Re: [Qemu-devel] [PATCH 02/10] qga: implement guest-pipe-open command

["Denis V. Lunev" <den@openvz.org>]

On 19/06/15 20:34, Eric Blake wrote:
> On 06/19/2015 10:57 AM, Denis V. Lunev wrote:
>> From: Olga Krishtal <okrishtal@virtuozzo.com>
>>
>> The command creates FIFO pair that can be used with existing file
>> read/write interfaces to communicate with processes spawned via the
>> forthcoming guest-file-exec interface.
>>
>> Signed-off-by: Olga Krishtal <okrishtal@virtuozzo.com>
>> Signed-off-by: Denis V. Lunev <den@openvz.org>
>> Acked-by: Roman Kagan <rkagan@virtuozzo.com>
>> CC: Eric Blake <eblake@redhat.com>
>> CC: Michael Roth <mdroth@linux.vnet.ibm.com>
>> ---
>>   qga/commands-posix.c | 96 +++++++++++++++++++++++++++++++++++++++++++++++++---
>>   qga/commands-win32.c |  8 ++++-
>>   qga/qapi-schema.json | 44 ++++++++++++++++++++++++
>>   3 files changed, 143 insertions(+), 5 deletions(-)
>>
>> +
>> +    if (pipe(fd) != 0) {
>> +        error_set_errno(errp, errno, QERR_QGA_COMMAND_FAILED, "pipe() failed");
>> +        return NULL;
>> +    }
>> +
>> +    this_end = (mode == GUEST_PIPE_MODE_WRITE);
>> +    other_end = !this_end;
>> +
>> +    qemu_set_nonblock(fd[this_end]);
>> +
>> +    qemu_set_cloexec(fd[this_end]);
>> +    qemu_set_cloexec(fd[other_end]);
> Would it be better to create a named FIFO somewhere in the file system,
> so that you can reopen the connection even if the qga daemon is
> restarted?  By using just pipe(), your fds are rather ephemeral, but if
> I understand correctly, the rest of the guest-file API tries to persist
> across qga restart.
>
this will not help IMHO

Let us assume that we have a FIFO on a filesystem
and we have a process which uses this FIFO as
stdout. If QGA will exit that process will blown up with
SIGPIPE on any write attempt.

Den