Message ID | 1432874550-10921-2-git-send-email-mukawa@igel.co.jp |
---|---|
State | Superseded |
Headers | show |
On Fri, May 29, 2015 at 01:42:27PM +0900, Tetsuya Mukawa wrote: > Current QEMU cannot detect vhost-user backend disconnection. The > patch adds ability to know it. > To know disconnection, add watcher to detect G_IO_HUP event. When > G_IO_HUP event is detected, the disconnected socket will be read > to cause a CHR_EVENT_CLOSED. > > Signed-off-by: Tetsuya Mukawa <mukawa@igel.co.jp> > --- > net/vhost-user.c | 14 ++++++++++++++ > 1 file changed, 14 insertions(+) It is unfortunate that qemu-chr.c doesn't have a built-in way to handle G_IO_HUP and raise the CHR_EVENT_CLOSED event. This is not really specific to vhost-user. CCing Paolo Bonzini (Character Devices). He is currently on vacation so let's not hold up this patch. It can be made generic later, if necessary. > diff --git a/net/vhost-user.c b/net/vhost-user.c > index 11899c5..1967ff4 100644 > --- a/net/vhost-user.c > +++ b/net/vhost-user.c > @@ -19,6 +19,7 @@ typedef struct VhostUserState { > NetClientState nc; > CharDriverState *chr; > VHostNetState *vhost_net; > + int watch; > } VhostUserState; > > typedef struct VhostUserChardevProps { > @@ -113,12 +114,23 @@ static void net_vhost_link_down(VhostUserState *s, bool link_down) > } > } > > +static gboolean net_vhost_user_watch(GIOChannel *chan, GIOCondition cond, > + void *opaque) > +{ > + VhostUserState *s = opaque; > + uint8_t buf[1]; > + > + qemu_chr_fe_read_all(s->chr, buf, sizeof(buf)); > + return FALSE; > +} If you respin this patch, please add a comment like: /* We don't actually want to read anything, but CHR_EVENT_CLOSED will be * raised as a side-effect of the read. */
On 2015/06/15 22:32, Stefan Hajnoczi wrote: > On Fri, May 29, 2015 at 01:42:27PM +0900, Tetsuya Mukawa wrote: >> Current QEMU cannot detect vhost-user backend disconnection. The >> patch adds ability to know it. >> To know disconnection, add watcher to detect G_IO_HUP event. When >> G_IO_HUP event is detected, the disconnected socket will be read >> to cause a CHR_EVENT_CLOSED. >> >> Signed-off-by: Tetsuya Mukawa <mukawa@igel.co.jp> >> --- >> net/vhost-user.c | 14 ++++++++++++++ >> 1 file changed, 14 insertions(+) > It is unfortunate that qemu-chr.c doesn't have a built-in way to handle > G_IO_HUP and raise the CHR_EVENT_CLOSED event. This is not really > specific to vhost-user. > > CCing Paolo Bonzini (Character Devices). He is currently on vacation so > let's not hold up this patch. It can be made generic later, if > necessary. > >> diff --git a/net/vhost-user.c b/net/vhost-user.c >> index 11899c5..1967ff4 100644 >> --- a/net/vhost-user.c >> +++ b/net/vhost-user.c >> @@ -19,6 +19,7 @@ typedef struct VhostUserState { >> NetClientState nc; >> CharDriverState *chr; >> VHostNetState *vhost_net; >> + int watch; >> } VhostUserState; >> >> typedef struct VhostUserChardevProps { >> @@ -113,12 +114,23 @@ static void net_vhost_link_down(VhostUserState *s, bool link_down) >> } >> } >> >> +static gboolean net_vhost_user_watch(GIOChannel *chan, GIOCondition cond, >> + void *opaque) >> +{ >> + VhostUserState *s = opaque; >> + uint8_t buf[1]; >> + >> + qemu_chr_fe_read_all(s->chr, buf, sizeof(buf)); >> + return FALSE; >> +} > If you respin this patch, please add a comment like: > > /* We don't actually want to read anything, but CHR_EVENT_CLOSED will be > * raised as a side-effect of the read. > */ I appreciate for your all comments. I will add above comments in next patches. Tetsuya
diff --git a/net/vhost-user.c b/net/vhost-user.c index 11899c5..1967ff4 100644 --- a/net/vhost-user.c +++ b/net/vhost-user.c @@ -19,6 +19,7 @@ typedef struct VhostUserState { NetClientState nc; CharDriverState *chr; VHostNetState *vhost_net; + int watch; } VhostUserState; typedef struct VhostUserChardevProps { @@ -113,12 +114,23 @@ static void net_vhost_link_down(VhostUserState *s, bool link_down) } } +static gboolean net_vhost_user_watch(GIOChannel *chan, GIOCondition cond, + void *opaque) +{ + VhostUserState *s = opaque; + uint8_t buf[1]; + + qemu_chr_fe_read_all(s->chr, buf, sizeof(buf)); + return FALSE; +} + static void net_vhost_user_event(void *opaque, int event) { VhostUserState *s = opaque; switch (event) { case CHR_EVENT_OPENED: + s->watch = qemu_chr_fe_add_watch(s->chr, G_IO_HUP, net_vhost_user_watch, s); vhost_user_start(s); net_vhost_link_down(s, false); error_report("chardev \"%s\" went up", s->chr->label); @@ -126,6 +138,8 @@ static void net_vhost_user_event(void *opaque, int event) case CHR_EVENT_CLOSED: net_vhost_link_down(s, true); vhost_user_stop(s); + g_source_remove(s->watch); + s->watch = 0; error_report("chardev \"%s\" went down", s->chr->label); break; }
Current QEMU cannot detect vhost-user backend disconnection. The patch adds ability to know it. To know disconnection, add watcher to detect G_IO_HUP event. When G_IO_HUP event is detected, the disconnected socket will be read to cause a CHR_EVENT_CLOSED. Signed-off-by: Tetsuya Mukawa <mukawa@igel.co.jp> --- net/vhost-user.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+)