util/memfd.c

   1 /*
   2  * memfd.c
   3  *
   4  * Copyright (c) 2015 Red Hat, Inc.
   5  *
   6  * QEMU library functions on POSIX which are shared between QEMU and
   7  * the QEMU tools.
   8  *
   9  * Permission is hereby granted, free of charge, to any person obtaining a copy
  10  * of this software and associated documentation files (the "Software"), to deal
  11  * in the Software without restriction, including without limitation the rights
  12  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  13  * copies of the Software, and to permit persons to whom the Software is
  14  * furnished to do so, subject to the following conditions:
  15  *
  16  * The above copyright notice and this permission notice shall be included in
  17  * all copies or substantial portions of the Software.
  18  *
  19  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  20  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  21  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  22  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  23  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  24  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  25  * THE SOFTWARE.
  26  */
  27
  28 #include "qemu/osdep.h"
  29
  30 #include "qapi/error.h"
  31 #include "qemu/memfd.h"
  32 #include "qemu/host-utils.h"
  33
  34 #if defined CONFIG_LINUX && !defined CONFIG_MEMFD
  35 #include <sys/syscall.h>
  36 #include <asm/unistd.h>
  37
  38 static int memfd_create(const char *name, unsigned int flags)
  39 {
  40 #ifdef __NR_memfd_create
  41     return syscall(__NR_memfd_create, name, flags);
  42 #else
  43     return -1;
  44 #endif
  45 }
  46 #endif
  47
  48 #ifndef MFD_CLOEXEC
  49 #define MFD_CLOEXEC 0x0001U
  50 #endif
  51
  52 #ifndef MFD_ALLOW_SEALING
  53 #define MFD_ALLOW_SEALING 0x0002U
  54 #endif
  55
  56 #ifndef MFD_HUGETLB
  57 #define MFD_HUGETLB 0x0004U
  58 #endif
  59
  60 #ifndef MFD_HUGE_SHIFT
  61 #define MFD_HUGE_SHIFT 26
  62 #endif
  63
  64 int qemu_memfd_create(const char *name, size_t size, bool hugetlb,
  65                       uint64_t hugetlbsize, unsigned int seals, Error **errp)
  66 {
  67     int htsize = hugetlbsize ? ctz64(hugetlbsize) : 0;
  68
  69     if (htsize && 1ULL << htsize != hugetlbsize) {
  70         error_setg(errp, "Hugepage size must be a power of 2");
  71         return -1;
  72     }
  73
  74     htsize = htsize << MFD_HUGE_SHIFT;
  75
  76 #ifdef CONFIG_LINUX
  77     int mfd = -1;
  78     unsigned int flags = MFD_CLOEXEC;
  79
  80     if (seals) {
  81         flags |= MFD_ALLOW_SEALING;
  82     }
  83     if (hugetlb) {
  84         flags |= MFD_HUGETLB;
  85         flags |= htsize;
  86     }
  87     mfd = memfd_create(name, flags);
  88     if (mfd < 0) {
  89         goto err;
  90     }
  91
  92     if (ftruncate(mfd, size) == -1) {
  93         goto err;
  94     }
  95
  96     if (seals && fcntl(mfd, F_ADD_SEALS, seals) == -1) {
  97         goto err;
  98     }
  99
 100     return mfd;
 101
 102 err:
 103     if (mfd >= 0) {
 104         close(mfd);
 105     }
 106 #endif
 107     error_setg_errno(errp, errno, "failed to create memfd");
 108     return -1;
 109 }
 110
 111 /*
 112  * This is a best-effort helper for shared memory allocation, with
 113  * optional sealing. The helper will do his best to allocate using
 114  * memfd with sealing, but may fallback on other methods without
 115  * sealing.
 116  */
 117 void *qemu_memfd_alloc(const char *name, size_t size, unsigned int seals,
 118                        int *fd, Error **errp)
 119 {
 120     void *ptr;
 121     int mfd = qemu_memfd_create(name, size, false, 0, seals, NULL);
 122
 123     /* some systems have memfd without sealing */
 124     if (mfd == -1) {
 125         mfd = qemu_memfd_create(name, size, false, 0, 0, NULL);
 126     }
 127
 128     if (mfd == -1) {
 129         const char *tmpdir = g_get_tmp_dir();
 130         gchar *fname;
 131
 132         fname = g_strdup_printf("%s/memfd-XXXXXX", tmpdir);
 133         mfd = mkstemp(fname);
 134         unlink(fname);
 135         g_free(fname);
 136
 137         if (mfd == -1 ||
 138             ftruncate(mfd, size) == -1) {
 139             goto err;
 140         }
 141     }
 142
 143     ptr = mmap(0, size, PROT_READ | PROT_WRITE, MAP_SHARED, mfd, 0);
 144     if (ptr == MAP_FAILED) {
 145         goto err;
 146     }
 147
 148     *fd = mfd;
 149     return ptr;
 150
 151 err:
 152     error_setg_errno(errp, errno, "failed to allocate shared memory");
 153     if (mfd >= 0) {
 154         close(mfd);
 155     }
 156     return NULL;
 157 }
 158
 159 void qemu_memfd_free(void *ptr, size_t size, int fd)
 160 {
 161     if (ptr) {
 162         munmap(ptr, size);
 163     }
 164
 165     if (fd != -1) {
 166         close(fd);
 167     }
 168 }
 169
 170 enum {
 171     MEMFD_KO,
 172     MEMFD_OK,
 173     MEMFD_TODO
 174 };
 175
 176 /**
 177  * qemu_memfd_alloc_check():
 178  *
 179  * Check if qemu_memfd_alloc() can allocate, including using a
 180  * fallback implementation when host doesn't support memfd.
 181  */
 182 bool qemu_memfd_alloc_check(void)
 183 {
 184     static int memfd_check = MEMFD_TODO;
 185
 186     if (memfd_check == MEMFD_TODO) {
 187         int fd;
 188         void *ptr;
 189
 190         ptr = qemu_memfd_alloc("test", 4096, 0, &fd, NULL);
 191         memfd_check = ptr ? MEMFD_OK : MEMFD_KO;
 192         qemu_memfd_free(ptr, 4096, fd);
 193     }
 194
 195     return memfd_check == MEMFD_OK;
 196 }
 197
 198 /**
 199  * qemu_memfd_check():
 200  *
 201  * Check if host supports memfd.
 202  */
 203 bool qemu_memfd_check(void)
 204 {
 205 #ifdef CONFIG_LINUX
 206     static int memfd_check = MEMFD_TODO;
 207
 208     if (memfd_check == MEMFD_TODO) {
 209         int mfd = memfd_create("test", 0);
 210         if (mfd >= 0) {
 211             memfd_check = MEMFD_OK;
 212             close(mfd);
 213         } else {
 214             memfd_check = MEMFD_KO;
 215         }
 216     }
 217
 218     return memfd_check == MEMFD_OK;
 219 #else
 220     return false;
 221 #endif
 222 }