xref: /openbmc/qemu/linux-user/i386/cpu_loop.c (revision 646b5378)
1 /*
2  *  qemu user cpu loop
3  *
4  *  Copyright (c) 2003-2008 Fabrice Bellard
5  *
6  *  This program is free software; you can redistribute it and/or modify
7  *  it under the terms of the GNU General Public License as published by
8  *  the Free Software Foundation; either version 2 of the License, or
9  *  (at your option) any later version.
10  *
11  *  This program is distributed in the hope that it will be useful,
12  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
13  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14  *  GNU General Public License for more details.
15  *
16  *  You should have received a copy of the GNU General Public License
17  *  along with this program; if not, see <http://www.gnu.org/licenses/>.
18  */
19 
20 #include "qemu/osdep.h"
21 #include "qemu.h"
22 #include "qemu/timer.h"
23 #include "user-internals.h"
24 #include "cpu_loop-common.h"
25 #include "signal-common.h"
26 #include "user-mmap.h"
27 
28 /***********************************************************/
29 /* CPUX86 core interface */
30 
31 uint64_t cpu_get_tsc(CPUX86State *env)
32 {
33     return cpu_get_host_ticks();
34 }
35 
36 static void write_dt(void *ptr, unsigned long addr, unsigned long limit,
37               int flags)
38 {
39     unsigned int e1, e2;
40     uint32_t *p;
41     e1 = (addr << 16) | (limit & 0xffff);
42     e2 = ((addr >> 16) & 0xff) | (addr & 0xff000000) | (limit & 0x000f0000);
43     e2 |= flags;
44     p = ptr;
45     p[0] = tswap32(e1);
46     p[1] = tswap32(e2);
47 }
48 
49 static uint64_t *idt_table;
50 
51 static void set_gate64(void *ptr, unsigned int type, unsigned int dpl,
52                        uint64_t addr, unsigned int sel)
53 {
54     uint32_t *p, e1, e2;
55     e1 = (addr & 0xffff) | (sel << 16);
56     e2 = (addr & 0xffff0000) | 0x8000 | (dpl << 13) | (type << 8);
57     p = ptr;
58     p[0] = tswap32(e1);
59     p[1] = tswap32(e2);
60     p[2] = tswap32(addr >> 32);
61     p[3] = 0;
62 }
63 
64 #ifdef TARGET_X86_64
65 /* only dpl matters as we do only user space emulation */
66 static void set_idt(int n, unsigned int dpl, bool is64)
67 {
68     set_gate64(idt_table + n * 2, 0, dpl, 0, 0);
69 }
70 #else
71 static void set_gate(void *ptr, unsigned int type, unsigned int dpl,
72                      uint32_t addr, unsigned int sel)
73 {
74     uint32_t *p, e1, e2;
75     e1 = (addr & 0xffff) | (sel << 16);
76     e2 = (addr & 0xffff0000) | 0x8000 | (dpl << 13) | (type << 8);
77     p = ptr;
78     p[0] = tswap32(e1);
79     p[1] = tswap32(e2);
80 }
81 
82 /* only dpl matters as we do only user space emulation */
83 static void set_idt(int n, unsigned int dpl, bool is64)
84 {
85     if (is64) {
86         set_gate64(idt_table + n * 2, 0, dpl, 0, 0);
87     } else {
88         set_gate(idt_table + n, 0, dpl, 0, 0);
89     }
90 }
91 #endif
92 
93 #ifdef TARGET_X86_64
94 static bool write_ok_or_segv(CPUX86State *env, abi_ptr addr, size_t len)
95 {
96     /*
97      * For all the vsyscalls, NULL means "don't write anything" not
98      * "write it at address 0".
99      */
100     if (addr == 0 || access_ok(env_cpu(env), VERIFY_WRITE, addr, len)) {
101         return true;
102     }
103 
104     env->error_code = PG_ERROR_W_MASK | PG_ERROR_U_MASK;
105     force_sig_fault(TARGET_SIGSEGV, TARGET_SEGV_MAPERR, addr);
106     return false;
107 }
108 
109 /*
110  * Since v3.1, the kernel traps and emulates the vsyscall page.
111  * Entry points other than the official generate SIGSEGV.
112  */
113 static void emulate_vsyscall(CPUX86State *env)
114 {
115     int syscall;
116     abi_ulong ret;
117     uint64_t caller;
118 
119     /*
120      * Validate the entry point.  We have already validated the page
121      * during translation to get here; now verify the offset.
122      */
123     switch (env->eip & ~TARGET_PAGE_MASK) {
124     case 0x000:
125         syscall = TARGET_NR_gettimeofday;
126         break;
127     case 0x400:
128         syscall = TARGET_NR_time;
129         break;
130     case 0x800:
131         syscall = TARGET_NR_getcpu;
132         break;
133     default:
134         goto sigsegv;
135     }
136 
137     /*
138      * Validate the return address.
139      * Note that the kernel treats this the same as an invalid entry point.
140      */
141     if (get_user_u64(caller, env->regs[R_ESP])) {
142         goto sigsegv;
143     }
144 
145     /*
146      * Validate the pointer arguments.
147      */
148     switch (syscall) {
149     case TARGET_NR_gettimeofday:
150         if (!write_ok_or_segv(env, env->regs[R_EDI],
151                               sizeof(struct target_timeval)) ||
152             !write_ok_or_segv(env, env->regs[R_ESI],
153                               sizeof(struct target_timezone))) {
154             return;
155         }
156         break;
157     case TARGET_NR_time:
158         if (!write_ok_or_segv(env, env->regs[R_EDI], sizeof(abi_long))) {
159             return;
160         }
161         break;
162     case TARGET_NR_getcpu:
163         if (!write_ok_or_segv(env, env->regs[R_EDI], sizeof(uint32_t)) ||
164             !write_ok_or_segv(env, env->regs[R_ESI], sizeof(uint32_t))) {
165             return;
166         }
167         break;
168     default:
169         g_assert_not_reached();
170     }
171 
172     /*
173      * Perform the syscall.  None of the vsyscalls should need restarting.
174      */
175     get_task_state(env_cpu(env))->orig_ax = syscall;
176     ret = do_syscall(env, syscall, env->regs[R_EDI], env->regs[R_ESI],
177                      env->regs[R_EDX], env->regs[10], env->regs[8],
178                      env->regs[9], 0, 0);
179     g_assert(ret != -QEMU_ERESTARTSYS);
180     g_assert(ret != -QEMU_ESIGRETURN);
181     if (ret == -TARGET_EFAULT) {
182         goto sigsegv;
183     }
184     env->regs[R_EAX] = ret;
185 
186     /* Emulate a ret instruction to leave the vsyscall page.  */
187     env->eip = caller;
188     env->regs[R_ESP] += 8;
189     return;
190 
191  sigsegv:
192     force_sig(TARGET_SIGSEGV);
193 }
194 #endif
195 
196 static bool maybe_handle_vm86_trap(CPUX86State *env, int trapnr)
197 {
198 #ifndef TARGET_X86_64
199     if (env->eflags & VM_MASK) {
200         handle_vm86_trap(env, trapnr);
201         return true;
202     }
203 #endif
204     return false;
205 }
206 
207 void cpu_loop(CPUX86State *env)
208 {
209     CPUState *cs = env_cpu(env);
210     int trapnr;
211     abi_ulong ret;
212 
213     for(;;) {
214         cpu_exec_start(cs);
215         trapnr = cpu_exec(cs);
216         cpu_exec_end(cs);
217         process_queued_cpu_work(cs);
218 
219         switch(trapnr) {
220         case 0x80:
221 #ifndef TARGET_X86_64
222         case EXCP_SYSCALL:
223 #endif
224             /* linux syscall from int $0x80 */
225             get_task_state(cs)->orig_ax = env->regs[R_EAX];
226             ret = do_syscall(env,
227                              env->regs[R_EAX],
228                              env->regs[R_EBX],
229                              env->regs[R_ECX],
230                              env->regs[R_EDX],
231                              env->regs[R_ESI],
232                              env->regs[R_EDI],
233                              env->regs[R_EBP],
234                              0, 0);
235             if (ret == -QEMU_ERESTARTSYS) {
236                 env->eip -= 2;
237             } else if (ret != -QEMU_ESIGRETURN) {
238                 env->regs[R_EAX] = ret;
239             }
240             break;
241 #ifdef TARGET_X86_64
242         case EXCP_SYSCALL:
243             /* linux syscall from syscall instruction.  */
244             get_task_state(cs)->orig_ax = env->regs[R_EAX];
245             ret = do_syscall(env,
246                              env->regs[R_EAX],
247                              env->regs[R_EDI],
248                              env->regs[R_ESI],
249                              env->regs[R_EDX],
250                              env->regs[10],
251                              env->regs[8],
252                              env->regs[9],
253                              0, 0);
254             if (ret == -QEMU_ERESTARTSYS) {
255                 env->eip -= 2;
256             } else if (ret != -QEMU_ESIGRETURN) {
257                 env->regs[R_EAX] = ret;
258             }
259             break;
260         case EXCP_VSYSCALL:
261             emulate_vsyscall(env);
262             break;
263 #endif
264         case EXCP0B_NOSEG:
265         case EXCP0C_STACK:
266             force_sig(TARGET_SIGBUS);
267             break;
268         case EXCP0D_GPF:
269             /* XXX: potential problem if ABI32 */
270             if (maybe_handle_vm86_trap(env, trapnr)) {
271                 break;
272             }
273             force_sig(TARGET_SIGSEGV);
274             break;
275         case EXCP0E_PAGE:
276             force_sig_fault(TARGET_SIGSEGV,
277                             (env->error_code & PG_ERROR_P_MASK ?
278                              TARGET_SEGV_ACCERR : TARGET_SEGV_MAPERR),
279                             env->cr[2]);
280             break;
281         case EXCP00_DIVZ:
282             if (maybe_handle_vm86_trap(env, trapnr)) {
283                 break;
284             }
285             force_sig_fault(TARGET_SIGFPE, TARGET_FPE_INTDIV, env->eip);
286             break;
287         case EXCP01_DB:
288             if (maybe_handle_vm86_trap(env, trapnr)) {
289                 break;
290             }
291             force_sig_fault(TARGET_SIGTRAP, TARGET_TRAP_BRKPT, env->eip);
292             break;
293         case EXCP03_INT3:
294             if (maybe_handle_vm86_trap(env, trapnr)) {
295                 break;
296             }
297             force_sig(TARGET_SIGTRAP);
298             break;
299         case EXCP04_INTO:
300         case EXCP05_BOUND:
301             if (maybe_handle_vm86_trap(env, trapnr)) {
302                 break;
303             }
304             force_sig(TARGET_SIGSEGV);
305             break;
306         case EXCP06_ILLOP:
307             force_sig_fault(TARGET_SIGILL, TARGET_ILL_ILLOPN, env->eip);
308             break;
309         case EXCP_INTERRUPT:
310             /* just indicate that signals should be handled asap */
311             break;
312         case EXCP_DEBUG:
313             force_sig_fault(TARGET_SIGTRAP, TARGET_TRAP_BRKPT, env->eip);
314             break;
315         case EXCP_ATOMIC:
316             cpu_exec_step_atomic(cs);
317             break;
318         default:
319             EXCP_DUMP(env, "qemu: unhandled CPU exception 0x%x - aborting\n",
320                       trapnr);
321             abort();
322         }
323         process_pending_signals(env);
324     }
325 }
326 
327 static void target_cpu_free(void *obj)
328 {
329     target_munmap(cpu_env(obj)->gdt.base,
330                   sizeof(uint64_t) * TARGET_GDT_ENTRIES);
331     g_free(obj);
332 }
333 
334 void target_cpu_copy_regs(CPUArchState *env, struct target_pt_regs *regs)
335 {
336     CPUState *cpu = env_cpu(env);
337     bool is64 = (env->features[FEAT_8000_0001_EDX] & CPUID_EXT2_LM) != 0;
338     int i;
339 
340     OBJECT(cpu)->free = target_cpu_free;
341     env->cr[0] = CR0_PG_MASK | CR0_WP_MASK | CR0_PE_MASK;
342     env->hflags |= HF_PE_MASK | HF_CPL_MASK;
343     if (env->features[FEAT_1_EDX] & CPUID_SSE) {
344         env->cr[4] |= CR4_OSFXSR_MASK;
345         env->hflags |= HF_OSFXSR_MASK;
346     }
347 
348     /* enable 64 bit mode if possible */
349     if (is64) {
350         env->cr[4] |= CR4_PAE_MASK;
351         env->efer |= MSR_EFER_LMA | MSR_EFER_LME;
352         env->hflags |= HF_LMA_MASK;
353     }
354 #ifndef TARGET_ABI32
355     else {
356         fprintf(stderr, "The selected x86 CPU does not support 64 bit mode\n");
357         exit(EXIT_FAILURE);
358     }
359 #endif
360 
361     /* flags setup : we activate the IRQs by default as in user mode */
362     env->eflags |= IF_MASK;
363 
364     /* linux register setup */
365 #ifndef TARGET_ABI32
366     env->regs[R_EAX] = regs->rax;
367     env->regs[R_EBX] = regs->rbx;
368     env->regs[R_ECX] = regs->rcx;
369     env->regs[R_EDX] = regs->rdx;
370     env->regs[R_ESI] = regs->rsi;
371     env->regs[R_EDI] = regs->rdi;
372     env->regs[R_EBP] = regs->rbp;
373     env->regs[R_ESP] = regs->rsp;
374     env->eip = regs->rip;
375 #else
376     env->regs[R_EAX] = regs->eax;
377     env->regs[R_EBX] = regs->ebx;
378     env->regs[R_ECX] = regs->ecx;
379     env->regs[R_EDX] = regs->edx;
380     env->regs[R_ESI] = regs->esi;
381     env->regs[R_EDI] = regs->edi;
382     env->regs[R_EBP] = regs->ebp;
383     env->regs[R_ESP] = regs->esp;
384     env->eip = regs->eip;
385 #endif
386 
387     /* linux interrupt setup */
388 #ifndef TARGET_ABI32
389     env->idt.limit = 511;
390 #else
391     env->idt.limit = 255;
392 #endif
393     env->idt.base = target_mmap(0, sizeof(uint64_t) * (env->idt.limit + 1),
394                                 PROT_READ|PROT_WRITE,
395                                 MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
396     idt_table = g2h_untagged(env->idt.base);
397     for (i = 0; i < 20; i++) {
398         set_idt(i, 0, is64);
399     }
400     set_idt(3, 3, is64);
401     set_idt(4, 3, is64);
402     set_idt(0x80, 3, is64);
403 
404     /* linux segment setup */
405     {
406         uint64_t *gdt_table;
407         env->gdt.base = target_mmap(0, sizeof(uint64_t) * TARGET_GDT_ENTRIES,
408                                     PROT_READ|PROT_WRITE,
409                                     MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
410         env->gdt.limit = sizeof(uint64_t) * TARGET_GDT_ENTRIES - 1;
411         gdt_table = g2h_untagged(env->gdt.base);
412 #ifdef TARGET_ABI32
413         write_dt(&gdt_table[__USER_CS >> 3], 0, 0xfffff,
414                  DESC_G_MASK | DESC_B_MASK | DESC_P_MASK | DESC_S_MASK |
415                  (3 << DESC_DPL_SHIFT) | (0xa << DESC_TYPE_SHIFT));
416 #else
417         /* 64 bit code segment */
418         write_dt(&gdt_table[__USER_CS >> 3], 0, 0xfffff,
419                  DESC_G_MASK | DESC_B_MASK | DESC_P_MASK | DESC_S_MASK |
420                  DESC_L_MASK |
421                  (3 << DESC_DPL_SHIFT) | (0xa << DESC_TYPE_SHIFT));
422 #endif
423         write_dt(&gdt_table[__USER_DS >> 3], 0, 0xfffff,
424                  DESC_G_MASK | DESC_B_MASK | DESC_P_MASK | DESC_S_MASK |
425                  (3 << DESC_DPL_SHIFT) | (0x2 << DESC_TYPE_SHIFT));
426     }
427     cpu_x86_load_seg(env, R_CS, __USER_CS);
428     cpu_x86_load_seg(env, R_SS, __USER_DS);
429 #ifdef TARGET_ABI32
430     cpu_x86_load_seg(env, R_DS, __USER_DS);
431     cpu_x86_load_seg(env, R_ES, __USER_DS);
432     cpu_x86_load_seg(env, R_FS, __USER_DS);
433     cpu_x86_load_seg(env, R_GS, __USER_DS);
434     /* This hack makes Wine work... */
435     env->segs[R_FS].selector = 0;
436 #else
437     cpu_x86_load_seg(env, R_DS, 0);
438     cpu_x86_load_seg(env, R_ES, 0);
439     cpu_x86_load_seg(env, R_FS, 0);
440     cpu_x86_load_seg(env, R_GS, 0);
441 #endif
442 }
443