Deleted Added
full compact
linux_sysvec.c (293495) linux_sysvec.c (293514)
1/*-
2 * Copyright (c) 1994-1996 S��ren Schmidt
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer
10 * in this position and unchanged.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 3. The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29#include <sys/cdefs.h>
1/*-
2 * Copyright (c) 1994-1996 S��ren Schmidt
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer
10 * in this position and unchanged.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 3. The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD: stable/10/sys/i386/linux/linux_sysvec.c 293495 2016-01-09 15:18:36Z dchagin $");
30__FBSDID("$FreeBSD: stable/10/sys/i386/linux/linux_sysvec.c 293514 2016-01-09 15:44:38Z dchagin $");
31
32#include <sys/param.h>
33#include <sys/systm.h>
34#include <sys/exec.h>
35#include <sys/fcntl.h>
36#include <sys/imgact.h>
37#include <sys/imgact_aout.h>
38#include <sys/imgact_elf.h>
39#include <sys/kernel.h>
40#include <sys/lock.h>
41#include <sys/malloc.h>
42#include <sys/module.h>
43#include <sys/mutex.h>
44#include <sys/proc.h>
45#include <sys/signalvar.h>
46#include <sys/syscallsubr.h>
47#include <sys/sysent.h>
48#include <sys/sysproto.h>
49#include <sys/vnode.h>
50#include <sys/eventhandler.h>
51
52#include <vm/vm.h>
53#include <vm/pmap.h>
54#include <vm/vm_extern.h>
55#include <vm/vm_map.h>
56#include <vm/vm_object.h>
57#include <vm/vm_page.h>
58#include <vm/vm_param.h>
59
60#include <machine/cpu.h>
61#include <machine/cputypes.h>
62#include <machine/md_var.h>
63#include <machine/pcb.h>
64
65#include <i386/linux/linux.h>
66#include <i386/linux/linux_proto.h>
67#include <compat/linux/linux_emul.h>
68#include <compat/linux/linux_futex.h>
69#include <compat/linux/linux_ioctl.h>
70#include <compat/linux/linux_mib.h>
71#include <compat/linux/linux_misc.h>
72#include <compat/linux/linux_signal.h>
73#include <compat/linux/linux_util.h>
31
32#include <sys/param.h>
33#include <sys/systm.h>
34#include <sys/exec.h>
35#include <sys/fcntl.h>
36#include <sys/imgact.h>
37#include <sys/imgact_aout.h>
38#include <sys/imgact_elf.h>
39#include <sys/kernel.h>
40#include <sys/lock.h>
41#include <sys/malloc.h>
42#include <sys/module.h>
43#include <sys/mutex.h>
44#include <sys/proc.h>
45#include <sys/signalvar.h>
46#include <sys/syscallsubr.h>
47#include <sys/sysent.h>
48#include <sys/sysproto.h>
49#include <sys/vnode.h>
50#include <sys/eventhandler.h>
51
52#include <vm/vm.h>
53#include <vm/pmap.h>
54#include <vm/vm_extern.h>
55#include <vm/vm_map.h>
56#include <vm/vm_object.h>
57#include <vm/vm_page.h>
58#include <vm/vm_param.h>
59
60#include <machine/cpu.h>
61#include <machine/cputypes.h>
62#include <machine/md_var.h>
63#include <machine/pcb.h>
64
65#include <i386/linux/linux.h>
66#include <i386/linux/linux_proto.h>
67#include <compat/linux/linux_emul.h>
68#include <compat/linux/linux_futex.h>
69#include <compat/linux/linux_ioctl.h>
70#include <compat/linux/linux_mib.h>
71#include <compat/linux/linux_misc.h>
72#include <compat/linux/linux_signal.h>
73#include <compat/linux/linux_util.h>
74#include <compat/linux/linux_vdso.h>
74
75MODULE_VERSION(linux, 1);
76
77MALLOC_DEFINE(M_LINUX, "linux", "Linux mode structures");
78
79#if BYTE_ORDER == LITTLE_ENDIAN
80#define SHELLMAGIC 0x2123 /* #! */
81#else
82#define SHELLMAGIC 0x2321
83#endif
84
85/*
86 * Allow the sendsig functions to use the ldebug() facility
87 * even though they are not syscalls themselves. Map them
88 * to syscall 0. This is slightly less bogus than using
89 * ldebug(sigreturn).
90 */
91#define LINUX_SYS_linux_rt_sendsig 0
92#define LINUX_SYS_linux_sendsig 0
93
94#define LINUX_PS_STRINGS (LINUX_USRSTACK - sizeof(struct ps_strings))
95
75
76MODULE_VERSION(linux, 1);
77
78MALLOC_DEFINE(M_LINUX, "linux", "Linux mode structures");
79
80#if BYTE_ORDER == LITTLE_ENDIAN
81#define SHELLMAGIC 0x2123 /* #! */
82#else
83#define SHELLMAGIC 0x2321
84#endif
85
86/*
87 * Allow the sendsig functions to use the ldebug() facility
88 * even though they are not syscalls themselves. Map them
89 * to syscall 0. This is slightly less bogus than using
90 * ldebug(sigreturn).
91 */
92#define LINUX_SYS_linux_rt_sendsig 0
93#define LINUX_SYS_linux_sendsig 0
94
95#define LINUX_PS_STRINGS (LINUX_USRSTACK - sizeof(struct ps_strings))
96
96extern char linux_sigcode[];
97extern int linux_szsigcode;
97static int linux_szsigcode;
98static vm_object_t linux_shared_page_obj;
99static char *linux_shared_page_mapping;
100extern char _binary_linux_locore_o_start;
101extern char _binary_linux_locore_o_end;
98
99extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL];
100
101SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler);
102SET_DECLARE(linux_device_handler_set, struct linux_device_handler);
103
104static int linux_fixup(register_t **stack_base,
105 struct image_params *iparams);
106static int elf_linux_fixup(register_t **stack_base,
107 struct image_params *iparams);
108static void linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask);
109static void exec_linux_setregs(struct thread *td,
110 struct image_params *imgp, u_long stack);
111static register_t *linux_copyout_strings(struct image_params *imgp);
112static boolean_t linux_trans_osrel(const Elf_Note *note, int32_t *osrel);
102
103extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL];
104
105SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler);
106SET_DECLARE(linux_device_handler_set, struct linux_device_handler);
107
108static int linux_fixup(register_t **stack_base,
109 struct image_params *iparams);
110static int elf_linux_fixup(register_t **stack_base,
111 struct image_params *iparams);
112static void linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask);
113static void exec_linux_setregs(struct thread *td,
114 struct image_params *imgp, u_long stack);
115static register_t *linux_copyout_strings(struct image_params *imgp);
116static boolean_t linux_trans_osrel(const Elf_Note *note, int32_t *osrel);
117static void linux_vdso_install(void *param);
118static void linux_vdso_deinstall(void *param);
113
114static int linux_szplatform;
115const char *linux_platform;
116
117static eventhandler_tag linux_exit_tag;
118static eventhandler_tag linux_exec_tag;
119static eventhandler_tag linux_thread_dtor_tag;
120
121/*
122 * Linux syscalls return negative errno's, we do positive and map them
123 * Reference:
124 * FreeBSD: src/sys/sys/errno.h
125 * Linux: linux-2.6.17.8/include/asm-generic/errno-base.h
126 * linux-2.6.17.8/include/asm-generic/errno.h
127 */
128static int bsd_to_linux_errno[ELAST + 1] = {
129 -0, -1, -2, -3, -4, -5, -6, -7, -8, -9,
130 -10, -35, -12, -13, -14, -15, -16, -17, -18, -19,
131 -20, -21, -22, -23, -24, -25, -26, -27, -28, -29,
132 -30, -31, -32, -33, -34, -11,-115,-114, -88, -89,
133 -90, -91, -92, -93, -94, -95, -96, -97, -98, -99,
134 -100,-101,-102,-103,-104,-105,-106,-107,-108,-109,
135 -110,-111, -40, -36,-112,-113, -39, -11, -87,-122,
136 -116, -66, -6, -6, -6, -6, -6, -37, -38, -9,
137 -6, -6, -43, -42, -75,-125, -84, -95, -16, -74,
138 -72, -67, -71
139};
140
141int bsd_to_linux_signal[LINUX_SIGTBLSZ] = {
142 LINUX_SIGHUP, LINUX_SIGINT, LINUX_SIGQUIT, LINUX_SIGILL,
143 LINUX_SIGTRAP, LINUX_SIGABRT, 0, LINUX_SIGFPE,
144 LINUX_SIGKILL, LINUX_SIGBUS, LINUX_SIGSEGV, LINUX_SIGSYS,
145 LINUX_SIGPIPE, LINUX_SIGALRM, LINUX_SIGTERM, LINUX_SIGURG,
146 LINUX_SIGSTOP, LINUX_SIGTSTP, LINUX_SIGCONT, LINUX_SIGCHLD,
147 LINUX_SIGTTIN, LINUX_SIGTTOU, LINUX_SIGIO, LINUX_SIGXCPU,
148 LINUX_SIGXFSZ, LINUX_SIGVTALRM, LINUX_SIGPROF, LINUX_SIGWINCH,
149 0, LINUX_SIGUSR1, LINUX_SIGUSR2
150};
151
152int linux_to_bsd_signal[LINUX_SIGTBLSZ] = {
153 SIGHUP, SIGINT, SIGQUIT, SIGILL,
154 SIGTRAP, SIGABRT, SIGBUS, SIGFPE,
155 SIGKILL, SIGUSR1, SIGSEGV, SIGUSR2,
156 SIGPIPE, SIGALRM, SIGTERM, SIGBUS,
157 SIGCHLD, SIGCONT, SIGSTOP, SIGTSTP,
158 SIGTTIN, SIGTTOU, SIGURG, SIGXCPU,
159 SIGXFSZ, SIGVTALRM, SIGPROF, SIGWINCH,
160 SIGIO, SIGURG, SIGSYS
161};
162
163#define LINUX_T_UNKNOWN 255
164static int _bsd_to_linux_trapcode[] = {
165 LINUX_T_UNKNOWN, /* 0 */
166 6, /* 1 T_PRIVINFLT */
167 LINUX_T_UNKNOWN, /* 2 */
168 3, /* 3 T_BPTFLT */
169 LINUX_T_UNKNOWN, /* 4 */
170 LINUX_T_UNKNOWN, /* 5 */
171 16, /* 6 T_ARITHTRAP */
172 254, /* 7 T_ASTFLT */
173 LINUX_T_UNKNOWN, /* 8 */
174 13, /* 9 T_PROTFLT */
175 1, /* 10 T_TRCTRAP */
176 LINUX_T_UNKNOWN, /* 11 */
177 14, /* 12 T_PAGEFLT */
178 LINUX_T_UNKNOWN, /* 13 */
179 17, /* 14 T_ALIGNFLT */
180 LINUX_T_UNKNOWN, /* 15 */
181 LINUX_T_UNKNOWN, /* 16 */
182 LINUX_T_UNKNOWN, /* 17 */
183 0, /* 18 T_DIVIDE */
184 2, /* 19 T_NMI */
185 4, /* 20 T_OFLOW */
186 5, /* 21 T_BOUND */
187 7, /* 22 T_DNA */
188 8, /* 23 T_DOUBLEFLT */
189 9, /* 24 T_FPOPFLT */
190 10, /* 25 T_TSSFLT */
191 11, /* 26 T_SEGNPFLT */
192 12, /* 27 T_STKFLT */
193 18, /* 28 T_MCHK */
194 19, /* 29 T_XMMFLT */
195 15 /* 30 T_RESERVED */
196};
197#define bsd_to_linux_trapcode(code) \
198 ((code)<sizeof(_bsd_to_linux_trapcode)/sizeof(*_bsd_to_linux_trapcode)? \
199 _bsd_to_linux_trapcode[(code)]: \
200 LINUX_T_UNKNOWN)
201
119
120static int linux_szplatform;
121const char *linux_platform;
122
123static eventhandler_tag linux_exit_tag;
124static eventhandler_tag linux_exec_tag;
125static eventhandler_tag linux_thread_dtor_tag;
126
127/*
128 * Linux syscalls return negative errno's, we do positive and map them
129 * Reference:
130 * FreeBSD: src/sys/sys/errno.h
131 * Linux: linux-2.6.17.8/include/asm-generic/errno-base.h
132 * linux-2.6.17.8/include/asm-generic/errno.h
133 */
134static int bsd_to_linux_errno[ELAST + 1] = {
135 -0, -1, -2, -3, -4, -5, -6, -7, -8, -9,
136 -10, -35, -12, -13, -14, -15, -16, -17, -18, -19,
137 -20, -21, -22, -23, -24, -25, -26, -27, -28, -29,
138 -30, -31, -32, -33, -34, -11,-115,-114, -88, -89,
139 -90, -91, -92, -93, -94, -95, -96, -97, -98, -99,
140 -100,-101,-102,-103,-104,-105,-106,-107,-108,-109,
141 -110,-111, -40, -36,-112,-113, -39, -11, -87,-122,
142 -116, -66, -6, -6, -6, -6, -6, -37, -38, -9,
143 -6, -6, -43, -42, -75,-125, -84, -95, -16, -74,
144 -72, -67, -71
145};
146
147int bsd_to_linux_signal[LINUX_SIGTBLSZ] = {
148 LINUX_SIGHUP, LINUX_SIGINT, LINUX_SIGQUIT, LINUX_SIGILL,
149 LINUX_SIGTRAP, LINUX_SIGABRT, 0, LINUX_SIGFPE,
150 LINUX_SIGKILL, LINUX_SIGBUS, LINUX_SIGSEGV, LINUX_SIGSYS,
151 LINUX_SIGPIPE, LINUX_SIGALRM, LINUX_SIGTERM, LINUX_SIGURG,
152 LINUX_SIGSTOP, LINUX_SIGTSTP, LINUX_SIGCONT, LINUX_SIGCHLD,
153 LINUX_SIGTTIN, LINUX_SIGTTOU, LINUX_SIGIO, LINUX_SIGXCPU,
154 LINUX_SIGXFSZ, LINUX_SIGVTALRM, LINUX_SIGPROF, LINUX_SIGWINCH,
155 0, LINUX_SIGUSR1, LINUX_SIGUSR2
156};
157
158int linux_to_bsd_signal[LINUX_SIGTBLSZ] = {
159 SIGHUP, SIGINT, SIGQUIT, SIGILL,
160 SIGTRAP, SIGABRT, SIGBUS, SIGFPE,
161 SIGKILL, SIGUSR1, SIGSEGV, SIGUSR2,
162 SIGPIPE, SIGALRM, SIGTERM, SIGBUS,
163 SIGCHLD, SIGCONT, SIGSTOP, SIGTSTP,
164 SIGTTIN, SIGTTOU, SIGURG, SIGXCPU,
165 SIGXFSZ, SIGVTALRM, SIGPROF, SIGWINCH,
166 SIGIO, SIGURG, SIGSYS
167};
168
169#define LINUX_T_UNKNOWN 255
170static int _bsd_to_linux_trapcode[] = {
171 LINUX_T_UNKNOWN, /* 0 */
172 6, /* 1 T_PRIVINFLT */
173 LINUX_T_UNKNOWN, /* 2 */
174 3, /* 3 T_BPTFLT */
175 LINUX_T_UNKNOWN, /* 4 */
176 LINUX_T_UNKNOWN, /* 5 */
177 16, /* 6 T_ARITHTRAP */
178 254, /* 7 T_ASTFLT */
179 LINUX_T_UNKNOWN, /* 8 */
180 13, /* 9 T_PROTFLT */
181 1, /* 10 T_TRCTRAP */
182 LINUX_T_UNKNOWN, /* 11 */
183 14, /* 12 T_PAGEFLT */
184 LINUX_T_UNKNOWN, /* 13 */
185 17, /* 14 T_ALIGNFLT */
186 LINUX_T_UNKNOWN, /* 15 */
187 LINUX_T_UNKNOWN, /* 16 */
188 LINUX_T_UNKNOWN, /* 17 */
189 0, /* 18 T_DIVIDE */
190 2, /* 19 T_NMI */
191 4, /* 20 T_OFLOW */
192 5, /* 21 T_BOUND */
193 7, /* 22 T_DNA */
194 8, /* 23 T_DOUBLEFLT */
195 9, /* 24 T_FPOPFLT */
196 10, /* 25 T_TSSFLT */
197 11, /* 26 T_SEGNPFLT */
198 12, /* 27 T_STKFLT */
199 18, /* 28 T_MCHK */
200 19, /* 29 T_XMMFLT */
201 15 /* 30 T_RESERVED */
202};
203#define bsd_to_linux_trapcode(code) \
204 ((code)<sizeof(_bsd_to_linux_trapcode)/sizeof(*_bsd_to_linux_trapcode)? \
205 _bsd_to_linux_trapcode[(code)]: \
206 LINUX_T_UNKNOWN)
207
208LINUX_VDSO_SYM_INTPTR(linux_sigcode);
209LINUX_VDSO_SYM_INTPTR(linux_rt_sigcode);
210LINUX_VDSO_SYM_INTPTR(linux_vsyscall);
211
202/*
203 * If FreeBSD & Linux have a difference of opinion about what a trap
204 * means, deal with it here.
205 *
206 * MPSAFE
207 */
208static int
209translate_traps(int signal, int trap_code)
210{
211 if (signal != SIGBUS)
212 return (signal);
213 switch (trap_code) {
214 case T_PROTFLT:
215 case T_TSSFLT:
216 case T_DOUBLEFLT:
217 case T_PAGEFLT:
218 return (SIGSEGV);
219 default:
220 return (signal);
221 }
222}
223
224static int
225linux_fixup(register_t **stack_base, struct image_params *imgp)
226{
227 register_t *argv, *envp;
228
229 argv = *stack_base;
230 envp = *stack_base + (imgp->args->argc + 1);
231 (*stack_base)--;
232 suword(*stack_base, (intptr_t)(void *)envp);
233 (*stack_base)--;
234 suword(*stack_base, (intptr_t)(void *)argv);
235 (*stack_base)--;
236 suword(*stack_base, imgp->args->argc);
237 return (0);
238}
239
240static int
241elf_linux_fixup(register_t **stack_base, struct image_params *imgp)
242{
243 struct proc *p;
244 Elf32_Auxargs *args;
245 Elf32_Addr *uplatform;
246 struct ps_strings *arginfo;
247 register_t *pos;
248
249 KASSERT(curthread->td_proc == imgp->proc,
250 ("unsafe elf_linux_fixup(), should be curproc"));
251
252 p = imgp->proc;
253 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
254 uplatform = (Elf32_Addr *)((caddr_t)arginfo - linux_szplatform);
255 args = (Elf32_Auxargs *)imgp->auxargs;
256 pos = *stack_base + (imgp->args->argc + imgp->args->envc + 2);
257
212/*
213 * If FreeBSD & Linux have a difference of opinion about what a trap
214 * means, deal with it here.
215 *
216 * MPSAFE
217 */
218static int
219translate_traps(int signal, int trap_code)
220{
221 if (signal != SIGBUS)
222 return (signal);
223 switch (trap_code) {
224 case T_PROTFLT:
225 case T_TSSFLT:
226 case T_DOUBLEFLT:
227 case T_PAGEFLT:
228 return (SIGSEGV);
229 default:
230 return (signal);
231 }
232}
233
234static int
235linux_fixup(register_t **stack_base, struct image_params *imgp)
236{
237 register_t *argv, *envp;
238
239 argv = *stack_base;
240 envp = *stack_base + (imgp->args->argc + 1);
241 (*stack_base)--;
242 suword(*stack_base, (intptr_t)(void *)envp);
243 (*stack_base)--;
244 suword(*stack_base, (intptr_t)(void *)argv);
245 (*stack_base)--;
246 suword(*stack_base, imgp->args->argc);
247 return (0);
248}
249
250static int
251elf_linux_fixup(register_t **stack_base, struct image_params *imgp)
252{
253 struct proc *p;
254 Elf32_Auxargs *args;
255 Elf32_Addr *uplatform;
256 struct ps_strings *arginfo;
257 register_t *pos;
258
259 KASSERT(curthread->td_proc == imgp->proc,
260 ("unsafe elf_linux_fixup(), should be curproc"));
261
262 p = imgp->proc;
263 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
264 uplatform = (Elf32_Addr *)((caddr_t)arginfo - linux_szplatform);
265 args = (Elf32_Auxargs *)imgp->auxargs;
266 pos = *stack_base + (imgp->args->argc + imgp->args->envc + 2);
267
268 AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR,
269 imgp->proc->p_sysent->sv_shared_page_base);
270 AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO, linux_vsyscall);
258 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature);
259
260 /*
261 * Do not export AT_CLKTCK when emulating Linux kernel prior to 2.4.0,
262 * as it has appeared in the 2.4.0-rc7 first time.
263 * Being exported, AT_CLKTCK is returned by sysconf(_SC_CLK_TCK),
264 * glibc falls back to the hard-coded CLK_TCK value when aux entry
265 * is not present.
266 * Also see linux_times() implementation.
267 */
268 if (linux_kernver(curthread) >= LINUX_KERNVER_2004000)
269 AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz);
270 AUXARGS_ENTRY(pos, AT_PHDR, args->phdr);
271 AUXARGS_ENTRY(pos, AT_PHENT, args->phent);
272 AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum);
273 AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz);
274 AUXARGS_ENTRY(pos, AT_FLAGS, args->flags);
275 AUXARGS_ENTRY(pos, AT_ENTRY, args->entry);
276 AUXARGS_ENTRY(pos, AT_BASE, args->base);
277 AUXARGS_ENTRY(pos, LINUX_AT_SECURE, 0);
278 AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid);
279 AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid);
280 AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid);
281 AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid);
282 AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(uplatform));
283 if (args->execfd != -1)
284 AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd);
285 AUXARGS_ENTRY(pos, AT_NULL, 0);
286
287 free(imgp->auxargs, M_TEMP);
288 imgp->auxargs = NULL;
289
290 (*stack_base)--;
291 suword(*stack_base, (register_t)imgp->args->argc);
292 return (0);
293}
294
295/*
296 * Copied from kern/kern_exec.c
297 */
298static register_t *
299linux_copyout_strings(struct image_params *imgp)
300{
301 int argc, envc;
302 char **vectp;
303 char *stringp, *destp;
304 register_t *stack_base;
305 struct ps_strings *arginfo;
306 struct proc *p;
307
308 /*
309 * Calculate string base and vector table pointers.
310 * Also deal with signal trampoline code for this exec type.
311 */
312 p = imgp->proc;
313 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
314 destp = (caddr_t)arginfo - SPARE_USRSPACE - linux_szplatform -
315 roundup((ARG_MAX - imgp->args->stringspace), sizeof(char *));
316
317 /*
318 * install LINUX_PLATFORM
319 */
320 copyout(linux_platform, ((caddr_t)arginfo - linux_szplatform),
321 linux_szplatform);
322
323 /*
324 * If we have a valid auxargs ptr, prepare some room
325 * on the stack.
326 */
327 if (imgp->auxargs) {
328 /*
329 * 'AT_COUNT*2' is size for the ELF Auxargs data. This is for
330 * lower compatibility.
331 */
332 imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size :
333 (LINUX_AT_COUNT * 2);
334 /*
335 * The '+ 2' is for the null pointers at the end of each of
336 * the arg and env vector sets,and imgp->auxarg_size is room
337 * for argument of Runtime loader.
338 */
339 vectp = (char **)(destp - (imgp->args->argc +
340 imgp->args->envc + 2 + imgp->auxarg_size) * sizeof(char *));
341 } else {
342 /*
343 * The '+ 2' is for the null pointers at the end of each of
344 * the arg and env vector sets
345 */
346 vectp = (char **)(destp - (imgp->args->argc + imgp->args->envc + 2) *
347 sizeof(char *));
348 }
349
350 /*
351 * vectp also becomes our initial stack base
352 */
353 stack_base = (register_t *)vectp;
354
355 stringp = imgp->args->begin_argv;
356 argc = imgp->args->argc;
357 envc = imgp->args->envc;
358
359 /*
360 * Copy out strings - arguments and environment.
361 */
362 copyout(stringp, destp, ARG_MAX - imgp->args->stringspace);
363
364 /*
365 * Fill in "ps_strings" struct for ps, w, etc.
366 */
367 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
368 suword(&arginfo->ps_nargvstr, argc);
369
370 /*
371 * Fill in argument portion of vector table.
372 */
373 for (; argc > 0; --argc) {
374 suword(vectp++, (long)(intptr_t)destp);
375 while (*stringp++ != 0)
376 destp++;
377 destp++;
378 }
379
380 /* a null vector table pointer separates the argp's from the envp's */
381 suword(vectp++, 0);
382
383 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
384 suword(&arginfo->ps_nenvstr, envc);
385
386 /*
387 * Fill in environment portion of vector table.
388 */
389 for (; envc > 0; --envc) {
390 suword(vectp++, (long)(intptr_t)destp);
391 while (*stringp++ != 0)
392 destp++;
393 destp++;
394 }
395
396 /* end of vector table is a null pointer */
397 suword(vectp, 0);
398
399 return (stack_base);
400}
401
271 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature);
272
273 /*
274 * Do not export AT_CLKTCK when emulating Linux kernel prior to 2.4.0,
275 * as it has appeared in the 2.4.0-rc7 first time.
276 * Being exported, AT_CLKTCK is returned by sysconf(_SC_CLK_TCK),
277 * glibc falls back to the hard-coded CLK_TCK value when aux entry
278 * is not present.
279 * Also see linux_times() implementation.
280 */
281 if (linux_kernver(curthread) >= LINUX_KERNVER_2004000)
282 AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz);
283 AUXARGS_ENTRY(pos, AT_PHDR, args->phdr);
284 AUXARGS_ENTRY(pos, AT_PHENT, args->phent);
285 AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum);
286 AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz);
287 AUXARGS_ENTRY(pos, AT_FLAGS, args->flags);
288 AUXARGS_ENTRY(pos, AT_ENTRY, args->entry);
289 AUXARGS_ENTRY(pos, AT_BASE, args->base);
290 AUXARGS_ENTRY(pos, LINUX_AT_SECURE, 0);
291 AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid);
292 AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid);
293 AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid);
294 AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid);
295 AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(uplatform));
296 if (args->execfd != -1)
297 AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd);
298 AUXARGS_ENTRY(pos, AT_NULL, 0);
299
300 free(imgp->auxargs, M_TEMP);
301 imgp->auxargs = NULL;
302
303 (*stack_base)--;
304 suword(*stack_base, (register_t)imgp->args->argc);
305 return (0);
306}
307
308/*
309 * Copied from kern/kern_exec.c
310 */
311static register_t *
312linux_copyout_strings(struct image_params *imgp)
313{
314 int argc, envc;
315 char **vectp;
316 char *stringp, *destp;
317 register_t *stack_base;
318 struct ps_strings *arginfo;
319 struct proc *p;
320
321 /*
322 * Calculate string base and vector table pointers.
323 * Also deal with signal trampoline code for this exec type.
324 */
325 p = imgp->proc;
326 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
327 destp = (caddr_t)arginfo - SPARE_USRSPACE - linux_szplatform -
328 roundup((ARG_MAX - imgp->args->stringspace), sizeof(char *));
329
330 /*
331 * install LINUX_PLATFORM
332 */
333 copyout(linux_platform, ((caddr_t)arginfo - linux_szplatform),
334 linux_szplatform);
335
336 /*
337 * If we have a valid auxargs ptr, prepare some room
338 * on the stack.
339 */
340 if (imgp->auxargs) {
341 /*
342 * 'AT_COUNT*2' is size for the ELF Auxargs data. This is for
343 * lower compatibility.
344 */
345 imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size :
346 (LINUX_AT_COUNT * 2);
347 /*
348 * The '+ 2' is for the null pointers at the end of each of
349 * the arg and env vector sets,and imgp->auxarg_size is room
350 * for argument of Runtime loader.
351 */
352 vectp = (char **)(destp - (imgp->args->argc +
353 imgp->args->envc + 2 + imgp->auxarg_size) * sizeof(char *));
354 } else {
355 /*
356 * The '+ 2' is for the null pointers at the end of each of
357 * the arg and env vector sets
358 */
359 vectp = (char **)(destp - (imgp->args->argc + imgp->args->envc + 2) *
360 sizeof(char *));
361 }
362
363 /*
364 * vectp also becomes our initial stack base
365 */
366 stack_base = (register_t *)vectp;
367
368 stringp = imgp->args->begin_argv;
369 argc = imgp->args->argc;
370 envc = imgp->args->envc;
371
372 /*
373 * Copy out strings - arguments and environment.
374 */
375 copyout(stringp, destp, ARG_MAX - imgp->args->stringspace);
376
377 /*
378 * Fill in "ps_strings" struct for ps, w, etc.
379 */
380 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
381 suword(&arginfo->ps_nargvstr, argc);
382
383 /*
384 * Fill in argument portion of vector table.
385 */
386 for (; argc > 0; --argc) {
387 suword(vectp++, (long)(intptr_t)destp);
388 while (*stringp++ != 0)
389 destp++;
390 destp++;
391 }
392
393 /* a null vector table pointer separates the argp's from the envp's */
394 suword(vectp++, 0);
395
396 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
397 suword(&arginfo->ps_nenvstr, envc);
398
399 /*
400 * Fill in environment portion of vector table.
401 */
402 for (; envc > 0; --envc) {
403 suword(vectp++, (long)(intptr_t)destp);
404 while (*stringp++ != 0)
405 destp++;
406 destp++;
407 }
408
409 /* end of vector table is a null pointer */
410 suword(vectp, 0);
411
412 return (stack_base);
413}
414
402
403
404extern unsigned long linux_sznonrtsigcode;
405
406static void
407linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
408{
409 struct thread *td = curthread;
410 struct proc *p = td->td_proc;
411 struct sigacts *psp;
412 struct trapframe *regs;
413 struct l_rt_sigframe *fp, frame;
414 int sig, code;
415 int oonstack;
416
417 sig = ksi->ksi_signo;
418 code = ksi->ksi_code;
419 PROC_LOCK_ASSERT(p, MA_OWNED);
420 psp = p->p_sigacts;
421 mtx_assert(&psp->ps_mtx, MA_OWNED);
422 regs = td->td_frame;
423 oonstack = sigonstack(regs->tf_esp);
424
425#ifdef DEBUG
426 if (ldebug(rt_sendsig))
427 printf(ARGS(rt_sendsig, "%p, %d, %p, %u"),
428 catcher, sig, (void*)mask, code);
429#endif
430 /*
431 * Allocate space for the signal handler context.
432 */
433 if ((td->td_pflags & TDP_ALTSTACK) && !oonstack &&
434 SIGISMEMBER(psp->ps_sigonstack, sig)) {
435 fp = (struct l_rt_sigframe *)(td->td_sigstk.ss_sp +
436 td->td_sigstk.ss_size - sizeof(struct l_rt_sigframe));
437 } else
438 fp = (struct l_rt_sigframe *)regs->tf_esp - 1;
439 mtx_unlock(&psp->ps_mtx);
440
441 /*
442 * Build the argument list for the signal handler.
443 */
444 if (p->p_sysent->sv_sigtbl)
445 if (sig <= p->p_sysent->sv_sigsize)
446 sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)];
447
448 bzero(&frame, sizeof(frame));
449
450 frame.sf_handler = catcher;
451 frame.sf_sig = sig;
452 frame.sf_siginfo = &fp->sf_si;
453 frame.sf_ucontext = &fp->sf_sc;
454
455 /* Fill in POSIX parts */
456 ksiginfo_to_lsiginfo(ksi, &frame.sf_si, sig);
457
458 /*
459 * Build the signal context to be used by sigreturn.
460 */
461 frame.sf_sc.uc_flags = 0; /* XXX ??? */
462 frame.sf_sc.uc_link = NULL; /* XXX ??? */
463
464 frame.sf_sc.uc_stack.ss_sp = td->td_sigstk.ss_sp;
465 frame.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size;
466 frame.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK)
467 ? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE;
468 PROC_UNLOCK(p);
469
470 bsd_to_linux_sigset(mask, &frame.sf_sc.uc_sigmask);
471
472 frame.sf_sc.uc_mcontext.sc_mask = frame.sf_sc.uc_sigmask.__bits[0];
473 frame.sf_sc.uc_mcontext.sc_gs = rgs();
474 frame.sf_sc.uc_mcontext.sc_fs = regs->tf_fs;
475 frame.sf_sc.uc_mcontext.sc_es = regs->tf_es;
476 frame.sf_sc.uc_mcontext.sc_ds = regs->tf_ds;
477 frame.sf_sc.uc_mcontext.sc_edi = regs->tf_edi;
478 frame.sf_sc.uc_mcontext.sc_esi = regs->tf_esi;
479 frame.sf_sc.uc_mcontext.sc_ebp = regs->tf_ebp;
480 frame.sf_sc.uc_mcontext.sc_ebx = regs->tf_ebx;
415static void
416linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
417{
418 struct thread *td = curthread;
419 struct proc *p = td->td_proc;
420 struct sigacts *psp;
421 struct trapframe *regs;
422 struct l_rt_sigframe *fp, frame;
423 int sig, code;
424 int oonstack;
425
426 sig = ksi->ksi_signo;
427 code = ksi->ksi_code;
428 PROC_LOCK_ASSERT(p, MA_OWNED);
429 psp = p->p_sigacts;
430 mtx_assert(&psp->ps_mtx, MA_OWNED);
431 regs = td->td_frame;
432 oonstack = sigonstack(regs->tf_esp);
433
434#ifdef DEBUG
435 if (ldebug(rt_sendsig))
436 printf(ARGS(rt_sendsig, "%p, %d, %p, %u"),
437 catcher, sig, (void*)mask, code);
438#endif
439 /*
440 * Allocate space for the signal handler context.
441 */
442 if ((td->td_pflags & TDP_ALTSTACK) && !oonstack &&
443 SIGISMEMBER(psp->ps_sigonstack, sig)) {
444 fp = (struct l_rt_sigframe *)(td->td_sigstk.ss_sp +
445 td->td_sigstk.ss_size - sizeof(struct l_rt_sigframe));
446 } else
447 fp = (struct l_rt_sigframe *)regs->tf_esp - 1;
448 mtx_unlock(&psp->ps_mtx);
449
450 /*
451 * Build the argument list for the signal handler.
452 */
453 if (p->p_sysent->sv_sigtbl)
454 if (sig <= p->p_sysent->sv_sigsize)
455 sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)];
456
457 bzero(&frame, sizeof(frame));
458
459 frame.sf_handler = catcher;
460 frame.sf_sig = sig;
461 frame.sf_siginfo = &fp->sf_si;
462 frame.sf_ucontext = &fp->sf_sc;
463
464 /* Fill in POSIX parts */
465 ksiginfo_to_lsiginfo(ksi, &frame.sf_si, sig);
466
467 /*
468 * Build the signal context to be used by sigreturn.
469 */
470 frame.sf_sc.uc_flags = 0; /* XXX ??? */
471 frame.sf_sc.uc_link = NULL; /* XXX ??? */
472
473 frame.sf_sc.uc_stack.ss_sp = td->td_sigstk.ss_sp;
474 frame.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size;
475 frame.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK)
476 ? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE;
477 PROC_UNLOCK(p);
478
479 bsd_to_linux_sigset(mask, &frame.sf_sc.uc_sigmask);
480
481 frame.sf_sc.uc_mcontext.sc_mask = frame.sf_sc.uc_sigmask.__bits[0];
482 frame.sf_sc.uc_mcontext.sc_gs = rgs();
483 frame.sf_sc.uc_mcontext.sc_fs = regs->tf_fs;
484 frame.sf_sc.uc_mcontext.sc_es = regs->tf_es;
485 frame.sf_sc.uc_mcontext.sc_ds = regs->tf_ds;
486 frame.sf_sc.uc_mcontext.sc_edi = regs->tf_edi;
487 frame.sf_sc.uc_mcontext.sc_esi = regs->tf_esi;
488 frame.sf_sc.uc_mcontext.sc_ebp = regs->tf_ebp;
489 frame.sf_sc.uc_mcontext.sc_ebx = regs->tf_ebx;
490 frame.sf_sc.uc_mcontext.sc_esp = regs->tf_esp;
481 frame.sf_sc.uc_mcontext.sc_edx = regs->tf_edx;
482 frame.sf_sc.uc_mcontext.sc_ecx = regs->tf_ecx;
483 frame.sf_sc.uc_mcontext.sc_eax = regs->tf_eax;
484 frame.sf_sc.uc_mcontext.sc_eip = regs->tf_eip;
485 frame.sf_sc.uc_mcontext.sc_cs = regs->tf_cs;
486 frame.sf_sc.uc_mcontext.sc_eflags = regs->tf_eflags;
487 frame.sf_sc.uc_mcontext.sc_esp_at_signal = regs->tf_esp;
488 frame.sf_sc.uc_mcontext.sc_ss = regs->tf_ss;
489 frame.sf_sc.uc_mcontext.sc_err = regs->tf_err;
490 frame.sf_sc.uc_mcontext.sc_cr2 = (register_t)ksi->ksi_addr;
491 frame.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code);
492
493#ifdef DEBUG
494 if (ldebug(rt_sendsig))
495 printf(LMSG("rt_sendsig flags: 0x%x, sp: %p, ss: 0x%x, mask: 0x%x"),
496 frame.sf_sc.uc_stack.ss_flags, td->td_sigstk.ss_sp,
497 td->td_sigstk.ss_size, frame.sf_sc.uc_mcontext.sc_mask);
498#endif
499
500 if (copyout(&frame, fp, sizeof(frame)) != 0) {
501 /*
502 * Process has trashed its stack; give it an illegal
503 * instruction to halt it in its tracks.
504 */
505#ifdef DEBUG
506 if (ldebug(rt_sendsig))
507 printf(LMSG("rt_sendsig: bad stack %p, oonstack=%x"),
508 fp, oonstack);
509#endif
510 PROC_LOCK(p);
511 sigexit(td, SIGILL);
512 }
513
514 /*
515 * Build context to run handler in.
516 */
517 regs->tf_esp = (int)fp;
491 frame.sf_sc.uc_mcontext.sc_edx = regs->tf_edx;
492 frame.sf_sc.uc_mcontext.sc_ecx = regs->tf_ecx;
493 frame.sf_sc.uc_mcontext.sc_eax = regs->tf_eax;
494 frame.sf_sc.uc_mcontext.sc_eip = regs->tf_eip;
495 frame.sf_sc.uc_mcontext.sc_cs = regs->tf_cs;
496 frame.sf_sc.uc_mcontext.sc_eflags = regs->tf_eflags;
497 frame.sf_sc.uc_mcontext.sc_esp_at_signal = regs->tf_esp;
498 frame.sf_sc.uc_mcontext.sc_ss = regs->tf_ss;
499 frame.sf_sc.uc_mcontext.sc_err = regs->tf_err;
500 frame.sf_sc.uc_mcontext.sc_cr2 = (register_t)ksi->ksi_addr;
501 frame.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code);
502
503#ifdef DEBUG
504 if (ldebug(rt_sendsig))
505 printf(LMSG("rt_sendsig flags: 0x%x, sp: %p, ss: 0x%x, mask: 0x%x"),
506 frame.sf_sc.uc_stack.ss_flags, td->td_sigstk.ss_sp,
507 td->td_sigstk.ss_size, frame.sf_sc.uc_mcontext.sc_mask);
508#endif
509
510 if (copyout(&frame, fp, sizeof(frame)) != 0) {
511 /*
512 * Process has trashed its stack; give it an illegal
513 * instruction to halt it in its tracks.
514 */
515#ifdef DEBUG
516 if (ldebug(rt_sendsig))
517 printf(LMSG("rt_sendsig: bad stack %p, oonstack=%x"),
518 fp, oonstack);
519#endif
520 PROC_LOCK(p);
521 sigexit(td, SIGILL);
522 }
523
524 /*
525 * Build context to run handler in.
526 */
527 regs->tf_esp = (int)fp;
518 regs->tf_eip = p->p_sysent->sv_sigcode_base + linux_sznonrtsigcode;
528 regs->tf_eip = linux_rt_sigcode;
519 regs->tf_eflags &= ~(PSL_T | PSL_VM | PSL_D);
520 regs->tf_cs = _ucodesel;
521 regs->tf_ds = _udatasel;
522 regs->tf_es = _udatasel;
523 regs->tf_fs = _udatasel;
524 regs->tf_ss = _udatasel;
525 PROC_LOCK(p);
526 mtx_lock(&psp->ps_mtx);
527}
528
529
530/*
531 * Send an interrupt to process.
532 *
533 * Stack is set up to allow sigcode stored
534 * in u. to call routine, followed by kcall
535 * to sigreturn routine below. After sigreturn
536 * resets the signal mask, the stack, and the
537 * frame pointer, it returns to the user
538 * specified pc, psl.
539 */
540static void
541linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
542{
543 struct thread *td = curthread;
544 struct proc *p = td->td_proc;
545 struct sigacts *psp;
546 struct trapframe *regs;
547 struct l_sigframe *fp, frame;
548 l_sigset_t lmask;
549 int sig, code;
550 int oonstack, i;
551
552 PROC_LOCK_ASSERT(p, MA_OWNED);
553 psp = p->p_sigacts;
554 sig = ksi->ksi_signo;
555 code = ksi->ksi_code;
556 mtx_assert(&psp->ps_mtx, MA_OWNED);
557 if (SIGISMEMBER(psp->ps_siginfo, sig)) {
558 /* Signal handler installed with SA_SIGINFO. */
559 linux_rt_sendsig(catcher, ksi, mask);
560 return;
561 }
562 regs = td->td_frame;
563 oonstack = sigonstack(regs->tf_esp);
564
565#ifdef DEBUG
566 if (ldebug(sendsig))
567 printf(ARGS(sendsig, "%p, %d, %p, %u"),
568 catcher, sig, (void*)mask, code);
569#endif
570
571 /*
572 * Allocate space for the signal handler context.
573 */
574 if ((td->td_pflags & TDP_ALTSTACK) && !oonstack &&
575 SIGISMEMBER(psp->ps_sigonstack, sig)) {
576 fp = (struct l_sigframe *)(td->td_sigstk.ss_sp +
577 td->td_sigstk.ss_size - sizeof(struct l_sigframe));
578 } else
579 fp = (struct l_sigframe *)regs->tf_esp - 1;
580 mtx_unlock(&psp->ps_mtx);
581 PROC_UNLOCK(p);
582
583 /*
584 * Build the argument list for the signal handler.
585 */
586 if (p->p_sysent->sv_sigtbl)
587 if (sig <= p->p_sysent->sv_sigsize)
588 sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)];
589
590 bzero(&frame, sizeof(frame));
591
592 frame.sf_handler = catcher;
593 frame.sf_sig = sig;
594
595 bsd_to_linux_sigset(mask, &lmask);
596
597 /*
598 * Build the signal context to be used by sigreturn.
599 */
600 frame.sf_sc.sc_mask = lmask.__bits[0];
601 frame.sf_sc.sc_gs = rgs();
602 frame.sf_sc.sc_fs = regs->tf_fs;
603 frame.sf_sc.sc_es = regs->tf_es;
604 frame.sf_sc.sc_ds = regs->tf_ds;
605 frame.sf_sc.sc_edi = regs->tf_edi;
606 frame.sf_sc.sc_esi = regs->tf_esi;
607 frame.sf_sc.sc_ebp = regs->tf_ebp;
608 frame.sf_sc.sc_ebx = regs->tf_ebx;
529 regs->tf_eflags &= ~(PSL_T | PSL_VM | PSL_D);
530 regs->tf_cs = _ucodesel;
531 regs->tf_ds = _udatasel;
532 regs->tf_es = _udatasel;
533 regs->tf_fs = _udatasel;
534 regs->tf_ss = _udatasel;
535 PROC_LOCK(p);
536 mtx_lock(&psp->ps_mtx);
537}
538
539
540/*
541 * Send an interrupt to process.
542 *
543 * Stack is set up to allow sigcode stored
544 * in u. to call routine, followed by kcall
545 * to sigreturn routine below. After sigreturn
546 * resets the signal mask, the stack, and the
547 * frame pointer, it returns to the user
548 * specified pc, psl.
549 */
550static void
551linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
552{
553 struct thread *td = curthread;
554 struct proc *p = td->td_proc;
555 struct sigacts *psp;
556 struct trapframe *regs;
557 struct l_sigframe *fp, frame;
558 l_sigset_t lmask;
559 int sig, code;
560 int oonstack, i;
561
562 PROC_LOCK_ASSERT(p, MA_OWNED);
563 psp = p->p_sigacts;
564 sig = ksi->ksi_signo;
565 code = ksi->ksi_code;
566 mtx_assert(&psp->ps_mtx, MA_OWNED);
567 if (SIGISMEMBER(psp->ps_siginfo, sig)) {
568 /* Signal handler installed with SA_SIGINFO. */
569 linux_rt_sendsig(catcher, ksi, mask);
570 return;
571 }
572 regs = td->td_frame;
573 oonstack = sigonstack(regs->tf_esp);
574
575#ifdef DEBUG
576 if (ldebug(sendsig))
577 printf(ARGS(sendsig, "%p, %d, %p, %u"),
578 catcher, sig, (void*)mask, code);
579#endif
580
581 /*
582 * Allocate space for the signal handler context.
583 */
584 if ((td->td_pflags & TDP_ALTSTACK) && !oonstack &&
585 SIGISMEMBER(psp->ps_sigonstack, sig)) {
586 fp = (struct l_sigframe *)(td->td_sigstk.ss_sp +
587 td->td_sigstk.ss_size - sizeof(struct l_sigframe));
588 } else
589 fp = (struct l_sigframe *)regs->tf_esp - 1;
590 mtx_unlock(&psp->ps_mtx);
591 PROC_UNLOCK(p);
592
593 /*
594 * Build the argument list for the signal handler.
595 */
596 if (p->p_sysent->sv_sigtbl)
597 if (sig <= p->p_sysent->sv_sigsize)
598 sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)];
599
600 bzero(&frame, sizeof(frame));
601
602 frame.sf_handler = catcher;
603 frame.sf_sig = sig;
604
605 bsd_to_linux_sigset(mask, &lmask);
606
607 /*
608 * Build the signal context to be used by sigreturn.
609 */
610 frame.sf_sc.sc_mask = lmask.__bits[0];
611 frame.sf_sc.sc_gs = rgs();
612 frame.sf_sc.sc_fs = regs->tf_fs;
613 frame.sf_sc.sc_es = regs->tf_es;
614 frame.sf_sc.sc_ds = regs->tf_ds;
615 frame.sf_sc.sc_edi = regs->tf_edi;
616 frame.sf_sc.sc_esi = regs->tf_esi;
617 frame.sf_sc.sc_ebp = regs->tf_ebp;
618 frame.sf_sc.sc_ebx = regs->tf_ebx;
619 frame.sf_sc.sc_esp = regs->tf_esp;
609 frame.sf_sc.sc_edx = regs->tf_edx;
610 frame.sf_sc.sc_ecx = regs->tf_ecx;
611 frame.sf_sc.sc_eax = regs->tf_eax;
612 frame.sf_sc.sc_eip = regs->tf_eip;
613 frame.sf_sc.sc_cs = regs->tf_cs;
614 frame.sf_sc.sc_eflags = regs->tf_eflags;
615 frame.sf_sc.sc_esp_at_signal = regs->tf_esp;
616 frame.sf_sc.sc_ss = regs->tf_ss;
617 frame.sf_sc.sc_err = regs->tf_err;
618 frame.sf_sc.sc_cr2 = (register_t)ksi->ksi_addr;
619 frame.sf_sc.sc_trapno = bsd_to_linux_trapcode(ksi->ksi_trapno);
620
621 for (i = 0; i < (LINUX_NSIG_WORDS-1); i++)
622 frame.sf_extramask[i] = lmask.__bits[i+1];
623
624 if (copyout(&frame, fp, sizeof(frame)) != 0) {
625 /*
626 * Process has trashed its stack; give it an illegal
627 * instruction to halt it in its tracks.
628 */
629 PROC_LOCK(p);
630 sigexit(td, SIGILL);
631 }
632
633 /*
634 * Build context to run handler in.
635 */
636 regs->tf_esp = (int)fp;
620 frame.sf_sc.sc_edx = regs->tf_edx;
621 frame.sf_sc.sc_ecx = regs->tf_ecx;
622 frame.sf_sc.sc_eax = regs->tf_eax;
623 frame.sf_sc.sc_eip = regs->tf_eip;
624 frame.sf_sc.sc_cs = regs->tf_cs;
625 frame.sf_sc.sc_eflags = regs->tf_eflags;
626 frame.sf_sc.sc_esp_at_signal = regs->tf_esp;
627 frame.sf_sc.sc_ss = regs->tf_ss;
628 frame.sf_sc.sc_err = regs->tf_err;
629 frame.sf_sc.sc_cr2 = (register_t)ksi->ksi_addr;
630 frame.sf_sc.sc_trapno = bsd_to_linux_trapcode(ksi->ksi_trapno);
631
632 for (i = 0; i < (LINUX_NSIG_WORDS-1); i++)
633 frame.sf_extramask[i] = lmask.__bits[i+1];
634
635 if (copyout(&frame, fp, sizeof(frame)) != 0) {
636 /*
637 * Process has trashed its stack; give it an illegal
638 * instruction to halt it in its tracks.
639 */
640 PROC_LOCK(p);
641 sigexit(td, SIGILL);
642 }
643
644 /*
645 * Build context to run handler in.
646 */
647 regs->tf_esp = (int)fp;
637 regs->tf_eip = p->p_sysent->sv_sigcode_base;
648 regs->tf_eip = linux_sigcode;
638 regs->tf_eflags &= ~(PSL_T | PSL_VM | PSL_D);
639 regs->tf_cs = _ucodesel;
640 regs->tf_ds = _udatasel;
641 regs->tf_es = _udatasel;
642 regs->tf_fs = _udatasel;
643 regs->tf_ss = _udatasel;
644 PROC_LOCK(p);
645 mtx_lock(&psp->ps_mtx);
646}
647
648/*
649 * System call to cleanup state after a signal
650 * has been taken. Reset signal mask and
651 * stack state from context left by sendsig (above).
652 * Return to previous pc and psl as specified by
653 * context left by sendsig. Check carefully to
654 * make sure that the user has not modified the
655 * psl to gain improper privileges or to cause
656 * a machine fault.
657 */
658int
659linux_sigreturn(struct thread *td, struct linux_sigreturn_args *args)
660{
661 struct l_sigframe frame;
662 struct trapframe *regs;
663 l_sigset_t lmask;
664 sigset_t bmask;
665 int eflags, i;
666 ksiginfo_t ksi;
667
668 regs = td->td_frame;
669
670#ifdef DEBUG
671 if (ldebug(sigreturn))
672 printf(ARGS(sigreturn, "%p"), (void *)args->sfp);
673#endif
674 /*
675 * The trampoline code hands us the sigframe.
676 * It is unsafe to keep track of it ourselves, in the event that a
677 * program jumps out of a signal handler.
678 */
679 if (copyin(args->sfp, &frame, sizeof(frame)) != 0)
680 return (EFAULT);
681
682 /*
683 * Check for security violations.
684 */
685#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
686 eflags = frame.sf_sc.sc_eflags;
687 if (!EFLAGS_SECURE(eflags, regs->tf_eflags))
688 return (EINVAL);
689
690 /*
691 * Don't allow users to load a valid privileged %cs. Let the
692 * hardware check for invalid selectors, excess privilege in
693 * other selectors, invalid %eip's and invalid %esp's.
694 */
695#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL)
696 if (!CS_SECURE(frame.sf_sc.sc_cs)) {
697 ksiginfo_init_trap(&ksi);
698 ksi.ksi_signo = SIGBUS;
699 ksi.ksi_code = BUS_OBJERR;
700 ksi.ksi_trapno = T_PROTFLT;
701 ksi.ksi_addr = (void *)regs->tf_eip;
702 trapsignal(td, &ksi);
703 return (EINVAL);
704 }
705
706 lmask.__bits[0] = frame.sf_sc.sc_mask;
707 for (i = 0; i < (LINUX_NSIG_WORDS-1); i++)
708 lmask.__bits[i+1] = frame.sf_extramask[i];
709 linux_to_bsd_sigset(&lmask, &bmask);
710 kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0);
711
712 /*
713 * Restore signal context.
714 */
715 /* %gs was restored by the trampoline. */
716 regs->tf_fs = frame.sf_sc.sc_fs;
717 regs->tf_es = frame.sf_sc.sc_es;
718 regs->tf_ds = frame.sf_sc.sc_ds;
719 regs->tf_edi = frame.sf_sc.sc_edi;
720 regs->tf_esi = frame.sf_sc.sc_esi;
721 regs->tf_ebp = frame.sf_sc.sc_ebp;
722 regs->tf_ebx = frame.sf_sc.sc_ebx;
723 regs->tf_edx = frame.sf_sc.sc_edx;
724 regs->tf_ecx = frame.sf_sc.sc_ecx;
725 regs->tf_eax = frame.sf_sc.sc_eax;
726 regs->tf_eip = frame.sf_sc.sc_eip;
727 regs->tf_cs = frame.sf_sc.sc_cs;
728 regs->tf_eflags = eflags;
729 regs->tf_esp = frame.sf_sc.sc_esp_at_signal;
730 regs->tf_ss = frame.sf_sc.sc_ss;
731
732 return (EJUSTRETURN);
733}
734
735/*
736 * System call to cleanup state after a signal
737 * has been taken. Reset signal mask and
738 * stack state from context left by rt_sendsig (above).
739 * Return to previous pc and psl as specified by
740 * context left by sendsig. Check carefully to
741 * make sure that the user has not modified the
742 * psl to gain improper privileges or to cause
743 * a machine fault.
744 */
745int
746linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args)
747{
748 struct l_ucontext uc;
749 struct l_sigcontext *context;
750 sigset_t bmask;
751 l_stack_t *lss;
752 stack_t ss;
753 struct trapframe *regs;
754 int eflags;
755 ksiginfo_t ksi;
756
757 regs = td->td_frame;
758
759#ifdef DEBUG
760 if (ldebug(rt_sigreturn))
761 printf(ARGS(rt_sigreturn, "%p"), (void *)args->ucp);
762#endif
763 /*
764 * The trampoline code hands us the ucontext.
765 * It is unsafe to keep track of it ourselves, in the event that a
766 * program jumps out of a signal handler.
767 */
768 if (copyin(args->ucp, &uc, sizeof(uc)) != 0)
769 return (EFAULT);
770
771 context = &uc.uc_mcontext;
772
773 /*
774 * Check for security violations.
775 */
776#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
777 eflags = context->sc_eflags;
778 if (!EFLAGS_SECURE(eflags, regs->tf_eflags))
779 return (EINVAL);
780
781 /*
782 * Don't allow users to load a valid privileged %cs. Let the
783 * hardware check for invalid selectors, excess privilege in
784 * other selectors, invalid %eip's and invalid %esp's.
785 */
786#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL)
787 if (!CS_SECURE(context->sc_cs)) {
788 ksiginfo_init_trap(&ksi);
789 ksi.ksi_signo = SIGBUS;
790 ksi.ksi_code = BUS_OBJERR;
791 ksi.ksi_trapno = T_PROTFLT;
792 ksi.ksi_addr = (void *)regs->tf_eip;
793 trapsignal(td, &ksi);
794 return (EINVAL);
795 }
796
797 linux_to_bsd_sigset(&uc.uc_sigmask, &bmask);
798 kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0);
799
800 /*
801 * Restore signal context
802 */
803 /* %gs was restored by the trampoline. */
804 regs->tf_fs = context->sc_fs;
805 regs->tf_es = context->sc_es;
806 regs->tf_ds = context->sc_ds;
807 regs->tf_edi = context->sc_edi;
808 regs->tf_esi = context->sc_esi;
809 regs->tf_ebp = context->sc_ebp;
810 regs->tf_ebx = context->sc_ebx;
811 regs->tf_edx = context->sc_edx;
812 regs->tf_ecx = context->sc_ecx;
813 regs->tf_eax = context->sc_eax;
814 regs->tf_eip = context->sc_eip;
815 regs->tf_cs = context->sc_cs;
816 regs->tf_eflags = eflags;
817 regs->tf_esp = context->sc_esp_at_signal;
818 regs->tf_ss = context->sc_ss;
819
820 /*
821 * call sigaltstack & ignore results..
822 */
823 lss = &uc.uc_stack;
824 ss.ss_sp = lss->ss_sp;
825 ss.ss_size = lss->ss_size;
826 ss.ss_flags = linux_to_bsd_sigaltstack(lss->ss_flags);
827
828#ifdef DEBUG
829 if (ldebug(rt_sigreturn))
830 printf(LMSG("rt_sigret flags: 0x%x, sp: %p, ss: 0x%x, mask: 0x%x"),
831 ss.ss_flags, ss.ss_sp, ss.ss_size, context->sc_mask);
832#endif
833 (void)kern_sigaltstack(td, &ss, NULL);
834
835 return (EJUSTRETURN);
836}
837
838static int
839linux_fetch_syscall_args(struct thread *td, struct syscall_args *sa)
840{
841 struct proc *p;
842 struct trapframe *frame;
843
844 p = td->td_proc;
845 frame = td->td_frame;
846
847 sa->code = frame->tf_eax;
848 sa->args[0] = frame->tf_ebx;
849 sa->args[1] = frame->tf_ecx;
850 sa->args[2] = frame->tf_edx;
851 sa->args[3] = frame->tf_esi;
852 sa->args[4] = frame->tf_edi;
853 sa->args[5] = frame->tf_ebp; /* Unconfirmed */
854
855 if (sa->code >= p->p_sysent->sv_size)
856 sa->callp = &p->p_sysent->sv_table[0];
857 else
858 sa->callp = &p->p_sysent->sv_table[sa->code];
859 sa->narg = sa->callp->sy_narg;
860
861 td->td_retval[0] = 0;
862 td->td_retval[1] = frame->tf_edx;
863
864 return (0);
865}
866
867/*
868 * If a linux binary is exec'ing something, try this image activator
869 * first. We override standard shell script execution in order to
870 * be able to modify the interpreter path. We only do this if a linux
871 * binary is doing the exec, so we do not create an EXEC module for it.
872 */
873static int exec_linux_imgact_try(struct image_params *iparams);
874
875static int
876exec_linux_imgact_try(struct image_params *imgp)
877{
878 const char *head = (const char *)imgp->image_header;
879 char *rpath;
880 int error = -1;
881
882 /*
883 * The interpreter for shell scripts run from a linux binary needs
884 * to be located in /compat/linux if possible in order to recursively
885 * maintain linux path emulation.
886 */
887 if (((const short *)head)[0] == SHELLMAGIC) {
888 /*
889 * Run our normal shell image activator. If it succeeds attempt
890 * to use the alternate path for the interpreter. If an alternate
891 * path is found, use our stringspace to store it.
892 */
893 if ((error = exec_shell_imgact(imgp)) == 0) {
894 linux_emul_convpath(FIRST_THREAD_IN_PROC(imgp->proc),
895 imgp->interpreter_name, UIO_SYSSPACE, &rpath, 0, AT_FDCWD);
896 if (rpath != NULL)
897 imgp->args->fname_buf =
898 imgp->interpreter_name = rpath;
899 }
900 }
901 return (error);
902}
903
904/*
905 * exec_setregs may initialize some registers differently than Linux
906 * does, thus potentially confusing Linux binaries. If necessary, we
907 * override the exec_setregs default(s) here.
908 */
909static void
910exec_linux_setregs(struct thread *td, struct image_params *imgp, u_long stack)
911{
912 struct pcb *pcb = td->td_pcb;
913
914 exec_setregs(td, imgp, stack);
915
916 /* Linux sets %gs to 0, we default to _udatasel */
917 pcb->pcb_gs = 0;
918 load_gs(0);
919
920 pcb->pcb_initial_npxcw = __LINUX_NPXCW__;
921}
922
923static void
924linux_get_machine(const char **dst)
925{
926
927 switch (cpu_class) {
928 case CPUCLASS_686:
929 *dst = "i686";
930 break;
931 case CPUCLASS_586:
932 *dst = "i586";
933 break;
934 case CPUCLASS_486:
935 *dst = "i486";
936 break;
937 default:
938 *dst = "i386";
939 }
940}
941
942struct sysentvec linux_sysvec = {
943 .sv_size = LINUX_SYS_MAXSYSCALL,
944 .sv_table = linux_sysent,
945 .sv_mask = 0,
946 .sv_sigsize = LINUX_SIGTBLSZ,
947 .sv_sigtbl = bsd_to_linux_signal,
948 .sv_errsize = ELAST + 1,
949 .sv_errtbl = bsd_to_linux_errno,
950 .sv_transtrap = translate_traps,
951 .sv_fixup = linux_fixup,
952 .sv_sendsig = linux_sendsig,
649 regs->tf_eflags &= ~(PSL_T | PSL_VM | PSL_D);
650 regs->tf_cs = _ucodesel;
651 regs->tf_ds = _udatasel;
652 regs->tf_es = _udatasel;
653 regs->tf_fs = _udatasel;
654 regs->tf_ss = _udatasel;
655 PROC_LOCK(p);
656 mtx_lock(&psp->ps_mtx);
657}
658
659/*
660 * System call to cleanup state after a signal
661 * has been taken. Reset signal mask and
662 * stack state from context left by sendsig (above).
663 * Return to previous pc and psl as specified by
664 * context left by sendsig. Check carefully to
665 * make sure that the user has not modified the
666 * psl to gain improper privileges or to cause
667 * a machine fault.
668 */
669int
670linux_sigreturn(struct thread *td, struct linux_sigreturn_args *args)
671{
672 struct l_sigframe frame;
673 struct trapframe *regs;
674 l_sigset_t lmask;
675 sigset_t bmask;
676 int eflags, i;
677 ksiginfo_t ksi;
678
679 regs = td->td_frame;
680
681#ifdef DEBUG
682 if (ldebug(sigreturn))
683 printf(ARGS(sigreturn, "%p"), (void *)args->sfp);
684#endif
685 /*
686 * The trampoline code hands us the sigframe.
687 * It is unsafe to keep track of it ourselves, in the event that a
688 * program jumps out of a signal handler.
689 */
690 if (copyin(args->sfp, &frame, sizeof(frame)) != 0)
691 return (EFAULT);
692
693 /*
694 * Check for security violations.
695 */
696#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
697 eflags = frame.sf_sc.sc_eflags;
698 if (!EFLAGS_SECURE(eflags, regs->tf_eflags))
699 return (EINVAL);
700
701 /*
702 * Don't allow users to load a valid privileged %cs. Let the
703 * hardware check for invalid selectors, excess privilege in
704 * other selectors, invalid %eip's and invalid %esp's.
705 */
706#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL)
707 if (!CS_SECURE(frame.sf_sc.sc_cs)) {
708 ksiginfo_init_trap(&ksi);
709 ksi.ksi_signo = SIGBUS;
710 ksi.ksi_code = BUS_OBJERR;
711 ksi.ksi_trapno = T_PROTFLT;
712 ksi.ksi_addr = (void *)regs->tf_eip;
713 trapsignal(td, &ksi);
714 return (EINVAL);
715 }
716
717 lmask.__bits[0] = frame.sf_sc.sc_mask;
718 for (i = 0; i < (LINUX_NSIG_WORDS-1); i++)
719 lmask.__bits[i+1] = frame.sf_extramask[i];
720 linux_to_bsd_sigset(&lmask, &bmask);
721 kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0);
722
723 /*
724 * Restore signal context.
725 */
726 /* %gs was restored by the trampoline. */
727 regs->tf_fs = frame.sf_sc.sc_fs;
728 regs->tf_es = frame.sf_sc.sc_es;
729 regs->tf_ds = frame.sf_sc.sc_ds;
730 regs->tf_edi = frame.sf_sc.sc_edi;
731 regs->tf_esi = frame.sf_sc.sc_esi;
732 regs->tf_ebp = frame.sf_sc.sc_ebp;
733 regs->tf_ebx = frame.sf_sc.sc_ebx;
734 regs->tf_edx = frame.sf_sc.sc_edx;
735 regs->tf_ecx = frame.sf_sc.sc_ecx;
736 regs->tf_eax = frame.sf_sc.sc_eax;
737 regs->tf_eip = frame.sf_sc.sc_eip;
738 regs->tf_cs = frame.sf_sc.sc_cs;
739 regs->tf_eflags = eflags;
740 regs->tf_esp = frame.sf_sc.sc_esp_at_signal;
741 regs->tf_ss = frame.sf_sc.sc_ss;
742
743 return (EJUSTRETURN);
744}
745
746/*
747 * System call to cleanup state after a signal
748 * has been taken. Reset signal mask and
749 * stack state from context left by rt_sendsig (above).
750 * Return to previous pc and psl as specified by
751 * context left by sendsig. Check carefully to
752 * make sure that the user has not modified the
753 * psl to gain improper privileges or to cause
754 * a machine fault.
755 */
756int
757linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args)
758{
759 struct l_ucontext uc;
760 struct l_sigcontext *context;
761 sigset_t bmask;
762 l_stack_t *lss;
763 stack_t ss;
764 struct trapframe *regs;
765 int eflags;
766 ksiginfo_t ksi;
767
768 regs = td->td_frame;
769
770#ifdef DEBUG
771 if (ldebug(rt_sigreturn))
772 printf(ARGS(rt_sigreturn, "%p"), (void *)args->ucp);
773#endif
774 /*
775 * The trampoline code hands us the ucontext.
776 * It is unsafe to keep track of it ourselves, in the event that a
777 * program jumps out of a signal handler.
778 */
779 if (copyin(args->ucp, &uc, sizeof(uc)) != 0)
780 return (EFAULT);
781
782 context = &uc.uc_mcontext;
783
784 /*
785 * Check for security violations.
786 */
787#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
788 eflags = context->sc_eflags;
789 if (!EFLAGS_SECURE(eflags, regs->tf_eflags))
790 return (EINVAL);
791
792 /*
793 * Don't allow users to load a valid privileged %cs. Let the
794 * hardware check for invalid selectors, excess privilege in
795 * other selectors, invalid %eip's and invalid %esp's.
796 */
797#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL)
798 if (!CS_SECURE(context->sc_cs)) {
799 ksiginfo_init_trap(&ksi);
800 ksi.ksi_signo = SIGBUS;
801 ksi.ksi_code = BUS_OBJERR;
802 ksi.ksi_trapno = T_PROTFLT;
803 ksi.ksi_addr = (void *)regs->tf_eip;
804 trapsignal(td, &ksi);
805 return (EINVAL);
806 }
807
808 linux_to_bsd_sigset(&uc.uc_sigmask, &bmask);
809 kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0);
810
811 /*
812 * Restore signal context
813 */
814 /* %gs was restored by the trampoline. */
815 regs->tf_fs = context->sc_fs;
816 regs->tf_es = context->sc_es;
817 regs->tf_ds = context->sc_ds;
818 regs->tf_edi = context->sc_edi;
819 regs->tf_esi = context->sc_esi;
820 regs->tf_ebp = context->sc_ebp;
821 regs->tf_ebx = context->sc_ebx;
822 regs->tf_edx = context->sc_edx;
823 regs->tf_ecx = context->sc_ecx;
824 regs->tf_eax = context->sc_eax;
825 regs->tf_eip = context->sc_eip;
826 regs->tf_cs = context->sc_cs;
827 regs->tf_eflags = eflags;
828 regs->tf_esp = context->sc_esp_at_signal;
829 regs->tf_ss = context->sc_ss;
830
831 /*
832 * call sigaltstack & ignore results..
833 */
834 lss = &uc.uc_stack;
835 ss.ss_sp = lss->ss_sp;
836 ss.ss_size = lss->ss_size;
837 ss.ss_flags = linux_to_bsd_sigaltstack(lss->ss_flags);
838
839#ifdef DEBUG
840 if (ldebug(rt_sigreturn))
841 printf(LMSG("rt_sigret flags: 0x%x, sp: %p, ss: 0x%x, mask: 0x%x"),
842 ss.ss_flags, ss.ss_sp, ss.ss_size, context->sc_mask);
843#endif
844 (void)kern_sigaltstack(td, &ss, NULL);
845
846 return (EJUSTRETURN);
847}
848
849static int
850linux_fetch_syscall_args(struct thread *td, struct syscall_args *sa)
851{
852 struct proc *p;
853 struct trapframe *frame;
854
855 p = td->td_proc;
856 frame = td->td_frame;
857
858 sa->code = frame->tf_eax;
859 sa->args[0] = frame->tf_ebx;
860 sa->args[1] = frame->tf_ecx;
861 sa->args[2] = frame->tf_edx;
862 sa->args[3] = frame->tf_esi;
863 sa->args[4] = frame->tf_edi;
864 sa->args[5] = frame->tf_ebp; /* Unconfirmed */
865
866 if (sa->code >= p->p_sysent->sv_size)
867 sa->callp = &p->p_sysent->sv_table[0];
868 else
869 sa->callp = &p->p_sysent->sv_table[sa->code];
870 sa->narg = sa->callp->sy_narg;
871
872 td->td_retval[0] = 0;
873 td->td_retval[1] = frame->tf_edx;
874
875 return (0);
876}
877
878/*
879 * If a linux binary is exec'ing something, try this image activator
880 * first. We override standard shell script execution in order to
881 * be able to modify the interpreter path. We only do this if a linux
882 * binary is doing the exec, so we do not create an EXEC module for it.
883 */
884static int exec_linux_imgact_try(struct image_params *iparams);
885
886static int
887exec_linux_imgact_try(struct image_params *imgp)
888{
889 const char *head = (const char *)imgp->image_header;
890 char *rpath;
891 int error = -1;
892
893 /*
894 * The interpreter for shell scripts run from a linux binary needs
895 * to be located in /compat/linux if possible in order to recursively
896 * maintain linux path emulation.
897 */
898 if (((const short *)head)[0] == SHELLMAGIC) {
899 /*
900 * Run our normal shell image activator. If it succeeds attempt
901 * to use the alternate path for the interpreter. If an alternate
902 * path is found, use our stringspace to store it.
903 */
904 if ((error = exec_shell_imgact(imgp)) == 0) {
905 linux_emul_convpath(FIRST_THREAD_IN_PROC(imgp->proc),
906 imgp->interpreter_name, UIO_SYSSPACE, &rpath, 0, AT_FDCWD);
907 if (rpath != NULL)
908 imgp->args->fname_buf =
909 imgp->interpreter_name = rpath;
910 }
911 }
912 return (error);
913}
914
915/*
916 * exec_setregs may initialize some registers differently than Linux
917 * does, thus potentially confusing Linux binaries. If necessary, we
918 * override the exec_setregs default(s) here.
919 */
920static void
921exec_linux_setregs(struct thread *td, struct image_params *imgp, u_long stack)
922{
923 struct pcb *pcb = td->td_pcb;
924
925 exec_setregs(td, imgp, stack);
926
927 /* Linux sets %gs to 0, we default to _udatasel */
928 pcb->pcb_gs = 0;
929 load_gs(0);
930
931 pcb->pcb_initial_npxcw = __LINUX_NPXCW__;
932}
933
934static void
935linux_get_machine(const char **dst)
936{
937
938 switch (cpu_class) {
939 case CPUCLASS_686:
940 *dst = "i686";
941 break;
942 case CPUCLASS_586:
943 *dst = "i586";
944 break;
945 case CPUCLASS_486:
946 *dst = "i486";
947 break;
948 default:
949 *dst = "i386";
950 }
951}
952
953struct sysentvec linux_sysvec = {
954 .sv_size = LINUX_SYS_MAXSYSCALL,
955 .sv_table = linux_sysent,
956 .sv_mask = 0,
957 .sv_sigsize = LINUX_SIGTBLSZ,
958 .sv_sigtbl = bsd_to_linux_signal,
959 .sv_errsize = ELAST + 1,
960 .sv_errtbl = bsd_to_linux_errno,
961 .sv_transtrap = translate_traps,
962 .sv_fixup = linux_fixup,
963 .sv_sendsig = linux_sendsig,
953 .sv_sigcode = linux_sigcode,
964 .sv_sigcode = &_binary_linux_locore_o_start,
954 .sv_szsigcode = &linux_szsigcode,
955 .sv_prepsyscall = NULL,
956 .sv_name = "Linux a.out",
957 .sv_coredump = NULL,
958 .sv_imgact_try = exec_linux_imgact_try,
959 .sv_minsigstksz = LINUX_MINSIGSTKSZ,
960 .sv_pagesize = PAGE_SIZE,
961 .sv_minuser = VM_MIN_ADDRESS,
962 .sv_maxuser = VM_MAXUSER_ADDRESS,
963 .sv_usrstack = LINUX_USRSTACK,
964 .sv_psstrings = PS_STRINGS,
965 .sv_stackprot = VM_PROT_ALL,
966 .sv_copyout_strings = exec_copyout_strings,
967 .sv_setregs = exec_linux_setregs,
968 .sv_fixlimit = NULL,
969 .sv_maxssiz = NULL,
970 .sv_flags = SV_ABI_LINUX | SV_AOUT | SV_IA32 | SV_ILP32,
971 .sv_set_syscall_retval = cpu_set_syscall_retval,
972 .sv_fetch_syscall_args = linux_fetch_syscall_args,
973 .sv_syscallnames = NULL,
974 .sv_shared_page_base = LINUX_SHAREDPAGE,
975 .sv_shared_page_len = PAGE_SIZE,
976 .sv_schedtail = linux_schedtail,
977 .sv_thread_detach = linux_thread_detach,
978};
979INIT_SYSENTVEC(aout_sysvec, &linux_sysvec);
980
981struct sysentvec elf_linux_sysvec = {
982 .sv_size = LINUX_SYS_MAXSYSCALL,
983 .sv_table = linux_sysent,
984 .sv_mask = 0,
985 .sv_sigsize = LINUX_SIGTBLSZ,
986 .sv_sigtbl = bsd_to_linux_signal,
987 .sv_errsize = ELAST + 1,
988 .sv_errtbl = bsd_to_linux_errno,
989 .sv_transtrap = translate_traps,
990 .sv_fixup = elf_linux_fixup,
991 .sv_sendsig = linux_sendsig,
965 .sv_szsigcode = &linux_szsigcode,
966 .sv_prepsyscall = NULL,
967 .sv_name = "Linux a.out",
968 .sv_coredump = NULL,
969 .sv_imgact_try = exec_linux_imgact_try,
970 .sv_minsigstksz = LINUX_MINSIGSTKSZ,
971 .sv_pagesize = PAGE_SIZE,
972 .sv_minuser = VM_MIN_ADDRESS,
973 .sv_maxuser = VM_MAXUSER_ADDRESS,
974 .sv_usrstack = LINUX_USRSTACK,
975 .sv_psstrings = PS_STRINGS,
976 .sv_stackprot = VM_PROT_ALL,
977 .sv_copyout_strings = exec_copyout_strings,
978 .sv_setregs = exec_linux_setregs,
979 .sv_fixlimit = NULL,
980 .sv_maxssiz = NULL,
981 .sv_flags = SV_ABI_LINUX | SV_AOUT | SV_IA32 | SV_ILP32,
982 .sv_set_syscall_retval = cpu_set_syscall_retval,
983 .sv_fetch_syscall_args = linux_fetch_syscall_args,
984 .sv_syscallnames = NULL,
985 .sv_shared_page_base = LINUX_SHAREDPAGE,
986 .sv_shared_page_len = PAGE_SIZE,
987 .sv_schedtail = linux_schedtail,
988 .sv_thread_detach = linux_thread_detach,
989};
990INIT_SYSENTVEC(aout_sysvec, &linux_sysvec);
991
992struct sysentvec elf_linux_sysvec = {
993 .sv_size = LINUX_SYS_MAXSYSCALL,
994 .sv_table = linux_sysent,
995 .sv_mask = 0,
996 .sv_sigsize = LINUX_SIGTBLSZ,
997 .sv_sigtbl = bsd_to_linux_signal,
998 .sv_errsize = ELAST + 1,
999 .sv_errtbl = bsd_to_linux_errno,
1000 .sv_transtrap = translate_traps,
1001 .sv_fixup = elf_linux_fixup,
1002 .sv_sendsig = linux_sendsig,
992 .sv_sigcode = linux_sigcode,
1003 .sv_sigcode = &_binary_linux_locore_o_start,
993 .sv_szsigcode = &linux_szsigcode,
994 .sv_prepsyscall = NULL,
995 .sv_name = "Linux ELF",
996 .sv_coredump = elf32_coredump,
997 .sv_imgact_try = exec_linux_imgact_try,
998 .sv_minsigstksz = LINUX_MINSIGSTKSZ,
999 .sv_pagesize = PAGE_SIZE,
1000 .sv_minuser = VM_MIN_ADDRESS,
1001 .sv_maxuser = VM_MAXUSER_ADDRESS,
1002 .sv_usrstack = LINUX_USRSTACK,
1003 .sv_psstrings = LINUX_PS_STRINGS,
1004 .sv_stackprot = VM_PROT_ALL,
1005 .sv_copyout_strings = linux_copyout_strings,
1006 .sv_setregs = exec_linux_setregs,
1007 .sv_fixlimit = NULL,
1008 .sv_maxssiz = NULL,
1009 .sv_flags = SV_ABI_LINUX | SV_IA32 | SV_ILP32 | SV_SHP,
1010 .sv_set_syscall_retval = cpu_set_syscall_retval,
1011 .sv_fetch_syscall_args = linux_fetch_syscall_args,
1012 .sv_syscallnames = NULL,
1013 .sv_shared_page_base = LINUX_SHAREDPAGE,
1014 .sv_shared_page_len = PAGE_SIZE,
1015 .sv_schedtail = linux_schedtail,
1016 .sv_thread_detach = linux_thread_detach,
1017};
1004 .sv_szsigcode = &linux_szsigcode,
1005 .sv_prepsyscall = NULL,
1006 .sv_name = "Linux ELF",
1007 .sv_coredump = elf32_coredump,
1008 .sv_imgact_try = exec_linux_imgact_try,
1009 .sv_minsigstksz = LINUX_MINSIGSTKSZ,
1010 .sv_pagesize = PAGE_SIZE,
1011 .sv_minuser = VM_MIN_ADDRESS,
1012 .sv_maxuser = VM_MAXUSER_ADDRESS,
1013 .sv_usrstack = LINUX_USRSTACK,
1014 .sv_psstrings = LINUX_PS_STRINGS,
1015 .sv_stackprot = VM_PROT_ALL,
1016 .sv_copyout_strings = linux_copyout_strings,
1017 .sv_setregs = exec_linux_setregs,
1018 .sv_fixlimit = NULL,
1019 .sv_maxssiz = NULL,
1020 .sv_flags = SV_ABI_LINUX | SV_IA32 | SV_ILP32 | SV_SHP,
1021 .sv_set_syscall_retval = cpu_set_syscall_retval,
1022 .sv_fetch_syscall_args = linux_fetch_syscall_args,
1023 .sv_syscallnames = NULL,
1024 .sv_shared_page_base = LINUX_SHAREDPAGE,
1025 .sv_shared_page_len = PAGE_SIZE,
1026 .sv_schedtail = linux_schedtail,
1027 .sv_thread_detach = linux_thread_detach,
1028};
1018INIT_SYSENTVEC(elf_sysvec, &elf_linux_sysvec);
1019
1029
1030static void
1031linux_vdso_install(void *param)
1032{
1033
1034 linux_szsigcode = (&_binary_linux_locore_o_end -
1035 &_binary_linux_locore_o_start);
1036
1037 if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len)
1038 panic("Linux invalid vdso size\n");
1039
1040 __elfN(linux_vdso_fixup)(&elf_linux_sysvec);
1041
1042 linux_shared_page_obj = __elfN(linux_shared_page_init)
1043 (&linux_shared_page_mapping);
1044
1045 __elfN(linux_vdso_reloc)(&elf_linux_sysvec, LINUX_SHAREDPAGE);
1046
1047 bcopy(elf_linux_sysvec.sv_sigcode, linux_shared_page_mapping,
1048 linux_szsigcode);
1049 elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj;
1050}
1051SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY,
1052 (sysinit_cfunc_t)linux_vdso_install, NULL);
1053
1054static void
1055linux_vdso_deinstall(void *param)
1056{
1057
1058 __elfN(linux_shared_page_fini)(linux_shared_page_obj);
1059};
1060SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST,
1061 (sysinit_cfunc_t)linux_vdso_deinstall, NULL);
1062
1020static char GNU_ABI_VENDOR[] = "GNU";
1021static int GNULINUX_ABI_DESC = 0;
1022
1023static boolean_t
1024linux_trans_osrel(const Elf_Note *note, int32_t *osrel)
1025{
1026 const Elf32_Word *desc;
1027 uintptr_t p;
1028
1029 p = (uintptr_t)(note + 1);
1030 p += roundup2(note->n_namesz, sizeof(Elf32_Addr));
1031
1032 desc = (const Elf32_Word *)p;
1033 if (desc[0] != GNULINUX_ABI_DESC)
1034 return (FALSE);
1035
1036 /*
1037 * For linux we encode osrel as follows (see linux_mib.c):
1038 * VVVMMMIII (version, major, minor), see linux_mib.c.
1039 */
1040 *osrel = desc[1] * 1000000 + desc[2] * 1000 + desc[3];
1041
1042 return (TRUE);
1043}
1044
1045static Elf_Brandnote linux_brandnote = {
1046 .hdr.n_namesz = sizeof(GNU_ABI_VENDOR),
1047 .hdr.n_descsz = 16, /* XXX at least 16 */
1048 .hdr.n_type = 1,
1049 .vendor = GNU_ABI_VENDOR,
1050 .flags = BN_TRANSLATE_OSREL,
1051 .trans_osrel = linux_trans_osrel
1052};
1053
1054static Elf32_Brandinfo linux_brand = {
1055 .brand = ELFOSABI_LINUX,
1056 .machine = EM_386,
1057 .compat_3_brand = "Linux",
1058 .emul_path = "/compat/linux",
1059 .interp_path = "/lib/ld-linux.so.1",
1060 .sysvec = &elf_linux_sysvec,
1061 .interp_newpath = NULL,
1062 .brand_note = &linux_brandnote,
1063 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
1064};
1065
1066static Elf32_Brandinfo linux_glibc2brand = {
1067 .brand = ELFOSABI_LINUX,
1068 .machine = EM_386,
1069 .compat_3_brand = "Linux",
1070 .emul_path = "/compat/linux",
1071 .interp_path = "/lib/ld-linux.so.2",
1072 .sysvec = &elf_linux_sysvec,
1073 .interp_newpath = NULL,
1074 .brand_note = &linux_brandnote,
1075 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
1076};
1077
1078Elf32_Brandinfo *linux_brandlist[] = {
1079 &linux_brand,
1080 &linux_glibc2brand,
1081 NULL
1082};
1083
1084static int
1085linux_elf_modevent(module_t mod, int type, void *data)
1086{
1087 Elf32_Brandinfo **brandinfo;
1088 int error;
1089 struct linux_ioctl_handler **lihp;
1090 struct linux_device_handler **ldhp;
1091
1092 error = 0;
1093
1094 switch(type) {
1095 case MOD_LOAD:
1096 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
1097 ++brandinfo)
1098 if (elf32_insert_brand_entry(*brandinfo) < 0)
1099 error = EINVAL;
1100 if (error == 0) {
1101 SET_FOREACH(lihp, linux_ioctl_handler_set)
1102 linux_ioctl_register_handler(*lihp);
1103 SET_FOREACH(ldhp, linux_device_handler_set)
1104 linux_device_register_handler(*ldhp);
1105 LIST_INIT(&futex_list);
1106 mtx_init(&futex_mtx, "ftllk", NULL, MTX_DEF);
1107 linux_exit_tag = EVENTHANDLER_REGISTER(process_exit, linux_proc_exit,
1108 NULL, 1000);
1109 linux_exec_tag = EVENTHANDLER_REGISTER(process_exec, linux_proc_exec,
1110 NULL, 1000);
1111 linux_thread_dtor_tag = EVENTHANDLER_REGISTER(thread_dtor,
1112 linux_thread_dtor, NULL, EVENTHANDLER_PRI_ANY);
1113 linux_get_machine(&linux_platform);
1114 linux_szplatform = roundup(strlen(linux_platform) + 1,
1115 sizeof(char *));
1116 linux_osd_jail_register();
1117 stclohz = (stathz ? stathz : hz);
1118 if (bootverbose)
1119 printf("Linux ELF exec handler installed\n");
1120 } else
1121 printf("cannot insert Linux ELF brand handler\n");
1122 break;
1123 case MOD_UNLOAD:
1124 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
1125 ++brandinfo)
1126 if (elf32_brand_inuse(*brandinfo))
1127 error = EBUSY;
1128 if (error == 0) {
1129 for (brandinfo = &linux_brandlist[0];
1130 *brandinfo != NULL; ++brandinfo)
1131 if (elf32_remove_brand_entry(*brandinfo) < 0)
1132 error = EINVAL;
1133 }
1134 if (error == 0) {
1135 SET_FOREACH(lihp, linux_ioctl_handler_set)
1136 linux_ioctl_unregister_handler(*lihp);
1137 SET_FOREACH(ldhp, linux_device_handler_set)
1138 linux_device_unregister_handler(*ldhp);
1139 mtx_destroy(&futex_mtx);
1140 EVENTHANDLER_DEREGISTER(process_exit, linux_exit_tag);
1141 EVENTHANDLER_DEREGISTER(process_exec, linux_exec_tag);
1142 EVENTHANDLER_DEREGISTER(thread_dtor, linux_thread_dtor_tag);
1143 linux_osd_jail_deregister();
1144 if (bootverbose)
1145 printf("Linux ELF exec handler removed\n");
1146 } else
1147 printf("Could not deinstall ELF interpreter entry\n");
1148 break;
1149 default:
1150 return (EOPNOTSUPP);
1151 }
1152 return (error);
1153}
1154
1155static moduledata_t linux_elf_mod = {
1156 "linuxelf",
1157 linux_elf_modevent,
1158 0
1159};
1160
1161DECLARE_MODULE_TIED(linuxelf, linux_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY);
1063static char GNU_ABI_VENDOR[] = "GNU";
1064static int GNULINUX_ABI_DESC = 0;
1065
1066static boolean_t
1067linux_trans_osrel(const Elf_Note *note, int32_t *osrel)
1068{
1069 const Elf32_Word *desc;
1070 uintptr_t p;
1071
1072 p = (uintptr_t)(note + 1);
1073 p += roundup2(note->n_namesz, sizeof(Elf32_Addr));
1074
1075 desc = (const Elf32_Word *)p;
1076 if (desc[0] != GNULINUX_ABI_DESC)
1077 return (FALSE);
1078
1079 /*
1080 * For linux we encode osrel as follows (see linux_mib.c):
1081 * VVVMMMIII (version, major, minor), see linux_mib.c.
1082 */
1083 *osrel = desc[1] * 1000000 + desc[2] * 1000 + desc[3];
1084
1085 return (TRUE);
1086}
1087
1088static Elf_Brandnote linux_brandnote = {
1089 .hdr.n_namesz = sizeof(GNU_ABI_VENDOR),
1090 .hdr.n_descsz = 16, /* XXX at least 16 */
1091 .hdr.n_type = 1,
1092 .vendor = GNU_ABI_VENDOR,
1093 .flags = BN_TRANSLATE_OSREL,
1094 .trans_osrel = linux_trans_osrel
1095};
1096
1097static Elf32_Brandinfo linux_brand = {
1098 .brand = ELFOSABI_LINUX,
1099 .machine = EM_386,
1100 .compat_3_brand = "Linux",
1101 .emul_path = "/compat/linux",
1102 .interp_path = "/lib/ld-linux.so.1",
1103 .sysvec = &elf_linux_sysvec,
1104 .interp_newpath = NULL,
1105 .brand_note = &linux_brandnote,
1106 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
1107};
1108
1109static Elf32_Brandinfo linux_glibc2brand = {
1110 .brand = ELFOSABI_LINUX,
1111 .machine = EM_386,
1112 .compat_3_brand = "Linux",
1113 .emul_path = "/compat/linux",
1114 .interp_path = "/lib/ld-linux.so.2",
1115 .sysvec = &elf_linux_sysvec,
1116 .interp_newpath = NULL,
1117 .brand_note = &linux_brandnote,
1118 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
1119};
1120
1121Elf32_Brandinfo *linux_brandlist[] = {
1122 &linux_brand,
1123 &linux_glibc2brand,
1124 NULL
1125};
1126
1127static int
1128linux_elf_modevent(module_t mod, int type, void *data)
1129{
1130 Elf32_Brandinfo **brandinfo;
1131 int error;
1132 struct linux_ioctl_handler **lihp;
1133 struct linux_device_handler **ldhp;
1134
1135 error = 0;
1136
1137 switch(type) {
1138 case MOD_LOAD:
1139 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
1140 ++brandinfo)
1141 if (elf32_insert_brand_entry(*brandinfo) < 0)
1142 error = EINVAL;
1143 if (error == 0) {
1144 SET_FOREACH(lihp, linux_ioctl_handler_set)
1145 linux_ioctl_register_handler(*lihp);
1146 SET_FOREACH(ldhp, linux_device_handler_set)
1147 linux_device_register_handler(*ldhp);
1148 LIST_INIT(&futex_list);
1149 mtx_init(&futex_mtx, "ftllk", NULL, MTX_DEF);
1150 linux_exit_tag = EVENTHANDLER_REGISTER(process_exit, linux_proc_exit,
1151 NULL, 1000);
1152 linux_exec_tag = EVENTHANDLER_REGISTER(process_exec, linux_proc_exec,
1153 NULL, 1000);
1154 linux_thread_dtor_tag = EVENTHANDLER_REGISTER(thread_dtor,
1155 linux_thread_dtor, NULL, EVENTHANDLER_PRI_ANY);
1156 linux_get_machine(&linux_platform);
1157 linux_szplatform = roundup(strlen(linux_platform) + 1,
1158 sizeof(char *));
1159 linux_osd_jail_register();
1160 stclohz = (stathz ? stathz : hz);
1161 if (bootverbose)
1162 printf("Linux ELF exec handler installed\n");
1163 } else
1164 printf("cannot insert Linux ELF brand handler\n");
1165 break;
1166 case MOD_UNLOAD:
1167 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
1168 ++brandinfo)
1169 if (elf32_brand_inuse(*brandinfo))
1170 error = EBUSY;
1171 if (error == 0) {
1172 for (brandinfo = &linux_brandlist[0];
1173 *brandinfo != NULL; ++brandinfo)
1174 if (elf32_remove_brand_entry(*brandinfo) < 0)
1175 error = EINVAL;
1176 }
1177 if (error == 0) {
1178 SET_FOREACH(lihp, linux_ioctl_handler_set)
1179 linux_ioctl_unregister_handler(*lihp);
1180 SET_FOREACH(ldhp, linux_device_handler_set)
1181 linux_device_unregister_handler(*ldhp);
1182 mtx_destroy(&futex_mtx);
1183 EVENTHANDLER_DEREGISTER(process_exit, linux_exit_tag);
1184 EVENTHANDLER_DEREGISTER(process_exec, linux_exec_tag);
1185 EVENTHANDLER_DEREGISTER(thread_dtor, linux_thread_dtor_tag);
1186 linux_osd_jail_deregister();
1187 if (bootverbose)
1188 printf("Linux ELF exec handler removed\n");
1189 } else
1190 printf("Could not deinstall ELF interpreter entry\n");
1191 break;
1192 default:
1193 return (EOPNOTSUPP);
1194 }
1195 return (error);
1196}
1197
1198static moduledata_t linux_elf_mod = {
1199 "linuxelf",
1200 linux_elf_modevent,
1201 0
1202};
1203
1204DECLARE_MODULE_TIED(linuxelf, linux_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY);