]> git.ipfire.org Git - thirdparty/kernel/linux.git/commit
cfi: Use RCU while invoking __module_address().
authorSebastian Andrzej Siewior <bigeasy@linutronix.de>
Wed, 8 Jan 2025 09:04:50 +0000 (10:04 +0100)
committerPetr Pavlu <petr.pavlu@suse.com>
Mon, 10 Mar 2025 10:54:45 +0000 (11:54 +0100)
commite151955bacf881e1f918613558e04f2f976169e4
tree61ad84173aa5c23a24d676902ba19bb2281b82e7
parentccf74e79ea3537bd697ca587192aae424d160235
cfi: Use RCU while invoking __module_address().

__module_address() can be invoked within a RCU section, there is no
requirement to have preemption disabled.

The _notrace() variant was introduced in commit 14c4c8e41511a ("cfi: Use
rcu_read_{un}lock_sched_notrace"). The recursive case where
__cfi_slowpath_diag() could end up calling itself is no longer present,
as all that logic is gone since commit 89245600941e ("cfi: Switch to
-fsanitize=kcfi").
Sami Tolvanen said that KCFI checks don't perform function calls.

Elliot Berman verified it with
| modprobe -a dummy_stm stm_ftrace stm_p_basic
| mkdir -p /sys/kernel/config/stp-policy/dummy_stm.0.my-policy/default
| echo function > /sys/kernel/tracing/current_tracer
| echo 1 > /sys/kernel/tracing/tracing_on
| echo dummy_stm.0 > /sys/class/stm_source/ftrace/stm_source_link

Replace the rcu_read_lock_sched_notrace() section around
__module_address() with RCU.

Cc: Elliot Berman <quic_eberman@quicinc.com>
Cc: Kees Cook <kees@kernel.org>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: Sami Tolvanen <samitolvanen@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: llvm@lists.linux.dev
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Elliot Berman <elliot.berman@oss.qualcomm.com> # sm8650-qrd [1]
Link: https://lore.kernel.org/all/20241230185812429-0800.eberman@hu-eberman-lv.qualcomm.com
Link: https://lore.kernel.org/r/20250108090457.512198-22-bigeasy@linutronix.de
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
kernel/cfi.c