Re: [REGRESSION][BISECTED] Scheduling errors with the AMD FX 8300 CPU

[Thomas Gleixner <tglx@linutronix.de>]

Tim!

On Mon, May 27 2024 at 18:36, Tim Teichmann wrote:
> Right here is the output of the
> Basic Leafs :
> ================
> 0x00000000: EAX=0x0000000d, EBX=0x68747541, ECX=0x444d4163, EDX=0x69746e65
> 0x00000001: EAX=0x00600f20, EBX=0x00080800, ECX=0x3e98320b, EDX=0x178bfbff
> 0x00000005: EAX=0x00000040, EBX=0x00000040, ECX=0x00000003, EDX=0x00000000
> 0x00000006: EAX=0x00000000, EBX=0x00000000, ECX=0x00000001, EDX=0x00000000
> 0x00000007: subleafs:
>   0: EAX=0x00000000, EBX=0x00000008, ECX=0x00000000, EDX=0x00000000
> 0x0000000d: subleafs:
>   0: EAX=0x00000007, EBX=0x00000340, ECX=0x000003c0, EDX=0x40000000
>   2: EAX=0x00000100, EBX=0x00000240, ECX=0x00000000, EDX=0x00000000
> Extended Leafs :
> ================
> 0x80000000: EAX=0x8000001e, EBX=0x68747541, ECX=0x444d4163, EDX=0x69746e65
> 0x80000001: EAX=0x00600f20, EBX=0x10000000, ECX=0x01eb3fff, EDX=0x2fd3fbff
> 0x80000002: EAX=0x20444d41, EBX=0x74285846, ECX=0x382d296d, EDX=0x20303033
> 0x80000003: EAX=0x68676945, EBX=0x6f432d74, ECX=0x50206572, EDX=0x65636f72
> 0x80000004: EAX=0x726f7373, EBX=0x20202020, ECX=0x20202020, EDX=0x00202020
> 0x80000005: EAX=0xff40ff18, EBX=0xff40ff30, ECX=0x10040140, EDX=0x40020140
> 0x80000006: EAX=0x64006400, EBX=0x64004200, ECX=0x08008140, EDX=0x0040c140
> 0x80000007: EAX=0x00000000, EBX=0x00000000, ECX=0x00000000, EDX=0x000007d9
> 0x80000008: EAX=0x00003030, EBX=0x00001000, ECX=0x00004007, EDX=0x00000000
> 0x8000000a: EAX=0x00000001, EBX=0x00010000, ECX=0x00000000, EDX=0x00001cff
> 0x80000019: EAX=0xf040f018, EBX=0x64006400, ECX=0x00000000, EDX=0x00000000
> 0x8000001a: EAX=0x00000003, EBX=0x00000000, ECX=0x00000000, EDX=0x00000000
> 0x8000001b: EAX=0x000000ff, EBX=0x00000000, ECX=0x00000000, EDX=0x00000000
> 0x8000001c: EAX=0x00000000, EBX=0x80032013, ECX=0x00010200, EDX=0x8000000f
> 0x8000001d: subleafs:
>   0: EAX=0x00000121, EBX=0x00c0003f, ECX=0x0000003f, EDX=0x00000000
>   1: EAX=0x00004122, EBX=0x0040003f, ECX=0x000001ff, EDX=0x00000000
>   2: EAX=0x00004143, EBX=0x03c0003f, ECX=0x000007ff, EDX=0x00000001
>   3: EAX=0x0001c163, EBX=0x0fc0003f, ECX=0x000007ff, EDX=0x00000001
> 0x8000001e: EAX=0x00000000, EBX=0x00000100, ECX=0x00000000, EDX=0x00000000

Duh. 0x8000001e claims that there are two SMT threads. The original code
only evaluates that part for family >= 0x17. I missed to forward that
condition.

Fix below.

Thanks,

        tglx
---
diff --git a/arch/x86/kernel/cpu/topology_amd.c b/arch/x86/kernel/cpu/topology_amd.c
index d419deed6a48..ebf1cefb1cb2 100644
--- a/arch/x86/kernel/cpu/topology_amd.c
+++ b/arch/x86/kernel/cpu/topology_amd.c
@@ -84,9 +84,9 @@ static bool parse_8000_001e(struct topo_scan *tscan, bool has_topoext)
 
 	/*
 	 * If leaf 0xb is available, then the domain shifts are set
-	 * already and nothing to do here.
+	 * already and nothing to do here. Only valid for family >= 0x17.
 	 */
-	if (!has_topoext) {
+	if (!has_topoext && c->x86 >= 0x17) {
 		/*
 		 * Leaf 0x80000008 set the CORE domain shift already.
 		 * Update the SMT domain, but do not propagate it.