Re: [PATCH 15/31] gpu: nova-core: Hopper/Blackwell: add FSP falcon EMEM operations

[John Hubbard <jhubbard@nvidia.com>]

On 1/21/26 8:06 AM, Gary Guo wrote:
> On Wed Dec 3, 2025 at 5:59 AM GMT, John Hubbard wrote:
>> Add external memory (EMEM) read/write operations to the GPU's FSP falcon
>> engine. These operations use Falcon PIO (Programmed I/O) to communicate
>> with the FSP through indirect memory access.
>>
>> Signed-off-by: John Hubbard <jhubbard@nvidia.com>
>> ---
>>  drivers/gpu/nova-core/falcon/fsp.rs | 60 ++++++++++++++++++++++++++++-
>>  drivers/gpu/nova-core/regs.rs       | 10 +++++
>>  2 files changed, 69 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/gpu/nova-core/falcon/fsp.rs b/drivers/gpu/nova-core/falcon/fsp.rs
>> index 7323ae2f2302..9e796e82e556 100644
>> --- a/drivers/gpu/nova-core/falcon/fsp.rs
>> +++ b/drivers/gpu/nova-core/falcon/fsp.rs
>> @@ -5,15 +5,27 @@
>>  //! The FSP falcon handles secure boot and Chain of Trust operations
>>  //! on Hopper and Blackwell architectures, replacing SEC2's role.
>>  
>> +use kernel::prelude::*;
>> +
>>  use crate::{
>> +    driver::Bar0,
>>      falcon::{
>> +        Falcon,
>>          FalconEngine,
>>          PFalcon2Base,
>>          PFalconBase, //
>>      },
>> -    regs::macros::RegisterBase,
>> +    regs::{
>> +        self,
>> +        macros::RegisterBase, //
>> +    },
>>  };
>>  
>> +/// EMEM control register bit 24: write mode.
>> +const EMEM_CTL_WRITE: u32 = 1 << 24;
>> +/// EMEM control register bit 25: read mode.
>> +const EMEM_CTL_READ: u32 = 1 << 25;
>> +
>>  /// Type specifying the `Fsp` falcon engine. Cannot be instantiated.
>>  #[allow(dead_code)]
>>  pub(crate) struct Fsp(());
>> @@ -30,3 +42,49 @@ impl RegisterBase<PFalcon2Base> for Fsp {
>>  impl FalconEngine for Fsp {
>>      const ID: Self = Fsp(());
>>  }
>> +
>> +impl Falcon<Fsp> {
>> +    /// Writes `data` to FSP external memory at byte `offset` using Falcon PIO.
>> +    ///
>> +    /// Returns `EINVAL` if offset or data length is not 4-byte aligned.
>> +    #[allow(dead_code)]
>> +    pub(crate) fn write_emem(&self, bar: &Bar0, offset: u32, data: &[u8]) -> Result {
>> +        if offset % 4 != 0 || data.len() % 4 != 0 {
> 
> I was about to suggest `is_multiple_of`, but then realize that it's only
> available in Rust 1.82+...

OK, I've added TODO's for that.

> 
>> +            return Err(EINVAL);
>> +        }
>> +
>> +        regs::NV_PFALCON_FALCON_EMEM_CTL::default()
>> +            .set_value(EMEM_CTL_WRITE | offset)
>> +            .write(bar, &Fsp::ID);
>> +
>> +        for chunk in data.chunks_exact(4) {
>> +            let word = u32::from_le_bytes([chunk[0], chunk[1], chunk[2], chunk[3]]);
> 
> Use `as_chunks()` can save you from doing this.
> 
> Also, a typical pattern is `chunk.try_into().unwrap()` -- yes it has an unwrap
> as `[T]` -> `[T; N]` can fail in general, but not with `chunks_exact`.
> 

Here, I recall having a discussion or two about it, and we concluded
that we'd rather avoid having to justify an "infallible .unwrap()",
in favor of just writing out the four chunks. I still like that in this
particular case, it's just one less "// PANIC: " to *not* have to deal
with. In other words, one line of code, instead of a line plus one or
two lines of special comments that someone will later try to turn into
a proof. haha :)


thanks,
-- 
John Hubbard