On Mon, Sep 02, 2019 at 09:53:58PM +0800, Zorro Lang wrote:
> The _xfs_check sometimes need to take too much memory, some quota
> related cases (e.g: xfs/442, generic/232 etc) always trigger OOM
> killer.
>
> [93334.020194] xfs_db invoked oom-killer: gfp_mask=0x6200ca(GFP_HIGHUSER_MOVABLE), nodemask=(null), order=0, oom_score_adj=-1000
> [93334.020206] xfs_db cpuset=/ mems_allowed=0
> [93334.020213] CPU: 2 PID: 977 Comm: xfs_db Kdump: loaded Not tainted 4.18.0-139.el8.ppc64le+debug #1
> [93334.020217] Call Trace:
> [93334.020223] [c0000001cb163640] [c000000001035754] dump_stack+0xe8/0x164 (unreliable)
> [93334.020229] [c0000001cb163690] [c0000000004a77dc] dump_header+0x7c/0x670
> [93334.020234] [c0000001cb1637b0] [c0000000004a872c] oom_kill_process+0x26c/0x3d0
> [93334.020239] [c0000001cb163800] [c0000000004aa438] out_of_memory+0x278/0x930
> [93334.020244] [c0000001cb1638a0] [c0000000004bd6c8] __alloc_pages_nodemask+0x1ab8/0x1b70
> [93334.020248] [c0000001cb163ae0] [c00000000058b8dc] alloc_pages_vma+0xec/0x660
> [93334.020253] [c0000001cb163b50] [c00000000052b544] do_anonymous_page+0x124/0xac0
> [93334.020258] [c0000001cb163bc0] [c000000000534918] __handle_mm_fault+0xda8/0x1b60
> [93334.020263] [c0000001cb163cb0] [c000000000535948] handle_mm_fault+0x278/0x5a0
> [93334.020268] [c0000001cb163d00] [c00000000008b30c] __do_page_fault+0x27c/0xe00
> [93334.020273] [c0000001cb163df0] [c00000000008bec8] do_page_fault+0x38/0xf0
> [93334.020278] [c0000001cb163e30] [c00000000000a904] handle_page_fault+0x18/0x38
> [93334.020281] Mem-Info:
> [93334.020286] active_anon:93769 inactive_anon:13540 isolated_anon:0
> [93334.020286] active_file:23 inactive_file:8 isolated_file:0
> [93334.020286] unevictable:0 dirty:0 writeback:0 unstable:0
> [93334.020286] slab_reclaimable:1543 slab_unreclaimable:8926
> [93334.020286] mapped:69 shmem:15 pagetables:76 bounce:0
> [93334.020286] free:2756 free_pcp:0 free_cma:0
>
> The xfs_check related code is old, and nearly won't be maintained
> too much, test xfs_repair and xfs_scrub is much more important than
> xfs_check. But the xfstests always do a POST-xfs_check at the end of
> each cases. So I'd like to add a new helper named
> _require_scratch_no_xfs_check to skip xfs_check from some sub-cases.
>
> Signed-off-by: Zorro Lang <zlang@xxxxxxxxxx>
> ---
>
> Hi,
>
> Maybe this patch not the best way to deal with this issue. Please feel free to
> tell me if you have better idea.
I would just get rid of it (and have tried several times) but have not
had the time to check that xfs_repair can detect every corruption that
xfs_check can.
(To get started, I think it involves forking the dangerous_repair group
tests to run _xfs_check and xfs_repair -n one after the other to compare
outputs.)
--D
> Thanks,
> Zorro
>
> common/rc | 1 +
> common/xfs | 26 ++++++++++++++++++++++++++
> tests/generic/232 | 2 ++
> tests/generic/233 | 2 ++
> tests/generic/234 | 2 ++
> tests/xfs/442 | 2 ++
> 6 files changed, 35 insertions(+)
>
> diff --git a/common/rc b/common/rc
> index e0b087c1..86eeda16 100644
> --- a/common/rc
> +++ b/common/rc
> @@ -1570,6 +1570,7 @@ _require_scratch()
> {
> _require_scratch_nocheck
> touch ${RESULT_DIR}/require_scratch
> + touch ${RESULT_DIR}/require_scratch.xfs_check
> }
>
> # require a scratch dev of a minimum size (in kb)
> diff --git a/common/xfs b/common/xfs
> index 1bce3c18..80e6d1e8 100644
> --- a/common/xfs
> +++ b/common/xfs
> @@ -168,11 +168,37 @@ _scratch_mkfs_xfs()
> return $mkfs_status
> }
>
> +# Usage: _require_scratch_no_xfs_check [limit_memsize]
> +#
> +# Due to xfs_check costs too much memory, sometimes it cause OOM when test on
> +# low memory machine. This function is used to skip xfs_check conditionally
> +# or unconditionally.
> +#
> +# Note: must be called after _require_scratch.
> +_require_scratch_no_xfs_check()
> +{
> + local bound=$1
> + local freemem
> +
> + freemem=`_free_memory_bytes`
> + if [ -n "$bound" ];then
> + if [ $freemem -lt $bound ];then
> + rm -f ${RESULT_DIR}/require_scratch.xfs_check
> + fi
> + else
> + rm -f ${RESULT_DIR}/require_scratch.xfs_check
> + fi
> +}
> +
> # xfs_check script is planned to be deprecated. But, we want to
> # be able to invoke "xfs_check" behavior in xfstests in order to
> # maintain the current verification levels.
> _xfs_check()
> {
> + if [ ! -f "${RESULT_DIR}/require_scratch.xfs_check" ];then
> + return 0
> + fi
> +
> OPTS=" "
> DBOPTS=" "
> USAGE="Usage: xfs_check [-fsvV] [-l logdev] [-i ino]... [-b bno]... special"
> diff --git a/tests/generic/232 b/tests/generic/232
> index d5c20249..20841313 100755
> --- a/tests/generic/232
> +++ b/tests/generic/232
> @@ -52,6 +52,8 @@ _fsstress()
> _supported_fs generic
> _supported_os Linux
> _require_scratch
> +# Do post xfs_check, if free memory size > 16G
> +_require_scratch_no_xfs_check $((16 * 1024 * 1024 * 1024))
> _require_quota
>
> _scratch_mkfs > $seqres.full 2>&1
> diff --git a/tests/generic/233 b/tests/generic/233
> index c49bf252..7933d4b8 100755
> --- a/tests/generic/233
> +++ b/tests/generic/233
> @@ -56,6 +56,8 @@ _fsstress()
> _supported_fs generic
> _supported_os Linux
> _require_scratch
> +# Do post xfs_check, if free memory size > 16G
> +_require_scratch_no_xfs_check $((16 * 1024 * 1024 * 1024))
> _require_quota
> _require_user
>
> diff --git a/tests/generic/234 b/tests/generic/234
> index dc296df8..e843dc10 100755
> --- a/tests/generic/234
> +++ b/tests/generic/234
> @@ -72,6 +72,8 @@ test_setting()
> _supported_fs generic
> _supported_os Linux
> _require_scratch
> +# Do post xfs_check, if free memory size > 16G
> +_require_scratch_no_xfs_check $((16 * 1024 * 1024 * 1024))
> _require_quota
>
> # real QA test starts here
> diff --git a/tests/xfs/442 b/tests/xfs/442
> index 7a5f2e8e..9e891b51 100755
> --- a/tests/xfs/442
> +++ b/tests/xfs/442
> @@ -36,6 +36,8 @@ _supported_fs xfs
> _supported_os Linux
>
> _require_scratch_reflink
> +# Do post xfs_check, if free memory size > 16G
> +_require_scratch_no_xfs_check $((16 * 1024 * 1024 * 1024))
> _require_quota
> _require_command "$KILLALL_PROG" "killall"
>
> --
> 2.17.2
>
