xref: /linux/drivers/spi/spi-sh-msiof.c (revision 8ccd54fe45713cd458015b5b08d6098545e70543)
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/of_device.h>
24 #include <linux/platform_device.h>
25 #include <linux/pm_runtime.h>
26 #include <linux/sh_dma.h>
27 
28 #include <linux/spi/sh_msiof.h>
29 #include <linux/spi/spi.h>
30 
31 #include <asm/unaligned.h>
32 
33 struct sh_msiof_chipdata {
34 	u32 bits_per_word_mask;
35 	u16 tx_fifo_size;
36 	u16 rx_fifo_size;
37 	u16 ctlr_flags;
38 	u16 min_div_pow;
39 };
40 
41 struct sh_msiof_spi_priv {
42 	struct spi_controller *ctlr;
43 	void __iomem *mapbase;
44 	struct clk *clk;
45 	struct platform_device *pdev;
46 	struct sh_msiof_spi_info *info;
47 	struct completion done;
48 	struct completion done_txdma;
49 	unsigned int tx_fifo_size;
50 	unsigned int rx_fifo_size;
51 	unsigned int min_div_pow;
52 	void *tx_dma_page;
53 	void *rx_dma_page;
54 	dma_addr_t tx_dma_addr;
55 	dma_addr_t rx_dma_addr;
56 	bool native_cs_inited;
57 	bool native_cs_high;
58 	bool slave_aborted;
59 };
60 
61 #define MAX_SS	3	/* Maximum number of native chip selects */
62 
63 #define SITMDR1	0x00	/* Transmit Mode Register 1 */
64 #define SITMDR2	0x04	/* Transmit Mode Register 2 */
65 #define SITMDR3	0x08	/* Transmit Mode Register 3 */
66 #define SIRMDR1	0x10	/* Receive Mode Register 1 */
67 #define SIRMDR2	0x14	/* Receive Mode Register 2 */
68 #define SIRMDR3	0x18	/* Receive Mode Register 3 */
69 #define SITSCR	0x20	/* Transmit Clock Select Register */
70 #define SIRSCR	0x22	/* Receive Clock Select Register (SH, A1, APE6) */
71 #define SICTR	0x28	/* Control Register */
72 #define SIFCTR	0x30	/* FIFO Control Register */
73 #define SISTR	0x40	/* Status Register */
74 #define SIIER	0x44	/* Interrupt Enable Register */
75 #define SITDR1	0x48	/* Transmit Control Data Register 1 (SH, A1) */
76 #define SITDR2	0x4c	/* Transmit Control Data Register 2 (SH, A1) */
77 #define SITFDR	0x50	/* Transmit FIFO Data Register */
78 #define SIRDR1	0x58	/* Receive Control Data Register 1 (SH, A1) */
79 #define SIRDR2	0x5c	/* Receive Control Data Register 2 (SH, A1) */
80 #define SIRFDR	0x60	/* Receive FIFO Data Register */
81 
82 /* SITMDR1 and SIRMDR1 */
83 #define SIMDR1_TRMD		BIT(31)		/* Transfer Mode (1 = Master mode) */
84 #define SIMDR1_SYNCMD_MASK	GENMASK(29, 28)	/* SYNC Mode */
85 #define SIMDR1_SYNCMD_SPI	(2 << 28)	/*   Level mode/SPI */
86 #define SIMDR1_SYNCMD_LR	(3 << 28)	/*   L/R mode */
87 #define SIMDR1_SYNCAC_SHIFT	25		/* Sync Polarity (1 = Active-low) */
88 #define SIMDR1_BITLSB_SHIFT	24		/* MSB/LSB First (1 = LSB first) */
89 #define SIMDR1_DTDL_SHIFT	20		/* Data Pin Bit Delay for MSIOF_SYNC */
90 #define SIMDR1_SYNCDL_SHIFT	16		/* Frame Sync Signal Timing Delay */
91 #define SIMDR1_FLD_MASK		GENMASK(3, 2)	/* Frame Sync Signal Interval (0-3) */
92 #define SIMDR1_FLD_SHIFT	2
93 #define SIMDR1_XXSTP		BIT(0)		/* Transmission/Reception Stop on FIFO */
94 /* SITMDR1 */
95 #define SITMDR1_PCON		BIT(30)		/* Transfer Signal Connection */
96 #define SITMDR1_SYNCCH_MASK	GENMASK(27, 26)	/* Sync Signal Channel Select */
97 #define SITMDR1_SYNCCH_SHIFT	26		/* 0=MSIOF_SYNC, 1=MSIOF_SS1, 2=MSIOF_SS2 */
98 
99 /* SITMDR2 and SIRMDR2 */
100 #define SIMDR2_BITLEN1(i)	(((i) - 1) << 24) /* Data Size (8-32 bits) */
101 #define SIMDR2_WDLEN1(i)	(((i) - 1) << 16) /* Word Count (1-64/256 (SH, A1))) */
102 #define SIMDR2_GRPMASK1		BIT(0)		/* Group Output Mask 1 (SH, A1) */
103 
104 /* SITSCR and SIRSCR */
105 #define SISCR_BRPS_MASK		GENMASK(12, 8)	/* Prescaler Setting (1-32) */
106 #define SISCR_BRPS(i)		(((i) - 1) << 8)
107 #define SISCR_BRDV_MASK		GENMASK(2, 0)	/* Baud Rate Generator's Division Ratio */
108 #define SISCR_BRDV_DIV_2	0
109 #define SISCR_BRDV_DIV_4	1
110 #define SISCR_BRDV_DIV_8	2
111 #define SISCR_BRDV_DIV_16	3
112 #define SISCR_BRDV_DIV_32	4
113 #define SISCR_BRDV_DIV_1	7
114 
115 /* SICTR */
116 #define SICTR_TSCKIZ_MASK	GENMASK(31, 30)	/* Transmit Clock I/O Polarity Select */
117 #define SICTR_TSCKIZ_SCK	BIT(31)		/*   Disable SCK when TX disabled */
118 #define SICTR_TSCKIZ_POL_SHIFT	30		/*   Transmit Clock Polarity */
119 #define SICTR_RSCKIZ_MASK	GENMASK(29, 28) /* Receive Clock Polarity Select */
120 #define SICTR_RSCKIZ_SCK	BIT(29)		/*   Must match CTR_TSCKIZ_SCK */
121 #define SICTR_RSCKIZ_POL_SHIFT	28		/*   Receive Clock Polarity */
122 #define SICTR_TEDG_SHIFT	27		/* Transmit Timing (1 = falling edge) */
123 #define SICTR_REDG_SHIFT	26		/* Receive Timing (1 = falling edge) */
124 #define SICTR_TXDIZ_MASK	GENMASK(23, 22)	/* Pin Output When TX is Disabled */
125 #define SICTR_TXDIZ_LOW		(0 << 22)	/*   0 */
126 #define SICTR_TXDIZ_HIGH	(1 << 22)	/*   1 */
127 #define SICTR_TXDIZ_HIZ		(2 << 22)	/*   High-impedance */
128 #define SICTR_TSCKE		BIT(15)		/* Transmit Serial Clock Output Enable */
129 #define SICTR_TFSE		BIT(14)		/* Transmit Frame Sync Signal Output Enable */
130 #define SICTR_TXE		BIT(9)		/* Transmit Enable */
131 #define SICTR_RXE		BIT(8)		/* Receive Enable */
132 #define SICTR_TXRST		BIT(1)		/* Transmit Reset */
133 #define SICTR_RXRST		BIT(0)		/* Receive Reset */
134 
135 /* SIFCTR */
136 #define SIFCTR_TFWM_MASK	GENMASK(31, 29)	/* Transmit FIFO Watermark */
137 #define SIFCTR_TFWM_64		(0 << 29)	/*  Transfer Request when 64 empty stages */
138 #define SIFCTR_TFWM_32		(1 << 29)	/*  Transfer Request when 32 empty stages */
139 #define SIFCTR_TFWM_24		(2 << 29)	/*  Transfer Request when 24 empty stages */
140 #define SIFCTR_TFWM_16		(3 << 29)	/*  Transfer Request when 16 empty stages */
141 #define SIFCTR_TFWM_12		(4 << 29)	/*  Transfer Request when 12 empty stages */
142 #define SIFCTR_TFWM_8		(5 << 29)	/*  Transfer Request when 8 empty stages */
143 #define SIFCTR_TFWM_4		(6 << 29)	/*  Transfer Request when 4 empty stages */
144 #define SIFCTR_TFWM_1		(7 << 29)	/*  Transfer Request when 1 empty stage */
145 #define SIFCTR_TFUA_MASK	GENMASK(26, 20) /* Transmit FIFO Usable Area */
146 #define SIFCTR_TFUA_SHIFT	20
147 #define SIFCTR_TFUA(i)		((i) << SIFCTR_TFUA_SHIFT)
148 #define SIFCTR_RFWM_MASK	GENMASK(15, 13)	/* Receive FIFO Watermark */
149 #define SIFCTR_RFWM_1		(0 << 13)	/*  Transfer Request when 1 valid stages */
150 #define SIFCTR_RFWM_4		(1 << 13)	/*  Transfer Request when 4 valid stages */
151 #define SIFCTR_RFWM_8		(2 << 13)	/*  Transfer Request when 8 valid stages */
152 #define SIFCTR_RFWM_16		(3 << 13)	/*  Transfer Request when 16 valid stages */
153 #define SIFCTR_RFWM_32		(4 << 13)	/*  Transfer Request when 32 valid stages */
154 #define SIFCTR_RFWM_64		(5 << 13)	/*  Transfer Request when 64 valid stages */
155 #define SIFCTR_RFWM_128		(6 << 13)	/*  Transfer Request when 128 valid stages */
156 #define SIFCTR_RFWM_256		(7 << 13)	/*  Transfer Request when 256 valid stages */
157 #define SIFCTR_RFUA_MASK	GENMASK(12, 4)	/* Receive FIFO Usable Area (0x40 = full) */
158 #define SIFCTR_RFUA_SHIFT	4
159 #define SIFCTR_RFUA(i)		((i) << SIFCTR_RFUA_SHIFT)
160 
161 /* SISTR */
162 #define SISTR_TFEMP		BIT(29) /* Transmit FIFO Empty */
163 #define SISTR_TDREQ		BIT(28) /* Transmit Data Transfer Request */
164 #define SISTR_TEOF		BIT(23) /* Frame Transmission End */
165 #define SISTR_TFSERR		BIT(21) /* Transmit Frame Synchronization Error */
166 #define SISTR_TFOVF		BIT(20) /* Transmit FIFO Overflow */
167 #define SISTR_TFUDF		BIT(19) /* Transmit FIFO Underflow */
168 #define SISTR_RFFUL		BIT(13) /* Receive FIFO Full */
169 #define SISTR_RDREQ		BIT(12) /* Receive Data Transfer Request */
170 #define SISTR_REOF		BIT(7)  /* Frame Reception End */
171 #define SISTR_RFSERR		BIT(5)  /* Receive Frame Synchronization Error */
172 #define SISTR_RFUDF		BIT(4)  /* Receive FIFO Underflow */
173 #define SISTR_RFOVF		BIT(3)  /* Receive FIFO Overflow */
174 
175 /* SIIER */
176 #define SIIER_TDMAE		BIT(31) /* Transmit Data DMA Transfer Req. Enable */
177 #define SIIER_TFEMPE		BIT(29) /* Transmit FIFO Empty Enable */
178 #define SIIER_TDREQE		BIT(28) /* Transmit Data Transfer Request Enable */
179 #define SIIER_TEOFE		BIT(23) /* Frame Transmission End Enable */
180 #define SIIER_TFSERRE		BIT(21) /* Transmit Frame Sync Error Enable */
181 #define SIIER_TFOVFE		BIT(20) /* Transmit FIFO Overflow Enable */
182 #define SIIER_TFUDFE		BIT(19) /* Transmit FIFO Underflow Enable */
183 #define SIIER_RDMAE		BIT(15) /* Receive Data DMA Transfer Req. Enable */
184 #define SIIER_RFFULE		BIT(13) /* Receive FIFO Full Enable */
185 #define SIIER_RDREQE		BIT(12) /* Receive Data Transfer Request Enable */
186 #define SIIER_REOFE		BIT(7)  /* Frame Reception End Enable */
187 #define SIIER_RFSERRE		BIT(5)  /* Receive Frame Sync Error Enable */
188 #define SIIER_RFUDFE		BIT(4)  /* Receive FIFO Underflow Enable */
189 #define SIIER_RFOVFE		BIT(3)  /* Receive FIFO Overflow Enable */
190 
191 
192 static u32 sh_msiof_read(struct sh_msiof_spi_priv *p, int reg_offs)
193 {
194 	switch (reg_offs) {
195 	case SITSCR:
196 	case SIRSCR:
197 		return ioread16(p->mapbase + reg_offs);
198 	default:
199 		return ioread32(p->mapbase + reg_offs);
200 	}
201 }
202 
203 static void sh_msiof_write(struct sh_msiof_spi_priv *p, int reg_offs,
204 			   u32 value)
205 {
206 	switch (reg_offs) {
207 	case SITSCR:
208 	case SIRSCR:
209 		iowrite16(value, p->mapbase + reg_offs);
210 		break;
211 	default:
212 		iowrite32(value, p->mapbase + reg_offs);
213 		break;
214 	}
215 }
216 
217 static int sh_msiof_modify_ctr_wait(struct sh_msiof_spi_priv *p,
218 				    u32 clr, u32 set)
219 {
220 	u32 mask = clr | set;
221 	u32 data;
222 
223 	data = sh_msiof_read(p, SICTR);
224 	data &= ~clr;
225 	data |= set;
226 	sh_msiof_write(p, SICTR, data);
227 
228 	return readl_poll_timeout_atomic(p->mapbase + SICTR, data,
229 					 (data & mask) == set, 1, 100);
230 }
231 
232 static irqreturn_t sh_msiof_spi_irq(int irq, void *data)
233 {
234 	struct sh_msiof_spi_priv *p = data;
235 
236 	/* just disable the interrupt and wake up */
237 	sh_msiof_write(p, SIIER, 0);
238 	complete(&p->done);
239 
240 	return IRQ_HANDLED;
241 }
242 
243 static void sh_msiof_spi_reset_regs(struct sh_msiof_spi_priv *p)
244 {
245 	u32 mask = SICTR_TXRST | SICTR_RXRST;
246 	u32 data;
247 
248 	data = sh_msiof_read(p, SICTR);
249 	data |= mask;
250 	sh_msiof_write(p, SICTR, data);
251 
252 	readl_poll_timeout_atomic(p->mapbase + SICTR, data, !(data & mask), 1,
253 				  100);
254 }
255 
256 static const u32 sh_msiof_spi_div_array[] = {
257 	SISCR_BRDV_DIV_1, SISCR_BRDV_DIV_2, SISCR_BRDV_DIV_4,
258 	SISCR_BRDV_DIV_8, SISCR_BRDV_DIV_16, SISCR_BRDV_DIV_32,
259 };
260 
261 static void sh_msiof_spi_set_clk_regs(struct sh_msiof_spi_priv *p,
262 				      struct spi_transfer *t)
263 {
264 	unsigned long parent_rate = clk_get_rate(p->clk);
265 	unsigned int div_pow = p->min_div_pow;
266 	u32 spi_hz = t->speed_hz;
267 	unsigned long div;
268 	u32 brps, scr;
269 
270 	if (!spi_hz || !parent_rate) {
271 		WARN(1, "Invalid clock rate parameters %lu and %u\n",
272 		     parent_rate, spi_hz);
273 		return;
274 	}
275 
276 	div = DIV_ROUND_UP(parent_rate, spi_hz);
277 	if (div <= 1024) {
278 		/* SISCR_BRDV_DIV_1 is valid only if BRPS is x 1/1 or x 1/2 */
279 		if (!div_pow && div <= 32 && div > 2)
280 			div_pow = 1;
281 
282 		if (div_pow)
283 			brps = (div + 1) >> div_pow;
284 		else
285 			brps = div;
286 
287 		for (; brps > 32; div_pow++)
288 			brps = (brps + 1) >> 1;
289 	} else {
290 		/* Set transfer rate composite divisor to 2^5 * 32 = 1024 */
291 		dev_err(&p->pdev->dev,
292 			"Requested SPI transfer rate %d is too low\n", spi_hz);
293 		div_pow = 5;
294 		brps = 32;
295 	}
296 
297 	t->effective_speed_hz = parent_rate / (brps << div_pow);
298 
299 	scr = sh_msiof_spi_div_array[div_pow] | SISCR_BRPS(brps);
300 	sh_msiof_write(p, SITSCR, scr);
301 	if (!(p->ctlr->flags & SPI_CONTROLLER_MUST_TX))
302 		sh_msiof_write(p, SIRSCR, scr);
303 }
304 
305 static u32 sh_msiof_get_delay_bit(u32 dtdl_or_syncdl)
306 {
307 	/*
308 	 * DTDL/SYNCDL bit	: p->info->dtdl or p->info->syncdl
309 	 * b'000		: 0
310 	 * b'001		: 100
311 	 * b'010		: 200
312 	 * b'011 (SYNCDL only)	: 300
313 	 * b'101		: 50
314 	 * b'110		: 150
315 	 */
316 	if (dtdl_or_syncdl % 100)
317 		return dtdl_or_syncdl / 100 + 5;
318 	else
319 		return dtdl_or_syncdl / 100;
320 }
321 
322 static u32 sh_msiof_spi_get_dtdl_and_syncdl(struct sh_msiof_spi_priv *p)
323 {
324 	u32 val;
325 
326 	if (!p->info)
327 		return 0;
328 
329 	/* check if DTDL and SYNCDL is allowed value */
330 	if (p->info->dtdl > 200 || p->info->syncdl > 300) {
331 		dev_warn(&p->pdev->dev, "DTDL or SYNCDL is too large\n");
332 		return 0;
333 	}
334 
335 	/* check if the sum of DTDL and SYNCDL becomes an integer value  */
336 	if ((p->info->dtdl + p->info->syncdl) % 100) {
337 		dev_warn(&p->pdev->dev, "the sum of DTDL/SYNCDL is not good\n");
338 		return 0;
339 	}
340 
341 	val = sh_msiof_get_delay_bit(p->info->dtdl) << SIMDR1_DTDL_SHIFT;
342 	val |= sh_msiof_get_delay_bit(p->info->syncdl) << SIMDR1_SYNCDL_SHIFT;
343 
344 	return val;
345 }
346 
347 static void sh_msiof_spi_set_pin_regs(struct sh_msiof_spi_priv *p, u32 ss,
348 				      u32 cpol, u32 cpha,
349 				      u32 tx_hi_z, u32 lsb_first, u32 cs_high)
350 {
351 	u32 tmp;
352 	int edge;
353 
354 	/*
355 	 * CPOL CPHA     TSCKIZ RSCKIZ TEDG REDG
356 	 *    0    0         10     10    1    1
357 	 *    0    1         10     10    0    0
358 	 *    1    0         11     11    0    0
359 	 *    1    1         11     11    1    1
360 	 */
361 	tmp = SIMDR1_SYNCMD_SPI | 1 << SIMDR1_FLD_SHIFT | SIMDR1_XXSTP;
362 	tmp |= !cs_high << SIMDR1_SYNCAC_SHIFT;
363 	tmp |= lsb_first << SIMDR1_BITLSB_SHIFT;
364 	tmp |= sh_msiof_spi_get_dtdl_and_syncdl(p);
365 	if (spi_controller_is_slave(p->ctlr)) {
366 		sh_msiof_write(p, SITMDR1, tmp | SITMDR1_PCON);
367 	} else {
368 		sh_msiof_write(p, SITMDR1,
369 			       tmp | SIMDR1_TRMD | SITMDR1_PCON |
370 			       (ss < MAX_SS ? ss : 0) << SITMDR1_SYNCCH_SHIFT);
371 	}
372 	if (p->ctlr->flags & SPI_CONTROLLER_MUST_TX) {
373 		/* These bits are reserved if RX needs TX */
374 		tmp &= ~0x0000ffff;
375 	}
376 	sh_msiof_write(p, SIRMDR1, tmp);
377 
378 	tmp = 0;
379 	tmp |= SICTR_TSCKIZ_SCK | cpol << SICTR_TSCKIZ_POL_SHIFT;
380 	tmp |= SICTR_RSCKIZ_SCK | cpol << SICTR_RSCKIZ_POL_SHIFT;
381 
382 	edge = cpol ^ !cpha;
383 
384 	tmp |= edge << SICTR_TEDG_SHIFT;
385 	tmp |= edge << SICTR_REDG_SHIFT;
386 	tmp |= tx_hi_z ? SICTR_TXDIZ_HIZ : SICTR_TXDIZ_LOW;
387 	sh_msiof_write(p, SICTR, tmp);
388 }
389 
390 static void sh_msiof_spi_set_mode_regs(struct sh_msiof_spi_priv *p,
391 				       const void *tx_buf, void *rx_buf,
392 				       u32 bits, u32 words)
393 {
394 	u32 dr2 = SIMDR2_BITLEN1(bits) | SIMDR2_WDLEN1(words);
395 
396 	if (tx_buf || (p->ctlr->flags & SPI_CONTROLLER_MUST_TX))
397 		sh_msiof_write(p, SITMDR2, dr2);
398 	else
399 		sh_msiof_write(p, SITMDR2, dr2 | SIMDR2_GRPMASK1);
400 
401 	if (rx_buf)
402 		sh_msiof_write(p, SIRMDR2, dr2);
403 }
404 
405 static void sh_msiof_reset_str(struct sh_msiof_spi_priv *p)
406 {
407 	sh_msiof_write(p, SISTR,
408 		       sh_msiof_read(p, SISTR) & ~(SISTR_TDREQ | SISTR_RDREQ));
409 }
410 
411 static void sh_msiof_spi_write_fifo_8(struct sh_msiof_spi_priv *p,
412 				      const void *tx_buf, int words, int fs)
413 {
414 	const u8 *buf_8 = tx_buf;
415 	int k;
416 
417 	for (k = 0; k < words; k++)
418 		sh_msiof_write(p, SITFDR, buf_8[k] << fs);
419 }
420 
421 static void sh_msiof_spi_write_fifo_16(struct sh_msiof_spi_priv *p,
422 				       const void *tx_buf, int words, int fs)
423 {
424 	const u16 *buf_16 = tx_buf;
425 	int k;
426 
427 	for (k = 0; k < words; k++)
428 		sh_msiof_write(p, SITFDR, buf_16[k] << fs);
429 }
430 
431 static void sh_msiof_spi_write_fifo_16u(struct sh_msiof_spi_priv *p,
432 					const void *tx_buf, int words, int fs)
433 {
434 	const u16 *buf_16 = tx_buf;
435 	int k;
436 
437 	for (k = 0; k < words; k++)
438 		sh_msiof_write(p, SITFDR, get_unaligned(&buf_16[k]) << fs);
439 }
440 
441 static void sh_msiof_spi_write_fifo_32(struct sh_msiof_spi_priv *p,
442 				       const void *tx_buf, int words, int fs)
443 {
444 	const u32 *buf_32 = tx_buf;
445 	int k;
446 
447 	for (k = 0; k < words; k++)
448 		sh_msiof_write(p, SITFDR, buf_32[k] << fs);
449 }
450 
451 static void sh_msiof_spi_write_fifo_32u(struct sh_msiof_spi_priv *p,
452 					const void *tx_buf, int words, int fs)
453 {
454 	const u32 *buf_32 = tx_buf;
455 	int k;
456 
457 	for (k = 0; k < words; k++)
458 		sh_msiof_write(p, SITFDR, get_unaligned(&buf_32[k]) << fs);
459 }
460 
461 static void sh_msiof_spi_write_fifo_s32(struct sh_msiof_spi_priv *p,
462 					const void *tx_buf, int words, int fs)
463 {
464 	const u32 *buf_32 = tx_buf;
465 	int k;
466 
467 	for (k = 0; k < words; k++)
468 		sh_msiof_write(p, SITFDR, swab32(buf_32[k] << fs));
469 }
470 
471 static void sh_msiof_spi_write_fifo_s32u(struct sh_msiof_spi_priv *p,
472 					 const void *tx_buf, int words, int fs)
473 {
474 	const u32 *buf_32 = tx_buf;
475 	int k;
476 
477 	for (k = 0; k < words; k++)
478 		sh_msiof_write(p, SITFDR, swab32(get_unaligned(&buf_32[k]) << fs));
479 }
480 
481 static void sh_msiof_spi_read_fifo_8(struct sh_msiof_spi_priv *p,
482 				     void *rx_buf, int words, int fs)
483 {
484 	u8 *buf_8 = rx_buf;
485 	int k;
486 
487 	for (k = 0; k < words; k++)
488 		buf_8[k] = sh_msiof_read(p, SIRFDR) >> fs;
489 }
490 
491 static void sh_msiof_spi_read_fifo_16(struct sh_msiof_spi_priv *p,
492 				      void *rx_buf, int words, int fs)
493 {
494 	u16 *buf_16 = rx_buf;
495 	int k;
496 
497 	for (k = 0; k < words; k++)
498 		buf_16[k] = sh_msiof_read(p, SIRFDR) >> fs;
499 }
500 
501 static void sh_msiof_spi_read_fifo_16u(struct sh_msiof_spi_priv *p,
502 				       void *rx_buf, int words, int fs)
503 {
504 	u16 *buf_16 = rx_buf;
505 	int k;
506 
507 	for (k = 0; k < words; k++)
508 		put_unaligned(sh_msiof_read(p, SIRFDR) >> fs, &buf_16[k]);
509 }
510 
511 static void sh_msiof_spi_read_fifo_32(struct sh_msiof_spi_priv *p,
512 				      void *rx_buf, int words, int fs)
513 {
514 	u32 *buf_32 = rx_buf;
515 	int k;
516 
517 	for (k = 0; k < words; k++)
518 		buf_32[k] = sh_msiof_read(p, SIRFDR) >> fs;
519 }
520 
521 static void sh_msiof_spi_read_fifo_32u(struct sh_msiof_spi_priv *p,
522 				       void *rx_buf, int words, int fs)
523 {
524 	u32 *buf_32 = rx_buf;
525 	int k;
526 
527 	for (k = 0; k < words; k++)
528 		put_unaligned(sh_msiof_read(p, SIRFDR) >> fs, &buf_32[k]);
529 }
530 
531 static void sh_msiof_spi_read_fifo_s32(struct sh_msiof_spi_priv *p,
532 				       void *rx_buf, int words, int fs)
533 {
534 	u32 *buf_32 = rx_buf;
535 	int k;
536 
537 	for (k = 0; k < words; k++)
538 		buf_32[k] = swab32(sh_msiof_read(p, SIRFDR) >> fs);
539 }
540 
541 static void sh_msiof_spi_read_fifo_s32u(struct sh_msiof_spi_priv *p,
542 				       void *rx_buf, int words, int fs)
543 {
544 	u32 *buf_32 = rx_buf;
545 	int k;
546 
547 	for (k = 0; k < words; k++)
548 		put_unaligned(swab32(sh_msiof_read(p, SIRFDR) >> fs), &buf_32[k]);
549 }
550 
551 static int sh_msiof_spi_setup(struct spi_device *spi)
552 {
553 	struct sh_msiof_spi_priv *p =
554 		spi_controller_get_devdata(spi->controller);
555 	u32 clr, set, tmp;
556 
557 	if (spi_get_csgpiod(spi, 0) || spi_controller_is_slave(p->ctlr))
558 		return 0;
559 
560 	if (p->native_cs_inited &&
561 	    (p->native_cs_high == !!(spi->mode & SPI_CS_HIGH)))
562 		return 0;
563 
564 	/* Configure native chip select mode/polarity early */
565 	clr = SIMDR1_SYNCMD_MASK;
566 	set = SIMDR1_SYNCMD_SPI;
567 	if (spi->mode & SPI_CS_HIGH)
568 		clr |= BIT(SIMDR1_SYNCAC_SHIFT);
569 	else
570 		set |= BIT(SIMDR1_SYNCAC_SHIFT);
571 	pm_runtime_get_sync(&p->pdev->dev);
572 	tmp = sh_msiof_read(p, SITMDR1) & ~clr;
573 	sh_msiof_write(p, SITMDR1, tmp | set | SIMDR1_TRMD | SITMDR1_PCON);
574 	tmp = sh_msiof_read(p, SIRMDR1) & ~clr;
575 	sh_msiof_write(p, SIRMDR1, tmp | set);
576 	pm_runtime_put(&p->pdev->dev);
577 	p->native_cs_high = spi->mode & SPI_CS_HIGH;
578 	p->native_cs_inited = true;
579 	return 0;
580 }
581 
582 static int sh_msiof_prepare_message(struct spi_controller *ctlr,
583 				    struct spi_message *msg)
584 {
585 	struct sh_msiof_spi_priv *p = spi_controller_get_devdata(ctlr);
586 	const struct spi_device *spi = msg->spi;
587 	u32 ss, cs_high;
588 
589 	/* Configure pins before asserting CS */
590 	if (spi_get_csgpiod(spi, 0)) {
591 		ss = ctlr->unused_native_cs;
592 		cs_high = p->native_cs_high;
593 	} else {
594 		ss = spi_get_chipselect(spi, 0);
595 		cs_high = !!(spi->mode & SPI_CS_HIGH);
596 	}
597 	sh_msiof_spi_set_pin_regs(p, ss, !!(spi->mode & SPI_CPOL),
598 				  !!(spi->mode & SPI_CPHA),
599 				  !!(spi->mode & SPI_3WIRE),
600 				  !!(spi->mode & SPI_LSB_FIRST), cs_high);
601 	return 0;
602 }
603 
604 static int sh_msiof_spi_start(struct sh_msiof_spi_priv *p, void *rx_buf)
605 {
606 	bool slave = spi_controller_is_slave(p->ctlr);
607 	int ret = 0;
608 
609 	/* setup clock and rx/tx signals */
610 	if (!slave)
611 		ret = sh_msiof_modify_ctr_wait(p, 0, SICTR_TSCKE);
612 	if (rx_buf && !ret)
613 		ret = sh_msiof_modify_ctr_wait(p, 0, SICTR_RXE);
614 	if (!ret)
615 		ret = sh_msiof_modify_ctr_wait(p, 0, SICTR_TXE);
616 
617 	/* start by setting frame bit */
618 	if (!ret && !slave)
619 		ret = sh_msiof_modify_ctr_wait(p, 0, SICTR_TFSE);
620 
621 	return ret;
622 }
623 
624 static int sh_msiof_spi_stop(struct sh_msiof_spi_priv *p, void *rx_buf)
625 {
626 	bool slave = spi_controller_is_slave(p->ctlr);
627 	int ret = 0;
628 
629 	/* shut down frame, rx/tx and clock signals */
630 	if (!slave)
631 		ret = sh_msiof_modify_ctr_wait(p, SICTR_TFSE, 0);
632 	if (!ret)
633 		ret = sh_msiof_modify_ctr_wait(p, SICTR_TXE, 0);
634 	if (rx_buf && !ret)
635 		ret = sh_msiof_modify_ctr_wait(p, SICTR_RXE, 0);
636 	if (!ret && !slave)
637 		ret = sh_msiof_modify_ctr_wait(p, SICTR_TSCKE, 0);
638 
639 	return ret;
640 }
641 
642 static int sh_msiof_slave_abort(struct spi_controller *ctlr)
643 {
644 	struct sh_msiof_spi_priv *p = spi_controller_get_devdata(ctlr);
645 
646 	p->slave_aborted = true;
647 	complete(&p->done);
648 	complete(&p->done_txdma);
649 	return 0;
650 }
651 
652 static int sh_msiof_wait_for_completion(struct sh_msiof_spi_priv *p,
653 					struct completion *x)
654 {
655 	if (spi_controller_is_slave(p->ctlr)) {
656 		if (wait_for_completion_interruptible(x) ||
657 		    p->slave_aborted) {
658 			dev_dbg(&p->pdev->dev, "interrupted\n");
659 			return -EINTR;
660 		}
661 	} else {
662 		if (!wait_for_completion_timeout(x, HZ)) {
663 			dev_err(&p->pdev->dev, "timeout\n");
664 			return -ETIMEDOUT;
665 		}
666 	}
667 
668 	return 0;
669 }
670 
671 static int sh_msiof_spi_txrx_once(struct sh_msiof_spi_priv *p,
672 				  void (*tx_fifo)(struct sh_msiof_spi_priv *,
673 						  const void *, int, int),
674 				  void (*rx_fifo)(struct sh_msiof_spi_priv *,
675 						  void *, int, int),
676 				  const void *tx_buf, void *rx_buf,
677 				  int words, int bits)
678 {
679 	int fifo_shift;
680 	int ret;
681 
682 	/* limit maximum word transfer to rx/tx fifo size */
683 	if (tx_buf)
684 		words = min_t(int, words, p->tx_fifo_size);
685 	if (rx_buf)
686 		words = min_t(int, words, p->rx_fifo_size);
687 
688 	/* the fifo contents need shifting */
689 	fifo_shift = 32 - bits;
690 
691 	/* default FIFO watermarks for PIO */
692 	sh_msiof_write(p, SIFCTR, 0);
693 
694 	/* setup msiof transfer mode registers */
695 	sh_msiof_spi_set_mode_regs(p, tx_buf, rx_buf, bits, words);
696 	sh_msiof_write(p, SIIER, SIIER_TEOFE | SIIER_REOFE);
697 
698 	/* write tx fifo */
699 	if (tx_buf)
700 		tx_fifo(p, tx_buf, words, fifo_shift);
701 
702 	reinit_completion(&p->done);
703 	p->slave_aborted = false;
704 
705 	ret = sh_msiof_spi_start(p, rx_buf);
706 	if (ret) {
707 		dev_err(&p->pdev->dev, "failed to start hardware\n");
708 		goto stop_ier;
709 	}
710 
711 	/* wait for tx fifo to be emptied / rx fifo to be filled */
712 	ret = sh_msiof_wait_for_completion(p, &p->done);
713 	if (ret)
714 		goto stop_reset;
715 
716 	/* read rx fifo */
717 	if (rx_buf)
718 		rx_fifo(p, rx_buf, words, fifo_shift);
719 
720 	/* clear status bits */
721 	sh_msiof_reset_str(p);
722 
723 	ret = sh_msiof_spi_stop(p, rx_buf);
724 	if (ret) {
725 		dev_err(&p->pdev->dev, "failed to shut down hardware\n");
726 		return ret;
727 	}
728 
729 	return words;
730 
731 stop_reset:
732 	sh_msiof_reset_str(p);
733 	sh_msiof_spi_stop(p, rx_buf);
734 stop_ier:
735 	sh_msiof_write(p, SIIER, 0);
736 	return ret;
737 }
738 
739 static void sh_msiof_dma_complete(void *arg)
740 {
741 	complete(arg);
742 }
743 
744 static int sh_msiof_dma_once(struct sh_msiof_spi_priv *p, const void *tx,
745 			     void *rx, unsigned int len)
746 {
747 	u32 ier_bits = 0;
748 	struct dma_async_tx_descriptor *desc_tx = NULL, *desc_rx = NULL;
749 	dma_cookie_t cookie;
750 	int ret;
751 
752 	/* First prepare and submit the DMA request(s), as this may fail */
753 	if (rx) {
754 		ier_bits |= SIIER_RDREQE | SIIER_RDMAE;
755 		desc_rx = dmaengine_prep_slave_single(p->ctlr->dma_rx,
756 					p->rx_dma_addr, len, DMA_DEV_TO_MEM,
757 					DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
758 		if (!desc_rx)
759 			return -EAGAIN;
760 
761 		desc_rx->callback = sh_msiof_dma_complete;
762 		desc_rx->callback_param = &p->done;
763 		cookie = dmaengine_submit(desc_rx);
764 		if (dma_submit_error(cookie))
765 			return cookie;
766 	}
767 
768 	if (tx) {
769 		ier_bits |= SIIER_TDREQE | SIIER_TDMAE;
770 		dma_sync_single_for_device(p->ctlr->dma_tx->device->dev,
771 					   p->tx_dma_addr, len, DMA_TO_DEVICE);
772 		desc_tx = dmaengine_prep_slave_single(p->ctlr->dma_tx,
773 					p->tx_dma_addr, len, DMA_MEM_TO_DEV,
774 					DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
775 		if (!desc_tx) {
776 			ret = -EAGAIN;
777 			goto no_dma_tx;
778 		}
779 
780 		desc_tx->callback = sh_msiof_dma_complete;
781 		desc_tx->callback_param = &p->done_txdma;
782 		cookie = dmaengine_submit(desc_tx);
783 		if (dma_submit_error(cookie)) {
784 			ret = cookie;
785 			goto no_dma_tx;
786 		}
787 	}
788 
789 	/* 1 stage FIFO watermarks for DMA */
790 	sh_msiof_write(p, SIFCTR, SIFCTR_TFWM_1 | SIFCTR_RFWM_1);
791 
792 	/* setup msiof transfer mode registers (32-bit words) */
793 	sh_msiof_spi_set_mode_regs(p, tx, rx, 32, len / 4);
794 
795 	sh_msiof_write(p, SIIER, ier_bits);
796 
797 	reinit_completion(&p->done);
798 	if (tx)
799 		reinit_completion(&p->done_txdma);
800 	p->slave_aborted = false;
801 
802 	/* Now start DMA */
803 	if (rx)
804 		dma_async_issue_pending(p->ctlr->dma_rx);
805 	if (tx)
806 		dma_async_issue_pending(p->ctlr->dma_tx);
807 
808 	ret = sh_msiof_spi_start(p, rx);
809 	if (ret) {
810 		dev_err(&p->pdev->dev, "failed to start hardware\n");
811 		goto stop_dma;
812 	}
813 
814 	if (tx) {
815 		/* wait for tx DMA completion */
816 		ret = sh_msiof_wait_for_completion(p, &p->done_txdma);
817 		if (ret)
818 			goto stop_reset;
819 	}
820 
821 	if (rx) {
822 		/* wait for rx DMA completion */
823 		ret = sh_msiof_wait_for_completion(p, &p->done);
824 		if (ret)
825 			goto stop_reset;
826 
827 		sh_msiof_write(p, SIIER, 0);
828 	} else {
829 		/* wait for tx fifo to be emptied */
830 		sh_msiof_write(p, SIIER, SIIER_TEOFE);
831 		ret = sh_msiof_wait_for_completion(p, &p->done);
832 		if (ret)
833 			goto stop_reset;
834 	}
835 
836 	/* clear status bits */
837 	sh_msiof_reset_str(p);
838 
839 	ret = sh_msiof_spi_stop(p, rx);
840 	if (ret) {
841 		dev_err(&p->pdev->dev, "failed to shut down hardware\n");
842 		return ret;
843 	}
844 
845 	if (rx)
846 		dma_sync_single_for_cpu(p->ctlr->dma_rx->device->dev,
847 					p->rx_dma_addr, len, DMA_FROM_DEVICE);
848 
849 	return 0;
850 
851 stop_reset:
852 	sh_msiof_reset_str(p);
853 	sh_msiof_spi_stop(p, rx);
854 stop_dma:
855 	if (tx)
856 		dmaengine_terminate_sync(p->ctlr->dma_tx);
857 no_dma_tx:
858 	if (rx)
859 		dmaengine_terminate_sync(p->ctlr->dma_rx);
860 	sh_msiof_write(p, SIIER, 0);
861 	return ret;
862 }
863 
864 static void copy_bswap32(u32 *dst, const u32 *src, unsigned int words)
865 {
866 	/* src or dst can be unaligned, but not both */
867 	if ((unsigned long)src & 3) {
868 		while (words--) {
869 			*dst++ = swab32(get_unaligned(src));
870 			src++;
871 		}
872 	} else if ((unsigned long)dst & 3) {
873 		while (words--) {
874 			put_unaligned(swab32(*src++), dst);
875 			dst++;
876 		}
877 	} else {
878 		while (words--)
879 			*dst++ = swab32(*src++);
880 	}
881 }
882 
883 static void copy_wswap32(u32 *dst, const u32 *src, unsigned int words)
884 {
885 	/* src or dst can be unaligned, but not both */
886 	if ((unsigned long)src & 3) {
887 		while (words--) {
888 			*dst++ = swahw32(get_unaligned(src));
889 			src++;
890 		}
891 	} else if ((unsigned long)dst & 3) {
892 		while (words--) {
893 			put_unaligned(swahw32(*src++), dst);
894 			dst++;
895 		}
896 	} else {
897 		while (words--)
898 			*dst++ = swahw32(*src++);
899 	}
900 }
901 
902 static void copy_plain32(u32 *dst, const u32 *src, unsigned int words)
903 {
904 	memcpy(dst, src, words * 4);
905 }
906 
907 static int sh_msiof_transfer_one(struct spi_controller *ctlr,
908 				 struct spi_device *spi,
909 				 struct spi_transfer *t)
910 {
911 	struct sh_msiof_spi_priv *p = spi_controller_get_devdata(ctlr);
912 	void (*copy32)(u32 *, const u32 *, unsigned int);
913 	void (*tx_fifo)(struct sh_msiof_spi_priv *, const void *, int, int);
914 	void (*rx_fifo)(struct sh_msiof_spi_priv *, void *, int, int);
915 	const void *tx_buf = t->tx_buf;
916 	void *rx_buf = t->rx_buf;
917 	unsigned int len = t->len;
918 	unsigned int bits = t->bits_per_word;
919 	unsigned int bytes_per_word;
920 	unsigned int words;
921 	int n;
922 	bool swab;
923 	int ret;
924 
925 	/* reset registers */
926 	sh_msiof_spi_reset_regs(p);
927 
928 	/* setup clocks (clock already enabled in chipselect()) */
929 	if (!spi_controller_is_slave(p->ctlr))
930 		sh_msiof_spi_set_clk_regs(p, t);
931 
932 	while (ctlr->dma_tx && len > 15) {
933 		/*
934 		 *  DMA supports 32-bit words only, hence pack 8-bit and 16-bit
935 		 *  words, with byte resp. word swapping.
936 		 */
937 		unsigned int l = 0;
938 
939 		if (tx_buf)
940 			l = min(round_down(len, 4), p->tx_fifo_size * 4);
941 		if (rx_buf)
942 			l = min(round_down(len, 4), p->rx_fifo_size * 4);
943 
944 		if (bits <= 8) {
945 			copy32 = copy_bswap32;
946 		} else if (bits <= 16) {
947 			copy32 = copy_wswap32;
948 		} else {
949 			copy32 = copy_plain32;
950 		}
951 
952 		if (tx_buf)
953 			copy32(p->tx_dma_page, tx_buf, l / 4);
954 
955 		ret = sh_msiof_dma_once(p, tx_buf, rx_buf, l);
956 		if (ret == -EAGAIN) {
957 			dev_warn_once(&p->pdev->dev,
958 				"DMA not available, falling back to PIO\n");
959 			break;
960 		}
961 		if (ret)
962 			return ret;
963 
964 		if (rx_buf) {
965 			copy32(rx_buf, p->rx_dma_page, l / 4);
966 			rx_buf += l;
967 		}
968 		if (tx_buf)
969 			tx_buf += l;
970 
971 		len -= l;
972 		if (!len)
973 			return 0;
974 	}
975 
976 	if (bits <= 8 && len > 15) {
977 		bits = 32;
978 		swab = true;
979 	} else {
980 		swab = false;
981 	}
982 
983 	/* setup bytes per word and fifo read/write functions */
984 	if (bits <= 8) {
985 		bytes_per_word = 1;
986 		tx_fifo = sh_msiof_spi_write_fifo_8;
987 		rx_fifo = sh_msiof_spi_read_fifo_8;
988 	} else if (bits <= 16) {
989 		bytes_per_word = 2;
990 		if ((unsigned long)tx_buf & 0x01)
991 			tx_fifo = sh_msiof_spi_write_fifo_16u;
992 		else
993 			tx_fifo = sh_msiof_spi_write_fifo_16;
994 
995 		if ((unsigned long)rx_buf & 0x01)
996 			rx_fifo = sh_msiof_spi_read_fifo_16u;
997 		else
998 			rx_fifo = sh_msiof_spi_read_fifo_16;
999 	} else if (swab) {
1000 		bytes_per_word = 4;
1001 		if ((unsigned long)tx_buf & 0x03)
1002 			tx_fifo = sh_msiof_spi_write_fifo_s32u;
1003 		else
1004 			tx_fifo = sh_msiof_spi_write_fifo_s32;
1005 
1006 		if ((unsigned long)rx_buf & 0x03)
1007 			rx_fifo = sh_msiof_spi_read_fifo_s32u;
1008 		else
1009 			rx_fifo = sh_msiof_spi_read_fifo_s32;
1010 	} else {
1011 		bytes_per_word = 4;
1012 		if ((unsigned long)tx_buf & 0x03)
1013 			tx_fifo = sh_msiof_spi_write_fifo_32u;
1014 		else
1015 			tx_fifo = sh_msiof_spi_write_fifo_32;
1016 
1017 		if ((unsigned long)rx_buf & 0x03)
1018 			rx_fifo = sh_msiof_spi_read_fifo_32u;
1019 		else
1020 			rx_fifo = sh_msiof_spi_read_fifo_32;
1021 	}
1022 
1023 	/* transfer in fifo sized chunks */
1024 	words = len / bytes_per_word;
1025 
1026 	while (words > 0) {
1027 		n = sh_msiof_spi_txrx_once(p, tx_fifo, rx_fifo, tx_buf, rx_buf,
1028 					   words, bits);
1029 		if (n < 0)
1030 			return n;
1031 
1032 		if (tx_buf)
1033 			tx_buf += n * bytes_per_word;
1034 		if (rx_buf)
1035 			rx_buf += n * bytes_per_word;
1036 		words -= n;
1037 
1038 		if (words == 0 && (len % bytes_per_word)) {
1039 			words = len % bytes_per_word;
1040 			bits = t->bits_per_word;
1041 			bytes_per_word = 1;
1042 			tx_fifo = sh_msiof_spi_write_fifo_8;
1043 			rx_fifo = sh_msiof_spi_read_fifo_8;
1044 		}
1045 	}
1046 
1047 	return 0;
1048 }
1049 
1050 static const struct sh_msiof_chipdata sh_data = {
1051 	.bits_per_word_mask = SPI_BPW_RANGE_MASK(8, 32),
1052 	.tx_fifo_size = 64,
1053 	.rx_fifo_size = 64,
1054 	.ctlr_flags = 0,
1055 	.min_div_pow = 0,
1056 };
1057 
1058 static const struct sh_msiof_chipdata rcar_gen2_data = {
1059 	.bits_per_word_mask = SPI_BPW_MASK(8) | SPI_BPW_MASK(16) |
1060 			      SPI_BPW_MASK(24) | SPI_BPW_MASK(32),
1061 	.tx_fifo_size = 64,
1062 	.rx_fifo_size = 64,
1063 	.ctlr_flags = SPI_CONTROLLER_MUST_TX,
1064 	.min_div_pow = 0,
1065 };
1066 
1067 static const struct sh_msiof_chipdata rcar_gen3_data = {
1068 	.bits_per_word_mask = SPI_BPW_MASK(8) | SPI_BPW_MASK(16) |
1069 			      SPI_BPW_MASK(24) | SPI_BPW_MASK(32),
1070 	.tx_fifo_size = 64,
1071 	.rx_fifo_size = 64,
1072 	.ctlr_flags = SPI_CONTROLLER_MUST_TX,
1073 	.min_div_pow = 1,
1074 };
1075 
1076 static const struct of_device_id sh_msiof_match[] __maybe_unused = {
1077 	{ .compatible = "renesas,sh-mobile-msiof", .data = &sh_data },
1078 	{ .compatible = "renesas,msiof-r8a7743",   .data = &rcar_gen2_data },
1079 	{ .compatible = "renesas,msiof-r8a7745",   .data = &rcar_gen2_data },
1080 	{ .compatible = "renesas,msiof-r8a7790",   .data = &rcar_gen2_data },
1081 	{ .compatible = "renesas,msiof-r8a7791",   .data = &rcar_gen2_data },
1082 	{ .compatible = "renesas,msiof-r8a7792",   .data = &rcar_gen2_data },
1083 	{ .compatible = "renesas,msiof-r8a7793",   .data = &rcar_gen2_data },
1084 	{ .compatible = "renesas,msiof-r8a7794",   .data = &rcar_gen2_data },
1085 	{ .compatible = "renesas,rcar-gen2-msiof", .data = &rcar_gen2_data },
1086 	{ .compatible = "renesas,msiof-r8a7796",   .data = &rcar_gen3_data },
1087 	{ .compatible = "renesas,rcar-gen3-msiof", .data = &rcar_gen3_data },
1088 	{ .compatible = "renesas,rcar-gen4-msiof", .data = &rcar_gen3_data },
1089 	{ .compatible = "renesas,sh-msiof",        .data = &sh_data }, /* Deprecated */
1090 	{},
1091 };
1092 MODULE_DEVICE_TABLE(of, sh_msiof_match);
1093 
1094 #ifdef CONFIG_OF
1095 static struct sh_msiof_spi_info *sh_msiof_spi_parse_dt(struct device *dev)
1096 {
1097 	struct sh_msiof_spi_info *info;
1098 	struct device_node *np = dev->of_node;
1099 	u32 num_cs = 1;
1100 
1101 	info = devm_kzalloc(dev, sizeof(struct sh_msiof_spi_info), GFP_KERNEL);
1102 	if (!info)
1103 		return NULL;
1104 
1105 	info->mode = of_property_read_bool(np, "spi-slave") ? MSIOF_SPI_SLAVE
1106 							    : MSIOF_SPI_MASTER;
1107 
1108 	/* Parse the MSIOF properties */
1109 	if (info->mode == MSIOF_SPI_MASTER)
1110 		of_property_read_u32(np, "num-cs", &num_cs);
1111 	of_property_read_u32(np, "renesas,tx-fifo-size",
1112 					&info->tx_fifo_override);
1113 	of_property_read_u32(np, "renesas,rx-fifo-size",
1114 					&info->rx_fifo_override);
1115 	of_property_read_u32(np, "renesas,dtdl", &info->dtdl);
1116 	of_property_read_u32(np, "renesas,syncdl", &info->syncdl);
1117 
1118 	info->num_chipselect = num_cs;
1119 
1120 	return info;
1121 }
1122 #else
1123 static struct sh_msiof_spi_info *sh_msiof_spi_parse_dt(struct device *dev)
1124 {
1125 	return NULL;
1126 }
1127 #endif
1128 
1129 static struct dma_chan *sh_msiof_request_dma_chan(struct device *dev,
1130 	enum dma_transfer_direction dir, unsigned int id, dma_addr_t port_addr)
1131 {
1132 	dma_cap_mask_t mask;
1133 	struct dma_chan *chan;
1134 	struct dma_slave_config cfg;
1135 	int ret;
1136 
1137 	dma_cap_zero(mask);
1138 	dma_cap_set(DMA_SLAVE, mask);
1139 
1140 	chan = dma_request_slave_channel_compat(mask, shdma_chan_filter,
1141 				(void *)(unsigned long)id, dev,
1142 				dir == DMA_MEM_TO_DEV ? "tx" : "rx");
1143 	if (!chan) {
1144 		dev_warn(dev, "dma_request_slave_channel_compat failed\n");
1145 		return NULL;
1146 	}
1147 
1148 	memset(&cfg, 0, sizeof(cfg));
1149 	cfg.direction = dir;
1150 	if (dir == DMA_MEM_TO_DEV) {
1151 		cfg.dst_addr = port_addr;
1152 		cfg.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
1153 	} else {
1154 		cfg.src_addr = port_addr;
1155 		cfg.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
1156 	}
1157 
1158 	ret = dmaengine_slave_config(chan, &cfg);
1159 	if (ret) {
1160 		dev_warn(dev, "dmaengine_slave_config failed %d\n", ret);
1161 		dma_release_channel(chan);
1162 		return NULL;
1163 	}
1164 
1165 	return chan;
1166 }
1167 
1168 static int sh_msiof_request_dma(struct sh_msiof_spi_priv *p)
1169 {
1170 	struct platform_device *pdev = p->pdev;
1171 	struct device *dev = &pdev->dev;
1172 	const struct sh_msiof_spi_info *info = p->info;
1173 	unsigned int dma_tx_id, dma_rx_id;
1174 	const struct resource *res;
1175 	struct spi_controller *ctlr;
1176 	struct device *tx_dev, *rx_dev;
1177 
1178 	if (dev->of_node) {
1179 		/* In the OF case we will get the slave IDs from the DT */
1180 		dma_tx_id = 0;
1181 		dma_rx_id = 0;
1182 	} else if (info && info->dma_tx_id && info->dma_rx_id) {
1183 		dma_tx_id = info->dma_tx_id;
1184 		dma_rx_id = info->dma_rx_id;
1185 	} else {
1186 		/* The driver assumes no error */
1187 		return 0;
1188 	}
1189 
1190 	/* The DMA engine uses the second register set, if present */
1191 	res = platform_get_resource(pdev, IORESOURCE_MEM, 1);
1192 	if (!res)
1193 		res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
1194 
1195 	ctlr = p->ctlr;
1196 	ctlr->dma_tx = sh_msiof_request_dma_chan(dev, DMA_MEM_TO_DEV,
1197 						 dma_tx_id, res->start + SITFDR);
1198 	if (!ctlr->dma_tx)
1199 		return -ENODEV;
1200 
1201 	ctlr->dma_rx = sh_msiof_request_dma_chan(dev, DMA_DEV_TO_MEM,
1202 						 dma_rx_id, res->start + SIRFDR);
1203 	if (!ctlr->dma_rx)
1204 		goto free_tx_chan;
1205 
1206 	p->tx_dma_page = (void *)__get_free_page(GFP_KERNEL | GFP_DMA);
1207 	if (!p->tx_dma_page)
1208 		goto free_rx_chan;
1209 
1210 	p->rx_dma_page = (void *)__get_free_page(GFP_KERNEL | GFP_DMA);
1211 	if (!p->rx_dma_page)
1212 		goto free_tx_page;
1213 
1214 	tx_dev = ctlr->dma_tx->device->dev;
1215 	p->tx_dma_addr = dma_map_single(tx_dev, p->tx_dma_page, PAGE_SIZE,
1216 					DMA_TO_DEVICE);
1217 	if (dma_mapping_error(tx_dev, p->tx_dma_addr))
1218 		goto free_rx_page;
1219 
1220 	rx_dev = ctlr->dma_rx->device->dev;
1221 	p->rx_dma_addr = dma_map_single(rx_dev, p->rx_dma_page, PAGE_SIZE,
1222 					DMA_FROM_DEVICE);
1223 	if (dma_mapping_error(rx_dev, p->rx_dma_addr))
1224 		goto unmap_tx_page;
1225 
1226 	dev_info(dev, "DMA available");
1227 	return 0;
1228 
1229 unmap_tx_page:
1230 	dma_unmap_single(tx_dev, p->tx_dma_addr, PAGE_SIZE, DMA_TO_DEVICE);
1231 free_rx_page:
1232 	free_page((unsigned long)p->rx_dma_page);
1233 free_tx_page:
1234 	free_page((unsigned long)p->tx_dma_page);
1235 free_rx_chan:
1236 	dma_release_channel(ctlr->dma_rx);
1237 free_tx_chan:
1238 	dma_release_channel(ctlr->dma_tx);
1239 	ctlr->dma_tx = NULL;
1240 	return -ENODEV;
1241 }
1242 
1243 static void sh_msiof_release_dma(struct sh_msiof_spi_priv *p)
1244 {
1245 	struct spi_controller *ctlr = p->ctlr;
1246 
1247 	if (!ctlr->dma_tx)
1248 		return;
1249 
1250 	dma_unmap_single(ctlr->dma_rx->device->dev, p->rx_dma_addr, PAGE_SIZE,
1251 			 DMA_FROM_DEVICE);
1252 	dma_unmap_single(ctlr->dma_tx->device->dev, p->tx_dma_addr, PAGE_SIZE,
1253 			 DMA_TO_DEVICE);
1254 	free_page((unsigned long)p->rx_dma_page);
1255 	free_page((unsigned long)p->tx_dma_page);
1256 	dma_release_channel(ctlr->dma_rx);
1257 	dma_release_channel(ctlr->dma_tx);
1258 }
1259 
1260 static int sh_msiof_spi_probe(struct platform_device *pdev)
1261 {
1262 	struct spi_controller *ctlr;
1263 	const struct sh_msiof_chipdata *chipdata;
1264 	struct sh_msiof_spi_info *info;
1265 	struct sh_msiof_spi_priv *p;
1266 	unsigned long clksrc;
1267 	int i;
1268 	int ret;
1269 
1270 	chipdata = of_device_get_match_data(&pdev->dev);
1271 	if (chipdata) {
1272 		info = sh_msiof_spi_parse_dt(&pdev->dev);
1273 	} else {
1274 		chipdata = (const void *)pdev->id_entry->driver_data;
1275 		info = dev_get_platdata(&pdev->dev);
1276 	}
1277 
1278 	if (!info) {
1279 		dev_err(&pdev->dev, "failed to obtain device info\n");
1280 		return -ENXIO;
1281 	}
1282 
1283 	if (info->mode == MSIOF_SPI_SLAVE)
1284 		ctlr = spi_alloc_slave(&pdev->dev,
1285 				       sizeof(struct sh_msiof_spi_priv));
1286 	else
1287 		ctlr = spi_alloc_master(&pdev->dev,
1288 					sizeof(struct sh_msiof_spi_priv));
1289 	if (ctlr == NULL)
1290 		return -ENOMEM;
1291 
1292 	p = spi_controller_get_devdata(ctlr);
1293 
1294 	platform_set_drvdata(pdev, p);
1295 	p->ctlr = ctlr;
1296 	p->info = info;
1297 	p->min_div_pow = chipdata->min_div_pow;
1298 
1299 	init_completion(&p->done);
1300 	init_completion(&p->done_txdma);
1301 
1302 	p->clk = devm_clk_get(&pdev->dev, NULL);
1303 	if (IS_ERR(p->clk)) {
1304 		dev_err(&pdev->dev, "cannot get clock\n");
1305 		ret = PTR_ERR(p->clk);
1306 		goto err1;
1307 	}
1308 
1309 	i = platform_get_irq(pdev, 0);
1310 	if (i < 0) {
1311 		ret = i;
1312 		goto err1;
1313 	}
1314 
1315 	p->mapbase = devm_platform_ioremap_resource(pdev, 0);
1316 	if (IS_ERR(p->mapbase)) {
1317 		ret = PTR_ERR(p->mapbase);
1318 		goto err1;
1319 	}
1320 
1321 	ret = devm_request_irq(&pdev->dev, i, sh_msiof_spi_irq, 0,
1322 			       dev_name(&pdev->dev), p);
1323 	if (ret) {
1324 		dev_err(&pdev->dev, "unable to request irq\n");
1325 		goto err1;
1326 	}
1327 
1328 	p->pdev = pdev;
1329 	pm_runtime_enable(&pdev->dev);
1330 
1331 	/* Platform data may override FIFO sizes */
1332 	p->tx_fifo_size = chipdata->tx_fifo_size;
1333 	p->rx_fifo_size = chipdata->rx_fifo_size;
1334 	if (p->info->tx_fifo_override)
1335 		p->tx_fifo_size = p->info->tx_fifo_override;
1336 	if (p->info->rx_fifo_override)
1337 		p->rx_fifo_size = p->info->rx_fifo_override;
1338 
1339 	/* init controller code */
1340 	ctlr->mode_bits = SPI_CPOL | SPI_CPHA | SPI_CS_HIGH;
1341 	ctlr->mode_bits |= SPI_LSB_FIRST | SPI_3WIRE;
1342 	clksrc = clk_get_rate(p->clk);
1343 	ctlr->min_speed_hz = DIV_ROUND_UP(clksrc, 1024);
1344 	ctlr->max_speed_hz = DIV_ROUND_UP(clksrc, 1 << p->min_div_pow);
1345 	ctlr->flags = chipdata->ctlr_flags;
1346 	ctlr->bus_num = pdev->id;
1347 	ctlr->num_chipselect = p->info->num_chipselect;
1348 	ctlr->dev.of_node = pdev->dev.of_node;
1349 	ctlr->setup = sh_msiof_spi_setup;
1350 	ctlr->prepare_message = sh_msiof_prepare_message;
1351 	ctlr->slave_abort = sh_msiof_slave_abort;
1352 	ctlr->bits_per_word_mask = chipdata->bits_per_word_mask;
1353 	ctlr->auto_runtime_pm = true;
1354 	ctlr->transfer_one = sh_msiof_transfer_one;
1355 	ctlr->use_gpio_descriptors = true;
1356 	ctlr->max_native_cs = MAX_SS;
1357 
1358 	ret = sh_msiof_request_dma(p);
1359 	if (ret < 0)
1360 		dev_warn(&pdev->dev, "DMA not available, using PIO\n");
1361 
1362 	ret = devm_spi_register_controller(&pdev->dev, ctlr);
1363 	if (ret < 0) {
1364 		dev_err(&pdev->dev, "devm_spi_register_controller error.\n");
1365 		goto err2;
1366 	}
1367 
1368 	return 0;
1369 
1370  err2:
1371 	sh_msiof_release_dma(p);
1372 	pm_runtime_disable(&pdev->dev);
1373  err1:
1374 	spi_controller_put(ctlr);
1375 	return ret;
1376 }
1377 
1378 static void sh_msiof_spi_remove(struct platform_device *pdev)
1379 {
1380 	struct sh_msiof_spi_priv *p = platform_get_drvdata(pdev);
1381 
1382 	sh_msiof_release_dma(p);
1383 	pm_runtime_disable(&pdev->dev);
1384 }
1385 
1386 static const struct platform_device_id spi_driver_ids[] = {
1387 	{ "spi_sh_msiof",	(kernel_ulong_t)&sh_data },
1388 	{},
1389 };
1390 MODULE_DEVICE_TABLE(platform, spi_driver_ids);
1391 
1392 #ifdef CONFIG_PM_SLEEP
1393 static int sh_msiof_spi_suspend(struct device *dev)
1394 {
1395 	struct sh_msiof_spi_priv *p = dev_get_drvdata(dev);
1396 
1397 	return spi_controller_suspend(p->ctlr);
1398 }
1399 
1400 static int sh_msiof_spi_resume(struct device *dev)
1401 {
1402 	struct sh_msiof_spi_priv *p = dev_get_drvdata(dev);
1403 
1404 	return spi_controller_resume(p->ctlr);
1405 }
1406 
1407 static SIMPLE_DEV_PM_OPS(sh_msiof_spi_pm_ops, sh_msiof_spi_suspend,
1408 			 sh_msiof_spi_resume);
1409 #define DEV_PM_OPS	(&sh_msiof_spi_pm_ops)
1410 #else
1411 #define DEV_PM_OPS	NULL
1412 #endif /* CONFIG_PM_SLEEP */
1413 
1414 static struct platform_driver sh_msiof_spi_drv = {
1415 	.probe		= sh_msiof_spi_probe,
1416 	.remove_new	= sh_msiof_spi_remove,
1417 	.id_table	= spi_driver_ids,
1418 	.driver		= {
1419 		.name		= "spi_sh_msiof",
1420 		.pm		= DEV_PM_OPS,
1421 		.of_match_table = of_match_ptr(sh_msiof_match),
1422 	},
1423 };
1424 module_platform_driver(sh_msiof_spi_drv);
1425 
1426 MODULE_DESCRIPTION("SuperH MSIOF SPI Controller Interface Driver");
1427 MODULE_AUTHOR("Magnus Damm");
1428 MODULE_LICENSE("GPL v2");
1429