mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH v1] perf arch x86: Zero-pad CPUID model for GenuineIntel in get_cpuid_str()
@ 2026-07-15 21:11 Chun-Tse Shao
  0 siblings, 0 replies; only message in thread
From: Chun-Tse Shao @ 2026-07-15 21:11 UTC (permalink / raw)
  To: irogers, acme, namhyung
  Cc: peterz, mingo, alexander.shishkin, jolsa, adrian.hunter,
	james.clark, afaerber, mani, dapeng1.mi, thomas.falcon,
	linux-perf-users, linux-kernel, Chun-Tse Shao

Change the model formatting in get_cpuid_str() for GenuineIntel CPUs from
"%X" to "%02X", while preserving unpadded "%X" for non-Intel vendors.

Intel vendor events in mapfile.csv (generated by create_perf_json.py
from the upstream perfmon repository) use 2-character hex model strings
(e.g. "GenuineIntel-6-9A", "GenuineIntel-6-BE", "GenuineIntel-18-01").
When an Intel model number is a single hex digit (e.g. 0x01 or 0x03),
formatting with "%X" yields "GenuineIntel-18-1", which fails to match
mapfile entries expecting "01".

Non-Intel vendors (such as AMD) use unpadded model numbers in mapfile.csv
(e.g. "AuthenticAMD-23-([12][0-9A-F]|[0-9A-F])"). Zero-padding non-Intel
vendors would break AMD CPUID matching by causing single-digit AMD models
(like Zen 1) to match "01" against "AuthenticAMD-23-[[:xdigit:]]+" (Zen 2).

Restricting "%02X" zero-padding specifically to GenuineIntel CPUs ensures
both Intel and AMD CPUID strings match their respective mapfile entries
without regressions.

Signed-off-by: Chun-Tse Shao <ctshao@google.com>
---
 tools/perf/arch/x86/util/header.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/tools/perf/arch/x86/util/header.c b/tools/perf/arch/x86/util/header.c
index 412977f8aa83..4b6d619fb8f9 100644
--- a/tools/perf/arch/x86/util/header.c
+++ b/tools/perf/arch/x86/util/header.c
@@ -66,8 +66,17 @@ get_cpuid(char *buffer, size_t sz, struct perf_cpu cpu __maybe_unused)
 char *get_cpuid_str(struct perf_cpu cpu __maybe_unused)
 {
 	char *buf = malloc(128);
+	char vendor[16];
+	unsigned int lvl;
+
+	if (!buf)
+		return NULL;
+
+	get_cpuid_0(vendor, &lvl);
+	/* Intel mapfile entries use 2-character hex model numbers */
+	const char *fmt = !strcmp(vendor, "GenuineIntel") ? "%s-%u-%02X-%X$" : "%s-%u-%X-%X$";
 
-	if (buf && __get_cpuid(buf, 128, "%s-%u-%X-%X$") < 0) {
+	if (__get_cpuid(buf, 128, fmt) < 0) {
 		free(buf);
 		return NULL;
 	}
-- 
2.55.0.141.g00534a21ce-goog


^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2026-07-15 21:11 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-07-15 21:11 [PATCH v1] perf arch x86: Zero-pad CPUID model for GenuineIntel in get_cpuid_str() Chun-Tse Shao

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox