1 /*- 2 * Copyright (c) 2010 Isilon Systems, Inc. 3 * Copyright (c) 2016 Matt Macy (mmacy@nextbsd.org) 4 * Copyright (c) 2017 Mellanox Technologies, Ltd. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice unmodified, this list of conditions, and the following 12 * disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 #include <sys/param.h> 33 #include <sys/systm.h> 34 #include <sys/malloc.h> 35 #include <sys/kernel.h> 36 #include <sys/sysctl.h> 37 #include <sys/lock.h> 38 #include <sys/mutex.h> 39 #include <sys/rwlock.h> 40 #include <sys/proc.h> 41 #include <sys/sched.h> 42 43 #include <machine/bus.h> 44 45 #include <vm/vm.h> 46 #include <vm/pmap.h> 47 #include <vm/vm_param.h> 48 #include <vm/vm_kern.h> 49 #include <vm/vm_object.h> 50 #include <vm/vm_map.h> 51 #include <vm/vm_page.h> 52 #include <vm/vm_pageout.h> 53 #include <vm/vm_pager.h> 54 #include <vm/vm_radix.h> 55 #include <vm/vm_reserv.h> 56 #include <vm/vm_extern.h> 57 58 #include <vm/uma.h> 59 #include <vm/uma_int.h> 60 61 #include <linux/gfp.h> 62 #include <linux/mm.h> 63 #include <linux/preempt.h> 64 65 #if defined(__amd64__) || defined(__aarch64__) || defined(__riscv__) 66 #define LINUXKPI_HAVE_DMAP 67 #else 68 #undef LINUXKPI_HAVE_DMAP 69 #endif 70 71 void * 72 linux_page_address(struct page *page) 73 { 74 #ifdef LINUXKPI_HAVE_DMAP 75 return ((void *)PHYS_TO_DMAP(VM_PAGE_TO_PHYS(page))); 76 #else 77 if (page->object != kmem_object && page->object != kernel_object) 78 return (NULL); 79 return ((void *)(uintptr_t)(VM_MIN_KERNEL_ADDRESS + 80 IDX_TO_OFF(page->pindex))); 81 #endif 82 } 83 84 vm_page_t 85 linux_alloc_pages(gfp_t flags, unsigned int order) 86 { 87 #ifdef LINUXKPI_HAVE_DMAP 88 unsigned long npages = 1UL << order; 89 int req = (flags & M_ZERO) ? (VM_ALLOC_ZERO | VM_ALLOC_NOOBJ | 90 VM_ALLOC_NORMAL) : (VM_ALLOC_NOOBJ | VM_ALLOC_NORMAL); 91 vm_page_t page; 92 93 if (order == 0 && (flags & GFP_DMA32) == 0) { 94 page = vm_page_alloc(NULL, 0, req); 95 if (page == NULL) 96 return (NULL); 97 } else { 98 vm_paddr_t pmax = (flags & GFP_DMA32) ? 99 BUS_SPACE_MAXADDR_32BIT : BUS_SPACE_MAXADDR; 100 retry: 101 page = vm_page_alloc_contig(NULL, 0, req, 102 npages, 0, pmax, PAGE_SIZE, 0, VM_MEMATTR_DEFAULT); 103 104 if (page == NULL) { 105 if (flags & M_WAITOK) { 106 if (!vm_page_reclaim_contig(req, 107 npages, 0, pmax, PAGE_SIZE, 0)) { 108 VM_WAIT; 109 } 110 flags &= ~M_WAITOK; 111 goto retry; 112 } 113 return (NULL); 114 } 115 } 116 if (flags & M_ZERO) { 117 unsigned long x; 118 119 for (x = 0; x != npages; x++) { 120 vm_page_t pgo = page + x; 121 122 if ((pgo->flags & PG_ZERO) == 0) 123 pmap_zero_page(pgo); 124 } 125 } 126 #else 127 vm_offset_t vaddr; 128 vm_page_t page; 129 130 vaddr = linux_alloc_kmem(flags, order); 131 if (vaddr == 0) 132 return (NULL); 133 134 page = PHYS_TO_VM_PAGE(vtophys((void *)vaddr)); 135 136 KASSERT(vaddr == (vm_offset_t)page_address(page), 137 ("Page address mismatch")); 138 #endif 139 return (page); 140 } 141 142 void 143 linux_free_pages(vm_page_t page, unsigned int order) 144 { 145 #ifdef LINUXKPI_HAVE_DMAP 146 unsigned long npages = 1UL << order; 147 unsigned long x; 148 149 for (x = 0; x != npages; x++) { 150 vm_page_t pgo = page + x; 151 152 vm_page_lock(pgo); 153 vm_page_free(pgo); 154 vm_page_unlock(pgo); 155 } 156 #else 157 vm_offset_t vaddr; 158 159 vaddr = (vm_offset_t)page_address(page); 160 161 linux_free_kmem(vaddr, order); 162 #endif 163 } 164 165 vm_offset_t 166 linux_alloc_kmem(gfp_t flags, unsigned int order) 167 { 168 size_t size = ((size_t)PAGE_SIZE) << order; 169 vm_offset_t addr; 170 171 if ((flags & GFP_DMA32) == 0) { 172 addr = kmem_malloc(kmem_arena, size, flags & GFP_NATIVE_MASK); 173 } else { 174 addr = kmem_alloc_contig(kmem_arena, size, 175 flags & GFP_NATIVE_MASK, 0, BUS_SPACE_MAXADDR_32BIT, 176 PAGE_SIZE, 0, VM_MEMATTR_DEFAULT); 177 } 178 return (addr); 179 } 180 181 void 182 linux_free_kmem(vm_offset_t addr, unsigned int order) 183 { 184 size_t size = ((size_t)PAGE_SIZE) << order; 185 186 kmem_free(kmem_arena, addr, size); 187 } 188 189 static int 190 linux_get_user_pages_internal(vm_map_t map, unsigned long start, int nr_pages, 191 int write, struct page **pages) 192 { 193 vm_prot_t prot; 194 size_t len; 195 int count; 196 int i; 197 198 prot = write ? (VM_PROT_READ | VM_PROT_WRITE) : VM_PROT_READ; 199 len = ((size_t)nr_pages) << PAGE_SHIFT; 200 count = vm_fault_quick_hold_pages(map, start, len, prot, pages, nr_pages); 201 if (count == -1) 202 return (-EFAULT); 203 204 for (i = 0; i != nr_pages; i++) { 205 struct page *pg = pages[i]; 206 207 vm_page_lock(pg); 208 vm_page_wire(pg); 209 vm_page_unlock(pg); 210 } 211 return (nr_pages); 212 } 213 214 int 215 __get_user_pages_fast(unsigned long start, int nr_pages, int write, 216 struct page **pages) 217 { 218 vm_map_t map; 219 vm_page_t *mp; 220 vm_offset_t va; 221 vm_offset_t end; 222 vm_prot_t prot; 223 int count; 224 225 if (nr_pages == 0 || in_interrupt()) 226 return (0); 227 228 MPASS(pages != NULL); 229 va = start; 230 map = &curthread->td_proc->p_vmspace->vm_map; 231 end = start + (((size_t)nr_pages) << PAGE_SHIFT); 232 if (start < vm_map_min(map) || end > vm_map_max(map)) 233 return (-EINVAL); 234 prot = write ? (VM_PROT_READ | VM_PROT_WRITE) : VM_PROT_READ; 235 for (count = 0, mp = pages, va = start; va < end; 236 mp++, va += PAGE_SIZE, count++) { 237 *mp = pmap_extract_and_hold(map->pmap, va, prot); 238 if (*mp == NULL) 239 break; 240 241 vm_page_lock(*mp); 242 vm_page_wire(*mp); 243 vm_page_unlock(*mp); 244 245 if ((prot & VM_PROT_WRITE) != 0 && 246 (*mp)->dirty != VM_PAGE_BITS_ALL) { 247 /* 248 * Explicitly dirty the physical page. Otherwise, the 249 * caller's changes may go unnoticed because they are 250 * performed through an unmanaged mapping or by a DMA 251 * operation. 252 * 253 * The object lock is not held here. 254 * See vm_page_clear_dirty_mask(). 255 */ 256 vm_page_dirty(*mp); 257 } 258 } 259 return (count); 260 } 261 262 long 263 get_user_pages_remote(struct task_struct *task, struct mm_struct *mm, 264 unsigned long start, unsigned long nr_pages, int gup_flags, 265 struct page **pages, struct vm_area_struct **vmas) 266 { 267 vm_map_t map; 268 269 map = &mm->vmspace->vm_map; 270 return (linux_get_user_pages_internal(map, start, nr_pages, 271 !!(gup_flags & FOLL_WRITE), pages)); 272 } 273 274 long 275 get_user_pages(unsigned long start, unsigned long nr_pages, int gup_flags, 276 struct page **pages, struct vm_area_struct **vmas) 277 { 278 vm_map_t map; 279 280 map = &curthread->td_proc->p_vmspace->vm_map; 281 return (linux_get_user_pages_internal(map, start, nr_pages, 282 !!(gup_flags & FOLL_WRITE), pages)); 283 } 284 285 int 286 is_vmalloc_addr(const void *addr) 287 { 288 return (vtoslab((vm_offset_t)addr & ~UMA_SLAB_MASK) != NULL); 289 } 290