Subversion Repositories shark

Rev

Rev 422 | Blame | Compare with Previous | Last modification | View Log | RSS feed

#ifndef __ASM_SYSTEM_H
#define __ASM_SYSTEM_H

#include <linux/config.h>
#include <linux/kernel.h>
#include <asm/segment.h>
#include <asm/cpufeature.h>
#include <linux/bitops.h> /* for LOCK_PREFIX */

#ifdef __KERNEL__

struct task_struct;     /* one of the stranger aspects of C forward declarations.. */
extern struct task_struct * FASTCALL(__switch_to(struct task_struct *prev, struct task_struct *next));

#define switch_to(prev,next,last) do {                                  \
        unsigned long esi,edi;                                          \
        asm volatile("pushfl\n\t"                                       \
                     "pushl %%ebp\n\t"                                  \
                     "movl %%esp,%0\n\t"        /* save ESP */          \
                     "movl %5,%%esp\n\t"        /* restore ESP */       \
                     "movl $1f,%1\n\t"          /* save EIP */          \
                     "pushl %6\n\t"             /* restore EIP */       \
                     "jmp __switch_to\n"                                \
                     "1:\t"                                             \
                     "popl %%ebp\n\t"                                   \
                     "popfl"                                            \
                     :"=m" (prev->thread.esp),"=m" (prev->thread.eip),  \
                      "=a" (last),"=S" (esi),"=D" (edi)                 \
                     :"m" (next->thread.esp),"m" (next->thread.eip),    \
                      "2" (prev), "d" (next));                          \
} while (0)


#define _set_base(addr,base) do { unsigned long __pr; \
__asm__ __volatile__ ("movw %%dx,%1\n\t" \
        "rorl $16,%%edx\n\t" \
        "movb %%dl,%2\n\t" \
        "movb %%dh,%3" \
        :"=&d" (__pr) \
        :"m" (*((addr)+2)), \
         "m" (*((addr)+4)), \
         "m" (*((addr)+7)), \
         "0" (base) \
        ); } while(0)


#define _set_limit(addr,limit) do { unsigned long __lr; \
__asm__ __volatile__ ("movw %%dx,%1\n\t" \
        "rorl $16,%%edx\n\t" \
        "movb %2,%%dh\n\t" \
        "andb $0xf0,%%dh\n\t" \
        "orb %%dh,%%dl\n\t" \
        "movb %%dl,%2" \
        :"=&d" (__lr) \
        :"m" (*(addr)), \
         "m" (*((addr)+6)), \
         "0" (limit) \
        ); } while(0)


#define set_base(ldt,base) _set_base( ((char *)&(ldt)) , (base) )
#define set_limit(ldt,limit) _set_limit( ((char *)&(ldt)) , ((limit)-1)>>12 )

static inline unsigned long _get_base(char * addr)
{
        unsigned long __base;
        __asm__("movb %3,%%dh\n\t"
                "movb %2,%%dl\n\t"
                "shll $16,%%edx\n\t"
                "movw %1,%%dx"
                :"=&d" (__base)
                :"m" (*((addr)+2)),
                 "m" (*((addr)+4)),
                 "m" (*((addr)+7)));
        return __base;
}

#define get_base(ldt) _get_base( ((char *)&(ldt)) )

/*
 * Load a segment. Fall back on loading the zero
 * segment if something goes wrong..
 */

#define loadsegment(seg,value)                  \
        asm volatile("\n"                       \
                "1:\t"                          \
                "movl %0,%%" #seg "\n"          \
                "2:\n"                          \
                ".section .fixup,\"ax\"\n"      \
                "3:\t"                          \
                "pushl $0\n\t"                  \
                "popl %%" #seg "\n\t"           \
                "jmp 2b\n"                      \
                ".previous\n"                   \
                ".section __ex_table,\"a\"\n\t" \
                ".align 4\n\t"                  \
                ".long 1b,3b\n"                 \
                ".previous"                     \
                : :"m" (*(unsigned int *)&(value)))


/*
 * Save a segment register away
 */

#define savesegment(seg, value) \
        asm volatile("movl %%" #seg ",%0":"=m" (*(int *)&(value)))


/*
 * Clear and set 'TS' bit respectively
 */

#define clts() __asm__ __volatile__ ("clts")
#define read_cr0() ({ \
        unsigned int __dummy; \
        __asm__( \
                "movl %%cr0,%0\n\t" \
                :"=r" (__dummy)); \
        __dummy; \
})

#define write_cr0(x) \
        __asm__("movl %0,%%cr0": :"r" (x));


#define read_cr4() ({ \
        unsigned int __dummy; \
        __asm__( \
                "movl %%cr4,%0\n\t" \
                :"=r" (__dummy)); \
        __dummy; \
})

#define write_cr4(x) \
        __asm__("movl %0,%%cr4": :"r" (x));

#define stts() write_cr0(8 | read_cr0())

#endif  /* __KERNEL__ */

#define wbinvd() \
        __asm__ __volatile__ ("wbinvd": : :"memory");


static inline unsigned long get_limit(unsigned long segment)
{
        unsigned long __limit;
        __asm__("lsll %1,%0"
                :"=r" (__limit):"r" (segment));
        return __limit+1;
}

#define nop() __asm__ __volatile__ ("nop")

#define xchg(ptr,v) ((__typeof__(*(ptr)))__xchg((unsigned long)(v),(ptr),sizeof(*(ptr))))

#define tas(ptr) (xchg((ptr),1))

struct __xchg_dummy { unsigned long a[100]; };
#define __xg(x) ((struct __xchg_dummy *)(x))


/*
 * The semantics of XCHGCMP8B are a bit strange, this is why
 * there is a loop and the loading of %%eax and %%edx has to
 * be inside. This inlines well in most cases, the cached
 * cost is around ~38 cycles. (in the future we might want
 * to do an SIMD/3DNOW!/MMX/FPU 64-bit store here, but that
 * might have an implicit FPU-save as a cost, so it's not
 * clear which path to go.)
 *
 * cmpxchg8b must be used with the lock prefix here to allow
 * the instruction to be executed atomically, see page 3-102
 * of the instruction set reference 24319102.pdf. We need
 * the reader side to see the coherent 64bit value.
 */

static inline void __set_64bit (unsigned long long * ptr,
                unsigned int low, unsigned int high)
{
        __asm__ __volatile__ (
                "\n1:\t"
                "movl (%0), %%eax\n\t"
                "movl 4(%0), %%edx\n\t"
                "lock cmpxchg8b (%0)\n\t"
                "jnz 1b"
                : /* no outputs */
                :       "D"(ptr),
                        "b"(low),
                        "c"(high)
                :       "ax","dx","memory");
}

static inline void __set_64bit_constant (unsigned long long *ptr,
                                                 unsigned long long value)
{
        __set_64bit(ptr,(unsigned int)(value), (unsigned int)((value)>>32ULL));
}
#define ll_low(x)       *(((unsigned int*)&(x))+0)
#define ll_high(x)      *(((unsigned int*)&(x))+1)

static inline void __set_64bit_var (unsigned long long *ptr,
                         unsigned long long value)
{
        __set_64bit(ptr,ll_low(value), ll_high(value));
}

#define set_64bit(ptr,value) \
(__builtin_constant_p(value) ? \
 __set_64bit_constant(ptr, value) : \
 __set_64bit_var(ptr, value) )


#define _set_64bit(ptr,value) \
(__builtin_constant_p(value) ? \
 __set_64bit(ptr, (unsigned int)(value), (unsigned int)((value)>>32ULL) ) : \
 __set_64bit(ptr, ll_low(value), ll_high(value)) )


/*
 * Note: no "lock" prefix even on SMP: xchg always implies lock anyway
 * Note 2: xchg has side effect, so that attribute volatile is necessary,
 *        but generally the primitive is invalid, *ptr is output argument. --ANK
 */

static inline unsigned long __xchg(unsigned long x, volatile void * ptr, int size)
{
        switch (size) {
                case 1:
                        __asm__ __volatile__("xchgb %b0,%1"
                                :"=q" (x)
                                :"m" (*__xg(ptr)), "0" (x)
                                :"memory");
                        break;
                case 2:
                        __asm__ __volatile__("xchgw %w0,%1"
                                :"=r" (x)
                                :"m" (*__xg(ptr)), "0" (x)
                                :"memory");
                        break;
                case 4:
                        __asm__ __volatile__("xchgl %0,%1"
                                :"=r" (x)
                                :"m" (*__xg(ptr)), "0" (x)
                                :"memory");
                        break;
        }
        return x;
}

/*
 * Atomic compare and exchange.  Compare OLD with MEM, if identical,
 * store NEW in MEM.  Return the initial value in MEM.  Success is
 * indicated by comparing RETURN with OLD.
 */


#ifdef CONFIG_X86_CMPXCHG
#define __HAVE_ARCH_CMPXCHG 1

static inline unsigned long __cmpxchg(volatile void *ptr, unsigned long old,
                                      unsigned long new, int size)
{
        unsigned long prev;
        switch (size) {
        case 1:
                __asm__ __volatile__(LOCK_PREFIX "cmpxchgb %b1,%2"
                                     : "=a"(prev)
                                     : "q"(new), "m"(*__xg(ptr)), "0"(old)
                                     : "memory");
                return prev;
        case 2:
                __asm__ __volatile__(LOCK_PREFIX "cmpxchgw %w1,%2"
                                     : "=a"(prev)
                                     : "q"(new), "m"(*__xg(ptr)), "0"(old)
                                     : "memory");
                return prev;
        case 4:
                __asm__ __volatile__(LOCK_PREFIX "cmpxchgl %1,%2"
                                     : "=a"(prev)
                                     : "q"(new), "m"(*__xg(ptr)), "0"(old)
                                     : "memory");
                return prev;
        }
        return old;
}

#define cmpxchg(ptr,o,n)\
        ((__typeof__(*(ptr)))__cmpxchg((ptr),(unsigned long)(o),\
                                        (unsigned long)(n),sizeof(*(ptr))))

   
#else
/* Compiling for a 386 proper.  Is it worth implementing via cli/sti?  */
#endif

#ifdef __KERNEL__
struct alt_instr {
        __u8 *instr;            /* original instruction */
        __u8 *replacement;
        __u8  cpuid;            /* cpuid bit set for replacement */
        __u8  instrlen;         /* length of original instruction */
        __u8  replacementlen;   /* length of new instruction, <= instrlen */
        __u8  pad;
};
#endif

/*
 * Alternative instructions for different CPU types or capabilities.
 *
 * This allows to use optimized instructions even on generic binary
 * kernels.
 *
 * length of oldinstr must be longer or equal the length of newinstr
 * It can be padded with nops as needed.
 *
 * For non barrier like inlines please define new variants
 * without volatile and memory clobber.
 */

#define alternative(oldinstr, newinstr, feature)        \
        asm volatile ("661:\n\t" oldinstr "\n662:\n"                 \
                      ".section .altinstructions,\"a\"\n"            \
                      "  .align 4\n"                                   \
                      "  .long 661b\n"            /* label */          \
                      "  .long 663f\n"            /* new instruction */         \
                      "  .byte %c0\n"             /* feature bit */    \
                      "  .byte 662b-661b\n"       /* sourcelen */      \
                      "  .byte 664f-663f\n"       /* replacementlen */ \
                      ".previous\n"                                             \
                      ".section .altinstr_replacement,\"ax\"\n"                 \
                      "663:\n\t" newinstr "\n664:\n"   /* replacement */    \
                      ".previous" :: "i" (feature) : "memory")  


/*
 * Alternative inline assembly with input.
 *
 * Pecularities:
 * No memory clobber here.
 * Argument numbers start with 1.
 * Best is to use constraints that are fixed size (like (%1) ... "r")
 * If you use variable sized constraints like "m" or "g" in the
 * replacement maake sure to pad to the worst case length.
 */

#define alternative_input(oldinstr, newinstr, feature, input)                   \
        asm volatile ("661:\n\t" oldinstr "\n662:\n"                            \
                      ".section .altinstructions,\"a\"\n"                       \
                      "  .align 4\n"                                            \
                      "  .long 661b\n"            /* label */                   \
                      "  .long 663f\n"            /* new instruction */         \
                      "  .byte %c0\n"             /* feature bit */             \
                      "  .byte 662b-661b\n"       /* sourcelen */               \
                      "  .byte 664f-663f\n"       /* replacementlen */          \
                      ".previous\n"                                             \
                      ".section .altinstr_replacement,\"ax\"\n"                 \
                      "663:\n\t" newinstr "\n664:\n"   /* replacement */        \
                      ".previous" :: "i" (feature), input)  


/*
 * Force strict CPU ordering.
 * And yes, this is required on UP too when we're talking
 * to devices.
 *
 * For now, "wmb()" doesn't actually do anything, as all
 * Intel CPU's follow what Intel calls a *Processor Order*,
 * in which all writes are seen in the program order even
 * outside the CPU.
 *
 * I expect future Intel CPU's to have a weaker ordering,
 * but I'd also expect them to finally get their act together
 * and add some real memory barriers if so.
 *
 * Some non intel clones support out of order store. wmb() ceases to be a
 * nop for these.
 */

 

/*
 * Actually only lfence would be needed for mb() because all stores done
 * by the kernel should be already ordered. But keep a full barrier for now.
 */


#define mb() alternative("lock; addl $0,0(%%esp)", "mfence", X86_FEATURE_XMM2)
#define rmb() alternative("lock; addl $0,0(%%esp)", "lfence", X86_FEATURE_XMM2)

/**
 * read_barrier_depends - Flush all pending reads that subsequents reads
 * depend on.
 *
 * No data-dependent reads from memory-like regions are ever reordered
 * over this barrier.  All reads preceding this primitive are guaranteed
 * to access memory (but not necessarily other CPUs' caches) before any
 * reads following this primitive that depend on the data return by
 * any of the preceding reads.  This primitive is much lighter weight than
 * rmb() on most CPUs, and is never heavier weight than is
 * rmb().
 *
 * These ordering constraints are respected by both the local CPU
 * and the compiler.
 *
 * Ordering is not guaranteed by anything other than these primitives,
 * not even by data dependencies.  See the documentation for
 * memory_barrier() for examples and URLs to more information.
 *
 * For example, the following code would force ordering (the initial
 * value of "a" is zero, "b" is one, and "p" is "&a"):
 *
 * <programlisting>
 *      CPU 0                           CPU 1
 *
 *      b = 2;
 *      memory_barrier();
 *      p = &b;                         q = p;
 *                                      read_barrier_depends();
 *                                      d = *q;
 * </programlisting>
 *
 * because the read of "*q" depends on the read of "p" and these
 * two reads are separated by a read_barrier_depends().  However,
 * the following code, with the same initial values for "a" and "b":
 *
 * <programlisting>
 *      CPU 0                           CPU 1
 *
 *      a = 2;
 *      memory_barrier();
 *      b = 3;                          y = b;
 *                                      read_barrier_depends();
 *                                      x = a;
 * </programlisting>
 *
 * does not enforce ordering, since there is no data dependency between
 * the read of "a" and the read of "b".  Therefore, on some CPUs, such
 * as Alpha, "y" could be set to 3 and "x" to 0.  Use rmb()
 * in cases like thiswhere there are no data dependencies.
 **/


#define read_barrier_depends()  do { } while(0)

#ifdef CONFIG_X86_OOSTORE
/* Actually there are no OOO store capable CPUs for now that do SSE,
   but make it already an possibility. */

#define wmb() alternative("lock; addl $0,0(%%esp)", "sfence", X86_FEATURE_XMM)
#else
#define wmb()   __asm__ __volatile__ ("": : :"memory")
#endif

#ifdef CONFIG_SMP
#define smp_mb()        mb()
#define smp_rmb()       rmb()
#define smp_wmb()       wmb()
#define smp_read_barrier_depends()      read_barrier_depends()
#define set_mb(var, value) do { xchg(&var, value); } while (0)
#else
#define smp_mb()        barrier()
#define smp_rmb()       barrier()
#define smp_wmb()       barrier()
#define smp_read_barrier_depends()      do { } while(0)
#define set_mb(var, value) do { var = value; barrier(); } while (0)
#endif

#define set_wmb(var, value) do { var = value; wmb(); } while (0)

/* interrupt control.. */
#define local_save_flags(x)     do { typecheck(unsigned long,x); __asm__ __volatile__("pushfl ; popl %0":"=g" (x): /* no input */); } while (0)
#define local_irq_restore(x)    do { typecheck(unsigned long,x); __asm__ __volatile__("pushl %0 ; popfl": /* no output */ :"g" (x):"memory", "cc"); } while (0)
#define local_irq_disable()     __asm__ __volatile__("cli": : :"memory")
#define local_irq_enable()      __asm__ __volatile__("sti": : :"memory")
/* used in the idle loop; sti takes one instruction cycle to complete */
#define safe_halt()             __asm__ __volatile__("sti; hlt": : :"memory")

#define irqs_disabled()                 \
({                                      \
        unsigned long flags;            \
        local_save_flags(flags);        \
        !(flags & (1<<9));              \
})


/* For spinlocks etc */
#define local_irq_save(x)       __asm__ __volatile__("pushfl ; popl %0 ; cli":"=g" (x): /* no input */ :"memory")

/*
 * disable hlt during certain critical i/o operations
 */

#define HAVE_DISABLE_HLT
void disable_hlt(void);
void enable_hlt(void);

extern unsigned long dmi_broken;
extern int is_sony_vaio_laptop;

#define BROKEN_ACPI_Sx          0x0001
#define BROKEN_INIT_AFTER_S1    0x0002
#define BROKEN_PNP_BIOS         0x0004
#define BROKEN_CPUFREQ          0x0008

#endif