From: takahiro.akashi@linaro.org (AKASHI Takahiro)
To: linux-arm-kernel@lists.infradead.org
Subject: [PATCH v5 6/6] arm64: ftrace: add a test of function prologue analyzer
Date: Fri, 6 Nov 2015 15:44:45 +0900 [thread overview]
Message-ID: <1446792285-1154-7-git-send-email-takahiro.akashi@linaro.org> (raw)
In-Reply-To: <1446792285-1154-1-git-send-email-takahiro.akashi@linaro.org>
The patch ("arm64: ftrace: add arch-specific stack tracer") introduced
a function prologue analyzer.
Given that there is no fixed template for a function prologue, at least
on gcc for aarch64, a function prologue analyzer may be rather heuristic.
So this patch adds a kernel command line option,
function_prologue_analyzer_test, in order to run a basic test at startup
by executing an analyzer against all the *traceable* functions.
For function_prologue_analyzer_test=2, the output looks like:
po sp fp symbol
== == == ======
0: 0 0x040 0x000 gic_handle_irq+0x20/0xa4
1: 0 0x040 0x000 gic_handle_irq+0x34/0x114
2: 0 0x030 0x000 run_init_process+0x14/0x48
3: 0 0x020 0x000 try_to_run_init_process+0x14/0x58
4: 0 0x080 0x000 do_one_initcall+0x1c/0x194
...
22959: 0 0x020 0x000 tty_lock_slave+0x14/0x3c
22960: 0 0x020 0x000 tty_unlock_slave+0x14/0x3c
function prologue analyzer test: 0 errors
"po" indicates a position of callsite of mcount(), and should be zero
if an analyzer has parsed a function prologue successfully and reached
a location where fp is properly updated.
"sp" is a final offset to a parent's fp at the exit of function prologue.
"fp" is also an ofset against sp at the exit of function prologue.
So normally,
<new sp> = <old fp> + <"sp">
<new fp> = <new sp> - <"fp">
And the last line shows the number of possible errors in the result.
For function_prologue_analyzer_test=1, only the last line will be shown.
Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org>
---
arch/arm64/kernel/stacktrace.c | 52 ++++++++++++++++++++++++++++++++++++++++
1 file changed, 52 insertions(+)
diff --git a/arch/arm64/kernel/stacktrace.c b/arch/arm64/kernel/stacktrace.c
index 4ee052a..d1a9e5b 100644
--- a/arch/arm64/kernel/stacktrace.c
+++ b/arch/arm64/kernel/stacktrace.c
@@ -18,6 +18,7 @@
#include <linux/kernel.h>
#include <linux/export.h>
#include <linux/ftrace.h>
+#include <linux/kallsyms.h>
#include <linux/sched.h>
#include <linux/stacktrace.h>
@@ -322,5 +323,56 @@ void save_stack_trace_sp(struct stack_trace *trace,
{
__save_stack_trace_tsk(current, trace, stack_dump_sp);
}
+
+static int start_analyzer_test __initdata;
+
+static int __init enable_analyzer_test(char *str)
+{
+ get_option(&str, &start_analyzer_test);
+ return 0;
+}
+early_param("function_prologue_analyzer_test", enable_analyzer_test);
+
+static void __init do_test_function_prologue_analyzer(void)
+{
+ extern unsigned long __start_mcount_loc[];
+ extern unsigned long __stop_mcount_loc[];
+ unsigned long count, i, errors;
+ int print_once;
+
+ count = __stop_mcount_loc - __start_mcount_loc;
+ errors = print_once = 0;
+ for (i = 0; i < count; i++) {
+ unsigned long addr, sp_off, fp_off;
+ int pos;
+ bool check;
+ char buf[60];
+
+ addr = __start_mcount_loc[i];
+ pos = analyze_function_prologue(addr, &sp_off, &fp_off);
+ check = ((pos != 0) || !sp_off || (sp_off <= fp_off));
+ if (check)
+ errors++;
+ if (check || (start_analyzer_test > 1)) {
+ if (!print_once) {
+ pr_debug(" po sp fp symbol\n");
+ pr_debug(" == == == ======\n");
+ print_once++;
+ }
+ sprint_symbol(buf, addr);
+ pr_debug("%5ld: %d 0x%03lx 0x%03lx %s\n",
+ i, pos, sp_off, fp_off, buf);
+ }
+ }
+ pr_debug("function prologue analyzer test: %ld errors\n", errors);
+}
+
+static int __init test_function_prologue_analyzer(void)
+{
+ if (start_analyzer_test)
+ do_test_function_prologue_analyzer();
+ return 0;
+}
+late_initcall(test_function_prologue_analyzer);
#endif /* CONFIG_STACK_TRACER */
#endif
--
1.7.9.5
next prev parent reply other threads:[~2015-11-06 6:44 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-11-06 6:44 [PATCH v5 0/6] arm64: ftrace: fix incorrect output from stack tracer AKASHI Takahiro
2015-11-06 6:44 ` [PATCH v5 1/6] arm64: ftrace: modify a stack frame in a safe way AKASHI Takahiro
2015-11-06 6:44 ` [PATCH v5 2/6] arm64: ftrace: fix a stack tracer's output under function graph tracer AKASHI Takahiro
2015-11-09 14:04 ` Jungseok Lee
2015-11-10 2:42 ` AKASHI Takahiro
2015-11-13 15:01 ` Jungseok Lee
2015-11-16 9:23 ` AKASHI Takahiro
2015-11-06 6:44 ` [PATCH v5 3/6] ftrace: allow arch-specific stack tracer AKASHI Takahiro
2015-11-06 13:39 ` Steven Rostedt
2015-11-06 6:44 ` [PATCH v5 4/6] arm64: insn: add instruction decoders for ldp/stp and add/sub AKASHI Takahiro
2015-11-10 13:40 ` Jungseok Lee
2015-11-11 4:54 ` AKASHI Takahiro
2015-11-06 6:44 ` [PATCH v5 5/6] arm64: ftrace: add arch-specific stack tracer AKASHI Takahiro
2015-11-10 14:05 ` Jungseok Lee
2015-11-11 5:03 ` AKASHI Takahiro
2015-11-11 22:56 ` Jungseok Lee
2015-11-06 6:44 ` AKASHI Takahiro [this message]
2015-11-09 14:24 ` [PATCH v5 0/6] arm64: ftrace: fix incorrect output from " Jungseok Lee
2015-11-10 2:58 ` AKASHI Takahiro
2015-11-10 13:32 ` Jungseok Lee
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=1446792285-1154-7-git-send-email-takahiro.akashi@linaro.org \
--to=takahiro.akashi@linaro.org \
--cc=linux-arm-kernel@lists.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;
as well as URLs for NNTP newsgroup(s).