1/*-
2 * Copyright (c) 2010 Isilon Systems, Inc.
3 * Copyright (c) 2016 Matthew Macy (mmacy@mattmacy.io)
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: stable/11/sys/compat/linuxkpi/common/src/linux_page.c 345927 2019-04-05 11:24:05Z hselasky $");
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#include <linux/fs.h>
65
66#if defined(__amd64__) || defined(__aarch64__) || defined(__riscv)
67#define	LINUXKPI_HAVE_DMAP
68#else
69#undef	LINUXKPI_HAVE_DMAP
70#endif
71
72void
73si_meminfo(struct sysinfo *si)
74{
75	si->totalram = physmem;
76	si->totalhigh = 0;
77	si->mem_unit = PAGE_SIZE;
78}
79
80void *
81linux_page_address(struct page *page)
82{
83
84	if (page->object != kmem_object && page->object != kernel_object) {
85#ifdef LINUXKPI_HAVE_DMAP
86		return ((void *)PHYS_TO_DMAP(VM_PAGE_TO_PHYS(page)));
87#else
88		return (NULL);
89#endif
90	}
91	return ((void *)(uintptr_t)(VM_MIN_KERNEL_ADDRESS +
92	    IDX_TO_OFF(page->pindex)));
93}
94
95vm_page_t
96linux_alloc_pages(gfp_t flags, unsigned int order)
97{
98#ifdef LINUXKPI_HAVE_DMAP
99	unsigned long npages = 1UL << order;
100	int req = (flags & M_ZERO) ? (VM_ALLOC_ZERO | VM_ALLOC_NOOBJ |
101	    VM_ALLOC_NORMAL) : (VM_ALLOC_NOOBJ | VM_ALLOC_NORMAL);
102	vm_page_t page;
103
104	if (order == 0 && (flags & GFP_DMA32) == 0) {
105		page = vm_page_alloc(NULL, 0, req);
106		if (page == NULL)
107			return (NULL);
108	} else {
109		vm_paddr_t pmax = (flags & GFP_DMA32) ?
110		    BUS_SPACE_MAXADDR_32BIT : BUS_SPACE_MAXADDR;
111retry:
112		page = vm_page_alloc_contig(NULL, 0, req,
113		    npages, 0, pmax, PAGE_SIZE, 0, VM_MEMATTR_DEFAULT);
114
115		if (page == NULL) {
116			if (flags & M_WAITOK) {
117				if (!vm_page_reclaim_contig(req,
118				    npages, 0, pmax, PAGE_SIZE, 0)) {
119					VM_WAIT;
120				}
121				flags &= ~M_WAITOK;
122				goto retry;
123			}
124			return (NULL);
125		}
126	}
127	if (flags & M_ZERO) {
128		unsigned long x;
129
130		for (x = 0; x != npages; x++) {
131			vm_page_t pgo = page + x;
132
133			if ((pgo->flags & PG_ZERO) == 0)
134				pmap_zero_page(pgo);
135		}
136	}
137#else
138	vm_offset_t vaddr;
139	vm_page_t page;
140
141	vaddr = linux_alloc_kmem(flags, order);
142	if (vaddr == 0)
143		return (NULL);
144
145	page = PHYS_TO_VM_PAGE(vtophys((void *)vaddr));
146
147	KASSERT(vaddr == (vm_offset_t)page_address(page),
148	    ("Page address mismatch"));
149#endif
150	return (page);
151}
152
153void
154linux_free_pages(vm_page_t page, unsigned int order)
155{
156#ifdef LINUXKPI_HAVE_DMAP
157	unsigned long npages = 1UL << order;
158	unsigned long x;
159
160	for (x = 0; x != npages; x++) {
161		vm_page_t pgo = page + x;
162
163		vm_page_lock(pgo);
164		vm_page_free(pgo);
165		vm_page_unlock(pgo);
166	}
167#else
168	vm_offset_t vaddr;
169
170	vaddr = (vm_offset_t)page_address(page);
171
172	linux_free_kmem(vaddr, order);
173#endif
174}
175
176vm_offset_t
177linux_alloc_kmem(gfp_t flags, unsigned int order)
178{
179	size_t size = ((size_t)PAGE_SIZE) << order;
180	vm_offset_t addr;
181
182	if ((flags & GFP_DMA32) == 0) {
183		addr = kmem_malloc(kmem_arena, size, flags & GFP_NATIVE_MASK);
184	} else {
185		addr = kmem_alloc_contig(kmem_arena, size,
186		    flags & GFP_NATIVE_MASK, 0, BUS_SPACE_MAXADDR_32BIT,
187		    PAGE_SIZE, 0, VM_MEMATTR_DEFAULT);
188	}
189	return (addr);
190}
191
192void
193linux_free_kmem(vm_offset_t addr, unsigned int order)
194{
195	size_t size = ((size_t)PAGE_SIZE) << order;
196
197	kmem_free(kmem_arena, addr, size);
198}
199
200static int
201linux_get_user_pages_internal(vm_map_t map, unsigned long start, int nr_pages,
202    int write, struct page **pages)
203{
204	vm_prot_t prot;
205	size_t len;
206	int count;
207	int i;
208
209	prot = write ? (VM_PROT_READ | VM_PROT_WRITE) : VM_PROT_READ;
210	len = ((size_t)nr_pages) << PAGE_SHIFT;
211	count = vm_fault_quick_hold_pages(map, start, len, prot, pages, nr_pages);
212	if (count == -1)
213		return (-EFAULT);
214
215	for (i = 0; i != nr_pages; i++) {
216		struct page *pg = pages[i];
217
218		vm_page_lock(pg);
219		vm_page_wire(pg);
220		vm_page_unhold(pg);
221		vm_page_unlock(pg);
222	}
223	return (nr_pages);
224}
225
226int
227__get_user_pages_fast(unsigned long start, int nr_pages, int write,
228    struct page **pages)
229{
230	vm_map_t map;
231	vm_page_t *mp;
232	vm_offset_t va;
233	vm_offset_t end;
234	vm_prot_t prot;
235	int count;
236
237	if (nr_pages == 0 || in_interrupt())
238		return (0);
239
240	MPASS(pages != NULL);
241	va = start;
242	map = &curthread->td_proc->p_vmspace->vm_map;
243	end = start + (((size_t)nr_pages) << PAGE_SHIFT);
244	if (start < vm_map_min(map) || end > vm_map_max(map))
245		return (-EINVAL);
246	prot = write ? (VM_PROT_READ | VM_PROT_WRITE) : VM_PROT_READ;
247	for (count = 0, mp = pages, va = start; va < end;
248	    mp++, va += PAGE_SIZE, count++) {
249		*mp = pmap_extract_and_hold(map->pmap, va, prot);
250		if (*mp == NULL)
251			break;
252
253		vm_page_lock(*mp);
254		vm_page_wire(*mp);
255		vm_page_unhold(*mp);
256		vm_page_unlock(*mp);
257
258		if ((prot & VM_PROT_WRITE) != 0 &&
259		    (*mp)->dirty != VM_PAGE_BITS_ALL) {
260			/*
261			 * Explicitly dirty the physical page.  Otherwise, the
262			 * caller's changes may go unnoticed because they are
263			 * performed through an unmanaged mapping or by a DMA
264			 * operation.
265			 *
266			 * The object lock is not held here.
267			 * See vm_page_clear_dirty_mask().
268			 */
269			vm_page_dirty(*mp);
270		}
271	}
272	return (count);
273}
274
275long
276get_user_pages_remote(struct task_struct *task, struct mm_struct *mm,
277    unsigned long start, unsigned long nr_pages, int gup_flags,
278    struct page **pages, struct vm_area_struct **vmas)
279{
280	vm_map_t map;
281
282	map = &task->task_thread->td_proc->p_vmspace->vm_map;
283	return (linux_get_user_pages_internal(map, start, nr_pages,
284	    !!(gup_flags & FOLL_WRITE), pages));
285}
286
287long
288get_user_pages(unsigned long start, unsigned long nr_pages, int gup_flags,
289    struct page **pages, struct vm_area_struct **vmas)
290{
291	vm_map_t map;
292
293	map = &curthread->td_proc->p_vmspace->vm_map;
294	return (linux_get_user_pages_internal(map, start, nr_pages,
295	    !!(gup_flags & FOLL_WRITE), pages));
296}
297
298int
299is_vmalloc_addr(const void *addr)
300{
301	return (vtoslab((vm_offset_t)addr & ~UMA_SLAB_MASK) != NULL);
302}
303
304struct page *
305linux_shmem_read_mapping_page_gfp(vm_object_t obj, int pindex, gfp_t gfp)
306{
307	vm_page_t page;
308	int rv;
309
310	if ((gfp & GFP_NOWAIT) != 0)
311		panic("GFP_NOWAIT is unimplemented");
312
313	VM_OBJECT_WLOCK(obj);
314	page = vm_page_grab(obj, pindex, VM_ALLOC_NORMAL | VM_ALLOC_NOBUSY |
315	    VM_ALLOC_WIRED);
316	if (page->valid != VM_PAGE_BITS_ALL) {
317		vm_page_xbusy(page);
318		if (vm_pager_has_page(obj, pindex, NULL, NULL)) {
319			rv = vm_pager_get_pages(obj, &page, 1, NULL, NULL);
320			if (rv != VM_PAGER_OK) {
321				vm_page_lock(page);
322				vm_page_unwire(page, PQ_NONE);
323				vm_page_free(page);
324				vm_page_unlock(page);
325				VM_OBJECT_WUNLOCK(obj);
326				return (ERR_PTR(-EINVAL));
327			}
328			MPASS(page->valid == VM_PAGE_BITS_ALL);
329		} else {
330			pmap_zero_page(page);
331			page->valid = VM_PAGE_BITS_ALL;
332			page->dirty = 0;
333		}
334		vm_page_xunbusy(page);
335	}
336	VM_OBJECT_WUNLOCK(obj);
337	return (page);
338}
339
340struct linux_file *
341linux_shmem_file_setup(const char *name, loff_t size, unsigned long flags)
342{
343	struct fileobj {
344		struct linux_file file __aligned(sizeof(void *));
345		struct vnode vnode __aligned(sizeof(void *));
346	};
347	struct fileobj *fileobj;
348	struct linux_file *filp;
349	struct vnode *vp;
350	int error;
351
352	fileobj = kzalloc(sizeof(*fileobj), GFP_KERNEL);
353	if (fileobj == NULL) {
354		error = -ENOMEM;
355		goto err_0;
356	}
357	filp = &fileobj->file;
358	vp = &fileobj->vnode;
359
360	filp->f_count = 1;
361	filp->f_vnode = vp;
362	filp->f_shmem = vm_pager_allocate(OBJT_DEFAULT, NULL, size,
363	    VM_PROT_READ | VM_PROT_WRITE, 0, curthread->td_ucred);
364	if (filp->f_shmem == NULL) {
365		error = -ENOMEM;
366		goto err_1;
367	}
368	return (filp);
369err_1:
370	kfree(filp);
371err_0:
372	return (ERR_PTR(error));
373}
374
375static vm_ooffset_t
376linux_invalidate_mapping_pages_sub(vm_object_t obj, vm_pindex_t start,
377    vm_pindex_t end, int flags)
378{
379	int start_count, end_count;
380
381	VM_OBJECT_WLOCK(obj);
382	start_count = obj->resident_page_count;
383	vm_object_page_remove(obj, start, end, flags);
384	end_count = obj->resident_page_count;
385	VM_OBJECT_WUNLOCK(obj);
386	return (start_count - end_count);
387}
388
389unsigned long
390linux_invalidate_mapping_pages(vm_object_t obj, pgoff_t start, pgoff_t end)
391{
392
393	return (linux_invalidate_mapping_pages_sub(obj, start, end, OBJPR_CLEANONLY));
394}
395
396void
397linux_shmem_truncate_range(vm_object_t obj, loff_t lstart, loff_t lend)
398{
399	vm_pindex_t start = OFF_TO_IDX(lstart + PAGE_SIZE - 1);
400	vm_pindex_t end = OFF_TO_IDX(lend + 1);
401
402	(void) linux_invalidate_mapping_pages_sub(obj, start, end, 0);
403}
404