[bpf-next] bpf: make the verifier trace the "not qeual" for regs

Commit Message

Menglong Dong Dec. 10, 2023, 1 p.m. UTC

We can derive some new information for BPF_JNE in regs_refine_cond_op().
Take following code for example:

  /* The type of "a" is u16 */
  if (a > 0 && a < 100) {
    /* the range of the register for a is [0, 99], not [1, 99],
     * and will cause the following error:
     *
     *   invalid zero-sized read
     *
     * as a can be 0.
     */
    bpf_skb_store_bytes(skb, xx, xx, a, 0);
  }

In the code above, "a > 0" will be compiled to "jmp xxx if a == 0". In the
TRUE branch, the dst_reg will be marked as known to 0. However, in the
fallthrough(FALSE) branch, the dst_reg will not be handled, which makes
the [min, max] for a is [0, 99], not [1, 99].

For BPF_JNE, we can reduce the range of the dst reg if the src reg is a
const and is exactly the edge of the dst reg.

Signed-off-by: Menglong Dong <[e-mail]>
---
 kernel/bpf/verifier.c | 45 ++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 44 insertions(+), 1 deletion(-)

Comments

Yonghong Song Dec. 11, 2023, 5:09 a.m. UTC | #1

On 12/10/23 5:00 AM, Menglong Dong wrote:
> We can derive some new information for BPF_JNE in regs_refine_cond_op().
> Take following code for example:
>
>    /* The type of "a" is u16 */
>    if (a > 0 && a < 100) {
>      /* the range of the register for a is [0, 99], not [1, 99],
>       * and will cause the following error:
>       *
>       *   invalid zero-sized read
>       *
>       * as a can be 0.
>       */
>      bpf_skb_store_bytes(skb, xx, xx, a, 0);
>    }

Could you have a C test to demonstrate this example?
Also, you should have a set of inline asm code (progs/verifier*.c)
to test various cases as in mark_reg32_not_equal() and
mark_reg_not_equal().

>
> In the code above, "a > 0" will be compiled to "jmp xxx if a == 0". In the
> TRUE branch, the dst_reg will be marked as known to 0. However, in the
> fallthrough(FALSE) branch, the dst_reg will not be handled, which makes
> the [min, max] for a is [0, 99], not [1, 99].
>
> For BPF_JNE, we can reduce the range of the dst reg if the src reg is a
> const and is exactly the edge of the dst reg.
>
> Signed-off-by: Menglong Dong <[e-mail]>
> ---
>   kernel/bpf/verifier.c | 45 ++++++++++++++++++++++++++++++++++++++++++-
>   1 file changed, 44 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> index 727a59e4a647..7b074ac93190 100644
> --- a/kernel/bpf/verifier.c
> +++ b/kernel/bpf/verifier.c
> @@ -1764,6 +1764,40 @@ static void __mark_reg_const_zero(struct bpf_reg_state *reg)
>   	reg->type = SCALAR_VALUE;
>   }
>   
> +#define CHECK_REG_MIN(value)			\
> +do {						\
> +	if ((value) == (typeof(value))imm)	\
> +		value++;			\
> +} while (0)
> +
> +#define CHECK_REG_MAX(value)			\
> +do {						\
> +	if ((value) == (typeof(value))imm)	\
> +		value--;			\
> +} while (0)
> +
> +static void mark_reg32_not_equal(struct bpf_reg_state *reg, u64 imm)
> +{

What if reg->s32_min_value == imm and reg->s32_max_value == imm?
Has this been handled in previous verifier logic?

> +		CHECK_REG_MIN(reg->s32_min_value);
> +		CHECK_REG_MAX(reg->s32_max_value);
> +		CHECK_REG_MIN(reg->u32_min_value);
> +		CHECK_REG_MAX(reg->u32_max_value);
> +}
> +
> +static void mark_reg_not_equal(struct bpf_reg_state *reg, u64 imm)
> +{
> +		CHECK_REG_MIN(reg->smin_value);
> +		CHECK_REG_MAX(reg->smax_value);
> +
> +		CHECK_REG_MIN(reg->umin_value);
> +		CHECK_REG_MAX(reg->umax_value);
> +
> +		CHECK_REG_MIN(reg->s32_min_value);
> +		CHECK_REG_MAX(reg->s32_max_value);
> +		CHECK_REG_MIN(reg->u32_min_value);
> +		CHECK_REG_MAX(reg->u32_max_value);
> +}
> +
>   static void mark_reg_known_zero(struct bpf_verifier_env *env,
>   				struct bpf_reg_state *regs, u32 regno)
>   {
> @@ -14332,7 +14366,16 @@ static void regs_refine_cond_op(struct bpf_reg_state *reg1, struct bpf_reg_state
>   		}
>   		break;
>   	case BPF_JNE:
> -		/* we don't derive any new information for inequality yet */
> +		/* try to recompute the bound of reg1 if reg2 is a const and
> +		 * is exactly the edge of reg1.
> +		 */
> +		if (is_reg_const(reg2, is_jmp32)) {
> +			val = reg_const_value(reg2, is_jmp32);
> +			if (is_jmp32)
> +				mark_reg32_not_equal(reg1, val);
> +			else
> +				mark_reg_not_equal(reg1, val);
> +		}
>   		break;
>   	case BPF_JSET:
>   		if (!is_reg_const(reg2, is_jmp32))

Menglong Dong Dec. 11, 2023, 9:39 a.m. UTC | #2

Hello,

On Mon, Dec 11, 2023 at 1:09 PM Yonghong Song <[e-mail]> wrote:
>
>
> On 12/10/23 5:00 AM, Menglong Dong wrote:
> > We can derive some new information for BPF_JNE in regs_refine_cond_op().
> > Take following code for example:
> >
> >    /* The type of "a" is u16 */
> >    if (a > 0 && a < 100) {
> >      /* the range of the register for a is [0, 99], not [1, 99],
> >       * and will cause the following error:
> >       *
> >       *   invalid zero-sized read
> >       *
> >       * as a can be 0.
> >       */
> >      bpf_skb_store_bytes(skb, xx, xx, a, 0);
> >    }
>
> Could you have a C test to demonstrate this example?
> Also, you should have a set of inline asm code (progs/verifier*.c)
> to test various cases as in mark_reg32_not_equal() and
> mark_reg_not_equal().
>

Yeah! I found that this part is tested in the test_progs/reg_bounds_crafted
too, and this commit failed that test case, which I should fix in the next
version.

> >
> > In the code above, "a > 0" will be compiled to "jmp xxx if a == 0". In the
> > TRUE branch, the dst_reg will be marked as known to 0. However, in the
> > fallthrough(FALSE) branch, the dst_reg will not be handled, which makes
> > the [min, max] for a is [0, 99], not [1, 99].
> >
> > For BPF_JNE, we can reduce the range of the dst reg if the src reg is a
> > const and is exactly the edge of the dst reg.
> >
> > Signed-off-by: Menglong Dong <[e-mail]>
> > ---
> >   kernel/bpf/verifier.c | 45 ++++++++++++++++++++++++++++++++++++++++++-
> >   1 file changed, 44 insertions(+), 1 deletion(-)
> >
> > diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> > index 727a59e4a647..7b074ac93190 100644
> > --- a/kernel/bpf/verifier.c
> > +++ b/kernel/bpf/verifier.c
> > @@ -1764,6 +1764,40 @@ static void __mark_reg_const_zero(struct bpf_reg_state *reg)
> >       reg->type = SCALAR_VALUE;
> >   }
> >
> > +#define CHECK_REG_MIN(value)                 \
> > +do {                                         \
> > +     if ((value) == (typeof(value))imm)      \
> > +             value++;                        \
> > +} while (0)
> > +
> > +#define CHECK_REG_MAX(value)                 \
> > +do {                                         \
> > +     if ((value) == (typeof(value))imm)      \
> > +             value--;                        \
> > +} while (0)
> > +
> > +static void mark_reg32_not_equal(struct bpf_reg_state *reg, u64 imm)
> > +{
>
> What if reg->s32_min_value == imm and reg->s32_max_value == imm?
> Has this been handled in previous verifier logic?

Will such a case happen? In current code path, the src reg is a const,
and the is_branch_taken() will return 0 or 1 if the
dst_reg->s32_min_value == dst_reg->s32_max_value.

Enn......maybe we can do more checking here in case that someone
calls this function in another place.

Thanks!
Menglong Dong

>
> > +             CHECK_REG_MIN(reg->s32_min_value);
> > +             CHECK_REG_MAX(reg->s32_max_value);
> > +             CHECK_REG_MIN(reg->u32_min_value);
> > +             CHECK_REG_MAX(reg->u32_max_value);
> > +}
> > +
> > +static void mark_reg_not_equal(struct bpf_reg_state *reg, u64 imm)
> > +{
> > +             CHECK_REG_MIN(reg->smin_value);
> > +             CHECK_REG_MAX(reg->smax_value);
> > +
> > +             CHECK_REG_MIN(reg->umin_value);
> > +             CHECK_REG_MAX(reg->umax_value);
> > +
> > +             CHECK_REG_MIN(reg->s32_min_value);
> > +             CHECK_REG_MAX(reg->s32_max_value);
> > +             CHECK_REG_MIN(reg->u32_min_value);
> > +             CHECK_REG_MAX(reg->u32_max_value);
> > +}
> > +
> >   static void mark_reg_known_zero(struct bpf_verifier_env *env,
> >                               struct bpf_reg_state *regs, u32 regno)
> >   {
> > @@ -14332,7 +14366,16 @@ static void regs_refine_cond_op(struct bpf_reg_state *reg1, struct bpf_reg_state
> >               }
> >               break;
> >       case BPF_JNE:
> > -             /* we don't derive any new information for inequality yet */
> > +             /* try to recompute the bound of reg1 if reg2 is a const and
> > +              * is exactly the edge of reg1.
> > +              */
> > +             if (is_reg_const(reg2, is_jmp32)) {
> > +                     val = reg_const_value(reg2, is_jmp32);
> > +                     if (is_jmp32)
> > +                             mark_reg32_not_equal(reg1, val);
> > +                     else
> > +                             mark_reg_not_equal(reg1, val);
> > +             }
> >               break;
> >       case BPF_JSET:
> >               if (!is_reg_const(reg2, is_jmp32))

Yonghong Song Dec. 11, 2023, 3:03 p.m. UTC | #3

On 12/11/23 1:39 AM, Menglong Dong wrote:
> Hello,
>
> On Mon, Dec 11, 2023 at 1:09 PM Yonghong Song <[e-mail]> wrote:
>>
>> On 12/10/23 5:00 AM, Menglong Dong wrote:
>>> We can derive some new information for BPF_JNE in regs_refine_cond_op().
>>> Take following code for example:
>>>
>>>     /* The type of "a" is u16 */
>>>     if (a > 0 && a < 100) {
>>>       /* the range of the register for a is [0, 99], not [1, 99],
>>>        * and will cause the following error:
>>>        *
>>>        *   invalid zero-sized read
>>>        *
>>>        * as a can be 0.
>>>        */
>>>       bpf_skb_store_bytes(skb, xx, xx, a, 0);
>>>     }
>> Could you have a C test to demonstrate this example?
>> Also, you should have a set of inline asm code (progs/verifier*.c)
>> to test various cases as in mark_reg32_not_equal() and
>> mark_reg_not_equal().
>>
> Yeah! I found that this part is tested in the test_progs/reg_bounds_crafted
> too, and this commit failed that test case, which I should fix in the next
> version.
>
>>> In the code above, "a > 0" will be compiled to "jmp xxx if a == 0". In the
>>> TRUE branch, the dst_reg will be marked as known to 0. However, in the
>>> fallthrough(FALSE) branch, the dst_reg will not be handled, which makes
>>> the [min, max] for a is [0, 99], not [1, 99].
>>>
>>> For BPF_JNE, we can reduce the range of the dst reg if the src reg is a
>>> const and is exactly the edge of the dst reg.
>>>
>>> Signed-off-by: Menglong Dong <[e-mail]>
>>> ---
>>>    kernel/bpf/verifier.c | 45 ++++++++++++++++++++++++++++++++++++++++++-
>>>    1 file changed, 44 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
>>> index 727a59e4a647..7b074ac93190 100644
>>> --- a/kernel/bpf/verifier.c
>>> +++ b/kernel/bpf/verifier.c
>>> @@ -1764,6 +1764,40 @@ static void __mark_reg_const_zero(struct bpf_reg_state *reg)
>>>        reg->type = SCALAR_VALUE;
>>>    }
>>>
>>> +#define CHECK_REG_MIN(value)                 \
>>> +do {                                         \
>>> +     if ((value) == (typeof(value))imm)      \
>>> +             value++;                        \
>>> +} while (0)
>>> +
>>> +#define CHECK_REG_MAX(value)                 \
>>> +do {                                         \
>>> +     if ((value) == (typeof(value))imm)      \
>>> +             value--;                        \
>>> +} while (0)
>>> +
>>> +static void mark_reg32_not_equal(struct bpf_reg_state *reg, u64 imm)
>>> +{
>> What if reg->s32_min_value == imm and reg->s32_max_value == imm?
>> Has this been handled in previous verifier logic?
> Will such a case happen? In current code path, the src reg is a const,
> and the is_branch_taken() will return 0 or 1 if the
> dst_reg->s32_min_value == dst_reg->s32_max_value.
>
> Enn......maybe we can do more checking here in case that someone
> calls this function in another place.

I double checked the source code as well. Indeed, 'reg' should
not be a constant as it has been handled in is_branch_taken()
properly. Ignore my comments above then. Thanks!

>
> Thanks!
> Menglong Dong
>
>>> +             CHECK_REG_MIN(reg->s32_min_value);
>>> +             CHECK_REG_MAX(reg->s32_max_value);
>>> +             CHECK_REG_MIN(reg->u32_min_value);
>>> +             CHECK_REG_MAX(reg->u32_max_value);
>>> +}
>>> +
>>> +static void mark_reg_not_equal(struct bpf_reg_state *reg, u64 imm)
>>> +{
>>> +             CHECK_REG_MIN(reg->smin_value);
>>> +             CHECK_REG_MAX(reg->smax_value);
>>> +
>>> +             CHECK_REG_MIN(reg->umin_value);
>>> +             CHECK_REG_MAX(reg->umax_value);
>>> +
>>> +             CHECK_REG_MIN(reg->s32_min_value);
>>> +             CHECK_REG_MAX(reg->s32_max_value);
>>> +             CHECK_REG_MIN(reg->u32_min_value);
>>> +             CHECK_REG_MAX(reg->u32_max_value);
>>> +}
>>> +
>>>    static void mark_reg_known_zero(struct bpf_verifier_env *env,
>>>                                struct bpf_reg_state *regs, u32 regno)
>>>    {
>>> @@ -14332,7 +14366,16 @@ static void regs_refine_cond_op(struct bpf_reg_state *reg1, struct bpf_reg_state
>>>                }
>>>                break;
>>>        case BPF_JNE:
>>> -             /* we don't derive any new information for inequality yet */
>>> +             /* try to recompute the bound of reg1 if reg2 is a const and
>>> +              * is exactly the edge of reg1.
>>> +              */
>>> +             if (is_reg_const(reg2, is_jmp32)) {
>>> +                     val = reg_const_value(reg2, is_jmp32);
>>> +                     if (is_jmp32)
>>> +                             mark_reg32_not_equal(reg1, val);
>>> +                     else
>>> +                             mark_reg_not_equal(reg1, val);
>>> +             }
>>>                break;
>>>        case BPF_JSET:
>>>                if (!is_reg_const(reg2, is_jmp32))

Andrii Nakryiko Dec. 11, 2023, 7:15 p.m. UTC | #4

On Sun, Dec 10, 2023 at 5:00 AM Menglong Dong <[e-mail]> wrote:
>
> We can derive some new information for BPF_JNE in regs_refine_cond_op().
> Take following code for example:
>
>   /* The type of "a" is u16 */
>   if (a > 0 && a < 100) {
>     /* the range of the register for a is [0, 99], not [1, 99],
>      * and will cause the following error:
>      *
>      *   invalid zero-sized read
>      *
>      * as a can be 0.
>      */
>     bpf_skb_store_bytes(skb, xx, xx, a, 0);
>   }
>
> In the code above, "a > 0" will be compiled to "jmp xxx if a == 0". In the
> TRUE branch, the dst_reg will be marked as known to 0. However, in the
> fallthrough(FALSE) branch, the dst_reg will not be handled, which makes
> the [min, max] for a is [0, 99], not [1, 99].
>
> For BPF_JNE, we can reduce the range of the dst reg if the src reg is a
> const and is exactly the edge of the dst reg.
>
> Signed-off-by: Menglong Dong <[e-mail]>
> ---
>  kernel/bpf/verifier.c | 45 ++++++++++++++++++++++++++++++++++++++++++-
>  1 file changed, 44 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> index 727a59e4a647..7b074ac93190 100644
> --- a/kernel/bpf/verifier.c
> +++ b/kernel/bpf/verifier.c
> @@ -1764,6 +1764,40 @@ static void __mark_reg_const_zero(struct bpf_reg_state *reg)
>         reg->type = SCALAR_VALUE;
>  }
>
> +#define CHECK_REG_MIN(value)                   \
> +do {                                           \
> +       if ((value) == (typeof(value))imm)      \
> +               value++;                        \
> +} while (0)
> +
> +#define CHECK_REG_MAX(value)                   \
> +do {                                           \
> +       if ((value) == (typeof(value))imm)      \
> +               value--;                        \
> +} while (0)
> +
> +static void mark_reg32_not_equal(struct bpf_reg_state *reg, u64 imm)
> +{
> +               CHECK_REG_MIN(reg->s32_min_value);
> +               CHECK_REG_MAX(reg->s32_max_value);
> +               CHECK_REG_MIN(reg->u32_min_value);
> +               CHECK_REG_MAX(reg->u32_max_value);
> +}
> +
> +static void mark_reg_not_equal(struct bpf_reg_state *reg, u64 imm)
> +{
> +               CHECK_REG_MIN(reg->smin_value);
> +               CHECK_REG_MAX(reg->smax_value);
> +
> +               CHECK_REG_MIN(reg->umin_value);
> +               CHECK_REG_MAX(reg->umax_value);
> +
> +               CHECK_REG_MIN(reg->s32_min_value);
> +               CHECK_REG_MAX(reg->s32_max_value);
> +               CHECK_REG_MIN(reg->u32_min_value);
> +               CHECK_REG_MAX(reg->u32_max_value);
> +}

please don't use macros for this, this code is tricky enough without
having to jump around double-checking what exactly macros are doing.
Just code it explicitly.

Also I don't see the need for mark_reg32_not_equal() and
mark_reg_not_equal() helper functions, there is just one place where
this logic is going to be called from, so let's add code right there.

> +
>  static void mark_reg_known_zero(struct bpf_verifier_env *env,
>                                 struct bpf_reg_state *regs, u32 regno)
>  {
> @@ -14332,7 +14366,16 @@ static void regs_refine_cond_op(struct bpf_reg_state *reg1, struct bpf_reg_state
>                 }
>                 break;
>         case BPF_JNE:
> -               /* we don't derive any new information for inequality yet */
> +               /* try to recompute the bound of reg1 if reg2 is a const and
> +                * is exactly the edge of reg1.
> +                */
> +               if (is_reg_const(reg2, is_jmp32)) {
> +                       val = reg_const_value(reg2, is_jmp32);
> +                       if (is_jmp32)
> +                               mark_reg32_not_equal(reg1, val);
> +                       else
> +                               mark_reg_not_equal(reg1, val);
> +               }
>                 break;
>         case BPF_JSET:
>                 if (!is_reg_const(reg2, is_jmp32))
> --
> 2.39.2
>

Menglong Dong Dec. 12, 2023, 2:15 a.m. UTC | #5

On Tue, Dec 12, 2023 at 3:16 AM Andrii Nakryiko
<[e-mail]> wrote:
>
> On Sun, Dec 10, 2023 at 5:00 AM Menglong Dong <[e-mail]> wrote:
> >
> > We can derive some new information for BPF_JNE in regs_refine_cond_op().
> > Take following code for example:
> >
> >   /* The type of "a" is u16 */
> >   if (a > 0 && a < 100) {
> >     /* the range of the register for a is [0, 99], not [1, 99],
> >      * and will cause the following error:
> >      *
> >      *   invalid zero-sized read
> >      *
> >      * as a can be 0.
> >      */
> >     bpf_skb_store_bytes(skb, xx, xx, a, 0);
> >   }
> >
> > In the code above, "a > 0" will be compiled to "jmp xxx if a == 0". In the
> > TRUE branch, the dst_reg will be marked as known to 0. However, in the
> > fallthrough(FALSE) branch, the dst_reg will not be handled, which makes
> > the [min, max] for a is [0, 99], not [1, 99].
> >
> > For BPF_JNE, we can reduce the range of the dst reg if the src reg is a
> > const and is exactly the edge of the dst reg.
> >
> > Signed-off-by: Menglong Dong <[e-mail]>
> > ---
> >  kernel/bpf/verifier.c | 45 ++++++++++++++++++++++++++++++++++++++++++-
> >  1 file changed, 44 insertions(+), 1 deletion(-)
> >
> > diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> > index 727a59e4a647..7b074ac93190 100644
> > --- a/kernel/bpf/verifier.c
> > +++ b/kernel/bpf/verifier.c
> > @@ -1764,6 +1764,40 @@ static void __mark_reg_const_zero(struct bpf_reg_state *reg)
> >         reg->type = SCALAR_VALUE;
> >  }
> >
> > +#define CHECK_REG_MIN(value)                   \
> > +do {                                           \
> > +       if ((value) == (typeof(value))imm)      \
> > +               value++;                        \
> > +} while (0)
> > +
> > +#define CHECK_REG_MAX(value)                   \
> > +do {                                           \
> > +       if ((value) == (typeof(value))imm)      \
> > +               value--;                        \
> > +} while (0)
> > +
> > +static void mark_reg32_not_equal(struct bpf_reg_state *reg, u64 imm)
> > +{
> > +               CHECK_REG_MIN(reg->s32_min_value);
> > +               CHECK_REG_MAX(reg->s32_max_value);
> > +               CHECK_REG_MIN(reg->u32_min_value);
> > +               CHECK_REG_MAX(reg->u32_max_value);
> > +}
> > +
> > +static void mark_reg_not_equal(struct bpf_reg_state *reg, u64 imm)
> > +{
> > +               CHECK_REG_MIN(reg->smin_value);
> > +               CHECK_REG_MAX(reg->smax_value);
> > +
> > +               CHECK_REG_MIN(reg->umin_value);
> > +               CHECK_REG_MAX(reg->umax_value);
> > +
> > +               CHECK_REG_MIN(reg->s32_min_value);
> > +               CHECK_REG_MAX(reg->s32_max_value);
> > +               CHECK_REG_MIN(reg->u32_min_value);
> > +               CHECK_REG_MAX(reg->u32_max_value);
> > +}
>
> please don't use macros for this, this code is tricky enough without
> having to jump around double-checking what exactly macros are doing.
> Just code it explicitly.
>

Okay!

> Also I don't see the need for mark_reg32_not_equal() and
> mark_reg_not_equal() helper functions, there is just one place where
> this logic is going to be called from, so let's add code right there.
>

Yeah, you are right. And I just found that you have already
implemented the test case for this logic in reg_bounds.c/range_cond().
I wonder why this logic is not implemented in the verifier yet?
Am I missing something?

Thanks!
Menglong Dong

> > +
> >  static void mark_reg_known_zero(struct bpf_verifier_env *env,
> >                                 struct bpf_reg_state *regs, u32 regno)
> >  {
> > @@ -14332,7 +14366,16 @@ static void regs_refine_cond_op(struct bpf_reg_state *reg1, struct bpf_reg_state
> >                 }
> >                 break;
> >         case BPF_JNE:
> > -               /* we don't derive any new information for inequality yet */
> > +               /* try to recompute the bound of reg1 if reg2 is a const and
> > +                * is exactly the edge of reg1.
> > +                */
> > +               if (is_reg_const(reg2, is_jmp32)) {
> > +                       val = reg_const_value(reg2, is_jmp32);
> > +                       if (is_jmp32)
> > +                               mark_reg32_not_equal(reg1, val);
> > +                       else
> > +                               mark_reg_not_equal(reg1, val);
> > +               }
> >                 break;
> >         case BPF_JSET:
> >                 if (!is_reg_const(reg2, is_jmp32))
> > --
> > 2.39.2
> >

Andrii Nakryiko Dec. 12, 2023, 3:51 a.m. UTC | #6

On Mon, Dec 11, 2023 at 6:16 PM Menglong Dong <[e-mail]> wrote:
>
> On Tue, Dec 12, 2023 at 3:16 AM Andrii Nakryiko
> <[e-mail]> wrote:
> >
> > On Sun, Dec 10, 2023 at 5:00 AM Menglong Dong <[e-mail]> wrote:
> > >
> > > We can derive some new information for BPF_JNE in regs_refine_cond_op().
> > > Take following code for example:
> > >
> > >   /* The type of "a" is u16 */
> > >   if (a > 0 && a < 100) {
> > >     /* the range of the register for a is [0, 99], not [1, 99],
> > >      * and will cause the following error:
> > >      *
> > >      *   invalid zero-sized read
> > >      *
> > >      * as a can be 0.
> > >      */
> > >     bpf_skb_store_bytes(skb, xx, xx, a, 0);
> > >   }
> > >
> > > In the code above, "a > 0" will be compiled to "jmp xxx if a == 0". In the
> > > TRUE branch, the dst_reg will be marked as known to 0. However, in the
> > > fallthrough(FALSE) branch, the dst_reg will not be handled, which makes
> > > the [min, max] for a is [0, 99], not [1, 99].
> > >
> > > For BPF_JNE, we can reduce the range of the dst reg if the src reg is a
> > > const and is exactly the edge of the dst reg.
> > >
> > > Signed-off-by: Menglong Dong <[e-mail]>
> > > ---
> > >  kernel/bpf/verifier.c | 45 ++++++++++++++++++++++++++++++++++++++++++-
> > >  1 file changed, 44 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> > > index 727a59e4a647..7b074ac93190 100644
> > > --- a/kernel/bpf/verifier.c
> > > +++ b/kernel/bpf/verifier.c
> > > @@ -1764,6 +1764,40 @@ static void __mark_reg_const_zero(struct bpf_reg_state *reg)
> > >         reg->type = SCALAR_VALUE;
> > >  }
> > >
> > > +#define CHECK_REG_MIN(value)                   \
> > > +do {                                           \
> > > +       if ((value) == (typeof(value))imm)      \
> > > +               value++;                        \
> > > +} while (0)
> > > +
> > > +#define CHECK_REG_MAX(value)                   \
> > > +do {                                           \
> > > +       if ((value) == (typeof(value))imm)      \
> > > +               value--;                        \
> > > +} while (0)
> > > +
> > > +static void mark_reg32_not_equal(struct bpf_reg_state *reg, u64 imm)
> > > +{
> > > +               CHECK_REG_MIN(reg->s32_min_value);
> > > +               CHECK_REG_MAX(reg->s32_max_value);
> > > +               CHECK_REG_MIN(reg->u32_min_value);
> > > +               CHECK_REG_MAX(reg->u32_max_value);
> > > +}
> > > +
> > > +static void mark_reg_not_equal(struct bpf_reg_state *reg, u64 imm)
> > > +{
> > > +               CHECK_REG_MIN(reg->smin_value);
> > > +               CHECK_REG_MAX(reg->smax_value);
> > > +
> > > +               CHECK_REG_MIN(reg->umin_value);
> > > +               CHECK_REG_MAX(reg->umax_value);
> > > +
> > > +               CHECK_REG_MIN(reg->s32_min_value);
> > > +               CHECK_REG_MAX(reg->s32_max_value);
> > > +               CHECK_REG_MIN(reg->u32_min_value);
> > > +               CHECK_REG_MAX(reg->u32_max_value);
> > > +}
> >
> > please don't use macros for this, this code is tricky enough without
> > having to jump around double-checking what exactly macros are doing.
> > Just code it explicitly.
> >
>
> Okay!
>
> > Also I don't see the need for mark_reg32_not_equal() and
> > mark_reg_not_equal() helper functions, there is just one place where
> > this logic is going to be called from, so let's add code right there.
> >
>
> Yeah, you are right. And I just found that you have already
> implemented the test case for this logic in reg_bounds.c/range_cond().
> I wonder why this logic is not implemented in the verifier yet?
> Am I missing something?

No, I just didn't want to add yet more verifier changes in my original
patch set on extending reg bounds logic.

>
> Thanks!
> Menglong Dong
>
> > > +
> > >  static void mark_reg_known_zero(struct bpf_verifier_env *env,
> > >                                 struct bpf_reg_state *regs, u32 regno)
> > >  {
> > > @@ -14332,7 +14366,16 @@ static void regs_refine_cond_op(struct bpf_reg_state *reg1, struct bpf_reg_state
> > >                 }
> > >                 break;
> > >         case BPF_JNE:
> > > -               /* we don't derive any new information for inequality yet */
> > > +               /* try to recompute the bound of reg1 if reg2 is a const and
> > > +                * is exactly the edge of reg1.
> > > +                */
> > > +               if (is_reg_const(reg2, is_jmp32)) {
> > > +                       val = reg_const_value(reg2, is_jmp32);
> > > +                       if (is_jmp32)
> > > +                               mark_reg32_not_equal(reg1, val);
> > > +                       else
> > > +                               mark_reg_not_equal(reg1, val);
> > > +               }
> > >                 break;
> > >         case BPF_JSET:
> > >                 if (!is_reg_const(reg2, is_jmp32))
> > > --
> > > 2.39.2
> > >

Patch

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 727a59e4a647..7b074ac93190 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -1764,6 +1764,40 @@  static void __mark_reg_const_zero(struct bpf_reg_state *reg)
 	reg->type = SCALAR_VALUE;
 }
 
+#define CHECK_REG_MIN(value)			\
+do {						\
+	if ((value) == (typeof(value))imm)	\
+		value++;			\
+} while (0)
+
+#define CHECK_REG_MAX(value)			\
+do {						\
+	if ((value) == (typeof(value))imm)	\
+		value--;			\
+} while (0)
+
+static void mark_reg32_not_equal(struct bpf_reg_state *reg, u64 imm)
+{
+		CHECK_REG_MIN(reg->s32_min_value);
+		CHECK_REG_MAX(reg->s32_max_value);
+		CHECK_REG_MIN(reg->u32_min_value);
+		CHECK_REG_MAX(reg->u32_max_value);
+}
+
+static void mark_reg_not_equal(struct bpf_reg_state *reg, u64 imm)
+{
+		CHECK_REG_MIN(reg->smin_value);
+		CHECK_REG_MAX(reg->smax_value);
+
+		CHECK_REG_MIN(reg->umin_value);
+		CHECK_REG_MAX(reg->umax_value);
+
+		CHECK_REG_MIN(reg->s32_min_value);
+		CHECK_REG_MAX(reg->s32_max_value);
+		CHECK_REG_MIN(reg->u32_min_value);
+		CHECK_REG_MAX(reg->u32_max_value);
+}
+
 static void mark_reg_known_zero(struct bpf_verifier_env *env,
 				struct bpf_reg_state *regs, u32 regno)
 {
@@ -14332,7 +14366,16 @@  static void regs_refine_cond_op(struct bpf_reg_state *reg1, struct bpf_reg_state
 		}
 		break;
 	case BPF_JNE:
-		/* we don't derive any new information for inequality yet */
+		/* try to recompute the bound of reg1 if reg2 is a const and
+		 * is exactly the edge of reg1.
+		 */
+		if (is_reg_const(reg2, is_jmp32)) {
+			val = reg_const_value(reg2, is_jmp32);
+			if (is_jmp32)
+				mark_reg32_not_equal(reg1, val);
+			else
+				mark_reg_not_equal(reg1, val);
+		}
 		break;
 	case BPF_JSET:
 		if (!is_reg_const(reg2, is_jmp32))