arch/s390/kernel/process.c

   1 /*
   2  * This file handles the architecture dependent parts of process handling.
   3  *
   4  *    Copyright IBM Corp. 1999,2009
   5  *    Author(s): Martin Schwidefsky <schwidefsky@de.ibm.com>,
   6  *               Hartmut Penner <hp@de.ibm.com>,
   7  *               Denis Joseph Barrow,
   8  */
   9
  10 #include <linux/compiler.h>
  11 #include <linux/cpu.h>
  12 #include <linux/errno.h>
  13 #include <linux/sched.h>
  14 #include <linux/kernel.h>
  15 #include <linux/mm.h>
  16 #include <linux/fs.h>
  17 #include <linux/smp.h>
  18 #include <linux/stddef.h>
  19 #include <linux/slab.h>
  20 #include <linux/unistd.h>
  21 #include <linux/ptrace.h>
  22 #include <linux/vmalloc.h>
  23 #include <linux/user.h>
  24 #include <linux/interrupt.h>
  25 #include <linux/delay.h>
  26 #include <linux/reboot.h>
  27 #include <linux/init.h>
  28 #include <linux/module.h>
  29 #include <linux/notifier.h>
  30 #include <linux/tick.h>
  31 #include <linux/elfcore.h>
  32 #include <linux/kernel_stat.h>
  33 #include <linux/personality.h>
  34 #include <linux/syscalls.h>
  35 #include <linux/compat.h>
  36 #include <linux/kprobes.h>
  37 #include <linux/random.h>
  38 #include <asm/compat.h>
  39 #include <asm/uaccess.h>
  40 #include <asm/pgtable.h>
  41 #include <asm/system.h>
  42 #include <asm/io.h>
  43 #include <asm/processor.h>
  44 #include <asm/irq.h>
  45 #include <asm/timer.h>
  46 #include <asm/nmi.h>
  47 #include <asm/smp.h>
  48 #include "entry.h"
  49
  50 asmlinkage void ret_from_fork(void) asm ("ret_from_fork");
  51
  52 /*
  53  * Return saved PC of a blocked thread. used in kernel/sched.
  54  * resume in entry.S does not create a new stack frame, it
  55  * just stores the registers %r6-%r15 to the frame given by
  56  * schedule. We want to return the address of the caller of
  57  * schedule, so we have to walk the backchain one time to
  58  * find the frame schedule() store its return address.
  59  */
  60 unsigned long thread_saved_pc(struct task_struct *tsk)
  61 {
  62         struct stack_frame *sf, *low, *high;
  63
  64         if (!tsk || !task_stack_page(tsk))
  65                 return 0;
  66         low = task_stack_page(tsk);
  67         high = (struct stack_frame *) task_pt_regs(tsk);
  68         sf = (struct stack_frame *) (tsk->thread.ksp & PSW_ADDR_INSN);
  69         if (sf <= low || sf > high)
  70                 return 0;
  71         sf = (struct stack_frame *) (sf->back_chain & PSW_ADDR_INSN);
  72         if (sf <= low || sf > high)
  73                 return 0;
  74         return sf->gprs[8];
  75 }
  76
  77 /*
  78  * The idle loop on a S390...
  79  */
  80 static void default_idle(void)
  81 {
  82         if (cpu_is_offline(smp_processor_id()))
  83                 cpu_die();
  84         local_irq_disable();
  85         if (need_resched()) {
  86                 local_irq_enable();
  87                 return;
  88         }
  89         local_mcck_disable();
  90         if (test_thread_flag(TIF_MCCK_PENDING)) {
  91                 local_mcck_enable();
  92                 local_irq_enable();
  93                 s390_handle_mcck();
  94                 return;
  95         }
  96         trace_hardirqs_on();
  97         /* Don't trace preempt off for idle. */
  98         stop_critical_timings();
  99         /* Stop virtual timer and halt the cpu. */
 100         vtime_stop_cpu();
 101         /* Reenable preemption tracer. */
 102         start_critical_timings();
 103 }
 104
 105 void cpu_idle(void)
 106 {
 107         for (;;) {
 108                 tick_nohz_stop_sched_tick(1);
 109                 while (!need_resched())
 110                         default_idle();
 111                 tick_nohz_restart_sched_tick();
 112                 preempt_enable_no_resched();
 113                 schedule();
 114                 preempt_disable();
 115         }
 116 }
 117
 118 extern void __kprobes kernel_thread_starter(void);
 119
 120 asm(
 121         ".section .kprobes.text, \"ax\"\n"
 122         ".global kernel_thread_starter\n"
 123         "kernel_thread_starter:\n"
 124         "    la    2,0(10)\n"
 125         "    basr  14,9\n"
 126         "    la    2,0\n"
 127         "    br    11\n"
 128         ".previous\n");
 129
 130 int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags)
 131 {
 132         struct pt_regs regs;
 133
 134         memset(&regs, 0, sizeof(regs));
 135         regs.psw.mask = psw_kernel_bits | PSW_MASK_IO | PSW_MASK_EXT;
 136         regs.psw.addr = (unsigned long) kernel_thread_starter | PSW_ADDR_AMODE;
 137         regs.gprs[9] = (unsigned long) fn;
 138         regs.gprs[10] = (unsigned long) arg;
 139         regs.gprs[11] = (unsigned long) do_exit;
 140         regs.orig_gpr2 = -1;
 141
 142         /* Ok, create the new process.. */
 143         return do_fork(flags | CLONE_VM | CLONE_UNTRACED,
 144                        0, &regs, 0, NULL, NULL);
 145 }
 146 EXPORT_SYMBOL(kernel_thread);
 147
 148 /*
 149  * Free current thread data structures etc..
 150  */
 151 void exit_thread(void)
 152 {
 153 }
 154
 155 void flush_thread(void)
 156 {
 157 }
 158
 159 void release_thread(struct task_struct *dead_task)
 160 {
 161 }
 162
 163 int copy_thread(unsigned long clone_flags, unsigned long new_stackp,
 164                 unsigned long unused,
 165                 struct task_struct *p, struct pt_regs *regs)
 166 {
 167         struct thread_info *ti;
 168         struct fake_frame
 169         {
 170                 struct stack_frame sf;
 171                 struct pt_regs childregs;
 172         } *frame;
 173
 174         frame = container_of(task_pt_regs(p), struct fake_frame, childregs);
 175         p->thread.ksp = (unsigned long) frame;
 176         /* Store access registers to kernel stack of new process. */
 177         frame->childregs = *regs;
 178         frame->childregs.gprs[2] = 0;   /* child returns 0 on fork. */
 179         frame->childregs.gprs[15] = new_stackp;
 180         frame->sf.back_chain = 0;
 181
 182         /* new return point is ret_from_fork */
 183         frame->sf.gprs[8] = (unsigned long) ret_from_fork;
 184
 185         /* fake return stack for resume(), don't go back to schedule */
 186         frame->sf.gprs[9] = (unsigned long) frame;
 187
 188         /* Save access registers to new thread structure. */
 189         save_access_regs(&p->thread.acrs[0]);
 190
 191 #ifndef CONFIG_64BIT
 192         /*
 193          * save fprs to current->thread.fp_regs to merge them with
 194          * the emulated registers and then copy the result to the child.
 195          */
 196         save_fp_regs(&current->thread.fp_regs);
 197         memcpy(&p->thread.fp_regs, &current->thread.fp_regs,
 198                sizeof(s390_fp_regs));
 199         /* Set a new TLS ?  */
 200         if (clone_flags & CLONE_SETTLS)
 201                 p->thread.acrs[0] = regs->gprs[6];
 202 #else /* CONFIG_64BIT */
 203         /* Save the fpu registers to new thread structure. */
 204         save_fp_regs(&p->thread.fp_regs);
 205         /* Set a new TLS ?  */
 206         if (clone_flags & CLONE_SETTLS) {
 207                 if (is_compat_task()) {
 208                         p->thread.acrs[0] = (unsigned int) regs->gprs[6];
 209                 } else {
 210                         p->thread.acrs[0] = (unsigned int)(regs->gprs[6] >> 32);
 211                         p->thread.acrs[1] = (unsigned int) regs->gprs[6];
 212                 }
 213         }
 214 #endif /* CONFIG_64BIT */
 215         /* start new process with ar4 pointing to the correct address space */
 216         p->thread.mm_segment = get_fs();
 217         /* Don't copy debug registers */
 218         memset(&p->thread.per_user, 0, sizeof(p->thread.per_user));
 219         memset(&p->thread.per_event, 0, sizeof(p->thread.per_event));
 220         clear_tsk_thread_flag(p, TIF_SINGLE_STEP);
 221         clear_tsk_thread_flag(p, TIF_PER_TRAP);
 222         /* Initialize per thread user and system timer values */
 223         ti = task_thread_info(p);
 224         ti->user_timer = 0;
 225         ti->system_timer = 0;
 226         return 0;
 227 }
 228
 229 SYSCALL_DEFINE0(fork)
 230 {
 231         struct pt_regs *regs = task_pt_regs(current);
 232         return do_fork(SIGCHLD, regs->gprs[15], regs, 0, NULL, NULL);
 233 }
 234
 235 SYSCALL_DEFINE4(clone, unsigned long, newsp, unsigned long, clone_flags,
 236                 int __user *, parent_tidptr, int __user *, child_tidptr)
 237 {
 238         struct pt_regs *regs = task_pt_regs(current);
 239
 240         if (!newsp)
 241                 newsp = regs->gprs[15];
 242         return do_fork(clone_flags, newsp, regs, 0,
 243                        parent_tidptr, child_tidptr);
 244 }
 245
 246 /*
 247  * This is trivial, and on the face of it looks like it
 248  * could equally well be done in user mode.
 249  *
 250  * Not so, for quite unobvious reasons - register pressure.
 251  * In user mode vfork() cannot have a stack frame, and if
 252  * done by calling the "clone()" system call directly, you
 253  * do not have enough call-clobbered registers to hold all
 254  * the information you need.
 255  */
 256 SYSCALL_DEFINE0(vfork)
 257 {
 258         struct pt_regs *regs = task_pt_regs(current);
 259         return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD,
 260                        regs->gprs[15], regs, 0, NULL, NULL);
 261 }
 262
 263 asmlinkage void execve_tail(void)
 264 {
 265         current->thread.fp_regs.fpc = 0;
 266         if (MACHINE_HAS_IEEE)
 267                 asm volatile("sfpc %0,%0" : : "d" (0));
 268 }
 269
 270 /*
 271  * sys_execve() executes a new program.
 272  */
 273 SYSCALL_DEFINE3(execve, const char __user *, name,
 274                 const char __user *const __user *, argv,
 275                 const char __user *const __user *, envp)
 276 {
 277         struct pt_regs *regs = task_pt_regs(current);
 278         char *filename;
 279         long rc;
 280
 281         filename = getname(name);
 282         rc = PTR_ERR(filename);
 283         if (IS_ERR(filename))
 284                 return rc;
 285         rc = do_execve(filename, argv, envp, regs);
 286         if (rc)
 287                 goto out;
 288         execve_tail();
 289         rc = regs->gprs[2];
 290 out:
 291         putname(filename);
 292         return rc;
 293 }
 294
 295 /*
 296  * fill in the FPU structure for a core dump.
 297  */
 298 int dump_fpu (struct pt_regs * regs, s390_fp_regs *fpregs)
 299 {
 300 #ifndef CONFIG_64BIT
 301         /*
 302          * save fprs to current->thread.fp_regs to merge them with
 303          * the emulated registers and then copy the result to the dump.
 304          */
 305         save_fp_regs(&current->thread.fp_regs);
 306         memcpy(fpregs, &current->thread.fp_regs, sizeof(s390_fp_regs));
 307 #else /* CONFIG_64BIT */
 308         save_fp_regs(fpregs);
 309 #endif /* CONFIG_64BIT */
 310         return 1;
 311 }
 312 EXPORT_SYMBOL(dump_fpu);
 313
 314 unsigned long get_wchan(struct task_struct *p)
 315 {
 316         struct stack_frame *sf, *low, *high;
 317         unsigned long return_address;
 318         int count;
 319
 320         if (!p || p == current || p->state == TASK_RUNNING || !task_stack_page(p))
 321                 return 0;
 322         low = task_stack_page(p);
 323         high = (struct stack_frame *) task_pt_regs(p);
 324         sf = (struct stack_frame *) (p->thread.ksp & PSW_ADDR_INSN);
 325         if (sf <= low || sf > high)
 326                 return 0;
 327         for (count = 0; count < 16; count++) {
 328                 sf = (struct stack_frame *) (sf->back_chain & PSW_ADDR_INSN);
 329                 if (sf <= low || sf > high)
 330                         return 0;
 331                 return_address = sf->gprs[8] & PSW_ADDR_INSN;
 332                 if (!in_sched_functions(return_address))
 333                         return return_address;
 334         }
 335         return 0;
 336 }
 337
 338 unsigned long arch_align_stack(unsigned long sp)
 339 {
 340         if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
 341                 sp -= get_random_int() & ~PAGE_MASK;
 342         return sp & ~0xf;
 343 }
 344
 345 static inline unsigned long brk_rnd(void)
 346 {
 347         /* 8MB for 32bit, 1GB for 64bit */
 348         if (is_32bit_task())
 349                 return (get_random_int() & 0x7ffUL) << PAGE_SHIFT;
 350         else
 351                 return (get_random_int() & 0x3ffffUL) << PAGE_SHIFT;
 352 }
 353
 354 unsigned long arch_randomize_brk(struct mm_struct *mm)
 355 {
 356         unsigned long ret = PAGE_ALIGN(mm->brk + brk_rnd());
 357
 358         if (ret < mm->brk)
 359                 return mm->brk;
 360         return ret;
 361 }
 362
 363 unsigned long randomize_et_dyn(unsigned long base)
 364 {
 365         unsigned long ret = PAGE_ALIGN(base + brk_rnd());
 366
 367         if (!(current->flags & PF_RANDOMIZE))
 368                 return base;
 369         if (ret < base)
 370                 return base;
 371         return ret;
 372 }