Re: Adding V-ext regs to signal context w/o expanding kernel struct sigcontext to avoid glibc ABI break

[Kito Cheng <kito.cheng@sifive.com>]

Hi Vineet:

> >>>> The new Kconfig CONFIG_RISCV_VSTATE_INIT_ALL seems like a
> >>>> hack bolted on top.
> >>> IIUC, most opinions suggested that we should keep the default Vector
> >>> state to ON in thread:
> >>> https://lore.kernel.org/all/20220921214439.1491510-17-stillson@rivosinc.com/T/#u
> >> Actually community feedback is that they *don't * want the default
> >> vector state to be on due to power implications, increased stack and
> >> memory usage for vector contents (in that thread and else where as
> >> well). So we should keep it disabled by default, but indeed we could
> >> have that Kconfig option to enable it. Granted distro kernels will keep
> >> it disabled by default, this lets vendors enable it selectively until
> >> the full userspace enabling bits are in place.
> > Should we punt this to the ELF (e.g., using a RISC-V specific
> > attribute) and take a per-process decision on whether to start in ON
> > or OFF?
> > I don't feel fully comfortable with a KCONFIG that could change and
> > invalidate the assumptions a userspace process could have made…
>
> The Kconfig is just a stop gap for vendors to enable V development while
> the full userspace stuff is sorted out.
>
> Indeed RISCV_ATTRIBUTES section has -march info, but we need to do some
> development around it to parse it and use it.

I don't think RISCV_ATTRIBUTES is the right place to check that - even if the
program compiles without V, it still can enable V and then get performance
benefit by ifunc in glibc, or even some 3rd party libraries might also be
optimized with V ext.

And don't forget other shared libraries in the system,
are we going to check all dependent libraries at program load time?
it will require resolving the library dependency at kernel.
Or we intend to enable V only if executable compiles with V?

> There are still corner cases such as non-V executable dlopen a dso - so
> kernel elf parser doing this might not cover all cases.
>
> Similar logic will need to be added to glibc loader - eventually.
>
> Adding the full plumbing is a chicken-and-egg problem.
>
>
> > Alternatively, we could establish the convention of having two stub
> > libraries that set up either enabled or disable state from their
> > .init_array to provide a mechanism for folks that want to make an
> > explicit assumption.  Although this may try to overdesign a solution
> > for a non-issue.
>
> I was thinking more along the lines of x86 GLIBC_TUNABLES to enable it
> via env/sub-shell on a per-task basis - the tunable hook could in turn
> verify that Vector support does exist - or it could invoke the prctl
> unconditionally (which would fail if V didn't exist etc).