Support for parsing the topology on AMD/Hygon processors using CPUID leaf 0xb was added in commit 3986a0a805e6 ("x86/CPU/AMD: Derive CPU topology from CPUID function 0xB when available"). In an effort to keep all the topology parsing bits in one place, this commit also introduced a pseudo dependency on the TOPOEXT feature to parse the CPUID leaf 0xb.
TOPOEXT feature (CPUID 0x80000001 ECX[22]) advertises the support for Cache Properties leaf 0x8000001d and the CPUID leaf 0x8000001e EAX for "Extended APIC ID" however support for 0xb was introduced alongside the x2APIC support not only on AMD [1], but also historically on x86 [2].
Similar to 0xb, the support for extended CPU topology leaf 0x80000026 too does not depend on the TOPOEXT feature.
The support for these leaves is expected to be confirmed by ensuring "leaf <= {extended_}cpuid_level" and then parsing the level 0 of the respective leaf to confirm EBX[15:0] (LogProcAtThisLevel) is non-zero as stated in the definition of "CPUID_Fn0000000B_EAX_x00 [Extended Topology Enumeration] (Core::X86::Cpuid::ExtTopEnumEax0)" in Processor Programming Reference (PPR) for AMD Family 19h Model 01h Rev B1 Vol1 [3] Sec. 2.1.15.1 "CPUID Instruction Functions".
This has not been a problem on baremetal platforms since support for TOPOEXT (Fam 0x15 and later) predates the support for CPUID leaf 0xb (Fam 0x17[Zen2] and later), however, for AMD guests on QEMU, "x2apic" feature can be enabled independent of the "topoext" feature where QEMU expects topology and the initial APICID to be parsed using the CPUID leaf 0xb (especially when number of cores > 255) which is populated independent of the "topoext" feature flag.
Unconditionally call cpu_parse_topology_ext() on AMD and Hygon processors to first parse the topology using the XTOPOLOGY leaves (0x80000026 / 0xb) before using the TOPOEXT leaf (0x8000001e).
While at it, break down the single large comment in parse_topology_amd() to better highlight the purpose of each CPUID leaf.
Cc: stable@vger.kernel.org # Only v6.9 and above; Depends on x86 topology rewrite Link: https://lore.kernel.org/lkml/1529686927-7665-1-git-send-email-suravee.suthik... [1] Link: https://lore.kernel.org/lkml/20080818181435.523309000@linux-os.sc.intel.com/ [2] Link: https://bugzilla.kernel.org/show_bug.cgi?id=206537 [3] Suggested-by: Naveen N Rao (AMD) naveen@kernel.org Fixes: 3986a0a805e6 ("x86/CPU/AMD: Derive CPU topology from CPUID function 0xB when available") Signed-off-by: K Prateek Nayak kprateek.nayak@amd.com --- Changelog v4..v5:
o Made a note on only targeting versions >= v6.9 for stable backports since the fix depends on the x86 topology rewrite. (Boris)
o Renamed "has_topoext" to "has_xtopology". (Boris)
o Broke down the large comment in parse_topology_amd() to better highlight the purpose of each leaf and the overall parsing flow. (Boris) --- arch/x86/kernel/cpu/topology_amd.c | 25 ++++++++++++++----------- 1 file changed, 14 insertions(+), 11 deletions(-)
diff --git a/arch/x86/kernel/cpu/topology_amd.c b/arch/x86/kernel/cpu/topology_amd.c index 827dd0dbb6e9..c79ebbb639cb 100644 --- a/arch/x86/kernel/cpu/topology_amd.c +++ b/arch/x86/kernel/cpu/topology_amd.c @@ -175,27 +175,30 @@ static void topoext_fixup(struct topo_scan *tscan)
static void parse_topology_amd(struct topo_scan *tscan) { - bool has_topoext = false; - /* - * If the extended topology leaf 0x8000_001e is available - * try to get SMT, CORE, TILE, and DIE shifts from extended + * Try to get SMT, CORE, TILE, and DIE shifts from extended * CPUID leaf 0x8000_0026 on supported processors first. If * extended CPUID leaf 0x8000_0026 is not supported, try to - * get SMT and CORE shift from leaf 0xb first, then try to - * get the CORE shift from leaf 0x8000_0008. + * get SMT and CORE shift from leaf 0xb. If either leaf is + * available, cpu_parse_topology_ext() will return true. */ - if (cpu_feature_enabled(X86_FEATURE_TOPOEXT)) - has_topoext = cpu_parse_topology_ext(tscan); + bool has_xtopology = cpu_parse_topology_ext(tscan);
if (cpu_feature_enabled(X86_FEATURE_AMD_HTR_CORES)) tscan->c->topo.cpu_type = cpuid_ebx(0x80000026);
- if (!has_topoext && !parse_8000_0008(tscan)) + /* + * If XTOPOLOGY leaves (0x26/0xb) are not available, try to + * get the CORE shift from leaf 0x8000_0008 first. + */ + if (!has_xtopology && !parse_8000_0008(tscan)) return;
- /* Prefer leaf 0x8000001e if available */ - if (parse_8000_001e(tscan, has_topoext)) + /* + * Prefer leaf 0x8000001e if available to get the SMT shift and + * the initial APIC ID if XTOPOLOGY leaves are not available. + */ + if (parse_8000_001e(tscan, has_xtopology)) return;
/* Try the NODEID MSR */
On Mon, Sep 01 2025 at 17:04, K. Prateek Nayak wrote:
Unconditionally call cpu_parse_topology_ext() on AMD and Hygon processors to first parse the topology using the XTOPOLOGY leaves (0x80000026 / 0xb) before using the TOPOEXT leaf (0x8000001e).
While at it, break down the single large comment in parse_topology_amd() to better highlight the purpose of each CPUID leaf.
Cc: stable@vger.kernel.org # Only v6.9 and above; Depends on x86 topology rewrite Link: https://lore.kernel.org/lkml/1529686927-7665-1-git-send-email-suravee.suthik... [1] Link: https://lore.kernel.org/lkml/20080818181435.523309000@linux-os.sc.intel.com/ [2] Link: https://bugzilla.kernel.org/show_bug.cgi?id=206537 [3] Suggested-by: Naveen N Rao (AMD) naveen@kernel.org Fixes: 3986a0a805e6 ("x86/CPU/AMD: Derive CPU topology from CPUID function 0xB when available") Signed-off-by: K Prateek Nayak kprateek.nayak@amd.com
Changelog v4..v5:
o Made a note on only targeting versions >= v6.9 for stable backports since the fix depends on the x86 topology rewrite. (Boris)
Shouldn't that be backported? I think so, so leave that v6.9 and above comment out. The stable folks will notice that it does not apply to pre 6.9 kernels and send you a nice email asking you to provide a solution for pre 6.9 stable kernels.
Thanks,
tglx
Hello Thomas,
On 9/11/2025 1:40 AM, Thomas Gleixner wrote:
On Mon, Sep 01 2025 at 17:04, K. Prateek Nayak wrote:
Unconditionally call cpu_parse_topology_ext() on AMD and Hygon processors to first parse the topology using the XTOPOLOGY leaves (0x80000026 / 0xb) before using the TOPOEXT leaf (0x8000001e).
While at it, break down the single large comment in parse_topology_amd() to better highlight the purpose of each CPUID leaf.
Cc: stable@vger.kernel.org # Only v6.9 and above; Depends on x86 topology rewrite Link: https://lore.kernel.org/lkml/1529686927-7665-1-git-send-email-suravee.suthik... [1] Link: https://lore.kernel.org/lkml/20080818181435.523309000@linux-os.sc.intel.com/ [2] Link: https://bugzilla.kernel.org/show_bug.cgi?id=206537 [3] Suggested-by: Naveen N Rao (AMD) naveen@kernel.org Fixes: 3986a0a805e6 ("x86/CPU/AMD: Derive CPU topology from CPUID function 0xB when available") Signed-off-by: K Prateek Nayak kprateek.nayak@amd.com
Changelog v4..v5:
o Made a note on only targeting versions >= v6.9 for stable backports since the fix depends on the x86 topology rewrite. (Boris)
Shouldn't that be backported? I think so, so leave that v6.9 and above comment out. The stable folks will notice that it does not apply to pre 6.9 kernels and send you a nice email asking you to provide a solution for pre 6.9 stable kernels.
Ack! Since this is already in tip:x86/urgent as commit cba4262a19af ("x86/cpu/topology: Always try cpu_parse_topology_ext() on AMD/Hygon") let me know if I should resend it or that comment can be zapped in-place.
I can also send out a separate patch targeting stable with the intended changes. Since we are on the topic, here is the patch I would have sent out to stable:
(Note: Only tested on top of v6.6.105 stable on a Zen3 machine; no changes found in /sys/devices/system/cpu/cpu*/topology/* with the patch applied on top)
From: K Prateek Nayak kprateek.nayak@amd.com Subject: [PATCH stable] x86/cpu/amd: Always try detect_extended_topology() on AMD processors
commit cba4262a19afae21665ee242b3404bcede5a94d7 upstream.
Support for parsing the topology on AMD/Hygon processors using CPUID leaf 0xb was added in
3986a0a805e6 ("x86/CPU/AMD: Derive CPU topology from CPUID function 0xB when available").
In an effort to keep all the topology parsing bits in one place, this commit also introduced a pseudo dependency on the TOPOEXT feature to parse the CPUID leaf 0xb.
The TOPOEXT feature (CPUID 0x80000001 ECX[22]) advertises the support for Cache Properties leaf 0x8000001d and the CPUID leaf 0x8000001e EAX for "Extended APIC ID" however support for 0xb was introduced alongside the x2APIC support not only on AMD [1], but also historically on x86 [2].
The support for the 0xb leaf is expected to be confirmed by ensuring
leaf <= max supported cpuid_level
and then parsing the level 0 of the leaf to confirm EBX[15:0] (LogProcAtThisLevel) is non-zero as stated in the definition of "CPUID_Fn0000000B_EAX_x00 [Extended Topology Enumeration] (Core::X86::Cpuid::ExtTopEnumEax0)" in Processor Programming Reference (PPR) for AMD Family 19h Model 01h Rev B1 Vol1 [3] Sec. 2.1.15.1 "CPUID Instruction Functions".
This has not been a problem on baremetal platforms since support for TOPOEXT (Fam 0x15 and later) predates the support for CPUID leaf 0xb (Fam 0x17[Zen2] and later), however, for AMD guests on QEMU, the "x2apic" feature can be enabled independent of the "topoext" feature where QEMU expects topology and the initial APICID to be parsed using the CPUID leaf 0xb (especially when number of cores > 255) which is populated independent of the "topoext" feature flag.
Unconditionally call detect_extended_topology() on AMD processors to first parse the topology using the extended topology leaf 0xb before using the TOPOEXT leaf (0x8000001e).
Fixes: 3986a0a805e6 ("x86/CPU/AMD: Derive CPU topology from CPUID function 0xB when available") Suggested-by: Naveen N Rao (AMD) naveen@kernel.org Signed-off-by: K Prateek Nayak kprateek.nayak@amd.com Link: https://lore.kernel.org/lkml/1529686927-7665-1-git-send-email-suravee.suthik... [1] Link: https://lore.kernel.org/lkml/20080818181435.523309000@linux-os.sc.intel.com/ [2] Link: https://bugzilla.kernel.org/show_bug.cgi?id=206537 [3] --- For stable maintainers,
The original changes from commit cba4262a19af ("x86/cpu/topology: Always try cpu_parse_topology_ext() on AMD/Hygon") cannot be easily backported due to the extensive x86 topology rewrite in v6.9.
This patch cleanly applies on top of all stable kernels from v6.6.y to v5.4.y. Boris' S-o-b from commit commit cba4262a19af has been dropped since the changes on top of the stable kernels are slightly different. --- arch/x86/kernel/cpu/amd.c | 37 ++++++++++++++++++++++--------------- 1 file changed, 22 insertions(+), 15 deletions(-)
diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c index 864d62e94614..33d8bbdd7b69 100644 --- a/arch/x86/kernel/cpu/amd.c +++ b/arch/x86/kernel/cpu/amd.c @@ -391,34 +391,41 @@ static void legacy_fixup_core_id(struct cpuinfo_x86 *c) */ static void amd_get_topology(struct cpuinfo_x86 *c) { + /* + * Try to get the topology information from the 0xb leaf first. + * If detect_extended_topology() returns 0, parsing was successful + * and APIC ID, cpu_core_id, cpu_die_id, phys_proc_id, and + * __max_die_per_package are already populated. + */ + bool has_extended_topology = !detect_extended_topology(c); int cpu = smp_processor_id();
+ if (has_extended_topology) + c->x86_coreid_bits = get_count_order(c->x86_max_cores); + /* get information required for multi-node processors */ if (boot_cpu_has(X86_FEATURE_TOPOEXT)) { - int err; u32 eax, ebx, ecx, edx;
cpuid(0x8000001e, &eax, &ebx, &ecx, &edx);
- c->cpu_die_id = ecx & 0xff; - if (c->x86 == 0x15) c->cu_id = ebx & 0xff;
- if (c->x86 >= 0x17) { - c->cpu_core_id = ebx & 0xff; - - if (smp_num_siblings > 1) - c->x86_max_cores /= smp_num_siblings; - } - /* - * In case leaf B is available, use it to derive - * topology information. + * If the extended topology leaf 0xb leaf doesn't exits, + * derive CORE and DIE information from the 0x8000001e leaf. */ - err = detect_extended_topology(c); - if (!err) - c->x86_coreid_bits = get_count_order(c->x86_max_cores); + if (!has_extended_topology) { + c->cpu_die_id = ecx & 0xff; + + if (c->x86 >= 0x17) { + c->cpu_core_id = ebx & 0xff; + + if (smp_num_siblings > 1) + c->x86_max_cores /= smp_num_siblings; + } + }
cacheinfo_amd_init_llc_id(c, cpu);
base-commit: fe9731e100041bb2cc186717bde3e05ca175623b --
Let me know what you think.
On 9/11/2025 10:07 AM, K Prateek Nayak wrote:
/*
* In case leaf B is available, use it to derive* topology information.
* If the extended topology leaf 0xb leaf doesn't exits,* derive CORE and DIE information from the 0x8000001e leaf.
err = detect_extended_topology(c);if (!err)c->x86_coreid_bits = get_count_order(c->x86_max_cores);
if (!has_extended_topology) {c->cpu_die_id = ecx & 0xff;
Just noticed that we still need that "cpu_die_id" from 0x8000001e since 0xb on AMD does not implement the LEVEL_TYPE of DIE_TYPE (5).
if (c->x86 >= 0x17) {c->cpu_core_id = ebx & 0xff;if (smp_num_siblings > 1)c->x86_max_cores /= smp_num_siblings;}}
linux-stable-mirror@lists.linaro.org
-
K Prateek Nayak -
Thomas Gleixner