From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:37067) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1dQUkw-0004xY-Dx for qemu-devel@nongnu.org; Thu, 29 Jun 2017 04:24:56 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1dQUkq-0007Nr-4C for qemu-devel@nongnu.org; Thu, 29 Jun 2017 04:24:54 -0400 Received: from smtp.eu.citrix.com ([185.25.65.24]:45989) by eggs.gnu.org with esmtps (TLS1.0:RSA_ARCFOUR_SHA1:16) (Exim 4.71) (envelope-from ) id 1dQUkp-0007L7-MD for qemu-devel@nongnu.org; Thu, 29 Jun 2017 04:24:48 -0400 From: Paul Durrant Date: Thu, 29 Jun 2017 08:24:41 +0000 Message-ID: <7ecfe394c646482faddf97a40bda944d@AMSPEX02CL03.citrite.net> References: In-Reply-To: Content-Language: en-US Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Subject: Re: [Qemu-devel] [PATCH v2] xenfb: remove xen_init_display "temporary" hack List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: 'Stefano Stabellini' , "xen-devel@lists.xenproject.org" , "qemu-devel@nongnu.org" Cc: "peter.maydell@linaro.org" , Anthony Perard , "kraxel@redhat.com" > -----Original Message----- > From: Stefano Stabellini [mailto:sstabellini@kernel.org] > Sent: 28 June 2017 19:37 > To: xen-devel@lists.xenproject.org; qemu-devel@nongnu.org > Cc: sstabellini@kernel.org; peter.maydell@linaro.org; Anthony Perard > ; kraxel@redhat.com; Paul Durrant > > Subject: [PATCH v2] xenfb: remove xen_init_display "temporary" hack >=20 > Initialize xenfb properly, as all other backends, from its own > "initialise" function. >=20 > Signed-off-by: Stefano Stabellini >=20 > --- > Changes in v2: > - remove xen_init_display from xen_backend.h > - handle cases where vkbd is missing >=20 > diff --git a/hw/display/xenfb.c b/hw/display/xenfb.c > index e76c0d8..3b0168b 100644 > --- a/hw/display/xenfb.c > +++ b/hw/display/xenfb.c > @@ -71,7 +71,6 @@ struct XenFB { > int fbpages; > int feature_update; > int bug_trigger; > - int have_console; > int do_resize; >=20 > struct { > @@ -80,6 +79,7 @@ struct XenFB { > int up_count; > int up_fullscreen; > }; > +static const GraphicHwOps xenfb_ops; >=20 > /* -------------------------------------------------------------------- = */ >=20 > @@ -855,6 +855,8 @@ static int fb_init(struct XenDevice *xendev) > static int fb_initialise(struct XenDevice *xendev) > { > struct XenFB *fb =3D container_of(xendev, struct XenFB, c.xendev); > + struct XenDevice *xin; > + struct XenInput *in; I think the scope of 'in' can be limited to the 'else' clause where it is u= sed below. > struct xenfb_page *fb_page; > int videoram; > int rc; > @@ -877,16 +879,16 @@ static int fb_initialise(struct XenDevice *xendev) > if (rc !=3D 0) > return rc; >=20 > -#if 0 /* handled in xen_init_display() for now */ > - if (!fb->have_console) { > - fb->c.ds =3D graphic_console_init(xenfb_update, > - xenfb_invalidate, > - NULL, > - NULL, > - fb); > - fb->have_console =3D 1; > + fb->c.con =3D graphic_console_init(NULL, 0, &xenfb_ops, fb); > + > + xin =3D xen_pv_find_xendev("vkbd", xen_domid, 0); > + if (xin =3D=3D NULL) { > + xen_pv_printf(xendev, 1, "xenfb is up, but vkbd is not present\n= "); > + } else { > + in =3D container_of(xin, struct XenInput, c.xendev); > + in->c.con =3D fb->c.con; I notice that in->c.con is tested in input_initialise() and it will fail if= NULL, so there is an inherent race between the frontends and also processi= ng of xenstore watches. That seems a little fragile at best, but the depend= ency of input_initialise() on in->c.con is also somewhat bogus as the vkbd = backend is created for xenfv machine types whereas xenfb is only there for = xenpv machine types. I think this needs to be fixed before xenfb can be rel= ied on to initialize correctly in all circumstances. Paul > + xen_be_check_state(xin); > } > -#endif >=20 > if (xenstore_read_fe_int(xendev, "feature-update", &fb- > >feature_update) =3D=3D -1) > fb->feature_update =3D 0; > @@ -972,42 +974,3 @@ static const GraphicHwOps xenfb_ops =3D { > .gfx_update =3D xenfb_update, > .update_interval =3D xenfb_update_interval, > }; > - > -/* > - * FIXME/TODO: Kill this. > - * Temporary needed while DisplayState reorganization is in flight. > - */ > -void xen_init_display(int domid) > -{ > - struct XenDevice *xfb, *xin; > - struct XenFB *fb; > - struct XenInput *in; > - int i =3D 0; > - > -wait_more: > - i++; > - main_loop_wait(true); > - xfb =3D xen_pv_find_xendev("vfb", domid, 0); > - xin =3D xen_pv_find_xendev("vkbd", domid, 0); > - if (!xfb || !xin) { > - if (i < 256) { > - usleep(10000); > - goto wait_more; > - } > - xen_pv_printf(NULL, 1, "displaystate setup failed\n"); > - return; > - } > - > - /* vfb */ > - fb =3D container_of(xfb, struct XenFB, c.xendev); > - fb->c.con =3D graphic_console_init(NULL, 0, &xenfb_ops, fb); > - fb->have_console =3D 1; > - > - /* vkbd */ > - in =3D container_of(xin, struct XenInput, c.xendev); > - in->c.con =3D fb->c.con; > - > - /* retry ->init() */ > - xen_be_check_state(xin); > - xen_be_check_state(xfb); > -} > diff --git a/hw/xenpv/xen_machine_pv.c b/hw/xenpv/xen_machine_pv.c > index 79aef4e..31d2f25 100644 > --- a/hw/xenpv/xen_machine_pv.c > +++ b/hw/xenpv/xen_machine_pv.c > @@ -94,9 +94,6 @@ static void xen_init_pv(MachineState *machine) >=20 > /* config cleanup hook */ > atexit(xen_config_cleanup); > - > - /* setup framebuffer */ > - xen_init_display(xen_domid); > } >=20 > static void xenpv_machine_init(MachineClass *mc) > diff --git a/include/hw/xen/xen_backend.h > b/include/hw/xen/xen_backend.h > index 852c2ea..8a6fbcb 100644 > --- a/include/hw/xen/xen_backend.h > +++ b/include/hw/xen/xen_backend.h > @@ -55,8 +55,6 @@ extern struct XenDevOps xen_netdev_ops; /* > xen_nic.c */ > extern struct XenDevOps xen_usb_ops; /* xen-usb.c */ > #endif >=20 > -void xen_init_display(int domid); > - > /* configuration (aka xenbus setup) */ > void xen_config_cleanup(void); > int xen_config_dev_blk(DriveInfo *disk);