On Wed, May 17, 2023 at 7:28 AM Masami Hiramatsu <mhiramat@xxxxxxxxxx> wrote:
>
> On Wed, 17 May 2023 11:45:07 +0800
> Ze Gao <zegao2021@xxxxxxxxx> wrote:
>
> > Current implementation calls kprobe related functions before doing
> > ftrace recursion check in fprobe_kprobe_handler, which opens door
> > to kernel crash due to stack recursion if preempt_count_{add, sub}
> > is traceable in kprobe_busy_{begin, end}.
> >
> > Things goes like this without this patch quoted from Steven:
> > "
> > fprobe_kprobe_handler() {
> > kprobe_busy_begin() {
> > preempt_disable() {
> > preempt_count_add() { <-- trace
> > fprobe_kprobe_handler() {
> > [ wash, rinse, repeat, CRASH!!! ]
> > "
> >
> > By refactoring the common part out of fprobe_kprobe_handler and
> > fprobe_handler and call ftrace recursion detection at the very beginning,
> > the whole fprobe_kprobe_handler is free from recursion.
> >
> > Signed-off-by: Ze Gao <zegao@xxxxxxxxxxx>
> > Acked-by: Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx>
> > Link: https://lore.kernel.org/linux-trace-kernel/20230516071830.8190-3-zegao@xxxxxxxxxxx
> > ---
> > kernel/trace/fprobe.c | 59 ++++++++++++++++++++++++++++++++-----------
> > 1 file changed, 44 insertions(+), 15 deletions(-)
> >
> > diff --git a/kernel/trace/fprobe.c b/kernel/trace/fprobe.c
> > index 9abb3905bc8e..097c740799ba 100644
> > --- a/kernel/trace/fprobe.c
> > +++ b/kernel/trace/fprobe.c
> > @@ -20,30 +20,22 @@ struct fprobe_rethook_node {
> > char data[];
> > };
> >
> > -static void fprobe_handler(unsigned long ip, unsigned long parent_ip,
> > - struct ftrace_ops *ops, struct ftrace_regs *fregs)
> > +static inline void __fprobe_handler(unsigned long ip, unsigned long
> > + parent_ip, struct ftrace_ops *ops, struct ftrace_regs *fregs)
>
> OK, I picked up this series to probes/fixes. Note that I fixed this line
> because the "unsigned long parent_ip" was split into 2 lines.
>
Hey Masami,
Regarding [0], I was bisecting BPF CI failures related to
multi-kprobes, and it turned out that [0] is the fix we need. It would
be great if you can make sure this fix gets into Linus' tree ASAP, so
that we can get it back into bpf/bpf-next trees and fix BPF selftests
for everyone (we mitigated this for BPF CI as a temporary workaround
for now). Thanks!
[0] https://lore.kernel.org/all/168100731160.79534.374827110083836722.stgit@devnote2/
> Thank you,
>
>
> > {
> > struct fprobe_rethook_node *fpr;
> > struct rethook_node *rh = NULL;
> > struct fprobe *fp;
> > void *entry_data = NULL;
> > - int bit, ret;
> > + int ret;
> >
> > fp = container_of(ops, struct fprobe, ops);
> > - if (fprobe_disabled(fp))
> > - return;
> > -
> > - bit = ftrace_test_recursion_trylock(ip, parent_ip);
> > - if (bit < 0) {
> > - fp->nmissed++;
> > - return;
> > - }
> >
> > if (fp->exit_handler) {
> > rh = rethook_try_get(fp->rethook);
> > if (!rh) {
> > fp->nmissed++;
> > - goto out;
> > + return;
> > }
> > fpr = container_of(rh, struct fprobe_rethook_node, node);
> > fpr->entry_ip = ip;
> > @@ -61,23 +53,60 @@ static void fprobe_handler(unsigned long ip, unsigned long parent_ip,
> > else
> > rethook_hook(rh, ftrace_get_regs(fregs), true);
> > }
> > -out:
> > +}
> > +
> > +static void fprobe_handler(unsigned long ip, unsigned long parent_ip,
> > + struct ftrace_ops *ops, struct ftrace_regs *fregs)
> > +{
> > + struct fprobe *fp;
> > + int bit;
> > +
> > + fp = container_of(ops, struct fprobe, ops);
> > + if (fprobe_disabled(fp))
> > + return;
> > +
> > + /* recursion detection has to go before any traceable function and
> > + * all functions before this point should be marked as notrace
> > + */
> > + bit = ftrace_test_recursion_trylock(ip, parent_ip);
> > + if (bit < 0) {
> > + fp->nmissed++;
> > + return;
> > + }
> > + __fprobe_handler(ip, parent_ip, ops, fregs);
> > ftrace_test_recursion_unlock(bit);
> > +
> > }
> > NOKPROBE_SYMBOL(fprobe_handler);
> >
> > static void fprobe_kprobe_handler(unsigned long ip, unsigned long parent_ip,
> > struct ftrace_ops *ops, struct ftrace_regs *fregs)
> > {
> > - struct fprobe *fp = container_of(ops, struct fprobe, ops);
> > + struct fprobe *fp;
> > + int bit;
> > +
> > + fp = container_of(ops, struct fprobe, ops);
> > + if (fprobe_disabled(fp))
> > + return;
> > +
> > + /* recursion detection has to go before any traceable function and
> > + * all functions called before this point should be marked as notrace
> > + */
> > + bit = ftrace_test_recursion_trylock(ip, parent_ip);
> > + if (bit < 0) {
> > + fp->nmissed++;
> > + return;
> > + }
> >
> > if (unlikely(kprobe_running())) {
> > fp->nmissed++;
> > return;
> > }
> > +
> > kprobe_busy_begin();
> > - fprobe_handler(ip, parent_ip, ops, fregs);
> > + __fprobe_handler(ip, parent_ip, ops, fregs);
> > kprobe_busy_end();
> > + ftrace_test_recursion_unlock(bit);
> > }
> >
> > static void fprobe_exit_handler(struct rethook_node *rh, void *data,
> > --
> > 2.40.1
> >
>
>
> --
> Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx>
>
