drivers/infiniband/hw/ipath/ipath_user_pages.c

   1 /*
   2  * Copyright (c) 2006 QLogic, Inc. All rights reserved.
   3  * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
   4  *
   5  * This software is available to you under a choice of one of two
   6  * licenses.  You may choose to be licensed under the terms of the GNU
   7  * General Public License (GPL) Version 2, available from the file
   8  * COPYING in the main directory of this source tree, or the
   9  * OpenIB.org BSD license below:
  10  *
  11  *     Redistribution and use in source and binary forms, with or
  12  *     without modification, are permitted provided that the following
  13  *     conditions are met:
  14  *
  15  *      - Redistributions of source code must retain the above
  16  *        copyright notice, this list of conditions and the following
  17  *        disclaimer.
  18  *
  19  *      - Redistributions in binary form must reproduce the above
  20  *        copyright notice, this list of conditions and the following
  21  *        disclaimer in the documentation and/or other materials
  22  *        provided with the distribution.
  23  *
  24  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  25  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  26  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  27  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
  28  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
  29  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  30  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  31  * SOFTWARE.
  32  */
  33
  34 #include <linux/mm.h>
  35 #include <linux/device.h>
  36
  37 #include "ipath_kernel.h"
  38
  39 static void __ipath_release_user_pages(struct page **p, size_t num_pages,
  40                                    int dirty)
  41 {
  42         size_t i;
  43
  44         for (i = 0; i < num_pages; i++) {
  45                 ipath_cdbg(MM, "%lu/%lu put_page %p\n", (unsigned long) i,
  46                            (unsigned long) num_pages, p[i]);
  47                 if (dirty)
  48                         set_page_dirty_lock(p[i]);
  49                 put_page(p[i]);
  50         }
  51 }
  52
  53 /* call with current->mm->mmap_sem held */
  54 static int __get_user_pages(unsigned long start_page, size_t num_pages,
  55                         struct page **p, struct vm_area_struct **vma)
  56 {
  57         unsigned long lock_limit;
  58         size_t got;
  59         int ret;
  60
  61 #if 0
  62         /*
  63          * XXX - causes MPI programs to fail, haven't had time to check
  64          * yet
  65          */
  66         if (!capable(CAP_IPC_LOCK)) {
  67                 ret = -EPERM;
  68                 goto bail;
  69         }
  70 #endif
  71
  72         lock_limit = current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur >>
  73                 PAGE_SHIFT;
  74
  75         if (num_pages > lock_limit) {
  76                 ret = -ENOMEM;
  77                 goto bail;
  78         }
  79
  80         ipath_cdbg(VERBOSE, "pin %lx pages from vaddr %lx\n",
  81                    (unsigned long) num_pages, start_page);
  82
  83         for (got = 0; got < num_pages; got += ret) {
  84                 ret = get_user_pages(current, current->mm,
  85                                      start_page + got * PAGE_SIZE,
  86                                      num_pages - got, 1, 1,
  87                                      p + got, vma);
  88                 if (ret < 0)
  89                         goto bail_release;
  90         }
  91
  92         current->mm->locked_vm += num_pages;
  93
  94         ret = 0;
  95         goto bail;
  96
  97 bail_release:
  98         __ipath_release_user_pages(p, got, 0);
  99 bail:
 100         return ret;
 101 }
 102
 103 /**
 104  * ipath_get_user_pages - lock user pages into memory
 105  * @start_page: the start page
 106  * @num_pages: the number of pages
 107  * @p: the output page structures
 108  *
 109  * This function takes a given start page (page aligned user virtual
 110  * address) and pins it and the following specified number of pages.  For
 111  * now, num_pages is always 1, but that will probably change at some point
 112  * (because caller is doing expected sends on a single virtually contiguous
 113  * buffer, so we can do all pages at once).
 114  */
 115 int ipath_get_user_pages(unsigned long start_page, size_t num_pages,
 116                          struct page **p)
 117 {
 118         int ret;
 119
 120         down_write(&current->mm->mmap_sem);
 121
 122         ret = __get_user_pages(start_page, num_pages, p, NULL);
 123
 124         up_write(&current->mm->mmap_sem);
 125
 126         return ret;
 127 }
 128
 129 /**
 130  * ipath_get_user_pages_nocopy - lock a single page for I/O and mark shared
 131  * @start_page: the page to lock
 132  * @p: the output page structure
 133  *
 134  * This is similar to ipath_get_user_pages, but it's always one page, and we
 135  * mark the page as locked for I/O, and shared.  This is used for the user
 136  * process page that contains the destination address for the rcvhdrq tail
 137  * update, so we need to have the vma. If we don't do this, the page can be
 138  * taken away from us on fork, even if the child never touches it, and then
 139  * the user process never sees the tail register updates.
 140  */
 141 int ipath_get_user_pages_nocopy(unsigned long page, struct page **p)
 142 {
 143         struct vm_area_struct *vma;
 144         int ret;
 145
 146         down_write(&current->mm->mmap_sem);
 147
 148         ret = __get_user_pages(page, 1, p, &vma);
 149
 150         up_write(&current->mm->mmap_sem);
 151
 152         return ret;
 153 }
 154
 155 void ipath_release_user_pages(struct page **p, size_t num_pages)
 156 {
 157         down_write(&current->mm->mmap_sem);
 158
 159         __ipath_release_user_pages(p, num_pages, 1);
 160
 161         current->mm->locked_vm -= num_pages;
 162
 163         up_write(&current->mm->mmap_sem);
 164 }
 165
 166 struct ipath_user_pages_work {
 167         struct work_struct work;
 168         struct mm_struct *mm;
 169         unsigned long num_pages;
 170 };
 171
 172 static void user_pages_account(void *ptr)
 173 {
 174         struct ipath_user_pages_work *work = ptr;
 175
 176         down_write(&work->mm->mmap_sem);
 177         work->mm->locked_vm -= work->num_pages;
 178         up_write(&work->mm->mmap_sem);
 179         mmput(work->mm);
 180         kfree(work);
 181 }
 182
 183 void ipath_release_user_pages_on_close(struct page **p, size_t num_pages)
 184 {
 185         struct ipath_user_pages_work *work;
 186         struct mm_struct *mm;
 187
 188         __ipath_release_user_pages(p, num_pages, 1);
 189
 190         mm = get_task_mm(current);
 191         if (!mm)
 192                 goto bail;
 193
 194         work = kmalloc(sizeof(*work), GFP_KERNEL);
 195         if (!work)
 196                 goto bail_mm;
 197
 198         goto bail;
 199
 200         INIT_WORK(&work->work, user_pages_account, work);
 201         work->mm = mm;
 202         work->num_pages = num_pages;
 203
 204 bail_mm:
 205         mmput(mm);
 206 bail:
 207         return;
 208 }