Re: [PATCH 3/4] fprobe: add recursion detection in fprobe_exit_handler
From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Date: 2023-05-16 04:25:53
Also in:
lkml
On Mon, 15 May 2023 11:26:40 +0800 Ze Gao [off-list ref] wrote:
fprobe_hander and fprobe_kprobe_handler has guarded ftrace recusion detection but fprobe_exit_handler has not, which possibly introduce recurive calls if the fprobe exit callback calls any traceable functions. Checking in fprobe_hander or fprobe_kprobe_handler is not enough and misses this case.
Good catch! Yes, this can fix such recursive call case because if
we put a fprobe to the exit of the "func()", recursive call happens
as below;
func() {
} => rethook
=> fprobe_exit_handler()
=> fp->exit_handler() {
func() {
} => rethook
=> fprobe_exit_handler()
=> fp->exit_handler() {
func() {
} => rethook ...
Note that this should not happen with fprobe-based events because
all the code (except for tests) under kernel/trace/ are marked
notrace automatically.
kretprobe avoids this by setting itself to current_kprobe, thus the
other kprobes recursively called from the rethook will be skipped.
So add recusion free guard the same way as fprobe_hander and also mark fprobe_exit_handler notrace. Since ftrace recursion check does not employ ips, so here use entry_ip and entry_parent_ip the same as fprobe_handler.
Looks good to me.
Fixes: 5b0ab78998e3 ("fprobe: Add exit_handler support")
Cc: stable@vger.kernel.org
Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
quoted hunk ↗ jump to hunk
Signed-off-by: Ze Gao <redacted> --- kernel/trace/fprobe.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-)diff --git a/kernel/trace/fprobe.c b/kernel/trace/fprobe.c index ad9a36c87ad9..cf982d4ab142 100644 --- a/kernel/trace/fprobe.c +++ b/kernel/trace/fprobe.c@@ -17,6 +17,7 @@ struct fprobe_rethook_node { struct rethook_node node; unsigned long entry_ip; + unsigned long entry_parent_ip; char data[]; };@@ -39,6 +40,7 @@ static inline notrace void __fprobe_handler(unsigned long ip, unsigned long } fpr = container_of(rh, struct fprobe_rethook_node, node); fpr->entry_ip = ip; + fpr->entry_parent_ip = parent_ip; if (fp->entry_data_size) entry_data = fpr->data; }@@ -109,19 +111,30 @@ static void notrace fprobe_kprobe_handler(unsigned long ip, unsigned long parent ftrace_test_recursion_unlock(bit); } -static void fprobe_exit_handler(struct rethook_node *rh, void *data, +static void notrace fprobe_exit_handler(struct rethook_node *rh, void *data, struct pt_regs *regs) { struct fprobe *fp = (struct fprobe *)data; struct fprobe_rethook_node *fpr; + int bit; if (!fp || fprobe_disabled(fp)) return; fpr = container_of(rh, struct fprobe_rethook_node, node); + /* we need to assure no calls to traceable functions in-between the + * end of fprobe_handler and the beginning of fprobe_exit_handler. + */ + bit = ftrace_test_recursion_trylock(fpr->entry_ip, fpr->entry_parent_ip); + if (bit < 0) { + fp->nmissed++; + return; + } + fp->exit_handler(fp, fpr->entry_ip, regs, fp->entry_data_size ? (void *)fpr->data : NULL); + ftrace_test_recursion_unlock(bit); } NOKPROBE_SYMBOL(fprobe_exit_handler);-- 2.40.1
-- Masami Hiramatsu (Google) [off-list ref]