[PATCH] RISC-V: Don't check text_mutex during stop_machine

Palmer Dabbelt palmer at dabbelt.com
Thu May 6 08:10:41 BST 2021

From: Palmer Dabbelt <palmerdabbelt at google.com>

We're currently using stop_machine() to update ftrace, which means that
the thread that takes text_mutex during ftrace_prepare() may not be the
same as the thread that eventually patches the code.  This isn't
actually a race because the lock is still held (preventing any other
concurrent accesses) and there is only one thread running during
stop_machine(), but it does trigger a lockdep failure.

This patch just elides the lockdep check during stop_machine.

Fixes: c15ac4fd60d5 ("riscv/ftrace: Add dynamic function tracer support")
Suggested-by: Steven Rostedt <rostedt at goodmis.org>
Reported-by: Changbin Du <changbin.du at gmail.com>
Signed-off-by: Palmer Dabbelt <palmerdabbelt at google.com>
In theory we should be able to avoid using stop_machine() with some
clever code sequences, but that's too big of a change to be considered a
fix.  I also can't find the text I thought was in the ISA manual about
the allowed behaviors for concurrent modification of the instruction
stream, so I might have just mis-remembered that.
 arch/riscv/include/asm/ftrace.h |  4 ++++
 arch/riscv/kernel/ftrace.c      | 15 +++++++++++++++
 arch/riscv/kernel/patch.c       | 10 +++++++++-
 3 files changed, 28 insertions(+), 1 deletion(-)

diff --git a/arch/riscv/include/asm/ftrace.h b/arch/riscv/include/asm/ftrace.h
index 04dad3380041..ee8f07b4dbee 100644
--- a/arch/riscv/include/asm/ftrace.h
+++ b/arch/riscv/include/asm/ftrace.h
@@ -85,4 +85,8 @@ int ftrace_init_nop(struct module *mod, struct dyn_ftrace *rec);
+#ifndef __ASSEMBLY__
+extern int riscv_ftrace_in_stop_machine;
 #endif /* _ASM_RISCV_FTRACE_H */
diff --git a/arch/riscv/kernel/ftrace.c b/arch/riscv/kernel/ftrace.c
index 7f1e5203de88..da2405652f1d 100644
--- a/arch/riscv/kernel/ftrace.c
+++ b/arch/riscv/kernel/ftrace.c
@@ -11,6 +11,8 @@
 #include <asm/cacheflush.h>
 #include <asm/patch.h>
+int riscv_ftrace_in_stop_machine;
 int ftrace_arch_code_modify_prepare(void) __acquires(&text_mutex)
@@ -232,3 +234,16 @@ int ftrace_disable_ftrace_graph_caller(void)
+void arch_ftrace_update_code(int command)
+	/*
+	 * The code sequences we use for ftrace can't be patched while the
+	 * kernel is running, so we need to use stop_machine() to modify them
+	 * for now.  This doesn't play nice with text_mutex, we use this flag
+	 * to elide the check.
+	 */
+	riscv_ftrace_in_stop_machine = true;
+	ftrace_run_stop_machine(command);
+	riscv_ftrace_in_stop_machine = false;
diff --git a/arch/riscv/kernel/patch.c b/arch/riscv/kernel/patch.c
index 0b552873a577..7983dba477f0 100644
--- a/arch/riscv/kernel/patch.c
+++ b/arch/riscv/kernel/patch.c
@@ -11,6 +11,7 @@
 #include <asm/kprobes.h>
 #include <asm/cacheflush.h>
 #include <asm/fixmap.h>
+#include <asm/ftrace.h>
 #include <asm/patch.h>
 struct patch_insn {
@@ -59,8 +60,15 @@ static int patch_insn_write(void *addr, const void *insn, size_t len)
 	 * Before reaching here, it was expected to lock the text_mutex
 	 * already, so we don't need to give another lock here and could
 	 * ensure that it was safe between each cores.
+	 *
+	 * We're currently using stop_machine() for ftrace, and while that
+	 * ensures text_mutex is held before installing the mappings it does
+	 * not ensure text_mutex is held by the calling thread.  That's safe
+	 * but triggers a lockdep failure, so just elide it for that specific
+	 * case.
-	lockdep_assert_held(&text_mutex);
+	if (!riscv_ftrace_in_stop_machine)
+		lockdep_assert_held(&text_mutex);
 	if (across_pages)
 		patch_map(addr + len, FIX_TEXT_POKE1);

More information about the linux-riscv mailing list