08.05.2013 17:30, Lisandro Damián Nicanor Pérez Meyer wrote: > Package: qemu-user-static > Version: 1.1.2+dfsg-6a > Severity: important > > Hi! I'm using qemu-mips-static to build some packages in a chroot. > Up to late last year I could do this without problems, using > a 64 bits wheezy host and a mips chroot. > > After updating the chroot to stable wheezy, I've found that running > Doxygen with dot graphs enabled will make qemu-mips-static hang the CPU. > > After further digging, I've found that using DOT_NUM_THREADS = 0 in > Doxygen's setup will use n+1 threads, where n is the number of available > processors (which, afaiu, it's the normal behaviour of doxygen). > > If I set this value to match the number of processors, everything is OK. > > I'm setting this bug as important as I can't even kill -9 the process and > have to reboot the host in order to get my CPU back.
So, this bug has been sitting with "moreinfo" tag on it for a very long time. Reehashing it again now. >From what I see here, there are 2 bugs actually. One is in qemu, qemu-linux-user does not support running multi-threaded applications. The fact some app worked in one version of qemu means nothing - it worked just by a pure luck or chance. Multi-threaded apps are not emulated properly, one qemu thread will step into another thread with all evil things happening. This is #769983 and is a well-known problem. Some work to address this is being planned upstream, but it is only plans, not even preliminary design (no code exists ofcourse). So from one side, this bugreport can be merged with #769983. And another, apparently, is in host kernel. qemu-linux-user is a regular userspace process. If it is able to make host kernel to stuck like this, requiring a reboot to recover, any other userspace process can do that and it is a very serious problem in kernel, and it has nothing to do with qemu at all (qemu is just a messenger in this case.) However, it looks like you're the only one user who is seeing this problem. So, basically, I don't see any value in keeping this bugreport open. One part of it is #769983, and another part does not belong to qemu, it is a kernel issue. If you want to keep the second part of it (host CPU hang), you can reassign it to kernel, providing necessary version information, and verifying if the prob still exists with current to-be-jessie kernel. Or I will just close it, since it makes no sense to keep it this way further. I'm sorry for being non-constructive here, but that's the best offer I can made out of this. Thanks, /mjt -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
