x86/microcode/AMD: Fix mixed steppings support

The AMD side of the loader has always claimed to support mixed
steppings. But somewhere along the way, it broke that by assuming that
the cached patch blob is a single one instead of it being one per
*node*.

So turn it into a per-node one so that each node can stash the blob
relevant for it.

  [ NB: Fixes tag is not really the exactly correct one but it is good
    enough. ]

Fixes: fe05589 ("x86/microcode: Merge the early microcode loader")
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Cc: <stable@kernel.org> # 2355370 ("x86/microcode/amd: Remove load_microcode_amd()'s bsp parameter")
Cc: <stable@kernel.org> # a5ad921 ("x86/microcode/AMD: Add a @cpu parameter to the reloading functions")
Link: https://lore.kernel.org/r/20230130161709.11615-4-bp@alien8.de

Author: bp3tk0v

arch/x86/kernel/cpu/microcode/amd.c

@@ -55,7 +55,9 @@ struct cont_desc {
 };
 
 static u32 ucode_new_rev;
-static u8 amd_ucode_patch[PATCH_MAX_SIZE];
+
+/* One blob per node. */
+static u8 amd_ucode_patch[MAX_NUMNODES][PATCH_MAX_SIZE];
 
 /*
  * Microcode patch container file is prepended to the initrd in cpio
@@ -428,7 +430,7 @@ static bool early_apply_microcode(u32 cpuid_1_eax, void *ucode, size_t size, boo
 	patch	= (u8 (*)[PATCH_MAX_SIZE])__pa_nodebug(&amd_ucode_patch);
 #else
 	new_rev = &ucode_new_rev;
-	patch	= &amd_ucode_patch;
+	patch	= &amd_ucode_patch[0];
 #endif
 
 	desc.cpuid_1_eax = cpuid_1_eax;
@@ -580,10 +582,10 @@ int __init save_microcode_in_initrd_amd(unsigned int cpuid_1_eax)
 
 void reload_ucode_amd(unsigned int cpu)
 {
-	struct microcode_amd *mc;
 	u32 rev, dummy __always_unused;
+	struct microcode_amd *mc;
 
-	mc = (struct microcode_amd *)amd_ucode_patch;
+	mc = (struct microcode_amd *)amd_ucode_patch[cpu_to_node(cpu)];
 
 	rdmsr(MSR_AMD64_PATCH_LEVEL, rev, dummy);
 
@@ -852,6 +854,8 @@ static enum ucode_state __load_microcode_amd(u8 family, const u8 *data,
 
 static enum ucode_state load_microcode_amd(u8 family, const u8 *data, size_t size)
 {
+	struct cpuinfo_x86 *c;
+	unsigned int nid, cpu;
 	struct ucode_patch *p;
 	enum ucode_state ret;
 
@@ -864,18 +868,22 @@ static enum ucode_state load_microcode_amd(u8 family, const u8 *data, size_t siz
 		return ret;
 	}
 
-	p = find_patch(0);
-	if (!p) {
-		return ret;
-	} else {
-		if (boot_cpu_data.microcode >= p->patch_id)
-			return ret;
+	for_each_node(nid) {
+		cpu = cpumask_first(cpumask_of_node(nid));
+		c = &cpu_data(cpu);
+
+		p = find_patch(cpu);
+		if (!p)
+			continue;
+
+		if (c->microcode >= p->patch_id)
+			continue;
 
 		ret = UCODE_NEW;
-	}
 
-	memset(amd_ucode_patch, 0, PATCH_MAX_SIZE);
-	memcpy(amd_ucode_patch, p->data, min_t(u32, p->size, PATCH_MAX_SIZE));
+		memset(&amd_ucode_patch[nid], 0, PATCH_MAX_SIZE);
+		memcpy(&amd_ucode_patch[nid], p->data, min_t(u32, p->size, PATCH_MAX_SIZE));
+	}
 
 	return ret;
 }