From: Peter Zijlstra <peterz@infradead.org>
To: Guo Ren <ren_guo@c-sky.com>
Cc: akpm@linux-foundation.org, arnd@arndb.de,
daniel.lezcano@linaro.org, davem@davemloft.net,
gregkh@linuxfoundation.org, hch@infradead.org,
marc.zyngier@arm.com, mark.rutland@arm.com, robh@kernel.org,
tglx@linutronix.de, linux-kernel@vger.kernel.org,
linux-arch@vger.kernel.org, devicetree@vger.kernel.org,
robh+dt@kernel.org, c-sky_gcc_upstream@c-sky.com,
Andrea Parri <andrea.parri@amarulasolutions.com>
Subject: Re: [PATCH V9 11/21] csky: Atomic operations
Date: Sun, 21 Oct 2018 22:55:08 +0200 [thread overview]
Message-ID: <20181021205508.GJ4931@worktop.programming.kicks-ass.net> (raw)
In-Reply-To: <efb6c9bb87f68c6f579ef497096086413effa5c2.1539655732.git.ren_guo@c-sky.com>
On Tue, Oct 16, 2018 at 10:58:30AM +0800, Guo Ren wrote:
> diff --git a/arch/csky/include/asm/spinlock.h b/arch/csky/include/asm/spinlock.h
> new file mode 100644
> index 0000000..0474603
> --- /dev/null
> +++ b/arch/csky/include/asm/spinlock.h
> @@ -0,0 +1,274 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +
> +#ifndef __ASM_CSKY_SPINLOCK_H
> +#define __ASM_CSKY_SPINLOCK_H
> +
> +#include <linux/spinlock_types.h>
> +#include <asm/barrier.h>
> +
> +#ifdef CONFIG_QUEUED_RWLOCKS
> +
> +/*
> + * Ticket-based spin-locking.
> + */
> +static inline void arch_spin_lock(arch_spinlock_t *lock)
> +{
> + arch_spinlock_t lockval;
> + u32 ticket_next = 1 << TICKET_NEXT;
> + u32 *p = &lock->lock;
> + u32 tmp;
> +
> + asm volatile (
> + "1: ldex.w %0, (%2) \n"
> + " mov %1, %0 \n"
> + " add %0, %3 \n"
> + " stex.w %0, (%2) \n"
> + " bez %0, 1b \n"
> + : "=&r" (tmp), "=&r" (lockval)
> + : "r"(p), "r"(ticket_next)
> + : "cc");
> +
> + while (lockval.tickets.next != lockval.tickets.owner)
> + lockval.tickets.owner = READ_ONCE(lock->tickets.owner);
> +
> + smp_mb();
> +}
> +
> +static inline int arch_spin_trylock(arch_spinlock_t *lock)
> +{
> + u32 tmp, contended, res;
> + u32 ticket_next = 1 << TICKET_NEXT;
> + u32 *p = &lock->lock;
> +
> + do {
> + asm volatile (
> + " ldex.w %0, (%3) \n"
> + " movi %2, 1 \n"
> + " rotli %1, %0, 16 \n"
> + " cmpne %1, %0 \n"
> + " bt 1f \n"
> + " movi %2, 0 \n"
> + " add %0, %0, %4 \n"
> + " stex.w %0, (%3) \n"
> + "1: \n"
> + : "=&r" (res), "=&r" (tmp), "=&r" (contended)
> + : "r"(p), "r"(ticket_next)
> + : "cc");
> + } while (!res);
> +
> + if (!contended)
> + smp_mb();
> +
> + return !contended;
> +}
> +
> +static inline void arch_spin_unlock(arch_spinlock_t *lock)
> +{
> + smp_mb();
> + lock->tickets.owner++;
WRITE_ONCE(lock->tickets.owner, lock->tickets.owner + 1);
> +}
> +
> +static inline int arch_spin_value_unlocked(arch_spinlock_t lock)
> +{
> + return lock.tickets.owner == lock.tickets.next;
> +}
> +
> +static inline int arch_spin_is_locked(arch_spinlock_t *lock)
> +{
> + return !arch_spin_value_unlocked(READ_ONCE(*lock));
> +}
> +
> +static inline int arch_spin_is_contended(arch_spinlock_t *lock)
> +{
> + struct __raw_tickets tickets = READ_ONCE(lock->tickets);
> +
> + return (tickets.next - tickets.owner) > 1;
> +}
> +#define arch_spin_is_contended arch_spin_is_contended
> +
> +#include <asm/qrwlock.h>
> +
> +/* See include/linux/spinlock.h */
> +#define smp_mb__after_spinlock() smp_mb()
> +
> +#else /* CONFIG_QUEUED_RWLOCKS */
> +
> +/*
> + * Test-and-set spin-locking.
> + */
I'm still not entirely sure why you want to have two spinlock
implementations; to me that is just extra maintenance overhead.
> +static inline void arch_spin_lock(arch_spinlock_t *lock)
> +{
> + u32 *p = &lock->lock;
> + u32 tmp;
> +
> + asm volatile (
> + "1: ldex.w %0, (%1) \n"
> + " bnez %0, 1b \n"
> + " movi %0, 1 \n"
> + " stex.w %0, (%1) \n"
> + " bez %0, 1b \n"
> + : "=&r" (tmp)
> + : "r"(p)
> + : "cc");
> + smp_mb();
> +}
> +
> +static inline void arch_spin_unlock(arch_spinlock_t *lock)
> +{
> + u32 *p = &lock->lock;
> + u32 tmp;
> +
> + smp_mb();
> + asm volatile (
> + " movi %0, 0 \n"
> + " stw %0, (%1) \n"
> + : "=&r" (tmp)
> + : "r"(p)
> + : "cc");
WRITE_ONCE(lock->lock, 0);
?
> +}
> +
> +static inline int arch_spin_trylock(arch_spinlock_t *lock)
> +{
> + u32 *p = &lock->lock;
> + u32 tmp;
> +
> + asm volatile (
> + "1: ldex.w %0, (%1) \n"
> + " bnez %0, 2f \n"
> + " movi %0, 1 \n"
> + " stex.w %0, (%1) \n"
> + " bez %0, 1b \n"
> + " movi %0, 0 \n"
> + "2: \n"
> + : "=&r" (tmp)
> + : "r"(p)
> + : "cc");
> +
> + if (!tmp)
> + smp_mb();
> +
> + return !tmp;
> +}
> +
> +#define arch_spin_is_locked(x) (READ_ONCE((x)->lock) != 0)
> +
> +/*
> + * read lock/unlock/trylock
> + */
Idem, why do you want a second rwlock_t implementation?
> +/*
> + * write lock/unlock/trylock
> + */
> +static inline void arch_write_lock(arch_rwlock_t *lock)
> +{
> + u32 *p = &lock->lock;
> + u32 tmp;
> +
> + asm volatile (
> + "1: ldex.w %0, (%1) \n"
> + " bnez %0, 1b \n"
> + " subi %0, 1 \n"
> + " stex.w %0, (%1) \n"
> + " bez %0, 1b \n"
> + : "=&r" (tmp)
> + : "r"(p)
> + : "cc");
> + smp_mb();
> +}
> +
> +static inline void arch_write_unlock(arch_rwlock_t *lock)
> +{
> + u32 *p = &lock->lock;
> + u32 tmp;
> +
> + smp_mb();
> + asm volatile (
> + "1: ldex.w %0, (%1) \n"
> + " movi %0, 0 \n"
> + " stex.w %0, (%1) \n"
> + " bez %0, 1b \n"
> + : "=&r" (tmp)
> + : "r"(p)
> + : "cc");
Isn't that:
WRITE_ONCE(lock->lock, 0);
> +}
> +
> +static inline int arch_write_trylock(arch_rwlock_t *lock)
> +{
> + u32 *p = &lock->lock;
> + u32 tmp;
> +
> + asm volatile (
> + "1: ldex.w %0, (%1) \n"
> + " bnez %0, 2f \n"
> + " subi %0, 1 \n"
> + " stex.w %0, (%1) \n"
> + " bez %0, 1b \n"
> + " movi %0, 0 \n"
> + "2: \n"
> + : "=&r" (tmp)
> + : "r"(p)
> + : "cc");
> +
> + if (!tmp)
> + smp_mb();
> +
> + return !tmp;
> +}
> diff --git a/arch/csky/kernel/atomic.S b/arch/csky/kernel/atomic.S
> new file mode 100644
> index 0000000..d2357c8
> --- /dev/null
> +++ b/arch/csky/kernel/atomic.S
> @@ -0,0 +1,87 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +// Copyright (C) 2018 Hangzhou C-SKY Microsystems co.,ltd.
> +
> +#include <linux/linkage.h>
> +#include <abi/entry.h>
> +
> +.text
> +
> +/*
> + * int csky_cmpxchg(int oldval, int newval, int *ptr)
> + *
> + * If *ptr != oldval && return 1,
> + * else *ptr = newval return 0.
> + */
> +#ifdef CONFIG_CPU_HAS_LDSTEX
> +ENTRY(csky_cmpxchg)
> + USPTOKSP
> + mfcr a3, epc
> + INCTRAP a3
> +
> + subi sp, 8
> + stw a3, (sp, 0)
> + mfcr a3, epsr
> + stw a3, (sp, 4)
> +
> + psrset ee
> +1:
> + ldex a3, (a2)
> + cmpne a0, a3
> + bt16 2f
> + mov a3, a1
> + stex a3, (a2)
> + bez a3, 1b
> +2:
> + sync.is
> + mvc a0
> + ldw a3, (sp, 0)
> + mtcr a3, epc
> + ldw a3, (sp, 4)
> + mtcr a3, epsr
> + addi sp, 8
> + KSPTOUSP
> + rte
> +END(csky_cmpxchg)
I don't understand why you have this; if the CPU has ll/sc, why do you
need syscall support?
In any case, nothing terminally broken; so I suppose that's good enough
for starters. I just really don't understand some decisions (like having
two lock implementations and having that cmpxchg syscall when you have
hardware ll/sc).
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
next prev parent reply other threads:[~2018-10-21 20:55 UTC|newest]
Thread overview: 68+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-10-16 2:58 [PATCH V9 00/21] C-SKY(csky) Linux Kernel Port Guo Ren
2018-10-16 2:58 ` [PATCH V9 01/21] csky: Build infrastructure Guo Ren
2018-10-23 0:08 ` Guo Ren
2018-10-24 22:53 ` Arnd Bergmann
2018-10-25 17:04 ` Guo Ren
2018-10-16 2:58 ` [PATCH V9 02/21] csky: defconfig Guo Ren
2018-10-17 14:56 ` Arnd Bergmann
2018-10-16 2:58 ` [PATCH V9 03/21] csky: Kernel booting Guo Ren
2018-10-17 14:58 ` Arnd Bergmann
2018-10-16 2:58 ` [PATCH V9 04/21] csky: Exception handling and mm-fault Guo Ren
2018-10-17 14:59 ` Arnd Bergmann
2018-10-16 2:58 ` [PATCH V9 05/21] csky: System Call Guo Ren
2018-10-17 15:02 ` Arnd Bergmann
2018-10-18 2:02 ` Guo Ren
2018-10-16 2:58 ` [PATCH V9 06/21] csky: Cache and TLB routines Guo Ren
2018-10-17 15:08 ` Arnd Bergmann
2018-10-16 2:58 ` [PATCH V9 07/21] csky: MMU and page table management Guo Ren
2018-10-17 15:06 ` Arnd Bergmann
2018-10-18 2:05 ` Guo Ren
2018-10-16 2:58 ` [PATCH V9 08/21] csky: Process management and Signal Guo Ren
2018-10-17 15:11 ` Arnd Bergmann
2018-10-18 2:37 ` Guo Ren
2018-10-16 2:58 ` [PATCH V9 09/21] csky: VDSO and rt_sigreturn Guo Ren
2018-10-17 15:13 ` Arnd Bergmann
2018-10-16 2:58 ` [PATCH V9 10/21] csky: IRQ handling Guo Ren
2018-10-17 15:14 ` Arnd Bergmann
2018-10-18 2:39 ` Guo Ren
2018-10-16 2:58 ` [PATCH V9 11/21] csky: Atomic operations Guo Ren
2018-10-17 15:17 ` Arnd Bergmann
2018-10-18 2:40 ` Guo Ren
2018-10-21 20:55 ` Peter Zijlstra [this message]
2018-10-22 1:52 ` Guo Ren
2018-10-16 2:58 ` [PATCH V9 12/21] csky: ELF and module probe Guo Ren
2018-10-17 15:18 ` Arnd Bergmann
2018-10-18 2:49 ` Guo Ren
2018-10-18 8:31 ` Arnd Bergmann
2018-10-16 2:58 ` [PATCH V9 13/21] csky: Library functions Guo Ren
2018-10-17 15:24 ` Arnd Bergmann
2018-10-18 3:10 ` Guo Ren
2018-10-16 2:58 ` [PATCH V9 14/21] csky: User access Guo Ren
2018-10-17 15:37 ` Arnd Bergmann
2018-10-16 2:58 ` [PATCH V9 15/21] csky: Debug and Ptrace GDB Guo Ren
2018-10-17 15:46 ` Arnd Bergmann
2018-10-18 3:17 ` Guo Ren
2018-10-16 2:58 ` [PATCH V9 16/21] csky: SMP support Guo Ren
2018-10-17 15:47 ` Arnd Bergmann
2018-10-16 2:58 ` [PATCH V9 17/21] csky: Misc headers Guo Ren
2018-10-17 15:49 ` Arnd Bergmann
2018-10-16 2:58 ` [PATCH V9 18/21] dt-bindings: csky CPU Bindings Guo Ren
2018-10-17 15:50 ` Arnd Bergmann
2018-10-18 3:21 ` Guo Ren
2018-10-18 3:45 ` Guo Ren
2018-10-18 14:31 ` Rob Herring
2018-10-19 2:19 ` Guo Ren
2018-10-16 2:58 ` [PATCH V9 19/21] dt-bindings: Add vendor prefix for csky Guo Ren
2018-10-16 2:58 ` [PATCH V9 20/21] MAINTAINERS: Add csky Guo Ren
2018-10-17 15:51 ` Arnd Bergmann
2018-10-16 5:48 ` [PATCH V9 21/21] csky: support dword access for get_user_size() Guo Ren
2018-10-17 15:44 ` Arnd Bergmann
2018-10-18 3:41 ` Guo Ren
2018-10-18 8:34 ` Arnd Bergmann
2018-10-18 8:57 ` Guo Ren
2018-10-24 7:17 ` Arnd Bergmann
2018-10-25 17:08 ` Guo Ren
2018-10-17 15:58 ` [PATCH V9 00/21] C-SKY(csky) Linux Kernel Port Arnd Bergmann
2018-10-18 4:10 ` Guo Ren
2018-10-18 8:36 ` Arnd Bergmann
2018-10-18 9:03 ` Guo Ren
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20181021205508.GJ4931@worktop.programming.kicks-ass.net \
--to=peterz@infradead.org \
--cc=akpm@linux-foundation.org \
--cc=andrea.parri@amarulasolutions.com \
--cc=arnd@arndb.de \
--cc=c-sky_gcc_upstream@c-sky.com \
--cc=daniel.lezcano@linaro.org \
--cc=davem@davemloft.net \
--cc=devicetree@vger.kernel.org \
--cc=gregkh@linuxfoundation.org \
--cc=hch@infradead.org \
--cc=linux-arch@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=marc.zyngier@arm.com \
--cc=mark.rutland@arm.com \
--cc=ren_guo@c-sky.com \
--cc=robh+dt@kernel.org \
--cc=robh@kernel.org \
--cc=tglx@linutronix.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).