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/error-report.h"
  32 #include "qemu/memfd.h"
  33 #include "qemu/host-utils.h"
  34
  35 #if defined CONFIG_LINUX && !defined CONFIG_MEMFD
  36 #include <sys/syscall.h>
  37 #include <asm/unistd.h>
  38
  39 int memfd_create(const char *name, unsigned int flags)
  40 {
  41 #ifdef __NR_memfd_create
  42     return syscall(__NR_memfd_create, name, flags);
  43 #else
  44     errno = ENOSYS;
  45     return -1;
  46 #endif
  47 }
  48 #endif
  49
  50 int qemu_memfd_create(const char *name, size_t size, bool hugetlb,
  51                       uint64_t hugetlbsize, unsigned int seals, Error **errp)
  52 {
  53     int htsize = hugetlbsize ? ctz64(hugetlbsize) : 0;
  54
  55     if (htsize && 1ULL << htsize != hugetlbsize) {
  56         error_setg(errp, "Hugepage size must be a power of 2");
  57         return -1;
  58     }
  59
  60     htsize = htsize << MFD_HUGE_SHIFT;
  61
  62 #ifdef CONFIG_LINUX
  63     int mfd = -1;
  64     unsigned int flags = MFD_CLOEXEC;
  65
  66     if (seals) {
  67         flags |= MFD_ALLOW_SEALING;
  68     }
  69     if (hugetlb) {
  70         flags |= MFD_HUGETLB;
  71         flags |= htsize;
  72     }
  73     mfd = memfd_create(name, flags);
  74     if (mfd < 0) {
  75         error_setg_errno(errp, errno,
  76                          "failed to create memfd with flags 0x%x", flags);
  77         goto err;
  78     }
  79
  80     if (ftruncate(mfd, size) == -1) {
  81         error_setg_errno(errp, errno, "failed to resize memfd to %zu", size);
  82         goto err;
  83     }
  84
  85     if (seals && fcntl(mfd, F_ADD_SEALS, seals) == -1) {
  86         error_setg_errno(errp, errno, "failed to add seals 0x%x", seals);
  87         goto err;
  88     }
  89
  90     return mfd;
  91
  92 err:
  93     if (mfd >= 0) {
  94         close(mfd);
  95     }
  96 #else
  97     error_setg_errno(errp, ENOSYS, "failed to create memfd");
  98 #endif
  99     return -1;
 100 }
 101
 102 /*
 103  * This is a best-effort helper for shared memory allocation, with
 104  * optional sealing. The helper will do his best to allocate using
 105  * memfd with sealing, but may fallback on other methods without
 106  * sealing.
 107  */
 108 void *qemu_memfd_alloc(const char *name, size_t size, unsigned int seals,
 109                        int *fd, Error **errp)
 110 {
 111     void *ptr;
 112     int mfd = qemu_memfd_create(name, size, false, 0, seals, NULL);
 113
 114     /* some systems have memfd without sealing */
 115     if (mfd == -1) {
 116         mfd = qemu_memfd_create(name, size, false, 0, 0, NULL);
 117     }
 118
 119     if (mfd == -1) {
 120         const char *tmpdir = g_get_tmp_dir();
 121         gchar *fname;
 122
 123         fname = g_strdup_printf("%s/memfd-XXXXXX", tmpdir);
 124         mfd = mkstemp(fname);
 125         unlink(fname);
 126         g_free(fname);
 127
 128         if (mfd == -1 ||
 129             ftruncate(mfd, size) == -1) {
 130             goto err;
 131         }
 132     }
 133
 134     ptr = mmap(0, size, PROT_READ | PROT_WRITE, MAP_SHARED, mfd, 0);
 135     if (ptr == MAP_FAILED) {
 136         goto err;
 137     }
 138
 139     *fd = mfd;
 140     return ptr;
 141
 142 err:
 143     error_setg_errno(errp, errno, "failed to allocate shared memory");
 144     if (mfd >= 0) {
 145         close(mfd);
 146     }
 147     return NULL;
 148 }
 149
 150 void qemu_memfd_free(void *ptr, size_t size, int fd)
 151 {
 152     if (ptr) {
 153         if (munmap(ptr, size) != 0) {
 154             error_report("memfd munmap() failed: %s", strerror(errno));
 155         }
 156     }
 157
 158     if (fd != -1) {
 159         if (close(fd) != 0) {
 160             error_report("memfd close() failed: %s", strerror(errno));
 161         }
 162     }
 163 }
 164
 165 enum {
 166     MEMFD_KO,
 167     MEMFD_OK,
 168     MEMFD_TODO
 169 };
 170
 171 /**
 172  * qemu_memfd_alloc_check():
 173  *
 174  * Check if qemu_memfd_alloc() can allocate, including using a
 175  * fallback implementation when host doesn't support memfd.
 176  */
 177 bool qemu_memfd_alloc_check(void)
 178 {
 179     static int memfd_check = MEMFD_TODO;
 180
 181     if (memfd_check == MEMFD_TODO) {
 182         int fd;
 183         void *ptr;
 184
 185         fd = -1;
 186         ptr = qemu_memfd_alloc("test", 4096, 0, &fd, NULL);
 187         memfd_check = ptr ? MEMFD_OK : MEMFD_KO;
 188         qemu_memfd_free(ptr, 4096, fd);
 189     }
 190
 191     return memfd_check == MEMFD_OK;
 192 }
 193
 194 /**
 195  * qemu_memfd_check():
 196  *
 197  * Check if host supports memfd.
 198  */
 199 bool qemu_memfd_check(unsigned int flags)
 200 {
 201 #ifdef CONFIG_LINUX
 202     int mfd = memfd_create("test", flags | MFD_CLOEXEC);
 203
 204     if (mfd >= 0) {
 205         close(mfd);
 206         return true;
 207     }
 208 #endif
 209
 210     return false;
 211 }