Re: [PATCH v1 0/1] tegra_mmc: get tap and trim from dts

2023-08-24 Thread Svyatoslav Ryhel



24 серпня 2023 р. 16:27:19 GMT+03:00, Thierry Reding  
написав(-ла):
>On Wed, Aug 23, 2023 at 02:30:48PM +0300, Svyatoslav Ryhel wrote:
>> 
>> 
>> 23 серпня 2023 р. 13:53:26 GMT+03:00, Thierry Reding 
>>  написав(-ла):
>> >On Sat, Aug 19, 2023 at 06:35:00PM +0300, Svyatoslav Ryhel wrote:
>> >> Default-tap and default-trim values are used for eMMC setup
>> >> mostly on T114+ devices. As for now, those values are hardcoded
>> >> for T210 and ignored for all other Tegra generations. Fix this
>> >> by passing tap and trim values from dts.
>> >> 
>> >> This commit will cause regression on T210 (emmc may not work),
>> >> fix is applied in tegra210.dtsi and will be sent next week.
>> >
>> >Heh... I don't think so. If you already know that this will cause a
>> >regression on Tegra210 you need to rework this. Adding regressions
>> >accidentally is already bad enough, but doing so knowingly is a big
>> >no-no.
>> 
>> DTS change for t210 was sent in "General tegra and board improvements"
>> patchset. This is why this pathset contains only 1 (one) patch and a
>> warning. It has a dependency.
>
>For U-Boot this may not matter as much because the control DTB is always
>linked into the binary, but for Linux for example we need to make sure
>that the kernel always remains backwards compatible with older device
>trees. I think that's good to follow in general, so making sure there
>are sensible defaults to fall back to if the DTB is missing some
>properties is good practice.
>
>Anyway, the way you posted these there was a future dependency, so if
>someone had applied the driver change without the DTB change having been
>applied first this would've caused a regression, which is always bad if
>it happens at random points because it makes things like bisections very
>painful.
>
>So unless the patch is fixed to fall back to defaults for Tegra210 we
>need to carefully coordinate in which order these patches go in. The DT
>change needs to go in first, followed by the driver change that relies
>on the DTB update.
>

You are right, it was my miss that I did not include tegra210 fix here. I will 
add it in v2. Thanks.

>Thierry


Re: [PATCH v1 0/1] tegra_mmc: get tap and trim from dts

2023-08-24 Thread Thierry Reding
On Wed, Aug 23, 2023 at 02:30:48PM +0300, Svyatoslav Ryhel wrote:
> 
> 
> 23 серпня 2023 р. 13:53:26 GMT+03:00, Thierry Reding 
>  написав(-ла):
> >On Sat, Aug 19, 2023 at 06:35:00PM +0300, Svyatoslav Ryhel wrote:
> >> Default-tap and default-trim values are used for eMMC setup
> >> mostly on T114+ devices. As for now, those values are hardcoded
> >> for T210 and ignored for all other Tegra generations. Fix this
> >> by passing tap and trim values from dts.
> >> 
> >> This commit will cause regression on T210 (emmc may not work),
> >> fix is applied in tegra210.dtsi and will be sent next week.
> >
> >Heh... I don't think so. If you already know that this will cause a
> >regression on Tegra210 you need to rework this. Adding regressions
> >accidentally is already bad enough, but doing so knowingly is a big
> >no-no.
> 
> DTS change for t210 was sent in "General tegra and board improvements"
> patchset. This is why this pathset contains only 1 (one) patch and a
> warning. It has a dependency.

For U-Boot this may not matter as much because the control DTB is always
linked into the binary, but for Linux for example we need to make sure
that the kernel always remains backwards compatible with older device
trees. I think that's good to follow in general, so making sure there
are sensible defaults to fall back to if the DTB is missing some
properties is good practice.

Anyway, the way you posted these there was a future dependency, so if
someone had applied the driver change without the DTB change having been
applied first this would've caused a regression, which is always bad if
it happens at random points because it makes things like bisections very
painful.

So unless the patch is fixed to fall back to defaults for Tegra210 we
need to carefully coordinate in which order these patches go in. The DT
change needs to go in first, followed by the driver change that relies
on the DTB update.

Thierry


signature.asc
Description: PGP signature


Re: [PATCH v1 0/1] tegra_mmc: get tap and trim from dts

2023-08-23 Thread Svyatoslav Ryhel



23 серпня 2023 р. 13:53:26 GMT+03:00, Thierry Reding  
написав(-ла):
>On Sat, Aug 19, 2023 at 06:35:00PM +0300, Svyatoslav Ryhel wrote:
>> Default-tap and default-trim values are used for eMMC setup
>> mostly on T114+ devices. As for now, those values are hardcoded
>> for T210 and ignored for all other Tegra generations. Fix this
>> by passing tap and trim values from dts.
>> 
>> This commit will cause regression on T210 (emmc may not work),
>> fix is applied in tegra210.dtsi and will be sent next week.
>
>Heh... I don't think so. If you already know that this will cause a
>regression on Tegra210 you need to rework this. Adding regressions
>accidentally is already bad enough, but doing so knowingly is a big
>no-no.

DTS change for t210 was sent in "General tegra and board improvements" 
patchset. This is why this pathset contains only 1 (one) patch and a warning. 
It has a dependency.

>Thierry


Re: [PATCH v1 0/1] tegra_mmc: get tap and trim from dts

2023-08-23 Thread Thierry Reding
On Sat, Aug 19, 2023 at 06:35:00PM +0300, Svyatoslav Ryhel wrote:
> Default-tap and default-trim values are used for eMMC setup
> mostly on T114+ devices. As for now, those values are hardcoded
> for T210 and ignored for all other Tegra generations. Fix this
> by passing tap and trim values from dts.
> 
> This commit will cause regression on T210 (emmc may not work),
> fix is applied in tegra210.dtsi and will be sent next week.

Heh... I don't think so. If you already know that this will cause a
regression on Tegra210 you need to rework this. Adding regressions
accidentally is already bad enough, but doing so knowingly is a big
no-no.

Thierry


signature.asc
Description: PGP signature


[PATCH v1 0/1] tegra_mmc: get tap and trim from dts

2023-08-19 Thread Svyatoslav Ryhel
Default-tap and default-trim values are used for eMMC setup
mostly on T114+ devices. As for now, those values are hardcoded
for T210 and ignored for all other Tegra generations. Fix this
by passing tap and trim values from dts.

This commit will cause regression on T210 (emmc may not work),
fix is applied in tegra210.dtsi and will be sent next week.

Svyatoslav Ryhel (1):
  mmc: tegra: get default-tap and default-trim from device tree

 arch/arm/include/asm/arch-tegra/tegra_mmc.h | 17 
 drivers/mmc/tegra_mmc.c | 46 ++---
 2 files changed, 30 insertions(+), 33 deletions(-)

-- 
2.39.2