On 06/28/2017 01:00 AM, John Ferlan wrote: > > > On 06/22/2017 12:18 PM, Michal Privoznik wrote: >> Currently, the only type of chardev that we create the backend >> for in the namespace is type='dev'. This is not enough, other >> backends might have files under /dev too. For instance channels >> might have a unix socket under /dev (well, bind mounted under >> /dev from a different place). >> >> Signed-off-by: Michal Privoznik <mpriv...@redhat.com> >> --- >> src/qemu/qemu_domain.c | 25 +++++++++++++++++-------- >> 1 file changed, 17 insertions(+), 8 deletions(-) >> >> diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c >> index 51779c535..65eec26dd 100644 >> --- a/src/qemu/qemu_domain.c >> +++ b/src/qemu/qemu_domain.c >> @@ -7732,7 +7732,7 @@ qemuDomainCreateDeviceRecursive(const char *device, >> >> isLink = S_ISLNK(sb.st_mode); >> isDev = S_ISCHR(sb.st_mode) || S_ISBLK(sb.st_mode); >> - isReg = S_ISREG(sb.st_mode); >> + isReg = S_ISREG(sb.st_mode) || S_ISFIFO(sb.st_mode) || >> S_ISSOCK(sb.st_mode); >> >> /* Here, @device might be whatever path in the system. We >> * should create the path in the namespace iff it's "/dev" >> @@ -8129,11 +8129,17 @@ qemuDomainSetupChardev(virDomainDefPtr def >> ATTRIBUTE_UNUSED, >> void *opaque) >> { >> const struct qemuDomainCreateDeviceData *data = opaque; >> + const char *path = NULL; >> >> - if (dev->source->type != VIR_DOMAIN_CHR_TYPE_DEV) >> + if (!(path = virDomainChrSourceDefPath(dev->source))) >> return 0; >> >> - return qemuDomainCreateDevice(dev->source->data.file.path, data, false); >> + /* Socket created by qemu. It doesn't exist upfront. */ >> + if (dev->source->type == VIR_DOMAIN_CHR_TYPE_UNIX && >> + dev->source->data.nix.listen) >> + return 0; >> + >> + return qemuDomainCreateDevice(path, data, true); >> } >> >> >> @@ -8531,7 +8537,7 @@ qemuDomainAttachDeviceMknodHelper(pid_t pid >> ATTRIBUTE_UNUSED, >> bool delDevice = false; >> bool isLink = S_ISLNK(data->sb.st_mode); >> bool isDev = S_ISCHR(data->sb.st_mode) || S_ISBLK(data->sb.st_mode); >> - bool isReg = S_ISREG(data->sb.st_mode); >> + bool isReg = S_ISREG(data->sb.st_mode) || S_ISFIFO(data->sb.st_mode) || >> S_ISSOCK(data->sb.st_mode); >> >> qemuSecurityPostFork(data->driver->securityManager); >> >> @@ -8576,7 +8582,7 @@ qemuDomainAttachDeviceMknodHelper(pid_t pid >> ATTRIBUTE_UNUSED, >> * as its backing chain. This might however clash here. >> * Therefore do the cleanup here. */ >> if (umount(data->file) < 0 && >> - errno != ENOENT) { >> + errno != ENOENT && errno != EINVAL) { > > Is this something that should be separate or as a result of the current > changes? Should it go with the previous patch since that's where it's > introduced?
It's a result of this change. umout() is not consistent in returned errno. Depending on type of file that we wanted to umount we might get either ENOENT or EINVAL. > > Reviewed-by: John Ferlan <jfer...@redhat.com> Thanks. Michal -- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list