From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id D5D84C36002 for ; Mon, 24 Mar 2025 13:59:56 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 9ACCB10E413; Mon, 24 Mar 2025 13:59:56 +0000 (UTC) Authentication-Results: gabe.freedesktop.org; dkim=pass (2048-bit key; unprotected) header.d=intel.com header.i=@intel.com header.b="C5VE/4BX"; dkim-atps=neutral Received: from mgamail.intel.com (mgamail.intel.com [192.198.163.9]) by gabe.freedesktop.org (Postfix) with ESMTPS id 64F4810E2FD; Mon, 24 Mar 2025 13:59:55 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1742824795; x=1774360795; h=from:to:cc:subject:in-reply-to:references:date: message-id:mime-version; bh=1bQ4DHykpNOMSEOJFNnc+aaWay084dnBaY3fuQ1lR90=; b=C5VE/4BXAYozisq32EKKA3Uw05gw7HdGY8vOq7GyXs+siPmrOw+j8Hiv /Gq6GXCQSAPYskw6lhel8Fe+mxR0TZ5I5Qp62z0SHEVLO4B5z81glSoeJ wLnBs2NmYEmZmpzP0To/UFah5uDeRoF/Yzmfg4CvBVxQUVVLTNqXWVo/P 1GMcoUoxQOvXaZm4Z08LXlu/bxIYLy35Stg+4DW8vhW0QDSeXiIj2zG0H arH93mX8eFp02Y32Q5yymgJt79nhZQwKAQW+ZHYk34Kz7URihsWTXAOy4 /byq4dON/JWz+btSzUkAnjxlZQMfAyzwXfIW6juixFyVTZGdO3WRjY4Mz w==; X-CSE-ConnectionGUID: Fh3/2RtiSjewa4kf7CqRpQ== X-CSE-MsgGUID: 9oQkSBwzQIW1vodGRw130g== X-IronPort-AV: E=McAfee;i="6700,10204,11383"; a="54665435" X-IronPort-AV: E=Sophos;i="6.14,272,1736841600"; d="scan'208";a="54665435" Received: from orviesa005.jf.intel.com ([10.64.159.145]) by fmvoesa103.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 24 Mar 2025 06:59:55 -0700 X-CSE-ConnectionGUID: CcTNAk5CSc6aKyyhwPTt+w== X-CSE-MsgGUID: ggPMw2+JRgq/jbXownNfdQ== X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="6.14,272,1736841600"; d="scan'208";a="129247550" Received: from kniemiec-mobl1.ger.corp.intel.com (HELO localhost) ([10.245.246.30]) by orviesa005-auth.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 24 Mar 2025 06:59:52 -0700 From: Jani Nikula To: imre.deak@intel.com Cc: intel-gfx@lists.freedesktop.org, intel-xe@lists.freedesktop.org, Ville Syrjala Subject: Re: [PATCH 1/3] drm/i915/pps: Add helpers to lock PPS for AUX transfers In-Reply-To: Organization: Intel Finland Oy - BIC 0357606-4 - Westendinkatu 7, 02160 Espoo References: <20250321145626.94101-1-imre.deak@intel.com> <20250321145626.94101-2-imre.deak@intel.com> <874izibtvx.fsf@intel.com> <87v7rya9zg.fsf@intel.com> Date: Mon, 24 Mar 2025 15:59:50 +0200 Message-ID: <87h63ia5rd.fsf@intel.com> MIME-Version: 1.0 Content-Type: text/plain X-BeenThere: intel-xe@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Intel Xe graphics driver List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: intel-xe-bounces@lists.freedesktop.org Sender: "Intel-xe" On Mon, 24 Mar 2025, Imre Deak wrote: > On Mon, Mar 24, 2025 at 02:28:35PM +0200, Jani Nikula wrote: >> On Mon, 24 Mar 2025, Imre Deak wrote: >> > On Mon, Mar 24, 2025 at 12:33:22PM +0200, Jani Nikula wrote: >> >> On Fri, 21 Mar 2025, Imre Deak wrote: >> >> > Factor out from the DP AUX transfer function the logic to lock/unlock >> >> > the Panel Power Sequencer state and enable/disable the VDD power >> >> > required for the AUX transfer, adding these to helpers in intel_pps.c . >> >> > This prepares for a follow-up change making these steps dependent on the >> >> > platform and output type. >> >> > >> >> > Signed-off-by: Imre Deak >> >> > --- >> >> > drivers/gpu/drm/i915/display/intel_dp_aux.c | 16 ++---------- >> >> > drivers/gpu/drm/i915/display/intel_pps.c | 29 ++++++++++++++++++++- >> >> > drivers/gpu/drm/i915/display/intel_pps.h | 3 ++- >> >> > 3 files changed, 32 insertions(+), 16 deletions(-) >> >> > >> >> > diff --git a/drivers/gpu/drm/i915/display/intel_dp_aux.c b/drivers/gpu/drm/i915/display/intel_dp_aux.c >> >> > index ec27bbd70bcf0..bf5ccfa24ca0b 100644 >> >> > --- a/drivers/gpu/drm/i915/display/intel_dp_aux.c >> >> > +++ b/drivers/gpu/drm/i915/display/intel_dp_aux.c >> >> > @@ -272,15 +272,7 @@ intel_dp_aux_xfer(struct intel_dp *intel_dp, >> >> > aux_domain = intel_aux_power_domain(dig_port); >> >> > >> >> > aux_wakeref = intel_display_power_get(display, aux_domain); >> >> > - pps_wakeref = intel_pps_lock(intel_dp); >> >> > - >> >> > - /* >> >> > - * We will be called with VDD already enabled for dpcd/edid/oui reads. >> >> > - * In such cases we want to leave VDD enabled and it's up to upper layers >> >> > - * to turn it off. But for eg. i2c-dev access we need to turn it on/off >> >> > - * ourselves. >> >> > - */ >> >> > - vdd = intel_pps_vdd_on_unlocked(intel_dp); >> >> > + pps_wakeref = intel_pps_lock_for_aux(intel_dp, &vdd); >> >> > >> >> > /* >> >> > * dp aux is extremely sensitive to irq latency, hence request the >> >> > @@ -289,8 +281,6 @@ intel_dp_aux_xfer(struct intel_dp *intel_dp, >> >> > */ >> >> > cpu_latency_qos_update_request(&intel_dp->pm_qos, 0); >> >> > >> >> > - intel_pps_check_power_unlocked(intel_dp); >> >> > - >> >> > /* >> >> > * FIXME PSR should be disabled here to prevent >> >> > * it using the same AUX CH simultaneously >> >> > @@ -427,10 +417,8 @@ intel_dp_aux_xfer(struct intel_dp *intel_dp, >> >> > out: >> >> > cpu_latency_qos_update_request(&intel_dp->pm_qos, PM_QOS_DEFAULT_VALUE); >> >> > >> >> > - if (vdd) >> >> > - intel_pps_vdd_off_unlocked(intel_dp, false); >> >> > + intel_pps_unlock_for_aux(intel_dp, pps_wakeref, vdd); >> >> > >> >> > - intel_pps_unlock(intel_dp, pps_wakeref); >> >> > intel_display_power_put_async(display, aux_domain, aux_wakeref); >> >> > out_unlock: >> >> > intel_digital_port_unlock(encoder); >> >> > diff --git a/drivers/gpu/drm/i915/display/intel_pps.c b/drivers/gpu/drm/i915/display/intel_pps.c >> >> > index 617ce49931726..3c078fd53fbfa 100644 >> >> > --- a/drivers/gpu/drm/i915/display/intel_pps.c >> >> > +++ b/drivers/gpu/drm/i915/display/intel_pps.c >> >> > @@ -571,7 +571,7 @@ static bool edp_have_panel_vdd(struct intel_dp *intel_dp) >> >> > return intel_de_read(display, _pp_ctrl_reg(intel_dp)) & EDP_FORCE_VDD; >> >> > } >> >> > >> >> > -void intel_pps_check_power_unlocked(struct intel_dp *intel_dp) >> >> > +static void intel_pps_check_power_unlocked(struct intel_dp *intel_dp) >> >> > { >> >> > struct intel_display *display = to_intel_display(intel_dp); >> >> > struct intel_digital_port *dig_port = dp_to_dig_port(intel_dp); >> >> > @@ -955,6 +955,33 @@ void intel_pps_vdd_off(struct intel_dp *intel_dp) >> >> > intel_pps_vdd_off_unlocked(intel_dp, false); >> >> > } >> >> > >> >> > +intel_wakeref_t intel_pps_lock_for_aux(struct intel_dp *intel_dp, bool *vdd_ref) >> >> > +{ >> >> > + intel_wakeref_t wakeref; >> >> > + >> >> > + wakeref = intel_pps_lock(intel_dp); >> >> > + >> >> > + /* >> >> > + * We will be called with VDD already enabled for dpcd/edid/oui reads. >> >> > + * In such cases we want to leave VDD enabled and it's up to upper layers >> >> > + * to turn it off. But for eg. i2c-dev access we need to turn it on/off >> >> > + * ourselves. >> >> > + */ >> >> > + *vdd_ref = intel_pps_vdd_on_unlocked(intel_dp); >> >> > + >> >> > + intel_pps_check_power_unlocked(intel_dp); >> >> > + >> >> > + return wakeref; >> >> > +} >> >> > + >> >> > +void intel_pps_unlock_for_aux(struct intel_dp *intel_dp, intel_wakeref_t wakeref, bool vdd_ref) >> >> > +{ >> >> > + if (vdd_ref) >> >> > + intel_pps_vdd_off_unlocked(intel_dp, false); >> >> > + >> >> > + intel_pps_unlock(intel_dp, wakeref); >> >> > +} >> >> >> >> It took me a while to pinpoint what exactly I don't like about this >> >> interface. >> >> >> >> And I mean the whole intel_pps.h interface is already really difficult >> >> to understand. >> >> >> >> This flips the lock/unlock and vdd on/off logic inside out. >> >> >> >> Normally you have functions for doing vdd or power or backlight, or >> >> anything PPS really, and they're either unlocked (assuming the caller >> >> handles PPS lock) or locked (the function itself takes the lock). >> > >> > The PPS and VDD handling steps are dependent (PPS must be locked for >> > enabling VDD) and both are skipped for the same reason during AUX >> > transfers. So I thought it makes sense to move these to a separate >> > function and skip both based on the same platform/output type check. >> >> On the contrary, I think the reasons are different. >> >> VDD is only needed for eDP. >> >> The PPS must be locked for VDD change (IOW for eDP) and for VLV/CHV pipe >> based PPS. But these two cases are independent. > > The case requiring VDD (eDP) is a subset of the cases requring PPS to be > locked (eDP or VLV/CHV). These are not independent cases. Logically, they are. VLV/CHV requires the PPS lock also for non-eDP. It's not a subset. BR, Jani. > >> >> This one purports to be an interface for lock/unlock, but in reality it >> >> also does VDD internally. And that feels really quite wrong to me. >> >> >> >> --- >> >> >> >> These are a single-use interface that I think make intel_pps.[ch] more >> >> difficult to understand. I'd suggest checking how you'd implement this >> >> logic inside intel_dp_aux_xfer() *without* changing the intel_pps.[ch] >> >> interface at all. >> >> >> >> Okay, took a quick stab at it, and unless I'm missing something it's >> >> super easy: >> > >> > I still think it'd be better to have a separate function for both >> > locking PPS and enabling VDD for the reason I described above, that is >> > to clarify that the PPS state must be locked to enable VDD. >> >> But there's no requirement that they must be done at the same time. > > There is also no reason not do them at the same time for AUX. A benefit > of doing that would be to clarify the dependency of VDD on PPS and also > simplify intel_dp_aux_xfer(). > >> The PPS lock could be held for a much longer period or for other >> things than just VDD. And in this case, the PPS lock may indeed >> protect *other* things than just VDD. Adding the separate function >> ties these unrelated cases together for IMO not good enough reason. >> intel_pps_vdd_on_unlocked() does check that it's called with the PPS >> lock held. >> >> But I realize it needs to be relaxed a bit like this: > > Yes, noticed this too. It was one reason I opted for skipping PPS > locking / VDD enabling from one spot. > >> diff --git a/drivers/gpu/drm/i915/display/intel_pps.c b/drivers/gpu/drm/i915/display/intel_pps.c >> index 617ce4993172..c883e872c9c8 100644 >> --- a/drivers/gpu/drm/i915/display/intel_pps.c >> +++ b/drivers/gpu/drm/i915/display/intel_pps.c >> @@ -744,11 +744,11 @@ bool intel_pps_vdd_on_unlocked(struct intel_dp *intel_dp) >> i915_reg_t pp_stat_reg, pp_ctrl_reg; >> bool need_to_disable = !intel_dp->pps.want_panel_vdd; >> >> - lockdep_assert_held(&display->pps.mutex); >> - >> if (!intel_dp_is_edp(intel_dp)) >> return false; >> >> + lockdep_assert_held(&display->pps.mutex); >> + >> cancel_delayed_work(&intel_dp->pps.panel_vdd_work); >> intel_dp->pps.want_panel_vdd = true; >> >> @@ -925,11 +925,11 @@ void intel_pps_vdd_off_unlocked(struct intel_dp *intel_dp, bool sync) >> { >> struct intel_display *display = to_intel_display(intel_dp); >> >> - lockdep_assert_held(&display->pps.mutex); >> - >> if (!intel_dp_is_edp(intel_dp)) >> return; >> >> + lockdep_assert_held(&display->pps.mutex); >> + >> INTEL_DISPLAY_STATE_WARN(display, !intel_dp->pps.want_panel_vdd, >> "[ENCODER:%d:%s] %s VDD not forced on", >> dp_to_dig_port(intel_dp)->base.base.base.id, >> >> >> > I guess the above could be done separately later in any case, so I can >> > inline the fix as you suggest. >> > >> >> diff --git a/drivers/gpu/drm/i915/display/intel_dp_aux.c b/drivers/gpu/drm/i915/display/intel_dp_aux.c >> >> index ec27bbd70bcf..a5608659df59 100644 >> >> --- a/drivers/gpu/drm/i915/display/intel_dp_aux.c >> >> +++ b/drivers/gpu/drm/i915/display/intel_dp_aux.c >> >> @@ -247,7 +247,7 @@ intel_dp_aux_xfer(struct intel_dp *intel_dp, >> >> u32 aux_clock_divider; >> >> enum intel_display_power_domain aux_domain; >> >> intel_wakeref_t aux_wakeref; >> >> - intel_wakeref_t pps_wakeref; >> >> + intel_wakeref_t pps_wakeref = NULL; >> >> int i, ret, recv_bytes; >> >> int try, clock = 0; >> >> u32 status; >> >> @@ -272,7 +272,10 @@ intel_dp_aux_xfer(struct intel_dp *intel_dp, >> >> aux_domain = intel_aux_power_domain(dig_port); >> >> >> >> aux_wakeref = intel_display_power_get(display, aux_domain); >> >> - pps_wakeref = intel_pps_lock(intel_dp); >> >> + >> >> + if (intel_dp_is_edp(intel_dp) || >> >> + (display->platform.valleyview || display->platform.cherryview)) >> >> + pps_wakeref = intel_pps_lock(intel_dp); >> >> >> >> /* >> >> * We will be called with VDD already enabled for dpcd/edid/oui reads. >> >> @@ -430,7 +433,8 @@ intel_dp_aux_xfer(struct intel_dp *intel_dp, >> >> if (vdd) >> >> intel_pps_vdd_off_unlocked(intel_dp, false); >> >> >> >> - intel_pps_unlock(intel_dp, pps_wakeref); >> >> + if (pps_wakeref) >> >> + intel_pps_unlock(intel_dp, pps_wakeref); >> >> intel_display_power_put_async(display, aux_domain, aux_wakeref); >> >> out_unlock: >> >> intel_digital_port_unlock(encoder); >> >> >> >> >> >> Please let's not make intel_pps.[ch] harder to understand. >> >> >> >> >> >> BR, >> >> Jani. >> >> >> >> >> >> > + >> >> > void intel_pps_on_unlocked(struct intel_dp *intel_dp) >> >> > { >> >> > struct intel_display *display = to_intel_display(intel_dp); >> >> > diff --git a/drivers/gpu/drm/i915/display/intel_pps.h b/drivers/gpu/drm/i915/display/intel_pps.h >> >> > index c83007152f07d..4390d05892325 100644 >> >> > --- a/drivers/gpu/drm/i915/display/intel_pps.h >> >> > +++ b/drivers/gpu/drm/i915/display/intel_pps.h >> >> > @@ -31,10 +31,11 @@ bool intel_pps_vdd_on_unlocked(struct intel_dp *intel_dp); >> >> > void intel_pps_vdd_off_unlocked(struct intel_dp *intel_dp, bool sync); >> >> > void intel_pps_on_unlocked(struct intel_dp *intel_dp); >> >> > void intel_pps_off_unlocked(struct intel_dp *intel_dp); >> >> > -void intel_pps_check_power_unlocked(struct intel_dp *intel_dp); >> >> > >> >> > void intel_pps_vdd_on(struct intel_dp *intel_dp); >> >> > void intel_pps_vdd_off(struct intel_dp *intel_dp); >> >> > +intel_wakeref_t intel_pps_lock_for_aux(struct intel_dp *intel_dp, bool *vdd_ref); >> >> > +void intel_pps_unlock_for_aux(struct intel_dp *intel_dp, intel_wakeref_t wakeref, bool vdd_ref); >> >> > void intel_pps_on(struct intel_dp *intel_dp); >> >> > void intel_pps_off(struct intel_dp *intel_dp); >> >> > void intel_pps_vdd_off_sync(struct intel_dp *intel_dp); >> >> >> >> -- >> >> Jani Nikula, Intel >> >> -- >> Jani Nikula, Intel -- Jani Nikula, Intel