kevin/linux
1
Fork 0
Code
547c178b72
linux/arch/ia64/kernel/patch.c
Chen, Kenneth W a0776ec8e9 [IA64] remove per-cpu ia64_phys_stacked_size_p8 
It's not efficient to use a per-cpu variable just to store
how many physical stack register a cpu has.  Ever since the
incarnation of ia64 up till upcoming Montecito processor, that
variable has "glued" to 96. Having a variable in memory means
that the kernel is burning an extra cacheline access on every
syscall and kernel exit path.  Such "static" value is better
served with the instruction patching utility exists today.
Convert ia64_phys_stacked_size_p8 into dynamic insn patching.

This also has a pleasant side effect of eliminating access to
per-cpu area while psr.ic=0 in the kernel exit path. (fixable
for per-cpu DTC work, but why bother?)

There are some concerns with the default value that the instruc-
tion encoded in the kernel image.  It shouldn't be concerned.
The reasons are:

(1) cpu_init() is called at CPU initialization.  In there, we
    find out physical stack register size from PAL and patch
    two instructions in kernel exit code.  The code in question
    can not be executed before the patching is done.

(2) current implementation stores zero in ia64_phys_stacked_size_p8,
    and that's what the current kernel exit path loads the value with.
    With the new code, it is equivalent that we store reg size 96
    in ia64_phys_stacked_size_p8, thus creating a better safety net.
    Given (1) above can never fail, having (2) is just a bonus.

All in all, this patch allow one less memory reference in the kernel
exit path, thus reducing syscall and interrupt return latency; and
avoid polluting potential useful data in the CPU cache.

Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
2007-02-06 15:04:18 -08:00

218 lines
6.0 KiB
C
Raw Blame History

/*
* Instruction-patching support.
*
* Copyright (C) 2003 Hewlett-Packard Co
* David Mosberger-Tang <davidm@hpl.hp.com>
*/
#include <linux/init.h>
#include <linux/string.h>
#include <asm/patch.h>
#include <asm/processor.h>
#include <asm/sections.h>
#include <asm/system.h>
#include <asm/unistd.h>
/*
* This was adapted from code written by Tony Luck:
*
* The 64-bit value in a "movl reg=value" is scattered between the two words of the bundle
* like this:
*
* 6 6 5 4 3 2 1
* 3210987654321098765432109876543210987654321098765432109876543210
* ABBBBBBBBBBBBBBBBBBBBBBBCCCCCCCCCCCCCCCCCCDEEEEEFFFFFFFFFGGGGGGG
*
* CCCCCCCCCCCCCCCCCCxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
* xxxxAFFFFFFFFFEEEEEDxGGGGGGGxxxxxxxxxxxxxBBBBBBBBBBBBBBBBBBBBBBB
*/
static u64
get_imm64 (u64 insn_addr)
{
u64 *p = (u64 *) (insn_addr & -16); /* mask out slot number */
return ( (p[1] & 0x0800000000000000UL) << 4) | /*A*/
((p[1] & 0x00000000007fffffUL) << 40) | /*B*/
((p[0] & 0xffffc00000000000UL) >> 24) | /*C*/
((p[1] & 0x0000100000000000UL) >> 23) | /*D*/
((p[1] & 0x0003e00000000000UL) >> 29) | /*E*/
((p[1] & 0x07fc000000000000UL) >> 43) | /*F*/
((p[1] & 0x000007f000000000UL) >> 36); /*G*/
}
/* Patch instruction with "val" where "mask" has 1 bits. */
void
ia64_patch (u64 insn_addr, u64 mask, u64 val)
{
u64 m0, m1, v0, v1, b0, b1, *b = (u64 *) (insn_addr & -16);
# define insn_mask ((1UL << 41) - 1)
unsigned long shift;
b0 = b[0]; b1 = b[1];
shift = 5 + 41 * (insn_addr % 16); /* 5 bits of template, then 3 x 41-bit instructions */
if (shift >= 64) {
m1 = mask << (shift - 64);
v1 = val << (shift - 64);
} else {
m0 = mask << shift; m1 = mask >> (64 - shift);
v0 = val << shift; v1 = val >> (64 - shift);
b[0] = (b0 & ~m0) | (v0 & m0);
}
b[1] = (b1 & ~m1) | (v1 & m1);
}
void
ia64_patch_imm64 (u64 insn_addr, u64 val)
{
/* The assembler may generate offset pointing to either slot 1
or slot 2 for a long (2-slot) instruction, occupying slots 1
and 2. */
insn_addr &= -16UL;
ia64_patch(insn_addr + 2,
0x01fffefe000UL, ( ((val & 0x8000000000000000UL) >> 27) /* bit 63 -> 36 */
| ((val & 0x0000000000200000UL) << 0) /* bit 21 -> 21 */
| ((val & 0x00000000001f0000UL) << 6) /* bit 16 -> 22 */
| ((val & 0x000000000000ff80UL) << 20) /* bit 7 -> 27 */
| ((val & 0x000000000000007fUL) << 13) /* bit 0 -> 13 */));
ia64_patch(insn_addr + 1, 0x1ffffffffffUL, val >> 22);
}
void
ia64_patch_imm60 (u64 insn_addr, u64 val)
{
/* The assembler may generate offset pointing to either slot 1
or slot 2 for a long (2-slot) instruction, occupying slots 1
and 2. */
insn_addr &= -16UL;
ia64_patch(insn_addr + 2,
0x011ffffe000UL, ( ((val & 0x0800000000000000UL) >> 23) /* bit 59 -> 36 */
| ((val & 0x00000000000fffffUL) << 13) /* bit 0 -> 13 */));
ia64_patch(insn_addr + 1, 0x1fffffffffcUL, val >> 18);
}
/*
* We need sometimes to load the physical address of a kernel
* object. Often we can convert the virtual address to physical
* at execution time, but sometimes (either for performance reasons
* or during error recovery) we cannot to this. Patch the marked
* bundles to load the physical address.
*/
void __init
ia64_patch_vtop (unsigned long start, unsigned long end)
{
s32 *offp = (s32 *) start;
u64 ip;
while (offp < (s32 *) end) {
ip = (u64) offp + *offp;
/* replace virtual address with corresponding physical address: */
ia64_patch_imm64(ip, ia64_tpa(get_imm64(ip)));
ia64_fc((void *) ip);
++offp;
}
ia64_sync_i();
ia64_srlz_i();
}
void __init
ia64_patch_mckinley_e9 (unsigned long start, unsigned long end)
{
static int first_time = 1;
int need_workaround;
s32 *offp = (s32 *) start;
u64 *wp;
need_workaround = (local_cpu_data->family == 0x1f && local_cpu_data->model == 0);
if (first_time) {
first_time = 0;
if (need_workaround)
printk(KERN_INFO "Leaving McKinley Errata 9 workaround enabled\n");
else
printk(KERN_INFO "McKinley Errata 9 workaround not needed; "
"disabling it\n");
}
if (need_workaround)
return;
while (offp < (s32 *) end) {
wp = (u64 *) ia64_imva((char *) offp + *offp);
wp[0] = 0x0000000100000000UL; /* nop.m 0; nop.i 0; nop.i 0 */
wp[1] = 0x0004000000000200UL;
wp[2] = 0x0000000100000011UL; /* nop.m 0; nop.i 0; br.ret.sptk.many b6 */
wp[3] = 0x0084006880000200UL;
ia64_fc(wp); ia64_fc(wp + 2);
++offp;
}
ia64_sync_i();
ia64_srlz_i();
}
static void __init
patch_fsyscall_table (unsigned long start, unsigned long end)
{
extern unsigned long fsyscall_table[NR_syscalls];
s32 *offp = (s32 *) start;
u64 ip;
while (offp < (s32 *) end) {
ip = (u64) ia64_imva((char *) offp + *offp);
ia64_patch_imm64(ip, (u64) fsyscall_table);
ia64_fc((void *) ip);
++offp;
}
ia64_sync_i();
ia64_srlz_i();
}
static void __init
patch_brl_fsys_bubble_down (unsigned long start, unsigned long end)
{
extern char fsys_bubble_down[];
s32 *offp = (s32 *) start;
u64 ip;
while (offp < (s32 *) end) {
ip = (u64) offp + *offp;
ia64_patch_imm60((u64) ia64_imva((void *) ip),
(u64) (fsys_bubble_down - (ip & -16)) / 16);
ia64_fc((void *) ip);
++offp;
}
ia64_sync_i();
ia64_srlz_i();
}
void __init
ia64_patch_gate (void)
{
# define START(name) ((unsigned long) __start_gate_##name##_patchlist)
# define END(name) ((unsigned long)__end_gate_##name##_patchlist)
patch_fsyscall_table(START(fsyscall), END(fsyscall));
patch_brl_fsys_bubble_down(START(brl_fsys_bubble_down), END(brl_fsys_bubble_down));
ia64_patch_vtop(START(vtop), END(vtop));
ia64_patch_mckinley_e9(START(mckinley_e9), END(mckinley_e9));
}
void ia64_patch_phys_stack_reg(unsigned long val)
{
s32 * offp = (s32 *) __start___phys_stack_reg_patchlist;
s32 * end = (s32 *) __end___phys_stack_reg_patchlist;
u64 ip, mask, imm;
/* see instruction format A4: adds r1 = imm13, r3 */
mask = (0x3fUL << 27) | (0x7f << 13);
imm = (((val >> 7) & 0x3f) << 27) | (val & 0x7f) << 13;
while (offp < end) {
ip = (u64) offp + *offp;
ia64_patch(ip, mask, imm);
ia64_fc(ip);
++offp;
}
ia64_sync_i();
ia64_srlz_i();
}
View Git Blame Copy Permalink