files: 0.957
device: 0.938
user-level: 0.909
performance: 0.895
graphic: 0.829
semantic: 0.799
VMM: 0.761
PID: 0.750
architecture: 0.745
x86: 0.718
risc-v: 0.715
vnc: 0.708
boot: 0.704
ppc: 0.698
virtual: 0.689
register: 0.680
network: 0.677
i386: 0.667
arm: 0.657
socket: 0.650
kernel: 0.648
KVM: 0.641
hypervisor: 0.627
TCG: 0.590
mistranslation: 0.572
permissions: 0.564
debug: 0.514
peripherals: 0.467
assembly: 0.432

qemu-img convert is freezing for some DMG files.

Recently, I created a file using hdiutil from MacOS (using Zlib compression):

$ hdiutil create -volname MyVolName -srcfolder /path/to/my/vol/ -ov -format UDZO myvolname.dmg

But, when I try to convert this volume using qemu-img convert, this command is freezing.

I'm using the upstream version to test it.

It is freezing inside the binary search method to retrieve the chunk.

But, I still don't know why.

I'm attaching the file as an example.

It can be mounted using MacOS or other Linux apps like hfsleuth and darling-dmg.



I have submitted a patch to prevent hanging in binary search.
See: http://lists.nongnu.org/archive/html/qemu-devel/2018-12/msg05453.html?fbclid=IwAR0ObsaZ4kVMVv6MWIdq0ZCAN5tGhDsd9GmFv8_v7HBTl94Cu5EkRZ3z4n4

Thanks,
Yu-Chen Lin

I'm almost sure that we need other improvements beyond fixing binary search deadlock.

It is returning read errors when you apply your patch.

Because of lacking zero chunk table, reading zero sector will return EIO.
I have submitted a series to fix this problem.

Please refer to this series: http://lists.nongnu.org/archive/html/qemu-devel/2018-12/msg05637.html

Thanks,
Yu-Chen Lin

Hi Yu-Chen.

The series looks good. It fixed the problems related to zero chunk. Nice work!

I added a comment that I reviewed your patch considering my scenarios here. At least, It contributes to merge your changes.

Additional info: tested with version 3.0.0. So, before pushing my commits. Conclusion: This is not a new bug.

Since QEMU 4.0 has been released now, I assume we can close this ticket as "Fix released"? Or is there still anything left to do here?

I re-test the dmg img with QEMU 4.0 again, and it works.

In my opinion, the bug can be closed as "Fix released".

Thanks.



Thanks for testing. So I'm marking this now as "Fix released".