linux-kernel - Re: [PATCH 2/2] clk: mediatek: mt8195-topckgen: Refactor parents for top

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: <CAGXv+5E7YYdkG7GtxG90KzdAG8Kke+74Amtbw4mmyVNZgDZHRA@mail.gmail.com>
Date:   Fri, 14 Jul 2023 12:19:00 +0800
From:   Chen-Yu Tsai <wenst@...omium.org>
To:     Alexandre Mergnat <amergnat@...libre.com>
Cc:     AngeloGioacchino Del Regno 
        <angelogioacchino.delregno@...labora.com>, sboyd@...nel.org,
        mturquette@...libre.com, matthias.bgg@...il.com, msp@...libre.com,
        yangyingliang@...wei.com, u.kleine-koenig@...gutronix.de,
        miles.chen@...iatek.com, linux-clk@...r.kernel.org,
        linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
        linux-mediatek@...ts.infradead.org, kernel@...labora.com,
        mripard@...nel.org
Subject: Re: [PATCH 2/2] clk: mediatek: mt8195-topckgen: Refactor parents for
 top_dp/edp muxes

On Thu, Jul 13, 2023 at 9:22 PM Alexandre Mergnat <amergnat@...libre.com> wrote:
> On 13/07/2023 09:21, AngeloGioacchino Del Regno wrote:
> > The top_dp and top_edp muxes can be both parented to either TVDPLL1
> > or TVDPLL2, two identically specced PLLs for the specific purpose of
> > giving out pixel clock: this becomes a problem when the MediaTek
> > DisplayPort Interface (DPI) driver tries to set the pixel clock rate.
> >
> > In the usecase of two simultaneous outputs (using two controllers),
> > it was seen that one of the displays would sometimes display garbled
> > output (if any at all) and this was because:
> >   - top_edp was set to TVDPLL1, outputting X GHz
> >   - top_dp was set to TVDPLL2, outputting Y GHz
> >     - mtk_dpi calls clk_set_rate(top_edp, Z GHz)
> >       - top_dp is switched to TVDPLL1
> >       - TVDPLL1 changes its rate, top_edp outputs the wrong rate.
> >       - eDP display is garbled
> >
> > To solve this issue, remove all TVDPLL1 parents from `top_dp` and
> > all TVDPLL2 parents from `top_edp`, plus, necessarily switch both
> > clocks to use the new MUX_GATE_CLR_SET_UPD_INDEXED() macro to be
> > able to use the right bit index for the new parents list.
> >
> > Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@...labora.com>
> > ---
> >   drivers/clk/mediatek/clk-mt8195-topckgen.c | 22 ++++++++++++++--------
> >   1 file changed, 14 insertions(+), 8 deletions(-)
> >
> > diff --git a/drivers/clk/mediatek/clk-mt8195-topckgen.c b/drivers/clk/mediatek/clk-mt8195-topckgen.c
> > index 81daa24cadde..abb3721f6e1b 100644
> > --- a/drivers/clk/mediatek/clk-mt8195-topckgen.c
> > +++ b/drivers/clk/mediatek/clk-mt8195-topckgen.c
> > @@ -417,15 +417,21 @@ static const char * const pwrmcu_parents[] = {
> >
> >   static const char * const dp_parents[] = {
> >       "clk26m",
> > -     "tvdpll1_d2",
> >       "tvdpll2_d2",
> > -     "tvdpll1_d4",
> >       "tvdpll2_d4",
> > -     "tvdpll1_d8",
> >       "tvdpll2_d8",
> > -     "tvdpll1_d16",
> >       "tvdpll2_d16"
> >   };
> > +static const u8 dp_parents_idx[] = { 0, 2, 4, 6, 8 };
> > +
> > +static const char * const edp_parents[] = {
> > +     "clk26m",
> > +     "tvdpll1_d2",
> > +     "tvdpll1_d4",
> > +     "tvdpll1_d8",
> > +     "tvdpll1_d16"
> > +};
> > +static const u8 edp_parents_idx[] = { 0, 1, 3, 5, 7 };
>
> AFAII your solution is to force a specific TVDPLLX for each display, and
> it isn't dynamic.
>
> Do you think it's possible to do that using the DTS ? I'm asking
> because, IMHO, this kind of setup is more friendly/readable/flexible in
> the DTS than hardcoded into the driver.

(CC-ing Maxime, who has some experience in the matter.)

assigned-parents doesn't prevent your system from reparenting the clocks
back to a conflicting configuration.

AFAIK the recommended way to deal with this is to use
clk_set_rate_exclusive() and co. in whatever consumer driver that needs
exclusive control on the clock rate. However I'm not sure if that works
for parents. It should, given the original use case was for the sunxi
platforms, which like the MediaTek platform here has 2 PLLs for video
related consumers, but I couldn't find code verifying it.


ChenYu

> >
> >   static const char * const disp_pwm_parents[] = {
> >       "clk26m",
> > @@ -957,11 +963,11 @@ static const struct mtk_mux top_mtk_muxes[] = {
> >       MUX_GATE_CLR_SET_UPD_FLAGS(CLK_TOP_PWRMCU, "top_pwrmcu",
> >               pwrmcu_parents, 0x08C, 0x090, 0x094, 16, 3, 23, 0x08, 6,
> >               CLK_IS_CRITICAL | CLK_SET_RATE_PARENT),
> > -     MUX_GATE_CLR_SET_UPD(CLK_TOP_DP, "top_dp",
> > -             dp_parents, 0x08C, 0x090, 0x094, 24, 4, 31, 0x08, 7),
> > +     MUX_GATE_CLR_SET_UPD_INDEXED(CLK_TOP_DP, "top_dp",
> > +             dp_parents, dp_parents_idx, 0x08C, 0x090, 0x094, 24, 4, 31, 0x08, 7),
> >       /* CLK_CFG_10 */
> > -     MUX_GATE_CLR_SET_UPD(CLK_TOP_EDP, "top_edp",
> > -             dp_parents, 0x098, 0x09C, 0x0A0, 0, 4, 7, 0x08, 8),
> > +     MUX_GATE_CLR_SET_UPD_INDEXED(CLK_TOP_EDP, "top_edp",
> > +             edp_parents, edp_parents_idx, 0x098, 0x09C, 0x0A0, 0, 4, 7, 0x08, 8),
> >       MUX_GATE_CLR_SET_UPD(CLK_TOP_DPI, "top_dpi",
> >               dp_parents, 0x098, 0x09C, 0x0A0, 8, 4, 15, 0x08, 9),
> >       MUX_GATE_CLR_SET_UPD(CLK_TOP_DISP_PWM0, "top_disp_pwm0",
>
> --
> Regards,
> Alexandre