1 /*- 2 * Copyright (c) 2004 Tim J. Robbins 3 * Copyright (c) 2002 Doug Rabson 4 * Copyright (c) 2000 Marcel Moolenaar 5 * Copyright (c) 1994-1995 Søren Schmidt 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer 13 * in this position and unchanged. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 3. The name of the author may not be used to endorse or promote products 18 * derived from this software without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 21 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 22 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 23 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 24 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 25 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 26 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 27 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 28 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 29 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 30 * 31 * $FreeBSD$ 32 */ 33 34 #include <sys/cdefs.h> 35 __FBSDID("$FreeBSD$"); 36 37 #include <sys/capsicum.h> 38 #include <sys/file.h> 39 #include <sys/imgact.h> 40 #include <sys/ktr.h> 41 #include <sys/mman.h> 42 #include <sys/proc.h> 43 #include <sys/resourcevar.h> 44 #include <sys/syscallsubr.h> 45 #include <sys/sysent.h> 46 #include <sys/sysproto.h> 47 48 #include <vm/pmap.h> 49 #include <vm/vm_extern.h> 50 #include <vm/vm_map.h> 51 52 #include <compat/linux/linux_emul.h> 53 #include <compat/linux/linux_mmap.h> 54 #include <compat/linux/linux_persona.h> 55 #include <compat/linux/linux_util.h> 56 57 58 #define STACK_SIZE (2 * 1024 * 1024) 59 #define GUARD_SIZE (4 * PAGE_SIZE) 60 61 #if defined(__amd64__) 62 static void linux_fixup_prot(struct thread *td, int *prot); 63 #endif 64 65 66 int 67 linux_mmap_common(struct thread *td, uintptr_t addr, size_t len, int prot, 68 int flags, int fd, off_t pos) 69 { 70 struct proc *p = td->td_proc; 71 struct vmspace *vms = td->td_proc->p_vmspace; 72 int bsd_flags, error; 73 struct file *fp; 74 75 LINUX_CTR6(mmap2, "0x%lx, %ld, %ld, 0x%08lx, %ld, 0x%lx", 76 addr, len, prot, flags, fd, pos); 77 78 error = 0; 79 bsd_flags = 0; 80 fp = NULL; 81 82 /* 83 * Linux mmap(2): 84 * You must specify exactly one of MAP_SHARED and MAP_PRIVATE 85 */ 86 if (!((flags & LINUX_MAP_SHARED) ^ (flags & LINUX_MAP_PRIVATE))) 87 return (EINVAL); 88 89 if (flags & LINUX_MAP_SHARED) 90 bsd_flags |= MAP_SHARED; 91 if (flags & LINUX_MAP_PRIVATE) 92 bsd_flags |= MAP_PRIVATE; 93 if (flags & LINUX_MAP_FIXED) 94 bsd_flags |= MAP_FIXED; 95 if (flags & LINUX_MAP_ANON) { 96 /* Enforce pos to be on page boundary, then ignore. */ 97 if ((pos & PAGE_MASK) != 0) 98 return (EINVAL); 99 pos = 0; 100 bsd_flags |= MAP_ANON; 101 } else 102 bsd_flags |= MAP_NOSYNC; 103 if (flags & LINUX_MAP_GROWSDOWN) 104 bsd_flags |= MAP_STACK; 105 106 /* 107 * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC 108 * on Linux/i386 if the binary requires executable stack. 109 * We do this only for IA32 emulation as on native i386 this is does not 110 * make sense without PAE. 111 * 112 * XXX. Linux checks that the file system is not mounted with noexec. 113 */ 114 #if defined(__amd64__) 115 linux_fixup_prot(td, &prot); 116 #endif 117 118 /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */ 119 fd = (bsd_flags & MAP_ANON) ? -1 : fd; 120 if (fd != -1) { 121 /* 122 * Linux follows Solaris mmap(2) description: 123 * The file descriptor fildes is opened with 124 * read permission, regardless of the 125 * protection options specified. 126 */ 127 128 error = fget(td, fd, &cap_mmap_rights, &fp); 129 if (error != 0) 130 return (error); 131 if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_DEV) { 132 fdrop(fp, td); 133 return (EINVAL); 134 } 135 136 /* Linux mmap() just fails for O_WRONLY files */ 137 if (!(fp->f_flag & FREAD)) { 138 fdrop(fp, td); 139 return (EACCES); 140 } 141 142 fdrop(fp, td); 143 } 144 145 if (flags & LINUX_MAP_GROWSDOWN) { 146 /* 147 * The Linux MAP_GROWSDOWN option does not limit auto 148 * growth of the region. Linux mmap with this option 149 * takes as addr the initial BOS, and as len, the initial 150 * region size. It can then grow down from addr without 151 * limit. However, Linux threads has an implicit internal 152 * limit to stack size of STACK_SIZE. Its just not 153 * enforced explicitly in Linux. But, here we impose 154 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack 155 * region, since we can do this with our mmap. 156 * 157 * Our mmap with MAP_STACK takes addr as the maximum 158 * downsize limit on BOS, and as len the max size of 159 * the region. It then maps the top SGROWSIZ bytes, 160 * and auto grows the region down, up to the limit 161 * in addr. 162 * 163 * If we don't use the MAP_STACK option, the effect 164 * of this code is to allocate a stack region of a 165 * fixed size of (STACK_SIZE - GUARD_SIZE). 166 */ 167 168 if ((caddr_t)addr + len > vms->vm_maxsaddr) { 169 /* 170 * Some Linux apps will attempt to mmap 171 * thread stacks near the top of their 172 * address space. If their TOS is greater 173 * than vm_maxsaddr, vm_map_growstack() 174 * will confuse the thread stack with the 175 * process stack and deliver a SEGV if they 176 * attempt to grow the thread stack past their 177 * current stacksize rlimit. To avoid this, 178 * adjust vm_maxsaddr upwards to reflect 179 * the current stacksize rlimit rather 180 * than the maximum possible stacksize. 181 * It would be better to adjust the 182 * mmap'ed region, but some apps do not check 183 * mmap's return value. 184 */ 185 PROC_LOCK(p); 186 vms->vm_maxsaddr = (char *)p->p_sysent->sv_usrstack - 187 lim_cur_proc(p, RLIMIT_STACK); 188 PROC_UNLOCK(p); 189 } 190 191 /* 192 * This gives us our maximum stack size and a new BOS. 193 * If we're using VM_STACK, then mmap will just map 194 * the top SGROWSIZ bytes, and let the stack grow down 195 * to the limit at BOS. If we're not using VM_STACK 196 * we map the full stack, since we don't have a way 197 * to autogrow it. 198 */ 199 if (len <= STACK_SIZE - GUARD_SIZE) { 200 addr = addr - (STACK_SIZE - GUARD_SIZE - len); 201 len = STACK_SIZE - GUARD_SIZE; 202 } 203 } 204 205 /* 206 * FreeBSD is free to ignore the address hint if MAP_FIXED wasn't 207 * passed. However, some Linux applications, like the ART runtime, 208 * depend on the hint. If the MAP_FIXED wasn't passed, but the 209 * address is not zero, try with MAP_FIXED and MAP_EXCL first, 210 * and fall back to the normal behaviour if that fails. 211 */ 212 if (addr != 0 && (bsd_flags & MAP_FIXED) == 0 && 213 (bsd_flags & MAP_EXCL) == 0) { 214 error = kern_mmap(td, addr, len, prot, 215 bsd_flags | MAP_FIXED | MAP_EXCL, fd, pos); 216 if (error == 0) 217 goto out; 218 } 219 220 error = kern_mmap(td, addr, len, prot, bsd_flags, fd, pos); 221 out: 222 LINUX_CTR2(mmap2, "return: %d (%p)", error, td->td_retval[0]); 223 224 return (error); 225 } 226 227 int 228 linux_mprotect_common(struct thread *td, uintptr_t addr, size_t len, int prot) 229 { 230 231 /* XXX Ignore PROT_GROWSDOWN and PROT_GROWSUP for now. */ 232 prot &= ~(LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP); 233 if ((prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC)) != 0) 234 return (EINVAL); 235 236 #if defined(__amd64__) 237 linux_fixup_prot(td, &prot); 238 #endif 239 return (kern_mprotect(td, addr, len, prot)); 240 } 241 242 #if defined(__amd64__) 243 static void 244 linux_fixup_prot(struct thread *td, int *prot) 245 { 246 struct linux_pemuldata *pem; 247 248 if (SV_PROC_FLAG(td->td_proc, SV_ILP32) && *prot & PROT_READ) { 249 pem = pem_find(td->td_proc); 250 if (pem->persona & LINUX_READ_IMPLIES_EXEC) 251 *prot |= PROT_EXEC; 252 } 253 254 } 255 #endif 256