On 10/30/26 12:57 AM, Eliot Courtney wrote: > On Mon Oct 26, 2026 at 1:24 PM JST, Dave Airlie via B4 Relay wrote: >> From: Dave Airlie <[email protected]> ... >> +++ b/drivers/gpu/nova-core/gpu.rs >> @@ -564,18 +564,19 @@ pub(crate) fn new<'a>( >> >> // Create GPU memory manager owning memory management resources. >> mm: { >> - let usable_vram = >> gsp_static_info.usable_fb_regions.first().ok_or(ENODEV)?; >> - let buddy_params = GpuBuddyParams { >> - base_offset: usable_vram.start, >> - size: usable_vram.end - usable_vram.start, >> - chunk_size: Alignment::new::<SZ_4K>(), >> - }; >> + let buddy_params = >> gsp_static_info.usable_fb_regions.first().map(|usable_vram| { >> + GpuBuddyParams { >> + base_offset: usable_vram.start, >> + size: usable_vram.end - usable_vram.start, >> + chunk_size: Alignment::new::<SZ_4K>(), >> + } >> + }); > > We should probably add a helper method on e.g. Chipset to say which > chips we expect FB on vs not, instead of proceeding with no fb regions > here for ones where we expect there to be one.
Open RM does approximately that, although Alex's point about using a HAL instead does of course apply. In Open RM, "bGpuArchIsZeroFb" is a per-chip HAL constant, true only for the Tegra iGPUs (GB10B, GB10C, GB20B, GB20C in the current public source). None of the chipsets that nova-core supports today is in that set, so an empty region list is still an error for all of them. thanks, -- John Hubbard
