Skip to contentSkip to conversation

btrfs: test active zone tracking

PATCH Superseded Naohiro Aota 3 participants 10 messages Sep 29, 2022 – Sep 30, 2022 Btrfs file system

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover btrfs: test active zone tracking

This series adds a test for checking if an active zone tracking feature of btrfs's zoned mode. The first patch introduces _zone_capacity() helper to get the zone capacity of a specified zone. It rewrites btrfs/237 with the helper and use the helper in a newly added test btrfs/292.

Changes:
- v2:
- Rename to common/zoned.
- Move _filter_blkzone_report() as well to common/zoned.
- Drop _require_fio as it was already unnecessary.

Naohiro Aota (2):
common: introduce zone_capacity() to return a zone capacity
btrfs: test active zone tracking

 common/filter       |  13 -----
 common/zoned        |  39 +++++++++++++
 tests/btrfs/237     |   8 +--
 tests/btrfs/292     | 136 ++++++++++++++++++++++++++++++++++++++++++++
 tests/btrfs/292.out |   2 +
 5 files changed, 179 insertions(+), 19 deletions(-)
 create mode 100644 common/zoned
 create mode 100755 tests/btrfs/292
 create mode 100644 tests/btrfs/292.out
Signature · 1 line
2.37.3
#1 lore raw

PATCH v2 1/2 common: introduce zone_capacity() to return a zone capacity

Introduce _zone_capacity() to return a zone capacity of the given address in the given device (optional). Move _filter_blkzone_report() for it, and rewrite btrfs/237 with it.

---
 common/filter   | 13 -------------
 common/zoned    | 28 ++++++++++++++++++++++++++++
 tests/btrfs/237 |  8 ++------
 3 files changed, 30 insertions(+), 19 deletions(-)
 create mode 100644 common/zoned
common/filter +0 -13
index 28dea64662dc..ac5c93422567 100644
@@ -651,18 +651,5 @@ _filter_bash()
 	sed -e "s/^bash: line 1: /bash: /"
 }
 
-#
-# blkzone report added zone capacity to be printed from v2.37.
-# This filter will add an extra column 'cap' with the same value of
-# 'len'(zone size) for blkzone version < 2.37
-#
-# Before: start: 0x000100000, len 0x040000, wptr 0x000000 ..
-# After: start: 0x000100000, len 0x040000, cap 0x040000, wptr 0x000000 ..
-_filter_blkzone_report()
-{
-	$AWK_PROG -F "," 'BEGIN{OFS=",";} $3 !~ /cap/ {$2=$2","$2;} {print;}' |\
-	sed -e 's/len/cap/2'
-}
-
 # make sure this script returns success
 /bin/true
common/zoned new file +28 -0
new file mode 100644
index 000000000000..d1bc60f784a1
@@ -0,0 +1,28 @@
+#
+# Common zoned block device specific functions
+#
+
+#
+# blkzone report added zone capacity to be printed from v2.37.
+# This filter will add an extra column 'cap' with the same value of
+# 'len'(zone size) for blkzone version < 2.37
+#
+# Before: start: 0x000100000, len 0x040000, wptr 0x000000 ..
+# After: start: 0x000100000, len 0x040000, cap 0x040000, wptr 0x000000 ..
+_filter_blkzone_report()
+{
+	$AWK_PROG -F "," 'BEGIN{OFS=",";} $3 !~ /cap/ {$2=$2","$2;} {print;}' |\
+	sed -e 's/len/cap/2'
+}
+
+_zone_capacity() {
+    local phy=$1
+    local dev=$2
+
+    [ -z "$dev" ] && dev=$SCRATCH_DEV
+
+    size=$($BLKZONE_PROG report -o $phy -l 1 $dev |\
+	       _filter_blkzone_report |\
+	       grep -Po "cap 0x[[:xdigit:]]+" | cut -d ' ' -f 2)
+    echo $((size << 9))
+}
tests/btrfs/237 +2 -6
index bc6522e2200a..101094b5ce70 100755
@@ -13,7 +13,7 @@
 _begin_fstest auto quick zone balance
 
 # Import common functions.
-. ./common/filter
+. ./common/zbd
 
 # real QA test starts here
 
@@ -56,11 +56,7 @@ fi
 
 start_data_bg_phy=$(get_data_bg_physical)
 start_data_bg_phy=$((start_data_bg_phy >> 9))
-
-size=$($BLKZONE_PROG report -o $start_data_bg_phy -l 1 $SCRATCH_DEV |\
-	_filter_blkzone_report |\
-	grep -Po "cap 0x[[:xdigit:]]+" | cut -d ' ' -f 2)
-size=$((size << 9))
+size=$(_zone_capacity $start_data_bg_phy)
 
 reclaim_threshold=75
 echo $reclaim_threshold > /sys/fs/btrfs/"$uuid"/bg_reclaim_threshold
Signature · 1 line
2.37.3
Reviewed-by on v2 1/2
#2 lore raw

Re: [PATCH v2 1/2] common: introduce zone_capacity() to return a zone capacity

Quoted 74 lines · Naohiro Aota

Introduce _zone_capacity() to return a zone capacity of the given address in the given device (optional). Move _filter_blkzone_report() for it, and rewrite btrfs/237 with it.

---
 common/filter   | 13 -------------
 common/zoned    | 28 ++++++++++++++++++++++++++++
 tests/btrfs/237 |  8 ++------
 3 files changed, 30 insertions(+), 19 deletions(-)
 create mode 100644 common/zoned
common/filter +0 -13
index 28dea64662dc..ac5c93422567 100644
@@ -651,18 +651,5 @@ _filter_bash()
 	sed -e "s/^bash: line 1: /bash: /"
 }
 
-#
-# blkzone report added zone capacity to be printed from v2.37.
-# This filter will add an extra column 'cap' with the same value of
-# 'len'(zone size) for blkzone version < 2.37
-#
-# Before: start: 0x000100000, len 0x040000, wptr 0x000000 ..
-# After: start: 0x000100000, len 0x040000, cap 0x040000, wptr 0x000000 ..
-_filter_blkzone_report()
-{
-	$AWK_PROG -F "," 'BEGIN{OFS=",";} $3 !~ /cap/ {$2=$2","$2;} {print;}' |\
-	sed -e 's/len/cap/2'
-}
-
 # make sure this script returns success
 /bin/true
common/zoned new file +28 -0
new file mode 100644
index 000000000000..d1bc60f784a1
@@ -0,0 +1,28 @@
+#
+# Common zoned block device specific functions
+#
+
+#
+# blkzone report added zone capacity to be printed from v2.37.
+# This filter will add an extra column 'cap' with the same value of
+# 'len'(zone size) for blkzone version < 2.37
+#
+# Before: start: 0x000100000, len 0x040000, wptr 0x000000 ..
+# After: start: 0x000100000, len 0x040000, cap 0x040000, wptr 0x000000 ..
+_filter_blkzone_report()
+{
+	$AWK_PROG -F "," 'BEGIN{OFS=",";} $3 !~ /cap/ {$2=$2","$2;} {print;}' |\
+	sed -e 's/len/cap/2'
+}
+
+_zone_capacity() {
+    local phy=$1
+    local dev=$2
+
+    [ -z "$dev" ] && dev=$SCRATCH_DEV
+
+    size=$($BLKZONE_PROG report -o $phy -l 1 $dev |\
+	       _filter_blkzone_report |\
+	       grep -Po "cap 0x[[:xdigit:]]+" | cut -d ' ' -f 2)
+    echo $((size << 9))
+}
tests/btrfs/237 +0 -0
index bc6522e2200a..101094b5ce70 100755

tests/btrfs/237 line 16

Quoted 5 lines
tests/btrfs/237 +1 -1
@@ -13,7 +13,7 @@
 _begin_fstest auto quick zone balance
 
 # Import common functions.
-. ./common/filter
+. ./common/zbd

I'm a little surprised this line doesn't report error :) Anyway, it should be common/zoned as above. Others look good to me. With this fix, you can add:

Quoted 19 lines
tests/btrfs/237 +1 -5
 # real QA test starts here
 
@@ -56,11 +56,7 @@ fi
 
 start_data_bg_phy=$(get_data_bg_physical)
 start_data_bg_phy=$((start_data_bg_phy >> 9))
-
-size=$($BLKZONE_PROG report -o $start_data_bg_phy -l 1 $SCRATCH_DEV |\
-	_filter_blkzone_report |\
-	grep -Po "cap 0x[[:xdigit:]]+" | cut -d ' ' -f 2)
-size=$((size << 9))
+size=$(_zone_capacity $start_data_bg_phy)
 
 reclaim_threshold=75
 echo $reclaim_threshold > /sys/fs/btrfs/"$uuid"/bg_reclaim_threshold
Signature · 1 line
2.37.3
#3 lore raw

Re: [PATCH v2 1/2] common: introduce zone_capacity() to return a zone capacity

Quoted 77 lines · Zorro Lang
Quoted 76 lines · Naohiro Aota

Introduce _zone_capacity() to return a zone capacity of the given address in the given device (optional). Move _filter_blkzone_report() for it, and rewrite btrfs/237 with it.

---
 common/filter   | 13 -------------
 common/zoned    | 28 ++++++++++++++++++++++++++++
 tests/btrfs/237 |  8 ++------
 3 files changed, 30 insertions(+), 19 deletions(-)
 create mode 100644 common/zoned
common/filter +0 -13
index 28dea64662dc..ac5c93422567 100644
@@ -651,18 +651,5 @@ _filter_bash()
 	sed -e "s/^bash: line 1: /bash: /"
 }
 
-#
-# blkzone report added zone capacity to be printed from v2.37.
-# This filter will add an extra column 'cap' with the same value of
-# 'len'(zone size) for blkzone version < 2.37
-#
-# Before: start: 0x000100000, len 0x040000, wptr 0x000000 ..
-# After: start: 0x000100000, len 0x040000, cap 0x040000, wptr 0x000000 ..
-_filter_blkzone_report()
-{
-	$AWK_PROG -F "," 'BEGIN{OFS=",";} $3 !~ /cap/ {$2=$2","$2;} {print;}' |\
-	sed -e 's/len/cap/2'
-}
-
 # make sure this script returns success
 /bin/true
common/zoned new file +28 -0
new file mode 100644
index 000000000000..d1bc60f784a1
@@ -0,0 +1,28 @@
+#
+# Common zoned block device specific functions
+#
+
+#
+# blkzone report added zone capacity to be printed from v2.37.
+# This filter will add an extra column 'cap' with the same value of
+# 'len'(zone size) for blkzone version < 2.37
+#
+# Before: start: 0x000100000, len 0x040000, wptr 0x000000 ..
+# After: start: 0x000100000, len 0x040000, cap 0x040000, wptr 0x000000 ..
+_filter_blkzone_report()
+{
+	$AWK_PROG -F "," 'BEGIN{OFS=",";} $3 !~ /cap/ {$2=$2","$2;} {print;}' |\
+	sed -e 's/len/cap/2'
+}
+
+_zone_capacity() {
+    local phy=$1
+    local dev=$2
+
+    [ -z "$dev" ] && dev=$SCRATCH_DEV
+
+    size=$($BLKZONE_PROG report -o $phy -l 1 $dev |\
+	       _filter_blkzone_report |\
+	       grep -Po "cap 0x[[:xdigit:]]+" | cut -d ' ' -f 2)
+    echo $((size << 9))
+}
tests/btrfs/237 +0 -0
index bc6522e2200a..101094b5ce70 100755
@@ -13,7 +13,7 @@
 _begin_fstest auto quick zone balance
 
Quoted 6 lines

tests/btrfs/237 line 16

Quoted 3 lines
tests/btrfs/237 +1 -1
 # Import common functions.
-. ./common/filter
+. ./common/zbd

I'm a little surprised this line doesn't report error :) Anyway, it should be common/zoned as above. Others look good to me. With this fix, you can add:

Oops, I forgot to run this one. Thank you for catching this.

Quoted 22 lines
Quoted 19 lines
tests/btrfs/237 +1 -5
 # real QA test starts here
 
@@ -56,11 +56,7 @@ fi
 
 start_data_bg_phy=$(get_data_bg_physical)
 start_data_bg_phy=$((start_data_bg_phy >> 9))
-
-size=$($BLKZONE_PROG report -o $start_data_bg_phy -l 1 $SCRATCH_DEV |\
-	_filter_blkzone_report |\
-	grep -Po "cap 0x[[:xdigit:]]+" | cut -d ' ' -f 2)
-size=$((size << 9))
+size=$(_zone_capacity $start_data_bg_phy)
 
 reclaim_threshold=75
 echo $reclaim_threshold > /sys/fs/btrfs/"$uuid"/bg_reclaim_threshold
Signature · 1 line
2.37.3
#4 lore raw

PATCH v2 2/2 btrfs: test active zone tracking

In reply to Naohiro Aota

A ZNS device limits the number of active zones, which is the number of zones can be written at the same time. To deal with the limit, btrfs's zoned mode tracks which zone (corresponds to a block group on the SINGLE profile) is active, and finish a zone if necessary.

This test checks if the active zone tracking and the finishing of zones works properly. First, it fills <number of max active zones> zones mostly. And, run some data/metadata stress workload to force btrfs to use a new zone.

This test fails on an older kernel (e.g, 5.18.2) like below.

btrfs/292
[failed, exit status 1]- output mismatch (see /host/btrfs/292.out.bad)

    --- tests/btrfs/292.out     2022-09-15 07:52:18.000000000 +0000
    +++ /host/btrfs/292.out.bad 2022-09-15 07:59:14.290967793 +0000
    @@ -1,2 +1,5 @@
     QA output created by 292
    -Silence is golden
    +stress_data_bgs failed
    +stress_data_bgs_2 failed
    +failed: '/bin/btrfs subvolume snapshot /mnt/scratch /mnt/scratch/snap825'
    +(see /host/btrfs/292.full for details)
    ...
    (Run 'diff -u /var/lib/xfstests/tests/btrfs/292.out /host/btrfs/292.out.bad'  to see the entire diff)

The failure is fixed with a series "btrfs: zoned: fix active zone tracking issues" [1] (upstream commits from 65ea1b66482f ("block: add bdev_max_segments() helper") to 2ce543f47843 ("btrfs: zoned: wait until zone is finished when allocation didn't progress")).

[1] https://lore.kernel.org/linux-btrfs/[email protected]/

---
 common/zoned        |  11 ++++
 tests/btrfs/292     | 136 ++++++++++++++++++++++++++++++++++++++++++++
 tests/btrfs/292.out |   2 +
 3 files changed, 149 insertions(+)
 create mode 100755 tests/btrfs/292
 create mode 100644 tests/btrfs/292.out
common/zoned +11 -0
index d1bc60f784a1..eed0082a15cf 100644
@@ -15,6 +15,17 @@ _filter_blkzone_report()
 	sed -e 's/len/cap/2'
 }
 
+_require_limited_active_zones() {
+    local dev=$1
+    local sysfs=$(_sysfs_dev ${dev})
+    local attr="${sysfs}/queue/max_active_zones"
+
+    [ -e "${attr}" ] || _notrun "cannot find queue/max_active_zones. Maybe non-zoned device?"
+    if [ $(cat "${attr}") == 0 ]; then
+	_notrun "this test requires limited active zones"
+    fi
+}
+
 _zone_capacity() {
     local phy=$1
     local dev=$2
tests/btrfs/292 new file +136 -0

136 lines not on this page · raw message

tests/btrfs/292.out new file +2 -0
new file mode 100644
index 000000000000..627309d3fbd2
@@ -0,0 +1,2 @@
+QA output created by 292
+Silence is golden
Signature · 1 line
2.37.3
#5 lore raw
Quoted 26 lines · Naohiro Aota

A ZNS device limits the number of active zones, which is the number of zones can be written at the same time. To deal with the limit, btrfs's zoned mode tracks which zone (corresponds to a block group on the SINGLE profile) is active, and finish a zone if necessary.

This test checks if the active zone tracking and the finishing of zones works properly. First, it fills <number of max active zones> zones mostly. And, run some data/metadata stress workload to force btrfs to use a new zone.

This test fails on an older kernel (e.g, 5.18.2) like below.

btrfs/292
[failed, exit status 1]- output mismatch (see /host/btrfs/292.out.bad)

    --- tests/btrfs/292.out     2022-09-15 07:52:18.000000000 +0000
    +++ /host/btrfs/292.out.bad 2022-09-15 07:59:14.290967793 +0000
    @@ -1,2 +1,5 @@
     QA output created by 292
    -Silence is golden
    +stress_data_bgs failed
    +stress_data_bgs_2 failed
    +failed: '/bin/btrfs subvolume snapshot /mnt/scratch /mnt/scratch/snap825'
    +(see /host/btrfs/292.full for details)
    ...
    (Run 'diff -u /var/lib/xfstests/tests/btrfs/292.out /host/btrfs/292.out.bad'  to see the entire diff)
Quoted 4 lines

The failure is fixed with a series "btrfs: zoned: fix active zone tracking issues" [1] (upstream commits from 65ea1b66482f ("block: add bdev_max_segments() helper") to 2ce543f47843 ("btrfs: zoned: wait until zone is finished when allocation didn't progress")).

If this's a regression test case for known fix, we'd better to use:
_fixed_by_kernel_commit 65ea1b66482f block: add bdev_max_segments (patchset)

to remind downstream testers about that known issue.

Quoted 44 lines

[1] https://lore.kernel.org/linux-btrfs/[email protected]/

---
 common/zoned        |  11 ++++
 tests/btrfs/292     | 136 ++++++++++++++++++++++++++++++++++++++++++++
 tests/btrfs/292.out |   2 +
 3 files changed, 149 insertions(+)
 create mode 100755 tests/btrfs/292
 create mode 100644 tests/btrfs/292.out
common/zoned +11 -0
index d1bc60f784a1..eed0082a15cf 100644
@@ -15,6 +15,17 @@ _filter_blkzone_report()
 	sed -e 's/len/cap/2'
 }
 
+_require_limited_active_zones() {
+    local dev=$1
+    local sysfs=$(_sysfs_dev ${dev})
+    local attr="${sysfs}/queue/max_active_zones"
+
+    [ -e "${attr}" ] || _notrun "cannot find queue/max_active_zones. Maybe non-zoned device?"
+    if [ $(cat "${attr}") == 0 ]; then
+	_notrun "this test requires limited active zones"
+    fi
+}
+
 _zone_capacity() {
     local phy=$1
     local dev=$2
tests/btrfs/292 new file +4 -0
new file mode 100755
index 000000000000..6cfd6b18c299
@@ -0,0 +1,136 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2022 Western Digital Corporation.  All Rights Reserved.
+#

tests/btrfs/292 line 14

Quoted 10 lines
tests/btrfs/292 new file +10 -0
+# FS QA Test 292
+#
+# Test that an active zone is properly reclaimed to allow the further
+# allocations, even if the active zones are mostly filled.
+#
+. ./common/preamble
+_begin_fstest auto quick snapshot zone
+
+# Import common functions.
+. ./common/btrfs

It's not necessary, due to fs specified common file is included automatically by _source_specific_fs() according to $FSTYP.

Quoted 16 lines
tests/btrfs/292 new file +16 -0
+. ./common/zoned
+
+# real QA test starts here
+
+_supported_fs btrfs
+_require_scratch
+_require_zoned_device "$SCRATCH_DEV"
+_require_limited_active_zones "$SCRATCH_DEV"
+
+_require_command "$BLKZONE_PROG" blkzone
+_require_btrfs_command inspect-internal dump-tree
+
+# This test requires specific data space usage, skip if we have compression
+# enabled.
+_require_no_compress
+

tests/btrfs/292 line 40

Quoted 10 lines
tests/btrfs/292 new file +10 -0
+max_active=$(cat $(_sysfs_dev ${SCRATCH_DEV})/queue/max_active_zones)
+
+# Fill the zones leaving the last 1MB
+fill_active_zones() {
+    # Asuumes we have the same capacity between zones.
+    local capacity=$(_zone_capacity 0)
+    local fill_size=$((capacity - 1024 * 1024))
+
+    for x in $(seq ${max_active}); do
+	dd if=/dev/zero of=${SCRATCH_MNT}/fill$(printf "%02d" $x) bs=${fill_size} \

What kind of indentation do you use? 4 spaces? 2 spaces? or a tab? Although that's not a big deal, 8 characters width tab is recommended in fstests :)

Quoted 33 lines
tests/btrfs/292 new file +33 -0
+	   count=1 oflag=direct 2>/dev/null
+	$BTRFS_UTIL_PROG filesystem sync ${SCRATCH_MNT}
+
+	local nactive=$($BLKZONE_PROG report ${SCRATCH_DEV} | grep oi | wc -l)
+	if [[ ${nactive} == ${max_active} ]]; then
+	    break
+	fi
+    done
+
+    echo "max active zones: ${max_active}" >> $seqres.full
+    $BLKZONE_PROG report ${SCRATCH_DEV} | grep oi | cat -n >> $seqres.full
+}
+
+workout() {
+    local func="$1"
+
+    _scratch_mkfs >/dev/null 2>&1
+    _scratch_mount
+
+    fill_active_zones
+    eval "$func"
+    local ret=$?
+
+    _scratch_unmount
+    _check_btrfs_filesystem ${SCRATCH_DEV}
+
+    return $ret
+}
+
+stress_data_bgs() {
+    # This dd fails with ENOSPC, which should not :(
+    dd if=/dev/zero of=${SCRATCH_MNT}/large bs=64M count=1 oflag=sync \
+       >>$seqres.full 2>&1

tests/btrfs/292 line 83

Quoted 10 lines
tests/btrfs/292 new file +10 -0
+}
+
+stress_data_bgs_2() {
+    # This dd fails with ENOSPC, which should not :(
+    dd if=/dev/zero of=${SCRATCH_MNT}/large bs=64M count=10 conv=fsync \
+       >>$seqres.full 2>&1 &
+    local pid1=$!
+
+    dd if=/dev/zero of=${SCRATCH_MNT}/large2 bs=64M count=10 conv=fsync \
+       >>$seqres.full 2>&1 &

If we used background processes, we'd better to deal with them in cleanup time. But above two background processes are simple enough, I think you can add a "wait" in _cleanup to make sure these backgroud processes are done. Or you'd like to remove the "local" for pid1 and pid2, then does below in _cleanup:

_cleanup()

{
	[ -n "$pid1" ] && kill $pid1
	[ -n "$pid2" ] && kill $pid2
	wait
	...
}

And ...

tests/btrfs/292 line 86

Quoted 3 lines
tests/btrfs/292 new file +3 -0
+    local pid2=$!
+
+    wait $pid1; local ret1=$?

unset pid1

tests/btrfs/292 line 87

Quoted 1 line
tests/btrfs/292 new file +1 -0
+    wait $pid2; local ret2=$?

unset pid2

At here

Quoted 33 lines
tests/btrfs/292 new file +33 -0
+
+    if [ $ret1 -ne 0 -o $ret2 -ne 0 ]; then
+	return 1
+    fi
+    return 0
+}
+
+get_meta_bgs() {
+    $BTRFS_UTIL_PROG inspect-internal dump-tree -t EXTENT ${SCRATCH_DEV} |
+        grep BLOCK_GROUP -A 1 |grep -B1 'METADATA|' |
+        grep -oP '\(\d+ BLOCK_GROUP_ITEM \d+\)'
+}
+
+# This test case does not return the result because
+# _run_btrfs_util_prog will call _fail() in the error case anyway.
+stress_metadata_bgs() {
+    local metabgs=$(get_meta_bgs)
+    local count=0
+
+    while : ; do
+        _run_btrfs_util_prog subvolume snapshot ${SCRATCH_MNT} ${SCRATCH_MNT}/snap$i
+        _run_btrfs_util_prog filesystem sync ${SCRATCH_MNT}
+        cur_metabgs=$(get_meta_bgs)
+        if [[ "${cur_metabgs}" != "${metabgs}" ]]; then
+            break
+        fi
+        i=$((i + 1))
+    done
+}
+
+WORKS=(
+    stress_data_bgs
+    stress_data_bgs_2

tests/btrfs/292 line 130

Quoted 10 lines
tests/btrfs/292 new file +10 -0
+    stress_metadata_bgs
+)
+
+status=0
+for work in "${WORKS[@]}"; do
+    if ! workout "${work}"; then
+	echo "${work} failed"
+	status=1
+    fi
+done

The $status is 1 at the beginning of each case, and we set it to 0 at the end of each case. If a test case _fail exit in the middle, then status keep be 1. For your case, I think you don't need to touch the $status, you _fail directly if anyone *worktout* returns failure. Or just let the "${work} failed" output breaks the golden image(.out) to cause a test failure (no matter the status=0/1).

tests/btrfs/292 line 135

Quoted 5 lines
tests/btrfs/292 new file +5 -0
+
+# success, all done
+if [ $status -eq 0 ]; then
+    echo "Silence is golden"
+fi

You can output "Silence is golden" at here directly, due to this case expect that "Silence".

I'm not the best reviewer for a zoned&btrfs related case, so I tried to review from fstests side. I saw Johannes Thumshirn <[email protected]> has given you a RVB last time, I think you can keep that as a review from btrfs and zbd side (except he has more review points now).

Thanks,
Zorro

Quoted 12 lines
tests/btrfs/292 new file +1 -0
+exit
tests/btrfs/292.out new file +2 -0
new file mode 100644
index 000000000000..627309d3fbd2
@@ -0,0 +1,2 @@
+QA output created by 292
+Silence is golden
Signature · 1 line
2.37.3
#6 lore raw
Quoted 27 lines · Zorro Lang
Quoted 26 lines · Naohiro Aota

A ZNS device limits the number of active zones, which is the number of zones can be written at the same time. To deal with the limit, btrfs's zoned mode tracks which zone (corresponds to a block group on the SINGLE profile) is active, and finish a zone if necessary.

This test checks if the active zone tracking and the finishing of zones works properly. First, it fills <number of max active zones> zones mostly. And, run some data/metadata stress workload to force btrfs to use a new zone.

This test fails on an older kernel (e.g, 5.18.2) like below.

btrfs/292
[failed, exit status 1]- output mismatch (see /host/btrfs/292.out.bad)

    --- tests/btrfs/292.out     2022-09-15 07:52:18.000000000 +0000
    +++ /host/btrfs/292.out.bad 2022-09-15 07:59:14.290967793 +0000
    @@ -1,2 +1,5 @@
     QA output created by 292
    -Silence is golden
    +stress_data_bgs failed
    +stress_data_bgs_2 failed
    +failed: '/bin/btrfs subvolume snapshot /mnt/scratch /mnt/scratch/snap825'
    +(see /host/btrfs/292.full for details)
    ...
    (Run 'diff -u /var/lib/xfstests/tests/btrfs/292.out /host/btrfs/292.out.bad'  to see the entire diff)
Quoted 7 lines
Quoted 4 lines

The failure is fixed with a series "btrfs: zoned: fix active zone tracking issues" [1] (upstream commits from 65ea1b66482f ("block: add bdev_max_segments() helper") to 2ce543f47843 ("btrfs: zoned: wait until zone is finished when allocation didn't progress")).

If this's a regression test case for known fix, we'd better to use:
_fixed_by_kernel_commit 65ea1b66482f block: add bdev_max_segments (patchset)

This is very misleading as the commit only adds a helper that's used in later commits. If anything, the last commit in the series should be mentioned.

#7 lore raw
Quoted 33 lines · David Sterba
Quoted 32 lines · Zorro Lang
Quoted 30 lines · Naohiro Aota

A ZNS device limits the number of active zones, which is the number of zones can be written at the same time. To deal with the limit, btrfs's zoned mode tracks which zone (corresponds to a block group on the SINGLE profile) is active, and finish a zone if necessary.

This test checks if the active zone tracking and the finishing of zones works properly. First, it fills <number of max active zones> zones mostly. And, run some data/metadata stress workload to force btrfs to use a new zone.

This test fails on an older kernel (e.g, 5.18.2) like below.

btrfs/292
[failed, exit status 1]- output mismatch (see /host/btrfs/292.out.bad)

    --- tests/btrfs/292.out     2022-09-15 07:52:18.000000000 +0000
    +++ /host/btrfs/292.out.bad 2022-09-15 07:59:14.290967793 +0000
    @@ -1,2 +1,5 @@
     QA output created by 292
    -Silence is golden
    +stress_data_bgs failed
    +stress_data_bgs_2 failed
    +failed: '/bin/btrfs subvolume snapshot /mnt/scratch /mnt/scratch/snap825'
    +(see /host/btrfs/292.full for details)
    ...
    (Run 'diff -u /var/lib/xfstests/tests/btrfs/292.out /host/btrfs/292.out.bad'  to see the entire diff)

The failure is fixed with a series "btrfs: zoned: fix active zone tracking issues" [1] (upstream commits from 65ea1b66482f ("block: add bdev_max_segments() helper") to 2ce543f47843 ("btrfs: zoned: wait until zone is finished when allocation didn't progress")).

Quoted 6 lines
Quoted 2 lines

If this's a regression test case for known fix, we'd better to use:
_fixed_by_kernel_commit 65ea1b66482f block: add bdev_max_segments (patchset)

This is very misleading as the commit only adds a helper that's used in later commits. If anything, the last commit in the series should be mentioned.

Sure, I just gave an example, you learned about that patchset more than me, so feel free to pick up a proper commit and description :)

Quoted 1 line
#8 lore raw
Quoted 37 lines · Zorro Lang
Quoted 36 lines · David Sterba
Quoted 34 lines · Zorro Lang
Quoted 26 lines · Naohiro Aota

A ZNS device limits the number of active zones, which is the number of zones can be written at the same time. To deal with the limit, btrfs's zoned mode tracks which zone (corresponds to a block group on the SINGLE profile) is active, and finish a zone if necessary.

This test checks if the active zone tracking and the finishing of zones works properly. First, it fills <number of max active zones> zones mostly. And, run some data/metadata stress workload to force btrfs to use a new zone.

This test fails on an older kernel (e.g, 5.18.2) like below.

btrfs/292
[failed, exit status 1]- output mismatch (see /host/btrfs/292.out.bad)

    --- tests/btrfs/292.out     2022-09-15 07:52:18.000000000 +0000
    +++ /host/btrfs/292.out.bad 2022-09-15 07:59:14.290967793 +0000
    @@ -1,2 +1,5 @@
     QA output created by 292
    -Silence is golden
    +stress_data_bgs failed
    +stress_data_bgs_2 failed
    +failed: '/bin/btrfs subvolume snapshot /mnt/scratch /mnt/scratch/snap825'
    +(see /host/btrfs/292.full for details)
    ...
    (Run 'diff -u /var/lib/xfstests/tests/btrfs/292.out /host/btrfs/292.out.bad'  to see the entire diff)
Quoted 4 lines

The failure is fixed with a series "btrfs: zoned: fix active zone tracking issues" [1] (upstream commits from 65ea1b66482f ("block: add bdev_max_segments() helper") to 2ce543f47843 ("btrfs: zoned: wait until zone is finished when allocation didn't progress")).

If this's a regression test case for known fix, we'd better to use:
_fixed_by_kernel_commit 65ea1b66482f block: add bdev_max_segments (patchset)

Quoted 6 lines
Quoted 3 lines

This is very misleading as the commit only adds a helper that's used in later commits. If anything, the last commit in the series should be mentioned.

Sure, I just gave an example, you learned about that patchset more than me, so feel free to pick up a proper commit and description :)

Indeed. I'll use the last commit here.

Quoted 2 lines
Quoted 1 line
#9 lore raw
In reply to Zorro Lang
Quoted 86 lines · Zorro Lang
Quoted 26 lines · Naohiro Aota

A ZNS device limits the number of active zones, which is the number of zones can be written at the same time. To deal with the limit, btrfs's zoned mode tracks which zone (corresponds to a block group on the SINGLE profile) is active, and finish a zone if necessary.

This test checks if the active zone tracking and the finishing of zones works properly. First, it fills <number of max active zones> zones mostly. And, run some data/metadata stress workload to force btrfs to use a new zone.

This test fails on an older kernel (e.g, 5.18.2) like below.

btrfs/292
[failed, exit status 1]- output mismatch (see /host/btrfs/292.out.bad)

    --- tests/btrfs/292.out     2022-09-15 07:52:18.000000000 +0000
    +++ /host/btrfs/292.out.bad 2022-09-15 07:59:14.290967793 +0000
    @@ -1,2 +1,5 @@
     QA output created by 292
    -Silence is golden
    +stress_data_bgs failed
    +stress_data_bgs_2 failed
    +failed: '/bin/btrfs subvolume snapshot /mnt/scratch /mnt/scratch/snap825'
    +(see /host/btrfs/292.full for details)
    ...
    (Run 'diff -u /var/lib/xfstests/tests/btrfs/292.out /host/btrfs/292.out.bad'  to see the entire diff)
Quoted 4 lines

The failure is fixed with a series "btrfs: zoned: fix active zone tracking issues" [1] (upstream commits from 65ea1b66482f ("block: add bdev_max_segments() helper") to 2ce543f47843 ("btrfs: zoned: wait until zone is finished when allocation didn't progress")).

If this's a regression test case for known fix, we'd better to use:
_fixed_by_kernel_commit 65ea1b66482f block: add bdev_max_segments (patchset)

to remind downstream testers about that known issue.

Quoted 49 lines

[1] https://lore.kernel.org/linux-btrfs/[email protected]/

---
 common/zoned        |  11 ++++
 tests/btrfs/292     | 136 ++++++++++++++++++++++++++++++++++++++++++++
 tests/btrfs/292.out |   2 +
 3 files changed, 149 insertions(+)
 create mode 100755 tests/btrfs/292
 create mode 100644 tests/btrfs/292.out
common/zoned +11 -0
index d1bc60f784a1..eed0082a15cf 100644
@@ -15,6 +15,17 @@ _filter_blkzone_report()
 	sed -e 's/len/cap/2'
 }
 
+_require_limited_active_zones() {
+    local dev=$1
+    local sysfs=$(_sysfs_dev ${dev})
+    local attr="${sysfs}/queue/max_active_zones"
+
+    [ -e "${attr}" ] || _notrun "cannot find queue/max_active_zones. Maybe non-zoned device?"
+    if [ $(cat "${attr}") == 0 ]; then
+	_notrun "this test requires limited active zones"
+    fi
+}
+
 _zone_capacity() {
     local phy=$1
     local dev=$2
tests/btrfs/292 new file +9 -0
new file mode 100755
index 000000000000..6cfd6b18c299
@@ -0,0 +1,136 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2022 Western Digital Corporation.  All Rights Reserved.
+#
+# FS QA Test 292
+#
+# Test that an active zone is properly reclaimed to allow the further
+# allocations, even if the active zones are mostly filled.
+#
Quoted 8 lines

tests/btrfs/292 line 14

Quoted 5 lines
tests/btrfs/292 new file +5 -0
+. ./common/preamble
+_begin_fstest auto quick snapshot zone
+
+# Import common functions.
+. ./common/btrfs

It's not necessary, due to fs specified common file is included automatically by _source_specific_fs() according to $FSTYP.

Yep. I'll drop this.

Quoted 21 lines
Quoted 21 lines
tests/btrfs/292 new file +21 -0
+. ./common/zoned
+
+# real QA test starts here
+
+_supported_fs btrfs
+_require_scratch
+_require_zoned_device "$SCRATCH_DEV"
+_require_limited_active_zones "$SCRATCH_DEV"
+
+_require_command "$BLKZONE_PROG" blkzone
+_require_btrfs_command inspect-internal dump-tree
+
+# This test requires specific data space usage, skip if we have compression
+# enabled.
+_require_no_compress
+
+max_active=$(cat $(_sysfs_dev ${SCRATCH_DEV})/queue/max_active_zones)
+
+# Fill the zones leaving the last 1MB
+fill_active_zones() {
+    # Asuumes we have the same capacity between zones.
Quoted 8 lines

tests/btrfs/292 line 40

Quoted 5 lines
tests/btrfs/292 new file +5 -0
+    local capacity=$(_zone_capacity 0)
+    local fill_size=$((capacity - 1024 * 1024))
+
+    for x in $(seq ${max_active}); do
+	dd if=/dev/zero of=${SCRATCH_MNT}/fill$(printf "%02d" $x) bs=${fill_size} \

What kind of indentation do you use? 4 spaces? 2 spaces? or a tab? Although that's not a big deal, 8 characters width tab is recommended in fstests :)

Ah, I use both vim and emacs and there must be confusing. I'll reformat.

Quoted 65 lines
Quoted 34 lines
tests/btrfs/292 new file +33 -0
+	   count=1 oflag=direct 2>/dev/null
+	$BTRFS_UTIL_PROG filesystem sync ${SCRATCH_MNT}
+
+	local nactive=$($BLKZONE_PROG report ${SCRATCH_DEV} | grep oi | wc -l)
+	if [[ ${nactive} == ${max_active} ]]; then
+	    break
+	fi
+    done
+
+    echo "max active zones: ${max_active}" >> $seqres.full
+    $BLKZONE_PROG report ${SCRATCH_DEV} | grep oi | cat -n >> $seqres.full
+}
+
+workout() {
+    local func="$1"
+
+    _scratch_mkfs >/dev/null 2>&1
+    _scratch_mount
+
+    fill_active_zones
+    eval "$func"
+    local ret=$?
+
+    _scratch_unmount
+    _check_btrfs_filesystem ${SCRATCH_DEV}
+
+    return $ret
+}
+
+stress_data_bgs() {
+    # This dd fails with ENOSPC, which should not :(
+    dd if=/dev/zero of=${SCRATCH_MNT}/large bs=64M count=1 oflag=sync \
+       >>$seqres.full 2>&1

tests/btrfs/292 line 83

Quoted 10 lines
tests/btrfs/292 new file +10 -0
+}
+
+stress_data_bgs_2() {
+    # This dd fails with ENOSPC, which should not :(
+    dd if=/dev/zero of=${SCRATCH_MNT}/large bs=64M count=10 conv=fsync \
+       >>$seqres.full 2>&1 &
+    local pid1=$!
+
+    dd if=/dev/zero of=${SCRATCH_MNT}/large2 bs=64M count=10 conv=fsync \
+       >>$seqres.full 2>&1 &

If we used background processes, we'd better to deal with them in cleanup time. But above two background processes are simple enough, I think you can add a "wait" in _cleanup to make sure these backgroud processes are done. Or you'd like to remove the "local" for pid1 and pid2, then does below in _cleanup:

_cleanup()

{
	[ -n "$pid1" ] && kill $pid1
	[ -n "$pid2" ] && kill $pid2
	wait
	...
}

And ...

tests/btrfs/292 line 86

Quoted 3 lines
tests/btrfs/292 new file +3 -0
+    local pid2=$!
+
+    wait $pid1; local ret1=$?
Quoted 7 lines

unset pid1

tests/btrfs/292 line 87

Quoted 1 line
tests/btrfs/292 new file +1 -0
+    wait $pid2; local ret2=$?

unset pid2

At here

Sure. I'll take this. Thank you.

Quoted 50 lines
Quoted 33 lines
tests/btrfs/292 new file +33 -0
+
+    if [ $ret1 -ne 0 -o $ret2 -ne 0 ]; then
+	return 1
+    fi
+    return 0
+}
+
+get_meta_bgs() {
+    $BTRFS_UTIL_PROG inspect-internal dump-tree -t EXTENT ${SCRATCH_DEV} |
+        grep BLOCK_GROUP -A 1 |grep -B1 'METADATA|' |
+        grep -oP '\(\d+ BLOCK_GROUP_ITEM \d+\)'
+}
+
+# This test case does not return the result because
+# _run_btrfs_util_prog will call _fail() in the error case anyway.
+stress_metadata_bgs() {
+    local metabgs=$(get_meta_bgs)
+    local count=0
+
+    while : ; do
+        _run_btrfs_util_prog subvolume snapshot ${SCRATCH_MNT} ${SCRATCH_MNT}/snap$i
+        _run_btrfs_util_prog filesystem sync ${SCRATCH_MNT}
+        cur_metabgs=$(get_meta_bgs)
+        if [[ "${cur_metabgs}" != "${metabgs}" ]]; then
+            break
+        fi
+        i=$((i + 1))
+    done
+}
+
+WORKS=(
+    stress_data_bgs
+    stress_data_bgs_2

tests/btrfs/292 line 130

Quoted 10 lines
tests/btrfs/292 new file +10 -0
+    stress_metadata_bgs
+)
+
+status=0
+for work in "${WORKS[@]}"; do
+    if ! workout "${work}"; then
+	echo "${work} failed"
+	status=1
+    fi
+done

The $status is 1 at the beginning of each case, and we set it to 0 at the end of each case. If a test case _fail exit in the middle, then status keep be 1. For your case, I think you don't need to touch the $status, you _fail directly if anyone *worktout* returns failure. Or just let the "${work} failed" output breaks the golden image(.out) to cause a test failure (no matter the status=0/1).

Quoted 8 lines

tests/btrfs/292 line 135

Quoted 5 lines
tests/btrfs/292 new file +5 -0
+
+# success, all done
+if [ $status -eq 0 ]; then
+    echo "Silence is golden"
+fi

You can output "Silence is golden" at here directly, due to this case expect that "Silence".

Ah, yes. I tried to run as many tests as possible, even if one test fails. But, I now think it's not much worthwhile. I'll just _fail any failed work.

Quoted 4 lines

I'm not the best reviewer for a zoned&btrfs related case, so I tried to review from fstests side. I saw Johannes Thumshirn <[email protected]> has given you a RVB last time, I think you can keep that as a review from btrfs and zbd side (except he has more review points now).

Well, I forgot to add the tag. Thanks.

Quoted 16 lines

Thanks,
Zorro

Quoted 12 lines
tests/btrfs/292 new file +1 -0
+exit
tests/btrfs/292.out new file +2 -0
new file mode 100644
index 000000000000..627309d3fbd2
@@ -0,0 +1,2 @@
+QA output created by 292
+Silence is golden
Signature · 1 line
2.37.3