Re: [PATCH 1/2] gpu: nova-core: gsp: retain the GSP-RM log buffers after unbind

[Vladislav Zaharov]

On Thu Aug 13, 2026, Danilo Krummrich wrote:
> I think those should use VVec.
> Let's move all the LogBuffer code into gsp/logbuffer.rs to keep gsp.rs clean.
> I think we can avoid this additional unsafe if we just create the retained dir
> right away in module_init().
> dev_dbg!() should be good enough.

All four make sense, thanks - v2 will have them. Creating the retained
directory in module_init() also removes the only reason retain() had to look
at DEBUGFS_ROOT, so the unsafe block goes away with it.

On Thu Aug 13, 2026, John Hubbard wrote:
> I'd *much* rather use a kernel parameter: keep_gsp_logs, instead of
> requiring a rebuild of the kernel.

Agreed, and it makes the patch smaller: with a module parameter the cfg gating
disappears and the code is simply always built. I used a Kconfig because of the
"the only Kconfig needed is for retaining the GSP log buffers after driver
unbind" remark in the earlier thread, which I took literally instead of asking.

That one needs a decision, though. The Rust module parameter abstraction has no
bool: rust/kernel/module_param.rs only instantiates param ops for i8..u64,
isize and usize, and rust/macros/module.rs panics on anything else. Nor is it
quite a one-liner to add, since bare bool parameters rely on
KERNEL_PARAM_OPS_FL_NOARG, which make_param_ops! cannot currently express.

I am happy to write that prerequisite patch, but it would pull this series into
rust/kernel review. So unless bool support is already in flight somewhere I
have not found, I propose v2 uses u8 for now and moves to bool once it exists.
Say the word if you would rather have it done properly first.

The re-test on top of the current tip is still owed; I will run it before v2
and report the result in its cover letter.

Thanks,
Vladislav