Re: [bpf-next PATCH 03/10] bpf: verifer, adjust_scalar_min_max_vals to always call update_reg_bounds()

Re: [bpf-next PATCH 03/10] bpf: verifer, adjust_scalar_min_max_vals to always call update_reg_bounds()

[Ma Xinjian]

[-- Attachment #1: Type: text/plain, Size: 3374 bytes --]

Hi, John.


During our test, we noticed this commit not only cause test_verifier error

"bpf: test_verifier, #70 error message updates for 32-bit right shift"

But also cause test_align error:

```

"Test   9: dubious pointer arithmetic ... Failed to find match 7: 
R5_w=inv(id=0,var_off=(0x2; 0xfffffffffffffffc))"

"Test  10: variable subtraction ... Failed to find match 12: 
R6=inv(id=0,smin_value=-1006,smax_value=1034,var_off=(0x2; 
0xfffffffffffffffc))"

```

clang/llvm version: 11.0.0


Whole error message please refer to  the attachment.

part of error info:

```

# Test   9: dubious pointer arithmetic ... Failed to find match 7: 
R5_w=inv(id=0,var_off=(0x2; 0xfffffffffffffffc))
# func#0 @0
# 0: R1=ctx(id=0,off=0,imm=0) R10=fp0
# 0: (61) r2 = *(u32 *)(r1 +76)
# 1: R1=ctx(id=0,off=0,imm=0) R2_w=pkt(id=0,off=0,r=0,imm=0) R10=fp0
# 1: (61) r3 = *(u32 *)(r1 +80)
# 2: R1=ctx(id=0,off=0,imm=0) R2_w=pkt(id=0,off=0,r=0,imm=0) 
R3_w=pkt_end(id=0,off=0,imm=0) R10=fp0
# 2: (b7) r0 = 0
# 3: R0_w=inv0 R1=ctx(id=0,off=0,imm=0) R2_w=pkt(id=0,off=0,r=0,imm=0) 
R3_w=pkt_end(id=0,off=0,imm=0) R10=fp0
# 3: (bf) r5 = r3
# 4: R0_w=inv0 R1=ctx(id=0,off=0,imm=0) R2_w=pkt(id=0,off=0,r=0,imm=0) 
R3_w=pkt_end(id=0,off=0,imm=0) R5_w=pkt_end(id=0,off=0,imm=0) R10=fp0
# 4: (1f) r5 -= r2
# 5: R0_w=inv0 R1=ctx(id=0,off=0,imm=0) R2_w=pkt(id=0,off=0,r=0,imm=0) 
R3_w=pkt_end(id=0,off=0,imm=0) R5_w=inv(id=0) R10=fp0
# 5: (67) r5 <<= 2
# 6: R0_w=inv0 R1=ctx(id=0,off=0,imm=0) R2_w=pkt(id=0,off=0,r=0,imm=0) 
R3_w=pkt_end(id=0,off=0,imm=0) 
R5_w=inv(id=0,smax_value=9223372036854775804,umax_value=18446744073709551612,var_off=(0x0; 
0xfffffffffffffffc)) R10=fp0
# 6: (07) r5 += 14
[snip]

# 15: (61) r4 = *(u32 *)(r6 +0)
# invalid access to packet, off=0 size=4, R6(id=1,off=0,r=0)
# R6 offset is outside of the packet
# processed 16 insns (limit 1000000) max_states_per_insn 0 total_states 
1 peak_states 1 mark_read 1
# FAIL
# Test  10: variable subtraction ... Failed to find match 12: 
R6=inv(id=0,smin_value=-1006,smax_value=1034,var_off=(0x2; 
0xfffffffffffffffc))
# func#0 @0
# 0: R1=ctx(id=0,off=0,imm=0) R10=fp0
# 0: (61) r2 = *(u32 *)(r1 +76)
# 1: R1=ctx(id=0,off=0,imm=0) R2_w=pkt(id=0,off=0,r=0,imm=0) R10=fp0
# 1: (61) r3 = *(u32 *)(r1 +80)
# 2: R1=ctx(id=0,off=0,imm=0) R2_w=pkt(id=0,off=0,r=0,imm=0) 
R3_w=pkt_end(id=0,off=0,imm=0) R10=fp0
# 2: (bf) r0 = r2
# 3: R0_w=pkt(id=0,off=0,r=0,imm=0) R1=ctx(id=0,off=0,imm=0) 
R2_w=pkt(id=0,off=0,r=0,imm=0) R3_w=pkt_end(id=0,off=0,imm=0) R10=fp0
# 3: (07) r0 += 8
# 4: R0_w=pkt(id=0,off=8,r=0,imm=0) R1=ctx(id=0,off=0,imm=0) 
R2_w=pkt(id=0,off=0,r=0,imm=0) R3_w=pkt_end(id=0,off=0,imm=0) R10=fp0
# 4: (3d) if r3 >= r0 goto pc+1
#  R0_w=pkt(id=0,off=8,r=0,imm=0) R1=ctx(id=0,off=0,imm=0) 
R2_w=pkt(id=0,off=0,r=0,imm=0) R3_w=pkt_end(id=0,off=0,imm=0) R10=fp0
# 5: R0_w=pkt(id=0,off=8,r=0,imm=0) R1=ctx(id=0,off=0,imm=0) 
R2_w=pkt(id=0,off=0,r=0,imm=0) R3_w=pkt_end(id=0,off=0,imm=0) R10=fp0
# 5: (95) exit
# 6: R0_w=pkt(id=0,off=8,r=8,imm=0) R1=ctx(id=0,off=0,imm=0) 
R2_w=pkt(id=0,off=0,r=8,imm=0) R3_w=pkt_end(id=0,off=0,imm=0) R10=fp0
# 6: (71) r6 = *(u8 *)(r2 +0)
# 7: R0_w=pkt(id=0,off=8,r=8,imm=0) R1=ctx(id=0,off=0,imm=0) 
R2_w=pkt(id=0,off=0,r=8,imm=0) R3_w=pkt_end(id=0,off=0,imm=0) 
R6_w=inv(id=0,umax_value=255,var_off=(0x0; 0xff)) R10=fp0
# 7: (bf) r7 = r6
[snip]

```

---
Best Regards.
Ma Xinjian






[-- Attachment #2: bpf_test_align_log.gz --]
[-- Type: application/gzip, Size: 1503 bytes --]

[bpf-next PATCH 03/10] bpf: verifer, adjust_scalar_min_max_vals to always call update_reg_bounds()

[John Fastabend]

Currently, for all op verification we call __red_deduce_bounds() and
__red_bound_offset() but we only call __update_reg_bounds() in bitwise
ops. However, we could benefit from calling __update_reg_bounds() in
BPF_ADD, BPF_SUB, and BPF_MUL cases as well.

For example, a register with state 'R1_w=invP0' when we subtract from
it,

 w1 -= 2

Before coerce we will now have an smin_value=S64_MIN, smax_value=U64_MAX
and unsigned bounds umin_value=0, umax_value=U64_MAX. These will then
be clamped to S32_MIN, U32_MAX values by coerce in the case of alu32 op
as done in above example. However tnum will be a constant because the
ALU op is done on a constant.

Without update_reg_bounds() we have a scenario where tnum is a const
but our unsigned bounds do not reflect this. By calling update_reg_bounds
after coerce to 32bit we further refine the umin_value to U64_MAX in the
alu64 case or U32_MAX in the alu32 case above.

Signed-off-by: John Fastabend <john.fastabend@gmail.com>
---
 kernel/bpf/verifier.c |    1 +
 1 file changed, 1 insertion(+)

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index f7a34b1..fea9725 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -5202,6 +5202,7 @@ static int adjust_scalar_min_max_vals(struct bpf_verifier_env *env,
 		coerce_reg_to_size(dst_reg, 4);
 	}
 
+	__update_reg_bounds(dst_reg);
 	__reg_deduce_bounds(dst_reg);
 	__reg_bound_offset(dst_reg);
 	return 0;