xref: /openbmc/linux/arch/x86/include/asm/thread_info.h (revision bc5aa3a0)
1 /* thread_info.h: low-level thread information
2  *
3  * Copyright (C) 2002  David Howells (dhowells@redhat.com)
4  * - Incorporating suggestions made by Linus Torvalds and Dave Miller
5  */
6 
7 #ifndef _ASM_X86_THREAD_INFO_H
8 #define _ASM_X86_THREAD_INFO_H
9 
10 #include <linux/compiler.h>
11 #include <asm/page.h>
12 #include <asm/percpu.h>
13 #include <asm/types.h>
14 
15 /*
16  * TOP_OF_KERNEL_STACK_PADDING is a number of unused bytes that we
17  * reserve at the top of the kernel stack.  We do it because of a nasty
18  * 32-bit corner case.  On x86_32, the hardware stack frame is
19  * variable-length.  Except for vm86 mode, struct pt_regs assumes a
20  * maximum-length frame.  If we enter from CPL 0, the top 8 bytes of
21  * pt_regs don't actually exist.  Ordinarily this doesn't matter, but it
22  * does in at least one case:
23  *
24  * If we take an NMI early enough in SYSENTER, then we can end up with
25  * pt_regs that extends above sp0.  On the way out, in the espfix code,
26  * we can read the saved SS value, but that value will be above sp0.
27  * Without this offset, that can result in a page fault.  (We are
28  * careful that, in this case, the value we read doesn't matter.)
29  *
30  * In vm86 mode, the hardware frame is much longer still, so add 16
31  * bytes to make room for the real-mode segments.
32  *
33  * x86_64 has a fixed-length stack frame.
34  */
35 #ifdef CONFIG_X86_32
36 # ifdef CONFIG_VM86
37 #  define TOP_OF_KERNEL_STACK_PADDING 16
38 # else
39 #  define TOP_OF_KERNEL_STACK_PADDING 8
40 # endif
41 #else
42 # define TOP_OF_KERNEL_STACK_PADDING 0
43 #endif
44 
45 /*
46  * low level task data that entry.S needs immediate access to
47  * - this struct should fit entirely inside of one cache line
48  * - this struct shares the supervisor stack pages
49  */
50 #ifndef __ASSEMBLY__
51 struct task_struct;
52 #include <asm/cpufeature.h>
53 #include <linux/atomic.h>
54 
55 struct thread_info {
56 	struct task_struct	*task;		/* main task structure */
57 	__u32			flags;		/* low level flags */
58 	__u32			status;		/* thread synchronous flags */
59 	__u32			cpu;		/* current CPU */
60 };
61 
62 #define INIT_THREAD_INFO(tsk)			\
63 {						\
64 	.task		= &tsk,			\
65 	.flags		= 0,			\
66 	.cpu		= 0,			\
67 }
68 
69 #define init_thread_info	(init_thread_union.thread_info)
70 #define init_stack		(init_thread_union.stack)
71 
72 #else /* !__ASSEMBLY__ */
73 
74 #include <asm/asm-offsets.h>
75 
76 #endif
77 
78 /*
79  * thread information flags
80  * - these are process state flags that various assembly files
81  *   may need to access
82  * - pending work-to-be-done flags are in LSW
83  * - other flags in MSW
84  * Warning: layout of LSW is hardcoded in entry.S
85  */
86 #define TIF_SYSCALL_TRACE	0	/* syscall trace active */
87 #define TIF_NOTIFY_RESUME	1	/* callback before returning to user */
88 #define TIF_SIGPENDING		2	/* signal pending */
89 #define TIF_NEED_RESCHED	3	/* rescheduling necessary */
90 #define TIF_SINGLESTEP		4	/* reenable singlestep on user return*/
91 #define TIF_SYSCALL_EMU		6	/* syscall emulation active */
92 #define TIF_SYSCALL_AUDIT	7	/* syscall auditing active */
93 #define TIF_SECCOMP		8	/* secure computing */
94 #define TIF_USER_RETURN_NOTIFY	11	/* notify kernel of userspace return */
95 #define TIF_UPROBE		12	/* breakpointed or singlestepping */
96 #define TIF_NOTSC		16	/* TSC is not accessible in userland */
97 #define TIF_IA32		17	/* IA32 compatibility process */
98 #define TIF_FORK		18	/* ret_from_fork */
99 #define TIF_NOHZ		19	/* in adaptive nohz mode */
100 #define TIF_MEMDIE		20	/* is terminating due to OOM killer */
101 #define TIF_POLLING_NRFLAG	21	/* idle is polling for TIF_NEED_RESCHED */
102 #define TIF_IO_BITMAP		22	/* uses I/O bitmap */
103 #define TIF_FORCED_TF		24	/* true if TF in eflags artificially */
104 #define TIF_BLOCKSTEP		25	/* set when we want DEBUGCTLMSR_BTF */
105 #define TIF_LAZY_MMU_UPDATES	27	/* task is updating the mmu lazily */
106 #define TIF_SYSCALL_TRACEPOINT	28	/* syscall tracepoint instrumentation */
107 #define TIF_ADDR32		29	/* 32-bit address space on 64 bits */
108 #define TIF_X32			30	/* 32-bit native x86-64 binary */
109 
110 #define _TIF_SYSCALL_TRACE	(1 << TIF_SYSCALL_TRACE)
111 #define _TIF_NOTIFY_RESUME	(1 << TIF_NOTIFY_RESUME)
112 #define _TIF_SIGPENDING		(1 << TIF_SIGPENDING)
113 #define _TIF_SINGLESTEP		(1 << TIF_SINGLESTEP)
114 #define _TIF_NEED_RESCHED	(1 << TIF_NEED_RESCHED)
115 #define _TIF_SYSCALL_EMU	(1 << TIF_SYSCALL_EMU)
116 #define _TIF_SYSCALL_AUDIT	(1 << TIF_SYSCALL_AUDIT)
117 #define _TIF_SECCOMP		(1 << TIF_SECCOMP)
118 #define _TIF_USER_RETURN_NOTIFY	(1 << TIF_USER_RETURN_NOTIFY)
119 #define _TIF_UPROBE		(1 << TIF_UPROBE)
120 #define _TIF_NOTSC		(1 << TIF_NOTSC)
121 #define _TIF_IA32		(1 << TIF_IA32)
122 #define _TIF_FORK		(1 << TIF_FORK)
123 #define _TIF_NOHZ		(1 << TIF_NOHZ)
124 #define _TIF_POLLING_NRFLAG	(1 << TIF_POLLING_NRFLAG)
125 #define _TIF_IO_BITMAP		(1 << TIF_IO_BITMAP)
126 #define _TIF_FORCED_TF		(1 << TIF_FORCED_TF)
127 #define _TIF_BLOCKSTEP		(1 << TIF_BLOCKSTEP)
128 #define _TIF_LAZY_MMU_UPDATES	(1 << TIF_LAZY_MMU_UPDATES)
129 #define _TIF_SYSCALL_TRACEPOINT	(1 << TIF_SYSCALL_TRACEPOINT)
130 #define _TIF_ADDR32		(1 << TIF_ADDR32)
131 #define _TIF_X32		(1 << TIF_X32)
132 
133 /*
134  * work to do in syscall_trace_enter().  Also includes TIF_NOHZ for
135  * enter_from_user_mode()
136  */
137 #define _TIF_WORK_SYSCALL_ENTRY	\
138 	(_TIF_SYSCALL_TRACE | _TIF_SYSCALL_EMU | _TIF_SYSCALL_AUDIT |	\
139 	 _TIF_SECCOMP | _TIF_SYSCALL_TRACEPOINT |	\
140 	 _TIF_NOHZ)
141 
142 /* work to do on any return to user space */
143 #define _TIF_ALLWORK_MASK						\
144 	((0x0000FFFF & ~_TIF_SECCOMP) | _TIF_SYSCALL_TRACEPOINT |	\
145 	_TIF_NOHZ)
146 
147 /* flags to check in __switch_to() */
148 #define _TIF_WORK_CTXSW							\
149 	(_TIF_IO_BITMAP|_TIF_NOTSC|_TIF_BLOCKSTEP)
150 
151 #define _TIF_WORK_CTXSW_PREV (_TIF_WORK_CTXSW|_TIF_USER_RETURN_NOTIFY)
152 #define _TIF_WORK_CTXSW_NEXT (_TIF_WORK_CTXSW)
153 
154 #define STACK_WARN		(THREAD_SIZE/8)
155 
156 /*
157  * macros/functions for gaining access to the thread information structure
158  *
159  * preempt_count needs to be 1 initially, until the scheduler is functional.
160  */
161 #ifndef __ASSEMBLY__
162 
163 static inline struct thread_info *current_thread_info(void)
164 {
165 	return (struct thread_info *)(current_top_of_stack() - THREAD_SIZE);
166 }
167 
168 static inline unsigned long current_stack_pointer(void)
169 {
170 	unsigned long sp;
171 #ifdef CONFIG_X86_64
172 	asm("mov %%rsp,%0" : "=g" (sp));
173 #else
174 	asm("mov %%esp,%0" : "=g" (sp));
175 #endif
176 	return sp;
177 }
178 
179 /*
180  * Walks up the stack frames to make sure that the specified object is
181  * entirely contained by a single stack frame.
182  *
183  * Returns:
184  *		 1 if within a frame
185  *		-1 if placed across a frame boundary (or outside stack)
186  *		 0 unable to determine (no frame pointers, etc)
187  */
188 static inline int arch_within_stack_frames(const void * const stack,
189 					   const void * const stackend,
190 					   const void *obj, unsigned long len)
191 {
192 #if defined(CONFIG_FRAME_POINTER)
193 	const void *frame = NULL;
194 	const void *oldframe;
195 
196 	oldframe = __builtin_frame_address(1);
197 	if (oldframe)
198 		frame = __builtin_frame_address(2);
199 	/*
200 	 * low ----------------------------------------------> high
201 	 * [saved bp][saved ip][args][local vars][saved bp][saved ip]
202 	 *                     ^----------------^
203 	 *               allow copies only within here
204 	 */
205 	while (stack <= frame && frame < stackend) {
206 		/*
207 		 * If obj + len extends past the last frame, this
208 		 * check won't pass and the next frame will be 0,
209 		 * causing us to bail out and correctly report
210 		 * the copy as invalid.
211 		 */
212 		if (obj + len <= frame)
213 			return obj >= oldframe + 2 * sizeof(void *) ? 1 : -1;
214 		oldframe = frame;
215 		frame = *(const void * const *)frame;
216 	}
217 	return -1;
218 #else
219 	return 0;
220 #endif
221 }
222 
223 #else /* !__ASSEMBLY__ */
224 
225 #ifdef CONFIG_X86_64
226 # define cpu_current_top_of_stack (cpu_tss + TSS_sp0)
227 #endif
228 
229 /*
230  * ASM operand which evaluates to a 'thread_info' address of
231  * the current task, if it is known that "reg" is exactly "off"
232  * bytes below the top of the stack currently.
233  *
234  * ( The kernel stack's size is known at build time, it is usually
235  *   2 or 4 pages, and the bottom  of the kernel stack contains
236  *   the thread_info structure. So to access the thread_info very
237  *   quickly from assembly code we can calculate down from the
238  *   top of the kernel stack to the bottom, using constant,
239  *   build-time calculations only. )
240  *
241  * For example, to fetch the current thread_info->flags value into %eax
242  * on x86-64 defconfig kernels, in syscall entry code where RSP is
243  * currently at exactly SIZEOF_PTREGS bytes away from the top of the
244  * stack:
245  *
246  *      mov ASM_THREAD_INFO(TI_flags, %rsp, SIZEOF_PTREGS), %eax
247  *
248  * will translate to:
249  *
250  *      8b 84 24 b8 c0 ff ff      mov    -0x3f48(%rsp), %eax
251  *
252  * which is below the current RSP by almost 16K.
253  */
254 #define ASM_THREAD_INFO(field, reg, off) ((field)+(off)-THREAD_SIZE)(reg)
255 
256 #endif
257 
258 /*
259  * Thread-synchronous status.
260  *
261  * This is different from the flags in that nobody else
262  * ever touches our thread-synchronous status, so we don't
263  * have to worry about atomic accesses.
264  */
265 #define TS_COMPAT		0x0002	/* 32bit syscall active (64BIT)*/
266 #ifdef CONFIG_COMPAT
267 #define TS_I386_REGS_POKED	0x0004	/* regs poked by 32-bit ptracer */
268 #endif
269 
270 #ifndef __ASSEMBLY__
271 
272 static inline bool in_ia32_syscall(void)
273 {
274 #ifdef CONFIG_X86_32
275 	return true;
276 #endif
277 #ifdef CONFIG_IA32_EMULATION
278 	if (current_thread_info()->status & TS_COMPAT)
279 		return true;
280 #endif
281 	return false;
282 }
283 
284 /*
285  * Force syscall return via IRET by making it look as if there was
286  * some work pending. IRET is our most capable (but slowest) syscall
287  * return path, which is able to restore modified SS, CS and certain
288  * EFLAGS values that other (fast) syscall return instructions
289  * are not able to restore properly.
290  */
291 #define force_iret() set_thread_flag(TIF_NOTIFY_RESUME)
292 
293 extern void arch_task_cache_init(void);
294 extern int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src);
295 extern void arch_release_task_struct(struct task_struct *tsk);
296 #endif	/* !__ASSEMBLY__ */
297 
298 #endif /* _ASM_X86_THREAD_INFO_H */
299