[PATCHv2,bpf-next,1/2] bpf: Fail uprobe multi link with negative offset

Commit Message

Jiri Olsa Dec. 17, 2023, 9:55 p.m. UTC

Currently the __uprobe_register will return 0 (success) when called with
negative offset. The reason is that the call to register_for_each_vma and
then build_map_info won't return error for negative offset. They just won't
do anything - no matching vma is found so there's no registered breakpoint
for the uprobe.

I don't think we can change the behaviour of __uprobe_register and fail
for negative uprobe offset, because apps might depend on that already.

But I think we can still make the change and check for it on bpf multi
link syscall level.

Also moving the __get_user call and check for the offsets to the top of
loop, to fail early without extra __get_user calls for ref_ctr_offset
and cookie arrays.

Signed-off-by: Jiri Olsa <[e-mail]>
---
 kernel/trace/bpf_trace.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

Comments

Song Liu Dec. 18, 2023, 4:12 p.m. UTC | #1

On Sun, Dec 17, 2023 at 1:55 PM Jiri Olsa <[e-mail]> wrote:
>
> Currently the __uprobe_register will return 0 (success) when called with
> negative offset. The reason is that the call to register_for_each_vma and
> then build_map_info won't return error for negative offset. They just won't
> do anything - no matching vma is found so there's no registered breakpoint
> for the uprobe.
>
> I don't think we can change the behaviour of __uprobe_register and fail
> for negative uprobe offset, because apps might depend on that already.
>
> But I think we can still make the change and check for it on bpf multi
> link syscall level.
>
> Also moving the __get_user call and check for the offsets to the top of
> loop, to fail early without extra __get_user calls for ref_ctr_offset
> and cookie arrays.
>
> Signed-off-by: Jiri Olsa <[e-mail]>

Acked-by: Song Liu <[e-mail]>

Andrii Nakryiko Dec. 18, 2023, 5:56 p.m. UTC | #2

On Sun, Dec 17, 2023 at 1:55 PM Jiri Olsa <[e-mail]> wrote:
>
> Currently the __uprobe_register will return 0 (success) when called with
> negative offset. The reason is that the call to register_for_each_vma and
> then build_map_info won't return error for negative offset. They just won't
> do anything - no matching vma is found so there's no registered breakpoint
> for the uprobe.
>
> I don't think we can change the behaviour of __uprobe_register and fail
> for negative uprobe offset, because apps might depend on that already.
>
> But I think we can still make the change and check for it on bpf multi
> link syscall level.
>
> Also moving the __get_user call and check for the offsets to the top of
> loop, to fail early without extra __get_user calls for ref_ctr_offset
> and cookie arrays.
>
> Signed-off-by: Jiri Olsa <[e-mail]>
> ---
>  kernel/trace/bpf_trace.c | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
> index 97c0c49c40a0..492d60e9c480 100644
> --- a/kernel/trace/bpf_trace.c
> +++ b/kernel/trace/bpf_trace.c
> @@ -3391,15 +3391,19 @@ int bpf_uprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *pr
>                 goto error_free;
>
>         for (i = 0; i < cnt; i++) {
> -               if (ucookies && __get_user(uprobes[i].cookie, ucookies + i)) {
> +               if (__get_user(uprobes[i].offset, uoffsets + i)) {
>                         err = -EFAULT;
>                         goto error_free;
>                 }
> +               if (uprobes[i].offset < 0) {
> +                       err = -EINVAL;
> +                       goto error_free;
> +               }

I applied this because it does fix the problem, but the whole
reshuffle of offsets in front of cookies is pointless, because of the
common for() loop. You are saving one or two __get_user() calls before
failing.

If we really want to do validation first, reading offsets should be in
its own for loop, then uref_ctr_offsets in its own, and then cookies
in its own loop as well. That way we read and validate the entire
array before reading another array. Please consider a follow up, if
you think it's important enough.


>                 if (uref_ctr_offsets && __get_user(uprobes[i].ref_ctr_offset, uref_ctr_offsets + i)) {
>                         err = -EFAULT;
>                         goto error_free;
>                 }
> -               if (__get_user(uprobes[i].offset, uoffsets + i)) {
> +               if (ucookies && __get_user(uprobes[i].cookie, ucookies + i)) {
>                         err = -EFAULT;
>                         goto error_free;
>                 }
> --
> 2.43.0
>

Jiri Olsa Dec. 19, 2023, 8:33 a.m. UTC | #3

On Mon, Dec 18, 2023 at 09:56:38AM -0800, Andrii Nakryiko wrote:
> On Sun, Dec 17, 2023 at 1:55 PM Jiri Olsa <[e-mail]> wrote:
> >
> > Currently the __uprobe_register will return 0 (success) when called with
> > negative offset. The reason is that the call to register_for_each_vma and
> > then build_map_info won't return error for negative offset. They just won't
> > do anything - no matching vma is found so there's no registered breakpoint
> > for the uprobe.
> >
> > I don't think we can change the behaviour of __uprobe_register and fail
> > for negative uprobe offset, because apps might depend on that already.
> >
> > But I think we can still make the change and check for it on bpf multi
> > link syscall level.
> >
> > Also moving the __get_user call and check for the offsets to the top of
> > loop, to fail early without extra __get_user calls for ref_ctr_offset
> > and cookie arrays.
> >
> > Signed-off-by: Jiri Olsa <[e-mail]>
> > ---
> >  kernel/trace/bpf_trace.c | 8 ++++++--
> >  1 file changed, 6 insertions(+), 2 deletions(-)
> >
> > diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
> > index 97c0c49c40a0..492d60e9c480 100644
> > --- a/kernel/trace/bpf_trace.c
> > +++ b/kernel/trace/bpf_trace.c
> > @@ -3391,15 +3391,19 @@ int bpf_uprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *pr
> >                 goto error_free;
> >
> >         for (i = 0; i < cnt; i++) {
> > -               if (ucookies && __get_user(uprobes[i].cookie, ucookies + i)) {
> > +               if (__get_user(uprobes[i].offset, uoffsets + i)) {
> >                         err = -EFAULT;
> >                         goto error_free;
> >                 }
> > +               if (uprobes[i].offset < 0) {
> > +                       err = -EINVAL;
> > +                       goto error_free;
> > +               }
> 
> I applied this because it does fix the problem, but the whole
> reshuffle of offsets in front of cookies is pointless, because of the
> common for() loop. You are saving one or two __get_user() calls before
> failing.
> 
> If we really want to do validation first, reading offsets should be in
> its own for loop, then uref_ctr_offsets in its own, and then cookies
> in its own loop as well. That way we read and validate the entire
> array before reading another array. Please consider a follow up, if
> you think it's important enough.

ok, thanks

jirka

> 
> 
> >                 if (uref_ctr_offsets && __get_user(uprobes[i].ref_ctr_offset, uref_ctr_offsets + i)) {
> >                         err = -EFAULT;
> >                         goto error_free;
> >                 }
> > -               if (__get_user(uprobes[i].offset, uoffsets + i)) {
> > +               if (ucookies && __get_user(uprobes[i].cookie, ucookies + i)) {
> >                         err = -EFAULT;
> >                         goto error_free;
> >                 }
> > --
> > 2.43.0
> >

Patch

diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index 97c0c49c40a0..492d60e9c480 100644
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
@@ -3391,15 +3391,19 @@  int bpf_uprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *pr
 		goto error_free;
 
 	for (i = 0; i < cnt; i++) {
-		if (ucookies && __get_user(uprobes[i].cookie, ucookies + i)) {
+		if (__get_user(uprobes[i].offset, uoffsets + i)) {
 			err = -EFAULT;
 			goto error_free;
 		}
+		if (uprobes[i].offset < 0) {
+			err = -EINVAL;
+			goto error_free;
+		}
 		if (uref_ctr_offsets && __get_user(uprobes[i].ref_ctr_offset, uref_ctr_offsets + i)) {
 			err = -EFAULT;
 			goto error_free;
 		}
-		if (__get_user(uprobes[i].offset, uoffsets + i)) {
+		if (ucookies && __get_user(uprobes[i].cookie, ucookies + i)) {
 			err = -EFAULT;
 			goto error_free;
 		}