aboutsummaryrefslogtreecommitdiff
path: root/arch/mips/mm/tlbex.c
diff options
context:
space:
mode:
Diffstat (limited to 'arch/mips/mm/tlbex.c')
-rw-r--r--arch/mips/mm/tlbex.c471
1 files changed, 227 insertions, 244 deletions
diff --git a/arch/mips/mm/tlbex.c b/arch/mips/mm/tlbex.c
index 9ab0f907a52..e80e10bafc8 100644
--- a/arch/mips/mm/tlbex.c
+++ b/arch/mips/mm/tlbex.c
@@ -26,10 +26,10 @@
#include <linux/types.h>
#include <linux/smp.h>
#include <linux/string.h>
-#include <linux/init.h>
#include <linux/cache.h>
#include <asm/cacheflush.h>
+#include <asm/cpu-type.h>
#include <asm/pgtable.h>
#include <asm/war.h>
#include <asm/uasm.h>
@@ -85,6 +85,7 @@ static int use_bbit_insns(void)
case CPU_CAVIUM_OCTEON:
case CPU_CAVIUM_OCTEON_PLUS:
case CPU_CAVIUM_OCTEON2:
+ case CPU_CAVIUM_OCTEON3:
return 1;
default:
return 0;
@@ -95,6 +96,7 @@ static int use_lwx_insns(void)
{
switch (current_cpu_type()) {
case CPU_CAVIUM_OCTEON2:
+ case CPU_CAVIUM_OCTEON3:
return 1;
default:
return 0;
@@ -136,7 +138,7 @@ static int scratchpad_offset(int i)
* why; it's not an issue caused by the core RTL.
*
*/
-static int __cpuinit m4kc_tlbp_war(void)
+static int m4kc_tlbp_war(void)
{
return (current_cpu_data.processor_id & 0xffff00) ==
(PRID_COMP_MIPS | PRID_IMP_4KC);
@@ -181,11 +183,9 @@ UASM_L_LA(_large_segbits_fault)
UASM_L_LA(_tlb_huge_update)
#endif
-static int __cpuinitdata hazard_instance;
+static int hazard_instance;
-static void __cpuinit uasm_bgezl_hazard(u32 **p,
- struct uasm_reloc **r,
- int instance)
+static void uasm_bgezl_hazard(u32 **p, struct uasm_reloc **r, int instance)
{
switch (instance) {
case 0 ... 7:
@@ -196,9 +196,7 @@ static void __cpuinit uasm_bgezl_hazard(u32 **p,
}
}
-static void __cpuinit uasm_bgezl_label(struct uasm_label **l,
- u32 **p,
- int instance)
+static void uasm_bgezl_label(struct uasm_label **l, u32 **p, int instance)
{
switch (instance) {
case 0 ... 7:
@@ -295,15 +293,15 @@ static inline void dump_handler(const char *symbol, const u32 *handler, int coun
* We deliberately chose a buffer size of 128, so we won't scribble
* over anything important on overflow before we panic.
*/
-static u32 tlb_handler[128] __cpuinitdata;
+static u32 tlb_handler[128];
/* simply assume worst case size for labels and relocs */
-static struct uasm_label labels[128] __cpuinitdata;
-static struct uasm_reloc relocs[128] __cpuinitdata;
+static struct uasm_label labels[128];
+static struct uasm_reloc relocs[128];
-static int check_for_high_segbits __cpuinitdata;
+static int check_for_high_segbits;
-static unsigned int kscratch_used_mask __cpuinitdata;
+static unsigned int kscratch_used_mask;
static inline int __maybe_unused c0_kscratch(void)
{
@@ -316,7 +314,7 @@ static inline int __maybe_unused c0_kscratch(void)
}
}
-static int __cpuinit allocate_kscratch(void)
+static int allocate_kscratch(void)
{
int r;
unsigned int a = cpu_data[0].kscratch_mask & ~kscratch_used_mask;
@@ -333,18 +331,14 @@ static int __cpuinit allocate_kscratch(void)
return r;
}
-static int scratch_reg __cpuinitdata;
-static int pgd_reg __cpuinitdata;
+static int scratch_reg;
+static int pgd_reg;
enum vmalloc64_mode {not_refill, refill_scratch, refill_noscratch};
-static struct work_registers __cpuinit build_get_work_registers(u32 **p)
+static struct work_registers build_get_work_registers(u32 **p)
{
struct work_registers r;
- int smp_processor_id_reg;
- int smp_processor_id_sel;
- int smp_processor_id_shift;
-
if (scratch_reg >= 0) {
/* Save in CPU local C0_KScratch? */
UASM_i_MTC0(p, 1, c0_kscratch(), scratch_reg);
@@ -355,25 +349,9 @@ static struct work_registers __cpuinit build_get_work_registers(u32 **p)
}
if (num_possible_cpus() > 1) {
-#ifdef CONFIG_MIPS_PGD_C0_CONTEXT
- smp_processor_id_shift = 51;
- smp_processor_id_reg = 20; /* XContext */
- smp_processor_id_sel = 0;
-#else
-# ifdef CONFIG_32BIT
- smp_processor_id_shift = 25;
- smp_processor_id_reg = 4; /* Context */
- smp_processor_id_sel = 0;
-# endif
-# ifdef CONFIG_64BIT
- smp_processor_id_shift = 26;
- smp_processor_id_reg = 4; /* Context */
- smp_processor_id_sel = 0;
-# endif
-#endif
/* Get smp_processor_id */
- UASM_i_MFC0(p, K0, smp_processor_id_reg, smp_processor_id_sel);
- UASM_i_SRL_SAFE(p, K0, K0, smp_processor_id_shift);
+ UASM_i_CPUID_MFC0(p, K0, SMP_CPUID_REG);
+ UASM_i_SRL_SAFE(p, K0, K0, SMP_CPUID_REGSHIFT);
/* handler_reg_save index in K0 */
UASM_i_SLL(p, K0, K0, ilog2(sizeof(struct tlb_reg_save)));
@@ -393,7 +371,7 @@ static struct work_registers __cpuinit build_get_work_registers(u32 **p)
return r;
}
-static void __cpuinit build_restore_work_registers(u32 **p)
+static void build_restore_work_registers(u32 **p)
{
if (scratch_reg >= 0) {
UASM_i_MFC0(p, 1, c0_kscratch(), scratch_reg);
@@ -418,7 +396,7 @@ extern unsigned long pgd_current[];
/*
* The R3000 TLB handler is simple.
*/
-static void __cpuinit build_r3000_tlb_refill_handler(void)
+static void build_r3000_tlb_refill_handler(void)
{
long pgdc = (long)pgd_current;
u32 *p;
@@ -463,7 +441,7 @@ static void __cpuinit build_r3000_tlb_refill_handler(void)
* other one.To keep things simple, we first assume linear space,
* then we relocate it to the final handler layout as needed.
*/
-static u32 final_handler[64] __cpuinitdata;
+static u32 final_handler[64];
/*
* Hazards
@@ -487,7 +465,7 @@ static u32 final_handler[64] __cpuinitdata;
*
* As if we MIPS hackers wouldn't know how to nop pipelines happy ...
*/
-static void __cpuinit __maybe_unused build_tlb_probe_entry(u32 **p)
+static void __maybe_unused build_tlb_probe_entry(u32 **p)
{
switch (current_cpu_type()) {
/* Found by experiment: R4600 v2.0/R4700 needs this, too. */
@@ -511,9 +489,9 @@ static void __cpuinit __maybe_unused build_tlb_probe_entry(u32 **p)
*/
enum tlb_write_entry { tlb_random, tlb_indexed };
-static void __cpuinit build_tlb_write_entry(u32 **p, struct uasm_label **l,
- struct uasm_reloc **r,
- enum tlb_write_entry wmode)
+static void build_tlb_write_entry(u32 **p, struct uasm_label **l,
+ struct uasm_reloc **r,
+ enum tlb_write_entry wmode)
{
void(*tlbw)(u32 **) = NULL;
@@ -531,6 +509,10 @@ static void __cpuinit build_tlb_write_entry(u32 **p, struct uasm_label **l,
switch (current_cpu_type()) {
case CPU_M14KC:
case CPU_74K:
+ case CPU_1074K:
+ case CPU_PROAPTIV:
+ case CPU_P5600:
+ case CPU_M5150:
break;
default:
@@ -600,6 +582,7 @@ static void __cpuinit build_tlb_write_entry(u32 **p, struct uasm_label **l,
case CPU_BMIPS4380:
case CPU_BMIPS5000:
case CPU_LOONGSON2:
+ case CPU_LOONGSON3:
case CPU_R5500:
if (m4kc_tlbp_war())
uasm_i_nop(p);
@@ -642,13 +625,13 @@ static void __cpuinit build_tlb_write_entry(u32 **p, struct uasm_label **l,
default:
panic("No TLB refill handler yet (CPU type: %d)",
- current_cpu_data.cputype);
+ current_cpu_type());
break;
}
}
-static __cpuinit __maybe_unused void build_convert_pte_to_entrylo(u32 **p,
- unsigned int reg)
+static __maybe_unused void build_convert_pte_to_entrylo(u32 **p,
+ unsigned int reg)
{
if (cpu_has_rixi) {
UASM_i_ROTR(p, reg, reg, ilog2(_PAGE_GLOBAL));
@@ -663,11 +646,9 @@ static __cpuinit __maybe_unused void build_convert_pte_to_entrylo(u32 **p,
#ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT
-static __cpuinit void build_restore_pagemask(u32 **p,
- struct uasm_reloc **r,
- unsigned int tmp,
- enum label_id lid,
- int restore_scratch)
+static void build_restore_pagemask(u32 **p, struct uasm_reloc **r,
+ unsigned int tmp, enum label_id lid,
+ int restore_scratch)
{
if (restore_scratch) {
/* Reset default page size */
@@ -706,12 +687,11 @@ static __cpuinit void build_restore_pagemask(u32 **p,
}
}
-static __cpuinit void build_huge_tlb_write_entry(u32 **p,
- struct uasm_label **l,
- struct uasm_reloc **r,
- unsigned int tmp,
- enum tlb_write_entry wmode,
- int restore_scratch)
+static void build_huge_tlb_write_entry(u32 **p, struct uasm_label **l,
+ struct uasm_reloc **r,
+ unsigned int tmp,
+ enum tlb_write_entry wmode,
+ int restore_scratch)
{
/* Set huge page tlb entry size */
uasm_i_lui(p, tmp, PM_HUGE_MASK >> 16);
@@ -726,9 +706,9 @@ static __cpuinit void build_huge_tlb_write_entry(u32 **p,
/*
* Check if Huge PTE is present, if so then jump to LABEL.
*/
-static void __cpuinit
+static void
build_is_huge_pte(u32 **p, struct uasm_reloc **r, unsigned int tmp,
- unsigned int pmd, int lid)
+ unsigned int pmd, int lid)
{
UASM_i_LW(p, tmp, 0, pmd);
if (use_bbit_insns()) {
@@ -739,9 +719,8 @@ build_is_huge_pte(u32 **p, struct uasm_reloc **r, unsigned int tmp,
}
}
-static __cpuinit void build_huge_update_entries(u32 **p,
- unsigned int pte,
- unsigned int tmp)
+static void build_huge_update_entries(u32 **p, unsigned int pte,
+ unsigned int tmp)
{
int small_sequence;
@@ -771,11 +750,10 @@ static __cpuinit void build_huge_update_entries(u32 **p,
UASM_i_MTC0(p, pte, C0_ENTRYLO1); /* load it */
}
-static __cpuinit void build_huge_handler_tail(u32 **p,
- struct uasm_reloc **r,
- struct uasm_label **l,
- unsigned int pte,
- unsigned int ptr)
+static void build_huge_handler_tail(u32 **p, struct uasm_reloc **r,
+ struct uasm_label **l,
+ unsigned int pte,
+ unsigned int ptr)
{
#ifdef CONFIG_SMP
UASM_i_SC(p, pte, 0, ptr);
@@ -794,7 +772,7 @@ static __cpuinit void build_huge_handler_tail(u32 **p,
* TMP and PTR are scratch.
* TMP will be clobbered, PTR will hold the pmd entry.
*/
-static void __cpuinit
+static void
build_get_pmde64(u32 **p, struct uasm_label **l, struct uasm_reloc **r,
unsigned int tmp, unsigned int ptr)
{
@@ -825,11 +803,11 @@ build_get_pmde64(u32 **p, struct uasm_label **l, struct uasm_reloc **r,
}
/* No uasm_i_nop needed here, since the next insn doesn't touch TMP. */
-#ifdef CONFIG_MIPS_PGD_C0_CONTEXT
if (pgd_reg != -1) {
/* pgd is in pgd_reg */
UASM_i_MFC0(p, ptr, c0_kscratch(), pgd_reg);
} else {
+#if defined(CONFIG_MIPS_PGD_C0_CONTEXT)
/*
* &pgd << 11 stored in CONTEXT [23..63].
*/
@@ -841,30 +819,18 @@ build_get_pmde64(u32 **p, struct uasm_label **l, struct uasm_reloc **r,
/* 1 0 1 0 1 << 6 xkphys cached */
uasm_i_ori(p, ptr, ptr, 0x540);
uasm_i_drotr(p, ptr, ptr, 11);
- }
#elif defined(CONFIG_SMP)
-# ifdef CONFIG_MIPS_MT_SMTC
- /*
- * SMTC uses TCBind value as "CPU" index
- */
- uasm_i_mfc0(p, ptr, C0_TCBIND);
- uasm_i_dsrl_safe(p, ptr, ptr, 19);
-# else
- /*
- * 64 bit SMP running in XKPHYS has smp_processor_id() << 3
- * stored in CONTEXT.
- */
- uasm_i_dmfc0(p, ptr, C0_CONTEXT);
- uasm_i_dsrl_safe(p, ptr, ptr, 23);
-# endif
- UASM_i_LA_mostly(p, tmp, pgdc);
- uasm_i_daddu(p, ptr, ptr, tmp);
- uasm_i_dmfc0(p, tmp, C0_BADVADDR);
- uasm_i_ld(p, ptr, uasm_rel_lo(pgdc), ptr);
+ UASM_i_CPUID_MFC0(p, ptr, SMP_CPUID_REG);
+ uasm_i_dsrl_safe(p, ptr, ptr, SMP_CPUID_PTRSHIFT);
+ UASM_i_LA_mostly(p, tmp, pgdc);
+ uasm_i_daddu(p, ptr, ptr, tmp);
+ uasm_i_dmfc0(p, tmp, C0_BADVADDR);
+ uasm_i_ld(p, ptr, uasm_rel_lo(pgdc), ptr);
#else
- UASM_i_LA_mostly(p, ptr, pgdc);
- uasm_i_ld(p, ptr, uasm_rel_lo(pgdc), ptr);
+ UASM_i_LA_mostly(p, ptr, pgdc);
+ uasm_i_ld(p, ptr, uasm_rel_lo(pgdc), ptr);
#endif
+ }
uasm_l_vmalloc_done(l, *p);
@@ -886,7 +852,7 @@ build_get_pmde64(u32 **p, struct uasm_label **l, struct uasm_reloc **r,
* BVADDR is the faulting address, PTR is scratch.
* PTR will hold the pgd for vmalloc.
*/
-static void __cpuinit
+static void
build_get_pgd_vmalloc64(u32 **p, struct uasm_label **l, struct uasm_reloc **r,
unsigned int bvaddr, unsigned int ptr,
enum vmalloc64_mode mode)
@@ -956,34 +922,28 @@ build_get_pgd_vmalloc64(u32 **p, struct uasm_label **l, struct uasm_reloc **r,
* TMP and PTR are scratch.
* TMP will be clobbered, PTR will hold the pgd entry.
*/
-static void __cpuinit __maybe_unused
+static void __maybe_unused
build_get_pgde32(u32 **p, unsigned int tmp, unsigned int ptr)
{
- long pgdc = (long)pgd_current;
+ if (pgd_reg != -1) {
+ /* pgd is in pgd_reg */
+ uasm_i_mfc0(p, ptr, c0_kscratch(), pgd_reg);
+ uasm_i_mfc0(p, tmp, C0_BADVADDR); /* get faulting address */
+ } else {
+ long pgdc = (long)pgd_current;
- /* 32 bit SMP has smp_processor_id() stored in CONTEXT. */
+ /* 32 bit SMP has smp_processor_id() stored in CONTEXT. */
#ifdef CONFIG_SMP
-#ifdef CONFIG_MIPS_MT_SMTC
- /*
- * SMTC uses TCBind value as "CPU" index
- */
- uasm_i_mfc0(p, ptr, C0_TCBIND);
- UASM_i_LA_mostly(p, tmp, pgdc);
- uasm_i_srl(p, ptr, ptr, 19);
-#else
- /*
- * smp_processor_id() << 2 is stored in CONTEXT.
- */
- uasm_i_mfc0(p, ptr, C0_CONTEXT);
- UASM_i_LA_mostly(p, tmp, pgdc);
- uasm_i_srl(p, ptr, ptr, 23);
-#endif
- uasm_i_addu(p, ptr, tmp, ptr);
+ uasm_i_mfc0(p, ptr, SMP_CPUID_REG);
+ UASM_i_LA_mostly(p, tmp, pgdc);
+ uasm_i_srl(p, ptr, ptr, SMP_CPUID_PTRSHIFT);
+ uasm_i_addu(p, ptr, tmp, ptr);
#else
- UASM_i_LA_mostly(p, ptr, pgdc);
+ UASM_i_LA_mostly(p, ptr, pgdc);
#endif
- uasm_i_mfc0(p, tmp, C0_BADVADDR); /* get faulting address */
- uasm_i_lw(p, ptr, uasm_rel_lo(pgdc), ptr);
+ uasm_i_mfc0(p, tmp, C0_BADVADDR); /* get faulting address */
+ uasm_i_lw(p, ptr, uasm_rel_lo(pgdc), ptr);
+ }
uasm_i_srl(p, tmp, tmp, PGDIR_SHIFT); /* get pgd only bits */
uasm_i_sll(p, tmp, tmp, PGD_T_LOG2);
uasm_i_addu(p, ptr, ptr, tmp); /* add in pgd offset */
@@ -991,7 +951,7 @@ build_get_pgde32(u32 **p, unsigned int tmp, unsigned int ptr)
#endif /* !CONFIG_64BIT */
-static void __cpuinit build_adjust_context(u32 **p, unsigned int ctx)
+static void build_adjust_context(u32 **p, unsigned int ctx)
{
unsigned int shift = 4 - (PTE_T_LOG2 + 1) + PAGE_SHIFT - 12;
unsigned int mask = (PTRS_PER_PTE / 2 - 1) << (PTE_T_LOG2 + 1);
@@ -1017,7 +977,7 @@ static void __cpuinit build_adjust_context(u32 **p, unsigned int ctx)
uasm_i_andi(p, ctx, ctx, mask);
}
-static void __cpuinit build_get_ptep(u32 **p, unsigned int tmp, unsigned int ptr)
+static void build_get_ptep(u32 **p, unsigned int tmp, unsigned int ptr)
{
/*
* Bug workaround for the Nevada. It seems as if under certain
@@ -1042,8 +1002,7 @@ static void __cpuinit build_get_ptep(u32 **p, unsigned int tmp, unsigned int ptr
UASM_i_ADDU(p, ptr, ptr, tmp); /* add in offset */
}
-static void __cpuinit build_update_entries(u32 **p, unsigned int tmp,
- unsigned int ptep)
+static void build_update_entries(u32 **p, unsigned int tmp, unsigned int ptep)
{
/*
* 64bit address support (36bit on a 32bit CPU) in a 32bit
@@ -1104,7 +1063,7 @@ struct mips_huge_tlb_info {
int restore_scratch;
};
-static struct mips_huge_tlb_info __cpuinit
+static struct mips_huge_tlb_info
build_fast_tlb_refill_handler (u32 **p, struct uasm_label **l,
struct uasm_reloc **r, unsigned int tmp,
unsigned int ptr, int c0_scratch_reg)
@@ -1282,7 +1241,7 @@ build_fast_tlb_refill_handler (u32 **p, struct uasm_label **l,
*/
#define MIPS64_REFILL_INSNS 32
-static void __cpuinit build_r4000_tlb_refill_handler(void)
+static void build_r4000_tlb_refill_handler(void)
{
u32 *p = tlb_handler;
struct uasm_label *l = labels;
@@ -1297,7 +1256,7 @@ static void __cpuinit build_r4000_tlb_refill_handler(void)
memset(relocs, 0, sizeof(relocs));
memset(final_handler, 0, sizeof(final_handler));
- if ((scratch_reg >= 0 || scratchpad_available()) && use_bbit_insns()) {
+ if (IS_ENABLED(CONFIG_64BIT) && (scratch_reg >= 0 || scratchpad_available()) && use_bbit_insns()) {
htlb_info = build_fast_tlb_refill_handler(&p, &l, &r, K0, K1,
scratch_reg);
vmalloc_mode = refill_scratch;
@@ -1356,95 +1315,100 @@ static void __cpuinit build_r4000_tlb_refill_handler(void)
* need three, with the second nop'ed and the third being
* unused.
*/
- /* Loongson2 ebase is different than r4k, we have more space */
-#if defined(CONFIG_32BIT) || defined(CONFIG_CPU_LOONGSON2)
- if ((p - tlb_handler) > 64)
- panic("TLB refill handler space exceeded");
-#else
- if (((p - tlb_handler) > (MIPS64_REFILL_INSNS * 2) - 1)
- || (((p - tlb_handler) > (MIPS64_REFILL_INSNS * 2) - 3)
- && uasm_insn_has_bdelay(relocs,
- tlb_handler + MIPS64_REFILL_INSNS - 3)))
- panic("TLB refill handler space exceeded");
-#endif
-
- /*
- * Now fold the handler in the TLB refill handler space.
- */
-#if defined(CONFIG_32BIT) || defined(CONFIG_CPU_LOONGSON2)
- f = final_handler;
- /* Simplest case, just copy the handler. */
- uasm_copy_handler(relocs, labels, tlb_handler, p, f);
- final_len = p - tlb_handler;
-#else /* CONFIG_64BIT */
- f = final_handler + MIPS64_REFILL_INSNS;
- if ((p - tlb_handler) <= MIPS64_REFILL_INSNS) {
- /* Just copy the handler. */
- uasm_copy_handler(relocs, labels, tlb_handler, p, f);
- final_len = p - tlb_handler;
- } else {
-#ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT
- const enum label_id ls = label_tlb_huge_update;
-#else
- const enum label_id ls = label_vmalloc;
-#endif
- u32 *split;
- int ov = 0;
- int i;
-
- for (i = 0; i < ARRAY_SIZE(labels) && labels[i].lab != ls; i++)
- ;
- BUG_ON(i == ARRAY_SIZE(labels));
- split = labels[i].addr;
-
- /*
- * See if we have overflown one way or the other.
- */
- if (split > tlb_handler + MIPS64_REFILL_INSNS ||
- split < p - MIPS64_REFILL_INSNS)
- ov = 1;
-
- if (ov) {
+ switch (boot_cpu_type()) {
+ default:
+ if (sizeof(long) == 4) {
+ case CPU_LOONGSON2:
+ /* Loongson2 ebase is different than r4k, we have more space */
+ if ((p - tlb_handler) > 64)
+ panic("TLB refill handler space exceeded");
/*
- * Split two instructions before the end. One
- * for the branch and one for the instruction
- * in the delay slot.
+ * Now fold the handler in the TLB refill handler space.
*/
- split = tlb_handler + MIPS64_REFILL_INSNS - 2;
-
+ f = final_handler;
+ /* Simplest case, just copy the handler. */
+ uasm_copy_handler(relocs, labels, tlb_handler, p, f);
+ final_len = p - tlb_handler;
+ break;
+ } else {
+ if (((p - tlb_handler) > (MIPS64_REFILL_INSNS * 2) - 1)
+ || (((p - tlb_handler) > (MIPS64_REFILL_INSNS * 2) - 3)
+ && uasm_insn_has_bdelay(relocs,
+ tlb_handler + MIPS64_REFILL_INSNS - 3)))
+ panic("TLB refill handler space exceeded");
/*
- * If the branch would fall in a delay slot,
- * we must back up an additional instruction
- * so that it is no longer in a delay slot.
+ * Now fold the handler in the TLB refill handler space.
*/
- if (uasm_insn_has_bdelay(relocs, split - 1))
- split--;
- }
- /* Copy first part of the handler. */
- uasm_copy_handler(relocs, labels, tlb_handler, split, f);
- f += split - tlb_handler;
-
- if (ov) {
- /* Insert branch. */
- uasm_l_split(&l, final_handler);
- uasm_il_b(&f, &r, label_split);
- if (uasm_insn_has_bdelay(relocs, split))
- uasm_i_nop(&f);
- else {
- uasm_copy_handler(relocs, labels,
- split, split + 1, f);
- uasm_move_labels(labels, f, f + 1, -1);
- f++;
- split++;
+ f = final_handler + MIPS64_REFILL_INSNS;
+ if ((p - tlb_handler) <= MIPS64_REFILL_INSNS) {
+ /* Just copy the handler. */
+ uasm_copy_handler(relocs, labels, tlb_handler, p, f);
+ final_len = p - tlb_handler;
+ } else {
+#ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT
+ const enum label_id ls = label_tlb_huge_update;
+#else
+ const enum label_id ls = label_vmalloc;
+#endif
+ u32 *split;
+ int ov = 0;
+ int i;
+
+ for (i = 0; i < ARRAY_SIZE(labels) && labels[i].lab != ls; i++)
+ ;
+ BUG_ON(i == ARRAY_SIZE(labels));
+ split = labels[i].addr;
+
+ /*
+ * See if we have overflown one way or the other.
+ */
+ if (split > tlb_handler + MIPS64_REFILL_INSNS ||
+ split < p - MIPS64_REFILL_INSNS)
+ ov = 1;
+
+ if (ov) {
+ /*
+ * Split two instructions before the end. One
+ * for the branch and one for the instruction
+ * in the delay slot.
+ */
+ split = tlb_handler + MIPS64_REFILL_INSNS - 2;
+
+ /*
+ * If the branch would fall in a delay slot,
+ * we must back up an additional instruction
+ * so that it is no longer in a delay slot.
+ */
+ if (uasm_insn_has_bdelay(relocs, split - 1))
+ split--;
+ }
+ /* Copy first part of the handler. */
+ uasm_copy_handler(relocs, labels, tlb_handler, split, f);
+ f += split - tlb_handler;
+
+ if (ov) {
+ /* Insert branch. */
+ uasm_l_split(&l, final_handler);
+ uasm_il_b(&f, &r, label_split);
+ if (uasm_insn_has_bdelay(relocs, split))
+ uasm_i_nop(&f);
+ else {
+ uasm_copy_handler(relocs, labels,
+ split, split + 1, f);
+ uasm_move_labels(labels, f, f + 1, -1);
+ f++;
+ split++;
+ }
+ }
+
+ /* Copy the rest of the handler. */
+ uasm_copy_handler(relocs, labels, split, p, final_handler);
+ final_len = (f - (final_handler + MIPS64_REFILL_INSNS)) +
+ (p - split);
}
}
-
- /* Copy the rest of the handler. */
- uasm_copy_handler(relocs, labels, split, p, final_handler);
- final_len = (f - (final_handler + MIPS64_REFILL_INSNS)) +
- (p - split);
+ break;
}
-#endif /* CONFIG_64BIT */
uasm_resolve_relocs(relocs, labels);
pr_debug("Wrote TLB refill handler (%u instructions).\n",
@@ -1458,28 +1422,31 @@ static void __cpuinit build_r4000_tlb_refill_handler(void)
extern u32 handle_tlbl[], handle_tlbl_end[];
extern u32 handle_tlbs[], handle_tlbs_end[];
extern u32 handle_tlbm[], handle_tlbm_end[];
+extern u32 tlbmiss_handler_setup_pgd_start[], tlbmiss_handler_setup_pgd[];
+extern u32 tlbmiss_handler_setup_pgd_end[];
-#ifdef CONFIG_MIPS_PGD_C0_CONTEXT
-extern u32 tlbmiss_handler_setup_pgd[], tlbmiss_handler_setup_pgd_end[];
-
-static void __cpuinit build_r4000_setup_pgd(void)
+static void build_setup_pgd(void)
{
const int a0 = 4;
- const int a1 = 5;
- u32 *p = tlbmiss_handler_setup_pgd_array;
+ const int __maybe_unused a1 = 5;
+ const int __maybe_unused a2 = 6;
+ u32 *p = tlbmiss_handler_setup_pgd_start;
const int tlbmiss_handler_setup_pgd_size =
- tlbmiss_handler_setup_pgd_end - tlbmiss_handler_setup_pgd;
- struct uasm_label *l = labels;
- struct uasm_reloc *r = relocs;
+ tlbmiss_handler_setup_pgd_end - tlbmiss_handler_setup_pgd_start;
+#ifndef CONFIG_MIPS_PGD_C0_CONTEXT
+ long pgdc = (long)pgd_current;
+#endif
memset(tlbmiss_handler_setup_pgd, 0, tlbmiss_handler_setup_pgd_size *
sizeof(tlbmiss_handler_setup_pgd[0]));
memset(labels, 0, sizeof(labels));
memset(relocs, 0, sizeof(relocs));
-
pgd_reg = allocate_kscratch();
-
+#ifdef CONFIG_MIPS_PGD_C0_CONTEXT
if (pgd_reg == -1) {
+ struct uasm_label *l = labels;
+ struct uasm_reloc *r = relocs;
+
/* PGD << 11 in c0_Context */
/*
* If it is a ckseg0 address, convert to a physical
@@ -1501,6 +1468,26 @@ static void __cpuinit build_r4000_setup_pgd(void)
uasm_i_jr(&p, 31);
UASM_i_MTC0(&p, a0, c0_kscratch(), pgd_reg);
}
+#else
+#ifdef CONFIG_SMP
+ /* Save PGD to pgd_current[smp_processor_id()] */
+ UASM_i_CPUID_MFC0(&p, a1, SMP_CPUID_REG);
+ UASM_i_SRL_SAFE(&p, a1, a1, SMP_CPUID_PTRSHIFT);
+ UASM_i_LA_mostly(&p, a2, pgdc);
+ UASM_i_ADDU(&p, a2, a2, a1);
+ UASM_i_SW(&p, a0, uasm_rel_lo(pgdc), a2);
+#else
+ UASM_i_LA_mostly(&p, a2, pgdc);
+ UASM_i_SW(&p, a0, uasm_rel_lo(pgdc), a2);
+#endif /* SMP */
+ uasm_i_jr(&p, 31);
+
+ /* if pgd_reg is allocated, save PGD also to scratch register */
+ if (pgd_reg != -1)
+ UASM_i_MTC0(&p, a0, c0_kscratch(), pgd_reg);
+ else
+ uasm_i_nop(&p);
+#endif
if (p >= tlbmiss_handler_setup_pgd_end)
panic("tlbmiss_handler_setup_pgd space exceeded");
@@ -1511,9 +1498,8 @@ static void __cpuinit build_r4000_setup_pgd(void)
dump_handler("tlbmiss_handler", tlbmiss_handler_setup_pgd,
tlbmiss_handler_setup_pgd_size);
}
-#endif
-static void __cpuinit
+static void
iPTE_LW(u32 **p, unsigned int pte, unsigned int ptr)
{
#ifdef CONFIG_SMP
@@ -1533,7 +1519,7 @@ iPTE_LW(u32 **p, unsigned int pte, unsigned int ptr)
#endif
}
-static void __cpuinit
+static void
iPTE_SW(u32 **p, struct uasm_reloc **r, unsigned int pte, unsigned int ptr,
unsigned int mode)
{
@@ -1593,7 +1579,7 @@ iPTE_SW(u32 **p, struct uasm_reloc **r, unsigned int pte, unsigned int ptr,
* the page table where this PTE is located, PTE will be re-loaded
* with it's original value.
*/
-static void __cpuinit
+static void
build_pte_present(u32 **p, struct uasm_reloc **r,
int pte, int ptr, int scratch, enum label_id lid)
{
@@ -1621,7 +1607,7 @@ build_pte_present(u32 **p, struct uasm_reloc **r,
}
/* Make PTE valid, store result in PTR. */
-static void __cpuinit
+static void
build_make_valid(u32 **p, struct uasm_reloc **r, unsigned int pte,
unsigned int ptr)
{
@@ -1634,7 +1620,7 @@ build_make_valid(u32 **p, struct uasm_reloc **r, unsigned int pte,
* Check if PTE can be written to, if not branch to LABEL. Regardless
* restore PTE with value from PTR when done.
*/
-static void __cpuinit
+static void
build_pte_writable(u32 **p, struct uasm_reloc **r,
unsigned int pte, unsigned int ptr, int scratch,
enum label_id lid)
@@ -1654,7 +1640,7 @@ build_pte_writable(u32 **p, struct uasm_reloc **r,
/* Make PTE writable, update software status bits as well, then store
* at PTR.
*/
-static void __cpuinit
+static void
build_make_write(u32 **p, struct uasm_reloc **r, unsigned int pte,
unsigned int ptr)
{
@@ -1668,7 +1654,7 @@ build_make_write(u32 **p, struct uasm_reloc **r, unsigned int pte,
* Check if PTE can be modified, if not branch to LABEL. Regardless
* restore PTE with value from PTR when done.
*/
-static void __cpuinit
+static void
build_pte_modifiable(u32 **p, struct uasm_reloc **r,
unsigned int pte, unsigned int ptr, int scratch,
enum label_id lid)
@@ -1697,7 +1683,7 @@ build_pte_modifiable(u32 **p, struct uasm_reloc **r,
* This places the pte into ENTRYLO0 and writes it with tlbwi.
* Then it returns.
*/
-static void __cpuinit
+static void
build_r3000_pte_reload_tlbwi(u32 **p, unsigned int pte, unsigned int tmp)
{
uasm_i_mtc0(p, pte, C0_ENTRYLO0); /* cp0 delay */
@@ -1713,7 +1699,7 @@ build_r3000_pte_reload_tlbwi(u32 **p, unsigned int pte, unsigned int tmp)
* may have the probe fail bit set as a result of a trap on a
* kseg2 access, i.e. without refill. Then it returns.
*/
-static void __cpuinit
+static void
build_r3000_tlb_reload_write(u32 **p, struct uasm_label **l,
struct uasm_reloc **r, unsigned int pte,
unsigned int tmp)
@@ -1731,7 +1717,7 @@ build_r3000_tlb_reload_write(u32 **p, struct uasm_label **l,
uasm_i_rfe(p); /* branch delay */
}
-static void __cpuinit
+static void
build_r3000_tlbchange_handler_head(u32 **p, unsigned int pte,
unsigned int ptr)
{
@@ -1751,7 +1737,7 @@ build_r3000_tlbchange_handler_head(u32 **p, unsigned int pte,
uasm_i_tlbp(p); /* load delay */
}
-static void __cpuinit build_r3000_tlb_load_handler(void)
+static void build_r3000_tlb_load_handler(void)
{
u32 *p = handle_tlbl;
const int handle_tlbl_size = handle_tlbl_end - handle_tlbl;
@@ -1782,7 +1768,7 @@ static void __cpuinit build_r3000_tlb_load_handler(void)
dump_handler("r3000_tlb_load", handle_tlbl, handle_tlbl_size);
}
-static void __cpuinit build_r3000_tlb_store_handler(void)
+static void build_r3000_tlb_store_handler(void)
{
u32 *p = handle_tlbs;
const int handle_tlbs_size = handle_tlbs_end - handle_tlbs;
@@ -1803,7 +1789,7 @@ static void __cpuinit build_r3000_tlb_store_handler(void)
uasm_i_j(&p, (unsigned long)tlb_do_page_fault_1 & 0x0fffffff);
uasm_i_nop(&p);
- if (p >= handle_tlbs)
+ if (p >= handle_tlbs_end)
panic("TLB store handler fastpath space exceeded");
uasm_resolve_relocs(relocs, labels);
@@ -1813,7 +1799,7 @@ static void __cpuinit build_r3000_tlb_store_handler(void)
dump_handler("r3000_tlb_store", handle_tlbs, handle_tlbs_size);
}
-static void __cpuinit build_r3000_tlb_modify_handler(void)
+static void build_r3000_tlb_modify_handler(void)
{
u32 *p = handle_tlbm;
const int handle_tlbm_size = handle_tlbm_end - handle_tlbm;
@@ -1848,7 +1834,7 @@ static void __cpuinit build_r3000_tlb_modify_handler(void)
/*
* R4000 style TLB load/store/modify handlers.
*/
-static struct work_registers __cpuinit
+static struct work_registers
build_r4000_tlbchange_handler_head(u32 **p, struct uasm_label **l,
struct uasm_reloc **r)
{
@@ -1884,7 +1870,7 @@ build_r4000_tlbchange_handler_head(u32 **p, struct uasm_label **l,
return wr;
}
-static void __cpuinit
+static void
build_r4000_tlbchange_handler_tail(u32 **p, struct uasm_label **l,
struct uasm_reloc **r, unsigned int tmp,
unsigned int ptr)
@@ -1902,7 +1888,7 @@ build_r4000_tlbchange_handler_tail(u32 **p, struct uasm_label **l,
#endif
}
-static void __cpuinit build_r4000_tlb_load_handler(void)
+static void build_r4000_tlb_load_handler(void)
{
u32 *p = handle_tlbl;
const int handle_tlbl_size = handle_tlbl_end - handle_tlbl;
@@ -2085,7 +2071,7 @@ static void __cpuinit build_r4000_tlb_load_handler(void)
dump_handler("r4000_tlb_load", handle_tlbl, handle_tlbl_size);
}
-static void __cpuinit build_r4000_tlb_store_handler(void)
+static void build_r4000_tlb_store_handler(void)
{
u32 *p = handle_tlbs;
const int handle_tlbs_size = handle_tlbs_end - handle_tlbs;
@@ -2140,7 +2126,7 @@ static void __cpuinit build_r4000_tlb_store_handler(void)
dump_handler("r4000_tlb_store", handle_tlbs, handle_tlbs_size);
}
-static void __cpuinit build_r4000_tlb_modify_handler(void)
+static void build_r4000_tlb_modify_handler(void)
{
u32 *p = handle_tlbm;
const int handle_tlbm_size = handle_tlbm_end - handle_tlbm;
@@ -2196,7 +2182,7 @@ static void __cpuinit build_r4000_tlb_modify_handler(void)
dump_handler("r4000_tlb_modify", handle_tlbm, handle_tlbm_size);
}
-static void __cpuinit flush_tlb_handlers(void)
+static void flush_tlb_handlers(void)
{
local_flush_icache_range((unsigned long)handle_tlbl,
(unsigned long)handle_tlbl_end);
@@ -2204,13 +2190,11 @@ static void __cpuinit flush_tlb_handlers(void)
(unsigned long)handle_tlbs_end);
local_flush_icache_range((unsigned long)handle_tlbm,
(unsigned long)handle_tlbm_end);
-#ifdef CONFIG_MIPS_PGD_C0_CONTEXT
local_flush_icache_range((unsigned long)tlbmiss_handler_setup_pgd,
(unsigned long)tlbmiss_handler_setup_pgd_end);
-#endif
}
-void __cpuinit build_tlb_refill_handler(void)
+void build_tlb_refill_handler(void)
{
/*
* The refill handler is generated per-CPU, multi-node systems
@@ -2239,6 +2223,7 @@ void __cpuinit build_tlb_refill_handler(void)
if (!run_once) {
if (!cpu_has_local_ebase)
build_r3000_tlb_refill_handler();
+ build_setup_pgd();
build_r3000_tlb_load_handler();
build_r3000_tlb_store_handler();
build_r3000_tlb_modify_handler();
@@ -2262,9 +2247,7 @@ void __cpuinit build_tlb_refill_handler(void)
default:
if (!run_once) {
scratch_reg = allocate_kscratch();
-#ifdef CONFIG_MIPS_PGD_C0_CONTEXT
- build_r4000_setup_pgd();
-#endif
+ build_setup_pgd();
build_r4000_tlb_load_handler();
build_r4000_tlb_store_handler();
build_r4000_tlb_modify_handler();
generated by cgit v1.2.3-70-g09d2 (git 2.46.0) at 2025-11-10 04:14:30 +0000