From: Imre Deak <imre.deak@intel.com>
To: "Souza, Jose" <jose.souza@intel.com>
Cc: "intel-gfx@lists.freedesktop.org" <intel-gfx@lists.freedesktop.org>
Subject: Re: [Intel-gfx] [PATCH 08/13] drm/i915/tc: Refactor TC-cold block/unblock helpers
Date: Tue, 28 Sep 2021 01:28:43 +0300 [thread overview]
Message-ID: <20210927222843.GC2043218@ideak-desk.fi.intel.com> (raw)
In-Reply-To: <e640e3f2a01ddc4e9f57a9ff7411468a938441df.camel@intel.com>
On Tue, Sep 28, 2021 at 01:21:21AM +0300, Souza, Jose wrote:
> On Tue, 2021-09-28 at 01:13 +0300, Imre Deak wrote:
> > On Tue, Sep 28, 2021 at 12:56:24AM +0300, Souza, Jose wrote:
> > > On Tue, 2021-09-21 at 03:23 +0300, Imre Deak wrote:
> > > > A follow-up change will select the TC-cold blocking power domain based
> > > > on the TypeC mode, prepare for that here.
> > > >
> > > > Also bring intel_tc_cold_requires_aux_pw() earlier to its logical place
> > > > for readability.
> > > >
> > > > No functional change.
> > > >
> > > > Cc: José Roberto de Souza <jose.souza@intel.com>
> > > > Signed-off-by: Imre Deak <imre.deak@intel.com>
> > > > ---
> > > > .../drm/i915/display/intel_display_types.h | 2 +
> > > > drivers/gpu/drm/i915/display/intel_tc.c | 63 +++++++++++--------
> > > > 2 files changed, 39 insertions(+), 26 deletions(-)
> > > >
> > > > diff --git a/drivers/gpu/drm/i915/display/intel_display_types.h b/drivers/gpu/drm/i915/display/intel_display_types.h
> > > > index e9e806d90eec4..08a73ffded957 100644
> > > > --- a/drivers/gpu/drm/i915/display/intel_display_types.h
> > > > +++ b/drivers/gpu/drm/i915/display/intel_display_types.h
> > > > @@ -1663,8 +1663,10 @@ struct intel_digital_port {
> > > > enum intel_display_power_domain ddi_io_power_domain;
> > > > intel_wakeref_t ddi_io_wakeref;
> > > > intel_wakeref_t aux_wakeref;
> > > > +
> > > > struct mutex tc_lock; /* protects the TypeC port mode */
> > > > intel_wakeref_t tc_lock_wakeref;
> > > > + enum intel_display_power_domain tc_lock_power_domain;
> > > > int tc_link_refcount;
> > > > bool tc_legacy_port:1;
> > > > char tc_port_name[8];
> > > > diff --git a/drivers/gpu/drm/i915/display/intel_tc.c b/drivers/gpu/drm/i915/display/intel_tc.c
> > > > index 77b16a7c43466..24d2dc2e19a7d 100644
> > > > --- a/drivers/gpu/drm/i915/display/intel_tc.c
> > > > +++ b/drivers/gpu/drm/i915/display/intel_tc.c
> > > > @@ -48,8 +48,16 @@ bool intel_tc_port_in_legacy_mode(struct intel_digital_port *dig_port)
> > > > return intel_tc_port_in_mode(dig_port, TC_PORT_LEGACY);
> > > > }
> > > >
> > > > +bool intel_tc_cold_requires_aux_pw(struct intel_digital_port *dig_port)
> > > > +{
> > > > + struct drm_i915_private *i915 = to_i915(dig_port->base.base.dev);
> > > > +
> > > > + return (DISPLAY_VER(i915) == 11 && dig_port->tc_legacy_port) ||
> > > > + IS_ALDERLAKE_P(i915);
> > > > +}
> > > > +
> > > > static enum intel_display_power_domain
> > > > -tc_cold_get_power_domain(struct intel_digital_port *dig_port)
> > > > +tc_cold_get_power_domain(struct intel_digital_port *dig_port, enum tc_port_mode mode)
> > > > {
> > > > if (intel_tc_cold_requires_aux_pw(dig_port))
> > > > return intel_legacy_aux_to_power_domain(dig_port->aux_ch);
> > > > @@ -58,23 +66,30 @@ tc_cold_get_power_domain(struct intel_digital_port *dig_port)
> > > > }
> > > >
> > > > static intel_wakeref_t
> > > > -tc_cold_block(struct intel_digital_port *dig_port)
> > > > +tc_cold_block_in_mode(struct intel_digital_port *dig_port, enum tc_port_mode mode,
> > > > + enum intel_display_power_domain *domain)
> > > > {
> > > > struct drm_i915_private *i915 = to_i915(dig_port->base.base.dev);
> > > > - enum intel_display_power_domain domain;
> > > >
> > > > if (DISPLAY_VER(i915) == 11 && !dig_port->tc_legacy_port)
> > > > return 0;
> > > >
> > > > - domain = tc_cold_get_power_domain(dig_port);
> > > > - return intel_display_power_get(i915, domain);
> > > > + *domain = tc_cold_get_power_domain(dig_port, mode);
> > >
> > > I see problems with this approach.
> > >
> > > If there is a TC alt-mode is connected and current software state is
> > > TBT or disconnected it will not get the power domain to exit TC cold,
> > > what could case invalid reads of registers.
> >
> > If the mode needs to be changed, like in the above cases, the power
> > domain required to exit TC cold will be taken.
>
> How?
> I see at least this 2 problematic cases.
>
> intel_tc_port_sanitize()
> tc_cold_block(mode=TBT)
It's in disconnected mode at this point.
>
> dig_port->tc_mode = intel_tc_port_get_current_mode(dig_port);
> ...
>
> tc_cold_unblock(dig_port, domain, tc_cold_wref);
>
> Other case
>
> intel_tc_port_connected()
>
> tc_cold_block(mode=TBT)
Reading PORT_TX_DFLEXDPSP while in TC-cold returns 0xffffffff, and it's
handled when reading it (which is also required by the spec).
> is_connected = tc_port_live_status_mask(dig_port) &
> BIT(dig_port->tc_mode);
>
> tc_cold_unblock(dig_port, domain, tc_cold_wref);
>
>
> >
> > > From the next patch commit message: 'For the ADL-P TBT mode the spec
> > > doesn't require blocking TC-cold by using the legacy AUX power
> > > domain'.
> > >
> > > It is not required for TBT but when there is nothing connected,
> > > hardware is not in TBT mode so it can still get into TC cold.
> >
> > If there is nothing connected the required power domain will be taken to
> > exit TC cold.
> >
> > > > +
> > > > + return intel_display_power_get(i915, *domain);
> > > > +}
> > > > +
> > > > +static intel_wakeref_t
> > > > +tc_cold_block(struct intel_digital_port *dig_port, enum intel_display_power_domain *domain)
> > > > +{
> > > > + return tc_cold_block_in_mode(dig_port, dig_port->tc_mode, domain);
> > > > }
> > > >
> > > > static void
> > > > -tc_cold_unblock(struct intel_digital_port *dig_port, intel_wakeref_t wakeref)
> > > > +tc_cold_unblock(struct intel_digital_port *dig_port, enum intel_display_power_domain domain,
> > > > + intel_wakeref_t wakeref)
> > > > {
> > > > struct drm_i915_private *i915 = to_i915(dig_port->base.base.dev);
> > > > - enum intel_display_power_domain domain;
> > > >
> > > > /*
> > > > * wakeref == -1, means some error happened saving save_depot_stack but
> > > > @@ -84,8 +99,7 @@ tc_cold_unblock(struct intel_digital_port *dig_port, intel_wakeref_t wakeref)
> > > > if (wakeref == 0)
> > > > return;
> > > >
> > > > - domain = tc_cold_get_power_domain(dig_port);
> > > > - intel_display_power_put_async(i915, domain, wakeref);
> > > > + intel_display_power_put(i915, domain, wakeref);
> > > > }
> > > >
> > > > static void
> > > > @@ -98,7 +112,8 @@ assert_tc_cold_blocked(struct intel_digital_port *dig_port)
> > > > return;
> > > >
> > > > enabled = intel_display_power_is_enabled(i915,
> > > > - tc_cold_get_power_domain(dig_port));
> > > > + tc_cold_get_power_domain(dig_port,
> > > > + dig_port->tc_mode));
> > > > drm_WARN_ON(&i915->drm, !enabled);
> > > > }
> > > >
> > > > @@ -634,7 +649,8 @@ void intel_tc_port_sanitize(struct intel_digital_port *dig_port)
> > > >
> > > > drm_WARN_ON(&i915->drm, dig_port->tc_mode != TC_PORT_DISCONNECTED);
> > > > if (active_links) {
> > > > - intel_wakeref_t tc_cold_wref = tc_cold_block(dig_port);
> > > > + enum intel_display_power_domain domain;
> > > > + intel_wakeref_t tc_cold_wref = tc_cold_block(dig_port, &domain);
> > > >
> > > > dig_port->tc_mode = intel_tc_port_get_current_mode(dig_port);
> > > >
> > > > @@ -644,7 +660,7 @@ void intel_tc_port_sanitize(struct intel_digital_port *dig_port)
> > > > dig_port->tc_port_name, active_links);
> > > > intel_tc_port_link_init_refcount(dig_port, active_links);
> > > >
> > > > - tc_cold_unblock(dig_port, tc_cold_wref);
> > > > + tc_cold_unblock(dig_port, domain, tc_cold_wref);
> > > > }
> > > >
> > > > drm_dbg_kms(&i915->drm, "Port %s: sanitize mode (%s)\n",
> > > > @@ -673,15 +689,16 @@ bool intel_tc_port_connected(struct intel_encoder *encoder)
> > > > {
> > > > struct intel_digital_port *dig_port = enc_to_dig_port(encoder);
> > > > bool is_connected;
> > > > + enum intel_display_power_domain domain;
> > > > intel_wakeref_t tc_cold_wref;
> > > >
> > > > intel_tc_port_lock(dig_port);
> > > > - tc_cold_wref = tc_cold_block(dig_port);
> > > > + tc_cold_wref = tc_cold_block(dig_port, &domain);
> > > >
> > > > is_connected = tc_port_live_status_mask(dig_port) &
> > > > BIT(dig_port->tc_mode);
> > > >
> > > > - tc_cold_unblock(dig_port, tc_cold_wref);
> > > > + tc_cold_unblock(dig_port, domain, tc_cold_wref);
> > > > intel_tc_port_unlock(dig_port);
> > > >
> > > > return is_connected;
> > > > @@ -698,15 +715,16 @@ static void __intel_tc_port_lock(struct intel_digital_port *dig_port,
> > > > mutex_lock(&dig_port->tc_lock);
> > > >
> > > > if (!dig_port->tc_link_refcount) {
> > > > + enum intel_display_power_domain domain;
> > > > intel_wakeref_t tc_cold_wref;
> > > >
> > > > - tc_cold_wref = tc_cold_block(dig_port);
> > > > + tc_cold_wref = tc_cold_block(dig_port, &domain);
> > > >
> > > > if (force_disconnect || intel_tc_port_needs_reset(dig_port))
> > > > intel_tc_port_reset_mode(dig_port, required_lanes,
> > > > force_disconnect);
> > > >
> > > > - tc_cold_unblock(dig_port, tc_cold_wref);
> > > > + tc_cold_unblock(dig_port, domain, tc_cold_wref);
> > > > }
> > > >
> > > > drm_WARN_ON(&i915->drm, dig_port->tc_mode == TC_PORT_DISCONNECTED);
> > > > @@ -775,6 +793,7 @@ void intel_tc_port_put_link(struct intel_digital_port *dig_port)
> > > > static bool
> > > > tc_has_modular_fia(struct drm_i915_private *i915, struct intel_digital_port *dig_port)
> > > > {
> > > > + enum intel_display_power_domain domain;
> > > > intel_wakeref_t wakeref;
> > > > u32 val;
> > > >
> > > > @@ -782,9 +801,9 @@ tc_has_modular_fia(struct drm_i915_private *i915, struct intel_digital_port *dig
> > > > return false;
> > > >
> > > > mutex_lock(&dig_port->tc_lock);
> > > > - wakeref = tc_cold_block(dig_port);
> > > > + wakeref = tc_cold_block(dig_port, &domain);
> > > > val = intel_uncore_read(&i915->uncore, PORT_TX_DFLEXDPSP(FIA1));
> > > > - tc_cold_unblock(dig_port, wakeref);
> > > > + tc_cold_unblock(dig_port, domain, wakeref);
> > > > mutex_unlock(&dig_port->tc_lock);
> > > >
> > > > drm_WARN_ON(&i915->drm, val == 0xffffffff);
> > > > @@ -829,11 +848,3 @@ void intel_tc_port_init(struct intel_digital_port *dig_port, bool is_legacy)
> > > > dig_port->tc_link_refcount = 0;
> > > > tc_port_load_fia_params(i915, dig_port);
> > > > }
> > > > -
> > > > -bool intel_tc_cold_requires_aux_pw(struct intel_digital_port *dig_port)
> > > > -{
> > > > - struct drm_i915_private *i915 = to_i915(dig_port->base.base.dev);
> > > > -
> > > > - return (DISPLAY_VER(i915) == 11 && dig_port->tc_legacy_port) ||
> > > > - IS_ALDERLAKE_P(i915);
> > > > -}
> > >
>
next prev parent reply other threads:[~2021-09-27 22:30 UTC|newest]
Thread overview: 72+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-09-21 0:23 [Intel-gfx] [PATCH 00/13] drm/i915/tc: Fix TypeC connect/disconnect sequences Imre Deak
2021-09-21 0:23 ` [Intel-gfx] [PATCH 01/13] drm/i915/tc: Fix TypeC port init/resume time sanitization Imre Deak
2021-09-23 23:10 ` Souza, Jose
2021-09-24 10:59 ` Jani Nikula
2021-09-24 11:06 ` Imre Deak
2021-09-29 13:28 ` [Intel-gfx] [PATCH v2 " Imre Deak
2021-09-29 19:19 ` Souza, Jose
2021-09-21 0:23 ` [Intel-gfx] [PATCH 02/13] drm/i915/adlp/tc: Fix PHY connected check for Thunderbolt mode Imre Deak
2021-09-23 23:18 ` Souza, Jose
2021-09-24 15:24 ` Imre Deak
2021-09-21 0:23 ` [Intel-gfx] [PATCH 03/13] drm/i915/tc: Remove waiting for PHY complete during releasing ownership Imre Deak
2021-09-24 0:17 ` Souza, Jose
2021-09-21 0:23 ` [Intel-gfx] [PATCH 04/13] drm/i915/tc: Check for DP-alt, legacy sinks before taking PHY ownership Imre Deak
2021-09-24 0:30 ` Souza, Jose
2021-09-24 15:31 ` Imre Deak
2021-09-29 13:28 ` [Intel-gfx] [PATCH v2 " Imre Deak
2021-09-21 0:23 ` [Intel-gfx] [PATCH 05/13] drm/i915/tc: Add/use helpers to retrieve TypeC port properties Imre Deak
2021-09-24 19:54 ` Souza, Jose
2021-09-21 0:23 ` [Intel-gfx] [PATCH 06/13] drm/i915/tc: Don't keep legacy TypeC ports in connected state w/o a sink Imre Deak
2021-09-24 19:57 ` Souza, Jose
2021-09-21 0:23 ` [Intel-gfx] [PATCH 07/13] drm/i915/tc: Add a mode for the TypeC PHY's disconnected state Imre Deak
2021-09-27 21:16 ` Souza, Jose
2021-09-27 21:46 ` Imre Deak
2021-09-28 19:18 ` Souza, Jose
2021-09-28 19:34 ` Imre Deak
2021-09-28 19:45 ` Souza, Jose
2021-09-28 19:55 ` Imre Deak
2021-09-28 20:02 ` Souza, Jose
2021-09-28 20:08 ` Imre Deak
2021-09-28 20:29 ` Souza, Jose
2021-09-28 20:38 ` Imre Deak
2021-09-28 20:56 ` Souza, Jose
2021-09-29 13:28 ` [Intel-gfx] [PATCH v2 " Imre Deak
2021-09-21 0:23 ` [Intel-gfx] [PATCH 08/13] drm/i915/tc: Refactor TC-cold block/unblock helpers Imre Deak
2021-09-27 21:56 ` Souza, Jose
2021-09-27 22:13 ` Imre Deak
2021-09-27 22:21 ` Souza, Jose
2021-09-27 22:28 ` Imre Deak [this message]
2021-09-27 23:33 ` Souza, Jose
2021-09-27 23:51 ` Imre Deak
2021-09-28 0:14 ` Souza, Jose
2021-09-28 0:45 ` Imre Deak
2021-09-28 1:03 ` Souza, Jose
2021-09-29 13:28 ` [Intel-gfx] [PATCH v2 " Imre Deak
2021-09-21 0:23 ` [Intel-gfx] [PATCH 09/13] drm/i915/tc: Avoid using legacy AUX PW in TBT mode Imre Deak
2021-09-28 20:31 ` Souza, Jose
2021-09-29 13:28 ` [Intel-gfx] [PATCH v2 " Imre Deak
2021-09-21 0:23 ` [Intel-gfx] [PATCH 10/13] drm/i915/icl/tc: Remove the ICL special casing during TC-cold blocking Imre Deak
2021-09-27 22:02 ` Souza, Jose
2021-09-28 10:52 ` Imre Deak
2021-09-28 20:50 ` Souza, Jose
2021-09-21 0:23 ` [Intel-gfx] [PATCH 11/13] drm/i915/tc: Fix TypeC PHY connect/disconnect logic on ADL-P Imre Deak
2021-09-28 20:51 ` Souza, Jose
2021-09-29 13:28 ` [Intel-gfx] [PATCH v2 " Imre Deak
2021-09-21 0:23 ` [Intel-gfx] [PATCH 12/13] drm/i915/tc: Drop extra TC cold blocking from intel_tc_port_connected() Imre Deak
2021-09-28 20:51 ` Souza, Jose
2021-09-21 0:23 ` [Intel-gfx] [PATCH 13/13] drm/i915/tc: Fix system hang on ADL-P during TypeC PHY disconnect Imre Deak
2021-09-28 20:55 ` Souza, Jose
2021-09-29 13:28 ` [Intel-gfx] [PATCH v2 " Imre Deak
2021-09-21 0:30 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for drm/i915/tc: Fix TypeC connect/disconnect sequences Patchwork
2021-09-21 0:32 ` [Intel-gfx] ✗ Fi.CI.SPARSE: " Patchwork
2021-09-21 1:01 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork
2021-09-21 3:01 ` [Intel-gfx] ✓ Fi.CI.IGT: " Patchwork
2021-09-29 13:42 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for drm/i915/tc: Fix TypeC connect/disconnect sequences (rev8) Patchwork
2021-09-29 13:43 ` [Intel-gfx] ✗ Fi.CI.SPARSE: " Patchwork
2021-09-29 14:11 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork
2021-09-29 16:58 ` [Intel-gfx] ✗ Fi.CI.IGT: failure " Patchwork
2021-09-29 21:16 ` Imre Deak
2021-09-29 22:23 ` Vudum, Lakshminarayana
2021-09-29 21:47 ` Patchwork
2021-09-29 21:54 ` Patchwork
2021-09-29 22:21 ` [Intel-gfx] ✓ Fi.CI.IGT: success " Patchwork
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=20210927222843.GC2043218@ideak-desk.fi.intel.com \
--to=imre.deak@intel.com \
--cc=intel-gfx@lists.freedesktop.org \
--cc=jose.souza@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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.