Re: [Xen-devel] [Xen-users] substantial shutdown delay for PV guests with PCI -passthrough

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

Atom2 writes ("Re: [Xen-devel] [Xen-users] substantial shutdown delay for PV 
guests with PCI -passthrough"):
> This file again contains a few annotations with regards to where the 
> delays happen. To my untrained eye it looks largely identical to the 
> last xl-output with the obvious change of domain-id, addresses, 
> line-numbers for debug output where changes in the sourve have happende 
> and the use of the new function libxl__wait_for_backend_deprecated 
> instead of libxl__wait_for_backend due to your patch, the latter of 
> which I take as proof that your patches have been applied.

Thanks.  I'm puzzled now.  I have another tiny patch to suggest,
which won't fix the problem but will produce more debugging output.

Can you run it again with this, on top of the previous patch, please ?

Thanks,
Ian.

diff --git a/tools/libxl/libxl_device.c b/tools/libxl/libxl_device.c
index 779b38b..eff452b 100644
--- a/tools/libxl/libxl_device.c
+++ b/tools/libxl/libxl_device.c
@@ -1131,7 +1131,13 @@ int libxl__wait_for_backend_deprecated(libxl__gc *gc, 
const char *be_path,
             watchdog--;
         }
     }
-    LIBXL__LOG(ctx, LIBXL__LOG_ERROR, "Backend %s not ready", be_path);
+    LOG(ERROR, "Backend %s not ready (state %s)", be_path, p);
+{
+const char *fe = libxl__xs_read(gc,0, GCSPRINTF("%s/frontend", be_path));
+const char *fe_state = !fe ? 0 : libxl__xs_read(gc,0, GCSPRINTF("%s/state", 
fe));
+LOG(ERROR, "FE %s state %s", fe, fe_state);
+}
+
 out:
     return rc;
 }



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel