]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/i386/trap.c
xnu-2422.1.72.tar.gz
[apple/xnu.git] / osfmk / i386 / trap.c
index c6f921c6f140692c407cdb03512deab581baaaa1..e15f40b059769587ba11c69b239b76900b4d4a72 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2010 Apple Inc. All rights reserved.
+ * Copyright (c) 2000-2012 Apple Inc. All rights reserved.
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
@@ -89,7 +89,9 @@
 #include <kern/spl.h>
 #include <kern/misc_protos.h>
 #include <kern/debug.h>
-
+#if CONFIG_TELEMETRY
+#include <kern/telemetry.h>
+#endif
 #include <sys/kdebug.h>
 
 #include <string.h>
@@ -113,14 +115,8 @@ extern void kprint_state(x86_saved_state64_t *saved_state);
  * Forward declarations
  */
 static void user_page_fault_continue(kern_return_t kret);
-#ifdef __i386__
-static void panic_trap(x86_saved_state32_t *saved_state);
-static void set_recovery_ip(x86_saved_state32_t *saved_state, vm_offset_t ip);
-extern void panic_64(x86_saved_state_t *, int, const char *, boolean_t);
-#else
 static void panic_trap(x86_saved_state64_t *saved_state);
 static void set_recovery_ip(x86_saved_state64_t *saved_state, vm_offset_t ip);
-#endif
 
 volatile perfCallback perfTrapHook = NULL; /* Pointer to CHUD trap hook routine */
 
@@ -193,7 +189,7 @@ thread_syscall_return(
                                ret);
 #endif
        }
-       throttle_lowpri_io(TRUE);
+       throttle_lowpri_io(1);
 
        thread_exception_return();
         /*NOTREACHED*/
@@ -335,6 +331,9 @@ void interrupt_populate_latency_stats(char *buf, unsigned bufsize) {
                snprintf(buf, bufsize, "0x%x 0x%x 0x%llx", tcpu, cpu_data_ptr[tcpu]->cpu_max_observed_int_latency_vector, cpu_data_ptr[tcpu]->cpu_max_observed_int_latency);
 }
 
+uint32_t interrupt_timer_coalescing_enabled = 1;
+uint64_t interrupt_coalesced_timers;
+
 /*
  * Handle interrupts:
  *  - local APIC interrupts (IPIs, timers, etc) are handled by the kernel,
@@ -349,6 +348,7 @@ interrupt(x86_saved_state_t *state)
        boolean_t       user_mode = FALSE;
        int             ipl;
        int             cnum = cpu_number();
+       cpu_data_t      *cdp = cpu_data_ptr[cnum];
        int             itype = 0;
 
        if (is_saved_state64(state) == TRUE) {
@@ -391,6 +391,17 @@ interrupt(x86_saved_state_t *state)
 
        SCHED_STATS_INTERRUPT(current_processor());
 
+#if CONFIG_TELEMETRY
+       if (telemetry_needs_record
+               && (current_task() != kernel_task)
+#if CONFIG_SCHED_IDLE_IN_PLACE
+               && ((current_thread()->state & TH_IDLE) == 0) /* idle-in-place should be treated like the idle thread */
+#endif
+               ) {
+               telemetry_mark_curthread(user_mode);
+       }
+#endif
+
        ipl = get_preemption_level();
        
        /*
@@ -405,21 +416,40 @@ interrupt(x86_saved_state_t *state)
        }
 
 
-       KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, 
-               MACHDBG_CODE(DBG_MACH_EXCP_INTR, 0) | DBG_FUNC_END,
-               interrupt_num, 0, 0, 0, 0);
-
-       if (cpu_data_ptr[cnum]->cpu_nested_istack) {
-               cpu_data_ptr[cnum]->cpu_nested_istack_events++;
+       if (__improbable(cdp->cpu_nested_istack)) {
+               cdp->cpu_nested_istack_events++;
        }
        else  {
-               uint64_t int_latency = mach_absolute_time() - cpu_data_ptr[cnum]->cpu_int_event_time;
-               if (ilat_assert && (int_latency > interrupt_latency_cap) && !machine_timeout_suspended()) {
-                       panic("Interrupt vector 0x%x exceeded interrupt latency threshold, 0x%llx absolute time delta, prior signals: 0x%x, current signals: 0x%x", interrupt_num, int_latency, cpu_data_ptr[cnum]->cpu_prior_signals, cpu_data_ptr[cnum]->cpu_signals);
+               uint64_t ctime = mach_absolute_time();
+               uint64_t int_latency = ctime - cdp->cpu_int_event_time;
+               uint64_t esdeadline, ehdeadline;
+               /* Attempt to process deferred timers in the context of
+                * this interrupt, unless interrupt time has already exceeded
+                * TCOAL_ILAT_THRESHOLD.
+                */
+#define TCOAL_ILAT_THRESHOLD (30000ULL)
+
+               if ((int_latency < TCOAL_ILAT_THRESHOLD) &&
+                   interrupt_timer_coalescing_enabled) {
+                       esdeadline = cdp->rtclock_timer.queue.earliest_soft_deadline;
+                       ehdeadline = cdp->rtclock_timer.deadline;
+                       if ((ctime >= esdeadline) && (ctime < ehdeadline)) {
+                               interrupt_coalesced_timers++;
+                               TCOAL_DEBUG(0x88880000 | DBG_FUNC_START, ctime, esdeadline, ehdeadline, interrupt_coalesced_timers, 0);
+                               rtclock_intr(state);
+                               TCOAL_DEBUG(0x88880000 | DBG_FUNC_END, ctime, esdeadline, interrupt_coalesced_timers, 0, 0);
+                       } else {
+                               TCOAL_DEBUG(0x77770000, ctime, cdp->rtclock_timer.queue.earliest_soft_deadline, cdp->rtclock_timer.deadline, interrupt_coalesced_timers, 0);
+                       }
                }
-               if (int_latency > cpu_data_ptr[cnum]->cpu_max_observed_int_latency) {
-                       cpu_data_ptr[cnum]->cpu_max_observed_int_latency = int_latency;
-                       cpu_data_ptr[cnum]->cpu_max_observed_int_latency_vector = interrupt_num;
+
+               if (__improbable(ilat_assert && (int_latency > interrupt_latency_cap) && !machine_timeout_suspended())) {
+                       panic("Interrupt vector 0x%x exceeded interrupt latency threshold, 0x%llx absolute time delta, prior signals: 0x%x, current signals: 0x%x", interrupt_num, int_latency, cdp->cpu_prior_signals, cdp->cpu_signals);
+               }
+
+               if (__improbable(int_latency > cdp->cpu_max_observed_int_latency)) {
+                       cdp->cpu_max_observed_int_latency = int_latency;
+                       cdp->cpu_max_observed_int_latency_vector = interrupt_num;
                }
        }
 
@@ -427,17 +457,22 @@ interrupt(x86_saved_state_t *state)
         * Having serviced the interrupt first, look at the interrupted stack depth.
         */
        if (!user_mode) {
-               uint64_t depth = cpu_data_ptr[cnum]->cpu_kernel_stack
+               uint64_t depth = cdp->cpu_kernel_stack
                                 + sizeof(struct x86_kernel_state)
                                 + sizeof(struct i386_exception_link *)
                                 - rsp;
-               if (depth > kernel_stack_depth_max) {
+               if (__improbable(depth > kernel_stack_depth_max)) {
                        kernel_stack_depth_max = (vm_offset_t)depth;
                        KERNEL_DEBUG_CONSTANT(
                                MACHDBG_CODE(DBG_MACH_SCHED, MACH_STACK_DEPTH),
                                (long) depth, (long) VM_KERNEL_UNSLIDE(rip), 0, 0, 0);
                }
        }
+       
+       KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, 
+               MACHDBG_CODE(DBG_MACH_EXCP_INTR, 0) | DBG_FUNC_END,
+               interrupt_num, 0, 0, 0, 0);
+
 }
 
 static inline void
@@ -463,11 +498,7 @@ kernel_trap(
        x86_saved_state_t       *state,
        uintptr_t *lo_spp)
 {
-#ifdef __i386__
-       x86_saved_state32_t     *saved_state;
-#else
        x86_saved_state64_t     *saved_state;
-#endif
        int                     code;
        user_addr_t             vaddr;
        int                     type;
@@ -486,22 +517,6 @@ kernel_trap(
        
        thread = current_thread();
 
-#ifdef __i386__
-       if (__improbable(is_saved_state64(state))) {
-               panic_64(state, 0, "Kernel trap with 64-bit state", FALSE);
-       }
-
-       saved_state = saved_state32(state);
-
-       /* Record cpu where state was captured (trampolines don't set this) */
-       saved_state->cpu = cpu_number();
-
-       vaddr = (user_addr_t)saved_state->cr2;
-       type  = saved_state->trapno;
-       code  = saved_state->err & 0xffff;
-       intr  = (saved_state->efl & EFL_IF) != 0;       /* state of ints at trap */
-       kern_ip = (vm_offset_t)saved_state->eip;
-#else
        if (__improbable(is_saved_state32(state)))
                panic("kernel_trap(%p) with 32-bit state", state);
        saved_state = saved_state64(state);
@@ -514,7 +529,6 @@ kernel_trap(
        code  = (int)(saved_state->isf.err & 0xffff);
        intr  = (saved_state->isf.rflags & EFL_IF) != 0;        /* state of ints at trap */
        kern_ip = (vm_offset_t)saved_state->isf.rip;
-#endif
 
        myast = ast_pending();
 
@@ -659,11 +673,7 @@ kernel_trap(
                fpSSEexterrflt();
                return;
            case T_DEBUG:
-#ifdef __i386__
-                   if ((saved_state->efl & EFL_TF) == 0 && NO_WATCHPOINTS)
-#else
                    if ((saved_state->isf.rflags & EFL_TF) == 0 && NO_WATCHPOINTS)
-#endif
                    {
                            /* We've somehow encountered a debug
                             * register match that does not belong
@@ -703,7 +713,8 @@ kernel_trap(
 #endif
 
                result = vm_fault(map,
-                                 vm_map_trunc_page(vaddr),
+                                 vm_map_trunc_page(vaddr,
+                                                   PAGE_MASK),
                                  prot,
                                  FALSE, 
                                  THREAD_UNINT, NULL, 0);
@@ -783,62 +794,13 @@ debugger_entry:
 }
 
 
-#ifdef __i386__
-static void
-set_recovery_ip(x86_saved_state32_t  *saved_state, vm_offset_t ip)
-{
-        saved_state->eip = ip;
-}
-#else
 static void
 set_recovery_ip(x86_saved_state64_t  *saved_state, vm_offset_t ip)
 {
         saved_state->isf.rip = ip;
 }
-#endif
 
 
-#ifdef __i386__
-static void
-panic_trap(x86_saved_state32_t *regs)
-{
-       const char *trapname = "Unknown";
-       pal_cr_t        cr0, cr2, cr3, cr4;
-
-       pal_get_control_registers( &cr0, &cr2, &cr3, &cr4 );
-
-       /*
-        * Issue an I/O port read if one has been requested - this is an
-        * event logic analyzers can use as a trigger point.
-        */
-       panic_io_port_read();
-
-       kprintf("panic trap number 0x%x, eip 0x%x\n", regs->trapno, regs->eip);
-       kprintf("cr0 0x%08x cr2 0x%08x cr3 0x%08x cr4 0x%08x\n",
-               cr0, cr2, cr3, cr4);
-
-       if (regs->trapno < TRAP_TYPES)
-               trapname = trap_type[regs->trapno];
-#undef panic
-       panic("Kernel trap at 0x%08x, type %d=%s, registers:\n"
-             "CR0: 0x%08x, CR2: 0x%08x, CR3: 0x%08x, CR4: 0x%08x\n"
-             "EAX: 0x%08x, EBX: 0x%08x, ECX: 0x%08x, EDX: 0x%08x\n"
-             "CR2: 0x%08x, EBP: 0x%08x, ESI: 0x%08x, EDI: 0x%08x\n"
-             "EFL: 0x%08x, EIP: 0x%08x, CS:  0x%08x, DS:  0x%08x\n"
-             "Error code: 0x%08x%s\n",
-             regs->eip, regs->trapno, trapname, cr0, cr2, cr3, cr4,
-             regs->eax,regs->ebx,regs->ecx,regs->edx,
-             regs->cr2,regs->ebp,regs->esi,regs->edi,
-             regs->efl,regs->eip,regs->cs & 0xFFFF, regs->ds & 0xFFFF, regs->err,
-             virtualized ? " VMM" : "");
-       /*
-        * This next statement is not executed,
-        * but it's needed to stop the compiler using tail call optimization
-        * for the panic call - which confuses the subsequent backtrace.
-        */
-       cr0 = 0;
-}
-#else
 
 
 static void
@@ -900,7 +862,6 @@ panic_trap(x86_saved_state64_t *regs)
         */
        cr0 = 0;
 }
-#endif
 
 #if CONFIG_DTRACE
 extern kern_return_t dtrace_user_probe(x86_saved_state_t *);
@@ -1110,7 +1071,7 @@ user_trap(
 
            case T_PAGE_FAULT:
            {
-               prot = VM_PROT_READ;
+                   prot = VM_PROT_READ;
 
                if (err & T_PF_WRITE)
                        prot |= VM_PROT_WRITE;
@@ -1118,13 +1079,15 @@ user_trap(
                if (__improbable(err & T_PF_EXECUTE))
                        prot |= VM_PROT_EXECUTE;
 #endif
-               kret = vm_fault(thread->map, vm_map_trunc_page(vaddr),
-                                prot, FALSE,
-                                THREAD_ABORTSAFE, NULL, 0);
+               kret = vm_fault(thread->map,
+                               vm_map_trunc_page(vaddr,
+                                                 PAGE_MASK),
+                               prot, FALSE,
+                               THREAD_ABORTSAFE, NULL, 0);
 
                if (__probable((kret == KERN_SUCCESS) || (kret == KERN_ABORTED))) {
                        thread_exception_return();
-               /* NOTREACHED */
+                       /*NOTREACHED*/
                }
 
                user_page_fault_continue(kret);
@@ -1232,27 +1195,11 @@ sync_iss_to_iks(x86_saved_state_t *saved_state)
                pal_get_kern_regs( saved_state );
 
        if ((kstack = current_thread()->kernel_stack) != 0) {
-#ifdef __i386__
-               x86_saved_state32_t     *regs = saved_state32(saved_state);
-#else
                x86_saved_state64_t     *regs = saved_state64(saved_state);
-#endif
 
                iks = STACK_IKS(kstack);
 
                /* Did we take the trap/interrupt in kernel mode? */
-#ifdef __i386__
-               if (regs == USER_REGS32(current_thread()))
-                       record_active_regs = TRUE;
-               else {
-                       iks->k_ebx = regs->ebx;
-                       iks->k_esp = (int)regs;
-                       iks->k_ebp = regs->ebp;
-                       iks->k_edi = regs->edi;
-                       iks->k_esi = regs->esi;
-                       iks->k_eip = regs->eip;
-               }
-#else
                if (regs == USER_REGS64(current_thread()))
                        record_active_regs = TRUE;
                else {
@@ -1265,20 +1212,9 @@ sync_iss_to_iks(x86_saved_state_t *saved_state)
                        iks->k_r15 = regs->r15;
                        iks->k_rip = regs->isf.rip;
                }
-#endif
        }
 
        if (record_active_regs == TRUE) {
-#ifdef __i386__
-               /* Show the trap handler path */
-               __asm__ volatile("movl %%ebx, %0" : "=m" (iks->k_ebx));
-               __asm__ volatile("movl %%esp, %0" : "=m" (iks->k_esp));
-               __asm__ volatile("movl %%ebp, %0" : "=m" (iks->k_ebp));
-               __asm__ volatile("movl %%edi, %0" : "=m" (iks->k_edi));
-               __asm__ volatile("movl %%esi, %0" : "=m" (iks->k_esi));
-               /* "Current" instruction pointer */
-               __asm__ volatile("movl $1f, %0\n1:" : "=m" (iks->k_eip));
-#else
                /* Show the trap handler path */
                __asm__ volatile("movq %%rbx, %0" : "=m" (iks->k_rbx));
                __asm__ volatile("movq %%rsp, %0" : "=m" (iks->k_rsp));
@@ -1292,7 +1228,6 @@ sync_iss_to_iks(x86_saved_state_t *saved_state)
                                 : "=m" (iks->k_rip)
                                 :
                                 : "rax");
-#endif
        }
 }
 
@@ -1309,16 +1244,6 @@ sync_iss_to_iks_unconditionally(__unused x86_saved_state_t *saved_state) {
 
        if ((kstack = current_thread()->kernel_stack) != 0) {
                iks = STACK_IKS(kstack);
-#ifdef __i386__
-               /* Display the trap handler path */
-               __asm__ volatile("movl %%ebx, %0" : "=m" (iks->k_ebx));
-               __asm__ volatile("movl %%esp, %0" : "=m" (iks->k_esp));
-               __asm__ volatile("movl %%ebp, %0" : "=m" (iks->k_ebp));
-               __asm__ volatile("movl %%edi, %0" : "=m" (iks->k_edi));
-               __asm__ volatile("movl %%esi, %0" : "=m" (iks->k_esi));
-               /* "Current" instruction pointer */
-               __asm__ volatile("movl $1f, %0\n1:" : "=m" (iks->k_eip));
-#else
                /* Display the trap handler path */
                __asm__ volatile("movq %%rbx, %0" : "=m" (iks->k_rbx));
                __asm__ volatile("movq %%rsp, %0" : "=m" (iks->k_rsp));
@@ -1329,6 +1254,5 @@ sync_iss_to_iks_unconditionally(__unused x86_saved_state_t *saved_state) {
                __asm__ volatile("movq %%r15, %0" : "=m" (iks->k_r15));
                /* "Current" instruction pointer */
                __asm__ volatile("leaq 1f(%%rip), %%rax; mov %%rax, %0\n1:" : "=m" (iks->k_rip)::"rax");
-#endif
        }
 }