Fix migration issue when the destination is loaded

[Dor Laor <dlaor@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 0 bytes --]



[-- Attachment #2: 0001-Fix-migration-issue-when-the-destination-is-loaded.patch --]
[-- Type: text/plain, Size: 1702 bytes --]

>From abbe3b57af6a28bb81e5fb8b09b10802a8ccc3fe Mon Sep 17 00:00:00 2001
From: Dor Laor <dor@redhat.com>
Date: Wed, 15 Jul 2009 17:53:16 +0300
Subject: [PATCH] Fix migration issue when the destination is loaded

If the migration socket is full, we get EAGAIN for the write.
The set_fd_handler2 defers the write for later on. The function
tries to wake up the iothread by qemu_kvm_notify_work.
Since this happens in a loop, multiple times, the pipe that emulates eventfd
becomes full and we get a deadlock.

It is solved by checking for write-readiness using select.
Note that I check for select only for full 8 byte write and not
for partial writes. This is because we'll break the reader otherwise.

Signed-off-by: Dor Laor <dor@redhat.com>
---
 qemu-kvm.c |   16 ++++++++++++++++
 1 files changed, 16 insertions(+), 0 deletions(-)

diff --git a/qemu-kvm.c b/qemu-kvm.c
index ed7e466..0ea67a7 100644
--- a/qemu-kvm.c
+++ b/qemu-kvm.c
@@ -2094,12 +2094,28 @@ void qemu_kvm_notify_work(void)
     uint64_t value = 1;
     char buffer[8];
     size_t offset = 0;
+    fd_set wfds;
+    struct timeval tv;
+    int retval;
 
     if (io_thread_fd == -1)
 	return;
 
     memcpy(buffer, &value, sizeof(value));
 
+    FD_ZERO(&wfds);
+    FD_SET(io_thread_fd, &wfds);
+    tv.tv_sec = tv.tv_usec = 0;
+    retval = select(io_thread_fd + 1, NULL, &wfds, NULL, &tv);
+    if (retval == -1) {
+        fprintf(stderr, "failed to notify io thread due to select error\n");
+        return;
+    } else if (retval == 0)
+        /* We probably ponded this pipe too much and it is full now */
+        return;
+
+    assert(FD_ISSET(io_thread_fd, &wfds));
+
     while (offset < 8) {
 	ssize_t len;
 
-- 
1.6.2.5