xref: /openbmc/linux/drivers/spi/spi-sh-msiof.c (revision 236a9bf2)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * SuperH MSIOF SPI Controller Interface
4  *
5  * Copyright (c) 2009 Magnus Damm
6  * Copyright (C) 2014 Renesas Electronics Corporation
7  * Copyright (C) 2014-2017 Glider bvba
8  */
9 
10 #include <linux/bitmap.h>
11 #include <linux/clk.h>
12 #include <linux/completion.h>
13 #include <linux/delay.h>
14 #include <linux/dma-mapping.h>
15 #include <linux/dmaengine.h>
16 #include <linux/err.h>
17 #include <linux/interrupt.h>
18 #include <linux/io.h>
19 #include <linux/iopoll.h>
20 #include <linux/kernel.h>
21 #include <linux/module.h>
22 #include <linux/of.h>
23 #include <linux/platform_device.h>
24 #include <linux/pm_runtime.h>
25 #include <linux/sh_dma.h>
26 
27 #include <linux/spi/sh_msiof.h>
28 #include <linux/spi/spi.h>
29 
30 #include <asm/unaligned.h>
31 
32 #define SH_MSIOF_FLAG_FIXED_DTDL_200	BIT(0)
33 
34 struct sh_msiof_chipdata {
35 	u32 bits_per_word_mask;
36 	u16 tx_fifo_size;
37 	u16 rx_fifo_size;
38 	u16 ctlr_flags;
39 	u16 min_div_pow;
40 	u32 flags;
41 };
42 
43 struct sh_msiof_spi_priv {
44 	struct spi_controller *ctlr;
45 	void __iomem *mapbase;
46 	struct clk *clk;
47 	struct platform_device *pdev;
48 	struct sh_msiof_spi_info *info;
49 	struct completion done;
50 	struct completion done_txdma;
51 	unsigned int tx_fifo_size;
52 	unsigned int rx_fifo_size;
53 	unsigned int min_div_pow;
54 	void *tx_dma_page;
55 	void *rx_dma_page;
56 	dma_addr_t tx_dma_addr;
57 	dma_addr_t rx_dma_addr;
58 	bool native_cs_inited;
59 	bool native_cs_high;
60 	bool target_aborted;
61 };
62 
63 #define MAX_SS	3	/* Maximum number of native chip selects */
64 
65 #define SITMDR1	0x00	/* Transmit Mode Register 1 */
66 #define SITMDR2	0x04	/* Transmit Mode Register 2 */
67 #define SITMDR3	0x08	/* Transmit Mode Register 3 */
68 #define SIRMDR1	0x10	/* Receive Mode Register 1 */
69 #define SIRMDR2	0x14	/* Receive Mode Register 2 */
70 #define SIRMDR3	0x18	/* Receive Mode Register 3 */
71 #define SITSCR	0x20	/* Transmit Clock Select Register */
72 #define SIRSCR	0x22	/* Receive Clock Select Register (SH, A1, APE6) */
73 #define SICTR	0x28	/* Control Register */
74 #define SIFCTR	0x30	/* FIFO Control Register */
75 #define SISTR	0x40	/* Status Register */
76 #define SIIER	0x44	/* Interrupt Enable Register */
77 #define SITDR1	0x48	/* Transmit Control Data Register 1 (SH, A1) */
78 #define SITDR2	0x4c	/* Transmit Control Data Register 2 (SH, A1) */
79 #define SITFDR	0x50	/* Transmit FIFO Data Register */
80 #define SIRDR1	0x58	/* Receive Control Data Register 1 (SH, A1) */
81 #define SIRDR2	0x5c	/* Receive Control Data Register 2 (SH, A1) */
82 #define SIRFDR	0x60	/* Receive FIFO Data Register */
83 
84 /* SITMDR1 and SIRMDR1 */
85 #define SIMDR1_TRMD		BIT(31)		/* Transfer Mode (1 = Master mode) */
86 #define SIMDR1_SYNCMD_MASK	GENMASK(29, 28)	/* SYNC Mode */
87 #define SIMDR1_SYNCMD_SPI	(2 << 28)	/*   Level mode/SPI */
88 #define SIMDR1_SYNCMD_LR	(3 << 28)	/*   L/R mode */
89 #define SIMDR1_SYNCAC_SHIFT	25		/* Sync Polarity (1 = Active-low) */
90 #define SIMDR1_BITLSB_SHIFT	24		/* MSB/LSB First (1 = LSB first) */
91 #define SIMDR1_DTDL_SHIFT	20		/* Data Pin Bit Delay for MSIOF_SYNC */
92 #define SIMDR1_SYNCDL_SHIFT	16		/* Frame Sync Signal Timing Delay */
93 #define SIMDR1_FLD_MASK		GENMASK(3, 2)	/* Frame Sync Signal Interval (0-3) */
94 #define SIMDR1_FLD_SHIFT	2
95 #define SIMDR1_XXSTP		BIT(0)		/* Transmission/Reception Stop on FIFO */
96 /* SITMDR1 */
97 #define SITMDR1_PCON		BIT(30)		/* Transfer Signal Connection */
98 #define SITMDR1_SYNCCH_MASK	GENMASK(27, 26)	/* Sync Signal Channel Select */
99 #define SITMDR1_SYNCCH_SHIFT	26		/* 0=MSIOF_SYNC, 1=MSIOF_SS1, 2=MSIOF_SS2 */
100 
101 /* SITMDR2 and SIRMDR2 */
102 #define SIMDR2_BITLEN1(i)	(((i) - 1) << 24) /* Data Size (8-32 bits) */
103 #define SIMDR2_WDLEN1(i)	(((i) - 1) << 16) /* Word Count (1-64/256 (SH, A1))) */
104 #define SIMDR2_GRPMASK1		BIT(0)		/* Group Output Mask 1 (SH, A1) */
105 
106 /* SITSCR and SIRSCR */
107 #define SISCR_BRPS_MASK		GENMASK(12, 8)	/* Prescaler Setting (1-32) */
108 #define SISCR_BRPS(i)		(((i) - 1) << 8)
109 #define SISCR_BRDV_MASK		GENMASK(2, 0)	/* Baud Rate Generator's Division Ratio */
110 #define SISCR_BRDV_DIV_2	0
111 #define SISCR_BRDV_DIV_4	1
112 #define SISCR_BRDV_DIV_8	2
113 #define SISCR_BRDV_DIV_16	3
114 #define SISCR_BRDV_DIV_32	4
115 #define SISCR_BRDV_DIV_1	7
116 
117 /* SICTR */
118 #define SICTR_TSCKIZ_MASK	GENMASK(31, 30)	/* Transmit Clock I/O Polarity Select */
119 #define SICTR_TSCKIZ_SCK	BIT(31)		/*   Disable SCK when TX disabled */
120 #define SICTR_TSCKIZ_POL_SHIFT	30		/*   Transmit Clock Polarity */
121 #define SICTR_RSCKIZ_MASK	GENMASK(29, 28) /* Receive Clock Polarity Select */
122 #define SICTR_RSCKIZ_SCK	BIT(29)		/*   Must match CTR_TSCKIZ_SCK */
123 #define SICTR_RSCKIZ_POL_SHIFT	28		/*   Receive Clock Polarity */
124 #define SICTR_TEDG_SHIFT	27		/* Transmit Timing (1 = falling edge) */
125 #define SICTR_REDG_SHIFT	26		/* Receive Timing (1 = falling edge) */
126 #define SICTR_TXDIZ_MASK	GENMASK(23, 22)	/* Pin Output When TX is Disabled */
127 #define SICTR_TXDIZ_LOW		(0 << 22)	/*   0 */
128 #define SICTR_TXDIZ_HIGH	(1 << 22)	/*   1 */
129 #define SICTR_TXDIZ_HIZ		(2 << 22)	/*   High-impedance */
130 #define SICTR_TSCKE		BIT(15)		/* Transmit Serial Clock Output Enable */
131 #define SICTR_TFSE		BIT(14)		/* Transmit Frame Sync Signal Output Enable */
132 #define SICTR_TXE		BIT(9)		/* Transmit Enable */
133 #define SICTR_RXE		BIT(8)		/* Receive Enable */
134 #define SICTR_TXRST		BIT(1)		/* Transmit Reset */
135 #define SICTR_RXRST		BIT(0)		/* Receive Reset */
136 
137 /* SIFCTR */
138 #define SIFCTR_TFWM_MASK	GENMASK(31, 29)	/* Transmit FIFO Watermark */
139 #define SIFCTR_TFWM_64		(0 << 29)	/*  Transfer Request when 64 empty stages */
140 #define SIFCTR_TFWM_32		(1 << 29)	/*  Transfer Request when 32 empty stages */
141 #define SIFCTR_TFWM_24		(2 << 29)	/*  Transfer Request when 24 empty stages */
142 #define SIFCTR_TFWM_16		(3 << 29)	/*  Transfer Request when 16 empty stages */
143 #define SIFCTR_TFWM_12		(4 << 29)	/*  Transfer Request when 12 empty stages */
144 #define SIFCTR_TFWM_8		(5 << 29)	/*  Transfer Request when 8 empty stages */
145 #define SIFCTR_TFWM_4		(6 << 29)	/*  Transfer Request when 4 empty stages */
146 #define SIFCTR_TFWM_1		(7 << 29)	/*  Transfer Request when 1 empty stage */
147 #define SIFCTR_TFUA_MASK	GENMASK(26, 20) /* Transmit FIFO Usable Area */
148 #define SIFCTR_TFUA_SHIFT	20
149 #define SIFCTR_TFUA(i)		((i) << SIFCTR_TFUA_SHIFT)
150 #define SIFCTR_RFWM_MASK	GENMASK(15, 13)	/* Receive FIFO Watermark */
151 #define SIFCTR_RFWM_1		(0 << 13)	/*  Transfer Request when 1 valid stages */
152 #define SIFCTR_RFWM_4		(1 << 13)	/*  Transfer Request when 4 valid stages */
153 #define SIFCTR_RFWM_8		(2 << 13)	/*  Transfer Request when 8 valid stages */
154 #define SIFCTR_RFWM_16		(3 << 13)	/*  Transfer Request when 16 valid stages */
155 #define SIFCTR_RFWM_32		(4 << 13)	/*  Transfer Request when 32 valid stages */
156 #define SIFCTR_RFWM_64		(5 << 13)	/*  Transfer Request when 64 valid stages */
157 #define SIFCTR_RFWM_128		(6 << 13)	/*  Transfer Request when 128 valid stages */
158 #define SIFCTR_RFWM_256		(7 << 13)	/*  Transfer Request when 256 valid stages */
159 #define SIFCTR_RFUA_MASK	GENMASK(12, 4)	/* Receive FIFO Usable Area (0x40 = full) */
160 #define SIFCTR_RFUA_SHIFT	4
161 #define SIFCTR_RFUA(i)		((i) << SIFCTR_RFUA_SHIFT)
162 
163 /* SISTR */
164 #define SISTR_TFEMP		BIT(29) /* Transmit FIFO Empty */
165 #define SISTR_TDREQ		BIT(28) /* Transmit Data Transfer Request */
166 #define SISTR_TEOF		BIT(23) /* Frame Transmission End */
167 #define SISTR_TFSERR		BIT(21) /* Transmit Frame Synchronization Error */
168 #define SISTR_TFOVF		BIT(20) /* Transmit FIFO Overflow */
169 #define SISTR_TFUDF		BIT(19) /* Transmit FIFO Underflow */
170 #define SISTR_RFFUL		BIT(13) /* Receive FIFO Full */
171 #define SISTR_RDREQ		BIT(12) /* Receive Data Transfer Request */
172 #define SISTR_REOF		BIT(7)  /* Frame Reception End */
173 #define SISTR_RFSERR		BIT(5)  /* Receive Frame Synchronization Error */
174 #define SISTR_RFUDF		BIT(4)  /* Receive FIFO Underflow */
175 #define SISTR_RFOVF		BIT(3)  /* Receive FIFO Overflow */
176 
177 /* SIIER */
178 #define SIIER_TDMAE		BIT(31) /* Transmit Data DMA Transfer Req. Enable */
179 #define SIIER_TFEMPE		BIT(29) /* Transmit FIFO Empty Enable */
180 #define SIIER_TDREQE		BIT(28) /* Transmit Data Transfer Request Enable */
181 #define SIIER_TEOFE		BIT(23) /* Frame Transmission End Enable */
182 #define SIIER_TFSERRE		BIT(21) /* Transmit Frame Sync Error Enable */
183 #define SIIER_TFOVFE		BIT(20) /* Transmit FIFO Overflow Enable */
184 #define SIIER_TFUDFE		BIT(19) /* Transmit FIFO Underflow Enable */
185 #define SIIER_RDMAE		BIT(15) /* Receive Data DMA Transfer Req. Enable */
186 #define SIIER_RFFULE		BIT(13) /* Receive FIFO Full Enable */
187 #define SIIER_RDREQE		BIT(12) /* Receive Data Transfer Request Enable */
188 #define SIIER_REOFE		BIT(7)  /* Frame Reception End Enable */
189 #define SIIER_RFSERRE		BIT(5)  /* Receive Frame Sync Error Enable */
190 #define SIIER_RFUDFE		BIT(4)  /* Receive FIFO Underflow Enable */
191 #define SIIER_RFOVFE		BIT(3)  /* Receive FIFO Overflow Enable */
192 
193 
194 static u32 sh_msiof_read(struct sh_msiof_spi_priv *p, int reg_offs)
195 {
196 	switch (reg_offs) {
197 	case SITSCR:
198 	case SIRSCR:
199 		return ioread16(p->mapbase + reg_offs);
200 	default:
201 		return ioread32(p->mapbase + reg_offs);
202 	}
203 }
204 
205 static void sh_msiof_write(struct sh_msiof_spi_priv *p, int reg_offs,
206 			   u32 value)
207 {
208 	switch (reg_offs) {
209 	case SITSCR:
210 	case SIRSCR:
211 		iowrite16(value, p->mapbase + reg_offs);
212 		break;
213 	default:
214 		iowrite32(value, p->mapbase + reg_offs);
215 		break;
216 	}
217 }
218 
219 static int sh_msiof_modify_ctr_wait(struct sh_msiof_spi_priv *p,
220 				    u32 clr, u32 set)
221 {
222 	u32 mask = clr | set;
223 	u32 data;
224 
225 	data = sh_msiof_read(p, SICTR);
226 	data &= ~clr;
227 	data |= set;
228 	sh_msiof_write(p, SICTR, data);
229 
230 	return readl_poll_timeout_atomic(p->mapbase + SICTR, data,
231 					 (data & mask) == set, 1, 100);
232 }
233 
234 static irqreturn_t sh_msiof_spi_irq(int irq, void *data)
235 {
236 	struct sh_msiof_spi_priv *p = data;
237 
238 	/* just disable the interrupt and wake up */
239 	sh_msiof_write(p, SIIER, 0);
240 	complete(&p->done);
241 
242 	return IRQ_HANDLED;
243 }
244 
245 static void sh_msiof_spi_reset_regs(struct sh_msiof_spi_priv *p)
246 {
247 	u32 mask = SICTR_TXRST | SICTR_RXRST;
248 	u32 data;
249 
250 	data = sh_msiof_read(p, SICTR);
251 	data |= mask;
252 	sh_msiof_write(p, SICTR, data);
253 
254 	readl_poll_timeout_atomic(p->mapbase + SICTR, data, !(data & mask), 1,
255 				  100);
256 }
257 
258 static const u32 sh_msiof_spi_div_array[] = {
259 	SISCR_BRDV_DIV_1, SISCR_BRDV_DIV_2, SISCR_BRDV_DIV_4,
260 	SISCR_BRDV_DIV_8, SISCR_BRDV_DIV_16, SISCR_BRDV_DIV_32,
261 };
262 
263 static void sh_msiof_spi_set_clk_regs(struct sh_msiof_spi_priv *p,
264 				      struct spi_transfer *t)
265 {
266 	unsigned long parent_rate = clk_get_rate(p->clk);
267 	unsigned int div_pow = p->min_div_pow;
268 	u32 spi_hz = t->speed_hz;
269 	unsigned long div;
270 	u32 brps, scr;
271 
272 	if (!spi_hz || !parent_rate) {
273 		WARN(1, "Invalid clock rate parameters %lu and %u\n",
274 		     parent_rate, spi_hz);
275 		return;
276 	}
277 
278 	div = DIV_ROUND_UP(parent_rate, spi_hz);
279 	if (div <= 1024) {
280 		/* SISCR_BRDV_DIV_1 is valid only if BRPS is x 1/1 or x 1/2 */
281 		if (!div_pow && div <= 32 && div > 2)
282 			div_pow = 1;
283 
284 		if (div_pow)
285 			brps = (div + 1) >> div_pow;
286 		else
287 			brps = div;
288 
289 		for (; brps > 32; div_pow++)
290 			brps = (brps + 1) >> 1;
291 	} else {
292 		/* Set transfer rate composite divisor to 2^5 * 32 = 1024 */
293 		dev_err(&p->pdev->dev,
294 			"Requested SPI transfer rate %d is too low\n", spi_hz);
295 		div_pow = 5;
296 		brps = 32;
297 	}
298 
299 	t->effective_speed_hz = parent_rate / (brps << div_pow);
300 
301 	scr = sh_msiof_spi_div_array[div_pow] | SISCR_BRPS(brps);
302 	sh_msiof_write(p, SITSCR, scr);
303 	if (!(p->ctlr->flags & SPI_CONTROLLER_MUST_TX))
304 		sh_msiof_write(p, SIRSCR, scr);
305 }
306 
307 static u32 sh_msiof_get_delay_bit(u32 dtdl_or_syncdl)
308 {
309 	/*
310 	 * DTDL/SYNCDL bit	: p->info->dtdl or p->info->syncdl
311 	 * b'000		: 0
312 	 * b'001		: 100
313 	 * b'010		: 200
314 	 * b'011 (SYNCDL only)	: 300
315 	 * b'101		: 50
316 	 * b'110		: 150
317 	 */
318 	if (dtdl_or_syncdl % 100)
319 		return dtdl_or_syncdl / 100 + 5;
320 	else
321 		return dtdl_or_syncdl / 100;
322 }
323 
324 static u32 sh_msiof_spi_get_dtdl_and_syncdl(struct sh_msiof_spi_priv *p)
325 {
326 	u32 val;
327 
328 	if (!p->info)
329 		return 0;
330 
331 	/* check if DTDL and SYNCDL is allowed value */
332 	if (p->info->dtdl > 200 || p->info->syncdl > 300) {
333 		dev_warn(&p->pdev->dev, "DTDL or SYNCDL is too large\n");
334 		return 0;
335 	}
336 
337 	/* check if the sum of DTDL and SYNCDL becomes an integer value  */
338 	if ((p->info->dtdl + p->info->syncdl) % 100) {
339 		dev_warn(&p->pdev->dev, "the sum of DTDL/SYNCDL is not good\n");
340 		return 0;
341 	}
342 
343 	val = sh_msiof_get_delay_bit(p->info->dtdl) << SIMDR1_DTDL_SHIFT;
344 	val |= sh_msiof_get_delay_bit(p->info->syncdl) << SIMDR1_SYNCDL_SHIFT;
345 
346 	return val;
347 }
348 
349 static void sh_msiof_spi_set_pin_regs(struct sh_msiof_spi_priv *p, u32 ss,
350 				      u32 cpol, u32 cpha,
351 				      u32 tx_hi_z, u32 lsb_first, u32 cs_high)
352 {
353 	u32 tmp;
354 	int edge;
355 
356 	/*
357 	 * CPOL CPHA     TSCKIZ RSCKIZ TEDG REDG
358 	 *    0    0         10     10    1    1
359 	 *    0    1         10     10    0    0
360 	 *    1    0         11     11    0    0
361 	 *    1    1         11     11    1    1
362 	 */
363 	tmp = SIMDR1_SYNCMD_SPI | 1 << SIMDR1_FLD_SHIFT | SIMDR1_XXSTP;
364 	tmp |= !cs_high << SIMDR1_SYNCAC_SHIFT;
365 	tmp |= lsb_first << SIMDR1_BITLSB_SHIFT;
366 	tmp |= sh_msiof_spi_get_dtdl_and_syncdl(p);
367 	if (spi_controller_is_target(p->ctlr)) {
368 		sh_msiof_write(p, SITMDR1, tmp | SITMDR1_PCON);
369 	} else {
370 		sh_msiof_write(p, SITMDR1,
371 			       tmp | SIMDR1_TRMD | SITMDR1_PCON |
372 			       (ss < MAX_SS ? ss : 0) << SITMDR1_SYNCCH_SHIFT);
373 	}
374 	if (p->ctlr->flags & SPI_CONTROLLER_MUST_TX) {
375 		/* These bits are reserved if RX needs TX */
376 		tmp &= ~0x0000ffff;
377 	}
378 	sh_msiof_write(p, SIRMDR1, tmp);
379 
380 	tmp = 0;
381 	tmp |= SICTR_TSCKIZ_SCK | cpol << SICTR_TSCKIZ_POL_SHIFT;
382 	tmp |= SICTR_RSCKIZ_SCK | cpol << SICTR_RSCKIZ_POL_SHIFT;
383 
384 	edge = cpol ^ !cpha;
385 
386 	tmp |= edge << SICTR_TEDG_SHIFT;
387 	tmp |= edge << SICTR_REDG_SHIFT;
388 	tmp |= tx_hi_z ? SICTR_TXDIZ_HIZ : SICTR_TXDIZ_LOW;
389 	sh_msiof_write(p, SICTR, tmp);
390 }
391 
392 static void sh_msiof_spi_set_mode_regs(struct sh_msiof_spi_priv *p,
393 				       const void *tx_buf, void *rx_buf,
394 				       u32 bits, u32 words)
395 {
396 	u32 dr2 = SIMDR2_BITLEN1(bits) | SIMDR2_WDLEN1(words);
397 
398 	if (tx_buf || (p->ctlr->flags & SPI_CONTROLLER_MUST_TX))
399 		sh_msiof_write(p, SITMDR2, dr2);
400 	else
401 		sh_msiof_write(p, SITMDR2, dr2 | SIMDR2_GRPMASK1);
402 
403 	if (rx_buf)
404 		sh_msiof_write(p, SIRMDR2, dr2);
405 }
406 
407 static void sh_msiof_reset_str(struct sh_msiof_spi_priv *p)
408 {
409 	sh_msiof_write(p, SISTR,
410 		       sh_msiof_read(p, SISTR) & ~(SISTR_TDREQ | SISTR_RDREQ));
411 }
412 
413 static void sh_msiof_spi_write_fifo_8(struct sh_msiof_spi_priv *p,
414 				      const void *tx_buf, int words, int fs)
415 {
416 	const u8 *buf_8 = tx_buf;
417 	int k;
418 
419 	for (k = 0; k < words; k++)
420 		sh_msiof_write(p, SITFDR, buf_8[k] << fs);
421 }
422 
423 static void sh_msiof_spi_write_fifo_16(struct sh_msiof_spi_priv *p,
424 				       const void *tx_buf, int words, int fs)
425 {
426 	const u16 *buf_16 = tx_buf;
427 	int k;
428 
429 	for (k = 0; k < words; k++)
430 		sh_msiof_write(p, SITFDR, buf_16[k] << fs);
431 }
432 
433 static void sh_msiof_spi_write_fifo_16u(struct sh_msiof_spi_priv *p,
434 					const void *tx_buf, int words, int fs)
435 {
436 	const u16 *buf_16 = tx_buf;
437 	int k;
438 
439 	for (k = 0; k < words; k++)
440 		sh_msiof_write(p, SITFDR, get_unaligned(&buf_16[k]) << fs);
441 }
442 
443 static void sh_msiof_spi_write_fifo_32(struct sh_msiof_spi_priv *p,
444 				       const void *tx_buf, int words, int fs)
445 {
446 	const u32 *buf_32 = tx_buf;
447 	int k;
448 
449 	for (k = 0; k < words; k++)
450 		sh_msiof_write(p, SITFDR, buf_32[k] << fs);
451 }
452 
453 static void sh_msiof_spi_write_fifo_32u(struct sh_msiof_spi_priv *p,
454 					const void *tx_buf, int words, int fs)
455 {
456 	const u32 *buf_32 = tx_buf;
457 	int k;
458 
459 	for (k = 0; k < words; k++)
460 		sh_msiof_write(p, SITFDR, get_unaligned(&buf_32[k]) << fs);
461 }
462 
463 static void sh_msiof_spi_write_fifo_s32(struct sh_msiof_spi_priv *p,
464 					const void *tx_buf, int words, int fs)
465 {
466 	const u32 *buf_32 = tx_buf;
467 	int k;
468 
469 	for (k = 0; k < words; k++)
470 		sh_msiof_write(p, SITFDR, swab32(buf_32[k] << fs));
471 }
472 
473 static void sh_msiof_spi_write_fifo_s32u(struct sh_msiof_spi_priv *p,
474 					 const void *tx_buf, int words, int fs)
475 {
476 	const u32 *buf_32 = tx_buf;
477 	int k;
478 
479 	for (k = 0; k < words; k++)
480 		sh_msiof_write(p, SITFDR, swab32(get_unaligned(&buf_32[k]) << fs));
481 }
482 
483 static void sh_msiof_spi_read_fifo_8(struct sh_msiof_spi_priv *p,
484 				     void *rx_buf, int words, int fs)
485 {
486 	u8 *buf_8 = rx_buf;
487 	int k;
488 
489 	for (k = 0; k < words; k++)
490 		buf_8[k] = sh_msiof_read(p, SIRFDR) >> fs;
491 }
492 
493 static void sh_msiof_spi_read_fifo_16(struct sh_msiof_spi_priv *p,
494 				      void *rx_buf, int words, int fs)
495 {
496 	u16 *buf_16 = rx_buf;
497 	int k;
498 
499 	for (k = 0; k < words; k++)
500 		buf_16[k] = sh_msiof_read(p, SIRFDR) >> fs;
501 }
502 
503 static void sh_msiof_spi_read_fifo_16u(struct sh_msiof_spi_priv *p,
504 				       void *rx_buf, int words, int fs)
505 {
506 	u16 *buf_16 = rx_buf;
507 	int k;
508 
509 	for (k = 0; k < words; k++)
510 		put_unaligned(sh_msiof_read(p, SIRFDR) >> fs, &buf_16[k]);
511 }
512 
513 static void sh_msiof_spi_read_fifo_32(struct sh_msiof_spi_priv *p,
514 				      void *rx_buf, int words, int fs)
515 {
516 	u32 *buf_32 = rx_buf;
517 	int k;
518 
519 	for (k = 0; k < words; k++)
520 		buf_32[k] = sh_msiof_read(p, SIRFDR) >> fs;
521 }
522 
523 static void sh_msiof_spi_read_fifo_32u(struct sh_msiof_spi_priv *p,
524 				       void *rx_buf, int words, int fs)
525 {
526 	u32 *buf_32 = rx_buf;
527 	int k;
528 
529 	for (k = 0; k < words; k++)
530 		put_unaligned(sh_msiof_read(p, SIRFDR) >> fs, &buf_32[k]);
531 }
532 
533 static void sh_msiof_spi_read_fifo_s32(struct sh_msiof_spi_priv *p,
534 				       void *rx_buf, int words, int fs)
535 {
536 	u32 *buf_32 = rx_buf;
537 	int k;
538 
539 	for (k = 0; k < words; k++)
540 		buf_32[k] = swab32(sh_msiof_read(p, SIRFDR) >> fs);
541 }
542 
543 static void sh_msiof_spi_read_fifo_s32u(struct sh_msiof_spi_priv *p,
544 				       void *rx_buf, int words, int fs)
545 {
546 	u32 *buf_32 = rx_buf;
547 	int k;
548 
549 	for (k = 0; k < words; k++)
550 		put_unaligned(swab32(sh_msiof_read(p, SIRFDR) >> fs), &buf_32[k]);
551 }
552 
553 static int sh_msiof_spi_setup(struct spi_device *spi)
554 {
555 	struct sh_msiof_spi_priv *p =
556 		spi_controller_get_devdata(spi->controller);
557 	u32 clr, set, tmp;
558 
559 	if (spi_get_csgpiod(spi, 0) || spi_controller_is_target(p->ctlr))
560 		return 0;
561 
562 	if (p->native_cs_inited &&
563 	    (p->native_cs_high == !!(spi->mode & SPI_CS_HIGH)))
564 		return 0;
565 
566 	/* Configure native chip select mode/polarity early */
567 	clr = SIMDR1_SYNCMD_MASK;
568 	set = SIMDR1_SYNCMD_SPI;
569 	if (spi->mode & SPI_CS_HIGH)
570 		clr |= BIT(SIMDR1_SYNCAC_SHIFT);
571 	else
572 		set |= BIT(SIMDR1_SYNCAC_SHIFT);
573 	pm_runtime_get_sync(&p->pdev->dev);
574 	tmp = sh_msiof_read(p, SITMDR1) & ~clr;
575 	sh_msiof_write(p, SITMDR1, tmp | set | SIMDR1_TRMD | SITMDR1_PCON);
576 	tmp = sh_msiof_read(p, SIRMDR1) & ~clr;
577 	sh_msiof_write(p, SIRMDR1, tmp | set);
578 	pm_runtime_put(&p->pdev->dev);
579 	p->native_cs_high = spi->mode & SPI_CS_HIGH;
580 	p->native_cs_inited = true;
581 	return 0;
582 }
583 
584 static int sh_msiof_prepare_message(struct spi_controller *ctlr,
585 				    struct spi_message *msg)
586 {
587 	struct sh_msiof_spi_priv *p = spi_controller_get_devdata(ctlr);
588 	const struct spi_device *spi = msg->spi;
589 	u32 ss, cs_high;
590 
591 	/* Configure pins before asserting CS */
592 	if (spi_get_csgpiod(spi, 0)) {
593 		ss = ctlr->unused_native_cs;
594 		cs_high = p->native_cs_high;
595 	} else {
596 		ss = spi_get_chipselect(spi, 0);
597 		cs_high = !!(spi->mode & SPI_CS_HIGH);
598 	}
599 	sh_msiof_spi_set_pin_regs(p, ss, !!(spi->mode & SPI_CPOL),
600 				  !!(spi->mode & SPI_CPHA),
601 				  !!(spi->mode & SPI_3WIRE),
602 				  !!(spi->mode & SPI_LSB_FIRST), cs_high);
603 	return 0;
604 }
605 
606 static int sh_msiof_spi_start(struct sh_msiof_spi_priv *p, void *rx_buf)
607 {
608 	bool target = spi_controller_is_target(p->ctlr);
609 	int ret = 0;
610 
611 	/* setup clock and rx/tx signals */
612 	if (!target)
613 		ret = sh_msiof_modify_ctr_wait(p, 0, SICTR_TSCKE);
614 	if (rx_buf && !ret)
615 		ret = sh_msiof_modify_ctr_wait(p, 0, SICTR_RXE);
616 	if (!ret)
617 		ret = sh_msiof_modify_ctr_wait(p, 0, SICTR_TXE);
618 
619 	/* start by setting frame bit */
620 	if (!ret && !target)
621 		ret = sh_msiof_modify_ctr_wait(p, 0, SICTR_TFSE);
622 
623 	return ret;
624 }
625 
626 static int sh_msiof_spi_stop(struct sh_msiof_spi_priv *p, void *rx_buf)
627 {
628 	bool target = spi_controller_is_target(p->ctlr);
629 	int ret = 0;
630 
631 	/* shut down frame, rx/tx and clock signals */
632 	if (!target)
633 		ret = sh_msiof_modify_ctr_wait(p, SICTR_TFSE, 0);
634 	if (!ret)
635 		ret = sh_msiof_modify_ctr_wait(p, SICTR_TXE, 0);
636 	if (rx_buf && !ret)
637 		ret = sh_msiof_modify_ctr_wait(p, SICTR_RXE, 0);
638 	if (!ret && !target)
639 		ret = sh_msiof_modify_ctr_wait(p, SICTR_TSCKE, 0);
640 
641 	return ret;
642 }
643 
644 static int sh_msiof_target_abort(struct spi_controller *ctlr)
645 {
646 	struct sh_msiof_spi_priv *p = spi_controller_get_devdata(ctlr);
647 
648 	p->target_aborted = true;
649 	complete(&p->done);
650 	complete(&p->done_txdma);
651 	return 0;
652 }
653 
654 static int sh_msiof_wait_for_completion(struct sh_msiof_spi_priv *p,
655 					struct completion *x)
656 {
657 	if (spi_controller_is_target(p->ctlr)) {
658 		if (wait_for_completion_interruptible(x) ||
659 		    p->target_aborted) {
660 			dev_dbg(&p->pdev->dev, "interrupted\n");
661 			return -EINTR;
662 		}
663 	} else {
664 		if (!wait_for_completion_timeout(x, HZ)) {
665 			dev_err(&p->pdev->dev, "timeout\n");
666 			return -ETIMEDOUT;
667 		}
668 	}
669 
670 	return 0;
671 }
672 
673 static int sh_msiof_spi_txrx_once(struct sh_msiof_spi_priv *p,
674 				  void (*tx_fifo)(struct sh_msiof_spi_priv *,
675 						  const void *, int, int),
676 				  void (*rx_fifo)(struct sh_msiof_spi_priv *,
677 						  void *, int, int),
678 				  const void *tx_buf, void *rx_buf,
679 				  int words, int bits)
680 {
681 	int fifo_shift;
682 	int ret;
683 
684 	/* limit maximum word transfer to rx/tx fifo size */
685 	if (tx_buf)
686 		words = min_t(int, words, p->tx_fifo_size);
687 	if (rx_buf)
688 		words = min_t(int, words, p->rx_fifo_size);
689 
690 	/* the fifo contents need shifting */
691 	fifo_shift = 32 - bits;
692 
693 	/* default FIFO watermarks for PIO */
694 	sh_msiof_write(p, SIFCTR, 0);
695 
696 	/* setup msiof transfer mode registers */
697 	sh_msiof_spi_set_mode_regs(p, tx_buf, rx_buf, bits, words);
698 	sh_msiof_write(p, SIIER, SIIER_TEOFE | SIIER_REOFE);
699 
700 	/* write tx fifo */
701 	if (tx_buf)
702 		tx_fifo(p, tx_buf, words, fifo_shift);
703 
704 	reinit_completion(&p->done);
705 	p->target_aborted = false;
706 
707 	ret = sh_msiof_spi_start(p, rx_buf);
708 	if (ret) {
709 		dev_err(&p->pdev->dev, "failed to start hardware\n");
710 		goto stop_ier;
711 	}
712 
713 	/* wait for tx fifo to be emptied / rx fifo to be filled */
714 	ret = sh_msiof_wait_for_completion(p, &p->done);
715 	if (ret)
716 		goto stop_reset;
717 
718 	/* read rx fifo */
719 	if (rx_buf)
720 		rx_fifo(p, rx_buf, words, fifo_shift);
721 
722 	/* clear status bits */
723 	sh_msiof_reset_str(p);
724 
725 	ret = sh_msiof_spi_stop(p, rx_buf);
726 	if (ret) {
727 		dev_err(&p->pdev->dev, "failed to shut down hardware\n");
728 		return ret;
729 	}
730 
731 	return words;
732 
733 stop_reset:
734 	sh_msiof_reset_str(p);
735 	sh_msiof_spi_stop(p, rx_buf);
736 stop_ier:
737 	sh_msiof_write(p, SIIER, 0);
738 	return ret;
739 }
740 
741 static void sh_msiof_dma_complete(void *arg)
742 {
743 	complete(arg);
744 }
745 
746 static int sh_msiof_dma_once(struct sh_msiof_spi_priv *p, const void *tx,
747 			     void *rx, unsigned int len)
748 {
749 	u32 ier_bits = 0;
750 	struct dma_async_tx_descriptor *desc_tx = NULL, *desc_rx = NULL;
751 	dma_cookie_t cookie;
752 	int ret;
753 
754 	/* First prepare and submit the DMA request(s), as this may fail */
755 	if (rx) {
756 		ier_bits |= SIIER_RDREQE | SIIER_RDMAE;
757 		desc_rx = dmaengine_prep_slave_single(p->ctlr->dma_rx,
758 					p->rx_dma_addr, len, DMA_DEV_TO_MEM,
759 					DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
760 		if (!desc_rx)
761 			return -EAGAIN;
762 
763 		desc_rx->callback = sh_msiof_dma_complete;
764 		desc_rx->callback_param = &p->done;
765 		cookie = dmaengine_submit(desc_rx);
766 		if (dma_submit_error(cookie))
767 			return cookie;
768 	}
769 
770 	if (tx) {
771 		ier_bits |= SIIER_TDREQE | SIIER_TDMAE;
772 		dma_sync_single_for_device(p->ctlr->dma_tx->device->dev,
773 					   p->tx_dma_addr, len, DMA_TO_DEVICE);
774 		desc_tx = dmaengine_prep_slave_single(p->ctlr->dma_tx,
775 					p->tx_dma_addr, len, DMA_MEM_TO_DEV,
776 					DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
777 		if (!desc_tx) {
778 			ret = -EAGAIN;
779 			goto no_dma_tx;
780 		}
781 
782 		desc_tx->callback = sh_msiof_dma_complete;
783 		desc_tx->callback_param = &p->done_txdma;
784 		cookie = dmaengine_submit(desc_tx);
785 		if (dma_submit_error(cookie)) {
786 			ret = cookie;
787 			goto no_dma_tx;
788 		}
789 	}
790 
791 	/* 1 stage FIFO watermarks for DMA */
792 	sh_msiof_write(p, SIFCTR, SIFCTR_TFWM_1 | SIFCTR_RFWM_1);
793 
794 	/* setup msiof transfer mode registers (32-bit words) */
795 	sh_msiof_spi_set_mode_regs(p, tx, rx, 32, len / 4);
796 
797 	sh_msiof_write(p, SIIER, ier_bits);
798 
799 	reinit_completion(&p->done);
800 	if (tx)
801 		reinit_completion(&p->done_txdma);
802 	p->target_aborted = false;
803 
804 	/* Now start DMA */
805 	if (rx)
806 		dma_async_issue_pending(p->ctlr->dma_rx);
807 	if (tx)
808 		dma_async_issue_pending(p->ctlr->dma_tx);
809 
810 	ret = sh_msiof_spi_start(p, rx);
811 	if (ret) {
812 		dev_err(&p->pdev->dev, "failed to start hardware\n");
813 		goto stop_dma;
814 	}
815 
816 	if (tx) {
817 		/* wait for tx DMA completion */
818 		ret = sh_msiof_wait_for_completion(p, &p->done_txdma);
819 		if (ret)
820 			goto stop_reset;
821 	}
822 
823 	if (rx) {
824 		/* wait for rx DMA completion */
825 		ret = sh_msiof_wait_for_completion(p, &p->done);
826 		if (ret)
827 			goto stop_reset;
828 
829 		sh_msiof_write(p, SIIER, 0);
830 	} else {
831 		/* wait for tx fifo to be emptied */
832 		sh_msiof_write(p, SIIER, SIIER_TEOFE);
833 		ret = sh_msiof_wait_for_completion(p, &p->done);
834 		if (ret)
835 			goto stop_reset;
836 	}
837 
838 	/* clear status bits */
839 	sh_msiof_reset_str(p);
840 
841 	ret = sh_msiof_spi_stop(p, rx);
842 	if (ret) {
843 		dev_err(&p->pdev->dev, "failed to shut down hardware\n");
844 		return ret;
845 	}
846 
847 	if (rx)
848 		dma_sync_single_for_cpu(p->ctlr->dma_rx->device->dev,
849 					p->rx_dma_addr, len, DMA_FROM_DEVICE);
850 
851 	return 0;
852 
853 stop_reset:
854 	sh_msiof_reset_str(p);
855 	sh_msiof_spi_stop(p, rx);
856 stop_dma:
857 	if (tx)
858 		dmaengine_terminate_sync(p->ctlr->dma_tx);
859 no_dma_tx:
860 	if (rx)
861 		dmaengine_terminate_sync(p->ctlr->dma_rx);
862 	sh_msiof_write(p, SIIER, 0);
863 	return ret;
864 }
865 
866 static void copy_bswap32(u32 *dst, const u32 *src, unsigned int words)
867 {
868 	/* src or dst can be unaligned, but not both */
869 	if ((unsigned long)src & 3) {
870 		while (words--) {
871 			*dst++ = swab32(get_unaligned(src));
872 			src++;
873 		}
874 	} else if ((unsigned long)dst & 3) {
875 		while (words--) {
876 			put_unaligned(swab32(*src++), dst);
877 			dst++;
878 		}
879 	} else {
880 		while (words--)
881 			*dst++ = swab32(*src++);
882 	}
883 }
884 
885 static void copy_wswap32(u32 *dst, const u32 *src, unsigned int words)
886 {
887 	/* src or dst can be unaligned, but not both */
888 	if ((unsigned long)src & 3) {
889 		while (words--) {
890 			*dst++ = swahw32(get_unaligned(src));
891 			src++;
892 		}
893 	} else if ((unsigned long)dst & 3) {
894 		while (words--) {
895 			put_unaligned(swahw32(*src++), dst);
896 			dst++;
897 		}
898 	} else {
899 		while (words--)
900 			*dst++ = swahw32(*src++);
901 	}
902 }
903 
904 static void copy_plain32(u32 *dst, const u32 *src, unsigned int words)
905 {
906 	memcpy(dst, src, words * 4);
907 }
908 
909 static int sh_msiof_transfer_one(struct spi_controller *ctlr,
910 				 struct spi_device *spi,
911 				 struct spi_transfer *t)
912 {
913 	struct sh_msiof_spi_priv *p = spi_controller_get_devdata(ctlr);
914 	void (*copy32)(u32 *, const u32 *, unsigned int);
915 	void (*tx_fifo)(struct sh_msiof_spi_priv *, const void *, int, int);
916 	void (*rx_fifo)(struct sh_msiof_spi_priv *, void *, int, int);
917 	const void *tx_buf = t->tx_buf;
918 	void *rx_buf = t->rx_buf;
919 	unsigned int len = t->len;
920 	unsigned int bits = t->bits_per_word;
921 	unsigned int bytes_per_word;
922 	unsigned int words;
923 	int n;
924 	bool swab;
925 	int ret;
926 
927 	/* reset registers */
928 	sh_msiof_spi_reset_regs(p);
929 
930 	/* setup clocks (clock already enabled in chipselect()) */
931 	if (!spi_controller_is_target(p->ctlr))
932 		sh_msiof_spi_set_clk_regs(p, t);
933 
934 	while (ctlr->dma_tx && len > 15) {
935 		/*
936 		 *  DMA supports 32-bit words only, hence pack 8-bit and 16-bit
937 		 *  words, with byte resp. word swapping.
938 		 */
939 		unsigned int l = 0;
940 
941 		if (tx_buf)
942 			l = min(round_down(len, 4), p->tx_fifo_size * 4);
943 		if (rx_buf)
944 			l = min(round_down(len, 4), p->rx_fifo_size * 4);
945 
946 		if (bits <= 8) {
947 			copy32 = copy_bswap32;
948 		} else if (bits <= 16) {
949 			copy32 = copy_wswap32;
950 		} else {
951 			copy32 = copy_plain32;
952 		}
953 
954 		if (tx_buf)
955 			copy32(p->tx_dma_page, tx_buf, l / 4);
956 
957 		ret = sh_msiof_dma_once(p, tx_buf, rx_buf, l);
958 		if (ret == -EAGAIN) {
959 			dev_warn_once(&p->pdev->dev,
960 				"DMA not available, falling back to PIO\n");
961 			break;
962 		}
963 		if (ret)
964 			return ret;
965 
966 		if (rx_buf) {
967 			copy32(rx_buf, p->rx_dma_page, l / 4);
968 			rx_buf += l;
969 		}
970 		if (tx_buf)
971 			tx_buf += l;
972 
973 		len -= l;
974 		if (!len)
975 			return 0;
976 	}
977 
978 	if (bits <= 8 && len > 15) {
979 		bits = 32;
980 		swab = true;
981 	} else {
982 		swab = false;
983 	}
984 
985 	/* setup bytes per word and fifo read/write functions */
986 	if (bits <= 8) {
987 		bytes_per_word = 1;
988 		tx_fifo = sh_msiof_spi_write_fifo_8;
989 		rx_fifo = sh_msiof_spi_read_fifo_8;
990 	} else if (bits <= 16) {
991 		bytes_per_word = 2;
992 		if ((unsigned long)tx_buf & 0x01)
993 			tx_fifo = sh_msiof_spi_write_fifo_16u;
994 		else
995 			tx_fifo = sh_msiof_spi_write_fifo_16;
996 
997 		if ((unsigned long)rx_buf & 0x01)
998 			rx_fifo = sh_msiof_spi_read_fifo_16u;
999 		else
1000 			rx_fifo = sh_msiof_spi_read_fifo_16;
1001 	} else if (swab) {
1002 		bytes_per_word = 4;
1003 		if ((unsigned long)tx_buf & 0x03)
1004 			tx_fifo = sh_msiof_spi_write_fifo_s32u;
1005 		else
1006 			tx_fifo = sh_msiof_spi_write_fifo_s32;
1007 
1008 		if ((unsigned long)rx_buf & 0x03)
1009 			rx_fifo = sh_msiof_spi_read_fifo_s32u;
1010 		else
1011 			rx_fifo = sh_msiof_spi_read_fifo_s32;
1012 	} else {
1013 		bytes_per_word = 4;
1014 		if ((unsigned long)tx_buf & 0x03)
1015 			tx_fifo = sh_msiof_spi_write_fifo_32u;
1016 		else
1017 			tx_fifo = sh_msiof_spi_write_fifo_32;
1018 
1019 		if ((unsigned long)rx_buf & 0x03)
1020 			rx_fifo = sh_msiof_spi_read_fifo_32u;
1021 		else
1022 			rx_fifo = sh_msiof_spi_read_fifo_32;
1023 	}
1024 
1025 	/* transfer in fifo sized chunks */
1026 	words = len / bytes_per_word;
1027 
1028 	while (words > 0) {
1029 		n = sh_msiof_spi_txrx_once(p, tx_fifo, rx_fifo, tx_buf, rx_buf,
1030 					   words, bits);
1031 		if (n < 0)
1032 			return n;
1033 
1034 		if (tx_buf)
1035 			tx_buf += n * bytes_per_word;
1036 		if (rx_buf)
1037 			rx_buf += n * bytes_per_word;
1038 		words -= n;
1039 
1040 		if (words == 0 && (len % bytes_per_word)) {
1041 			words = len % bytes_per_word;
1042 			bits = t->bits_per_word;
1043 			bytes_per_word = 1;
1044 			tx_fifo = sh_msiof_spi_write_fifo_8;
1045 			rx_fifo = sh_msiof_spi_read_fifo_8;
1046 		}
1047 	}
1048 
1049 	return 0;
1050 }
1051 
1052 static const struct sh_msiof_chipdata sh_data = {
1053 	.bits_per_word_mask = SPI_BPW_RANGE_MASK(8, 32),
1054 	.tx_fifo_size = 64,
1055 	.rx_fifo_size = 64,
1056 	.ctlr_flags = 0,
1057 	.min_div_pow = 0,
1058 };
1059 
1060 static const struct sh_msiof_chipdata rcar_gen2_data = {
1061 	.bits_per_word_mask = SPI_BPW_MASK(8) | SPI_BPW_MASK(16) |
1062 			      SPI_BPW_MASK(24) | SPI_BPW_MASK(32),
1063 	.tx_fifo_size = 64,
1064 	.rx_fifo_size = 64,
1065 	.ctlr_flags = SPI_CONTROLLER_MUST_TX,
1066 	.min_div_pow = 0,
1067 };
1068 
1069 static const struct sh_msiof_chipdata rcar_gen3_data = {
1070 	.bits_per_word_mask = SPI_BPW_MASK(8) | SPI_BPW_MASK(16) |
1071 			      SPI_BPW_MASK(24) | SPI_BPW_MASK(32),
1072 	.tx_fifo_size = 64,
1073 	.rx_fifo_size = 64,
1074 	.ctlr_flags = SPI_CONTROLLER_MUST_TX,
1075 	.min_div_pow = 1,
1076 };
1077 
1078 static const struct sh_msiof_chipdata rcar_r8a7795_data = {
1079 	.bits_per_word_mask = SPI_BPW_MASK(8) | SPI_BPW_MASK(16) |
1080 			      SPI_BPW_MASK(24) | SPI_BPW_MASK(32),
1081 	.tx_fifo_size = 64,
1082 	.rx_fifo_size = 64,
1083 	.ctlr_flags = SPI_CONTROLLER_MUST_TX,
1084 	.min_div_pow = 1,
1085 	.flags = SH_MSIOF_FLAG_FIXED_DTDL_200,
1086 };
1087 
1088 static const struct of_device_id sh_msiof_match[] __maybe_unused = {
1089 	{ .compatible = "renesas,sh-mobile-msiof", .data = &sh_data },
1090 	{ .compatible = "renesas,msiof-r8a7743",   .data = &rcar_gen2_data },
1091 	{ .compatible = "renesas,msiof-r8a7745",   .data = &rcar_gen2_data },
1092 	{ .compatible = "renesas,msiof-r8a7790",   .data = &rcar_gen2_data },
1093 	{ .compatible = "renesas,msiof-r8a7791",   .data = &rcar_gen2_data },
1094 	{ .compatible = "renesas,msiof-r8a7792",   .data = &rcar_gen2_data },
1095 	{ .compatible = "renesas,msiof-r8a7793",   .data = &rcar_gen2_data },
1096 	{ .compatible = "renesas,msiof-r8a7794",   .data = &rcar_gen2_data },
1097 	{ .compatible = "renesas,rcar-gen2-msiof", .data = &rcar_gen2_data },
1098 	{ .compatible = "renesas,msiof-r8a7795",   .data = &rcar_r8a7795_data },
1099 	{ .compatible = "renesas,msiof-r8a7796",   .data = &rcar_gen3_data },
1100 	{ .compatible = "renesas,rcar-gen3-msiof", .data = &rcar_gen3_data },
1101 	{ .compatible = "renesas,rcar-gen4-msiof", .data = &rcar_gen3_data },
1102 	{ .compatible = "renesas,sh-msiof",        .data = &sh_data }, /* Deprecated */
1103 	{},
1104 };
1105 MODULE_DEVICE_TABLE(of, sh_msiof_match);
1106 
1107 #ifdef CONFIG_OF
1108 static struct sh_msiof_spi_info *sh_msiof_spi_parse_dt(struct device *dev)
1109 {
1110 	struct sh_msiof_spi_info *info;
1111 	struct device_node *np = dev->of_node;
1112 	u32 num_cs = 1;
1113 
1114 	info = devm_kzalloc(dev, sizeof(struct sh_msiof_spi_info), GFP_KERNEL);
1115 	if (!info)
1116 		return NULL;
1117 
1118 	info->mode = of_property_read_bool(np, "spi-slave") ? MSIOF_SPI_TARGET
1119 							    : MSIOF_SPI_HOST;
1120 
1121 	/* Parse the MSIOF properties */
1122 	if (info->mode == MSIOF_SPI_HOST)
1123 		of_property_read_u32(np, "num-cs", &num_cs);
1124 	of_property_read_u32(np, "renesas,tx-fifo-size",
1125 					&info->tx_fifo_override);
1126 	of_property_read_u32(np, "renesas,rx-fifo-size",
1127 					&info->rx_fifo_override);
1128 	of_property_read_u32(np, "renesas,dtdl", &info->dtdl);
1129 	of_property_read_u32(np, "renesas,syncdl", &info->syncdl);
1130 
1131 	info->num_chipselect = num_cs;
1132 
1133 	return info;
1134 }
1135 #else
1136 static struct sh_msiof_spi_info *sh_msiof_spi_parse_dt(struct device *dev)
1137 {
1138 	return NULL;
1139 }
1140 #endif
1141 
1142 static struct dma_chan *sh_msiof_request_dma_chan(struct device *dev,
1143 	enum dma_transfer_direction dir, unsigned int id, dma_addr_t port_addr)
1144 {
1145 	dma_cap_mask_t mask;
1146 	struct dma_chan *chan;
1147 	struct dma_slave_config cfg;
1148 	int ret;
1149 
1150 	dma_cap_zero(mask);
1151 	dma_cap_set(DMA_SLAVE, mask);
1152 
1153 	chan = dma_request_slave_channel_compat(mask, shdma_chan_filter,
1154 				(void *)(unsigned long)id, dev,
1155 				dir == DMA_MEM_TO_DEV ? "tx" : "rx");
1156 	if (!chan) {
1157 		dev_warn(dev, "dma_request_slave_channel_compat failed\n");
1158 		return NULL;
1159 	}
1160 
1161 	memset(&cfg, 0, sizeof(cfg));
1162 	cfg.direction = dir;
1163 	if (dir == DMA_MEM_TO_DEV) {
1164 		cfg.dst_addr = port_addr;
1165 		cfg.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
1166 	} else {
1167 		cfg.src_addr = port_addr;
1168 		cfg.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
1169 	}
1170 
1171 	ret = dmaengine_slave_config(chan, &cfg);
1172 	if (ret) {
1173 		dev_warn(dev, "dmaengine_slave_config failed %d\n", ret);
1174 		dma_release_channel(chan);
1175 		return NULL;
1176 	}
1177 
1178 	return chan;
1179 }
1180 
1181 static int sh_msiof_request_dma(struct sh_msiof_spi_priv *p)
1182 {
1183 	struct platform_device *pdev = p->pdev;
1184 	struct device *dev = &pdev->dev;
1185 	const struct sh_msiof_spi_info *info = p->info;
1186 	unsigned int dma_tx_id, dma_rx_id;
1187 	const struct resource *res;
1188 	struct spi_controller *ctlr;
1189 	struct device *tx_dev, *rx_dev;
1190 
1191 	if (dev->of_node) {
1192 		/* In the OF case we will get the slave IDs from the DT */
1193 		dma_tx_id = 0;
1194 		dma_rx_id = 0;
1195 	} else if (info && info->dma_tx_id && info->dma_rx_id) {
1196 		dma_tx_id = info->dma_tx_id;
1197 		dma_rx_id = info->dma_rx_id;
1198 	} else {
1199 		/* The driver assumes no error */
1200 		return 0;
1201 	}
1202 
1203 	/* The DMA engine uses the second register set, if present */
1204 	res = platform_get_resource(pdev, IORESOURCE_MEM, 1);
1205 	if (!res)
1206 		res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
1207 
1208 	ctlr = p->ctlr;
1209 	ctlr->dma_tx = sh_msiof_request_dma_chan(dev, DMA_MEM_TO_DEV,
1210 						 dma_tx_id, res->start + SITFDR);
1211 	if (!ctlr->dma_tx)
1212 		return -ENODEV;
1213 
1214 	ctlr->dma_rx = sh_msiof_request_dma_chan(dev, DMA_DEV_TO_MEM,
1215 						 dma_rx_id, res->start + SIRFDR);
1216 	if (!ctlr->dma_rx)
1217 		goto free_tx_chan;
1218 
1219 	p->tx_dma_page = (void *)__get_free_page(GFP_KERNEL | GFP_DMA);
1220 	if (!p->tx_dma_page)
1221 		goto free_rx_chan;
1222 
1223 	p->rx_dma_page = (void *)__get_free_page(GFP_KERNEL | GFP_DMA);
1224 	if (!p->rx_dma_page)
1225 		goto free_tx_page;
1226 
1227 	tx_dev = ctlr->dma_tx->device->dev;
1228 	p->tx_dma_addr = dma_map_single(tx_dev, p->tx_dma_page, PAGE_SIZE,
1229 					DMA_TO_DEVICE);
1230 	if (dma_mapping_error(tx_dev, p->tx_dma_addr))
1231 		goto free_rx_page;
1232 
1233 	rx_dev = ctlr->dma_rx->device->dev;
1234 	p->rx_dma_addr = dma_map_single(rx_dev, p->rx_dma_page, PAGE_SIZE,
1235 					DMA_FROM_DEVICE);
1236 	if (dma_mapping_error(rx_dev, p->rx_dma_addr))
1237 		goto unmap_tx_page;
1238 
1239 	dev_info(dev, "DMA available");
1240 	return 0;
1241 
1242 unmap_tx_page:
1243 	dma_unmap_single(tx_dev, p->tx_dma_addr, PAGE_SIZE, DMA_TO_DEVICE);
1244 free_rx_page:
1245 	free_page((unsigned long)p->rx_dma_page);
1246 free_tx_page:
1247 	free_page((unsigned long)p->tx_dma_page);
1248 free_rx_chan:
1249 	dma_release_channel(ctlr->dma_rx);
1250 free_tx_chan:
1251 	dma_release_channel(ctlr->dma_tx);
1252 	ctlr->dma_tx = NULL;
1253 	return -ENODEV;
1254 }
1255 
1256 static void sh_msiof_release_dma(struct sh_msiof_spi_priv *p)
1257 {
1258 	struct spi_controller *ctlr = p->ctlr;
1259 
1260 	if (!ctlr->dma_tx)
1261 		return;
1262 
1263 	dma_unmap_single(ctlr->dma_rx->device->dev, p->rx_dma_addr, PAGE_SIZE,
1264 			 DMA_FROM_DEVICE);
1265 	dma_unmap_single(ctlr->dma_tx->device->dev, p->tx_dma_addr, PAGE_SIZE,
1266 			 DMA_TO_DEVICE);
1267 	free_page((unsigned long)p->rx_dma_page);
1268 	free_page((unsigned long)p->tx_dma_page);
1269 	dma_release_channel(ctlr->dma_rx);
1270 	dma_release_channel(ctlr->dma_tx);
1271 }
1272 
1273 static int sh_msiof_spi_probe(struct platform_device *pdev)
1274 {
1275 	struct spi_controller *ctlr;
1276 	const struct sh_msiof_chipdata *chipdata;
1277 	struct sh_msiof_spi_info *info;
1278 	struct sh_msiof_spi_priv *p;
1279 	unsigned long clksrc;
1280 	int i;
1281 	int ret;
1282 
1283 	chipdata = of_device_get_match_data(&pdev->dev);
1284 	if (chipdata) {
1285 		info = sh_msiof_spi_parse_dt(&pdev->dev);
1286 	} else {
1287 		chipdata = (const void *)pdev->id_entry->driver_data;
1288 		info = dev_get_platdata(&pdev->dev);
1289 	}
1290 
1291 	if (!info) {
1292 		dev_err(&pdev->dev, "failed to obtain device info\n");
1293 		return -ENXIO;
1294 	}
1295 
1296 	if (chipdata->flags & SH_MSIOF_FLAG_FIXED_DTDL_200)
1297 		info->dtdl = 200;
1298 
1299 	if (info->mode == MSIOF_SPI_TARGET)
1300 		ctlr = spi_alloc_target(&pdev->dev,
1301 				        sizeof(struct sh_msiof_spi_priv));
1302 	else
1303 		ctlr = spi_alloc_host(&pdev->dev,
1304 				      sizeof(struct sh_msiof_spi_priv));
1305 	if (ctlr == NULL)
1306 		return -ENOMEM;
1307 
1308 	p = spi_controller_get_devdata(ctlr);
1309 
1310 	platform_set_drvdata(pdev, p);
1311 	p->ctlr = ctlr;
1312 	p->info = info;
1313 	p->min_div_pow = chipdata->min_div_pow;
1314 
1315 	init_completion(&p->done);
1316 	init_completion(&p->done_txdma);
1317 
1318 	p->clk = devm_clk_get(&pdev->dev, NULL);
1319 	if (IS_ERR(p->clk)) {
1320 		dev_err(&pdev->dev, "cannot get clock\n");
1321 		ret = PTR_ERR(p->clk);
1322 		goto err1;
1323 	}
1324 
1325 	i = platform_get_irq(pdev, 0);
1326 	if (i < 0) {
1327 		ret = i;
1328 		goto err1;
1329 	}
1330 
1331 	p->mapbase = devm_platform_ioremap_resource(pdev, 0);
1332 	if (IS_ERR(p->mapbase)) {
1333 		ret = PTR_ERR(p->mapbase);
1334 		goto err1;
1335 	}
1336 
1337 	ret = devm_request_irq(&pdev->dev, i, sh_msiof_spi_irq, 0,
1338 			       dev_name(&pdev->dev), p);
1339 	if (ret) {
1340 		dev_err(&pdev->dev, "unable to request irq\n");
1341 		goto err1;
1342 	}
1343 
1344 	p->pdev = pdev;
1345 	pm_runtime_enable(&pdev->dev);
1346 
1347 	/* Platform data may override FIFO sizes */
1348 	p->tx_fifo_size = chipdata->tx_fifo_size;
1349 	p->rx_fifo_size = chipdata->rx_fifo_size;
1350 	if (p->info->tx_fifo_override)
1351 		p->tx_fifo_size = p->info->tx_fifo_override;
1352 	if (p->info->rx_fifo_override)
1353 		p->rx_fifo_size = p->info->rx_fifo_override;
1354 
1355 	/* init controller code */
1356 	ctlr->mode_bits = SPI_CPOL | SPI_CPHA | SPI_CS_HIGH;
1357 	ctlr->mode_bits |= SPI_LSB_FIRST | SPI_3WIRE;
1358 	clksrc = clk_get_rate(p->clk);
1359 	ctlr->min_speed_hz = DIV_ROUND_UP(clksrc, 1024);
1360 	ctlr->max_speed_hz = DIV_ROUND_UP(clksrc, 1 << p->min_div_pow);
1361 	ctlr->flags = chipdata->ctlr_flags;
1362 	ctlr->bus_num = pdev->id;
1363 	ctlr->num_chipselect = p->info->num_chipselect;
1364 	ctlr->dev.of_node = pdev->dev.of_node;
1365 	ctlr->setup = sh_msiof_spi_setup;
1366 	ctlr->prepare_message = sh_msiof_prepare_message;
1367 	ctlr->target_abort = sh_msiof_target_abort;
1368 	ctlr->bits_per_word_mask = chipdata->bits_per_word_mask;
1369 	ctlr->auto_runtime_pm = true;
1370 	ctlr->transfer_one = sh_msiof_transfer_one;
1371 	ctlr->use_gpio_descriptors = true;
1372 	ctlr->max_native_cs = MAX_SS;
1373 
1374 	ret = sh_msiof_request_dma(p);
1375 	if (ret < 0)
1376 		dev_warn(&pdev->dev, "DMA not available, using PIO\n");
1377 
1378 	ret = devm_spi_register_controller(&pdev->dev, ctlr);
1379 	if (ret < 0) {
1380 		dev_err(&pdev->dev, "devm_spi_register_controller error.\n");
1381 		goto err2;
1382 	}
1383 
1384 	return 0;
1385 
1386  err2:
1387 	sh_msiof_release_dma(p);
1388 	pm_runtime_disable(&pdev->dev);
1389  err1:
1390 	spi_controller_put(ctlr);
1391 	return ret;
1392 }
1393 
1394 static void sh_msiof_spi_remove(struct platform_device *pdev)
1395 {
1396 	struct sh_msiof_spi_priv *p = platform_get_drvdata(pdev);
1397 
1398 	sh_msiof_release_dma(p);
1399 	pm_runtime_disable(&pdev->dev);
1400 }
1401 
1402 static const struct platform_device_id spi_driver_ids[] = {
1403 	{ "spi_sh_msiof",	(kernel_ulong_t)&sh_data },
1404 	{},
1405 };
1406 MODULE_DEVICE_TABLE(platform, spi_driver_ids);
1407 
1408 #ifdef CONFIG_PM_SLEEP
1409 static int sh_msiof_spi_suspend(struct device *dev)
1410 {
1411 	struct sh_msiof_spi_priv *p = dev_get_drvdata(dev);
1412 
1413 	return spi_controller_suspend(p->ctlr);
1414 }
1415 
1416 static int sh_msiof_spi_resume(struct device *dev)
1417 {
1418 	struct sh_msiof_spi_priv *p = dev_get_drvdata(dev);
1419 
1420 	return spi_controller_resume(p->ctlr);
1421 }
1422 
1423 static SIMPLE_DEV_PM_OPS(sh_msiof_spi_pm_ops, sh_msiof_spi_suspend,
1424 			 sh_msiof_spi_resume);
1425 #define DEV_PM_OPS	(&sh_msiof_spi_pm_ops)
1426 #else
1427 #define DEV_PM_OPS	NULL
1428 #endif /* CONFIG_PM_SLEEP */
1429 
1430 static struct platform_driver sh_msiof_spi_drv = {
1431 	.probe		= sh_msiof_spi_probe,
1432 	.remove_new	= sh_msiof_spi_remove,
1433 	.id_table	= spi_driver_ids,
1434 	.driver		= {
1435 		.name		= "spi_sh_msiof",
1436 		.pm		= DEV_PM_OPS,
1437 		.of_match_table = of_match_ptr(sh_msiof_match),
1438 	},
1439 };
1440 module_platform_driver(sh_msiof_spi_drv);
1441 
1442 MODULE_DESCRIPTION("SuperH MSIOF SPI Controller Interface Driver");
1443 MODULE_AUTHOR("Magnus Damm");
1444 MODULE_LICENSE("GPL v2");
1445