Received-SPF: pass (google.com: domain of linux-kernel+bounces-207377-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
From: Thomas Gleixner <tglx@linutronix.de>
To: Steven Rostedt <rostedt@goodmis.org>, Sam Sun <samsun1006219@gmail.com>
Cc: linux-kernel@vger.kernel.org, x86@kernel.org,
 syzkaller-bugs@googlegroups.com, peterz@infradead.org,
 jpoimboe@kernel.org, jbaron@akamai.com, ardb@kernel.org, mingo@redhat.com,
 Borislav Petkov <bp@alien8.de>, dave.hansen@linux.intel.com,
 hpa@zytor.com, xrivendell7@gmail.com, Greg Kroah-Hartman
 <gregkh@linuxfoundation.org>, Tejun Heo <tj@kernel.org>
Subject: Re: [Linux kernel bug] WARNING in static_key_slow_inc_cpuslocked
In-Reply-To: <20240609090431.3af238bc@gandalf.local.home>
Date: Sun, 09 Jun 2024 16:06:05 +0200
Message-ID: <87o78axlbm.ffs@tglx>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain

On Sun, Jun 09 2024 at 09:04, Steven Rostedt wrote:
> On Sun, 9 Jun 2024 14:33:01 +0800
> Sam Sun <samsun1006219@gmail.com> wrote:
>> [   82.310798][ T8020] ------------[ cut here ]------------
>> [   82.311236][ T8020] kernel BUG at arch/x86/kernel/jump_label.c:73!
>
> This is not a bug with jump labels. It's a bug with whatever is using jump
> labels. Looks like something tried to modify a jump label that no longer
> exists.

The jump label exists.

>> [   82.331873][ T8020]  set_attr_rdpmc+0x193/0x270
>> [   82.332179][ T8020]  ? get_attr_rdpmc+0x30/0x30
>> [   82.332511][ T8020]  ? sysfs_kf_write+0x18d/0x2b0
>> [   82.332832][ T8020]  ? sysfs_kf_read+0x370/0x370
>> [   82.333159][ T8020]  kernfs_fop_write_iter+0x3ab/0x500
>
> So, something in kernfs modified a jump label location that was freed?

No. What happens is:

CPU 0                           	CPU 1

kernfs_fop_write_iter()			kernfs_fop_write_iter()
  set_attr_rdpmc()		  	  set_attr_rdpmc()
    arch_jump_label_transform_queue()       arch_jump_label_transform_queue()
     mutex_lock(text_mutex)                   mutex_lock(text_mutex)
     __jump_label_patch()
     text_poke_queue()
     mutex_unlokc(text_mutex)
                                              __jump_label_patch()

CPU 1 sees the original text and not the expected because CPU 0 did not
yet invoke arch_jump_label_transform_apply().

So clearly set_attr_rdpmc() lacks serialization, no?

Thanks,

        tglx