lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20181025181017.GL154160@art_vandelay>
Date:   Thu, 25 Oct 2018 14:10:17 -0400
From:   Sean Paul <sean@...rly.run>
To:     Douglas Anderson <dianders@...omium.org>
Cc:     Sean Paul <seanpaul@...omium.org>,
        Thierry Reding <thierry.reding@...il.com>,
        Sandeep Panda <spanda@...eaurora.org>,
        linux-arm-msm@...r.kernel.org,
        Laurent Pinchart <laurent.pinchart@...asonboard.com>,
        jsanka@...eaurora.org, ryandcase@...omium.org,
        Andrzej Hajda <a.hajda@...sung.com>,
        Archit Taneja <architt@...eaurora.org>,
        linux-kernel@...r.kernel.org, dri-devel@...ts.freedesktop.org,
        David Airlie <airlied@...ux.ie>
Subject: Re: [PATCH 4/6] drm/bridge: ti-sn65dsi86: Remove the mystery delay

On Mon, Oct 22, 2018 at 01:46:37PM -0700, Douglas Anderson wrote:
> Let's solve the mystery of commit bf1178c98930 ("drm/bridge:
> ti-sn65dsi86: Add mystery delay to enable()").  Specifically the
> reason we needed that mystery delay is that we weren't paying
> attention to HPD.
> 
> Looking at the datasheet for the same panel that was tested for the
> original commit, I see there's a timing "t3" that times from power on
> to the aux channel being operational.  This time is specced as 0 - 200
> ms.  The datasheet says that the aux channel is operational at exactly
> the same time that HPD is asserted.
> 
> Scoping the signals on this board showed that HPD was asserted 84 ms
> after power was asserted.  That very closely matches the magic 70 ms
> delay that we had.  ...and actually, in my testing the 70 ms wasn't
> quite enough of a delay and some percentage of the time the display
> didn't come up until I bumped it to 100 ms (presumably 84 ms would
> have worked too).
> 
> To solve this, we tried to hook up the HPD signal in the bridge.
> ...but in doing so we found that that the bridge didn't report that
> HPD was asserted until ~280 ms after we powered it (!).  This is
> explained by looking at the sn65dsi86 datasheet section "8.4.5.1 HPD
> (Hot Plug/Unplug Detection)".  Reading there we see that the bridge
> isn't even intended to report HPD until 100 ms after it's asserted.
> ...but that would have left us at 184 ms.  The extra 100 ms
> (presumably) comes from this part in the datasheet:
> 
> > The HPD state machine operates off an internal ring oscillator. The
> > ring oscillator frequency will vary [ ... ]. The min/max range in
> > the HPD State Diagram refers to the possible times based off
> > variation in the ring oscillator frequency.
> 
> Given that the 280 ms we'll end up delaying if we hook up HPD is
> _slower_ than the 200 ms we could just hardcode, for now we'll solve
> the problem by just hardcoding a 200 ms delay in the panel driver
> using the patch in this series ("drm/panel: simple: Support panels
> with HPD where HPD isn't connected").
> 
> If we later find a panel that needs to use this bridge where we need
> HPD then we'll have to come up with some new code to handle it.  Given
> the silly debouncing in the bridge chip, though, it seems unlikely.
> 
> One last note is that I tried to solve this through another way: In
> ti_sn_bridge_enable() I tried to use various combinations of
> dp_dpcd_writeb() and dp_dpcd_readb() to detect when the aux channel
> was up.  In theory that would let me detect _exactly_ when I could
> continue and do link training.  Unfortunately even if I did an aux
> transfer w/out waiting I couldn't see any errors.  Possibly I could
> keep looping over link training until it came back with success, but
> that seemed a little overly hacky to me.
> 
> Signed-off-by: Douglas Anderson <dianders@...omium.org>

Awesome commit message and comment, thanks for solving the mystery!

Reviewed-by: Sean Paul <sean@...rly.run>


> ---
> 
>  drivers/gpu/drm/bridge/ti-sn65dsi86.c | 29 +++++++++++++++------------
>  1 file changed, 16 insertions(+), 13 deletions(-)
> 
> diff --git a/drivers/gpu/drm/bridge/ti-sn65dsi86.c b/drivers/gpu/drm/bridge/ti-sn65dsi86.c
> index f8a931cf3665..680566d97adc 100644
> --- a/drivers/gpu/drm/bridge/ti-sn65dsi86.c
> +++ b/drivers/gpu/drm/bridge/ti-sn65dsi86.c
> @@ -458,18 +458,6 @@ static void ti_sn_bridge_enable(struct drm_bridge *bridge)
>  	unsigned int val;
>  	int ret;
>  
> -	/*
> -	 * FIXME:
> -	 * This 70ms was found necessary by experimentation. If it's not
> -	 * present, link training fails. It seems like it can go anywhere from
> -	 * pre_enable() up to semi-auto link training initiation below.
> -	 *
> -	 * Neither the datasheet for the bridge nor the panel tested mention a
> -	 * delay of this magnitude in the timing requirements. So for now, add
> -	 * the mystery delay until someone figures out a better fix.
> -	 */
> -	msleep(70);
> -
>  	/* DSI_A lane config */
>  	val = CHA_DSI_LANES(4 - pdata->dsi->lanes);
>  	regmap_update_bits(pdata->regmap, SN_DSI_LANES_REG,
> @@ -536,7 +524,22 @@ static void ti_sn_bridge_pre_enable(struct drm_bridge *bridge)
>  	/* configure bridge ref_clk */
>  	ti_sn_bridge_set_refclk_freq(pdata);
>  
> -	/* in case drm_panel is connected then HPD is not supported */
> +	/*
> +	 * HPD on this bridge chip is a bit useless.  This is an eDP bridge
> +	 * so the HPD is an internal signal that's only there to signal that
> +	 * the panel is done powering up.  ...but the bridge chip debounces
> +	 * this signal by between 100 ms and 400 ms (depending on process,
> +	 * voltage, and temperate--I measured it at about 200 ms).  One
> +	 * particular panel asserted HPD 84 ms after it was powered on meaning
> +	 * that we saw HPD 284 ms after power on.  ...but the same panel said
> +	 * that instead of looking at HPD you could just hardcode a delay of
> +	 * 200 ms.  We'll assume that the panel driver will have the hardcoded
> +	 * delay in its prepare and always disable HPD.
> +	 *
> +	 * If HPD somehow makes sense on some future panel we'll have to
> +	 * change this to be conditional on someone specifying that HPD should
> +	 * be used.
> +	 */
>  	regmap_update_bits(pdata->regmap, SN_HPD_DISABLE_REG, HPD_DISABLE,
>  			   HPD_DISABLE);
>  
> -- 
> 2.19.1.568.g152ad8e336-goog
> 

-- 
Sean Paul, Software Engineer, Google / Chromium OS

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ