Bryan O'Sullivan | 7f510b4 | 2006-03-29 15:23:31 -0800 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved. |
| 3 | * |
| 4 | * This software is available to you under a choice of one of two |
| 5 | * licenses. You may choose to be licensed under the terms of the GNU |
| 6 | * General Public License (GPL) Version 2, available from the file |
| 7 | * COPYING in the main directory of this source tree, or the |
| 8 | * OpenIB.org BSD license below: |
| 9 | * |
| 10 | * Redistribution and use in source and binary forms, with or |
| 11 | * without modification, are permitted provided that the following |
| 12 | * conditions are met: |
| 13 | * |
| 14 | * - Redistributions of source code must retain the above |
| 15 | * copyright notice, this list of conditions and the following |
| 16 | * disclaimer. |
| 17 | * |
| 18 | * - Redistributions in binary form must reproduce the above |
| 19 | * copyright notice, this list of conditions and the following |
| 20 | * disclaimer in the documentation and/or other materials |
| 21 | * provided with the distribution. |
| 22 | * |
| 23 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, |
| 24 | * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF |
| 25 | * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND |
| 26 | * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS |
| 27 | * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN |
| 28 | * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN |
| 29 | * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
| 30 | * SOFTWARE. |
| 31 | */ |
| 32 | |
| 33 | #include <linux/mm.h> |
| 34 | #include <linux/device.h> |
| 35 | |
| 36 | #include "ipath_kernel.h" |
| 37 | |
| 38 | static void __ipath_release_user_pages(struct page **p, size_t num_pages, |
| 39 | int dirty) |
| 40 | { |
| 41 | size_t i; |
| 42 | |
| 43 | for (i = 0; i < num_pages; i++) { |
| 44 | ipath_cdbg(MM, "%lu/%lu put_page %p\n", (unsigned long) i, |
| 45 | (unsigned long) num_pages, p[i]); |
| 46 | if (dirty) |
| 47 | set_page_dirty_lock(p[i]); |
| 48 | put_page(p[i]); |
| 49 | } |
| 50 | } |
| 51 | |
| 52 | /* call with current->mm->mmap_sem held */ |
| 53 | static int __get_user_pages(unsigned long start_page, size_t num_pages, |
| 54 | struct page **p, struct vm_area_struct **vma) |
| 55 | { |
| 56 | unsigned long lock_limit; |
| 57 | size_t got; |
| 58 | int ret; |
| 59 | |
| 60 | #if 0 |
| 61 | /* |
| 62 | * XXX - causes MPI programs to fail, haven't had time to check |
| 63 | * yet |
| 64 | */ |
| 65 | if (!capable(CAP_IPC_LOCK)) { |
| 66 | ret = -EPERM; |
| 67 | goto bail; |
| 68 | } |
| 69 | #endif |
| 70 | |
| 71 | lock_limit = current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur >> |
| 72 | PAGE_SHIFT; |
| 73 | |
| 74 | if (num_pages > lock_limit) { |
| 75 | ret = -ENOMEM; |
| 76 | goto bail; |
| 77 | } |
| 78 | |
| 79 | ipath_cdbg(VERBOSE, "pin %lx pages from vaddr %lx\n", |
| 80 | (unsigned long) num_pages, start_page); |
| 81 | |
| 82 | for (got = 0; got < num_pages; got += ret) { |
| 83 | ret = get_user_pages(current, current->mm, |
| 84 | start_page + got * PAGE_SIZE, |
| 85 | num_pages - got, 1, 1, |
| 86 | p + got, vma); |
| 87 | if (ret < 0) |
| 88 | goto bail_release; |
| 89 | } |
| 90 | |
| 91 | current->mm->locked_vm += num_pages; |
| 92 | |
| 93 | ret = 0; |
| 94 | goto bail; |
| 95 | |
| 96 | bail_release: |
| 97 | __ipath_release_user_pages(p, got, 0); |
| 98 | bail: |
| 99 | return ret; |
| 100 | } |
| 101 | |
| 102 | /** |
| 103 | * ipath_get_user_pages - lock user pages into memory |
| 104 | * @start_page: the start page |
| 105 | * @num_pages: the number of pages |
| 106 | * @p: the output page structures |
| 107 | * |
| 108 | * This function takes a given start page (page aligned user virtual |
| 109 | * address) and pins it and the following specified number of pages. For |
| 110 | * now, num_pages is always 1, but that will probably change at some point |
| 111 | * (because caller is doing expected sends on a single virtually contiguous |
| 112 | * buffer, so we can do all pages at once). |
| 113 | */ |
| 114 | int ipath_get_user_pages(unsigned long start_page, size_t num_pages, |
| 115 | struct page **p) |
| 116 | { |
| 117 | int ret; |
| 118 | |
| 119 | down_write(¤t->mm->mmap_sem); |
| 120 | |
| 121 | ret = __get_user_pages(start_page, num_pages, p, NULL); |
| 122 | |
| 123 | up_write(¤t->mm->mmap_sem); |
| 124 | |
| 125 | return ret; |
| 126 | } |
| 127 | |
| 128 | /** |
| 129 | * ipath_get_user_pages_nocopy - lock a single page for I/O and mark shared |
| 130 | * @start_page: the page to lock |
| 131 | * @p: the output page structure |
| 132 | * |
| 133 | * This is similar to ipath_get_user_pages, but it's always one page, and we |
| 134 | * mark the page as locked for I/O, and shared. This is used for the user |
| 135 | * process page that contains the destination address for the rcvhdrq tail |
| 136 | * update, so we need to have the vma. If we don't do this, the page can be |
| 137 | * taken away from us on fork, even if the child never touches it, and then |
| 138 | * the user process never sees the tail register updates. |
| 139 | */ |
| 140 | int ipath_get_user_pages_nocopy(unsigned long page, struct page **p) |
| 141 | { |
| 142 | struct vm_area_struct *vma; |
| 143 | int ret; |
| 144 | |
| 145 | down_write(¤t->mm->mmap_sem); |
| 146 | |
| 147 | ret = __get_user_pages(page, 1, p, &vma); |
| 148 | |
| 149 | up_write(¤t->mm->mmap_sem); |
| 150 | |
| 151 | return ret; |
| 152 | } |
| 153 | |
| 154 | void ipath_release_user_pages(struct page **p, size_t num_pages) |
| 155 | { |
| 156 | down_write(¤t->mm->mmap_sem); |
| 157 | |
| 158 | __ipath_release_user_pages(p, num_pages, 1); |
| 159 | |
| 160 | current->mm->locked_vm -= num_pages; |
| 161 | |
| 162 | up_write(¤t->mm->mmap_sem); |
| 163 | } |
| 164 | |
| 165 | struct ipath_user_pages_work { |
| 166 | struct work_struct work; |
| 167 | struct mm_struct *mm; |
| 168 | unsigned long num_pages; |
| 169 | }; |
| 170 | |
| 171 | static void user_pages_account(void *ptr) |
| 172 | { |
| 173 | struct ipath_user_pages_work *work = ptr; |
| 174 | |
| 175 | down_write(&work->mm->mmap_sem); |
| 176 | work->mm->locked_vm -= work->num_pages; |
| 177 | up_write(&work->mm->mmap_sem); |
| 178 | mmput(work->mm); |
| 179 | kfree(work); |
| 180 | } |
| 181 | |
| 182 | void ipath_release_user_pages_on_close(struct page **p, size_t num_pages) |
| 183 | { |
| 184 | struct ipath_user_pages_work *work; |
| 185 | struct mm_struct *mm; |
| 186 | |
| 187 | __ipath_release_user_pages(p, num_pages, 1); |
| 188 | |
| 189 | mm = get_task_mm(current); |
| 190 | if (!mm) |
| 191 | goto bail; |
| 192 | |
| 193 | work = kmalloc(sizeof(*work), GFP_KERNEL); |
| 194 | if (!work) |
| 195 | goto bail_mm; |
| 196 | |
| 197 | goto bail; |
| 198 | |
| 199 | INIT_WORK(&work->work, user_pages_account, work); |
| 200 | work->mm = mm; |
| 201 | work->num_pages = num_pages; |
| 202 | |
| 203 | bail_mm: |
| 204 | mmput(mm); |
| 205 | bail: |
| 206 | return; |
| 207 | } |