[deliverable/linux.git] / arch / um / os-Linux / skas / process.c

/*
 * Copyright (C) 2002- 2004 Jeff Dike (jdike@addtoit.com)
 * Licensed under the GPL
 */

#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <errno.h>
#include <signal.h>
#include <sched.h>
#include "ptrace_user.h"
#include <sys/wait.h>
#include <sys/mman.h>
#include <sys/user.h>
#include <sys/time.h>
#include <sys/syscall.h>
#include <asm/types.h>
#include "user.h"
#include "sysdep/ptrace.h"
#include "kern_util.h"
#include "skas.h"
#include "stub-data.h"
#include "mm_id.h"
#include "sysdep/sigcontext.h"
#include "sysdep/stub.h"
#include "os.h"
#include "proc_mm.h"
#include "skas_ptrace.h"
#include "chan_user.h"
#include "registers.h"
#include "mem.h"
#include "uml-config.h"
#include "process.h"
#include "longjmp.h"
#include "kern_constants.h"
#include "as-layout.h"

int is_skas_winch(int pid, int fd, void *data)
{
	if(pid != os_getpgrp())
		return(0);

	register_winch_irq(-1, fd, -1, data);
	return(1);
}

static int ptrace_dump_regs(int pid)
{
        unsigned long regs[MAX_REG_NR];
        int i;

        if(ptrace(PTRACE_GETREGS, pid, 0, regs) < 0)
                return -errno;
        else {
                printk("Stub registers -\n");
                for(i = 0; i < ARRAY_SIZE(regs); i++)
                        printk("\t%d - %lx\n", i, regs[i]);
        }

        return 0;
}

/*
 * Signals that are OK to receive in the stub - we'll just continue it.
 * SIGWINCH will happen when UML is inside a detached screen.
 */
#define STUB_SIG_MASK ((1 << SIGVTALRM) | (1 << SIGWINCH))

/* Signals that the stub will finish with - anything else is an error */
#define STUB_DONE_MASK ((1 << SIGUSR1) | (1 << SIGTRAP))

void wait_stub_done(int pid)
{
	int n, status, err;

	while(1){
		CATCH_EINTR(n = waitpid(pid, &status, WUNTRACED));
		if((n < 0) || !WIFSTOPPED(status))
			goto bad_wait;

		if(((1 << WSTOPSIG(status)) & STUB_SIG_MASK) == 0)
			break;

		err = ptrace(PTRACE_CONT, pid, 0, 0);
		if(err)
			panic("wait_stub_done : continue failed, errno = %d\n",
			      errno);
	}

	if(((1 << WSTOPSIG(status)) & STUB_DONE_MASK) != 0)
		return;

bad_wait:
	err = ptrace_dump_regs(pid);
	if(err)
		printk("Failed to get registers from stub, errno = %d\n", -err);
	panic("wait_stub_done : failed to wait for SIGUSR1/SIGTRAP, pid = %d, "
	      "n = %d, errno = %d, status = 0x%x\n", pid, n, errno, status);
}

extern unsigned long current_stub_stack(void);

void get_skas_faultinfo(int pid, struct faultinfo * fi)
{
	int err;

	if(ptrace_faultinfo){
		err = ptrace(PTRACE_FAULTINFO, pid, 0, fi);
		if(err)
			panic("get_skas_faultinfo - PTRACE_FAULTINFO failed, "
			      "errno = %d\n", errno);

		/* Special handling for i386, which has different structs */
		if (sizeof(struct ptrace_faultinfo) < sizeof(struct faultinfo))
			memset((char *)fi + sizeof(struct ptrace_faultinfo), 0,
			       sizeof(struct faultinfo) -
			       sizeof(struct ptrace_faultinfo));
	}
	else {
		err = ptrace(PTRACE_CONT, pid, 0, SIGSEGV);
		if(err)
			panic("Failed to continue stub, pid = %d, errno = %d\n",
			      pid, errno);
		wait_stub_done(pid);

		/* faultinfo is prepared by the stub-segv-handler at start of
		 * the stub stack page. We just have to copy it.
		 */
		memcpy(fi, (void *)current_stub_stack(), sizeof(*fi));
	}
}

static void handle_segv(int pid, union uml_pt_regs * regs)
{
	get_skas_faultinfo(pid, &regs->skas.faultinfo);
	segv(regs->skas.faultinfo, 0, 1, NULL);
}

/*To use the same value of using_sysemu as the caller, ask it that value (in local_using_sysemu)*/
static void handle_trap(int pid, union uml_pt_regs *regs, int local_using_sysemu)
{
	int err, status;

	/* Mark this as a syscall */
	UPT_SYSCALL_NR(regs) = PT_SYSCALL_NR(regs->skas.regs);

	if (!local_using_sysemu)
	{
		err = ptrace(PTRACE_POKEUSR, pid, PT_SYSCALL_NR_OFFSET,
			     __NR_getpid);
		if(err < 0)
			panic("handle_trap - nullifying syscall failed errno = %d\n",
			      errno);

		err = ptrace(PTRACE_SYSCALL, pid, 0, 0);
		if(err < 0)
			panic("handle_trap - continuing to end of syscall failed, "
			      "errno = %d\n", errno);

		CATCH_EINTR(err = waitpid(pid, &status, WUNTRACED));
		if((err < 0) || !WIFSTOPPED(status) ||
		   (WSTOPSIG(status) != SIGTRAP + 0x80)){
                        err = ptrace_dump_regs(pid);
                        if(err)
                                printk("Failed to get registers from process, "
                                       "errno = %d\n", -err);
			panic("handle_trap - failed to wait at end of syscall, "
			      "errno = %d, status = %d\n", errno, status);
                }
	}

	handle_syscall(regs);
}

extern int __syscall_stub_start;

static int userspace_tramp(void *stack)
{
	void *addr;
	int err;

	ptrace(PTRACE_TRACEME, 0, 0, 0);

	init_new_thread_signals();
	err = set_interval(1);
	if(err)
		panic("userspace_tramp - setting timer failed, errno = %d\n",
		      err);

	if(!proc_mm){
		/* This has a pte, but it can't be mapped in with the usual
		 * tlb_flush mechanism because this is part of that mechanism
		 */
		int fd;
		__u64 offset;
		fd = phys_mapping(to_phys(&__syscall_stub_start), &offset);
		addr = mmap64((void *) UML_CONFIG_STUB_CODE, UM_KERN_PAGE_SIZE,
			      PROT_EXEC, MAP_FIXED | MAP_PRIVATE, fd, offset);
		if(addr == MAP_FAILED){
			printk("mapping mmap stub failed, errno = %d\n",
			       errno);
			exit(1);
		}

		if(stack != NULL){
			fd = phys_mapping(to_phys(stack), &offset);
			addr = mmap((void *) UML_CONFIG_STUB_DATA,
				    UM_KERN_PAGE_SIZE, PROT_READ | PROT_WRITE,
				    MAP_FIXED | MAP_SHARED, fd, offset);
			if(addr == MAP_FAILED){
				printk("mapping segfault stack failed, "
				       "errno = %d\n", errno);
				exit(1);
			}
		}
	}
	if(!ptrace_faultinfo && (stack != NULL)){
		struct sigaction sa;

		unsigned long v = UML_CONFIG_STUB_CODE +
				  (unsigned long) stub_segv_handler -
				  (unsigned long) &__syscall_stub_start;

		set_sigstack((void *) UML_CONFIG_STUB_DATA, UM_KERN_PAGE_SIZE);
		sigemptyset(&sa.sa_mask);
		sigaddset(&sa.sa_mask, SIGIO);
		sigaddset(&sa.sa_mask, SIGWINCH);
		sigaddset(&sa.sa_mask, SIGALRM);
		sigaddset(&sa.sa_mask, SIGVTALRM);
		sigaddset(&sa.sa_mask, SIGUSR1);
		sa.sa_flags = SA_ONSTACK;
		sa.sa_handler = (void *) v;
		sa.sa_restorer = NULL;
		if(sigaction(SIGSEGV, &sa, NULL) < 0)
			panic("userspace_tramp - setting SIGSEGV handler "
			      "failed - errno = %d\n", errno);
	}

	os_stop_process(os_getpid());
	return(0);
}

/* Each element set once, and only accessed by a single processor anyway */
#undef NR_CPUS
#define NR_CPUS 1
int userspace_pid[NR_CPUS];

int start_userspace(unsigned long stub_stack)
{
	void *stack;
	unsigned long sp;
	int pid, status, n, flags;

	stack = mmap(NULL, UM_KERN_PAGE_SIZE,
		     PROT_READ | PROT_WRITE | PROT_EXEC,
		     MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
	if(stack == MAP_FAILED)
		panic("start_userspace : mmap failed, errno = %d", errno);
	sp = (unsigned long) stack + UM_KERN_PAGE_SIZE - sizeof(void *);

	flags = CLONE_FILES | SIGCHLD;
	if(proc_mm) flags |= CLONE_VM;
	pid = clone(userspace_tramp, (void *) sp, flags, (void *) stub_stack);
	if(pid < 0)
		panic("start_userspace : clone failed, errno = %d", errno);

	do {
		CATCH_EINTR(n = waitpid(pid, &status, WUNTRACED));
		if(n < 0)
			panic("start_userspace : wait failed, errno = %d",
			      errno);
	} while(WIFSTOPPED(status) && (WSTOPSIG(status) == SIGVTALRM));

	if(!WIFSTOPPED(status) || (WSTOPSIG(status) != SIGSTOP))
		panic("start_userspace : expected SIGSTOP, got status = %d",
		      status);

	if (ptrace(PTRACE_OLDSETOPTIONS, pid, NULL, (void *)PTRACE_O_TRACESYSGOOD) < 0)
		panic("start_userspace : PTRACE_OLDSETOPTIONS failed, errno=%d\n",
		      errno);

	if(munmap(stack, UM_KERN_PAGE_SIZE) < 0)
		panic("start_userspace : munmap failed, errno = %d\n", errno);

	return(pid);
}

void userspace(union uml_pt_regs *regs)
{
	int err, status, op, pid = userspace_pid[0];
	/* To prevent races if using_sysemu changes under us.*/
	int local_using_sysemu;

	while(1){
		restore_registers(pid, regs);

		/* Now we set local_using_sysemu to be used for one loop */
		local_using_sysemu = get_using_sysemu();

		op = SELECT_PTRACE_OPERATION(local_using_sysemu,
					     singlestepping(NULL));

		err = ptrace(op, pid, 0, 0);
		if(err)
			panic("userspace - could not resume userspace process, "
			      "pid=%d, ptrace operation = %d, errno = %d\n",
			      pid, op, errno);

		CATCH_EINTR(err = waitpid(pid, &status, WUNTRACED));
		if(err < 0)
			panic("userspace - waitpid failed, errno = %d\n",
			      errno);

		regs->skas.is_user = 1;
		save_registers(pid, regs);
		UPT_SYSCALL_NR(regs) = -1; /* Assume: It's not a syscall */

		if(WIFSTOPPED(status)){
			int sig = WSTOPSIG(status);
		  	switch(sig){
			case SIGSEGV:
				if(PTRACE_FULL_FAULTINFO || !ptrace_faultinfo){
					get_skas_faultinfo(pid, &regs->skas.faultinfo);
					(*sig_info[SIGSEGV])(SIGSEGV, regs);
				}
				else handle_segv(pid, regs);
				break;
			case SIGTRAP + 0x80:
			        handle_trap(pid, regs, local_using_sysemu);
				break;
			case SIGTRAP:
				relay_signal(SIGTRAP, regs);
				break;
			case SIGIO:
			case SIGVTALRM:
			case SIGILL:
			case SIGBUS:
			case SIGFPE:
			case SIGWINCH:
				block_signals();
				(*sig_info[sig])(sig, regs);
				unblock_signals();
				break;
			default:
			        printk("userspace - child stopped with signal "
				       "%d\n", sig);
			}
			pid = userspace_pid[0];
			interrupt_end();

			/* Avoid -ERESTARTSYS handling in host */
			if(PT_SYSCALL_NR_OFFSET != PT_SYSCALL_RET_OFFSET)
				PT_SYSCALL_NR(regs->skas.regs) = -1;
		}
	}
}

static unsigned long thread_regs[MAX_REG_NR];
static unsigned long thread_fp_regs[HOST_FP_SIZE];

static int __init init_thread_regs(void)
{
	get_safe_registers(thread_regs, thread_fp_regs);
	/* Set parent's instruction pointer to start of clone-stub */
	thread_regs[REGS_IP_INDEX] = UML_CONFIG_STUB_CODE +
				(unsigned long) stub_clone_handler -
				(unsigned long) &__syscall_stub_start;
	thread_regs[REGS_SP_INDEX] = UML_CONFIG_STUB_DATA + UM_KERN_PAGE_SIZE -
		sizeof(void *);
#ifdef __SIGNAL_FRAMESIZE
	thread_regs[REGS_SP_INDEX] -= __SIGNAL_FRAMESIZE;
#endif
	return 0;
}

__initcall(init_thread_regs);

int copy_context_skas0(unsigned long new_stack, int pid)
{
	int err;
	unsigned long current_stack = current_stub_stack();
	struct stub_data *data = (struct stub_data *) current_stack;
	struct stub_data *child_data = (struct stub_data *) new_stack;
	__u64 new_offset;
	int new_fd = phys_mapping(to_phys((void *)new_stack), &new_offset);

	/* prepare offset and fd of child's stack as argument for parent's
	 * and child's mmap2 calls
	 */
	*data = ((struct stub_data) { .offset	= MMAP_OFFSET(new_offset),
				      .fd	= new_fd,
				      .timer    = ((struct itimerval)
					            { { 0, 1000000 / hz() },
						      { 0, 1000000 / hz() }})});
	err = ptrace_setregs(pid, thread_regs);
	if(err < 0)
		panic("copy_context_skas0 : PTRACE_SETREGS failed, "
		      "pid = %d, errno = %d\n", pid, -err);

	err = ptrace_setfpregs(pid, thread_fp_regs);
	if(err < 0)
		panic("copy_context_skas0 : PTRACE_SETFPREGS failed, "
		      "pid = %d, errno = %d\n", pid, -err);

	/* set a well known return code for detection of child write failure */
	child_data->err = 12345678;

	/* Wait, until parent has finished its work: read child's pid from
	 * parent's stack, and check, if bad result.
	 */
	err = ptrace(PTRACE_CONT, pid, 0, 0);
	if(err)
		panic("Failed to continue new process, pid = %d, "
		      "errno = %d\n", pid, errno);
	wait_stub_done(pid);

	pid = data->err;
	if(pid < 0)
		panic("copy_context_skas0 - stub-parent reports error %d\n",
		      -pid);

	/* Wait, until child has finished too: read child's result from
	 * child's stack and check it.
	 */
	wait_stub_done(pid);
	if (child_data->err != UML_CONFIG_STUB_DATA)
		panic("copy_context_skas0 - stub-child reports error %ld\n",
		      child_data->err);

	if (ptrace(PTRACE_OLDSETOPTIONS, pid, NULL,
		   (void *)PTRACE_O_TRACESYSGOOD) < 0)
		panic("copy_context_skas0 : PTRACE_OLDSETOPTIONS failed, "
		      "errno = %d\n", errno);

	return pid;
}

/*
 * This is used only, if stub pages are needed, while proc_mm is
 * available. Opening /proc/mm creates a new mm_context, which lacks
 * the stub-pages. Thus, we map them using /proc/mm-fd
 */
void map_stub_pages(int fd, unsigned long code,
		    unsigned long data, unsigned long stack)
{
	struct proc_mm_op mmop;
	int n;
	__u64 code_offset;
	int code_fd = phys_mapping(to_phys((void *) &__syscall_stub_start),
				   &code_offset);

	mmop = ((struct proc_mm_op) { .op        = MM_MMAP,
				      .u         =
				      { .mmap    =
					{ .addr    = code,
					  .len     = UM_KERN_PAGE_SIZE,
					  .prot    = PROT_EXEC,
					  .flags   = MAP_FIXED | MAP_PRIVATE,
					  .fd      = code_fd,
					  .offset  = code_offset
	} } });
	CATCH_EINTR(n = write(fd, &mmop, sizeof(mmop)));
	if(n != sizeof(mmop)){
		n = errno;
		printk("mmap args - addr = 0x%lx, fd = %d, offset = %llx\n",
		       code, code_fd, (unsigned long long) code_offset);
		panic("map_stub_pages : /proc/mm map for code failed, "
		      "err = %d\n", n);
	}

	if ( stack ) {
		__u64 map_offset;
		int map_fd = phys_mapping(to_phys((void *)stack), &map_offset);
		mmop = ((struct proc_mm_op)
				{ .op        = MM_MMAP,
				  .u         =
				  { .mmap    =
				    { .addr    = data,
				      .len     = UM_KERN_PAGE_SIZE,
				      .prot    = PROT_READ | PROT_WRITE,
				      .flags   = MAP_FIXED | MAP_SHARED,
				      .fd      = map_fd,
				      .offset  = map_offset
		} } });
		CATCH_EINTR(n = write(fd, &mmop, sizeof(mmop)));
		if(n != sizeof(mmop))
			panic("map_stub_pages : /proc/mm map for data failed, "
			      "err = %d\n", errno);
	}
}

void new_thread(void *stack, jmp_buf *buf, void (*handler)(void))
{
	(*buf)[0].JB_IP = (unsigned long) handler;
	(*buf)[0].JB_SP = (unsigned long) stack + UM_THREAD_SIZE -
		sizeof(void *);
}

#define INIT_JMP_NEW_THREAD 0
#define INIT_JMP_CALLBACK 1
#define INIT_JMP_HALT 2
#define INIT_JMP_REBOOT 3

void switch_threads(jmp_buf *me, jmp_buf *you)
{
	if(UML_SETJMP(me) == 0)
		UML_LONGJMP(you, 1);
}

static jmp_buf initial_jmpbuf;

/* XXX Make these percpu */
static void (*cb_proc)(void *arg);
static void *cb_arg;
static jmp_buf *cb_back;

int start_idle_thread(void *stack, jmp_buf *switch_buf)
{
	int n;

	set_handler(SIGWINCH, (__sighandler_t) sig_handler,
		    SA_ONSTACK | SA_RESTART, SIGUSR1, SIGIO, SIGALRM,
		    SIGVTALRM, -1);

	/*
	 * Can't use UML_SETJMP or UML_LONGJMP here because they save
	 * and restore signals, with the possible side-effect of
	 * trying to handle any signals which came when they were
	 * blocked, which can't be done on this stack.
	 * Signals must be blocked when jumping back here and restored
	 * after returning to the jumper.
	 */
	n = setjmp(initial_jmpbuf);
	switch(n){
	case INIT_JMP_NEW_THREAD:
		(*switch_buf)[0].JB_IP = (unsigned long) new_thread_handler;
		(*switch_buf)[0].JB_SP = (unsigned long) stack +
			UM_THREAD_SIZE - sizeof(void *);
		break;
	case INIT_JMP_CALLBACK:
		(*cb_proc)(cb_arg);
		longjmp(*cb_back, 1);
		break;
	case INIT_JMP_HALT:
		kmalloc_ok = 0;
		return(0);
	case INIT_JMP_REBOOT:
		kmalloc_ok = 0;
		return(1);
	default:
		panic("Bad sigsetjmp return in start_idle_thread - %d\n", n);
	}
	longjmp(*switch_buf, 1);
}

void initial_thread_cb_skas(void (*proc)(void *), void *arg)
{
	jmp_buf here;

	cb_proc = proc;
	cb_arg = arg;
	cb_back = &here;

	block_signals();
	if(UML_SETJMP(&here) == 0)
		UML_LONGJMP(&initial_jmpbuf, INIT_JMP_CALLBACK);
	unblock_signals();

	cb_proc = NULL;
	cb_arg = NULL;
	cb_back = NULL;
}

void halt_skas(void)
{
	block_signals();
	UML_LONGJMP(&initial_jmpbuf, INIT_JMP_HALT);
}

void reboot_skas(void)
{
	block_signals();
	UML_LONGJMP(&initial_jmpbuf, INIT_JMP_REBOOT);
}

void switch_mm_skas(struct mm_id *mm_idp)
{
	int err;

	/* FIXME: need cpu pid in switch_mm_skas */
	if(proc_mm){
		err = ptrace(PTRACE_SWITCH_MM, userspace_pid[0], 0,
			     mm_idp->u.mm_fd);
		if(err)
			panic("switch_mm_skas - PTRACE_SWITCH_MM failed, "
			      "errno = %d\n", errno);
	}
	else userspace_pid[0] = mm_idp->u.pid;
}
Commit	Line	Data
abaf6977 GS	1	/*
	2	* Copyright (C) 2002- 2004 Jeff Dike (jdike@addtoit.com)
	3	* Licensed under the GPL
	4	*/
	5
	6	#include <stdlib.h>
	7	#include <string.h>
	8	#include <unistd.h>
	9	#include <errno.h>
	10	#include <signal.h>
abaf6977 GS	11	#include <sched.h>
	12	#include "ptrace_user.h"
	13	#include <sys/wait.h>
	14	#include <sys/mman.h>
	15	#include <sys/user.h>
	16	#include <sys/time.h>
1f6f6164	17	#include <sys/syscall.h>
abaf6977 GS	18	#include <asm/types.h>
	19	#include "user.h"
	20	#include "sysdep/ptrace.h"
abaf6977 GS	21	#include "kern_util.h"
	22	#include "skas.h"
	23	#include "stub-data.h"
	24	#include "mm_id.h"
	25	#include "sysdep/sigcontext.h"
	26	#include "sysdep/stub.h"
	27	#include "os.h"
	28	#include "proc_mm.h"
	29	#include "skas_ptrace.h"
	30	#include "chan_user.h"
	31	#include "registers.h"
	32	#include "mem.h"
	33	#include "uml-config.h"
	34	#include "process.h"
1d7173ba	35	#include "longjmp.h"
1ffb9164	36	#include "kern_constants.h"
16dd07bc	37	#include "as-layout.h"
abaf6977 GS	38
	39	int is_skas_winch(int pid, int fd, void *data)
	40	{
	41	if(pid != os_getpgrp())
	42	return(0);
	43
	44	register_winch_irq(-1, fd, -1, data);
	45	return(1);
	46	}
	47
f30c2c98 JD	48	static int ptrace_dump_regs(int pid)
	49	{
	50	unsigned long regs[MAX_REG_NR];
	51	int i;
	52
	53	if(ptrace(PTRACE_GETREGS, pid, 0, regs) < 0)
	54	return -errno;
	55	else {
	56	printk("Stub registers -\n");
	57	for(i = 0; i < ARRAY_SIZE(regs); i++)
	58	printk("\t%d - %lx\n", i, regs[i]);
	59	}
	60
	61	return 0;
	62	}
	63
16dd07bc JD	64	/*
	65	* Signals that are OK to receive in the stub - we'll just continue it.
	66	* SIGWINCH will happen when UML is inside a detached screen.
	67	*/
	68	#define STUB_SIG_MASK ((1 << SIGVTALRM) \| (1 << SIGWINCH))
	69
	70	/* Signals that the stub will finish with - anything else is an error */
	71	#define STUB_DONE_MASK ((1 << SIGUSR1) \| (1 << SIGTRAP))
	72
	73	void wait_stub_done(int pid)
abaf6977 GS	74	{
	75	int n, status, err;
	76
16dd07bc	77	while(1){
abaf6977	78	CATCH_EINTR(n = waitpid(pid, &status, WUNTRACED));
16dd07bc JD	79	if((n < 0) \|\| !WIFSTOPPED(status))
	80	goto bad_wait;
	81
	82	if(((1 << WSTOPSIG(status)) & STUB_SIG_MASK) == 0)
	83	break;
	84
	85	err = ptrace(PTRACE_CONT, pid, 0, 0);
f30c2c98	86	if(err)
16dd07bc JD	87	panic("wait_stub_done : continue failed, errno = %d\n",
16dd07bc JD	88	errno);
abaf6977	89	}
16dd07bc JD	90
	91	if(((1 << WSTOPSIG(status)) & STUB_DONE_MASK) != 0)
	92	return;
	93
	94	bad_wait:
	95	err = ptrace_dump_regs(pid);
	96	if(err)
	97	printk("Failed to get registers from stub, errno = %d\n", -err);
	98	panic("wait_stub_done : failed to wait for SIGUSR1/SIGTRAP, pid = %d, "
	99	"n = %d, errno = %d, status = 0x%x\n", pid, n, errno, status);
abaf6977 GS	100	}
	101
	102	extern unsigned long current_stub_stack(void);
	103
	104	void get_skas_faultinfo(int pid, struct faultinfo * fi)
	105	{
	106	int err;
	107
	108	if(ptrace_faultinfo){
	109	err = ptrace(PTRACE_FAULTINFO, pid, 0, fi);
	110	if(err)
	111	panic("get_skas_faultinfo - PTRACE_FAULTINFO failed, "
	112	"errno = %d\n", errno);
	113
	114	/* Special handling for i386, which has different structs */
	115	if (sizeof(struct ptrace_faultinfo) < sizeof(struct faultinfo))
	116	memset((char *)fi + sizeof(struct ptrace_faultinfo), 0,
	117	sizeof(struct faultinfo) -
	118	sizeof(struct ptrace_faultinfo));
	119	}
	120	else {
16dd07bc JD	121	err = ptrace(PTRACE_CONT, pid, 0, SIGSEGV);
	122	if(err)
	123	panic("Failed to continue stub, pid = %d, errno = %d\n",
	124	pid, errno);
	125	wait_stub_done(pid);
abaf6977 GS	126
	127	/* faultinfo is prepared by the stub-segv-handler at start of
	128	* the stub stack page. We just have to copy it.
	129	*/
	130	memcpy(fi, (void )current_stub_stack(), sizeof(fi));
	131	}
	132	}
	133
	134	static void handle_segv(int pid, union uml_pt_regs * regs)
	135	{
	136	get_skas_faultinfo(pid, &regs->skas.faultinfo);
	137	segv(regs->skas.faultinfo, 0, 1, NULL);
	138	}
	139
	140	/To use the same value of using_sysemu as the caller, ask it that value (in local_using_sysemu)/
	141	static void handle_trap(int pid, union uml_pt_regs *regs, int local_using_sysemu)
	142	{
	143	int err, status;
	144
	145	/* Mark this as a syscall */
	146	UPT_SYSCALL_NR(regs) = PT_SYSCALL_NR(regs->skas.regs);
	147
	148	if (!local_using_sysemu)
	149	{
	150	err = ptrace(PTRACE_POKEUSR, pid, PT_SYSCALL_NR_OFFSET,
	151	__NR_getpid);
	152	if(err < 0)
	153	panic("handle_trap - nullifying syscall failed errno = %d\n",
	154	errno);
	155
	156	err = ptrace(PTRACE_SYSCALL, pid, 0, 0);
	157	if(err < 0)
	158	panic("handle_trap - continuing to end of syscall failed, "
	159	"errno = %d\n", errno);
	160
	161	CATCH_EINTR(err = waitpid(pid, &status, WUNTRACED));
	162	if((err < 0) \|\| !WIFSTOPPED(status) \|\|
f30c2c98 JD	163	(WSTOPSIG(status) != SIGTRAP + 0x80)){
	164	err = ptrace_dump_regs(pid);
	165	if(err)
	166	printk("Failed to get registers from process, "
	167	"errno = %d\n", -err);
abaf6977 GS	168	panic("handle_trap - failed to wait at end of syscall, "
abaf6977 GS	169	"errno = %d, status = %d\n", errno, status);
f30c2c98	170	}
abaf6977 GS	171	}
	172
	173	handle_syscall(regs);
	174	}
	175
	176	extern int __syscall_stub_start;
abaf6977 GS	177
	178	static int userspace_tramp(void *stack)
	179	{
	180	void *addr;
537ae946	181	int err;
abaf6977 GS	182
	183	ptrace(PTRACE_TRACEME, 0, 0, 0);
	184
e64bd134	185	init_new_thread_signals();
537ae946 JD	186	err = set_interval(1);
	187	if(err)
	188	panic("userspace_tramp - setting timer failed, errno = %d\n",
	189	err);
abaf6977 GS	190
	191	if(!proc_mm){
	192	/* This has a pte, but it can't be mapped in with the usual
	193	* tlb_flush mechanism because this is part of that mechanism
	194	*/
09ee011e JD	195	int fd;
	196	__u64 offset;
	197	fd = phys_mapping(to_phys(&__syscall_stub_start), &offset);
1ffb9164	198	addr = mmap64((void *) UML_CONFIG_STUB_CODE, UM_KERN_PAGE_SIZE,
09ee011e	199	PROT_EXEC, MAP_FIXED \| MAP_PRIVATE, fd, offset);
abaf6977	200	if(addr == MAP_FAILED){
09ee011e	201	printk("mapping mmap stub failed, errno = %d\n",
abaf6977 GS	202	errno);
	203	exit(1);
	204	}
	205
	206	if(stack != NULL){
abaf6977	207	fd = phys_mapping(to_phys(stack), &offset);
1ffb9164 JD	208	addr = mmap((void *) UML_CONFIG_STUB_DATA,
1ffb9164 JD	209	UM_KERN_PAGE_SIZE, PROT_READ \| PROT_WRITE,
abaf6977 GS	210	MAP_FIXED \| MAP_SHARED, fd, offset);
abaf6977 GS	211	if(addr == MAP_FAILED){
09ee011e	212	printk("mapping segfault stack failed, "
abaf6977 GS	213	"errno = %d\n", errno);
	214	exit(1);
	215	}
	216	}
	217	}
09ee011e	218	if(!ptrace_faultinfo && (stack != NULL)){
4b84c69b JD	219	struct sigaction sa;
4b84c69b JD	220
abaf6977 GS	221	unsigned long v = UML_CONFIG_STUB_CODE +
	222	(unsigned long) stub_segv_handler -
	223	(unsigned long) &__syscall_stub_start;
	224
1ffb9164	225	set_sigstack((void *) UML_CONFIG_STUB_DATA, UM_KERN_PAGE_SIZE);
4b84c69b JD	226	sigemptyset(&sa.sa_mask);
	227	sigaddset(&sa.sa_mask, SIGIO);
	228	sigaddset(&sa.sa_mask, SIGWINCH);
	229	sigaddset(&sa.sa_mask, SIGALRM);
	230	sigaddset(&sa.sa_mask, SIGVTALRM);
	231	sigaddset(&sa.sa_mask, SIGUSR1);
	232	sa.sa_flags = SA_ONSTACK;
	233	sa.sa_handler = (void *) v;
	234	sa.sa_restorer = NULL;
	235	if(sigaction(SIGSEGV, &sa, NULL) < 0)
	236	panic("userspace_tramp - setting SIGSEGV handler "
	237	"failed - errno = %d\n", errno);
abaf6977 GS	238	}
	239
	240	os_stop_process(os_getpid());
	241	return(0);
	242	}
	243
	244	/* Each element set once, and only accessed by a single processor anyway */
	245	#undef NR_CPUS
	246	#define NR_CPUS 1
	247	int userspace_pid[NR_CPUS];
	248
	249	int start_userspace(unsigned long stub_stack)
	250	{
	251	void *stack;
	252	unsigned long sp;
	253	int pid, status, n, flags;
	254
c539ab73 JD	255	stack = mmap(NULL, UM_KERN_PAGE_SIZE,
c539ab73 JD	256	PROT_READ \| PROT_WRITE \| PROT_EXEC,
abaf6977 GS	257	MAP_PRIVATE \| MAP_ANONYMOUS, -1, 0);
	258	if(stack == MAP_FAILED)
	259	panic("start_userspace : mmap failed, errno = %d", errno);
c539ab73	260	sp = (unsigned long) stack + UM_KERN_PAGE_SIZE - sizeof(void *);
abaf6977 GS	261
	262	flags = CLONE_FILES \| SIGCHLD;
	263	if(proc_mm) flags \|= CLONE_VM;
	264	pid = clone(userspace_tramp, (void ) sp, flags, (void ) stub_stack);
	265	if(pid < 0)
	266	panic("start_userspace : clone failed, errno = %d", errno);
	267
	268	do {
	269	CATCH_EINTR(n = waitpid(pid, &status, WUNTRACED));
	270	if(n < 0)
	271	panic("start_userspace : wait failed, errno = %d",
	272	errno);
	273	} while(WIFSTOPPED(status) && (WSTOPSIG(status) == SIGVTALRM));
	274
	275	if(!WIFSTOPPED(status) \|\| (WSTOPSIG(status) != SIGSTOP))
	276	panic("start_userspace : expected SIGSTOP, got status = %d",
	277	status);
	278
	279	if (ptrace(PTRACE_OLDSETOPTIONS, pid, NULL, (void *)PTRACE_O_TRACESYSGOOD) < 0)
	280	panic("start_userspace : PTRACE_OLDSETOPTIONS failed, errno=%d\n",
	281	errno);
	282
c539ab73	283	if(munmap(stack, UM_KERN_PAGE_SIZE) < 0)
abaf6977 GS	284	panic("start_userspace : munmap failed, errno = %d\n", errno);
	285
	286	return(pid);
	287	}
	288
	289	void userspace(union uml_pt_regs *regs)
	290	{
	291	int err, status, op, pid = userspace_pid[0];
2ea5bc5e JD	292	/* To prevent races if using_sysemu changes under us.*/
2ea5bc5e JD	293	int local_using_sysemu;
abaf6977 GS	294
	295	while(1){
	296	restore_registers(pid, regs);
	297
	298	/* Now we set local_using_sysemu to be used for one loop */
	299	local_using_sysemu = get_using_sysemu();
	300
2ea5bc5e JD	301	op = SELECT_PTRACE_OPERATION(local_using_sysemu,
2ea5bc5e JD	302	singlestepping(NULL));
abaf6977 GS	303
	304	err = ptrace(op, pid, 0, 0);
	305	if(err)
	306	panic("userspace - could not resume userspace process, "
	307	"pid=%d, ptrace operation = %d, errno = %d\n",
802e3077	308	pid, op, errno);
abaf6977 GS	309
	310	CATCH_EINTR(err = waitpid(pid, &status, WUNTRACED));
	311	if(err < 0)
	312	panic("userspace - waitpid failed, errno = %d\n",
	313	errno);
	314
	315	regs->skas.is_user = 1;
	316	save_registers(pid, regs);
	317	UPT_SYSCALL_NR(regs) = -1; /* Assume: It's not a syscall */
	318
	319	if(WIFSTOPPED(status)){
16dd07bc JD	320	int sig = WSTOPSIG(status);
16dd07bc JD	321	switch(sig){
abaf6977	322	case SIGSEGV:
16dd07bc JD	323	if(PTRACE_FULL_FAULTINFO \|\| !ptrace_faultinfo){
	324	get_skas_faultinfo(pid, &regs->skas.faultinfo);
	325	(*sig_info[SIGSEGV])(SIGSEGV, regs);
	326	}
abaf6977 GS	327	else handle_segv(pid, regs);
	328	break;
	329	case SIGTRAP + 0x80:
	330	handle_trap(pid, regs, local_using_sysemu);
	331	break;
	332	case SIGTRAP:
	333	relay_signal(SIGTRAP, regs);
	334	break;
	335	case SIGIO:
	336	case SIGVTALRM:
	337	case SIGILL:
	338	case SIGBUS:
	339	case SIGFPE:
	340	case SIGWINCH:
16dd07bc JD	341	block_signals();
	342	(*sig_info[sig])(sig, regs);
	343	unblock_signals();
abaf6977 GS	344	break;
	345	default:
	346	printk("userspace - child stopped with signal "
16dd07bc	347	"%d\n", sig);
abaf6977	348	}
abaf6977 GS	349	pid = userspace_pid[0];
	350	interrupt_end();
	351
	352	/* Avoid -ERESTARTSYS handling in host */
	353	if(PT_SYSCALL_NR_OFFSET != PT_SYSCALL_RET_OFFSET)
	354	PT_SYSCALL_NR(regs->skas.regs) = -1;
	355	}
	356	}
	357	}
abaf6977	358
16dd07bc JD	359	static unsigned long thread_regs[MAX_REG_NR];
	360	static unsigned long thread_fp_regs[HOST_FP_SIZE];
	361
	362	static int __init init_thread_regs(void)
	363	{
	364	get_safe_registers(thread_regs, thread_fp_regs);
	365	/* Set parent's instruction pointer to start of clone-stub */
	366	thread_regs[REGS_IP_INDEX] = UML_CONFIG_STUB_CODE +
	367	(unsigned long) stub_clone_handler -
	368	(unsigned long) &__syscall_stub_start;
c539ab73	369	thread_regs[REGS_SP_INDEX] = UML_CONFIG_STUB_DATA + UM_KERN_PAGE_SIZE -
16dd07bc JD	370	sizeof(void *);
	371	#ifdef __SIGNAL_FRAMESIZE
	372	thread_regs[REGS_SP_INDEX] -= __SIGNAL_FRAMESIZE;
	373	#endif
	374	return 0;
	375	}
	376
	377	__initcall(init_thread_regs);
	378
abaf6977 GS	379	int copy_context_skas0(unsigned long new_stack, int pid)
	380	{
	381	int err;
abaf6977 GS	382	unsigned long current_stack = current_stub_stack();
	383	struct stub_data data = (struct stub_data ) current_stack;
	384	struct stub_data child_data = (struct stub_data ) new_stack;
	385	__u64 new_offset;
	386	int new_fd = phys_mapping(to_phys((void *)new_stack), &new_offset);
	387
	388	/* prepare offset and fd of child's stack as argument for parent's
	389	* and child's mmap2 calls
	390	*/
	391	*data = ((struct stub_data) { .offset = MMAP_OFFSET(new_offset),
	392	.fd = new_fd,
	393	.timer = ((struct itimerval)
	394	{ { 0, 1000000 / hz() },
	395	{ 0, 1000000 / hz() }})});
16dd07bc	396	err = ptrace_setregs(pid, thread_regs);
abaf6977 GS	397	if(err < 0)
abaf6977 GS	398	panic("copy_context_skas0 : PTRACE_SETREGS failed, "
2ace87b9	399	"pid = %d, errno = %d\n", pid, -err);
abaf6977	400
16dd07bc	401	err = ptrace_setfpregs(pid, thread_fp_regs);
e2216feb JD	402	if(err < 0)
e2216feb JD	403	panic("copy_context_skas0 : PTRACE_SETFPREGS failed, "
2ace87b9	404	"pid = %d, errno = %d\n", pid, -err);
e2216feb	405
abaf6977 GS	406	/* set a well known return code for detection of child write failure */
	407	child_data->err = 12345678;
	408
	409	/* Wait, until parent has finished its work: read child's pid from
	410	* parent's stack, and check, if bad result.
	411	*/
16dd07bc JD	412	err = ptrace(PTRACE_CONT, pid, 0, 0);
	413	if(err)
	414	panic("Failed to continue new process, pid = %d, "
	415	"errno = %d\n", pid, errno);
	416	wait_stub_done(pid);
abaf6977 GS	417
	418	pid = data->err;
	419	if(pid < 0)
	420	panic("copy_context_skas0 - stub-parent reports error %d\n",
2ace87b9	421	-pid);
abaf6977 GS	422
	423	/* Wait, until child has finished too: read child's result from
	424	* child's stack and check it.
	425	*/
16dd07bc	426	wait_stub_done(pid);
abaf6977	427	if (child_data->err != UML_CONFIG_STUB_DATA)
802e3077	428	panic("copy_context_skas0 - stub-child reports error %ld\n",
abaf6977 GS	429	child_data->err);
	430
	431	if (ptrace(PTRACE_OLDSETOPTIONS, pid, NULL,
	432	(void *)PTRACE_O_TRACESYSGOOD) < 0)
	433	panic("copy_context_skas0 : PTRACE_OLDSETOPTIONS failed, "
	434	"errno = %d\n", errno);
	435
	436	return pid;
	437	}
	438
	439	/*
	440	* This is used only, if stub pages are needed, while proc_mm is
ef0470c0	441	* available. Opening /proc/mm creates a new mm_context, which lacks
abaf6977 GS	442	* the stub-pages. Thus, we map them using /proc/mm-fd
	443	*/
	444	void map_stub_pages(int fd, unsigned long code,
	445	unsigned long data, unsigned long stack)
	446	{
	447	struct proc_mm_op mmop;
	448	int n;
09ee011e JD	449	__u64 code_offset;
	450	int code_fd = phys_mapping(to_phys((void *) &__syscall_stub_start),
	451	&code_offset);
abaf6977 GS	452
	453	mmop = ((struct proc_mm_op) { .op = MM_MMAP,
	454	.u =
	455	{ .mmap =
	456	{ .addr = code,
c539ab73	457	.len = UM_KERN_PAGE_SIZE,
abaf6977 GS	458	.prot = PROT_EXEC,
abaf6977 GS	459	.flags = MAP_FIXED \| MAP_PRIVATE,
09ee011e JD	460	.fd = code_fd,
09ee011e JD	461	.offset = code_offset
abaf6977	462	} } });
a61f334f	463	CATCH_EINTR(n = write(fd, &mmop, sizeof(mmop)));
b4cf95c6	464	if(n != sizeof(mmop)){
a61f334f	465	n = errno;
b4cf95c6 JD	466	printk("mmap args - addr = 0x%lx, fd = %d, offset = %llx\n",
b4cf95c6 JD	467	code, code_fd, (unsigned long long) code_offset);
abaf6977	468	panic("map_stub_pages : /proc/mm map for code failed, "
a61f334f	469	"err = %d\n", n);
b4cf95c6	470	}
abaf6977 GS	471
	472	if ( stack ) {
	473	__u64 map_offset;
	474	int map_fd = phys_mapping(to_phys((void *)stack), &map_offset);
	475	mmop = ((struct proc_mm_op)
	476	{ .op = MM_MMAP,
	477	.u =
	478	{ .mmap =
	479	{ .addr = data,
c539ab73	480	.len = UM_KERN_PAGE_SIZE,
abaf6977 GS	481	.prot = PROT_READ \| PROT_WRITE,
	482	.flags = MAP_FIXED \| MAP_SHARED,
	483	.fd = map_fd,
	484	.offset = map_offset
	485	} } });
a61f334f	486	CATCH_EINTR(n = write(fd, &mmop, sizeof(mmop)));
abaf6977 GS	487	if(n != sizeof(mmop))
abaf6977 GS	488	panic("map_stub_pages : /proc/mm map for data failed, "
a61f334f	489	"err = %d\n", errno);
abaf6977 GS	490	}
	491	}
	492
3c917350	493	void new_thread(void stack, jmp_buf buf, void (*handler)(void))
abaf6977	494	{
3c917350	495	(*buf)[0].JB_IP = (unsigned long) handler;
e1a79c40 JD	496	(*buf)[0].JB_SP = (unsigned long) stack + UM_THREAD_SIZE -
e1a79c40 JD	497	sizeof(void *);
abaf6977 GS	498	}
abaf6977 GS	499
e2216feb	500	#define INIT_JMP_NEW_THREAD 0
3c917350 JD	501	#define INIT_JMP_CALLBACK 1
	502	#define INIT_JMP_HALT 2
	503	#define INIT_JMP_REBOOT 3
abaf6977	504
3c917350	505	void switch_threads(jmp_buf me, jmp_buf you)
abaf6977	506	{
3c917350 JD	507	if(UML_SETJMP(me) == 0)
3c917350 JD	508	UML_LONGJMP(you, 1);
abaf6977 GS	509	}
abaf6977 GS	510
ad28e029	511	static jmp_buf initial_jmpbuf;
abaf6977 GS	512
	513	/* XXX Make these percpu */
	514	static void (cb_proc)(void arg);
	515	static void *cb_arg;
ad28e029	516	static jmp_buf *cb_back;
abaf6977	517
3c917350	518	int start_idle_thread(void stack, jmp_buf switch_buf)
abaf6977	519	{
a5df0d1a	520	int n;
abaf6977 GS	521
	522	set_handler(SIGWINCH, (__sighandler_t) sig_handler,
	523	SA_ONSTACK \| SA_RESTART, SIGUSR1, SIGIO, SIGALRM,
	524	SIGVTALRM, -1);
	525
77f6af77 JD	526	/*
	527	* Can't use UML_SETJMP or UML_LONGJMP here because they save
	528	* and restore signals, with the possible side-effect of
	529	* trying to handle any signals which came when they were
	530	* blocked, which can't be done on this stack.
	531	* Signals must be blocked when jumping back here and restored
	532	* after returning to the jumper.
	533	*/
	534	n = setjmp(initial_jmpbuf);
abaf6977 GS	535	switch(n){
abaf6977 GS	536	case INIT_JMP_NEW_THREAD:
3c917350 JD	537	(*switch_buf)[0].JB_IP = (unsigned long) new_thread_handler;
3c917350 JD	538	(*switch_buf)[0].JB_SP = (unsigned long) stack +
e1a79c40	539	UM_THREAD_SIZE - sizeof(void *);
abaf6977 GS	540	break;
	541	case INIT_JMP_CALLBACK:
	542	(*cb_proc)(cb_arg);
77f6af77	543	longjmp(*cb_back, 1);
abaf6977 GS	544	break;
	545	case INIT_JMP_HALT:
	546	kmalloc_ok = 0;
	547	return(0);
	548	case INIT_JMP_REBOOT:
	549	kmalloc_ok = 0;
	550	return(1);
	551	default:
	552	panic("Bad sigsetjmp return in start_idle_thread - %d\n", n);
	553	}
77f6af77	554	longjmp(*switch_buf, 1);
abaf6977 GS	555	}
	556
	557	void initial_thread_cb_skas(void (proc)(void ), void *arg)
	558	{
ad28e029	559	jmp_buf here;
abaf6977 GS	560
	561	cb_proc = proc;
	562	cb_arg = arg;
	563	cb_back = &here;
	564
	565	block_signals();
a5df0d1a	566	if(UML_SETJMP(&here) == 0)
ad28e029	567	UML_LONGJMP(&initial_jmpbuf, INIT_JMP_CALLBACK);
abaf6977 GS	568	unblock_signals();
	569
	570	cb_proc = NULL;
	571	cb_arg = NULL;
	572	cb_back = NULL;
	573	}
	574
	575	void halt_skas(void)
	576	{
	577	block_signals();
ad28e029	578	UML_LONGJMP(&initial_jmpbuf, INIT_JMP_HALT);
abaf6977 GS	579	}
	580
	581	void reboot_skas(void)
	582	{
	583	block_signals();
ad28e029	584	UML_LONGJMP(&initial_jmpbuf, INIT_JMP_REBOOT);
abaf6977 GS	585	}
	586
	587	void switch_mm_skas(struct mm_id *mm_idp)
	588	{
	589	int err;
	590
0e7d18b5	591	/* FIXME: need cpu pid in switch_mm_skas */
abaf6977 GS	592	if(proc_mm){
	593	err = ptrace(PTRACE_SWITCH_MM, userspace_pid[0], 0,
	594	mm_idp->u.mm_fd);
	595	if(err)
	596	panic("switch_mm_skas - PTRACE_SWITCH_MM failed, "
	597	"errno = %d\n", errno);
	598	}
	599	else userspace_pid[0] = mm_idp->u.pid;
	600	}