Re: [PATCH hazptr v2 2/4] ptreq.h: Introduce ptr_eq() to preserve address dependency

[Mathieu Desnoyers]

On 2026-10-09 17:55, Gary Guo wrote:
> On Fri Oct 9, 2026 at 7:21 PM BST, Mathieu Desnoyers wrote:
> > Compiler CSE and SSA GVN optimizations can cause the address dependency
> > of addresses returned by rcu_dereference to be lost when comparing those
> > pointers with either constants or previously loaded pointers.
> >
> > Introduce ptr_eq() to compare two addresses while preserving the address
> > dependencies for later use of the address. It should be used when
> > comparing an address returned by rcu_dereference().
> >
> > This is needed to prevent the compiler CSE and SSA GVN optimizations
> > from using @a (or @b) in places where the source refers to @b (or @a)
> > based on the fact that after the comparison, the two are known to be
> > equal, which does not preserve address dependencies and allows the
> > following misordering speculations:
> >
> > - If @b is a constant, the compiler can issue the loads which depend
> >    on @a before loading @a.
> > - If @b is a register populated by a prior load, weakly-ordered
> >    CPUs can speculate loads which depend on @a before loading @a.
> >
> > The same logic applies with @a and @b swapped.
> >
> > The header "linux/ptreq.h" is the expected include target. It
> > contains the documentation of the ptr_eq() API.
> >
> > An architecture specific implementation of the pointer comparison
> > can be implemented by each architecture as asm/ptreq.h. The x86
> > implementation is provided initially. If no implementation header is
> > present for the architecture, an arch-agnostic fallback based on
> > OPTIMIZER_HIDE_VAR is included from asm-generic.
> >
> > Suggested-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
> > Suggested-by: Boqun Feng <boqun.feng@xxxxxxxxx>
> > Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@xxxxxxxxxxxx>
> > Cc: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> > Cc: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
> > Cc: "Paul E. McKenney" <paulmck@xxxxxxxxxx>
> > Cc: Will Deacon <will@xxxxxxxxxx>
> > Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> > Cc: Boqun Feng <boqun.feng@xxxxxxxxx>
> > Cc: Alan Stern <stern@xxxxxxxxxxxxxxxxxxx>
> > Cc: John Stultz <jstultz@xxxxxxxxxx>
> > Cc: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
> > Cc: Boqun Feng <boqun.feng@xxxxxxxxx>
> > Cc: Frederic Weisbecker <frederic@xxxxxxxxxx>
> > Cc: Joel Fernandes <joel@xxxxxxxxxxxxxxxxx>
> > Cc: Josh Triplett <josh@xxxxxxxxxxxxxxxx>
> > Cc: Uladzislau Rezki <urezki@xxxxxxxxx>
> > Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>
> > Cc: Lai Jiangshan <jiangshanlai@xxxxxxxxx>
> > Cc: Zqiang <qiang.zhang1211@xxxxxxxxx>
> > Cc: Ingo Molnar <mingo@xxxxxxxxxx>
> > Cc: Waiman Long <longman@xxxxxxxxxx>
> > Cc: Mark Rutland <mark.rutland@xxxxxxx>
> > Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> > Cc: Vlastimil Babka <vbabka@xxxxxxx>
> > Cc: maged.michael@xxxxxxxxx
> > Cc: Mateusz Guzik <mjguzik@xxxxxxxxx>
> > Cc: Gary Guo <gary@xxxxxxxxxxx>
> > Cc: rcu@xxxxxxxxxxxxxxx
> > Cc: linux-mm@xxxxxxxxx
> > Cc: lkmm@xxxxxxxxxxxxxxx
> > Cc: Nikita Popov <github@xxxxxxxxxx>
> > Cc: llvm@xxxxxxxxxxxxxxx
> > Cc: Lian Wang <lianux.mm@xxxxxxxxx>
> > Cc: Kunwu Chan <kunwu.chan@xxxxxxxxx>
> > ---
> > Changes since v1:
> > - Move to linux/ptreq.h, implement x86-specific comparison
> >    and asm-generic header (fallback).
> >
> > Changes since v0:
> > - Include feedback from Alan Stern.
> > ---
> >   arch/x86/include/asm/ptreq.h | 16 +++++++++
> >   include/asm-generic/Kbuild   |  1 +
> >   include/asm-generic/ptreq.h  | 16 +++++++++
>
> I agree with Boqun that this should be ptr_eq. There is no need to save that
> character :)
>
> >   include/linux/ptreq.h        | 63 ++++++++++++++++++++++++++++++++++++
> >   4 files changed, 96 insertions(+)
> >   create mode 100644 arch/x86/include/asm/ptreq.h
> >   create mode 100644 include/asm-generic/ptreq.h
> >   create mode 100644 include/linux/ptreq.h
> >
> > diff --git a/arch/x86/include/asm/ptreq.h b/arch/x86/include/asm/ptreq.h
> > new file mode 100644
> > index 000000000000..593b45f80451
> > --- /dev/null
> > +++ b/arch/x86/include/asm/ptreq.h
> > @@ -0,0 +1,16 @@
> > +/* SPDX-License-Identifier: GPL-2.0 */
> > +#ifndef _ASM_X86_PTREQ_H
> > +#define _ASM_X86_PTREQ_H
> > +
> > +#include <linux/types.h>
> > +#include <asm/asm.h>
> > +
> > +static __always_inline
> > +bool ptr_eq(const volatile void *a, const volatile void *b)
> > +{
> > +	bool ret;
> > +	asm(__ASM_SIZE(cmp) " %1,%2" : "=@ccz" (ret) : "r" (a), "r" (b));
> > +	return ret;
> > +}
> > +
> > +#endif	/* _ASM_X86_PTREQ_H */
> > diff --git a/include/asm-generic/Kbuild b/include/asm-generic/Kbuild
> > index 2bc00c67dc54..e1d95dea09d5 100644
> > --- a/include/asm-generic/Kbuild
> > +++ b/include/asm-generic/Kbuild
> > @@ -47,6 +47,7 @@ mandatory-y += percpu.h
> >   mandatory-y += percpu_types.h
> >   mandatory-y += pgalloc.h
> >   mandatory-y += preempt.h
> > +mandatory-y += ptreq.h
> >   mandatory-y += rqspinlock.h
> >   mandatory-y += runtime-const.h
> >   mandatory-y += rwonce.h
> > diff --git a/include/asm-generic/ptreq.h b/include/asm-generic/ptreq.h
> > new file mode 100644
> > index 000000000000..db5a88628b1e
> > --- /dev/null
> > +++ b/include/asm-generic/ptreq.h
> > @@ -0,0 +1,16 @@
> > +/* SPDX-License-Identifier: GPL-2.0 */
> > +#ifndef __ASM_GENERIC_PTREQ_H
> > +#define __ASM_GENERIC_PTREQ_H
> > +
> > +#include <linux/types.h>
> > +#include <linux/compiler.h>
> > +
> > +static __always_inline
> > +bool ptr_eq(const volatile void *a, const volatile void *b)
>
> I suppose this is `const volatile` to take advantage of implicit conversion that
> adds cv-qualifiers to be added implicitly. Might worth a mention in somewhere
> (commit message maybe) because I was confused for a second on why this needs
> to be volatile.
>
> > +{
> > +	OPTIMIZER_HIDE_VAR(a);
> > +	OPTIMIZER_HIDE_VAR(b);
> > +	return a == b;
> > +}
>
> One thought on inline assembly. For RISC-V, branching with comparison is a
> single instruction, but there's no instruction for obtaining the boolean for
> a == b.
>
> One option is to do
>
>      static __always_inline
>      bool ptr_eq_likely(const volatile void *a, const volatile void *b) {
>          asm goto ("bne %0, %1, %l2" : : "r" (a), "r" (b) :: not_eq);
>          return true;
>      not_eq:
>          return likely(false);
>      }
>
> and then the optimizer can turn a ptr_eq call + a branch to a single
> instruction. This, however, would require likely + unlikely variants.
>
> Alternatively, we can just not have all the complexity, and just emit one extra
> instruction:
>
>      static __always_inline
>      bool ptr_eq(const volatile void *a, const volatile void *b) {
>          unsigned long diff;
>          asm("xor %0, %1, %2" : "=r" (diff) : "r" (a), "r" (b));
>          return diff == 0;
>      }
>
> Thoughts?

If architectures end up needing the likely/unlikely variants we can
do that if it saves instructions on the hot path. How about we do this
change when we introduce optimized support for riscv ?

Thanks,

Mathieu


--
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com