Re: [PATCH RFC] riscv/barrier: Define __smp_{mb,rmb,wmb}

From: Andrea Parri
Date: Mon Feb 26 2018 - 21:27:58 EST


On Mon, Feb 26, 2018 at 05:28:53PM -0800, Palmer Dabbelt wrote:
> On Mon, 26 Feb 2018 02:35:52 PST (-0800), parri.andrea@xxxxxxxxx wrote:
> >On Thu, Feb 22, 2018 at 03:14:52PM -0800, Palmer Dabbelt wrote:
> >>On Tue, 20 Feb 2018 02:17:28 PST (-0800), parri.andrea@xxxxxxxxx wrote:
> >>>Introduce __smp_{mb,rmb,wmb}, and rely on the generic definitions
> >>>for smp_{mb,rmb,wmb}. A first consequence is that smp_{mb,rmb,wmb}
> >>>map to a compiler barrier on !SMP (while their definition remains
> >>>unchanged on SMP). As a further consequence, smp_load_acquire and
> >>>smp_store_release have "fence rw,rw" instead of "fence iorw,iorw".
> >>>
> >>>Signed-off-by: Andrea Parri <parri.andrea@xxxxxxxxx>
> >>>---
> >>> arch/riscv/include/asm/barrier.h | 6 +++---
> >>> 1 file changed, 3 insertions(+), 3 deletions(-)
> >>>
> >>>diff --git a/arch/riscv/include/asm/barrier.h b/arch/riscv/include/asm/barrier.h
> >>>index c0319cbf1eec5..5510366d169ae 100644
> >>>--- a/arch/riscv/include/asm/barrier.h
> >>>+++ b/arch/riscv/include/asm/barrier.h
> >>>@@ -34,9 +34,9 @@
> >>> #define wmb() RISCV_FENCE(ow,ow)
> >>>
> >>> /* These barriers do not need to enforce ordering on devices, just memory. */
> >>>-#define smp_mb() RISCV_FENCE(rw,rw)
> >>>-#define smp_rmb() RISCV_FENCE(r,r)
> >>>-#define smp_wmb() RISCV_FENCE(w,w)
> >>>+#define __smp_mb() RISCV_FENCE(rw,rw)
> >>>+#define __smp_rmb() RISCV_FENCE(r,r)
> >>>+#define __smp_wmb() RISCV_FENCE(w,w)
> >>>
> >>> /*
> >>> * This is a very specific barrier: it's currently only used in two places in
> >>
> >>Thanks! I'm going to take this for the next RC.
> >
> >Thank you, Palmer. I'm planning to post more changes to the file,
> >but I'd like to build on top of this change: could you point me to
> >the appropriate branch/repo for this?
>
> Here's the canonical RISC-V Linux git repo
>
> https://git.kernel.org/pub/scm/linux/kernel/git/palmer/riscv-linux.git/
>
> Your branch is now the HEAD of the "for-linus" branch, which means it'll be
> sent to Linus the next time I send patches. I generate and tag "for-linus"
> on Monday mornings and then send it out on Wednesday mornings, just to make
> sure everything has time to bake.
>
> https://git.kernel.org/pub/scm/linux/kernel/git/palmer/riscv-linux.git/
>
> Additionally, I mantain a "for-next" branch that contains everything that's
> been sufficiently reviewed to be made part of Linux, but that is being
> staged for a bit longer than what's in for-linus for one reason or another
> (usually it's just not RC material and is targeted for the next merge
> window). There is also a RISC-V integration branch named "riscv-all" that
> contains all our work in progress patches. This is likely to be unstable,
> but it's best to check there to see if anything interesting is going on
> related to what you're working on to avoid duplicating work.
>
> These branches are all generated from my personal git tree
>
> https://git.kernel.org/pub/scm/linux/kernel/git/palmer/linux.git/
>
> There's a bunch of branches in here tracking each change set (yours is
> called "fix-smp_mb", to indicate it can go in during an RC) that's still in
> flight. There's some scripts to generate some of these branches, but the
> commits I actually send upstream are merged by hand
>
> https://github.com/riscv/riscv-linux-infra
>
> "for-next" and "riscv-all" are rebased regularly, so it's probably best to
> track commits back to their original WIP branch and work from there to avoid
> major headaches.

Thank you for the info. I've just sent one more patch on top of your
'for-linus' (there appeared to be no conflict with 'riscv-all').

Andrea