[libvirt] [PATCH] Fix the style of argument("cpumap") in op_pincpu()
Daniel Veillard
veillard at redhat.com
Tue May 12 14:03:57 UTC 2009
On Tue, May 12, 2009 at 10:50:02AM +0200, Daniel Veillard wrote:
> On Fri, May 08, 2009 at 06:45:59PM +0900, Tatsuro Enokura wrote:
> > > > Okay, but I'm still worrying.
> > > > The fact that we fail to detect the error NG1 is a bug, and that
> > > > bug should be fixed. Seems to me the change may just replace one error
> > > > by another one but in the end we should instead aim at fixing the NG1
> > > > problem with the old format, not substituing it with something else.
> > > >
> > > > I would prefer if the patch did some checking about the current xend
> > > > version running, but unfortunately priv->xendConfigVersion won't be
> > > > precise enough.
> > > >
> > > > Sorry I don't know how to handle this more correctly right now
> >
> > I see.
> > priv->xendConfigVersion isn't appropriate for tracing the xend's
> > version. We should request to the xen community for new interface
> > of to get the xend's version.
> >
> > On other hand,
> > the new xend and libvirt without the cpumap patch occur error(NG2).
> > Behavior of the old xend and libvirt with the cpumap patch is the same
> > behavior of libvirt without the cpumap patch.
> > Moreover, there is no work for libvirt of this issue any further at present.
> >
> > I make the patch that added the foregoing content as TODO comment.
> >
> > Signed-off-by: Tatsuro Enokura <fj2026af at aa.jp.fujitsu.com>
>
> Okay, here is my suggested patch to minimize impact on old xen setups
> maybe someone locally fixed the xend side, and we should at least try to
> not break it on old setups.
> So I updated the comment
> And activate the change only for version of xend >= 3 which won't
> cover the full range but should still protect some of the older setups,
Urgh, I wasn't fully awake, that one does compile :-\ !
Daniel
--
Daniel Veillard | libxml Gnome XML XSLT toolkit http://xmlsoft.org/
daniel at veillard.com | Rpmfind RPM search engine http://rpmfind.net/
http://veillard.com/ | virtualization library http://libvirt.org/
-------------- next part --------------
Index: src/xend_internal.c
===================================================================
RCS file: /data/cvs/libxen/src/xend_internal.c,v
retrieving revision 1.262
diff -u -r1.262 xend_internal.c
--- src/xend_internal.c 8 May 2009 09:58:46 -0000 1.262
+++ src/xend_internal.c 12 May 2009 14:00:48 -0000
@@ -3765,6 +3765,11 @@
* @maplen: length of cpumap in bytes
*
* Dynamically change the real CPUs which can be allocated to a virtual CPU.
+ * NOTE: The XenD cpu affinity map format changed from "[0,1,2]" to
+ * "0,1,2"
+ * the XenD cpu affinity works only after cset 19579.
+ * there is no fine grained xend version detection possible, so we
+ * use the old format for anything before version 3
*
* Returns 0 for success; -1 (with errno) on error
*/
@@ -3772,8 +3777,9 @@
xenDaemonDomainPinVcpu(virDomainPtr domain, unsigned int vcpu,
unsigned char *cpumap, int maplen)
{
- char buf[VIR_UUID_BUFLEN], mapstr[sizeof(cpumap_t) * 64] = "[";
+ char buf[VIR_UUID_BUFLEN], mapstr[sizeof(cpumap_t) * 64];
int i, j;
+ xenUnifiedPrivatePtr priv;
if ((domain == NULL) || (domain->conn == NULL) || (domain->name == NULL)
|| (cpumap == NULL) || (maplen < 1) || (maplen > (int)sizeof(cpumap_t))) {
@@ -3782,13 +3788,25 @@
return (-1);
}
+ priv = (xenUnifiedPrivatePtr) domain->conn->privateData;
+ if (priv->xendConfigVersion < 3) {
+ buf[0] = ']';
+ buf[1] = 0;
+ } else {
+ buf[0] = 0;
+ }
+
/* from bit map, build character string of mapped CPU numbers */
for (i = 0; i < maplen; i++) for (j = 0; j < 8; j++)
if (cpumap[i] & (1 << j)) {
snprintf(buf, sizeof(buf), "%d,", (8 * i) + j);
strcat(mapstr, buf);
}
- mapstr[strlen(mapstr) - 1] = ']';
+ if (priv->xendConfigVersion < 3)
+ mapstr[strlen(mapstr) - 1] = ']';
+ else
+ mapstr[strlen(mapstr) - 1] = 0;
+
snprintf(buf, sizeof(buf), "%d", vcpu);
return(xend_op(domain->conn, domain->name, "op", "pincpu", "vcpu", buf,
"cpumap", mapstr, NULL));
More information about the libvir-list
mailing list