ftrace/x86: Anotate text_mutex split between ftrace_arch_code_modify_post_process...
authorJiri Kosina <jkosina@suse.cz>
Sat, 29 Jun 2019 21:22:33 +0000 (23:22 +0200)
committerSteven Rostedt (VMware) <rostedt@goodmis.org>
Tue, 2 Jul 2019 19:41:35 +0000 (15:41 -0400)
ftrace_arch_code_modify_prepare() is acquiring text_mutex, while the
corresponding release is happening in ftrace_arch_code_modify_post_process().

This has already been documented in the code, but let's also make the fact
that this is intentional clear to the semantic analysis tools such as sparse.

Link: http://lkml.kernel.org/r/nycvar.YFH.7.76.1906292321170.27227@cbobk.fhfr.pm
Fixes: 39611265edc1a ("ftrace/x86: Add a comment to why we take text_mutex in ftrace_arch_code_modify_prepare()")
Fixes: d5b844a2cf507 ("ftrace/x86: Remove possible deadlock between register_kprobe() and ftrace_run_update_code()")
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
arch/x86/kernel/ftrace.c

index d7e93b2..7622852 100644 (file)
@@ -35,6 +35,7 @@
 #ifdef CONFIG_DYNAMIC_FTRACE
 
 int ftrace_arch_code_modify_prepare(void)
+    __acquires(&text_mutex)
 {
        /*
         * Need to grab text_mutex to prevent a race from module loading
@@ -48,6 +49,7 @@ int ftrace_arch_code_modify_prepare(void)
 }
 
 int ftrace_arch_code_modify_post_process(void)
+    __releases(&text_mutex)
 {
        set_all_modules_text_ro();
        set_kernel_text_ro();