base fs supports SEEK_HOLE
> + return 1
> + fi
> + ;;
> + *)
> + # by default fstyp does not support SEEK_HOLE;
> + # if your fs does support SEEK_HOLE, add it to whitelist above!
> + return 1
> + ;;
> + esac
> +}
> +
> +# Run seek sanity test with predefined expectation for SEEK_DATA/HOLE behavior
> +_run_seek_sanity_test()
> +{
> + local testseekargs
> + if _fstyp_support_seek_data_hole; then
> + testseekargs+="-f"
> + fi
> + $here/src/seek_sanity_test $testseekargs $*
> +}
> +
> # Check if the file system supports seek_data/hole
> _require_seek_data_hole()
> {
> diff --git a/src/seek_sanity_test.c b/src/seek_sanity_test.c
> index e9938d1b..1eefc74d 100644
> --- a/src/seek_sanity_test.c
> +++ b/src/seek_sanity_test.c
> @@ -23,6 +23,7 @@
> #endif
>
> static blksize_t alloc_size;
> +int allow_default_behavior = 1;
> int default_behavior = 0;
> int unwritten_extents = 0;
> char *base_file_path;
> @@ -1115,6 +1116,12 @@ static int test_basic_support(void)
> }
>
> if (pos == filsz) {
> + if (!allow_default_behavior) {
> + fprintf(stderr, "File syste does not support llseek(2) extension "
^^^^^ system
> + "SEEK_HOLE. Aborting.\n");
> + ret = -1;
> + goto out;
> + }
I'd prefer putting this check after the checking 'default_behavior'
block.
> default_behavior = 1;
> fprintf(stderr, "File system supports the default behavior.\n");
> }
i.e. here:
if (default_behavior && !allow_default_behavior) {
fprintf(stderr, "Default behavior is not allowed. Aborting.\n");
ret = -1;
got out;
}
Thanks,
Eryu
> @@ -1148,7 +1155,7 @@ out:
>
> void usage(char *cmd)
> {
> - fprintf(stdout, "Usage: %s [-t] [-s <starttest>] [-e <endtest>] base_file_path\n", cmd);
> + fprintf(stdout, "Usage: %s [-tf] [-s <starttest>] [-e <endtest>] base_file_path\n", cmd);
> exit(1);
> }
>
> @@ -1169,11 +1176,14 @@ int main(int argc, char **argv)
> teststart = 1;
> testend = 12;
>
> - while ((opt = getopt(argc, argv, "ts:e:")) != -1) {
> + while ((opt = getopt(argc, argv, "tfs:e:")) != -1) {
> switch (opt) {
> case 't':
> check_support++;
> break;
> + case 'f':
> + allow_default_behavior = 0;
> + break;
> case 's':
> teststart = strtol(optarg, NULL, 10);
> if (teststart <= 0 || teststart > numtests) {
> diff --git a/tests/generic/285 b/tests/generic/285
> index b53b3acc..20e89e5b 100755
> --- a/tests/generic/285
> +++ b/tests/generic/285
> @@ -37,7 +37,7 @@ _cleanup()
> eval "rm -f $BASE_TEST_FILE.*"
> }
>
> -$here/src/seek_sanity_test $BASE_TEST_FILE > $seqres.full 2>&1 ||
> +_run_seek_sanity_test $BASE_TEST_FILE > $seqres.full 2>&1 ||
> _fail "seek sanity check failed!"
>
> # success, all done
> diff --git a/tests/generic/436 b/tests/generic/436
> index a3ef70e5..ada60b2e 100755
> --- a/tests/generic/436
> +++ b/tests/generic/436
> @@ -36,7 +36,7 @@ _cleanup()
> rm -f $tmp.* $BASE_TEST_FILE.*
> }
>
> -$here/src/seek_sanity_test -s 13 -e 16 $BASE_TEST_FILE > $seqres.full 2>&1 ||
> +_run_seek_sanity_test -s 13 -e 16 $BASE_TEST_FILE > $seqres.full 2>&1 ||
> _fail "seek sanity check failed!"
>
> # success, all done
> diff --git a/tests/generic/445 b/tests/generic/445
> index 694913ae..248f321b 100755
> --- a/tests/generic/445
> +++ b/tests/generic/445
> @@ -36,7 +36,7 @@ _cleanup()
> rm -f $tmp.* $BASE_TEST_FILE.*
> }
>
> -$here/src/seek_sanity_test -s 17 -e 17 $BASE_TEST_FILE > $seqres.full 2>&1 ||
> +_run_seek_sanity_test -s 17 -e 17 $BASE_TEST_FILE > $seqres.full 2>&1 ||
> _fail "seek sanity check failed!"
>
> # success, all done
> diff --git a/tests/generic/448 b/tests/generic/448
> index d3ec5528..dada499b 100755
> --- a/tests/generic/448
> +++ b/tests/generic/448
> @@ -34,7 +34,7 @@ BASE_TEST_FILE=$TEST_DIR/seek_sanity_testfile_$seq
>
> _require_test_program "seek_sanity_test"
>
> -$here/src/seek_sanity_test -s 18 -e 18 $BASE_TEST_FILE > $seqres.full 2>&1 ||
> +_run_seek_sanity_test -s 18 -e 18 $BASE_TEST_FILE > $seqres.full 2>&1 ||
> _fail "seek sanity check failed!"
>
> _check_dmesg
> diff --git a/tests/generic/490 b/tests/generic/490
> index 330c2b60..07b08939 100755
> --- a/tests/generic/490
> +++ b/tests/generic/490
> @@ -39,7 +39,7 @@ _cleanup()
> rm -f $base_test_file*
> }
>
> -$here/src/seek_sanity_test -s 19 -e 20 $base_test_file > $seqres.full 2>&1 ||
> +_run_seek_sanity_test -s 19 -e 20 $base_test_file > $seqres.full 2>&1 ||
> _fail "seek sanity check failed!"
>
> # success, all done
> --
> 2.17.1
>
prev parent reply other threads:[~2019-03-06 12:28 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-03-03 15:09 [PATCH v2 1/2] fstests: Add more sanity to seek_sanity_test Amir Goldstein
2019-03-03 15:09 ` [PATCH v2 2/2] generic: Test that SEEK_HOLE can find a punched hole Amir Goldstein
2019-03-06 12:30 ` Eryu Guan
2019-03-06 12:28 ` Eryu Guan [this message]
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=20190306122826.GH2824@desktop \
--to=guaneryu@gmail.com \
--cc=amir73il@gmail.com \
--cc=eddiehorng.tw@gmail.com \
--cc=fstests@vger.kernel.org \
--cc=linux-unionfs@vger.kernel.org \
--cc=miklos@szeredi.hu \
/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;
as well as URLs for NNTP newsgroup(s).