[PATCH] powerpc: Merge align.c

Becky Bruce becky.bruce at freescale.com
Tue Nov 15 06:53:19 EST 2005


Ben,

I talked to Kumar about this a little bit (I had started a merge of 
this file, but got distracted!) and he doesn't have any test cases.  
I'll put something together and test this out on some of the 32-bit 
systems I have here in my lab.  It won't be complete, but it will be 
something.......

Cheers,
B

On Nov 14, 2005, at 2:00 AM, Benjamin Herrenschmidt wrote:

> Need testing !!!
>
> This patch merges align.c, the result isn't quite what was in ppc64 nor
> what was in ppc32 :) It should implement all the functionalities of 
> both
> though. Kumar, since you played with that in the past, I suppose you
> have some test cases for verifying that it works properly before I dig
> out the 601 machine ? :)
>
> Since it's likely that I won't be able to test all scenario, code
> inspection is much welcome.
>
> Signed-off-by: Benjamin Herrenschmidt <benh at kernel.crashing.org>
>
>
> Index: linux-work/arch/powerpc/kernel/Makefile
> ===================================================================
> --- linux-work.orig/arch/powerpc/kernel/Makefile        2005-11-14 
> 15:17:57.000000000 +1100
> +++ linux-work/arch/powerpc/kernel/Makefile     2005-11-14 
> 17:18:14.000000000 +1100
> @@ -12,7 +12,7 @@
>  endif
>  
>  obj-y                          := semaphore.o cputable.o ptrace.o 
> syscalls.o \
> -                                  irq.o signal_32.o pmc.o vdso.o
> +                                  irq.o align.o signal_32.o pmc.o 
> vdso.o
>  obj-y                          += vdso32/
>  obj-$(CONFIG_PPC64)            += setup_64.o binfmt_elf32.o 
> sys_ppc32.o \
>                                    signal_64.o ptrace32.o systbl.o \
> Index: linux-work/arch/powerpc/kernel/align.c
> ===================================================================
> --- /dev/null   1970-01-01 00:00:00.000000000 +0000
> +++ linux-work/arch/powerpc/kernel/align.c      2005-11-14 
> 18:41:22.000000000 +1100
> @@ -0,0 +1,513 @@
> +/* align.c - handle alignment exceptions for the Power PC.
> + *
> + * Copyright (c) 1996 Paul Mackerras <paulus at cs.anu.edu.au>
> + * Copyright (c) 1998-1999 TiVo, Inc.
> + *   PowerPC 403GCX modifications.
> + * Copyright (c) 1999 Grant Erickson <grant at lcse.umn.edu>
> + *   PowerPC 403GCX/405GP modifications.
> + * Copyright (c) 2001-2002 PPC64 team, IBM Corp
> + *   64-bit and Power4 support
> + * Copyright (c) 2005 Benjamin Herrenschmidt, IBM Corp
> + *                    <benh at kernel.crashing.org>
> + *   Merge ppc32 and ppc64 implementations
> + *
> + * 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.
> + */
> +
> +#include <linux/kernel.h>
> +#include <linux/mm.h>
> +#include <asm/processor.h>
> +#include <asm/uaccess.h>
> +#include <asm/system.h>
> +#include <asm/cache.h>
> +#include <asm/cputable.h>
> +
> +struct aligninfo {
> +       unsigned char len;
> +       unsigned char flags;
> +};
> +
> +#define IS_XFORM(inst) (((inst) >> 26) == 31)
> +#define IS_DSFORM(inst)        (((inst) >> 26) >= 56)
> +
> +#define INVALID        { 0, 0 }
> +
> +#define LD     1       /* load */
> +#define ST     2       /* store */
> +#define        SE      4       /* sign-extend value */
> +#define F      8       /* to/from fp regs */
> +#define U      0x10    /* update index register */
> +#define M      0x20    /* multiple load/store */
> +#define SW     0x40    /* byte swap int or ... */
> +#define S      0x40    /* ... single-precision fp */
> +#define SX     0x40    /* byte count in XER */
> +#define HARD   0x80    /* string, stwcx. */
> +
> +#define DCBZ   0x5f    /* 8xx/82xx dcbz faults when cache not enabled 
> */
> +
> +#define SWAP(a, b)     (t = (a), (a) = (b), (b) = t)
> +
> +/*
> + * The PowerPC stores certain bits of the instruction that caused the
> + * alignment exception in the DSISR register.  This array maps those
> + * bits to information about the operand length and what the
> + * instruction would do.
> + */
> +static struct aligninfo aligninfo[128] = {
> +       { 4, LD },              /* 00 0 0000: lwz / lwarx */
> +       INVALID,                /* 00 0 0001 */
> +       { 4, ST },              /* 00 0 0010: stw */
> +       INVALID,                /* 00 0 0011 */
> +       { 2, LD },              /* 00 0 0100: lhz */
> +       { 2, LD+SE },           /* 00 0 0101: lha */
> +       { 2, ST },              /* 00 0 0110: sth */
> +       { 4, LD+M },            /* 00 0 0111: lmw */
> +       { 4, LD+F+S },          /* 00 0 1000: lfs */
> +       { 8, LD+F },            /* 00 0 1001: lfd */
> +       { 4, ST+F+S },          /* 00 0 1010: stfs */
> +       { 8, ST+F },            /* 00 0 1011: stfd */
> +       INVALID,                /* 00 0 1100 */
> +       { 8, LD },              /* 00 0 1101: ld/ldu/lwa */
> +       INVALID,                /* 00 0 1110 */
> +       { 8, ST },              /* 00 0 1111: std/stdu */
> +       { 4, LD+U },            /* 00 1 0000: lwzu */
> +       INVALID,                /* 00 1 0001 */
> +       { 4, ST+U },            /* 00 1 0010: stwu */
> +       INVALID,                /* 00 1 0011 */
> +       { 2, LD+U },            /* 00 1 0100: lhzu */
> +       { 2, LD+SE+U },         /* 00 1 0101: lhau */
> +       { 2, ST+U },            /* 00 1 0110: sthu */
> +       { 4, ST+M },            /* 00 1 0111: stmw */
> +       { 4, LD+F+S+U },        /* 00 1 1000: lfsu */
> +       { 8, LD+F+U },          /* 00 1 1001: lfdu */
> +       { 4, ST+F+S+U },        /* 00 1 1010: stfsu */
> +       { 8, ST+F+U },          /* 00 1 1011: stfdu */
> +       INVALID,                /* 00 1 1100 */
> +       INVALID,                /* 00 1 1101 */
> +       INVALID,                /* 00 1 1110 */
> +       INVALID,                /* 00 1 1111 */
> +       { 8, LD },              /* 01 0 0000: ldx */
> +       INVALID,                /* 01 0 0001 */
> +       { 8, ST },              /* 01 0 0010: stdx */
> +       INVALID,                /* 01 0 0011 */
> +       INVALID,                /* 01 0 0100 */
> +       { 4, LD+SE },           /* 01 0 0101: lwax */
> +       INVALID,                /* 01 0 0110 */
> +       INVALID,                /* 01 0 0111 */
> +       { 4, LD+M+HARD+SX },    /* 01 0 1000: lswx */
> +       { 4, LD+M+HARD },       /* 01 0 1001: lswi */
> +       { 4, ST+M+HARD+SX },    /* 01 0 1010: stswx */
> +       { 4, ST+M+HARD },       /* 01 0 1011: stswi */
> +       INVALID,                /* 01 0 1100 */
> +       { 8, LD+U },            /* 01 0 1101: ldu */
> +       INVALID,                /* 01 0 1110 */
> +       { 8, ST+U },            /* 01 0 1111: stdu */
> +       { 8, LD+U },            /* 01 1 0000: ldux */
> +       INVALID,                /* 01 1 0001 */
> +       { 8, ST+U },            /* 01 1 0010: stdux */
> +       INVALID,                /* 01 1 0011 */
> +       INVALID,                /* 01 1 0100 */
> +       { 4, LD+SE+U },         /* 01 1 0101: lwaux */
> +       INVALID,                /* 01 1 0110 */
> +       INVALID,                /* 01 1 0111 */
> +       INVALID,                /* 01 1 1000 */
> +       INVALID,                /* 01 1 1001 */
> +       INVALID,                /* 01 1 1010 */
> +       INVALID,                /* 01 1 1011 */
> +       INVALID,                /* 01 1 1100 */
> +       INVALID,                /* 01 1 1101 */
> +       INVALID,                /* 01 1 1110 */
> +       INVALID,                /* 01 1 1111 */
> +       INVALID,                /* 10 0 0000 */
> +       INVALID,                /* 10 0 0001 */
> +       INVALID,                /* 10 0 0010: stwcx. */
> +       INVALID,                /* 10 0 0011 */
> +       INVALID,                /* 10 0 0100 */
> +       INVALID,                /* 10 0 0101 */
> +       INVALID,                /* 10 0 0110 */
> +       INVALID,                /* 10 0 0111 */
> +       { 4, LD+SW },           /* 10 0 1000: lwbrx */
> +       INVALID,                /* 10 0 1001 */
> +       { 4, ST+SW },           /* 10 0 1010: stwbrx */
> +       INVALID,                /* 10 0 1011 */
> +       { 2, LD+SW },           /* 10 0 1100: lhbrx */
> +       { 4, LD+SE },           /* 10 0 1101  lwa */
> +       { 2, ST+SW },           /* 10 0 1110: sthbrx */
> +       INVALID,                /* 10 0 1111 */
> +       INVALID,                /* 10 1 0000 */
> +       INVALID,                /* 10 1 0001 */
> +       INVALID,                /* 10 1 0010 */
> +       INVALID,                /* 10 1 0011 */
> +       INVALID,                /* 10 1 0100 */
> +       INVALID,                /* 10 1 0101 */
> +       INVALID,                /* 10 1 0110 */
> +       INVALID,                /* 10 1 0111 */
> +       INVALID,                /* 10 1 1000 */
> +       INVALID,                /* 10 1 1001 */
> +       INVALID,                /* 10 1 1010 */
> +       INVALID,                /* 10 1 1011 */
> +       INVALID,                /* 10 1 1100 */
> +       INVALID,                /* 10 1 1101 */
> +       INVALID,                /* 10 1 1110 */
> +       { 0, ST+HARD },         /* 10 1 1111: dcbz */
> +       { 4, LD },              /* 11 0 0000: lwzx */
> +       INVALID,                /* 11 0 0001 */
> +       { 4, ST },              /* 11 0 0010: stwx */
> +       INVALID,                /* 11 0 0011 */
> +       { 2, LD },              /* 11 0 0100: lhzx */
> +       { 2, LD+SE },           /* 11 0 0101: lhax */
> +       { 2, ST },              /* 11 0 0110: sthx */
> +       INVALID,                /* 11 0 0111 */
> +       { 4, LD+F+S },          /* 11 0 1000: lfsx */
> +       { 8, LD+F },            /* 11 0 1001: lfdx */
> +       { 4, ST+F+S },          /* 11 0 1010: stfsx */
> +       { 8, ST+F },            /* 11 0 1011: stfdx */
> +       INVALID,                /* 11 0 1100 */
> +       { 8, LD+M },            /* 11 0 1101: lmd */
> +       INVALID,                /* 11 0 1110 */
> +       { 8, ST+M },            /* 11 0 1111: stmd */
> +       { 4, LD+U },            /* 11 1 0000: lwzux */
> +       INVALID,                /* 11 1 0001 */
> +       { 4, ST+U },            /* 11 1 0010: stwux */
> +       INVALID,                /* 11 1 0011 */
> +       { 2, LD+U },            /* 11 1 0100: lhzux */
> +       { 2, LD+SE+U },         /* 11 1 0101: lhaux */
> +       { 2, ST+U },            /* 11 1 0110: sthux */
> +       INVALID,                /* 11 1 0111 */
> +       { 4, LD+F+S+U },        /* 11 1 1000: lfsux */
> +       { 8, LD+F+U },          /* 11 1 1001: lfdux */
> +       { 4, ST+F+S+U },        /* 11 1 1010: stfsux */
> +       { 8, ST+F+U },          /* 11 1 1011: stfdux */
> +       INVALID,                /* 11 1 1100 */
> +       INVALID,                /* 11 1 1101 */
> +       INVALID,                /* 11 1 1110 */
> +       INVALID,                /* 11 1 1111 */
> +};
> +
> +/*
> + * Create a DSISR value from the instruction
> + */
> +static inline unsigned make_dsisr(unsigned instr)
> +{
> +       unsigned dsisr;
> +
> +
> +       /* bits  6:15 --> 22:31 */
> +       dsisr = (instr & 0x03ff0000) >> 16;
> +
> +       if ( IS_XFORM(instr) ) {
> +               /* bits 29:30 --> 15:16 */
> +               dsisr |= (instr & 0x00000006) << 14;
> +               /* bit     25 -->    17 */
> +               dsisr |= (instr & 0x00000040) << 8;
> +               /* bits 21:24 --> 18:21 */
> +               dsisr |= (instr & 0x00000780) << 3;
> +       }
> +       else {
> +               /* bit      5 -->    17 */
> +               dsisr |= (instr & 0x04000000) >> 12;
> +               /* bits  1: 4 --> 18:21 */
> +               dsisr |= (instr & 0x78000000) >> 17;
> +               /* bits 30:31 --> 12:13 */
> +               if ( IS_DSFORM(instr) )
> +                       dsisr |= (instr & 0x00000003) << 18;
> +       }
> +
> +       return dsisr;
> +}
> +
> +/*
> + * The dcbz (data cache block zero) instruction
> + * gives an alignment fault if used on non-cacheable
> + * memory.  We handle the fault mainly for the
> + * case when we are running with the cache disabled
> + * for debugging.
> + */
> +static int emulate_dcbz(struct pt_regs *regs, unsigned char __user 
> *addr)
> +{
> +       long __user *p;
> +       int i, size;
> +
> +#ifdef __powerpc64__
> +       size = ppc64_caches.dline_size;
> +#else
> +       size = L1_CACHE_BYTES;
> +#endif
> +       p = (long __user *) (regs->dar & -size);
> +       if (user_mode(regs) && !access_ok(VERIFY_WRITE, p, size))
> +               return -EFAULT;
> +       for (i = 0; i < size / sizeof(long); ++i)
> +               if (__put_user(0, p+i))
> +                       return -EFAULT;
> +       return 1;
> +}
> +
> +/*
> + * Emulate load & store multiple instructions
> + */
> +static int emulate_multiple(struct pt_regs *regs, unsigned char 
> __user *addr,
> +                           unsigned int reg, unsigned int nb,
> +                           unsigned int flags, unsigned int instr)
> +{
> +       unsigned char *rptr;
> +       int nb0, i;
> +
> +       /*
> +        * We do not try to emulate 8 bytes multiple as they aren't 
> really
> +        * available in our operating environments and we don't try to
> +        * emulate multiples operations in kernel land as they should 
> never
> +        * be used/generated there at least not on unaligned boundaries
> +        */
> +       if (unlikely((nb > 4) || !user_mode(regs)))
> +               return 0;
> +
> +       /* lmw, stmw, lswi/x, stswi/x */
> +       nb0 = 0;
> +       if (flags & HARD) {
> +               if (flags & SX) {
> +                       nb = regs->xer & 127;
> +                       if (nb == 0)
> +                               return 1;
> +               } else {
> +                       if (__get_user(instr,
> +                                      (unsigned int __user 
> *)regs->nip))
> +                               return -EFAULT;
> +                       nb = (instr >> 11) & 0x1f;
> +                       if (nb == 0)
> +                               nb = 32;
> +               }
> +               if (nb + reg * 4 > 128) {
> +                       nb0 = nb + reg * 4 - 128;
> +                       nb = 128 - reg * 4;
> +               }
> +       } else {
> +               /* lwm, stmw */
> +               nb = (32 - reg) * 4;
> +       }
> +
> +       if (!access_ok((flags & ST ? VERIFY_WRITE: VERIFY_READ), addr, 
> nb+nb0))
> +               return -EFAULT; /* bad address */
> +
> +       rptr = (unsigned char *) &regs->gpr[reg];
> +       if (flags & LD) {
> +               for (i = 0; i < nb; ++i)
> +                       if (__get_user(rptr[i], addr + i))
> +                               return -EFAULT;
> +               if (nb0 > 0) {
> +                       rptr = (unsigned char *) &regs->gpr[0];
> +                       addr += nb;
> +                       for (i = 0; i < nb0; ++i)
> +                               if (__get_user(rptr[i], addr + i))
> +                                       return -EFAULT;
> +               }
> +               for (; (i & 3) != 0; ++i)
> +                       rptr[i] = 0;
> +       } else {
> +               for (i = 0; i < nb; ++i)
> +                       if (__put_user(rptr[i], addr + i))
> +                               return -EFAULT;
> +               if (nb0 > 0) {
> +                       rptr = (unsigned char *) &regs->gpr[0];
> +                       addr += nb;
> +                       for (i = 0; i < nb0; ++i)
> +                               if (__put_user(rptr[i], addr + i))
> +                                       return -EFAULT;
> +               }
> +       }
> +       return 1;
> +}
> +
> +
> +/*
> + * Called on alignment exception. Attempts to fixup
> + *
> + * Return 1 on success
> + * Return 0 if unable to handle the interrupt
> + * Return -EFAULT if data address is bad
> + */
> +
> +int fix_alignment(struct pt_regs *regs)
> +{
> +       unsigned int instr, nb, flags;
> +       unsigned int reg, areg;
> +       unsigned int dsisr;
> +       unsigned char __user *addr;
> +       unsigned char __user *p;
> +       int ret, t;
> +       union {
> +               long ll;
> +               double dd;
> +               unsigned char v[8];
> +               struct {
> +                       unsigned hi32;
> +                       int      low32;
> +               } x32;
> +               struct {
> +                       unsigned char hi48[6];
> +                       short         low16;
> +               } x16;
> +       } data;
> +
> +       /*
> +        * We require a complete register set, if not, then our 
> assembly
> +        * is broken
> +        */
> +       CHECK_FULL_REGS(regs);
> +
> +       dsisr = regs->dsisr;
> +
> +       /* Some processors don't provide us with a DSISR we can use 
> here,
> +        * let's make one up from the instruction
> +        */
> +       if (cpu_has_feature(CPU_FTR_NODSISRALIGN)) {
> +               unsigned int real_instr;
> +               if (unlikely(__get_user(real_instr,
> +                                       (unsigned int __user 
> *)regs->nip)))
> +                       return -EFAULT;
> +               dsisr = make_dsisr(real_instr);
> +       }
> +
> +       /* extract the operation and registers from the dsisr */
> +       reg = (dsisr >> 5) & 0x1f;      /* source/dest register */
> +       areg = dsisr & 0x1f;            /* register to update */
> +       instr = (dsisr >> 10) & 0x7f;
> +       instr |= (dsisr >> 13) & 0x60;
> +
> +       /* Lookup the operation in our table */
> +       nb = aligninfo[instr].len;
> +       flags = aligninfo[instr].flags;
> +
> +       /* DAR has the operand effective address */
> +       addr = (unsigned char __user *)regs->dar;
> +
> +       /* A size of 0 indicates an instruction we don't support, with
> +        * the exception of DCBZ which is handled as a special case 
> here
> +        */
> +       if (instr == DCBZ)
> +               return emulate_dcbz(regs, addr);
> +       if (unlikely(nb == 0))
> +               return 0;
> +
> +       /* Load/Store Multiple instructions are handled in their own
> +        * function
> +        */
> +       if (flags & M)
> +               return emulate_multiple(regs, addr, reg, nb, flags, 
> instr);
> +
> +       /* Verify the address of the operand */
> +       if (unlikely(user_mode(regs) &&
> +                    !access_ok((flags & ST ? VERIFY_WRITE : 
> VERIFY_READ),
> +                               addr, nb)))
> +               return -EFAULT;
> +
> +       /* Force the fprs into the save area so we can reference them 
> */
> +       if (flags & F) {
> +               /* userland only */
> +               if (unlikely(!user_mode(regs)))
> +                       return 0;
> +               flush_fp_to_thread(current);
> +       }
> +
> +       /* If we are loading, get the data from user space, else
> +        * get it from register values
> +        */
> +       if (flags & LD) {
> +               data.ll = 0;
> +               ret = 0;
> +               p = addr;
> +               switch (nb) {
> +               case 8:
> +                       ret |= __get_user(data.v[0], p++);
> +                       ret |= __get_user(data.v[1], p++);
> +                       ret |= __get_user(data.v[2], p++);
> +                       ret |= __get_user(data.v[3], p++);
> +               case 4:
> +                       ret |= __get_user(data.v[4], p++);
> +                       ret |= __get_user(data.v[5], p++);
> +               case 2:
> +                       ret |= __get_user(data.v[6], p++);
> +                       ret |= __get_user(data.v[7], p++);
> +                       if (unlikely(ret))
> +                               return -EFAULT;
> +               }
> +       } else if (flags & F)
> +               data.dd = current->thread.fpr[reg];
> +       else
> +               data.ll = regs->gpr[reg];
> +
> +       /* Perform other misc operations like sign extension, byteswap,
> +        * or floating point single precision conversion
> +        */
> +       switch (flags & ~U) {
> +       case LD+SE:     /* sign extend */
> +               if ( nb == 2 )
> +                       data.ll = data.x16.low16;
> +               else    /* nb must be 4 */
> +                       data.ll = data.x32.low32;
> +               break;
> +       case LD+S:      /* byte-swap */
> +       case ST+S:
> +               if (nb == 2) {
> +                       SWAP(data.v[6], data.v[7]);
> +               } else {
> +                       SWAP(data.v[4], data.v[7]);
> +                       SWAP(data.v[5], data.v[6]);
> +               }
> +               break;
> +
> +       /* Single-precision FP load and store require conversions... */
> +       case LD+F+S:
> +#ifdef CONFIG_PPC_FPU
> +               preempt_disable();
> +               enable_kernel_fp();
> +               cvt_fd((float *)&data.v[4], &data.dd, 
> &current->thread);
> +               preempt_enable();
> +#else
> +               return 0;
> +#endif
> +               break;
> +       case ST+F+S:
> +#ifdef CONFIG_PPC_FPU
> +               preempt_disable();
> +               enable_kernel_fp();
> +               cvt_df(&data.dd, (float *)&data.v[4], 
> &current->thread);
> +               preempt_enable();
> +#else
> +               return 0;
> +#endif
> +               break;
> +       }
> +
> +       /* Store result to memory or update registers */
> +       if (flags & ST) {
> +               ret = 0;
> +               p = addr;
> +               switch (nb) {
> +               case 8:
> +                       ret |= __put_user(data.v[0], p++);
> +                       ret |= __put_user(data.v[1], p++);
> +                       ret |= __put_user(data.v[2], p++);
> +                       ret |= __put_user(data.v[3], p++);
> +               case 4:
> +                       ret |= __put_user(data.v[4], p++);
> +                       ret |= __put_user(data.v[5], p++);
> +               case 2:
> +                       ret |= __put_user(data.v[6], p++);
> +                       ret |= __put_user(data.v[7], p++);
> +               }
> +               if (unlikely(ret))
> +                       return -EFAULT;
> +       } else if (flags & F)
> +               current->thread.fpr[reg] = data.dd;
> +       else
> +               regs->gpr[reg] = data.ll;
> +
> +       /* Update RA as needed */
> +       if (flags & U)
> +               regs->gpr[areg] = regs->dar;
> +
> +       return 1;
> +}
> Index: linux-work/arch/ppc/kernel/Makefile
> ===================================================================
> --- linux-work.orig/arch/ppc/kernel/Makefile    2005-11-11 
> 10:14:48.000000000 +1100
> +++ linux-work/arch/ppc/kernel/Makefile 2005-11-14 18:42:30.000000000 
> +1100
> @@ -13,7 +13,7 @@
>  extra-y                                += vmlinux.lds
>  
>  obj-y                          := entry.o traps.o idle.o time.o 
> misc.o \
> -                                       process.o align.o \
> +                                       process.o \
>                                         setup.o \
>                                         ppc_htab.o
>  obj-$(CONFIG_6xx)              += l2cr.o cpu_setup_6xx.o
> Index: linux-work/arch/ppc64/kernel/Makefile
> ===================================================================
> --- linux-work.orig/arch/ppc64/kernel/Makefile  2005-11-14 
> 15:20:05.000000000 +1100
> +++ linux-work/arch/ppc64/kernel/Makefile       2005-11-14 
> 18:42:12.000000000 +1100
> @@ -11,9 +11,7 @@
>  
>  endif
>  
> -obj-y               += idle.o dma.o \
> -                       align.o \
> -                       iommu.o
> +obj-y               += idle.o dma.o iommu.o
>  
>  pci-obj-$(CONFIG_PPC_MULTIPLATFORM)    += pci_dn.o pci_direct_iommu.o
>  
> Index: linux-work/include/asm-powerpc/cputable.h
> ===================================================================
> --- linux-work.orig/include/asm-powerpc/cputable.h      2005-11-11 
> 10:14:49.000000000 +1100
> +++ linux-work/include/asm-powerpc/cputable.h   2005-11-14 
> 18:33:42.000000000 +1100
> @@ -90,6 +90,7 @@
>  #define CPU_FTR_NEED_COHERENT          ASM_CONST(0x0000000000020000)
>  #define CPU_FTR_NO_BTIC                        
> ASM_CONST(0x0000000000040000)
>  #define CPU_FTR_BIG_PHYS               ASM_CONST(0x0000000000080000)
> +#define CPU_FTR_NODSISRALIGN           ASM_CONST(0x0000000000100000)
>  
>  #ifdef __powerpc64__
>  /* Add the 64b processor unique features in the top half of the word 
> */
> @@ -97,7 +98,6 @@
>  #define CPU_FTR_16M_PAGE               ASM_CONST(0x0000000200000000)
>  #define CPU_FTR_TLBIEL                         
> ASM_CONST(0x0000000400000000)
>  #define CPU_FTR_NOEXECUTE              ASM_CONST(0x0000000800000000)
> -#define CPU_FTR_NODSISRALIGN           ASM_CONST(0x0000001000000000)
>  #define CPU_FTR_IABR                   ASM_CONST(0x0000002000000000)
>  #define CPU_FTR_MMCRA                          
> ASM_CONST(0x0000004000000000)
>  #define CPU_FTR_CTRL                   ASM_CONST(0x0000008000000000)
> @@ -113,7 +113,6 @@
>  #define CPU_FTR_16M_PAGE               ASM_CONST(0x0)
>  #define CPU_FTR_TLBIEL                         ASM_CONST(0x0)
>  #define CPU_FTR_NOEXECUTE              ASM_CONST(0x0)
> -#define CPU_FTR_NODSISRALIGN           ASM_CONST(0x0)
>  #define CPU_FTR_IABR                   ASM_CONST(0x0)
>  #define CPU_FTR_MMCRA                          ASM_CONST(0x0)
>  #define CPU_FTR_CTRL                   ASM_CONST(0x0)
> @@ -273,18 +272,21 @@
>         CPU_FTRS_POWER3_32 = CPU_FTR_COMMON | CPU_FTR_SPLIT_ID_CACHE |
>             CPU_FTR_USE_TB | CPU_FTR_HPTE_TABLE,
>         CPU_FTRS_POWER4_32 = CPU_FTR_COMMON | CPU_FTR_SPLIT_ID_CACHE |
> -           CPU_FTR_USE_TB | CPU_FTR_HPTE_TABLE,
> +           CPU_FTR_USE_TB | CPU_FTR_HPTE_TABLE | CPU_FTR_NODSISRALIGN,
>         CPU_FTRS_970_32 = CPU_FTR_COMMON | CPU_FTR_SPLIT_ID_CACHE |
>             CPU_FTR_USE_TB | CPU_FTR_HPTE_TABLE | CPU_FTR_ALTIVEC_COMP 
> |
> -           CPU_FTR_MAYBE_CAN_NAP,
> +           CPU_FTR_MAYBE_CAN_NAP | CPU_FTR_NODSISRALIGN,
>         CPU_FTRS_8XX = CPU_FTR_SPLIT_ID_CACHE | CPU_FTR_USE_TB,
> -       CPU_FTRS_40X = CPU_FTR_SPLIT_ID_CACHE | CPU_FTR_USE_TB,
> -       CPU_FTRS_44X = CPU_FTR_SPLIT_ID_CACHE | CPU_FTR_USE_TB,
> -       CPU_FTRS_E200 = CPU_FTR_USE_TB,
> -       CPU_FTRS_E500 = CPU_FTR_SPLIT_ID_CACHE | CPU_FTR_USE_TB,
> +       CPU_FTRS_40X = CPU_FTR_SPLIT_ID_CACHE | CPU_FTR_USE_TB |
> +           CPU_FTR_NODSISRALIGN,
> +       CPU_FTRS_44X = CPU_FTR_SPLIT_ID_CACHE | CPU_FTR_USE_TB |
> +           CPU_FTR_NODSISRALIGN,
> +       CPU_FTRS_E200 = CPU_FTR_USE_TB | CPU_FTR_NODSISRALIGN,
> +       CPU_FTRS_E500 = CPU_FTR_SPLIT_ID_CACHE | CPU_FTR_USE_TB |
> +           CPU_FTR_NODSISRALIGN,
>         CPU_FTRS_E500_2 = CPU_FTR_SPLIT_ID_CACHE | CPU_FTR_USE_TB |
> -           CPU_FTR_BIG_PHYS,
> -       CPU_FTRS_GENERIC_32 = CPU_FTR_COMMON,
> +           CPU_FTR_BIG_PHYS | CPU_FTR_NODSISRALIGN,
> +       CPU_FTRS_GENERIC_32 = CPU_FTR_COMMON | CPU_FTR_NODSISRALIGN,
>  #ifdef __powerpc64__
>         CPU_FTRS_POWER3 = CPU_FTR_SPLIT_ID_CACHE | CPU_FTR_USE_TB |
>             CPU_FTR_HPTE_TABLE | CPU_FTR_IABR,
>
>
> _______________________________________________
> Linuxppc-dev mailing list
> Linuxppc-dev at ozlabs.org
> https://ozlabs.org/mailman/listinfo/linuxppc-dev




More information about the Linuxppc-dev mailing list