From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from sipsolutions.net (s3.sipsolutions.net [168.119.38.16]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id EC7DE36B92E; Mon, 21 Sep 2026 10:15:47 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=168.119.38.16 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1789985750; cv=none; b=g6+ZY5VpCRiUCIDQbct+YSOEUE03dkC1KqX9dywXcKo9aGkrytdHHGVwHjjTpL/qm5anbjgf0q5sTpi1cxNtnBzX5kLfY+FrRCUOGGNf7K6jUj1k+uT3vaSLgIqN+6uto3DpHyagJZ4cA/Mxh6tEIh+6SasluYRvMuU4e1K7htI= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1789985750; c=relaxed/simple; bh=zZrDW3SHHmLudM5KCemxSzMyjiOJhBFvYJ3SWRiP5d4=; h=From:To:Cc:Subject:Date:Message-ID:In-Reply-To:References: MIME-Version; b=CWqROrNzFI3UUuJCI7gXkG7NAeKnHd8OJQCNRpM93MR1QLg4GAuq1zq0+0ntg4UL7WqDm0LGCUMb7nGcWNKVIAJHvqYLfh3Agl8VIrLTuUyAVtkdeClPl2iVXcOCcSi0Zf8mTrMkVOGdLoiUXYXg+uVKsH7v/cieaXgjT+sEnhM= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dmarc=permerror header.from=sipsolutions.net; spf=pass smtp.mailfrom=sipsolutions.net; dkim=pass (2048-bit key) header.d=sipsolutions.net header.i=@sipsolutions.net header.b=ZMh+bOFq; arc=none smtp.client-ip=168.119.38.16 Authentication-Results: smtp.subspace.kernel.org; dmarc=permerror header.from=sipsolutions.net Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=sipsolutions.net Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=sipsolutions.net header.i=@sipsolutions.net header.b="ZMh+bOFq" DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=sipsolutions.net; s=mail; h=Content-Transfer-Encoding:MIME-Version: References:In-Reply-To:Message-ID:Date:Subject:Cc:To:From:Content-Type:Sender :Reply-To:Content-ID:Content-Description:Resent-Date:Resent-From:Resent-To: Resent-Cc:Resent-Message-ID; bh=A8GueV0q9qDzWaxCC5p2cjn5HRkzxejPacA4CAQ/AA4=; t=1789985748; x=1791195348; b=ZMh+bOFqqO8vUUCfg3Sl4Y9NMqNiEfIVG2AqqoarBII710u DVkATePu+HaKMraDGXlyBqwcv86PkIrKfpdA8WPu16BSnXXBbErKx/1r89abA/cYDVJSEijrvYBfp 8B4fN/PYEThWckrdbXRocz3YeNJjkU99lAALEaldprUzsLtazcuyUPoHFiDwMVED+AS3L9vg4vaIa JWFC+5kl8NOGmfP4Wu/A1kjoxm+b8yRINIJE0jnfao4TMWxTIeV1ZtNufFEq67k2mh7LF5DKIJ+pD lnGEnNR7SNMug9Tz7O6IX/pn9y8EiHDIKl4wf7UljaHtL9FQktQGbmrGyz3OkmKg==; Received: by sipsolutions.net with esmtpsa (TLS1.3:ECDHE_X25519__ECDSA_SECP256R1_SHA256__AES_256_GCM:256) (Exim 4.98.2) (envelope-from ) id 1x8b3m-00000005jvH-2m4F; Mon, 21 Sep 2026 12:15:42 +0200 From: Johannes Berg To: linux-trace-kernel@vger.kernel.org, linux-kernel@vger.kernel.org, netdev@vger.kernel.org Cc: Steven Rostedt , Masami Hiramatsu , Mathieu Desnoyers , Johannes Berg Subject: [PATCH v6 1/4] tracing: add __print_sym() to replace __print_symbolic() Date: Mon, 21 Sep 2026 12:06:32 +0200 Message-ID: <20260921101538.97286-7-johannes@sipsolutions.net> X-Mailer: git-send-email 2.55.0 In-Reply-To: <20260921101538.97286-6-johannes@sipsolutions.net> References: <20260921101538.97286-6-johannes@sipsolutions.net> Precedence: bulk X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit From: Johannes Berg The way __print_symbolic() works is limited and inefficient in multiple ways: - you can only use it with a static list of symbols, but e.g. the SKB dropreasons are now a dynamic list - it builds the list in memory _three_ times, so it takes a lot of memory: - The print_fmt contains the list (since it's passed to the macro there). This actually contains the names _twice_, which is fixed up at runtime. - TRACE_DEFINE_ENUM() puts a 24-byte struct trace_eval_map for every entry, plus the string pointed to by it, which cannot be deduplicated with the strings in the print_fmt - The in-kernel symbolic printing creates yet another list of struct trace_print_flags for trace_print_symbols_seq() - it also requires runtime fixup during init, which is a lot of string parsing due to the print_fmt fixup Introduce __print_sym() to - over time - replace the old one. We can easily extend this also to __print_flags later, but I cared only about the SKB dropreasons for now, which has only __print_symbolic(). This new __print_sym() requires only a single list of items, created by TRACE_DEFINE_SYM_LIST(), or can even use another already existing list by using TRACE_DEFINE_SYM_FNS() with lookup and show methods. Then, instead of doing an init-time fixup, just do this at the time when userspace reads the print_fmt. This way, dynamically updated lists are possible. For userspace, nothing actually changes, because the print_fmt is shown exactly the same way the old __print_symbolic() was. This adds about 4k .text in my test builds, but that'll be more than paid for by the actual conversions. Signed-off-by: Johannes Berg --- v6: don't place data needed at runtime in .init sections, fixing the crash Steven reported with v1: https://lore.kernel.org/r/20240819182340.3bd23d67@gandalf.local.home/ (yes, it's been over 2 years ... my bad) --- include/asm-generic/vmlinux.lds.h | 3 + include/linux/module.h | 2 + include/linux/trace_events.h | 7 ++ include/linux/tracepoint.h | 20 +++++ include/trace/stages/init.h | 55 ++++++++++++ include/trace/stages/stage2_data_offsets.h | 6 ++ include/trace/stages/stage3_trace_output.h | 9 ++ include/trace/stages/stage7_class_define.h | 3 + kernel/module/main.c | 3 + kernel/trace/trace_events.c | 99 +++++++++++++++++++++- kernel/trace/trace_output.c | 45 ++++++++++ 11 files changed, 250 insertions(+), 2 deletions(-) diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h index b2988aa12f66..e48327fafdfe 100644 --- a/include/asm-generic/vmlinux.lds.h +++ b/include/asm-generic/vmlinux.lds.h @@ -276,8 +276,10 @@ . = ALIGN(8); \ BOUNDED_SECTION(_ftrace_events) \ BOUNDED_SECTION_BY(_ftrace_eval_map, _ftrace_eval_maps) +#define FTRACE_SYM_DEFS() BOUNDED_SECTION(_ftrace_sym_defs) #else #define FTRACE_EVENTS() +#define FTRACE_SYM_DEFS() #endif #ifdef CONFIG_TRACING @@ -391,6 +393,7 @@ TRACE_PRINTKS() \ BPF_RAW_TP() \ TRACEPOINT_STR() \ + FTRACE_SYM_DEFS() \ KUNIT_TABLE() /* diff --git a/include/linux/module.h b/include/linux/module.h index 96cc98568eea..a02833ac421b 100644 --- a/include/linux/module.h +++ b/include/linux/module.h @@ -516,6 +516,8 @@ struct module { unsigned int num_trace_events; struct trace_eval_map **trace_evals; unsigned int num_trace_evals; + struct trace_sym_def **trace_sym_defs; + unsigned int num_trace_sym_defs; #endif #ifdef CONFIG_DYNAMIC_FTRACE unsigned int num_ftrace_callsites; diff --git a/include/linux/trace_events.h b/include/linux/trace_events.h index 5cbd09c8be8d..80c34ef89feb 100644 --- a/include/linux/trace_events.h +++ b/include/linux/trace_events.h @@ -29,6 +29,13 @@ const char *trace_print_symbols_seq(struct trace_seq *p, unsigned long val, const struct trace_print_flags *symbol_array, size_t symbol_array_size); +const char *trace_print_sym_seq(struct trace_seq *p, unsigned long long val, + const char *(*lookup)(unsigned long long val)); +const char *trace_sym_lookup(const struct trace_sym_entry *list, + size_t len, unsigned long long value); +void trace_sym_show(struct seq_file *m, + const struct trace_sym_entry *list, size_t len); + #if BITS_PER_LONG == 32 const char *trace_print_flags_seq_u64(struct trace_seq *p, const char *delim, unsigned long long flags, diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h index e0d838c9ce93..05c45a72756c 100644 --- a/include/linux/tracepoint.h +++ b/include/linux/tracepoint.h @@ -32,6 +32,24 @@ struct trace_eval_map { unsigned long eval_value; }; +struct trace_sym_def { + const char *system; + const char *symbol_id; + /* may return NULL, called under rcu_read_lock() */ + const char * (*lookup)(unsigned long long); + /* + * Must print the list: ', { val, "name"}, ...' + * with no trailing comma, but with the leading ', ' + * to simplify things: + */ + void (*show)(struct seq_file *); +}; + +struct trace_sym_entry { + unsigned long long value; + const char *name; +}; + #define TRACEPOINT_DEFAULT_PRIO 10 extern int @@ -163,6 +181,8 @@ extern void syscall_unregfunc(void); #define TRACE_DEFINE_ENUM(x) #define TRACE_DEFINE_SIZEOF(x) +#define TRACE_DEFINE_SYM_FNS(...) +#define TRACE_DEFINE_SYM_LIST(...) #ifdef CONFIG_HAVE_ARCH_PREL32_RELOCATIONS static inline struct tracepoint *tracepoint_ptr_deref(tracepoint_ptr_t *p) diff --git a/include/trace/stages/init.h b/include/trace/stages/init.h index 000bcfc8dd2e..6285a03a52b2 100644 --- a/include/trace/stages/init.h +++ b/include/trace/stages/init.h @@ -23,6 +23,61 @@ TRACE_MAKE_SYSTEM_STR(); __section("_ftrace_eval_map") \ *TRACE_SYSTEM##_##a = &__##TRACE_SYSTEM##_##a +/* + * Define a symbol for __print_sym by giving lookup and + * show functions. See &struct trace_sym_def. + */ +#undef TRACE_DEFINE_SYM_FNS +#define TRACE_DEFINE_SYM_FNS(_symbol_id, _lookup, _show) \ + _TRACE_DEFINE_SYM_FNS(TRACE_SYSTEM, _symbol_id, _lookup, _show) +#define _TRACE_DEFINE_SYM_FNS(_system, _symbol_id, _lookup, _show) \ + __TRACE_DEFINE_SYM_FNS(_system, _symbol_id, _lookup, _show) +#define __TRACE_DEFINE_SYM_FNS(_system, _symbol_id, _lookup, _show) \ + ___TRACE_DEFINE_SYM_FNS(_system ## _ ## _symbol_id, _symbol_id, \ + _lookup, _show) +#define ___TRACE_DEFINE_SYM_FNS(_name, _symbol_id, _lookup, _show) \ + static struct trace_sym_def \ + __trace_sym_def_ ## _name = { \ + .system = TRACE_SYSTEM_STRING, \ + .symbol_id = #_symbol_id, \ + .lookup = _lookup, \ + .show = _show, \ + }; \ + static struct trace_sym_def __used \ + __section("_ftrace_sym_defs") \ + *__trace_sym_def_p_ ## _name = &__trace_sym_def_ ## _name + +/* + * Define a symbol for __print_sym by giving lookup and + * show functions. See &struct trace_sym_def. + */ +#undef TRACE_DEFINE_SYM_LIST +#define TRACE_DEFINE_SYM_LIST(_symbol_id, ...) \ + _TRACE_DEFINE_SYM_LIST(TRACE_SYSTEM, _symbol_id, __VA_ARGS__) +#define _TRACE_DEFINE_SYM_LIST(_system, _symbol_id, ...) \ + __TRACE_DEFINE_SYM_LIST(_system, _symbol_id, __VA_ARGS__) +#define __TRACE_DEFINE_SYM_LIST(_system, _symbol_id, ...) \ + ___TRACE_DEFINE_SYM_LIST(_system ## _ ## _symbol_id, _symbol_id,\ + __VA_ARGS__) +#define ___TRACE_DEFINE_SYM_LIST(_name, _symbol_id, ...) \ + static struct trace_sym_entry \ + __trace_sym_list_ ## _name[] = { __VA_ARGS__ }; \ + static const char * \ + __trace_sym_lookup_ ## _name(unsigned long long value) \ + { \ + return trace_sym_lookup(__trace_sym_list_ ## _name, \ + ARRAY_SIZE(__trace_sym_list_ ## _name), value); \ + } \ + static void \ + __trace_sym_show_ ## _name(struct seq_file *m) \ + { \ + trace_sym_show(m, __trace_sym_list_ ## _name, \ + ARRAY_SIZE(__trace_sym_list_ ## _name)); \ + } \ + ___TRACE_DEFINE_SYM_FNS(_name, _symbol_id, \ + __trace_sym_lookup_ ## _name, \ + __trace_sym_show_ ## _name) + #undef TRACE_DEFINE_SIZEOF #define TRACE_DEFINE_SIZEOF(a) \ static struct trace_eval_map __used __initdata \ diff --git a/include/trace/stages/stage2_data_offsets.h b/include/trace/stages/stage2_data_offsets.h index 8b0cff06d346..5afd9de7deb3 100644 --- a/include/trace/stages/stage2_data_offsets.h +++ b/include/trace/stages/stage2_data_offsets.h @@ -5,6 +5,12 @@ #undef TRACE_DEFINE_ENUM #define TRACE_DEFINE_ENUM(a) +#undef TRACE_DEFINE_SYM_FNS +#define TRACE_DEFINE_SYM_FNS(_symbol_id, _lookup, _show) + +#undef TRACE_DEFINE_SYM_LIST +#define TRACE_DEFINE_SYM_LIST(_symbol_id, ...) + #undef TRACE_DEFINE_SIZEOF #define TRACE_DEFINE_SIZEOF(a) diff --git a/include/trace/stages/stage3_trace_output.h b/include/trace/stages/stage3_trace_output.h index 181b81335781..09304259a8c1 100644 --- a/include/trace/stages/stage3_trace_output.h +++ b/include/trace/stages/stage3_trace_output.h @@ -79,6 +79,15 @@ trace_print_symbols_seq(p, value, symbols, ARRAY_SIZE(symbols)); \ }) +#undef __print_sym +#define __print_sym(value, symbol_id) \ + ___print_sym(TRACE_SYSTEM, value, symbol_id) +#define ___print_sym(sys, value, symbol_id) \ + ____print_sym(sys, value, symbol_id) +#define ____print_sym(sys, value, symbol_id) \ + trace_print_sym_seq(p, value, \ + __trace_sym_def_p_ ## sys ## _ ## symbol_id->lookup) + #undef __print_flags_u64 #undef __print_symbolic_u64 #if BITS_PER_LONG == 32 diff --git a/include/trace/stages/stage7_class_define.h b/include/trace/stages/stage7_class_define.h index 47008897a795..936b9fa08cea 100644 --- a/include/trace/stages/stage7_class_define.h +++ b/include/trace/stages/stage7_class_define.h @@ -45,6 +45,9 @@ #define __event_in_softirq() (REC->common_flags & 0x10) #define __event_in_irq() (REC->common_flags & 0x18) +#undef __print_sym +#define __print_sym(value, symbol_id) __print_sym(value:symbol_id) + /* * The below is not executed in the kernel. It is only what is * displayed in the print format for userspace to parse. diff --git a/kernel/module/main.c b/kernel/module/main.c index d0e1e0bd2ad0..8ab386a37376 100644 --- a/kernel/module/main.c +++ b/kernel/module/main.c @@ -2735,6 +2735,9 @@ static int find_module_sections(struct module *mod, struct load_info *info) mod->trace_evals = section_objs(info, "_ftrace_eval_map", sizeof(*mod->trace_evals), &mod->num_trace_evals); + mod->trace_sym_defs = section_objs(info, "_ftrace_sym_defs", + sizeof(*mod->trace_sym_defs), + &mod->num_trace_sym_defs); #endif #ifdef CONFIG_TRACING mod->trace_bprintk_fmt_start = section_objs(info, "__trace_printk_fmt", diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c index 1d39eaf6a0f7..abf38e3095c9 100644 --- a/kernel/trace/trace_events.c +++ b/kernel/trace/trace_events.c @@ -2109,6 +2109,102 @@ static void *f_next(struct seq_file *m, void *v, loff_t *pos) return node; } +extern struct trace_sym_def *__start_ftrace_sym_defs[]; +extern struct trace_sym_def *__stop_ftrace_sym_defs[]; + +/* note: @name is not NUL-terminated */ +static void show_sym_list(struct seq_file *m, struct trace_event_call *call, + const char *name, unsigned int name_len) +{ + struct trace_sym_def **sym_defs; + unsigned int n_sym_defs, i; + + if ((call->flags & TRACE_EVENT_FL_DYNAMIC) || !call->module) { + sym_defs = __start_ftrace_sym_defs; + n_sym_defs = __stop_ftrace_sym_defs - __start_ftrace_sym_defs; + } else { +#ifdef CONFIG_MODULES + struct module *mod = call->module; + + sym_defs = mod->trace_sym_defs; + n_sym_defs = mod->num_trace_sym_defs; +#else + return; +#endif /* CONFIG_MODULES */ + } + + for (i = 0; i < n_sym_defs; i++) { + unsigned int sym_len; + + if (!sym_defs[i]) + continue; + if (sym_defs[i]->system != call->class->system) + continue; + sym_len = strlen(sym_defs[i]->symbol_id); + if (name_len != sym_len) + continue; + if (strncmp(sym_defs[i]->symbol_id, name, sym_len)) + continue; + if (sym_defs[i]->show) + sym_defs[i]->show(m); + break; + } +} + +static void show_print_fmt(struct seq_file *m, struct trace_event_call *call) +{ + char *ptr = call->print_fmt; + bool in_print_sym = false; + int quote = 0; + + seq_puts(m, "\nprint fmt: "); + while (*ptr) { + if (*ptr == '\\') { + seq_putc(m, *ptr); + ptr++; + /* paranoid */ + if (!*ptr) + break; + goto next; + } + if (*ptr == '"') { + quote ^= 1; + goto next; + } + if (quote) + goto next; + + if (in_print_sym && *ptr != ':') + goto next; + + if (in_print_sym && *ptr == ':') { + const char *name; + + ptr++; + name = ptr; + /* skip the name */ + while (*ptr && *ptr != ')') + ptr++; + /* and show the actual list inline now */ + show_sym_list(m, call, name, ptr - name); + in_print_sym = false; + continue; + } + + if (strncmp(ptr, "__print_sym(", 12) == 0) { + ptr += 12; + seq_puts(m, "__print_symbolic("); + in_print_sym = true; + continue; + } +next: + seq_putc(m, *ptr); + ptr++; + } + + seq_putc(m, '\n'); +} + static int f_show(struct seq_file *m, void *v) { struct trace_event_file *file = event_file_data(m->private); @@ -2128,8 +2224,7 @@ static int f_show(struct seq_file *m, void *v) return 0; case FORMAT_PRINTFMT: - seq_printf(m, "\nprint fmt: %s\n", - call->print_fmt); + show_print_fmt(m, call); return 0; } diff --git a/kernel/trace/trace_output.c b/kernel/trace/trace_output.c index a5ad76175d10..0850b3ad5707 100644 --- a/kernel/trace/trace_output.c +++ b/kernel/trace/trace_output.c @@ -131,6 +131,51 @@ trace_print_symbols_seq(struct trace_seq *p, unsigned long val, } EXPORT_SYMBOL(trace_print_symbols_seq); +const char *trace_sym_lookup(const struct trace_sym_entry *list, + size_t len, unsigned long long value) +{ + size_t i; + + for (i = 0; i < len; i++) { + if (list[i].value == value) + return list[i].name; + } + return NULL; +} +EXPORT_SYMBOL(trace_sym_lookup); + +void trace_sym_show(struct seq_file *m, + const struct trace_sym_entry *list, size_t len) +{ + size_t i; + + for (i = 0; i < len; i++) + seq_printf(m, ", { %lld, \"%s\" }", + list[i].value, list[i].name); +} +EXPORT_SYMBOL(trace_sym_show); + +const char * +trace_print_sym_seq(struct trace_seq *p, unsigned long long val, + const char *(*lookup)(unsigned long long val)) +{ + const char *ret = trace_seq_buffer_ptr(p); + const char *name; + + rcu_read_lock(); + name = lookup(val); + if (name) + trace_seq_puts(p, name); + else + trace_seq_printf(p, "0x%llx", val); + rcu_read_unlock(); + + trace_seq_putc(p, 0); + + return ret; +} +EXPORT_SYMBOL(trace_print_sym_seq); + #if BITS_PER_LONG == 32 const char * trace_print_flags_seq_u64(struct trace_seq *p, const char *delim, -- 2.55.0