Skip to contentSkip to conversation

x86: pmu: fix failures on 32-bit due to wrong masks

PATCH In review Nadav Amit 3 participants 4 messages Jun 19, 2020 – Jun 29, 2020 KVM

lore
Share & tools
lore

Nadav Amit

author
#0 lore raw

PATCH 1/1 x86: pmu: fix failures on 32-bit due to wrong masks

Some mask computation are using long constants instead of long long constants, which causes test failures on x86-32.

Signed-off-by Nadav Amit <[email protected]>
---
 x86/pmu.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
x86/pmu.c +4 -4
index 91a6fb4..5a3d55b 100644
@@ -324,11 +324,11 @@ static void check_counter_overflow(void)
 
 		cnt.count = 1 - count;
 		if (gp_counter_base == MSR_IA32_PMC0)
-			cnt.count &= (1ul << eax.split.bit_width) - 1;
+			cnt.count &= (1ull << eax.split.bit_width) - 1;
 
 		if (i == num_counters) {
 			cnt.ctr = fixed_events[0].unit_sel;
-			cnt.count &= (1ul << edx.split.bit_width_fixed) - 1;
+			cnt.count &= (1ull << edx.split.bit_width_fixed) - 1;
 		}
 
 		if (i % 2)

@@ -456,7 +456,7 @@ static void check_running_counter_wrmsr(void)
 
 	count = -1;
 	if (gp_counter_base == MSR_IA32_PMC0)
-		count &= (1ul << eax.split.bit_width) - 1;
+		count &= (1ull << eax.split.bit_width) - 1;
 
 	wrmsr(gp_counter_base, count);
 

@@ -488,7 +488,7 @@ static void  check_gp_counters_write_width(void)
 {
 	u64 val_64 = 0xffffff0123456789ull;
 	u64 val_32 = val_64 & ((1ull << 32) - 1);
-	u64 val_max_width = val_64 & ((1ul << eax.split.bit_width) - 1);
+	u64 val_max_width = val_64 & ((1ull << eax.split.bit_width) - 1);
 	int i;
 
 	/*
Signature · 1 line
2.20.1
author
#1 lore raw
Quoted 7 lines · Nadav Amit

Some mask computation are using long constants instead of long long constants, which causes test failures on x86-32.

Signed-off-by Nadav Amit <[email protected]>

Paolo,

As you were so quick to respond to the other patches that I recent sent (despite some defections), I presume you missed this one.

(Otherwise, no rush).

maintainer
#2 lore raw
Quoted 8 lines · Nadav Amit
Quoted 7 lines · Nadav Amit

Some mask computation are using long constants instead of long long constants, which causes test failures on x86-32.

Signed-off-by Nadav Amit <[email protected]>
Quoted 7 lines

Paolo,

As you were so quick to respond to the other patches that I recent sent (despite some defections), I presume you missed this one.

(Otherwise, no rush).

Yes, indeed. Applied now, thanks!

Paolo