Re: [PATCH net-next v3 0/3] Rust abstractions for network PHY drivers
From: Greg KH <gregkh@linuxfoundation.org>
Date: 2023-10-09 15:04:08
Also in:
rust-for-linux
On Mon, Oct 09, 2023 at 04:56:36PM +0200, Andrew Lunn wrote:
On Mon, Oct 09, 2023 at 04:21:09PM +0200, Andrea Righi wrote:quoted
On Mon, Oct 09, 2023 at 02:53:00PM +0200, Miguel Ojeda wrote:quoted
On Mon, Oct 9, 2023 at 2:48 PM Andrew Lunn [off-list ref] wrote:quoted
Any ideas?That is `RETHUNK` and `X86_KERNEL_IBT`. Since this will keep confusing people, I will make it a `depends on !` as discussed in the past. I hope it is OK for e.g. Andrea.Disabling RETHUNK or IBT is not acceptable for a general-purpose kernel. If that constraint is introduced we either need to revert that patch in the Ubuntu kernel or disable Rust support. It would be nice to have a least something like CONFIG_RUST_IS_BROKEN_BUT_IM_HAPPY, off by default, and have `RUST_IS_BROKEN_BUT_IM_HAPPY || depends on !`.Should this actually be CONFIG_RUST_IS_BROKEN_ON_X86_BUT_IM_HAPPY ?
Just do the proper dependency on RETHUNK and you should be fine, it will be able to be enabled on arches that do not require RETHUNK for proper security.
At lest for networking, the code is architecture independent. For a driver to be useful, it needs to compile for most architectures. So i hope Rust will quickly make it to other architectures. And for PHY drivers, ARM and MIPS are probably more important than x86.
Is MIPS a proper target for rust yet? thanks, greg k-h