| Message ID | 20201111031213.25109-2-alexei.starovoitov@gmail.com |
|---|---|
| State | Superseded |
| Headers | show |
| Series | bpf: Pointers beyond packet end. | expand |
Alexei Starovoitov wrote: > From: Alexei Starovoitov <ast@kernel.org> > > This patch adds the verifier support to recognize inlined branch conditions. > The LLVM knows that the branch evaluates to the same value, but the verifier > couldn't track it. Hence causing valid programs to be rejected. > The potential LLVM workaround: https://reviews.llvm.org/D87428 > can have undesired side effects, since LLVM doesn't know that > skb->data/data_end are being compared. LLVM has to introduce extra boolean > variable and use inline_asm trick to force easier for the verifier assembly. > > Instead teach the verifier to recognize that > r1 = skb->data; > r1 += 10; > r2 = skb->data_end; > if (r1 > r2) { > here r1 points beyond packet_end and > subsequent > if (r1 > r2) // always evaluates to "true". > } > > Tested-by: Jiri Olsa <jolsa@redhat.com> > Signed-off-by: Alexei Starovoitov <ast@kernel.org> > --- > include/linux/bpf_verifier.h | 2 +- > kernel/bpf/verifier.c | 129 +++++++++++++++++++++++++++++------ > 2 files changed, 108 insertions(+), 23 deletions(-) > Thanks, we can remove another set of inline asm logic. Acked-by: John Fastabend <john.fastabend@gmail.com> > if (pred >= 0) { > @@ -7517,7 +7601,8 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env, > */ > if (!__is_pointer_value(false, dst_reg)) > err = mark_chain_precision(env, insn->dst_reg); > - if (BPF_SRC(insn->code) == BPF_X && !err) > + if (BPF_SRC(insn->code) == BPF_X && !err && > + !__is_pointer_value(false, src_reg)) This could have been more specific with !type_is_pkt_pointer() correct? I think its fine as is though. > err = mark_chain_precision(env, insn->src_reg); > if (err) > return err; > -- > 2.24.1 >
On 11/12/20 8:16 PM, John Fastabend wrote: > Alexei Starovoitov wrote: >> From: Alexei Starovoitov <ast@kernel.org> >> >> This patch adds the verifier support to recognize inlined branch conditions. >> The LLVM knows that the branch evaluates to the same value, but the verifier >> couldn't track it. Hence causing valid programs to be rejected. >> The potential LLVM workaround: https://reviews.llvm.org/D87428 >> can have undesired side effects, since LLVM doesn't know that >> skb->data/data_end are being compared. LLVM has to introduce extra boolean >> variable and use inline_asm trick to force easier for the verifier assembly. >> >> Instead teach the verifier to recognize that >> r1 = skb->data; >> r1 += 10; >> r2 = skb->data_end; >> if (r1 > r2) { >> here r1 points beyond packet_end and >> subsequent >> if (r1 > r2) // always evaluates to "true". >> } >> >> Tested-by: Jiri Olsa <jolsa@redhat.com> >> Signed-off-by: Alexei Starovoitov <ast@kernel.org> >> --- >> include/linux/bpf_verifier.h | 2 +- >> kernel/bpf/verifier.c | 129 +++++++++++++++++++++++++++++------ >> 2 files changed, 108 insertions(+), 23 deletions(-) >> > > Thanks, we can remove another set of inline asm logic. > > Acked-by: John Fastabend <john.fastabend@gmail.com> > >> if (pred >= 0) { >> @@ -7517,7 +7601,8 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env, >> */ >> if (!__is_pointer_value(false, dst_reg)) >> err = mark_chain_precision(env, insn->dst_reg); >> - if (BPF_SRC(insn->code) == BPF_X && !err) >> + if (BPF_SRC(insn->code) == BPF_X && !err && >> + !__is_pointer_value(false, src_reg)) > > This could have been more specific with !type_is_pkt_pointer() correct? I > think its fine as is though. > >> err = mark_chain_precision(env, insn->src_reg); >> if (err) >> return err; Given the reg->range could now be negative, I wonder whether for the regsafe() pruning logic we should now better add a >=0 sanity check in there before we attempt to test on rold->range > rcur->range? Thanks, Daniel
On Thu, Nov 12, 2020 at 11:16:11AM -0800, John Fastabend wrote: > Alexei Starovoitov wrote: > > From: Alexei Starovoitov <ast@kernel.org> > > > > This patch adds the verifier support to recognize inlined branch conditions. > > The LLVM knows that the branch evaluates to the same value, but the verifier > > couldn't track it. Hence causing valid programs to be rejected. > > The potential LLVM workaround: https://reviews.llvm.org/D87428 > > can have undesired side effects, since LLVM doesn't know that > > skb->data/data_end are being compared. LLVM has to introduce extra boolean > > variable and use inline_asm trick to force easier for the verifier assembly. > > > > Instead teach the verifier to recognize that > > r1 = skb->data; > > r1 += 10; > > r2 = skb->data_end; > > if (r1 > r2) { > > here r1 points beyond packet_end and > > subsequent > > if (r1 > r2) // always evaluates to "true". > > } > > > > Tested-by: Jiri Olsa <jolsa@redhat.com> > > Signed-off-by: Alexei Starovoitov <ast@kernel.org> > > --- > > include/linux/bpf_verifier.h | 2 +- > > kernel/bpf/verifier.c | 129 +++++++++++++++++++++++++++++------ > > 2 files changed, 108 insertions(+), 23 deletions(-) > > > > Thanks, we can remove another set of inline asm logic. Awesome! Please contribute your C examples to selftests when possible. > Acked-by: John Fastabend <john.fastabend@gmail.com> > > > if (pred >= 0) { > > @@ -7517,7 +7601,8 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env, > > */ > > if (!__is_pointer_value(false, dst_reg)) > > err = mark_chain_precision(env, insn->dst_reg); > > - if (BPF_SRC(insn->code) == BPF_X && !err) > > + if (BPF_SRC(insn->code) == BPF_X && !err && > > + !__is_pointer_value(false, src_reg)) > > This could have been more specific with !type_is_pkt_pointer() correct? I > think its fine as is though. I actually meant to use __is_pointer_value() here for two reasons: 1. to match dst_reg check just few lines above. 2. mark_chain_precision() is for scalars only. If in the future is_*_branch_taken() will support other kinds of pointers the more precise !type_is_pkt_pointer() check would need to be modified. That would be unnecessary code churn. Thanks for the review!
On Fri, Nov 13, 2020 at 12:56:52AM +0100, Daniel Borkmann wrote: > On 11/12/20 8:16 PM, John Fastabend wrote: > > Alexei Starovoitov wrote: > > > From: Alexei Starovoitov <ast@kernel.org> > > > > > > This patch adds the verifier support to recognize inlined branch conditions. > > > The LLVM knows that the branch evaluates to the same value, but the verifier > > > couldn't track it. Hence causing valid programs to be rejected. > > > The potential LLVM workaround: https://reviews.llvm.org/D87428 > > > can have undesired side effects, since LLVM doesn't know that > > > skb->data/data_end are being compared. LLVM has to introduce extra boolean > > > variable and use inline_asm trick to force easier for the verifier assembly. > > > > > > Instead teach the verifier to recognize that > > > r1 = skb->data; > > > r1 += 10; > > > r2 = skb->data_end; > > > if (r1 > r2) { > > > here r1 points beyond packet_end and > > > subsequent > > > if (r1 > r2) // always evaluates to "true". > > > } > > > > > > Tested-by: Jiri Olsa <jolsa@redhat.com> > > > Signed-off-by: Alexei Starovoitov <ast@kernel.org> > > > --- > > > include/linux/bpf_verifier.h | 2 +- > > > kernel/bpf/verifier.c | 129 +++++++++++++++++++++++++++++------ > > > 2 files changed, 108 insertions(+), 23 deletions(-) > > > > > > > Thanks, we can remove another set of inline asm logic. > > > > Acked-by: John Fastabend <john.fastabend@gmail.com> > > > if (pred >= 0) { > > > @@ -7517,7 +7601,8 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env, > > > */ > > > if (!__is_pointer_value(false, dst_reg)) > > > err = mark_chain_precision(env, insn->dst_reg); > > > - if (BPF_SRC(insn->code) == BPF_X && !err) > > > + if (BPF_SRC(insn->code) == BPF_X && !err && > > > + !__is_pointer_value(false, src_reg)) > > > > This could have been more specific with !type_is_pkt_pointer() correct? I > > think its fine as is though. > > > > > err = mark_chain_precision(env, insn->src_reg); > > > if (err) > > > return err; > > Given the reg->range could now be negative, I wonder whether for the regsafe() > pruning logic we should now better add a >=0 sanity check in there before we > attempt to test on rold->range > rcur->range? I thought about it and specifically picked negative range value to keep regsafe() check as-is. The check is this: if (rold->range > rcur->range) return false; rold is the one that was safe in the past. If rold was positive and the current is negative we fail here which is ok. State pruning is conservative. If rold was negative it means the previous state was safe even though that pointer was pointing beyond packet end. So it's ok for rcur->range to be anything. Whether rcur is positive or negative doesn't matter. Everything is still ok. If rold->range == -1 and rcur->range == -2 we fail here. It's minor annoyance. State pruning is tiny bit more conservative than necessary. So I think no extra checks in regsafe() are neeeded. Does it make sense?
Alexei Starovoitov wrote: > On Thu, Nov 12, 2020 at 11:16:11AM -0800, John Fastabend wrote: > > Alexei Starovoitov wrote: > > > From: Alexei Starovoitov <ast@kernel.org> > > > > > > This patch adds the verifier support to recognize inlined branch conditions. > > > The LLVM knows that the branch evaluates to the same value, but the verifier > > > couldn't track it. Hence causing valid programs to be rejected. > > > The potential LLVM workaround: https://reviews.llvm.org/D87428 > > > can have undesired side effects, since LLVM doesn't know that > > > skb->data/data_end are being compared. LLVM has to introduce extra boolean > > > variable and use inline_asm trick to force easier for the verifier assembly. > > > > > > Instead teach the verifier to recognize that > > > r1 = skb->data; > > > r1 += 10; > > > r2 = skb->data_end; > > > if (r1 > r2) { > > > here r1 points beyond packet_end and > > > subsequent > > > if (r1 > r2) // always evaluates to "true". > > > } > > > > > > Tested-by: Jiri Olsa <jolsa@redhat.com> > > > Signed-off-by: Alexei Starovoitov <ast@kernel.org> > > > --- > > > include/linux/bpf_verifier.h | 2 +- > > > kernel/bpf/verifier.c | 129 +++++++++++++++++++++++++++++------ > > > 2 files changed, 108 insertions(+), 23 deletions(-) > > > > > > > Thanks, we can remove another set of inline asm logic. > > Awesome! Please contribute your C examples to selftests when possible. Sure will do, its just some mundane header parsing iirc. > > > Acked-by: John Fastabend <john.fastabend@gmail.com> > > > > > if (pred >= 0) { > > > @@ -7517,7 +7601,8 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env, > > > */ > > > if (!__is_pointer_value(false, dst_reg)) > > > err = mark_chain_precision(env, insn->dst_reg); > > > - if (BPF_SRC(insn->code) == BPF_X && !err) > > > + if (BPF_SRC(insn->code) == BPF_X && !err && > > > + !__is_pointer_value(false, src_reg)) > > > > This could have been more specific with !type_is_pkt_pointer() correct? I > > think its fine as is though. > > I actually meant to use __is_pointer_value() here for two reasons: > 1. to match dst_reg check just few lines above. Agree. > 2. mark_chain_precision() is for scalars only. If in the future > is_*_branch_taken() will support other kinds of pointers the more > precise !type_is_pkt_pointer() check would need to be modified. > That would be unnecessary code churn. Agree.
On 11/13/20 1:09 AM, Alexei Starovoitov wrote: > On Fri, Nov 13, 2020 at 12:56:52AM +0100, Daniel Borkmann wrote: >> On 11/12/20 8:16 PM, John Fastabend wrote: >>> Alexei Starovoitov wrote: >>>> From: Alexei Starovoitov <ast@kernel.org> >>>> >>>> This patch adds the verifier support to recognize inlined branch conditions. >>>> The LLVM knows that the branch evaluates to the same value, but the verifier >>>> couldn't track it. Hence causing valid programs to be rejected. >>>> The potential LLVM workaround: https://reviews.llvm.org/D87428 >>>> can have undesired side effects, since LLVM doesn't know that >>>> skb->data/data_end are being compared. LLVM has to introduce extra boolean >>>> variable and use inline_asm trick to force easier for the verifier assembly. >>>> >>>> Instead teach the verifier to recognize that >>>> r1 = skb->data; >>>> r1 += 10; >>>> r2 = skb->data_end; >>>> if (r1 > r2) { >>>> here r1 points beyond packet_end and >>>> subsequent >>>> if (r1 > r2) // always evaluates to "true". >>>> } >>>> >>>> Tested-by: Jiri Olsa <jolsa@redhat.com> >>>> Signed-off-by: Alexei Starovoitov <ast@kernel.org> >>>> --- >>>> include/linux/bpf_verifier.h | 2 +- >>>> kernel/bpf/verifier.c | 129 +++++++++++++++++++++++++++++------ >>>> 2 files changed, 108 insertions(+), 23 deletions(-) >>> >>> Thanks, we can remove another set of inline asm logic. >>> >>> Acked-by: John Fastabend <john.fastabend@gmail.com> >>>> if (pred >= 0) { >>>> @@ -7517,7 +7601,8 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env, >>>> */ >>>> if (!__is_pointer_value(false, dst_reg)) >>>> err = mark_chain_precision(env, insn->dst_reg); >>>> - if (BPF_SRC(insn->code) == BPF_X && !err) >>>> + if (BPF_SRC(insn->code) == BPF_X && !err && >>>> + !__is_pointer_value(false, src_reg)) >>> >>> This could have been more specific with !type_is_pkt_pointer() correct? I >>> think its fine as is though. >>> >>>> err = mark_chain_precision(env, insn->src_reg); >>>> if (err) >>>> return err; >> >> Given the reg->range could now be negative, I wonder whether for the regsafe() >> pruning logic we should now better add a >=0 sanity check in there before we >> attempt to test on rold->range > rcur->range? > > I thought about it and specifically picked negative range value to keep > regsafe() check as-is. > The check is this: > if (rold->range > rcur->range) > return false; > rold is the one that was safe in the past. > If rold was positive and the current is negative we fail here > which is ok. State pruning is conservative. > > If rold was negative it means the previous state was safe even though that pointer > was pointing beyond packet end. So it's ok for rcur->range to be anything. > Whether rcur is positive or negative doesn't matter. Everything is still ok. > If rold->range == -1 and rcur->range == -2 we fail here. > It's minor annoyance. State pruning is tiny bit more conservative than necessary. > > So I think no extra checks in regsafe() are neeeded. > Does it make sense? Yeah, same conclusion here. We still might want to add more BPF asm based tests on this in general, but either way logic lgtm, so applied, thanks.
diff --git a/include/linux/bpf_verifier.h b/include/linux/bpf_verifier.h index e83ef6f6bf43..306869d4743b 100644 --- a/include/linux/bpf_verifier.h +++ b/include/linux/bpf_verifier.h @@ -45,7 +45,7 @@ struct bpf_reg_state { enum bpf_reg_type type; union { /* valid when type == PTR_TO_PACKET */ - u16 range; + int range; /* valid when type == CONST_PTR_TO_MAP | PTR_TO_MAP_VALUE | * PTR_TO_MAP_VALUE_OR_NULL diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 10da26e55130..7b1f85aa9741 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -2739,7 +2739,9 @@ static int check_packet_access(struct bpf_verifier_env *env, u32 regno, int off, regno); return -EACCES; } - err = __check_mem_access(env, regno, off, size, reg->range, + + err = reg->range < 0 ? -EINVAL : + __check_mem_access(env, regno, off, size, reg->range, zero_size_allowed); if (err) { verbose(env, "R%d offset is outside of the packet\n", regno); @@ -4697,6 +4699,32 @@ static void clear_all_pkt_pointers(struct bpf_verifier_env *env) __clear_all_pkt_pointers(env, vstate->frame[i]); } +enum { + AT_PKT_END = -1, + BEYOND_PKT_END = -2, +}; + +static void mark_pkt_end(struct bpf_verifier_state *vstate, int regn, bool range_open) +{ + struct bpf_func_state *state = vstate->frame[vstate->curframe]; + struct bpf_reg_state *reg = &state->regs[regn]; + + if (reg->type != PTR_TO_PACKET) + /* PTR_TO_PACKET_META is not supported yet */ + return; + + /* The 'reg' is pkt > pkt_end or pkt >= pkt_end. + * How far beyond pkt_end it goes is unknown. + * if (!range_open) it's the case of pkt >= pkt_end + * if (range_open) it's the case of pkt > pkt_end + * hence this pointer is at least 1 byte bigger than pkt_end + */ + if (range_open) + reg->range = BEYOND_PKT_END; + else + reg->range = AT_PKT_END; +} + static void release_reg_references(struct bpf_verifier_env *env, struct bpf_func_state *state, int ref_obj_id) @@ -6708,7 +6736,7 @@ static int check_alu_op(struct bpf_verifier_env *env, struct bpf_insn *insn) static void __find_good_pkt_pointers(struct bpf_func_state *state, struct bpf_reg_state *dst_reg, - enum bpf_reg_type type, u16 new_range) + enum bpf_reg_type type, int new_range) { struct bpf_reg_state *reg; int i; @@ -6733,8 +6761,7 @@ static void find_good_pkt_pointers(struct bpf_verifier_state *vstate, enum bpf_reg_type type, bool range_right_open) { - u16 new_range; - int i; + int new_range, i; if (dst_reg->off < 0 || (dst_reg->off == 0 && range_right_open)) @@ -6985,6 +7012,67 @@ static int is_branch_taken(struct bpf_reg_state *reg, u64 val, u8 opcode, return is_branch64_taken(reg, val, opcode); } +static int flip_opcode(u32 opcode) +{ + /* How can we transform "a <op> b" into "b <op> a"? */ + static const u8 opcode_flip[16] = { + /* these stay the same */ + [BPF_JEQ >> 4] = BPF_JEQ, + [BPF_JNE >> 4] = BPF_JNE, + [BPF_JSET >> 4] = BPF_JSET, + /* these swap "lesser" and "greater" (L and G in the opcodes) */ + [BPF_JGE >> 4] = BPF_JLE, + [BPF_JGT >> 4] = BPF_JLT, + [BPF_JLE >> 4] = BPF_JGE, + [BPF_JLT >> 4] = BPF_JGT, + [BPF_JSGE >> 4] = BPF_JSLE, + [BPF_JSGT >> 4] = BPF_JSLT, + [BPF_JSLE >> 4] = BPF_JSGE, + [BPF_JSLT >> 4] = BPF_JSGT + }; + return opcode_flip[opcode >> 4]; +} + +static int is_pkt_ptr_branch_taken(struct bpf_reg_state *dst_reg, + struct bpf_reg_state *src_reg, + u8 opcode) +{ + struct bpf_reg_state *pkt; + + if (src_reg->type == PTR_TO_PACKET_END) { + pkt = dst_reg; + } else if (dst_reg->type == PTR_TO_PACKET_END) { + pkt = src_reg; + opcode = flip_opcode(opcode); + } else { + return -1; + } + + if (pkt->range >= 0) + return -1; + + switch (opcode) { + case BPF_JLE: + /* pkt <= pkt_end */ + fallthrough; + case BPF_JGT: + /* pkt > pkt_end */ + if (pkt->range == BEYOND_PKT_END) + /* pkt has at last one extra byte beyond pkt_end */ + return opcode == BPF_JGT; + break; + case BPF_JLT: + /* pkt < pkt_end */ + fallthrough; + case BPF_JGE: + /* pkt >= pkt_end */ + if (pkt->range == BEYOND_PKT_END || pkt->range == AT_PKT_END) + return opcode == BPF_JGE; + break; + } + return -1; +} + /* Adjusts the register min/max values in the case that the dst_reg is the * variable register that we are working on, and src_reg is a constant or we're * simply doing a BPF_K check. @@ -7148,23 +7236,7 @@ static void reg_set_min_max_inv(struct bpf_reg_state *true_reg, u64 val, u32 val32, u8 opcode, bool is_jmp32) { - /* How can we transform "a <op> b" into "b <op> a"? */ - static const u8 opcode_flip[16] = { - /* these stay the same */ - [BPF_JEQ >> 4] = BPF_JEQ, - [BPF_JNE >> 4] = BPF_JNE, - [BPF_JSET >> 4] = BPF_JSET, - /* these swap "lesser" and "greater" (L and G in the opcodes) */ - [BPF_JGE >> 4] = BPF_JLE, - [BPF_JGT >> 4] = BPF_JLT, - [BPF_JLE >> 4] = BPF_JGE, - [BPF_JLT >> 4] = BPF_JGT, - [BPF_JSGE >> 4] = BPF_JSLE, - [BPF_JSGT >> 4] = BPF_JSLT, - [BPF_JSLE >> 4] = BPF_JSGE, - [BPF_JSLT >> 4] = BPF_JSGT - }; - opcode = opcode_flip[opcode >> 4]; + opcode = flip_opcode(opcode); /* This uses zero as "not present in table"; luckily the zero opcode, * BPF_JA, can't get here. */ @@ -7346,6 +7418,7 @@ static bool try_match_pkt_pointers(const struct bpf_insn *insn, /* pkt_data' > pkt_end, pkt_meta' > pkt_data */ find_good_pkt_pointers(this_branch, dst_reg, dst_reg->type, false); + mark_pkt_end(other_branch, insn->dst_reg, true); } else if ((dst_reg->type == PTR_TO_PACKET_END && src_reg->type == PTR_TO_PACKET) || (reg_is_init_pkt_pointer(dst_reg, PTR_TO_PACKET) && @@ -7353,6 +7426,7 @@ static bool try_match_pkt_pointers(const struct bpf_insn *insn, /* pkt_end > pkt_data', pkt_data > pkt_meta' */ find_good_pkt_pointers(other_branch, src_reg, src_reg->type, true); + mark_pkt_end(this_branch, insn->src_reg, false); } else { return false; } @@ -7365,6 +7439,7 @@ static bool try_match_pkt_pointers(const struct bpf_insn *insn, /* pkt_data' < pkt_end, pkt_meta' < pkt_data */ find_good_pkt_pointers(other_branch, dst_reg, dst_reg->type, true); + mark_pkt_end(this_branch, insn->dst_reg, false); } else if ((dst_reg->type == PTR_TO_PACKET_END && src_reg->type == PTR_TO_PACKET) || (reg_is_init_pkt_pointer(dst_reg, PTR_TO_PACKET) && @@ -7372,6 +7447,7 @@ static bool try_match_pkt_pointers(const struct bpf_insn *insn, /* pkt_end < pkt_data', pkt_data > pkt_meta' */ find_good_pkt_pointers(this_branch, src_reg, src_reg->type, false); + mark_pkt_end(other_branch, insn->src_reg, true); } else { return false; } @@ -7384,6 +7460,7 @@ static bool try_match_pkt_pointers(const struct bpf_insn *insn, /* pkt_data' >= pkt_end, pkt_meta' >= pkt_data */ find_good_pkt_pointers(this_branch, dst_reg, dst_reg->type, true); + mark_pkt_end(other_branch, insn->dst_reg, false); } else if ((dst_reg->type == PTR_TO_PACKET_END && src_reg->type == PTR_TO_PACKET) || (reg_is_init_pkt_pointer(dst_reg, PTR_TO_PACKET) && @@ -7391,6 +7468,7 @@ static bool try_match_pkt_pointers(const struct bpf_insn *insn, /* pkt_end >= pkt_data', pkt_data >= pkt_meta' */ find_good_pkt_pointers(other_branch, src_reg, src_reg->type, false); + mark_pkt_end(this_branch, insn->src_reg, true); } else { return false; } @@ -7403,6 +7481,7 @@ static bool try_match_pkt_pointers(const struct bpf_insn *insn, /* pkt_data' <= pkt_end, pkt_meta' <= pkt_data */ find_good_pkt_pointers(other_branch, dst_reg, dst_reg->type, false); + mark_pkt_end(this_branch, insn->dst_reg, true); } else if ((dst_reg->type == PTR_TO_PACKET_END && src_reg->type == PTR_TO_PACKET) || (reg_is_init_pkt_pointer(dst_reg, PTR_TO_PACKET) && @@ -7410,6 +7489,7 @@ static bool try_match_pkt_pointers(const struct bpf_insn *insn, /* pkt_end <= pkt_data', pkt_data <= pkt_meta' */ find_good_pkt_pointers(this_branch, src_reg, src_reg->type, true); + mark_pkt_end(other_branch, insn->src_reg, false); } else { return false; } @@ -7509,6 +7589,10 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env, src_reg->var_off.value, opcode, is_jmp32); + } else if (reg_is_pkt_pointer_any(dst_reg) && + reg_is_pkt_pointer_any(src_reg) && + !is_jmp32) { + pred = is_pkt_ptr_branch_taken(dst_reg, src_reg, opcode); } if (pred >= 0) { @@ -7517,7 +7601,8 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env, */ if (!__is_pointer_value(false, dst_reg)) err = mark_chain_precision(env, insn->dst_reg); - if (BPF_SRC(insn->code) == BPF_X && !err) + if (BPF_SRC(insn->code) == BPF_X && !err && + !__is_pointer_value(false, src_reg)) err = mark_chain_precision(env, insn->src_reg); if (err) return err;