diff options
Diffstat (limited to 'arch/mips/mm/tlbex.c')
| -rw-r--r-- | arch/mips/mm/tlbex.c | 1130 |
1 files changed, 666 insertions, 464 deletions
diff --git a/arch/mips/mm/tlbex.c b/arch/mips/mm/tlbex.c index 083d3412d0b..e80e10bafc8 100644 --- a/arch/mips/mm/tlbex.c +++ b/arch/mips/mm/tlbex.c @@ -5,10 +5,11 @@ * * Synthesize TLB refill handlers at runtime. * - * Copyright (C) 2004, 2005, 2006, 2008 Thiemo Seufer - * Copyright (C) 2005, 2007, 2008, 2009 Maciej W. Rozycki + * Copyright (C) 2004, 2005, 2006, 2008 Thiemo Seufer + * Copyright (C) 2005, 2007, 2008, 2009 Maciej W. Rozycki * Copyright (C) 2006 Ralf Baechle (ralf@linux-mips.org) * Copyright (C) 2008, 2009 Cavium Networks, Inc. + * Copyright (C) 2011 MIPS Technologies, Inc. * * ... and the days got worse and worse and now you see * I've gone completly out of my mind. @@ -25,13 +26,14 @@ #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> +#include <asm/setup.h> /* * TLB load/store/modify handlers. @@ -42,6 +44,18 @@ extern void tlb_do_page_fault_0(void); extern void tlb_do_page_fault_1(void); +struct work_registers { + int r1; + int r2; + int r3; +}; + +struct tlb_reg_save { + unsigned long a; + unsigned long b; +} ____cacheline_aligned_in_smp; + +static struct tlb_reg_save handler_reg_save[NR_CPUS]; static inline int r45k_bvahwbug(void) { @@ -71,6 +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; @@ -81,6 +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; @@ -109,6 +125,8 @@ static bool scratchpad_available(void) static int scratchpad_offset(int i) { BUG(); + /* Really unreachable, but evidently some GCC want this. */ + return 0; } #endif /* @@ -120,7 +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); @@ -132,8 +150,8 @@ enum label_id { label_leave, label_vmalloc, label_vmalloc_done, - label_tlbw_hazard, - label_split, + label_tlbw_hazard_0, + label_split = label_tlbw_hazard_0 + 8, label_tlbl_goaround1, label_tlbl_goaround2, label_nopage_tlbl, @@ -142,7 +160,7 @@ enum label_id { label_smp_pgtable_change, label_r3000_write_probe_fail, label_large_segbits_fault, -#ifdef CONFIG_HUGETLB_PAGE +#ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT label_tlb_huge_update, #endif }; @@ -151,7 +169,7 @@ UASM_L_LA(_second_part) UASM_L_LA(_leave) UASM_L_LA(_vmalloc) UASM_L_LA(_vmalloc_done) -UASM_L_LA(_tlbw_hazard) +/* _tlbw_hazard_x is handled differently. */ UASM_L_LA(_split) UASM_L_LA(_tlbl_goaround1) UASM_L_LA(_tlbl_goaround2) @@ -161,24 +179,88 @@ UASM_L_LA(_nopage_tlbm) UASM_L_LA(_smp_pgtable_change) UASM_L_LA(_r3000_write_probe_fail) UASM_L_LA(_large_segbits_fault) -#ifdef CONFIG_HUGETLB_PAGE +#ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT UASM_L_LA(_tlb_huge_update) #endif +static int hazard_instance; + +static void uasm_bgezl_hazard(u32 **p, struct uasm_reloc **r, int instance) +{ + switch (instance) { + case 0 ... 7: + uasm_il_bgezl(p, r, 0, label_tlbw_hazard_0 + instance); + return; + default: + BUG(); + } +} + +static void uasm_bgezl_label(struct uasm_label **l, u32 **p, int instance) +{ + switch (instance) { + case 0 ... 7: + uasm_build_label(l, *p, label_tlbw_hazard_0 + instance); + break; + default: + BUG(); + } +} + /* - * For debug purposes. + * pgtable bits are assigned dynamically depending on processor feature + * and statically based on kernel configuration. This spits out the actual + * values the kernel is using. Required to make sense from disassembled + * TLB exception handlers. */ -static inline void dump_handler(const u32 *handler, int count) +static void output_pgtable_bits_defines(void) +{ +#define pr_define(fmt, ...) \ + pr_debug("#define " fmt, ##__VA_ARGS__) + + pr_debug("#include <asm/asm.h>\n"); + pr_debug("#include <asm/regdef.h>\n"); + pr_debug("\n"); + + pr_define("_PAGE_PRESENT_SHIFT %d\n", _PAGE_PRESENT_SHIFT); + pr_define("_PAGE_READ_SHIFT %d\n", _PAGE_READ_SHIFT); + pr_define("_PAGE_WRITE_SHIFT %d\n", _PAGE_WRITE_SHIFT); + pr_define("_PAGE_ACCESSED_SHIFT %d\n", _PAGE_ACCESSED_SHIFT); + pr_define("_PAGE_MODIFIED_SHIFT %d\n", _PAGE_MODIFIED_SHIFT); +#ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT + pr_define("_PAGE_HUGE_SHIFT %d\n", _PAGE_HUGE_SHIFT); + pr_define("_PAGE_SPLITTING_SHIFT %d\n", _PAGE_SPLITTING_SHIFT); +#endif + if (cpu_has_rixi) { +#ifdef _PAGE_NO_EXEC_SHIFT + pr_define("_PAGE_NO_EXEC_SHIFT %d\n", _PAGE_NO_EXEC_SHIFT); +#endif +#ifdef _PAGE_NO_READ_SHIFT + pr_define("_PAGE_NO_READ_SHIFT %d\n", _PAGE_NO_READ_SHIFT); +#endif + } + pr_define("_PAGE_GLOBAL_SHIFT %d\n", _PAGE_GLOBAL_SHIFT); + pr_define("_PAGE_VALID_SHIFT %d\n", _PAGE_VALID_SHIFT); + pr_define("_PAGE_DIRTY_SHIFT %d\n", _PAGE_DIRTY_SHIFT); + pr_define("_PFN_SHIFT %d\n", _PFN_SHIFT); + pr_debug("\n"); +} + +static inline void dump_handler(const char *symbol, const u32 *handler, int count) { int i; + pr_debug("LEAF(%s)\n", symbol); + pr_debug("\t.set push\n"); pr_debug("\t.set noreorder\n"); for (i = 0; i < count; i++) - pr_debug("\t%p\t.word 0x%08x\n", &handler[i], handler[i]); + pr_debug("\t.word\t0x%08x\t\t# %p\n", handler[i], &handler[i]); - pr_debug("\t.set pop\n"); + pr_debug("\t.set\tpop\n"); + + pr_debug("\tEND(%s)\n", symbol); } /* The only general purpose registers allowed in TLB handlers. */ @@ -211,21 +293,28 @@ static inline void dump_handler(const u32 *handler, int count) * 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]; -#ifdef CONFIG_64BIT -static int check_for_high_segbits __cpuinitdata; -#endif +static int check_for_high_segbits; -static int check_for_high_segbits __cpuinitdata; +static unsigned int kscratch_used_mask; -static unsigned int kscratch_used_mask __cpuinitdata; +static inline int __maybe_unused c0_kscratch(void) +{ + switch (current_cpu_type()) { + case CPU_XLP: + case CPU_XLR: + return 22; + default: + return 31; + } +} -static int __cpuinit allocate_kscratch(void) +static int allocate_kscratch(void) { int r; unsigned int a = cpu_data[0].kscratch_mask & ~kscratch_used_mask; @@ -242,10 +331,57 @@ 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 build_get_work_registers(u32 **p) +{ + struct work_registers r; + + if (scratch_reg >= 0) { + /* Save in CPU local C0_KScratch? */ + UASM_i_MTC0(p, 1, c0_kscratch(), scratch_reg); + r.r1 = K0; + r.r2 = K1; + r.r3 = 1; + return r; + } + + if (num_possible_cpus() > 1) { + /* Get smp_processor_id */ + 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))); + + UASM_i_LA(p, K1, (long)&handler_reg_save); + UASM_i_ADDU(p, K0, K0, K1); + } else { + UASM_i_LA(p, K0, (long)&handler_reg_save); + } + /* K0 now points to save area, save $1 and $2 */ + UASM_i_SW(p, 1, offsetof(struct tlb_reg_save, a), K0); + UASM_i_SW(p, 2, offsetof(struct tlb_reg_save, b), K0); + + r.r1 = K1; + r.r2 = 1; + r.r3 = 2; + return r; +} + +static void build_restore_work_registers(u32 **p) +{ + if (scratch_reg >= 0) { + UASM_i_MFC0(p, 1, c0_kscratch(), scratch_reg); + return; + } + /* K0 already points to save area, restore $1 and $2 */ + UASM_i_LW(p, 1, offsetof(struct tlb_reg_save, a), K0); + UASM_i_LW(p, 2, offsetof(struct tlb_reg_save, b), K0); +} + #ifndef CONFIG_MIPS_PGD_C0_CONTEXT /* @@ -260,7 +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; @@ -294,7 +430,7 @@ static void __cpuinit build_r3000_tlb_refill_handler(void) memcpy((void *)ebase, tlb_handler, 0x80); - dump_handler((u32 *)ebase, 32); + dump_handler("r3000_tlb_refill", (u32 *)ebase, 32); } #endif /* CONFIG_MIPS_PGD_C0_CONTEXT */ @@ -305,7 +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 @@ -313,8 +449,8 @@ static u32 final_handler[64] __cpuinitdata; * From the IDT errata for the QED RM5230 (Nevada), processor revision 1.0: * 2. A timing hazard exists for the TLBP instruction. * - * stalling_instruction - * TLBP + * stalling_instruction + * TLBP * * The JTLB is being read for the TLBP throughout the stall generated by the * previous instruction. This is not really correct as the stalling instruction @@ -325,18 +461,17 @@ static u32 final_handler[64] __cpuinitdata; * The software work-around is to not allow the instruction preceding the TLBP * to stall - make it an NOP or some other instruction guaranteed not to stall. * - * Errata 2 will not be fixed. This errata is also on the R5000. + * Errata 2 will not be fixed. This errata is also on the R5000. * * 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. */ case CPU_R4600: case CPU_R4700: case CPU_R5000: - case CPU_R5000A: case CPU_NEVADA: uasm_i_nop(p); uasm_i_tlbp(p); @@ -350,13 +485,13 @@ static void __cpuinit __maybe_unused build_tlb_probe_entry(u32 **p) /* * Write random or indexed TLB entry, and care about the hazards from - * the preceeding mtc0 and for the following eret. + * the preceding mtc0 and for the following eret. */ 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; @@ -366,8 +501,24 @@ static void __cpuinit build_tlb_write_entry(u32 **p, struct uasm_label **l, } if (cpu_has_mips_r2) { - if (cpu_has_mips_r2_exec_hazard) + /* + * The architecture spec says an ehb is required here, + * but a number of cores do not have the hazard and + * using an ehb causes an expensive pipeline stall. + */ + switch (current_cpu_type()) { + case CPU_M14KC: + case CPU_74K: + case CPU_1074K: + case CPU_PROAPTIV: + case CPU_P5600: + case CPU_M5150: + break; + + default: uasm_i_ehb(p); + break; + } tlbw(p); return; } @@ -383,25 +534,32 @@ static void __cpuinit build_tlb_write_entry(u32 **p, struct uasm_label **l, * This branch uses up a mtc0 hazard nop slot and saves * two nops after the tlbw instruction. */ - uasm_il_bgezl(p, r, 0, label_tlbw_hazard); + uasm_bgezl_hazard(p, r, hazard_instance); tlbw(p); - uasm_l_tlbw_hazard(l, *p); + uasm_bgezl_label(l, p, hazard_instance); + hazard_instance++; uasm_i_nop(p); break; case CPU_R4600: case CPU_R4700: - case CPU_R5000: - case CPU_R5000A: uasm_i_nop(p); tlbw(p); uasm_i_nop(p); break; + case CPU_R5000: + case CPU_NEVADA: + uasm_i_nop(p); /* QED specifies 2 nops hazard */ + uasm_i_nop(p); /* QED specifies 2 nops hazard */ + tlbw(p); + break; + case CPU_R4300: case CPU_5KC: case CPU_TX49XX: case CPU_PR4450: + case CPU_XLR: uasm_i_nop(p); tlbw(p); break; @@ -411,6 +569,8 @@ static void __cpuinit build_tlb_write_entry(u32 **p, struct uasm_label **l, case CPU_R14000: case CPU_4KC: case CPU_4KEC: + case CPU_M14KC: + case CPU_M14KEC: case CPU_SB1: case CPU_SB1A: case CPU_4KSC: @@ -422,6 +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); @@ -429,17 +590,6 @@ static void __cpuinit build_tlb_write_entry(u32 **p, struct uasm_label **l, tlbw(p); break; - case CPU_NEVADA: - uasm_i_nop(p); /* QED specifies 2 nops hazard */ - /* - * This branch uses up a mtc0 hazard nop slot and saves - * a nop after the tlbw instruction. - */ - uasm_il_bgezl(p, r, 0, label_tlbw_hazard); - tlbw(p); - uasm_l_tlbw_hazard(l, *p); - break; - case CPU_RM7000: uasm_i_nop(p); uasm_i_nop(p); @@ -448,24 +598,6 @@ static void __cpuinit build_tlb_write_entry(u32 **p, struct uasm_label **l, tlbw(p); break; - case CPU_RM9000: - /* - * When the JTLB is updated by tlbwi or tlbwr, a subsequent - * use of the JTLB for instructions should not occur for 4 - * cpu cycles and use for data translations should not occur - * for 3 cpu cycles. - */ - uasm_i_ssnop(p); - uasm_i_ssnop(p); - uasm_i_ssnop(p); - uasm_i_ssnop(p); - tlbw(p); - uasm_i_ssnop(p); - uasm_i_ssnop(p); - uasm_i_ssnop(p); - uasm_i_ssnop(p); - break; - case CPU_VR4111: case CPU_VR4121: case CPU_VR4122: @@ -493,17 +625,16 @@ 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 (kernel_uses_smartmips_rixi) { - UASM_i_SRL(p, reg, reg, ilog2(_PAGE_NO_EXEC)); - UASM_i_ROTR(p, reg, reg, ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC)); + if (cpu_has_rixi) { + UASM_i_ROTR(p, reg, reg, ilog2(_PAGE_GLOBAL)); } else { #ifdef CONFIG_64BIT_PHYS_ADDR uasm_i_dsrl_safe(p, reg, reg, ilog2(_PAGE_GLOBAL)); @@ -513,13 +644,11 @@ static __cpuinit __maybe_unused void build_convert_pte_to_entrylo(u32 **p, } } -#ifdef CONFIG_HUGETLB_PAGE +#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 */ @@ -536,8 +665,8 @@ static __cpuinit void build_restore_pagemask(u32 **p, uasm_i_mtc0(p, 0, C0_PAGEMASK); uasm_il_b(p, r, lid); } - if (scratch_reg > 0) - UASM_i_MFC0(p, 1, 31, scratch_reg); + if (scratch_reg >= 0) + UASM_i_MFC0(p, 1, c0_kscratch(), scratch_reg); else UASM_i_LW(p, 1, scratchpad_offset(0), 0); } else { @@ -558,12 +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); @@ -578,9 +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()) { @@ -591,9 +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; @@ -608,7 +735,7 @@ static __cpuinit void build_huge_update_entries(u32 **p, */ small_sequence = (HPAGE_SIZE >> 7) < 0x10000; - /* We can clobber tmp. It isn't used after this.*/ + /* We can clobber tmp. It isn't used after this.*/ if (!small_sequence) uasm_i_lui(p, tmp, HPAGE_SIZE >> (7 + 16)); @@ -623,11 +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); @@ -639,14 +765,14 @@ static __cpuinit void build_huge_handler_tail(u32 **p, build_huge_update_entries(p, pte, ptr); build_huge_tlb_write_entry(p, l, r, pte, tlb_indexed, 0); } -#endif /* CONFIG_HUGETLB_PAGE */ +#endif /* CONFIG_MIPS_HUGE_TLB_SUPPORT */ #ifdef CONFIG_64BIT /* * 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) { @@ -677,11 +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, 31, 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]. */ @@ -690,33 +816,21 @@ build_get_pmde64(u32 **p, struct uasm_label **l, struct uasm_reloc **r, /* Clear lower 23 bits of context. */ uasm_i_dins(p, ptr, 0, 0, 23); - /* 1 0 1 0 1 << 6 xkphys cached */ + /* 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); @@ -738,7 +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) @@ -792,8 +906,8 @@ build_get_pgd_vmalloc64(u32 **p, struct uasm_label **l, struct uasm_reloc **r, uasm_i_jr(p, ptr); if (mode == refill_scratch) { - if (scratch_reg > 0) - UASM_i_MFC0(p, 1, 31, scratch_reg); + if (scratch_reg >= 0) + UASM_i_MFC0(p, 1, c0_kscratch(), scratch_reg); else UASM_i_LW(p, 1, scratchpad_offset(0), 0); } else { @@ -808,34 +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() << 3 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 */ @@ -843,7 +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); @@ -869,7 +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 @@ -894,8 +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 @@ -905,12 +1012,10 @@ static void __cpuinit build_update_entries(u32 **p, unsigned int tmp, if (cpu_has_64bits) { uasm_i_ld(p, tmp, 0, ptep); /* get even pte */ uasm_i_ld(p, ptep, sizeof(pte_t), ptep); /* get odd pte */ - if (kernel_uses_smartmips_rixi) { - UASM_i_SRL(p, tmp, tmp, ilog2(_PAGE_NO_EXEC)); - UASM_i_SRL(p, ptep, ptep, ilog2(_PAGE_NO_EXEC)); - UASM_i_ROTR(p, tmp, tmp, ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC)); + if (cpu_has_rixi) { + UASM_i_ROTR(p, tmp, tmp, ilog2(_PAGE_GLOBAL)); UASM_i_MTC0(p, tmp, C0_ENTRYLO0); /* load it */ - UASM_i_ROTR(p, ptep, ptep, ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC)); + UASM_i_ROTR(p, ptep, ptep, ilog2(_PAGE_GLOBAL)); } else { uasm_i_dsrl_safe(p, tmp, tmp, ilog2(_PAGE_GLOBAL)); /* convert to entrylo0 */ UASM_i_MTC0(p, tmp, C0_ENTRYLO0); /* load it */ @@ -932,14 +1037,12 @@ static void __cpuinit build_update_entries(u32 **p, unsigned int tmp, UASM_i_LW(p, ptep, sizeof(pte_t), ptep); /* get odd pte */ if (r45k_bvahwbug()) build_tlb_probe_entry(p); - if (kernel_uses_smartmips_rixi) { - UASM_i_SRL(p, tmp, tmp, ilog2(_PAGE_NO_EXEC)); - UASM_i_SRL(p, ptep, ptep, ilog2(_PAGE_NO_EXEC)); - UASM_i_ROTR(p, tmp, tmp, ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC)); + if (cpu_has_rixi) { + UASM_i_ROTR(p, tmp, tmp, ilog2(_PAGE_GLOBAL)); if (r4k_250MHZhwbug()) UASM_i_MTC0(p, 0, C0_ENTRYLO0); UASM_i_MTC0(p, tmp, C0_ENTRYLO0); /* load it */ - UASM_i_ROTR(p, ptep, ptep, ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC)); + UASM_i_ROTR(p, ptep, ptep, ilog2(_PAGE_GLOBAL)); } else { UASM_i_SRL(p, tmp, tmp, ilog2(_PAGE_GLOBAL)); /* convert to entrylo0 */ if (r4k_250MHZhwbug()) @@ -960,10 +1063,10 @@ 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) + unsigned int ptr, int c0_scratch_reg) { struct mips_huge_tlb_info rv; unsigned int even, odd; @@ -977,12 +1080,12 @@ build_fast_tlb_refill_handler (u32 **p, struct uasm_label **l, UASM_i_MFC0(p, tmp, C0_BADVADDR); if (pgd_reg != -1) - UASM_i_MFC0(p, ptr, 31, pgd_reg); + UASM_i_MFC0(p, ptr, c0_kscratch(), pgd_reg); else UASM_i_MFC0(p, ptr, C0_CONTEXT); - if (c0_scratch >= 0) - UASM_i_MTC0(p, scratch, 31, c0_scratch); + if (c0_scratch_reg >= 0) + UASM_i_MTC0(p, scratch, c0_kscratch(), c0_scratch_reg); else UASM_i_SW(p, scratch, scratchpad_offset(0), 0); @@ -997,14 +1100,14 @@ build_fast_tlb_refill_handler (u32 **p, struct uasm_label **l, } } else { if (pgd_reg != -1) - UASM_i_MFC0(p, ptr, 31, pgd_reg); + UASM_i_MFC0(p, ptr, c0_kscratch(), pgd_reg); else UASM_i_MFC0(p, ptr, C0_CONTEXT); UASM_i_MFC0(p, tmp, C0_BADVADDR); - if (c0_scratch >= 0) - UASM_i_MTC0(p, scratch, 31, c0_scratch); + if (c0_scratch_reg >= 0) + UASM_i_MTC0(p, scratch, c0_kscratch(), c0_scratch_reg); else UASM_i_SW(p, scratch, scratchpad_offset(0), 0); @@ -1017,7 +1120,7 @@ build_fast_tlb_refill_handler (u32 **p, struct uasm_label **l, if (pgd_reg == -1) { vmalloc_branch_delay_filled = 1; - /* 1 0 1 0 1 << 6 xkphys cached */ + /* 1 0 1 0 1 << 6 xkphys cached */ uasm_i_ori(p, ptr, ptr, 0x540); uasm_i_drotr(p, ptr, ptr, 11); } @@ -1035,9 +1138,9 @@ build_fast_tlb_refill_handler (u32 **p, struct uasm_label **l, uasm_l_vmalloc_done(l, *p); /* - * tmp ptr - * fall-through case = badvaddr *pgd_current - * vmalloc case = badvaddr swapper_pg_dir + * tmp ptr + * fall-through case = badvaddr *pgd_current + * vmalloc case = badvaddr swapper_pg_dir */ if (vmalloc_branch_delay_filled) @@ -1072,16 +1175,16 @@ build_fast_tlb_refill_handler (u32 **p, struct uasm_label **l, /* Adjust the context during the load latency. */ build_adjust_context(p, tmp); -#ifdef CONFIG_HUGETLB_PAGE +#ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT uasm_il_bbit1(p, r, scratch, ilog2(_PAGE_HUGE), label_tlb_huge_update); /* * The in the LWX case we don't want to do the load in the - * delay slot. It cannot issue in the same cycle and may be + * delay slot. It cannot issue in the same cycle and may be * speculative and unneeded. */ if (use_lwx_insns()) uasm_i_nop(p); -#endif /* CONFIG_HUGETLB_PAGE */ +#endif /* CONFIG_MIPS_HUGE_TLB_SUPPORT */ /* build_update_entries */ @@ -1098,14 +1201,10 @@ build_fast_tlb_refill_handler (u32 **p, struct uasm_label **l, UASM_i_LW(p, even, 0, ptr); /* get even pte */ UASM_i_LW(p, odd, sizeof(pte_t), ptr); /* get odd pte */ } - if (kernel_uses_smartmips_rixi) { - uasm_i_dsrl_safe(p, even, even, ilog2(_PAGE_NO_EXEC)); - uasm_i_dsrl_safe(p, odd, odd, ilog2(_PAGE_NO_EXEC)); - uasm_i_drotr(p, even, even, - ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC)); + if (cpu_has_rixi) { + uasm_i_drotr(p, even, even, ilog2(_PAGE_GLOBAL)); UASM_i_MTC0(p, even, C0_ENTRYLO0); /* load it */ - uasm_i_drotr(p, odd, odd, - ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC)); + uasm_i_drotr(p, odd, odd, ilog2(_PAGE_GLOBAL)); } else { uasm_i_dsrl_safe(p, even, even, ilog2(_PAGE_GLOBAL)); UASM_i_MTC0(p, even, C0_ENTRYLO0); /* load it */ @@ -1113,8 +1212,8 @@ build_fast_tlb_refill_handler (u32 **p, struct uasm_label **l, } UASM_i_MTC0(p, odd, C0_ENTRYLO1); /* load it */ - if (c0_scratch >= 0) { - UASM_i_MFC0(p, scratch, 31, c0_scratch); + if (c0_scratch_reg >= 0) { + UASM_i_MFC0(p, scratch, c0_kscratch(), c0_scratch_reg); build_tlb_write_entry(p, l, r, tlb_random); uasm_l_leave(l, *p); rv.restore_scratch = 1; @@ -1142,25 +1241,22 @@ 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; struct uasm_reloc *r = relocs; u32 *f; unsigned int final_len; - struct mips_huge_tlb_info htlb_info; - enum vmalloc64_mode vmalloc_mode; + struct mips_huge_tlb_info htlb_info __maybe_unused; + enum vmalloc64_mode vmalloc_mode __maybe_unused; memset(tlb_handler, 0, sizeof(tlb_handler)); memset(labels, 0, sizeof(labels)); memset(relocs, 0, sizeof(relocs)); memset(final_handler, 0, sizeof(final_handler)); - if (scratch_reg == 0) - scratch_reg = allocate_kscratch(); - - 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; @@ -1191,7 +1287,7 @@ static void __cpuinit build_r4000_tlb_refill_handler(void) build_get_pgde32(&p, K0, K1); /* get pgd in K1 */ #endif -#ifdef CONFIG_HUGETLB_PAGE +#ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT build_is_huge_pte(&p, &r, K0, K1, label_tlb_huge_update); #endif @@ -1201,7 +1297,7 @@ static void __cpuinit build_r4000_tlb_refill_handler(void) uasm_l_leave(&l, p); uasm_i_eret(&p); /* return from trap */ } -#ifdef CONFIG_HUGETLB_PAGE +#ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT uasm_l_tlb_huge_update(&l, p); build_huge_update_entries(&p, htlb_info.huge_pte, K1); build_huge_tlb_write_entry(&p, &l, &r, K0, tlb_random, @@ -1219,95 +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 { -#if defined(CONFIG_HUGETLB_PAGE) - 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", @@ -1315,36 +1416,37 @@ static void __cpuinit build_r4000_tlb_refill_handler(void) memcpy((void *)ebase, final_handler, 0x100); - dump_handler((u32 *)ebase, 64); + dump_handler("r4000_tlb_refill", (u32 *)ebase, 64); } -/* - * 128 instructions for the fastpath handler is generous and should - * never be exceeded. - */ -#define FASTPATH_SIZE 128 - -u32 handle_tlbl[FASTPATH_SIZE] __cacheline_aligned; -u32 handle_tlbs[FASTPATH_SIZE] __cacheline_aligned; -u32 handle_tlbm[FASTPATH_SIZE] __cacheline_aligned; -#ifdef CONFIG_MIPS_PGD_C0_CONTEXT -u32 tlbmiss_handler_setup_pgd[16] __cacheline_aligned; +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[]; -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; - struct uasm_label *l = labels; - struct uasm_reloc *r = relocs; + 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_start; +#ifndef CONFIG_MIPS_PGD_C0_CONTEXT + long pgdc = (long)pgd_current; +#endif - memset(tlbmiss_handler_setup_pgd, 0, sizeof(tlbmiss_handler_setup_pgd)); + 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 @@ -1364,20 +1466,40 @@ static void __cpuinit build_r4000_setup_pgd(void) } else { /* PGD in c0_KScratch */ uasm_i_jr(&p, 31); - UASM_i_MTC0(&p, a0, 31, pgd_reg); + UASM_i_MTC0(&p, a0, c0_kscratch(), pgd_reg); } - if (p - tlbmiss_handler_setup_pgd > ARRAY_SIZE(tlbmiss_handler_setup_pgd)) +#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"); + uasm_resolve_relocs(relocs, labels); pr_debug("Wrote tlbmiss_handler_setup_pgd (%u instructions).\n", (unsigned int)(p - tlbmiss_handler_setup_pgd)); - dump_handler(tlbmiss_handler_setup_pgd, - ARRAY_SIZE(tlbmiss_handler_setup_pgd)); + 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 @@ -1397,7 +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) { @@ -1457,29 +1579,35 @@ 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, - unsigned int pte, unsigned int ptr, enum label_id lid) + int pte, int ptr, int scratch, enum label_id lid) { - if (kernel_uses_smartmips_rixi) { + int t = scratch >= 0 ? scratch : pte; + + if (cpu_has_rixi) { if (use_bbit_insns()) { uasm_il_bbit0(p, r, pte, ilog2(_PAGE_PRESENT), lid); uasm_i_nop(p); } else { - uasm_i_andi(p, pte, pte, _PAGE_PRESENT); - uasm_il_beqz(p, r, pte, lid); - iPTE_LW(p, pte, ptr); + uasm_i_andi(p, t, pte, _PAGE_PRESENT); + uasm_il_beqz(p, r, t, lid); + if (pte == t) + /* You lose the SMP race :-(*/ + iPTE_LW(p, pte, ptr); } } else { - uasm_i_andi(p, pte, pte, _PAGE_PRESENT | _PAGE_READ); - uasm_i_xori(p, pte, pte, _PAGE_PRESENT | _PAGE_READ); - uasm_il_bnez(p, r, pte, lid); - iPTE_LW(p, pte, ptr); + uasm_i_andi(p, t, pte, _PAGE_PRESENT | _PAGE_READ); + uasm_i_xori(p, t, t, _PAGE_PRESENT | _PAGE_READ); + uasm_il_bnez(p, r, t, lid); + if (pte == t) + /* You lose the SMP race :-(*/ + iPTE_LW(p, pte, ptr); } } /* 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) { @@ -1492,27 +1620,27 @@ 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, enum label_id lid) + unsigned int pte, unsigned int ptr, int scratch, + enum label_id lid) { - if (use_bbit_insns()) { - uasm_il_bbit0(p, r, pte, ilog2(_PAGE_PRESENT), lid); - uasm_i_nop(p); - uasm_il_bbit0(p, r, pte, ilog2(_PAGE_WRITE), lid); - uasm_i_nop(p); - } else { - uasm_i_andi(p, pte, pte, _PAGE_PRESENT | _PAGE_WRITE); - uasm_i_xori(p, pte, pte, _PAGE_PRESENT | _PAGE_WRITE); - uasm_il_bnez(p, r, pte, lid); + int t = scratch >= 0 ? scratch : pte; + + uasm_i_andi(p, t, pte, _PAGE_PRESENT | _PAGE_WRITE); + uasm_i_xori(p, t, t, _PAGE_PRESENT | _PAGE_WRITE); + uasm_il_bnez(p, r, t, lid); + if (pte == t) + /* You lose the SMP race :-(*/ iPTE_LW(p, pte, ptr); - } + else + uasm_i_nop(p); } /* 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) { @@ -1526,17 +1654,21 @@ 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, enum label_id lid) + unsigned int pte, unsigned int ptr, int scratch, + enum label_id lid) { if (use_bbit_insns()) { uasm_il_bbit0(p, r, pte, ilog2(_PAGE_WRITE), lid); uasm_i_nop(p); } else { - uasm_i_andi(p, pte, pte, _PAGE_WRITE); - uasm_il_beqz(p, r, pte, lid); - iPTE_LW(p, pte, ptr); + int t = scratch >= 0 ? scratch : pte; + uasm_i_andi(p, t, pte, _PAGE_WRITE); + uasm_il_beqz(p, r, t, lid); + if (pte == t) + /* You lose the SMP race :-(*/ + iPTE_LW(p, pte, ptr); } } @@ -1551,7 +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 */ @@ -1567,7 +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) @@ -1585,7 +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) { @@ -1605,18 +1737,19 @@ 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; struct uasm_label *l = labels; struct uasm_reloc *r = relocs; - memset(handle_tlbl, 0, sizeof(handle_tlbl)); + memset(handle_tlbl, 0, handle_tlbl_size * sizeof(handle_tlbl[0])); memset(labels, 0, sizeof(labels)); memset(relocs, 0, sizeof(relocs)); build_r3000_tlbchange_handler_head(&p, K0, K1); - build_pte_present(&p, &r, K0, K1, label_nopage_tlbl); + build_pte_present(&p, &r, K0, K1, -1, label_nopage_tlbl); uasm_i_nop(&p); /* load delay */ build_make_valid(&p, &r, K0, K1); build_r3000_tlb_reload_write(&p, &l, &r, K0, K1); @@ -1625,28 +1758,29 @@ static void __cpuinit build_r3000_tlb_load_handler(void) uasm_i_j(&p, (unsigned long)tlb_do_page_fault_0 & 0x0fffffff); uasm_i_nop(&p); - if ((p - handle_tlbl) > FASTPATH_SIZE) + if (p >= handle_tlbl_end) panic("TLB load handler fastpath space exceeded"); uasm_resolve_relocs(relocs, labels); pr_debug("Wrote TLB load handler fastpath (%u instructions).\n", (unsigned int)(p - handle_tlbl)); - dump_handler(handle_tlbl, ARRAY_SIZE(handle_tlbl)); + 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; struct uasm_label *l = labels; struct uasm_reloc *r = relocs; - memset(handle_tlbs, 0, sizeof(handle_tlbs)); + memset(handle_tlbs, 0, handle_tlbs_size * sizeof(handle_tlbs[0])); memset(labels, 0, sizeof(labels)); memset(relocs, 0, sizeof(relocs)); build_r3000_tlbchange_handler_head(&p, K0, K1); - build_pte_writable(&p, &r, K0, K1, label_nopage_tlbs); + build_pte_writable(&p, &r, K0, K1, -1, label_nopage_tlbs); uasm_i_nop(&p); /* load delay */ build_make_write(&p, &r, K0, K1); build_r3000_tlb_reload_write(&p, &l, &r, K0, K1); @@ -1655,28 +1789,29 @@ 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) > FASTPATH_SIZE) + if (p >= handle_tlbs_end) panic("TLB store handler fastpath space exceeded"); uasm_resolve_relocs(relocs, labels); pr_debug("Wrote TLB store handler fastpath (%u instructions).\n", (unsigned int)(p - handle_tlbs)); - dump_handler(handle_tlbs, ARRAY_SIZE(handle_tlbs)); + 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; struct uasm_label *l = labels; struct uasm_reloc *r = relocs; - memset(handle_tlbm, 0, sizeof(handle_tlbm)); + memset(handle_tlbm, 0, handle_tlbm_size * sizeof(handle_tlbm[0])); memset(labels, 0, sizeof(labels)); memset(relocs, 0, sizeof(relocs)); build_r3000_tlbchange_handler_head(&p, K0, K1); - build_pte_modifiable(&p, &r, K0, K1, label_nopage_tlbm); + build_pte_modifiable(&p, &r, K0, K1, -1, label_nopage_tlbm); uasm_i_nop(&p); /* load delay */ build_make_write(&p, &r, K0, K1); build_r3000_pte_reload_tlbwi(&p, K0, K1); @@ -1685,55 +1820,57 @@ static void __cpuinit build_r3000_tlb_modify_handler(void) uasm_i_j(&p, (unsigned long)tlb_do_page_fault_1 & 0x0fffffff); uasm_i_nop(&p); - if ((p - handle_tlbm) > FASTPATH_SIZE) + if (p >= handle_tlbm_end) panic("TLB modify handler fastpath space exceeded"); uasm_resolve_relocs(relocs, labels); pr_debug("Wrote TLB modify handler fastpath (%u instructions).\n", (unsigned int)(p - handle_tlbm)); - dump_handler(handle_tlbm, ARRAY_SIZE(handle_tlbm)); + dump_handler("r3000_tlb_modify", handle_tlbm, handle_tlbm_size); } #endif /* CONFIG_MIPS_PGD_C0_CONTEXT */ /* * R4000 style TLB load/store/modify handlers. */ -static void __cpuinit +static struct work_registers build_r4000_tlbchange_handler_head(u32 **p, struct uasm_label **l, - struct uasm_reloc **r, unsigned int pte, - unsigned int ptr) + struct uasm_reloc **r) { + struct work_registers wr = build_get_work_registers(p); + #ifdef CONFIG_64BIT - build_get_pmde64(p, l, r, pte, ptr); /* get pmd in ptr */ + build_get_pmde64(p, l, r, wr.r1, wr.r2); /* get pmd in ptr */ #else - build_get_pgde32(p, pte, ptr); /* get pgd in ptr */ + build_get_pgde32(p, wr.r1, wr.r2); /* get pgd in ptr */ #endif -#ifdef CONFIG_HUGETLB_PAGE +#ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT /* * For huge tlb entries, pmd doesn't contain an address but * instead contains the tlb pte. Check the PAGE_HUGE bit and * see if we need to jump to huge tlb processing. */ - build_is_huge_pte(p, r, pte, ptr, label_tlb_huge_update); + build_is_huge_pte(p, r, wr.r1, wr.r2, label_tlb_huge_update); #endif - UASM_i_MFC0(p, pte, C0_BADVADDR); - UASM_i_LW(p, ptr, 0, ptr); - UASM_i_SRL(p, pte, pte, PAGE_SHIFT + PTE_ORDER - PTE_T_LOG2); - uasm_i_andi(p, pte, pte, (PTRS_PER_PTE - 1) << PTE_T_LOG2); - UASM_i_ADDU(p, ptr, ptr, pte); + UASM_i_MFC0(p, wr.r1, C0_BADVADDR); + UASM_i_LW(p, wr.r2, 0, wr.r2); + UASM_i_SRL(p, wr.r1, wr.r1, PAGE_SHIFT + PTE_ORDER - PTE_T_LOG2); + uasm_i_andi(p, wr.r1, wr.r1, (PTRS_PER_PTE - 1) << PTE_T_LOG2); + UASM_i_ADDU(p, wr.r2, wr.r2, wr.r1); #ifdef CONFIG_SMP uasm_l_smp_pgtable_change(l, *p); #endif - iPTE_LW(p, pte, ptr); /* get even pte */ + iPTE_LW(p, wr.r1, wr.r2); /* get even pte */ if (!m4kc_tlbp_war()) build_tlb_probe_entry(p); + 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) @@ -1743,6 +1880,7 @@ build_r4000_tlbchange_handler_tail(u32 **p, struct uasm_label **l, build_update_entries(p, tmp, ptr); build_tlb_write_entry(p, l, r, tlb_indexed); uasm_l_leave(l, *p); + build_restore_work_registers(p); uasm_i_eret(p); /* return from trap */ #ifdef CONFIG_64BIT @@ -1750,13 +1888,15 @@ 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; struct uasm_label *l = labels; struct uasm_reloc *r = relocs; + struct work_registers wr; - memset(handle_tlbl, 0, sizeof(handle_tlbl)); + memset(handle_tlbl, 0, handle_tlbl_size * sizeof(handle_tlbl[0])); memset(labels, 0, sizeof(labels)); memset(relocs, 0, sizeof(relocs)); @@ -1774,221 +1914,287 @@ static void __cpuinit build_r4000_tlb_load_handler(void) /* No need for uasm_i_nop */ } - build_r4000_tlbchange_handler_head(&p, &l, &r, K0, K1); - build_pte_present(&p, &r, K0, K1, label_nopage_tlbl); + wr = build_r4000_tlbchange_handler_head(&p, &l, &r); + build_pte_present(&p, &r, wr.r1, wr.r2, wr.r3, label_nopage_tlbl); if (m4kc_tlbp_war()) build_tlb_probe_entry(&p); - if (kernel_uses_smartmips_rixi) { + if (cpu_has_rixi) { /* * If the page is not _PAGE_VALID, RI or XI could not * have triggered it. Skip the expensive test.. */ if (use_bbit_insns()) { - uasm_il_bbit0(&p, &r, K0, ilog2(_PAGE_VALID), + uasm_il_bbit0(&p, &r, wr.r1, ilog2(_PAGE_VALID), label_tlbl_goaround1); } else { - uasm_i_andi(&p, K0, K0, _PAGE_VALID); - uasm_il_beqz(&p, &r, K0, label_tlbl_goaround1); + uasm_i_andi(&p, wr.r3, wr.r1, _PAGE_VALID); + uasm_il_beqz(&p, &r, wr.r3, label_tlbl_goaround1); } uasm_i_nop(&p); uasm_i_tlbr(&p); + + switch (current_cpu_type()) { + default: + if (cpu_has_mips_r2) { + uasm_i_ehb(&p); + + case CPU_CAVIUM_OCTEON: + case CPU_CAVIUM_OCTEON_PLUS: + case CPU_CAVIUM_OCTEON2: + break; + } + } + /* Examine entrylo 0 or 1 based on ptr. */ if (use_bbit_insns()) { - uasm_i_bbit0(&p, K1, ilog2(sizeof(pte_t)), 8); + uasm_i_bbit0(&p, wr.r2, ilog2(sizeof(pte_t)), 8); } else { - uasm_i_andi(&p, K0, K1, sizeof(pte_t)); - uasm_i_beqz(&p, K0, 8); + uasm_i_andi(&p, wr.r3, wr.r2, sizeof(pte_t)); + uasm_i_beqz(&p, wr.r3, 8); } - - UASM_i_MFC0(&p, K0, C0_ENTRYLO0); /* load it in the delay slot*/ - UASM_i_MFC0(&p, K0, C0_ENTRYLO1); /* load it if ptr is odd */ + /* load it in the delay slot*/ + UASM_i_MFC0(&p, wr.r3, C0_ENTRYLO0); + /* load it if ptr is odd */ + UASM_i_MFC0(&p, wr.r3, C0_ENTRYLO1); /* - * If the entryLo (now in K0) is valid (bit 1), RI or + * If the entryLo (now in wr.r3) is valid (bit 1), RI or * XI must have triggered it. */ if (use_bbit_insns()) { - uasm_il_bbit1(&p, &r, K0, 1, label_nopage_tlbl); - /* Reload the PTE value */ - iPTE_LW(&p, K0, K1); + uasm_il_bbit1(&p, &r, wr.r3, 1, label_nopage_tlbl); + uasm_i_nop(&p); uasm_l_tlbl_goaround1(&l, p); } else { - uasm_i_andi(&p, K0, K0, 2); - uasm_il_bnez(&p, &r, K0, label_nopage_tlbl); - uasm_l_tlbl_goaround1(&l, p); - /* Reload the PTE value */ - iPTE_LW(&p, K0, K1); + uasm_i_andi(&p, wr.r3, wr.r3, 2); + uasm_il_bnez(&p, &r, wr.r3, label_nopage_tlbl); + uasm_i_nop(&p); } + uasm_l_tlbl_goaround1(&l, p); } - build_make_valid(&p, &r, K0, K1); - build_r4000_tlbchange_handler_tail(&p, &l, &r, K0, K1); + build_make_valid(&p, &r, wr.r1, wr.r2); + build_r4000_tlbchange_handler_tail(&p, &l, &r, wr.r1, wr.r2); -#ifdef CONFIG_HUGETLB_PAGE +#ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT /* * This is the entry point when build_r4000_tlbchange_handler_head * spots a huge page. */ uasm_l_tlb_huge_update(&l, p); - iPTE_LW(&p, K0, K1); - build_pte_present(&p, &r, K0, K1, label_nopage_tlbl); + iPTE_LW(&p, wr.r1, wr.r2); + build_pte_present(&p, &r, wr.r1, wr.r2, wr.r3, label_nopage_tlbl); build_tlb_probe_entry(&p); - if (kernel_uses_smartmips_rixi) { + if (cpu_has_rixi) { /* * If the page is not _PAGE_VALID, RI or XI could not * have triggered it. Skip the expensive test.. */ if (use_bbit_insns()) { - uasm_il_bbit0(&p, &r, K0, ilog2(_PAGE_VALID), + uasm_il_bbit0(&p, &r, wr.r1, ilog2(_PAGE_VALID), label_tlbl_goaround2); } else { - uasm_i_andi(&p, K0, K0, _PAGE_VALID); - uasm_il_beqz(&p, &r, K0, label_tlbl_goaround2); + uasm_i_andi(&p, wr.r3, wr.r1, _PAGE_VALID); + uasm_il_beqz(&p, &r, wr.r3, label_tlbl_goaround2); } uasm_i_nop(&p); uasm_i_tlbr(&p); + + switch (current_cpu_type()) { + default: + if (cpu_has_mips_r2) { + uasm_i_ehb(&p); + + case CPU_CAVIUM_OCTEON: + case CPU_CAVIUM_OCTEON_PLUS: + case CPU_CAVIUM_OCTEON2: + break; + } + } + /* Examine entrylo 0 or 1 based on ptr. */ if (use_bbit_insns()) { - uasm_i_bbit0(&p, K1, ilog2(sizeof(pte_t)), 8); + uasm_i_bbit0(&p, wr.r2, ilog2(sizeof(pte_t)), 8); } else { - uasm_i_andi(&p, K0, K1, sizeof(pte_t)); - uasm_i_beqz(&p, K0, 8); + uasm_i_andi(&p, wr.r3, wr.r2, sizeof(pte_t)); + uasm_i_beqz(&p, wr.r3, 8); } - UASM_i_MFC0(&p, K0, C0_ENTRYLO0); /* load it in the delay slot*/ - UASM_i_MFC0(&p, K0, C0_ENTRYLO1); /* load it if ptr is odd */ + /* load it in the delay slot*/ + UASM_i_MFC0(&p, wr.r3, C0_ENTRYLO0); + /* load it if ptr is odd */ + UASM_i_MFC0(&p, wr.r3, C0_ENTRYLO1); /* - * If the entryLo (now in K0) is valid (bit 1), RI or + * If the entryLo (now in wr.r3) is valid (bit 1), RI or * XI must have triggered it. */ if (use_bbit_insns()) { - uasm_il_bbit0(&p, &r, K0, 1, label_tlbl_goaround2); + uasm_il_bbit0(&p, &r, wr.r3, 1, label_tlbl_goaround2); } else { - uasm_i_andi(&p, K0, K0, 2); - uasm_il_beqz(&p, &r, K0, label_tlbl_goaround2); + uasm_i_andi(&p, wr.r3, wr.r3, 2); + uasm_il_beqz(&p, &r, wr.r3, label_tlbl_goaround2); } - /* Reload the PTE value */ - iPTE_LW(&p, K0, K1); - + if (PM_DEFAULT_MASK == 0) + uasm_i_nop(&p); /* * We clobbered C0_PAGEMASK, restore it. On the other branch * it is restored in build_huge_tlb_write_entry. */ - build_restore_pagemask(&p, &r, K0, label_nopage_tlbl, 0); + build_restore_pagemask(&p, &r, wr.r3, label_nopage_tlbl, 0); uasm_l_tlbl_goaround2(&l, p); } - uasm_i_ori(&p, K0, K0, (_PAGE_ACCESSED | _PAGE_VALID)); - build_huge_handler_tail(&p, &r, &l, K0, K1); + uasm_i_ori(&p, wr.r1, wr.r1, (_PAGE_ACCESSED | _PAGE_VALID)); + build_huge_handler_tail(&p, &r, &l, wr.r1, wr.r2); #endif uasm_l_nopage_tlbl(&l, p); + build_restore_work_registers(&p); +#ifdef CONFIG_CPU_MICROMIPS + if ((unsigned long)tlb_do_page_fault_0 & 1) { + uasm_i_lui(&p, K0, uasm_rel_hi((long)tlb_do_page_fault_0)); + uasm_i_addiu(&p, K0, K0, uasm_rel_lo((long)tlb_do_page_fault_0)); + uasm_i_jr(&p, K0); + } else +#endif uasm_i_j(&p, (unsigned long)tlb_do_page_fault_0 & 0x0fffffff); uasm_i_nop(&p); - if ((p - handle_tlbl) > FASTPATH_SIZE) + if (p >= handle_tlbl_end) panic("TLB load handler fastpath space exceeded"); uasm_resolve_relocs(relocs, labels); pr_debug("Wrote TLB load handler fastpath (%u instructions).\n", (unsigned int)(p - handle_tlbl)); - dump_handler(handle_tlbl, ARRAY_SIZE(handle_tlbl)); + 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; struct uasm_label *l = labels; struct uasm_reloc *r = relocs; + struct work_registers wr; - memset(handle_tlbs, 0, sizeof(handle_tlbs)); + memset(handle_tlbs, 0, handle_tlbs_size * sizeof(handle_tlbs[0])); memset(labels, 0, sizeof(labels)); memset(relocs, 0, sizeof(relocs)); - build_r4000_tlbchange_handler_head(&p, &l, &r, K0, K1); - build_pte_writable(&p, &r, K0, K1, label_nopage_tlbs); + wr = build_r4000_tlbchange_handler_head(&p, &l, &r); + build_pte_writable(&p, &r, wr.r1, wr.r2, wr.r3, label_nopage_tlbs); if (m4kc_tlbp_war()) build_tlb_probe_entry(&p); - build_make_write(&p, &r, K0, K1); - build_r4000_tlbchange_handler_tail(&p, &l, &r, K0, K1); + build_make_write(&p, &r, wr.r1, wr.r2); + build_r4000_tlbchange_handler_tail(&p, &l, &r, wr.r1, wr.r2); -#ifdef CONFIG_HUGETLB_PAGE +#ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT /* * This is the entry point when * build_r4000_tlbchange_handler_head spots a huge page. */ uasm_l_tlb_huge_update(&l, p); - iPTE_LW(&p, K0, K1); - build_pte_writable(&p, &r, K0, K1, label_nopage_tlbs); + iPTE_LW(&p, wr.r1, wr.r2); + build_pte_writable(&p, &r, wr.r1, wr.r2, wr.r3, label_nopage_tlbs); build_tlb_probe_entry(&p); - uasm_i_ori(&p, K0, K0, + uasm_i_ori(&p, wr.r1, wr.r1, _PAGE_ACCESSED | _PAGE_MODIFIED | _PAGE_VALID | _PAGE_DIRTY); - build_huge_handler_tail(&p, &r, &l, K0, K1); + build_huge_handler_tail(&p, &r, &l, wr.r1, wr.r2); #endif uasm_l_nopage_tlbs(&l, p); + build_restore_work_registers(&p); +#ifdef CONFIG_CPU_MICROMIPS + if ((unsigned long)tlb_do_page_fault_1 & 1) { + uasm_i_lui(&p, K0, uasm_rel_hi((long)tlb_do_page_fault_1)); + uasm_i_addiu(&p, K0, K0, uasm_rel_lo((long)tlb_do_page_fault_1)); + uasm_i_jr(&p, K0); + } else +#endif uasm_i_j(&p, (unsigned long)tlb_do_page_fault_1 & 0x0fffffff); uasm_i_nop(&p); - if ((p - handle_tlbs) > FASTPATH_SIZE) + if (p >= handle_tlbs_end) panic("TLB store handler fastpath space exceeded"); uasm_resolve_relocs(relocs, labels); pr_debug("Wrote TLB store handler fastpath (%u instructions).\n", (unsigned int)(p - handle_tlbs)); - dump_handler(handle_tlbs, ARRAY_SIZE(handle_tlbs)); + 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; struct uasm_label *l = labels; struct uasm_reloc *r = relocs; + struct work_registers wr; - memset(handle_tlbm, 0, sizeof(handle_tlbm)); + memset(handle_tlbm, 0, handle_tlbm_size * sizeof(handle_tlbm[0])); memset(labels, 0, sizeof(labels)); memset(relocs, 0, sizeof(relocs)); - build_r4000_tlbchange_handler_head(&p, &l, &r, K0, K1); - build_pte_modifiable(&p, &r, K0, K1, label_nopage_tlbm); + wr = build_r4000_tlbchange_handler_head(&p, &l, &r); + build_pte_modifiable(&p, &r, wr.r1, wr.r2, wr.r3, label_nopage_tlbm); if (m4kc_tlbp_war()) build_tlb_probe_entry(&p); /* Present and writable bits set, set accessed and dirty bits. */ - build_make_write(&p, &r, K0, K1); - build_r4000_tlbchange_handler_tail(&p, &l, &r, K0, K1); + build_make_write(&p, &r, wr.r1, wr.r2); + build_r4000_tlbchange_handler_tail(&p, &l, &r, wr.r1, wr.r2); -#ifdef CONFIG_HUGETLB_PAGE +#ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT /* * This is the entry point when * build_r4000_tlbchange_handler_head spots a huge page. */ uasm_l_tlb_huge_update(&l, p); - iPTE_LW(&p, K0, K1); - build_pte_modifiable(&p, &r, K0, K1, label_nopage_tlbm); + iPTE_LW(&p, wr.r1, wr.r2); + build_pte_modifiable(&p, &r, wr.r1, wr.r2, wr.r3, label_nopage_tlbm); build_tlb_probe_entry(&p); - uasm_i_ori(&p, K0, K0, + uasm_i_ori(&p, wr.r1, wr.r1, _PAGE_ACCESSED | _PAGE_MODIFIED | _PAGE_VALID | _PAGE_DIRTY); - build_huge_handler_tail(&p, &r, &l, K0, K1); + build_huge_handler_tail(&p, &r, &l, wr.r1, wr.r2); #endif uasm_l_nopage_tlbm(&l, p); + build_restore_work_registers(&p); +#ifdef CONFIG_CPU_MICROMIPS + if ((unsigned long)tlb_do_page_fault_1 & 1) { + uasm_i_lui(&p, K0, uasm_rel_hi((long)tlb_do_page_fault_1)); + uasm_i_addiu(&p, K0, K0, uasm_rel_lo((long)tlb_do_page_fault_1)); + uasm_i_jr(&p, K0); + } else +#endif uasm_i_j(&p, (unsigned long)tlb_do_page_fault_1 & 0x0fffffff); uasm_i_nop(&p); - if ((p - handle_tlbm) > FASTPATH_SIZE) + if (p >= handle_tlbm_end) panic("TLB modify handler fastpath space exceeded"); uasm_resolve_relocs(relocs, labels); pr_debug("Wrote TLB modify handler fastpath (%u instructions).\n", (unsigned int)(p - handle_tlbm)); - dump_handler(handle_tlbm, ARRAY_SIZE(handle_tlbm)); + dump_handler("r4000_tlb_modify", handle_tlbm, handle_tlbm_size); +} + +static void flush_tlb_handlers(void) +{ + local_flush_icache_range((unsigned long)handle_tlbl, + (unsigned long)handle_tlbl_end); + local_flush_icache_range((unsigned long)handle_tlbs, + (unsigned long)handle_tlbs_end); + local_flush_icache_range((unsigned long)handle_tlbm, + (unsigned long)handle_tlbm_end); + local_flush_icache_range((unsigned long)tlbmiss_handler_setup_pgd, + (unsigned long)tlbmiss_handler_setup_pgd_end); } -void __cpuinit build_tlb_refill_handler(void) +void build_tlb_refill_handler(void) { /* * The refill handler is generated per-CPU, multi-node systems @@ -1997,6 +2203,8 @@ void __cpuinit build_tlb_refill_handler(void) */ static int run_once = 0; + output_pgtable_bits_defines(); + #ifdef CONFIG_64BIT check_for_high_segbits = current_cpu_data.vmbits > (PGDIR_SHIFT + PGD_ORDER + PAGE_SHIFT - 3); #endif @@ -2010,11 +2218,16 @@ void __cpuinit build_tlb_refill_handler(void) case CPU_TX3922: case CPU_TX3927: #ifndef CONFIG_MIPS_PGD_C0_CONTEXT - build_r3000_tlb_refill_handler(); + if (cpu_has_local_ebase) + build_r3000_tlb_refill_handler(); 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(); + flush_tlb_handlers(); run_once++; } #else @@ -2033,28 +2246,17 @@ void __cpuinit build_tlb_refill_handler(void) default: if (!run_once) { -#ifdef CONFIG_MIPS_PGD_C0_CONTEXT - build_r4000_setup_pgd(); -#endif + scratch_reg = allocate_kscratch(); + build_setup_pgd(); build_r4000_tlb_load_handler(); build_r4000_tlb_store_handler(); build_r4000_tlb_modify_handler(); + if (!cpu_has_local_ebase) + build_r4000_tlb_refill_handler(); + flush_tlb_handlers(); run_once++; } - build_r4000_tlb_refill_handler(); + if (cpu_has_local_ebase) + build_r4000_tlb_refill_handler(); } } - -void __cpuinit flush_tlb_handlers(void) -{ - local_flush_icache_range((unsigned long)handle_tlbl, - (unsigned long)handle_tlbl + sizeof(handle_tlbl)); - local_flush_icache_range((unsigned long)handle_tlbs, - (unsigned long)handle_tlbs + sizeof(handle_tlbs)); - local_flush_icache_range((unsigned long)handle_tlbm, - (unsigned long)handle_tlbm + sizeof(handle_tlbm)); -#ifdef CONFIG_MIPS_PGD_C0_CONTEXT - local_flush_icache_range((unsigned long)tlbmiss_handler_setup_pgd, - (unsigned long)tlbmiss_handler_setup_pgd + sizeof(handle_tlbm)); -#endif -} |
