On Tue Dec 30, 2025 at 10:42 PM CET, Timur Tabi wrote:
> On Sun, 2025-12-28 at 18:45 +0100, Ewan Chorynski wrote:
>> 
>> [    3.623000] NovaCore 0000:01:00.0: GSP MBOX0: 0xffffe000, MBOX1: 0x0
>> [    3.623026] NovaCore 0000:01:00.0: Using SEC2 to load and run the 
>> booter_load firmware...
>> [    3.626236] NovaCore 0000:01:00.0: SEC2 MBOX0: 0x31, MBOX10x0
>> [    3.626265] NovaCore 0000:01:00.0: Booter-load failed with error 0x31
>> 
>> Is this expected to happen ? Here is the full device description from
>> lspci :
>> 
>> 01:00.0 VGA compatible controller: NVIDIA Corporation TU117M [GeForce GTX 
>> 1650 Mobile / Max-Q]
>> (rev a1)
>
> This should work on your GPU with my patches.  Error 0x31 is a generic 
> failure error code that
> indicates that booter_load failed to initialize, and that could be for any 
> number of reasons. 
> Unfortunately, the only way I've been able to debug such issues is to 
> reproduce them in-house with a
> custom build of booter_load.
>
> I suggest that for now, you hold off until these commits are merged, along 
> with my debugfs patches,
> and then I can try to find a similar GPU internally that exhibits the failure 
> and see if there's
> something else missing.  Turing is the most complicated of all GSP-capable 
> GPUs to boot, partly
> because the firmware does a terrible job of reporting errors.
>
> I'm assuming that Nouveau boots just fine with the same firmware images?  If 
> you turn on debug
> logging in Nouveau, it should say that it's booting with 570.144.  You can 
> force it by deleting all
> the *535.113.01* images in /lib/firmware/nvidia/

Nouveau is not probing either with 570.114.

[    5.619048] nouveau 0000:01:00.0: gsp: firmware 
"nvidia/tu117/gsp/gsp-570.144.bin" loaded - 28542040 byte(s)
[    5.619431] nouveau 0000:01:00.0: gsp: firmware 
"nvidia/tu117/gsp/bootloader-570.144.bin" loaded - 4196 byte(s)
[    5.619944] nouveau 0000:01:00.0: gsp: firmware 
"nvidia/tu117/gsp/booter_load-570.144.bin" loaded - 59272 byte(s)
[    5.620351] nouveau 0000:01:00.0: gsp: firmware 
"nvidia/tu117/gsp/booter_unload-570.144.bin" loaded - 39304 byte(s)
[    5.620356] nouveau 0000:01:00.0: gsp: RM version: 570.144
...
[    5.860631] nouveau 0000:01:00.0: sec2(gsp):booter-load: booting
[    5.860943] nouveau 0000:01:00.0: sec2(gsp): mbox 00000031 00000000
[    5.860961] nouveau 0000:01:00.0: sec2(gsp):booter-load: boot failed: -5
[    5.860982] nouveau 0000:01:00.0: gsp: released sec2 falcon

This suggests that the firmware itself is part of the issue. I can send
the full boot logs from nouveau in debug if needed.

I also tried the fix given by John at [1] but got the same error so it
may be a different issue.

[1]: 
https://lore.kernel.org/rust-for-linux/[email protected]

Reply via email to