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]
Message-ID: <c7952df9-5830-45d3-89bb-b45f2b030e24@gmail.com>
Date: Thu, 25 Jul 2024 14:05:34 +0800
From: Leon Hwang <hffilwlqm@...il.com>
To: Yonghong Song <yonghong.song@...ux.dev>, Zheao Li <me@...jusaka.me>,
 Alexei Starovoitov <ast@...nel.org>, Daniel Borkmann <daniel@...earbox.net>,
 John Fastabend <john.fastabend@...il.com>,
 Andrii Nakryiko <andrii@...nel.org>, Martin KaFai Lau
 <martin.lau@...ux.dev>, Eduard Zingerman <eddyz87@...il.com>,
 Song Liu <song@...nel.org>, KP Singh <kpsingh@...nel.org>,
 Stanislav Fomichev <sdf@...ichev.me>, Hao Luo <haoluo@...gle.com>,
 Jiri Olsa <jolsa@...nel.org>
Cc: bpf@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH bpf-next v2] bpf: Add bpf_check_attach_target_with_klog
 method to output failure logs to kernel



On 25/7/24 13:54, Yonghong Song wrote:
> 
> On 7/24/24 10:15 PM, Zheao Li wrote:
>> This is a v2 patch, previous Link:
>> https://lore.kernel.org/bpf/20240724152521.20546-1-me@manjusaka.me/T/#u
>>
>> Compare with v1:
>>
>> 1. Format the code style and signed-off field
>> 2. Use a shorter name bpf_check_attach_target_with_klog instead of
>> original name bpf_check_attach_target_with_kernel_log
>>
>> When attaching a freplace hook, failures can occur,
>> but currently, no output is provided to help developers diagnose the
>> root cause.
>>
>> This commit adds a new method, bpf_check_attach_target_with_klog,
>> which outputs the verifier log to the kernel.
>> Developers can then use dmesg to obtain more detailed information
>> about the failure.
>>
>> For an example of eBPF code,
>> Link:
>> https://github.com/Asphaltt/learn-by-example/blob/main/ebpf/freplace/main.go
>>
>> Co-developed-by: Leon Hwang <hffilwlqm@...il.com>
>> Signed-off-by: Leon Hwang <hffilwlqm@...il.com>
>> Signed-off-by: Zheao Li <me@...jusaka.me>
>> ---
>>   include/linux/bpf_verifier.h |  5 +++++
>>   kernel/bpf/syscall.c         |  5 +++--
>>   kernel/bpf/trampoline.c      |  6 +++---
>>   kernel/bpf/verifier.c        | 19 +++++++++++++++++++
>>   4 files changed, 30 insertions(+), 5 deletions(-)
>>
>> diff --git a/include/linux/bpf_verifier.h b/include/linux/bpf_verifier.h
>> index 5cea15c81b8a..8eddba62c194 100644
>> --- a/include/linux/bpf_verifier.h
>> +++ b/include/linux/bpf_verifier.h
>> @@ -848,6 +848,11 @@ static inline void bpf_trampoline_unpack_key(u64
>> key, u32 *obj_id, u32 *btf_id)
>>           *btf_id = key & 0x7FFFFFFF;
>>   }
>>   +int bpf_check_attach_target_with_klog(const struct bpf_prog *prog,
>> +                        const struct bpf_prog *tgt_prog,
>> +                        u32 btf_id,
>> +                        struct bpf_attach_target_info *tgt_info);
> 
> format issue in the above. Same code alignment is needed for arguments
> in different lines.
> 
>> +
>>   int bpf_check_attach_target(struct bpf_verifier_log *log,
>>                   const struct bpf_prog *prog,
>>                   const struct bpf_prog *tgt_prog,
>> diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
>> index 869265852d51..bf826fcc8cf4 100644
>> --- a/kernel/bpf/syscall.c
>> +++ b/kernel/bpf/syscall.c
>> @@ -3464,8 +3464,9 @@ static int bpf_tracing_prog_attach(struct
>> bpf_prog *prog,
>>            */
>>           struct bpf_attach_target_info tgt_info = {};
>>   -        err = bpf_check_attach_target(NULL, prog, tgt_prog, btf_id,
>> -                          &tgt_info);
>> +        err = bpf_check_attach_target_with_klog(prog, NULL,
>> +                                  prog->aux->attach_btf_id,
>> +                                  &tgt_info);
> 
> code alignment issue here as well.
> Also, the argument should be 'prog, tgt_prog, btf_id, &tgt_info', right?
> 
>>           if (err)
>>               goto out_unlock;
>>   diff --git a/kernel/bpf/trampoline.c b/kernel/bpf/trampoline.c
>> index f8302a5ca400..8862adaa7302 100644
>> --- a/kernel/bpf/trampoline.c
>> +++ b/kernel/bpf/trampoline.c
>> @@ -699,9 +699,9 @@ int bpf_trampoline_link_cgroup_shim(struct
>> bpf_prog *prog,
>>       u64 key;
>>       int err;
>>   -    err = bpf_check_attach_target(NULL, prog, NULL,
>> -                      prog->aux->attach_btf_id,
>> -                      &tgt_info);
>> +    err = bpf_check_attach_target_with_klog(prog, NULL,
>> +                              prog->aux->attach_btf_id,
>> +                              &tgt_info);
> 
> code alignment issue here
> 
>>       if (err)
>>           return err;
>>   diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
>> index 1f5302fb0957..4873b72f5a9a 100644
>> --- a/kernel/bpf/verifier.c
>> +++ b/kernel/bpf/verifier.c
>> @@ -21643,6 +21643,25 @@ static int
>> check_non_sleepable_error_inject(u32 btf_id)
>>       return btf_id_set_contains(&btf_non_sleepable_error_inject,
>> btf_id);
>>   }
>>   +int bpf_check_attach_target_with_klog(const struct bpf_prog *prog,
>> +                        const struct bpf_prog *tgt_prog,
>> +                        u32 btf_id,
>> +                        struct bpf_attach_target_info *tgt_info);
> 
> code alignment issue here.
> 
>> +{
>> +    struct bpf_verifier_log *log;
>> +    int err;
>> +
>> +    log = kzalloc(sizeof(*log), GFP_KERNEL | __GFP_NOWARN);
> 
> __GFP_NOWARN is unnecessary here.
> 
>> +    if (!log) {
>> +        err = -ENOMEM;
>> +        return err;
>> +    }
>> +    log->level = BPF_LOG_KERNEL;
>> +    err = bpf_check_attach_target(log, prog, tgt_prog, btf_id,
>> tgt_info);
>> +    kfree(log);
>> +    return err;
>> +}
>> +
>>   int bpf_check_attach_target(struct bpf_verifier_log *log,
>>                   const struct bpf_prog *prog,
>>                   const struct bpf_prog *tgt_prog,
> 
> More importantly, Andrii has implemented retsnoop, which intends to locate
> precise location in the kernel where err happens. The link is
>   https://github.com/anakryiko/retsnoop
> 
> Maybe you want to take a look and see whether it can resolve your issue.
> We should really avoid putting more stuff in dmesg whenever possible.
> 

retsnoop is really cool.

However, when something wrong in bpf_check_attach_target(), retsnoop
only gets its return value -EINVAL, without any bpf_log() in it. It's
hard to figure out the reason why bpf_check_attach_target() returns -EINVAL.

How about adding a tracepoint in bpf_check_attach_target_with_klog()?
It's to avoid putting stuff in dmesg.

Thanks,
Leon


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ