Re: [PATCH 5/6] bpf: Allow to resolve bpf trampoline and dispatcher in unwind

[Jiri Olsa <jolsa@redhat.com>]

On Tue, Jan 21, 2020 at 12:55:10AM +0100, Daniel Borkmann wrote:
> On 1/18/20 2:49 PM, Jiri Olsa wrote:
> > When unwinding the stack we need to identify each address
> > to successfully continue. Adding latch tree to keep trampolines
> > for quick lookup during the unwind.
> > 
> > The patch uses first 48 bytes for latch tree node, leaving 4048
> > bytes from the rest of the page for trampoline or dispatcher
> > generated code.
> > 
> > It's still enough not to affect trampoline and dispatcher progs
> > maximum counts.
> > 
> > Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> > ---
> >   include/linux/bpf.h     | 12 ++++++-
> >   kernel/bpf/core.c       |  2 ++
> >   kernel/bpf/dispatcher.c |  4 +--
> >   kernel/bpf/trampoline.c | 76 +++++++++++++++++++++++++++++++++++++----
> >   4 files changed, 84 insertions(+), 10 deletions(-)
> > 
> > diff --git a/include/linux/bpf.h b/include/linux/bpf.h
> > index 8e3b8f4ad183..41eb0cf663e8 100644
> > --- a/include/linux/bpf.h
> > +++ b/include/linux/bpf.h
> > @@ -519,7 +519,6 @@ struct bpf_trampoline *bpf_trampoline_lookup(u64 key);
> >   int bpf_trampoline_link_prog(struct bpf_prog *prog);
> >   int bpf_trampoline_unlink_prog(struct bpf_prog *prog);
> >   void bpf_trampoline_put(struct bpf_trampoline *tr);
> > -void *bpf_jit_alloc_exec_page(void);
> >   #define BPF_DISPATCHER_INIT(name) {			\
> >   	.mutex = __MUTEX_INITIALIZER(name.mutex),	\
> >   	.func = &name##func,				\
> > @@ -551,6 +550,13 @@ void *bpf_jit_alloc_exec_page(void);
> >   #define BPF_DISPATCHER_PTR(name) (&name)
> >   void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from,
> >   				struct bpf_prog *to);
> > +struct bpf_image {
> > +	struct latch_tree_node tnode;
> > +	unsigned char data[];
> > +};
> > +#define BPF_IMAGE_SIZE (PAGE_SIZE - sizeof(struct bpf_image))
> > +bool is_bpf_image(void *addr);
> > +void *bpf_image_alloc(void);
> >   #else
> >   static inline struct bpf_trampoline *bpf_trampoline_lookup(u64 key)
> >   {
> > @@ -572,6 +578,10 @@ static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {}
> >   static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d,
> >   					      struct bpf_prog *from,
> >   					      struct bpf_prog *to) {}
> > +static inline bool is_bpf_image(void *addr)
> > +{
> > +	return false;
> > +}
> >   #endif
> >   struct bpf_func_info_aux {
> > diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
> > index 29d47aae0dd1..b3299dc9adda 100644
> > --- a/kernel/bpf/core.c
> > +++ b/kernel/bpf/core.c
> > @@ -704,6 +704,8 @@ bool is_bpf_text_address(unsigned long addr)
> >   	rcu_read_lock();
> >   	ret = bpf_prog_kallsyms_find(addr) != NULL;
> > +	if (!ret)
> > +		ret = is_bpf_image((void *) addr);
> >   	rcu_read_unlock();
> 
> Btw, shouldn't this be a separate entity entirely to avoid unnecessary inclusion
> in bpf_arch_text_poke() for the is_bpf_text_address() check there?

right, we dont want poking in trampolines/dispatchers.. I'll change that

> 
> Did you drop the bpf_{trampoline,dispatcher}_<...> entry addition in kallsyms?

working on that, will send it separately

jirka