Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Huacai Chen <chenhuacai@loongson.cn>
To:     Arnd Bergmann <arnd@arndb.de>, Huacai Chen <chenhuacai@kernel.org>,
        Thomas Bogendoerfer <tsbogend@alpha.franken.de>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        Andy Lutomirski <luto@kernel.org>,
        Peter Zijlstra <peterz@infradead.org>,
        Catalin Marinas <catalin.marinas@arm.com>,
        Will Deacon <will@kernel.org>,
        Dinh Nguyen <dinguyen@kernel.org>
Cc:     loongarch@lists.linux.dev, linux-arch@vger.kernel.org,
        Xuefeng Li <lixuefeng@loongson.cn>,
        Guo Ren <guoren@kernel.org>, Xuerui Wang <kernel@xen0n.name>,
        Jiaxun Yang <jiaxun.yang@flygoat.com>,
        Andrew Morton <akpm@linux-foundation.org>, linux-mm@kvack.org,
        linux-mips@vger.kernel.org, linux-kernel@vger.kernel.org,
        linux-arm-kernel@lists.infradead.org,
        Feiyang Chen <chenfeiyang@loongson.cn>,
        Min Zhou <zhoumin@loongson.cn>,
        Huacai Chen <chenhuacai@loongson.cn>
Subject: [PATCH V13 2/4] LoongArch: Add sparse memory vmemmap support
Date:   Mon, 24 Oct 2022 15:01:03 +0800
Message-Id: <20221024070105.306280-3-chenhuacai@loongson.cn>
In-Reply-To: <20221024070105.306280-1-chenhuacai@loongson.cn>
References: <20221024070105.306280-1-chenhuacai@loongson.cn>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

From: Feiyang Chen <chenfeiyang@loongson.cn>

Add sparse memory vmemmap support for LoongArch. SPARSEMEM_VMEMMAP
uses a virtually mapped memmap to optimise pfn_to_page and page_to_pfn
operations. This is the most efficient option when sufficient kernel
resources are available.

Signed-off-by: Min Zhou <zhoumin@loongson.cn>
Signed-off-by: Feiyang Chen <chenfeiyang@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
---
 arch/loongarch/Kconfig                 |  1 +
 arch/loongarch/include/asm/pgtable.h   |  7 ++-
 arch/loongarch/include/asm/sparsemem.h |  8 +++
 arch/loongarch/mm/init.c               | 72 ++++++++++++++++++++++++--
 include/linux/mm.h                     |  2 +
 mm/sparse-vmemmap.c                    | 10 ++++
 6 files changed, 96 insertions(+), 4 deletions(-)

diff --git a/arch/loongarch/Kconfig b/arch/loongarch/Kconfig
index 903096bd87f8..6f7fa0c0ca08 100644
--- a/arch/loongarch/Kconfig
+++ b/arch/loongarch/Kconfig
@@ -487,6 +487,7 @@ config ARCH_FLATMEM_ENABLE
 
 config ARCH_SPARSEMEM_ENABLE
 	def_bool y
+	select SPARSEMEM_VMEMMAP_ENABLE
 	help
 	  Say Y to support efficient handling of sparse physical memory,
 	  for architectures which are either NUMA (Non-Uniform Memory Access)
diff --git a/arch/loongarch/include/asm/pgtable.h b/arch/loongarch/include/asm/pgtable.h
index 72eb8bc352fb..2e635dca2420 100644
--- a/arch/loongarch/include/asm/pgtable.h
+++ b/arch/loongarch/include/asm/pgtable.h
@@ -11,6 +11,7 @@
 
 #include <linux/compiler.h>
 #include <asm/addrspace.h>
+#include <asm/page.h>
 #include <asm/pgtable-bits.h>
 
 #if CONFIG_PGTABLE_LEVELS == 2
@@ -59,6 +60,7 @@
 #include <linux/mm_types.h>
 #include <linux/mmzone.h>
 #include <asm/fixmap.h>
+#include <asm/sparsemem.h>
 
 struct mm_struct;
 struct vm_area_struct;
@@ -86,7 +88,10 @@ extern unsigned long zero_page_mask;
 #define VMALLOC_START	MODULES_END
 #define VMALLOC_END	\
 	(vm_map_base +	\
-	 min(PTRS_PER_PGD * PTRS_PER_PUD * PTRS_PER_PMD * PTRS_PER_PTE * PAGE_SIZE, (1UL << cpu_vabits)) - PMD_SIZE)
+	 min(PTRS_PER_PGD * PTRS_PER_PUD * PTRS_PER_PMD * PTRS_PER_PTE * PAGE_SIZE, (1UL << cpu_vabits)) - PMD_SIZE - VMEMMAP_SIZE)
+
+#define vmemmap		((struct page *)((VMALLOC_END + PMD_SIZE) & PMD_MASK))
+#define VMEMMAP_END	((unsigned long)vmemmap + VMEMMAP_SIZE - 1)
 
 #define pte_ERROR(e) \
 	pr_err("%s:%d: bad pte %016lx.\n", __FILE__, __LINE__, pte_val(e))
diff --git a/arch/loongarch/include/asm/sparsemem.h b/arch/loongarch/include/asm/sparsemem.h
index 3d18cdf1b069..8d4af6aff8a8 100644
--- a/arch/loongarch/include/asm/sparsemem.h
+++ b/arch/loongarch/include/asm/sparsemem.h
@@ -11,8 +11,16 @@
 #define SECTION_SIZE_BITS	29 /* 2^29 = Largest Huge Page Size */
 #define MAX_PHYSMEM_BITS	48
 
+#ifdef CONFIG_SPARSEMEM_VMEMMAP
+#define VMEMMAP_SIZE	(sizeof(struct page) * (1UL << (cpu_pabits + 1 - PAGE_SHIFT)))
+#endif
+
 #endif /* CONFIG_SPARSEMEM */
 
+#ifndef VMEMMAP_SIZE
+#define VMEMMAP_SIZE	0	/* 1, For FLATMEM; 2, For SPARSEMEM without VMEMMAP. */
+#endif
+
 #ifdef CONFIG_MEMORY_HOTPLUG
 int memory_add_physaddr_to_nid(u64 addr);
 #define memory_add_physaddr_to_nid memory_add_physaddr_to_nid
diff --git a/arch/loongarch/mm/init.c b/arch/loongarch/mm/init.c
index 080061793c85..451d93667bcc 100644
--- a/arch/loongarch/mm/init.c
+++ b/arch/loongarch/mm/init.c
@@ -22,7 +22,7 @@
 #include <linux/pfn.h>
 #include <linux/hardirq.h>
 #include <linux/gfp.h>
-#include <linux/initrd.h>
+#include <linux/hugetlb.h>
 #include <linux/mmzone.h>
 
 #include <asm/asm-offsets.h>
@@ -152,6 +152,72 @@ EXPORT_SYMBOL_GPL(memory_add_physaddr_to_nid);
 #endif
 #endif
 
+#ifdef CONFIG_SPARSEMEM_VMEMMAP
+static int __meminit vmemmap_populate_hugepages(unsigned long start, unsigned long end,
+						int node, struct vmem_altmap *altmap)
+{
+	unsigned long addr = start;
+	unsigned long next;
+	pgd_t *pgd;
+	p4d_t *p4d;
+	pud_t *pud;
+	pmd_t *pmd;
+
+	for (addr = start; addr < end; addr = next) {
+		next = pmd_addr_end(addr, end);
+
+		pgd = vmemmap_pgd_populate(addr, node);
+		if (!pgd)
+			return -ENOMEM;
+		p4d = vmemmap_p4d_populate(pgd, addr, node);
+		if (!p4d)
+			return -ENOMEM;
+		pud = vmemmap_pud_populate(p4d, addr, node);
+		if (!pud)
+			return -ENOMEM;
+
+		pmd = pmd_offset(pud, addr);
+		if (pmd_none(*pmd)) {
+			void *p = NULL;
+
+			p = vmemmap_alloc_block_buf(PMD_SIZE, node, NULL);
+			if (p) {
+				pmd_t entry;
+
+				entry = pfn_pmd(virt_to_pfn(p), PAGE_KERNEL);
+				pmd_val(entry) |= _PAGE_HUGE | _PAGE_HGLOBAL;
+				set_pmd_at(&init_mm, addr, pmd, entry);
+
+				continue;
+			}
+		} else if (pmd_val(*pmd) & _PAGE_HUGE) {
+			vmemmap_verify((pte_t *)pmd, node, addr, next);
+			continue;
+		}
+		if (vmemmap_populate_basepages(addr, next, node, NULL))
+			return -ENOMEM;
+	}
+
+	return 0;
+}
+
+int __meminit vmemmap_populate(unsigned long start, unsigned long end,
+			       int node, struct vmem_altmap *altmap)
+{
+#if CONFIG_PGTABLE_LEVELS == 2
+	return vmemmap_populate_basepages(start, end, node, NULL);
+#else
+	return vmemmap_populate_hugepages(start, end, node, NULL);
+#endif
+}
+
+#ifdef CONFIG_MEMORY_HOTPLUG
+void vmemmap_free(unsigned long start, unsigned long end, struct vmem_altmap *altmap)
+{
+}
+#endif
+#endif
+
 static pte_t *fixmap_pte(unsigned long addr)
 {
 	pgd_t *pgd;
@@ -168,7 +234,7 @@ static pte_t *fixmap_pte(unsigned long addr)
 		new = memblock_alloc_low(PAGE_SIZE, PAGE_SIZE);
 		pgd_populate(&init_mm, pgd, new);
 #ifndef __PAGETABLE_PUD_FOLDED
-		pud_init((unsigned long)new, (unsigned long)invalid_pmd_table);
+		pud_init(new);
 #endif
 	}
 
@@ -179,7 +245,7 @@ static pte_t *fixmap_pte(unsigned long addr)
 		new = memblock_alloc_low(PAGE_SIZE, PAGE_SIZE);
 		pud_populate(&init_mm, pud, new);
 #ifndef __PAGETABLE_PMD_FOLDED
-		pmd_init((unsigned long)new, (unsigned long)invalid_pte_table);
+		pmd_init(new);
 #endif
 	}
 
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 8bbcccbc5565..45eb04c43c6b 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -3233,6 +3233,8 @@ void *sparse_buffer_alloc(unsigned long size);
 struct page * __populate_section_memmap(unsigned long pfn,
 		unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
 		struct dev_pagemap *pgmap);
+void pmd_init(void *addr);
+void pud_init(void *addr);
 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
diff --git a/mm/sparse-vmemmap.c b/mm/sparse-vmemmap.c
index 46ae542118c0..797b30e9050c 100644
--- a/mm/sparse-vmemmap.c
+++ b/mm/sparse-vmemmap.c
@@ -196,6 +196,10 @@ pmd_t * __meminit vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node)
 	return pmd;
 }
 
+void __weak __meminit pmd_init(void *addr)
+{
+}
+
 pud_t * __meminit vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node)
 {
 	pud_t *pud = pud_offset(p4d, addr);
@@ -203,11 +207,16 @@ pud_t * __meminit vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node)
 		void *p = vmemmap_alloc_block_zero(PAGE_SIZE, node);
 		if (!p)
 			return NULL;
+		pmd_init(p);
 		pud_populate(&init_mm, pud, p);
 	}
 	return pud;
 }
 
+void __weak __meminit pud_init(void *addr)
+{
+}
+
 p4d_t * __meminit vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node)
 {
 	p4d_t *p4d = p4d_offset(pgd, addr);
@@ -215,6 +224,7 @@ p4d_t * __meminit vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node)
 		void *p = vmemmap_alloc_block_zero(PAGE_SIZE, node);
 		if (!p)
 			return NULL;
+		pud_init(p);
 		p4d_populate(&init_mm, p4d, p);
 	}
 	return p4d;
-- 
2.31.1