[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <2025060126-various-greasily-61df@gregkh>
Date: Sun, 1 Jun 2025 09:48:11 +0200
From: Greg KH <gregkh@...uxfoundation.org>
To: Timur Tabi <timur@...nel.org>
Cc: Alexandre Courbot <acourbot@...dia.com>,
Danilo Krummrich <dakr@...nel.org>,
John Hubbard <jhubbard@...dia.com>, Miguel Ojeda <ojeda@...nel.org>,
Alex Gaynor <alex.gaynor@...il.com>,
Boqun Feng <boqun.feng@...il.com>, Gary Guo <gary@...yguo.net>,
Björn Roy Baron <bjorn3_gh@...tonmail.com>,
Benno Lossin <benno.lossin@...ton.me>,
Andreas Hindborg <a.hindborg@...nel.org>,
Alice Ryhl <aliceryhl@...gle.com>, Trevor Gross <tmgross@...ch.edu>,
linux-kernel@...r.kernel.org, rust-for-linux@...r.kernel.org,
Timur Tabi <ttabi@...dia.com>
Subject: Re: [PATCH] rust: add basic ELF sections parser
On Sat, May 31, 2025 at 09:38:24AM -0500, Timur Tabi wrote:
> On Sat, May 31, 2025 at 7:25 AM Greg KH <gregkh@...uxfoundation.org> wrote:
> >
> > What exactly do you mean by this? That is what I have been asking, what
> > is the specific reason why this can't be done in userspace? What
> > hardware "thing" can't be read by userspace, and why not? Userspace has
> > access to PCI devices directly, surely there is nothing "secret" here.
>
> Why in the world would you want user space to read hardware registers,
> when the driver is already doing it???????
>
> And please note that the driver has to read and parse a lot of other
> register in order to know which register contains the fuse settings,
> and even whether that register exists, and at what address. The fuse
> register is hardware-specific. It doesn't exist on Turing, it does
> exist on Ampere and Ada (but just GA10x, not GA100), and it's not used
> on Hopper and Blackwell. You want to duplicate all this code in
> user-space (assuming that registers really are accessible in user
> space), just avoid a 12-line function that already exists and works in
> Nouveau?????????
Because you are not proposing a 12 line function here, you all were
attempting to add an elf parser to the core kernel that everyone would
always load into their memory space.
> Please make it make sense, Greg.
Please read my thread where I asked for detailed specifics as to exactly
what is required here for you all to determine what is needed to load
the firmware.
have a good weekend,
greg k-h
Powered by blists - more mailing lists