From: Daniel Vetter <daniel@ffwll.ch>
To: "Ville Syrjälä" <ville.syrjala@linux.intel.com>
Cc: Thomas Richter <richter@rus.uni-stuttgart.de>,
intel-gfx@lists.freedesktop.org
Subject: Re: [PATCH] drm/i915: Avoid double mutex lock applying pipe A quirk during sanitize_crtc()
Date: Tue, 10 Jun 2014 09:02:07 +0200 [thread overview]
Message-ID: <20140610070207.GD5821@phenom.ffwll.local> (raw)
In-Reply-To: <20140609083025.GL27580@intel.com>
On Mon, Jun 09, 2014 at 11:30:26AM +0300, Ville Syrjälä wrote:
> On Mon, Jun 09, 2014 at 07:47:10AM +0100, Chris Wilson wrote:
> > Thomas found that his machine would deadlock reloading the i915.ko
> > module after resume. He identified that this was caused by the
> > reacquisition of the connection mutex inside intel_enable_pipe_a()
> > during the CRTC sanitization routine. This will only affect machines
> > that quirk PIPE A, i.e. the original 830m chipsets.
> >
> > This patch move the locking into a wrapper function so that
> > intel_enable_pipe_a() can bypass the locking knowing that it already
> > holds the correct locks.
>
> It can still try to grab crtc->mutex twice. Looks like Danial undid my
> fix to not take all the modeset locks around
> intel_modeset_setup_hw_state().
Hm, I didn't find anything in git logs and we've had places where we used
modeset_lock_all since a long time. And I don't see how Chris' patch
wouldn't address this here. Can you please explain?
Thanks, Daniel
>
> >
> > Reported-by: Thomas Richter <richter@rus.uni-stuttgart.de>
> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> > Cc: Thomas Richter <richter@rus.uni-stuttgart.de>
> > Cc: Daniel Vetter <daniel@ffwll.ch>
> > Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > ---
> > drivers/gpu/drm/i915/intel_display.c | 54 ++++++++++++++++++++++------------
> > 1 file changed, 35 insertions(+), 19 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> > index c1f79a1..26d3424 100644
> > --- a/drivers/gpu/drm/i915/intel_display.c
> > +++ b/drivers/gpu/drm/i915/intel_display.c
> > @@ -8387,9 +8387,10 @@ mode_fits_in_fbdev(struct drm_device *dev,
> > #endif
> > }
> >
> > -bool intel_get_load_detect_pipe(struct drm_connector *connector,
> > - struct drm_display_mode *mode,
> > - struct intel_load_detect_pipe *old)
> > +static bool
> > +__intel_get_load_detect_pipe(struct drm_connector *connector,
> > + struct drm_display_mode *mode,
> > + struct intel_load_detect_pipe *old)
> > {
> > struct intel_crtc *intel_crtc;
> > struct intel_encoder *intel_encoder =
> > @@ -8405,7 +8406,7 @@ bool intel_get_load_detect_pipe(struct drm_connector *connector,
> > connector->base.id, connector->name,
> > encoder->base.id, encoder->name);
> >
> > - mutex_lock(&dev->mode_config.connection_mutex);
> > + lockdep_assert_held(&dev->mode_config.connection_mutex);
> >
> > /*
> > * Algorithm gets a little messy:
> > @@ -8449,7 +8450,7 @@ bool intel_get_load_detect_pipe(struct drm_connector *connector,
> > */
> > if (!crtc) {
> > DRM_DEBUG_KMS("no pipe available for load-detect\n");
> > - goto fail_unlock_connector;
> > + return false;
> > }
> >
> > mutex_lock(&crtc->mutex);
> > @@ -8503,14 +8504,13 @@ bool intel_get_load_detect_pipe(struct drm_connector *connector,
> > else
> > intel_crtc->new_config = NULL;
> > mutex_unlock(&crtc->mutex);
> > -fail_unlock_connector:
> > - mutex_unlock(&dev->mode_config.connection_mutex);
> >
> > return false;
> > }
> >
> > -void intel_release_load_detect_pipe(struct drm_connector *connector,
> > - struct intel_load_detect_pipe *old)
> > +static void
> > +__intel_release_load_detect_pipe(struct drm_connector *connector,
> > + struct intel_load_detect_pipe *old)
> > {
> > struct intel_encoder *intel_encoder =
> > intel_attached_encoder(connector);
> > @@ -8518,6 +8518,8 @@ void intel_release_load_detect_pipe(struct drm_connector *connector,
> > struct drm_crtc *crtc = encoder->crtc;
> > struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
> >
> > + lockdep_assert_held(&connector->dev->mode_config.connection_mutex);
> > +
> > DRM_DEBUG_KMS("[CONNECTOR:%d:%s], [ENCODER:%d:%s]\n",
> > connector->base.id, connector->name,
> > encoder->base.id, encoder->name);
> > @@ -8533,17 +8535,32 @@ void intel_release_load_detect_pipe(struct drm_connector *connector,
> > drm_framebuffer_unregister_private(old->release_fb);
> > drm_framebuffer_unreference(old->release_fb);
> > }
> > + } else {
> > + /* Switch crtc and encoder back off if necessary */
> > + if (old->dpms_mode != DRM_MODE_DPMS_ON)
> > + connector->funcs->dpms(connector, old->dpms_mode);
> > + }
> > + mutex_unlock(&crtc->mutex);
> > +}
> >
> > - mutex_unlock(&crtc->mutex);
> > +bool intel_get_load_detect_pipe(struct drm_connector *connector,
> > + struct drm_display_mode *mode,
> > + struct intel_load_detect_pipe *old)
> > +{
> > + mutex_lock(&connector->dev->mode_config.connection_mutex);
> > + if (!__intel_get_load_detect_pipe(connector, mode, old)) {
> > mutex_unlock(&connector->dev->mode_config.connection_mutex);
> > - return;
> > + return false;
> > }
> >
> > - /* Switch crtc and encoder back off if necessary */
> > - if (old->dpms_mode != DRM_MODE_DPMS_ON)
> > - connector->funcs->dpms(connector, old->dpms_mode);
> > + /* lock will be released by intel_release_load_detect_pipe() */
> > + return true;
> > +}
> >
> > - mutex_unlock(&crtc->mutex);
> > +void intel_release_load_detect_pipe(struct drm_connector *connector,
> > + struct intel_load_detect_pipe *old)
> > +{
> > + __intel_release_load_detect_pipe(connector, old);
> > mutex_unlock(&connector->dev->mode_config.connection_mutex);
> > }
> >
> > @@ -12354,10 +12371,9 @@ static void intel_enable_pipe_a(struct drm_device *dev)
> > if (!crt)
> > return;
> >
> > - if (intel_get_load_detect_pipe(crt, NULL, &load_detect_temp))
> > - intel_release_load_detect_pipe(crt, &load_detect_temp);
> > -
> > -
> > + lockdep_assert_held(&dev->mode_config.connection_mutex);
> > + if (__intel_get_load_detect_pipe(crt, NULL, &load_detect_temp))
> > + __intel_release_load_detect_pipe(crt, &load_detect_temp);
> > }
> >
> > static bool
> > --
> > 1.7.9.5
>
> --
> Ville Syrjälä
> Intel OTC
--
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
next prev parent reply other threads:[~2014-06-10 7:02 UTC|newest]
Thread overview: 70+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-06-05 16:15 [PATCH 00/15] drm/i915: Fix 830M/ns2501 for real, well almost ville.syrjala
2014-06-05 16:15 ` [PATCH 01/15] drm/i915: Use named initializers for gmch wm params ville.syrjala
2014-06-05 20:43 ` Chris Wilson
2014-06-05 21:02 ` Thomas Richter
2014-06-05 21:33 ` Bug reports on 830MG patches (thanks, but more trouble) Thomas Richter
2014-06-06 8:46 ` Ville Syrjälä
2014-06-06 17:24 ` Thomas Richter
2014-06-06 20:08 ` Ville Syrjälä
2014-06-06 21:09 ` Thomas Richter
2014-06-06 21:41 ` Ville Syrjälä
2014-06-08 21:29 ` [PATCH] Check for a min level when computing the watermark Thomas Richter
2014-06-06 16:38 ` [PATCH 01/15] drm/i915: Use named initializers for gmch wm params Daniel Vetter
2014-06-05 16:15 ` [PATCH 02/15] drm/i915: Fix gen2 planes B and C max watermark value ville.syrjala
2014-06-05 16:15 ` [PATCH 03/15] drm/i915: Don't get hw state from DVO chip unless DVO is enabled ville.syrjala
2014-06-06 16:39 ` Daniel Vetter
2014-06-05 16:15 ` [PATCH 04/15] drm/i915: ns2501 is on DVOB ville.syrjala
2014-06-06 16:57 ` Daniel Vetter
2014-06-06 21:46 ` Ville Syrjälä
2014-06-05 16:15 ` [PATCH 05/15] drm/i915: Enable DVO between mode_set and dpms hooks ville.syrjala
2014-06-05 16:15 ` [PATCH 06/15] drm/i915: Don't call DVO mode_set hook on DPMS changes ville.syrjala
2014-06-05 16:15 ` [PATCH 07/15] drm/i915: Kill useless ns2501_dump_regs ville.syrjala
2014-06-05 16:15 ` [PATCH 08/15] drm/i915: Rewrite ns2501 driver a bit ville.syrjala
2014-06-05 16:15 ` [PATCH 09/15] drm/i915: Ignore VBT int_crt_support on 830M ville.syrjala
2014-06-06 17:00 ` Daniel Vetter
2014-06-06 19:44 ` [PATCH v2 " ville.syrjala
2014-06-06 20:13 ` Daniel Vetter
2014-06-07 20:37 ` [Patch] Add minimum watermark level for I830 Thomas Richter
2014-06-06 21:15 ` [PATCH v2 09/15] drm/i915: Ignore VBT int_crt_support on 830M Bob Paauwe
2014-06-06 22:23 ` Daniel Vetter
2014-06-06 22:51 ` Jesse Barnes
[not found] ` <2094_1402093395_53923F53_2094_10301_1_CAKMK7uGAnNP4VR9+zXd0KD5v0Vo=XuDS=NhRNFRqHKcae7T4XQ@mail.gmail.com>
2014-06-07 17:32 ` Thomas Richter
2014-10-24 13:23 ` Jani Nikula
2014-10-24 14:11 ` Ville Syrjälä
2014-06-05 16:15 ` [PATCH 10/15] drm/i915: Fix DVO 2x clock enable " ville.syrjala
2014-06-05 16:16 ` [PATCH 11/15] Revert "drm/i915: Nuke pipe A quirk on i830M" ville.syrjala
2014-06-05 16:16 ` [PATCH 12/15] drm/i915: Add pipe B force quirk for 830M ville.syrjala
2014-06-05 16:16 ` [PATCH 13/15] drm/i915: Eliminate rmw from .update_primary_plane() ville.syrjala
2014-06-06 0:02 ` Matt Roper
2014-06-06 19:45 ` [PATCH v2 " ville.syrjala
2014-06-05 16:16 ` [PATCH 14/15] drm/i915: Call .update_primary_plane in intel_{enable, disable}_primary_hw_plane() ville.syrjala
2014-06-06 0:02 ` Matt Roper
2014-06-06 8:40 ` Ville Syrjälä
2014-06-06 19:46 ` [PATCH v2 " ville.syrjala
2014-06-05 16:16 ` [PATCH 15/15] drm/i915: Check pixel clock in ns2501 mode_valid hook ville.syrjala
2014-06-06 19:47 ` [PATCH 16/15] drm/i915: Pass intel_crtc to intel_disable_pipe() and intel_wait_for_pipe_off() ville.syrjala
2014-06-06 19:47 ` [PATCH 17/15] drm/i915: Disable double wide even when leaving the pipe on ville.syrjala
2014-06-06 22:09 ` [PATCH v2 " ville.syrjala
2014-06-08 23:14 ` Deadlock in intel_enable_pipe_a() Thomas Richter
2014-06-09 6:47 ` [PATCH] drm/i915: Avoid double mutex lock applying pipe A quirk during sanitize_crtc() Chris Wilson
2014-06-09 8:30 ` Ville Syrjälä
2014-06-09 8:50 ` Chris Wilson
[not found] ` <28223_1402303866_5395757A_28223_3428_1_20140609085045.GE16767@nuc-i3427.alporthouse.com>
2014-06-09 10:57 ` Partial success - Fixing resume from s2ram on S6010 Thomas Richter
2014-06-09 11:08 ` Ville Syrjälä
[not found] ` <28223_1402312148_539595D3_28223_4884_1_20140609110857.GM27580@intel.com>
2014-06-09 11:19 ` Thomas Richter
2014-06-09 11:31 ` Ville Syrjälä
[not found] ` <2086_1402313568_53959B5F_2086_895_1_20140609113155.GN27580@intel.com>
2014-06-09 12:33 ` Thomas Richter
2014-06-09 12:57 ` Thomas Richter
2014-06-09 18:41 ` Thomas Richter
2014-06-09 19:46 ` [PATCH] drm/i915: Init important ns2501 registers ville.syrjala
[not found] ` <28223_1402343538_53961072_28223_7661_1_1402343204-28608-1-git-send-email-ville.syrjala@linux.intel.com>
2014-06-09 20:58 ` Thomas Richter
2014-06-09 22:29 ` Thomas Richter
2014-06-10 14:04 ` Ville Syrjälä
[not found] ` <29040_1402409145_539710B9_29040_2220_1_20140610140430.GD27580@intel.com>
2014-06-10 16:38 ` Thomas Richter
2014-06-18 16:03 ` i830GM on IBM R31 works with alm_fixes5 repository Thomas Richter
2014-06-10 7:02 ` Daniel Vetter [this message]
2014-06-10 8:53 ` [PATCH] drm/i915: Avoid double mutex lock applying pipe A quirk during sanitize_crtc() Ville Syrjälä
2014-06-10 9:22 ` Daniel Vetter
2014-06-10 6:59 ` Daniel Vetter
2014-06-10 7:13 ` Chris Wilson
2014-06-06 19:47 ` [PATCH 18/15] drm/i915: Preserve VGACNTR bits from the BIOS ville.syrjala
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20140610070207.GD5821@phenom.ffwll.local \
--to=daniel@ffwll.ch \
--cc=intel-gfx@lists.freedesktop.org \
--cc=richter@rus.uni-stuttgart.de \
--cc=ville.syrjala@linux.intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox