Re: [libvirt] [PATCH 00/16] snapshot refactoring (incremental backup saga)

On Fri, 2019-03-22 at 06:50 -0400, John Ferlan wrote: > On 3/22/19 4:46 AM, Andrea Bolognani wrote: >> This causes libvirtd to crash at startup on my machine. >> >> Have a backtrace: >> >> Thread 19 (Thread 0x7fffaa4e0700 (LWP 31651)): >> #0 0x00007ffff72ecd31 in open64 () from /lib64/libc.so.6 >> #1 0x00007ffff727d3f6 in _IO_file_open () from /lib64/libc.so.6 >> #2 0x00007ffff727d5ad in __GI__IO_file_fopen () from /lib64/libc.so.6 >> #3 0x00007ffff727132d in __fopen_internal () from /lib64/libc.so.6 >> #4 0x00007ffff71199d8 in ?? () from /lib64/libudev.so.1 >> #5 0x00007ffff71146dd in ?? () from /lib64/libudev.so.1 >> #6 0x00007ffff71173ed in ?? () from /lib64/libudev.so.1 >> #7 0x00007ffff71179d9 in ?? () from /lib64/libudev.so.1 >> #8 0x00007ffff710be77 in udev_device_get_property_value () from /lib64/libudev.so.1 >> #9 0x00007fffc2e8a3ef in udevGetDeviceProperty (udev_device=udev_device@entry=0x7fff900606e0, property_key=property_key@entry=0x7fffc2ea3504 "DRIVER") at node_device/node_device_udev.c:140 >> #10 0x00007fffc2e8a459 in udevGetStringProperty (udev_device=udev_device@entry=0x7fff900606e0, property_key=property_key@entry=0x7fffc2ea3504 "DRIVER", value=0x7fff900f98d8) at node_device/node_device_udev.c:154 >> #11 0x00007fffc2e8b3ad in udevAddOneDevice (device=device@entry=0x7fff900606e0) at node_device/node_device_udev.c:1369 >> #12 0x00007fffc2e8d27e in udevProcessDeviceListEntry (list_entry=0x7fff9004f350, udev=0x7fff98120f00) at node_device/node_device_udev.c:1435 >> #13 udevEnumerateDevices (udev=0x7fff98120f00) at node_device/node_device_udev.c:1489 >> #14 nodeStateInitializeEnumerate (opaque=0x7fff98120f00) at node_device/node_device_udev.c:1798 >> #15 0x00007ffff7c5a2c2 in virThreadHelper (data=<optimized out>) at util/virthread.c:206 >> #16 0x00007ffff740d58e in start_thread () from /lib64/libpthread.so.0 >> #17 0x00007ffff72fc6a3 in clone () from /lib64/libc.so.6 > > Very strange - I don't recall the patches touching any of the > virnodedeviceobj* code. Is it possible to bisect to something after the > below? > > Although there were changes by Nikolay right at the top of tree that > did... Can you go back to top, revert Nikolay's changes and see if that > does it? You will need Michal's change to virDomainMomentAssignDef. Michal's patch seems to do the trick: I built from 5e752513d802 and the daemon starts fine even when I have guests with snapshots.