[PATCH 1/2] clk: rockchip: mark rk3399 GIC clocks as critical

[PATCH 1/2] clk: rockchip: mark rk3399 GIC clocks as critical

[Brian Norris]

We never want to kill the GIC.

Noticed when making other clock fixups, and seeing the newly-constructed
clock tree try to disable cpll, where we had this parent structure:

  aclk_gic <------\
                  |--- aclk_gic_pre <-- cpll <-- pll_cpll
  aclk_gic_noc <--/

Signed-off-by: Brian Norris <briannorris@chromium.org>
---
 drivers/clk/rockchip/clk-rk3399.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/clk/rockchip/clk-rk3399.c b/drivers/clk/rockchip/clk-rk3399.c
index 291543f52caa..145756c4f3c8 100644
--- a/drivers/clk/rockchip/clk-rk3399.c
+++ b/drivers/clk/rockchip/clk-rk3399.c
@@ -1466,6 +1466,8 @@ static struct rockchip_clk_branch rk3399_clk_pmu_branches[] __initdata = {
 
 static const char *const rk3399_cru_critical_clocks[] __initconst = {
 	"aclk_cci_pre",
+	"aclk_gic",
+	"aclk_gic_noc",
 	"pclk_perilp0",
 	"pclk_perilp0",
 	"hclk_perilp0",
-- 
2.8.0.rc3.226.g39d4020

[PATCH 2/2] clk: rockchip: fix incorrect parent for rk3399's {c,g}pll_aclk_perihp_src

[Brian Norris]

From: Xing Zheng <zhengxing@rock-chips.com>

There was a typo, swapping 'c' <--> 'g'.

Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
Signed-off-by: Brian Norris <briannorris@chromium.org>
---
 drivers/clk/rockchip/clk-rk3399.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/clk/rockchip/clk-rk3399.c b/drivers/clk/rockchip/clk-rk3399.c
index 145756c4f3c8..9f86bfef70f7 100644
--- a/drivers/clk/rockchip/clk-rk3399.c
+++ b/drivers/clk/rockchip/clk-rk3399.c
@@ -832,9 +832,9 @@ static struct rockchip_clk_branch rk3399_clk_branches[] __initdata = {
 			RK3399_CLKGATE_CON(13), 1, GFLAGS),
 
 	/* perihp */
-	GATE(0, "cpll_aclk_perihp_src", "gpll", CLK_IGNORE_UNUSED,
+	GATE(0, "cpll_aclk_perihp_src", "cpll", CLK_IGNORE_UNUSED,
 			RK3399_CLKGATE_CON(5), 0, GFLAGS),
-	GATE(0, "gpll_aclk_perihp_src", "cpll", CLK_IGNORE_UNUSED,
+	GATE(0, "gpll_aclk_perihp_src", "gpll", CLK_IGNORE_UNUSED,
 			RK3399_CLKGATE_CON(5), 1, GFLAGS),
 	COMPOSITE(ACLK_PERIHP, "aclk_perihp", mux_aclk_perihp_p, CLK_IGNORE_UNUSED,
 			RK3399_CLKSEL_CON(14), 7, 1, MFLAGS, 0, 5, DFLAGS,
-- 
2.8.0.rc3.226.g39d4020

Re: [PATCH 1/2] clk: rockchip: mark rk3399 GIC clocks as critical

[Doug Anderson]

Brian,

On Fri, May 13, 2016 at 11:42 AM, Brian Norris <briannorris@chromium.org> wrote:
> We never want to kill the GIC.
>
> Noticed when making other clock fixups, and seeing the newly-constructed
> clock tree try to disable cpll, where we had this parent structure:
>
>   aclk_gic <------\
>                   |--- aclk_gic_pre <-- cpll <-- pll_cpll
>   aclk_gic_noc <--/
>
> Signed-off-by: Brian Norris <briannorris@chromium.org>
> ---
>  drivers/clk/rockchip/clk-rk3399.c | 2 ++
>  1 file changed, 2 insertions(+)

Reviewed-by: Douglas Anderson <dianders@chromium.org>

Re: [PATCH 2/2] clk: rockchip: fix incorrect parent for rk3399's {c,g}pll_aclk_perihp_src

[Doug Anderson]

Hi,

On Fri, May 13, 2016 at 11:42 AM, Brian Norris <briannorris@chromium.org> wrote:
> From: Xing Zheng <zhengxing@rock-chips.com>
>
> There was a typo, swapping 'c' <--> 'g'.
>
> Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
> Signed-off-by: Brian Norris <briannorris@chromium.org>
> ---
>  drivers/clk/rockchip/clk-rk3399.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/clk/rockchip/clk-rk3399.c b/drivers/clk/rockchip/clk-rk3399.c
> index 145756c4f3c8..9f86bfef70f7 100644
> --- a/drivers/clk/rockchip/clk-rk3399.c
> +++ b/drivers/clk/rockchip/clk-rk3399.c
> @@ -832,9 +832,9 @@ static struct rockchip_clk_branch rk3399_clk_branches[] __initdata = {
>                         RK3399_CLKGATE_CON(13), 1, GFLAGS),
>
>         /* perihp */
> -       GATE(0, "cpll_aclk_perihp_src", "gpll", CLK_IGNORE_UNUSED,
> +       GATE(0, "cpll_aclk_perihp_src", "cpll", CLK_IGNORE_UNUSED,
>                         RK3399_CLKGATE_CON(5), 0, GFLAGS),
> -       GATE(0, "gpll_aclk_perihp_src", "cpll", CLK_IGNORE_UNUSED,
> +       GATE(0, "gpll_aclk_perihp_src", "gpll", CLK_IGNORE_UNUSED,
>                         RK3399_CLKGATE_CON(5), 1, GFLAGS),
>         COMPOSITE(ACLK_PERIHP, "aclk_perihp", mux_aclk_perihp_p, CLK_IGNORE_UNUSED,
>                         RK3399_CLKSEL_CON(14), 7, 1, MFLAGS, 0, 5, DFLAGS,

Definitely there was a bug since this table itself was inconsistent.
...and I _think_ this fix is correct, but I'll note that the TRM has
more inconsistency here.

In the big clock table 'CRU Clock Architecture Diagram', I see:
  CLK 4 is CPLL
  CLK 5 is GPLL
  CLK 125 is aclk_perihp_cpll_src and has 4 (CPLL) as source, with
g5[0] as the bit
  CLK 126 is aclk_perihp_gpll_src and has 5 (GPLL) as source, with
g5[1] as the bit

In the definition of CRU_CLKGATE_CON5:
  bit 0 shows aclk_perihp_gpll_src_en
  bit 1 shows aclk_perihp_cpll_src_en


Thus the table shows CPLL as gate5[0] and GPLL as gate5[1].  The
register definition shows the opposite.  I'll tend to believe the
table over the register definition, but I figured I'd bring it up
anyway.


Xing Zheng: can you confirm that the table is correct and ask
documentation folks to fix the register definition for
CRU_CLKGATE_CON5?



Thanks!

-Doug

Re: [PATCH 2/2] clk: rockchip: fix incorrect parent for rk3399's {c,g}pll_aclk_perihp_src

[Xing Zheng]

Hi Doug,

On 2016年05月14日 04:10, Doug Anderson wrote:
> Hi,
>
> On Fri, May 13, 2016 at 11:42 AM, Brian Norris <briannorris@chromium.org> wrote:
>> From: Xing Zheng <zhengxing@rock-chips.com>
>>
>> There was a typo, swapping 'c' <--> 'g'.
>>
>> Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
>> Signed-off-by: Brian Norris <briannorris@chromium.org>
>> ---
>>   drivers/clk/rockchip/clk-rk3399.c | 4 ++--
>>   1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/clk/rockchip/clk-rk3399.c b/drivers/clk/rockchip/clk-rk3399.c
>> index 145756c4f3c8..9f86bfef70f7 100644
>> --- a/drivers/clk/rockchip/clk-rk3399.c
>> +++ b/drivers/clk/rockchip/clk-rk3399.c
>> @@ -832,9 +832,9 @@ static struct rockchip_clk_branch rk3399_clk_branches[] __initdata = {
>>                          RK3399_CLKGATE_CON(13), 1, GFLAGS),
>>
>>          /* perihp */
>> -       GATE(0, "cpll_aclk_perihp_src", "gpll", CLK_IGNORE_UNUSED,
>> +       GATE(0, "cpll_aclk_perihp_src", "cpll", CLK_IGNORE_UNUSED,
>>                          RK3399_CLKGATE_CON(5), 0, GFLAGS),
>> -       GATE(0, "gpll_aclk_perihp_src", "cpll", CLK_IGNORE_UNUSED,
>> +       GATE(0, "gpll_aclk_perihp_src", "gpll", CLK_IGNORE_UNUSED,
>>                          RK3399_CLKGATE_CON(5), 1, GFLAGS),
>>          COMPOSITE(ACLK_PERIHP, "aclk_perihp", mux_aclk_perihp_p, CLK_IGNORE_UNUSED,
>>                          RK3399_CLKSEL_CON(14), 7, 1, MFLAGS, 0, 5, DFLAGS,
> Definitely there was a bug since this table itself was inconsistent.
> ...and I _think_ this fix is correct, but I'll note that the TRM has
> more inconsistency here.
>
> In the big clock table 'CRU Clock Architecture Diagram', I see:
>    CLK 4 is CPLL
>    CLK 5 is GPLL
>    CLK 125 is aclk_perihp_cpll_src and has 4 (CPLL) as source, with
> g5[0] as the bit
>    CLK 126 is aclk_perihp_gpll_src and has 5 (GPLL) as source, with
> g5[1] as the bit
>
> In the definition of CRU_CLKGATE_CON5:
>    bit 0 shows aclk_perihp_gpll_src_en
>    bit 1 shows aclk_perihp_cpll_src_en
>
>
> Thus the table shows CPLL as gate5[0] and GPLL as gate5[1].  The
> register definition shows the opposite.  I'll tend to believe the
> table over the register definition, but I figured I'd bring it up
> anyway.
>
>
> Xing Zheng: can you confirm that the table is correct and ask
> documentation folks to fix the register definition for
> CRU_CLKGATE_CON5?
Yes, previously, our IC & DOC partner confirmed that the definition of 
CRU_CLKGATE_CON5 should be:
   bit 0 shows aclk_perihp_cpll_src_en
   bit 1 shows aclk_perihp_gpll_src_en

Sorry to the incorrect register definition, we will fix them and review 
the TRM again.

Thanks.

-- 
- Xing Zheng

Re: [PATCH 2/2] clk: rockchip: fix incorrect parent for rk3399's {c,g}pll_aclk_perihp_src

[Doug Anderson]

Hi,

On Fri, May 13, 2016 at 8:36 PM, Xing Zheng <zhengxing@rock-chips.com> wrote:
> Hi Doug,
>
>
> On 2016年05月14日 04:10, Doug Anderson wrote:
>>
>> Hi,
>>
>> On Fri, May 13, 2016 at 11:42 AM, Brian Norris <briannorris@chromium.org>
>> wrote:
>>>
>>> From: Xing Zheng <zhengxing@rock-chips.com>
>>>
>>> There was a typo, swapping 'c' <--> 'g'.
>>>
>>> Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
>>> Signed-off-by: Brian Norris <briannorris@chromium.org>
>>> ---
>>>   drivers/clk/rockchip/clk-rk3399.c | 4 ++--
>>>   1 file changed, 2 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/clk/rockchip/clk-rk3399.c
>>> b/drivers/clk/rockchip/clk-rk3399.c
>>> index 145756c4f3c8..9f86bfef70f7 100644
>>> --- a/drivers/clk/rockchip/clk-rk3399.c
>>> +++ b/drivers/clk/rockchip/clk-rk3399.c
>>> @@ -832,9 +832,9 @@ static struct rockchip_clk_branch
>>> rk3399_clk_branches[] __initdata = {
>>>                          RK3399_CLKGATE_CON(13), 1, GFLAGS),
>>>
>>>          /* perihp */
>>> -       GATE(0, "cpll_aclk_perihp_src", "gpll", CLK_IGNORE_UNUSED,
>>> +       GATE(0, "cpll_aclk_perihp_src", "cpll", CLK_IGNORE_UNUSED,
>>>                          RK3399_CLKGATE_CON(5), 0, GFLAGS),
>>> -       GATE(0, "gpll_aclk_perihp_src", "cpll", CLK_IGNORE_UNUSED,
>>> +       GATE(0, "gpll_aclk_perihp_src", "gpll", CLK_IGNORE_UNUSED,
>>>                          RK3399_CLKGATE_CON(5), 1, GFLAGS),
>>>          COMPOSITE(ACLK_PERIHP, "aclk_perihp", mux_aclk_perihp_p,
>>> CLK_IGNORE_UNUSED,
>>>                          RK3399_CLKSEL_CON(14), 7, 1, MFLAGS, 0, 5,
>>> DFLAGS,
>>
>> Definitely there was a bug since this table itself was inconsistent.
>> ...and I _think_ this fix is correct, but I'll note that the TRM has
>> more inconsistency here.
>>
>> In the big clock table 'CRU Clock Architecture Diagram', I see:
>>    CLK 4 is CPLL
>>    CLK 5 is GPLL
>>    CLK 125 is aclk_perihp_cpll_src and has 4 (CPLL) as source, with
>> g5[0] as the bit
>>    CLK 126 is aclk_perihp_gpll_src and has 5 (GPLL) as source, with
>> g5[1] as the bit
>>
>> In the definition of CRU_CLKGATE_CON5:
>>    bit 0 shows aclk_perihp_gpll_src_en
>>    bit 1 shows aclk_perihp_cpll_src_en
>>
>>
>> Thus the table shows CPLL as gate5[0] and GPLL as gate5[1].  The
>> register definition shows the opposite.  I'll tend to believe the
>> table over the register definition, but I figured I'd bring it up
>> anyway.
>>
>>
>> Xing Zheng: can you confirm that the table is correct and ask
>> documentation folks to fix the register definition for
>> CRU_CLKGATE_CON5?
>
> Yes, previously, our IC & DOC partner confirmed that the definition of
> CRU_CLKGATE_CON5 should be:
>   bit 0 shows aclk_perihp_cpll_src_en
>   bit 1 shows aclk_perihp_gpll_src_en
>
> Sorry to the incorrect register definition, we will fix them and review the
> TRM again.

Great!

Since we now have extra confirmation that Brian's patch is indeed correct:

Reviewed-by: Douglas Anderson <dianders@chromium.org>

Re: [PATCH 1/2] clk: rockchip: mark rk3399 GIC clocks as critical

[Heiko Stuebner]

Am Freitag, 13. Mai 2016, 11:42:16 schrieb Brian Norris:
> We never want to kill the GIC.
> 
> Noticed when making other clock fixups, and seeing the newly-constructed
> clock tree try to disable cpll, where we had this parent structure:
> 
>   aclk_gic <------\
> 
>                   |--- aclk_gic_pre <-- cpll <-- pll_cpll
> 
>   aclk_gic_noc <--/
> 
> Signed-off-by: Brian Norris <briannorris@chromium.org>

applied both patches to my clk-fixes branch with Dougs review-tags.


Thanks
Heiko

Re: [PATCH 2/2] clk: rockchip: fix incorrect parent for rk3399's {c,g}pll_aclk_perihp_src

[Xing Zheng]

Hi Doug,
     It is so sorry that our IC folks re-correct these bits, they need 
to be recovered like these:
----
In the definition of CRU_CLKGATE_CON5:
     bit 1 - shows aclk_perihp_cpll_src_en
     bit 0 - shows aclk_perihp_gpll_src_en
----
     And the new TRM will be updated.

     FYI.

Thanks.

On 2016年05月16日 23:49, Doug Anderson wrote:
> Hi,
>
> On Fri, May 13, 2016 at 8:36 PM, Xing Zheng <zhengxing@rock-chips.com> wrote:
>> Hi Doug,
>>
>>
>> On 2016年05月14日 04:10, Doug Anderson wrote:
>>> Hi,
>>>
>>> On Fri, May 13, 2016 at 11:42 AM, Brian Norris <briannorris@chromium.org>
>>> wrote:
>>>> From: Xing Zheng <zhengxing@rock-chips.com>
>>>>
>>>> There was a typo, swapping 'c' <--> 'g'.
>>>>
>>>> Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
>>>> Signed-off-by: Brian Norris <briannorris@chromium.org>
>>>> ---
>>>>    drivers/clk/rockchip/clk-rk3399.c | 4 ++--
>>>>    1 file changed, 2 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/drivers/clk/rockchip/clk-rk3399.c
>>>> b/drivers/clk/rockchip/clk-rk3399.c
>>>> index 145756c4f3c8..9f86bfef70f7 100644
>>>> --- a/drivers/clk/rockchip/clk-rk3399.c
>>>> +++ b/drivers/clk/rockchip/clk-rk3399.c
>>>> @@ -832,9 +832,9 @@ static struct rockchip_clk_branch
>>>> rk3399_clk_branches[] __initdata = {
>>>>                           RK3399_CLKGATE_CON(13), 1, GFLAGS),
>>>>
>>>>           /* perihp */
>>>> -       GATE(0, "cpll_aclk_perihp_src", "gpll", CLK_IGNORE_UNUSED,
>>>> +       GATE(0, "cpll_aclk_perihp_src", "cpll", CLK_IGNORE_UNUSED,
>>>>                           RK3399_CLKGATE_CON(5), 0, GFLAGS),
>>>> -       GATE(0, "gpll_aclk_perihp_src", "cpll", CLK_IGNORE_UNUSED,
>>>> +       GATE(0, "gpll_aclk_perihp_src", "gpll", CLK_IGNORE_UNUSED,
>>>>                           RK3399_CLKGATE_CON(5), 1, GFLAGS),
>>>>           COMPOSITE(ACLK_PERIHP, "aclk_perihp", mux_aclk_perihp_p,
>>>> CLK_IGNORE_UNUSED,
>>>>                           RK3399_CLKSEL_CON(14), 7, 1, MFLAGS, 0, 5,
>>>> DFLAGS,
>>> Definitely there was a bug since this table itself was inconsistent.
>>> ...and I _think_ this fix is correct, but I'll note that the TRM has
>>> more inconsistency here.
>>>
>>> In the big clock table 'CRU Clock Architecture Diagram', I see:
>>>     CLK 4 is CPLL
>>>     CLK 5 is GPLL
>>>     CLK 125 is aclk_perihp_cpll_src and has 4 (CPLL) as source, with
>>> g5[0] as the bit
>>>     CLK 126 is aclk_perihp_gpll_src and has 5 (GPLL) as source, with
>>> g5[1] as the bit
>>>
>>> In the definition of CRU_CLKGATE_CON5:
>>>     bit 0 shows aclk_perihp_gpll_src_en
>>>     bit 1 shows aclk_perihp_cpll_src_en
>>>
>>>
>>> Thus the table shows CPLL as gate5[0] and GPLL as gate5[1].  The
>>> register definition shows the opposite.  I'll tend to believe the
>>> table over the register definition, but I figured I'd bring it up
>>> anyway.
>>>
>>>
>>> Xing Zheng: can you confirm that the table is correct and ask
>>> documentation folks to fix the register definition for
>>> CRU_CLKGATE_CON5?
>> Yes, previously, our IC & DOC partner confirmed that the definition of
>> CRU_CLKGATE_CON5 should be:
>>    bit 0 shows aclk_perihp_cpll_src_en
>>    bit 1 shows aclk_perihp_gpll_src_en
>>
>> Sorry to the incorrect register definition, we will fix them and review the
>> TRM again.
> Great!
>
> Since we now have extra confirmation that Brian's patch is indeed correct:
>
> Reviewed-by: Douglas Anderson <dianders@chromium.org>
>
>
>

-- 
- Xing Zheng