From: Richard Jones <rjones@redhat.com>
Date: Thu, 8 Apr 2010 18:07:45 +0000 (+0100)
Subject: If qemu dies during launch in "null vmchannel" mode, don't hang (RHBZ#579155).
X-Git-Tag: 1.3.2~14
X-Git-Url: http://git.annexia.org/?p=libguestfs.git;a=commitdiff_plain;h=e35debd642bea240609775610b68145dab0a634b

If qemu dies during launch in "null vmchannel" mode, don't hang (RHBZ#579155).

Detect if qemu dies during launch by wait(2)-ing for it, and
then getting EOF on the qemu pipe.  This was broken in null
vmchannel mode, causing a hang.
---

diff --git a/src/guestfs.c b/src/guestfs.c
index d8c856a..fe08cb5 100644
--- a/src/guestfs.c
+++ b/src/guestfs.c
@@ -2366,8 +2366,16 @@ accept_from_daemon (guestfs_h *g)
   int sock = -1;
 
   while (sock == -1) {
+    /* If the qemu process has died, clean up the zombie (RHBZ#579155).
+     * By partially polling in the select below we ensure that this
+     * function will be called eventually.
+     */
+    waitpid (g->pid, NULL, WNOHANG);
+
     rset2 = rset;
-    int r = select (max_fd+1, &rset2, NULL, NULL, NULL);
+
+    struct timeval tv = { .tv_sec = 1, .tv_usec = 0 };
+    int r = select (max_fd+1, &rset2, NULL, NULL, &tv);
     if (r == -1) {
       if (errno == EINTR || errno == EAGAIN)
         continue;