[PATCH 00/13] Add IPU & DSP remoteprocs on OMAP4 and OMAP5

[Suman Anna]

On 7/10/20 12:17 PM, Suman Anna wrote:
> On 7/10/20 11:58 AM, Tony Lindgren wrote:
>> Hi,
>>
>> * Suman Anna <s-anna at ti.com> [200709 16:20]:
>>> Hi Tony,
>>>
>>> The following series contains all the necessary DT pieces to boot the
>>> IPU and DSP remote processors on OMAP4 and OMAP5 SoCs. They are
>>> enabled specifically on the TI OMAP4 PandaBoard and OMAP5 uEVM boards.
>>> This is the last DT piece that now completes the support for IPUs and
>>> DSPs on all OMAP4+ SoCs, similar patches were merged for 5.8 covering
>>> the DRA7xx/AM57xx SoCs. Appreciate it if you can pick up the series for
>>> 5.9 if it isn't too late.
>>
>> Great and good to hear things are working with only dts changes now :)
>> Yes let's try to get these merged.
> 
> Thanks.
> 
>>
>>> There is one issue that I have run into while testing this series on
>>> the latest kernel. I am seeing a l3_noc error for OMAP4 DSP when it
>>> attempts to auto-suspend or stop after it is booted. The issue is a
>>> L4CFG read error that happens in the sysc_disable_module() function
>>> in ti-sysc code.
>>>
>>> I do not have any issues on my downstream 5.4 based SDK kernel. I have
>>> root-caused this to the OMAP4 voltage controller patches you added for
>>> 5.5 kernel through your omap-for-v5.5/pm branch, specifically the
>>> commit 4873843718f9 ("ARM: OMAP2+: Initialize voltage controller for 
>>> omap4").
>>> The VOLTCTRL register value is 0x300 before that patch, and modifying
>>> this register either through  omap4_vc_init_pmic_signaling() or
>>> omap4_vc_set_pmic_signaling() will trigger this. A debug print in
>>> sysc_disable_module() also seems to help.
>>
>> Hmm interesting, not sure how the VOLTCTRL register affects this.
>>
>> I wonder the following commit in v5.8-rc3 might help with this though:
>>
>> 5ce8aee81be6 ("bus: ti-sysc: Flush posted write on enable and disable")
>>
> 
> I had already tested on v5.8-rc4 when I posted the patches, so this 
> patch doesn't help. OMAP5 DSP is fine, because Think it has to do with 
> this automated
> 
> So, I am looking at the TRM, and the three 
> VDD_{IVA,MPU,CORE}_I2C_DISABLE bits in VOLTCTRL are marked debug-purpose 
> only, so I don't think we should be setting those to begin with. Any 
> reason why you want to set those? Anyway, these bits were not an issue, 
> I have specifically tried that already.

FYI, the following one-line removal is enough for me to not see the error.

diff --git a/arch/arm/mach-omap2/vc.c b/arch/arm/mach-omap2/vc.c
index 86f1ac4c2412..b80c9dff81c4 100644
--- a/arch/arm/mach-omap2/vc.c
+++ b/arch/arm/mach-omap2/vc.c
@@ -44,7 +44,6 @@
  #define OMAP4_VDD_DEFAULT_VAL  \
         (OMAP4430_VDD_I2C_DISABLE_MASK | \
          OMAP4430_VDD_IVA_PRESENCE | OMAP4430_VDD_MPU_PRESENCE | \
-        OMAP4430_AUTO_CTRL_VDD_IVA(OMAP4430_AUTO_CTRL_VDD_RET) | \
          OMAP4430_AUTO_CTRL_VDD_MPU(OMAP4430_AUTO_CTRL_VDD_RET) | \
          OMAP4430_AUTO_CTRL_VDD_CORE(OMAP4430_AUTO_CTRL_VDD_RET))

regards
Suman

> 
>> I was seeing that occasionally with mcspi, but never had anything
>> reproducable.
>>
>> Regards,
>>
>> Tony
>>
>