mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Ian Rogers <irogers@google.com>
To: irogers@google.com, acme@kernel.org, namhyung@kernel.org
Cc: adrian.hunter@intel.com, haghdoost@uber.com,
	james.clark@linaro.org,  jolsa@kernel.org,
	linux-kernel@vger.kernel.org,  linux-perf-users@vger.kernel.org,
	mingo@redhat.com, peterz@infradead.org
Subject: [PATCH v2 02/18] perf demangle: Add fast-path prefix checks for Rust and Java demanglers
Date: Tue, 27 Oct 2026 16:14:56 -0700	[thread overview]
Message-ID: <20261027231513.816148-3-irogers@google.com> (raw)
In-Reply-To: <20261027231513.816148-1-irogers@google.com>

During DSO symbol loading in dso__load_sym_internal(), dso__demangle_sym()
is called for every ELF symbol. Both rust_demangle_demangle() and
java_demangle_sym() previously performed string searches or buffer
allocations before checking whether the symbol name could possibly be a
mangled Rust or Java symbol, adding unnecessary overhead on plain C
symbols.

Add fast-path prefix checks in dso__demangle_sym() before calling
rust_demangle_demangle() (using strstarts() for '_ZN', 'ZN', '__ZN',
'_R', 'R', or '__R', keeping demangle-rust-v0.c unmodified relative to
upstream rustc-demangle) and in java_demangle_sym() (requiring 'L', ';',
'(', and ')') to reject non-Rust and non-Java symbol names immediately.

Using 'perf bench internals symbols':

  Before (perf binary):
    Average DSO load + sort: 7.348 msec (+- 0.069 msec)
  After (perf binary, -4.2%):
    Average DSO load + sort: 7.041 msec (+- 0.051 msec)

  Before (libLLVM.so.21.1):
    Average DSO load + sort: 81.196 msec (+- 0.780 msec)
  After (libLLVM.so.21.1, -3.2%):
    Average DSO load + sort: 78.599 msec (+- 0.857 msec)

Signed-off-by: Ian Rogers <irogers@google.com>
---
 tools/perf/util/demangle-java.c | 23 +++++++++++++++++++++--
 tools/perf/util/symbol.c        | 15 ++++++++++++++-
 2 files changed, 35 insertions(+), 3 deletions(-)

diff --git a/tools/perf/util/demangle-java.c b/tools/perf/util/demangle-java.c
index c3cb327ed562..5625c4de39eb 100644
--- a/tools/perf/util/demangle-java.c
+++ b/tools/perf/util/demangle-java.c
@@ -158,14 +158,33 @@ char *
 java_demangle_sym(const char *str, int flags)
 {
 	char *buf, *ptr;
-	const char *p;
+	const char *l, *semi, *open, *p;
 	size_t len, l1 = 0;
 
 	if (!str)
 		return NULL;
 
+	/*
+	 * OpenJDK JIT symbols have the form:
+	 *   [prefix]L<class>;<method>(<args>)<return_type>
+	 * where the return type (such as 'I' for int or 'V' for void) is at the
+	 * end after ')'. In __demangle_java_sym(), the parser starts in
+	 * MODE_PREFIX and only reaches MODE_TYPE (required when ')' is reached)
+	 * via 'L' (MODE_CLASS) -> ';' (MODE_FUNC) -> '(' (MODE_TYPE). Verify
+	 * those delimiters appear in order before allocating the output buffer.
+	 */
+	l = strchr(str, 'L');
+	if (!l)
+		return NULL;
+	semi = strchr(l + 1, ';');
+	if (!semi)
+		return NULL;
+	open = strchr(semi + 1, '(');
+	if (!open)
+		return NULL;
+
 	/* find start of return type */
-	p = strrchr(str, ')');
+	p = strrchr(open + 1, ')');
 	if (!p)
 		return NULL;
 
diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
index b61a574ce9a3..dc70bbe8514f 100644
--- a/tools/perf/util/symbol.c
+++ b/tools/perf/util/symbol.c
@@ -2738,7 +2738,20 @@ char *dso__demangle_sym(struct dso *dso, int kmodule, const char *elf_name)
 	if (!want_demangle((dso && dso__kernel(dso)) || kmodule))
 		return demangled;
 
-	rust_demangle_demangle(elf_name, &rust_demangle);
+	/*
+	 * Rust mangled symbols start with one of the prefixes checked by
+	 * rust_demangle_legacy_demangle() ("_ZN", Windows dbghelp "ZN", or
+	 * macOS "__ZN") or rust_demangle_v0_demangle() ("_R", Windows dbghelp
+	 * "R", or macOS "__R") in demangle-rust-v0.c. Check the prefix here so
+	 * we avoid calling rust_demangle_demangle() (which scans the entire
+	 * symbol name with strstr() and strlen() for ThinLTO ".llvm." suffixes)
+	 * on non-Rust symbols, while keeping demangle-rust-v0.c unmodified
+	 * relative to upstream rustc-demangle.
+	 */
+	if (strstarts(elf_name, "_ZN") || strstarts(elf_name, "ZN") ||
+	    strstarts(elf_name, "__ZN") || strstarts(elf_name, "_R") ||
+	    strstarts(elf_name, "R") || strstarts(elf_name, "__R"))
+		rust_demangle_demangle(elf_name, &rust_demangle);
 	if (rust_demangle_is_known(&rust_demangle)) {
 		/* A rust mangled name. */
 		if (rust_demangle.mangled_len == 0)
-- 
2.56.0.313.g2479322e68-goog


  parent reply	other threads:[~2026-10-27 23:15 UTC|newest]

Thread overview: 36+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-09-28  7:52 [PATCH v1 0/7] perf symbol: Reference counting, flat array storage, and LRU shrinking Ian Rogers
2026-09-28  7:52 ` [PATCH v1 1/7] perf symbol: Add accessor functions for struct symbol fields Ian Rogers
2026-09-28  7:52 ` [PATCH v1 2/7] perf symbol: Remove symbol_conf.priv_size and negative-offset allocations Ian Rogers
2026-09-28  7:52 ` [PATCH v1 3/7] perf symbol: Switch backing storage from rbtree to struct symbols array Ian Rogers
2026-09-28  7:52 ` [PATCH v1 4/7] perf symbol: Add reference counting and DECLARE_RC_STRUCT(symbol) Ian Rogers
2026-09-28  7:52 ` [PATCH v1 5/7] perf symbol: Add LRU memory shrinking for symbols, DSOs, and machines Ian Rogers
2026-09-28  7:52 ` [PATCH v1 6/7] perf session: Periodically shrink symbols and DSOs during event processing Ian Rogers
2026-09-28  7:52 ` [PATCH v1 7/7] perf test symbols: Add tests for symbol and DSO LRU shrinking Ian Rogers
2026-09-28 15:37 ` [PATCH v1 0/7] perf symbol: Reference counting, flat array storage, and " Ian Rogers
2026-09-28 19:45 ` Alireza Haghdoost
2026-09-28 20:55   ` Ian Rogers
2026-09-28 21:51     ` Namhyung Kim
2026-09-28 22:04       ` Ian Rogers
2026-09-29  0:12         ` Alireza Haghdoost
2026-09-29 18:09           ` Namhyung Kim
2026-09-29 21:24             ` Alireza Haghdoost
2026-09-29 18:06         ` Namhyung Kim
2026-10-27 23:14 ` [PATCH v2 00/18] perf symbol: Reference counting, sorted arrays, and memory shrinking Ian Rogers
2026-10-27 23:14   ` [PATCH v2 01/18] perf bench internals: Add symbols and maps benchmarks Ian Rogers
2026-10-27 23:14   ` Ian Rogers [this message]
2026-10-27 23:14   ` [PATCH v2 03/18] perf symbol: Add accessor functions for struct symbol fields Ian Rogers
2026-10-27 23:14   ` [PATCH v2 04/18] perf map_browser: Compute symbol browser index on demand Ian Rogers
2026-10-27 23:14   ` [PATCH v2 05/18] perf db-export: Store symbol database IDs in a hashmap Ian Rogers
2026-10-27 23:15   ` [PATCH v2 06/18] perf annotate: Move struct annotation to a hashmap and remove symbol_conf.priv_size Ian Rogers
2026-10-27 23:15   ` [PATCH v2 07/18] perf symbol: Switch backing storage from rbtree to struct symbols array Ian Rogers
2026-10-27 23:15   ` [PATCH v2 08/18] perf symbol: Add reference counting and DECLARE_RC_STRUCT(symbol) Ian Rogers
2026-10-27 23:15   ` [PATCH v2 09/18] perf symbol: Optimize symbol sorting with natural-run merge sort and binary insertion Ian Rogers
2026-10-27 23:15   ` [PATCH v2 10/18] perf maps: Optimize tail insertion and small-append sorting Ian Rogers
2026-10-27 23:15   ` [PATCH v2 11/18] perf symbol-elf: Cache ELF program/section headers and defer symbol name lookup Ian Rogers
2026-10-27 23:15   ` [PATCH v2 12/18] perf rwsem: Add down_read_trylock() and down_write_trylock() Ian Rogers
2026-10-27 23:15   ` [PATCH v2 13/18] perf vdso: Constify dso parameter in dso__is_vdso() Ian Rogers
2026-10-27 23:15   ` [PATCH v2 14/18] perf machine: Protect machines->guests rbtree with a rw_semaphore Ian Rogers
2026-10-27 23:15   ` [PATCH v2 15/18] perf symbol: Add memory shrinking for symbols, DSOs, and machines Ian Rogers
2026-10-27 23:15   ` [PATCH v2 16/18] perf session: Add dso__session() and session-scoped allocation wrappers Ian Rogers
2026-10-27 23:15   ` [PATCH v2 17/18] perf symbol: Use session allocation wrappers in symbol, DSO, thread, and hist paths Ian Rogers
2026-10-27 23:15   ` [PATCH v2 18/18] perf test symbols: Add tests for symbol shrinking and session alloc wrappers Ian Rogers

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20261027231513.816148-3-irogers@google.com \
    --to=irogers@google.com \
    --cc=acme@kernel.org \
    --cc=adrian.hunter@intel.com \
    --cc=haghdoost@uber.com \
    --cc=james.clark@linaro.org \
    --cc=jolsa@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-perf-users@vger.kernel.org \
    --cc=mingo@redhat.com \
    --cc=namhyung@kernel.org \
    --cc=peterz@infradead.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®