From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([208.118.235.92]:45254) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1UhoYK-0002uo-9P for qemu-devel@nongnu.org; Wed, 29 May 2013 18:09:10 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1UhoYE-0004Tr-8I for qemu-devel@nongnu.org; Wed, 29 May 2013 18:09:04 -0400 Received: from vms173025pub.verizon.net ([206.46.173.25]:41905) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1UhoYE-0004T0-2I for qemu-devel@nongnu.org; Wed, 29 May 2013 18:08:58 -0400 Received: from wf-rch.minyard.home ([unknown] [173.74.121.95]) by vms173025.mailsrvcs.net (Sun Java(tm) System Messaging Server 7u2-7.02 32bit (built Apr 16 2009)) with ESMTPA id <0MNK00IC7YU345C0@vms173025.mailsrvcs.net> for qemu-devel@nongnu.org; Wed, 29 May 2013 17:08:35 -0500 (CDT) From: minyard@acm.org Date: Wed, 29 May 2013 17:07:59 -0500 Message-id: <1369865296-19584-4-git-send-email-minyard@acm.org> In-reply-to: <1369865296-19584-1-git-send-email-minyard@acm.org> References: <1369865296-19584-1-git-send-email-minyard@acm.org> Subject: [Qemu-devel] [PATCH 03/20] qemu-char: Fix a race reporting opens and closes List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: Corey Minyard , openipmi-developer@lists.sourceforge.net From: Corey Minyard When an open event is reported, it is done through a bh. But close events are reported immediately. So if an open event is in the bh and a close occurs, an extraneous open happens, which can confuse a user. To fix this, this patch sets the "opened" flag immediately instead of in the bh handler and checks to make sure the opened flag is set before reporting an open event. This also modifies the spice code to call qemu_chr_generic_open to report an open, to keep things consistent. Signed-off-by: Corey Minyard --- qemu-char.c | 18 ++++++++++++++++-- spice-qemu-char.c | 7 +++++-- 2 files changed, 21 insertions(+), 4 deletions(-) diff --git a/qemu-char.c b/qemu-char.c index 2c34224..76cddd9 100644 --- a/qemu-char.c +++ b/qemu-char.c @@ -98,10 +98,13 @@ void qemu_chr_be_event(CharDriverState *s, int event) /* Keep track if the char device is open */ switch (event) { case CHR_EVENT_OPENED: + /* + * See the comment in qemu_chr_generic_open_bh() for why + * 's->opened = 1' is not here. + */ if (s->recon_timer) { qemu_del_timer(s->recon_timer); } - s->be_open = 1; break; case CHR_EVENT_CLOSED: if (s->recon_timer) { @@ -126,13 +129,24 @@ void qemu_chr_be_event(CharDriverState *s, int event) static gboolean qemu_chr_be_generic_open_bh(gpointer opaque) { CharDriverState *s = opaque; - qemu_chr_be_event(s, CHR_EVENT_OPENED); + /* + * Since the "close" event doesn't go through a bh, there is a + * possible race condition if a close comes in after an open, but + * the open is in the bh queue. So we double-check here, and we + * set opened in qemu_chr_generic_open() instead of + * qemu_chr_be_event(). + */ + if (s->be_open) { + qemu_chr_be_event(s, CHR_EVENT_OPENED); + } s->idle_tag = 0; return FALSE; } void qemu_chr_be_generic_open(CharDriverState *s) { + /* See the comment in qemu_chr_generic_open_bh() for why this is here */ + s->be_open = 1; if (s->idle_tag == 0) { s->idle_tag = g_idle_add(qemu_chr_be_generic_open_bh, s); } diff --git a/spice-qemu-char.c b/spice-qemu-char.c index 69c5938..436ab21 100644 --- a/spice-qemu-char.c +++ b/spice-qemu-char.c @@ -97,8 +97,11 @@ static void vmc_state(SpiceCharDeviceInstance *sin, int connected) return; } - qemu_chr_be_event(scd->chr, - connected ? CHR_EVENT_OPENED : CHR_EVENT_CLOSED); + if (connected) { + qemu_chr_generic_open(scd->chr); + } else { + qemu_chr_be_event(scd->chr, CHR_EVENT_CLOSED); + } } static SpiceCharDeviceInterface vmc_interface = { -- 1.7.9.5