From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([208.118.235.92]:54647) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1UTDIV-0005Ep-9W for qemu-devel@nongnu.org; Fri, 19 Apr 2013 11:32:24 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1UTDIQ-0002XA-Eh for qemu-devel@nongnu.org; Fri, 19 Apr 2013 11:32:23 -0400 Received: from mail-ea0-x235.google.com ([2a00:1450:4013:c01::235]:46173) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1UTDIQ-0002Ww-8v for qemu-devel@nongnu.org; Fri, 19 Apr 2013 11:32:18 -0400 Received: by mail-ea0-f181.google.com with SMTP id z10so1776664ead.26 for ; Fri, 19 Apr 2013 08:32:17 -0700 (PDT) Sender: Paolo Bonzini From: Paolo Bonzini Date: Fri, 19 Apr 2013 17:32:07 +0200 Message-Id: <1366385529-10329-3-git-send-email-pbonzini@redhat.com> In-Reply-To: <1366385529-10329-1-git-send-email-pbonzini@redhat.com> References: <1366385529-10329-1-git-send-email-pbonzini@redhat.com> Subject: [Qemu-devel] [PATCH v2 2/4] qemu-char: simplify pty polling List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: amit.shah@redhat.com, aliguori@us.ibm.com, kraxel@redhat.com There is no need to use a timer and pty_chr_read to detect a connected pty. It is simpler to just call g_poll periodically and check for POLLHUP. It is done once per second, and only if the pty is disconnected, so it is cheap enough. Tested with "-monitor pty" and "-serial mon:pty", both of which work correctly and do not freeze QEMU. (How to test ptys? "socat -,raw,echo=0 /dev/pts/4,raw"). Signed-off-by: Paolo Bonzini --- qemu-char.c | 41 +++++++++++++++++------------------------ 1 files changed, 17 insertions(+), 24 deletions(-) diff --git a/qemu-char.c b/qemu-char.c index 552a498..d14888d 100644 --- a/qemu-char.c +++ b/qemu-char.c @@ -1028,7 +1028,6 @@ typedef struct { GIOChannel *fd; guint fd_tag; int connected; - int polling; int read_bytes; guint timer_tag; } PtyCharDriver; @@ -1044,12 +1043,6 @@ static gboolean pty_chr_timer(gpointer opaque) if (s->connected) { goto out; } - if (s->polling) { - /* If we arrive here without polling being cleared due - * read returning -EIO, then we are (re-)connected */ - pty_chr_state(chr, 1); - goto out; - } /* Next poll ... */ pty_chr_update_read_handler(chr); @@ -1128,22 +1121,17 @@ static gboolean pty_chr_read(GIOChannel *chan, GIOCondition cond, void *opaque) static void pty_chr_update_read_handler(CharDriverState *chr) { PtyCharDriver *s = chr->opaque; + GPollFD pfd; - if (s->fd_tag) { - g_source_remove(s->fd_tag); + pfd.fd = g_io_channel_unix_get_fd(s->fd); + pfd.events = G_IO_OUT; + pfd.revents = 0; + g_poll(&pfd, 1, 0); + if (pfd.revents & G_IO_HUP) { + pty_chr_state(chr, 0); + } else { + pty_chr_state(chr, 1); } - - s->fd_tag = io_add_watch_poll(s->fd, pty_chr_read_poll, pty_chr_read, chr); - s->polling = 1; - /* - * Short timeout here: just need wait long enougth that qemu makes - * it through the poll loop once. When reconnected we want a - * short timeout so we notice it almost instantly. Otherwise - * read() gives us -EIO instantly, making pty_chr_state() reset the - * timeout to the normal (much longer) poll interval before the - * timer triggers. - */ - pty_chr_rearm_timer(chr, 10); } static void pty_chr_state(CharDriverState *chr, int connected) @@ -1156,15 +1144,20 @@ static void pty_chr_state(CharDriverState *chr, int connected) s->fd_tag = 0; } s->connected = 0; - s->polling = 0; /* (re-)connect poll interval for idle guests: once per second. * We check more frequently in case the guests sends data to * the virtual device linked to our pty. */ pty_chr_rearm_timer(chr, 1000); } else { - if (!s->connected) + if (s->timer_tag) { + g_source_remove(s->timer_tag); + s->timer_tag = 0; + } + if (!s->connected) { qemu_chr_be_generic_open(chr); - s->connected = 1; + s->connected = 1; + s->fd_tag = io_add_watch_poll(s->fd, pty_chr_read_poll, pty_chr_read, chr); + } } } -- 1.7.1