From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([208.118.235.92]:48959) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1Tss6w-0006Y1-B5 for qemu-devel@nongnu.org; Wed, 09 Jan 2013 04:38:16 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1Tss6u-0000F0-Ti for qemu-devel@nongnu.org; Wed, 09 Jan 2013 04:38:14 -0500 Received: from mx1.redhat.com ([209.132.183.28]:15450) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1Tss6u-0000Ev-LT for qemu-devel@nongnu.org; Wed, 09 Jan 2013 04:38:12 -0500 Message-ID: <50ED3A76.6030503@redhat.com> Date: Wed, 09 Jan 2013 10:37:58 +0100 From: Paolo Bonzini MIME-Version: 1.0 References: <1357659056-2027-1-git-send-email-chouteau@adacore.com> In-Reply-To: <1357659056-2027-1-git-send-email-chouteau@adacore.com> Content-Type: text/plain; charset=ISO-8859-15 Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH] Check return values from g_poll and select List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Fabien Chouteau Cc: blauwirbel@gmail.com, sw@weilnetz.de, aliguori@us.ibm.com, qemu-devel@nongnu.org, mdroth@linux.vnet.ibm.com Il 08/01/2013 16:30, Fabien Chouteau ha scritto: > The current implementation of os_host_main_loop_wait() on Windows, > returns 1 only when a g_poll() event occurs because the return value of > select() is overridden. This is wrong as we may skip a socket event, as > shown in this example: > > 1. select() returns 0 > 2. g_poll() returns 1 (socket event occurs) > 3. os_host_main_loop_wait() returns 1 > 4. qemu_iohandler_poll() sees no socket event because select() has > return before the event occurs > 5. select() returns 1 > 6. g_poll() returns 0 (g_poll overrides select's return value) > 7. os_host_main_loop_wait() returns 0 > 8. qemu_iohandler_poll() doesn't check for socket events because the > return value of os_host_main_loop_wait() is zero. > 9. goto 5 > > This patch use one variable for each of these return values, so we don't > miss a select() event anymore. > > Also move the call to select() after g_poll(), this will improve latency > as we don't have to go through two os_host_main_loop_wait() calls to > detect a socket event. > > Signed-off-by: Fabien Chouteau Reviewed-by: Paolo Bonzini Paolo > --- > main-loop.c | 27 +++++++++++++-------------- > 1 file changed, 13 insertions(+), 14 deletions(-) > > diff --git a/main-loop.c b/main-loop.c > index 54f38ae..6f52ac3 100644 > --- a/main-loop.c > +++ b/main-loop.c > @@ -330,7 +330,7 @@ void qemu_fd_register(int fd) > static int os_host_main_loop_wait(uint32_t timeout) > { > GMainContext *context = g_main_context_default(); > - int ret, i; > + int select_ret, g_poll_ret, ret, i; > PollingEntry *pe; > WaitObjects *w = &wait_objects; > gint poll_timeout; > @@ -345,13 +345,6 @@ static int os_host_main_loop_wait(uint32_t timeout) > return ret; > } > > - if (nfds >= 0) { > - ret = select(nfds + 1, &rfds, &wfds, &xfds, &tv0); > - if (ret != 0) { > - timeout = 0; > - } > - } > - > g_main_context_prepare(context, &max_priority); > n_poll_fds = g_main_context_query(context, max_priority, &poll_timeout, > poll_fds, ARRAY_SIZE(poll_fds)); > @@ -367,9 +360,9 @@ static int os_host_main_loop_wait(uint32_t timeout) > } > > qemu_mutex_unlock_iothread(); > - ret = g_poll(poll_fds, n_poll_fds + w->num, poll_timeout); > + g_poll_ret = g_poll(poll_fds, n_poll_fds + w->num, poll_timeout); > qemu_mutex_lock_iothread(); > - if (ret > 0) { > + if (g_poll_ret > 0) { > for (i = 0; i < w->num; i++) { > w->revents[i] = poll_fds[n_poll_fds + i].revents; > } > @@ -384,12 +377,18 @@ static int os_host_main_loop_wait(uint32_t timeout) > g_main_context_dispatch(context); > } > > - /* If an edge-triggered socket event occurred, select will return a > - * positive result on the next iteration. We do not need to do anything > - * here. > + /* Call select after g_poll to avoid a useless iteration and therefore > + * improve socket latency. > */ > > - return ret; > + if (nfds >= 0) { > + select_ret = select(nfds + 1, &rfds, &wfds, &xfds, &tv0); > + if (select_ret != 0) { > + timeout = 0; > + } > + } > + > + return select_ret || g_poll_ret; > } > #endif > >