Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753432AbcDSJS3 (ORCPT ); Tue, 19 Apr 2016 05:18:29 -0400 Received: from mail-ig0-f175.google.com ([209.85.213.175]:35077 "EHLO mail-ig0-f175.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753159AbcDSJS1 convert rfc822-to-8bit (ORCPT ); Tue, 19 Apr 2016 05:18:27 -0400 Date: Tue, 19 Apr 2016 17:18:18 +0800 From: Boqun Feng To: Pan Xinhui Cc: linux-kernel@vger.kernel.org, linuxppc-dev@lists.ozlabs.org, benh@kernel.crashing.org, paulus@samba.org, mpe@ellerman.id.au, peterz@infradead.org, paulmck@linux.vnet.ibm.com, tglx@linutronix.de Subject: Re: [PATCH V2] powerpc: Implement {cmp}xchg for u8 and u16 Message-ID: <20160419091818.GA69264@boquns-mbp.cn.ibm.com> References: <5715D04E.9050009@linux.vnet.ibm.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: 8BIT In-Reply-To: <5715D04E.9050009@linux.vnet.ibm.com> User-Agent: Mutt/1.6.0 (2016-04-01) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 6249 Lines: 208 Hi Xinhui, On Tue, Apr 19, 2016 at 02:29:34PM +0800, Pan Xinhui wrote: > From: Pan Xinhui > > Implement xchg{u8,u16}{local,relaxed}, and > cmpxchg{u8,u16}{,local,acquire,relaxed}. > > It works on all ppc. > Nice work! AFAICT, your work doesn't depend on anything that ppc-specific, right? So maybe we can use it as a general approach for a fallback implementation on the archs without u8/u16 atomics. ;-) > Suggested-by: Peter Zijlstra (Intel) > Signed-off-by: Pan Xinhui > --- > change from V1: > rework totally. > --- > arch/powerpc/include/asm/cmpxchg.h | 83 ++++++++++++++++++++++++++++++++++++++ > 1 file changed, 83 insertions(+) > > diff --git a/arch/powerpc/include/asm/cmpxchg.h b/arch/powerpc/include/asm/cmpxchg.h > index 44efe73..79a1f45 100644 > --- a/arch/powerpc/include/asm/cmpxchg.h > +++ b/arch/powerpc/include/asm/cmpxchg.h > @@ -7,6 +7,37 @@ > #include > #include > > +#ifdef __BIG_ENDIAN > +#define BITOFF_CAL(size, off) ((sizeof(u32) - size - off) * BITS_PER_BYTE) > +#else > +#define BITOFF_CAL(size, off) (off * BITS_PER_BYTE) > +#endif > + > +static __always_inline unsigned long > +__cmpxchg_u32_local(volatile unsigned int *p, unsigned long old, > + unsigned long new); > + > +#define __XCHG_GEN(cmp, type, sfx, u32sfx, skip, v) \ > +static __always_inline u32 \ > +__##cmp##xchg_##type##sfx(v void *ptr, u32 old, u32 new) \ > +{ \ > + int size = sizeof (type); \ > + int off = (unsigned long)ptr % sizeof(u32); \ > + volatile u32 *p = ptr - off; \ > + int bitoff = BITOFF_CAL(size, off); \ > + u32 bitmask = ((0x1 << size * BITS_PER_BYTE) - 1) << bitoff; \ > + u32 oldv, newv; \ > + u32 ret; \ > + do { \ > + oldv = READ_ONCE(*p); \ > + ret = (oldv & bitmask) >> bitoff; \ > + if (skip && ret != old) \ > + break; \ > + newv = (oldv & ~bitmask) | (new << bitoff); \ > + } while (__cmpxchg_u32##u32sfx((v void*)p, oldv, newv) != oldv);\ Forgive me if this is too paranoid, but I think we can save the READ_ONCE() in the loop if we change the code into the following, because cmpxchg will return the "new" value, if the cmp part fails. newv = READ_ONCE(*p); do { oldv = newv; ret = (oldv & bitmask) >> bitoff; if (skip && ret != old) break; newv = (oldv & ~bitmask) | (new << bitoff); newv = __cmpxchg_u32##u32sfx((void *)p, oldv, newv); } while(newv != oldv); > + return ret; \ > +} > + > /* > * Atomic exchange > * > @@ -14,6 +45,19 @@ > * the previous value stored there. > */ > > +#define XCHG_GEN(type, sfx, v) \ > + __XCHG_GEN(_, type, sfx, _local, 0, v) \ ^^^^^^^ This should be sfx, right? Otherwise, all the newly added xchg will call __cmpxchg_u32_local, this will result in wrong ordering guarantees. > +static __always_inline u32 __xchg_##type##sfx(v void *p, u32 n) \ > +{ \ > + return ___xchg_##type##sfx(p, 0, n); \ > +} > + > +XCHG_GEN(u8, _local, volatile); I don't think we need the "volatile" modifier here, because READ_ONCE() and __cmpxchg_u32_* all have "volatile" semantics IIUC, so maybe we can save a paramter for the __XCHG_GEN macro. Regards, Boqun > +XCHG_GEN(u8, _relaxed, ); > +XCHG_GEN(u16, _local, volatile); > +XCHG_GEN(u16, _relaxed, ); > +#undef XCHG_GEN > + > static __always_inline unsigned long > __xchg_u32_local(volatile void *p, unsigned long val) > { > @@ -88,6 +132,10 @@ static __always_inline unsigned long > __xchg_local(volatile void *ptr, unsigned long x, unsigned int size) > { > switch (size) { > + case 1: > + return __xchg_u8_local(ptr, x); > + case 2: > + return __xchg_u16_local(ptr, x); > case 4: > return __xchg_u32_local(ptr, x); > #ifdef CONFIG_PPC64 > @@ -103,6 +151,10 @@ static __always_inline unsigned long > __xchg_relaxed(void *ptr, unsigned long x, unsigned int size) > { > switch (size) { > + case 1: > + return __xchg_u8_relaxed(ptr, x); > + case 2: > + return __xchg_u16_relaxed(ptr, x); > case 4: > return __xchg_u32_relaxed(ptr, x); > #ifdef CONFIG_PPC64 > @@ -226,6 +278,21 @@ __cmpxchg_u32_acquire(u32 *p, unsigned long old, unsigned long new) > return prev; > } > > + > +#define CMPXCHG_GEN(type, sfx, v) \ > + __XCHG_GEN(cmp, type, sfx, sfx, 1, v) > + > +CMPXCHG_GEN(u8, , volatile); > +CMPXCHG_GEN(u8, _local, volatile); > +CMPXCHG_GEN(u8, _relaxed, ); > +CMPXCHG_GEN(u8, _acquire, ); > +CMPXCHG_GEN(u16, , volatile); > +CMPXCHG_GEN(u16, _local, volatile); > +CMPXCHG_GEN(u16, _relaxed, ); > +CMPXCHG_GEN(u16, _acquire, ); > +#undef CMPXCHG_GEN > +#undef __XCHG_GEN > + > #ifdef CONFIG_PPC64 > static __always_inline unsigned long > __cmpxchg_u64(volatile unsigned long *p, unsigned long old, unsigned long new) > @@ -316,6 +383,10 @@ __cmpxchg(volatile void *ptr, unsigned long old, unsigned long new, > unsigned int size) > { > switch (size) { > + case 1: > + return __cmpxchg_u8(ptr, old, new); > + case 2: > + return __cmpxchg_u16(ptr, old, new); > case 4: > return __cmpxchg_u32(ptr, old, new); > #ifdef CONFIG_PPC64 > @@ -332,6 +403,10 @@ __cmpxchg_local(volatile void *ptr, unsigned long old, unsigned long new, > unsigned int size) > { > switch (size) { > + case 1: > + return __cmpxchg_u8_local(ptr, old, new); > + case 2: > + return __cmpxchg_u16_local(ptr, old, new); > case 4: > return __cmpxchg_u32_local(ptr, old, new); > #ifdef CONFIG_PPC64 > @@ -348,6 +423,10 @@ __cmpxchg_relaxed(void *ptr, unsigned long old, unsigned long new, > unsigned int size) > { > switch (size) { > + case 1: > + return __cmpxchg_u8_relaxed(ptr, old, new); > + case 2: > + return __cmpxchg_u16_relaxed(ptr, old, new); > case 4: > return __cmpxchg_u32_relaxed(ptr, old, new); > #ifdef CONFIG_PPC64 > @@ -364,6 +443,10 @@ __cmpxchg_acquire(void *ptr, unsigned long old, unsigned long new, > unsigned int size) > { > switch (size) { > + case 1: > + return __cmpxchg_u8_acquire(ptr, old, new); > + case 2: > + return __cmpxchg_u16_acquire(ptr, old, new); > case 4: > return __cmpxchg_u32_acquire(ptr, old, new); > #ifdef CONFIG_PPC64 > -- > 2.4.3 >