Dear community,
we are getting file system errors on an ext4 partition (that wasn't shut down properly in linux before) after writing to it from u-boot. At the moment we are using v2019.01 with some application specific code that communicates over text files with the update code in the main application running on linux.
But we can reproduce the behaviour on the current version (2022.04) in the sandbox with these steps: -------- [1] create an empty test partition (our partition in the application has 100MB): ~> dd if=/dev/zero of=test.bin bs=8M count=1 1+0 records in 1+0 records out 8388608 bytes (8.4 MB, 8.0 MiB) copied, 0.00804182 s, 1.0 GB/s
[2] create the filesystem: ~> mkfs.ext4 -O ^metadata_csum test.bin mke2fs 1.45.5 (07-Jan-2020) Discarding device blocks: done Creating filesystem with 2048 4k blocks and 2048 inodes
Allocating group tables: done Writing inode tables: done Creating journal (1024 blocks): done Writing superblocks and filesystem accounting information: done
[3] mount the filesystem: ~> mount -o loop,rw test.bin mnt
[4] create a test file and make sure it is written ~> echo test > mnt/test.txt ~> sync
[5] simulate power loss by dd'ing the file while it is mounted ~> dd if=test.bin of=test_notclean.bin 16384+0 records in 16384+0 records out 8388608 bytes (8.4 MB, 8.0 MiB) copied, 0.0307814 s, 273 MB/s
[6] a fsck (without -f option) shows and fixes some minor issues. A fsck with -f finds no more issues: ~> cp test_notclean.bin test_check.bin ~> e2fsck test_check.bin e2fsck 1.45.5 (07-Jan-2020) test_check.bin: recovering journal Setting free inodes count to 2036 (was 2037) Setting free blocks count to 949 (was 950) test_check.bin: clean, 12/2048 files, 1099/2048 blocks
~> e2fsck -f test_check.bin e2fsck 1.45.5 (07-Jan-2020) Pass 1: Checking inodes, blocks, and sizes Pass 2: Checking directory structure Pass 3: Checking directory connectivity Pass 4: Checking reference counts Pass 5: Checking group summary information test_check.bin: 12/2048 files (8.3% non-contiguous), 1099/2048 blocks
[7] Write to the partition in u-boot sandbox and abort with CTRL-C: ~> ./u-boot U-Boot 2022.04 (Apr 08 2022 - 15:20:04 +0200)
DRAM: 128 MiB Core: 16 devices, 12 uclasses, devicetree: board MMC: Loading Environment from nowhere... OK Warning: device tree node '/config/environment' not found In: serial Out: serial Err: serial SCSI: Net: No ethernet found. Hit any key to stop autoboot: 0 ## Error: "distro_bootcmd" not defined => host bind 0 test_notclean.bin => ext4write host 0 0x0 /test.txt 1 Journal Scan Completed Recovery required Journal Recovery Completed file found, deleting update journal finished File System is consistent update journal finished 1 bytes written in 8 ms (0 Bytes/s) =>
[8] Similar to before, the quick fsck corrects the number of free inodes: ~> cp test_notclean.bin test_check.bin ~> e2fsck test_check.bin e2fsck 1.45.5 (07-Jan-2020) Setting free inodes count to 2036 (was 2037) test_check.bin: clean, 12/2048 files, 1100/2048 blocks
[9] But now the -f fsck finds some more issues: ~> e2fsck -fy test_check.bin e2fsck 1.45.5 (07-Jan-2020) Pass 1: Checking inodes, blocks, and sizes Pass 2: Checking directory structure Pass 3: Checking directory connectivity Pass 4: Checking reference counts Pass 5: Checking group summary information Block bitmap differences: -1098 Fix? yes
Free blocks count wrong for group #0 (948, counted=949). Fix? yes
Free blocks count wrong (948, counted=949). Fix? yes
test_check.bin: ***** FILE SYSTEM WAS MODIFIED ***** test_check.bin: 12/2048 files (8.3% non-contiguous), 1099/2048 blocks --------
Note: If I don't execute step [7] and mount the prepared partition in linux, write to it and unmount it, fsck shows no errors afterwards.
I hope this information can help to reproduce the problem and fix it. As a short-term workaround we are doing a -f fsck on every boot in our application.
Regards, Christian Theiss