X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/c0fea4742e91338fffdcf79f86a7c1d5e2b97eb1..527f99514973766e9c0382a4d8550dfb00f54939:/osfmk/i386/pcb.c diff --git a/osfmk/i386/pcb.c b/osfmk/i386/pcb.c index 6ca7bff6b..5fb9112ea 100644 --- a/osfmk/i386/pcb.c +++ b/osfmk/i386/pcb.c @@ -1,23 +1,29 @@ /* - * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved. + * Copyright (c) 2000-2016 Apple Inc. All rights reserved. * - * @APPLE_LICENSE_HEADER_START@ + * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ * - * The contents of this file constitute Original Code as defined in and - * are subject to the Apple Public Source License Version 1.1 (the - * "License"). You may not use this file except in compliance with the - * License. Please obtain a copy of the License at - * http://www.apple.com/publicsource and read it before using this file. + * This file contains Original Code and/or Modifications of Original Code + * as defined in and that are subject to the Apple Public Source License + * Version 2.0 (the 'License'). You may not use this file except in + * compliance with the License. The rights granted to you under the License + * may not be used to create, or enable the creation or redistribution of, + * unlawful or unlicensed copies of an Apple operating system, or to + * circumvent, violate, or enable the circumvention or violation of, any + * terms of an Apple operating system software license agreement. * - * This Original Code and all software distributed under the License are - * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER + * Please obtain a copy of the License at + * http://www.opensource.apple.com/apsl/ and read it before using this file. + * + * The Original Code and all software distributed under the License are + * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the - * License for the specific language governing rights and limitations - * under the License. + * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. + * Please see the License for the specific language governing rights and + * limitations under the License. * - * @APPLE_LICENSE_HEADER_END@ + * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ */ /* * @OSF_COPYRIGHT@ @@ -48,7 +54,6 @@ * the rights to redistribute these changes. */ -#include #include #include @@ -58,9 +63,6 @@ #include #include -#include -#include - #include #include #include @@ -74,68 +76,62 @@ #include #include #include +#include #include #include #include #include #include -#include +#include +#include #include #include -#include -#include -#include #include -#include +#include #include -#include +#include #include +#include /* LAPIC_PMC_SWI_VECTOR */ + +#if HYPERVISOR +#include +#endif /* * Maps state flavor to number of words in the state: */ unsigned int _MachineStateCount[] = { - /* FLAVOR_LIST */ - 0, - x86_THREAD_STATE32_COUNT, - x86_FLOAT_STATE32_COUNT, - x86_EXCEPTION_STATE32_COUNT, - x86_THREAD_STATE64_COUNT, - x86_FLOAT_STATE64_COUNT, - x86_EXCEPTION_STATE64_COUNT, - x86_THREAD_STATE_COUNT, - x86_FLOAT_STATE_COUNT, - x86_EXCEPTION_STATE_COUNT, - 0, - x86_SAVED_STATE32_COUNT, - x86_SAVED_STATE64_COUNT, - x86_DEBUG_STATE32_COUNT, - x86_DEBUG_STATE64_COUNT, - x86_DEBUG_STATE_COUNT + [x86_THREAD_STATE32] = x86_THREAD_STATE32_COUNT, + [x86_THREAD_STATE64] = x86_THREAD_STATE64_COUNT, + [x86_THREAD_STATE] = x86_THREAD_STATE_COUNT, + [x86_FLOAT_STATE32] = x86_FLOAT_STATE32_COUNT, + [x86_FLOAT_STATE64] = x86_FLOAT_STATE64_COUNT, + [x86_FLOAT_STATE] = x86_FLOAT_STATE_COUNT, + [x86_EXCEPTION_STATE32] = x86_EXCEPTION_STATE32_COUNT, + [x86_EXCEPTION_STATE64] = x86_EXCEPTION_STATE64_COUNT, + [x86_EXCEPTION_STATE] = x86_EXCEPTION_STATE_COUNT, + [x86_DEBUG_STATE32] = x86_DEBUG_STATE32_COUNT, + [x86_DEBUG_STATE64] = x86_DEBUG_STATE64_COUNT, + [x86_DEBUG_STATE] = x86_DEBUG_STATE_COUNT, + [x86_AVX_STATE32] = x86_AVX_STATE32_COUNT, + [x86_AVX_STATE64] = x86_AVX_STATE64_COUNT, + [x86_AVX_STATE] = x86_AVX_STATE_COUNT, +#if !defined(RC_HIDE_XNU_J137) + [x86_AVX512_STATE32] = x86_AVX512_STATE32_COUNT, + [x86_AVX512_STATE64] = x86_AVX512_STATE64_COUNT, + [x86_AVX512_STATE] = x86_AVX512_STATE_COUNT, +#endif /* not RC_HIDE_XNU_J137 */ }; -zone_t iss_zone32; /* zone for 32bit saved_state area */ -zone_t iss_zone64; /* zone for 64bit saved_state area */ -zone_t ids_zone32; /* zone for 32bit debug_state area */ -zone_t ids_zone64; /* zone for 64bit debug_state area */ - +zone_t iss_zone; /* zone for saved_state area */ +zone_t ids_zone; /* zone for debug_state area */ /* Forward */ -void act_machine_throughcall(thread_t thr_act); -user_addr_t get_useraddr(void); -void act_machine_return(int); -void act_machine_sv_free(thread_t, int); - -extern thread_t Switch_context( - thread_t old, - thread_continue_t cont, - thread_t new); extern void Thread_continue(void); extern void Load_context( - thread_t thread); - + thread_t thread) __attribute__((noreturn)); static void get_exception_state32(thread_t thread, x86_exception_state32_t *es); @@ -155,14 +151,26 @@ set_thread_state32(thread_t thread, x86_thread_state32_t *ts); static int set_thread_state64(thread_t thread, x86_thread_state64_t *ts); +#if HYPERVISOR +static inline void +ml_hv_cswitch(thread_t old, thread_t new) +{ + if (old->hv_thread_target) + hv_callbacks.preempt(old->hv_thread_target); + + if (new->hv_thread_target) + hv_callbacks.dispatch(new->hv_thread_target); +} +#endif + /* - * Don't let an illegal value for dr7 get set. Specifically, - * check for undefined settings. Setting these bit patterns + * Don't let an illegal value for the lower 32-bits of dr7 get set. + * Specifically, check for undefined settings. Setting these bit patterns * result in undefined behaviour and can lead to an unexpected * TRCTRAP. */ static boolean_t -dr7_is_valid(uint32_t *dr7) +dr7d_is_valid(uint32_t *dr7d) { int i; uint32_t mask1, mask2; @@ -174,16 +182,7 @@ dr7_is_valid(uint32_t *dr7) if (!(get_cr4() & CR4_DE)) for (i = 0, mask1 = 0x3<<16, mask2 = 0x2<<16; i < 4; i++, mask1 <<= 4, mask2 <<= 4) - if ((*dr7 & mask1) == mask2) - return (FALSE); - - /* - * len0-3 pattern "10B" is ok for len on 64-bit. - */ - if (current_cpu_datap()->cpu_is64bit == TRUE) - for (i = 0, mask1 = 0x3<<18, mask2 = 0x2<<18; i < 4; - i++, mask1 <<= 4, mask2 <<= 4) - if ((*dr7 & mask1) == mask2) + if ((*dr7d & mask1) == mask2) return (FALSE); /* @@ -192,73 +191,97 @@ dr7_is_valid(uint32_t *dr7) * to "00B" */ for (i = 0; i < 4; i++) - if (((((*dr7 >> (16 + i*4))) & 0x3) == 0) && - ((((*dr7 >> (18 + i*4))) & 0x3) != 0)) + if (((((*dr7d >> (16 + i*4))) & 0x3) == 0) && + ((((*dr7d >> (18 + i*4))) & 0x3) != 0)) return (FALSE); /* * Intel docs have these bits fixed. */ - *dr7 |= 0x1 << 10; /* set bit 10 to 1 */ - *dr7 &= ~(0x1 << 11); /* set bit 11 to 0 */ - *dr7 &= ~(0x1 << 12); /* set bit 12 to 0 */ - *dr7 &= ~(0x1 << 14); /* set bit 14 to 0 */ - *dr7 &= ~(0x1 << 15); /* set bit 15 to 0 */ + *dr7d |= 0x1 << 10; /* set bit 10 to 1 */ + *dr7d &= ~(0x1 << 11); /* set bit 11 to 0 */ + *dr7d &= ~(0x1 << 12); /* set bit 12 to 0 */ + *dr7d &= ~(0x1 << 14); /* set bit 14 to 0 */ + *dr7d &= ~(0x1 << 15); /* set bit 15 to 0 */ /* * We don't allow anything to set the global breakpoints. */ - if (*dr7 & 0x2) + if (*dr7d & 0x2) return (FALSE); - if (*dr7 & (0x2<<2)) + if (*dr7d & (0x2<<2)) return (FALSE); - if (*dr7 & (0x2<<4)) + if (*dr7d & (0x2<<4)) return (FALSE); - if (*dr7 & (0x2<<6)) + if (*dr7d & (0x2<<6)) return (FALSE); return (TRUE); } -static inline void -set_live_debug_state32(cpu_data_t *cdp, x86_debug_state32_t *ds) +extern void set_64bit_debug_regs(x86_debug_state64_t *ds); + +boolean_t +debug_state_is_valid32(x86_debug_state32_t *ds) { - __asm__ volatile ("movl %0,%%db0" : :"r" (ds->dr0)); - __asm__ volatile ("movl %0,%%db1" : :"r" (ds->dr1)); - __asm__ volatile ("movl %0,%%db2" : :"r" (ds->dr2)); - __asm__ volatile ("movl %0,%%db3" : :"r" (ds->dr3)); - if (cpu_mode_is64bit()) - cdp->cpu_dr7 = ds->dr7; -} + if (!dr7d_is_valid(&ds->dr7)) + return FALSE; -extern void set_64bit_debug_regs(x86_debug_state64_t *ds); + return TRUE; +} -static inline void -set_live_debug_state64(cpu_data_t *cdp, x86_debug_state64_t *ds) +boolean_t +debug_state_is_valid64(x86_debug_state64_t *ds) { + if (!dr7d_is_valid((uint32_t *)&ds->dr7)) + return FALSE; + /* - * We need to enter 64-bit mode in order to set the full - * width of these registers + * Don't allow the user to set debug addresses above their max + * value */ - set_64bit_debug_regs(ds); - cdp->cpu_dr7 = ds->dr7; + if (ds->dr7 & 0x1) + if (ds->dr0 >= VM_MAX_PAGE_ADDRESS) + return FALSE; + + if (ds->dr7 & (0x1<<2)) + if (ds->dr1 >= VM_MAX_PAGE_ADDRESS) + return FALSE; + + if (ds->dr7 & (0x1<<4)) + if (ds->dr2 >= VM_MAX_PAGE_ADDRESS) + return FALSE; + + if (ds->dr7 & (0x1<<6)) + if (ds->dr3 >= VM_MAX_PAGE_ADDRESS) + return FALSE; + + /* For x86-64, we must ensure the upper 32-bits of DR7 are clear */ + ds->dr7 &= 0xffffffffULL; + + return TRUE; } + static kern_return_t set_debug_state32(thread_t thread, x86_debug_state32_t *ds) { x86_debug_state32_t *ids; pcb_t pcb; - pcb = thread->machine.pcb; + pcb = THREAD_TO_PCB(thread); ids = pcb->ids; + if (debug_state_is_valid32(ds) != TRUE) { + return KERN_INVALID_ARGUMENT; + } + if (ids == NULL) { - ids = zalloc(ids_zone32); + ids = zalloc(ids_zone); bzero(ids, sizeof *ids); simple_lock(&pcb->lock); @@ -268,45 +291,14 @@ set_debug_state32(thread_t thread, x86_debug_state32_t *ds) simple_unlock(&pcb->lock); } else { simple_unlock(&pcb->lock); - zfree(ids_zone32, ids); + zfree(ids_zone, ids); } } - if (!dr7_is_valid(&ds->dr7)) - goto err; - - /* - * Only allow local breakpoints and make sure they are not - * in the trampoline code. - */ - - if (ds->dr7 & 0x1) - if (ds->dr0 >= (unsigned long)HIGH_MEM_BASE) - goto err; - - if (ds->dr7 & (0x1<<2)) - if (ds->dr1 >= (unsigned long)HIGH_MEM_BASE) - goto err; - - if (ds->dr7 & (0x1<<4)) - if (ds->dr2 >= (unsigned long)HIGH_MEM_BASE) - goto err; - - if (ds->dr7 & (0x1<<6)) - if (ds->dr3 >= (unsigned long)HIGH_MEM_BASE) - goto err; - ids->dr0 = ds->dr0; - ids->dr1 = ds->dr1; - ids->dr2 = ds->dr2; - ids->dr3 = ds->dr3; - ids->dr6 = ds->dr6; - ids->dr7 = ds->dr7; + copy_debug_state32(ds, ids, FALSE); return (KERN_SUCCESS); - -err: - return (KERN_INVALID_ARGUMENT); } static kern_return_t @@ -315,13 +307,24 @@ set_debug_state64(thread_t thread, x86_debug_state64_t *ds) x86_debug_state64_t *ids; pcb_t pcb; - pcb = thread->machine.pcb; + pcb = THREAD_TO_PCB(thread); ids = pcb->ids; + if (debug_state_is_valid64(ds) != TRUE) { + return KERN_INVALID_ARGUMENT; + } + if (ids == NULL) { - ids = zalloc(ids_zone64); + ids = zalloc(ids_zone); bzero(ids, sizeof *ids); +#if HYPERVISOR + if (thread->hv_thread_target) { + hv_callbacks.volatile_state(thread->hv_thread_target, + HV_DEBUG_STATE); + } +#endif + simple_lock(&pcb->lock); /* make sure it wasn't already alloc()'d elsewhere */ if (pcb->ids == NULL) { @@ -329,44 +332,13 @@ set_debug_state64(thread_t thread, x86_debug_state64_t *ds) simple_unlock(&pcb->lock); } else { simple_unlock(&pcb->lock); - zfree(ids_zone64, ids); + zfree(ids_zone, ids); } } - if (!dr7_is_valid((uint32_t *)&ds->dr7)) - goto err; - - /* - * Don't allow the user to set debug addresses above their max - * value - */ - if (ds->dr7 & 0x1) - if (ds->dr0 >= VM_MAX_PAGE_ADDRESS) - goto err; - - if (ds->dr7 & (0x1<<2)) - if (ds->dr1 >= VM_MAX_PAGE_ADDRESS) - goto err; - - if (ds->dr7 & (0x1<<4)) - if (ds->dr2 >= VM_MAX_PAGE_ADDRESS) - goto err; - - if (ds->dr7 & (0x1<<6)) - if (ds->dr3 >= VM_MAX_PAGE_ADDRESS) - goto err; - - ids->dr0 = ds->dr0; - ids->dr1 = ds->dr1; - ids->dr2 = ds->dr2; - ids->dr3 = ds->dr3; - ids->dr6 = ds->dr6; - ids->dr7 = ds->dr7; + copy_debug_state64(ds, ids, FALSE); return (KERN_SUCCESS); - -err: - return (KERN_INVALID_ARGUMENT); } static void @@ -374,17 +346,10 @@ get_debug_state32(thread_t thread, x86_debug_state32_t *ds) { x86_debug_state32_t *saved_state; - saved_state = thread->machine.pcb->ids; + saved_state = thread->machine.ids; if (saved_state) { - ds->dr0 = saved_state->dr0; - ds->dr1 = saved_state->dr1; - ds->dr2 = saved_state->dr2; - ds->dr3 = saved_state->dr3; - ds->dr4 = saved_state->dr4; - ds->dr5 = saved_state->dr5; - ds->dr6 = saved_state->dr6; - ds->dr7 = saved_state->dr7; + copy_debug_state32(saved_state, ds, TRUE); } else bzero(ds, sizeof *ds); } @@ -394,17 +359,10 @@ get_debug_state64(thread_t thread, x86_debug_state64_t *ds) { x86_debug_state64_t *saved_state; - saved_state = (x86_debug_state64_t *)thread->machine.pcb->ids; + saved_state = (x86_debug_state64_t *)thread->machine.ids; if (saved_state) { - ds->dr0 = saved_state->dr0; - ds->dr1 = saved_state->dr1; - ds->dr2 = saved_state->dr2; - ds->dr3 = saved_state->dr3; - ds->dr4 = saved_state->dr4; - ds->dr5 = saved_state->dr5; - ds->dr6 = saved_state->dr6; - ds->dr7 = saved_state->dr7; + copy_debug_state64(saved_state, ds, TRUE); } else bzero(ds, sizeof *ds); } @@ -424,154 +382,6 @@ consider_machine_adjust(void) { } - - -static void -act_machine_switch_pcb( thread_t new ) -{ - pcb_t pcb = new->machine.pcb; - struct real_descriptor *ldtp; - vm_offset_t pcb_stack_top; - vm_offset_t hi_pcb_stack_top; - vm_offset_t hi_iss; - cpu_data_t *cdp = current_cpu_datap(); - - assert(new->kernel_stack != 0); - STACK_IEL(new->kernel_stack)->saved_state = pcb->iss; - - if (!cpu_mode_is64bit()) { - x86_saved_state32_tagged_t *hi_iss32; - - /* - * Save a pointer to the top of the "kernel" stack - - * actually the place in the PCB where a trap into - * kernel mode will push the registers. - */ - hi_iss = (vm_offset_t)((unsigned long) - pmap_cpu_high_map_vaddr(cpu_number(), HIGH_CPU_ISS0) | - ((unsigned long)pcb->iss & PAGE_MASK)); - - cdp->cpu_hi_iss = (void *)hi_iss; - - pmap_high_map(pcb->iss_pte0, HIGH_CPU_ISS0); - pmap_high_map(pcb->iss_pte1, HIGH_CPU_ISS1); - - hi_iss32 = (x86_saved_state32_tagged_t *) hi_iss; - assert(hi_iss32->tag == x86_SAVED_STATE32); - - hi_pcb_stack_top = (int) (hi_iss32 + 1); - - /* - * For fast syscall, top of interrupt stack points to pcb stack - */ - *(vm_offset_t *) current_sstk() = hi_pcb_stack_top; - - current_ktss()->esp0 = hi_pcb_stack_top; -/* XXX: This check is performed against the thread save state flavor rather than the - * task's 64-bit feature flag because of the thread/task 64-bit state divergence - * that can arise in task_set_64bit() on x86. When that is addressed, we can - * revert to checking the task 64 bit feature flag. The assert below is retained - * for that reason. - */ - } else if (is_saved_state64(pcb->iss)) { - x86_saved_state64_tagged_t *iss64; - vm_offset_t isf; - - assert(is_saved_state64(pcb->iss)); - - iss64 = (x86_saved_state64_tagged_t *) pcb->iss; - - /* - * Set pointer to PCB's interrupt stack frame in cpu data. - * Used by syscall and double-fault trap handlers. - */ - isf = (vm_offset_t) &iss64->state.isf; - cdp->cpu_uber.cu_isf = UBER64(isf); - pcb_stack_top = (vm_offset_t) (iss64 + 1); - /* require 16-byte alignment */ - assert((pcb_stack_top & 0xF) == 0); - /* Interrupt stack is pcb */ - current_ktss64()->rsp0 = UBER64(pcb_stack_top); - - /* - * Top of temporary sysenter stack points to pcb stack. - * Although this is not normally used by 64-bit users, - * it needs to be set in case a sysenter is attempted. - */ - *current_sstk64() = UBER64(pcb_stack_top); - - cdp->cpu_task_map = new->map->pmap->pm_kernel_cr3 ? - TASK_MAP_64BIT_SHARED : TASK_MAP_64BIT; - - /* - * Enable the 64-bit user code segment, USER64_CS. - */ - ldt_desc_p(USER64_CS)->access |= ACC_PL_U; - - } else { - x86_saved_state_compat32_t *iss32compat; - vm_offset_t isf; - - assert(is_saved_state32(pcb->iss)); - iss32compat = (x86_saved_state_compat32_t *) pcb->iss; - - pcb_stack_top = (int) (iss32compat + 1); - /* require 16-byte alignment */ - assert((pcb_stack_top & 0xF) == 0); - - /* - * Set pointer to PCB's interrupt stack frame in cpu data. - * Used by debug trap handler. - */ - isf = (vm_offset_t) &iss32compat->isf64; - cdp->cpu_uber.cu_isf = UBER64(isf); - - /* Top of temporary sysenter stack points to pcb stack */ - *current_sstk64() = UBER64(pcb_stack_top); - - /* Interrupt stack is pcb */ - current_ktss64()->rsp0 = UBER64(pcb_stack_top); - - cdp->cpu_task_map = TASK_MAP_32BIT; - - /* - * Disable USER64_CS - */ - ldt_desc_p(USER64_CS)->access &= ~ACC_PL_U; - } - - /* - * Set the thread`s cthread (a.k.a pthread) - * For 32-bit user this involves setting the USER_CTHREAD - * descriptor in the LDT to point to the cthread data. - * The involves copying in the pre-initialized descriptor. - */ - ldtp = (struct real_descriptor *)current_ldt(); - ldtp[sel_idx(USER_CTHREAD)] = pcb->cthread_desc; - if (pcb->uldt_selector != 0) - ldtp[sel_idx(pcb->uldt_selector)] = pcb->uldt_desc; - /* - * For 64-bit, we additionally set the 64-bit User GS base - * address. On return to 64-bit user, the GS.Base MSR will be written. - */ - cdp->cpu_uber.cu_user_gs_base = pcb->cthread_self; - - /* - * Set the thread`s LDT or LDT entry. - */ - if (new->task == TASK_NULL || new->task->i386_ldt == 0) { - /* - * Use system LDT. - */ - ml_cpu_set_ldt(KERNEL_LDT); - } else { - /* - * Task has its own LDT. - */ - user_ldt_set(new); - } -} - /* * Switch to the first thread on a CPU. */ @@ -580,10 +390,19 @@ machine_load_context( thread_t new) { new->machine.specFlags |= OnProc; - act_machine_switch_pcb(new); + act_machine_switch_pcb(NULL, new); Load_context(new); } +static inline void pmap_switch_context(thread_t ot, thread_t nt, int cnum) { + pmap_assert(ml_get_interrupts_enabled() == FALSE); + vm_map_t nmap = nt->map, omap = ot->map; + if ((omap != nmap) || (nmap->pmap->pagezero_accessible)) { + PMAP_DEACTIVATE_MAP(omap, ot, cnum); + PMAP_ACTIVATE_MAP(nmap, nt, cnum); + } +} + /* * Switch to a new thread. * Save the old thread`s kernel state or continuation, @@ -595,42 +414,62 @@ machine_switch_context( thread_continue_t continuation, thread_t new) { -#if MACH_RT - assert(current_cpu_datap()->cpu_active_stack == old->kernel_stack); -#endif + assert(current_cpu_datap()->cpu_active_stack == old->kernel_stack); + +#if KPC + kpc_off_cpu(old); +#endif /* KPC */ /* * Save FP registers if in use. */ - fpu_save_context(old); + fpu_switch_context(old, new); old->machine.specFlags &= ~OnProc; new->machine.specFlags |= OnProc; + /* + * Monitor the stack depth and report new max, + * not worrying about races. + */ + vm_offset_t depth = current_stack_depth(); + if (depth > kernel_stack_depth_max) { + kernel_stack_depth_max = depth; + KERNEL_DEBUG_CONSTANT( + MACHDBG_CODE(DBG_MACH_SCHED, MACH_STACK_DEPTH), + (long) depth, 0, 0, 0, 0); + } + /* * Switch address maps if need be, even if not switching tasks. * (A server activation may be "borrowing" a client map.) */ - PMAP_SWITCH_CONTEXT(old, new, cpu_number()) + pmap_switch_context(old, new, cpu_number()); /* * Load the rest of the user state for the new thread */ - act_machine_switch_pcb(new); - KERNEL_DEBUG_CONSTANT( - MACHDBG_CODE(DBG_MACH_SCHED,MACH_SCHED) | DBG_FUNC_NONE, - (int)old, (int)new, old->sched_pri, new->sched_pri, 0); + act_machine_switch_pcb(old, new); + +#if HYPERVISOR + ml_hv_cswitch(old, new); +#endif + return(Switch_context(old, continuation, new)); } -/* - * act_machine_sv_free - * release saveareas associated with an act. if flag is true, release - * user level savearea(s) too, else don't - */ -void -act_machine_sv_free(__unused thread_t act, __unused int flag) +thread_t +machine_processor_shutdown( + thread_t thread, + void (*doshutdown)(processor_t), + processor_t processor) { +#if CONFIG_VMX + vmx_suspend(); +#endif + fpu_switch_context(thread, NULL); + pmap_switch_context(thread, processor->idle_thread, cpu_number()); + return(Shutdown_context(thread, doshutdown, processor)); } @@ -642,18 +481,24 @@ kern_return_t machine_thread_state_initialize( thread_t thread) { - /* - * If there's an fpu save area, free it. - * The initialized state will then be lazily faulted-in, if required. - * And if we're target, re-arm the no-fpu trap. - */ - if (thread->machine.pcb->ifps) { - (void) fpu_set_fxstate(thread, NULL); + /* + * If there's an fpu save area, free it. + * The initialized state will then be lazily faulted-in, if required. + * And if we're target, re-arm the no-fpu trap. + */ + if (thread->machine.ifps) { + (void) fpu_set_fxstate(thread, NULL, x86_FLOAT_STATE64); if (thread == current_thread()) - clear_fpu(); + clear_fpu(); } - return KERN_SUCCESS; + + if (thread->machine.ids) { + zfree(ids_zone, thread->machine.ids); + thread->machine.ids = NULL; + } + + return KERN_SUCCESS; } uint32_t @@ -687,7 +532,7 @@ get_eflags_exportmask(void) * for either 32bit or 64bit tasks */ - + static void get_exception_state64(thread_t thread, x86_exception_state64_t *es) { @@ -696,7 +541,8 @@ get_exception_state64(thread_t thread, x86_exception_state64_t *es) saved_state = USER_REGS64(thread); es->trapno = saved_state->isf.trapno; - es->err = saved_state->isf.err; + es->cpu = saved_state->isf.cpu; + es->err = (typeof(es->err))saved_state->isf.err; es->faultvaddr = saved_state->cr2; } @@ -708,6 +554,7 @@ get_exception_state32(thread_t thread, x86_exception_state32_t *es) saved_state = USER_REGS32(thread); es->trapno = saved_state->trapno; + es->cpu = saved_state->cpu; es->err = saved_state->err; es->faultvaddr = saved_state->cr2; } @@ -718,8 +565,36 @@ set_thread_state32(thread_t thread, x86_thread_state32_t *ts) { x86_saved_state32_t *saved_state; + pal_register_cache_state(thread, DIRTY); + saved_state = USER_REGS32(thread); + /* + * Scrub segment selector values: + */ + ts->cs = USER_CS; + /* + * On a 64 bit kernel, we always override the data segments, + * as the actual selector numbers have changed. This also + * means that we don't support setting the data segments + * manually any more. + */ + ts->ss = USER_DS; + ts->ds = USER_DS; + ts->es = USER_DS; + + /* Set GS to CTHREAD only if's been established */ + ts->gs = thread->machine.cthread_self ? USER_CTHREAD : NULL_SEG; + + /* Check segment selectors are safe */ + if (!valid_user_segment_selectors(ts->cs, + ts->ss, + ts->ds, + ts->es, + ts->fs, + ts->gs)) + return(KERN_INVALID_ARGUMENT); + saved_state->eax = ts->eax; saved_state->ebx = ts->ebx; saved_state->ecx = ts->ecx; @@ -730,10 +605,10 @@ set_thread_state32(thread_t thread, x86_thread_state32_t *ts) saved_state->uesp = ts->esp; saved_state->efl = (ts->eflags & ~EFL_USER_CLEAR) | EFL_USER_SET; saved_state->eip = ts->eip; - saved_state->cs = ts->cs ? ts->cs : USER_CS; - saved_state->ss = ts->ss ? ts->ss : USER_DS; - saved_state->ds = ts->ds ? ts->ds : USER_DS; - saved_state->es = ts->es ? ts->es : USER_DS; + saved_state->cs = ts->cs; + saved_state->ss = ts->ss; + saved_state->ds = ts->ds; + saved_state->es = ts->es; saved_state->fs = ts->fs; saved_state->gs = ts->gs; @@ -753,6 +628,8 @@ set_thread_state64(thread_t thread, x86_thread_state64_t *ts) { x86_saved_state64_t *saved_state; + pal_register_cache_state(thread, DIRTY); + saved_state = USER_REGS64(thread); if (!IS_USERADDR64_CANONICAL(ts->rsp) || @@ -768,7 +645,6 @@ set_thread_state64(thread_t thread, x86_thread_state64_t *ts) saved_state->r14 = ts->r14; saved_state->r15 = ts->r15; saved_state->rax = ts->rax; - saved_state->rax = ts->rax; saved_state->rbx = ts->rbx; saved_state->rcx = ts->rcx; saved_state->rdx = ts->rdx; @@ -779,8 +655,8 @@ set_thread_state64(thread_t thread, x86_thread_state64_t *ts) saved_state->isf.rflags = (ts->rflags & ~EFL_USER_CLEAR) | EFL_USER_SET; saved_state->isf.rip = ts->rip; saved_state->isf.cs = USER64_CS; - saved_state->fs = ts->fs; - saved_state->gs = ts->gs; + saved_state->fs = (uint32_t)ts->fs; + saved_state->gs = (uint32_t)ts->gs; return(KERN_SUCCESS); } @@ -792,6 +668,8 @@ get_thread_state32(thread_t thread, x86_thread_state32_t *ts) { x86_saved_state32_t *saved_state; + pal_register_cache_state(thread, VALID); + saved_state = USER_REGS32(thread); ts->eax = saved_state->eax; @@ -818,6 +696,8 @@ get_thread_state64(thread_t thread, x86_thread_state64_t *ts) { x86_saved_state64_t *saved_state; + pal_register_cache_state(thread, VALID); + saved_state = USER_REGS64(thread); ts->r8 = saved_state->r8; @@ -844,7 +724,6 @@ get_thread_state64(thread_t thread, x86_thread_state64_t *ts) } - /* * act_machine_set_state: * @@ -858,28 +737,31 @@ machine_thread_set_state( thread_state_t tstate, mach_msg_type_number_t count) { - - switch (flavor) + switch (flavor) { + case x86_SAVED_STATE32: { - case x86_SAVED_STATE32: - { x86_saved_state32_t *state; x86_saved_state32_t *saved_state; if (count < x86_SAVED_STATE32_COUNT) - return(KERN_INVALID_ARGUMENT); + return(KERN_INVALID_ARGUMENT); + + if (thread_is_64bit(thr_act)) + return(KERN_INVALID_ARGUMENT); state = (x86_saved_state32_t *) tstate; /* Check segment selectors are safe */ if (!valid_user_segment_selectors(state->cs, - state->ss, - state->ds, - state->es, - state->fs, - state->gs)) - return KERN_INVALID_ARGUMENT; - + state->ss, + state->ds, + state->es, + state->fs, + state->gs)) + return KERN_INVALID_ARGUMENT; + + pal_register_cache_state(thr_act, DIRTY); + saved_state = USER_REGS32(thr_act); /* @@ -917,23 +799,31 @@ machine_thread_set_state( saved_state->es = state->es; saved_state->fs = state->fs; saved_state->gs = state->gs; + break; - } + } - case x86_SAVED_STATE64: - { + case x86_SAVED_STATE64: + { x86_saved_state64_t *state; x86_saved_state64_t *saved_state; if (count < x86_SAVED_STATE64_COUNT) - return(KERN_INVALID_ARGUMENT); + return(KERN_INVALID_ARGUMENT); + + if (!thread_is_64bit(thr_act)) + return(KERN_INVALID_ARGUMENT); state = (x86_saved_state64_t *) tstate; - /* Check segment selectors are safe XXX gs/fs? */ - if (!valid_user_code_selector(state->isf.cs) || - !valid_user_data_selector(state->fs) || - !valid_user_data_selector(state->gs)) + /* Verify that the supplied code segment selector is + * valid. In 64-bit mode, the FS and GS segment overrides + * use the FS.base and GS.base MSRs to calculate + * base addresses, and the trampolines don't directly + * restore the segment registers--hence they are no + * longer relevant for validation. + */ + if (!valid_user_code_selector(state->isf.cs)) return KERN_INVALID_ARGUMENT; /* Check pc and stack are canonical addresses */ @@ -941,6 +831,8 @@ machine_thread_set_state( !IS_USERADDR64_CANONICAL(state->isf.rip)) return KERN_INVALID_ARGUMENT; + pal_register_cache_state(thr_act, DIRTY); + saved_state = USER_REGS64(thr_act); /* @@ -966,107 +858,142 @@ machine_thread_set_state( saved_state->isf.rflags = (state->isf.rflags & ~EFL_USER_CLEAR) | EFL_USER_SET; - /* + /* * User setting segment registers. * Code and stack selectors have already been * checked. Others will be reset by 'sys' * if they are not valid. */ - saved_state->isf.cs = state->isf.cs; + saved_state->isf.cs = state->isf.cs; saved_state->isf.ss = state->isf.ss; saved_state->fs = state->fs; saved_state->gs = state->gs; break; - } + } - case x86_FLOAT_STATE32: - { - if (count != x86_FLOAT_STATE32_COUNT) + case x86_FLOAT_STATE32: + case x86_AVX_STATE32: +#if !defined(RC_HIDE_XNU_J137) + case x86_AVX512_STATE32: +#endif /* not RC_HIDE_XNU_J137 */ + { + if (count != _MachineStateCount[flavor]) return(KERN_INVALID_ARGUMENT); if (thread_is_64bit(thr_act)) return(KERN_INVALID_ARGUMENT); - return fpu_set_fxstate(thr_act, tstate); - } + return fpu_set_fxstate(thr_act, tstate, flavor); + } - case x86_FLOAT_STATE64: - { - if (count != x86_FLOAT_STATE64_COUNT) + case x86_FLOAT_STATE64: + case x86_AVX_STATE64: +#if !defined(RC_HIDE_XNU_J137) + case x86_AVX512_STATE64: +#endif /* not RC_HIDE_XNU_J137 */ + { + if (count != _MachineStateCount[flavor]) return(KERN_INVALID_ARGUMENT); - if ( !thread_is_64bit(thr_act)) + if (!thread_is_64bit(thr_act)) return(KERN_INVALID_ARGUMENT); - return fpu_set_fxstate(thr_act, tstate); - } + return fpu_set_fxstate(thr_act, tstate, flavor); + } - case x86_FLOAT_STATE: - { - x86_float_state_t *state; + case x86_FLOAT_STATE: + { + x86_float_state_t *state; if (count != x86_FLOAT_STATE_COUNT) return(KERN_INVALID_ARGUMENT); state = (x86_float_state_t *)tstate; - if (state->fsh.flavor == x86_FLOAT_STATE64 && state->fsh.count == x86_FLOAT_STATE64_COUNT && thread_is_64bit(thr_act)) { - return fpu_set_fxstate(thr_act, (thread_state_t)&state->ufs.fs64); + return fpu_set_fxstate(thr_act, (thread_state_t)&state->ufs.fs64, x86_FLOAT_STATE64); } if (state->fsh.flavor == x86_FLOAT_STATE32 && state->fsh.count == x86_FLOAT_STATE32_COUNT && !thread_is_64bit(thr_act)) { - return fpu_set_fxstate(thr_act, (thread_state_t)&state->ufs.fs32); + return fpu_set_fxstate(thr_act, (thread_state_t)&state->ufs.fs32, x86_FLOAT_STATE32); } return(KERN_INVALID_ARGUMENT); - } + } + case x86_AVX_STATE: +#if !defined(RC_HIDE_XNU_J137) + case x86_AVX512_STATE: +#endif + { + x86_avx_state_t *state; + if (count != _MachineStateCount[flavor]) + return(KERN_INVALID_ARGUMENT); - case OLD_i386_THREAD_STATE: - case x86_THREAD_STATE32: - { + state = (x86_avx_state_t *)tstate; + /* Flavors are defined to have sequential values: 32-bit, 64-bit, non-specific */ + /* 64-bit flavor? */ + if (state->ash.flavor == (flavor - 1) && + state->ash.count == _MachineStateCount[flavor - 1] && + thread_is_64bit(thr_act)) { + return fpu_set_fxstate(thr_act, + (thread_state_t)&state->ufs.as64, + flavor - 1); + } + /* 32-bit flavor? */ + if (state->ash.flavor == (flavor - 2) && + state->ash.count == _MachineStateCount[flavor - 2] && + !thread_is_64bit(thr_act)) { + return fpu_set_fxstate(thr_act, + (thread_state_t)&state->ufs.as32, + flavor - 2); + } + return(KERN_INVALID_ARGUMENT); + } + + case x86_THREAD_STATE32: + { if (count != x86_THREAD_STATE32_COUNT) return(KERN_INVALID_ARGUMENT); if (thread_is_64bit(thr_act)) return(KERN_INVALID_ARGUMENT); - + return set_thread_state32(thr_act, (x86_thread_state32_t *)tstate); - } + } - case x86_THREAD_STATE64: - { + case x86_THREAD_STATE64: + { if (count != x86_THREAD_STATE64_COUNT) return(KERN_INVALID_ARGUMENT); - if ( !thread_is_64bit(thr_act)) + if (!thread_is_64bit(thr_act)) return(KERN_INVALID_ARGUMENT); return set_thread_state64(thr_act, (x86_thread_state64_t *)tstate); - } - case x86_THREAD_STATE: - { - x86_thread_state_t *state; + } + case x86_THREAD_STATE: + { + x86_thread_state_t *state; if (count != x86_THREAD_STATE_COUNT) return(KERN_INVALID_ARGUMENT); state = (x86_thread_state_t *)tstate; - if (state->tsh.flavor == x86_THREAD_STATE64 && state->tsh.count == x86_THREAD_STATE64_COUNT && + if (state->tsh.flavor == x86_THREAD_STATE64 && + state->tsh.count == x86_THREAD_STATE64_COUNT && thread_is_64bit(thr_act)) { - return set_thread_state64(thr_act, &state->uts.ts64); - } else if (state->tsh.flavor == x86_THREAD_STATE32 && state->tsh.count == x86_THREAD_STATE32_COUNT && + return set_thread_state64(thr_act, &state->uts.ts64); + } else if (state->tsh.flavor == x86_THREAD_STATE32 && + state->tsh.count == x86_THREAD_STATE32_COUNT && !thread_is_64bit(thr_act)) { - return set_thread_state32(thr_act, &state->uts.ts32); + return set_thread_state32(thr_act, &state->uts.ts32); } else - return(KERN_INVALID_ARGUMENT); - - break; - } + return(KERN_INVALID_ARGUMENT); + } case x86_DEBUG_STATE32: { x86_debug_state32_t *state; @@ -1139,6 +1066,7 @@ machine_thread_get_state( thread_state_t tstate, mach_msg_type_number_t *count) { + switch (flavor) { case THREAD_STATE_FLAVOR_LIST: @@ -1168,6 +1096,39 @@ machine_thread_get_state( break; } + case THREAD_STATE_FLAVOR_LIST_10_9: + { + if (*count < 5) + return (KERN_INVALID_ARGUMENT); + + tstate[0] = x86_THREAD_STATE; + tstate[1] = x86_FLOAT_STATE; + tstate[2] = x86_EXCEPTION_STATE; + tstate[3] = x86_DEBUG_STATE; + tstate[4] = x86_AVX_STATE; + + *count = 5; + break; + } + +#if !defined(RC_HIDE_XNU_J137) + case THREAD_STATE_FLAVOR_LIST_10_13: + { + if (*count < 6) + return (KERN_INVALID_ARGUMENT); + + tstate[0] = x86_THREAD_STATE; + tstate[1] = x86_FLOAT_STATE; + tstate[2] = x86_EXCEPTION_STATE; + tstate[3] = x86_DEBUG_STATE; + tstate[4] = x86_AVX_STATE; + tstate[5] = x86_AVX512_STATE; + + *count = 6; + break; + } + +#endif case x86_SAVED_STATE32: { x86_saved_state32_t *state; @@ -1176,6 +1137,9 @@ machine_thread_get_state( if (*count < x86_SAVED_STATE32_COUNT) return(KERN_INVALID_ARGUMENT); + if (thread_is_64bit(thr_act)) + return(KERN_INVALID_ARGUMENT); + state = (x86_saved_state32_t *) tstate; saved_state = USER_REGS32(thr_act); @@ -1200,6 +1164,9 @@ machine_thread_get_state( if (*count < x86_SAVED_STATE64_COUNT) return(KERN_INVALID_ARGUMENT); + if (!thread_is_64bit(thr_act)) + return(KERN_INVALID_ARGUMENT); + state = (x86_saved_state64_t *)tstate; saved_state = USER_REGS64(thr_act); @@ -1224,7 +1191,7 @@ machine_thread_get_state( *count = x86_FLOAT_STATE32_COUNT; - return fpu_get_fxstate(thr_act, tstate); + return fpu_get_fxstate(thr_act, tstate, flavor); } case x86_FLOAT_STATE64: @@ -1237,7 +1204,7 @@ machine_thread_get_state( *count = x86_FLOAT_STATE64_COUNT; - return fpu_get_fxstate(thr_act, tstate); + return fpu_get_fxstate(thr_act, tstate, flavor); } case x86_FLOAT_STATE: @@ -1258,20 +1225,79 @@ machine_thread_get_state( state->fsh.flavor = x86_FLOAT_STATE64; state->fsh.count = x86_FLOAT_STATE64_COUNT; - kret = fpu_get_fxstate(thr_act, (thread_state_t)&state->ufs.fs64); + kret = fpu_get_fxstate(thr_act, (thread_state_t)&state->ufs.fs64, x86_FLOAT_STATE64); } else { state->fsh.flavor = x86_FLOAT_STATE32; state->fsh.count = x86_FLOAT_STATE32_COUNT; - kret = fpu_get_fxstate(thr_act, (thread_state_t)&state->ufs.fs32); + kret = fpu_get_fxstate(thr_act, (thread_state_t)&state->ufs.fs32, x86_FLOAT_STATE32); } *count = x86_FLOAT_STATE_COUNT; return(kret); } + case x86_AVX_STATE32: +#if !defined(RC_HIDE_XNU_J137) + case x86_AVX512_STATE32: +#endif + { + if (*count != _MachineStateCount[flavor]) + return(KERN_INVALID_ARGUMENT); + + if (thread_is_64bit(thr_act)) + return(KERN_INVALID_ARGUMENT); + + *count = _MachineStateCount[flavor]; + + return fpu_get_fxstate(thr_act, tstate, flavor); + } + + case x86_AVX_STATE64: +#if !defined(RC_HIDE_XNU_J137) + case x86_AVX512_STATE64: +#endif + { + if (*count != _MachineStateCount[flavor]) + return(KERN_INVALID_ARGUMENT); + + if ( !thread_is_64bit(thr_act)) + return(KERN_INVALID_ARGUMENT); + + *count = _MachineStateCount[flavor]; + + return fpu_get_fxstate(thr_act, tstate, flavor); + } + + case x86_AVX_STATE: +#if !defined(RC_HIDE_XNU_J137) + case x86_AVX512_STATE: +#endif + { + x86_avx_state_t *state; + thread_state_t fstate; + + if (*count < _MachineStateCount[flavor]) + return(KERN_INVALID_ARGUMENT); + + *count = _MachineStateCount[flavor]; + state = (x86_avx_state_t *)tstate; + + bzero((char *)state, *count * sizeof(int)); + + if (thread_is_64bit(thr_act)) { + flavor -= 1; /* 64-bit flavor */ + fstate = (thread_state_t) &state->ufs.as64; + } else { + flavor -= 2; /* 32-bit flavor */ + fstate = (thread_state_t) &state->ufs.as32; + } + state->ash.flavor = flavor; + state->ash.count = _MachineStateCount[flavor]; + + return fpu_get_fxstate(thr_act, fstate, flavor); + } - case OLD_i386_THREAD_STATE: case x86_THREAD_STATE32: { if (*count < x86_THREAD_STATE32_COUNT) @@ -1339,6 +1365,11 @@ machine_thread_get_state( *count = x86_EXCEPTION_STATE32_COUNT; get_exception_state32(thr_act, (x86_exception_state32_t *)tstate); + /* + * Suppress the cpu number for binary compatibility + * of this deprecated state. + */ + ((x86_exception_state32_t *)tstate)->cpu = 0; break; } @@ -1353,6 +1384,11 @@ machine_thread_get_state( *count = x86_EXCEPTION_STATE64_COUNT; get_exception_state64(thr_act, (x86_exception_state64_t *)tstate); + /* + * Suppress the cpu number for binary compatibility + * of this deprecated state. + */ + ((x86_exception_state64_t *)tstate)->cpu = 0; break; } @@ -1435,7 +1471,7 @@ machine_thread_get_state( *count = x86_DEBUG_STATE_COUNT; break; } - default: + default: return(KERN_INVALID_ARGUMENT); } @@ -1449,74 +1485,109 @@ machine_thread_get_kern_state( thread_state_t tstate, mach_msg_type_number_t *count) { + x86_saved_state_t *int_state = current_cpu_datap()->cpu_int_state; /* * This works only for an interrupted kernel thread */ - if (thread != current_thread() || current_cpu_datap()->cpu_int_state == NULL) + if (thread != current_thread() || int_state == NULL) return KERN_FAILURE; - switch(flavor) { - - case x86_THREAD_STATE32: - { + switch (flavor) { + case x86_THREAD_STATE32: { + x86_thread_state32_t *state; + x86_saved_state32_t *saved_state; - x86_thread_state32_t *state; - x86_saved_state32_t *saved_state; + if (!is_saved_state32(int_state) || + *count < x86_THREAD_STATE32_COUNT) + return (KERN_INVALID_ARGUMENT); - if (*count < x86_THREAD_STATE32_COUNT) - return(KERN_INVALID_ARGUMENT); - - state = (x86_thread_state32_t *)tstate; + state = (x86_thread_state32_t *) tstate; - assert(is_saved_state32(current_cpu_datap()->cpu_int_state)); - saved_state = saved_state32(current_cpu_datap()->cpu_int_state); - /* - * General registers. - */ - state->eax = saved_state->eax; - state->ebx = saved_state->ebx; - state->ecx = saved_state->ecx; - state->edx = saved_state->edx; - state->edi = saved_state->edi; - state->esi = saved_state->esi; - state->ebp = saved_state->ebp; - state->esp = saved_state->uesp; - state->eflags = saved_state->efl; - state->eip = saved_state->eip; - state->cs = saved_state->cs; - state->ss = saved_state->ss; - state->ds = saved_state->ds & 0xffff; - state->es = saved_state->es & 0xffff; - state->fs = saved_state->fs & 0xffff; - state->gs = saved_state->gs & 0xffff; - - *count = x86_THREAD_STATE32_COUNT; - - return KERN_SUCCESS; - } - break; // for completeness + saved_state = saved_state32(int_state); + /* + * General registers. + */ + state->eax = saved_state->eax; + state->ebx = saved_state->ebx; + state->ecx = saved_state->ecx; + state->edx = saved_state->edx; + state->edi = saved_state->edi; + state->esi = saved_state->esi; + state->ebp = saved_state->ebp; + state->esp = saved_state->uesp; + state->eflags = saved_state->efl; + state->eip = saved_state->eip; + state->cs = saved_state->cs; + state->ss = saved_state->ss; + state->ds = saved_state->ds & 0xffff; + state->es = saved_state->es & 0xffff; + state->fs = saved_state->fs & 0xffff; + state->gs = saved_state->gs & 0xffff; + + *count = x86_THREAD_STATE32_COUNT; + + return KERN_SUCCESS; + } + + case x86_THREAD_STATE64: { + x86_thread_state64_t *state; + x86_saved_state64_t *saved_state; + + if (!is_saved_state64(int_state) || + *count < x86_THREAD_STATE64_COUNT) + return (KERN_INVALID_ARGUMENT); - case x86_THREAD_STATE: - { - // wrap a 32 bit thread state into a 32/64bit clean thread state - x86_thread_state_t *state; - x86_saved_state32_t *saved_state; + state = (x86_thread_state64_t *) tstate; - if(*count < x86_THREAD_STATE_COUNT) - return (KERN_INVALID_ARGUMENT); + saved_state = saved_state64(int_state); + /* + * General registers. + */ + state->rax = saved_state->rax; + state->rbx = saved_state->rbx; + state->rcx = saved_state->rcx; + state->rdx = saved_state->rdx; + state->rdi = saved_state->rdi; + state->rsi = saved_state->rsi; + state->rbp = saved_state->rbp; + state->rsp = saved_state->isf.rsp; + state->r8 = saved_state->r8; + state->r9 = saved_state->r9; + state->r10 = saved_state->r10; + state->r11 = saved_state->r11; + state->r12 = saved_state->r12; + state->r13 = saved_state->r13; + state->r14 = saved_state->r14; + state->r15 = saved_state->r15; + + state->rip = saved_state->isf.rip; + state->rflags = saved_state->isf.rflags; + state->cs = saved_state->isf.cs; + state->fs = saved_state->fs & 0xffff; + state->gs = saved_state->gs & 0xffff; + *count = x86_THREAD_STATE64_COUNT; - state = (x86_thread_state_t *)tstate; - assert(is_saved_state32(current_cpu_datap()->cpu_int_state)); - saved_state = saved_state32(current_cpu_datap()->cpu_int_state); + return KERN_SUCCESS; + } + + case x86_THREAD_STATE: { + x86_thread_state_t *state = NULL; + + if (*count < x86_THREAD_STATE_COUNT) + return (KERN_INVALID_ARGUMENT); + + state = (x86_thread_state_t *) tstate; + + if (is_saved_state32(int_state)) { + x86_saved_state32_t *saved_state = saved_state32(int_state); state->tsh.flavor = x86_THREAD_STATE32; state->tsh.count = x86_THREAD_STATE32_COUNT; - - /* + + /* * General registers. */ - state->uts.ts32.eax = saved_state->eax; state->uts.ts32.ebx = saved_state->ebx; state->uts.ts32.ecx = saved_state->ecx; @@ -1533,172 +1604,74 @@ machine_thread_get_kern_state( state->uts.ts32.es = saved_state->es & 0xffff; state->uts.ts32.fs = saved_state->fs & 0xffff; state->uts.ts32.gs = saved_state->gs & 0xffff; - - *count = x86_THREAD_STATE_COUNT; - return KERN_SUCCESS; - } - break; - } - return KERN_FAILURE; -} + } else if (is_saved_state64(int_state)) { + x86_saved_state64_t *saved_state = saved_state64(int_state); + state->tsh.flavor = x86_THREAD_STATE64; + state->tsh.count = x86_THREAD_STATE64_COUNT; -/* - * Initialize the machine-dependent state for a new thread. - */ -kern_return_t -machine_thread_create( - thread_t thread, - task_t task) -{ - pcb_t pcb = &thread->machine.xxx_pcb; - struct real_descriptor *ldtp; - pmap_paddr_t paddr; - x86_saved_state_t *iss; - - inval_copy_windows(thread); - - thread->machine.physwindow_pte = 0; - thread->machine.physwindow_busy = 0; - - if (task_has_64BitAddr(task)) { - x86_sframe64_t *sf64; - - sf64 = (x86_sframe64_t *)zalloc(iss_zone64); - - if (sf64 == NULL) - panic("iss_zone64"); - pcb->sf = (void *)sf64; - - bzero((char *)sf64, sizeof(x86_sframe64_t)); - - iss = (x86_saved_state_t *) &sf64->ssf; - iss->flavor = x86_SAVED_STATE64; - /* - * Guarantee that the bootstrapped thread will be in user - * mode. - */ - iss->ss_64.isf.rflags = EFL_USER_SET; - iss->ss_64.isf.cs = USER64_CS; - iss->ss_64.isf.ss = USER_DS; - iss->ss_64.fs = USER_DS; - iss->ss_64.gs = USER_DS; - } else { - if (cpu_mode_is64bit()) { - x86_sframe_compat32_t *sfc32; - - sfc32 = (x86_sframe_compat32_t *)zalloc(iss_zone32); - if (sfc32 == NULL) - panic("iss_zone32"); - pcb->sf = (void *)sfc32; - - bzero((char *)sfc32, sizeof(x86_sframe_compat32_t)); - - iss = (x86_saved_state_t *) &sfc32->ssf.iss32; - iss->flavor = x86_SAVED_STATE32; -#if DEBUG - { - x86_saved_state_compat32_t *xssc; - - xssc = (x86_saved_state_compat32_t *) iss; - xssc->pad_for_16byte_alignment[0] = 0x64326432; - xssc->pad_for_16byte_alignment[1] = 0x64326432; - } -#endif + /* + * General registers. + */ + state->uts.ts64.rax = saved_state->rax; + state->uts.ts64.rbx = saved_state->rbx; + state->uts.ts64.rcx = saved_state->rcx; + state->uts.ts64.rdx = saved_state->rdx; + state->uts.ts64.rdi = saved_state->rdi; + state->uts.ts64.rsi = saved_state->rsi; + state->uts.ts64.rbp = saved_state->rbp; + state->uts.ts64.rsp = saved_state->isf.rsp; + state->uts.ts64.r8 = saved_state->r8; + state->uts.ts64.r9 = saved_state->r9; + state->uts.ts64.r10 = saved_state->r10; + state->uts.ts64.r11 = saved_state->r11; + state->uts.ts64.r12 = saved_state->r12; + state->uts.ts64.r13 = saved_state->r13; + state->uts.ts64.r14 = saved_state->r14; + state->uts.ts64.r15 = saved_state->r15; + + state->uts.ts64.rip = saved_state->isf.rip; + state->uts.ts64.rflags = saved_state->isf.rflags; + state->uts.ts64.cs = saved_state->isf.cs; + state->uts.ts64.fs = saved_state->fs & 0xffff; + state->uts.ts64.gs = saved_state->gs & 0xffff; } else { - x86_sframe32_t *sf32; - - sf32 = (x86_sframe32_t *)zalloc(iss_zone32); - - if (sf32 == NULL) - panic("iss_zone32"); - pcb->sf = (void *)sf32; - - bzero((char *)sf32, sizeof(x86_sframe32_t)); - - iss = (x86_saved_state_t *) &sf32->ssf; - iss->flavor = x86_SAVED_STATE32; + panic("unknown thread state"); } - /* - * Guarantee that the bootstrapped thread will be in user - * mode. - */ - iss->ss_32.cs = USER_CS; - iss->ss_32.ss = USER_DS; - iss->ss_32.ds = USER_DS; - iss->ss_32.es = USER_DS; - iss->ss_32.fs = USER_DS; - iss->ss_32.gs = USER_DS; - iss->ss_32.efl = EFL_USER_SET; - } - pcb->iss = iss; - - thread->machine.pcb = pcb; - simple_lock_init(&pcb->lock, 0); - - ldtp = (struct real_descriptor *)pmap_index_to_virt(HIGH_FIXED_LDT_BEGIN); - pcb->cthread_desc = ldtp[sel_idx(USER_DS)]; - pcb->uldt_desc = ldtp[sel_idx(USER_DS)]; - pcb->uldt_selector = 0; - - pcb->iss_pte0 = (uint64_t)pte_kernel_rw(kvtophys((vm_offset_t)pcb->iss)); - - if (0 == (paddr = pa_to_pte(kvtophys((vm_offset_t)(pcb->iss) + PAGE_SIZE)))) - pcb->iss_pte1 = INTEL_PTE_INVALID; - else - pcb->iss_pte1 = (uint64_t)pte_kernel_rw(paddr); - - return(KERN_SUCCESS); -} - -/* - * Machine-dependent cleanup prior to destroying a thread - */ -void -machine_thread_destroy( - thread_t thread) -{ - register pcb_t pcb = thread->machine.pcb; - assert(pcb); - - if (pcb->ifps != 0) - fpu_free(pcb->ifps); - if (pcb->sf != 0) { - if (thread_is_64bit(thread)) - zfree(iss_zone64, pcb->sf); - else - zfree(iss_zone32, pcb->sf); - pcb->sf = 0; - } - if (pcb->ids) { - if (thread_is_64bit(thread)) - zfree(ids_zone64, pcb->ids); - else - zfree(ids_zone32, pcb->ids); + *count = x86_THREAD_STATE_COUNT; + return KERN_SUCCESS; + } } - thread->machine.pcb = (pcb_t)0; - + return KERN_FAILURE; } + void -machine_thread_switch_addrmode(thread_t thread, int oldmode_is64bit) +machine_thread_switch_addrmode(thread_t thread) { - register pcb_t pcb = thread->machine.pcb; + /* + * We don't want to be preempted until we're done + * - particularly if we're switching the current thread + */ + disable_preemption(); - assert(pcb); - - if (pcb->sf != 0) { - if (oldmode_is64bit) - zfree(iss_zone64, pcb->sf); - else - zfree(iss_zone32, pcb->sf); - } + /* + * Reset the state saveareas. As we're resetting, we anticipate no + * memory allocations in this path. + */ machine_thread_create(thread, thread->task); + /* Adjust FPU state */ + fpu_switch_addrmode(thread, task_has_64BitAddr(thread->task)); + /* If we're switching ourselves, reset the pcb addresses etc. */ - if (thread == current_thread()) - act_machine_switch_pcb(thread); + if (thread == current_thread()) { + boolean_t istate = ml_set_interrupts_enabled(FALSE); + act_machine_switch_pcb(NULL, thread); + ml_set_interrupts_enabled(istate); + } + enable_preemption(); } @@ -1708,48 +1681,11 @@ machine_thread_switch_addrmode(thread_t thread, int oldmode_is64bit) * when starting up a new processor */ void -machine_set_current_thread( thread_t thread ) +machine_set_current_thread(thread_t thread) { current_cpu_datap()->cpu_active_thread = thread; } -/* - * This is called when a task is termianted. - */ -void -machine_thread_terminate_self(void) -{ - task_t self_task = current_task(); - if (self_task) { - user_ldt_t user_ldt = self_task->i386_ldt; - if (user_ldt != 0) { - self_task->i386_ldt = 0; - user_ldt_free(user_ldt); - } - } -} - -void -act_machine_return(int code) -{ - /* - * This code is called with nothing locked. - * It also returns with nothing locked, if it returns. - * - * This routine terminates the current thread activation. - * If this is the only activation associated with its - * thread shuttle, then the entire thread (shuttle plus - * activation) is terminated. - */ - assert( code == KERN_TERMINATED ); - - thread_terminate_self(); - - /*NOTREACHED*/ - - panic("act_machine_return(%d): TALKING ZOMBIE! (1)", code); -} - /* * Perform machine-dependent per-thread initializations @@ -1757,139 +1693,26 @@ act_machine_return(int code) void machine_thread_init(void) { - if (cpu_mode_is64bit()) { - iss_zone64 = zinit(sizeof(x86_sframe64_t), - THREAD_MAX * sizeof(x86_sframe64_t), - THREAD_CHUNK * sizeof(x86_sframe64_t), - "x86_64 saved state"); - - assert(sizeof(x86_sframe_compat32_t) % 16 == 0); - iss_zone32 = zinit(sizeof(x86_sframe_compat32_t), - THREAD_MAX * sizeof(x86_sframe_compat32_t), - THREAD_CHUNK * sizeof(x86_sframe_compat32_t), - "x86_32 saved state"); - - ids_zone32 = zinit(sizeof(x86_debug_state32_t), - THREAD_MAX * (sizeof(x86_debug_state32_t)), - THREAD_CHUNK * (sizeof(x86_debug_state32_t)), - "x86_32 debug state"); - ids_zone64 = zinit(sizeof(x86_debug_state64_t), - THREAD_MAX * sizeof(x86_debug_state64_t), - THREAD_CHUNK * sizeof(x86_debug_state64_t), - "x86_64 debug state"); - - } else { - iss_zone32 = zinit(sizeof(x86_sframe32_t), - THREAD_MAX * sizeof(x86_sframe32_t), - THREAD_CHUNK * sizeof(x86_sframe32_t), - "x86 saved state"); - ids_zone32 = zinit(sizeof(x86_debug_state32_t), - THREAD_MAX * (sizeof(x86_debug_state32_t)), - THREAD_CHUNK * (sizeof(x86_debug_state32_t)), - "x86 debug state"); - } - fpu_module_init(); - iopb_init(); -} - -/* - * Some routines for debugging activation code - */ -static void dump_handlers(thread_t); -void dump_regs(thread_t); -int dump_act(thread_t thr_act); - -static void -dump_handlers(thread_t thr_act) -{ - ReturnHandler *rhp = thr_act->handlers; - int counter = 0; - - printf("\t"); - while (rhp) { - if (rhp == &thr_act->special_handler){ - if (rhp->next) - printf("[NON-Zero next ptr(%x)]", rhp->next); - printf("special_handler()->"); - break; - } - printf("hdlr_%d(%x)->",counter,rhp->handler); - rhp = rhp->next; - if (++counter > 32) { - printf("Aborting: HUGE handler chain\n"); - break; - } - } - printf("HLDR_NULL\n"); -} - -void -dump_regs(thread_t thr_act) -{ - if (thr_act->machine.pcb == NULL) - return; - - if (thread_is_64bit(thr_act)) { - x86_saved_state64_t *ssp; - - ssp = USER_REGS64(thr_act); - - panic("dump_regs: 64bit tasks not yet supported"); - - } else { - x86_saved_state32_t *ssp; + iss_zone = zinit(sizeof(x86_saved_state_t), + thread_max * sizeof(x86_saved_state_t), + THREAD_CHUNK * sizeof(x86_saved_state_t), + "x86_64 saved state"); - ssp = USER_REGS32(thr_act); + ids_zone = zinit(sizeof(x86_debug_state64_t), + thread_max * sizeof(x86_debug_state64_t), + THREAD_CHUNK * sizeof(x86_debug_state64_t), + "x86_64 debug state"); - /* - * Print out user register state - */ - printf("\tRegs:\tedi=%x esi=%x ebp=%x ebx=%x edx=%x\n", - ssp->edi, ssp->esi, ssp->ebp, ssp->ebx, ssp->edx); - - printf("\t\tecx=%x eax=%x eip=%x efl=%x uesp=%x\n", - ssp->ecx, ssp->eax, ssp->eip, ssp->efl, ssp->uesp); - - printf("\t\tcs=%x ss=%x\n", ssp->cs, ssp->ss); - } + fpu_module_init(); } -int -dump_act(thread_t thr_act) -{ - if (!thr_act) - return(0); - - printf("thread(0x%x)(%d): task=%x(%d)\n", - thr_act, thr_act->ref_count, - thr_act->task, thr_act->task ? thr_act->task->ref_count : 0); - printf("\tsusp=%d user_stop=%d active=%x ast=%x\n", - thr_act->suspend_count, thr_act->user_stop_count, - thr_act->active, thr_act->ast); - printf("\tpcb=%x\n", thr_act->machine.pcb); - - if (thr_act->kernel_stack) { - vm_offset_t stack = thr_act->kernel_stack; - - printf("\tk_stk %x eip %x ebx %x esp %x iss %x\n", - stack, STACK_IKS(stack)->k_eip, STACK_IKS(stack)->k_ebx, - STACK_IKS(stack)->k_esp, STACK_IEL(stack)->saved_state); - } - - dump_handlers(thr_act); - dump_regs(thr_act); - return((int)thr_act); -} user_addr_t get_useraddr(void) { thread_t thr_act = current_thread(); - if (thr_act->machine.pcb == NULL) - return (0); - if (thread_is_64bit(thr_act)) { x86_saved_state64_t *iss64; @@ -1915,7 +1738,7 @@ machine_stack_detach(thread_t thread) vm_offset_t stack; KERNEL_DEBUG(MACHDBG_CODE(DBG_MACH_SCHED, MACH_STACK_DETACH), - thread, thread->priority, + (uintptr_t)thread_tid(thread), thread->priority, thread->sched_pri, 0, 0); @@ -1934,19 +1757,26 @@ machine_stack_attach( thread_t thread, vm_offset_t stack) { - struct x86_kernel_state32 *statep; + struct x86_kernel_state *statep; KERNEL_DEBUG(MACHDBG_CODE(DBG_MACH_SCHED, MACH_STACK_ATTACH), - thread, thread->priority, + (uintptr_t)thread_tid(thread), thread->priority, thread->sched_pri, 0, 0); assert(stack); thread->kernel_stack = stack; + thread_initialize_kernel_state(thread); statep = STACK_IKS(stack); +#if defined(__x86_64__) + statep->k_rip = (unsigned long) Thread_continue; + statep->k_rbx = (unsigned long) thread_continue; + statep->k_rsp = (unsigned long) STACK_IKS(stack); +#else statep->k_eip = (unsigned long) Thread_continue; statep->k_ebx = (unsigned long) thread_continue; - statep->k_esp = (unsigned long) STACK_IEL(stack); + statep->k_esp = (unsigned long) STACK_IKS(stack); +#endif return; } @@ -1964,6 +1794,8 @@ machine_stack_handoff(thread_t old, assert(new); assert(old); + kpc_off_cpu(old); + stack = old->kernel_stack; if (stack == old->reserved_stack) { assert(new->reserved_stack); @@ -1977,19 +1809,20 @@ machine_stack_handoff(thread_t old, */ new->kernel_stack = stack; - fpu_save_context(old); - + fpu_switch_context(old, new); + old->machine.specFlags &= ~OnProc; new->machine.specFlags |= OnProc; - PMAP_SWITCH_CONTEXT(old, new, cpu_number()); - act_machine_switch_pcb(new); + pmap_switch_context(old, new, cpu_number()); + act_machine_switch_pcb(old, new); - KERNEL_DEBUG_CONSTANT( - MACHDBG_CODE(DBG_MACH_SCHED, MACH_STACK_HANDOFF)|DBG_FUNC_NONE, - old->reason, (int)new, old->sched_pri, new->sched_pri, 0); +#if HYPERVISOR + ml_hv_cswitch(old, new); +#endif machine_set_current_thread(new); + thread_initialize_kernel_state(new); return; } @@ -2014,31 +1847,30 @@ struct x86_act_context64 { void * act_thread_csave(void) { - kern_return_t kret; + kern_return_t kret; mach_msg_type_number_t val; - thread_t thr_act = current_thread(); + thread_t thr_act = current_thread(); - if (thread_is_64bit(thr_act)) { - struct x86_act_context64 *ic64; + if (thread_is_64bit(thr_act)) { + struct x86_act_context64 *ic64; - ic64 = (struct x86_act_context64 *)kalloc(sizeof(struct x86_act_context64)); + ic64 = (struct x86_act_context64 *)kalloc(sizeof(struct x86_act_context64)); if (ic64 == (struct x86_act_context64 *)NULL) - return((void *)0); + return((void *)0); val = x86_SAVED_STATE64_COUNT; kret = machine_thread_get_state(thr_act, x86_SAVED_STATE64, - (thread_state_t) &ic64->ss, &val); + (thread_state_t) &ic64->ss, &val); if (kret != KERN_SUCCESS) { - kfree(ic64, sizeof(struct x86_act_context64)); + kfree(ic64, sizeof(struct x86_act_context64)); return((void *)0); } val = x86_FLOAT_STATE64_COUNT; kret = machine_thread_get_state(thr_act, x86_FLOAT_STATE64, - (thread_state_t) &ic64->fs, &val); - + (thread_state_t) &ic64->fs, &val); if (kret != KERN_SUCCESS) { - kfree(ic64, sizeof(struct x86_act_context64)); + kfree(ic64, sizeof(struct x86_act_context64)); return((void *)0); } @@ -2054,25 +1886,25 @@ act_thread_csave(void) return(ic64); } else { - struct x86_act_context32 *ic32; + struct x86_act_context32 *ic32; - ic32 = (struct x86_act_context32 *)kalloc(sizeof(struct x86_act_context32)); + ic32 = (struct x86_act_context32 *)kalloc(sizeof(struct x86_act_context32)); if (ic32 == (struct x86_act_context32 *)NULL) - return((void *)0); + return((void *)0); val = x86_SAVED_STATE32_COUNT; kret = machine_thread_get_state(thr_act, x86_SAVED_STATE32, - (thread_state_t) &ic32->ss, &val); + (thread_state_t) &ic32->ss, &val); if (kret != KERN_SUCCESS) { - kfree(ic32, sizeof(struct x86_act_context32)); + kfree(ic32, sizeof(struct x86_act_context32)); return((void *)0); } val = x86_FLOAT_STATE32_COUNT; kret = machine_thread_get_state(thr_act, x86_FLOAT_STATE32, - (thread_state_t) &ic32->fs, &val); + (thread_state_t) &ic32->fs, &val); if (kret != KERN_SUCCESS) { - kfree(ic32, sizeof(struct x86_act_context32)); + kfree(ic32, sizeof(struct x86_act_context32)); return((void *)0); } @@ -2097,7 +1929,7 @@ act_thread_catt(void *ctx) kern_return_t kret; if (ctx == (void *)NULL) - return; + return; if (thread_is_64bit(thr_act)) { struct x86_act_context64 *ic64; @@ -2107,8 +1939,8 @@ act_thread_catt(void *ctx) kret = machine_thread_set_state(thr_act, x86_SAVED_STATE64, (thread_state_t) &ic64->ss, x86_SAVED_STATE64_COUNT); if (kret == KERN_SUCCESS) { - machine_thread_set_state(thr_act, x86_FLOAT_STATE64, - (thread_state_t) &ic64->fs, x86_FLOAT_STATE64_COUNT); + machine_thread_set_state(thr_act, x86_FLOAT_STATE64, + (thread_state_t) &ic64->fs, x86_FLOAT_STATE64_COUNT); } kfree(ic64, sizeof(struct x86_act_context64)); } else { @@ -2119,13 +1951,8 @@ act_thread_catt(void *ctx) kret = machine_thread_set_state(thr_act, x86_SAVED_STATE32, (thread_state_t) &ic32->ss, x86_SAVED_STATE32_COUNT); if (kret == KERN_SUCCESS) { - kret = machine_thread_set_state(thr_act, x86_FLOAT_STATE32, + (void) machine_thread_set_state(thr_act, x86_FLOAT_STATE32, (thread_state_t) &ic32->fs, x86_FLOAT_STATE32_COUNT); - if (kret == KERN_SUCCESS && thr_act->machine.pcb->ids) - machine_thread_set_state(thr_act, - x86_DEBUG_STATE32, - (thread_state_t)&ic32->ds, - x86_DEBUG_STATE32_COUNT); } kfree(ic32, sizeof(struct x86_act_context32)); } @@ -2136,3 +1963,53 @@ void act_thread_cfree(__unused void *ctx) { /* XXX - Unused */ } + +/* + * Duplicate one x86_debug_state32_t to another. "all" parameter + * chooses whether dr4 and dr5 are copied (they are never meant + * to be installed when we do machine_task_set_state() or + * machine_thread_set_state()). + */ +void +copy_debug_state32( + x86_debug_state32_t *src, + x86_debug_state32_t *target, + boolean_t all) +{ + if (all) { + target->dr4 = src->dr4; + target->dr5 = src->dr5; + } + + target->dr0 = src->dr0; + target->dr1 = src->dr1; + target->dr2 = src->dr2; + target->dr3 = src->dr3; + target->dr6 = src->dr6; + target->dr7 = src->dr7; +} + +/* + * Duplicate one x86_debug_state64_t to another. "all" parameter + * chooses whether dr4 and dr5 are copied (they are never meant + * to be installed when we do machine_task_set_state() or + * machine_thread_set_state()). + */ +void +copy_debug_state64( + x86_debug_state64_t *src, + x86_debug_state64_t *target, + boolean_t all) +{ + if (all) { + target->dr4 = src->dr4; + target->dr5 = src->dr5; + } + + target->dr0 = src->dr0; + target->dr1 = src->dr1; + target->dr2 = src->dr2; + target->dr3 = src->dr3; + target->dr6 = src->dr6; + target->dr7 = src->dr7; +}