1
linux/arch/ia64/kernel/patch.c

238 lines
6.4 KiB
C
Raw Normal View 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();
}
[IA64] Workaround for RSE issue Problem: An application violating the architectural rules regarding operation dependencies and having specific Register Stack Engine (RSE) state at the time of the violation, may result in an illegal operation fault and invalid RSE state. Such faults may initiate a cascade of repeated illegal operation faults within OS interruption handlers. The specific behavior is OS dependent. Implication: An application causing an illegal operation fault with specific RSE state may result in a series of illegal operation faults and an eventual OS stack overflow condition. Workaround: OS interruption handlers that switch to kernel backing store implement a check for invalid RSE state to avoid the series of illegal operation faults. The core of the workaround is the RSE_WORKAROUND code sequence inserted into each invocation of the SAVE_MIN_WITH_COVER and SAVE_MIN_WITH_COVER_R19 macros. This sequence includes hard-coded constants that depend on the number of stacked physical registers being 96. The rest of this patch consists of code to disable this workaround should this not be the case (with the presumption that if a future Itanium processor increases the number of registers, it would also remove the need for this patch). Move the start of the RBS up to a mod32 boundary to avoid some corner cases. The dispatch_illegal_op_fault code outgrew the spot it was squatting in when built with this patch and CONFIG_VIRT_CPU_ACCOUNTING=y Move it out to the end of the ivt. Signed-off-by: Tony Luck <tony.luck@intel.com>
2008-05-27 13:23:16 -07:00
/*
* Disable the RSE workaround by turning the conditional branch
* that we tagged in each place the workaround was used into an
* unconditional branch.
*/
void __init
ia64_patch_rse (unsigned long start, unsigned long end)
{
s32 *offp = (s32 *) start;
u64 ip, *b;
while (offp < (s32 *) end) {
ip = (u64) offp + *offp;
b = (u64 *)(ip & -16);
b[1] &= ~0xf800000L;
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");
}
if (need_workaround)
return;
while (offp < (s32 *) end) {
wp = (u64 *) ia64_imva((char *) offp + *offp);
wp[0] = 0x0000000100000011UL; /* nop.m 0; nop.i 0; br.ret.sptk.many b6 */
wp[1] = 0x0084006880000200UL;
wp[2] = 0x0000000100000000UL; /* nop.m 0; nop.i 0; nop.i 0 */
wp[3] = 0x0004000000000200UL;
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));
}
[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>
2006-10-13 10:05:45 -07:00
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();
}