I have noticed this. And all register accesses in my patch are done through __raw_readl/__raw_writel. 2011/9/26, wu zhangjin <wuzhangjin@xxxxxxxxx>: > On Sun, Sep 25, 2011 at 9:40 PM, Shane McDonald > <mcdonald.shane@xxxxxxxxx> wrote: >>>> diff --git a/arch/mips/include/asm/mach-loongson1/regs-clk.h >>>> b/arch/mips/include/asm/mach-loongson1/regs-clk.h >>>> new file mode 100644 >>>> index 0000000..7a09d6a >>>> --- /dev/null >>>> +++ b/arch/mips/include/asm/mach-loongson1/regs-clk.h >>>> @@ -0,0 +1,32 @@ >>>> +/* >>>> + * Copyright (c) 2011 Zhang, Keguang <keguang.zhang@xxxxxxxxx> >>>> + * >>>> + * Loongson1 Clock Register Definitions. >>>> + * >>>> + * This program is free software; you can redistribute it and/or >>>> modify it >>>> + * under the terms of the GNU General Public License as published by >>>> the >>>> + * Free Software Foundation; either version 2 of the License, or (at >>>> your >>>> + * option) any later version. >>>> + */ >>>> + >>>> +#ifndef __ASM_MACH_LOONGSON1_REGS_CLK_H >>>> +#define __ASM_MACH_LOONGSON1_REGS_CLK_H >>>> + >>>> +#define LS1_CLK_REG(x) ((void __iomem *)(LOONGSON1_CLK_BASE + >>>> (x))) >>> >>> "volatile" keyword may be required for __iomem access, the same to the >>> following similar usage. >>> >>> Considering a scene is(LS1_XXX_REG(X) doesn't really exist): >>> >>> LS1_XXX_REG(X) = 0; /* put cpu into idle and wait interrupt >>> */ >>> LS1_XXX_REG(X) = 7; /* recover the cpu frequency to the >>> highest */ >>> >>> If no "volatile" keyword indicated, the first line will be >>> intelligently but wrongly removed by compiler. >> >> No -- please see Documentation/volatile-considered-harmful.txt, >> particularly the paragraph starting at line 49. This macro >> is only being used as an argument to __raw_readl, >> as it should be. > > Yeah, __raw_readl/writel() will use volatile to prevent it from > optimization, thanks ;) > > "within the kernel, I/O memory accesses are always done through > accessor functions; > accessing I/O memory directly through pointers is frowned upon and > does not work on all > architectures. Those accessors are written to prevent unwanted > optimization, > .... > - The above-mentioned accessor functions might use volatile on > architectures where direct I/O memory access does work. Essentially, > each accessor call becomes a little critical section on its own and > ensures that the access happens as expected by the programmer. > ... > Patches to remove volatile variables are generally welcome - as long as > they come with a justification which shows that the concurrency issues have > been properly thought through. > " > > Thanks & Regards, > Wu Zhangjin > >> >> Shane >> > -- Best Regards! Kelvin
