Re: [PATCH v4] RISC-V: Show accurate per-hart isa in /proc/cpuinfo

[Andrew Jones]

Hi Evan,

Here's my stab at new wording.

On Tue, Jul 11, 2023 at 01:18:30PM -0700, Evan Green wrote:
...
> diff --git a/Documentation/riscv/uabi.rst b/Documentation/riscv/uabi.rst
> index 8960fac42c40..afdda580e5a2 100644
> --- a/Documentation/riscv/uabi.rst
> +++ b/Documentation/riscv/uabi.rst
> @@ -42,6 +42,16 @@ An example string following the order is::
>  
>     rv64imadc_zifoo_zigoo_zafoo_sbar_scar_zxmbaz_xqux_xrux
>  
> +"isa" vs "hart isa" lines in /proc/cpuinfo
> +------------------------------------------
> +
> +The "isa" line in /proc/cpuinfo describes the lowest common denominator of
> +RISC-V ISA extensions understood by the kernel and implemented on all harts. The
> +"hart isa" line, in contrast, describes the set of extensions understood by the
> +kernel on the particular hart being described, even if those extensions may not
> +be present on all harts in the system. The "hart isa" line is consistent with
> +what's returned by __riscv_hwprobe() when querying for that specific CPU.
> +

The "isa" and "hart isa" lines in /proc/cpuinfo list RISC-V ISA extensions
which the kernel can identify (the kernel recognizes the extension's name)
and have not been filtered out due to effectively not being present.  An
extension is effectively not present when it is unusable, either due to
defects (which the kernel is aware of), due to missing information which
is necessary to complete the extension's description, or due to being
explicitly "hidden", such as when a kernel command line parameter
instructs the kernel to pretend the extension is not present.  Note, an
extension's presence in a list does not imply the kernel is using the
extension, nor does it imply that userspace or guest kernels may use the
extension (__riscv_hwprobe() should be queried for userspace usability.
The hypervisor should be queried, using hypervisor-specific APIs, to
check guest kernel usability.)

The "isa" line describes the lowest common denominator of extensions,
which are the extensions implemented on all harts.  In contrast, the
extensions listed in the "hart isa" line need not be implemented by
any other hart than the hart corresponding to the line.

---

I've specifically dropped the 'The "hart isa" line is consistent with
what's returned by __riscv_hwprobe()...' part because I suspect hwprobe
could at least lag what gets put in "hart isa", since the kernel may be
taught about an extension for a different purpose first, neglecting
hwprobe. And, there may be cases that hwprobe never enumerates an
extension which the kernel does.

Thanks,
drew