[libvirt] [PATCH] Don't fail qemu driver intialization if we can't determine hugepage size

[Guido Günther]

On Mon, Aug 11, 2014 at 11:17:10AM +0200, Michal Privoznik wrote:
> On 10.08.2014 13:51, Guido Günther wrote:
> >Otherwise we fail like
> >
> >   libvirt version: 1.2.7, package: 6 (root 2014-08-08-16:09:22 bogon)
> >   virAuditOpen:62 : Unable to initialize audit layer: Protocol not supported
> >   virFileGetDefaultHugepageSize:2958 : internal error: Unable to parse /proc/meminfo
> >   virStateInitialize:749 : Initialization of QEMU state driver failed: internal error: Unable to parse /proc/meminfo
> >   daemonRunStateInit:922 : Driver state initialization failed
> >
> >if the data can't be determined.
> >
> >Reference: http://bugs.debian.org/757609
> >---
> >  src/util/virfile.c | 5 +++--
> >  1 file changed, 3 insertions(+), 2 deletions(-)
> >
> >diff --git a/src/util/virfile.c b/src/util/virfile.c
> >index f9efc65..b6f5e3f 100644
> >--- a/src/util/virfile.c
> >+++ b/src/util/virfile.c
> >@@ -2953,8 +2953,9 @@ virFileGetDefaultHugepageSize(unsigned long long *size)
> >          goto cleanup;
> >
> >      if (!(c = strstr(meminfo, HUGEPAGESIZE_STR))) {
> >-        virReportError(VIR_ERR_INTERNAL_ERROR,
> >-                       _("Unable to parse %s"),
> >+        virReportError(VIR_ERR_NO_SUPPORT,
> >+                       _("%s not found in %s"),
> >+                       HUGEPAGESIZE_STR,
> >                         PROC_MEMINFO);
> >          goto cleanup;
> >      }
> >
> 
> This merely changes the error code and error message. But the initialization
> will fail anyway. Well, it would up till d26e81083. But the error message
> you suggests is more verbose and describes the origin of fault more
> accurately.

Initialization doesn't fail anymore with this since we have:

    if (privileged &&
        virFileFindHugeTLBFS(&cfg->hugetlbfs, &cfg->nhugetlbfs) < 0) {
        /* This however is not implemented on all platforms. */
        virErrorPtr err = virGetLastError();
        if (err && err->code != VIR_ERR_NO_SUPPORT)
            goto error;
    }

in virQEMUDriverConfigNew so we only warn in this case. In case
someone has hugetlbfs_mount in qemu.conf we still fail hard which
is IMHO correct in this case.

> ACK

Thanks. I'll wait a bit more before pushding in case this isn't the
desired behviour!  
Cheers,
   -- Guido


> 
> Michal
>