Re: [PATCH v2] arm64: dts: qcom: agatti: enable FastRPC on the ADSP

[Konrad Dybcio]

On 3/30/26 6:54 AM, Sumit Garg wrote:
> On Tue, Jan 13, 2026 at 06:41:59PM +0200, Dmitry Baryshkov wrote:
>> On Agatti platform the ADSP provides FastRPC support. Add corresponding
>> device node, in order to be able to utilize the DSP offload from the
>> Linux side.
>>
>> Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@xxxxxxxxxxxxxxxx>
>> ---
>> Changes in v2:
>> - Add more FastRPC context banks (6, 7)
>> - Link to v1: https://lore.kernel.org/r/20260113-agatti-fastrpc-v1-1-e210903ffcb1@xxxxxxxxxxxxxxxx
>> ---
>>  arch/arm64/boot/dts/qcom/agatti.dtsi | 41 ++++++++++++++++++++++++++++++++++++
>>  1 file changed, 41 insertions(+)
>>
> 
> This patch causes a critical boot regression on RB1 as per the CI test
> report here [1]. I see this change already landed in this merge window.
> So, either this commit has to be reverted for v7.0 release or an
> appropriate fix is required.
> 
> [1] https://github.com/qualcomm-linux/kernel/issues/359

Why has that issue been opened 2 weeks ago with no action since?

FWIW, copying the broader log from the LAVA job:


[[0;32m  OK  [0m] Reached target [0;1;39mBasic System[0m.
         Starting [0;1;39maDSP RPC daemon[0m...
         Starting [0;1;39maudiopd aDSP RPC daemon[0m...
         Starting [0;1;39mAndroid Debug Bridge[0m...
         Starting [0;1;39mAvahi mDNS/DNS-SD Stack[0m...
         Starting [0;1;39mBluetooth service[0m...
         Starting [0;1;39mD-Bus System Message Bus[0m...
[   17.318840] qcom,fastrpc-cb ab00000.remoteproc:glink-edge:fastrpc:compute-cb@4: mem mmap error, fd 11, vaddr ffffae380000, size 262144
[   17.327877] qcom_q6v5_pas ab00000.remoteproc: fatal error received: SFR Init: wdog or kernel error suspected.
[   17.343248] remoteproc remoteproc1: crash detected in adsp: type fatal error
[   17.350567] qcom_q6v5_pas ab00000.remoteproc: Handover signaled, but it already happened
[   17.358869] remoteproc remoteproc1: handling crash #1 in adsp
[   17.364730] remoteproc remoteproc1: recovering adsp

Format: Log Type - Time(microsec) - Message - Optional Info
Log Type: B - Since Boot(Power On Reset),  D - Delta,  S - Statistic
<ded>

It seems like this only ends up being a problem once the userland
daemon starting poking at the channel - +Ekansh, have you seen this
before?

Konrad