On Thu, May 30, 2024 at 02:35:51PM -0700, Palmer Dabbelt wrote:
On Thu, 09 May 2024 09:26:56 PDT (-0700), andy.chiu@sifive.com wrote:
diff --git a/arch/riscv/kernel/sys_hwprobe.c b/arch/riscv/kernel/sys_hwprobe.c index 969ef3d59dbe..35390b4a5a17 100644 --- a/arch/riscv/kernel/sys_hwprobe.c +++ b/arch/riscv/kernel/sys_hwprobe.c @@ -114,6 +114,11 @@ static void hwprobe_isa_ext0(struct riscv_hwprobe *pair, EXT_KEY(ZIHINTPAUSE);
if (has_vector()) {
EXT_KEY(ZVE32X);EXT_KEY(ZVE32F);EXT_KEY(ZVE64X);EXT_KEY(ZVE64F);EXT_KEY(ZVE64D); EXT_KEY(ZVBB); EXT_KEY(ZVBC); EXT_KEY(ZVKB);Conor left a comment over here https://lore.kernel.org/all/20240510-zve-detection-v5-6-0711bdd26c12@sifive.com/.
This link is to the patch you're replying to, not anything from me. I commented on a bunch of stuff in v4, but not this patch - generally I ignore hwprobe to be honest...
I think the best bet is to just merge this v5 on for-next now, though -- there's a bunch of patch sets touching ISA string parsing and IIUC that sub-extension parsing stuff is a pre-existing issue, and Clement's patch set still has some outstanding feedback to address.
So I think if we just go with this we're not regressing anything, we just have a bit more to clean up. Maybe it's a little uglier now that userspace can see the sub-extensions, but I'd bet wacky ISA strings will be able to confuse us for a while.
I wanna do some cleanup stuff w/ Clements series applied, if that's what you were talking about, but I don't see much point starting that until the cpufeature stuff has calmed down - Charlie's and Clement's series really need to be in for-next for it to be worth doing.
I staged this so I can throw it at the tester, LMK if anyone has issues otherwise it'll show up on for-next.