Re: [Libguestfs] [nbdkit PATCH 3/4] server: Close backends if a filter's .open fails
On 10/3/19 3:48 PM, Eric Blake wrote:
On 10/3/19 2:38 PM, Richard W.M. Jones wrote:
>
> $ ./nbdkit -s memory 1M < fuzzing/testcase_dir/newstyle-cflags
> NBDMAGICIHAVEOPTnbdkit: plugins.c:274: plugin_close: Assertion
> `connection_get_handle (conn, 0)' failed.
> Aborted (core dumped)
>
> git bisect implicates this patch:
>
> 2f80ce1209d5898cb9a567c0b29e7736ff4d03eb is the first bad commit
Yep, I see it. More patches coming (we need to better track when .open
and .prepare have succeeded, so that .finalize and .close are only
called on valid handles).
As a stop-gap while working on my larger patches, I'm pushing this.
From 679c72d651631f5711852b7395271da834265961 Mon Sep 17 00:00:00 2001
From: Eric Blake <eblake(a)redhat.com>
Date: Thu, 3 Oct 2019 15:55:00 -0500
Subject: [nbdkit PATCH] server: Allow plugin_close
With the retry filter, we've proven that a failed reopen can leave the
plugin handle unallocated. Just as filters already check for a NULL
handle (see commit fd2deeb1), the plugin must now do likewise.
Fixes: 2f80ce12
Reported-by: Richard W.M. Jones <rjones(a)redhat.com>
Signed-off-by: Eric Blake <eblake(a)redhat.com>
---
server/plugins.c | 7 +++----
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/server/plugins.c b/server/plugins.c
index 727fb0e0..120e3cc2 100644
--- a/server/plugins.c
+++ b/server/plugins.c
@@ -270,11 +270,10 @@ static void
plugin_close (struct backend *b, struct connection *conn)
{
struct backend_plugin *p = container_of (b, struct backend_plugin,
backend);
+ void *handle = connection_get_handle (conn, 0);
- assert (connection_get_handle (conn, 0));
-
- if (p->plugin.close)
- p->plugin.close (connection_get_handle (conn, 0));
+ if (handle && p->plugin.close)
+ p->plugin.close (handle);
}
static int64_t
--
2.21.0
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3226
Virtualization: qemu.org | libvirt.org