[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20260123122502.006a01c3.zhiw@nvidia.com>
Date: Fri, 23 Jan 2026 12:25:02 +0200
From: Zhi Wang <zhiw@...dia.com>
To: Joel Fernandes <joelagnelf@...dia.com>
CC: <rust-for-linux@...r.kernel.org>, <linux-pci@...r.kernel.org>,
<linux-kernel@...r.kernel.org>, <dakr@...nel.org>, <aliceryhl@...gle.com>,
<jgg@...dia.com>
Subject: Re: [PATCH v2 1/2] rust: introduce abstractions for fwctl
On Thu, 22 Jan 2026 16:17:55 -0500
Joel Fernandes <joelagnelf@...dia.com> wrote:
> On Thu, Jan 22, 2026 at 10:42:30PM +0200, Zhi Wang wrote:
> > + /// Called when a userspace RPC request is received.
> > + fn fw_rpc(
> > + uctx: &mut UserCtx<Self::UserCtx>,
> > + scope: u32,
> > + rpc_in: &mut [u8],
> > + out_len: *mut usize,
> > + ) -> Result<Option<KVec<u8>>, Error>;
>
> Exposing a raw pointer in the trait API means drivers that want to
> "reuse input buffer" need to write unsafe code right?
>
> unsafe { *out_len = ... };
>
> I believe the unsafe code should be confined to the abstraction layer
> instead, not in the driver.
>
> How about using using an enum return type instead to properly wrap
> inplace versus driver allocated outputs?
>
> pub enum RpcOutput {
> Allocated(KVec<u8>),
> InPlace(usize),
> }
>
> fn fw_rpc(
> uctx: &mut UserCtx<Self::UserCtx>,
> scope: u32,
> rpc_in: &mut [u8],
> ) -> Result<RpcOutput, Error>;
>
> Then the abstraction handles the pointer write:
>
> match T::fw_rpc(ctx, scope, rpc_in_slice) {
> Ok(RpcOutput::Allocated(kvec)) => {
> ...
> }
> Ok(RpcOutput::InPlace(len)) => {
> unsafe { *out_len = len };
> ...
> }
> }
>
> fw_rpc() as a bonus also gets 1 less function parameter and cleaner
> return signature.
>
This is a good idea. Noted. :)
> --
> Joel Fernandes
>
Powered by blists - more mailing lists