Re: -EINTR return in domain_relinquish_resources

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Thu, Jan 22, 2015 at 10:00:35AM +0000, Jan Beulich wrote:
> >>> On 21.01.15 at 22:27, <konrad.wilk@oracle.com> wrote:
> > As I was looking at some of the XSA I realized that the
> > call-chain of:
> > 
> >  domain_relinquish_resources
> >    ->vcpu_destroy_pagetables
> >      -> put_page_and_type_preemptible
> >         -> __put_page_type
> > 		returns -EINTR
> > [...]
> >  b). Or should the hypervisor convert the -EINTR to -ERESTART
> >      (or -EAGAIN) - which most of the code (see users of
> >      get_page_type_preemptible) do right now?
> 
> This one, in vcpu_destroy_pagetables(). I'm afraid I overlooked
> this when adding the preemption capability here.

OK. Would this RFC patch be OK? (I can send it off as normal - just
want to make sure you are OK with this being put there)

>From 1558a9870f438df949a8ad09a27825bd35a9f4ea Mon Sep 17 00:00:00 2001
From: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Date: Thu, 22 Jan 2015 11:34:21 -0500
Subject: [PATCH] domain: In vcpu_destroy_pagetables we can return -EINTR
 instead of -EAGAIN

which has the side effect that domain_relinquish_resources will stop
and return to user-space -EINTR - which it is not equipped to deal with.
The preemption mechanism we have to domain destruction is to return
-EAGAIN and as such we need to catch the case of:

domain_relinquish_resources
  ->vcpu_destroy_pagetables
    -> put_page_and_type_preemptible
       -> __put_page_type
           returns -EINTR

and convert it to the proper type.

Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
---
 xen/arch/x86/mm.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/xen/arch/x86/mm.c b/xen/arch/x86/mm.c
index 6e9c2c0..3ac2ae5 100644
--- a/xen/arch/x86/mm.c
+++ b/xen/arch/x86/mm.c
@@ -2677,6 +2677,9 @@ int vcpu_destroy_pagetables(struct vcpu *v)
 
     v->arch.cr3 = 0;
 
+    if ( rc == -EINTR )
+        rc = -EAGAIN;
+
     return rc;
 }
 
-- 
2.1.0