xnu-3789.51.2.tar.gz

[apple/xnu.git] / osfmk / i386 / machine_routines.c

diff --git a/osfmk/i386/machine_routines.c b/osfmk/i386/machine_routines.c
index 06c57561c6401af357db22da96236f2a68d89799..76652946c7e98a76dce65f850c51356d66228ddb 100644 (file)
--- a/osfmk/i386/machine_routines.c
+++ b/osfmk/i386/machine_routines.c
@@ -33,13 +33,16 @@
 #include <mach/processor.h>
 #include <kern/processor.h>
 #include <kern/machine.h>
 #include <mach/processor.h>
 #include <kern/processor.h>
 #include <kern/machine.h>

-#include <kern/cpu_data.h>
+

 #include <kern/cpu_number.h>
 #include <kern/thread.h>
 #include <kern/thread_call.h>
 #include <kern/cpu_number.h>
 #include <kern/thread.h>
 #include <kern/thread_call.h>

+#include <kern/policy_internal.h>
+
+#include <prng/random.h>

 #include <i386/machine_cpu.h>
 #include <i386/lapic.h>
 #include <i386/machine_cpu.h>
 #include <i386/lapic.h>

-#include <i386/lock.h>
+#include <i386/bit_routines.h>

 #include <i386/mp_events.h>
 #include <i386/pmCPU.h>
 #include <i386/trap.h>
 #include <i386/mp_events.h>
 #include <i386/pmCPU.h>
 #include <i386/trap.h>


@@ -54,7 +57,8 @@
 #if KPC
 #include <kern/kpc.h>
 #endif
 #if KPC
 #include <kern/kpc.h>
 #endif

-
+#include <architecture/i386/pio.h>
+#include <i386/cpu_data.h>

 #if DEBUG
 #define DBG(x...)      kprintf("DBG: " x)
 #else
 #if DEBUG
 #define DBG(x...)      kprintf("DBG: " x)
 #else


@@ -65,9 +69,11 @@ extern void  wakeup(void *);
 
 static int max_cpus_initialized = 0;
 
 
 static int max_cpus_initialized = 0;
 

-unsigned int   LockTimeOut;
-unsigned int   LockTimeOutTSC;
-unsigned int   MutexSpin;
+uint64_t       LockTimeOut;
+uint64_t       TLBTimeOut;
+uint64_t       LockTimeOutTSC;
+uint32_t       LockTimeOutUsec;
+uint64_t       MutexSpin;

 uint64_t       LastDebuggerEntryAllowance;
 uint64_t       delay_spin_threshold;
 
 uint64_t       LastDebuggerEntryAllowance;
 uint64_t       delay_spin_threshold;
 


@@ -150,14 +156,17 @@ ml_static_mfree(
                        }
                        pmap_remove(kernel_pmap, vaddr_cur, vaddr_cur+PAGE_SIZE);
                        assert(pmap_valid_page(ppn));
                        }
                        pmap_remove(kernel_pmap, vaddr_cur, vaddr_cur+PAGE_SIZE);
                        assert(pmap_valid_page(ppn));

-

                        if (IS_MANAGED_PAGE(ppn)) {
                                vm_page_create(ppn,(ppn+1));
                        if (IS_MANAGED_PAGE(ppn)) {
                                vm_page_create(ppn,(ppn+1));

-                               vm_page_wire_count--;

                                freed_pages++;
                        }
                }
        }
                                freed_pages++;
                        }
                }
        }

+       vm_page_lockspin_queues();
+       vm_page_wire_count -= freed_pages;
+       vm_page_wire_count_initial -= freed_pages;
+       vm_page_unlock_queues();
+

 #if    DEBUG   
        kprintf("ml_static_mfree: Released 0x%x pages at VA %p, size:0x%llx, last ppn: 0x%x\n", freed_pages, (void *)vaddr, (uint64_t)size, ppn);
 #endif
 #if    DEBUG   
        kprintf("ml_static_mfree: Released 0x%x pages at VA %p, size:0x%llx, last ppn: 0x%x\n", freed_pages, (void *)vaddr, (uint64_t)size, ppn);
 #endif


@@ -277,7 +286,7 @@ boolean_t ml_set_interrupts_enabled(boolean_t enable)
                __asm__ volatile("sti;nop");
 
                if ((get_preemption_level() == 0) && (*ast_pending() & AST_URGENT))
                __asm__ volatile("sti;nop");
 
                if ((get_preemption_level() == 0) && (*ast_pending() & AST_URGENT))

-                       __asm__ volatile ("int $0xff");
+                       __asm__ volatile ("int %0" :: "N" (T_PREEMPT));

        }
        else {
                if (istate)
        }
        else {
                if (istate)


@@ -308,20 +317,25 @@ void ml_cause_interrupt(void)
        panic("ml_cause_interrupt not defined yet on Intel");
 }
 
        panic("ml_cause_interrupt not defined yet on Intel");
 }
 

+/*
+ * TODO: transition users of this to kernel_thread_start_priority
+ * ml_thread_policy is an unsupported KPI
+ */

 void ml_thread_policy(
        thread_t thread,
 __unused       unsigned policy_id,
        unsigned policy_info)
 {
        if (policy_info & MACHINE_NETWORK_WORKLOOP) {
 void ml_thread_policy(
        thread_t thread,
 __unused       unsigned policy_id,
        unsigned policy_info)
 {
        if (policy_info & MACHINE_NETWORK_WORKLOOP) {

-               spl_t           s = splsched();
+               thread_precedence_policy_data_t info;
+               __assert_only kern_return_t kret;

 
 

-               thread_lock(thread);
+               info.importance = 1;

 
 

-               set_priority(thread, thread->priority + 1);
-
-               thread_unlock(thread);
-               splx(s);
+               kret = thread_policy_set_internal(thread, THREAD_PRECEDENCE_POLICY,
+                                                               (thread_policy_t)&info,
+                                                               THREAD_PRECEDENCE_POLICY_COUNT);
+               assert(kret == KERN_SUCCESS);

        }
 }
 
        }
 }
 


@@ -353,6 +367,20 @@ machine_signal_idle(
        cpu_interrupt(processor->cpu_id);
 }
 
        cpu_interrupt(processor->cpu_id);
 }
 

+void
+machine_signal_idle_deferred(
+       __unused processor_t processor)
+{
+       panic("Unimplemented");
+}
+
+void
+machine_signal_idle_cancel(
+       __unused processor_t processor)
+{
+       panic("Unimplemented");
+}
+

 static kern_return_t
 register_cpu(
         uint32_t        lapic_id,
 static kern_return_t
 register_cpu(
         uint32_t        lapic_id,


@@ -386,19 +414,7 @@ register_cpu(
                goto failed;
 
 #if KPC
                goto failed;
 
 #if KPC

-       this_cpu_datap->cpu_kpc_buf[0] = kpc_counterbuf_alloc();
-       if(this_cpu_datap->cpu_kpc_buf[0] == NULL )
-               goto failed;
-       this_cpu_datap->cpu_kpc_buf[1] = kpc_counterbuf_alloc();
-       if(this_cpu_datap->cpu_kpc_buf[1] == NULL )
-               goto failed;
-
-       this_cpu_datap->cpu_kpc_shadow = kpc_counterbuf_alloc();
-       if(this_cpu_datap->cpu_kpc_shadow == NULL )
-               goto failed;
-
-       this_cpu_datap->cpu_kpc_reload = kpc_counterbuf_alloc();
-       if(this_cpu_datap->cpu_kpc_reload == NULL )
+       if (kpc_register_cpu(this_cpu_datap) != TRUE)

                goto failed;
 #endif
 
                goto failed;
 #endif
 


@@ -435,10 +451,7 @@ failed:
        chudxnu_cpu_free(this_cpu_datap->cpu_chud);
        console_cpu_free(this_cpu_datap->cpu_console_buf);
 #if KPC
        chudxnu_cpu_free(this_cpu_datap->cpu_chud);
        console_cpu_free(this_cpu_datap->cpu_console_buf);
 #if KPC

-       kpc_counterbuf_free(this_cpu_datap->cpu_kpc_buf[0]);
-       kpc_counterbuf_free(this_cpu_datap->cpu_kpc_buf[1]);
-       kpc_counterbuf_free(this_cpu_datap->cpu_kpc_shadow);
-       kpc_counterbuf_free(this_cpu_datap->cpu_kpc_reload);
+       kpc_unregister_cpu(this_cpu_datap);

 #endif
 
        return KERN_FAILURE;
 #endif
 
        return KERN_FAILURE;


@@ -490,6 +503,12 @@ ml_processor_register(
     /* fix the CPU id */
     this_cpu_datap->cpu_id = cpu_id;
 
     /* fix the CPU id */
     this_cpu_datap->cpu_id = cpu_id;
 

+    /* allocate and initialize other per-cpu structures */
+    if (!boot_cpu) {
+       mp_cpus_call_cpu_init(cpunum);
+       prng_cpu_init(cpunum);
+    }
+

     /* output arg */
     *processor_out = this_cpu_datap->cpu_processor;
 
     /* output arg */
     *processor_out = this_cpu_datap->cpu_processor;
 


@@ -593,7 +612,6 @@ ml_get_max_cpus(void)
         (void) ml_set_interrupts_enabled(current_state);
         return(machine_info.max_cpus);
 }
         (void) ml_set_interrupts_enabled(current_state);
         return(machine_info.max_cpus);
 }

-

 /*
  *     Routine:        ml_init_lock_timeout
  *     Function:
 /*
  *     Routine:        ml_init_lock_timeout
  *     Function:


@@ -603,17 +621,48 @@ ml_init_lock_timeout(void)
 {
        uint64_t        abstime;
        uint32_t        mtxspin;
 {
        uint64_t        abstime;
        uint32_t        mtxspin;

+#if DEVELOPMENT || DEBUG

        uint64_t        default_timeout_ns = NSEC_PER_SEC>>2;
        uint64_t        default_timeout_ns = NSEC_PER_SEC>>2;

+#else
+       uint64_t        default_timeout_ns = NSEC_PER_SEC>>1;
+#endif

        uint32_t        slto;
        uint32_t        prt;
 
        if (PE_parse_boot_argn("slto_us", &slto, sizeof (slto)))
                default_timeout_ns = slto * NSEC_PER_USEC;
 
        uint32_t        slto;
        uint32_t        prt;
 
        if (PE_parse_boot_argn("slto_us", &slto, sizeof (slto)))
                default_timeout_ns = slto * NSEC_PER_USEC;
 

-       /* LockTimeOut is absolutetime, LockTimeOutTSC is in TSC ticks */
+       /*
+        * LockTimeOut is absolutetime, LockTimeOutTSC is in TSC ticks,
+        * and LockTimeOutUsec is in microseconds and it's 32-bits.
+        */
+       LockTimeOutUsec = (uint32_t) (default_timeout_ns / NSEC_PER_USEC);

        nanoseconds_to_absolutetime(default_timeout_ns, &abstime);
        nanoseconds_to_absolutetime(default_timeout_ns, &abstime);

-       LockTimeOut = (uint32_t) abstime;
-       LockTimeOutTSC = (uint32_t) tmrCvt(abstime, tscFCvtn2t);
+       LockTimeOut = abstime;
+       LockTimeOutTSC = tmrCvt(abstime, tscFCvtn2t);
+
+       /*
+        * TLBTimeOut dictates the TLB flush timeout period. It defaults to
+        * LockTimeOut but can be overriden separately. In particular, a
+        * zero value inhibits the timeout-panic and cuts a trace evnt instead
+        * - see pmap_flush_tlbs().
+        */
+       if (PE_parse_boot_argn("tlbto_us", &slto, sizeof (slto))) {
+               default_timeout_ns = slto * NSEC_PER_USEC;
+               nanoseconds_to_absolutetime(default_timeout_ns, &abstime);
+               TLBTimeOut = (uint32_t) abstime;
+       } else {
+               TLBTimeOut = LockTimeOut;
+       }
+
+#if DEVELOPMENT || DEBUG
+       reportphyreaddelayabs = LockTimeOut;
+#endif
+       if (PE_parse_boot_argn("phyreadmaxus", &slto, sizeof (slto))) {
+               default_timeout_ns = slto * NSEC_PER_USEC;
+               nanoseconds_to_absolutetime(default_timeout_ns, &abstime);
+               reportphyreaddelayabs = abstime;
+       }

 
        if (PE_parse_boot_argn("mtxspin", &mtxspin, sizeof (mtxspin))) {
                if (mtxspin > USEC_PER_SEC>>4)
 
        if (PE_parse_boot_argn("mtxspin", &mtxspin, sizeof (mtxspin))) {
                if (mtxspin > USEC_PER_SEC>>4)


@@ -627,7 +676,38 @@ ml_init_lock_timeout(void)
        nanoseconds_to_absolutetime(4ULL * NSEC_PER_SEC, &LastDebuggerEntryAllowance);
        if (PE_parse_boot_argn("panic_restart_timeout", &prt, sizeof (prt)))
                nanoseconds_to_absolutetime(prt * NSEC_PER_SEC, &panic_restart_timeout);
        nanoseconds_to_absolutetime(4ULL * NSEC_PER_SEC, &LastDebuggerEntryAllowance);
        if (PE_parse_boot_argn("panic_restart_timeout", &prt, sizeof (prt)))
                nanoseconds_to_absolutetime(prt * NSEC_PER_SEC, &panic_restart_timeout);

+

        virtualized = ((cpuid_features() & CPUID_FEATURE_VMM) != 0);
        virtualized = ((cpuid_features() & CPUID_FEATURE_VMM) != 0);

+       if (virtualized) {
+               int     vti;
+               
+               if (!PE_parse_boot_argn("vti", &vti, sizeof (vti)))
+                       vti = 6;
+               printf("Timeouts adjusted for virtualization (<<%d)\n", vti);
+               kprintf("Timeouts adjusted for virtualization (<<%d):\n", vti);
+#define VIRTUAL_TIMEOUT_INFLATE64(_timeout)                    \
+MACRO_BEGIN                                                    \
+       kprintf("%24s: 0x%016llx ", #_timeout, _timeout);       \
+       _timeout <<= vti;                                       \
+       kprintf("-> 0x%016llx\n",  _timeout);                   \
+MACRO_END
+#define VIRTUAL_TIMEOUT_INFLATE32(_timeout)                    \
+MACRO_BEGIN                                                    \
+       kprintf("%24s:         0x%08x ", #_timeout, _timeout);  \
+       if ((_timeout <<vti) >> vti == _timeout)                \
+               _timeout <<= vti;                               \
+       else                                                    \
+               _timeout = ~0; /* cap rather than overflow */   \
+       kprintf("-> 0x%08x\n",  _timeout);                      \
+MACRO_END
+               VIRTUAL_TIMEOUT_INFLATE32(LockTimeOutUsec);
+               VIRTUAL_TIMEOUT_INFLATE64(LockTimeOut);
+               VIRTUAL_TIMEOUT_INFLATE64(LockTimeOutTSC);
+               VIRTUAL_TIMEOUT_INFLATE64(TLBTimeOut);
+               VIRTUAL_TIMEOUT_INFLATE64(MutexSpin);
+               VIRTUAL_TIMEOUT_INFLATE64(reportphyreaddelayabs);
+       }
+

        interrupt_latency_tracker_setup();
        simple_lock_init(&ml_timer_evaluation_slock, 0);
 }
        interrupt_latency_tracker_setup();
        simple_lock_init(&ml_timer_evaluation_slock, 0);
 }


@@ -769,7 +849,7 @@ kernel_preempt_check(void)
 }
 
 boolean_t machine_timeout_suspended(void) {
 }
 
 boolean_t machine_timeout_suspended(void) {

-       return (virtualized || pmap_tlb_flush_timeout || spinlock_timed_out || panic_active() || mp_recent_debugger_activity());
+       return (pmap_tlb_flush_timeout || spinlock_timed_out || panic_active() || mp_recent_debugger_activity() || ml_recent_wake());

 }
 
 /* Eagerly evaluate all pending timer and thread callouts
 }
 
 /* Eagerly evaluate all pending timer and thread callouts


@@ -795,3 +875,57 @@ boolean_t
 ml_timer_forced_evaluation(void) {
        return ml_timer_evaluation_in_progress;
 }
 ml_timer_forced_evaluation(void) {
        return ml_timer_evaluation_in_progress;
 }

+
+/* 32-bit right-rotate n bits */
+static inline uint32_t ror32(uint32_t val, const unsigned int n)
+{      
+       __asm__ volatile("rorl %%cl,%0" : "=r" (val) : "0" (val), "c" (n));
+       return val;
+}
+
+void
+ml_entropy_collect(void)
+{
+       uint32_t        tsc_lo, tsc_hi;
+       uint32_t        *ep;
+
+       assert(cpu_number() == master_cpu);
+
+       /* update buffer pointer cyclically */
+       if (EntropyData.index_ptr - EntropyData.buffer == ENTROPY_BUFFER_SIZE)
+               ep = EntropyData.index_ptr = EntropyData.buffer;
+       else
+               ep = EntropyData.index_ptr++;
+
+       rdtsc_nofence(tsc_lo, tsc_hi);
+       *ep = ror32(*ep, 9) ^ tsc_lo;
+}
+
+uint64_t
+ml_energy_stat(__unused thread_t t) {
+       return 0;
+}
+
+void
+ml_gpu_stat_update(uint64_t gpu_ns_delta) {
+       current_thread()->machine.thread_gpu_ns += gpu_ns_delta;
+}
+
+uint64_t
+ml_gpu_stat(thread_t t) {
+       return t->machine.thread_gpu_ns;
+}
+
+int plctrace_enabled = 0;
+
+void _disable_preemption(void) {
+       disable_preemption_internal();
+}
+
+void _enable_preemption(void) {
+       enable_preemption_internal();
+}
+
+void plctrace_disable(void) {
+       plctrace_enabled = 0;
+}