Re: amlogic: Synchronous Abort with v6.15-rc6

[Vignesh Raman <vignesh.raman@collabora.com>]

Hi Neil,

On 13/05/25 21:25, Neil Armstrong wrote:
> On 13/05/2025 16:26, Vignesh Raman wrote:
>> Hi Neil,
>>
>> On 13/05/25 15:25, neil.armstrong@linaro.org wrote:
>>> Hi,
>>>
>>> On 13/05/2025 11:41, Vignesh Raman wrote:
>>>> Hi,
>>>>
>>>> The Amlogic Meson G12B (A311D) SoC board (meson-g12b-a311d-khadas- 
>>>> vim3) fails to boot with the following logs,
>>>>
>>>> https://gitlab.freedesktop.org/vigneshraman/linux/-/jobs/76210200
>>>>
>>>> 02:58:27.819: Starting kernel ...
>>>> 02:58:27.819: "Synchronous Abort" handler, esr 0x96000010, far 
>>>> 0x51000d8
>>>> 02:58:27.819: "Synchronous Abort" handler, esr 0x96000010, far 
>>>> 0x51000c8
>>>> 02:58:27.819: elr: 0000000001002a48 lr : 0000000001002c88 (reloc)
>>>> 02:58:27.819: elr: 00000000f2f26a48 lr : 00000000f2f26c88
>>>> 02:58:27.819: x0 : 00000000050fffc0 x1 : 0000000001004383
>>>> 02:58:27.819: x2 : 00000000f2f6a358 x3 : 000000000000003f
>>>> 02:58:27.819: x4 : 00000000f4fefffc x5 : 0000000000000010
>>>> 02:58:27.819: x6 : 00000000f4e5b000 x7 : 00000000050fffa0
>>>> 02:58:27.819: x8 : 0000000000000008 x9 : 0000000000000001
>>>> 02:58:27.820: x10: 0000000000000010 x11: 0000000000000004
>>>> 02:58:27.820: x12: 00000000f4fefc00 x13: 0000000000000000
>>>> 02:58:27.820: x14: 00000000f4fefc00 x15: 0000000000000000
>>>> 02:58:27.820: x16: 00000000f2f6d0e0 x17: 0000000000c0c0c0
>>>> 02:58:27.820: x18: 00000000f0f21d90 x19: 00000000050fffc0
>>>> 02:58:27.820: x20: 0000000000000000 x21: 000000007ffe9000
>>>> 02:58:27.820: x22: 0000000000001710 x23: 00000000f0f0e8c0
>>>> 02:58:27.820: x24: 0000000000000000 x25: 0000000000000000
>>>> 02:58:27.820: x26: 0000000000000000 x27: 0000000000000000
>>>> 02:58:27.820: x28: 00000000f0f4f790 x29: 00000000050fff60
>>>> 02:58:27.820: Code: a9025bf5 f9001bf7 f9400641 36000141 (f9408663)
>>>> 02:58:27.820: Resetting CPU ...
>>>>
>>>
>>> Could you boot with earlycon enabled to at least have a clue of 
>>> what's happening ?
>>
>> There is a synchronous abort right after 'Starting kernel...'. Maybe 
>> this prevents any early console logs from appearing.
>>
>> https://gitlab.freedesktop.org/vigneshraman/linux/-/jobs/76243455
>>
>> 13:54:02.013: => setenv bootargs 'console=ttyAML0,115200n8 root=/dev/ 
>> nfs rw nfsroot=192.168.201.1:/var/lib/lava/dispatcher/tmp/18641665/ 
>> extract-nfsrootfs-h2fjsmdc,tcp,hard,v3 init=/init rootwait 
>> usbcore.quirks=0bda:8153:k earlycon=meson,0xff803000 keep_bootcon  
>> ip=dhcp'
>> 13:54:02.014: $ booti 0x1080000 - 0x1000000
>> 13:54:02.014: => booti 0x1080000 - 0x1000000
>> 13:54:02.014: Moving Image from 0x1080000 to 0x1200000, end=5160000
>> 13:54:02.014: ## Flattened Device Tree blob at 01000000
>> 13:54:02.014:    Booting using the fdt blob at 0x1000000
>> 13:54:02.014: Working FDT set to 1000000
>> 13:54:02.014: ERROR: reserving fdt memory region failed (addr=5000000 
>> size=300000 flags=4)
>> 13:54:02.014:    Loading Device Tree to 000000007ffe9000, end 
>> 000000007ffff57c ... OK
>> 13:54:02.014: Working FDT set to 7ffe9000
>> 13:54:02.014: Starting kernel ...
>> 13:54:02.014: "Synchronous Abort" handler, esr 0x96000010, far 0x51000d8
>> 13:54:02.014: "Synchronous Abort" handler, esr 0x96000010, far 0x51000c8
>> 13:54:02.014: elr: 0000000001002a48 lr : 0000000001002c88 (reloc)
>> 13:54:02.014: elr: 00000000f2f26a48 lr : 00000000f2f26c88
>> 13:54:02.014: x0 : 00000000050fffc0 x1 : 0000000001004383
>> 13:54:02.014: x2 : 00000000f2f6a358 x3 : 000000000000003f
>> 13:54:02.014: x4 : 00000000f4fefffc x5 : 0000000000000010
>> 13:54:02.014: x6 : 00000000f4e5b000 x7 : 00000000050fffa0
>> 13:54:02.014: x8 : 0000000000000008 x9 : 0000000000000001
>> 13:54:02.014: x10: 0000000000000010 x11: 0000000000000004
>> 13:54:02.014: x12: 00000000f4fefc00 x13: 0000000000000000
>> 13:54:02.014: x14: 00000000f4fefc00 x15: 0000000000000000
>> 13:54:02.014: x16: 00000000f2f6d0e0 x17: 0000000000c0c0c0
>> 13:54:02.014: x18: 00000000f0f21d90 x19: 00000000050fffc0
>> 13:54:02.014: x20: 0000000000000000 x21: 000000007ffe9000
>> 13:54:02.014: x22: 0000000000001710 x23: 00000000f0f0e8c0
>> 13:54:02.014: x24: 0000000000000000 x25: 0000000000000000
>> 13:54:02.014: x26: 0000000000000000 x27: 0000000000000000
>> 13:54:02.014: x28: 00000000f0f566a0 x29: 00000000050fff60
>> 13:54:02.014: Code: a9025bf5 f9001bf7 f9400641 36000141 (f9408663)
>> 13:54:02.014: Resetting CPU ...
> 
> In fact, this is an U-boot fault, not Linux. I guess perhaps the Linux 
> kernel is too big, can you check the size difference since v6.14 ?

Yes, looks like the kernel image size causes this issue. I disabled a 
few configs in drm-ci, and the kernel image size was around 49M. It 
booted fine with v6.15-rc1.

In previous tests, v6.14-rc7 kernel image size was 52M, and v6.15-rc1 
kernel image size was 53M.

Thanks for the pointers.

Regards,
Vignesh

> 
>>
>> Successful logs with v6.14-rc7,
>> https://gitlab.freedesktop.org/vigneshraman/linux/-/jobs/76248143
>>
>>>
>>> BPI-M2S (A311D) is booting fine on v6.15-rc, see https://gitlab.com/ 
>>> amlogic-foss/abcd-linux-test/-/jobs/10004997032
>>>
>>> But perhaps something specific to the VIM3 is breaking, do you change 
>>> the kernel config for the DRM ?
>>
>> Yes, we change the config. Please see the config used to build the 
>> kernel,
>> https://gitlab.freedesktop.org/vigneshraman/linux/-/jobs/76243439/ 
>> artifacts/file/artifacts/testing:arm64_config
>>
>> Regards,
>> Vignesh
>>
>>>
>>> Neil
>>>
>>>
>>>> The board boots successfully in v6.14-rc7, but the issue appears 
>>>> starting from v6.15-rc1. These tests were run in drm-ci.
>>>>
>>>> On bisecting the kernel, the first bad commit identified by git 
>>>> biscet is,
>>>>
>>>> 59c35416f4246aee66b5f5523fdc950b83325d82 is the first bad commit
>>>> commit 59c35416f4246aee66b5f5523fdc950b83325d82
>>>> Merge: 472863ab2aca e988adcb5dee
>>>> Author: Linus Torvalds <torvalds@linux-foundation.org>
>>>> Date:   Sat Mar 29 17:23:34 2025 -0700
>>>>
>>>>      Merge tag 'clk-for-linus' of git://git.kernel.org/pub/scm/ 
>>>> linux/ kernel/git/clk/linux
>>>>
>>>> Please could you look into this issue. Thank you.
>>>>
>>>> Regards,
>>>> Vignesh
>>>
>>
>