linux/include/asm-generic/io-64-nonatomic-hi-lo.h
Jason Baron 3a044178cc readq/writeq: Add explicit lo_hi_[read|write]_q and hi_lo_[read|write]_q
Even on x86-64, I've found the need to break up a readq() into 2 readl()
calls. According to the Intel datasheet for the E3-1200 processor:

"
Software must not access B0/D0/F0 32-bit memory-mapped registers with
requests that cross a DW boundary.
"

(http://www.intel.com/content/www/us/en/processors/xeon/xeon-e3-1200-family-vol-2-datasheet.html p. 16)

I can confirm this is true via several hard machine lockups.

Thus, add explicit hi_lo_[readq|write]_q and lo_hi_[read|write]_q so that these
uses are spelled out.

Signed-off-by: Jason Baron <jbaron@akamai.com>
Link: http://lkml.kernel.org/r/281f09da7ad01e5cea99737ec34d2399bdbbbf63.1403818526.git.jbaron@akamai.com
Signed-off-by: Borislav Petkov <bp@suse.de>
2014-07-04 13:27:30 +02:00

33 lines
604 B
C

#ifndef _ASM_IO_64_NONATOMIC_HI_LO_H_
#define _ASM_IO_64_NONATOMIC_HI_LO_H_
#include <linux/io.h>
#include <asm-generic/int-ll64.h>
static inline __u64 hi_lo_readq(const volatile void __iomem *addr)
{
const volatile u32 __iomem *p = addr;
u32 low, high;
high = readl(p + 1);
low = readl(p);
return low + ((u64)high << 32);
}
static inline void hi_lo_writeq(__u64 val, volatile void __iomem *addr)
{
writel(val >> 32, addr + 4);
writel(val, addr);
}
#ifndef readq
#define readq hi_lo_readq
#endif
#ifndef writeq
#define writeq hi_lo_writeq
#endif
#endif /* _ASM_IO_64_NONATOMIC_HI_LO_H_ */