RE: [PATCH bpf-next v1 1/2] bpf: Add support for per-parameter trusted args

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



> From: Kumar Kartikeya Dwivedi [mailto:memxor@xxxxxxxxx]
> Sent: Thursday, July 28, 2022 10:45 AM
> On Thu, 28 Jul 2022 at 10:18, Roberto Sassu <roberto.sassu@xxxxxxxxxx>
> wrote:
> >
> > > From: Roberto Sassu [mailto:roberto.sassu@xxxxxxxxxx]
> > > Sent: Thursday, July 28, 2022 9:46 AM
> > > > From: Kumar Kartikeya Dwivedi [mailto:memxor@xxxxxxxxx]
> > > > Sent: Wednesday, July 27, 2022 10:16 AM
> > > > Similar to how we detect mem, size pairs in kfunc, teach verifier to
> > > > treat __ref suffix on argument name to imply that it must be a trusted
> > > > arg when passed to kfunc, similar to the effect of KF_TRUSTED_ARGS flag
> > > > but limited to the specific parameter. This is required to ensure that
> > > > kfunc that operate on some object only work on acquired pointers and not
> > > > normal PTR_TO_BTF_ID with same type which can be obtained by pointer
> > > > walking. Release functions need not specify such suffix on release
> > > > arguments as they are already expected to receive one referenced
> > > > argument.
> > >
> > > Thanks, Kumar. I will try it.
> >
> > Uhm. I realized that I was already using another suffix,
> > __maybe_null, to indicate that a caller can pass NULL as
> > argument.
> >
> > Wouldn't probably work well with two suffixes.
> >
> 
> Then you can maybe extend it to parse two suffixes at most (for now atleast)?
> 
> > Have you considered to extend BTF_ID_FLAGS to take five
> > extra arguments, to set flags for each kfunc parameter?
> >
> 
> I didn't understand this. Flags parameter is an OR of the flags you
> set, why would we want to extend it to take 5 args?
> You can just or f1 | f2 | f3 | f4 | f5, as many as you want.

Instead of using the suffix as mechanism to set per-parameter
flags, I would do:

BTF_ID_FLAGS(func, bpf_verify_pkcs7_signature, KF_SLEEPABLE, 0, 0, KF_REF | KF_MAYBE_NULL, 0, 0)

where the first argument after the kfunc name is used for
function-wide flags, the second for first parameter flags, ...

Roberto

> > Thanks
> >
> > Roberto
> >
> > > Roberto
> > >
> > > > Cc: Roberto Sassu <roberto.sassu@xxxxxxxxxx>
> > > > Signed-off-by: Kumar Kartikeya Dwivedi <memxor@xxxxxxxxx>
> > > > ---
> > > >  Documentation/bpf/kfuncs.rst | 18 +++++++++++++++++
> > > >  kernel/bpf/btf.c             | 39 ++++++++++++++++++++++++------------
> > > >  net/bpf/test_run.c           |  9 +++++++--
> > > >  3 files changed, 51 insertions(+), 15 deletions(-)
> > > >
> > > > diff --git a/Documentation/bpf/kfuncs.rst b/Documentation/bpf/kfuncs.rst
> > > > index c0b7dae6dbf5..41dff6337446 100644
> > > > --- a/Documentation/bpf/kfuncs.rst
> > > > +++ b/Documentation/bpf/kfuncs.rst
> > > > @@ -72,6 +72,24 @@ argument as its size. By default, without __sz
> > > annotation,
> > > > the size of the type
> > > >  of the pointer is used. Without __sz annotation, a kfunc cannot accept a
> void
> > > >  pointer.
> > > >
> > > > +2.2.2 __ref Annotation
> > > > +----------------------
> > > > +
> > > > +This annotation is used to indicate that the argument is trusted, i.e. it will
> > > > +be a pointer from an acquire function (defined later), and its offset will be
> > > > +zero. This annotation has the same effect as the KF_TRUSTED_ARGS
> kfunc
> > > flag
> > > > but
> > > > +only on the parameter it is applied to. An example is shown below::
> > > > +
> > > > +        void bpf_task_send_signal(struct task_struct *task__ref, int signal)
> > > > +        {
> > > > +        ...
> > > > +        }
> > > > +
> > > > +Here, bpf_task_send_signal will only act on trusted task_struct pointers,
> and
> > > > +cannot be used on pointers obtained using pointer walking. This ensures
> that
> > > > +caller always calls this kfunc on a task whose lifetime is guaranteed for
> the
> > > > +duration of the call.
> > > > +
> > > >  .. _BPF_kfunc_nodef:
> > > >
> > > >  2.3 Using an existing kernel function
> > > > diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
> > > > index 7ac971ea98d1..3ce9b2deef9c 100644
> > > > --- a/kernel/bpf/btf.c
> > > > +++ b/kernel/bpf/btf.c
> > > > @@ -6141,18 +6141,13 @@ static bool __btf_type_is_scalar_struct(struct
> > > > bpf_verifier_log *log,
> > > >     return true;
> > > >  }
> > > >
> > > > -static bool is_kfunc_arg_mem_size(const struct btf *btf,
> > > > -                             const struct btf_param *arg,
> > > > -                             const struct bpf_reg_state *reg)
> > > > +static bool btf_param_match_suffix(const struct btf *btf,
> > > > +                              const struct btf_param *arg,
> > > > +                              const char *suffix)
> > > >  {
> > > > -   int len, sfx_len = sizeof("__sz") - 1;
> > > > -   const struct btf_type *t;
> > > > +   int len, sfx_len = strlen(suffix);
> > > >     const char *param_name;
> > > >
> > > > -   t = btf_type_skip_modifiers(btf, arg->type, NULL);
> > > > -   if (!btf_type_is_scalar(t) || reg->type != SCALAR_VALUE)
> > > > -           return false;
> > > > -
> > > >     /* In the future, this can be ported to use BTF tagging */
> > > >     param_name = btf_name_by_offset(btf, arg->name_off);
> > > >     if (str_is_empty(param_name))
> > > > @@ -6161,10 +6156,26 @@ static bool is_kfunc_arg_mem_size(const
> struct
> > > btf
> > > > *btf,
> > > >     if (len < sfx_len)
> > > >             return false;
> > > >     param_name += len - sfx_len;
> > > > -   if (strncmp(param_name, "__sz", sfx_len))
> > > > +   return !strncmp(param_name, suffix, sfx_len);
> > > > +}
> > > > +
> > > > +static bool is_kfunc_arg_ref(const struct btf *btf,
> > > > +                        const struct btf_param *arg)
> > > > +{
> > > > +   return btf_param_match_suffix(btf, arg, "__ref");
> > > > +}
> > > > +
> > > > +static bool is_kfunc_arg_mem_size(const struct btf *btf,
> > > > +                             const struct btf_param *arg,
> > > > +                             const struct bpf_reg_state *reg)
> > > > +{
> > > > +   const struct btf_type *t;
> > > > +
> > > > +   t = btf_type_skip_modifiers(btf, arg->type, NULL);
> > > > +   if (!btf_type_is_scalar(t) || reg->type != SCALAR_VALUE)
> > > >             return false;
> > > >
> > > > -   return true;
> > > > +   return btf_param_match_suffix(btf, arg, "__sz");
> > > >  }
> > > >
> > > >  static int btf_check_func_arg_match(struct bpf_verifier_env *env,
> > > > @@ -6174,7 +6185,7 @@ static int btf_check_func_arg_match(struct
> > > > bpf_verifier_env *env,
> > > >                                 u32 kfunc_flags)
> > > >  {
> > > >     enum bpf_prog_type prog_type = resolve_prog_type(env->prog);
> > > > -   bool rel = false, kptr_get = false, trusted_arg = false;
> > > > +   bool rel = false, kptr_get = false, kf_trusted_args = false;
> > > >     struct bpf_verifier_log *log = &env->log;
> > > >     u32 i, nargs, ref_id, ref_obj_id = 0;
> > > >     bool is_kfunc = btf_is_kernel(btf);
> > > > @@ -6211,7 +6222,7 @@ static int btf_check_func_arg_match(struct
> > > > bpf_verifier_env *env,
> > > >             /* Only kfunc can be release func */
> > > >             rel = kfunc_flags & KF_RELEASE;
> > > >             kptr_get = kfunc_flags & KF_KPTR_GET;
> > > > -           trusted_arg = kfunc_flags & KF_TRUSTED_ARGS;
> > > > +           kf_trusted_args = kfunc_flags & KF_TRUSTED_ARGS;
> > > >     }
> > > >
> > > >     /* check that BTF function arguments match actual types that the
> > > > @@ -6221,6 +6232,7 @@ static int btf_check_func_arg_match(struct
> > > > bpf_verifier_env *env,
> > > >             enum bpf_arg_type arg_type = ARG_DONTCARE;
> > > >             u32 regno = i + 1;
> > > >             struct bpf_reg_state *reg = &regs[regno];
> > > > +           bool trusted_arg = false;
> > > >
> > > >             t = btf_type_skip_modifiers(btf, args[i].type, NULL);
> > > >             if (btf_type_is_scalar(t)) {
> > > > @@ -6239,6 +6251,7 @@ static int btf_check_func_arg_match(struct
> > > > bpf_verifier_env *env,
> > > >             /* Check if argument must be a referenced pointer, args + i has
> > > >              * been verified to be a pointer (after skipping modifiers).
> > > >              */
> > > > +           trusted_arg = kf_trusted_args || is_kfunc_arg_ref(btf, args + i);
> > > >             if (is_kfunc && trusted_arg && !reg->ref_obj_id) {
> > > >                     bpf_log(log, "R%d must be referenced\n", regno);
> > > >                     return -EINVAL;
> > > > diff --git a/net/bpf/test_run.c b/net/bpf/test_run.c
> > > > index cbc9cd5058cb..247bfe52e585 100644
> > > > --- a/net/bpf/test_run.c
> > > > +++ b/net/bpf/test_run.c
> > > > @@ -691,7 +691,11 @@ noinline void
> > > bpf_kfunc_call_test_mem_len_fail2(u64
> > > > *mem, int len)
> > > >  {
> > > >  }
> > > >
> > > > -noinline void bpf_kfunc_call_test_ref(struct prog_test_ref_kfunc *p)
> > > > +noinline void bpf_kfunc_call_test_trusted(struct prog_test_ref_kfunc *p)
> > > > +{
> > > > +}
> > > > +
> > > > +noinline void bpf_kfunc_call_test_ref(struct prog_test_ref_kfunc *p__ref)
> > > >  {
> > > >  }
> > > >
> > > > @@ -718,7 +722,8 @@ BTF_ID_FLAGS(func, bpf_kfunc_call_test_fail3)
> > > >  BTF_ID_FLAGS(func, bpf_kfunc_call_test_mem_len_pass1)
> > > >  BTF_ID_FLAGS(func, bpf_kfunc_call_test_mem_len_fail1)
> > > >  BTF_ID_FLAGS(func, bpf_kfunc_call_test_mem_len_fail2)
> > > > -BTF_ID_FLAGS(func, bpf_kfunc_call_test_ref, KF_TRUSTED_ARGS)
> > > > +BTF_ID_FLAGS(func, bpf_kfunc_call_test_trusted, KF_TRUSTED_ARGS)
> > > > +BTF_ID_FLAGS(func, bpf_kfunc_call_test_ref)
> > > >  BTF_SET8_END(test_sk_check_kfunc_ids)
> > > >
> > > >  static void *bpf_test_init(const union bpf_attr *kattr, u32 user_size,
> > > > --
> > > > 2.34.1
> >




[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux