git: 7e7d4e711ff9 - main - x86: Handle when MPERF/APERF MSRs aren't writable

The branch main has been updated by ziaee:

URL: https://cgit.FreeBSD.org/src/commit/?id=7e7d4e711ff94d114c93fd522d4125aa9bd9f5cd

commit 7e7d4e711ff94d114c93fd522d4125aa9bd9f5cd
Author:     Matt Delco <delco@google.com>
AuthorDate: 2026-03-26 17:30:31 +0000
Commit:     Alexander Ziaee <ziaee@FreeBSD.org>
CommitDate: 2026-03-26 19:02:22 +0000

    x86: Handle when MPERF/APERF MSRs aren't writable
    
    For performance and/or correct reasons some hypervisors allow
    MPERF/APERF MSRs to be read but not written to. This change
    modifies the handling of these MSRs to not rely on writes.
    This patch is part of Google Cloud Engine (GCE) C4-LSSD turnup.
    
    Sponsored by:           Google
    Tested by:              NetApp (previous)
    PR:                     292808
    MFC after:              3 days
    Co-authored-by:         Jim Mattson <jmattson@google.com>
    Reviewed by:            jrtc27, imp, kib, markj, olce, obiwac
    Differential Revision:  https://reviews.freebsd.org/D55996
---
 sys/x86/x86/cpu_machdep.c | 14 +++++++-------
 sys/x86/x86/tsc.c         |  9 ++++++---
 2 files changed, 13 insertions(+), 10 deletions(-)

diff --git a/sys/x86/x86/cpu_machdep.c b/sys/x86/x86/cpu_machdep.c
index 775fee97d3cd..422bdf9cfb0d 100644
--- a/sys/x86/x86/cpu_machdep.c
+++ b/sys/x86/x86/cpu_machdep.c
@@ -423,7 +423,7 @@ int
 cpu_est_clockrate(int cpu_id, uint64_t *rate)
 {
 	uint64_t tsc1, tsc2;
-	uint64_t acnt, mcnt, perf;
+	uint64_t acnt_start, acnt_end, mcnt_start, mcnt_end, perf;
 	register_t reg;
 	int error = 0;
 
@@ -453,20 +453,20 @@ cpu_est_clockrate(int cpu_id, uint64_t *rate)
 	/* Calibrate by measuring a short delay. */
 	reg = intr_disable();
 	if (tsc_is_invariant) {
-		wrmsr(MSR_MPERF, 0);
-		wrmsr(MSR_APERF, 0);
+		mcnt_start = rdmsr(MSR_MPERF);
+		acnt_start = rdmsr(MSR_APERF);
 		tsc1 = rdtsc();
 		DELAY(1000);
-		mcnt = rdmsr(MSR_MPERF);
-		acnt = rdmsr(MSR_APERF);
+		mcnt_end = rdmsr(MSR_MPERF);
+		acnt_end = rdmsr(MSR_APERF);
 		tsc2 = rdtsc();
 		intr_restore(reg);
-		if (mcnt == 0) {
+		if (mcnt_end == mcnt_start) {
 			tsc_perf_stat = 0;
 			error = EOPNOTSUPP;
 			goto err;
 		}
-		perf = 1000 * acnt / mcnt;
+		perf = 1000 * (acnt_end - acnt_start) / (mcnt_end - mcnt_start);
 		*rate = (tsc2 - tsc1) * perf;
 	} else {
 		tsc1 = rdtsc();
diff --git a/sys/x86/x86/tsc.c b/sys/x86/x86/tsc.c
index 3b873d9dae73..f88ce60c6319 100644
--- a/sys/x86/x86/tsc.c
+++ b/sys/x86/x86/tsc.c
@@ -433,6 +433,8 @@ probe_tsc_freq_late(void)
 void
 start_TSC(void)
 {
+	uint64_t mperf, aperf;
+
 	if ((cpu_feature & CPUID_TSC) == 0 || tsc_disabled)
 		return;
 
@@ -442,11 +444,12 @@ start_TSC(void)
 		/*
 		 * XXX Some emulators expose host CPUID without actual support
 		 * for these MSRs.  We must test whether they really work.
+		 * They may also be read-only, so test for increment.
 		 */
-		wrmsr(MSR_MPERF, 0);
-		wrmsr(MSR_APERF, 0);
+		mperf = rdmsr(MSR_MPERF);
+		aperf = rdmsr(MSR_APERF);
 		DELAY(10);
-		if (rdmsr(MSR_MPERF) > 0 && rdmsr(MSR_APERF) > 0)
+		if (rdmsr(MSR_MPERF) != mperf && rdmsr(MSR_APERF) != aperf)
 			tsc_perf_stat = 1;
 	}