1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2015 Thomas Meyer (thomas@m3y3r.de)
4 * Copyright (C) 2002- 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
14 #include <asm/unistd.h>
15 #include <as-layout.h>
17 #include <kern_util.h>
20 #include <ptrace_user.h>
21 #include <registers.h>
23 #include <sysdep/stub.h>
24 #include <linux/threads.h>
26 int is_skas_winch(int pid, int fd, void *data)
28 return pid == getpgrp();
31 static const char *ptrace_reg_name(int idx)
33 #define R(n) case HOST_##n: return #n
56 #elif defined(__i386__)
79 static int ptrace_dump_regs(int pid)
81 unsigned long regs[MAX_REG_NR];
84 if (ptrace(PTRACE_GETREGS, pid, 0, regs) < 0)
87 printk(UM_KERN_ERR "Stub registers -\n");
88 for (i = 0; i < ARRAY_SIZE(regs); i++) {
89 const char *regname = ptrace_reg_name(i);
91 printk(UM_KERN_ERR "\t%s\t(%2d): %lx\n", regname, i, regs[i]);
98 * Signals that are OK to receive in the stub - we'll just continue it.
99 * SIGWINCH will happen when UML is inside a detached screen.
101 #define STUB_SIG_MASK ((1 << SIGALRM) | (1 << SIGWINCH))
103 /* Signals that the stub will finish with - anything else is an error */
104 #define STUB_DONE_MASK (1 << SIGTRAP)
106 void wait_stub_done(int pid)
111 CATCH_EINTR(n = waitpid(pid, &status, WUNTRACED | __WALL));
112 if ((n < 0) || !WIFSTOPPED(status))
115 if (((1 << WSTOPSIG(status)) & STUB_SIG_MASK) == 0)
118 err = ptrace(PTRACE_CONT, pid, 0, 0);
120 printk(UM_KERN_ERR "wait_stub_done : continue failed, "
121 "errno = %d\n", errno);
126 if (((1 << WSTOPSIG(status)) & STUB_DONE_MASK) != 0)
130 err = ptrace_dump_regs(pid);
132 printk(UM_KERN_ERR "Failed to get registers from stub, "
133 "errno = %d\n", -err);
134 printk(UM_KERN_ERR "wait_stub_done : failed to wait for SIGTRAP, "
135 "pid = %d, n = %d, errno = %d, status = 0x%x\n", pid, n, errno,
140 extern unsigned long current_stub_stack(void);
142 static void get_skas_faultinfo(int pid, struct faultinfo *fi, unsigned long *aux_fp_regs)
146 err = get_fp_registers(pid, aux_fp_regs);
148 printk(UM_KERN_ERR "save_fp_registers returned %d\n",
152 err = ptrace(PTRACE_CONT, pid, 0, SIGSEGV);
154 printk(UM_KERN_ERR "Failed to continue stub, pid = %d, "
155 "errno = %d\n", pid, errno);
161 * faultinfo is prepared by the stub_segv_handler at start of
162 * the stub stack page. We just have to copy it.
164 memcpy(fi, (void *)current_stub_stack(), sizeof(*fi));
166 err = put_fp_registers(pid, aux_fp_regs);
168 printk(UM_KERN_ERR "put_fp_registers returned %d\n",
174 static void handle_segv(int pid, struct uml_pt_regs *regs, unsigned long *aux_fp_regs)
176 get_skas_faultinfo(pid, ®s->faultinfo, aux_fp_regs);
177 segv(regs->faultinfo, 0, 1, NULL);
181 * To use the same value of using_sysemu as the caller, ask it that value
182 * (in local_using_sysemu
184 static void handle_trap(int pid, struct uml_pt_regs *regs,
185 int local_using_sysemu)
189 if ((UPT_IP(regs) >= STUB_START) && (UPT_IP(regs) < STUB_END))
192 if (!local_using_sysemu)
194 err = ptrace(PTRACE_POKEUSER, pid, PT_SYSCALL_NR_OFFSET,
197 printk(UM_KERN_ERR "handle_trap - nullifying syscall "
198 "failed, errno = %d\n", errno);
202 err = ptrace(PTRACE_SYSCALL, pid, 0, 0);
204 printk(UM_KERN_ERR "handle_trap - continuing to end of "
205 "syscall failed, errno = %d\n", errno);
209 CATCH_EINTR(err = waitpid(pid, &status, WUNTRACED | __WALL));
210 if ((err < 0) || !WIFSTOPPED(status) ||
211 (WSTOPSIG(status) != SIGTRAP + 0x80)) {
212 err = ptrace_dump_regs(pid);
214 printk(UM_KERN_ERR "Failed to get registers "
215 "from process, errno = %d\n", -err);
216 printk(UM_KERN_ERR "handle_trap - failed to wait at "
217 "end of syscall, errno = %d, status = %d\n",
223 handle_syscall(regs);
226 extern char __syscall_stub_start[];
229 * userspace_tramp() - userspace trampoline
230 * @stack: pointer to the new userspace stack page, can be NULL, if? FIXME:
232 * The userspace trampoline is used to setup a new userspace process in start_userspace() after it was clone()'ed.
233 * This function will run on a temporary stack page.
234 * It ptrace()'es itself, then
235 * Two pages are mapped into the userspace address space:
236 * - STUB_CODE (with EXEC), which contains the skas stub code
237 * - STUB_DATA (with R/W), which contains a data page that is used to transfer certain data between the UML userspace process and the UML kernel.
238 * Also for the userspace process a SIGSEGV handler is installed to catch pagefaults in the userspace process.
239 * And last the process stops itself to give control to the UML kernel for this userspace process.
241 * Return: Always zero, otherwise the current userspace process is ended with non null exit() call
243 static int userspace_tramp(void *stack)
247 unsigned long long offset;
249 ptrace(PTRACE_TRACEME, 0, 0, 0);
251 signal(SIGTERM, SIG_DFL);
252 signal(SIGWINCH, SIG_IGN);
254 fd = phys_mapping(to_phys(__syscall_stub_start), &offset);
255 addr = mmap64((void *) STUB_CODE, UM_KERN_PAGE_SIZE,
256 PROT_EXEC, MAP_FIXED | MAP_PRIVATE, fd, offset);
257 if (addr == MAP_FAILED) {
258 printk(UM_KERN_ERR "mapping mmap stub at 0x%lx failed, "
259 "errno = %d\n", STUB_CODE, errno);
264 fd = phys_mapping(to_phys(stack), &offset);
265 addr = mmap((void *) STUB_DATA,
266 UM_KERN_PAGE_SIZE, PROT_READ | PROT_WRITE,
267 MAP_FIXED | MAP_SHARED, fd, offset);
268 if (addr == MAP_FAILED) {
269 printk(UM_KERN_ERR "mapping segfault stack "
270 "at 0x%lx failed, errno = %d\n",
278 unsigned long v = STUB_CODE +
279 (unsigned long) stub_segv_handler -
280 (unsigned long) __syscall_stub_start;
282 set_sigstack((void *) STUB_DATA, UM_KERN_PAGE_SIZE);
283 sigemptyset(&sa.sa_mask);
284 sa.sa_flags = SA_ONSTACK | SA_NODEFER | SA_SIGINFO;
285 sa.sa_sigaction = (void *) v;
286 sa.sa_restorer = NULL;
287 if (sigaction(SIGSEGV, &sa, NULL) < 0) {
288 printk(UM_KERN_ERR "userspace_tramp - setting SIGSEGV "
289 "handler failed - errno = %d\n", errno);
294 kill(os_getpid(), SIGSTOP);
298 int userspace_pid[NR_CPUS];
299 int kill_userspace_mm[NR_CPUS];
302 * start_userspace() - prepare a new userspace process
303 * @stub_stack: pointer to the stub stack. Can be NULL, if? FIXME:
305 * Setups a new temporary stack page that is used while userspace_tramp() runs
306 * Clones the kernel process into a new userspace process, with FDs only.
308 * Return: When positive: the process id of the new userspace process,
309 * when negative: an error number.
310 * FIXME: can PIDs become negative?!
312 int start_userspace(unsigned long stub_stack)
316 int pid, status, n, flags, err;
318 /* setup a temporary stack page */
319 stack = mmap(NULL, UM_KERN_PAGE_SIZE,
320 PROT_READ | PROT_WRITE | PROT_EXEC,
321 MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
322 if (stack == MAP_FAILED) {
324 printk(UM_KERN_ERR "start_userspace : mmap failed, "
325 "errno = %d\n", errno);
329 /* set stack pointer to the end of the stack page, so it can grow downwards */
330 sp = (unsigned long)stack + UM_KERN_PAGE_SIZE;
332 flags = CLONE_FILES | SIGCHLD;
334 /* clone into new userspace process */
335 pid = clone(userspace_tramp, (void *) sp, flags, (void *) stub_stack);
338 printk(UM_KERN_ERR "start_userspace : clone failed, "
339 "errno = %d\n", errno);
344 CATCH_EINTR(n = waitpid(pid, &status, WUNTRACED | __WALL));
347 printk(UM_KERN_ERR "start_userspace : wait failed, "
348 "errno = %d\n", errno);
351 } while (WIFSTOPPED(status) && (WSTOPSIG(status) == SIGALRM));
353 if (!WIFSTOPPED(status) || (WSTOPSIG(status) != SIGSTOP)) {
355 printk(UM_KERN_ERR "start_userspace : expected SIGSTOP, got "
356 "status = %d\n", status);
360 if (ptrace(PTRACE_OLDSETOPTIONS, pid, NULL,
361 (void *) PTRACE_O_TRACESYSGOOD) < 0) {
363 printk(UM_KERN_ERR "start_userspace : PTRACE_OLDSETOPTIONS "
364 "failed, errno = %d\n", errno);
368 if (munmap(stack, UM_KERN_PAGE_SIZE) < 0) {
370 printk(UM_KERN_ERR "start_userspace : munmap failed, "
371 "errno = %d\n", errno);
378 os_kill_ptraced_process(pid, 1);
382 void userspace(struct uml_pt_regs *regs, unsigned long *aux_fp_regs)
384 int err, status, op, pid = userspace_pid[0];
385 /* To prevent races if using_sysemu changes under us.*/
386 int local_using_sysemu;
389 /* Handle any immediate reschedules or signals */
393 if (kill_userspace_mm[0])
397 * This can legitimately fail if the process loads a
398 * bogus value into a segment register. It will
399 * segfault and PTRACE_GETREGS will read that value
400 * out of the process. However, PTRACE_SETREGS will
401 * fail. In this case, there is nothing to do but
402 * just kill the process.
404 if (ptrace(PTRACE_SETREGS, pid, 0, regs->gp)) {
405 printk(UM_KERN_ERR "userspace - ptrace set regs "
406 "failed, errno = %d\n", errno);
410 if (put_fp_registers(pid, regs->fp)) {
411 printk(UM_KERN_ERR "userspace - ptrace set fp regs "
412 "failed, errno = %d\n", errno);
416 /* Now we set local_using_sysemu to be used for one loop */
417 local_using_sysemu = get_using_sysemu();
419 op = SELECT_PTRACE_OPERATION(local_using_sysemu,
420 singlestepping(NULL));
422 if (ptrace(op, pid, 0, 0)) {
423 printk(UM_KERN_ERR "userspace - ptrace continue "
424 "failed, op = %d, errno = %d\n", op, errno);
428 CATCH_EINTR(err = waitpid(pid, &status, WUNTRACED | __WALL));
430 printk(UM_KERN_ERR "userspace - wait failed, "
431 "errno = %d\n", errno);
436 if (ptrace(PTRACE_GETREGS, pid, 0, regs->gp)) {
437 printk(UM_KERN_ERR "userspace - PTRACE_GETREGS failed, "
438 "errno = %d\n", errno);
442 if (get_fp_registers(pid, regs->fp)) {
443 printk(UM_KERN_ERR "userspace - get_fp_registers failed, "
444 "errno = %d\n", errno);
448 UPT_SYSCALL_NR(regs) = -1; /* Assume: It's not a syscall */
450 if (WIFSTOPPED(status)) {
451 int sig = WSTOPSIG(status);
453 /* These signal handlers need the si argument.
454 * The SIGIO and SIGALARM handlers which constitute the
455 * majority of invocations, do not use it.
464 ptrace(PTRACE_GETSIGINFO, pid, 0, (struct siginfo *)&si);
470 if (PTRACE_FULL_FAULTINFO) {
471 get_skas_faultinfo(pid,
472 ®s->faultinfo, aux_fp_regs);
473 (*sig_info[SIGSEGV])(SIGSEGV, (struct siginfo *)&si,
476 else handle_segv(pid, regs, aux_fp_regs);
479 handle_trap(pid, regs, local_using_sysemu);
482 relay_signal(SIGTRAP, (struct siginfo *)&si, regs);
491 block_signals_trace();
492 (*sig_info[sig])(sig, (struct siginfo *)&si, regs);
493 unblock_signals_trace();
496 printk(UM_KERN_ERR "userspace - child stopped "
497 "with signal %d\n", sig);
500 pid = userspace_pid[0];
503 /* Avoid -ERESTARTSYS handling in host */
504 if (PT_SYSCALL_NR_OFFSET != PT_SYSCALL_RET_OFFSET)
505 PT_SYSCALL_NR(regs->gp) = -1;
510 static unsigned long thread_regs[MAX_REG_NR];
511 static unsigned long thread_fp_regs[FP_SIZE];
513 static int __init init_thread_regs(void)
515 get_safe_registers(thread_regs, thread_fp_regs);
516 /* Set parent's instruction pointer to start of clone-stub */
517 thread_regs[REGS_IP_INDEX] = STUB_CODE +
518 (unsigned long) stub_clone_handler -
519 (unsigned long) __syscall_stub_start;
520 thread_regs[REGS_SP_INDEX] = STUB_DATA + UM_KERN_PAGE_SIZE -
522 #ifdef __SIGNAL_FRAMESIZE
523 thread_regs[REGS_SP_INDEX] -= __SIGNAL_FRAMESIZE;
528 __initcall(init_thread_regs);
530 int copy_context_skas0(unsigned long new_stack, int pid)
533 unsigned long current_stack = current_stub_stack();
534 struct stub_data *data = (struct stub_data *) current_stack;
535 struct stub_data *child_data = (struct stub_data *) new_stack;
536 unsigned long long new_offset;
537 int new_fd = phys_mapping(to_phys((void *)new_stack), &new_offset);
540 * prepare offset and fd of child's stack as argument for parent's
541 * and child's mmap2 calls
543 *data = ((struct stub_data) {
544 .offset = MMAP_OFFSET(new_offset),
546 .parent_err = -ESRCH,
550 *child_data = ((struct stub_data) {
554 err = ptrace_setregs(pid, thread_regs);
557 printk(UM_KERN_ERR "copy_context_skas0 : PTRACE_SETREGS "
558 "failed, pid = %d, errno = %d\n", pid, -err);
562 err = put_fp_registers(pid, thread_fp_regs);
564 printk(UM_KERN_ERR "copy_context_skas0 : put_fp_registers "
565 "failed, pid = %d, err = %d\n", pid, err);
570 * Wait, until parent has finished its work: read child's pid from
571 * parent's stack, and check, if bad result.
573 err = ptrace(PTRACE_CONT, pid, 0, 0);
576 printk(UM_KERN_ERR "Failed to continue new process, pid = %d, "
577 "errno = %d\n", pid, errno);
583 pid = data->parent_err;
585 printk(UM_KERN_ERR "copy_context_skas0 - stub-parent reports "
591 * Wait, until child has finished too: read child's result from
592 * child's stack and check it.
595 if (child_data->child_err != STUB_DATA) {
596 printk(UM_KERN_ERR "copy_context_skas0 - stub-child %d reports "
597 "error %ld\n", pid, data->child_err);
598 err = data->child_err;
602 if (ptrace(PTRACE_OLDSETOPTIONS, pid, NULL,
603 (void *)PTRACE_O_TRACESYSGOOD) < 0) {
605 printk(UM_KERN_ERR "copy_context_skas0 : PTRACE_OLDSETOPTIONS "
606 "failed, errno = %d\n", errno);
613 os_kill_ptraced_process(pid, 1);
617 void new_thread(void *stack, jmp_buf *buf, void (*handler)(void))
619 (*buf)[0].JB_IP = (unsigned long) handler;
620 (*buf)[0].JB_SP = (unsigned long) stack + UM_THREAD_SIZE -
624 #define INIT_JMP_NEW_THREAD 0
625 #define INIT_JMP_CALLBACK 1
626 #define INIT_JMP_HALT 2
627 #define INIT_JMP_REBOOT 3
629 void switch_threads(jmp_buf *me, jmp_buf *you)
631 if (UML_SETJMP(me) == 0)
635 static jmp_buf initial_jmpbuf;
637 /* XXX Make these percpu */
638 static void (*cb_proc)(void *arg);
640 static jmp_buf *cb_back;
642 int start_idle_thread(void *stack, jmp_buf *switch_buf)
646 set_handler(SIGWINCH);
649 * Can't use UML_SETJMP or UML_LONGJMP here because they save
650 * and restore signals, with the possible side-effect of
651 * trying to handle any signals which came when they were
652 * blocked, which can't be done on this stack.
653 * Signals must be blocked when jumping back here and restored
654 * after returning to the jumper.
656 n = setjmp(initial_jmpbuf);
658 case INIT_JMP_NEW_THREAD:
659 (*switch_buf)[0].JB_IP = (unsigned long) uml_finishsetup;
660 (*switch_buf)[0].JB_SP = (unsigned long) stack +
661 UM_THREAD_SIZE - sizeof(void *);
663 case INIT_JMP_CALLBACK:
665 longjmp(*cb_back, 1);
670 case INIT_JMP_REBOOT:
674 printk(UM_KERN_ERR "Bad sigsetjmp return in "
675 "start_idle_thread - %d\n", n);
678 longjmp(*switch_buf, 1);
681 printk(UM_KERN_ERR "impossible long jump!");
686 void initial_thread_cb_skas(void (*proc)(void *), void *arg)
694 block_signals_trace();
695 if (UML_SETJMP(&here) == 0)
696 UML_LONGJMP(&initial_jmpbuf, INIT_JMP_CALLBACK);
697 unblock_signals_trace();
706 block_signals_trace();
707 UML_LONGJMP(&initial_jmpbuf, INIT_JMP_HALT);
710 void reboot_skas(void)
712 block_signals_trace();
713 UML_LONGJMP(&initial_jmpbuf, INIT_JMP_REBOOT);
716 void __switch_mm(struct mm_id *mm_idp)
718 userspace_pid[0] = mm_idp->u.pid;
719 kill_userspace_mm[0] = mm_idp->kill;