xnu-792.25.20.tar.gz

[apple/xnu.git] / bsd / dev / ppc / systemcalls.c

diff --git a/bsd/dev/ppc/systemcalls.c b/bsd/dev/ppc/systemcalls.c
index 4bd050722c5da318655991e1b5202958de3c04b5..4d0a3d397e8b590185f18d85bd1e907d8a2ada0f 100644 (file)
--- a/bsd/dev/ppc/systemcalls.c
+++ b/bsd/dev/ppc/systemcalls.c
@@ -1,5 +1,5 @@
 /*
 /*

- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2005 Apple Computer, Inc. All rights reserved.

  *
  * @APPLE_LICENSE_HEADER_START@
  * 
  *
  * @APPLE_LICENSE_HEADER_START@
  * 


@@ -19,150 +19,425 @@
  * 
  * @APPLE_LICENSE_HEADER_END@
  */
  * 
  * @APPLE_LICENSE_HEADER_END@
  */

-/*
- * Copyright (c) 1997 Apple Computer, Inc.
- *
- * PowerPC Family:     System Call handlers.
- *
- * HISTORY
- * 27-July-97  A. Ramesh  
- *     Adopted for Common Core.
- */
- 
-#include <mach/mach_types.h>
-#include <mach/error.h>

 
 

-#include <kern/syscall_sw.h>
-#include <kern/kdp.h>
+#include <kern/task.h>
+#include <kern/thread.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 <machdep/ppc/frame.h>
-#include <machdep/ppc/thread.h>
-#include <machdep/ppc/asm.h>
-#include <machdep/ppc/proc_reg.h>
-#include <machdep/ppc/trap.h>
-#include <machdep/ppc/exception.h>
+#include <sys/kernel.h>
+#include <sys/vm.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>

 
 

-#define        ERESTART        -1              /* restart syscall */
-#define        EJUSTRETURN     -2              /* don't modify regs, just return */
+extern void
+unix_syscall(struct savearea *regs);
+void
+unix_syscall_return(int error);

 
 

+extern struct savearea * 
+find_user_regs(
+       thread_t act);

 
 

-struct unix_syscallargs {
-       int flavor;
-       int r3;
-       int arg1, arg2,arg3,arg4,arg5,arg6,arg7,arg8,arg9;
-};
-extern struct sysent {         /* system call table */
-       int16_t         sy_narg;        /* number of args */
-       int16_t         sy_parallel;/* can execute in parallel */
-       int32_t         (*sy_call)();   /* implementing function */
-} sysent[];
+extern void enter_funnel_section(funnel_t *funnel_lock);
+extern void exit_funnel_section(void);

 
 /*
 
 /*

-** Function:   unix_syscall
-**
-** Inputs:     pcb     - pointer to Process Control Block
-**             arg1    - arguments to mach system calls
-**             arg2
-**             arg3
-**             arg4
-**             arg5
-**             arg6
-**             arg7
-**
-** Outputs:    none
-*/
+ * Function:   unix_syscall
+ *
+ * Inputs:     regs    - pointer to Process Control Block
+ *
+ * Outputs:    none
+ */

 void
 void

-unix_syscall(
-    struct pcb * pcb,
-    int arg1,
-    int arg2,
-    int arg3,
-    int arg4,
-    int arg5,
-    int arg6,
-    int arg7 
-    )
+unix_syscall(struct savearea   *regs)

 {
 {

-    struct ppc_saved_state     *regs;
-    thread_t                   thread;
-    struct proc                        *p;
-   struct sysent               *callp;
-    int                                nargs, error;
-    unsigned short             code;
-    int                                rval[2];
-       struct unix_syscallargs sarg;
+       thread_t                        thread_act;
+       struct uthread          *uthread;
+       struct proc                     *proc;
+       struct sysent           *callp;
+       int                                     error;
+       unsigned short          code;
+       boolean_t                       flavor;
+       int funnel_type;
+       unsigned int cancel_enable;
+
+       flavor = (((unsigned int)regs->save_r0) == 0)? 1: 0;

 
 

-    if (!USERMODE(pcb->ss.srr1))
-       panic("unix_syscall");
+       if (flavor)
+               code = regs->save_r3;
+       else
+               code = regs->save_r0;

 
 

-    regs = &pcb->ss;
-    thread = current_thread();
+       if (kdebug_enable && (code != 180)) {
+               if (flavor)
+                       KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
+                               regs->save_r4, regs->save_r5, regs->save_r6, regs->save_r7, 0);
+               else
+                       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_thread();
+       uthread = get_bsdthread_info(thread_act);

 
 

+       if (!(uthread->uu_flag & UT_VFORK))
+               proc = (struct proc *)get_bsdtask_info(current_task());
+       else
+               proc = current_proc();

 
 

-    /*
-    ** Get index into sysent table
-    */   
-    code = regs->r0;
+       /* Make sure there is a process associated with this task */
+       if (proc == NULL) {
+               regs->save_r3 = (long long)EPERM;
+               /* set the "pc" to execute cerror routine */
+               regs->save_srr0 -= 4;
+               task_terminate_internal(current_task());
+               thread_exception_return();
+               /* NOTREACHED */
+       }

 
 

-    

        /*
        /*

-       ** Set up call pointer
-       */
+        * 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;
+               uthread->uu_ucred = kauth_cred_proc_ref(proc);
+               if (IS_VALID_CRED(old))
+                       kauth_cred_unref(&old);
+       }
+

        callp = (code >= nsysent) ? &sysent[63] : &sysent[code];
 
        callp = (code >= nsysent) ? &sysent[63] : &sysent[code];
 

-       sarg. flavor = (callp == sysent): 1: 0;
-       sarg. r3 = regs->r3;
-       sarg. arg1 = arg1;
-       sarg. arg2 = arg2;
-       sarg. arg3 = arg3;
-       sarg. arg4 = arg4;
-       sarg. arg5 = arg5;
-       sarg. arg6 = arg6;
-       sarg. arg7 = arg7;
+       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) {
+                       regsp = (void *) &regs->save_r3;
+               } else {
+                       /* 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]);
+       }
+
+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);
+       
+       uthread->uu_rval[0] = 0;
+
+       /*
+        * r4 is volatile, if we set it to regs->save_r4 here the child
+        * will have parents r4 after execve
+        */
+       uthread->uu_rval[1] = 0;
+
+       error = 0;
+
+       /*
+        * PPC runtime calls cerror after every unix system call, so
+        * assume no error and adjust the "pc" to skip this call.
+        * It will be set back to the cerror call if an error is detected.
+        */
+       regs->save_srr0 += 4;
+
+       if (KTRPOINT(proc, KTR_SYSCALL))
+               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);

 
 

-    set_bsduthreadargs(thread,pcb,&sarg);
+#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) {
+               regs->save_srr0 -= 8;
+       } else if (error != EJUSTRETURN) {
+               if (error) {
+                       regs->save_r3 = (long long)error;
+                       /* set the "pc" to execute cerror routine */
+                       regs->save_srr0 -= 4;
+               } else { /* (not error) */
+                       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 (callp->sy_narg > 8)
-       panic("unix_syscall: max arg count exceeded");

 
 

-       rval[0] = 0;
+       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;
+               }
+       }

 
 

-       /* r4 is volatile, if we set it to regs->r4 here the child
-        * will have parents r4 after execve */
-       rval[1] = 0;
+       if (cancel_enable == _SYSCALL_CANCEL_NONE)
+                uthread->uu_flag &= ~UT_NOTCANCELPT;

 
 

-       error = 0; /* Start with a good value */
+       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)) {
+
+               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 */
+}
+
+void
+unix_syscall_return(int error)
+{
+       thread_t                                        thread_act;
+       struct uthread                          *uthread;
+       struct proc                                     *proc;
+       struct savearea                         *regs;
+       unsigned short                          code;
+       struct sysent                           *callp;
+       unsigned int cancel_enable;
+
+       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];

 
        /*
 
        /*

-       ** the PPC runtime calls cerror after every unix system call, so
-       ** assume no error and adjust the "pc" to skip this call.
-       ** It will be set back to the cerror call if an error is detected.
-       */
-       regs->srr0 += 4;
-       vt = get_bsduthreadarg(thread);
-       p = ((struct proc *)get_bsdtask_info(current_task()));
-       error = (*(callp->sy_call))(p, (caddr_t)vt, rval);
-
-    if (error == ERESTART) {
-       regs->srr0 -= 8;
-    }
-    else if (error != EJUSTRETURN) {
-       if (error)
-       {
-           regs->r3 = error;
-           /* set the "pc" to execute cerror routine */
-           regs->srr0 -= 4;
-       } else { /* (not error) */
-           regs->r3 = rval[0];
-           regs->r4 = rval[1];
-       } 
-    }
-    /* else  (error == EJUSTRETURN) { nothing } */
-
-    thread_exception_return();
-    /* NOTREACHED */
-       
+        * Get index into sysent table
+        */   
+       if (error == ERESTART) {
+               regs->save_srr0 -= 8;
+       } else if (error != EJUSTRETURN) {
+               if (error) {
+                       regs->save_r3 = (long long)error;
+                       /* set the "pc" to execute cerror routine */
+                       regs->save_srr0 -= 4;
+               } else { /* (not error) */
+                       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 (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;
+               }
+       }
+
+       cancel_enable = callp->sy_cancel;
+
+       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)) {
+               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 */

 }
 
 }
 

+#ifdef JOE_DEBUG
+joe_debug(char *p) {
+
+        printf("%s\n", p);
+}
+#endif