RE: [PATCH] Convert struct pid count to refcount_t

From: David Laight <David.Laight@ACULAB.COM>
To: 'Alan Stern' <stern@rowland.harvard.edu>,
	Oleg Nesterov <oleg@redhat.com>
Cc: "Paul E. McKenney" <paulmck@linux.ibm.com>,
	Jann Horn <jannh@google.com>,
	Joel Fernandes <joel@joelfernandes.org>,
	Kees Cook <keescook@chromium.org>,
	"Eric W. Biederman" <ebiederm@xmission.com>,
	LKML <linux-kernel@vger.kernel.org>,
	Android Kernel Team <kernel-team@android.com>,
	Kernel Hardening <kernel-hardening@lists.openwall.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	Matthew Wilcox <willy@infradead.org>,
	"Michal Hocko" <mhocko@suse.com>,
	"Reshetova, Elena" <elena.reshetova@intel.com>
Subject: RE: [PATCH] Convert struct pid count to refcount_t
Date: Mon, 1 Apr 2019 15:28:35 +0000	[thread overview]
Message-ID: <76f3b894980d4a79a2c0ae815425d6a0@AcuMS.aculab.com> (raw)
In-Reply-To: <Pine.LNX.4.44L0.1903291519060.1497-100000@iolanthe.rowland.org>

From: Alan Stern
> Sent: 29 March 2019 19:45
...
> There is a big difference between WRITE_ONCE() and plain assignment.
> Given "WRITE_ONCE(X, 2)", the compiler will emit a simple store
> instruction.  But given "X = 2", the compiler is allowed to emit
> instructions equivalent to:
> 
> 	if (X != 2)
> 		X = 2;

Worse for you, it can also emit:
	X = 0;
	X = 2;

Many years ago I fell foul of a compiler (not C) that implemented
a write to a 2 bit wide bitfield as:
	X &= ~3
	X |= value
even when 'value' was a compile time constant of 3.
Took a while to find out why the linked list got f*cked.

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)