lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Mon, 28 Nov 2022 20:40:06 +0800
From:   Yang Jihong <yangjihong1@...wei.com>
To:     Alexei Starovoitov <alexei.starovoitov@...il.com>
CC:     <ast@...nel.org>, <daniel@...earbox.net>, <andrii@...nel.org>,
        <martin.lau@...ux.dev>, <song@...nel.org>, <yhs@...com>,
        <john.fastabend@...il.com>, <kpsingh@...nel.org>, <sdf@...gle.com>,
        <haoluo@...gle.com>, <jolsa@...nel.org>,
        <illusionist.neo@...il.com>, <linux@...linux.org.uk>,
        <davem@...emloft.net>, <edumazet@...gle.com>, <kuba@...nel.org>,
        <pabeni@...hat.com>, <mykolal@...com>, <shuah@...nel.org>,
        <benjamin.tissoires@...hat.com>, <memxor@...il.com>,
        <colin.i.king@...il.com>, <asavkov@...hat.com>, <delyank@...com>,
        <bpf@...r.kernel.org>, <linux-arm-kernel@...ts.infradead.org>,
        <linux-kernel@...r.kernel.org>, <netdev@...r.kernel.org>,
        <linux-kselftest@...r.kernel.org>
Subject: Re: [PATCH bpf-next v3 1/4] bpf: Adapt 32-bit return value kfunc for
 32-bit ARM when zext extension



On 2022/11/28 9:57, Alexei Starovoitov wrote:
> On Sat, Nov 26, 2022 at 05:45:27PM +0800, Yang Jihong wrote:
>> For ARM32 architecture, if data width of kfunc return value is 32 bits,
>> need to do explicit zero extension for high 32-bit, insn_def_regno should
>> return dst_reg for BPF_JMP type of BPF_PSEUDO_KFUNC_CALL. Otherwise,
>> opt_subreg_zext_lo32_rnd_hi32 returns -EFAULT, resulting in BPF failure.
>>
>> Signed-off-by: Yang Jihong <yangjihong1@...wei.com>
>> ---
>>   kernel/bpf/verifier.c | 44 ++++++++++++++++++++++++++++++++++++++++---
>>   1 file changed, 41 insertions(+), 3 deletions(-)
>>
>> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
>> index 264b3dc714cc..193ea927aa69 100644
>> --- a/kernel/bpf/verifier.c
>> +++ b/kernel/bpf/verifier.c
>> @@ -1927,6 +1927,21 @@ find_kfunc_desc(const struct bpf_prog *prog, u32 func_id, u16 offset)
>>   		       sizeof(tab->descs[0]), kfunc_desc_cmp_by_id_off);
>>   }
>>   
>> +static int kfunc_desc_cmp_by_imm(const void *a, const void *b);
>> +
>> +static const struct bpf_kfunc_desc *
>> +find_kfunc_desc_by_imm(const struct bpf_prog *prog, s32 imm)
>> +{
>> +	struct bpf_kfunc_desc desc = {
>> +		.imm = imm,
>> +	};
>> +	struct bpf_kfunc_desc_tab *tab;
>> +
>> +	tab = prog->aux->kfunc_tab;
>> +	return bsearch(&desc, tab->descs, tab->nr_descs,
>> +		       sizeof(tab->descs[0]), kfunc_desc_cmp_by_imm);
>> +}
>> +
>>   static struct btf *__find_kfunc_desc_btf(struct bpf_verifier_env *env,
>>   					 s16 offset)
>>   {
>> @@ -2342,6 +2357,13 @@ static bool is_reg64(struct bpf_verifier_env *env, struct bpf_insn *insn,
>>   			 */
>>   			if (insn->src_reg == BPF_PSEUDO_CALL)
>>   				return false;
>> +
>> +			/* Kfunc call will reach here because of insn_has_def32,
>> +			 * conservatively return TRUE.
>> +			 */
>> +			if (insn->src_reg == BPF_PSEUDO_KFUNC_CALL)
>> +				return true;
>> +
>>   			/* Helper call will reach here because of arg type
>>   			 * check, conservatively return TRUE.
>>   			 */
>> @@ -2405,10 +2427,26 @@ static bool is_reg64(struct bpf_verifier_env *env, struct bpf_insn *insn,
>>   }
>>   
>>   /* Return the regno defined by the insn, or -1. */
>> -static int insn_def_regno(const struct bpf_insn *insn)
>> +static int insn_def_regno(struct bpf_verifier_env *env, const struct bpf_insn *insn)
>>   {
>>   	switch (BPF_CLASS(insn->code)) {
>>   	case BPF_JMP:
>> +		if (insn->src_reg == BPF_PSEUDO_KFUNC_CALL) {
>> +			const struct bpf_kfunc_desc *desc;
>> +
>> +			/* The value of desc cannot be NULL */
>> +			desc = find_kfunc_desc_by_imm(env->prog, insn->imm);
>> +
>> +			/* A kfunc can return void.
>> +			 * The btf type of the kfunc's return value needs
>> +			 * to be checked against "void" first
>> +			 */
>> +			if (desc->func_model.ret_size == 0)
>> +				return -1;
>> +			else
>> +				return insn->dst_reg;
>> +		}
>> +		fallthrough;
> 
> I cannot make any sense of this patch.
> insn->dst_reg above is 0.
> The kfunc call doesn't define a register from insn_def_regno() pov.
> 
> Are you hacking insn_def_regno() to return 0 so that
> if (WARN_ON(load_reg == -1)) {
>    verbose(env, "verifier bug. zext_dst is set, but no reg is defined\n");
>    return -EFAULT;
> }
> in opt_subreg_zext_lo32_rnd_hi32() doesn't trigger ?
> 
> But this verifier message should have been a hint that you need
> to analyze why zext_dst is set on this kfunc call.
> Maybe it shouldn't ?
> Did you analyze the logic of mark_btf_func_reg_size() ?
make r0 zext is not caused by mark_btf_func_reg_size.

This problem occurs when running the kfunc_call_test_ref_btf_id test 
case in the 32-bit ARM environment.
The bpf prog is as follows:
int kfunc_call_test_ref_btf_id(struct __sk_buff *skb)
{
struct prog_test_ref_kfunc *pt;
unsigned long s = 0;
int ret = 0;

pt = bpf_kfunc_call_test_acquire(&s);
if (pt) {
      // here, do_check clears the upper 32bits of r0 through:
      // check_alu_op
      //   ->check_reg_arg
      //    ->mark_insn_zext
if (pt->a != 42 || pt->b != 108)
ret = -1;
bpf_kfunc_call_test_release(pt);
}
return ret;
}

> 
> Before producing any patches please understand the logic fully.
> Your commit log
> "insn_def_regno should
>   return dst_reg for BPF_JMP type of BPF_PSEUDO_KFUNC_CALL."
> 
> Makes no sense to me, since dst_reg is unused in JMP insn.
> There is no concept of a src or dst register in a JMP insn.
> 
> 32-bit x86 supports calling kfuncs. See emit_kfunc_call().
> And we don't have this "verifier bug. zext_dst is set" issue there, right?
> But what you're saying in the commit log:
> "if data width of kfunc return value is 32 bits"
> should have been applicable to x86-32 as well.
> So please start with a test that demonstrates the issue on x86-32 and
> then we can discuss the way to fix it.
> 
> The patch 2 sort-of makes sense.
> 
> For patch 3 pls add new test funcs to bpf_testmod.
> We will move all of them from net/bpf/test_run.c to bpf_testmod eventually.
> .
> 

Powered by blists - more mailing lists