xref: /kernel/linux/linux-6.6/arch/x86/entry/calling.h (revision 62306a36)
1/* SPDX-License-Identifier: GPL-2.0 */
2#include <linux/jump_label.h>
3#include <asm/unwind_hints.h>
4#include <asm/cpufeatures.h>
5#include <asm/page_types.h>
6#include <asm/percpu.h>
7#include <asm/asm-offsets.h>
8#include <asm/processor-flags.h>
9#include <asm/ptrace-abi.h>
10#include <asm/msr.h>
11#include <asm/nospec-branch.h>
12
13/*
14
15 x86 function call convention, 64-bit:
16 -------------------------------------
17  arguments           |  callee-saved      | extra caller-saved | return
18 [callee-clobbered]   |                    | [callee-clobbered] |
19 ---------------------------------------------------------------------------
20 rdi rsi rdx rcx r8-9 | rbx rbp [*] r12-15 | r10-11             | rax, rdx [**]
21
22 ( rsp is obviously invariant across normal function calls. (gcc can 'merge'
23   functions when it sees tail-call optimization possibilities) rflags is
24   clobbered. Leftover arguments are passed over the stack frame.)
25
26 [*]  In the frame-pointers case rbp is fixed to the stack frame.
27
28 [**] for struct return values wider than 64 bits the return convention is a
29      bit more complex: up to 128 bits width we return small structures
30      straight in rax, rdx. For structures larger than that (3 words or
31      larger) the caller puts a pointer to an on-stack return struct
32      [allocated in the caller's stack frame] into the first argument - i.e.
33      into rdi. All other arguments shift up by one in this case.
34      Fortunately this case is rare in the kernel.
35
36For 32-bit we have the following conventions - kernel is built with
37-mregparm=3 and -freg-struct-return:
38
39 x86 function calling convention, 32-bit:
40 ----------------------------------------
41  arguments         | callee-saved        | extra caller-saved | return
42 [callee-clobbered] |                     | [callee-clobbered] |
43 -------------------------------------------------------------------------
44 eax edx ecx        | ebx edi esi ebp [*] | <none>             | eax, edx [**]
45
46 ( here too esp is obviously invariant across normal function calls. eflags
47   is clobbered. Leftover arguments are passed over the stack frame. )
48
49 [*]  In the frame-pointers case ebp is fixed to the stack frame.
50
51 [**] We build with -freg-struct-return, which on 32-bit means similar
52      semantics as on 64-bit: edx can be used for a second return value
53      (i.e. covering integer and structure sizes up to 64 bits) - after that
54      it gets more complex and more expensive: 3-word or larger struct returns
55      get done in the caller's frame and the pointer to the return struct goes
56      into regparm0, i.e. eax - the other arguments shift up and the
57      function's register parameters degenerate to regparm=2 in essence.
58
59*/
60
61#ifdef CONFIG_X86_64
62
63/*
64 * 64-bit system call stack frame layout defines and helpers,
65 * for assembly code:
66 */
67
68.macro PUSH_REGS rdx=%rdx rcx=%rcx rax=%rax save_ret=0
69	.if \save_ret
70	pushq	%rsi		/* pt_regs->si */
71	movq	8(%rsp), %rsi	/* temporarily store the return address in %rsi */
72	movq	%rdi, 8(%rsp)	/* pt_regs->di (overwriting original return address) */
73	.else
74	pushq   %rdi		/* pt_regs->di */
75	pushq   %rsi		/* pt_regs->si */
76	.endif
77	pushq	\rdx		/* pt_regs->dx */
78	pushq   \rcx		/* pt_regs->cx */
79	pushq   \rax		/* pt_regs->ax */
80	pushq   %r8		/* pt_regs->r8 */
81	pushq   %r9		/* pt_regs->r9 */
82	pushq   %r10		/* pt_regs->r10 */
83	pushq   %r11		/* pt_regs->r11 */
84	pushq	%rbx		/* pt_regs->rbx */
85	pushq	%rbp		/* pt_regs->rbp */
86	pushq	%r12		/* pt_regs->r12 */
87	pushq	%r13		/* pt_regs->r13 */
88	pushq	%r14		/* pt_regs->r14 */
89	pushq	%r15		/* pt_regs->r15 */
90	UNWIND_HINT_REGS
91
92	.if \save_ret
93	pushq	%rsi		/* return address on top of stack */
94	.endif
95.endm
96
97.macro CLEAR_REGS
98	/*
99	 * Sanitize registers of values that a speculation attack might
100	 * otherwise want to exploit. The lower registers are likely clobbered
101	 * well before they could be put to use in a speculative execution
102	 * gadget.
103	 */
104	xorl	%esi,  %esi	/* nospec si  */
105	xorl	%edx,  %edx	/* nospec dx  */
106	xorl	%ecx,  %ecx	/* nospec cx  */
107	xorl	%r8d,  %r8d	/* nospec r8  */
108	xorl	%r9d,  %r9d	/* nospec r9  */
109	xorl	%r10d, %r10d	/* nospec r10 */
110	xorl	%r11d, %r11d	/* nospec r11 */
111	xorl	%ebx,  %ebx	/* nospec rbx */
112	xorl	%ebp,  %ebp	/* nospec rbp */
113	xorl	%r12d, %r12d	/* nospec r12 */
114	xorl	%r13d, %r13d	/* nospec r13 */
115	xorl	%r14d, %r14d	/* nospec r14 */
116	xorl	%r15d, %r15d	/* nospec r15 */
117
118.endm
119
120.macro PUSH_AND_CLEAR_REGS rdx=%rdx rcx=%rcx rax=%rax save_ret=0
121	PUSH_REGS rdx=\rdx, rcx=\rcx, rax=\rax, save_ret=\save_ret
122	CLEAR_REGS
123.endm
124
125.macro POP_REGS pop_rdi=1
126	popq %r15
127	popq %r14
128	popq %r13
129	popq %r12
130	popq %rbp
131	popq %rbx
132	popq %r11
133	popq %r10
134	popq %r9
135	popq %r8
136	popq %rax
137	popq %rcx
138	popq %rdx
139	popq %rsi
140	.if \pop_rdi
141	popq %rdi
142	.endif
143.endm
144
145#ifdef CONFIG_PAGE_TABLE_ISOLATION
146
147/*
148 * PAGE_TABLE_ISOLATION PGDs are 8k.  Flip bit 12 to switch between the two
149 * halves:
150 */
151#define PTI_USER_PGTABLE_BIT		PAGE_SHIFT
152#define PTI_USER_PGTABLE_MASK		(1 << PTI_USER_PGTABLE_BIT)
153#define PTI_USER_PCID_BIT		X86_CR3_PTI_PCID_USER_BIT
154#define PTI_USER_PCID_MASK		(1 << PTI_USER_PCID_BIT)
155#define PTI_USER_PGTABLE_AND_PCID_MASK  (PTI_USER_PCID_MASK | PTI_USER_PGTABLE_MASK)
156
157.macro SET_NOFLUSH_BIT	reg:req
158	bts	$X86_CR3_PCID_NOFLUSH_BIT, \reg
159.endm
160
161.macro ADJUST_KERNEL_CR3 reg:req
162	ALTERNATIVE "", "SET_NOFLUSH_BIT \reg", X86_FEATURE_PCID
163	/* Clear PCID and "PAGE_TABLE_ISOLATION bit", point CR3 at kernel pagetables: */
164	andq    $(~PTI_USER_PGTABLE_AND_PCID_MASK), \reg
165.endm
166
167.macro SWITCH_TO_KERNEL_CR3 scratch_reg:req
168	ALTERNATIVE "jmp .Lend_\@", "", X86_FEATURE_PTI
169	mov	%cr3, \scratch_reg
170	ADJUST_KERNEL_CR3 \scratch_reg
171	mov	\scratch_reg, %cr3
172.Lend_\@:
173.endm
174
175#define THIS_CPU_user_pcid_flush_mask   \
176	PER_CPU_VAR(cpu_tlbstate) + TLB_STATE_user_pcid_flush_mask
177
178.macro SWITCH_TO_USER_CR3_NOSTACK scratch_reg:req scratch_reg2:req
179	ALTERNATIVE "jmp .Lend_\@", "", X86_FEATURE_PTI
180	mov	%cr3, \scratch_reg
181
182	ALTERNATIVE "jmp .Lwrcr3_\@", "", X86_FEATURE_PCID
183
184	/*
185	 * Test if the ASID needs a flush.
186	 */
187	movq	\scratch_reg, \scratch_reg2
188	andq	$(0x7FF), \scratch_reg		/* mask ASID */
189	bt	\scratch_reg, THIS_CPU_user_pcid_flush_mask
190	jnc	.Lnoflush_\@
191
192	/* Flush needed, clear the bit */
193	btr	\scratch_reg, THIS_CPU_user_pcid_flush_mask
194	movq	\scratch_reg2, \scratch_reg
195	jmp	.Lwrcr3_pcid_\@
196
197.Lnoflush_\@:
198	movq	\scratch_reg2, \scratch_reg
199	SET_NOFLUSH_BIT \scratch_reg
200
201.Lwrcr3_pcid_\@:
202	/* Flip the ASID to the user version */
203	orq	$(PTI_USER_PCID_MASK), \scratch_reg
204
205.Lwrcr3_\@:
206	/* Flip the PGD to the user version */
207	orq     $(PTI_USER_PGTABLE_MASK), \scratch_reg
208	mov	\scratch_reg, %cr3
209.Lend_\@:
210.endm
211
212.macro SWITCH_TO_USER_CR3_STACK	scratch_reg:req
213	pushq	%rax
214	SWITCH_TO_USER_CR3_NOSTACK scratch_reg=\scratch_reg scratch_reg2=%rax
215	popq	%rax
216.endm
217
218.macro SAVE_AND_SWITCH_TO_KERNEL_CR3 scratch_reg:req save_reg:req
219	ALTERNATIVE "jmp .Ldone_\@", "", X86_FEATURE_PTI
220	movq	%cr3, \scratch_reg
221	movq	\scratch_reg, \save_reg
222	/*
223	 * Test the user pagetable bit. If set, then the user page tables
224	 * are active. If clear CR3 already has the kernel page table
225	 * active.
226	 */
227	bt	$PTI_USER_PGTABLE_BIT, \scratch_reg
228	jnc	.Ldone_\@
229
230	ADJUST_KERNEL_CR3 \scratch_reg
231	movq	\scratch_reg, %cr3
232
233.Ldone_\@:
234.endm
235
236.macro RESTORE_CR3 scratch_reg:req save_reg:req
237	ALTERNATIVE "jmp .Lend_\@", "", X86_FEATURE_PTI
238
239	ALTERNATIVE "jmp .Lwrcr3_\@", "", X86_FEATURE_PCID
240
241	/*
242	 * KERNEL pages can always resume with NOFLUSH as we do
243	 * explicit flushes.
244	 */
245	bt	$PTI_USER_PGTABLE_BIT, \save_reg
246	jnc	.Lnoflush_\@
247
248	/*
249	 * Check if there's a pending flush for the user ASID we're
250	 * about to set.
251	 */
252	movq	\save_reg, \scratch_reg
253	andq	$(0x7FF), \scratch_reg
254	bt	\scratch_reg, THIS_CPU_user_pcid_flush_mask
255	jnc	.Lnoflush_\@
256
257	btr	\scratch_reg, THIS_CPU_user_pcid_flush_mask
258	jmp	.Lwrcr3_\@
259
260.Lnoflush_\@:
261	SET_NOFLUSH_BIT \save_reg
262
263.Lwrcr3_\@:
264	/*
265	 * The CR3 write could be avoided when not changing its value,
266	 * but would require a CR3 read *and* a scratch register.
267	 */
268	movq	\save_reg, %cr3
269.Lend_\@:
270.endm
271
272#else /* CONFIG_PAGE_TABLE_ISOLATION=n: */
273
274.macro SWITCH_TO_KERNEL_CR3 scratch_reg:req
275.endm
276.macro SWITCH_TO_USER_CR3_NOSTACK scratch_reg:req scratch_reg2:req
277.endm
278.macro SWITCH_TO_USER_CR3_STACK scratch_reg:req
279.endm
280.macro SAVE_AND_SWITCH_TO_KERNEL_CR3 scratch_reg:req save_reg:req
281.endm
282.macro RESTORE_CR3 scratch_reg:req save_reg:req
283.endm
284
285#endif
286
287/*
288 * IBRS kernel mitigation for Spectre_v2.
289 *
290 * Assumes full context is established (PUSH_REGS, CR3 and GS) and it clobbers
291 * the regs it uses (AX, CX, DX). Must be called before the first RET
292 * instruction (NOTE! UNTRAIN_RET includes a RET instruction)
293 *
294 * The optional argument is used to save/restore the current value,
295 * which is used on the paranoid paths.
296 *
297 * Assumes x86_spec_ctrl_{base,current} to have SPEC_CTRL_IBRS set.
298 */
299.macro IBRS_ENTER save_reg
300#ifdef CONFIG_CPU_IBRS_ENTRY
301	ALTERNATIVE "jmp .Lend_\@", "", X86_FEATURE_KERNEL_IBRS
302	movl	$MSR_IA32_SPEC_CTRL, %ecx
303
304.ifnb \save_reg
305	rdmsr
306	shl	$32, %rdx
307	or	%rdx, %rax
308	mov	%rax, \save_reg
309	test	$SPEC_CTRL_IBRS, %eax
310	jz	.Ldo_wrmsr_\@
311	lfence
312	jmp	.Lend_\@
313.Ldo_wrmsr_\@:
314.endif
315
316	movq	PER_CPU_VAR(x86_spec_ctrl_current), %rdx
317	movl	%edx, %eax
318	shr	$32, %rdx
319	wrmsr
320.Lend_\@:
321#endif
322.endm
323
324/*
325 * Similar to IBRS_ENTER, requires KERNEL GS,CR3 and clobbers (AX, CX, DX)
326 * regs. Must be called after the last RET.
327 */
328.macro IBRS_EXIT save_reg
329#ifdef CONFIG_CPU_IBRS_ENTRY
330	ALTERNATIVE "jmp .Lend_\@", "", X86_FEATURE_KERNEL_IBRS
331	movl	$MSR_IA32_SPEC_CTRL, %ecx
332
333.ifnb \save_reg
334	mov	\save_reg, %rdx
335.else
336	movq	PER_CPU_VAR(x86_spec_ctrl_current), %rdx
337	andl	$(~SPEC_CTRL_IBRS), %edx
338.endif
339
340	movl	%edx, %eax
341	shr	$32, %rdx
342	wrmsr
343.Lend_\@:
344#endif
345.endm
346
347/*
348 * Mitigate Spectre v1 for conditional swapgs code paths.
349 *
350 * FENCE_SWAPGS_USER_ENTRY is used in the user entry swapgs code path, to
351 * prevent a speculative swapgs when coming from kernel space.
352 *
353 * FENCE_SWAPGS_KERNEL_ENTRY is used in the kernel entry non-swapgs code path,
354 * to prevent the swapgs from getting speculatively skipped when coming from
355 * user space.
356 */
357.macro FENCE_SWAPGS_USER_ENTRY
358	ALTERNATIVE "", "lfence", X86_FEATURE_FENCE_SWAPGS_USER
359.endm
360.macro FENCE_SWAPGS_KERNEL_ENTRY
361	ALTERNATIVE "", "lfence", X86_FEATURE_FENCE_SWAPGS_KERNEL
362.endm
363
364.macro STACKLEAK_ERASE_NOCLOBBER
365#ifdef CONFIG_GCC_PLUGIN_STACKLEAK
366	PUSH_AND_CLEAR_REGS
367	call stackleak_erase
368	POP_REGS
369#endif
370.endm
371
372.macro SAVE_AND_SET_GSBASE scratch_reg:req save_reg:req
373	rdgsbase \save_reg
374	GET_PERCPU_BASE \scratch_reg
375	wrgsbase \scratch_reg
376.endm
377
378#else /* CONFIG_X86_64 */
379# undef		UNWIND_HINT_IRET_REGS
380# define	UNWIND_HINT_IRET_REGS
381#endif /* !CONFIG_X86_64 */
382
383.macro STACKLEAK_ERASE
384#ifdef CONFIG_GCC_PLUGIN_STACKLEAK
385	call stackleak_erase
386#endif
387.endm
388
389#ifdef CONFIG_SMP
390
391/*
392 * CPU/node NR is loaded from the limit (size) field of a special segment
393 * descriptor entry in GDT.
394 */
395.macro LOAD_CPU_AND_NODE_SEG_LIMIT reg:req
396	movq	$__CPUNODE_SEG, \reg
397	lsl	\reg, \reg
398.endm
399
400/*
401 * Fetch the per-CPU GSBASE value for this processor and put it in @reg.
402 * We normally use %gs for accessing per-CPU data, but we are setting up
403 * %gs here and obviously can not use %gs itself to access per-CPU data.
404 *
405 * Do not use RDPID, because KVM loads guest's TSC_AUX on vm-entry and
406 * may not restore the host's value until the CPU returns to userspace.
407 * Thus the kernel would consume a guest's TSC_AUX if an NMI arrives
408 * while running KVM's run loop.
409 */
410.macro GET_PERCPU_BASE reg:req
411	LOAD_CPU_AND_NODE_SEG_LIMIT \reg
412	andq	$VDSO_CPUNODE_MASK, \reg
413	movq	__per_cpu_offset(, \reg, 8), \reg
414.endm
415
416#else
417
418.macro GET_PERCPU_BASE reg:req
419	movq	pcpu_unit_offsets(%rip), \reg
420.endm
421
422#endif /* CONFIG_SMP */
423