LoongArch: Introduce hardware page table walker

Loongson-3A6000 and newer processors have hardware page table walker
(PTW) support. PTW can handle all fastpaths of TLBI/TLBL/TLBS/TLBM
exceptions by hardware, software only need to handle slowpaths (page
faults).

BTW, PTW doesn't append _PAGE_MODIFIED for page table entries, so we
change pmd_dirty() and pte_dirty() to also check _PAGE_DIRTY for the
"dirty" attribute.

Signed-off-by: Liang Gao <[email protected]>
Signed-off-by: Jun Yi <[email protected]>
Signed-off-by: Huacai Chen <[email protected]>

Author: chenhuacai

arch/loongarch/include/asm/cpu-features.h

@@ -64,6 +64,6 @@
 #define cpu_has_eiodecode	cpu_opt(LOONGARCH_CPU_EIODECODE)
 #define cpu_has_guestid		cpu_opt(LOONGARCH_CPU_GUESTID)
 #define cpu_has_hypervisor	cpu_opt(LOONGARCH_CPU_HYPERVISOR)
-
+#define cpu_has_ptw		cpu_opt(LOONGARCH_CPU_PTW)
 
 #endif /* __ASM_CPU_FEATURES_H */

arch/loongarch/include/asm/cpu.h

@@ -98,6 +98,7 @@ enum cpu_type_enum {
 #define CPU_FEATURE_EIODECODE		23	/* CPU has EXTIOI interrupt pin decode mode */
 #define CPU_FEATURE_GUESTID		24	/* CPU has GuestID feature */
 #define CPU_FEATURE_HYPERVISOR		25	/* CPU has hypervisor (running in VM) */
+#define CPU_FEATURE_PTW			26	/* CPU has hardware page table walker */
 
 #define LOONGARCH_CPU_CPUCFG		BIT_ULL(CPU_FEATURE_CPUCFG)
 #define LOONGARCH_CPU_LAM		BIT_ULL(CPU_FEATURE_LAM)
@@ -125,5 +126,6 @@ enum cpu_type_enum {
 #define LOONGARCH_CPU_EIODECODE		BIT_ULL(CPU_FEATURE_EIODECODE)
 #define LOONGARCH_CPU_GUESTID		BIT_ULL(CPU_FEATURE_GUESTID)
 #define LOONGARCH_CPU_HYPERVISOR	BIT_ULL(CPU_FEATURE_HYPERVISOR)
+#define LOONGARCH_CPU_PTW		BIT_ULL(CPU_FEATURE_PTW)
 
 #endif /* _ASM_CPU_H */

arch/loongarch/include/asm/loongarch.h

@@ -135,6 +135,7 @@ __asm__(".macro	parse_r var r\n\t"
 #define  CPUCFG2_MIPSBT			BIT(20)
 #define  CPUCFG2_LSPW			BIT(21)
 #define  CPUCFG2_LAM			BIT(22)
+#define  CPUCFG2_PTW			BIT(24)
 
 #define LOONGARCH_CPUCFG3		0x3
 #define  CPUCFG3_CCDMA			BIT(0)
@@ -412,6 +413,9 @@ __asm__(".macro	parse_r var r\n\t"
 #define  CSR_PWCTL0_PTBASE		(_ULCAST_(0x1f) << CSR_PWCTL0_PTBASE_SHIFT)
 
 #define LOONGARCH_CSR_PWCTL1		0x1d	/* PWCtl1 */
+#define  CSR_PWCTL1_PTW_SHIFT		24
+#define  CSR_PWCTL1_PTW_WIDTH		1
+#define  CSR_PWCTL1_PTW			(_ULCAST_(0x1) << CSR_PWCTL1_PTW_SHIFT)
 #define  CSR_PWCTL1_DIR3WIDTH_SHIFT	18
 #define  CSR_PWCTL1_DIR3WIDTH_WIDTH	5
 #define  CSR_PWCTL1_DIR3WIDTH		(_ULCAST_(0x1f) << CSR_PWCTL1_DIR3WIDTH_SHIFT)

arch/loongarch/include/asm/pgtable.h

@@ -362,7 +362,7 @@ extern pgd_t invalid_pg_dir[];
  */
 static inline int pte_write(pte_t pte)	{ return pte_val(pte) & _PAGE_WRITE; }
 static inline int pte_young(pte_t pte)	{ return pte_val(pte) & _PAGE_ACCESSED; }
-static inline int pte_dirty(pte_t pte)	{ return pte_val(pte) & _PAGE_MODIFIED; }
+static inline int pte_dirty(pte_t pte)	{ return pte_val(pte) & (_PAGE_DIRTY | _PAGE_MODIFIED); }
 
 static inline pte_t pte_mkold(pte_t pte)
 {
@@ -506,7 +506,7 @@ static inline pmd_t pmd_wrprotect(pmd_t pmd)
 
 static inline int pmd_dirty(pmd_t pmd)
 {
-	return !!(pmd_val(pmd) & _PAGE_MODIFIED);
+	return !!(pmd_val(pmd) & (_PAGE_DIRTY | _PAGE_MODIFIED));
 }
 
 static inline pmd_t pmd_mkclean(pmd_t pmd)

arch/loongarch/include/asm/tlb.h

@@ -158,6 +158,9 @@ extern void handle_tlb_store(void);
 extern void handle_tlb_modify(void);
 extern void handle_tlb_refill(void);
 extern void handle_tlb_protect(void);
+extern void handle_tlb_load_ptw(void);
+extern void handle_tlb_store_ptw(void);
+extern void handle_tlb_modify_ptw(void);
 
 extern void dump_tlb_all(void);
 extern void dump_tlb_regs(void);

arch/loongarch/include/uapi/asm/hwcap.h

@@ -16,5 +16,6 @@
 #define HWCAP_LOONGARCH_LBT_X86		(1 << 10)
 #define HWCAP_LOONGARCH_LBT_ARM		(1 << 11)
 #define HWCAP_LOONGARCH_LBT_MIPS	(1 << 12)
+#define HWCAP_LOONGARCH_PTW		(1 << 13)
 
 #endif /* _UAPI_ASM_HWCAP_H */

arch/loongarch/kernel/cpu-probe.c

@@ -136,6 +136,10 @@ static void cpu_probe_common(struct cpuinfo_loongarch *c)
 		c->options |= LOONGARCH_CPU_CRYPTO;
 		elf_hwcap |= HWCAP_LOONGARCH_CRYPTO;
 	}
+	if (config & CPUCFG2_PTW) {
+		c->options |= LOONGARCH_CPU_PTW;
+		elf_hwcap |= HWCAP_LOONGARCH_PTW;
+	}
 	if (config & CPUCFG2_LVZP) {
 		c->options |= LOONGARCH_CPU_LVZ;
 		elf_hwcap |= HWCAP_LOONGARCH_LVZ;

arch/loongarch/kernel/proc.c

@@ -80,6 +80,7 @@ static int show_cpuinfo(struct seq_file *m, void *v)
 	if (cpu_has_crc32)	seq_printf(m, " crc32");
 	if (cpu_has_complex)	seq_printf(m, " complex");
 	if (cpu_has_crypto)	seq_printf(m, " crypto");
+	if (cpu_has_ptw)	seq_printf(m, " ptw");
 	if (cpu_has_lvz)	seq_printf(m, " lvz");
 	if (cpu_has_lbt_x86)	seq_printf(m, " lbt_x86");
 	if (cpu_has_lbt_arm)	seq_printf(m, " lbt_arm");

arch/loongarch/mm/tlb.c

@@ -167,6 +167,9 @@ void __update_tlb(struct vm_area_struct *vma, unsigned long address, pte_t *ptep
 	int idx;
 	unsigned long flags;
 
+	if (cpu_has_ptw)
+		return;
+
 	/*
 	 * Handle debugger faulting in for debugee.
 	 */
@@ -222,6 +225,9 @@ static void setup_ptwalker(void)
 	pwctl0 = pte_i | pte_w << 5 | pmd_i << 10 | pmd_w << 15 | pud_i << 20 | pud_w << 25;
 	pwctl1 = pgd_i | pgd_w << 6;
 
+	if (cpu_has_ptw)
+		pwctl1 |= CSR_PWCTL1_PTW;
+
 	csr_write64(pwctl0, LOONGARCH_CSR_PWCTL0);
 	csr_write64(pwctl1, LOONGARCH_CSR_PWCTL1);
 	csr_write64((long)swapper_pg_dir, LOONGARCH_CSR_PGDH);
@@ -264,10 +270,17 @@ void setup_tlb_handler(int cpu)
 	if (cpu == 0) {
 		memcpy((void *)tlbrentry, handle_tlb_refill, 0x80);
 		local_flush_icache_range(tlbrentry, tlbrentry + 0x80);
-		set_handler(EXCCODE_TLBI * VECSIZE, handle_tlb_load, VECSIZE);
-		set_handler(EXCCODE_TLBL * VECSIZE, handle_tlb_load, VECSIZE);
-		set_handler(EXCCODE_TLBS * VECSIZE, handle_tlb_store, VECSIZE);
-		set_handler(EXCCODE_TLBM * VECSIZE, handle_tlb_modify, VECSIZE);
+		if (!cpu_has_ptw) {
+			set_handler(EXCCODE_TLBI * VECSIZE, handle_tlb_load, VECSIZE);
+			set_handler(EXCCODE_TLBL * VECSIZE, handle_tlb_load, VECSIZE);
+			set_handler(EXCCODE_TLBS * VECSIZE, handle_tlb_store, VECSIZE);
+			set_handler(EXCCODE_TLBM * VECSIZE, handle_tlb_modify, VECSIZE);
+		} else {
+			set_handler(EXCCODE_TLBI * VECSIZE, handle_tlb_load_ptw, VECSIZE);
+			set_handler(EXCCODE_TLBL * VECSIZE, handle_tlb_load_ptw, VECSIZE);
+			set_handler(EXCCODE_TLBS * VECSIZE, handle_tlb_store_ptw, VECSIZE);
+			set_handler(EXCCODE_TLBM * VECSIZE, handle_tlb_modify_ptw, VECSIZE);
+		}
 		set_handler(EXCCODE_TLBNR * VECSIZE, handle_tlb_protect, VECSIZE);
 		set_handler(EXCCODE_TLBNX * VECSIZE, handle_tlb_protect, VECSIZE);
 		set_handler(EXCCODE_TLBPE * VECSIZE, handle_tlb_protect, VECSIZE);

arch/loongarch/mm/tlbex.S

@@ -190,6 +190,13 @@ nopage_tlb_load:
 	jr		t0
 SYM_FUNC_END(handle_tlb_load)
 
+SYM_FUNC_START(handle_tlb_load_ptw)
+	csrwr		t0, LOONGARCH_CSR_KS0
+	csrwr		t1, LOONGARCH_CSR_KS1
+	la_abs		t0, tlb_do_page_fault_0
+	jr		t0
+SYM_FUNC_END(handle_tlb_load_ptw)
+
 SYM_FUNC_START(handle_tlb_store)
 	csrwr		t0, EXCEPTION_KS0
 	csrwr		t1, EXCEPTION_KS1
@@ -339,6 +346,13 @@ nopage_tlb_store:
 	jr		t0
 SYM_FUNC_END(handle_tlb_store)
 
+SYM_FUNC_START(handle_tlb_store_ptw)
+	csrwr		t0, LOONGARCH_CSR_KS0
+	csrwr		t1, LOONGARCH_CSR_KS1
+	la_abs		t0, tlb_do_page_fault_1
+	jr		t0
+SYM_FUNC_END(handle_tlb_store_ptw)
+
 SYM_FUNC_START(handle_tlb_modify)
 	csrwr		t0, EXCEPTION_KS0
 	csrwr		t1, EXCEPTION_KS1
@@ -486,6 +500,13 @@ nopage_tlb_modify:
 	jr		t0
 SYM_FUNC_END(handle_tlb_modify)
 
+SYM_FUNC_START(handle_tlb_modify_ptw)
+	csrwr		t0, LOONGARCH_CSR_KS0
+	csrwr		t1, LOONGARCH_CSR_KS1
+	la_abs		t0, tlb_do_page_fault_1
+	jr		t0
+SYM_FUNC_END(handle_tlb_modify_ptw)
+
 SYM_FUNC_START(handle_tlb_refill)
 	csrwr		t0, LOONGARCH_CSR_TLBRSAVE
 	csrrd		t0, LOONGARCH_CSR_PGD