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  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:	Tue, 16 Nov 2010 19:57:20 -0800
From:	Andrew Morton <>
To:	Huang Ying <>
Cc:	Len Brown <>,
	"" <>,
	Andi Kleen <>,
	"" <>,
	Linus Torvalds <>,
	Thomas Gleixner <>,
	Ingo Molnar <>,
	Mauro Carvalho Chehab <>,
	Peter Zijlstra <>,
	Steven Rostedt <>
Subject: Re: [PATCH -v4 1/2] lib, Make gen_pool memory allocator lockless

On Wed, 17 Nov 2010 11:03:25 +0800 Huang Ying <> wrote:

> It seems that Steven thinks many architectures without NMI-safe cmpxchg
> have no real NMI too.

Could be.

Really, we should nail this down and work out the matrix of
what-works-with-what, and then arrange for the things which _won't_
work to be either non-Kconfigurable or non-compilable.

The worst thing we could do would be to advertise some code as
"nmi-safe!!", then to have someone go and use it on that basis, only
for their users to later discover ghastly rare races.

> In the patch description and comments, it is said that on architectures
> without NMI-safe cmpxchg, gen_pool can not be used in NMI handler
> safely.
> Or do you think it is better to use a spin_trylock based fallback if
> NMI-safe cmpxchg is not available? Or require cmpxchg implementation
> uses spin_trylock instead of spin_lock?

As a first step, a typical thing to do would be to create
CONFIG_ARCH_HAVE_NMI_SAFE_CMPXCHG, define that in the appropriate
architectures, make ftrace and perf and genpool and anything else
dependent upon that at Kconfig-time.

A spin_trylock_irqsave() implementation would do what?  Rarely fail the
memory allocation attempt if the trylock failed?  I guess that's
acceptable in the context of gen_pool, because memory allocators are
expected to fail, and everyone carefully tests the allocation-failed
error paths (lol).

But rare failures may not be useful within the context of future
clients of the "lockless" list implementation so I'd say that a safer
approach would be to make the list implementation require

So that's all pretty simple so far.  However...

The list implementation is still useful to
non-CONFIG_ARCH_HAVE_NMI_SAFE_CMPXCHG clients, as long as they aren't
using it from NMI context, yes?

In which case I suppose we could add a library of lockless_list_foo()
functions which are usable from non-NMI contexts and which are
available to all configs.  And on top of that implement a library of
nmi_safe_lockless_list_foo() functions which are just wrappers around
lockless_list_foo(), but which are only available if

Which is getting to be a bit of a pain, so you might choose to
disregard everything after "However..." ;)
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

Powered by blists - more mailing lists