linux-kernel - Re: WARNING with LBR + precise_ip=2 + bpf_get

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: <20200618223555.GB793265@tassilo.jf.intel.com>
Date:   Thu, 18 Jun 2020 15:35:55 -0700
From:   Andi Kleen <ak@...ux.intel.com>
To:     Song Liu <songliubraving@...com>
Cc:     linux-kernel <linux-kernel@...r.kernel.org>,
        Kan Liang <kan.liang@...ux.intel.com>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Josh Poimboeuf <jpoimboe@...hat.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Tejun Heo <tj@...nel.org>, Andrii Nakryiko <andriin@...com>,
        Francis Ricci <fjricci@...com>
Subject: Re: WARNING with LBR + precise_ip=2 + bpf_get_stackid()

> We noticed that this only happens with precise_ip >= 2. This is caused by 
> setup_pebs_fixed_sample_data() using pens->real_ip:
> 
> 	set_linear_ip(regs, pebs->real_ip);

With precise_ip == 1 the IP will be one instruction behind.
That's the only difference to 2 or 3.

So something about the actual instruction confuses the unwinder.

I would check if there is anything special about these IPs:

> 0xfffffe00004d1f78 entry_SYSCALL_64
> 0xfffffe00004d1fa0 entry_SYSCALL_64
> 0xfffffe00004d1fd8 entry_SYSCALL_64

> 
> For our use case, we do need precise_ip=2. So we would like suggestions to fix the
> warning and/or to avoid double fault. 

As a minimum, I would just add an && !in_nmi() to the warning. Clearly
it doesn't make sense to print a warning that overflows the stack.

-Andi