[PATCH] fs-backend: fix gnttab unmap

[PATCH] fs-backend: fix gnttab unmap

[Stefano Stabellini]

Hi all,
when a stubdom is destroyed, fs-backend does not correctly unmap all the
pages mapped for the shared ring.
This causes some pages not to be properly freed by xen, hence after a
cycle of 50 save\restore we don't have enough spare memory to start 2
VMs any more.
This patch fixes the issue properly unmapping all the pages needed by
the shared ring.

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>

---

diff -r 633e08774dab tools/fs-back/fs-backend.c
--- a/tools/fs-back/fs-backend.c	Thu Mar 19 14:01:56 2009 +0000
+++ b/tools/fs-back/fs-backend.c	Thu Mar 19 16:33:34 2009 +0000
@@ -161,7 +161,7 @@
     while (!xenbus_frontend_state_changed(mount, STATE_CLOSING));
     xenbus_write_backend_state(mount, STATE_CLOSED);
 
-    xc_gnttab_munmap(mount->gnth, mount->ring.sring, 1);
+    xc_gnttab_munmap(mount->gnth, mount->ring.sring, mount->shared_ring_size);
     xc_gnttab_close(mount->gnth);
     xc_evtchn_unbind(mount->evth, mount->local_evtchn);
     xc_evtchn_close(mount->evth);

Re: [PATCH] fs-backend: fix gnttab unmap

[Keith Coleman]

On Thu, Mar 19, 2009 at 12:58 PM, Stefano Stabellini
<stefano.stabellini@eu.citrix.com> wrote:
> Hi all,
> when a stubdom is destroyed, fs-backend does not correctly unmap all the
> pages mapped for the shared ring.
> This causes some pages not to be properly freed by xen, hence after a
> cycle of 50 save\restore we don't have enough spare memory to start 2
> VMs any more.
> This patch fixes the issue properly unmapping all the pages needed by
> the shared ring.
>

I can confirm that fs-backend is more stable. I have performed
hundreds of save/restore cycles without issue.


Keith Coleman

Re: [PATCH] fs-backend: fix gnttab unmap

[Stefano Stabellini]

Keith Coleman wrote:

> On Thu, Mar 19, 2009 at 12:58 PM, Stefano Stabellini
> <stefano.stabellini@eu.citrix.com> wrote:
>> Hi all,
>> when a stubdom is destroyed, fs-backend does not correctly unmap all the
>> pages mapped for the shared ring.
>> This causes some pages not to be properly freed by xen, hence after a
>> cycle of 50 save\restore we don't have enough spare memory to start 2
>> VMs any more.
>> This patch fixes the issue properly unmapping all the pages needed by
>> the shared ring.
>>
> 
> I can confirm that fs-backend is more stable. I have performed
> hundreds of save/restore cycles without issue.
> 

good, thanks again for testing.