Re: [PATCH v3 27/33] gpu: nova-core: switch to the r000 GSP firmware

[Timur Tabi]

On Mon, 2026-09-21 at 19:44 -0700, John Hubbard wrote:
> On 9/18/26 3:48 PM, Timur Tabi wrote:
> > On Thu, 2026-09-17 at 18:07 -0700, John Hubbard wrote:
> > > 
> > > --- a/drivers/gpu/nova-core/fb/hal/gb100.rs
> > > +++ b/drivers/gpu/nova-core/fb/hal/gb100.rs
> > > @@ -79,10 +79,28 @@ fn write_sysmem_flush_page_gb100(hshub0: Mmio<'_, regs::Hshub0Registers>, addr:
> > >      hshub0.write_reg(regs::NV_PFB_HSHUB_EG_PCIE_FLUSH_SYSMEM_ADDR_LO::zeroed().with_adr(addr_lo));
> > >  }
> > >  
> > > -// This PMU reservation size is r570-specific.
> > > +/// PMU backing store.
> > > +const PMU_BACKING_STORE_SIZE: usize = 9 * SZ_1M;
> > 
> > I don't see this comment or value removed/changed in this patchset, so is it really 570-
> > specific?
> 
> Yes, it was r570-specific, and this patch is where it changes: the
> hunk removes the comment and replaces the value.
> 
> In 570.144 the PMU region is:
> 
>     * 8 MiB of backing store
>     * 16 MiB of surfaces
>     * 4 KiB of miscellaneous memory
>     * all aligned to 128 KiB. 
> 
> In 615.71.09 the backing store is 9 MiB and the surfaces are 16 MiB + 256 KiB.

On 615, the backing store is 0x950000, not 9 MiB:

NvU32
kpmuReservedMemoryBackingStoreSizeGet_IMPL
(
    KernelPmu *pKernelPmu
)
{
    if (kpmuGetIsSelfInit(pKernelPmu))
    {
        //
        // MMINTS-TODO: cross-reference with this when reserving memory
        // in pmu_20.c
        //
        return 0x950000;
    }

    if (kpmuGetIsSelfInit(pKernelPmu))
    {
        //
        // MMINTS-TODO: cross-reference with this when reserving memory
        // in pmu_20.c
        //
        return 0x900000;
    }

    return 0U;
}

kpmuGetIsSelfInit() returns True on GB100 and later, and False otherwise.  The second if-
statement is never entered, because this is just some #ifdef weirdness when OpenRM was
generated.