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 static int 66 linux_mmap_check_fp(struct file *fp, int flags, int prot, int maxprot) 67 { 68 69 /* Linux mmap() just fails for O_WRONLY files */ 70 if ((fp->f_flag & FREAD) == 0) 71 return (EACCES); 72 73 return (0); 74 } 75 76 int 77 linux_mmap_common(struct thread *td, uintptr_t addr, size_t len, int prot, 78 int flags, int fd, off_t pos) 79 { 80 struct proc *p = td->td_proc; 81 struct vmspace *vms = td->td_proc->p_vmspace; 82 int bsd_flags, error; 83 struct file *fp; 84 85 LINUX_CTR6(mmap2, "0x%lx, %ld, %ld, 0x%08lx, %ld, 0x%lx", 86 addr, len, prot, flags, fd, pos); 87 88 error = 0; 89 bsd_flags = 0; 90 fp = NULL; 91 92 /* 93 * Linux mmap(2): 94 * You must specify exactly one of MAP_SHARED and MAP_PRIVATE 95 */ 96 if (!((flags & LINUX_MAP_SHARED) ^ (flags & LINUX_MAP_PRIVATE))) 97 return (EINVAL); 98 99 if (flags & LINUX_MAP_SHARED) 100 bsd_flags |= MAP_SHARED; 101 if (flags & LINUX_MAP_PRIVATE) 102 bsd_flags |= MAP_PRIVATE; 103 if (flags & LINUX_MAP_FIXED) 104 bsd_flags |= MAP_FIXED; 105 if (flags & LINUX_MAP_ANON) { 106 /* Enforce pos to be on page boundary, then ignore. */ 107 if ((pos & PAGE_MASK) != 0) 108 return (EINVAL); 109 pos = 0; 110 bsd_flags |= MAP_ANON; 111 } else 112 bsd_flags |= MAP_NOSYNC; 113 if (flags & LINUX_MAP_GROWSDOWN) 114 bsd_flags |= MAP_STACK; 115 116 #if defined(__amd64__) 117 /* 118 * According to the Linux mmap(2) man page, "MAP_32BIT flag 119 * is ignored when MAP_FIXED is set." 120 */ 121 if ((flags & LINUX_MAP_32BIT) && (flags & LINUX_MAP_FIXED) == 0) 122 bsd_flags |= MAP_32BIT; 123 124 /* 125 * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC 126 * on Linux/i386 if the binary requires executable stack. 127 * We do this only for IA32 emulation as on native i386 this is does not 128 * make sense without PAE. 129 * 130 * XXX. Linux checks that the file system is not mounted with noexec. 131 */ 132 linux_fixup_prot(td, &prot); 133 #endif 134 135 /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */ 136 fd = (bsd_flags & MAP_ANON) ? -1 : fd; 137 if (flags & LINUX_MAP_GROWSDOWN) { 138 /* 139 * The Linux MAP_GROWSDOWN option does not limit auto 140 * growth of the region. Linux mmap with this option 141 * takes as addr the initial BOS, and as len, the initial 142 * region size. It can then grow down from addr without 143 * limit. However, Linux threads has an implicit internal 144 * limit to stack size of STACK_SIZE. Its just not 145 * enforced explicitly in Linux. But, here we impose 146 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack 147 * region, since we can do this with our mmap. 148 * 149 * Our mmap with MAP_STACK takes addr as the maximum 150 * downsize limit on BOS, and as len the max size of 151 * the region. It then maps the top SGROWSIZ bytes, 152 * and auto grows the region down, up to the limit 153 * in addr. 154 * 155 * If we don't use the MAP_STACK option, the effect 156 * of this code is to allocate a stack region of a 157 * fixed size of (STACK_SIZE - GUARD_SIZE). 158 */ 159 160 if ((caddr_t)addr + len > vms->vm_maxsaddr) { 161 /* 162 * Some Linux apps will attempt to mmap 163 * thread stacks near the top of their 164 * address space. If their TOS is greater 165 * than vm_maxsaddr, vm_map_growstack() 166 * will confuse the thread stack with the 167 * process stack and deliver a SEGV if they 168 * attempt to grow the thread stack past their 169 * current stacksize rlimit. To avoid this, 170 * adjust vm_maxsaddr upwards to reflect 171 * the current stacksize rlimit rather 172 * than the maximum possible stacksize. 173 * It would be better to adjust the 174 * mmap'ed region, but some apps do not check 175 * mmap's return value. 176 */ 177 PROC_LOCK(p); 178 vms->vm_maxsaddr = (char *)p->p_sysent->sv_usrstack - 179 lim_cur_proc(p, RLIMIT_STACK); 180 PROC_UNLOCK(p); 181 } 182 183 /* 184 * This gives us our maximum stack size and a new BOS. 185 * If we're using VM_STACK, then mmap will just map 186 * the top SGROWSIZ bytes, and let the stack grow down 187 * to the limit at BOS. If we're not using VM_STACK 188 * we map the full stack, since we don't have a way 189 * to autogrow it. 190 */ 191 if (len <= STACK_SIZE - GUARD_SIZE) { 192 addr = addr - (STACK_SIZE - GUARD_SIZE - len); 193 len = STACK_SIZE - GUARD_SIZE; 194 } 195 } 196 197 /* 198 * FreeBSD is free to ignore the address hint if MAP_FIXED wasn't 199 * passed. However, some Linux applications, like the ART runtime, 200 * depend on the hint. If the MAP_FIXED wasn't passed, but the 201 * address is not zero, try with MAP_FIXED and MAP_EXCL first, 202 * and fall back to the normal behaviour if that fails. 203 */ 204 if (addr != 0 && (bsd_flags & MAP_FIXED) == 0 && 205 (bsd_flags & MAP_EXCL) == 0) { 206 error = kern_mmap_fpcheck(td, addr, len, prot, 207 bsd_flags | MAP_FIXED | MAP_EXCL, fd, pos, 208 linux_mmap_check_fp); 209 if (error == 0) 210 goto out; 211 } 212 213 error = kern_mmap_fpcheck(td, addr, len, prot, bsd_flags, fd, pos, 214 linux_mmap_check_fp); 215 out: 216 LINUX_CTR2(mmap2, "return: %d (%p)", error, td->td_retval[0]); 217 218 return (error); 219 } 220 221 int 222 linux_mprotect_common(struct thread *td, uintptr_t addr, size_t len, int prot) 223 { 224 225 /* XXX Ignore PROT_GROWSDOWN and PROT_GROWSUP for now. */ 226 prot &= ~(LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP); 227 if ((prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC)) != 0) 228 return (EINVAL); 229 230 #if defined(__amd64__) 231 linux_fixup_prot(td, &prot); 232 #endif 233 return (kern_mprotect(td, addr, len, prot)); 234 } 235 236 #if defined(__amd64__) 237 static void 238 linux_fixup_prot(struct thread *td, int *prot) 239 { 240 struct linux_pemuldata *pem; 241 242 if (SV_PROC_FLAG(td->td_proc, SV_ILP32) && *prot & PROT_READ) { 243 pem = pem_find(td->td_proc); 244 if (pem->persona & LINUX_READ_IMPLIES_EXEC) 245 *prot |= PROT_EXEC; 246 } 247 248 } 249 #endif 250