LoongArch: Add memory management

Add memory management support for LoongArch, including: cache and tlb
management, page fault handling and ioremap/mmap support.

Reviewed-by: WANG Xuerui <git@xen0n.name>
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>

12 files changed, 1811 insertions, 0 deletions

diff --git a/arch/loongarch/mm/cache.c b/arch/loongarch/mm/cache.c
new file mode 100644
index 000000000000..9e5ce5aa73f7
--- /dev/null
+++ b/arch/loongarch/mm/cache.c
@@ -0,0 +1,141 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2020-2022 Loongson Technology Corporation Limited
+ *
+ * Derived from MIPS:
+ * Copyright (C) 1994 - 2003, 06, 07 by Ralf Baechle (ralf@linux-mips.org)
+ * Copyright (C) 2007 MIPS Technologies, Inc.
+ */
+#include <linux/export.h>
+#include <linux/fcntl.h>
+#include <linux/fs.h>
+#include <linux/highmem.h>
+#include <linux/kernel.h>
+#include <linux/linkage.h>
+#include <linux/mm.h>
+#include <linux/sched.h>
+#include <linux/syscalls.h>
+
+#include <asm/cacheflush.h>
+#include <asm/cpu.h>
+#include <asm/cpu-features.h>
+#include <asm/dma.h>
+#include <asm/loongarch.h>
+#include <asm/processor.h>
+#include <asm/setup.h>
+
+/*
+ * LoongArch maintains ICache/DCache coherency by hardware,
+ * we just need "ibar" to avoid instruction hazard here.
+ */
+void local_flush_icache_range(unsigned long start, unsigned long end)
+{
+	asm volatile ("\tibar 0\n"::);
+}
+EXPORT_SYMBOL(local_flush_icache_range);
+
+void cache_error_setup(void)
+{
+	extern char __weak except_vec_cex;
+	set_merr_handler(0x0, &except_vec_cex, 0x80);
+}
+
+static unsigned long icache_size __read_mostly;
+static unsigned long dcache_size __read_mostly;
+static unsigned long vcache_size __read_mostly;
+static unsigned long scache_size __read_mostly;
+
+static char *way_string[] = { NULL, "direct mapped", "2-way",
+	"3-way", "4-way", "5-way", "6-way", "7-way", "8-way",
+	"9-way", "10-way", "11-way", "12-way",
+	"13-way", "14-way", "15-way", "16-way",
+};
+
+static void probe_pcache(void)
+{
+	struct cpuinfo_loongarch *c = &current_cpu_data;
+	unsigned int lsize, sets, ways;
+	unsigned int config;
+
+	config = read_cpucfg(LOONGARCH_CPUCFG17);
+	lsize = 1 << ((config & CPUCFG17_L1I_SIZE_M) >> CPUCFG17_L1I_SIZE);
+	sets  = 1 << ((config & CPUCFG17_L1I_SETS_M) >> CPUCFG17_L1I_SETS);
+	ways  = ((config & CPUCFG17_L1I_WAYS_M) >> CPUCFG17_L1I_WAYS) + 1;
+
+	c->icache.linesz = lsize;
+	c->icache.sets = sets;
+	c->icache.ways = ways;
+	icache_size = sets * ways * lsize;
+	c->icache.waysize = icache_size / c->icache.ways;
+
+	config = read_cpucfg(LOONGARCH_CPUCFG18);
+	lsize = 1 << ((config & CPUCFG18_L1D_SIZE_M) >> CPUCFG18_L1D_SIZE);
+	sets  = 1 << ((config & CPUCFG18_L1D_SETS_M) >> CPUCFG18_L1D_SETS);
+	ways  = ((config & CPUCFG18_L1D_WAYS_M) >> CPUCFG18_L1D_WAYS) + 1;
+
+	c->dcache.linesz = lsize;
+	c->dcache.sets = sets;
+	c->dcache.ways = ways;
+	dcache_size = sets * ways * lsize;
+	c->dcache.waysize = dcache_size / c->dcache.ways;
+
+	c->options |= LOONGARCH_CPU_PREFETCH;
+
+	pr_info("Primary instruction cache %ldkB, %s, %s, linesize %d bytes.\n",
+		icache_size >> 10, way_string[c->icache.ways], "VIPT", c->icache.linesz);
+
+	pr_info("Primary data cache %ldkB, %s, %s, %s, linesize %d bytes\n",
+		dcache_size >> 10, way_string[c->dcache.ways], "VIPT", "no aliases", c->dcache.linesz);
+}
+
+static void probe_vcache(void)
+{
+	struct cpuinfo_loongarch *c = &current_cpu_data;
+	unsigned int lsize, sets, ways;
+	unsigned int config;
+
+	config = read_cpucfg(LOONGARCH_CPUCFG19);
+	lsize = 1 << ((config & CPUCFG19_L2_SIZE_M) >> CPUCFG19_L2_SIZE);
+	sets  = 1 << ((config & CPUCFG19_L2_SETS_M) >> CPUCFG19_L2_SETS);
+	ways  = ((config & CPUCFG19_L2_WAYS_M) >> CPUCFG19_L2_WAYS) + 1;
+
+	c->vcache.linesz = lsize;
+	c->vcache.sets = sets;
+	c->vcache.ways = ways;
+	vcache_size = lsize * sets * ways;
+	c->vcache.waysize = vcache_size / c->vcache.ways;
+
+	pr_info("Unified victim cache %ldkB %s, linesize %d bytes.\n",
+		vcache_size >> 10, way_string[c->vcache.ways], c->vcache.linesz);
+}
+
+static void probe_scache(void)
+{
+	struct cpuinfo_loongarch *c = &current_cpu_data;
+	unsigned int lsize, sets, ways;
+	unsigned int config;
+
+	config = read_cpucfg(LOONGARCH_CPUCFG20);
+	lsize = 1 << ((config & CPUCFG20_L3_SIZE_M) >> CPUCFG20_L3_SIZE);
+	sets  = 1 << ((config & CPUCFG20_L3_SETS_M) >> CPUCFG20_L3_SETS);
+	ways  = ((config & CPUCFG20_L3_WAYS_M) >> CPUCFG20_L3_WAYS) + 1;
+
+	c->scache.linesz = lsize;
+	c->scache.sets = sets;
+	c->scache.ways = ways;
+	/* 4 cores. scaches are shared */
+	scache_size = lsize * sets * ways;
+	c->scache.waysize = scache_size / c->scache.ways;
+
+	pr_info("Unified secondary cache %ldkB %s, linesize %d bytes.\n",
+		scache_size >> 10, way_string[c->scache.ways], c->scache.linesz);
+}
+
+void cpu_cache_init(void)
+{
+	probe_pcache();
+	probe_vcache();
+	probe_scache();
+
+	shm_align_mask = PAGE_SIZE - 1;
+}
diff --git a/arch/loongarch/mm/extable.c b/arch/loongarch/mm/extable.c
new file mode 100644
index 000000000000..bc20988f2b87
--- /dev/null
+++ b/arch/loongarch/mm/extable.c
@@ -0,0 +1,22 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2020-2022 Loongson Technology Corporation Limited
+ */
+#include <linux/extable.h>
+#include <linux/spinlock.h>
+#include <asm/branch.h>
+#include <linux/uaccess.h>
+
+int fixup_exception(struct pt_regs *regs)
+{
+	const struct exception_table_entry *fixup;
+
+	fixup = search_exception_tables(exception_era(regs));
+	if (fixup) {
+		regs->csr_era = fixup->fixup;
+
+		return 1;
+	}
+
+	return 0;
+}
diff --git a/arch/loongarch/mm/fault.c b/arch/loongarch/mm/fault.c
new file mode 100644
index 000000000000..605579b19a00
--- /dev/null
+++ b/arch/loongarch/mm/fault.c
@@ -0,0 +1,261 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2020-2022 Loongson Technology Corporation Limited
+ *
+ * Derived from MIPS:
+ * Copyright (C) 1995 - 2000 by Ralf Baechle
+ */
+#include <linux/context_tracking.h>
+#include <linux/signal.h>
+#include <linux/sched.h>
+#include <linux/interrupt.h>
+#include <linux/kernel.h>
+#include <linux/entry-common.h>
+#include <linux/errno.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/ptrace.h>
+#include <linux/ratelimit.h>
+#include <linux/mman.h>
+#include <linux/mm.h>
+#include <linux/smp.h>
+#include <linux/kdebug.h>
+#include <linux/kprobes.h>
+#include <linux/perf_event.h>
+#include <linux/uaccess.h>
+
+#include <asm/branch.h>
+#include <asm/mmu_context.h>
+#include <asm/ptrace.h>
+
+int show_unhandled_signals = 1;
+
+static void __kprobes no_context(struct pt_regs *regs, unsigned long address)
+{
+	const int field = sizeof(unsigned long) * 2;
+
+	/* Are we prepared to handle this kernel fault?	 */
+	if (fixup_exception(regs))
+		return;
+
+	/*
+	 * Oops. The kernel tried to access some bad page. We'll have to
+	 * terminate things with extreme prejudice.
+	 */
+	bust_spinlocks(1);
+
+	pr_alert("CPU %d Unable to handle kernel paging request at "
+	       "virtual address %0*lx, era == %0*lx, ra == %0*lx\n",
+	       raw_smp_processor_id(), field, address, field, regs->csr_era,
+	       field,  regs->regs[1]);
+	die("Oops", regs);
+}
+
+static void __kprobes do_out_of_memory(struct pt_regs *regs, unsigned long address)
+{
+	/*
+	 * We ran out of memory, call the OOM killer, and return the userspace
+	 * (which will retry the fault, or kill us if we got oom-killed).
+	 */
+	if (!user_mode(regs)) {
+		no_context(regs, address);
+		return;
+	}
+	pagefault_out_of_memory();
+}
+
+static void __kprobes do_sigbus(struct pt_regs *regs,
+		unsigned long write, unsigned long address, int si_code)
+{
+	/* Kernel mode? Handle exceptions or die */
+	if (!user_mode(regs)) {
+		no_context(regs, address);
+		return;
+	}
+
+	/*
+	 * Send a sigbus, regardless of whether we were in kernel
+	 * or user mode.
+	 */
+	current->thread.csr_badvaddr = address;
+	current->thread.trap_nr = read_csr_excode();
+	force_sig_fault(SIGBUS, BUS_ADRERR, (void __user *)address);
+}
+
+static void __kprobes do_sigsegv(struct pt_regs *regs,
+		unsigned long write, unsigned long address, int si_code)
+{
+	const int field = sizeof(unsigned long) * 2;
+	static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
+
+	/* Kernel mode? Handle exceptions or die */
+	if (!user_mode(regs)) {
+		no_context(regs, address);
+		return;
+	}
+
+	/* User mode accesses just cause a SIGSEGV */
+	current->thread.csr_badvaddr = address;
+	if (!write)
+		current->thread.error_code = 1;
+	else
+		current->thread.error_code = 2;
+	current->thread.trap_nr = read_csr_excode();
+
+	if (show_unhandled_signals &&
+	    unhandled_signal(current, SIGSEGV) && __ratelimit(&ratelimit_state)) {
+		pr_info("do_page_fault(): sending SIGSEGV to %s for invalid %s %0*lx\n",
+			current->comm,
+			write ? "write access to" : "read access from",
+			field, address);
+		pr_info("era = %0*lx in", field,
+			(unsigned long) regs->csr_era);
+		print_vma_addr(KERN_CONT " ", regs->csr_era);
+		pr_cont("\n");
+		pr_info("ra  = %0*lx in", field,
+			(unsigned long) regs->regs[1]);
+		print_vma_addr(KERN_CONT " ", regs->regs[1]);
+		pr_cont("\n");
+	}
+	force_sig_fault(SIGSEGV, si_code, (void __user *)address);
+}
+
+/*
+ * This routine handles page faults.  It determines the address,
+ * and the problem, and then passes it off to one of the appropriate
+ * routines.
+ */
+static void __kprobes __do_page_fault(struct pt_regs *regs,
+			unsigned long write, unsigned long address)
+{
+	int si_code = SEGV_MAPERR;
+	unsigned int flags = FAULT_FLAG_DEFAULT;
+	struct task_struct *tsk = current;
+	struct mm_struct *mm = tsk->mm;
+	struct vm_area_struct *vma = NULL;
+	vm_fault_t fault;
+
+	/*
+	 * We fault-in kernel-space virtual memory on-demand. The
+	 * 'reference' page table is init_mm.pgd.
+	 *
+	 * NOTE! We MUST NOT take any locks for this case. We may
+	 * be in an interrupt or a critical region, and should
+	 * only copy the information from the master page table,
+	 * nothing more.
+	 */
+	if (address & __UA_LIMIT) {
+		if (!user_mode(regs))
+			no_context(regs, address);
+		else
+			do_sigsegv(regs, write, address, si_code);
+		return;
+	}
+
+	/*
+	 * If we're in an interrupt or have no user
+	 * context, we must not take the fault..
+	 */
+	if (faulthandler_disabled() || !mm) {
+		do_sigsegv(regs, write, address, si_code);
+		return;
+	}
+
+	if (user_mode(regs))
+		flags |= FAULT_FLAG_USER;
+
+	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
+retry:
+	mmap_read_lock(mm);
+	vma = find_vma(mm, address);
+	if (!vma)
+		goto bad_area;
+	if (vma->vm_start <= address)
+		goto good_area;
+	if (!(vma->vm_flags & VM_GROWSDOWN))
+		goto bad_area;
+	if (!expand_stack(vma, address))
+		goto good_area;
+/*
+ * Something tried to access memory that isn't in our memory map..
+ * Fix it, but check if it's kernel or user first..
+ */
+bad_area:
+	mmap_read_unlock(mm);
+	do_sigsegv(regs, write, address, si_code);
+	return;
+
+/*
+ * Ok, we have a good vm_area for this memory access, so
+ * we can handle it..
+ */
+good_area:
+	si_code = SEGV_ACCERR;
+
+	if (write) {
+		flags |= FAULT_FLAG_WRITE;
+		if (!(vma->vm_flags & VM_WRITE))
+			goto bad_area;
+	} else {
+		if (!(vma->vm_flags & VM_READ) && address != exception_era(regs))
+			goto bad_area;
+		if (!(vma->vm_flags & VM_EXEC) && address == exception_era(regs))
+			goto bad_area;
+	}
+
+	/*
+	 * If for any reason at all we couldn't handle the fault,
+	 * make sure we exit gracefully rather than endlessly redo
+	 * the fault.
+	 */
+	fault = handle_mm_fault(vma, address, flags, regs);
+
+	if (fault_signal_pending(fault, regs)) {
+		if (!user_mode(regs))
+			no_context(regs, address);
+		return;
+	}
+
+	if (unlikely(fault & VM_FAULT_RETRY)) {
+		flags |= FAULT_FLAG_TRIED;
+
+		/*
+		 * No need to mmap_read_unlock(mm) as we would
+		 * have already released it in __lock_page_or_retry
+		 * in mm/filemap.c.
+		 */
+		goto retry;
+	}
+	if (unlikely(fault & VM_FAULT_ERROR)) {
+		mmap_read_unlock(mm);
+		if (fault & VM_FAULT_OOM) {
+			do_out_of_memory(regs, address);
+			return;
+		} else if (fault & VM_FAULT_SIGSEGV) {
+			do_sigsegv(regs, write, address, si_code);
+			return;
+		} else if (fault & (VM_FAULT_SIGBUS|VM_FAULT_HWPOISON|VM_FAULT_HWPOISON_LARGE)) {
+			do_sigbus(regs, write, address, si_code);
+			return;
+		}
+		BUG();
+	}
+
+	mmap_read_unlock(mm);
+}
+
+asmlinkage void __kprobes do_page_fault(struct pt_regs *regs,
+			unsigned long write, unsigned long address)
+{
+	irqentry_state_t state = irqentry_enter(regs);
+
+	/* Enable interrupt if enabled in parent context */
+	if (likely(regs->csr_prmd & CSR_PRMD_PIE))
+		local_irq_enable();
+
+	__do_page_fault(regs, write, address);
+
+	local_irq_disable();
+
+	irqentry_exit(regs, state);
+}
diff --git a/arch/loongarch/mm/hugetlbpage.c b/arch/loongarch/mm/hugetlbpage.c
new file mode 100644
index 000000000000..ba138117b124
--- /dev/null
+++ b/arch/loongarch/mm/hugetlbpage.c
@@ -0,0 +1,87 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2020-2022 Loongson Technology Corporation Limited
+ */
+
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/hugetlb.h>
+#include <linux/pagemap.h>
+#include <linux/err.h>
+#include <linux/sysctl.h>
+#include <asm/mman.h>
+#include <asm/tlb.h>
+#include <asm/tlbflush.h>
+
+pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
+		      unsigned long addr, unsigned long sz)
+{
+	pgd_t *pgd;
+	p4d_t *p4d;
+	pud_t *pud;
+	pte_t *pte = NULL;
+
+	pgd = pgd_offset(mm, addr);
+	p4d = p4d_alloc(mm, pgd, addr);
+	pud = pud_alloc(mm, p4d, addr);
+	if (pud)
+		pte = (pte_t *)pmd_alloc(mm, pud, addr);
+
+	return pte;
+}
+
+pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr,
+		       unsigned long sz)
+{
+	pgd_t *pgd;
+	p4d_t *p4d;
+	pud_t *pud;
+	pmd_t *pmd = NULL;
+
+	pgd = pgd_offset(mm, addr);
+	if (pgd_present(*pgd)) {
+		p4d = p4d_offset(pgd, addr);
+		if (p4d_present(*p4d)) {
+			pud = pud_offset(p4d, addr);
+			if (pud_present(*pud))
+				pmd = pmd_offset(pud, addr);
+		}
+	}
+	return (pte_t *) pmd;
+}
+
+/*
+ * This function checks for proper alignment of input addr and len parameters.
+ */
+int is_aligned_hugepage_range(unsigned long addr, unsigned long len)
+{
+	if (len & ~HPAGE_MASK)
+		return -EINVAL;
+	if (addr & ~HPAGE_MASK)
+		return -EINVAL;
+	return 0;
+}
+
+int pmd_huge(pmd_t pmd)
+{
+	return (pmd_val(pmd) & _PAGE_HUGE) != 0;
+}
+
+int pud_huge(pud_t pud)
+{
+	return (pud_val(pud) & _PAGE_HUGE) != 0;
+}
+
+uint64_t pmd_to_entrylo(unsigned long pmd_val)
+{
+	uint64_t val;
+	/* PMD as PTE. Must be huge page */
+	if (!pmd_huge(__pmd(pmd_val)))
+		panic("%s", __func__);
+
+	val = pmd_val ^ _PAGE_HUGE;
+	val |= ((val & _PAGE_HGLOBAL) >>
+		(_PAGE_HGLOBAL_SHIFT - _PAGE_GLOBAL_SHIFT));
+
+	return val;
+}
diff --git a/arch/loongarch/mm/init.c b/arch/loongarch/mm/init.c
new file mode 100644
index 000000000000..afd6634ce171
--- /dev/null
+++ b/arch/loongarch/mm/init.c
@@ -0,0 +1,165 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2020-2022 Loongson Technology Corporation Limited
+ */
+#include <linux/init.h>
+#include <linux/export.h>
+#include <linux/signal.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/pagemap.h>
+#include <linux/memblock.h>
+#include <linux/memremap.h>
+#include <linux/mm.h>
+#include <linux/mman.h>
+#include <linux/highmem.h>
+#include <linux/swap.h>
+#include <linux/proc_fs.h>
+#include <linux/pfn.h>
+#include <linux/hardirq.h>
+#include <linux/gfp.h>
+#include <linux/initrd.h>
+#include <linux/mmzone.h>
+
+#include <asm/asm-offsets.h>
+#include <asm/bootinfo.h>
+#include <asm/cpu.h>
+#include <asm/dma.h>
+#include <asm/mmu_context.h>
+#include <asm/sections.h>
+#include <asm/pgtable.h>
+#include <asm/pgalloc.h>
+#include <asm/tlb.h>
+
+/*
+ * We have up to 8 empty zeroed pages so we can map one of the right colour
+ * when needed.	 Since page is never written to after the initialization we
+ * don't have to care about aliases on other CPUs.
+ */
+unsigned long empty_zero_page, zero_page_mask;
+EXPORT_SYMBOL_GPL(empty_zero_page);
+EXPORT_SYMBOL(zero_page_mask);
+
+void setup_zero_pages(void)
+{
+	unsigned int order, i;
+	struct page *page;
+
+	order = 0;
+
+	empty_zero_page = __get_free_pages(GFP_KERNEL | __GFP_ZERO, order);
+	if (!empty_zero_page)
+		panic("Oh boy, that early out of memory?");
+
+	page = virt_to_page((void *)empty_zero_page);
+	split_page(page, order);
+	for (i = 0; i < (1 << order); i++, page++)
+		mark_page_reserved(page);
+
+	zero_page_mask = ((PAGE_SIZE << order) - 1) & PAGE_MASK;
+}
+
+void copy_user_highpage(struct page *to, struct page *from,
+	unsigned long vaddr, struct vm_area_struct *vma)
+{
+	void *vfrom, *vto;
+
+	vto = kmap_atomic(to);
+	vfrom = kmap_atomic(from);
+	copy_page(vto, vfrom);
+	kunmap_atomic(vfrom);
+	kunmap_atomic(vto);
+	/* Make sure this page is cleared on other CPU's too before using it */
+	smp_wmb();
+}
+
+int __ref page_is_ram(unsigned long pfn)
+{
+	unsigned long addr = PFN_PHYS(pfn);
+
+	return memblock_is_memory(addr) && !memblock_is_reserved(addr);
+}
+
+void __init paging_init(void)
+{
+	unsigned long max_zone_pfns[MAX_NR_ZONES];
+
+#ifdef CONFIG_ZONE_DMA
+	max_zone_pfns[ZONE_DMA] = MAX_DMA_PFN;
+#endif
+#ifdef CONFIG_ZONE_DMA32
+	max_zone_pfns[ZONE_DMA32] = MAX_DMA32_PFN;
+#endif
+	max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
+
+	free_area_init(max_zone_pfns);
+}
+
+void __init mem_init(void)
+{
+	max_mapnr = max_low_pfn;
+	high_memory = (void *) __va(max_low_pfn << PAGE_SHIFT);
+
+	memblock_free_all();
+	setup_zero_pages();	/* Setup zeroed pages.  */
+}
+
+void __ref free_initmem(void)
+{
+	free_initmem_default(POISON_FREE_INITMEM);
+}
+
+#ifdef CONFIG_MEMORY_HOTPLUG
+int arch_add_memory(int nid, u64 start, u64 size, struct mhp_params *params)
+{
+	unsigned long start_pfn = start >> PAGE_SHIFT;
+	unsigned long nr_pages = size >> PAGE_SHIFT;
+	int ret;
+
+	ret = __add_pages(nid, start_pfn, nr_pages, params);
+
+	if (ret)
+		pr_warn("%s: Problem encountered in __add_pages() as ret=%d\n",
+				__func__,  ret);
+
+	return ret;
+}
+
+#ifdef CONFIG_MEMORY_HOTREMOVE
+void arch_remove_memory(u64 start, u64 size, struct vmem_altmap *altmap)
+{
+	unsigned long start_pfn = start >> PAGE_SHIFT;
+	unsigned long nr_pages = size >> PAGE_SHIFT;
+	struct page *page = pfn_to_page(start_pfn);
+
+	/* With altmap the first mapped page is offset from @start */
+	if (altmap)
+		page += vmem_altmap_offset(altmap);
+	__remove_pages(start_pfn, nr_pages, altmap);
+}
+#endif
+#endif
+
+/*
+ * Align swapper_pg_dir in to 64K, allows its address to be loaded
+ * with a single LUI instruction in the TLB handlers.  If we used
+ * __aligned(64K), its size would get rounded up to the alignment
+ * size, and waste space.  So we place it in its own section and align
+ * it in the linker script.
+ */
+pgd_t swapper_pg_dir[_PTRS_PER_PGD] __section(".bss..swapper_pg_dir");
+
+pgd_t invalid_pg_dir[_PTRS_PER_PGD] __page_aligned_bss;
+#ifndef __PAGETABLE_PUD_FOLDED
+pud_t invalid_pud_table[PTRS_PER_PUD] __page_aligned_bss;
+#endif
+#ifndef __PAGETABLE_PMD_FOLDED
+pmd_t invalid_pmd_table[PTRS_PER_PMD] __page_aligned_bss;
+EXPORT_SYMBOL_GPL(invalid_pmd_table);
+#endif
+pte_t invalid_pte_table[PTRS_PER_PTE] __page_aligned_bss;
+EXPORT_SYMBOL(invalid_pte_table);
diff --git a/arch/loongarch/mm/ioremap.c b/arch/loongarch/mm/ioremap.c
new file mode 100644
index 000000000000..73b0980ab6f5
--- /dev/null
+++ b/arch/loongarch/mm/ioremap.c
@@ -0,0 +1,27 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2020-2022 Loongson Technology Corporation Limited
+ */
+
+#include <asm/io.h>
+
+void __init __iomem *early_ioremap(u64 phys_addr, unsigned long size)
+{
+	return ((void __iomem *)TO_CACHE(phys_addr));
+}
+
+void __init early_iounmap(void __iomem *addr, unsigned long size)
+{
+
+}
+
+void *early_memremap_ro(resource_size_t phys_addr, unsigned long size)
+{
+	return early_memremap(phys_addr, size);
+}
+
+void *early_memremap_prot(resource_size_t phys_addr, unsigned long size,
+		    unsigned long prot_val)
+{
+	return early_memremap(phys_addr, size);
+}
diff --git a/arch/loongarch/mm/maccess.c b/arch/loongarch/mm/maccess.c
new file mode 100644
index 000000000000..58173842c6be
--- /dev/null
+++ b/arch/loongarch/mm/maccess.c
@@ -0,0 +1,10 @@
+// SPDX-License-Identifier: GPL-2.0-only
+
+#include <linux/uaccess.h>
+#include <linux/kernel.h>
+
+bool copy_from_kernel_nofault_allowed(const void *unsafe_src, size_t size)
+{
+	/* highest bit set means kernel space */
+	return (unsigned long)unsafe_src >> (BITS_PER_LONG - 1);
+}
diff --git a/arch/loongarch/mm/mmap.c b/arch/loongarch/mm/mmap.c
new file mode 100644
index 000000000000..52e40f0ba732
--- /dev/null
+++ b/arch/loongarch/mm/mmap.c
@@ -0,0 +1,125 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2020-2022 Loongson Technology Corporation Limited
+ */
+#include <linux/compiler.h>
+#include <linux/elf-randomize.h>
+#include <linux/errno.h>
+#include <linux/mm.h>
+#include <linux/mman.h>
+#include <linux/export.h>
+#include <linux/personality.h>
+#include <linux/random.h>
+#include <linux/sched/signal.h>
+#include <linux/sched/mm.h>
+
+unsigned long shm_align_mask = PAGE_SIZE - 1;	/* Sane caches */
+EXPORT_SYMBOL(shm_align_mask);
+
+#define COLOUR_ALIGN(addr, pgoff)				\
+	((((addr) + shm_align_mask) & ~shm_align_mask) +	\
+	 (((pgoff) << PAGE_SHIFT) & shm_align_mask))
+
+enum mmap_allocation_direction {UP, DOWN};
+
+static unsigned long arch_get_unmapped_area_common(struct file *filp,
+	unsigned long addr0, unsigned long len, unsigned long pgoff,
+	unsigned long flags, enum mmap_allocation_direction dir)
+{
+	struct mm_struct *mm = current->mm;
+	struct vm_area_struct *vma;
+	unsigned long addr = addr0;
+	int do_color_align;
+	struct vm_unmapped_area_info info;
+
+	if (unlikely(len > TASK_SIZE))
+		return -ENOMEM;
+
+	if (flags & MAP_FIXED) {
+		/* Even MAP_FIXED mappings must reside within TASK_SIZE */
+		if (TASK_SIZE - len < addr)
+			return -EINVAL;
+
+		/*
+		 * We do not accept a shared mapping if it would violate
+		 * cache aliasing constraints.
+		 */
+		if ((flags & MAP_SHARED) &&
+		    ((addr - (pgoff << PAGE_SHIFT)) & shm_align_mask))
+			return -EINVAL;
+		return addr;
+	}
+
+	do_color_align = 0;
+	if (filp || (flags & MAP_SHARED))
+		do_color_align = 1;
+
+	/* requesting a specific address */
+	if (addr) {
+		if (do_color_align)
+			addr = COLOUR_ALIGN(addr, pgoff);
+		else
+			addr = PAGE_ALIGN(addr);
+
+		vma = find_vma(mm, addr);
+		if (TASK_SIZE - len >= addr &&
+		    (!vma || addr + len <= vm_start_gap(vma)))
+			return addr;
+	}
+
+	info.length = len;
+	info.align_mask = do_color_align ? (PAGE_MASK & shm_align_mask) : 0;
+	info.align_offset = pgoff << PAGE_SHIFT;
+
+	if (dir == DOWN) {
+		info.flags = VM_UNMAPPED_AREA_TOPDOWN;
+		info.low_limit = PAGE_SIZE;
+		info.high_limit = mm->mmap_base;
+		addr = vm_unmapped_area(&info);
+
+		if (!(addr & ~PAGE_MASK))
+			return addr;
+
+		/*
+		 * A failed mmap() very likely causes application failure,
+		 * so fall back to the bottom-up function here. This scenario
+		 * can happen with large stack limits and large mmap()
+		 * allocations.
+		 */
+	}
+
+	info.flags = 0;
+	info.low_limit = mm->mmap_base;
+	info.high_limit = TASK_SIZE;
+	return vm_unmapped_area(&info);
+}
+
+unsigned long arch_get_unmapped_area(struct file *filp, unsigned long addr0,
+	unsigned long len, unsigned long pgoff, unsigned long flags)
+{
+	return arch_get_unmapped_area_common(filp,
+			addr0, len, pgoff, flags, UP);
+}
+
+/*
+ * There is no need to export this but sched.h declares the function as
+ * extern so making it static here results in an error.
+ */
+unsigned long arch_get_unmapped_area_topdown(struct file *filp,
+	unsigned long addr0, unsigned long len, unsigned long pgoff,
+	unsigned long flags)
+{
+	return arch_get_unmapped_area_common(filp,
+			addr0, len, pgoff, flags, DOWN);
+}
+
+int __virt_addr_valid(volatile void *kaddr)
+{
+	unsigned long vaddr = (unsigned long)kaddr;
+
+	if ((vaddr < PAGE_OFFSET) || (vaddr >= vm_map_base))
+		return 0;
+
+	return pfn_valid(PFN_DOWN(virt_to_phys(kaddr)));
+}
+EXPORT_SYMBOL_GPL(__virt_addr_valid);
diff --git a/arch/loongarch/mm/page.S b/arch/loongarch/mm/page.S
new file mode 100644
index 000000000000..ddc78ab33c7b
--- /dev/null
+++ b/arch/loongarch/mm/page.S
@@ -0,0 +1,84 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2020-2022 Loongson Technology Corporation Limited
+ */
+#include <linux/linkage.h>
+#include <asm/asm.h>
+#include <asm/export.h>
+#include <asm/page.h>
+#include <asm/regdef.h>
+
+	.align 5
+SYM_FUNC_START(clear_page)
+	lu12i.w  t0, 1 << (PAGE_SHIFT - 12)
+	add.d    t0, t0, a0
+1:
+	st.d     zero, a0, 0
+	st.d     zero, a0, 8
+	st.d     zero, a0, 16
+	st.d     zero, a0, 24
+	st.d     zero, a0, 32
+	st.d     zero, a0, 40
+	st.d     zero, a0, 48
+	st.d     zero, a0, 56
+	addi.d   a0,   a0, 128
+	st.d     zero, a0, -64
+	st.d     zero, a0, -56
+	st.d     zero, a0, -48
+	st.d     zero, a0, -40
+	st.d     zero, a0, -32
+	st.d     zero, a0, -24
+	st.d     zero, a0, -16
+	st.d     zero, a0, -8
+	bne      t0,   a0, 1b
+
+	jirl     $r0, ra, 0
+SYM_FUNC_END(clear_page)
+EXPORT_SYMBOL(clear_page)
+
+.align 5
+SYM_FUNC_START(copy_page)
+	lu12i.w  t8, 1 << (PAGE_SHIFT - 12)
+	add.d    t8, t8, a0
+1:
+	ld.d     t0, a1,  0
+	ld.d     t1, a1,  8
+	ld.d     t2, a1,  16
+	ld.d     t3, a1,  24
+	ld.d     t4, a1,  32
+	ld.d     t5, a1,  40
+	ld.d     t6, a1,  48
+	ld.d     t7, a1,  56
+
+	st.d     t0, a0,  0
+	st.d     t1, a0,  8
+	ld.d     t0, a1,  64
+	ld.d     t1, a1,  72
+	st.d     t2, a0,  16
+	st.d     t3, a0,  24
+	ld.d     t2, a1,  80
+	ld.d     t3, a1,  88
+	st.d     t4, a0,  32
+	st.d     t5, a0,  40
+	ld.d     t4, a1,  96
+	ld.d     t5, a1,  104
+	st.d     t6, a0,  48
+	st.d     t7, a0,  56
+	ld.d     t6, a1,  112
+	ld.d     t7, a1,  120
+	addi.d   a0, a0,  128
+	addi.d   a1, a1,  128
+
+	st.d     t0, a0,  -64
+	st.d     t1, a0,  -56
+	st.d     t2, a0,  -48
+	st.d     t3, a0,  -40
+	st.d     t4, a0,  -32
+	st.d     t5, a0,  -24
+	st.d     t6, a0,  -16
+	st.d     t7, a0,  -8
+
+	bne      t8, a0, 1b
+	jirl     $r0, ra, 0
+SYM_FUNC_END(copy_page)
+EXPORT_SYMBOL(copy_page)
diff --git a/arch/loongarch/mm/pgtable.c b/arch/loongarch/mm/pgtable.c
new file mode 100644
index 000000000000..0569647152e9
--- /dev/null
+++ b/arch/loongarch/mm/pgtable.c
@@ -0,0 +1,130 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2020-2022 Loongson Technology Corporation Limited
+ */
+#include <linux/init.h>
+#include <linux/export.h>
+#include <linux/mm.h>
+#include <asm/pgalloc.h>
+#include <asm/pgtable.h>
+#include <asm/tlbflush.h>
+
+pgd_t *pgd_alloc(struct mm_struct *mm)
+{
+	pgd_t *ret, *init;
+
+	ret = (pgd_t *) __get_free_pages(GFP_KERNEL, PGD_ORDER);
+	if (ret) {
+		init = pgd_offset(&init_mm, 0UL);
+		pgd_init((unsigned long)ret);
+		memcpy(ret + USER_PTRS_PER_PGD, init + USER_PTRS_PER_PGD,
+		       (PTRS_PER_PGD - USER_PTRS_PER_PGD) * sizeof(pgd_t));
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(pgd_alloc);
+
+void pgd_init(unsigned long page)
+{
+	unsigned long *p, *end;
+	unsigned long entry;
+
+#if !defined(__PAGETABLE_PUD_FOLDED)
+	entry = (unsigned long)invalid_pud_table;
+#elif !defined(__PAGETABLE_PMD_FOLDED)
+	entry = (unsigned long)invalid_pmd_table;
+#else
+	entry = (unsigned long)invalid_pte_table;
+#endif
+
+	p = (unsigned long *) page;
+	end = p + PTRS_PER_PGD;
+
+	do {
+		p[0] = entry;
+		p[1] = entry;
+		p[2] = entry;
+		p[3] = entry;
+		p[4] = entry;
+		p += 8;
+		p[-3] = entry;
+		p[-2] = entry;
+		p[-1] = entry;
+	} while (p != end);
+}
+EXPORT_SYMBOL_GPL(pgd_init);
+
+#ifndef __PAGETABLE_PMD_FOLDED
+void pmd_init(unsigned long addr, unsigned long pagetable)
+{
+	unsigned long *p, *end;
+
+	p = (unsigned long *) addr;
+	end = p + PTRS_PER_PMD;
+
+	do {
+		p[0] = pagetable;
+		p[1] = pagetable;
+		p[2] = pagetable;
+		p[3] = pagetable;
+		p[4] = pagetable;
+		p += 8;
+		p[-3] = pagetable;
+		p[-2] = pagetable;
+		p[-1] = pagetable;
+	} while (p != end);
+}
+EXPORT_SYMBOL_GPL(pmd_init);
+#endif
+
+#ifndef __PAGETABLE_PUD_FOLDED
+void pud_init(unsigned long addr, unsigned long pagetable)
+{
+	unsigned long *p, *end;
+
+	p = (unsigned long *)addr;
+	end = p + PTRS_PER_PUD;
+
+	do {
+		p[0] = pagetable;
+		p[1] = pagetable;
+		p[2] = pagetable;
+		p[3] = pagetable;
+		p[4] = pagetable;
+		p += 8;
+		p[-3] = pagetable;
+		p[-2] = pagetable;
+		p[-1] = pagetable;
+	} while (p != end);
+}
+#endif
+
+pmd_t mk_pmd(struct page *page, pgprot_t prot)
+{
+	pmd_t pmd;
+
+	pmd_val(pmd) = (page_to_pfn(page) << _PFN_SHIFT) | pgprot_val(prot);
+
+	return pmd;
+}
+
+void set_pmd_at(struct mm_struct *mm, unsigned long addr,
+		pmd_t *pmdp, pmd_t pmd)
+{
+	*pmdp = pmd;
+	flush_tlb_all();
+}
+
+void __init pagetable_init(void)
+{
+	/* Initialize the entire pgd.  */
+	pgd_init((unsigned long)swapper_pg_dir);
+	pgd_init((unsigned long)invalid_pg_dir);
+#ifndef __PAGETABLE_PUD_FOLDED
+	pud_init((unsigned long)invalid_pud_table, (unsigned long)invalid_pmd_table);
+#endif
+#ifndef __PAGETABLE_PMD_FOLDED
+	pmd_init((unsigned long)invalid_pmd_table, (unsigned long)invalid_pte_table);
+#endif
+}
diff --git a/arch/loongarch/mm/tlb.c b/arch/loongarch/mm/tlb.c
new file mode 100644
index 000000000000..78dd328cec77
--- /dev/null
+++ b/arch/loongarch/mm/tlb.c
@@ -0,0 +1,282 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2020-2022 Loongson Technology Corporation Limited
+ */
+#include <linux/init.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/mm.h>
+#include <linux/hugetlb.h>
+#include <linux/export.h>
+
+#include <asm/cpu.h>
+#include <asm/bootinfo.h>
+#include <asm/mmu_context.h>
+#include <asm/pgtable.h>
+#include <asm/tlb.h>
+
+void local_flush_tlb_all(void)
+{
+	invtlb_all(INVTLB_CURRENT_ALL, 0, 0);
+}
+EXPORT_SYMBOL(local_flush_tlb_all);
+
+void local_flush_tlb_user(void)
+{
+	invtlb_all(INVTLB_CURRENT_GFALSE, 0, 0);
+}
+EXPORT_SYMBOL(local_flush_tlb_user);
+
+void local_flush_tlb_kernel(void)
+{
+	invtlb_all(INVTLB_CURRENT_GTRUE, 0, 0);
+}
+EXPORT_SYMBOL(local_flush_tlb_kernel);
+
+/*
+ * All entries common to a mm share an asid. To effectively flush
+ * these entries, we just bump the asid.
+ */
+void local_flush_tlb_mm(struct mm_struct *mm)
+{
+	int cpu;
+
+	preempt_disable();
+
+	cpu = smp_processor_id();
+
+	if (asid_valid(mm, cpu))
+		drop_mmu_context(mm, cpu);
+	else
+		cpumask_clear_cpu(cpu, mm_cpumask(mm));
+
+	preempt_enable();
+}
+
+void local_flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
+	unsigned long end)
+{
+	struct mm_struct *mm = vma->vm_mm;
+	int cpu = smp_processor_id();
+
+	if (asid_valid(mm, cpu)) {
+		unsigned long size, flags;
+
+		local_irq_save(flags);
+		start = round_down(start, PAGE_SIZE << 1);
+		end = round_up(end, PAGE_SIZE << 1);
+		size = (end - start) >> (PAGE_SHIFT + 1);
+		if (size <= (current_cpu_data.tlbsizestlbsets ?
+			     current_cpu_data.tlbsize / 8 :
+			     current_cpu_data.tlbsize / 2)) {
+			int asid = cpu_asid(cpu, mm);
+
+			while (start < end) {
+				invtlb(INVTLB_ADDR_GFALSE_AND_ASID, asid, start);
+				start += (PAGE_SIZE << 1);
+			}
+		} else {
+			drop_mmu_context(mm, cpu);
+		}
+		local_irq_restore(flags);
+	} else {
+		cpumask_clear_cpu(cpu, mm_cpumask(mm));
+	}
+}
+
+void local_flush_tlb_kernel_range(unsigned long start, unsigned long end)
+{
+	unsigned long size, flags;
+
+	local_irq_save(flags);
+	size = (end - start + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
+	size = (size + 1) >> 1;
+	if (size <= (current_cpu_data.tlbsizestlbsets ?
+		     current_cpu_data.tlbsize / 8 :
+		     current_cpu_data.tlbsize / 2)) {
+
+		start &= (PAGE_MASK << 1);
+		end += ((PAGE_SIZE << 1) - 1);
+		end &= (PAGE_MASK << 1);
+
+		while (start < end) {
+			invtlb_addr(INVTLB_ADDR_GTRUE_OR_ASID, 0, start);
+			start += (PAGE_SIZE << 1);
+		}
+	} else {
+		local_flush_tlb_kernel();
+	}
+	local_irq_restore(flags);
+}
+
+void local_flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
+{
+	int cpu = smp_processor_id();
+
+	if (asid_valid(vma->vm_mm, cpu)) {
+		int newpid;
+
+		newpid = cpu_asid(cpu, vma->vm_mm);
+		page &= (PAGE_MASK << 1);
+		invtlb(INVTLB_ADDR_GFALSE_AND_ASID, newpid, page);
+	} else {
+		cpumask_clear_cpu(cpu, mm_cpumask(vma->vm_mm));
+	}
+}
+
+/*
+ * This one is only used for pages with the global bit set so we don't care
+ * much about the ASID.
+ */
+void local_flush_tlb_one(unsigned long page)
+{
+	page &= (PAGE_MASK << 1);
+	invtlb_addr(INVTLB_ADDR_GTRUE_OR_ASID, 0, page);
+}
+
+static void __update_hugetlb(struct vm_area_struct *vma, unsigned long address, pte_t *ptep)
+{
+#ifdef CONFIG_HUGETLB_PAGE
+	int idx;
+	unsigned long lo;
+	unsigned long flags;
+
+	local_irq_save(flags);
+
+	address &= (PAGE_MASK << 1);
+	write_csr_entryhi(address);
+	tlb_probe();
+	idx = read_csr_tlbidx();
+	write_csr_pagesize(PS_HUGE_SIZE);
+	lo = pmd_to_entrylo(pte_val(*ptep));
+	write_csr_entrylo0(lo);
+	write_csr_entrylo1(lo + (HPAGE_SIZE >> 1));
+
+	if (idx < 0)
+		tlb_write_random();
+	else
+		tlb_write_indexed();
+	write_csr_pagesize(PS_DEFAULT_SIZE);
+
+	local_irq_restore(flags);
+#endif
+}
+
+void __update_tlb(struct vm_area_struct *vma, unsigned long address, pte_t *ptep)
+{
+	int idx;
+	unsigned long flags;
+
+	/*
+	 * Handle debugger faulting in for debugee.
+	 */
+	if (current->active_mm != vma->vm_mm)
+		return;
+
+	if (pte_val(*ptep) & _PAGE_HUGE) {
+		__update_hugetlb(vma, address, ptep);
+		return;
+	}
+
+	local_irq_save(flags);
+
+	if ((unsigned long)ptep & sizeof(pte_t))
+		ptep--;
+
+	address &= (PAGE_MASK << 1);
+	write_csr_entryhi(address);
+	tlb_probe();
+	idx = read_csr_tlbidx();
+	write_csr_pagesize(PS_DEFAULT_SIZE);
+	write_csr_entrylo0(pte_val(*ptep++));
+	write_csr_entrylo1(pte_val(*ptep));
+	if (idx < 0)
+		tlb_write_random();
+	else
+		tlb_write_indexed();
+
+	local_irq_restore(flags);
+}
+
+static void setup_ptwalker(void)
+{
+	unsigned long pwctl0, pwctl1;
+	unsigned long pgd_i = 0, pgd_w = 0;
+	unsigned long pud_i = 0, pud_w = 0;
+	unsigned long pmd_i = 0, pmd_w = 0;
+	unsigned long pte_i = 0, pte_w = 0;
+
+	pgd_i = PGDIR_SHIFT;
+	pgd_w = PAGE_SHIFT - 3;
+#if CONFIG_PGTABLE_LEVELS > 3
+	pud_i = PUD_SHIFT;
+	pud_w = PAGE_SHIFT - 3;
+#endif
+#if CONFIG_PGTABLE_LEVELS > 2
+	pmd_i = PMD_SHIFT;
+	pmd_w = PAGE_SHIFT - 3;
+#endif
+	pte_i = PAGE_SHIFT;
+	pte_w = PAGE_SHIFT - 3;
+
+	pwctl0 = pte_i | pte_w << 5 | pmd_i << 10 | pmd_w << 15 | pud_i << 20 | pud_w << 25;
+	pwctl1 = pgd_i | pgd_w << 6;
+
+	csr_write64(pwctl0, LOONGARCH_CSR_PWCTL0);
+	csr_write64(pwctl1, LOONGARCH_CSR_PWCTL1);
+	csr_write64((long)swapper_pg_dir, LOONGARCH_CSR_PGDH);
+	csr_write64((long)invalid_pg_dir, LOONGARCH_CSR_PGDL);
+	csr_write64((long)smp_processor_id(), LOONGARCH_CSR_TMID);
+}
+
+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_VALID_SHIFT %d\n", _PAGE_VALID_SHIFT);
+	pr_define("_PAGE_DIRTY_SHIFT %d\n", _PAGE_DIRTY_SHIFT);
+	pr_define("_PAGE_HUGE_SHIFT %d\n", _PAGE_HUGE_SHIFT);
+	pr_define("_PAGE_GLOBAL_SHIFT %d\n", _PAGE_GLOBAL_SHIFT);
+	pr_define("_PAGE_PRESENT_SHIFT %d\n", _PAGE_PRESENT_SHIFT);
+	pr_define("_PAGE_WRITE_SHIFT %d\n", _PAGE_WRITE_SHIFT);
+	pr_define("_PAGE_NO_READ_SHIFT %d\n", _PAGE_NO_READ_SHIFT);
+	pr_define("_PAGE_NO_EXEC_SHIFT %d\n", _PAGE_NO_EXEC_SHIFT);
+	pr_define("_PFN_SHIFT %d\n", _PFN_SHIFT);
+	pr_debug("\n");
+}
+
+void setup_tlb_handler(void)
+{
+	static int run_once = 0;
+
+	setup_ptwalker();
+	output_pgtable_bits_defines();
+
+	/* The tlb handlers are generated only once */
+	if (!run_once) {
+		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);
+		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);
+		run_once++;
+	}
+}
+
+void tlb_init(void)
+{
+	write_csr_pagesize(PS_DEFAULT_SIZE);
+	write_csr_stlbpgsize(PS_DEFAULT_SIZE);
+	write_csr_tlbrefill_pagesize(PS_DEFAULT_SIZE);
+	setup_tlb_handler();
+	local_flush_tlb_all();
+}
diff --git a/arch/loongarch/mm/tlbex.S b/arch/loongarch/mm/tlbex.S
new file mode 100644
index 000000000000..bef740710a3b
--- /dev/null
+++ b/arch/loongarch/mm/tlbex.S
@@ -0,0 +1,477 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2020-2022 Loongson Technology Corporation Limited
+ */
+#include <asm/asm.h>
+#include <asm/export.h>
+#include <asm/loongarch.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/regdef.h>
+#include <asm/stackframe.h>
+
+	.macro tlb_do_page_fault, write
+	SYM_FUNC_START(tlb_do_page_fault_\write)
+	SAVE_ALL
+	csrrd	a2, LOONGARCH_CSR_BADV
+	move	a0, sp
+	REG_S	a2, sp, PT_BVADDR
+	li.w	a1, \write
+	la.abs	t0, do_page_fault
+	jirl    ra, t0, 0
+	RESTORE_ALL_AND_RET
+	SYM_FUNC_END(tlb_do_page_fault_\write)
+	.endm
+
+	tlb_do_page_fault 0
+	tlb_do_page_fault 1
+
+SYM_FUNC_START(handle_tlb_protect)
+	BACKUP_T0T1
+	SAVE_ALL
+	move	a0, sp
+	move	a1, zero
+	csrrd	a2, LOONGARCH_CSR_BADV
+	REG_S	a2, sp, PT_BVADDR
+	la.abs	t0, do_page_fault
+	jirl    ra, t0, 0
+	RESTORE_ALL_AND_RET
+SYM_FUNC_END(handle_tlb_protect)
+
+SYM_FUNC_START(handle_tlb_load)
+	csrwr	t0, EXCEPTION_KS0
+	csrwr	t1, EXCEPTION_KS1
+	csrwr	ra, EXCEPTION_KS2
+
+	/*
+	 * The vmalloc handling is not in the hotpath.
+	 */
+	csrrd	t0, LOONGARCH_CSR_BADV
+	blt	t0, $r0, vmalloc_load
+	csrrd	t1, LOONGARCH_CSR_PGDL
+
+vmalloc_done_load:
+	/* Get PGD offset in bytes */
+	srli.d	t0, t0, PGDIR_SHIFT
+	andi	t0, t0, (PTRS_PER_PGD - 1)
+	slli.d	t0, t0, 3
+	add.d	t1, t1, t0
+#if CONFIG_PGTABLE_LEVELS > 3
+	csrrd	t0, LOONGARCH_CSR_BADV
+	ld.d	t1, t1, 0
+	srli.d	t0, t0, PUD_SHIFT
+	andi	t0, t0, (PTRS_PER_PUD - 1)
+	slli.d	t0, t0, 3
+	add.d	t1, t1, t0
+#endif
+#if CONFIG_PGTABLE_LEVELS > 2
+	csrrd	t0, LOONGARCH_CSR_BADV
+	ld.d	t1, t1, 0
+	srli.d	t0, t0, PMD_SHIFT
+	andi	t0, t0, (PTRS_PER_PMD - 1)
+	slli.d	t0, t0, 3
+	add.d	t1, t1, t0
+#endif
+	ld.d	ra, t1, 0
+
+	/*
+	 * For huge tlb entries, pmde 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.
+	 */
+	andi	t0, ra, _PAGE_HUGE
+	bne	t0, $r0, tlb_huge_update_load
+
+	csrrd	t0, LOONGARCH_CSR_BADV
+	srli.d	t0, t0, (PAGE_SHIFT + PTE_ORDER)
+	andi	t0, t0, (PTRS_PER_PTE - 1)
+	slli.d	t0, t0, _PTE_T_LOG2
+	add.d	t1, ra, t0
+
+	ld.d	t0, t1, 0
+	tlbsrch
+
+	srli.d	ra, t0, _PAGE_PRESENT_SHIFT
+	andi	ra, ra, 1
+	beq	ra, $r0, nopage_tlb_load
+
+	ori	t0, t0, _PAGE_VALID
+	st.d	t0, t1, 0
+	ori	t1, t1, 8
+	xori	t1, t1, 8
+	ld.d	t0, t1, 0
+	ld.d	t1, t1, 8
+	csrwr	t0, LOONGARCH_CSR_TLBELO0
+	csrwr	t1, LOONGARCH_CSR_TLBELO1
+	tlbwr
+leave_load:
+	csrrd	t0, EXCEPTION_KS0
+	csrrd	t1, EXCEPTION_KS1
+	csrrd	ra, EXCEPTION_KS2
+	ertn
+#ifdef CONFIG_64BIT
+vmalloc_load:
+	la.abs	t1, swapper_pg_dir
+	b	vmalloc_done_load
+#endif
+
+	/*
+	 * This is the entry point when build_tlbchange_handler_head
+	 * spots a huge page.
+	 */
+tlb_huge_update_load:
+	ld.d	t0, t1, 0
+	srli.d	ra, t0, _PAGE_PRESENT_SHIFT
+	andi	ra, ra, 1
+	beq	ra, $r0, nopage_tlb_load
+	tlbsrch
+
+	ori	t0, t0, _PAGE_VALID
+	st.d	t0, t1, 0
+	addu16i.d	t1, $r0, -(CSR_TLBIDX_EHINV >> 16)
+	addi.d	ra, t1, 0
+	csrxchg	ra, t1, LOONGARCH_CSR_TLBIDX
+	tlbwr
+
+	csrxchg	$r0, t1, LOONGARCH_CSR_TLBIDX
+
+	/*
+	 * A huge PTE describes an area the size of the
+	 * configured huge page size. This is twice the
+	 * of the large TLB entry size we intend to use.
+	 * A TLB entry half the size of the configured
+	 * huge page size is configured into entrylo0
+	 * and entrylo1 to cover the contiguous huge PTE
+	 * address space.
+	 */
+	/* Huge page: Move Global bit */
+	xori	t0, t0, _PAGE_HUGE
+	lu12i.w	t1, _PAGE_HGLOBAL >> 12
+	and	t1, t0, t1
+	srli.d	t1, t1, (_PAGE_HGLOBAL_SHIFT - _PAGE_GLOBAL_SHIFT)
+	or	t0, t0, t1
+
+	addi.d	ra, t0, 0
+	csrwr	t0, LOONGARCH_CSR_TLBELO0
+	addi.d	t0, ra, 0
+
+	/* Convert to entrylo1 */
+	addi.d	t1, $r0, 1
+	slli.d	t1, t1, (HPAGE_SHIFT - 1)
+	add.d	t0, t0, t1
+	csrwr	t0, LOONGARCH_CSR_TLBELO1
+
+	/* Set huge page tlb entry size */
+	addu16i.d	t0, $r0, (CSR_TLBIDX_PS >> 16)
+	addu16i.d	t1, $r0, (PS_HUGE_SIZE << (CSR_TLBIDX_PS_SHIFT - 16))
+	csrxchg		t1, t0, LOONGARCH_CSR_TLBIDX
+
+	tlbfill
+
+	addu16i.d	t0, $r0, (CSR_TLBIDX_PS >> 16)
+	addu16i.d	t1, $r0, (PS_DEFAULT_SIZE << (CSR_TLBIDX_PS_SHIFT - 16))
+	csrxchg		t1, t0, LOONGARCH_CSR_TLBIDX
+
+nopage_tlb_load:
+	csrrd	ra, EXCEPTION_KS2
+	la.abs	t0, tlb_do_page_fault_0
+	jirl	$r0, t0, 0
+SYM_FUNC_END(handle_tlb_load)
+
+SYM_FUNC_START(handle_tlb_store)
+	csrwr	t0, EXCEPTION_KS0
+	csrwr	t1, EXCEPTION_KS1
+	csrwr	ra, EXCEPTION_KS2
+
+	/*
+	 * The vmalloc handling is not in the hotpath.
+	 */
+	csrrd	t0, LOONGARCH_CSR_BADV
+	blt	t0, $r0, vmalloc_store
+	csrrd	t1, LOONGARCH_CSR_PGDL
+
+vmalloc_done_store:
+	/* Get PGD offset in bytes */
+	srli.d	t0, t0, PGDIR_SHIFT
+	andi	t0, t0, (PTRS_PER_PGD - 1)
+	slli.d	t0, t0, 3
+	add.d	t1, t1, t0
+
+#if CONFIG_PGTABLE_LEVELS > 3
+	csrrd	t0, LOONGARCH_CSR_BADV
+	ld.d	t1, t1, 0
+	srli.d	t0, t0, PUD_SHIFT
+	andi	t0, t0, (PTRS_PER_PUD - 1)
+	slli.d	t0, t0, 3
+	add.d	t1, t1, t0
+#endif
+#if CONFIG_PGTABLE_LEVELS > 2
+	csrrd	t0, LOONGARCH_CSR_BADV
+	ld.d	t1, t1, 0
+	srli.d	t0, t0, PMD_SHIFT
+	andi	t0, t0, (PTRS_PER_PMD - 1)
+	slli.d	t0, t0, 3
+	add.d	t1, t1, t0
+#endif
+	ld.d	ra, t1, 0
+
+	/*
+	 * For huge tlb entries, pmde 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.
+	 */
+	andi	t0, ra, _PAGE_HUGE
+	bne	t0, $r0, tlb_huge_update_store
+
+	csrrd	t0, LOONGARCH_CSR_BADV
+	srli.d	t0, t0, (PAGE_SHIFT + PTE_ORDER)
+	andi	t0, t0, (PTRS_PER_PTE - 1)
+	slli.d	t0, t0, _PTE_T_LOG2
+	add.d	t1, ra, t0
+
+	ld.d	t0, t1, 0
+	tlbsrch
+
+	srli.d	ra, t0, _PAGE_PRESENT_SHIFT
+	andi	ra, ra, ((_PAGE_PRESENT | _PAGE_WRITE) >> _PAGE_PRESENT_SHIFT)
+	xori	ra, ra, ((_PAGE_PRESENT | _PAGE_WRITE) >> _PAGE_PRESENT_SHIFT)
+	bne	ra, $r0, nopage_tlb_store
+
+	ori	t0, t0, (_PAGE_VALID | _PAGE_DIRTY | _PAGE_MODIFIED)
+	st.d	t0, t1, 0
+
+	ori	t1, t1, 8
+	xori	t1, t1, 8
+	ld.d	t0, t1, 0
+	ld.d	t1, t1, 8
+	csrwr	t0, LOONGARCH_CSR_TLBELO0
+	csrwr	t1, LOONGARCH_CSR_TLBELO1
+	tlbwr
+leave_store:
+	csrrd	t0, EXCEPTION_KS0
+	csrrd	t1, EXCEPTION_KS1
+	csrrd	ra, EXCEPTION_KS2
+	ertn
+#ifdef CONFIG_64BIT
+vmalloc_store:
+	la.abs	t1, swapper_pg_dir
+	b	vmalloc_done_store
+#endif
+
+	/*
+	 * This is the entry point when build_tlbchange_handler_head
+	 * spots a huge page.
+	 */
+tlb_huge_update_store:
+	ld.d	t0, t1, 0
+	srli.d	ra, t0, _PAGE_PRESENT_SHIFT
+	andi	ra, ra, ((_PAGE_PRESENT | _PAGE_WRITE) >> _PAGE_PRESENT_SHIFT)
+	xori	ra, ra, ((_PAGE_PRESENT | _PAGE_WRITE) >> _PAGE_PRESENT_SHIFT)
+	bne	ra, $r0, nopage_tlb_store
+
+	tlbsrch
+	ori	t0, t0, (_PAGE_VALID | _PAGE_DIRTY | _PAGE_MODIFIED)
+
+	st.d	t0, t1, 0
+	addu16i.d	t1, $r0, -(CSR_TLBIDX_EHINV >> 16)
+	addi.d	ra, t1, 0
+	csrxchg	ra, t1, LOONGARCH_CSR_TLBIDX
+	tlbwr
+
+	csrxchg	$r0, t1, LOONGARCH_CSR_TLBIDX
+	/*
+	 * A huge PTE describes an area the size of the
+	 * configured huge page size. This is twice the
+	 * of the large TLB entry size we intend to use.
+	 * A TLB entry half the size of the configured
+	 * huge page size is configured into entrylo0
+	 * and entrylo1 to cover the contiguous huge PTE
+	 * address space.
+	 */
+	/* Huge page: Move Global bit */
+	xori	t0, t0, _PAGE_HUGE
+	lu12i.w	t1, _PAGE_HGLOBAL >> 12
+	and	t1, t0, t1
+	srli.d	t1, t1, (_PAGE_HGLOBAL_SHIFT - _PAGE_GLOBAL_SHIFT)
+	or	t0, t0, t1
+
+	addi.d	ra, t0, 0
+	csrwr	t0, LOONGARCH_CSR_TLBELO0
+	addi.d	t0, ra, 0
+
+	/* Convert to entrylo1 */
+	addi.d	t1, $r0, 1
+	slli.d	t1, t1, (HPAGE_SHIFT - 1)
+	add.d	t0, t0, t1
+	csrwr	t0, LOONGARCH_CSR_TLBELO1
+
+	/* Set huge page tlb entry size */
+	addu16i.d	t0, $r0, (CSR_TLBIDX_PS >> 16)
+	addu16i.d	t1, $r0, (PS_HUGE_SIZE << (CSR_TLBIDX_PS_SHIFT - 16))
+	csrxchg		t1, t0, LOONGARCH_CSR_TLBIDX
+
+	tlbfill
+
+	/* Reset default page size */
+	addu16i.d	t0, $r0, (CSR_TLBIDX_PS >> 16)
+	addu16i.d	t1, $r0, (PS_DEFAULT_SIZE << (CSR_TLBIDX_PS_SHIFT - 16))
+	csrxchg		t1, t0, LOONGARCH_CSR_TLBIDX
+
+nopage_tlb_store:
+	csrrd	ra, EXCEPTION_KS2
+	la.abs	t0, tlb_do_page_fault_1
+	jirl	$r0, t0, 0
+SYM_FUNC_END(handle_tlb_store)
+
+SYM_FUNC_START(handle_tlb_modify)
+	csrwr	t0, EXCEPTION_KS0
+	csrwr	t1, EXCEPTION_KS1
+	csrwr	ra, EXCEPTION_KS2
+
+	/*
+	 * The vmalloc handling is not in the hotpath.
+	 */
+	csrrd	t0, LOONGARCH_CSR_BADV
+	blt	t0, $r0, vmalloc_modify
+	csrrd	t1, LOONGARCH_CSR_PGDL
+
+vmalloc_done_modify:
+	/* Get PGD offset in bytes */
+	srli.d	t0, t0, PGDIR_SHIFT
+	andi	t0, t0, (PTRS_PER_PGD - 1)
+	slli.d	t0, t0, 3
+	add.d	t1, t1, t0
+#if CONFIG_PGTABLE_LEVELS > 3
+	csrrd	t0, LOONGARCH_CSR_BADV
+	ld.d	t1, t1, 0
+	srli.d	t0, t0, PUD_SHIFT
+	andi	t0, t0, (PTRS_PER_PUD - 1)
+	slli.d	t0, t0, 3
+	add.d	t1, t1, t0
+#endif
+#if CONFIG_PGTABLE_LEVELS > 2
+	csrrd	t0, LOONGARCH_CSR_BADV
+	ld.d	t1, t1, 0
+	srli.d	t0, t0, PMD_SHIFT
+	andi	t0, t0, (PTRS_PER_PMD - 1)
+	slli.d	t0, t0, 3
+	add.d	t1, t1, t0
+#endif
+	ld.d	ra, t1, 0
+
+	/*
+	 * For huge tlb entries, pmde 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.
+	 */
+	andi	t0, ra, _PAGE_HUGE
+	bne	t0, $r0, tlb_huge_update_modify
+
+	csrrd	t0, LOONGARCH_CSR_BADV
+	srli.d	t0, t0, (PAGE_SHIFT + PTE_ORDER)
+	andi	t0, t0, (PTRS_PER_PTE - 1)
+	slli.d	t0, t0, _PTE_T_LOG2
+	add.d	t1, ra, t0
+
+	ld.d	t0, t1, 0
+	tlbsrch
+
+	srli.d	ra, t0, _PAGE_WRITE_SHIFT
+	andi	ra, ra, 1
+	beq	ra, $r0, nopage_tlb_modify
+
+	ori	t0, t0, (_PAGE_VALID | _PAGE_DIRTY | _PAGE_MODIFIED)
+	st.d	t0, t1, 0
+	ori	t1, t1, 8
+	xori	t1, t1, 8
+	ld.d	t0, t1, 0
+	ld.d	t1, t1, 8
+	csrwr	t0, LOONGARCH_CSR_TLBELO0
+	csrwr	t1, LOONGARCH_CSR_TLBELO1
+	tlbwr
+leave_modify:
+	csrrd	t0, EXCEPTION_KS0
+	csrrd	t1, EXCEPTION_KS1
+	csrrd	ra, EXCEPTION_KS2
+	ertn
+#ifdef CONFIG_64BIT
+vmalloc_modify:
+	la.abs  t1, swapper_pg_dir
+	b	vmalloc_done_modify
+#endif
+
+	/*
+	 * This is the entry point when
+	 * build_tlbchange_handler_head spots a huge page.
+	 */
+tlb_huge_update_modify:
+	ld.d	t0, t1, 0
+
+	srli.d	ra, t0, _PAGE_WRITE_SHIFT
+	andi	ra, ra, 1
+	beq	ra, $r0, nopage_tlb_modify
+
+	tlbsrch
+	ori	t0, t0, (_PAGE_VALID | _PAGE_DIRTY | _PAGE_MODIFIED)
+
+	st.d	t0, t1, 0
+	/*
+	 * A huge PTE describes an area the size of the
+	 * configured huge page size. This is twice the
+	 * of the large TLB entry size we intend to use.
+	 * A TLB entry half the size of the configured
+	 * huge page size is configured into entrylo0
+	 * and entrylo1 to cover the contiguous huge PTE
+	 * address space.
+	 */
+	/* Huge page: Move Global bit */
+	xori	t0, t0, _PAGE_HUGE
+	lu12i.w	t1, _PAGE_HGLOBAL >> 12
+	and	t1, t0, t1
+	srli.d	t1, t1, (_PAGE_HGLOBAL_SHIFT - _PAGE_GLOBAL_SHIFT)
+	or	t0, t0, t1
+
+	addi.d	ra, t0, 0
+	csrwr	t0, LOONGARCH_CSR_TLBELO0
+	addi.d	t0, ra, 0
+
+	/* Convert to entrylo1 */
+	addi.d	t1, $r0, 1
+	slli.d	t1, t1, (HPAGE_SHIFT - 1)
+	add.d	t0, t0, t1
+	csrwr	t0, LOONGARCH_CSR_TLBELO1
+
+	/* Set huge page tlb entry size */
+	addu16i.d	t0, $r0, (CSR_TLBIDX_PS >> 16)
+	addu16i.d	t1, $r0, (PS_HUGE_SIZE << (CSR_TLBIDX_PS_SHIFT - 16))
+	csrxchg	t1, t0, LOONGARCH_CSR_TLBIDX
+
+	tlbwr
+
+	/* Reset default page size */
+	addu16i.d	t0, $r0, (CSR_TLBIDX_PS >> 16)
+	addu16i.d	t1, $r0, (PS_DEFAULT_SIZE << (CSR_TLBIDX_PS_SHIFT - 16))
+	csrxchg	t1, t0, LOONGARCH_CSR_TLBIDX
+
+nopage_tlb_modify:
+	csrrd	ra, EXCEPTION_KS2
+	la.abs	t0, tlb_do_page_fault_1
+	jirl	$r0, t0, 0
+SYM_FUNC_END(handle_tlb_modify)
+
+SYM_FUNC_START(handle_tlb_refill)
+	csrwr	t0, LOONGARCH_CSR_TLBRSAVE
+	csrrd	t0, LOONGARCH_CSR_PGD
+	lddir	t0, t0, 3
+#if CONFIG_PGTABLE_LEVELS > 3
+	lddir	t0, t0, 2
+#endif
+#if CONFIG_PGTABLE_LEVELS > 2
+	lddir	t0, t0, 1
+#endif
+	ldpte	t0, 0
+	ldpte	t0, 1
+	tlbfill
+	csrrd	t0, LOONGARCH_CSR_TLBRSAVE
+	ertn
+SYM_FUNC_END(handle_tlb_refill)