Skip to contentSkip to conversation

x86/microcode: Check microcode revision before updating sibling threads

PATCH In review Ashok Raj 3 participants 5 messages Feb 16, 2018 – Feb 17, 2018

Subsystems (2)
lore
Share & tools
lore

Ashok Raj

author
#0 lore raw

PATCH 1/1 x86/microcode: Check microcode revision before updating sibling threads

After updating microcode on one of the threads in the core, the thread sibling automatically gets the update since the microcode resources are shared. Check the ucode revision on the cpu before performing a ucode update.

---
 arch/x86/kernel/cpu/microcode/intel.c | 11 +++++++++++
 1 file changed, 11 insertions(+)
arch/x86/kernel/cpu/microcode/intel.c +11 -0
index 09b95a7..5802c2f 100644
@@ -786,11 +786,21 @@ static enum ucode_state apply_microcode_intel(int cpu)
 
 	uci = ucode_cpu_info + cpu;
 	mc = uci->mc;
+
 	if (!mc) {
 		/* Look for a newer patch in our cache: */
 		mc = find_patch(uci);
 		if (!mc)
 			return UCODE_NFOUND;
+	} else {
+		rev = intel_get_microcode_revision();
+		/*
+		 * Its possible the microcode got udpated
+		 * because its sibling update was done earlier.
+		 * Skip the udpate in that case.
+		 */
+		if (rev == mc->hdr.rev)
+			goto done;
 	}
 
 	/* write microcode via MSR 0x79 */
@@ -813,6 +823,7 @@ static enum ucode_state apply_microcode_intel(int cpu)
 		prev_rev = rev;
 	}
 
+done:
 	c = &cpu_data(cpu);
 
 	uci->cpu_sig.rev = rev;
Signature · 1 line
2.7.4
#1 lore raw
Quoted 26 lines · Ashok Raj

After updating microcode on one of the threads in the core, the thread sibling automatically gets the update since the microcode resources are shared. Check the ucode revision on the cpu before performing a ucode update.

---
 arch/x86/kernel/cpu/microcode/intel.c | 11 +++++++++++
 1 file changed, 11 insertions(+)
arch/x86/kernel/cpu/microcode/intel.c +1 -0
index 09b95a7..5802c2f 100644
@@ -786,11 +786,21 @@ static enum ucode_state apply_microcode_intel(int cpu)
 
 	uci = ucode_cpu_info + cpu;
 	mc = uci->mc;
+
 	if (!mc) {
 		/* Look for a newer patch in our cache: */
 		mc = find_patch(uci);
 		if (!mc)
 			return UCODE_NFOUND;

arch/x86/kernel/cpu/microcode/intel.c static enum ucode_state apply_microcode_intel(int cpu) · line 804

Quoted 10 lines
arch/x86/kernel/cpu/microcode/intel.c +9 -0
+	} else {
+		rev = intel_get_microcode_revision();
+		/*
+		 * Its possible the microcode got udpated
+		 * because its sibling update was done earlier.
+		 * Skip the udpate in that case.
+		 */
+		if (rev == mc->hdr.rev)
+			goto done;
 	}

Make that like the AMD version:

		/*
		 * Its possible the microcode got udpated
		 * because its sibling update was done earlier.
		 * Skip the udpate in that case.
		 */
		rev = intel_get_microcode_revision();
		if (rev >= mc->hdr.rev) {
			uci->cpu_sig.rev = rev;
			c->microcode = rev;

			return UCODE_OK;
		}

and make the subject prefix "x86/microcode/intel:" for your next submission.

Thx.

Signature · 5 lines
Regards/Gruss,
    Boris.

SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
-- 
#2 lore raw
In reply to Ashok Raj
Quoted 4 lines · Ashok Raj

After updating microcode on one of the threads in the core, the thread sibling automatically gets the update since the microcode resources are shared. Check the ucode revision on the cpu before performing a ucode update.

I wonder when this broke, because it used to do that just fine.

I even had to explain to a couple Debian users that no, nothing was broken and yes, while it looked like the kernel was skipping some "CPUs", it was actually updating all of them...

Signature · 1 line
  Henrique Holschuh
#3 lore raw
Quoted 1 line · Henrique de Moraes Holschuh

I wonder when this broke,

It didn't break - we're just printing the update only once.

Signature · 5 lines
Regards/Gruss,
    Boris.

SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
-- 
#4 lore raw
Quoted 4 lines · Borislav Petkov
Quoted 1 line · Henrique de Moraes Holschuh

I wonder when this broke,

It didn't break - we're just printing the update only once.

It was before that change, when the microcode subsystem still printed a message for every CPU that got updated or early-updated.

It was fixed to properly refresh the microcode revision in the percpu area right before deciding whether it should update that cpu or not, and thus it would skip the second thread on Intel cores because it would notice it was already up-to-date.

Maybe it was just the late loader, though... I'd need to do some git log digging.

Signature · 1 line
  Henrique Holschuh