xref: /openbmc/qemu/util/memfd.c (revision a77d20ba)
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 
memfd_create(const char * name,unsigned int flags)38 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     errno = ENOSYS;
44     return -1;
45 #endif
46 }
47 #endif
48 
qemu_memfd_create(const char * name,size_t size,bool hugetlb,uint64_t hugetlbsize,unsigned int seals,Error ** errp)49 int qemu_memfd_create(const char *name, size_t size, bool hugetlb,
50                       uint64_t hugetlbsize, unsigned int seals, Error **errp)
51 {
52     int htsize = hugetlbsize ? ctz64(hugetlbsize) : 0;
53 
54     if (htsize && 1ULL << htsize != hugetlbsize) {
55         error_setg(errp, "Hugepage size must be a power of 2");
56         return -1;
57     }
58 
59     htsize = htsize << MFD_HUGE_SHIFT;
60 
61 #ifdef CONFIG_LINUX
62     int mfd = -1;
63     unsigned int flags = MFD_CLOEXEC;
64 
65     if (seals) {
66         flags |= MFD_ALLOW_SEALING;
67     }
68     if (hugetlb) {
69         flags |= MFD_HUGETLB;
70         flags |= htsize;
71     }
72     mfd = memfd_create(name, flags);
73     if (mfd < 0) {
74         error_setg_errno(errp, errno,
75                          "failed to create memfd with flags 0x%x", flags);
76         goto err;
77     }
78 
79     if (ftruncate(mfd, size) == -1) {
80         error_setg_errno(errp, errno, "failed to resize memfd to %zu", size);
81         goto err;
82     }
83 
84     if (seals && fcntl(mfd, F_ADD_SEALS, seals) == -1) {
85         error_setg_errno(errp, errno, "failed to add seals 0x%x", seals);
86         goto err;
87     }
88 
89     return mfd;
90 
91 err:
92     if (mfd >= 0) {
93         close(mfd);
94     }
95 #else
96     error_setg_errno(errp, ENOSYS, "failed to create memfd");
97 #endif
98     return -1;
99 }
100 
101 /*
102  * This is a best-effort helper for shared memory allocation, with
103  * optional sealing. The helper will do his best to allocate using
104  * memfd with sealing, but may fallback on other methods without
105  * sealing.
106  */
qemu_memfd_alloc(const char * name,size_t size,unsigned int seals,int * fd,Error ** errp)107 void *qemu_memfd_alloc(const char *name, size_t size, unsigned int seals,
108                        int *fd, Error **errp)
109 {
110     void *ptr;
111     int mfd = qemu_memfd_create(name, size, false, 0, seals, NULL);
112 
113     /* some systems have memfd without sealing */
114     if (mfd == -1) {
115         mfd = qemu_memfd_create(name, size, false, 0, 0, NULL);
116     }
117 
118     if (mfd == -1) {
119         const char *tmpdir = g_get_tmp_dir();
120         gchar *fname;
121 
122         fname = g_strdup_printf("%s/memfd-XXXXXX", tmpdir);
123         mfd = mkstemp(fname);
124         unlink(fname);
125         g_free(fname);
126 
127         if (mfd == -1 ||
128             ftruncate(mfd, size) == -1) {
129             goto err;
130         }
131     }
132 
133     ptr = mmap(0, size, PROT_READ | PROT_WRITE, MAP_SHARED, mfd, 0);
134     if (ptr == MAP_FAILED) {
135         goto err;
136     }
137 
138     *fd = mfd;
139     return ptr;
140 
141 err:
142     error_setg_errno(errp, errno, "failed to allocate shared memory");
143     if (mfd >= 0) {
144         close(mfd);
145     }
146     return NULL;
147 }
148 
qemu_memfd_free(void * ptr,size_t size,int fd)149 void qemu_memfd_free(void *ptr, size_t size, int fd)
150 {
151     if (ptr) {
152         munmap(ptr, size);
153     }
154 
155     if (fd != -1) {
156         close(fd);
157     }
158 }
159 
160 enum {
161     MEMFD_KO,
162     MEMFD_OK,
163     MEMFD_TODO
164 };
165 
166 /**
167  * qemu_memfd_alloc_check():
168  *
169  * Check if qemu_memfd_alloc() can allocate, including using a
170  * fallback implementation when host doesn't support memfd.
171  */
qemu_memfd_alloc_check(void)172 bool qemu_memfd_alloc_check(void)
173 {
174     static int memfd_check = MEMFD_TODO;
175 
176     if (memfd_check == MEMFD_TODO) {
177         int fd;
178         void *ptr;
179 
180         fd = -1;
181         ptr = qemu_memfd_alloc("test", 4096, 0, &fd, NULL);
182         memfd_check = ptr ? MEMFD_OK : MEMFD_KO;
183         qemu_memfd_free(ptr, 4096, fd);
184     }
185 
186     return memfd_check == MEMFD_OK;
187 }
188 
189 /**
190  * qemu_memfd_check():
191  *
192  * Check if host supports memfd.
193  */
qemu_memfd_check(unsigned int flags)194 bool qemu_memfd_check(unsigned int flags)
195 {
196 #ifdef CONFIG_LINUX
197     int mfd = memfd_create("test", flags | MFD_CLOEXEC);
198 
199     if (mfd >= 0) {
200         close(mfd);
201         return true;
202     }
203 #endif
204 
205     return false;
206 }
207