linux-kernel - Re: [PATCH] tracing: Fix tracing_marker may trigger page fault during preempt

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] [day] [month] [year] [list]

Message-ID: <aLLQ-43Ll8rF7kon@arm.com>
Date: Sat, 30 Aug 2025 11:22:51 +0100
From: Catalin Marinas <catalin.marinas@....com>
To: Steven Rostedt <rostedt@...dmis.org>
Cc: Luo Gengkun <luogengkun@...weicloud.com>, mhiramat@...nel.org,
	mathieu.desnoyers@...icios.com, linux-kernel@...r.kernel.org,
	linux-trace-kernel@...r.kernel.org, Will Deacon <will@...nel.org>,
	linux-arm-kernel@...ts.infradead.org,
	Mark Rutland <mark.rutland@....com>,
	Al Viro <viro@...iv.linux.org.uk>
Subject: Re: [PATCH] tracing: Fix tracing_marker may trigger page fault
 during preempt_disable

On Fri, Aug 29, 2025 at 06:13:11PM -0400, Steven Rostedt wrote:
> On Fri, 29 Aug 2025 20:53:40 +0100
> Catalin Marinas <catalin.marinas@....com> wrote:
> valid user address.
> > BTW, arm64 also bails out early in do_page_fault() if in_atomic() but I
> > suspect that's not the case here.
> > 
> > Adding Al Viro since since he wrote a large part of uaccess.h.
> 
> So, __copy_from_user_inatomic() is supposed to be called if
> pagefault_disable() has already been called? If this is the case, can we
> add more comments to this code? I've been using the inatomic() version this
> way in preempt disabled locations since 2016.

This should work as long as in_atomic() returns true as it's checked in
the arm64 fault code. With PREEMPT_NONE, however, I don't think this
works. __copy_from_user_inatomic() could be changed to call
pagefault_disable() if !in_atomic() but you might as well call
copy_from_user_nofault() in the trace code directly as per Luo's patch.

> I just wanted to figure out why __copy_from_user_inatomic() wasn't atomic.
> If anything, it needs to be better documented.

Yeah, I had no idea until I looked at the code. I guess it means it can
be safely used if in_atomic() == true (well, making it up, not sure what
the intention was).

-- 
Catalin