* [PATCH] x86_64: Add memcpy32_toio to write to PCI MMIO
@ 2014-09-08 8:14 Subhransu S. Prusty
2014-09-09 8:01 ` Clemens Ladisch
0 siblings, 1 reply; 3+ messages in thread
From: Subhransu S. Prusty @ 2014-09-08 8:14 UTC (permalink / raw)
To: hpa; +Cc: tglx, mingo, x86, linux-kernel, vinod.koul, Subhransu S. Prusty
This is needed because the hardware does not support 64-bit moveq
insructions while writing to PCI MMIO.
Signed-off-by: Subhransu S. Prusty <subhransu.s.prusty@intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
---
arch/x86/include/asm/io.h | 22 ++++++++++++++++++++++
1 file changed, 22 insertions(+)
diff --git a/arch/x86/include/asm/io.h b/arch/x86/include/asm/io.h
index b8237d8..fa095fb 100644
--- a/arch/x86/include/asm/io.h
+++ b/arch/x86/include/asm/io.h
@@ -218,6 +218,28 @@ memcpy_toio(volatile void __iomem *dst, const void *src, size_t count)
memcpy((void __force *)dst, src, count);
}
+#ifndef CONFIG_X86_64
+#define MEMCPY_TOIO memcpy_toio
+#else
+#define MEMCPY_TOIO memcpy32_toio
+#endif
+
+/**
+ * memcpy32_toio: Copy using writel commands
+ *
+ * This is needed because the hardware does not support
+ * 64-bit moveq insructions while writing to PCI MMIO
+ */
+static inline void memcpy32_toio(void *dst, const void *src, int count)
+{
+ int i;
+ const u32 *src_32 = src;
+ u32 *dst_32 = dst;
+
+ for (i = 0; i < count/sizeof(u32); i++)
+ writel(*src_32++, dst_32++);
+}
+
/*
* ISA space is 'always mapped' on a typical x86 system, no need to
* explicitly ioremap() it. The fact that the ISA IO space is mapped
--
1.9.0
^ permalink raw reply related [flat|nested] 3+ messages in thread* Re: [PATCH] x86_64: Add memcpy32_toio to write to PCI MMIO
2014-09-08 8:14 [PATCH] x86_64: Add memcpy32_toio to write to PCI MMIO Subhransu S. Prusty
@ 2014-09-09 8:01 ` Clemens Ladisch
2014-09-16 15:29 ` Vinod Koul
0 siblings, 1 reply; 3+ messages in thread
From: Clemens Ladisch @ 2014-09-09 8:01 UTC (permalink / raw)
To: Subhransu S. Prusty; +Cc: hpa, tglx, mingo, x86, linux-kernel, vinod.koul
Subhransu S. Prusty wrote:
> This is needed because the hardware
Which hardware? Every x86-64 CPU ever built by AMD, Intel, and VIA?
> does not support 64-bit moveq insructions while writing to PCI MMIO.
> +#ifndef CONFIG_X86_64
> +#define MEMCPY_TOIO memcpy_toio
> +#else
> +#define MEMCPY_TOIO memcpy32_toio
> +#endif
This does not change any code that uses memcpy_toio().
> + for (i = 0; i < count/sizeof(u32); i++)
> + writel(*src_32++, dst_32++);
This breaks when count is not a multiple of four.
Regards,
Clemens
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] x86_64: Add memcpy32_toio to write to PCI MMIO
2014-09-09 8:01 ` Clemens Ladisch
@ 2014-09-16 15:29 ` Vinod Koul
0 siblings, 0 replies; 3+ messages in thread
From: Vinod Koul @ 2014-09-16 15:29 UTC (permalink / raw)
To: Clemens Ladisch; +Cc: Subhransu S. Prusty, hpa, tglx, mingo, x86, linux-kernel
On Tue, Sep 09, 2014 at 10:01:18AM +0200, Clemens Ladisch wrote:
> Subhransu S. Prusty wrote:
> > This is needed because the hardware
>
> Which hardware? Every x86-64 CPU ever built by AMD, Intel, and VIA?
Baytrail and few other SoCs
>
> > does not support 64-bit moveq insructions while writing to PCI MMIO.
>
> > +#ifndef CONFIG_X86_64
> > +#define MEMCPY_TOIO memcpy_toio
> > +#else
> > +#define MEMCPY_TOIO memcpy32_toio
> > +#endif
>
> This does not change any code that uses memcpy_toio().
So the devices cant do 64bit memcpy_toio so we are forced to use 32bit
copies only. This given us such an option
>
> > + for (i = 0; i < count/sizeof(u32); i++)
> > + writel(*src_32++, dst_32++);
>
> This breaks when count is not a multiple of four.
Ah yes, we can fix that
--
~Vinod
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2014-09-16 15:54 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2014-09-08 8:14 [PATCH] x86_64: Add memcpy32_toio to write to PCI MMIO Subhransu S. Prusty
2014-09-09 8:01 ` Clemens Ladisch
2014-09-16 15:29 ` Vinod Koul
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox