Expand PMF_FN_* macros.
[netbsd-mini2440.git] / sys / compat / linux / arch / i386 / linux_exec_machdep.c
blobf512b80fb0f1ed07dce56b0f89ee65a066feac05
1 /* $NetBSD: linux_exec_machdep.c,v 1.13 2009/09/20 10:29:30 taca Exp $ */
3 /*-
4 * Copyright (c) 2004 The NetBSD Foundation, Inc.
5 * All rights reserved.
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Christos Zoulas.
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.
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.
32 #include <sys/cdefs.h>
33 __KERNEL_RCSID(0, "$NetBSD: linux_exec_machdep.c,v 1.13 2009/09/20 10:29:30 taca Exp $");
35 #if defined(_KERNEL_OPT)
36 #include "opt_vm86.h"
37 #include "opt_user_ldt.h"
38 #endif
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>
50 #include <sys/cpu.h>
51 #include <machine/vmparam.h>
53 #include <uvm/uvm.h>
55 #include <sys/syscallargs.h>
57 #ifndef DEBUG_LINUX
58 #define DPRINTF(a)
59 #else
60 #define DPRINTF(a) uprintf a
61 #endif
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>
74 int
75 linux_exec_setup_stack(struct lwp *l, struct exec_package *epp)
77 u_long max_stack_size;
78 u_long access_linear_min, access_size;
79 u_long noaccess_linear_min, noaccess_size;
81 #ifndef USRSTACK32
82 #define USRSTACK32 (0x00000000ffffffffL&~PGOFSET)
83 #endif
85 if (epp->ep_flags & EXEC_32) {
86 epp->ep_minsaddr = USRSTACK32;
87 max_stack_size = MAXSSIZ;
88 } else {
89 epp->ep_minsaddr = USRSTACK;
90 max_stack_size = MAXSSIZ;
93 if (epp->ep_minsaddr > LINUX_USRSTACK)
94 epp->ep_minsaddr = LINUX_USRSTACK;
95 #ifdef DEBUG_LINUX
96 else {
98 * Someone needs to make KERNBASE and TEXTADDR
99 * java versions < 1.4.2 need the stack to be
100 * at 0xC0000000
102 uprintf("Cannot setup stack to 0xC0000000, "
103 "java will not work properly\n");
105 #endif
106 epp->ep_maxsaddr = (u_long)STACK_GROW(epp->ep_minsaddr,
107 max_stack_size);
108 epp->ep_ssize = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
111 * set up commands for stack. note that this takes *two*, one to
112 * map the part of the stack which we can access, and one to map
113 * the part which we can't.
115 * arguably, it could be made into one, but that would require the
116 * addition of another mapping proc, which is unnecessary
118 access_size = epp->ep_ssize;
119 access_linear_min = (u_long)STACK_ALLOC(epp->ep_minsaddr, access_size);
120 noaccess_size = max_stack_size - access_size;
121 noaccess_linear_min = (u_long)STACK_ALLOC(STACK_GROW(epp->ep_minsaddr,
122 access_size), noaccess_size);
123 if (noaccess_size > 0) {
124 NEW_VMCMD2(&epp->ep_vmcmds, vmcmd_map_zero, noaccess_size,
125 noaccess_linear_min, NULLVP, 0, VM_PROT_NONE, VMCMD_STACK);
127 KASSERT(access_size > 0);
128 NEW_VMCMD2(&epp->ep_vmcmds, vmcmd_map_zero, access_size,
129 access_linear_min, NULLVP, 0, VM_PROT_READ | VM_PROT_WRITE,
130 VMCMD_STACK);
132 return 0;
136 #ifdef LINUX_NPTL
137 static __inline void
138 load_gs(u_int sel)
140 __asm __volatile("movw %0,%%gs" : : "rm" ((unsigned short)sel));
145 linux_init_thread_area(struct lwp *l, struct lwp *l2)
147 struct trapframe *tf = l->l_md.md_regs, *tf2 = l2->l_md.md_regs;
148 struct pcb *pcb2 = lwp_getpcb(l2);
149 struct linux_user_desc info;
150 struct segment_descriptor sd;
151 int error, idx, a[2];
153 error = copyin((void *)tf->tf_esi, &info, sizeof(info));
154 if (error)
155 return error;
156 idx = info.entry_number;
159 * looks like we're getting the idx we returned
160 * in the set_thread_area() syscall
162 if (idx != LINUX_GLIBC_TLS_SEL && idx != GUGS_SEL) {
163 DPRINTF(("resetting idx %d to GUGS_SEL", idx));
164 idx = GUGS_SEL;
167 /* this doesnt happen in practice */
168 if (idx == LINUX_GLIBC_TLS_SEL) {
169 /* we might copy out the entry_number as 3 */
170 info.entry_number = GUGS_SEL;
171 error = copyout(&info, (void *)tf->tf_esi, sizeof(info));
172 if (error)
173 return error;
176 a[0] = LINUX_LDT_entry_a(&info);
177 a[1] = LINUX_LDT_entry_b(&info);
179 (void)memcpy(&sd, &a, sizeof(a));
180 KASSERT(ISMEMSDP((&sd)));
181 DPRINTF(("Segment created in clone with CLONE_SETTLS: lobase: %x, "
182 "hibase: %x, lolimit: %x, hilimit: %x, type: %i, dpl: %i, p: %i, "
183 "xx: %i, def32: %i, gran: %i\n", sd.sd_lobase,
184 sd.sd_hibase, sd.sd_lolimit, sd.sd_hilimit, sd.sd_type, sd.sd_dpl,
185 sd.sd_p, sd.sd_xx, sd.sd_def32, sd.sd_gran));
187 (void)memcpy(&pcb2->pcb_gsd, &sd, sizeof(sd));
188 tf2->tf_gs = GSEL(GUGS_SEL, SEL_UPL);
190 return 0;
195 linux_sys_set_thread_area(struct lwp *l,
196 const struct linux_sys_set_thread_area_args *uap, register_t *retval)
198 struct pcb *pcb = lwp_getpcb(l);
199 struct linux_user_desc info;
200 struct segment_descriptor sd;
201 int error, idx, a[2];
203 *retval = 0;
204 error = copyin(SCARG(uap, desc), &info, sizeof(info));
205 if (error)
206 return error;
208 DPRINTF(("set thread area: %i, %x, %x, %i, %i, %i, %i, %i, %i\n",
209 info.entry_number, info.base_addr, info.limit, info.seg_32bit,
210 info.contents, info.read_exec_only, info.limit_in_pages,
211 info.seg_not_present, info.useable));
213 idx = info.entry_number;
215 * Semantics of linux version: every thread in the system has array of
216 * 3 tls descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown. This
217 * syscall loads one of the selected tls decriptors with a value and
218 * also loads GDT descriptors 6, 7 and 8 with the content of the
219 * per-thread descriptors.
221 * Semantics of fbsd version: I think we can ignore that linux has 3
222 * per-thread descriptors and use just the 1st one. The tls_array[]
223 * is used only in set/get-thread_area() syscalls and for loading the
224 * GDT descriptors. In fbsd we use just one GDT descriptor for TLS so
225 * we will load just one.
227 * XXX: this doesn't work when a user space process tries to use more
228 * than 1 TLS segment. Comment in the linux sources says wine might do
229 * this.
233 * we support just GLIBC TLS now
234 * we should let 3 proceed as well because we use this segment so
235 * if code does two subsequent calls it should succeed
237 if (idx != LINUX_GLIBC_TLS_SEL && idx != -1 && idx != GUGS_SEL)
238 return EINVAL;
241 * we have to copy out the GDT entry we use
242 * FreeBSD uses GDT entry #3 for storing %gs so load that
244 * XXX: what if a user space program doesn't check this value and tries
245 * to use 6, 7 or 8?
247 idx = info.entry_number = GUGS_SEL;
248 error = copyout(&info, SCARG(uap, desc), sizeof(info));
249 if (error)
250 return error;
252 if (LINUX_LDT_empty(&info)) {
253 a[0] = 0;
254 a[1] = 0;
255 } else {
256 a[0] = LINUX_LDT_entry_a(&info);
257 a[1] = LINUX_LDT_entry_b(&info);
260 (void)memcpy(&sd, &a, sizeof(a));
261 KASSERT(ISMEMSDP((&sd)));
262 DPRINTF(("Segment created in set_thread_area: lobase: %x, hibase: %x, "
263 "lolimit: %x, hilimit: %x, type: %i, dpl: %i, p: %i, xx: %i, "
264 "def32: %i, gran: %i\n", sd.sd_lobase, sd.sd_hibase, sd.sd_lolimit,
265 sd.sd_hilimit, sd.sd_type, sd.sd_dpl, sd.sd_p, sd.sd_xx,
266 sd.sd_def32, sd.sd_gran));
268 kpreempt_disable();
269 (void)memcpy(&pcb->pcb_gsd, &sd, sizeof(sd));
270 (void)memcpy(&curcpu()->ci_gdt[GUGS_SEL], &sd, sizeof(sd));
271 load_gs(GSEL(GUGS_SEL, SEL_UPL));
272 kpreempt_enable();
273 return 0;
277 linux_sys_get_thread_area(struct lwp *l,
278 const struct linux_sys_get_thread_area_args *uap, register_t *retval)
280 struct pcb *pcb = lwp_getpcb(l);
281 struct linux_user_desc info;
282 struct linux_desc_struct desc;
283 struct segment_descriptor sd;
284 int error, idx;
286 *retval = 0;
287 error = copyin(SCARG(uap, desc), &info, sizeof(info));
288 if (error)
289 return error;
291 idx = info.entry_number;
292 /* XXX: I am not sure if we want 3 to be allowed too. */
293 if (idx != LINUX_GLIBC_TLS_SEL && idx != GUGS_SEL)
294 return EINVAL;
296 idx = GUGS_SEL;
298 (void)memset(&info, 0, sizeof(info));
299 (void)memcpy(&sd, pcb->pcb_gsd, sizeof(sd));
300 (void)memcpy(&desc, &sd, sizeof(desc));
302 info.entry_number = idx;
303 info.base_addr = LINUX_GET_BASE(&desc);
304 info.limit = LINUX_GET_LIMIT(&desc);
305 info.seg_32bit = LINUX_GET_32BIT(&desc);
306 info.contents = LINUX_GET_CONTENTS(&desc);
307 info.read_exec_only = !LINUX_GET_WRITABLE(&desc);
308 info.limit_in_pages = LINUX_GET_LIMIT_PAGES(&desc);
309 info.seg_not_present = !LINUX_GET_PRESENT(&desc);
310 info.useable = LINUX_GET_USEABLE(&desc);
312 return copyout(&info, SCARG(uap, desc), sizeof(info));
315 #endif