]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/dev/ppc/systemcalls.c
xnu-792.tar.gz
[apple/xnu.git] / bsd / dev / ppc / systemcalls.c
index 79dcb99f5025cee4dad89e817d226046a90a68d9..a20314a54707efe69c4424041063cd38b17bb0b6 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2003 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
  *
  * @APPLE_LICENSE_HEADER_START@
  * 
 
 #include <kern/task.h>
 #include <kern/thread.h>
-#include <kern/thread_act.h>
 #include <kern/assert.h>
 #include <kern/clock.h>
+#include <kern/locks.h>
+#include <kern/sched_prim.h>
 #include <mach/machine/thread_status.h>
 #include <ppc/savearea.h>
 
 #include <sys/kernel.h>
 #include <sys/vm.h>
-#include <sys/proc.h>
+#include <sys/proc_internal.h>
 #include <sys/syscall.h>
 #include <sys/systm.h>
 #include <sys/user.h>
 #include <sys/errno.h>
 #include <sys/ktrace.h>
 #include <sys/kdebug.h>
+#include <sys/sysent.h>
+#include <sys/sysproto.h>
+#include <sys/kauth.h>
 
 #include <bsm/audit_kernel.h>
 
 extern void
-unix_syscall(
-       struct savearea *regs
-);
+unix_syscall(struct savearea *regs);
+void
+unix_syscall_return(int error);
 
 extern struct savearea * 
 find_user_regs(
-       thread_act_t act);
+       thread_t act);
 
 extern void enter_funnel_section(funnel_t *funnel_lock);
 extern void exit_funnel_section(void);
@@ -60,11 +64,9 @@ extern void exit_funnel_section(void);
  * Outputs:    none
  */
 void
-unix_syscall(
-       struct savearea *regs
-)
+unix_syscall(struct savearea   *regs)
 {
-       thread_act_t            thread_act;
+       thread_t                        thread_act;
        struct uthread          *uthread;
        struct proc                     *proc;
        struct sysent           *callp;
@@ -72,8 +74,9 @@ unix_syscall(
        unsigned short          code;
        boolean_t                       flavor;
        int funnel_type;
+       unsigned int cancel_enable;
 
-       flavor = (((unsigned int)regs->save_r0) == NULL)? 1: 0;
+       flavor = (((unsigned int)regs->save_r0) == 0)? 1: 0;
 
        if (flavor)
                code = regs->save_r3;
@@ -88,49 +91,83 @@ unix_syscall(
                        KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
                                regs->save_r3, regs->save_r4, regs->save_r5, regs->save_r6, 0);
        }
-       thread_act = current_act();
+       thread_act = current_thread();
        uthread = get_bsdthread_info(thread_act);
 
-       if (!(uthread->uu_flag & P_VFORK))
+       if (!(uthread->uu_flag & UT_VFORK))
                proc = (struct proc *)get_bsdtask_info(current_task());
        else
                proc = current_proc();
 
+       /*
+        * Delayed binding of thread credential to process credential, if we
+        * are not running with an explicitly set thread credential.
+        */
+       if (uthread->uu_ucred != proc->p_ucred &&
+           (uthread->uu_flag & UT_SETUID) == 0) {
+               kauth_cred_t old = uthread->uu_ucred;
+               proc_lock(proc);
+               uthread->uu_ucred = proc->p_ucred;
+               kauth_cred_ref(uthread->uu_ucred);
+               proc_unlock(proc);
+               if (old != NOCRED)
+                       kauth_cred_rele(old);
+       }
+
        uthread->uu_ar0 = (int *)regs;
 
        callp = (code >= nsysent) ? &sysent[63] : &sysent[code];
 
-#ifdef DEBUG
-       if (callp->sy_narg > 8)
-               panic("unix_syscall: max arg count exceeded");
-#endif
-
        if (callp->sy_narg != 0) {
+               void            *regsp;
+               sy_munge_t      *mungerp;
+               
+               if (IS_64BIT_PROCESS(proc)) {
+                       /* XXX Turn 64 bit unsafe calls into nosys() */
+                       if (callp->sy_funnel & UNSAFE_64BIT) {
+                               callp = &sysent[63];
+                               goto unsafe;
+                       }
+                       mungerp = callp->sy_arg_munge64;
+               }
+               else {
+                       mungerp = callp->sy_arg_munge32;
+               }
                if ( !flavor) {
-                       uthread->uu_arg[0] = regs->save_r3;
-                       uthread->uu_arg[1] = regs->save_r4;
-                       uthread->uu_arg[2] = regs->save_r5;
-                       uthread->uu_arg[3] = regs->save_r6;
-                       uthread->uu_arg[4] = regs->save_r7;
-                       uthread->uu_arg[5] = regs->save_r8;
-                       uthread->uu_arg[6] = regs->save_r9;
-                       uthread->uu_arg[7] = regs->save_r10;
+                       regsp = (void *) &regs->save_r3;
                } else {
-                       uthread->uu_arg[0] = regs->save_r4;
-                       uthread->uu_arg[1] = regs->save_r5;
-                       uthread->uu_arg[2] = regs->save_r6;
-                       uthread->uu_arg[3] = regs->save_r7;
-                       uthread->uu_arg[4] = regs->save_r8;
-                       uthread->uu_arg[5] = regs->save_r9;
-                       uthread->uu_arg[7] = regs->save_r10;
+                       /* indirect system call consumes an argument so only 7 are supported */
+                       if (callp->sy_narg > 7) {
+                               callp = &sysent[63];
+                               goto unsafe;
+                       }
+                       regsp = (void *) &regs->save_r4;
                }
+               /* call syscall argument munger to copy in arguments (see xnu/bsd/dev/ppc/munge.s) */
+               (*mungerp)(regsp, (void *) &uthread->uu_arg[0]);
        }
 
-       funnel_type = (int)callp->sy_funnel;
+unsafe:
+       cancel_enable = callp->sy_cancel;
+       
+       if (cancel_enable == _SYSCALL_CANCEL_NONE) {
+                       uthread->uu_flag |= UT_NOTCANCELPT;
+       } else {
+               if((uthread->uu_flag & (UT_CANCELDISABLE | UT_CANCEL | UT_CANCELED)) == UT_CANCEL) {
+                       if (cancel_enable == _SYSCALL_CANCEL_PRE) {
+                                       /* system call cancelled; return to handle cancellation */
+                                       regs->save_r3 = (long long)EINTR;
+                                       thread_exception_return();
+                                       /* NOTREACHED */
+                       } else {
+                        thread_abort_safely(thread_act);
+                       }
+               }
+       }
+
+       funnel_type = (int)(callp->sy_funnel & FUNNEL_MASK);
        if (funnel_type == KERNEL_FUNNEL) 
                 enter_funnel_section(kernel_flock);
-       else if (funnel_type == NETWORK_FUNNEL)
-                enter_funnel_section(network_flock);
        
        uthread->uu_rval[0] = 0;
 
@@ -150,12 +187,20 @@ unix_syscall(
        regs->save_srr0 += 4;
 
        if (KTRPOINT(proc, KTR_SYSCALL))
-               ktrsyscall(proc, code, callp->sy_narg, uthread->uu_arg, funnel_type);
+               ktrsyscall(proc, code, callp->sy_narg, uthread->uu_arg);
 
+#ifdef JOE_DEBUG
+       uthread->uu_iocount = 0;
+       uthread->uu_vpindex = 0;
+#endif
        AUDIT_SYSCALL_ENTER(code, proc, uthread);
        error = (*(callp->sy_call))(proc, (void *)uthread->uu_arg, &(uthread->uu_rval[0]));
        AUDIT_SYSCALL_EXIT(error, proc, uthread);
 
+#ifdef JOE_DEBUG
+       if (uthread->uu_iocount)
+               joe_debug("system call returned with uu_iocount != 0");
+#endif
        regs = find_user_regs(thread_act);
 
        if (error == ERESTART) {
@@ -166,42 +211,124 @@ unix_syscall(
                        /* set the "pc" to execute cerror routine */
                        regs->save_srr0 -= 4;
                } else { /* (not error) */
-                       regs->save_r3 = uthread->uu_rval[0];
-                       regs->save_r4 = uthread->uu_rval[1];
+                       switch (callp->sy_return_type) {
+                       case _SYSCALL_RET_INT_T:
+                               regs->save_r3 = uthread->uu_rval[0];
+                               regs->save_r4 = uthread->uu_rval[1];
+                               break;
+                       case _SYSCALL_RET_UINT_T:
+                               regs->save_r3 = ((u_int)uthread->uu_rval[0]);
+                               regs->save_r4 = ((u_int)uthread->uu_rval[1]);
+                               break;
+                       case _SYSCALL_RET_OFF_T:
+                               /* off_t returns 64 bits split across two registers for 32 bit */
+                               /* process and in one register for 64 bit process */
+                               if (IS_64BIT_PROCESS(proc)) {
+                                       u_int64_t       *retp = (u_int64_t *)&uthread->uu_rval[0];
+                                       regs->save_r3 = *retp;
+                                       regs->save_r4 = 0;
+                               }
+                               else {
+                                       regs->save_r3 = uthread->uu_rval[0];
+                                       regs->save_r4 = uthread->uu_rval[1];
+                               }
+                               break;
+                       case _SYSCALL_RET_ADDR_T:
+                       case _SYSCALL_RET_SIZE_T:
+                       case _SYSCALL_RET_SSIZE_T:
+                               /* the variable length return types (user_addr_t, user_ssize_t, 
+                                * and user_size_t) are always the largest possible size in the 
+                                * kernel (we use uu_rval[0] and [1] as one 64 bit value).
+                                */
+                               {
+                                       user_addr_t *retp = (user_addr_t *)&uthread->uu_rval[0];
+                                       regs->save_r3 = *retp;
+                                       regs->save_r4 = 0;
+                               }
+                               break;
+                       case _SYSCALL_RET_NONE:
+                               break;
+                       default:
+                               panic("unix_syscall: unknown return type");
+                               break;
+                       }
                } 
        }
        /* else  (error == EJUSTRETURN) { nothing } */
 
-       if (KTRPOINT(proc, KTR_SYSRET))
-               ktrsysret(proc, code, error, uthread->uu_rval[0], funnel_type);
 
-        exit_funnel_section();
+       if (KTRPOINT(proc, KTR_SYSRET)) {
+               switch(callp->sy_return_type) {
+               case _SYSCALL_RET_ADDR_T:
+               case _SYSCALL_RET_SIZE_T:
+               case _SYSCALL_RET_SSIZE_T:
+                       /*
+                        * Trace the value of the least significant bits,
+                        * until we can revise the ktrace API safely.
+                        */
+                       ktrsysret(proc, code, error, uthread->uu_rval[1]);
+                       break;
+               default:
+                       ktrsysret(proc, code, error, uthread->uu_rval[0]);
+                       break;
+               }
+       }
 
+       if (cancel_enable == _SYSCALL_CANCEL_NONE)
+                uthread->uu_flag &= ~UT_NOTCANCELPT;
+
+       exit_funnel_section();
+
+       if (uthread->uu_lowpri_delay) {
+               /*
+                * task is marked as a low priority I/O type
+                * and the I/O we issued while in this system call
+                * collided with normal I/O operations... we'll
+                * delay in order to mitigate the impact of this
+                * task on the normal operation of the system
+                */
+               IOSleep(uthread->uu_lowpri_delay);
+               uthread->uu_lowpri_delay = 0;
+       }
        if (kdebug_enable && (code != 180)) {
-               KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
-                       error, uthread->uu_rval[0], uthread->uu_rval[1], 0, 0);
+
+               if (callp->sy_return_type == _SYSCALL_RET_SSIZE_T)
+                       KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
+                                             error, uthread->uu_rval[1], 0, 0, 0);
+               else
+                       KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
+                                             error, uthread->uu_rval[0], uthread->uu_rval[1], 0, 0);
        }
 
        thread_exception_return();
        /* NOTREACHED */
 }
 
-unix_syscall_return(error)
+void
+unix_syscall_return(int error)
 {
-       thread_act_t                            thread_act;
+       thread_t                                        thread_act;
        struct uthread                          *uthread;
        struct proc                                     *proc;
        struct savearea                         *regs;
        unsigned short                          code;
        struct sysent                           *callp;
        int funnel_type;
+       unsigned int cancel_enable;
 
-       thread_act = current_act();
+       thread_act = current_thread();
        proc = current_proc();
        uthread = get_bsdthread_info(thread_act);
 
        regs = find_user_regs(thread_act);
 
+       if (regs->save_r0 != 0)
+               code = regs->save_r0;
+       else
+               code = regs->save_r3;
+
+       callp = (code >= nsysent) ? &sysent[63] : &sysent[code];
+
        /*
         * Get index into sysent table
         */   
@@ -213,29 +340,91 @@ unix_syscall_return(error)
                        /* set the "pc" to execute cerror routine */
                        regs->save_srr0 -= 4;
                } else { /* (not error) */
-                       regs->save_r3 = uthread->uu_rval[0];
-                       regs->save_r4 = uthread->uu_rval[1];
+                       switch (callp->sy_return_type) {
+                       case _SYSCALL_RET_INT_T:
+                               regs->save_r3 = uthread->uu_rval[0];
+                               regs->save_r4 = uthread->uu_rval[1];
+                               break;
+                       case _SYSCALL_RET_UINT_T:
+                               regs->save_r3 = ((u_int)uthread->uu_rval[0]);
+                               regs->save_r4 = ((u_int)uthread->uu_rval[1]);
+                               break;
+                       case _SYSCALL_RET_OFF_T:
+                               /* off_t returns 64 bits split across two registers for 32 bit */
+                               /* process and in one register for 64 bit process */
+                               if (IS_64BIT_PROCESS(proc)) {
+                                       u_int64_t       *retp = (u_int64_t *)&uthread->uu_rval[0];
+                                       regs->save_r3 = *retp;
+                               }
+                               else {
+                                       regs->save_r3 = uthread->uu_rval[0];
+                                       regs->save_r4 = uthread->uu_rval[1];
+                               }
+                               break;
+                       case _SYSCALL_RET_ADDR_T:
+                       case _SYSCALL_RET_SIZE_T:
+                       case _SYSCALL_RET_SSIZE_T:
+                               /* the variable length return types (user_addr_t, user_ssize_t, 
+                                * and user_size_t) are always the largest possible size in the 
+                                * kernel (we use uu_rval[0] and [1] as one 64 bit value).
+                                */
+                               {
+                                       u_int64_t       *retp = (u_int64_t *)&uthread->uu_rval[0];
+                                       regs->save_r3 = *retp;
+                               }
+                               break;
+                       case _SYSCALL_RET_NONE:
+                               break;
+                       default:
+                               panic("unix_syscall: unknown return type");
+                               break;
+                       }
                } 
        }
        /* else  (error == EJUSTRETURN) { nothing } */
 
-       if (regs->save_r0 != NULL)
-               code = regs->save_r0;
-       else
-               code = regs->save_r3;
-
-       callp = (code >= nsysent) ? &sysent[63] : &sysent[code];
+       if (KTRPOINT(proc, KTR_SYSRET)) {
+               switch(callp->sy_return_type) {
+               case _SYSCALL_RET_ADDR_T:
+               case _SYSCALL_RET_SIZE_T:
+               case _SYSCALL_RET_SSIZE_T:
+                       /*
+                        * Trace the value of the least significant bits,
+                        * until we can revise the ktrace API safely.
+                        */
+                       ktrsysret(proc, code, error, uthread->uu_rval[1]);
+                       break;
+               default:
+                       ktrsysret(proc, code, error, uthread->uu_rval[0]);
+                       break;
+               }
+       }
 
-       funnel_type = (int)callp->sy_funnel;
+       cancel_enable = callp->sy_cancel;
 
-       if (KTRPOINT(proc, KTR_SYSRET))
-               ktrsysret(proc, code, error, uthread->uu_rval[0], funnel_type);
+       if (cancel_enable == _SYSCALL_CANCEL_NONE)
+                uthread->uu_flag &= ~UT_NOTCANCELPT;
 
-        exit_funnel_section();
+       exit_funnel_section();
 
+       if (uthread->uu_lowpri_delay) {
+               /*
+                * task is marked as a low priority I/O type
+                * and the I/O we issued while in this system call
+                * collided with normal I/O operations... we'll
+                * delay in order to mitigate the impact of this
+                * task on the normal operation of the system
+                */
+               IOSleep(uthread->uu_lowpri_delay);
+               uthread->uu_lowpri_delay = 0;
+       }
        if (kdebug_enable && (code != 180)) {
-               KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
-                       error, uthread->uu_rval[0], uthread->uu_rval[1], 0, 0);
+               if (callp->sy_return_type == _SYSCALL_RET_SSIZE_T)
+                       KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
+                                             error, uthread->uu_rval[1], 0, 0, 0);
+               else
+                       KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
+                                             error, uthread->uu_rval[0], uthread->uu_rval[1], 0, 0);
        }
 
        thread_exception_return();
@@ -251,35 +440,81 @@ unix_syscall_return(error)
  * and decrementing interval timers, optionally reloading the interval
  * timers when they expire.
  */
-struct gettimeofday_args{
-       struct timeval *tp;
-       struct timezone *tzp;
-};
 /*  NOTE THIS implementation is for  ppc architectures only.
  *  It is infrequently called, since the commpage intercepts
  *  most calls in user mode.
+ *
+ * XXX Y2038 bug because of assumed return of 32 bit seconds value, and
+ * XXX first parameter to clock_gettimeofday()
  */
 int
-ppc_gettimeofday(p, uap, retval)
-       struct proc *p;
-       register struct gettimeofday_args *uap;
-       register_t *retval;
+ppc_gettimeofday(__unused struct proc *p, 
+                                register struct ppc_gettimeofday_args *uap, 
+                                register_t *retval)
 {
        int error = 0;
+       extern lck_spin_t * tz_slock;
 
        if (uap->tp)
                clock_gettimeofday(&retval[0], &retval[1]);
        
        if (uap->tzp) {
                struct timezone ltz;
-               extern simple_lock_data_t tz_slock;
 
-               usimple_lock(&tz_slock);
+               lck_spin_lock(tz_slock);
                ltz = tz;
-               usimple_unlock(&tz_slock);
-               error = copyout((caddr_t)&ltz, (caddr_t)uap->tzp, sizeof (tz));
+               lck_spin_unlock(tz_slock);
+               error = copyout((caddr_t)&ltz, uap->tzp, sizeof (tz));
        }
 
        return (error);
 }
 
+#ifdef JOE_DEBUG
+joe_debug(char *p) {
+
+        printf("%s\n", p);
+}
+#endif
+
+
+/* 
+ * WARNING - this is a temporary workaround for binary compatibility issues
+ * with anti-piracy software that relies on patching ptrace (3928003).
+ * This KPI will be removed in the system release after Tiger.
+ */
+uintptr_t temp_patch_ptrace(uintptr_t new_ptrace)
+{
+       struct sysent *         callp;
+       sy_call_t *                     old_ptrace;
+
+       if (new_ptrace == 0)
+               return(0);
+               
+       enter_funnel_section(kernel_flock);
+       callp = &sysent[26];
+       old_ptrace = callp->sy_call;
+       
+       /* only allow one patcher of ptrace */
+       if (old_ptrace == (sy_call_t *) ptrace) {
+               callp->sy_call = (sy_call_t *) new_ptrace;
+       }
+       else {
+               old_ptrace = NULL;
+       }
+       exit_funnel_section( );
+       
+       return((uintptr_t)old_ptrace);
+}
+
+void temp_unpatch_ptrace(void)
+{
+       struct sysent *         callp;
+               
+       enter_funnel_section(kernel_flock);
+       callp = &sysent[26];
+       callp->sy_call = (sy_call_t *) ptrace;
+       exit_funnel_section( );
+       
+       return;
+}