On Sat, Mar 02, 2013 at 12:30:10PM -0600, Doug Goldstein wrote:
On Sat, Mar 2, 2013 at 8:30 AM, Christophe Fergeau <cfergeau@redhat.com> wrote:
diff --git a/src/qemu/qemu_conf.c b/src/qemu/qemu_conf.c index f3b09cf..3ef3499 100644 --- a/src/qemu/qemu_conf.c +++ b/src/qemu/qemu_conf.c @@ -129,8 +129,8 @@ virQEMUDriverConfigPtr virQEMUDriverConfigNew(bool privileged) if (virGetGroupID(QEMU_GROUP, &cfg->group) < 0) goto error; } else { - cfg->user = 0; - cfg->group = 0; + cfg->user = (uid_t)-1; + cfg->group = (gid_t)-1; } cfg->dynamicOwnership = privileged;
I'll agree this change should fix it from a code inspection. So ACK this hunk. Really starting to think we need some tests for this. Given the late phase in the 1.0.3 release cycle (post freeze), do you have a specific test case I can use to verify this?
My test case was running GNOME Boxes and seeing it fail ;) I've just extracted a more minimal test: <domain type='kvm' id='3'> <name>win7-test</name> <uuid>3e660fc9-fe67-950b-019f-1aab94e4abb1</uuid> <title>Microsoft Windows 7 2</title> <memory unit='KiB'>2097152</memory> <os> <type arch='x86_64' machine='pc-1.2'>hvm</type> <boot dev='cdrom'/> </os> <devices> <disk type='file' device='cdrom'> <driver name='qemu' type='raw'/> <source file='/home/teuf/isos/msdn/win7/fr_windows_7_home_premium_n_with_sp1_x64_dvd_u_676833.iso' startupPolicy='mandatory'/> <target dev='hdc' bus='ide'/> <readonly/> </disk> </devices> </domain> which fails to start without this patch, and starts correctly with it. Most of this is boilerplate (some parts are probably even optional), the important bit is the <source file='someiso.iso' startupPolicy='mandatory'/> Christophe