Re: clk/clk-next boot bisection: v5.4-rc1-201-ga228ae437aa5 on tegra124-nyan-big

[Dmitry Osipenko <digetx@gmail.com>]

14.11.2019 15:15, Thierry Reding пишет:
> On Thu, Nov 14, 2019 at 02:56:20PM +0300, Dmitry Osipenko wrote:
>> 14.11.2019 07:59, kernelci.org bot пишет:
>>> * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
>>> * This automated bisection report was sent to you on the basis  *
>>> * that you may be involved with the breaking commit it has      *
>>> * found.  No manual investigation has been done to verify it,   *
>>> * and the root cause of the problem may be somewhere else.      *
>>> *                                                               *
>>> * If you do send a fix, please include this trailer:            *
>>> *   Reported-by: "kernelci.org bot" <bot@kernelci.org>          *
>>> *                                                               *
>>> * Hope this helps!                                              *
>>> * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
>>>
>>> clk/clk-next boot bisection: v5.4-rc1-201-ga228ae437aa5 on tegra124-nyan-big
>>>
>>> Summary:
>>>   Start:      a228ae437aa5 Merge branch 'clk-unused' into clk-next
>>>   Details:    https://kernelci.org/boot/id/5dcc99e959b514100f138e14
>>>   Plain log:  https://storage.kernelci.org//clk/clk-next/v5.4-rc1-201-ga228ae437aa5/arm/tegra_defconfig/gcc-8/lab-collabora/boot-tegra124-nyan-big.txt
>>>   HTML log:   https://storage.kernelci.org//clk/clk-next/v5.4-rc1-201-ga228ae437aa5/arm/tegra_defconfig/gcc-8/lab-collabora/boot-tegra124-nyan-big.html
>>>   Result:     25175c806a68 clk: tegra: Reimplement SOR clock on Tegra124
>>>
>>> Checks:
>>>   revert:     PASS
>>>   verify:     PASS
>>>
>>> Parameters:
>>>   Tree:       clk
>>>   URL:        https://git.kernel.org/pub/scm/linux/kernel/git/clk/linux.git
>>>   Branch:     clk-next
>>>   Target:     tegra124-nyan-big
>>>   CPU arch:   arm
>>>   Lab:        lab-collabora
>>>   Compiler:   gcc-8
>>>   Config:     tegra_defconfig
>>>   Test suite: boot
>>>
>>> Breaking commit found:
>>>
>>> -------------------------------------------------------------------------------
>>> commit 25175c806a6841149abe46168e0af12593141612
>>> Author: Thierry Reding <treding@nvidia.com>
>>> Date:   Thu Jul 25 18:19:00 2019 +0200
>>>
>>>     clk: tegra: Reimplement SOR clock on Tegra124
>>>     
>>>     In order to allow the display driver to deal uniformly with all SOR
>>>     generations, implement the SOR clocks in a way that is compatible with
>>>     Tegra186 and later.
>>>     
>>>     Acked-by: Stephen Boyd <sboyd@kernel.org>
>>>     Signed-off-by: Thierry Reding <treding@nvidia.com>
>>>
>>> diff --git a/drivers/clk/tegra/clk-tegra124.c b/drivers/clk/tegra/clk-tegra124.c
>>> index 7d231529c3a5..b3110d5b5a6c 100644
>>> --- a/drivers/clk/tegra/clk-tegra124.c
>>> +++ b/drivers/clk/tegra/clk-tegra124.c
>>> @@ -1005,20 +1005,24 @@ static struct tegra_devclk devclks[] __initdata = {
>>>  	{ .con_id = "hda2hdmi", .dt_id = TEGRA124_CLK_HDA2HDMI },
>>>  };
>>>  
>>> -static const char *mux_pllp_pllm_plld_plla_pllc_plld2_clkm[] = {
>>> -	"pll_p", "pll_m", "pll_d_out0", "pll_a_out0", "pll_c",
>>> -	"pll_d2_out0", "clk_m"
>>> +static const char * const sor0_parents[] = {
>>> +	"pll_p_out0", "pll_m_out0", "pll_d_out0", "pll_a_out0", "pll_c_out0",
>>> +	"pll_d2_out0", "clk_m",
>>>  };
>>
>> I'm not sure how it supposed to work because looks like some of these
>> "_out0" parent clocks just do not exists in the upstream clk driver.
>>
>> Either some other patch that adds "_out0" aliases is missed or this
>> patch was based on top of a different codebase (downstream kernel?).
> 
> The pll_d{,2}_out0 clocks are registered as fixed-factor clocks. The
> others I'm not exactly sure, but I think I ended up taking these
> directly from the register tables in the TRM (the clock mux fields).
> Looks like we ended up naming these differently in the kernel driver
> when it was merged initially.
> 
> The reason why this works is because the common clock framework will not
> flag these clocks as missing since they may appear at some later point
> (perhaps by some different provider).

Right.

> In this case it doesn't really matter because the only parents that are
> ever used or the clk_m, pll_d_out0 and pll_d2_out0.

Okay.

> It might be a good idea to clean some of this up at some point, though.

Indeed.

>> What's also a bit odd is that commit says Tegra186.. but AFAIK there is
>> no driver for Tegra186 clocks in upstream.
> 
> Clocks on Tegra186 are provided by the clk-bpmp driver. Basically those
> are mostly the same clocks but the implementation is provided by the
> BPMP and the clk-bpmp driver talks to it via an IPC mechanism.

I see how it works now, thanks for the clarification.

> So basically the bottom line here is that the SOR clocks were not
> correctly implemented on chips prior to Tegra186 which leads to obscure
> bugs with HDMI and DP outputs.

Thank you for the explanation.

>> BTW, seems T114 clk driver also uses a non-existent pll_m/p/c_out0 for
>> the emc_mux.
>>
>> Thierry / Peter, could you please explain what's going on with this
>> patch and why clocks are missed on T114? Maybe I'm missing something?
>>
>>> -#define mux_pllp_pllm_plld_plla_pllc_plld2_clkm_idx NULL
>>>  
>>> -static const char *mux_clkm_plldp_sor0out[] = {
>>> -	"clk_m", "pll_dp", "sor0_out",
>>> +static const char * const sor0_out_parents[] = {
>>> +	"clk_m", "sor0_pad_clkout",
>>>  };
>>> -#define mux_clkm_plldp_sor0out_idx NULL
>>>  
>>>  static struct tegra_periph_init_data tegra124_periph[] = {
>>> -	MUX8_NOGATE_LOCK("sor0_out", mux_pllp_pllm_plld_plla_pllc_plld2_clkm, CLK_SOURCE_SOR0, tegra_clk_sor0_out, &sor0_lock),
>>> -	NODIV("sor0", mux_clkm_plldp_sor0out, CLK_SOURCE_SOR0, 14, 3, 182, 0, tegra_clk_sor0, &sor0_lock),
>>> +	TEGRA_INIT_DATA_TABLE("sor0", NULL, NULL, sor0_parents,
>>> +			      CLK_SOURCE_SOR0, 29, 0x7, 0, 0, 0, 0,
>>> +			      0, 182, 0, tegra_clk_sor0, NULL, 0,
>>> +			      &sor0_lock),
>>> +	TEGRA_INIT_DATA_TABLE("sor0_out", NULL, NULL, sor0_out_parents,
>>> +			      CLK_SOURCE_SOR0, 14, 0x1, 0, 0, 0, 0,
>>> +			      0, 0, TEGRA_PERIPH_NO_GATE, tegra_clk_sor0_out,
>>> +			      NULL, 0, &sor0_lock),
>>>  };
>>>  
>>>  static struct clk **clks;
>>> -------------------------------------------------------------------------------
>>>
>>>
>>> Git bisection log:
>>>
>>> -------------------------------------------------------------------------------
>>> git bisect start
>>> # good: [dafbb1e6473788cb3068eaeddc58f9d88e9c7a62] Merge branch 'clk-ti' into clk-next
>>> git bisect good dafbb1e6473788cb3068eaeddc58f9d88e9c7a62
>>> # bad: [a228ae437aa553736058cbbd58d2d0e191635cdc] Merge branch 'clk-unused' into clk-next
>>> git bisect bad a228ae437aa553736058cbbd58d2d0e191635cdc
>>> # bad: [3214be6cb1e487b0f8c3bb2eac9b06df07a07e06] clk: tegra: Share clk and rst register defines with Tegra clock driver
>>> git bisect bad 3214be6cb1e487b0f8c3bb2eac9b06df07a07e06
>>> # bad: [05308d7e7bbc932025f1dafc401c73ce83c6f414] clk: tegra: Reimplement SOR clocks on Tegra210
>>> git bisect bad 05308d7e7bbc932025f1dafc401c73ce83c6f414
>>> # good: [d1ee3173a139ed2eb8d87e06216f0426b16084d8] Merge branch 'for-5.5/dt-bindings' into for-5.5/clk
>>> git bisect good d1ee3173a139ed2eb8d87e06216f0426b16084d8
>>> # good: [e5f8a107d92db30a7ad7d8d95aee59f5ad76206a] clk: tegra: Move SOR0 implementation to Tegra124
>>> git bisect good e5f8a107d92db30a7ad7d8d95aee59f5ad76206a
>>> # bad: [25175c806a6841149abe46168e0af12593141612] clk: tegra: Reimplement SOR clock on Tegra124
>>> git bisect bad 25175c806a6841149abe46168e0af12593141612
>>> # good: [da8d1a3555406275650b366460c6235f1696bf8b] clk: tegra: Rename sor0_lvds to sor0_out
>>> git bisect good da8d1a3555406275650b366460c6235f1696bf8b
>>> # first bad commit: [25175c806a6841149abe46168e0af12593141612] clk: tegra: Reimplement SOR clock on Tegra124
>>> -------------------------------------------------------------------------------
>>>
>>