xref: /linux/drivers/accel/ivpu/ivpu_mmu.c (revision b563e47957af4ff71736c5cc4120a59b055ab583)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Copyright (C) 2020-2023 Intel Corporation
4  */
5 
6 #include <linux/circ_buf.h>
7 #include <linux/highmem.h>
8 
9 #include "ivpu_drv.h"
10 #include "ivpu_hw_mtl_reg.h"
11 #include "ivpu_hw_reg_io.h"
12 #include "ivpu_mmu.h"
13 #include "ivpu_mmu_context.h"
14 #include "ivpu_pm.h"
15 
16 #define IVPU_MMU_IDR0_REF		0x080f3e0f
17 #define IVPU_MMU_IDR0_REF_SIMICS	0x080f3e1f
18 #define IVPU_MMU_IDR1_REF		0x0e739d18
19 #define IVPU_MMU_IDR3_REF		0x0000003c
20 #define IVPU_MMU_IDR5_REF		0x00040070
21 #define IVPU_MMU_IDR5_REF_SIMICS	0x00000075
22 #define IVPU_MMU_IDR5_REF_FPGA		0x00800075
23 
24 #define IVPU_MMU_CDTAB_ENT_SIZE		64
25 #define IVPU_MMU_CDTAB_ENT_COUNT_LOG2	8 /* 256 entries */
26 #define IVPU_MMU_CDTAB_ENT_COUNT	((u32)1 << IVPU_MMU_CDTAB_ENT_COUNT_LOG2)
27 
28 #define IVPU_MMU_STREAM_ID0		0
29 #define IVPU_MMU_STREAM_ID3		3
30 
31 #define IVPU_MMU_STRTAB_ENT_SIZE	64
32 #define IVPU_MMU_STRTAB_ENT_COUNT	4
33 #define IVPU_MMU_STRTAB_CFG_LOG2SIZE	2
34 #define IVPU_MMU_STRTAB_CFG		IVPU_MMU_STRTAB_CFG_LOG2SIZE
35 
36 #define IVPU_MMU_Q_COUNT_LOG2		4 /* 16 entries */
37 #define IVPU_MMU_Q_COUNT		((u32)1 << IVPU_MMU_Q_COUNT_LOG2)
38 #define IVPU_MMU_Q_WRAP_BIT		(IVPU_MMU_Q_COUNT << 1)
39 #define IVPU_MMU_Q_WRAP_MASK		(IVPU_MMU_Q_WRAP_BIT - 1)
40 #define IVPU_MMU_Q_IDX_MASK		(IVPU_MMU_Q_COUNT - 1)
41 #define IVPU_MMU_Q_IDX(val)		((val) & IVPU_MMU_Q_IDX_MASK)
42 
43 #define IVPU_MMU_CMDQ_CMD_SIZE		16
44 #define IVPU_MMU_CMDQ_SIZE		(IVPU_MMU_Q_COUNT * IVPU_MMU_CMDQ_CMD_SIZE)
45 
46 #define IVPU_MMU_EVTQ_CMD_SIZE		32
47 #define IVPU_MMU_EVTQ_SIZE		(IVPU_MMU_Q_COUNT * IVPU_MMU_EVTQ_CMD_SIZE)
48 
49 #define IVPU_MMU_CMD_OPCODE		GENMASK(7, 0)
50 
51 #define IVPU_MMU_CMD_SYNC_0_CS		GENMASK(13, 12)
52 #define IVPU_MMU_CMD_SYNC_0_MSH		GENMASK(23, 22)
53 #define IVPU_MMU_CMD_SYNC_0_MSI_ATTR	GENMASK(27, 24)
54 #define IVPU_MMU_CMD_SYNC_0_MSI_ATTR	GENMASK(27, 24)
55 #define IVPU_MMU_CMD_SYNC_0_MSI_DATA	GENMASK(63, 32)
56 
57 #define IVPU_MMU_CMD_CFGI_0_SSEC	BIT(10)
58 #define IVPU_MMU_CMD_CFGI_0_SSV		BIT(11)
59 #define IVPU_MMU_CMD_CFGI_0_SSID	GENMASK(31, 12)
60 #define IVPU_MMU_CMD_CFGI_0_SID		GENMASK(63, 32)
61 #define IVPU_MMU_CMD_CFGI_1_RANGE	GENMASK(4, 0)
62 
63 #define IVPU_MMU_CMD_TLBI_0_ASID	GENMASK(63, 48)
64 #define IVPU_MMU_CMD_TLBI_0_VMID	GENMASK(47, 32)
65 
66 #define CMD_PREFETCH_CFG		0x1
67 #define CMD_CFGI_STE			0x3
68 #define CMD_CFGI_ALL			0x4
69 #define CMD_CFGI_CD			0x5
70 #define CMD_CFGI_CD_ALL			0x6
71 #define CMD_TLBI_NH_ASID		0x11
72 #define CMD_TLBI_EL2_ALL		0x20
73 #define CMD_TLBI_NSNH_ALL		0x30
74 #define CMD_SYNC			0x46
75 
76 #define IVPU_MMU_EVT_F_UUT		0x01
77 #define IVPU_MMU_EVT_C_BAD_STREAMID	0x02
78 #define IVPU_MMU_EVT_F_STE_FETCH	0x03
79 #define IVPU_MMU_EVT_C_BAD_STE		0x04
80 #define IVPU_MMU_EVT_F_BAD_ATS_TREQ	0x05
81 #define IVPU_MMU_EVT_F_STREAM_DISABLED	0x06
82 #define IVPU_MMU_EVT_F_TRANSL_FORBIDDEN	0x07
83 #define IVPU_MMU_EVT_C_BAD_SUBSTREAMID	0x08
84 #define IVPU_MMU_EVT_F_CD_FETCH		0x09
85 #define IVPU_MMU_EVT_C_BAD_CD		0x0a
86 #define IVPU_MMU_EVT_F_WALK_EABT	0x0b
87 #define IVPU_MMU_EVT_F_TRANSLATION	0x10
88 #define IVPU_MMU_EVT_F_ADDR_SIZE	0x11
89 #define IVPU_MMU_EVT_F_ACCESS		0x12
90 #define IVPU_MMU_EVT_F_PERMISSION	0x13
91 #define IVPU_MMU_EVT_F_TLB_CONFLICT	0x20
92 #define IVPU_MMU_EVT_F_CFG_CONFLICT	0x21
93 #define IVPU_MMU_EVT_E_PAGE_REQUEST	0x24
94 #define IVPU_MMU_EVT_F_VMS_FETCH	0x25
95 
96 #define IVPU_MMU_EVT_OP_MASK		GENMASK_ULL(7, 0)
97 #define IVPU_MMU_EVT_SSID_MASK		GENMASK_ULL(31, 12)
98 
99 #define IVPU_MMU_Q_BASE_RWA		BIT(62)
100 #define IVPU_MMU_Q_BASE_ADDR_MASK	GENMASK_ULL(51, 5)
101 #define IVPU_MMU_STRTAB_BASE_RA		BIT(62)
102 #define IVPU_MMU_STRTAB_BASE_ADDR_MASK	GENMASK_ULL(51, 6)
103 
104 #define IVPU_MMU_IRQ_EVTQ_EN		BIT(2)
105 #define IVPU_MMU_IRQ_GERROR_EN		BIT(0)
106 
107 #define IVPU_MMU_CR0_ATSCHK		BIT(4)
108 #define IVPU_MMU_CR0_CMDQEN		BIT(3)
109 #define IVPU_MMU_CR0_EVTQEN		BIT(2)
110 #define IVPU_MMU_CR0_PRIQEN		BIT(1)
111 #define IVPU_MMU_CR0_SMMUEN		BIT(0)
112 
113 #define IVPU_MMU_CR1_TABLE_SH		GENMASK(11, 10)
114 #define IVPU_MMU_CR1_TABLE_OC		GENMASK(9, 8)
115 #define IVPU_MMU_CR1_TABLE_IC		GENMASK(7, 6)
116 #define IVPU_MMU_CR1_QUEUE_SH		GENMASK(5, 4)
117 #define IVPU_MMU_CR1_QUEUE_OC		GENMASK(3, 2)
118 #define IVPU_MMU_CR1_QUEUE_IC		GENMASK(1, 0)
119 #define IVPU_MMU_CACHE_NC		0
120 #define IVPU_MMU_CACHE_WB		1
121 #define IVPU_MMU_CACHE_WT		2
122 #define IVPU_MMU_SH_NSH			0
123 #define IVPU_MMU_SH_OSH			2
124 #define IVPU_MMU_SH_ISH			3
125 
126 #define IVPU_MMU_CMDQ_OP		GENMASK_ULL(7, 0)
127 
128 #define IVPU_MMU_CD_0_TCR_T0SZ		GENMASK_ULL(5, 0)
129 #define IVPU_MMU_CD_0_TCR_TG0		GENMASK_ULL(7, 6)
130 #define IVPU_MMU_CD_0_TCR_IRGN0		GENMASK_ULL(9, 8)
131 #define IVPU_MMU_CD_0_TCR_ORGN0		GENMASK_ULL(11, 10)
132 #define IVPU_MMU_CD_0_TCR_SH0		GENMASK_ULL(13, 12)
133 #define IVPU_MMU_CD_0_TCR_EPD0		BIT_ULL(14)
134 #define IVPU_MMU_CD_0_TCR_EPD1		BIT_ULL(30)
135 #define IVPU_MMU_CD_0_ENDI		BIT(15)
136 #define IVPU_MMU_CD_0_V			BIT(31)
137 #define IVPU_MMU_CD_0_TCR_IPS		GENMASK_ULL(34, 32)
138 #define IVPU_MMU_CD_0_TCR_TBI0		BIT_ULL(38)
139 #define IVPU_MMU_CD_0_AA64		BIT(41)
140 #define IVPU_MMU_CD_0_S			BIT(44)
141 #define IVPU_MMU_CD_0_R			BIT(45)
142 #define IVPU_MMU_CD_0_A			BIT(46)
143 #define IVPU_MMU_CD_0_ASET		BIT(47)
144 #define IVPU_MMU_CD_0_ASID		GENMASK_ULL(63, 48)
145 
146 #define IVPU_MMU_CD_1_TTB0_MASK		GENMASK_ULL(51, 4)
147 
148 #define IVPU_MMU_STE_0_S1CDMAX		GENMASK_ULL(63, 59)
149 #define IVPU_MMU_STE_0_S1FMT		GENMASK_ULL(5, 4)
150 #define IVPU_MMU_STE_0_S1FMT_LINEAR	0
151 #define IVPU_MMU_STE_DWORDS		8
152 #define IVPU_MMU_STE_0_CFG_S1_TRANS	5
153 #define IVPU_MMU_STE_0_CFG		GENMASK_ULL(3, 1)
154 #define IVPU_MMU_STE_0_S1CTXPTR_MASK	GENMASK_ULL(51, 6)
155 #define IVPU_MMU_STE_0_V			BIT(0)
156 
157 #define IVPU_MMU_STE_1_STRW_NSEL1	0ul
158 #define IVPU_MMU_STE_1_CONT		GENMASK_ULL(16, 13)
159 #define IVPU_MMU_STE_1_STRW		GENMASK_ULL(31, 30)
160 #define IVPU_MMU_STE_1_PRIVCFG		GENMASK_ULL(49, 48)
161 #define IVPU_MMU_STE_1_PRIVCFG_UNPRIV	2ul
162 #define IVPU_MMU_STE_1_INSTCFG		GENMASK_ULL(51, 50)
163 #define IVPU_MMU_STE_1_INSTCFG_DATA	2ul
164 #define IVPU_MMU_STE_1_MEV		BIT(19)
165 #define IVPU_MMU_STE_1_S1STALLD		BIT(27)
166 #define IVPU_MMU_STE_1_S1C_CACHE_NC	0ul
167 #define IVPU_MMU_STE_1_S1C_CACHE_WBRA	1ul
168 #define IVPU_MMU_STE_1_S1C_CACHE_WT	2ul
169 #define IVPU_MMU_STE_1_S1C_CACHE_WB	3ul
170 #define IVPU_MMU_STE_1_S1CIR		GENMASK_ULL(3, 2)
171 #define IVPU_MMU_STE_1_S1COR		GENMASK_ULL(5, 4)
172 #define IVPU_MMU_STE_1_S1CSH		GENMASK_ULL(7, 6)
173 #define IVPU_MMU_STE_1_S1DSS		GENMASK_ULL(1, 0)
174 #define IVPU_MMU_STE_1_S1DSS_TERMINATE	0x0
175 
176 #define IVPU_MMU_REG_TIMEOUT_US		(10 * USEC_PER_MSEC)
177 #define IVPU_MMU_QUEUE_TIMEOUT_US	(100 * USEC_PER_MSEC)
178 
179 #define IVPU_MMU_GERROR_ERR_MASK ((REG_FLD(MTL_VPU_HOST_MMU_GERROR, CMDQ)) | \
180 				  (REG_FLD(MTL_VPU_HOST_MMU_GERROR, EVTQ_ABT)) | \
181 				  (REG_FLD(MTL_VPU_HOST_MMU_GERROR, PRIQ_ABT)) | \
182 				  (REG_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_CMDQ_ABT)) | \
183 				  (REG_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_EVTQ_ABT)) | \
184 				  (REG_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_PRIQ_ABT)) | \
185 				  (REG_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_ABT)))
186 
187 static char *ivpu_mmu_event_to_str(u32 cmd)
188 {
189 	switch (cmd) {
190 	case IVPU_MMU_EVT_F_UUT:
191 		return "Unsupported Upstream Transaction";
192 	case IVPU_MMU_EVT_C_BAD_STREAMID:
193 		return "Transaction StreamID out of range";
194 	case IVPU_MMU_EVT_F_STE_FETCH:
195 		return "Fetch of STE caused external abort";
196 	case IVPU_MMU_EVT_C_BAD_STE:
197 		return "Used STE invalid";
198 	case IVPU_MMU_EVT_F_BAD_ATS_TREQ:
199 		return "Address Request disallowed for a StreamID";
200 	case IVPU_MMU_EVT_F_STREAM_DISABLED:
201 		return "Transaction marks non-substream disabled";
202 	case IVPU_MMU_EVT_F_TRANSL_FORBIDDEN:
203 		return "MMU bypass is disallowed for this StreamID";
204 	case IVPU_MMU_EVT_C_BAD_SUBSTREAMID:
205 		return "Invalid StreamID";
206 	case IVPU_MMU_EVT_F_CD_FETCH:
207 		return "Fetch of CD caused external abort";
208 	case IVPU_MMU_EVT_C_BAD_CD:
209 		return "Fetched CD invalid";
210 	case IVPU_MMU_EVT_F_WALK_EABT:
211 		return " An external abort occurred fetching a TLB";
212 	case IVPU_MMU_EVT_F_TRANSLATION:
213 		return "Translation fault";
214 	case IVPU_MMU_EVT_F_ADDR_SIZE:
215 		return " Output address caused address size fault";
216 	case IVPU_MMU_EVT_F_ACCESS:
217 		return "Access flag fault";
218 	case IVPU_MMU_EVT_F_PERMISSION:
219 		return "Permission fault occurred on page access";
220 	case IVPU_MMU_EVT_F_TLB_CONFLICT:
221 		return "A TLB conflict";
222 	case IVPU_MMU_EVT_F_CFG_CONFLICT:
223 		return "A configuration cache conflict";
224 	case IVPU_MMU_EVT_E_PAGE_REQUEST:
225 		return "Page request hint from a client device";
226 	case IVPU_MMU_EVT_F_VMS_FETCH:
227 		return "Fetch of VMS caused external abort";
228 	default:
229 		return "Unknown CMDQ command";
230 	}
231 }
232 
233 static void ivpu_mmu_config_check(struct ivpu_device *vdev)
234 {
235 	u32 val_ref;
236 	u32 val;
237 
238 	if (ivpu_is_simics(vdev))
239 		val_ref = IVPU_MMU_IDR0_REF_SIMICS;
240 	else
241 		val_ref = IVPU_MMU_IDR0_REF;
242 
243 	val = REGV_RD32(MTL_VPU_HOST_MMU_IDR0);
244 	if (val != val_ref)
245 		ivpu_dbg(vdev, MMU, "IDR0 0x%x != IDR0_REF 0x%x\n", val, val_ref);
246 
247 	val = REGV_RD32(MTL_VPU_HOST_MMU_IDR1);
248 	if (val != IVPU_MMU_IDR1_REF)
249 		ivpu_dbg(vdev, MMU, "IDR1 0x%x != IDR1_REF 0x%x\n", val, IVPU_MMU_IDR1_REF);
250 
251 	val = REGV_RD32(MTL_VPU_HOST_MMU_IDR3);
252 	if (val != IVPU_MMU_IDR3_REF)
253 		ivpu_dbg(vdev, MMU, "IDR3 0x%x != IDR3_REF 0x%x\n", val, IVPU_MMU_IDR3_REF);
254 
255 	if (ivpu_is_simics(vdev))
256 		val_ref = IVPU_MMU_IDR5_REF_SIMICS;
257 	else if (ivpu_is_fpga(vdev))
258 		val_ref = IVPU_MMU_IDR5_REF_FPGA;
259 	else
260 		val_ref = IVPU_MMU_IDR5_REF;
261 
262 	val = REGV_RD32(MTL_VPU_HOST_MMU_IDR5);
263 	if (val != val_ref)
264 		ivpu_dbg(vdev, MMU, "IDR5 0x%x != IDR5_REF 0x%x\n", val, val_ref);
265 }
266 
267 static int ivpu_mmu_cdtab_alloc(struct ivpu_device *vdev)
268 {
269 	struct ivpu_mmu_info *mmu = vdev->mmu;
270 	struct ivpu_mmu_cdtab *cdtab = &mmu->cdtab;
271 	size_t size = IVPU_MMU_CDTAB_ENT_COUNT * IVPU_MMU_CDTAB_ENT_SIZE;
272 
273 	cdtab->base = dmam_alloc_coherent(vdev->drm.dev, size, &cdtab->dma, GFP_KERNEL);
274 	if (!cdtab->base)
275 		return -ENOMEM;
276 
277 	ivpu_dbg(vdev, MMU, "CDTAB alloc: dma=%pad size=%zu\n", &cdtab->dma, size);
278 
279 	return 0;
280 }
281 
282 static int ivpu_mmu_strtab_alloc(struct ivpu_device *vdev)
283 {
284 	struct ivpu_mmu_info *mmu = vdev->mmu;
285 	struct ivpu_mmu_strtab *strtab = &mmu->strtab;
286 	size_t size = IVPU_MMU_STRTAB_ENT_COUNT * IVPU_MMU_STRTAB_ENT_SIZE;
287 
288 	strtab->base = dmam_alloc_coherent(vdev->drm.dev, size, &strtab->dma, GFP_KERNEL);
289 	if (!strtab->base)
290 		return -ENOMEM;
291 
292 	strtab->base_cfg = IVPU_MMU_STRTAB_CFG;
293 	strtab->dma_q = IVPU_MMU_STRTAB_BASE_RA;
294 	strtab->dma_q |= strtab->dma & IVPU_MMU_STRTAB_BASE_ADDR_MASK;
295 
296 	ivpu_dbg(vdev, MMU, "STRTAB alloc: dma=%pad dma_q=%pad size=%zu\n",
297 		 &strtab->dma, &strtab->dma_q, size);
298 
299 	return 0;
300 }
301 
302 static int ivpu_mmu_cmdq_alloc(struct ivpu_device *vdev)
303 {
304 	struct ivpu_mmu_info *mmu = vdev->mmu;
305 	struct ivpu_mmu_queue *q = &mmu->cmdq;
306 
307 	q->base = dmam_alloc_coherent(vdev->drm.dev, IVPU_MMU_CMDQ_SIZE, &q->dma, GFP_KERNEL);
308 	if (!q->base)
309 		return -ENOMEM;
310 
311 	q->dma_q = IVPU_MMU_Q_BASE_RWA;
312 	q->dma_q |= q->dma & IVPU_MMU_Q_BASE_ADDR_MASK;
313 	q->dma_q |= IVPU_MMU_Q_COUNT_LOG2;
314 
315 	ivpu_dbg(vdev, MMU, "CMDQ alloc: dma=%pad dma_q=%pad size=%u\n",
316 		 &q->dma, &q->dma_q, IVPU_MMU_CMDQ_SIZE);
317 
318 	return 0;
319 }
320 
321 static int ivpu_mmu_evtq_alloc(struct ivpu_device *vdev)
322 {
323 	struct ivpu_mmu_info *mmu = vdev->mmu;
324 	struct ivpu_mmu_queue *q = &mmu->evtq;
325 
326 	q->base = dmam_alloc_coherent(vdev->drm.dev, IVPU_MMU_EVTQ_SIZE, &q->dma, GFP_KERNEL);
327 	if (!q->base)
328 		return -ENOMEM;
329 
330 	q->dma_q = IVPU_MMU_Q_BASE_RWA;
331 	q->dma_q |= q->dma & IVPU_MMU_Q_BASE_ADDR_MASK;
332 	q->dma_q |= IVPU_MMU_Q_COUNT_LOG2;
333 
334 	ivpu_dbg(vdev, MMU, "EVTQ alloc: dma=%pad dma_q=%pad size=%u\n",
335 		 &q->dma, &q->dma_q, IVPU_MMU_EVTQ_SIZE);
336 
337 	return 0;
338 }
339 
340 static int ivpu_mmu_structs_alloc(struct ivpu_device *vdev)
341 {
342 	int ret;
343 
344 	ret = ivpu_mmu_cdtab_alloc(vdev);
345 	if (ret) {
346 		ivpu_err(vdev, "Failed to allocate cdtab: %d\n", ret);
347 		return ret;
348 	}
349 
350 	ret = ivpu_mmu_strtab_alloc(vdev);
351 	if (ret) {
352 		ivpu_err(vdev, "Failed to allocate strtab: %d\n", ret);
353 		return ret;
354 	}
355 
356 	ret = ivpu_mmu_cmdq_alloc(vdev);
357 	if (ret) {
358 		ivpu_err(vdev, "Failed to allocate cmdq: %d\n", ret);
359 		return ret;
360 	}
361 
362 	ret = ivpu_mmu_evtq_alloc(vdev);
363 	if (ret)
364 		ivpu_err(vdev, "Failed to allocate evtq: %d\n", ret);
365 
366 	return ret;
367 }
368 
369 static int ivpu_mmu_reg_write(struct ivpu_device *vdev, u32 reg, u32 val)
370 {
371 	u32 reg_ack = reg + 4; /* ACK register is 4B after base register */
372 	u32 val_ack;
373 	int ret;
374 
375 	REGV_WR32(reg, val);
376 
377 	ret = REGV_POLL(reg_ack, val_ack, (val == val_ack), IVPU_MMU_REG_TIMEOUT_US);
378 	if (ret)
379 		ivpu_err(vdev, "Failed to write register 0x%x\n", reg);
380 
381 	return ret;
382 }
383 
384 static int ivpu_mmu_irqs_setup(struct ivpu_device *vdev)
385 {
386 	u32 irq_ctrl = IVPU_MMU_IRQ_EVTQ_EN | IVPU_MMU_IRQ_GERROR_EN;
387 	int ret;
388 
389 	ret = ivpu_mmu_reg_write(vdev, MTL_VPU_HOST_MMU_IRQ_CTRL, 0);
390 	if (ret)
391 		return ret;
392 
393 	return ivpu_mmu_reg_write(vdev, MTL_VPU_HOST_MMU_IRQ_CTRL, irq_ctrl);
394 }
395 
396 static int ivpu_mmu_cmdq_wait_for_cons(struct ivpu_device *vdev)
397 {
398 	struct ivpu_mmu_queue *cmdq = &vdev->mmu->cmdq;
399 
400 	return REGV_POLL(MTL_VPU_HOST_MMU_CMDQ_CONS, cmdq->cons, (cmdq->prod == cmdq->cons),
401 			 IVPU_MMU_QUEUE_TIMEOUT_US);
402 }
403 
404 static int ivpu_mmu_cmdq_cmd_write(struct ivpu_device *vdev, const char *name, u64 data0, u64 data1)
405 {
406 	struct ivpu_mmu_queue *q = &vdev->mmu->cmdq;
407 	u64 *queue_buffer = q->base;
408 	int idx = IVPU_MMU_Q_IDX(q->prod) * (IVPU_MMU_CMDQ_CMD_SIZE / sizeof(*queue_buffer));
409 
410 	if (!CIRC_SPACE(IVPU_MMU_Q_IDX(q->prod), IVPU_MMU_Q_IDX(q->cons), IVPU_MMU_Q_COUNT)) {
411 		ivpu_err(vdev, "Failed to write MMU CMD %s\n", name);
412 		return -EBUSY;
413 	}
414 
415 	queue_buffer[idx] = data0;
416 	queue_buffer[idx + 1] = data1;
417 	q->prod = (q->prod + 1) & IVPU_MMU_Q_WRAP_MASK;
418 
419 	ivpu_dbg(vdev, MMU, "CMD write: %s data: 0x%llx 0x%llx\n", name, data0, data1);
420 
421 	return 0;
422 }
423 
424 static int ivpu_mmu_cmdq_sync(struct ivpu_device *vdev)
425 {
426 	struct ivpu_mmu_queue *q = &vdev->mmu->cmdq;
427 	u64 val;
428 	int ret;
429 
430 	val = FIELD_PREP(IVPU_MMU_CMD_OPCODE, CMD_SYNC) |
431 	      FIELD_PREP(IVPU_MMU_CMD_SYNC_0_CS, 0x2) |
432 	      FIELD_PREP(IVPU_MMU_CMD_SYNC_0_MSH, 0x3) |
433 	      FIELD_PREP(IVPU_MMU_CMD_SYNC_0_MSI_ATTR, 0xf);
434 
435 	ret = ivpu_mmu_cmdq_cmd_write(vdev, "SYNC", val, 0);
436 	if (ret)
437 		return ret;
438 
439 	clflush_cache_range(q->base, IVPU_MMU_CMDQ_SIZE);
440 	REGV_WR32(MTL_VPU_HOST_MMU_CMDQ_PROD, q->prod);
441 
442 	ret = ivpu_mmu_cmdq_wait_for_cons(vdev);
443 	if (ret)
444 		ivpu_err(vdev, "Timed out waiting for consumer: %d\n", ret);
445 
446 	return ret;
447 }
448 
449 static int ivpu_mmu_cmdq_write_cfgi_all(struct ivpu_device *vdev)
450 {
451 	u64 data0 = FIELD_PREP(IVPU_MMU_CMD_OPCODE, CMD_CFGI_ALL);
452 	u64 data1 = FIELD_PREP(IVPU_MMU_CMD_CFGI_1_RANGE, 0x1f);
453 
454 	return ivpu_mmu_cmdq_cmd_write(vdev, "CFGI_ALL", data0, data1);
455 }
456 
457 static int ivpu_mmu_cmdq_write_tlbi_nh_asid(struct ivpu_device *vdev, u16 ssid)
458 {
459 	u64 val = FIELD_PREP(IVPU_MMU_CMD_OPCODE, CMD_TLBI_NH_ASID) |
460 		  FIELD_PREP(IVPU_MMU_CMD_TLBI_0_ASID, ssid);
461 
462 	return ivpu_mmu_cmdq_cmd_write(vdev, "TLBI_NH_ASID", val, 0);
463 }
464 
465 static int ivpu_mmu_cmdq_write_tlbi_nsnh_all(struct ivpu_device *vdev)
466 {
467 	u64 val = FIELD_PREP(IVPU_MMU_CMD_OPCODE, CMD_TLBI_NSNH_ALL);
468 
469 	return ivpu_mmu_cmdq_cmd_write(vdev, "TLBI_NSNH_ALL", val, 0);
470 }
471 
472 static int ivpu_mmu_reset(struct ivpu_device *vdev)
473 {
474 	struct ivpu_mmu_info *mmu = vdev->mmu;
475 	u32 val;
476 	int ret;
477 
478 	memset(mmu->cmdq.base, 0, IVPU_MMU_CMDQ_SIZE);
479 	clflush_cache_range(mmu->cmdq.base, IVPU_MMU_CMDQ_SIZE);
480 	mmu->cmdq.prod = 0;
481 	mmu->cmdq.cons = 0;
482 
483 	memset(mmu->evtq.base, 0, IVPU_MMU_EVTQ_SIZE);
484 	clflush_cache_range(mmu->evtq.base, IVPU_MMU_EVTQ_SIZE);
485 	mmu->evtq.prod = 0;
486 	mmu->evtq.cons = 0;
487 
488 	ret = ivpu_mmu_reg_write(vdev, MTL_VPU_HOST_MMU_CR0, 0);
489 	if (ret)
490 		return ret;
491 
492 	val = FIELD_PREP(IVPU_MMU_CR1_TABLE_SH, IVPU_MMU_SH_ISH) |
493 	      FIELD_PREP(IVPU_MMU_CR1_TABLE_OC, IVPU_MMU_CACHE_WB) |
494 	      FIELD_PREP(IVPU_MMU_CR1_TABLE_IC, IVPU_MMU_CACHE_WB) |
495 	      FIELD_PREP(IVPU_MMU_CR1_QUEUE_SH, IVPU_MMU_SH_ISH) |
496 	      FIELD_PREP(IVPU_MMU_CR1_QUEUE_OC, IVPU_MMU_CACHE_WB) |
497 	      FIELD_PREP(IVPU_MMU_CR1_QUEUE_IC, IVPU_MMU_CACHE_WB);
498 	REGV_WR32(MTL_VPU_HOST_MMU_CR1, val);
499 
500 	REGV_WR64(MTL_VPU_HOST_MMU_STRTAB_BASE, mmu->strtab.dma_q);
501 	REGV_WR32(MTL_VPU_HOST_MMU_STRTAB_BASE_CFG, mmu->strtab.base_cfg);
502 
503 	REGV_WR64(MTL_VPU_HOST_MMU_CMDQ_BASE, mmu->cmdq.dma_q);
504 	REGV_WR32(MTL_VPU_HOST_MMU_CMDQ_PROD, 0);
505 	REGV_WR32(MTL_VPU_HOST_MMU_CMDQ_CONS, 0);
506 
507 	val = IVPU_MMU_CR0_CMDQEN;
508 	ret = ivpu_mmu_reg_write(vdev, MTL_VPU_HOST_MMU_CR0, val);
509 	if (ret)
510 		return ret;
511 
512 	ret = ivpu_mmu_cmdq_write_cfgi_all(vdev);
513 	if (ret)
514 		return ret;
515 
516 	ret = ivpu_mmu_cmdq_write_tlbi_nsnh_all(vdev);
517 	if (ret)
518 		return ret;
519 
520 	ret = ivpu_mmu_cmdq_sync(vdev);
521 	if (ret)
522 		return ret;
523 
524 	REGV_WR64(MTL_VPU_HOST_MMU_EVTQ_BASE, mmu->evtq.dma_q);
525 	REGV_WR32(MTL_VPU_HOST_MMU_EVTQ_PROD_SEC, 0);
526 	REGV_WR32(MTL_VPU_HOST_MMU_EVTQ_CONS_SEC, 0);
527 
528 	val |= IVPU_MMU_CR0_EVTQEN;
529 	ret = ivpu_mmu_reg_write(vdev, MTL_VPU_HOST_MMU_CR0, val);
530 	if (ret)
531 		return ret;
532 
533 	val |= IVPU_MMU_CR0_ATSCHK;
534 	ret = ivpu_mmu_reg_write(vdev, MTL_VPU_HOST_MMU_CR0, val);
535 	if (ret)
536 		return ret;
537 
538 	ret = ivpu_mmu_irqs_setup(vdev);
539 	if (ret)
540 		return ret;
541 
542 	val |= IVPU_MMU_CR0_SMMUEN;
543 	return ivpu_mmu_reg_write(vdev, MTL_VPU_HOST_MMU_CR0, val);
544 }
545 
546 static void ivpu_mmu_strtab_link_cd(struct ivpu_device *vdev, u32 sid)
547 {
548 	struct ivpu_mmu_info *mmu = vdev->mmu;
549 	struct ivpu_mmu_strtab *strtab = &mmu->strtab;
550 	struct ivpu_mmu_cdtab *cdtab = &mmu->cdtab;
551 	u64 *entry = strtab->base + (sid * IVPU_MMU_STRTAB_ENT_SIZE);
552 	u64 str[2];
553 
554 	str[0] = FIELD_PREP(IVPU_MMU_STE_0_CFG, IVPU_MMU_STE_0_CFG_S1_TRANS) |
555 		 FIELD_PREP(IVPU_MMU_STE_0_S1CDMAX, IVPU_MMU_CDTAB_ENT_COUNT_LOG2) |
556 		 FIELD_PREP(IVPU_MMU_STE_0_S1FMT, IVPU_MMU_STE_0_S1FMT_LINEAR) |
557 		 IVPU_MMU_STE_0_V |
558 		 (cdtab->dma & IVPU_MMU_STE_0_S1CTXPTR_MASK);
559 
560 	str[1] = FIELD_PREP(IVPU_MMU_STE_1_S1DSS, IVPU_MMU_STE_1_S1DSS_TERMINATE) |
561 		 FIELD_PREP(IVPU_MMU_STE_1_S1CIR, IVPU_MMU_STE_1_S1C_CACHE_NC) |
562 		 FIELD_PREP(IVPU_MMU_STE_1_S1COR, IVPU_MMU_STE_1_S1C_CACHE_NC) |
563 		 FIELD_PREP(IVPU_MMU_STE_1_S1CSH, IVPU_MMU_SH_NSH) |
564 		 FIELD_PREP(IVPU_MMU_STE_1_PRIVCFG, IVPU_MMU_STE_1_PRIVCFG_UNPRIV) |
565 		 FIELD_PREP(IVPU_MMU_STE_1_INSTCFG, IVPU_MMU_STE_1_INSTCFG_DATA) |
566 		 FIELD_PREP(IVPU_MMU_STE_1_STRW, IVPU_MMU_STE_1_STRW_NSEL1) |
567 		 FIELD_PREP(IVPU_MMU_STE_1_CONT, IVPU_MMU_STRTAB_CFG_LOG2SIZE) |
568 		 IVPU_MMU_STE_1_MEV |
569 		 IVPU_MMU_STE_1_S1STALLD;
570 
571 	WRITE_ONCE(entry[1], str[1]);
572 	WRITE_ONCE(entry[0], str[0]);
573 
574 	clflush_cache_range(entry, IVPU_MMU_STRTAB_ENT_SIZE);
575 
576 	ivpu_dbg(vdev, MMU, "STRTAB write entry (SSID=%u): 0x%llx, 0x%llx\n", sid, str[0], str[1]);
577 }
578 
579 static int ivpu_mmu_strtab_init(struct ivpu_device *vdev)
580 {
581 	ivpu_mmu_strtab_link_cd(vdev, IVPU_MMU_STREAM_ID0);
582 	ivpu_mmu_strtab_link_cd(vdev, IVPU_MMU_STREAM_ID3);
583 
584 	return 0;
585 }
586 
587 int ivpu_mmu_invalidate_tlb(struct ivpu_device *vdev, u16 ssid)
588 {
589 	struct ivpu_mmu_info *mmu = vdev->mmu;
590 	int ret = 0;
591 
592 	mutex_lock(&mmu->lock);
593 	if (!mmu->on)
594 		goto unlock;
595 
596 	ret = ivpu_mmu_cmdq_write_tlbi_nh_asid(vdev, ssid);
597 	if (ret)
598 		goto unlock;
599 
600 	ret = ivpu_mmu_cmdq_sync(vdev);
601 unlock:
602 	mutex_unlock(&mmu->lock);
603 	return ret;
604 }
605 
606 static int ivpu_mmu_cd_add(struct ivpu_device *vdev, u32 ssid, u64 cd_dma)
607 {
608 	struct ivpu_mmu_info *mmu = vdev->mmu;
609 	struct ivpu_mmu_cdtab *cdtab = &mmu->cdtab;
610 	u64 *entry;
611 	u64 cd[4];
612 	int ret = 0;
613 
614 	if (ssid > IVPU_MMU_CDTAB_ENT_COUNT)
615 		return -EINVAL;
616 
617 	entry = cdtab->base + (ssid * IVPU_MMU_CDTAB_ENT_SIZE);
618 
619 	if (cd_dma != 0) {
620 		cd[0] = FIELD_PREP(IVPU_MMU_CD_0_TCR_T0SZ, 26) |
621 			FIELD_PREP(IVPU_MMU_CD_0_TCR_TG0, 0) |
622 			FIELD_PREP(IVPU_MMU_CD_0_TCR_IRGN0, 0) |
623 			FIELD_PREP(IVPU_MMU_CD_0_TCR_ORGN0, 0) |
624 			FIELD_PREP(IVPU_MMU_CD_0_TCR_SH0, 0) |
625 			FIELD_PREP(IVPU_MMU_CD_0_TCR_IPS, 3) |
626 			FIELD_PREP(IVPU_MMU_CD_0_ASID, ssid) |
627 			IVPU_MMU_CD_0_TCR_EPD1 |
628 			IVPU_MMU_CD_0_AA64 |
629 			IVPU_MMU_CD_0_R |
630 			IVPU_MMU_CD_0_ASET |
631 			IVPU_MMU_CD_0_V;
632 		cd[1] = cd_dma & IVPU_MMU_CD_1_TTB0_MASK;
633 		cd[2] = 0;
634 		cd[3] = 0x0000000000007444;
635 
636 		/* For global context generate memory fault on VPU */
637 		if (ssid == IVPU_GLOBAL_CONTEXT_MMU_SSID)
638 			cd[0] |= IVPU_MMU_CD_0_A;
639 	} else {
640 		memset(cd, 0, sizeof(cd));
641 	}
642 
643 	WRITE_ONCE(entry[1], cd[1]);
644 	WRITE_ONCE(entry[2], cd[2]);
645 	WRITE_ONCE(entry[3], cd[3]);
646 	WRITE_ONCE(entry[0], cd[0]);
647 
648 	clflush_cache_range(entry, IVPU_MMU_CDTAB_ENT_SIZE);
649 
650 	ivpu_dbg(vdev, MMU, "CDTAB %s entry (SSID=%u, dma=%pad): 0x%llx, 0x%llx, 0x%llx, 0x%llx\n",
651 		 cd_dma ? "write" : "clear", ssid, &cd_dma, cd[0], cd[1], cd[2], cd[3]);
652 
653 	mutex_lock(&mmu->lock);
654 	if (!mmu->on)
655 		goto unlock;
656 
657 	ret = ivpu_mmu_cmdq_write_cfgi_all(vdev);
658 	if (ret)
659 		goto unlock;
660 
661 	ret = ivpu_mmu_cmdq_sync(vdev);
662 unlock:
663 	mutex_unlock(&mmu->lock);
664 	return ret;
665 }
666 
667 static int ivpu_mmu_cd_add_gbl(struct ivpu_device *vdev)
668 {
669 	int ret;
670 
671 	ret = ivpu_mmu_cd_add(vdev, 0, vdev->gctx.pgtable.pgd_dma);
672 	if (ret)
673 		ivpu_err(vdev, "Failed to add global CD entry: %d\n", ret);
674 
675 	return ret;
676 }
677 
678 static int ivpu_mmu_cd_add_user(struct ivpu_device *vdev, u32 ssid, dma_addr_t cd_dma)
679 {
680 	int ret;
681 
682 	if (ssid == 0) {
683 		ivpu_err(vdev, "Invalid SSID: %u\n", ssid);
684 		return -EINVAL;
685 	}
686 
687 	ret = ivpu_mmu_cd_add(vdev, ssid, cd_dma);
688 	if (ret)
689 		ivpu_err(vdev, "Failed to add CD entry SSID=%u: %d\n", ssid, ret);
690 
691 	return ret;
692 }
693 
694 int ivpu_mmu_init(struct ivpu_device *vdev)
695 {
696 	struct ivpu_mmu_info *mmu = vdev->mmu;
697 	int ret;
698 
699 	ivpu_dbg(vdev, MMU, "Init..\n");
700 
701 	drmm_mutex_init(&vdev->drm, &mmu->lock);
702 	ivpu_mmu_config_check(vdev);
703 
704 	ret = ivpu_mmu_structs_alloc(vdev);
705 	if (ret)
706 		return ret;
707 
708 	ret = ivpu_mmu_strtab_init(vdev);
709 	if (ret) {
710 		ivpu_err(vdev, "Failed to initialize strtab: %d\n", ret);
711 		return ret;
712 	}
713 
714 	ret = ivpu_mmu_cd_add_gbl(vdev);
715 	if (ret) {
716 		ivpu_err(vdev, "Failed to initialize strtab: %d\n", ret);
717 		return ret;
718 	}
719 
720 	ret = ivpu_mmu_enable(vdev);
721 	if (ret) {
722 		ivpu_err(vdev, "Failed to resume MMU: %d\n", ret);
723 		return ret;
724 	}
725 
726 	ivpu_dbg(vdev, MMU, "Init done\n");
727 
728 	return 0;
729 }
730 
731 int ivpu_mmu_enable(struct ivpu_device *vdev)
732 {
733 	struct ivpu_mmu_info *mmu = vdev->mmu;
734 	int ret;
735 
736 	mutex_lock(&mmu->lock);
737 
738 	mmu->on = true;
739 
740 	ret = ivpu_mmu_reset(vdev);
741 	if (ret) {
742 		ivpu_err(vdev, "Failed to reset MMU: %d\n", ret);
743 		goto err;
744 	}
745 
746 	ret = ivpu_mmu_cmdq_write_cfgi_all(vdev);
747 	if (ret)
748 		goto err;
749 
750 	ret = ivpu_mmu_cmdq_write_tlbi_nsnh_all(vdev);
751 	if (ret)
752 		goto err;
753 
754 	ret = ivpu_mmu_cmdq_sync(vdev);
755 	if (ret)
756 		goto err;
757 
758 	mutex_unlock(&mmu->lock);
759 
760 	return 0;
761 err:
762 	mmu->on = false;
763 	mutex_unlock(&mmu->lock);
764 	return ret;
765 }
766 
767 void ivpu_mmu_disable(struct ivpu_device *vdev)
768 {
769 	struct ivpu_mmu_info *mmu = vdev->mmu;
770 
771 	mutex_lock(&mmu->lock);
772 	mmu->on = false;
773 	mutex_unlock(&mmu->lock);
774 }
775 
776 static void ivpu_mmu_dump_event(struct ivpu_device *vdev, u32 *event)
777 {
778 	u32 ssid = FIELD_GET(IVPU_MMU_EVT_SSID_MASK, event[0]);
779 	u32 op = FIELD_GET(IVPU_MMU_EVT_OP_MASK, event[0]);
780 	u64 fetch_addr = ((u64)event[7]) << 32 | event[6];
781 	u64 in_addr = ((u64)event[5]) << 32 | event[4];
782 	u32 sid = event[1];
783 
784 	ivpu_err(vdev, "MMU EVTQ: 0x%x (%s) SSID: %d SID: %d, e[2] %08x, e[3] %08x, in addr: 0x%llx, fetch addr: 0x%llx\n",
785 		 op, ivpu_mmu_event_to_str(op), ssid, sid, event[2], event[3], in_addr, fetch_addr);
786 }
787 
788 static u32 *ivpu_mmu_get_event(struct ivpu_device *vdev)
789 {
790 	struct ivpu_mmu_queue *evtq = &vdev->mmu->evtq;
791 	u32 idx = IVPU_MMU_Q_IDX(evtq->cons);
792 	u32 *evt = evtq->base + (idx * IVPU_MMU_EVTQ_CMD_SIZE);
793 
794 	evtq->prod = REGV_RD32(MTL_VPU_HOST_MMU_EVTQ_PROD_SEC);
795 	if (!CIRC_CNT(IVPU_MMU_Q_IDX(evtq->prod), IVPU_MMU_Q_IDX(evtq->cons), IVPU_MMU_Q_COUNT))
796 		return NULL;
797 
798 	clflush_cache_range(evt, IVPU_MMU_EVTQ_CMD_SIZE);
799 
800 	evtq->cons = (evtq->cons + 1) & IVPU_MMU_Q_WRAP_MASK;
801 	REGV_WR32(MTL_VPU_HOST_MMU_EVTQ_CONS_SEC, evtq->cons);
802 
803 	return evt;
804 }
805 
806 void ivpu_mmu_irq_evtq_handler(struct ivpu_device *vdev)
807 {
808 	bool schedule_recovery = false;
809 	u32 *event;
810 	u32 ssid;
811 
812 	ivpu_dbg(vdev, IRQ, "MMU event queue\n");
813 
814 	while ((event = ivpu_mmu_get_event(vdev)) != NULL) {
815 		ivpu_mmu_dump_event(vdev, event);
816 
817 		ssid = FIELD_GET(IVPU_MMU_EVT_SSID_MASK, event[0]);
818 		if (ssid == IVPU_GLOBAL_CONTEXT_MMU_SSID)
819 			schedule_recovery = true;
820 		else
821 			ivpu_mmu_user_context_mark_invalid(vdev, ssid);
822 	}
823 
824 	if (schedule_recovery)
825 		ivpu_pm_schedule_recovery(vdev);
826 }
827 
828 void ivpu_mmu_irq_gerr_handler(struct ivpu_device *vdev)
829 {
830 	u32 gerror_val, gerrorn_val, active;
831 
832 	ivpu_dbg(vdev, IRQ, "MMU error\n");
833 
834 	gerror_val = REGV_RD32(MTL_VPU_HOST_MMU_GERROR);
835 	gerrorn_val = REGV_RD32(MTL_VPU_HOST_MMU_GERRORN);
836 
837 	active = gerror_val ^ gerrorn_val;
838 	if (!(active & IVPU_MMU_GERROR_ERR_MASK))
839 		return;
840 
841 	if (REG_TEST_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_ABT, active))
842 		ivpu_warn_ratelimited(vdev, "MMU MSI ABT write aborted\n");
843 
844 	if (REG_TEST_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_PRIQ_ABT, active))
845 		ivpu_warn_ratelimited(vdev, "MMU PRIQ MSI ABT write aborted\n");
846 
847 	if (REG_TEST_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_EVTQ_ABT, active))
848 		ivpu_warn_ratelimited(vdev, "MMU EVTQ MSI ABT write aborted\n");
849 
850 	if (REG_TEST_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_CMDQ_ABT, active))
851 		ivpu_warn_ratelimited(vdev, "MMU CMDQ MSI ABT write aborted\n");
852 
853 	if (REG_TEST_FLD(MTL_VPU_HOST_MMU_GERROR, PRIQ_ABT, active))
854 		ivpu_err_ratelimited(vdev, "MMU PRIQ write aborted\n");
855 
856 	if (REG_TEST_FLD(MTL_VPU_HOST_MMU_GERROR, EVTQ_ABT, active))
857 		ivpu_err_ratelimited(vdev, "MMU EVTQ write aborted\n");
858 
859 	if (REG_TEST_FLD(MTL_VPU_HOST_MMU_GERROR, CMDQ, active))
860 		ivpu_err_ratelimited(vdev, "MMU CMDQ write aborted\n");
861 
862 	REGV_WR32(MTL_VPU_HOST_MMU_GERRORN, gerror_val);
863 }
864 
865 int ivpu_mmu_set_pgtable(struct ivpu_device *vdev, int ssid, struct ivpu_mmu_pgtable *pgtable)
866 {
867 	return ivpu_mmu_cd_add_user(vdev, ssid, pgtable->pgd_dma);
868 }
869 
870 void ivpu_mmu_clear_pgtable(struct ivpu_device *vdev, int ssid)
871 {
872 	ivpu_mmu_cd_add_user(vdev, ssid, 0); /* 0 will clear CD entry */
873 }
874