On Wed, May 10, 2017 at 04:31:58PM +0200, Paolo Bonzini wrote: > Since the last patch in v1 didn't work, I bit the bullet and converted > the whole thing to coroutines (patches 4-6). This in turns allows a more > elegant solution to wait for CURLStates to get free (patch 7). > > I tested this by lowering CURL_NUM_STATES to 2. With this change, the > buggy case triggers a couple times while booting a Fedora netinst image.
This series fixes the original bug, so: Tested-by: Richard W.M. Jones <rjo...@redhat.com> I think the Reported-by in patch 3 should credit Kun Wei for finding the bug, and we should probably mention the BZ too: Reported-by: Kun Wei <ku...@redhat.com> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1447590 A nit pick perhaps but in patch 5 you say "This was broken before for disks > 2TB, but now it would break at 4GB.". I understand after reading it a few times that you mean it would be broken at 4GB, if you hadn't changed size_t -> uint64_t (on 32 bit platforms). Perhaps better to clarify that sentence. --- I also ran some performance and stability testing. I used virt-ls for this. The following command will iterate over every file in a remote guest image and print an md5sum: LIBGUESTFS_BACKEND=direct \ LIBGUESTFS_HV=~/d/qemu/x86_64-softmmu/qemu-system-x86_64 \ virt-ls -a http://somehost/rhel-guest-image-7.1-20150224.0.x86_64.qcow2 \ -lR --checksum / I timed this with and without your patches, but there was no significant difference (but note that virt-ls is a fundamentally sequential program). It didn't crash or hang at any time during my testing. Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming and virtualization blog: http://rwmj.wordpress.com virt-p2v converts physical machines to virtual machines. Boot with a live CD or over the network (PXE) and turn machines into KVM guests. http://libguestfs.org/virt-v2v