diff mbox

[1/7] asm-generic/io.h: add io{read,write}64 accessors

Message ID C2D7FE5348E1B147BCA15975FBA23075F4EA07DD@us01wembx1.internal.synopsys.com
State New
Headers show

Commit Message

Vineet Gupta May 5, 2016, 8:16 a.m. UTC
On Thursday 05 May 2016 04:06 AM, Arnd Bergmann wrote:
> On Wednesday 04 May 2016 20:16:19 Horia Geantă wrote:
>> @@ -625,6 +645,16 @@ static inline u32 ioread32be(const volatile void __iomem *addr)
>>  }
>>  #endif
>>  
>> +#ifdef CONFIG_64BIT
>> +#ifndef ioread64be
>> +#define ioread64be ioread64be
>> +static inline u64 ioread64be(const volatile void __iomem *addr)
>> +{
>> +       return __be64_to_cpu(__raw_readq(addr));
>> +}
>> +#endif
>> +#endif /* CONFIG_64BIT */
>> +
>>  #ifndef iowrite16be
>>  #define iowrite16be iowrite16be
>>  static inline void iowrite16be(u16 value, void volatile __iomem *addr)
>> @@ -641,6 +671,16 @@ static inline void iowrite32be(u32 value, volatile void __iomem *addr)
>>  }
>>  #endif
>>  
>> +#ifdef CONFIG_64BIT
>> +#ifndef iowrite64be
>> +#define iowrite64be iowrite64be
>> +static inline void iowrite64be(u64 value, volatile void __iomem *addr)
>> +{
>> +       __raw_writeq(__cpu_to_be64(value), addr);
>> +}
>> +#endif
>> +#endif /* CONFIG_64BIT */
>> +
>>
> I just noticed that these two are both a bit wrong, but they copy the
> mistake that already exists in the 16 and 32 bit versions: If an
> architecture overrides readq/writeq to have barriers but does not override
> ioread64be/iowrite64be, this will lack the barriers and behave differently
> from the little-endian version. I think the only affected architecture
> is ARC, since ARM and ARM64 both override the big-endian accessors to
> have the correct barriers, and all others don't use barriers at all.
>
> Maybe you can add a patch before this one to replace the 16/32-bit accessors
> with ones that do a
>
> static inline void iowrite32be(u32 value, volatile void __iomem *addr)
> {
> 	writel(swab32(value), addr);
> }
>
> This will lead to a double-swap on architectures that don't override it,
> but it will work correctly on all architectures without them having
> to override the big-endian accessors.

Thx for noticing this Arnd and the heads up. Does the patch below look ok to you ?

----------->
rom b7e719831c389ab4fa338b2e2e7c0d1ff90dabb0 Mon Sep 17 00:00:00 2001
From: Vineet Gupta <vgupta@synopsys.com>
Date: Thu, 5 May 2016 13:32:34 +0530
Subject: [PATCH] ARC: Add missing io barriers to io{read,write}{16,32}be()

While reviewing a different change to asm-generic/io.h Arnd spotted that
ARC ioread32 and ioread32be both of which come from asm-generic versions
are not symmetrical in terms of calling the io barriers.

generic ioread32   -> ARC readl()                  [ has barriers]
generic ioread32be -> __be32_to_cpu(__raw_readl()) [ lacks barriers]

While generic ioread32be is being remediated to call readl(), that involves
a swab32(), causing double swaps on ioread32be() on Big Endian systems.

So provide our versions of big endian IO accessors to ensure io barrier
calls while also keeping them optimal

Suggested-by: Arnd Bergmann <arnd@arndb.de>
Cc: stable@vger.kernel.org  [4.2+]
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
---
 arch/arc/include/asm/io.h | 27 ++++++++++++++++++---------
 1 file changed, 18 insertions(+), 9 deletions(-)

+
 /* Change struct page to physical address */
 #define page_to_phys(page)        (page_to_pfn(page) << PAGE_SHIFT)
 
@@ -108,15 +126,6 @@ static inline void __raw_writel(u32 w, volatile void __iomem
*addr)
 
 }
 
-#ifdef CONFIG_ISA_ARCV2
-#include <asm/barrier.h>
-#define __iormb()        rmb()
-#define __iowmb()        wmb()
-#else
-#define __iormb()        do { } while (0)
-#define __iowmb()        do { } while (0)
-#endif
-
 /*
  * MMIO can also get buffered/optimized in micro-arch, so barriers needed
  * Based on ARM model for the typical use case

Comments

Arnd Bergmann May 5, 2016, 10:56 a.m. UTC | #1
On Thursday 05 May 2016 08:16:47 Vineet Gupta wrote:
> Thx for noticing this Arnd and the heads up. Does the patch below look ok to you ?
> 
> ----------->
> rom b7e719831c389ab4fa338b2e2e7c0d1ff90dabb0 Mon Sep 17 00:00:00 2001
> From: Vineet Gupta <vgupta@synopsys.com>
> Date: Thu, 5 May 2016 13:32:34 +0530
> Subject: [PATCH] ARC: Add missing io barriers to io{read,write}{16,32}be()
> 
> While reviewing a different change to asm-generic/io.h Arnd spotted that
> ARC ioread32 and ioread32be both of which come from asm-generic versions
> are not symmetrical in terms of calling the io barriers.
> 
> generic ioread32   -> ARC readl()                  [ has barriers]
> generic ioread32be -> __be32_to_cpu(__raw_readl()) [ lacks barriers]
> 
> While generic ioread32be is being remediated to call readl(), that involves
> a swab32(), causing double swaps on ioread32be() on Big Endian systems.
> 
> So provide our versions of big endian IO accessors to ensure io barrier
> calls while also keeping them optimal
> 
> Suggested-by: Arnd Bergmann <arnd@arndb.de>
> Cc: stable@vger.kernel.org  [4.2+]
> Signed-off-by: Vineet Gupta <vgupta@synopsys.com>

Yes, that looks correct. We probably want this regardless of the change
I proposed for the generic file, to avoid the double swap.

Acked-by: Arnd Bergmann <arnd@arndb.de>
Vineet Gupta May 5, 2016, 11:05 a.m. UTC | #2
On Thursday 05 May 2016 04:26 PM, Arnd Bergmann wrote:
> On Thursday 05 May 2016 08:16:47 Vineet Gupta wrote:
>> > Thx for noticing this Arnd and the heads up. Does the patch below look ok to you ?
>> > 
>> > ----------->
>> > rom b7e719831c389ab4fa338b2e2e7c0d1ff90dabb0 Mon Sep 17 00:00:00 2001
>> > From: Vineet Gupta <vgupta@synopsys.com>
>> > Date: Thu, 5 May 2016 13:32:34 +0530
>> > Subject: [PATCH] ARC: Add missing io barriers to io{read,write}{16,32}be()
>> > 
>> > While reviewing a different change to asm-generic/io.h Arnd spotted that
>> > ARC ioread32 and ioread32be both of which come from asm-generic versions
>> > are not symmetrical in terms of calling the io barriers.
>> > 
>> > generic ioread32   -> ARC readl()                  [ has barriers]
>> > generic ioread32be -> __be32_to_cpu(__raw_readl()) [ lacks barriers]
>> > 
>> > While generic ioread32be is being remediated to call readl(), that involves
>> > a swab32(), causing double swaps on ioread32be() on Big Endian systems.
>> > 
>> > So provide our versions of big endian IO accessors to ensure io barrier
>> > calls while also keeping them optimal
>> > 
>> > Suggested-by: Arnd Bergmann <arnd@arndb.de>
>> > Cc: stable@vger.kernel.org  [4.2+]
>> > Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
>
> Yes, that looks correct. We probably want this regardless of the change
> I proposed for the generic file, to avoid the double swap.

Indeed so - I've queued this for 4.6 fixes !

> Acked-by: Arnd Bergmann <arnd@arndb.de>

Thx !
diff mbox

Patch

diff --git a/arch/arc/include/asm/io.h b/arch/arc/include/asm/io.h
index 17f85c9c73cf..c22b181e8206 100644
--- a/arch/arc/include/asm/io.h
+++ b/arch/arc/include/asm/io.h
@@ -13,6 +13,15 @@ 
 #include <asm/byteorder.h>
 #include <asm/page.h>
 
+#ifdef CONFIG_ISA_ARCV2
+#include <asm/barrier.h>
+#define __iormb()        rmb()
+#define __iowmb()        wmb()
+#else
+#define __iormb()        do { } while (0)
+#define __iowmb()        do { } while (0)
+#endif
+
 extern void __iomem *ioremap(phys_addr_t paddr, unsigned long size);
 extern void __iomem *ioremap_prot(phys_addr_t paddr, unsigned long size,
                   unsigned long flags);
@@ -31,6 +40,15 @@  extern void iounmap(const void __iomem *addr);
 #define ioremap_wc(phy, sz)        ioremap(phy, sz)
 #define ioremap_wt(phy, sz)        ioremap(phy, sz)
 
+/*
+ * io{read,write}{16,32}be() macros
+ */
+#define ioread16be(p)        ({ u16 __v = be16_to_cpu((__force
__be16)__raw_readw(p)); __iormb(); __v; })
+#define ioread32be(p)        ({ u32 __v = be32_to_cpu((__force
__be32)__raw_readl(p)); __iormb(); __v; })
+
+#define iowrite16be(v,p)    ({ __iowmb(); __raw_writew((__force
u16)cpu_to_be16(v), p); })
+#define iowrite32be(v,p)    ({ __iowmb(); __raw_writel((__force
u32)cpu_to_be32(v), p); })