blktests scsi/011 zbd/010 failed on ppc64le

[yizhanglinux]

Hello
I found scsi/011 zbd/010 failed with 7.0.0-rc4 on ppc64le, please help check it, thanks.

# uname -r
7.0.0-0.rc4.260320g0e4f8f1a3d08.40.fc45.ppc64le

# ./check scsi/011 zbd/010
scsi/011 (test data lifetime propagation)                    [failed]
    runtime  1.282s  ...  1.252s
    --- tests/scsi/011.out	2026-03-21 10:20:05.935610925 -0400
    +++ /root/blktests/results/nodev/scsi/011.out.bad	2026-03-21 10:34:12.353519535 -0400
    @@ -1,6 +1,5 @@
     Running scsi/011
    +mount: /root/blktests/results/tmpdir.scsi.011.m9C/mnt: wrong fs type, bad option, bad superblock on /dev/sdb, missing codepage or helper program, or other error.
    +       dmesg(1) may have more information after failed mount system call.
     0
    -1
    -2
    -3
    ...
    (Run 'diff -u tests/scsi/011.out /root/blktests/results/nodev/scsi/011.out.bad' to see the entire diff)
zbd/010 (test gap zone support with F2FS)                    [failed]
    runtime  1.980s  ...  1.690s
    --- tests/zbd/010.out	2026-03-21 10:20:05.939251897 -0400
    +++ /root/blktests/results/nodev/zbd/010.out.bad	2026-03-21 10:34:15.933518559 -0400
    @@ -1,2 +1,4 @@
     Running zbd/010
    -Test complete
    +mount: /root/blktests/results/tmpdir.zbd.010.0SQ/mnt: wrong fs type, bad option, bad superblock on /dev/nullb0, missing codepage or helper program, or other error.
    +       dmesg(1) may have more information after failed mount system call.
    +Test failed

# dmesg
[ 4131.947663] run blktests scsi/011 at 2026-03-21 10:34:11
[ 4132.240177] sd 104:0:0:0: [sdb] Synchronizing SCSI cache
[ 4132.506663] scsi_debug:sdebug_driver_probe: scsi_debug: trim poll_queues to 0. poll_q/nr_hw = (0/1)
[ 4132.506686] scsi host105: scsi_debug: version 0191 [20210520]
                 dev_size_mb=1024, opts=0x0, submit_queues=1, statistics=0
[ 4132.506946] scsi 105:0:0:0: Direct-Access     Linux    scsi_debug       0191 PQ: 0 ANSI: 7
[ 4132.507125] scsi 105:0:0:0: Power-on or device reset occurred
[ 4132.507366] sd 105:0:0:0: Attached scsi generic sg1 type 0
[ 4132.507374] sd 105:0:0:0: [sdb] 262144 4096-byte logical blocks: (1.07 GB/1.00 GiB)
[ 4132.507394] sd 105:0:0:0: [sdb] Write Protect is off
[ 4132.507402] sd 105:0:0:0: [sdb] Mode Sense: 73 00 10 08
[ 4132.507431] sd 105:0:0:0: [sdb] Write cache: enabled, read cache: enabled, supports DPO and FUA
[ 4132.507488] sd 105:0:0:0: [sdb] permanent stream count = 5
[ 4132.507510] sd 105:0:0:0: [sdb] Preferred minimum I/O size 4096 bytes
[ 4132.507513] sd 105:0:0:0: [sdb] Optimal transfer size 4194304 bytes
[ 4132.660001] sd 105:0:0:0: [sdb] Attached SCSI disk
[ 4132.920060] F2FS-fs (sdb): Invalid log_blocksize (12), supports only 16
[ 4132.920069] F2FS-fs (sdb): Can't find valid F2FS filesystem in 1th superblock
[ 4132.920125] F2FS-fs (sdb): Magic Mismatch, valid(0xf2f52010) - read(0x0)
[ 4132.920128] F2FS-fs (sdb): Can't find valid F2FS filesystem in 2th superblock
[ 4133.060145] sd 105:0:0:0: [sdb] Synchronizing SCSI cache
[ 4134.435693] null_blk: disk nullb0 created
[ 4134.435702] null_blk: module loaded
[ 4134.680514] scsi_debug:sdebug_driver_probe: scsi_debug: trim poll_queues to 0. poll_q/nr_hw = (0/1)
[ 4134.680535] scsi host106: scsi_debug: version 0191 [20210520]
                 dev_size_mb=8, opts=0x0, submit_queues=1, statistics=0
[ 4134.680824] scsi 106:0:0:0: Direct-Access     Linux    scsi_debug       0191 PQ: 0 ANSI: 7
[ 4134.680996] scsi 106:0:0:0: Power-on or device reset occurred
[ 4134.711275] sd 106:0:0:0: Attached scsi generic sg1 type 0
[ 4134.731292] sd 106:0:0:0: [sdb] 16384 512-byte logical blocks: (8.39 MB/8.00 MiB)
[ 4134.741313] sd 106:0:0:0: [sdb] Write Protect is off
[ 4134.741316] sd 106:0:0:0: [sdb] Mode Sense: 73 00 10 08
[ 4134.761337] sd 106:0:0:0: [sdb] Write cache: enabled, read cache: enabled, supports DPO and FUA
[ 4134.821398] sd 106:0:0:0: [sdb] permanent stream count = 5
[ 4134.851435] sd 106:0:0:0: [sdb] Preferred minimum I/O size 512 bytes
[ 4134.851439] sd 106:0:0:0: [sdb] Optimal transfer size 524288 bytes
[ 4135.069993] sd 106:0:0:0: [sdb] Attached SCSI disk
[ 4135.089208] run blktests zbd/010 at 2026-03-21 10:34:14
[ 4135.363815] null_blk: module loaded
[ 4135.441930] null_blk: disk nullb0 created
[ 4135.520136] sd 106:0:0:0: [sdb] Synchronizing SCSI cache
[ 4135.746750] scsi_debug:sdebug_driver_probe: scsi_debug: trim poll_queues to 0. poll_q/nr_hw = (0/1)
[ 4135.746772] scsi host107: scsi_debug: version 0191 [20210520]
                 dev_size_mb=1024, opts=0x0, submit_queues=1, statistics=0
[ 4135.747033] scsi 107:0:0:0: Direct-Access-ZBC Linux    scsi_debug       0191 PQ: 0 ANSI: 7
[ 4135.747192] scsi 107:0:0:0: Power-on or device reset occurred
[ 4135.747440] sd 107:0:0:0: [sdb] Host-managed zoned block device
[ 4135.747446] sd 107:0:0:0: Attached scsi generic sg1 type 20
[ 4135.747504] sd 107:0:0:0: [sdb] 262144 4096-byte logical blocks: (1.07 GB/1.00 GiB)
[ 4135.747527] sd 107:0:0:0: [sdb] Write Protect is off
[ 4135.747535] sd 107:0:0:0: [sdb] Mode Sense: 5b 00 10 08
[ 4135.747554] sd 107:0:0:0: [sdb] Write cache: enabled, read cache: enabled, supports DPO and FUA
[ 4135.747596] sd 107:0:0:0: [sdb] permanent stream count = 5
[ 4135.747618] sd 107:0:0:0: [sdb] Preferred minimum I/O size 4096 bytes
[ 4135.747622] sd 107:0:0:0: [sdb] Optimal transfer size 4194304 bytes
[ 4135.747720] sd 107:0:0:0: [sdb] 256 zones of 1024 logical blocks
[ 4135.860016] sd 107:0:0:0: [sdb] Attached SCSI disk
[ 4135.993957] F2FS-fs (nullb0): Invalid log_blocksize (12), supports only 16
[ 4135.993965] F2FS-fs (nullb0): Can't find valid F2FS filesystem in 1th superblock
[ 4135.993985] F2FS-fs (nullb0): Magic Mismatch, valid(0xf2f52010) - read(0x0)
[ 4135.993988] F2FS-fs (nullb0): Can't find valid F2FS filesystem in 2th superblock
[ 4136.230173] sd 107:0:0:0: [sdb] Synchronizing SCSI cache

[kawasaki]

@yizhanglinux Thanks for the report. I think the kernel message "F2FS-fs (sdb): Invalid log_blocksize (12), supports only 16" is the key. It looks that simple f2fs mkfs and mount are failing.

I suggest to try out the commands below. Here, /dev/sdd is the device created by the scsi_debug modprobe.

$ sudo modprobe scsi_debug delay=0 dev_size_mb=1024 sector=4096
$ sudo mkfs.f2fs /dev/sdd
$ sudo mount /dev/sdd /mnt

If the mount fails, something is broken in the ppc environment: f2fs or scsi_debug.

[yizhanglinux]

Yes, the mount operation failed again. Will try to figure out the working version.

# modprobe scsi_debug delay=0 dev_size_mb=1024 sector_size=4096
# mkfs.f2fs /dev/sdb

    F2FS-tools: mkfs.f2fs Ver: 1.16.0 (2023-04-11)

Info: Disable heap-based policy
Info: Debug level = 0
Info: Trim is enabled
Info: [/dev/sdb] Disk Model: scsi_debug
Info: Segments per section = 1
Info: Sections per zone = 1
Info: sector size = 4096
Info: total sectors = 262144 (1024 MB)
Info: zone aligned segment0 blkaddr: 512
Info: format version with
  "Linux version 7.0.0-0.rc5.260327g46b513250491.46.fc45.ppc64le (mockbuild@bd3252f9c84a44e4825f52b5b49fd294) (gcc (GCC) 16.0.1 20260321 (Red Hat 16.0.1-0), GNU ld version 2.46.50.20260316) #1 SMP PREEMPT_DYNAMIC Fri Mar 27 15:18:04 UTC 2026"
Info: [/dev/sdb] Discarding device
Info: This device doesn't support BLKSECDISCARD
Info: This device doesn't support BLKDISCARD
Info: Overprovision ratio = 5.370%
Info: Overprovision segments = 27 (GC reserved = 25)
Info: format successful

# mount /dev/sdb /mnt
mount: /mnt: wrong fs type, bad option, bad superblock on /dev/sdb, missing codepage or helper program, or other error.
       dmesg(1) may have more information after failed mount system call.

# dmesg
[ 6790.651305] scsi_debug:sdebug_driver_probe: scsi_debug: trim poll_queues to 0. poll_q/nr_hw = (0/1)
[ 6790.651321] scsi host18: scsi_debug: version 0191 [20210520]
                 dev_size_mb=1024, opts=0x0, submit_queues=1, statistics=0
[ 6790.651576] scsi 18:0:0:0: Direct-Access     Linux    scsi_debug       0191 PQ: 0 ANSI: 7
[ 6790.656936] scsi 18:0:0:0: Power-on or device reset occurred
[ 6790.657242] sd 18:0:0:0: Attached scsi generic sg1 type 0
[ 6790.657256] sd 18:0:0:0: [sdb] 262144 4096-byte logical blocks: (1.07 GB/1.00 GiB)
[ 6790.657282] sd 18:0:0:0: [sdb] Write Protect is off
[ 6790.657290] sd 18:0:0:0: [sdb] Mode Sense: 73 00 10 08
[ 6790.657326] sd 18:0:0:0: [sdb] Write cache: enabled, read cache: enabled, supports DPO and FUA
[ 6790.657415] sd 18:0:0:0: [sdb] permanent stream count = 5
[ 6790.657458] sd 18:0:0:0: [sdb] Preferred minimum I/O size 4096 bytes
[ 6790.657466] sd 18:0:0:0: [sdb] Optimal transfer size 4194304 bytes
[ 6790.806602] sd 18:0:0:0: [sdb] Attached SCSI disk
[ 6837.908506] F2FS-fs (sdb): Invalid log_blocksize (12), supports only 16
[ 6837.908520] F2FS-fs (sdb): Can't find valid F2FS filesystem in 1th superblock
[ 6837.908545] F2FS-fs (sdb): Magic Mismatch, valid(0xf2f52010) - read(0x0)
[ 6837.908551] F2FS-fs (sdb): Can't find valid F2FS filesystem in 2th superblock

[yizhanglinux]

I tried kernel 6.18 and 6.19, and it can also be reproduced.

[yizhanglinux]

Add the two cases' author @bvanassche

[bvanassche]

How about skipping the tests that use mkfs.f2fs if PAGE_SIZE != 4 KiB?

[kawasaki]

I found this article, and learned that F2FS has the restriction that "F2FS requires block devices to have block size to be the same as the system memory page size". While the test case uses 4KB block scsi_debug, the PPC system has larger page size than 4KB. Hence the failure.

(I think the error message can be more user friendly. I also could not find the restriction described in Documentation/filesystems/f2fs.rst, but these are side stories)

I agree that Bart's idea will be a solution. Just skip the test case if the PAGE_SIZE of the test system is not 4KiB. Another idea is to get PAGE_SIZE of the system, and use that block size of the scsi_debug device. This is more ideal, since the test coverage will not be reduced on PPC systems, or other future systems with PAGE_SIZE > 4KiB.

[bvanassche]

From an LLM:

---

What is the current state of 16 KiB block size support in F2FS?

As of early 2026, the current state of 16 KiB block size support in F2FS is functional but specifically constrained to environments where the system's Page Size also equals 16 KiB.

While F2FS was historically locked to a 4 KiB block size, recent upstreaming efforts (driven largely by the Android ecosystem's move toward 16 KiB pages for performance gains) have relaxed this.

Current Support Status

• Block Size == Page Size: You can now format and mount an F2FS partition with a 16 KiB block size, provided you are running on a kernel/architecture configured with a 16 KiB PAGE_SIZE (common in newer ARM64 profiles).

• The 4 KiB Legacy Barrier: Standard F2FS still relies heavily on the assumption that one filesystem block fits exactly into one memory page. Because of this, you cannot currently mount a 16 KiB block size F2FS volume on a system using 4 KiB pages.

• Large Sector Support: Support for 16 KiB blocks is integrated into f2fs-tools. When formatting, the tool can now negotiate the larger block size, which helps reduce metadata overhead and improves sequential throughput on modern flash storage that uses larger physical NAND pages.

---

In other words, don't expect 64 KiB logical blocks to be supported by F2FS.

[yizhanglinux]

Thanks for the info, just created #239 to fix the failure.

[kawasaki]

@bvanassche Thanks. According to what LLM says, f2fs supports 4KiB and 16KiB block sizes, when the kernel has the same value of page size. In that sense, I think the two test cases 1) should support both 4Kib and 16KiB page sizes, and 2) should skip 64KiB page size. The PR that @yizhanglinux created covers only the 2) part, but I think it is good. I think we can work on the 1) part later.

[yizhanglinux]

@kawasaki
For 1) support both 4Kib and 16KiB page sizes,
How about updating the _page_size_equals function as [1] and the require part as _page_size_equals 4096 16384

[1]

_page_size_equals() {
	local current_size
	current_size=$(_get_page_size)

	# Iterate through all provided page size arguments
	for required_size in "$@"; do
		if [ "$current_size" -eq "$required_size" ]; then
			# Return success immediately if any match is found
			return 0
		fi
	done

	# If no match is found after the loop, the current system page size is not supported.
	# Concatenate all passed arguments into a single string for the skip message.
	local supported_sizes="$*"
	SKIP_REASONS+=("This test requires page size to be one of [${supported_sizes}] bytes, but system has ${current_size} bytes")
	return 1
}

[kawasaki]

@yizhanglinux Thanks. I agree that the suggested approach covers both 4KiB and 16KiB and it will be the better.

Said that, it will need a bit more work.

• Do you have 16k page size system? If you have it and can confirm the test case pass on it, it will be appreciated. I have one ARM test system, and it can boot 16k page size kernel probably. If you do not have such system, I will do the test run on my test system, but it will take some time.
• Currently, scsi/011 and zbd/010 create scsi_debug with 4KiB sector size. I'm not sure if scsi/011 and zbd/010 need additional change to adjust the scsi_debug sector size to the system page size or not.

If you can do and/or wait for these works, we can do 1) and 2) at the same time. But if you can not wait, I'm okay to apply the PR with current change only for 2).