Re: [PATCH v19 00/27] riscv control-flow integrity for usermode

[Deepak Gupta]

On Fri, Aug 08, 2025 at 12:48:31PM +0100, Mark Brown wrote:
> On Fri, Aug 08, 2025 at 01:23:15AM -0700, Deepak Gupta wrote:
> > On Thu, Aug 07, 2025 at 01:28:36PM +0100, Mark Brown wrote:
>
> > > Do you have an update for my clone3() shadow
>
> > No I don't.
>
> > > stack series that I could roll in for when I repost that after the merge
> > > window, and/or instructions for how to run this stuff for RISC-V on some
> > > emulated platform?
>
> > I would want to write-up instructions. But I don't want you to go through
> > a lot of hassle of building toolchain and bunch of other stuff.
> > Let me see how I can make it easy for you. Will report back.
>
> Thanks.  FWIW I should already be sorted for the kernel build, unless
> there's a super new or specialist toolchain required for this feature

Unlike x86 shadow stack and arm's GCS, push on RISC-V shadow stack is done
using dedicated `sspush` instruction and pop is done using instruction
`sspopchk`. RISC-V vDSO has certain `C` files and they would need shadow stack
push and pop (along with landing pad on them). So that's why kernel compile
would require the toolchain. I'll point you to it.

> (I'd guess I should be fine for the shadow stacks bit?) - it's userspace
> and emulation for the extension I'm missing.

Qemu changes are upstream. I'll provide the command line option.
On userspace, I'll point you to something pre-compiled rootfs.