1/*
2 * Copyright (c) 2000-2012 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28/*
29 * @OSF_COPYRIGHT@
30 *
31 */
32
33#ifndef	I386_CPU_DATA
34#define I386_CPU_DATA
35
36#include <mach_assert.h>
37
38#include <kern/assert.h>
39#include <kern/kern_types.h>
40#include <kern/queue.h>
41#include <kern/processor.h>
42#include <kern/pms.h>
43#include <pexpert/pexpert.h>
44#include <mach/i386/thread_status.h>
45#include <mach/i386/vm_param.h>
46#include <i386/rtclock_protos.h>
47#include <i386/pmCPU.h>
48#include <i386/cpu_topology.h>
49
50#if CONFIG_VMX
51#include <i386/vmx/vmx_cpu.h>
52#endif
53
54#include <machine/pal_routines.h>
55
56/*
57 * Data structures referenced (anonymously) from per-cpu data:
58 */
59struct cpu_cons_buffer;
60struct cpu_desc_table;
61struct mca_state;
62
63/*
64 * Data structures embedded in per-cpu data:
65 */
66typedef struct rtclock_timer {
67	mpqueue_head_t		queue;
68	uint64_t		deadline;
69	uint64_t		when_set;
70	boolean_t		has_expired;
71} rtclock_timer_t;
72
73
74typedef struct {
75	struct x86_64_tss	*cdi_ktss;
76	struct __attribute__((packed)) {
77		uint16_t size;
78		void *ptr;
79	} cdi_gdt, cdi_idt;
80	struct fake_descriptor	*cdi_ldt;
81	vm_offset_t		cdi_sstk;
82} cpu_desc_index_t;
83
84typedef enum {
85	TASK_MAP_32BIT,			/* 32-bit user, compatibility mode */
86	TASK_MAP_64BIT,			/* 64-bit user thread, shared space */
87} task_map_t;
88
89
90/*
91 * This structure is used on entry into the (uber-)kernel on syscall from
92 * a 64-bit user. It contains the address of the machine state save area
93 * for the current thread and a temporary place to save the user's rsp
94 * before loading this address into rsp.
95 */
96typedef struct {
97	addr64_t	cu_isf;		/* thread->pcb->iss.isf */
98	uint64_t	cu_tmp;		/* temporary scratch */
99	addr64_t	cu_user_gs_base;
100} cpu_uber_t;
101
102typedef	uint16_t	pcid_t;
103typedef	uint8_t		pcid_ref_t;
104
105#define CPU_RTIME_BINS (12)
106#define CPU_ITIME_BINS (CPU_RTIME_BINS)
107
108/*
109 * Per-cpu data.
110 *
111 * Each processor has a per-cpu data area which is dereferenced through the
112 * current_cpu_datap() macro. For speed, the %gs segment is based here, and
113 * using this, inlines provides single-instruction access to frequently used
114 * members - such as get_cpu_number()/cpu_number(), and get_active_thread()/
115 * current_thread().
116 *
117 * Cpu data owned by another processor can be accessed using the
118 * cpu_datap(cpu_number) macro which uses the cpu_data_ptr[] array of per-cpu
119 * pointers.
120 */
121typedef struct cpu_data
122{
123	struct pal_cpu_data	cpu_pal_data;		/* PAL-specific data */
124#define				cpu_pd cpu_pal_data	/* convenience alias */
125	struct cpu_data		*cpu_this;		/* pointer to myself */
126	thread_t		cpu_active_thread;
127	thread_t		cpu_nthread;
128	volatile int		cpu_preemption_level;
129	int			cpu_number;		/* Logical CPU */
130	void			*cpu_int_state;		/* interrupt state */
131	vm_offset_t		cpu_active_stack;	/* kernel stack base */
132	vm_offset_t		cpu_kernel_stack;	/* kernel stack top */
133	vm_offset_t		cpu_int_stack_top;
134	int			cpu_interrupt_level;
135	int			cpu_phys_number;	/* Physical CPU */
136	cpu_id_t		cpu_id;			/* Platform Expert */
137	volatile int		cpu_signals;		/* IPI events */
138	volatile int		cpu_prior_signals;	/* Last set of events,
139							 * debugging
140							 */
141	ast_t			cpu_pending_ast;
142	volatile int		cpu_running;
143	boolean_t		cpu_fixed_pmcs_enabled;
144	rtclock_timer_t		rtclock_timer;
145	volatile addr64_t	cpu_active_cr3 __attribute((aligned(64)));
146	union {
147		volatile uint32_t cpu_tlb_invalid;
148		struct {
149			volatile uint16_t cpu_tlb_invalid_local;
150			volatile uint16_t cpu_tlb_invalid_global;
151		};
152	};
153	volatile task_map_t	cpu_task_map;
154	volatile addr64_t	cpu_task_cr3;
155	addr64_t		cpu_kernel_cr3;
156	cpu_uber_t		cpu_uber;
157	void			*cpu_chud;
158	void			*cpu_console_buf;
159	struct x86_lcpu		lcpu;
160	struct processor	*cpu_processor;
161#if NCOPY_WINDOWS > 0
162	struct cpu_pmap		*cpu_pmap;
163#endif
164	struct cpu_desc_table	*cpu_desc_tablep;
165	struct fake_descriptor	*cpu_ldtp;
166	cpu_desc_index_t	cpu_desc_index;
167	int			cpu_ldt;
168#if NCOPY_WINDOWS > 0
169	vm_offset_t		cpu_copywindow_base;
170	uint64_t		*cpu_copywindow_pdp;
171
172	vm_offset_t		cpu_physwindow_base;
173	uint64_t		*cpu_physwindow_ptep;
174#endif
175
176#define HWINTCNT_SIZE 256
177	uint32_t		cpu_hwIntCnt[HWINTCNT_SIZE];	/* Interrupt counts */
178 	uint64_t		cpu_hwIntpexits[HWINTCNT_SIZE];
179	uint64_t		cpu_hwIntcexits[HWINTCNT_SIZE];
180	uint64_t		cpu_dr7; /* debug control register */
181	uint64_t		cpu_int_event_time;	/* intr entry/exit time */
182	pal_rtc_nanotime_t	*cpu_nanotime;		/* Nanotime info */
183#if	CONFIG_COUNTERS
184	thread_t		csw_old_thread;
185	thread_t		csw_new_thread;
186#endif /* CONFIG COUNTERS */
187#if KPC
188	/* double-buffered performance counter data */
189	uint64_t                *cpu_kpc_buf[2];
190	/* PMC shadow and reload value buffers */
191	uint64_t                *cpu_kpc_shadow;
192	uint64_t                *cpu_kpc_reload;
193#endif
194	uint32_t		cpu_pmap_pcid_enabled;
195	pcid_t			cpu_active_pcid;
196	pcid_t			cpu_last_pcid;
197	volatile pcid_ref_t	*cpu_pmap_pcid_coherentp;
198	volatile pcid_ref_t	*cpu_pmap_pcid_coherentp_kernel;
199#define	PMAP_PCID_MAX_PCID      (0x1000)
200	pcid_t			cpu_pcid_free_hint;
201	pcid_ref_t		cpu_pcid_refcounts[PMAP_PCID_MAX_PCID];
202	pmap_t			cpu_pcid_last_pmap_dispatched[PMAP_PCID_MAX_PCID];
203#ifdef	PCID_STATS
204	uint64_t		cpu_pmap_pcid_flushes;
205	uint64_t		cpu_pmap_pcid_preserves;
206#endif
207	uint64_t		cpu_aperf;
208	uint64_t		cpu_mperf;
209	uint64_t		cpu_c3res;
210	uint64_t		cpu_c6res;
211	uint64_t		cpu_c7res;
212	uint64_t		cpu_itime_total;
213	uint64_t		cpu_rtime_total;
214	uint64_t		cpu_ixtime;
215	uint64_t                cpu_idle_exits;
216 	uint64_t		cpu_rtimes[CPU_RTIME_BINS];
217 	uint64_t		cpu_itimes[CPU_ITIME_BINS];
218 	uint64_t		cpu_cur_insns;
219 	uint64_t		cpu_cur_ucc;
220 	uint64_t		cpu_cur_urc;
221	uint64_t                cpu_max_observed_int_latency;
222	int                     cpu_max_observed_int_latency_vector;
223	volatile boolean_t	cpu_NMI_acknowledged;
224	uint64_t		debugger_entry_time;
225	uint64_t		debugger_ipi_time;
226	/* A separate nested interrupt stack flag, to account
227	 * for non-nested interrupts arriving while on the interrupt stack
228	 * Currently only occurs when AICPM enables interrupts on the
229	 * interrupt stack during processor offlining.
230	 */
231	uint32_t		cpu_nested_istack;
232	uint32_t		cpu_nested_istack_events;
233	x86_saved_state64_t	*cpu_fatal_trap_state;
234	x86_saved_state64_t	*cpu_post_fatal_trap_state;
235#if CONFIG_VMX
236	vmx_cpu_t		cpu_vmx;		/* wonderful world of virtualization */
237#endif
238#if CONFIG_MCA
239	struct mca_state	*cpu_mca_state;		/* State at MC fault */
240#endif
241 	int			cpu_type;
242 	int			cpu_subtype;
243 	int			cpu_threadtype;
244 	boolean_t		cpu_iflag;
245 	boolean_t		cpu_boot_complete;
246 	int			cpu_hibernate;
247} cpu_data_t;
248
249extern cpu_data_t	*cpu_data_ptr[];
250
251/* Macro to generate inline bodies to retrieve per-cpu data fields. */
252#if defined(__clang__)
253#define GS_RELATIVE volatile __attribute__((address_space(256)))
254#ifndef offsetof
255#define offsetof(TYPE,MEMBER) __builtin_offsetof(TYPE,MEMBER)
256#endif
257
258#define CPU_DATA_GET(member,type)										\
259	cpu_data_t GS_RELATIVE *cpu_data =							\
260		(cpu_data_t GS_RELATIVE *)0UL;									\
261	type ret;															\
262	ret = cpu_data->member;												\
263	return ret;
264
265#define CPU_DATA_GET_INDEX(member,index,type)							\
266	cpu_data_t GS_RELATIVE *cpu_data =							\
267		(cpu_data_t GS_RELATIVE *)0UL;									\
268	type ret;															\
269	ret = cpu_data->member[index];										\
270	return ret;
271
272#define CPU_DATA_SET(member,value)										\
273	cpu_data_t GS_RELATIVE *cpu_data =							\
274		(cpu_data_t GS_RELATIVE *)0UL;									\
275	cpu_data->member = value;
276
277#define CPU_DATA_XCHG(member,value,type)								\
278	cpu_data_t GS_RELATIVE *cpu_data =							\
279		(cpu_data_t GS_RELATIVE *)0UL;									\
280	type ret;															\
281	ret = cpu_data->member;												\
282	cpu_data->member = value;											\
283	return ret;
284
285#else /* !defined(__clang__) */
286
287#ifndef offsetof
288#define offsetof(TYPE,MEMBER) ((size_t) &((TYPE *)0)->MEMBER)
289#endif /* offsetof */
290#define CPU_DATA_GET(member,type)					\
291	type ret;							\
292	__asm__ volatile ("mov %%gs:%P1,%0"				\
293		: "=r" (ret)						\
294		: "i" (offsetof(cpu_data_t,member)));			\
295	return ret;
296
297#define CPU_DATA_GET_INDEX(member,index,type)	\
298	type ret;							\
299	__asm__ volatile ("mov %%gs:(%1),%0"				\
300		: "=r" (ret)						\
301		: "r" (offsetof(cpu_data_t,member[index])));			\
302	return ret;
303
304#define CPU_DATA_SET(member,value)					\
305	__asm__ volatile ("mov %0,%%gs:%P1"				\
306		:							\
307		: "r" (value), "i" (offsetof(cpu_data_t,member)));
308
309#define CPU_DATA_XCHG(member,value,type)				\
310	type ret;							\
311	__asm__ volatile ("xchg %0,%%gs:%P1"				\
312		: "=r" (ret)						\
313		: "i" (offsetof(cpu_data_t,member)), "0" (value));	\
314	return ret;
315
316#endif /* !defined(__clang__) */
317
318/*
319 * Everyone within the osfmk part of the kernel can use the fast
320 * inline versions of these routines.  Everyone outside, must call
321 * the real thing,
322 */
323static inline thread_t
324get_active_thread(void)
325{
326	CPU_DATA_GET(cpu_active_thread,thread_t)
327}
328#define current_thread_fast()		get_active_thread()
329#define current_thread()		current_thread_fast()
330
331#define cpu_mode_is64bit()		TRUE
332
333static inline int
334get_preemption_level(void)
335{
336	CPU_DATA_GET(cpu_preemption_level,int)
337}
338static inline int
339get_interrupt_level(void)
340{
341	CPU_DATA_GET(cpu_interrupt_level,int)
342}
343static inline int
344get_cpu_number(void)
345{
346	CPU_DATA_GET(cpu_number,int)
347}
348static inline int
349get_cpu_phys_number(void)
350{
351	CPU_DATA_GET(cpu_phys_number,int)
352}
353
354
355static inline void
356disable_preemption(void)
357{
358#if defined(__clang__)
359	cpu_data_t GS_RELATIVE *cpu_data = (cpu_data_t GS_RELATIVE *)0UL;
360	cpu_data->cpu_preemption_level++;
361#else
362	__asm__ volatile ("incl %%gs:%P0"
363			:
364			: "i" (offsetof(cpu_data_t, cpu_preemption_level)));
365#endif
366}
367
368static inline void
369enable_preemption(void)
370{
371	assert(get_preemption_level() > 0);
372
373#if defined(__clang__)
374	cpu_data_t GS_RELATIVE *cpu_data = (cpu_data_t GS_RELATIVE *)0UL;
375	if (0 == --cpu_data->cpu_preemption_level)
376		kernel_preempt_check();
377#else
378	__asm__ volatile ("decl %%gs:%P0		\n\t"
379			  "jne 1f			\n\t"
380			  "call _kernel_preempt_check	\n\t"
381			  "1:"
382			: /* no outputs */
383			: "i" (offsetof(cpu_data_t, cpu_preemption_level))
384			: "eax", "ecx", "edx", "cc", "memory");
385#endif
386}
387
388static inline void
389enable_preemption_no_check(void)
390{
391	assert(get_preemption_level() > 0);
392
393#if defined(__clang__)
394	cpu_data_t GS_RELATIVE *cpu_data = (cpu_data_t GS_RELATIVE *)0UL;
395	cpu_data->cpu_preemption_level--;
396#else
397	__asm__ volatile ("decl %%gs:%P0"
398			: /* no outputs */
399			: "i" (offsetof(cpu_data_t, cpu_preemption_level))
400			: "cc", "memory");
401#endif
402}
403
404static inline void
405mp_disable_preemption(void)
406{
407	disable_preemption();
408}
409
410static inline void
411mp_enable_preemption(void)
412{
413	enable_preemption();
414}
415
416static inline void
417mp_enable_preemption_no_check(void)
418{
419	enable_preemption_no_check();
420}
421
422static inline cpu_data_t *
423current_cpu_datap(void)
424{
425	CPU_DATA_GET(cpu_this, cpu_data_t *);
426}
427
428static inline cpu_data_t *
429cpu_datap(int cpu)
430{
431	return cpu_data_ptr[cpu];
432}
433
434extern cpu_data_t *cpu_data_alloc(boolean_t is_boot_cpu);
435extern void cpu_data_realloc(void);
436
437#endif	/* I386_CPU_DATA */
438