mbox series

[0/4] bpf powerpc: Add BPF_PROBE_MEM support for 64bit JIT

Message ID 20210706073211.349889-1-ravi.bangoria@linux.ibm.com
Headers show
Series bpf powerpc: Add BPF_PROBE_MEM support for 64bit JIT | expand

Message

Ravi Bangoria July 6, 2021, 7:32 a.m. UTC
Patch #1, #2 are simple cleanup patches. Patch #3 adds
BPF_PROBE_MEM support with PowerPC 64bit JIT compiler.
Patch #4 adds explicit addr > TASK_SIZE_MAX check to
handle bad userspace pointers.

Ravi Bangoria (4):
  bpf powerpc: Remove unused SEEN_STACK
  bpf powerpc: Remove extra_pass from bpf_jit_build_body()
  bpf powerpc: Add BPF_PROBE_MEM support for 64bit JIT
  bpf powerpc: Add addr > TASK_SIZE_MAX explicit check

 arch/powerpc/net/bpf_jit.h        |   8 ++-
 arch/powerpc/net/bpf_jit_comp.c   |  25 ++++++--
 arch/powerpc/net/bpf_jit_comp32.c |   4 +-
 arch/powerpc/net/bpf_jit_comp64.c | 100 +++++++++++++++++++++++++++++-
 4 files changed, 124 insertions(+), 13 deletions(-)

Comments

Christophe Leroy July 6, 2021, 10 a.m. UTC | #1
Le 06/07/2021 à 09:32, Ravi Bangoria a écrit :
> On PowerPC with KUAP enabled, any kernel code which wants to
> access userspace needs to be surrounded by disable-enable KUAP.
> But that is not happening for BPF_PROBE_MEM load instruction.
> So, when BPF program tries to access invalid userspace address,
> page-fault handler considers it as bad KUAP fault:
> 
>    Kernel attempted to read user page (d0000000) - exploit attempt? (uid: 0)
> 
> Considering the fact that PTR_TO_BTF_ID (which uses BPF_PROBE_MEM
> mode) could either be a valid kernel pointer or NULL but should
> never be a pointer to userspace address, execute BPF_PROBE_MEM load
> only if addr > TASK_SIZE_MAX, otherwise set dst_reg=0 and move on.
> 
> This will catch NULL, valid or invalid userspace pointers. Only bad
> kernel pointer will be handled by BPF exception table.
> 
> [Alexei suggested for x86]
> Suggested-by: Alexei Starovoitov <ast@kernel.org>
> Signed-off-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
> ---
>   arch/powerpc/net/bpf_jit_comp64.c | 38 +++++++++++++++++++++++++++++++
>   1 file changed, 38 insertions(+)
> 
> diff --git a/arch/powerpc/net/bpf_jit_comp64.c b/arch/powerpc/net/bpf_jit_comp64.c
> index 1884c6dca89a..46becae76210 100644
> --- a/arch/powerpc/net/bpf_jit_comp64.c
> +++ b/arch/powerpc/net/bpf_jit_comp64.c
> @@ -753,6 +753,14 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, struct codegen_context *
>   		/* dst = *(u8 *)(ul) (src + off) */
>   		case BPF_LDX | BPF_MEM | BPF_B:
>   		case BPF_LDX | BPF_PROBE_MEM | BPF_B:
> +			if (BPF_MODE(code) == BPF_PROBE_MEM) {
> +				EMIT(PPC_RAW_ADDI(b2p[TMP_REG_1], src_reg, off));
> +				PPC_LI64(b2p[TMP_REG_2], TASK_SIZE_MAX);
> +				EMIT(PPC_RAW_CMPLD(b2p[TMP_REG_1], b2p[TMP_REG_2]));
> +				PPC_BCC(COND_GT, (ctx->idx + 4) * 4);
> +				EMIT(PPC_RAW_XOR(dst_reg, dst_reg, dst_reg));

Prefered way to clear a register is to do 'li reg, 0'

> +				PPC_JMP((ctx->idx + 2) * 4);
> +			}
>   			EMIT(PPC_RAW_LBZ(dst_reg, src_reg, off));
>   			if (insn_is_zext(&insn[i + 1]))
>   				addrs[++i] = ctx->idx * 4;
> @@ -763,6 +771,14 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, struct codegen_context *
>   		/* dst = *(u16 *)(ul) (src + off) */
>   		case BPF_LDX | BPF_MEM | BPF_H:
>   		case BPF_LDX | BPF_PROBE_MEM | BPF_H:
> +			if (BPF_MODE(code) == BPF_PROBE_MEM) {
> +				EMIT(PPC_RAW_ADDI(b2p[TMP_REG_1], src_reg, off));
> +				PPC_LI64(b2p[TMP_REG_2], TASK_SIZE_MAX);
> +				EMIT(PPC_RAW_CMPLD(b2p[TMP_REG_1], b2p[TMP_REG_2]));
> +				PPC_BCC(COND_GT, (ctx->idx + 4) * 4);
> +				EMIT(PPC_RAW_XOR(dst_reg, dst_reg, dst_reg));
> +				PPC_JMP((ctx->idx + 2) * 4);
> +			}

That code seems strictly identical to the previous one and the next one.
Can you refactor in a function ?

>   			EMIT(PPC_RAW_LHZ(dst_reg, src_reg, off));
>   			if (insn_is_zext(&insn[i + 1]))
>   				addrs[++i] = ctx->idx * 4;
> @@ -773,6 +789,14 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, struct codegen_context *
>   		/* dst = *(u32 *)(ul) (src + off) */
>   		case BPF_LDX | BPF_MEM | BPF_W:
>   		case BPF_LDX | BPF_PROBE_MEM | BPF_W:
> +			if (BPF_MODE(code) == BPF_PROBE_MEM) {
> +				EMIT(PPC_RAW_ADDI(b2p[TMP_REG_1], src_reg, off));
> +				PPC_LI64(b2p[TMP_REG_2], TASK_SIZE_MAX);
> +				EMIT(PPC_RAW_CMPLD(b2p[TMP_REG_1], b2p[TMP_REG_2]));
> +				PPC_BCC(COND_GT, (ctx->idx + 4) * 4);
> +				EMIT(PPC_RAW_XOR(dst_reg, dst_reg, dst_reg));
> +				PPC_JMP((ctx->idx + 2) * 4);
> +			}
>   			EMIT(PPC_RAW_LWZ(dst_reg, src_reg, off));
>   			if (insn_is_zext(&insn[i + 1]))
>   				addrs[++i] = ctx->idx * 4;
> @@ -783,6 +807,20 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, struct codegen_context *
>   		/* dst = *(u64 *)(ul) (src + off) */
>   		case BPF_LDX | BPF_MEM | BPF_DW:
>   		case BPF_LDX | BPF_PROBE_MEM | BPF_DW:
> +			if (BPF_MODE(code) == BPF_PROBE_MEM) {
> +				EMIT(PPC_RAW_ADDI(b2p[TMP_REG_1], src_reg, off));
> +				PPC_LI64(b2p[TMP_REG_2], TASK_SIZE_MAX);
> +				EMIT(PPC_RAW_CMPLD(b2p[TMP_REG_1], b2p[TMP_REG_2]));
> +				if (off % 4)

That test is worth a comment.

And I'd prefer

	if (off & 3) {
		PPC_BCC(COND_GT, (ctx->idx + 5) * 4);
		EMIT(PPC_RAW_XOR(dst_reg, dst_reg, dst_reg));
		PPC_JMP((ctx->idx + 3) * 4);
	} else {
		PPC_BCC(COND_GT, (ctx->idx + 4) * 4);
		EMIT(PPC_RAW_XOR(dst_reg, dst_reg, dst_reg));
		PPC_JMP((ctx->idx + 2) * 4);
	}

> +					PPC_BCC(COND_GT, (ctx->idx + 5) * 4);
> +				else
> +					PPC_BCC(COND_GT, (ctx->idx + 4) * 4);
> +				EMIT(PPC_RAW_XOR(dst_reg, dst_reg, dst_reg));

Use PPC_RAW_LI(dst_reg, 0);

> +				if (off % 4)
> +					PPC_JMP((ctx->idx + 3) * 4);
> +				else
> +					PPC_JMP((ctx->idx + 2) * 4);
> +			}
>   			PPC_BPF_LL(dst_reg, src_reg, off);
>   			ret = add_extable_entry(fp, image, pass, code, ctx, dst_reg);
>   			if (ret)
>
Ravi Bangoria July 7, 2021, 4:06 a.m. UTC | #2
>> @@ -763,6 +771,14 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, struct codegen_context *

>>           /* dst = *(u16 *)(ul) (src + off) */

>>           case BPF_LDX | BPF_MEM | BPF_H:

>>           case BPF_LDX | BPF_PROBE_MEM | BPF_H:

>> +            if (BPF_MODE(code) == BPF_PROBE_MEM) {

>> +                EMIT(PPC_RAW_ADDI(b2p[TMP_REG_1], src_reg, off));

>> +                PPC_LI64(b2p[TMP_REG_2], TASK_SIZE_MAX);

>> +                EMIT(PPC_RAW_CMPLD(b2p[TMP_REG_1], b2p[TMP_REG_2]));

>> +                PPC_BCC(COND_GT, (ctx->idx + 4) * 4);

>> +                EMIT(PPC_RAW_XOR(dst_reg, dst_reg, dst_reg));

>> +                PPC_JMP((ctx->idx + 2) * 4);

>> +            }

> 

> That code seems strictly identical to the previous one and the next one.

> Can you refactor in a function ?


I'll check this.

> 

>>               EMIT(PPC_RAW_LHZ(dst_reg, src_reg, off));

>>               if (insn_is_zext(&insn[i + 1]))

>>                   addrs[++i] = ctx->idx * 4;

>> @@ -773,6 +789,14 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, struct codegen_context *

>>           /* dst = *(u32 *)(ul) (src + off) */

>>           case BPF_LDX | BPF_MEM | BPF_W:

>>           case BPF_LDX | BPF_PROBE_MEM | BPF_W:

>> +            if (BPF_MODE(code) == BPF_PROBE_MEM) {

>> +                EMIT(PPC_RAW_ADDI(b2p[TMP_REG_1], src_reg, off));

>> +                PPC_LI64(b2p[TMP_REG_2], TASK_SIZE_MAX);

>> +                EMIT(PPC_RAW_CMPLD(b2p[TMP_REG_1], b2p[TMP_REG_2]));

>> +                PPC_BCC(COND_GT, (ctx->idx + 4) * 4);

>> +                EMIT(PPC_RAW_XOR(dst_reg, dst_reg, dst_reg));

>> +                PPC_JMP((ctx->idx + 2) * 4);

>> +            }

>>               EMIT(PPC_RAW_LWZ(dst_reg, src_reg, off));

>>               if (insn_is_zext(&insn[i + 1]))

>>                   addrs[++i] = ctx->idx * 4;

>> @@ -783,6 +807,20 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, struct codegen_context *

>>           /* dst = *(u64 *)(ul) (src + off) */

>>           case BPF_LDX | BPF_MEM | BPF_DW:

>>           case BPF_LDX | BPF_PROBE_MEM | BPF_DW:

>> +            if (BPF_MODE(code) == BPF_PROBE_MEM) {

>> +                EMIT(PPC_RAW_ADDI(b2p[TMP_REG_1], src_reg, off));

>> +                PPC_LI64(b2p[TMP_REG_2], TASK_SIZE_MAX);

>> +                EMIT(PPC_RAW_CMPLD(b2p[TMP_REG_1], b2p[TMP_REG_2]));

>> +                if (off % 4)

> 

> That test is worth a comment.


(off % 4) test is based on how PPC_BPF_LL() emits instruction.

> 

> And I'd prefer

> 

>      if (off & 3) {

>          PPC_BCC(COND_GT, (ctx->idx + 5) * 4);

>          EMIT(PPC_RAW_XOR(dst_reg, dst_reg, dst_reg));

>          PPC_JMP((ctx->idx + 3) * 4);

>      } else {

>          PPC_BCC(COND_GT, (ctx->idx + 4) * 4);

>          EMIT(PPC_RAW_XOR(dst_reg, dst_reg, dst_reg));

>          PPC_JMP((ctx->idx + 2) * 4);

>      }


Yes this is neat.

Thanks for the review,
Ravi