danluu / fs-errors Goto Github PK

We currently call it with sudo so that it has permissions to mount and unmount, but this means that (for example) the results file ends up being owned by root.

I'm working on adding ZFS support.

The initial roadblock I've run into is that ZFS does not support zpools smaller than 64M. Should I just make a 64M pool? Would it affect the results if the pools are different sizes?

I got the following output:

images/ext3.img.gz,./mmap_read,error,overlay,-7,,
images/ext3.img.gz,./mmap_write,error,overlay,-7,,
images/ext3.img.gz,./pread,error,overlay,255,read fail. errno: Input/output error,
images/ext3.img.gz,./pwrite,error,overlay,255,open fail Input/output error,
images/ext4.img.gz,./mmap_read,error,overlay,-7,,
images/ext4.img.gz,./mmap_write,error,overlay,-7,,
images/ext4.img.gz,./pread,error,overlay,255,read fail. errno: Input/output error,
images/ext4.img.gz,./pwrite,error,overlay,255,open fail Input/output error,
images/exfat.img.gz,./mmap_read,error,overlay,-7,,
images/exfat.img.gz,./mmap_write,error,overlay,-7,,
images/exfat.img.gz,./pread,error,overlay,255,read fail. errno: Transport endpoint is not connected,
images/exfat.img.gz,./pwrite,error,overlay,255,open fail Transport endpoint is not connected,
images/jfs.img.gz,./mmap_read,error,overlay,-7,,
images/jfs.img.gz,./mmap_write,error,overlay,-7,,
images/jfs.img.gz,./pread,error,overlay,255,read fail. errno: Input/output error,
images/jfs.img.gz,./pwrite,error,overlay,255,open fail Input/output error,
images/xfs.img.gz,./mmap_read,error,overlay,-7,,
images/xfs.img.gz,./mmap_write,error,overlay,-7,,
images/xfs.img.gz,./pread,error,overlay,255,read fail. errno: Input/output error,
images/xfs.img.gz,./pwrite,error,overlay,255,open fail Input/output error,
images/ext3-largefile.img.gz,./mmap_read,error,overlay,-7,,
images/ext3-largefile.img.gz,./mmap_write,error,overlay,-7,,
images/ext3-largefile.img.gz,./pread,error,overlay,255,read fail. errno: Input/output error,
images/ext3-largefile.img.gz,./pwrite,error,overlay,255,open fail Input/output error,
images/ext4-largefile.img.gz,./mmap_read,error,overlay,-7,,
images/ext4-largefile.img.gz,./mmap_write,error,overlay,-7,,
images/ext4-largefile.img.gz,./pread,error,overlay,255,read fail. errno: Input/output error,
images/ext4-largefile.img.gz,./pwrite,error,overlay,255,open fail Input/output error,
images/jfs-largefile.img.gz,./mmap_read,error,overlay,-7,,
images/jfs-largefile.img.gz,./mmap_write,error,overlay,-7,,
images/jfs-largefile.img.gz,./pread,error,overlay,255,read fail. errno: Input/output error,
images/jfs-largefile.img.gz,./pwrite,error,overlay,255,open fail Input/output error,

Would you be able to explain what these errors mean and whether it means there are errors in the filesystem? I couldn't figure out to interpret these based on the README. Thank you!

$ btrfs fi df /mnt/test/
Data, single: total=8.00MiB, used=0.00B
System, DUP: total=8.00MiB, used=16.00KiB
Metadata, DUP: total=32.00MiB, used=112.00KiB
GlobalReserve, single: total=16.00MiB, used=0.00B

The DUP profile here may be the image was created on @WesleyAC 's machine? It's not clear that our read corruption test is really valid? OTOH, if we had (for example) a flash read disturb error, having duplicated metadata would give us the correct result.

We currently only have one directory, which means that we dump things into the source directory when we build.

This is generally bad and will get worse once we start generating data and have data, executables, and source in the same directory.

Steps to (maybe) manually get error:

Where the file test is:

0 2389 linear /dev/loop0 0
2389 1 error
2390 5422 linear /dev/loop0 0

losetup /dev/loop0 ext4.img
cat test | sudo dmsetup create test
sudo mkdir /mnt/test
sudo mount /dev/mapper/test /mnt/test
cat /mnt/test/test.txt

pread has changes that aren't in pwrite. They should really just be merged into a single program.