Summary: | IO error on machine shutdown if connected through unix fd socket | ||
---|---|---|---|
Product: | Spice | Reporter: | Zeeshan Ali <zeenix> |
Component: | spice-gtk | Assignee: | Spice Bug List <spice-bugs> |
Status: | RESOLVED FIXED | QA Contact: | |
Severity: | normal | ||
Priority: | medium | ||
Version: | unspecified | ||
Hardware: | Other | ||
OS: | All | ||
Whiteboard: | |||
i915 platform: | i915 features: |
Description
Zeeshan Ali
2015-04-13 15:41:40 UTC
Not spice-gtk fault, On regular direct spice server socket, we get clean shutdown with poll G_IO_IN and read=0. with a libvirt provided fd, it gets G_IO_IN | G_IO_HUP. Further investigation required on libvirt side. libvirt used a socketpair, and passes them to qemu/spice using spice_server_add_client it looks like spice server could use more shutdown() calls here and there. (In reply to Marc-Andre Lureau from comment #3) > it looks like spice server could use more shutdown() calls here and there. Is that a side-comment or would that help with this bug? After further research, I think it's expected to get HUP, it doesn't get it with tcp sockets, only local sockets (that's annoying..) It happens on regular close() conditions, so I would just ignore it for now. A clean shutdown would involve a new message to tell the client to first close its end, in order to avoid the HUP. That way the client could distinguish normal termination from unexpected ones. http://lists.freedesktop.org/archives/spice-devel/2015-April/019461.html |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.