qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Paolo Bonzini <pbonzini@redhat.com>
To: Stefano Stabellini <stefano.stabellini@eu.citrix.com>,
	Lan Tianyu <tianyu.lan@intel.com>
Cc: mjt@tls.msk.ru, qemu-devel@nongnu.org, jbeulich@suse.com,
	konrad.wilk@oracle.com
Subject: Re: [Qemu-devel] [Fix PATCH] Qemu/Xen: Fix early freeing MSIX MMIO memory region
Date: Tue, 6 Oct 2015 15:33:56 +0200	[thread overview]
Message-ID: <5613CDC4.1070602@redhat.com> (raw)
In-Reply-To: <alpine.DEB.2.02.1510051713580.1179@kaball.uk.xensource.com>



On 05/10/2015 18:53, Stefano Stabellini wrote:
>> This patch is to fix the issue via moving MSIX MMIO memory region into
>> struct XenPCIPassthroughState and free it together with pt device's obj.
> 
> Given that all the MSI-X related info are in XenPTMSIX, I would prefer
> to keep the mmio memory region there, if possible.
> 
> Couldn't you just unhook msix->mmio from XenPCIPassthroughState's object
> in xen_pt_msix_delete?  Calling object_property_del_child or
> object_unparent?

This is the right thing to do, but there are two separate things to fix.

One is the use-after-free of msix->mmio, the other is that freeing 
s->msix and in general xen_pt_config_delete should be done from the 
.instance_finalize callback.  This is documented in docs/memory.txt.

This is an attempt at a patch (not even compiled):

diff --git a/hw/xen/xen_pt_msi.c b/hw/xen/xen_pt_msi.c
index e3d7194..e476bac 100644
--- a/hw/xen/xen_pt_msi.c
+++ b/hw/xen/xen_pt_msi.c
@@ -610,7 +610,7 @@ error_out:
     return rc;
 }
 
-void xen_pt_msix_delete(XenPCIPassthroughState *s)
+void xen_pt_msix_unmap(XenPCIPassthroughState *s)
 {
     XenPTMSIX *msix = s->msix;
 
@@ -627,6 +627,17 @@ void xen_pt_msix_delete(XenPCIPassthroughState *s)
     }
 
     memory_region_del_subregion(&s->bar[msix->bar_index], &msix->mmio);
+}
+
+void xen_pt_msix_delete(XenPCIPassthroughState *s)
+{
+    XenPTMSIX *msix = s->msix;
+
+    if (!msix) {
+        return;
+    }
+
+    object_unparent(&msix->mmio);
 
     g_free(s->msix);
     s->msix = NULL;

where xen_pt_config_unmap would be called from xen_pt_destroy, and the call
to xen_pt_config_delete would be moved to xen_pci_passthrough_info's
instance_finalize member.

Paolo

  reply	other threads:[~2015-10-06 13:34 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-09-30  6:51 [Qemu-devel] [Fix PATCH] Qemu/Xen: Fix early freeing MSIX MMIO memory region Lan Tianyu
2015-10-05 16:53 ` Stefano Stabellini
2015-10-06 13:33   ` Paolo Bonzini [this message]
2015-10-06 13:49     ` Stefano Stabellini
2015-10-06 15:18       ` Lan, Tianyu

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=5613CDC4.1070602@redhat.com \
    --to=pbonzini@redhat.com \
    --cc=jbeulich@suse.com \
    --cc=konrad.wilk@oracle.com \
    --cc=mjt@tls.msk.ru \
    --cc=qemu-devel@nongnu.org \
    --cc=stefano.stabellini@eu.citrix.com \
    --cc=tianyu.lan@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).