diff mbox series

[v8,bpf-next,07/13] bpf: Add btf_struct_ids_match function

Message ID 20200722211223.1055107-8-jolsa@kernel.org
State Changes Requested
Delegated to: BPF Maintainers
Headers show
Series bpf: Add d_path helper | expand

Commit Message

Jiri Olsa July 22, 2020, 9:12 p.m. UTC
Adding btf_struct_ids_match function to check if given address provided
by BTF object + offset is also address of another nested BTF object.

This allows to pass an argument to helper, which is defined via parent
BTF object + offset, like for bpf_d_path (added in following changes):

  SEC("fentry/filp_close")
  int BPF_PROG(prog_close, struct file *file, void *id)
  {
    ...
    ret = bpf_d_path(&file->f_path, ...

The first bpf_d_path argument is hold by verifier as BTF file object
plus offset of f_path member.

The btf_struct_ids_match function will walk the struct file object and
check if there's nested struct path object on the given offset.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 include/linux/bpf.h   |  2 ++
 kernel/bpf/btf.c      | 29 +++++++++++++++++++++++++++++
 kernel/bpf/verifier.c | 18 ++++++++++++------
 3 files changed, 43 insertions(+), 6 deletions(-)

Comments

Andrii Nakryiko July 28, 2020, 11:35 p.m. UTC | #1
On Wed, Jul 22, 2020 at 2:13 PM Jiri Olsa <jolsa@kernel.org> wrote:
>
> Adding btf_struct_ids_match function to check if given address provided
> by BTF object + offset is also address of another nested BTF object.
>
> This allows to pass an argument to helper, which is defined via parent
> BTF object + offset, like for bpf_d_path (added in following changes):
>
>   SEC("fentry/filp_close")
>   int BPF_PROG(prog_close, struct file *file, void *id)
>   {
>     ...
>     ret = bpf_d_path(&file->f_path, ...
>
> The first bpf_d_path argument is hold by verifier as BTF file object
> plus offset of f_path member.
>
> The btf_struct_ids_match function will walk the struct file object and
> check if there's nested struct path object on the given offset.
>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  include/linux/bpf.h   |  2 ++
>  kernel/bpf/btf.c      | 29 +++++++++++++++++++++++++++++
>  kernel/bpf/verifier.c | 18 ++++++++++++------
>  3 files changed, 43 insertions(+), 6 deletions(-)
>
> diff --git a/include/linux/bpf.h b/include/linux/bpf.h
> index bae557ff2da8..c981e258fed3 100644
> --- a/include/linux/bpf.h
> +++ b/include/linux/bpf.h
> @@ -1306,6 +1306,8 @@ int btf_struct_access(struct bpf_verifier_log *log,
>                       const struct btf_type *t, int off, int size,
>                       enum bpf_access_type atype,
>                       u32 *next_btf_id);
> +bool btf_struct_ids_match(struct bpf_verifier_log *log,
> +                         int off, u32 id, u32 mid);
>  int btf_resolve_helper_id(struct bpf_verifier_log *log,
>                           const struct bpf_func_proto *fn, int);
>
> diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
> index 1ab5fd5bf992..562d4453fad3 100644
> --- a/kernel/bpf/btf.c
> +++ b/kernel/bpf/btf.c
> @@ -4140,6 +4140,35 @@ int btf_struct_access(struct bpf_verifier_log *log,
>         return -EINVAL;
>  }
>
> +bool btf_struct_ids_match(struct bpf_verifier_log *log,
> +                         int off, u32 id, u32 mid)
> +{
> +       const struct btf_type *type;
> +       u32 nid;
> +       int err;
> +

mid and nid are terrible names, especially as an input argument name.
mid == need_type_id? nid == cur_type_id or something along those
lines?

> +       do {
> +               type = btf_type_by_id(btf_vmlinux, id);
> +               if (!type)
> +                       return false;
> +               err = btf_struct_walk(log, type, off, 1, &nid);
> +               if (err < 0)
> +                       return false;
> +
> +               /* We found nested struct object. If it matches
> +                * the requested ID, we're done. Otherwise let's
> +                * continue the search with offset 0 in the new
> +                * type.
> +                */
> +               if (err == walk_struct && mid == nid)
> +                       return true;
> +               off = 0;
> +               id = nid;
> +       } while (err == walk_struct);

This seems like a slightly more obvious control flow:

again:

   ...

   if (err != walk_struct)
      return false;

   if (mid != nid) {
      off = 0;
      id = nid;
      goto again;
   }

   return true;

> +
> +       return false;
> +}
> +
>  int btf_resolve_helper_id(struct bpf_verifier_log *log,
>                           const struct bpf_func_proto *fn, int arg)
>  {

[...]
Jiri Olsa July 29, 2020, 4:04 p.m. UTC | #2
On Tue, Jul 28, 2020 at 04:35:16PM -0700, Andrii Nakryiko wrote:

SNIP

> > diff --git a/include/linux/bpf.h b/include/linux/bpf.h
> > index bae557ff2da8..c981e258fed3 100644
> > --- a/include/linux/bpf.h
> > +++ b/include/linux/bpf.h
> > @@ -1306,6 +1306,8 @@ int btf_struct_access(struct bpf_verifier_log *log,
> >                       const struct btf_type *t, int off, int size,
> >                       enum bpf_access_type atype,
> >                       u32 *next_btf_id);
> > +bool btf_struct_ids_match(struct bpf_verifier_log *log,
> > +                         int off, u32 id, u32 mid);
> >  int btf_resolve_helper_id(struct bpf_verifier_log *log,
> >                           const struct bpf_func_proto *fn, int);
> >
> > diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
> > index 1ab5fd5bf992..562d4453fad3 100644
> > --- a/kernel/bpf/btf.c
> > +++ b/kernel/bpf/btf.c
> > @@ -4140,6 +4140,35 @@ int btf_struct_access(struct bpf_verifier_log *log,
> >         return -EINVAL;
> >  }
> >
> > +bool btf_struct_ids_match(struct bpf_verifier_log *log,
> > +                         int off, u32 id, u32 mid)
> > +{
> > +       const struct btf_type *type;
> > +       u32 nid;
> > +       int err;
> > +
> 
> mid and nid are terrible names, especially as an input argument name.
> mid == need_type_id? nid == cur_type_id or something along those
> lines?

'mid' was for matching id, 'nid' for nested id ;-)
need_type_id/cur_type_id sound good

> 
> > +       do {
> > +               type = btf_type_by_id(btf_vmlinux, id);
> > +               if (!type)
> > +                       return false;
> > +               err = btf_struct_walk(log, type, off, 1, &nid);
> > +               if (err < 0)
> > +                       return false;
> > +
> > +               /* We found nested struct object. If it matches
> > +                * the requested ID, we're done. Otherwise let's
> > +                * continue the search with offset 0 in the new
> > +                * type.
> > +                */
> > +               if (err == walk_struct && mid == nid)
> > +                       return true;
> > +               off = 0;
> > +               id = nid;
> > +       } while (err == walk_struct);
> 
> This seems like a slightly more obvious control flow:
> 
> again:
> 
>    ...
> 
>    if (err != walk_struct)
>       return false;

ok, and perhaps use in here the switch(err) as in the previous patch?

thanks,
jirka

> 
>    if (mid != nid) {
>       off = 0;
>       id = nid;
>       goto again;
>    }
> 
>    return true;
> 
> > +
> > +       return false;
> > +}
> > +
> >  int btf_resolve_helper_id(struct bpf_verifier_log *log,
> >                           const struct bpf_func_proto *fn, int arg)
> >  {
> 
> [...]
>
Andrii Nakryiko July 29, 2020, 5:51 p.m. UTC | #3
On Wed, Jul 29, 2020 at 9:04 AM Jiri Olsa <jolsa@redhat.com> wrote:
>
> On Tue, Jul 28, 2020 at 04:35:16PM -0700, Andrii Nakryiko wrote:
>
> SNIP
>
> > > diff --git a/include/linux/bpf.h b/include/linux/bpf.h
> > > index bae557ff2da8..c981e258fed3 100644
> > > --- a/include/linux/bpf.h
> > > +++ b/include/linux/bpf.h
> > > @@ -1306,6 +1306,8 @@ int btf_struct_access(struct bpf_verifier_log *log,
> > >                       const struct btf_type *t, int off, int size,
> > >                       enum bpf_access_type atype,
> > >                       u32 *next_btf_id);
> > > +bool btf_struct_ids_match(struct bpf_verifier_log *log,
> > > +                         int off, u32 id, u32 mid);
> > >  int btf_resolve_helper_id(struct bpf_verifier_log *log,
> > >                           const struct bpf_func_proto *fn, int);
> > >
> > > diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
> > > index 1ab5fd5bf992..562d4453fad3 100644
> > > --- a/kernel/bpf/btf.c
> > > +++ b/kernel/bpf/btf.c
> > > @@ -4140,6 +4140,35 @@ int btf_struct_access(struct bpf_verifier_log *log,
> > >         return -EINVAL;
> > >  }
> > >
> > > +bool btf_struct_ids_match(struct bpf_verifier_log *log,
> > > +                         int off, u32 id, u32 mid)

just realized that if id == mid and off == 0, btf_struct_ids_match()
will return false. Right now verifier is careful to not call
btf_struct_ids_match in such case, but I wonder if it's better to make
that (common) case also work?

> > > +{
> > > +       const struct btf_type *type;
> > > +       u32 nid;
> > > +       int err;
> > > +
> >
> > mid and nid are terrible names, especially as an input argument name.
> > mid == need_type_id? nid == cur_type_id or something along those
> > lines?
>
> 'mid' was for matching id, 'nid' for nested id ;-)
> need_type_id/cur_type_id sound good

nested I guessed, mid was a mystery to me :))

>
> >
> > > +       do {
> > > +               type = btf_type_by_id(btf_vmlinux, id);
> > > +               if (!type)
> > > +                       return false;
> > > +               err = btf_struct_walk(log, type, off, 1, &nid);
> > > +               if (err < 0)
> > > +                       return false;
> > > +
> > > +               /* We found nested struct object. If it matches
> > > +                * the requested ID, we're done. Otherwise let's
> > > +                * continue the search with offset 0 in the new
> > > +                * type.
> > > +                */
> > > +               if (err == walk_struct && mid == nid)
> > > +                       return true;
> > > +               off = 0;
> > > +               id = nid;
> > > +       } while (err == walk_struct);
> >
> > This seems like a slightly more obvious control flow:
> >
> > again:
> >
> >    ...
> >
> >    if (err != walk_struct)
> >       return false;
>
> ok, and perhaps use in here the switch(err) as in the previous patch?

I think straightforward if is better than switch here, because
anything but walk_struct is not what we expect.

>
> thanks,
> jirka
>
> >
> >    if (mid != nid) {
> >       off = 0;
> >       id = nid;
> >       goto again;
> >    }
> >
> >    return true;
> >
> > > +
> > > +       return false;
> > > +}
> > > +
> > >  int btf_resolve_helper_id(struct bpf_verifier_log *log,
> > >                           const struct bpf_func_proto *fn, int arg)
> > >  {
> >
> > [...]
> >
>
Jiri Olsa July 29, 2020, 6:55 p.m. UTC | #4
On Wed, Jul 29, 2020 at 10:51:26AM -0700, Andrii Nakryiko wrote:
> On Wed, Jul 29, 2020 at 9:04 AM Jiri Olsa <jolsa@redhat.com> wrote:
> >
> > On Tue, Jul 28, 2020 at 04:35:16PM -0700, Andrii Nakryiko wrote:
> >
> > SNIP
> >
> > > > diff --git a/include/linux/bpf.h b/include/linux/bpf.h
> > > > index bae557ff2da8..c981e258fed3 100644
> > > > --- a/include/linux/bpf.h
> > > > +++ b/include/linux/bpf.h
> > > > @@ -1306,6 +1306,8 @@ int btf_struct_access(struct bpf_verifier_log *log,
> > > >                       const struct btf_type *t, int off, int size,
> > > >                       enum bpf_access_type atype,
> > > >                       u32 *next_btf_id);
> > > > +bool btf_struct_ids_match(struct bpf_verifier_log *log,
> > > > +                         int off, u32 id, u32 mid);
> > > >  int btf_resolve_helper_id(struct bpf_verifier_log *log,
> > > >                           const struct bpf_func_proto *fn, int);
> > > >
> > > > diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
> > > > index 1ab5fd5bf992..562d4453fad3 100644
> > > > --- a/kernel/bpf/btf.c
> > > > +++ b/kernel/bpf/btf.c
> > > > @@ -4140,6 +4140,35 @@ int btf_struct_access(struct bpf_verifier_log *log,
> > > >         return -EINVAL;
> > > >  }
> > > >
> > > > +bool btf_struct_ids_match(struct bpf_verifier_log *log,
> > > > +                         int off, u32 id, u32 mid)
> 
> just realized that if id == mid and off == 0, btf_struct_ids_match()
> will return false. Right now verifier is careful to not call
> btf_struct_ids_match in such case, but I wonder if it's better to make
> that (common) case also work?

right, also we should call btf_struct_ids_match when
IDs are equal and off != 0, which we don't do now

jirka
diff mbox series

Patch

diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index bae557ff2da8..c981e258fed3 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -1306,6 +1306,8 @@  int btf_struct_access(struct bpf_verifier_log *log,
 		      const struct btf_type *t, int off, int size,
 		      enum bpf_access_type atype,
 		      u32 *next_btf_id);
+bool btf_struct_ids_match(struct bpf_verifier_log *log,
+			  int off, u32 id, u32 mid);
 int btf_resolve_helper_id(struct bpf_verifier_log *log,
 			  const struct bpf_func_proto *fn, int);
 
diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
index 1ab5fd5bf992..562d4453fad3 100644
--- a/kernel/bpf/btf.c
+++ b/kernel/bpf/btf.c
@@ -4140,6 +4140,35 @@  int btf_struct_access(struct bpf_verifier_log *log,
 	return -EINVAL;
 }
 
+bool btf_struct_ids_match(struct bpf_verifier_log *log,
+			  int off, u32 id, u32 mid)
+{
+	const struct btf_type *type;
+	u32 nid;
+	int err;
+
+	do {
+		type = btf_type_by_id(btf_vmlinux, id);
+		if (!type)
+			return false;
+		err = btf_struct_walk(log, type, off, 1, &nid);
+		if (err < 0)
+			return false;
+
+		/* We found nested struct object. If it matches
+		 * the requested ID, we're done. Otherwise let's
+		 * continue the search with offset 0 in the new
+		 * type.
+		 */
+		if (err == walk_struct && mid == nid)
+			return true;
+		off = 0;
+		id = nid;
+	} while (err == walk_struct);
+
+	return false;
+}
+
 int btf_resolve_helper_id(struct bpf_verifier_log *log,
 			  const struct bpf_func_proto *fn, int arg)
 {
diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 9a6703bc3f36..39922fa07154 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -3887,16 +3887,21 @@  static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 				goto err_type;
 		}
 	} else if (arg_type == ARG_PTR_TO_BTF_ID) {
+		bool ids_match = false;
+
 		expected_type = PTR_TO_BTF_ID;
 		if (type != expected_type)
 			goto err_type;
 		if (!fn->check_btf_id) {
 			if (reg->btf_id != meta->btf_id) {
-				verbose(env, "Helper has type %s got %s in R%d\n",
-					kernel_type_name(meta->btf_id),
-					kernel_type_name(reg->btf_id), regno);
-
-				return -EACCES;
+				ids_match = btf_struct_ids_match(&env->log, reg->off, reg->btf_id,
+								 meta->btf_id);
+				if (!ids_match) {
+					verbose(env, "Helper has type %s got %s in R%d\n",
+						kernel_type_name(meta->btf_id),
+						kernel_type_name(reg->btf_id), regno);
+					return -EACCES;
+				}
 			}
 		} else if (!fn->check_btf_id(reg->btf_id, arg)) {
 			verbose(env, "Helper does not support %s in R%d\n",
@@ -3904,7 +3909,8 @@  static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 
 			return -EACCES;
 		}
-		if (!tnum_is_const(reg->var_off) || reg->var_off.value || reg->off) {
+		if (!ids_match &&
+		    (!tnum_is_const(reg->var_off) || reg->var_off.value || reg->off)) {
 			verbose(env, "R%d is a pointer to in-kernel struct with non-zero offset\n",
 				regno);
 			return -EACCES;