Skip to contentSkip to conversation

x86/TSC: Use RDTSCP

RFC PATCH Alexey Dobriyan 2 participants 3 messages Dec 11, 2018 – Dec 12, 2018

lore
Share & tools
lore
#0 lore raw
Quoted 3 lines

And I had a bit of a WTF moment, as in "WTF does RDTSC have to do with a speculation protection barrier". Does it actually make sense?

It doesn't. There was too much s/lfence/barrier_nospec/ apparently.

Review

Quoted 7 lines
(quoted diff) +7 -0
+	asm volatile(ALTERNATIVE_3("rdtsc",
+				   "mfence; rdtsc", X86_FEATURE_MFENCE_RDTSC,
+				   "lfence; rdtsc", X86_FEATURE_LFENCE_RDTSC,
+				   "rdtscp", X86_FEATURE_RDTSCP)
+			: EAX_EDX_RET(val, low, high)
+			/* RDTSCP clobbers ECX with MSR_TSC_AUX. */
+			:: "ecx");

I have a question: does alternatives ordering matter? CPU can have both features.

And other code in this file uses "c" for clobber.

#1 lore raw
Quoted 1 line · Alexey Dobriyan

I have a question: does alternatives ordering matter?

Yes.

Quoted 1 line

CPU can have both features.

The last present feature wins.

Quoted 1 line

And other code in this file uses "c" for clobber.

Ok.

Signature · 4 lines
Regards/Gruss,
    Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.
#2 lore raw
Quoted 3 lines · Borislav Petkov
Quoted 1 line

And other code in this file uses "c" for clobber.

Ok.

Except that clobbers need to be full register names:

./arch/x86/include/asm/msr.h: In function ‘do_hres’:
./arch/x86/include/asm/msr.h:236:2: error: unknown register name ‘c’ in ‘asm’
asm volatile(ALTERNATIVE_3("rdtsc",
^~~

...

Signature · 4 lines
Regards/Gruss,
    Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.