On Thu, Oct 31, 2019 at 10:36:38AM +0000, Dmitry Monakhov wrote:
> From: Dmitry Monakhov <dmtrmonakhov@xxxxxxxxxxxxxx>
>
> Quotasync may livelock if others tasks generate enough dirty dquots in parallel
> This test case pefrorm fchown to produce dirty quotas
>
> This test known to detect livelock non-journaled quota for kernels prior to v5.4
>
> Signed-off-by: Dmitry Monakhov <dmtrmonakhov@xxxxxxxxxxxxxx>
> ---
> src/Makefile | 2 +-
> src/chowner.c | 107 ++++++++++++++++++++++++++++++++++++++++++++++++++
Missing an entry in .gitingore
> tests/generic/585 | 94 ++++++++++++++++++++++++++++++++++++++++++++
> tests/generic/585.out | 2 +
> tests/generic/group | 1 +
> 5 files changed, 205 insertions(+), 1 deletion(-)
> create mode 100644 src/chowner.c
> create mode 100755 tests/generic/585
> create mode 100644 tests/generic/585.out
>
> diff --git a/src/Makefile b/src/Makefile
> index ce6d861..b3ab7b4 100644
> --- a/src/Makefile
> +++ b/src/Makefile
> @@ -28,7 +28,7 @@ LINUX_TARGETS = xfsctl bstat t_mtab getdevicesize preallo_rw_pattern_reader \
> attr-list-by-handle-cursor-test listxattr dio-interleaved t_dir_type \
> dio-invalidate-cache stat_test t_encrypted_d_revalidate \
> attr_replace_test swapon mkswap t_attr_corruption t_open_tmpfiles \
> - fscrypt-crypt-util bulkstat_null_ocount
> + fscrypt-crypt-util bulkstat_null_ocount chowner
>
> SUBDIRS = log-writes perf
>
> diff --git a/src/chowner.c b/src/chowner.c
> new file mode 100644
> index 0000000..043cd06
> --- /dev/null
> +++ b/src/chowner.c
> @@ -0,0 +1,107 @@
> +// SPDX-License-Identifier: GPL-2.0
> +//
> +// Copyright (c) 2019 YANDEX LLC. All Rights Reserved.
> +// Author: Dmitry Monakhov <dmtrmonakhov@xxxxxxxxxxxxxx>
> +
> +#include <unistd.h>
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <time.h>
> +#include <sys/time.h>
> +#include <fcntl.h>
> +#include <limits.h>
> +
> +char *progname;
> +
> +void usage(void)
> +{
> + printf("Usage %s: [-b uid] [-e uid] [-g] [-P] [-t timeout] FNAME\n", progname);
> + printf("\t\t -b: begin of uid range\n");
> + printf("\t\t -e: length of uid range\n");
^^^^^^ should be end of uid range?
> + printf("\t\t -g: change group uid\n");
> + printf("\t\t -t: loop timeout\n");
> + printf("\t\t -P: Do not run chowner loop, only prepare files\n");
> + exit(1);
> +}
> +int main(int argc, char *const *argv)
> +{
> + char *filename;
> + int fd;
> + int c;
> + uid_t begin = 2000;
> + uid_t end = 12000;
> + unsigned int timeout = 10;
> + struct timeval start, now, delta = { 0, 0 };
> + uid_t uid;
> + gid_t gid = getegid();
> + int do_group = 0;
> + int do_prepare = 0;
> +
> + progname = argv[0];
> + while ((c = getopt(argc, argv, "b:e:gPt:")) != -1) {
> + switch (c) {
> + case 'b':
> + begin = atoi(optarg);
> + break;
> + case 'e':
> + end = atoi(optarg);
> + break;
> + case 'g':
> + do_group = 1;
> + break;
> + case 'P':
> + do_prepare = 1;
> + break;
> + case 't':
> + timeout = atoi(optarg);
> + break;
> + default:
> + usage();
> + }
> + }
> + if (optind == argc-1)
> + filename = argv[optind];
> + else
> + usage();
> + if (do_prepare) {
> + char path[PATH_MAX];
> + for (uid = begin; uid < end;uid++) {
> + sprintf(path, "%s.%d", filename, uid);
So in do_prepare case, expect filename to be a filename prefix? Or a
directory? From the test code, it seems to be a dir, but this c code
treats as an filename prefix.
# Preparation step: Create all files with uid in range
# to cache quota in kernel memory
mkdir -p $SCRATCH_MNT/q
$here/src/chowner $SCRATCH_MNT/q -b $begin -e $end -P
> + fd = open(path, O_RDWR|O_CREAT, 0666);
> + if (fd < 0) {
> + perror("open");
> + exit(1);
> + }
> + if (do_group)
> + gid = uid;
> + if (fchown(fd, uid, gid)) {
> + perror("chown");
> + exit(1);
> + }
> + close(fd);
> + }
> + return 0;
> + }
> + fd = open(filename, O_RDWR|O_CREAT, 0666);
> + if (fd < 0) {
> + perror("open");
> + exit(1);
> + }
> + gettimeofday(&start, NULL);
> +
> + while (1) {
> + for (uid = begin; uid < end;uid++) {
> + if (do_group)
> + gid = uid;
> + if (fchown(fd, uid, gid)) {
> + perror("chown");
> + exit(1);
> + }
> + }
> + gettimeofday(&now, NULL);
> + timersub(&now, &start, &delta);
> + if (delta.tv_sec >= timeout)
> + break;
> + }
> + return 0;
> +}
> diff --git a/tests/generic/585 b/tests/generic/585
> new file mode 100755
> index 0000000..d37522a
> --- /dev/null
> +++ b/tests/generic/585
> @@ -0,0 +1,94 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2019 YANDEX LLC. All Rights Reserved.
> +#
> +# FS QA Test 585
> +#
> +# Check livelock during quota-sync while other tasks dirty quotas in parallel.
> +# Run fchown(2) in a loop is the fastest way to produce dirty quotas
> +#
> +seq=`basename $0`
> +seqres=$RESULT_DIR/$seq
> +echo "QA output created by $seq"
> +
> +here=`pwd`
> +tmp=/tmp/$$
> +status=1 # failure is the default!
> +trap "_cleanup; exit \$status" 0 1 2 3 15
> +
> +_cleanup()
> +{
> + cd /
> + rm -f $tmp.*
> +}
> +# get standard environment, filters and checks
> +. ./common/rc
> +. ./common/filter
> +. ./common/quota
> +
> +
> +begin=2000
> +end=10000
> +nr_proc=$((16 * LOAD_FACTOR))
> +deadline=$((100 * TIME_FACTOR))
So this test has two "_workout"s, and one takes around 120s, so whole
test takes around 4 minutes, which seems a bit long for a generic test.
What's the reproducibility if you lower the runtime? I'd be good to find
a good banalce between reproducibility and test run time. e.g. 20s for
60% reproducibility is better than 240s for 90%.
> +
> +_workout()
Local functions doesn't need "_" prefix.
> +{
> + for ((i=0; i < nr_proc; i++))
> + do
Use tab for indention. And the indentions are not consistent, some lines
use tab and some lines use spaces.
> + # Spread files to isolated dirs to minimize locking contention
> + mkdir -p $SCRATCH_MNT/chowner/$i
> + $here/src/chowner $SCRATCH_MNT/chowner/$i/test -b $begin -e $end \
> + -t $((deadline + 10)) &
> + pids="$pids $!"
> + done
> + # Let chowners warm up ...
> + sleep 5
> + start=$(date +%s)
> + for ((i=0;i<3;i++))
> + do
> + s=$(date +%s)
> + # In normal situation command should finish in ~1sec,
> + # but in case of livelock it will spin until chowners exits
> + $*
> + e=$(date +%s)
> + echo "loop $i: $* runtime: $((e-s))" >> $seqres.full
> + sleep 2
> + done
> + end=$(date +%s)
> + runtime=$((end-start))
> + echo "DONE: $* total runtime: $runtime" >> $seqres.full
> + kill -TERM $pids 2> /dev/null
> + wait $pids
> +
> + [ $runtime -le $deadline ] || \
> + _fail "Live lock detected, $* runtime: $runtime, deadline: $deadline"
> +}
> +
> +# real QA test starts here
> +_supported_fs generic
> +_supported_os Linux
> +_require_quota
> +#_require_user
Could be removed.
> +_require_scratch
> +_require_command "$KILLALL_PROG" killall
killall is not used, can be removed.
And we need
_require_test_program "chowner"
> +
> +rm -f $seqres.full
> +_scratch_mkfs_sized $((512 * 1024 * 1024)) >> $seqres.full 2>&1
Any reason to use sized mkfs?
> +_scratch_mount "-o quota,user"
> +chmod 777 $SCRATCH_MNT
> +quotacheck -u $SCRATCH_MNT 2>/dev/null
> +quotaon -v -u $SCRATCH_MNT >> $seqres.full 2>&1
Above steps could be replaced by:
_qmount_option "usrquota"
_qmount
Thanks,
Eryu
> +
> +# Preparation step: Create all files with uid in range
> +# to cache quota in kernel memory
> +mkdir -p $SCRATCH_MNT/q
> +$here/src/chowner $SCRATCH_MNT/q -b $begin -e $end -P
> +
> +_workout quotasync $SCRATCH_MNT
> +_workout sync
> +
> +echo "Silence is golden"
> +# success, all done
> +status=0
> +exit
> diff --git a/tests/generic/585.out b/tests/generic/585.out
> new file mode 100644
> index 0000000..e4dd43b
> --- /dev/null
> +++ b/tests/generic/585.out
> @@ -0,0 +1,2 @@
> +QA output created by 585
> +Silence is golden
> diff --git a/tests/generic/group b/tests/generic/group
> index 42ca2b9..703a1b4 100644
> --- a/tests/generic/group
> +++ b/tests/generic/group
> @@ -587,3 +587,4 @@
> 582 auto quick encrypt
> 583 auto quick encrypt
> 584 auto quick encrypt
> +585 auto quota rw stress
> --
> 2.7.4
>
