Re: [Intel-gfx] [PATCH] drm/i915: Fix lock dropping in intel_tv_detect()

From: Ville Syrjälä
Date: Mon Sep 01 2014 - 05:46:26 EST


On Mon, Sep 01, 2014 at 09:50:22AM +0100, Chris Wilson wrote:
> On Mon, Sep 01, 2014 at 11:09:46AM +0300, ville.syrjala@xxxxxxxxxxxxxxx wrote:
> > From: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx>
> >
> > When intel_tv_detect() fails to do load detection it would forget to
> > drop the locks and clean up the acquire context. Fix it up.
> >
> > This is a regression from:
> > commit 208bf9fdcd3575aa4a5d48b3e0295f7cdaf6fc44
> > Author: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx>
> > Date: Mon Aug 11 13:15:35 2014 +0300
> >
> > drm/i915: Fix locking for intel_enable_pipe_a()
> >
> > Cc: Tibor Billes <tbilles@xxxxxxx>
> > Signed-off-by: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx>
> > ---
> > drivers/gpu/drm/i915/intel_tv.c | 15 +++++++++------
> > 1 file changed, 9 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/intel_tv.c b/drivers/gpu/drm/i915/intel_tv.c
> > index 32186a6..abbf0ea 100644
> > --- a/drivers/gpu/drm/i915/intel_tv.c
> > +++ b/drivers/gpu/drm/i915/intel_tv.c
> > @@ -1311,7 +1311,8 @@ intel_tv_detect(struct drm_connector *connector, bool force)
> > {
> > struct drm_display_mode mode;
> > struct intel_tv *intel_tv = intel_attached_tv(connector);
> > - int type;
> > + enum drm_connector_status status = connector->status;
> > + int type = -1;
> >
> > DRM_DEBUG_KMS("[CONNECTOR:%d:%s] force=%d\n",
> > connector->base.id, connector->name,
> > @@ -1328,21 +1329,23 @@ intel_tv_detect(struct drm_connector *connector, bool force)
> > if (intel_get_load_detect_pipe(connector, &mode, &tmp, &ctx)) {
> > type = intel_tv_detect_type(intel_tv, connector);
> > intel_release_load_detect_pipe(connector, &tmp);
> > + status = type < 0 ?
> > + connector_status_disconnected :
> > + connector_status_connected;
> > } else
> > - return connector_status_unknown;
> > + status = connector_status_unknown;
> >
> > drm_modeset_drop_locks(&ctx);
> > drm_modeset_acquire_fini(&ctx);
> > - } else
> > - return connector->status;
> > + }
> >
> > if (type < 0)
> > - return connector_status_disconnected;
> > + return status;
> >
> > intel_tv->type = type;
> > intel_tv_find_better_format(connector);
> >
> > - return connector_status_connected;
> > + return status;
>
> Hmm, this makes the code unclear. if type != -1, status should always be
> connected.
>
> I think:
>
> if (status != connector_status_connected)
> return status;
>
> if (WARN_ON(type < 0))
> return connector_status_disconnected;
>
> intel_tv->type = type;
> find_better_format();
> return connector_status_connected;
>
> And leave type unset to encourage gcc.
>
>
>
> diff --git a/drivers/gpu/drm/i915/intel_tv.c b/drivers/gpu/drm/i915/intel_tv.c
> index 32186a6..a109b7b 100644
> --- a/drivers/gpu/drm/i915/intel_tv.c
> +++ b/drivers/gpu/drm/i915/intel_tv.c
> @@ -1311,6 +1311,7 @@ intel_tv_detect(struct drm_connector *connector, bool force)
> {
> struct drm_display_mode mode;
> struct intel_tv *intel_tv = intel_attached_tv(connector);
> + enum drm_connector_status status = connector->status;
> int type;
>
> DRM_DEBUG_KMS("[CONNECTOR:%d:%s] force=%d\n",
> @@ -1327,17 +1328,20 @@ intel_tv_detect(struct drm_connector *connector, bool force)
>
> if (intel_get_load_detect_pipe(connector, &mode, &tmp, &ctx)) {
> type = intel_tv_detect_type(intel_tv, connector);
> + status = type < 0 ?
> + connector_status_disconnected :
> + connector_status_connected;
> +
> intel_release_load_detect_pipe(connector, &tmp);
> } else
> - return connector_status_unknown;
> + status = connector_status_unknown;
>
> drm_modeset_drop_locks(&ctx);
> drm_modeset_acquire_fini(&ctx);
> - } else
> - return connector->status;
> + }
>
> - if (type < 0)
> - return connector_status_disconnected;
> + if (status != connector_status_connected)
> + return status;
>
> intel_tv->type = type;
> intel_tv_find_better_format(connector);

With this approach we're going to have to keep the !force else branch
to avoid populating intel_tv->type with stack garbage. But I suppose
the resulting code might still be a bit clearer.

--
Ville Syrjälä
Intel OTC
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/