1// SPDX-License-Identifier: GPL-2.0-only
2// Copyright (C) 2021 ARM Limited.
3//
4// Assembly portion of the syscall ABI test
5
6//
7// Load values from memory into registers, invoke a syscall and save the
8// register values back to memory for later checking.  The syscall to be
9// invoked is configured in x8 of the input GPR data.
10//
11// x0:	SVE VL, 0 for FP only
12// x1:	SME VL
13//
14//	GPRs:	gpr_in, gpr_out
15//	FPRs:	fpr_in, fpr_out
16//	Zn:	z_in, z_out
17//	Pn:	p_in, p_out
18//	FFR:	ffr_in, ffr_out
19//	ZA:	za_in, za_out
20//	SVCR:	svcr_in, svcr_out
21
22#include "syscall-abi.h"
23
24.arch_extension sve
25
26/*
27 * LDR (vector to ZA array):
28 *	LDR ZA[\nw, #\offset], [X\nxbase, #\offset, MUL VL]
29 */
30.macro _ldr_za nw, nxbase, offset=0
31	.inst	0xe1000000			\
32		| (((\nw) & 3) << 13)		\
33		| ((\nxbase) << 5)		\
34		| ((\offset) & 7)
35.endm
36
37/*
38 * STR (vector from ZA array):
39 *	STR ZA[\nw, #\offset], [X\nxbase, #\offset, MUL VL]
40 */
41.macro _str_za nw, nxbase, offset=0
42	.inst	0xe1200000			\
43		| (((\nw) & 3) << 13)		\
44		| ((\nxbase) << 5)		\
45		| ((\offset) & 7)
46.endm
47
48.globl do_syscall
49do_syscall:
50	// Store callee saved registers x19-x29 (80 bytes) plus x0 and x1
51	stp	x29, x30, [sp, #-112]!
52	mov	x29, sp
53	stp	x0, x1, [sp, #16]
54	stp	x19, x20, [sp, #32]
55	stp	x21, x22, [sp, #48]
56	stp	x23, x24, [sp, #64]
57	stp	x25, x26, [sp, #80]
58	stp	x27, x28, [sp, #96]
59
60	// Set SVCR if we're doing SME
61	cbz	x1, 1f
62	adrp	x2, svcr_in
63	ldr	x2, [x2, :lo12:svcr_in]
64	msr	S3_3_C4_C2_2, x2
651:
66
67	// Load ZA if it's enabled - uses x12 as scratch due to SME LDR
68	tbz	x2, #SVCR_ZA_SHIFT, 1f
69	mov	w12, #0
70	ldr	x2, =za_in
712:	_ldr_za 12, 2
72	add	x2, x2, x1
73	add	x12, x12, #1
74	cmp	x1, x12
75	bne	2b
761:
77
78	// Load GPRs x8-x28, and save our SP/FP for later comparison
79	ldr	x2, =gpr_in
80	add	x2, x2, #64
81	ldp	x8, x9, [x2], #16
82	ldp	x10, x11, [x2], #16
83	ldp	x12, x13, [x2], #16
84	ldp	x14, x15, [x2], #16
85	ldp	x16, x17, [x2], #16
86	ldp	x18, x19, [x2], #16
87	ldp	x20, x21, [x2], #16
88	ldp	x22, x23, [x2], #16
89	ldp	x24, x25, [x2], #16
90	ldp	x26, x27, [x2], #16
91	ldr	x28, [x2], #8
92	str	x29, [x2], #8		// FP
93	str	x30, [x2], #8		// LR
94
95	// Load FPRs if we're not doing SVE
96	cbnz	x0, 1f
97	ldr	x2, =fpr_in
98	ldp	q0, q1, [x2]
99	ldp	q2, q3, [x2, #16 * 2]
100	ldp	q4, q5, [x2, #16 * 4]
101	ldp	q6, q7, [x2, #16 * 6]
102	ldp	q8, q9, [x2, #16 * 8]
103	ldp	q10, q11, [x2, #16 * 10]
104	ldp	q12, q13, [x2, #16 * 12]
105	ldp	q14, q15, [x2, #16 * 14]
106	ldp	q16, q17, [x2, #16 * 16]
107	ldp	q18, q19, [x2, #16 * 18]
108	ldp	q20, q21, [x2, #16 * 20]
109	ldp	q22, q23, [x2, #16 * 22]
110	ldp	q24, q25, [x2, #16 * 24]
111	ldp	q26, q27, [x2, #16 * 26]
112	ldp	q28, q29, [x2, #16 * 28]
113	ldp	q30, q31, [x2, #16 * 30]
1141:
115
116	// Load the SVE registers if we're doing SVE/SME
117	cbz	x0, 1f
118
119	ldr	x2, =z_in
120	ldr	z0, [x2, #0, MUL VL]
121	ldr	z1, [x2, #1, MUL VL]
122	ldr	z2, [x2, #2, MUL VL]
123	ldr	z3, [x2, #3, MUL VL]
124	ldr	z4, [x2, #4, MUL VL]
125	ldr	z5, [x2, #5, MUL VL]
126	ldr	z6, [x2, #6, MUL VL]
127	ldr	z7, [x2, #7, MUL VL]
128	ldr	z8, [x2, #8, MUL VL]
129	ldr	z9, [x2, #9, MUL VL]
130	ldr	z10, [x2, #10, MUL VL]
131	ldr	z11, [x2, #11, MUL VL]
132	ldr	z12, [x2, #12, MUL VL]
133	ldr	z13, [x2, #13, MUL VL]
134	ldr	z14, [x2, #14, MUL VL]
135	ldr	z15, [x2, #15, MUL VL]
136	ldr	z16, [x2, #16, MUL VL]
137	ldr	z17, [x2, #17, MUL VL]
138	ldr	z18, [x2, #18, MUL VL]
139	ldr	z19, [x2, #19, MUL VL]
140	ldr	z20, [x2, #20, MUL VL]
141	ldr	z21, [x2, #21, MUL VL]
142	ldr	z22, [x2, #22, MUL VL]
143	ldr	z23, [x2, #23, MUL VL]
144	ldr	z24, [x2, #24, MUL VL]
145	ldr	z25, [x2, #25, MUL VL]
146	ldr	z26, [x2, #26, MUL VL]
147	ldr	z27, [x2, #27, MUL VL]
148	ldr	z28, [x2, #28, MUL VL]
149	ldr	z29, [x2, #29, MUL VL]
150	ldr	z30, [x2, #30, MUL VL]
151	ldr	z31, [x2, #31, MUL VL]
152
153	// Only set a non-zero FFR, test patterns must be zero since the
154	// syscall should clear it - this lets us handle FA64.
155	ldr	x2, =ffr_in
156	ldr	p0, [x2, #0]
157	ldr	x2, [x2, #0]
158	cbz	x2, 2f
159	wrffr	p0.b
1602:
161
162	ldr	x2, =p_in
163	ldr	p0, [x2, #0, MUL VL]
164	ldr	p1, [x2, #1, MUL VL]
165	ldr	p2, [x2, #2, MUL VL]
166	ldr	p3, [x2, #3, MUL VL]
167	ldr	p4, [x2, #4, MUL VL]
168	ldr	p5, [x2, #5, MUL VL]
169	ldr	p6, [x2, #6, MUL VL]
170	ldr	p7, [x2, #7, MUL VL]
171	ldr	p8, [x2, #8, MUL VL]
172	ldr	p9, [x2, #9, MUL VL]
173	ldr	p10, [x2, #10, MUL VL]
174	ldr	p11, [x2, #11, MUL VL]
175	ldr	p12, [x2, #12, MUL VL]
176	ldr	p13, [x2, #13, MUL VL]
177	ldr	p14, [x2, #14, MUL VL]
178	ldr	p15, [x2, #15, MUL VL]
1791:
180
181	// Do the syscall
182	svc	#0
183
184	// Save GPRs x8-x30
185	ldr	x2, =gpr_out
186	add	x2, x2, #64
187	stp	x8, x9, [x2], #16
188	stp	x10, x11, [x2], #16
189	stp	x12, x13, [x2], #16
190	stp	x14, x15, [x2], #16
191	stp	x16, x17, [x2], #16
192	stp	x18, x19, [x2], #16
193	stp	x20, x21, [x2], #16
194	stp	x22, x23, [x2], #16
195	stp	x24, x25, [x2], #16
196	stp	x26, x27, [x2], #16
197	stp	x28, x29, [x2], #16
198	str	x30, [x2]
199
200	// Restore x0 and x1 for feature checks
201	ldp	x0, x1, [sp, #16]
202
203	// Save FPSIMD state
204	ldr	x2, =fpr_out
205	stp	q0, q1, [x2]
206	stp	q2, q3, [x2, #16 * 2]
207	stp	q4, q5, [x2, #16 * 4]
208	stp	q6, q7, [x2, #16 * 6]
209	stp	q8, q9, [x2, #16 * 8]
210	stp	q10, q11, [x2, #16 * 10]
211	stp	q12, q13, [x2, #16 * 12]
212	stp	q14, q15, [x2, #16 * 14]
213	stp	q16, q17, [x2, #16 * 16]
214	stp	q18, q19, [x2, #16 * 18]
215	stp	q20, q21, [x2, #16 * 20]
216	stp	q22, q23, [x2, #16 * 22]
217	stp	q24, q25, [x2, #16 * 24]
218	stp	q26, q27, [x2, #16 * 26]
219	stp	q28, q29, [x2, #16 * 28]
220	stp	q30, q31, [x2, #16 * 30]
221
222	// Save SVCR if we're doing SME
223	cbz	x1, 1f
224	mrs	x2, S3_3_C4_C2_2
225	adrp	x3, svcr_out
226	str	x2, [x3, :lo12:svcr_out]
2271:
228
229	// Save ZA if it's enabled - uses x12 as scratch due to SME STR
230	tbz	x2, #SVCR_ZA_SHIFT, 1f
231	mov	w12, #0
232	ldr	x2, =za_out
2332:	_str_za 12, 2
234	add	x2, x2, x1
235	add	x12, x12, #1
236	cmp	x1, x12
237	bne	2b
2381:
239
240	// Save the SVE state if we have some
241	cbz	x0, 1f
242
243	ldr	x2, =z_out
244	str	z0, [x2, #0, MUL VL]
245	str	z1, [x2, #1, MUL VL]
246	str	z2, [x2, #2, MUL VL]
247	str	z3, [x2, #3, MUL VL]
248	str	z4, [x2, #4, MUL VL]
249	str	z5, [x2, #5, MUL VL]
250	str	z6, [x2, #6, MUL VL]
251	str	z7, [x2, #7, MUL VL]
252	str	z8, [x2, #8, MUL VL]
253	str	z9, [x2, #9, MUL VL]
254	str	z10, [x2, #10, MUL VL]
255	str	z11, [x2, #11, MUL VL]
256	str	z12, [x2, #12, MUL VL]
257	str	z13, [x2, #13, MUL VL]
258	str	z14, [x2, #14, MUL VL]
259	str	z15, [x2, #15, MUL VL]
260	str	z16, [x2, #16, MUL VL]
261	str	z17, [x2, #17, MUL VL]
262	str	z18, [x2, #18, MUL VL]
263	str	z19, [x2, #19, MUL VL]
264	str	z20, [x2, #20, MUL VL]
265	str	z21, [x2, #21, MUL VL]
266	str	z22, [x2, #22, MUL VL]
267	str	z23, [x2, #23, MUL VL]
268	str	z24, [x2, #24, MUL VL]
269	str	z25, [x2, #25, MUL VL]
270	str	z26, [x2, #26, MUL VL]
271	str	z27, [x2, #27, MUL VL]
272	str	z28, [x2, #28, MUL VL]
273	str	z29, [x2, #29, MUL VL]
274	str	z30, [x2, #30, MUL VL]
275	str	z31, [x2, #31, MUL VL]
276
277	ldr	x2, =p_out
278	str	p0, [x2, #0, MUL VL]
279	str	p1, [x2, #1, MUL VL]
280	str	p2, [x2, #2, MUL VL]
281	str	p3, [x2, #3, MUL VL]
282	str	p4, [x2, #4, MUL VL]
283	str	p5, [x2, #5, MUL VL]
284	str	p6, [x2, #6, MUL VL]
285	str	p7, [x2, #7, MUL VL]
286	str	p8, [x2, #8, MUL VL]
287	str	p9, [x2, #9, MUL VL]
288	str	p10, [x2, #10, MUL VL]
289	str	p11, [x2, #11, MUL VL]
290	str	p12, [x2, #12, MUL VL]
291	str	p13, [x2, #13, MUL VL]
292	str	p14, [x2, #14, MUL VL]
293	str	p15, [x2, #15, MUL VL]
294
295	// Only save FFR if we wrote a value for SME
296	ldr	x2, =ffr_in
297	ldr	x2, [x2, #0]
298	cbz	x2, 1f
299	ldr	x2, =ffr_out
300	rdffr	p0.b
301	str	p0, [x2, #0]
3021:
303
304	// Restore callee saved registers x19-x30
305	ldp	x19, x20, [sp, #32]
306	ldp	x21, x22, [sp, #48]
307	ldp	x23, x24, [sp, #64]
308	ldp	x25, x26, [sp, #80]
309	ldp	x27, x28, [sp, #96]
310	ldp	x29, x30, [sp], #112
311
312	// Clear SVCR if we were doing SME so future tests don't have ZA
313	cbz	x1, 1f
314	msr	S3_3_C4_C2_2, xzr
3151:
316
317	ret
318