[libvirt] [PATCH] qemu: fix monitor socket reconnection

[Cole Robinson]

On 07/29/2009 08:35 AM, Daniel P. Berrange wrote:
> On Wed, Jul 29, 2009 at 12:09:44AM +0900, Ryota Ozaki wrote:
>> Hi,
>>
>> This patch fixes a handling of a qemu monitor unix socket. Current code
>> addresses 'Socket may not have shown up yet' case with catching EACCES
>> of connect(2), however, in that case, an errno which should be catched
>> is ENOENT, not EACCES.
>>
>> Thanks,
>>   ozaki-r
>>
>>
>> >From 5a4b7ee2296ca6756127c26c6bfbbe1f49ab8722 Mon Sep 17 00:00:00 2001
>> From: Ryota Ozaki <ozaki.ryota at gmail.com>
>> Date: Tue, 28 Jul 2009 23:48:22 +0900
>> Subject: [PATCH] qemu: fix monitor socket reconnection
>>
>> ---
>>  src/qemu_driver.c |    4 ++--
>>  1 files changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/src/qemu_driver.c b/src/qemu_driver.c
>> index 52a77be..9fb8506 100644
>> --- a/src/qemu_driver.c
>> +++ b/src/qemu_driver.c
>> @@ -917,8 +917,8 @@ qemudOpenMonitorUnix(virConnectPtr conn,
>>          if (ret == 0)
>>              break;
>>
>> -        if (errno == EACCES || errno == ECONNREFUSED) {
>> -            /* EACCES       : Socket may not have shown up yet
>> +        if (errno == ENOENT || errno == ECONNREFUSED) {
>> +            /* ENOENT       : Socket may not have shown up yet
>>               * ECONNREFUSED : Leftover socket hasn't been removed yet */
>>              continue;
>>          }
> 
> 
> ACK, that existing code is clearly bogus
> 

Hmm, I thoroughly tested this for both of those issues listed in the
comments, and the original code seemed to do the job. But it certainly
appears to be bogus.

I'll try to reproduce the original issue and make sure this code works
as expected.

- Cole