xref: /linux/drivers/spi/spi-sh-msiof.c (revision 63307d015b91e626c97bb82e88054af3d0b74643)
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/gpio.h>
18 #include <linux/gpio/consumer.h>
19 #include <linux/interrupt.h>
20 #include <linux/io.h>
21 #include <linux/iopoll.h>
22 #include <linux/kernel.h>
23 #include <linux/module.h>
24 #include <linux/of.h>
25 #include <linux/of_device.h>
26 #include <linux/platform_device.h>
27 #include <linux/pm_runtime.h>
28 #include <linux/sh_dma.h>
29 
30 #include <linux/spi/sh_msiof.h>
31 #include <linux/spi/spi.h>
32 
33 #include <asm/unaligned.h>
34 
35 struct sh_msiof_chipdata {
36 	u32 bits_per_word_mask;
37 	u16 tx_fifo_size;
38 	u16 rx_fifo_size;
39 	u16 ctlr_flags;
40 	u16 min_div_pow;
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 	unsigned short unused_ss;
59 	bool native_cs_inited;
60 	bool native_cs_high;
61 	bool slave_aborted;
62 };
63 
64 #define MAX_SS	3	/* Maximum number of native chip selects */
65 
66 #define TMDR1	0x00	/* Transmit Mode Register 1 */
67 #define TMDR2	0x04	/* Transmit Mode Register 2 */
68 #define TMDR3	0x08	/* Transmit Mode Register 3 */
69 #define RMDR1	0x10	/* Receive Mode Register 1 */
70 #define RMDR2	0x14	/* Receive Mode Register 2 */
71 #define RMDR3	0x18	/* Receive Mode Register 3 */
72 #define TSCR	0x20	/* Transmit Clock Select Register */
73 #define RSCR	0x22	/* Receive Clock Select Register (SH, A1, APE6) */
74 #define CTR	0x28	/* Control Register */
75 #define FCTR	0x30	/* FIFO Control Register */
76 #define STR	0x40	/* Status Register */
77 #define IER	0x44	/* Interrupt Enable Register */
78 #define TDR1	0x48	/* Transmit Control Data Register 1 (SH, A1) */
79 #define TDR2	0x4c	/* Transmit Control Data Register 2 (SH, A1) */
80 #define TFDR	0x50	/* Transmit FIFO Data Register */
81 #define RDR1	0x58	/* Receive Control Data Register 1 (SH, A1) */
82 #define RDR2	0x5c	/* Receive Control Data Register 2 (SH, A1) */
83 #define RFDR	0x60	/* Receive FIFO Data Register */
84 
85 /* TMDR1 and RMDR1 */
86 #define MDR1_TRMD	   BIT(31)  /* Transfer Mode (1 = Master mode) */
87 #define MDR1_SYNCMD_MASK   GENMASK(29, 28) /* SYNC Mode */
88 #define MDR1_SYNCMD_SPI	   (2 << 28)/*   Level mode/SPI */
89 #define MDR1_SYNCMD_LR	   (3 << 28)/*   L/R mode */
90 #define MDR1_SYNCAC_SHIFT  25       /* Sync Polarity (1 = Active-low) */
91 #define MDR1_BITLSB_SHIFT  24       /* MSB/LSB First (1 = LSB first) */
92 #define MDR1_DTDL_SHIFT	   20       /* Data Pin Bit Delay for MSIOF_SYNC */
93 #define MDR1_SYNCDL_SHIFT  16       /* Frame Sync Signal Timing Delay */
94 #define MDR1_FLD_MASK	   GENMASK(3, 2) /* Frame Sync Signal Interval (0-3) */
95 #define MDR1_FLD_SHIFT	   2
96 #define MDR1_XXSTP	   BIT(0)   /* Transmission/Reception Stop on FIFO */
97 /* TMDR1 */
98 #define TMDR1_PCON	   BIT(30)  /* Transfer Signal Connection */
99 #define TMDR1_SYNCCH_MASK  GENMASK(27, 26) /* Sync Signal Channel Select */
100 #define TMDR1_SYNCCH_SHIFT 26       /* 0=MSIOF_SYNC, 1=MSIOF_SS1, 2=MSIOF_SS2 */
101 
102 /* TMDR2 and RMDR2 */
103 #define MDR2_BITLEN1(i)	(((i) - 1) << 24) /* Data Size (8-32 bits) */
104 #define MDR2_WDLEN1(i)	(((i) - 1) << 16) /* Word Count (1-64/256 (SH, A1))) */
105 #define MDR2_GRPMASK1	BIT(0)      /* Group Output Mask 1 (SH, A1) */
106 
107 /* TSCR and RSCR */
108 #define SCR_BRPS_MASK	GENMASK(12, 8) /* Prescaler Setting (1-32) */
109 #define SCR_BRPS(i)	(((i) - 1) << 8)
110 #define SCR_BRDV_MASK	GENMASK(2, 0) /* Baud Rate Generator's Division Ratio */
111 #define SCR_BRDV_DIV_2	0
112 #define SCR_BRDV_DIV_4	1
113 #define SCR_BRDV_DIV_8	2
114 #define SCR_BRDV_DIV_16	3
115 #define SCR_BRDV_DIV_32	4
116 #define SCR_BRDV_DIV_1	7
117 
118 /* CTR */
119 #define CTR_TSCKIZ_MASK	GENMASK(31, 30) /* Transmit Clock I/O Polarity Select */
120 #define CTR_TSCKIZ_SCK	BIT(31)   /*   Disable SCK when TX disabled */
121 #define CTR_TSCKIZ_POL_SHIFT 30   /*   Transmit Clock Polarity */
122 #define CTR_RSCKIZ_MASK	GENMASK(29, 28) /* Receive Clock Polarity Select */
123 #define CTR_RSCKIZ_SCK	BIT(29)   /*   Must match CTR_TSCKIZ_SCK */
124 #define CTR_RSCKIZ_POL_SHIFT 28   /*   Receive Clock Polarity */
125 #define CTR_TEDG_SHIFT	     27   /* Transmit Timing (1 = falling edge) */
126 #define CTR_REDG_SHIFT	     26   /* Receive Timing (1 = falling edge) */
127 #define CTR_TXDIZ_MASK	GENMASK(23, 22) /* Pin Output When TX is Disabled */
128 #define CTR_TXDIZ_LOW	(0 << 22) /*   0 */
129 #define CTR_TXDIZ_HIGH	(1 << 22) /*   1 */
130 #define CTR_TXDIZ_HIZ	(2 << 22) /*   High-impedance */
131 #define CTR_TSCKE	BIT(15)   /* Transmit Serial Clock Output Enable */
132 #define CTR_TFSE	BIT(14)   /* Transmit Frame Sync Signal Output Enable */
133 #define CTR_TXE		BIT(9)    /* Transmit Enable */
134 #define CTR_RXE		BIT(8)    /* Receive Enable */
135 #define CTR_TXRST	BIT(1)    /* Transmit Reset */
136 #define CTR_RXRST	BIT(0)    /* Receive Reset */
137 
138 /* FCTR */
139 #define FCTR_TFWM_MASK	GENMASK(31, 29) /* Transmit FIFO Watermark */
140 #define FCTR_TFWM_64	(0 << 29) /*  Transfer Request when 64 empty stages */
141 #define FCTR_TFWM_32	(1 << 29) /*  Transfer Request when 32 empty stages */
142 #define FCTR_TFWM_24	(2 << 29) /*  Transfer Request when 24 empty stages */
143 #define FCTR_TFWM_16	(3 << 29) /*  Transfer Request when 16 empty stages */
144 #define FCTR_TFWM_12	(4 << 29) /*  Transfer Request when 12 empty stages */
145 #define FCTR_TFWM_8	(5 << 29) /*  Transfer Request when 8 empty stages */
146 #define FCTR_TFWM_4	(6 << 29) /*  Transfer Request when 4 empty stages */
147 #define FCTR_TFWM_1	(7 << 29) /*  Transfer Request when 1 empty stage */
148 #define FCTR_TFUA_MASK	GENMASK(26, 20) /* Transmit FIFO Usable Area */
149 #define FCTR_TFUA_SHIFT	20
150 #define FCTR_TFUA(i)	((i) << FCTR_TFUA_SHIFT)
151 #define FCTR_RFWM_MASK	GENMASK(15, 13) /* Receive FIFO Watermark */
152 #define FCTR_RFWM_1	(0 << 13) /*  Transfer Request when 1 valid stages */
153 #define FCTR_RFWM_4	(1 << 13) /*  Transfer Request when 4 valid stages */
154 #define FCTR_RFWM_8	(2 << 13) /*  Transfer Request when 8 valid stages */
155 #define FCTR_RFWM_16	(3 << 13) /*  Transfer Request when 16 valid stages */
156 #define FCTR_RFWM_32	(4 << 13) /*  Transfer Request when 32 valid stages */
157 #define FCTR_RFWM_64	(5 << 13) /*  Transfer Request when 64 valid stages */
158 #define FCTR_RFWM_128	(6 << 13) /*  Transfer Request when 128 valid stages */
159 #define FCTR_RFWM_256	(7 << 13) /*  Transfer Request when 256 valid stages */
160 #define FCTR_RFUA_MASK	GENMASK(12, 4) /* Receive FIFO Usable Area (0x40 = full) */
161 #define FCTR_RFUA_SHIFT	4
162 #define FCTR_RFUA(i)	((i) << FCTR_RFUA_SHIFT)
163 
164 /* STR */
165 #define STR_TFEMP	BIT(29) /* Transmit FIFO Empty */
166 #define STR_TDREQ	BIT(28) /* Transmit Data Transfer Request */
167 #define STR_TEOF	BIT(23) /* Frame Transmission End */
168 #define STR_TFSERR	BIT(21) /* Transmit Frame Synchronization Error */
169 #define STR_TFOVF	BIT(20) /* Transmit FIFO Overflow */
170 #define STR_TFUDF	BIT(19) /* Transmit FIFO Underflow */
171 #define STR_RFFUL	BIT(13) /* Receive FIFO Full */
172 #define STR_RDREQ	BIT(12) /* Receive Data Transfer Request */
173 #define STR_REOF	BIT(7)  /* Frame Reception End */
174 #define STR_RFSERR	BIT(5)  /* Receive Frame Synchronization Error */
175 #define STR_RFUDF	BIT(4)  /* Receive FIFO Underflow */
176 #define STR_RFOVF	BIT(3)  /* Receive FIFO Overflow */
177 
178 /* IER */
179 #define IER_TDMAE	BIT(31) /* Transmit Data DMA Transfer Req. Enable */
180 #define IER_TFEMPE	BIT(29) /* Transmit FIFO Empty Enable */
181 #define IER_TDREQE	BIT(28) /* Transmit Data Transfer Request Enable */
182 #define IER_TEOFE	BIT(23) /* Frame Transmission End Enable */
183 #define IER_TFSERRE	BIT(21) /* Transmit Frame Sync Error Enable */
184 #define IER_TFOVFE	BIT(20) /* Transmit FIFO Overflow Enable */
185 #define IER_TFUDFE	BIT(19) /* Transmit FIFO Underflow Enable */
186 #define IER_RDMAE	BIT(15) /* Receive Data DMA Transfer Req. Enable */
187 #define IER_RFFULE	BIT(13) /* Receive FIFO Full Enable */
188 #define IER_RDREQE	BIT(12) /* Receive Data Transfer Request Enable */
189 #define IER_REOFE	BIT(7)  /* Frame Reception End Enable */
190 #define IER_RFSERRE	BIT(5)  /* Receive Frame Sync Error Enable */
191 #define IER_RFUDFE	BIT(4)  /* Receive FIFO Underflow Enable */
192 #define IER_RFOVFE	BIT(3)  /* Receive FIFO Overflow Enable */
193 
194 
195 static u32 sh_msiof_read(struct sh_msiof_spi_priv *p, int reg_offs)
196 {
197 	switch (reg_offs) {
198 	case TSCR:
199 	case RSCR:
200 		return ioread16(p->mapbase + reg_offs);
201 	default:
202 		return ioread32(p->mapbase + reg_offs);
203 	}
204 }
205 
206 static void sh_msiof_write(struct sh_msiof_spi_priv *p, int reg_offs,
207 			   u32 value)
208 {
209 	switch (reg_offs) {
210 	case TSCR:
211 	case RSCR:
212 		iowrite16(value, p->mapbase + reg_offs);
213 		break;
214 	default:
215 		iowrite32(value, p->mapbase + reg_offs);
216 		break;
217 	}
218 }
219 
220 static int sh_msiof_modify_ctr_wait(struct sh_msiof_spi_priv *p,
221 				    u32 clr, u32 set)
222 {
223 	u32 mask = clr | set;
224 	u32 data;
225 
226 	data = sh_msiof_read(p, CTR);
227 	data &= ~clr;
228 	data |= set;
229 	sh_msiof_write(p, CTR, data);
230 
231 	return readl_poll_timeout_atomic(p->mapbase + CTR, data,
232 					 (data & mask) == set, 10, 1000);
233 }
234 
235 static irqreturn_t sh_msiof_spi_irq(int irq, void *data)
236 {
237 	struct sh_msiof_spi_priv *p = data;
238 
239 	/* just disable the interrupt and wake up */
240 	sh_msiof_write(p, IER, 0);
241 	complete(&p->done);
242 
243 	return IRQ_HANDLED;
244 }
245 
246 static void sh_msiof_spi_reset_regs(struct sh_msiof_spi_priv *p)
247 {
248 	u32 mask = CTR_TXRST | CTR_RXRST;
249 	u32 data;
250 
251 	data = sh_msiof_read(p, CTR);
252 	data |= mask;
253 	sh_msiof_write(p, CTR, data);
254 
255 	readl_poll_timeout_atomic(p->mapbase + CTR, data, !(data & mask), 1,
256 				  100);
257 }
258 
259 static const u32 sh_msiof_spi_div_array[] = {
260 	SCR_BRDV_DIV_1, SCR_BRDV_DIV_2,	 SCR_BRDV_DIV_4,
261 	SCR_BRDV_DIV_8,	SCR_BRDV_DIV_16, SCR_BRDV_DIV_32,
262 };
263 
264 static void sh_msiof_spi_set_clk_regs(struct sh_msiof_spi_priv *p,
265 				      unsigned long parent_rate, u32 spi_hz)
266 {
267 	unsigned long div;
268 	u32 brps, scr;
269 	unsigned int div_pow = p->min_div_pow;
270 
271 	if (!spi_hz || !parent_rate) {
272 		WARN(1, "Invalid clock rate parameters %lu and %u\n",
273 		     parent_rate, spi_hz);
274 		return;
275 	}
276 
277 	div = DIV_ROUND_UP(parent_rate, spi_hz);
278 	if (div <= 1024) {
279 		/* SCR_BRDV_DIV_1 is valid only if BRPS is x 1/1 or x 1/2 */
280 		if (!div_pow && div <= 32 && div > 2)
281 			div_pow = 1;
282 
283 		if (div_pow)
284 			brps = (div + 1) >> div_pow;
285 		else
286 			brps = div;
287 
288 		for (; brps > 32; div_pow++)
289 			brps = (brps + 1) >> 1;
290 	} else {
291 		/* Set transfer rate composite divisor to 2^5 * 32 = 1024 */
292 		dev_err(&p->pdev->dev,
293 			"Requested SPI transfer rate %d is too low\n", spi_hz);
294 		div_pow = 5;
295 		brps = 32;
296 	}
297 
298 	scr = sh_msiof_spi_div_array[div_pow] | SCR_BRPS(brps);
299 	sh_msiof_write(p, TSCR, scr);
300 	if (!(p->ctlr->flags & SPI_CONTROLLER_MUST_TX))
301 		sh_msiof_write(p, RSCR, scr);
302 }
303 
304 static u32 sh_msiof_get_delay_bit(u32 dtdl_or_syncdl)
305 {
306 	/*
307 	 * DTDL/SYNCDL bit	: p->info->dtdl or p->info->syncdl
308 	 * b'000		: 0
309 	 * b'001		: 100
310 	 * b'010		: 200
311 	 * b'011 (SYNCDL only)	: 300
312 	 * b'101		: 50
313 	 * b'110		: 150
314 	 */
315 	if (dtdl_or_syncdl % 100)
316 		return dtdl_or_syncdl / 100 + 5;
317 	else
318 		return dtdl_or_syncdl / 100;
319 }
320 
321 static u32 sh_msiof_spi_get_dtdl_and_syncdl(struct sh_msiof_spi_priv *p)
322 {
323 	u32 val;
324 
325 	if (!p->info)
326 		return 0;
327 
328 	/* check if DTDL and SYNCDL is allowed value */
329 	if (p->info->dtdl > 200 || p->info->syncdl > 300) {
330 		dev_warn(&p->pdev->dev, "DTDL or SYNCDL is too large\n");
331 		return 0;
332 	}
333 
334 	/* check if the sum of DTDL and SYNCDL becomes an integer value  */
335 	if ((p->info->dtdl + p->info->syncdl) % 100) {
336 		dev_warn(&p->pdev->dev, "the sum of DTDL/SYNCDL is not good\n");
337 		return 0;
338 	}
339 
340 	val = sh_msiof_get_delay_bit(p->info->dtdl) << MDR1_DTDL_SHIFT;
341 	val |= sh_msiof_get_delay_bit(p->info->syncdl) << MDR1_SYNCDL_SHIFT;
342 
343 	return val;
344 }
345 
346 static void sh_msiof_spi_set_pin_regs(struct sh_msiof_spi_priv *p, u32 ss,
347 				      u32 cpol, u32 cpha,
348 				      u32 tx_hi_z, u32 lsb_first, u32 cs_high)
349 {
350 	u32 tmp;
351 	int edge;
352 
353 	/*
354 	 * CPOL CPHA     TSCKIZ RSCKIZ TEDG REDG
355 	 *    0    0         10     10    1    1
356 	 *    0    1         10     10    0    0
357 	 *    1    0         11     11    0    0
358 	 *    1    1         11     11    1    1
359 	 */
360 	tmp = MDR1_SYNCMD_SPI | 1 << MDR1_FLD_SHIFT | MDR1_XXSTP;
361 	tmp |= !cs_high << MDR1_SYNCAC_SHIFT;
362 	tmp |= lsb_first << MDR1_BITLSB_SHIFT;
363 	tmp |= sh_msiof_spi_get_dtdl_and_syncdl(p);
364 	if (spi_controller_is_slave(p->ctlr)) {
365 		sh_msiof_write(p, TMDR1, tmp | TMDR1_PCON);
366 	} else {
367 		sh_msiof_write(p, TMDR1,
368 			       tmp | MDR1_TRMD | TMDR1_PCON |
369 			       (ss < MAX_SS ? ss : 0) << TMDR1_SYNCCH_SHIFT);
370 	}
371 	if (p->ctlr->flags & SPI_CONTROLLER_MUST_TX) {
372 		/* These bits are reserved if RX needs TX */
373 		tmp &= ~0x0000ffff;
374 	}
375 	sh_msiof_write(p, RMDR1, tmp);
376 
377 	tmp = 0;
378 	tmp |= CTR_TSCKIZ_SCK | cpol << CTR_TSCKIZ_POL_SHIFT;
379 	tmp |= CTR_RSCKIZ_SCK | cpol << CTR_RSCKIZ_POL_SHIFT;
380 
381 	edge = cpol ^ !cpha;
382 
383 	tmp |= edge << CTR_TEDG_SHIFT;
384 	tmp |= edge << CTR_REDG_SHIFT;
385 	tmp |= tx_hi_z ? CTR_TXDIZ_HIZ : CTR_TXDIZ_LOW;
386 	sh_msiof_write(p, CTR, tmp);
387 }
388 
389 static void sh_msiof_spi_set_mode_regs(struct sh_msiof_spi_priv *p,
390 				       const void *tx_buf, void *rx_buf,
391 				       u32 bits, u32 words)
392 {
393 	u32 dr2 = MDR2_BITLEN1(bits) | MDR2_WDLEN1(words);
394 
395 	if (tx_buf || (p->ctlr->flags & SPI_CONTROLLER_MUST_TX))
396 		sh_msiof_write(p, TMDR2, dr2);
397 	else
398 		sh_msiof_write(p, TMDR2, dr2 | MDR2_GRPMASK1);
399 
400 	if (rx_buf)
401 		sh_msiof_write(p, RMDR2, dr2);
402 }
403 
404 static void sh_msiof_reset_str(struct sh_msiof_spi_priv *p)
405 {
406 	sh_msiof_write(p, STR,
407 		       sh_msiof_read(p, STR) & ~(STR_TDREQ | STR_RDREQ));
408 }
409 
410 static void sh_msiof_spi_write_fifo_8(struct sh_msiof_spi_priv *p,
411 				      const void *tx_buf, int words, int fs)
412 {
413 	const u8 *buf_8 = tx_buf;
414 	int k;
415 
416 	for (k = 0; k < words; k++)
417 		sh_msiof_write(p, TFDR, buf_8[k] << fs);
418 }
419 
420 static void sh_msiof_spi_write_fifo_16(struct sh_msiof_spi_priv *p,
421 				       const void *tx_buf, int words, int fs)
422 {
423 	const u16 *buf_16 = tx_buf;
424 	int k;
425 
426 	for (k = 0; k < words; k++)
427 		sh_msiof_write(p, TFDR, buf_16[k] << fs);
428 }
429 
430 static void sh_msiof_spi_write_fifo_16u(struct sh_msiof_spi_priv *p,
431 					const void *tx_buf, int words, int fs)
432 {
433 	const u16 *buf_16 = tx_buf;
434 	int k;
435 
436 	for (k = 0; k < words; k++)
437 		sh_msiof_write(p, TFDR, get_unaligned(&buf_16[k]) << fs);
438 }
439 
440 static void sh_msiof_spi_write_fifo_32(struct sh_msiof_spi_priv *p,
441 				       const void *tx_buf, int words, int fs)
442 {
443 	const u32 *buf_32 = tx_buf;
444 	int k;
445 
446 	for (k = 0; k < words; k++)
447 		sh_msiof_write(p, TFDR, buf_32[k] << fs);
448 }
449 
450 static void sh_msiof_spi_write_fifo_32u(struct sh_msiof_spi_priv *p,
451 					const void *tx_buf, int words, int fs)
452 {
453 	const u32 *buf_32 = tx_buf;
454 	int k;
455 
456 	for (k = 0; k < words; k++)
457 		sh_msiof_write(p, TFDR, get_unaligned(&buf_32[k]) << fs);
458 }
459 
460 static void sh_msiof_spi_write_fifo_s32(struct sh_msiof_spi_priv *p,
461 					const void *tx_buf, int words, int fs)
462 {
463 	const u32 *buf_32 = tx_buf;
464 	int k;
465 
466 	for (k = 0; k < words; k++)
467 		sh_msiof_write(p, TFDR, swab32(buf_32[k] << fs));
468 }
469 
470 static void sh_msiof_spi_write_fifo_s32u(struct sh_msiof_spi_priv *p,
471 					 const void *tx_buf, int words, int fs)
472 {
473 	const u32 *buf_32 = tx_buf;
474 	int k;
475 
476 	for (k = 0; k < words; k++)
477 		sh_msiof_write(p, TFDR, swab32(get_unaligned(&buf_32[k]) << fs));
478 }
479 
480 static void sh_msiof_spi_read_fifo_8(struct sh_msiof_spi_priv *p,
481 				     void *rx_buf, int words, int fs)
482 {
483 	u8 *buf_8 = rx_buf;
484 	int k;
485 
486 	for (k = 0; k < words; k++)
487 		buf_8[k] = sh_msiof_read(p, RFDR) >> fs;
488 }
489 
490 static void sh_msiof_spi_read_fifo_16(struct sh_msiof_spi_priv *p,
491 				      void *rx_buf, int words, int fs)
492 {
493 	u16 *buf_16 = rx_buf;
494 	int k;
495 
496 	for (k = 0; k < words; k++)
497 		buf_16[k] = sh_msiof_read(p, RFDR) >> fs;
498 }
499 
500 static void sh_msiof_spi_read_fifo_16u(struct sh_msiof_spi_priv *p,
501 				       void *rx_buf, int words, int fs)
502 {
503 	u16 *buf_16 = rx_buf;
504 	int k;
505 
506 	for (k = 0; k < words; k++)
507 		put_unaligned(sh_msiof_read(p, RFDR) >> fs, &buf_16[k]);
508 }
509 
510 static void sh_msiof_spi_read_fifo_32(struct sh_msiof_spi_priv *p,
511 				      void *rx_buf, int words, int fs)
512 {
513 	u32 *buf_32 = rx_buf;
514 	int k;
515 
516 	for (k = 0; k < words; k++)
517 		buf_32[k] = sh_msiof_read(p, RFDR) >> fs;
518 }
519 
520 static void sh_msiof_spi_read_fifo_32u(struct sh_msiof_spi_priv *p,
521 				       void *rx_buf, int words, int fs)
522 {
523 	u32 *buf_32 = rx_buf;
524 	int k;
525 
526 	for (k = 0; k < words; k++)
527 		put_unaligned(sh_msiof_read(p, RFDR) >> fs, &buf_32[k]);
528 }
529 
530 static void sh_msiof_spi_read_fifo_s32(struct sh_msiof_spi_priv *p,
531 				       void *rx_buf, int words, int fs)
532 {
533 	u32 *buf_32 = rx_buf;
534 	int k;
535 
536 	for (k = 0; k < words; k++)
537 		buf_32[k] = swab32(sh_msiof_read(p, RFDR) >> fs);
538 }
539 
540 static void sh_msiof_spi_read_fifo_s32u(struct sh_msiof_spi_priv *p,
541 				       void *rx_buf, int words, int fs)
542 {
543 	u32 *buf_32 = rx_buf;
544 	int k;
545 
546 	for (k = 0; k < words; k++)
547 		put_unaligned(swab32(sh_msiof_read(p, RFDR) >> fs), &buf_32[k]);
548 }
549 
550 static int sh_msiof_spi_setup(struct spi_device *spi)
551 {
552 	struct sh_msiof_spi_priv *p =
553 		spi_controller_get_devdata(spi->controller);
554 	u32 clr, set, tmp;
555 
556 	if (spi->cs_gpiod || spi_controller_is_slave(p->ctlr))
557 		return 0;
558 
559 	if (p->native_cs_inited &&
560 	    (p->native_cs_high == !!(spi->mode & SPI_CS_HIGH)))
561 		return 0;
562 
563 	/* Configure native chip select mode/polarity early */
564 	clr = MDR1_SYNCMD_MASK;
565 	set = MDR1_SYNCMD_SPI;
566 	if (spi->mode & SPI_CS_HIGH)
567 		clr |= BIT(MDR1_SYNCAC_SHIFT);
568 	else
569 		set |= BIT(MDR1_SYNCAC_SHIFT);
570 	pm_runtime_get_sync(&p->pdev->dev);
571 	tmp = sh_msiof_read(p, TMDR1) & ~clr;
572 	sh_msiof_write(p, TMDR1, tmp | set | MDR1_TRMD | TMDR1_PCON);
573 	tmp = sh_msiof_read(p, RMDR1) & ~clr;
574 	sh_msiof_write(p, RMDR1, tmp | set);
575 	pm_runtime_put(&p->pdev->dev);
576 	p->native_cs_high = spi->mode & SPI_CS_HIGH;
577 	p->native_cs_inited = true;
578 	return 0;
579 }
580 
581 static int sh_msiof_prepare_message(struct spi_controller *ctlr,
582 				    struct spi_message *msg)
583 {
584 	struct sh_msiof_spi_priv *p = spi_controller_get_devdata(ctlr);
585 	const struct spi_device *spi = msg->spi;
586 	u32 ss, cs_high;
587 
588 	/* Configure pins before asserting CS */
589 	if (spi->cs_gpiod) {
590 		ss = p->unused_ss;
591 		cs_high = p->native_cs_high;
592 	} else {
593 		ss = spi->chip_select;
594 		cs_high = !!(spi->mode & SPI_CS_HIGH);
595 	}
596 	sh_msiof_spi_set_pin_regs(p, ss, !!(spi->mode & SPI_CPOL),
597 				  !!(spi->mode & SPI_CPHA),
598 				  !!(spi->mode & SPI_3WIRE),
599 				  !!(spi->mode & SPI_LSB_FIRST), cs_high);
600 	return 0;
601 }
602 
603 static int sh_msiof_spi_start(struct sh_msiof_spi_priv *p, void *rx_buf)
604 {
605 	bool slave = spi_controller_is_slave(p->ctlr);
606 	int ret = 0;
607 
608 	/* setup clock and rx/tx signals */
609 	if (!slave)
610 		ret = sh_msiof_modify_ctr_wait(p, 0, CTR_TSCKE);
611 	if (rx_buf && !ret)
612 		ret = sh_msiof_modify_ctr_wait(p, 0, CTR_RXE);
613 	if (!ret)
614 		ret = sh_msiof_modify_ctr_wait(p, 0, CTR_TXE);
615 
616 	/* start by setting frame bit */
617 	if (!ret && !slave)
618 		ret = sh_msiof_modify_ctr_wait(p, 0, CTR_TFSE);
619 
620 	return ret;
621 }
622 
623 static int sh_msiof_spi_stop(struct sh_msiof_spi_priv *p, void *rx_buf)
624 {
625 	bool slave = spi_controller_is_slave(p->ctlr);
626 	int ret = 0;
627 
628 	/* shut down frame, rx/tx and clock signals */
629 	if (!slave)
630 		ret = sh_msiof_modify_ctr_wait(p, CTR_TFSE, 0);
631 	if (!ret)
632 		ret = sh_msiof_modify_ctr_wait(p, CTR_TXE, 0);
633 	if (rx_buf && !ret)
634 		ret = sh_msiof_modify_ctr_wait(p, CTR_RXE, 0);
635 	if (!ret && !slave)
636 		ret = sh_msiof_modify_ctr_wait(p, CTR_TSCKE, 0);
637 
638 	return ret;
639 }
640 
641 static int sh_msiof_slave_abort(struct spi_controller *ctlr)
642 {
643 	struct sh_msiof_spi_priv *p = spi_controller_get_devdata(ctlr);
644 
645 	p->slave_aborted = true;
646 	complete(&p->done);
647 	complete(&p->done_txdma);
648 	return 0;
649 }
650 
651 static int sh_msiof_wait_for_completion(struct sh_msiof_spi_priv *p,
652 					struct completion *x)
653 {
654 	if (spi_controller_is_slave(p->ctlr)) {
655 		if (wait_for_completion_interruptible(x) ||
656 		    p->slave_aborted) {
657 			dev_dbg(&p->pdev->dev, "interrupted\n");
658 			return -EINTR;
659 		}
660 	} else {
661 		if (!wait_for_completion_timeout(x, HZ)) {
662 			dev_err(&p->pdev->dev, "timeout\n");
663 			return -ETIMEDOUT;
664 		}
665 	}
666 
667 	return 0;
668 }
669 
670 static int sh_msiof_spi_txrx_once(struct sh_msiof_spi_priv *p,
671 				  void (*tx_fifo)(struct sh_msiof_spi_priv *,
672 						  const void *, int, int),
673 				  void (*rx_fifo)(struct sh_msiof_spi_priv *,
674 						  void *, int, int),
675 				  const void *tx_buf, void *rx_buf,
676 				  int words, int bits)
677 {
678 	int fifo_shift;
679 	int ret;
680 
681 	/* limit maximum word transfer to rx/tx fifo size */
682 	if (tx_buf)
683 		words = min_t(int, words, p->tx_fifo_size);
684 	if (rx_buf)
685 		words = min_t(int, words, p->rx_fifo_size);
686 
687 	/* the fifo contents need shifting */
688 	fifo_shift = 32 - bits;
689 
690 	/* default FIFO watermarks for PIO */
691 	sh_msiof_write(p, FCTR, 0);
692 
693 	/* setup msiof transfer mode registers */
694 	sh_msiof_spi_set_mode_regs(p, tx_buf, rx_buf, bits, words);
695 	sh_msiof_write(p, IER, IER_TEOFE | IER_REOFE);
696 
697 	/* write tx fifo */
698 	if (tx_buf)
699 		tx_fifo(p, tx_buf, words, fifo_shift);
700 
701 	reinit_completion(&p->done);
702 	p->slave_aborted = false;
703 
704 	ret = sh_msiof_spi_start(p, rx_buf);
705 	if (ret) {
706 		dev_err(&p->pdev->dev, "failed to start hardware\n");
707 		goto stop_ier;
708 	}
709 
710 	/* wait for tx fifo to be emptied / rx fifo to be filled */
711 	ret = sh_msiof_wait_for_completion(p, &p->done);
712 	if (ret)
713 		goto stop_reset;
714 
715 	/* read rx fifo */
716 	if (rx_buf)
717 		rx_fifo(p, rx_buf, words, fifo_shift);
718 
719 	/* clear status bits */
720 	sh_msiof_reset_str(p);
721 
722 	ret = sh_msiof_spi_stop(p, rx_buf);
723 	if (ret) {
724 		dev_err(&p->pdev->dev, "failed to shut down hardware\n");
725 		return ret;
726 	}
727 
728 	return words;
729 
730 stop_reset:
731 	sh_msiof_reset_str(p);
732 	sh_msiof_spi_stop(p, rx_buf);
733 stop_ier:
734 	sh_msiof_write(p, IER, 0);
735 	return ret;
736 }
737 
738 static void sh_msiof_dma_complete(void *arg)
739 {
740 	complete(arg);
741 }
742 
743 static int sh_msiof_dma_once(struct sh_msiof_spi_priv *p, const void *tx,
744 			     void *rx, unsigned int len)
745 {
746 	u32 ier_bits = 0;
747 	struct dma_async_tx_descriptor *desc_tx = NULL, *desc_rx = NULL;
748 	dma_cookie_t cookie;
749 	int ret;
750 
751 	/* First prepare and submit the DMA request(s), as this may fail */
752 	if (rx) {
753 		ier_bits |= IER_RDREQE | IER_RDMAE;
754 		desc_rx = dmaengine_prep_slave_single(p->ctlr->dma_rx,
755 					p->rx_dma_addr, len, DMA_DEV_TO_MEM,
756 					DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
757 		if (!desc_rx)
758 			return -EAGAIN;
759 
760 		desc_rx->callback = sh_msiof_dma_complete;
761 		desc_rx->callback_param = &p->done;
762 		cookie = dmaengine_submit(desc_rx);
763 		if (dma_submit_error(cookie))
764 			return cookie;
765 	}
766 
767 	if (tx) {
768 		ier_bits |= IER_TDREQE | IER_TDMAE;
769 		dma_sync_single_for_device(p->ctlr->dma_tx->device->dev,
770 					   p->tx_dma_addr, len, DMA_TO_DEVICE);
771 		desc_tx = dmaengine_prep_slave_single(p->ctlr->dma_tx,
772 					p->tx_dma_addr, len, DMA_MEM_TO_DEV,
773 					DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
774 		if (!desc_tx) {
775 			ret = -EAGAIN;
776 			goto no_dma_tx;
777 		}
778 
779 		desc_tx->callback = sh_msiof_dma_complete;
780 		desc_tx->callback_param = &p->done_txdma;
781 		cookie = dmaengine_submit(desc_tx);
782 		if (dma_submit_error(cookie)) {
783 			ret = cookie;
784 			goto no_dma_tx;
785 		}
786 	}
787 
788 	/* 1 stage FIFO watermarks for DMA */
789 	sh_msiof_write(p, FCTR, FCTR_TFWM_1 | FCTR_RFWM_1);
790 
791 	/* setup msiof transfer mode registers (32-bit words) */
792 	sh_msiof_spi_set_mode_regs(p, tx, rx, 32, len / 4);
793 
794 	sh_msiof_write(p, IER, ier_bits);
795 
796 	reinit_completion(&p->done);
797 	if (tx)
798 		reinit_completion(&p->done_txdma);
799 	p->slave_aborted = false;
800 
801 	/* Now start DMA */
802 	if (rx)
803 		dma_async_issue_pending(p->ctlr->dma_rx);
804 	if (tx)
805 		dma_async_issue_pending(p->ctlr->dma_tx);
806 
807 	ret = sh_msiof_spi_start(p, rx);
808 	if (ret) {
809 		dev_err(&p->pdev->dev, "failed to start hardware\n");
810 		goto stop_dma;
811 	}
812 
813 	if (tx) {
814 		/* wait for tx DMA completion */
815 		ret = sh_msiof_wait_for_completion(p, &p->done_txdma);
816 		if (ret)
817 			goto stop_reset;
818 	}
819 
820 	if (rx) {
821 		/* wait for rx DMA completion */
822 		ret = sh_msiof_wait_for_completion(p, &p->done);
823 		if (ret)
824 			goto stop_reset;
825 
826 		sh_msiof_write(p, IER, 0);
827 	} else {
828 		/* wait for tx fifo to be emptied */
829 		sh_msiof_write(p, IER, IER_TEOFE);
830 		ret = sh_msiof_wait_for_completion(p, &p->done);
831 		if (ret)
832 			goto stop_reset;
833 	}
834 
835 	/* clear status bits */
836 	sh_msiof_reset_str(p);
837 
838 	ret = sh_msiof_spi_stop(p, rx);
839 	if (ret) {
840 		dev_err(&p->pdev->dev, "failed to shut down hardware\n");
841 		return ret;
842 	}
843 
844 	if (rx)
845 		dma_sync_single_for_cpu(p->ctlr->dma_rx->device->dev,
846 					p->rx_dma_addr, len, DMA_FROM_DEVICE);
847 
848 	return 0;
849 
850 stop_reset:
851 	sh_msiof_reset_str(p);
852 	sh_msiof_spi_stop(p, rx);
853 stop_dma:
854 	if (tx)
855 		dmaengine_terminate_all(p->ctlr->dma_tx);
856 no_dma_tx:
857 	if (rx)
858 		dmaengine_terminate_all(p->ctlr->dma_rx);
859 	sh_msiof_write(p, IER, 0);
860 	return ret;
861 }
862 
863 static void copy_bswap32(u32 *dst, const u32 *src, unsigned int words)
864 {
865 	/* src or dst can be unaligned, but not both */
866 	if ((unsigned long)src & 3) {
867 		while (words--) {
868 			*dst++ = swab32(get_unaligned(src));
869 			src++;
870 		}
871 	} else if ((unsigned long)dst & 3) {
872 		while (words--) {
873 			put_unaligned(swab32(*src++), dst);
874 			dst++;
875 		}
876 	} else {
877 		while (words--)
878 			*dst++ = swab32(*src++);
879 	}
880 }
881 
882 static void copy_wswap32(u32 *dst, const u32 *src, unsigned int words)
883 {
884 	/* src or dst can be unaligned, but not both */
885 	if ((unsigned long)src & 3) {
886 		while (words--) {
887 			*dst++ = swahw32(get_unaligned(src));
888 			src++;
889 		}
890 	} else if ((unsigned long)dst & 3) {
891 		while (words--) {
892 			put_unaligned(swahw32(*src++), dst);
893 			dst++;
894 		}
895 	} else {
896 		while (words--)
897 			*dst++ = swahw32(*src++);
898 	}
899 }
900 
901 static void copy_plain32(u32 *dst, const u32 *src, unsigned int words)
902 {
903 	memcpy(dst, src, words * 4);
904 }
905 
906 static int sh_msiof_transfer_one(struct spi_controller *ctlr,
907 				 struct spi_device *spi,
908 				 struct spi_transfer *t)
909 {
910 	struct sh_msiof_spi_priv *p = spi_controller_get_devdata(ctlr);
911 	void (*copy32)(u32 *, const u32 *, unsigned int);
912 	void (*tx_fifo)(struct sh_msiof_spi_priv *, const void *, int, int);
913 	void (*rx_fifo)(struct sh_msiof_spi_priv *, void *, int, int);
914 	const void *tx_buf = t->tx_buf;
915 	void *rx_buf = t->rx_buf;
916 	unsigned int len = t->len;
917 	unsigned int bits = t->bits_per_word;
918 	unsigned int bytes_per_word;
919 	unsigned int words;
920 	int n;
921 	bool swab;
922 	int ret;
923 
924 	/* reset registers */
925 	sh_msiof_spi_reset_regs(p);
926 
927 	/* setup clocks (clock already enabled in chipselect()) */
928 	if (!spi_controller_is_slave(p->ctlr))
929 		sh_msiof_spi_set_clk_regs(p, clk_get_rate(p->clk), t->speed_hz);
930 
931 	while (ctlr->dma_tx && len > 15) {
932 		/*
933 		 *  DMA supports 32-bit words only, hence pack 8-bit and 16-bit
934 		 *  words, with byte resp. word swapping.
935 		 */
936 		unsigned int l = 0;
937 
938 		if (tx_buf)
939 			l = min(round_down(len, 4), p->tx_fifo_size * 4);
940 		if (rx_buf)
941 			l = min(round_down(len, 4), p->rx_fifo_size * 4);
942 
943 		if (bits <= 8) {
944 			copy32 = copy_bswap32;
945 		} else if (bits <= 16) {
946 			copy32 = copy_wswap32;
947 		} else {
948 			copy32 = copy_plain32;
949 		}
950 
951 		if (tx_buf)
952 			copy32(p->tx_dma_page, tx_buf, l / 4);
953 
954 		ret = sh_msiof_dma_once(p, tx_buf, rx_buf, l);
955 		if (ret == -EAGAIN) {
956 			dev_warn_once(&p->pdev->dev,
957 				"DMA not available, falling back to PIO\n");
958 			break;
959 		}
960 		if (ret)
961 			return ret;
962 
963 		if (rx_buf) {
964 			copy32(rx_buf, p->rx_dma_page, l / 4);
965 			rx_buf += l;
966 		}
967 		if (tx_buf)
968 			tx_buf += l;
969 
970 		len -= l;
971 		if (!len)
972 			return 0;
973 	}
974 
975 	if (bits <= 8 && len > 15) {
976 		bits = 32;
977 		swab = true;
978 	} else {
979 		swab = false;
980 	}
981 
982 	/* setup bytes per word and fifo read/write functions */
983 	if (bits <= 8) {
984 		bytes_per_word = 1;
985 		tx_fifo = sh_msiof_spi_write_fifo_8;
986 		rx_fifo = sh_msiof_spi_read_fifo_8;
987 	} else if (bits <= 16) {
988 		bytes_per_word = 2;
989 		if ((unsigned long)tx_buf & 0x01)
990 			tx_fifo = sh_msiof_spi_write_fifo_16u;
991 		else
992 			tx_fifo = sh_msiof_spi_write_fifo_16;
993 
994 		if ((unsigned long)rx_buf & 0x01)
995 			rx_fifo = sh_msiof_spi_read_fifo_16u;
996 		else
997 			rx_fifo = sh_msiof_spi_read_fifo_16;
998 	} else if (swab) {
999 		bytes_per_word = 4;
1000 		if ((unsigned long)tx_buf & 0x03)
1001 			tx_fifo = sh_msiof_spi_write_fifo_s32u;
1002 		else
1003 			tx_fifo = sh_msiof_spi_write_fifo_s32;
1004 
1005 		if ((unsigned long)rx_buf & 0x03)
1006 			rx_fifo = sh_msiof_spi_read_fifo_s32u;
1007 		else
1008 			rx_fifo = sh_msiof_spi_read_fifo_s32;
1009 	} else {
1010 		bytes_per_word = 4;
1011 		if ((unsigned long)tx_buf & 0x03)
1012 			tx_fifo = sh_msiof_spi_write_fifo_32u;
1013 		else
1014 			tx_fifo = sh_msiof_spi_write_fifo_32;
1015 
1016 		if ((unsigned long)rx_buf & 0x03)
1017 			rx_fifo = sh_msiof_spi_read_fifo_32u;
1018 		else
1019 			rx_fifo = sh_msiof_spi_read_fifo_32;
1020 	}
1021 
1022 	/* transfer in fifo sized chunks */
1023 	words = len / bytes_per_word;
1024 
1025 	while (words > 0) {
1026 		n = sh_msiof_spi_txrx_once(p, tx_fifo, rx_fifo, tx_buf, rx_buf,
1027 					   words, bits);
1028 		if (n < 0)
1029 			return n;
1030 
1031 		if (tx_buf)
1032 			tx_buf += n * bytes_per_word;
1033 		if (rx_buf)
1034 			rx_buf += n * bytes_per_word;
1035 		words -= n;
1036 
1037 		if (words == 0 && (len % bytes_per_word)) {
1038 			words = len % bytes_per_word;
1039 			bits = t->bits_per_word;
1040 			bytes_per_word = 1;
1041 			tx_fifo = sh_msiof_spi_write_fifo_8;
1042 			rx_fifo = sh_msiof_spi_read_fifo_8;
1043 		}
1044 	}
1045 
1046 	return 0;
1047 }
1048 
1049 static const struct sh_msiof_chipdata sh_data = {
1050 	.bits_per_word_mask = SPI_BPW_RANGE_MASK(8, 32),
1051 	.tx_fifo_size = 64,
1052 	.rx_fifo_size = 64,
1053 	.ctlr_flags = 0,
1054 	.min_div_pow = 0,
1055 };
1056 
1057 static const struct sh_msiof_chipdata rcar_gen2_data = {
1058 	.bits_per_word_mask = SPI_BPW_MASK(8) | SPI_BPW_MASK(16) |
1059 			      SPI_BPW_MASK(24) | SPI_BPW_MASK(32),
1060 	.tx_fifo_size = 64,
1061 	.rx_fifo_size = 64,
1062 	.ctlr_flags = SPI_CONTROLLER_MUST_TX,
1063 	.min_div_pow = 0,
1064 };
1065 
1066 static const struct sh_msiof_chipdata rcar_gen3_data = {
1067 	.bits_per_word_mask = SPI_BPW_MASK(8) | SPI_BPW_MASK(16) |
1068 			      SPI_BPW_MASK(24) | SPI_BPW_MASK(32),
1069 	.tx_fifo_size = 64,
1070 	.rx_fifo_size = 64,
1071 	.ctlr_flags = SPI_CONTROLLER_MUST_TX,
1072 	.min_div_pow = 1,
1073 };
1074 
1075 static const struct of_device_id sh_msiof_match[] = {
1076 	{ .compatible = "renesas,sh-mobile-msiof", .data = &sh_data },
1077 	{ .compatible = "renesas,msiof-r8a7743",   .data = &rcar_gen2_data },
1078 	{ .compatible = "renesas,msiof-r8a7745",   .data = &rcar_gen2_data },
1079 	{ .compatible = "renesas,msiof-r8a7790",   .data = &rcar_gen2_data },
1080 	{ .compatible = "renesas,msiof-r8a7791",   .data = &rcar_gen2_data },
1081 	{ .compatible = "renesas,msiof-r8a7792",   .data = &rcar_gen2_data },
1082 	{ .compatible = "renesas,msiof-r8a7793",   .data = &rcar_gen2_data },
1083 	{ .compatible = "renesas,msiof-r8a7794",   .data = &rcar_gen2_data },
1084 	{ .compatible = "renesas,rcar-gen2-msiof", .data = &rcar_gen2_data },
1085 	{ .compatible = "renesas,msiof-r8a7796",   .data = &rcar_gen3_data },
1086 	{ .compatible = "renesas,rcar-gen3-msiof", .data = &rcar_gen3_data },
1087 	{ .compatible = "renesas,sh-msiof",        .data = &sh_data }, /* Deprecated */
1088 	{},
1089 };
1090 MODULE_DEVICE_TABLE(of, sh_msiof_match);
1091 
1092 #ifdef CONFIG_OF
1093 static struct sh_msiof_spi_info *sh_msiof_spi_parse_dt(struct device *dev)
1094 {
1095 	struct sh_msiof_spi_info *info;
1096 	struct device_node *np = dev->of_node;
1097 	u32 num_cs = 1;
1098 
1099 	info = devm_kzalloc(dev, sizeof(struct sh_msiof_spi_info), GFP_KERNEL);
1100 	if (!info)
1101 		return NULL;
1102 
1103 	info->mode = of_property_read_bool(np, "spi-slave") ? MSIOF_SPI_SLAVE
1104 							    : MSIOF_SPI_MASTER;
1105 
1106 	/* Parse the MSIOF properties */
1107 	if (info->mode == MSIOF_SPI_MASTER)
1108 		of_property_read_u32(np, "num-cs", &num_cs);
1109 	of_property_read_u32(np, "renesas,tx-fifo-size",
1110 					&info->tx_fifo_override);
1111 	of_property_read_u32(np, "renesas,rx-fifo-size",
1112 					&info->rx_fifo_override);
1113 	of_property_read_u32(np, "renesas,dtdl", &info->dtdl);
1114 	of_property_read_u32(np, "renesas,syncdl", &info->syncdl);
1115 
1116 	info->num_chipselect = num_cs;
1117 
1118 	return info;
1119 }
1120 #else
1121 static struct sh_msiof_spi_info *sh_msiof_spi_parse_dt(struct device *dev)
1122 {
1123 	return NULL;
1124 }
1125 #endif
1126 
1127 static int sh_msiof_get_cs_gpios(struct sh_msiof_spi_priv *p)
1128 {
1129 	struct device *dev = &p->pdev->dev;
1130 	unsigned int used_ss_mask = 0;
1131 	unsigned int cs_gpios = 0;
1132 	unsigned int num_cs, i;
1133 	int ret;
1134 
1135 	ret = gpiod_count(dev, "cs");
1136 	if (ret <= 0)
1137 		return 0;
1138 
1139 	num_cs = max_t(unsigned int, ret, p->ctlr->num_chipselect);
1140 	for (i = 0; i < num_cs; i++) {
1141 		struct gpio_desc *gpiod;
1142 
1143 		gpiod = devm_gpiod_get_index(dev, "cs", i, GPIOD_ASIS);
1144 		if (!IS_ERR(gpiod)) {
1145 			devm_gpiod_put(dev, gpiod);
1146 			cs_gpios++;
1147 			continue;
1148 		}
1149 
1150 		if (PTR_ERR(gpiod) != -ENOENT)
1151 			return PTR_ERR(gpiod);
1152 
1153 		if (i >= MAX_SS) {
1154 			dev_err(dev, "Invalid native chip select %d\n", i);
1155 			return -EINVAL;
1156 		}
1157 		used_ss_mask |= BIT(i);
1158 	}
1159 	p->unused_ss = ffz(used_ss_mask);
1160 	if (cs_gpios && p->unused_ss >= MAX_SS) {
1161 		dev_err(dev, "No unused native chip select available\n");
1162 		return -EINVAL;
1163 	}
1164 	return 0;
1165 }
1166 
1167 static struct dma_chan *sh_msiof_request_dma_chan(struct device *dev,
1168 	enum dma_transfer_direction dir, unsigned int id, dma_addr_t port_addr)
1169 {
1170 	dma_cap_mask_t mask;
1171 	struct dma_chan *chan;
1172 	struct dma_slave_config cfg;
1173 	int ret;
1174 
1175 	dma_cap_zero(mask);
1176 	dma_cap_set(DMA_SLAVE, mask);
1177 
1178 	chan = dma_request_slave_channel_compat(mask, shdma_chan_filter,
1179 				(void *)(unsigned long)id, dev,
1180 				dir == DMA_MEM_TO_DEV ? "tx" : "rx");
1181 	if (!chan) {
1182 		dev_warn(dev, "dma_request_slave_channel_compat failed\n");
1183 		return NULL;
1184 	}
1185 
1186 	memset(&cfg, 0, sizeof(cfg));
1187 	cfg.direction = dir;
1188 	if (dir == DMA_MEM_TO_DEV) {
1189 		cfg.dst_addr = port_addr;
1190 		cfg.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
1191 	} else {
1192 		cfg.src_addr = port_addr;
1193 		cfg.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
1194 	}
1195 
1196 	ret = dmaengine_slave_config(chan, &cfg);
1197 	if (ret) {
1198 		dev_warn(dev, "dmaengine_slave_config failed %d\n", ret);
1199 		dma_release_channel(chan);
1200 		return NULL;
1201 	}
1202 
1203 	return chan;
1204 }
1205 
1206 static int sh_msiof_request_dma(struct sh_msiof_spi_priv *p)
1207 {
1208 	struct platform_device *pdev = p->pdev;
1209 	struct device *dev = &pdev->dev;
1210 	const struct sh_msiof_spi_info *info = p->info;
1211 	unsigned int dma_tx_id, dma_rx_id;
1212 	const struct resource *res;
1213 	struct spi_controller *ctlr;
1214 	struct device *tx_dev, *rx_dev;
1215 
1216 	if (dev->of_node) {
1217 		/* In the OF case we will get the slave IDs from the DT */
1218 		dma_tx_id = 0;
1219 		dma_rx_id = 0;
1220 	} else if (info && info->dma_tx_id && info->dma_rx_id) {
1221 		dma_tx_id = info->dma_tx_id;
1222 		dma_rx_id = info->dma_rx_id;
1223 	} else {
1224 		/* The driver assumes no error */
1225 		return 0;
1226 	}
1227 
1228 	/* The DMA engine uses the second register set, if present */
1229 	res = platform_get_resource(pdev, IORESOURCE_MEM, 1);
1230 	if (!res)
1231 		res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
1232 
1233 	ctlr = p->ctlr;
1234 	ctlr->dma_tx = sh_msiof_request_dma_chan(dev, DMA_MEM_TO_DEV,
1235 						 dma_tx_id, res->start + TFDR);
1236 	if (!ctlr->dma_tx)
1237 		return -ENODEV;
1238 
1239 	ctlr->dma_rx = sh_msiof_request_dma_chan(dev, DMA_DEV_TO_MEM,
1240 						 dma_rx_id, res->start + RFDR);
1241 	if (!ctlr->dma_rx)
1242 		goto free_tx_chan;
1243 
1244 	p->tx_dma_page = (void *)__get_free_page(GFP_KERNEL | GFP_DMA);
1245 	if (!p->tx_dma_page)
1246 		goto free_rx_chan;
1247 
1248 	p->rx_dma_page = (void *)__get_free_page(GFP_KERNEL | GFP_DMA);
1249 	if (!p->rx_dma_page)
1250 		goto free_tx_page;
1251 
1252 	tx_dev = ctlr->dma_tx->device->dev;
1253 	p->tx_dma_addr = dma_map_single(tx_dev, p->tx_dma_page, PAGE_SIZE,
1254 					DMA_TO_DEVICE);
1255 	if (dma_mapping_error(tx_dev, p->tx_dma_addr))
1256 		goto free_rx_page;
1257 
1258 	rx_dev = ctlr->dma_rx->device->dev;
1259 	p->rx_dma_addr = dma_map_single(rx_dev, p->rx_dma_page, PAGE_SIZE,
1260 					DMA_FROM_DEVICE);
1261 	if (dma_mapping_error(rx_dev, p->rx_dma_addr))
1262 		goto unmap_tx_page;
1263 
1264 	dev_info(dev, "DMA available");
1265 	return 0;
1266 
1267 unmap_tx_page:
1268 	dma_unmap_single(tx_dev, p->tx_dma_addr, PAGE_SIZE, DMA_TO_DEVICE);
1269 free_rx_page:
1270 	free_page((unsigned long)p->rx_dma_page);
1271 free_tx_page:
1272 	free_page((unsigned long)p->tx_dma_page);
1273 free_rx_chan:
1274 	dma_release_channel(ctlr->dma_rx);
1275 free_tx_chan:
1276 	dma_release_channel(ctlr->dma_tx);
1277 	ctlr->dma_tx = NULL;
1278 	return -ENODEV;
1279 }
1280 
1281 static void sh_msiof_release_dma(struct sh_msiof_spi_priv *p)
1282 {
1283 	struct spi_controller *ctlr = p->ctlr;
1284 
1285 	if (!ctlr->dma_tx)
1286 		return;
1287 
1288 	dma_unmap_single(ctlr->dma_rx->device->dev, p->rx_dma_addr, PAGE_SIZE,
1289 			 DMA_FROM_DEVICE);
1290 	dma_unmap_single(ctlr->dma_tx->device->dev, p->tx_dma_addr, PAGE_SIZE,
1291 			 DMA_TO_DEVICE);
1292 	free_page((unsigned long)p->rx_dma_page);
1293 	free_page((unsigned long)p->tx_dma_page);
1294 	dma_release_channel(ctlr->dma_rx);
1295 	dma_release_channel(ctlr->dma_tx);
1296 }
1297 
1298 static int sh_msiof_spi_probe(struct platform_device *pdev)
1299 {
1300 	struct resource	*r;
1301 	struct spi_controller *ctlr;
1302 	const struct sh_msiof_chipdata *chipdata;
1303 	struct sh_msiof_spi_info *info;
1304 	struct sh_msiof_spi_priv *p;
1305 	int i;
1306 	int ret;
1307 
1308 	chipdata = of_device_get_match_data(&pdev->dev);
1309 	if (chipdata) {
1310 		info = sh_msiof_spi_parse_dt(&pdev->dev);
1311 	} else {
1312 		chipdata = (const void *)pdev->id_entry->driver_data;
1313 		info = dev_get_platdata(&pdev->dev);
1314 	}
1315 
1316 	if (!info) {
1317 		dev_err(&pdev->dev, "failed to obtain device info\n");
1318 		return -ENXIO;
1319 	}
1320 
1321 	if (info->mode == MSIOF_SPI_SLAVE)
1322 		ctlr = spi_alloc_slave(&pdev->dev,
1323 				       sizeof(struct sh_msiof_spi_priv));
1324 	else
1325 		ctlr = spi_alloc_master(&pdev->dev,
1326 					sizeof(struct sh_msiof_spi_priv));
1327 	if (ctlr == NULL)
1328 		return -ENOMEM;
1329 
1330 	p = spi_controller_get_devdata(ctlr);
1331 
1332 	platform_set_drvdata(pdev, p);
1333 	p->ctlr = ctlr;
1334 	p->info = info;
1335 	p->min_div_pow = chipdata->min_div_pow;
1336 
1337 	init_completion(&p->done);
1338 	init_completion(&p->done_txdma);
1339 
1340 	p->clk = devm_clk_get(&pdev->dev, NULL);
1341 	if (IS_ERR(p->clk)) {
1342 		dev_err(&pdev->dev, "cannot get clock\n");
1343 		ret = PTR_ERR(p->clk);
1344 		goto err1;
1345 	}
1346 
1347 	i = platform_get_irq(pdev, 0);
1348 	if (i < 0) {
1349 		dev_err(&pdev->dev, "cannot get IRQ\n");
1350 		ret = i;
1351 		goto err1;
1352 	}
1353 
1354 	r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
1355 	p->mapbase = devm_ioremap_resource(&pdev->dev, r);
1356 	if (IS_ERR(p->mapbase)) {
1357 		ret = PTR_ERR(p->mapbase);
1358 		goto err1;
1359 	}
1360 
1361 	ret = devm_request_irq(&pdev->dev, i, sh_msiof_spi_irq, 0,
1362 			       dev_name(&pdev->dev), p);
1363 	if (ret) {
1364 		dev_err(&pdev->dev, "unable to request irq\n");
1365 		goto err1;
1366 	}
1367 
1368 	p->pdev = pdev;
1369 	pm_runtime_enable(&pdev->dev);
1370 
1371 	/* Platform data may override FIFO sizes */
1372 	p->tx_fifo_size = chipdata->tx_fifo_size;
1373 	p->rx_fifo_size = chipdata->rx_fifo_size;
1374 	if (p->info->tx_fifo_override)
1375 		p->tx_fifo_size = p->info->tx_fifo_override;
1376 	if (p->info->rx_fifo_override)
1377 		p->rx_fifo_size = p->info->rx_fifo_override;
1378 
1379 	/* Setup GPIO chip selects */
1380 	ctlr->num_chipselect = p->info->num_chipselect;
1381 	ret = sh_msiof_get_cs_gpios(p);
1382 	if (ret)
1383 		goto err1;
1384 
1385 	/* init controller code */
1386 	ctlr->mode_bits = SPI_CPOL | SPI_CPHA | SPI_CS_HIGH;
1387 	ctlr->mode_bits |= SPI_LSB_FIRST | SPI_3WIRE;
1388 	ctlr->flags = chipdata->ctlr_flags;
1389 	ctlr->bus_num = pdev->id;
1390 	ctlr->dev.of_node = pdev->dev.of_node;
1391 	ctlr->setup = sh_msiof_spi_setup;
1392 	ctlr->prepare_message = sh_msiof_prepare_message;
1393 	ctlr->slave_abort = sh_msiof_slave_abort;
1394 	ctlr->bits_per_word_mask = chipdata->bits_per_word_mask;
1395 	ctlr->auto_runtime_pm = true;
1396 	ctlr->transfer_one = sh_msiof_transfer_one;
1397 	ctlr->use_gpio_descriptors = true;
1398 
1399 	ret = sh_msiof_request_dma(p);
1400 	if (ret < 0)
1401 		dev_warn(&pdev->dev, "DMA not available, using PIO\n");
1402 
1403 	ret = devm_spi_register_controller(&pdev->dev, ctlr);
1404 	if (ret < 0) {
1405 		dev_err(&pdev->dev, "devm_spi_register_controller error.\n");
1406 		goto err2;
1407 	}
1408 
1409 	return 0;
1410 
1411  err2:
1412 	sh_msiof_release_dma(p);
1413 	pm_runtime_disable(&pdev->dev);
1414  err1:
1415 	spi_controller_put(ctlr);
1416 	return ret;
1417 }
1418 
1419 static int sh_msiof_spi_remove(struct platform_device *pdev)
1420 {
1421 	struct sh_msiof_spi_priv *p = platform_get_drvdata(pdev);
1422 
1423 	sh_msiof_release_dma(p);
1424 	pm_runtime_disable(&pdev->dev);
1425 	return 0;
1426 }
1427 
1428 static const struct platform_device_id spi_driver_ids[] = {
1429 	{ "spi_sh_msiof",	(kernel_ulong_t)&sh_data },
1430 	{},
1431 };
1432 MODULE_DEVICE_TABLE(platform, spi_driver_ids);
1433 
1434 #ifdef CONFIG_PM_SLEEP
1435 static int sh_msiof_spi_suspend(struct device *dev)
1436 {
1437 	struct sh_msiof_spi_priv *p = dev_get_drvdata(dev);
1438 
1439 	return spi_controller_suspend(p->ctlr);
1440 }
1441 
1442 static int sh_msiof_spi_resume(struct device *dev)
1443 {
1444 	struct sh_msiof_spi_priv *p = dev_get_drvdata(dev);
1445 
1446 	return spi_controller_resume(p->ctlr);
1447 }
1448 
1449 static SIMPLE_DEV_PM_OPS(sh_msiof_spi_pm_ops, sh_msiof_spi_suspend,
1450 			 sh_msiof_spi_resume);
1451 #define DEV_PM_OPS	&sh_msiof_spi_pm_ops
1452 #else
1453 #define DEV_PM_OPS	NULL
1454 #endif /* CONFIG_PM_SLEEP */
1455 
1456 static struct platform_driver sh_msiof_spi_drv = {
1457 	.probe		= sh_msiof_spi_probe,
1458 	.remove		= sh_msiof_spi_remove,
1459 	.id_table	= spi_driver_ids,
1460 	.driver		= {
1461 		.name		= "spi_sh_msiof",
1462 		.pm		= DEV_PM_OPS,
1463 		.of_match_table = of_match_ptr(sh_msiof_match),
1464 	},
1465 };
1466 module_platform_driver(sh_msiof_spi_drv);
1467 
1468 MODULE_DESCRIPTION("SuperH MSIOF SPI Controller Interface Driver");
1469 MODULE_AUTHOR("Magnus Damm");
1470 MODULE_LICENSE("GPL v2");
1471 MODULE_ALIAS("platform:spi_sh_msiof");
1472