linux_exec_machdep.c revision 1.6
1 /* $NetBSD: linux_exec_machdep.c,v 1.6 2008/10/25 23:38:28 christos Exp $ */ 2 3 /*- 4 * Copyright (c) 2004 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Christos Zoulas. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 #include <sys/cdefs.h> 33 __KERNEL_RCSID(0, "$NetBSD: linux_exec_machdep.c,v 1.6 2008/10/25 23:38:28 christos Exp $"); 34 35 #if defined(_KERNEL_OPT) 36 #include "opt_vm86.h" 37 #include "opt_user_ldt.h" 38 #endif 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/resource.h> 43 #include <sys/proc.h> 44 #include <sys/conf.h> 45 #include <sys/exec.h> 46 #include <sys/exec_elf.h> 47 #include <sys/vnode.h> 48 #include <sys/lwp.h> 49 50 #include <sys/cpu.h> 51 #include <machine/vmparam.h> 52 53 #include <uvm/uvm.h> 54 55 #include <sys/syscallargs.h> 56 57 #ifndef DEBUG_LINUX 58 #define DPRINTF(a) 59 #else 60 #define DPRINTF(a) uprintf a 61 #endif 62 63 #include <compat/linux/common/linux_types.h> 64 #include <compat/linux/common/linux_signal.h> 65 #include <compat/linux/common/linux_machdep.h> 66 #include <compat/linux/common/linux_util.h> 67 #include <compat/linux/common/linux_ioctl.h> 68 #include <compat/linux/common/linux_hdio.h> 69 #include <compat/linux/common/linux_exec.h> 70 #include <compat/linux/common/linux_errno.h> 71 #include <compat/linux//linux_syscallargs.h> 72 73 int 74 linux_exec_setup_stack(struct lwp *l, struct exec_package *epp) 75 { 76 u_long max_stack_size; 77 u_long access_linear_min, access_size; 78 u_long noaccess_linear_min, noaccess_size; 79 80 #ifndef USRSTACK32 81 #define USRSTACK32 (0x00000000ffffffffL&~PGOFSET) 82 #endif 83 84 if (epp->ep_flags & EXEC_32) { 85 epp->ep_minsaddr = USRSTACK32; 86 max_stack_size = MAXSSIZ; 87 } else { 88 epp->ep_minsaddr = USRSTACK; 89 max_stack_size = MAXSSIZ; 90 } 91 92 if (epp->ep_minsaddr > LINUX_USRSTACK) 93 epp->ep_minsaddr = LINUX_USRSTACK; 94 #ifdef DEBUG_LINUX 95 else { 96 /* 97 * Someone needs to make KERNBASE and TEXTADDR 98 * java versions < 1.4.2 need the stack to be 99 * at 0xC0000000 100 */ 101 uprintf("Cannot setup stack to 0xC0000000, " 102 "java will not work properly\n"); 103 } 104 #endif 105 epp->ep_maxsaddr = (u_long)STACK_GROW(epp->ep_minsaddr, 106 max_stack_size); 107 epp->ep_ssize = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur; 108 109 /* 110 * set up commands for stack. note that this takes *two*, one to 111 * map the part of the stack which we can access, and one to map 112 * the part which we can't. 113 * 114 * arguably, it could be made into one, but that would require the 115 * addition of another mapping proc, which is unnecessary 116 */ 117 access_size = epp->ep_ssize; 118 access_linear_min = (u_long)STACK_ALLOC(epp->ep_minsaddr, access_size); 119 noaccess_size = max_stack_size - access_size; 120 noaccess_linear_min = (u_long)STACK_ALLOC(STACK_GROW(epp->ep_minsaddr, 121 access_size), noaccess_size); 122 if (noaccess_size > 0) { 123 NEW_VMCMD(&epp->ep_vmcmds, vmcmd_map_zero, noaccess_size, 124 noaccess_linear_min, NULLVP, 0, VM_PROT_NONE); 125 } 126 KASSERT(access_size > 0); 127 NEW_VMCMD(&epp->ep_vmcmds, vmcmd_map_zero, access_size, 128 access_linear_min, NULLVP, 0, VM_PROT_READ | VM_PROT_WRITE); 129 130 return 0; 131 } 132 133 134 #ifdef LINUX_NPTL 135 136 int 137 linux_init_thread_area(struct lwp *l, struct lwp *l2) 138 { 139 struct trapframe *tf = l->l_md.md_regs, *tf2 = l2->l_md.md_regs; 140 struct pcb *pcb2 = &l2->l_addr->u_pcb; 141 struct linux_user_desc info; 142 struct segment_descriptor sd; 143 int error, idx, a[2]; 144 145 error = copyin((void *)tf->tf_esi, &info, sizeof(info)); 146 if (error) 147 return error; 148 idx = info.entry_number; 149 150 /* 151 * looks like we're getting the idx we returned 152 * in the set_thread_area() syscall 153 */ 154 if (idx != 6 && idx != 3) { 155 printf("resetting idx %d to 3", idx); 156 idx = 3; 157 } 158 159 /* this doesnt happen in practice */ 160 if (idx == 6) { 161 /* we might copy out the entry_number as 3 */ 162 info.entry_number = 3; 163 error = copyout(&info, (void *)tf->tf_esi, sizeof(info)); 164 if (error) 165 return error; 166 } 167 168 a[0] = LINUX_LDT_entry_a(&info); 169 a[1] = LINUX_LDT_entry_b(&info); 170 171 (void)memcpy(&sd, &a, sizeof(a)); 172 DPRINTF(("Segment created in clone with CLONE_SETTLS: lobase: %x, " 173 "hibase: %x, lolimit: %x, hilimit: %x, type: %i, dpl: %i, p: %i, " 174 "xx: %i, def32: %i, gran: %i\n", sd.sd_lobase, 175 sd.sd_hibase, sd.sd_lolimit, sd.sd_hilimit, sd.sd_type, sd.sd_dpl, 176 sd.sd_p, sd.sd_xx, sd.sd_def32, sd.sd_gran)); 177 178 (void)memcpy(pcb2->pcb_gsd, &sd, sizeof(pcb2->pcb_gsd)); 179 tf2->tf_gs = GSEL(GUGS_SEL, SEL_UPL); 180 return 0; 181 } 182 183 184 int 185 linux_sys_set_thread_area(struct lwp *l, 186 const struct linux_sys_set_thread_area_args *uap, register_t *retval) 187 { 188 struct trapframe *tf = l->l_md.md_regs; 189 struct pcb *pcb = &l->l_addr->u_pcb; 190 struct linux_user_desc info; 191 struct segment_descriptor sd; 192 int error, idx, a[2]; 193 194 *retval = 0; 195 error = copyin(SCARG(uap, desc), &info, sizeof(info)); 196 if (error) 197 return error; 198 199 DPRINTF(("set thread area: %i, %x, %x, %i, %i, %i, %i, %i, %i\n", 200 info.entry_number, info.base_addr, info.limit, info.seg_32bit, 201 info.contents, info.read_exec_only, info.limit_in_pages, 202 info.seg_not_present, info.useable)); 203 204 idx = info.entry_number; 205 /* 206 * Semantics of linux version: every thread in the system has array of 207 * 3 tls descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown. This 208 * syscall loads one of the selected tls decriptors with a value and 209 * also loads GDT descriptors 6, 7 and 8 with the content of the 210 * per-thread descriptors. 211 * 212 * Semantics of fbsd version: I think we can ignore that linux has 3 213 * per-thread descriptors and use just the 1st one. The tls_array[] 214 * is used only in set/get-thread_area() syscalls and for loading the 215 * GDT descriptors. In fbsd we use just one GDT descriptor for TLS so 216 * we will load just one. 217 * 218 * XXX: this doesn't work when a user space process tries to use more 219 * than 1 TLS segment. Comment in the linux sources says wine might do 220 * this. 221 */ 222 223 /* 224 * we support just GLIBC TLS now 225 * we should let 3 proceed as well because we use this segment so 226 * if code does two subsequent calls it should succeed 227 */ 228 if (idx != 6 && idx != -1 && idx != 3) 229 return EINVAL; 230 231 /* 232 * we have to copy out the GDT entry we use 233 * FreeBSD uses GDT entry #3 for storing %gs so load that 234 * 235 * XXX: what if a user space program doesn't check this value and tries 236 * to use 6, 7 or 8? 237 */ 238 idx = info.entry_number = 3; 239 error = copyout(&info, SCARG(uap, desc), sizeof(info)); 240 if (error) 241 return error; 242 243 if (LINUX_LDT_empty(&info)) { 244 a[0] = 0; 245 a[1] = 0; 246 } else { 247 a[0] = LINUX_LDT_entry_a(&info); 248 a[1] = LINUX_LDT_entry_b(&info); 249 } 250 251 (void)memcpy(&sd, &a, sizeof(a)); 252 DPRINTF(("Segment created in set_thread_area: lobase: %x, hibase: %x, " 253 "lolimit: %x, hilimit: %x, type: %i, dpl: %i, p: %i, xx: %i, " 254 "def32: %i, gran: %i\n", sd.sd_lobase, sd.sd_hibase, sd.sd_lolimit, 255 sd.sd_hilimit, sd.sd_type, sd.sd_dpl, sd.sd_p, sd.sd_xx, 256 sd.sd_def32, sd.sd_gran)); 257 258 (void)memcpy(pcb->pcb_gsd, &sd, sizeof(pcb->pcb_gsd)); 259 tf->tf_gs = GSEL(GUGS_SEL, SEL_UPL); 260 return 0; 261 } 262 263 int 264 linux_sys_get_thread_area(struct lwp *l, 265 const struct linux_sys_get_thread_area_args *uap, register_t *retval) 266 { 267 struct pcb *pcb = &l->l_addr->u_pcb; 268 struct linux_user_desc info; 269 struct linux_desc_struct desc; 270 struct segment_descriptor sd; 271 int error, idx; 272 273 *retval = 0; 274 error = copyin(SCARG(uap, desc), &info, sizeof(info)); 275 if (error) 276 return error; 277 278 idx = info.entry_number; 279 /* XXX: I am not sure if we want 3 to be allowed too. */ 280 if (idx != 6 && idx != 3) 281 return EINVAL; 282 283 idx = 3; 284 285 (void)memset(&info, 0, sizeof(info)); 286 (void)memcpy(&sd, pcb->pcb_gsd, sizeof(sd)); 287 (void)memcpy(&desc, &sd, sizeof(desc)); 288 289 info.entry_number = idx; 290 info.base_addr = LINUX_GET_BASE(&desc); 291 info.limit = LINUX_GET_LIMIT(&desc); 292 info.seg_32bit = LINUX_GET_32BIT(&desc); 293 info.contents = LINUX_GET_CONTENTS(&desc); 294 info.read_exec_only = !LINUX_GET_WRITABLE(&desc); 295 info.limit_in_pages = LINUX_GET_LIMIT_PAGES(&desc); 296 info.seg_not_present = !LINUX_GET_PRESENT(&desc); 297 info.useable = LINUX_GET_USEABLE(&desc); 298 299 return copyout(&info, SCARG(uap, desc), sizeof(info)); 300 } 301 302 #endif 303
Indexes created Tue Sep 23 07:09:52 GMT 2025