1 // SPDX-License-Identifier: GPL-2.0-only
2 //
3 // Driver for Cadence QSPI Controller
4 //
5 // Copyright Altera Corporation (C) 2012-2014. All rights reserved.
6 // Copyright Intel Corporation (C) 2019-2020. All rights reserved.
7 // Copyright (C) 2020 Texas Instruments Incorporated - http://www.ti.com
8 
9 #include <linux/clk.h>
10 #include <linux/completion.h>
11 #include <linux/delay.h>
12 #include <linux/dma-mapping.h>
13 #include <linux/dmaengine.h>
14 #include <linux/err.h>
15 #include <linux/errno.h>
16 #include <linux/interrupt.h>
17 #include <linux/io.h>
18 #include <linux/iopoll.h>
19 #include <linux/jiffies.h>
20 #include <linux/kernel.h>
21 #include <linux/module.h>
22 #include <linux/of_device.h>
23 #include <linux/of.h>
24 #include <linux/platform_device.h>
25 #include <linux/pm_runtime.h>
26 #include <linux/reset.h>
27 #include <linux/sched.h>
28 #include <linux/spi/spi.h>
29 #include <linux/spi/spi-mem.h>
30 #include <linux/timer.h>
31 
32 #define CQSPI_NAME			"cadence-qspi"
33 #define CQSPI_MAX_CHIPSELECT		16
34 
35 /* Quirks */
36 #define CQSPI_NEEDS_WR_DELAY		BIT(0)
37 #define CQSPI_DISABLE_DAC_MODE		BIT(1)
38 
39 /* Capabilities */
40 #define CQSPI_SUPPORTS_OCTAL		BIT(0)
41 
42 struct cqspi_st;
43 
44 struct cqspi_flash_pdata {
45 	struct cqspi_st	*cqspi;
46 	u32		clk_rate;
47 	u32		read_delay;
48 	u32		tshsl_ns;
49 	u32		tsd2d_ns;
50 	u32		tchsh_ns;
51 	u32		tslch_ns;
52 	u8		inst_width;
53 	u8		addr_width;
54 	u8		data_width;
55 	bool		dtr;
56 	u8		cs;
57 };
58 
59 struct cqspi_st {
60 	struct platform_device	*pdev;
61 
62 	struct clk		*clk;
63 	unsigned int		sclk;
64 
65 	void __iomem		*iobase;
66 	void __iomem		*ahb_base;
67 	resource_size_t		ahb_size;
68 	struct completion	transfer_complete;
69 
70 	struct dma_chan		*rx_chan;
71 	struct completion	rx_dma_complete;
72 	dma_addr_t		mmap_phys_base;
73 
74 	int			current_cs;
75 	unsigned long		master_ref_clk_hz;
76 	bool			is_decoded_cs;
77 	u32			fifo_depth;
78 	u32			fifo_width;
79 	u32			num_chipselect;
80 	bool			rclk_en;
81 	u32			trigger_address;
82 	u32			wr_delay;
83 	bool			use_direct_mode;
84 	struct cqspi_flash_pdata f_pdata[CQSPI_MAX_CHIPSELECT];
85 };
86 
87 struct cqspi_driver_platdata {
88 	u32 hwcaps_mask;
89 	u8 quirks;
90 };
91 
92 /* Operation timeout value */
93 #define CQSPI_TIMEOUT_MS			500
94 #define CQSPI_READ_TIMEOUT_MS			10
95 
96 /* Instruction type */
97 #define CQSPI_INST_TYPE_SINGLE			0
98 #define CQSPI_INST_TYPE_DUAL			1
99 #define CQSPI_INST_TYPE_QUAD			2
100 #define CQSPI_INST_TYPE_OCTAL			3
101 
102 #define CQSPI_DUMMY_CLKS_PER_BYTE		8
103 #define CQSPI_DUMMY_BYTES_MAX			4
104 #define CQSPI_DUMMY_CLKS_MAX			31
105 
106 #define CQSPI_STIG_DATA_LEN_MAX			8
107 
108 /* Register map */
109 #define CQSPI_REG_CONFIG			0x00
110 #define CQSPI_REG_CONFIG_ENABLE_MASK		BIT(0)
111 #define CQSPI_REG_CONFIG_ENB_DIR_ACC_CTRL	BIT(7)
112 #define CQSPI_REG_CONFIG_DECODE_MASK		BIT(9)
113 #define CQSPI_REG_CONFIG_CHIPSELECT_LSB		10
114 #define CQSPI_REG_CONFIG_DMA_MASK		BIT(15)
115 #define CQSPI_REG_CONFIG_BAUD_LSB		19
116 #define CQSPI_REG_CONFIG_DTR_PROTO		BIT(24)
117 #define CQSPI_REG_CONFIG_DUAL_OPCODE		BIT(30)
118 #define CQSPI_REG_CONFIG_IDLE_LSB		31
119 #define CQSPI_REG_CONFIG_CHIPSELECT_MASK	0xF
120 #define CQSPI_REG_CONFIG_BAUD_MASK		0xF
121 
122 #define CQSPI_REG_RD_INSTR			0x04
123 #define CQSPI_REG_RD_INSTR_OPCODE_LSB		0
124 #define CQSPI_REG_RD_INSTR_TYPE_INSTR_LSB	8
125 #define CQSPI_REG_RD_INSTR_TYPE_ADDR_LSB	12
126 #define CQSPI_REG_RD_INSTR_TYPE_DATA_LSB	16
127 #define CQSPI_REG_RD_INSTR_MODE_EN_LSB		20
128 #define CQSPI_REG_RD_INSTR_DUMMY_LSB		24
129 #define CQSPI_REG_RD_INSTR_TYPE_INSTR_MASK	0x3
130 #define CQSPI_REG_RD_INSTR_TYPE_ADDR_MASK	0x3
131 #define CQSPI_REG_RD_INSTR_TYPE_DATA_MASK	0x3
132 #define CQSPI_REG_RD_INSTR_DUMMY_MASK		0x1F
133 
134 #define CQSPI_REG_WR_INSTR			0x08
135 #define CQSPI_REG_WR_INSTR_OPCODE_LSB		0
136 #define CQSPI_REG_WR_INSTR_TYPE_ADDR_LSB	12
137 #define CQSPI_REG_WR_INSTR_TYPE_DATA_LSB	16
138 
139 #define CQSPI_REG_DELAY				0x0C
140 #define CQSPI_REG_DELAY_TSLCH_LSB		0
141 #define CQSPI_REG_DELAY_TCHSH_LSB		8
142 #define CQSPI_REG_DELAY_TSD2D_LSB		16
143 #define CQSPI_REG_DELAY_TSHSL_LSB		24
144 #define CQSPI_REG_DELAY_TSLCH_MASK		0xFF
145 #define CQSPI_REG_DELAY_TCHSH_MASK		0xFF
146 #define CQSPI_REG_DELAY_TSD2D_MASK		0xFF
147 #define CQSPI_REG_DELAY_TSHSL_MASK		0xFF
148 
149 #define CQSPI_REG_READCAPTURE			0x10
150 #define CQSPI_REG_READCAPTURE_BYPASS_LSB	0
151 #define CQSPI_REG_READCAPTURE_DELAY_LSB		1
152 #define CQSPI_REG_READCAPTURE_DELAY_MASK	0xF
153 
154 #define CQSPI_REG_SIZE				0x14
155 #define CQSPI_REG_SIZE_ADDRESS_LSB		0
156 #define CQSPI_REG_SIZE_PAGE_LSB			4
157 #define CQSPI_REG_SIZE_BLOCK_LSB		16
158 #define CQSPI_REG_SIZE_ADDRESS_MASK		0xF
159 #define CQSPI_REG_SIZE_PAGE_MASK		0xFFF
160 #define CQSPI_REG_SIZE_BLOCK_MASK		0x3F
161 
162 #define CQSPI_REG_SRAMPARTITION			0x18
163 #define CQSPI_REG_INDIRECTTRIGGER		0x1C
164 
165 #define CQSPI_REG_DMA				0x20
166 #define CQSPI_REG_DMA_SINGLE_LSB		0
167 #define CQSPI_REG_DMA_BURST_LSB			8
168 #define CQSPI_REG_DMA_SINGLE_MASK		0xFF
169 #define CQSPI_REG_DMA_BURST_MASK		0xFF
170 
171 #define CQSPI_REG_REMAP				0x24
172 #define CQSPI_REG_MODE_BIT			0x28
173 
174 #define CQSPI_REG_SDRAMLEVEL			0x2C
175 #define CQSPI_REG_SDRAMLEVEL_RD_LSB		0
176 #define CQSPI_REG_SDRAMLEVEL_WR_LSB		16
177 #define CQSPI_REG_SDRAMLEVEL_RD_MASK		0xFFFF
178 #define CQSPI_REG_SDRAMLEVEL_WR_MASK		0xFFFF
179 
180 #define CQSPI_REG_WR_COMPLETION_CTRL		0x38
181 #define CQSPI_REG_WR_DISABLE_AUTO_POLL		BIT(14)
182 
183 #define CQSPI_REG_IRQSTATUS			0x40
184 #define CQSPI_REG_IRQMASK			0x44
185 
186 #define CQSPI_REG_INDIRECTRD			0x60
187 #define CQSPI_REG_INDIRECTRD_START_MASK		BIT(0)
188 #define CQSPI_REG_INDIRECTRD_CANCEL_MASK	BIT(1)
189 #define CQSPI_REG_INDIRECTRD_DONE_MASK		BIT(5)
190 
191 #define CQSPI_REG_INDIRECTRDWATERMARK		0x64
192 #define CQSPI_REG_INDIRECTRDSTARTADDR		0x68
193 #define CQSPI_REG_INDIRECTRDBYTES		0x6C
194 
195 #define CQSPI_REG_CMDCTRL			0x90
196 #define CQSPI_REG_CMDCTRL_EXECUTE_MASK		BIT(0)
197 #define CQSPI_REG_CMDCTRL_INPROGRESS_MASK	BIT(1)
198 #define CQSPI_REG_CMDCTRL_DUMMY_LSB		7
199 #define CQSPI_REG_CMDCTRL_WR_BYTES_LSB		12
200 #define CQSPI_REG_CMDCTRL_WR_EN_LSB		15
201 #define CQSPI_REG_CMDCTRL_ADD_BYTES_LSB		16
202 #define CQSPI_REG_CMDCTRL_ADDR_EN_LSB		19
203 #define CQSPI_REG_CMDCTRL_RD_BYTES_LSB		20
204 #define CQSPI_REG_CMDCTRL_RD_EN_LSB		23
205 #define CQSPI_REG_CMDCTRL_OPCODE_LSB		24
206 #define CQSPI_REG_CMDCTRL_WR_BYTES_MASK		0x7
207 #define CQSPI_REG_CMDCTRL_ADD_BYTES_MASK	0x3
208 #define CQSPI_REG_CMDCTRL_RD_BYTES_MASK		0x7
209 #define CQSPI_REG_CMDCTRL_DUMMY_MASK		0x1F
210 
211 #define CQSPI_REG_INDIRECTWR			0x70
212 #define CQSPI_REG_INDIRECTWR_START_MASK		BIT(0)
213 #define CQSPI_REG_INDIRECTWR_CANCEL_MASK	BIT(1)
214 #define CQSPI_REG_INDIRECTWR_DONE_MASK		BIT(5)
215 
216 #define CQSPI_REG_INDIRECTWRWATERMARK		0x74
217 #define CQSPI_REG_INDIRECTWRSTARTADDR		0x78
218 #define CQSPI_REG_INDIRECTWRBYTES		0x7C
219 
220 #define CQSPI_REG_CMDADDRESS			0x94
221 #define CQSPI_REG_CMDREADDATALOWER		0xA0
222 #define CQSPI_REG_CMDREADDATAUPPER		0xA4
223 #define CQSPI_REG_CMDWRITEDATALOWER		0xA8
224 #define CQSPI_REG_CMDWRITEDATAUPPER		0xAC
225 
226 #define CQSPI_REG_POLLING_STATUS		0xB0
227 #define CQSPI_REG_POLLING_STATUS_DUMMY_LSB	16
228 
229 #define CQSPI_REG_OP_EXT_LOWER			0xE0
230 #define CQSPI_REG_OP_EXT_READ_LSB		24
231 #define CQSPI_REG_OP_EXT_WRITE_LSB		16
232 #define CQSPI_REG_OP_EXT_STIG_LSB		0
233 
234 /* Interrupt status bits */
235 #define CQSPI_REG_IRQ_MODE_ERR			BIT(0)
236 #define CQSPI_REG_IRQ_UNDERFLOW			BIT(1)
237 #define CQSPI_REG_IRQ_IND_COMP			BIT(2)
238 #define CQSPI_REG_IRQ_IND_RD_REJECT		BIT(3)
239 #define CQSPI_REG_IRQ_WR_PROTECTED_ERR		BIT(4)
240 #define CQSPI_REG_IRQ_ILLEGAL_AHB_ERR		BIT(5)
241 #define CQSPI_REG_IRQ_WATERMARK			BIT(6)
242 #define CQSPI_REG_IRQ_IND_SRAM_FULL		BIT(12)
243 
244 #define CQSPI_IRQ_MASK_RD		(CQSPI_REG_IRQ_WATERMARK	| \
245 					 CQSPI_REG_IRQ_IND_SRAM_FULL	| \
246 					 CQSPI_REG_IRQ_IND_COMP)
247 
248 #define CQSPI_IRQ_MASK_WR		(CQSPI_REG_IRQ_IND_COMP		| \
249 					 CQSPI_REG_IRQ_WATERMARK	| \
250 					 CQSPI_REG_IRQ_UNDERFLOW)
251 
252 #define CQSPI_IRQ_STATUS_MASK		0x1FFFF
253 
254 static int cqspi_wait_for_bit(void __iomem *reg, const u32 mask, bool clr)
255 {
256 	u32 val;
257 
258 	return readl_relaxed_poll_timeout(reg, val,
259 					  (((clr ? ~val : val) & mask) == mask),
260 					  10, CQSPI_TIMEOUT_MS * 1000);
261 }
262 
263 static bool cqspi_is_idle(struct cqspi_st *cqspi)
264 {
265 	u32 reg = readl(cqspi->iobase + CQSPI_REG_CONFIG);
266 
267 	return reg & (1 << CQSPI_REG_CONFIG_IDLE_LSB);
268 }
269 
270 static u32 cqspi_get_rd_sram_level(struct cqspi_st *cqspi)
271 {
272 	u32 reg = readl(cqspi->iobase + CQSPI_REG_SDRAMLEVEL);
273 
274 	reg >>= CQSPI_REG_SDRAMLEVEL_RD_LSB;
275 	return reg & CQSPI_REG_SDRAMLEVEL_RD_MASK;
276 }
277 
278 static irqreturn_t cqspi_irq_handler(int this_irq, void *dev)
279 {
280 	struct cqspi_st *cqspi = dev;
281 	unsigned int irq_status;
282 
283 	/* Read interrupt status */
284 	irq_status = readl(cqspi->iobase + CQSPI_REG_IRQSTATUS);
285 
286 	/* Clear interrupt */
287 	writel(irq_status, cqspi->iobase + CQSPI_REG_IRQSTATUS);
288 
289 	irq_status &= CQSPI_IRQ_MASK_RD | CQSPI_IRQ_MASK_WR;
290 
291 	if (irq_status)
292 		complete(&cqspi->transfer_complete);
293 
294 	return IRQ_HANDLED;
295 }
296 
297 static unsigned int cqspi_calc_rdreg(struct cqspi_flash_pdata *f_pdata)
298 {
299 	u32 rdreg = 0;
300 
301 	rdreg |= f_pdata->inst_width << CQSPI_REG_RD_INSTR_TYPE_INSTR_LSB;
302 	rdreg |= f_pdata->addr_width << CQSPI_REG_RD_INSTR_TYPE_ADDR_LSB;
303 	rdreg |= f_pdata->data_width << CQSPI_REG_RD_INSTR_TYPE_DATA_LSB;
304 
305 	return rdreg;
306 }
307 
308 static unsigned int cqspi_calc_dummy(const struct spi_mem_op *op, bool dtr)
309 {
310 	unsigned int dummy_clk;
311 
312 	dummy_clk = op->dummy.nbytes * (8 / op->dummy.buswidth);
313 	if (dtr)
314 		dummy_clk /= 2;
315 
316 	return dummy_clk;
317 }
318 
319 static int cqspi_set_protocol(struct cqspi_flash_pdata *f_pdata,
320 			      const struct spi_mem_op *op)
321 {
322 	f_pdata->inst_width = CQSPI_INST_TYPE_SINGLE;
323 	f_pdata->addr_width = CQSPI_INST_TYPE_SINGLE;
324 	f_pdata->data_width = CQSPI_INST_TYPE_SINGLE;
325 	f_pdata->dtr = op->data.dtr && op->cmd.dtr && op->addr.dtr;
326 
327 	switch (op->data.buswidth) {
328 	case 0:
329 		break;
330 	case 1:
331 		f_pdata->data_width = CQSPI_INST_TYPE_SINGLE;
332 		break;
333 	case 2:
334 		f_pdata->data_width = CQSPI_INST_TYPE_DUAL;
335 		break;
336 	case 4:
337 		f_pdata->data_width = CQSPI_INST_TYPE_QUAD;
338 		break;
339 	case 8:
340 		f_pdata->data_width = CQSPI_INST_TYPE_OCTAL;
341 		break;
342 	default:
343 		return -EINVAL;
344 	}
345 
346 	/* Right now we only support 8-8-8 DTR mode. */
347 	if (f_pdata->dtr) {
348 		switch (op->cmd.buswidth) {
349 		case 0:
350 			break;
351 		case 8:
352 			f_pdata->inst_width = CQSPI_INST_TYPE_OCTAL;
353 			break;
354 		default:
355 			return -EINVAL;
356 		}
357 
358 		switch (op->addr.buswidth) {
359 		case 0:
360 			break;
361 		case 8:
362 			f_pdata->addr_width = CQSPI_INST_TYPE_OCTAL;
363 			break;
364 		default:
365 			return -EINVAL;
366 		}
367 
368 		switch (op->data.buswidth) {
369 		case 0:
370 			break;
371 		case 8:
372 			f_pdata->data_width = CQSPI_INST_TYPE_OCTAL;
373 			break;
374 		default:
375 			return -EINVAL;
376 		}
377 	}
378 
379 	return 0;
380 }
381 
382 static int cqspi_wait_idle(struct cqspi_st *cqspi)
383 {
384 	const unsigned int poll_idle_retry = 3;
385 	unsigned int count = 0;
386 	unsigned long timeout;
387 
388 	timeout = jiffies + msecs_to_jiffies(CQSPI_TIMEOUT_MS);
389 	while (1) {
390 		/*
391 		 * Read few times in succession to ensure the controller
392 		 * is indeed idle, that is, the bit does not transition
393 		 * low again.
394 		 */
395 		if (cqspi_is_idle(cqspi))
396 			count++;
397 		else
398 			count = 0;
399 
400 		if (count >= poll_idle_retry)
401 			return 0;
402 
403 		if (time_after(jiffies, timeout)) {
404 			/* Timeout, in busy mode. */
405 			dev_err(&cqspi->pdev->dev,
406 				"QSPI is still busy after %dms timeout.\n",
407 				CQSPI_TIMEOUT_MS);
408 			return -ETIMEDOUT;
409 		}
410 
411 		cpu_relax();
412 	}
413 }
414 
415 static int cqspi_exec_flash_cmd(struct cqspi_st *cqspi, unsigned int reg)
416 {
417 	void __iomem *reg_base = cqspi->iobase;
418 	int ret;
419 
420 	/* Write the CMDCTRL without start execution. */
421 	writel(reg, reg_base + CQSPI_REG_CMDCTRL);
422 	/* Start execute */
423 	reg |= CQSPI_REG_CMDCTRL_EXECUTE_MASK;
424 	writel(reg, reg_base + CQSPI_REG_CMDCTRL);
425 
426 	/* Polling for completion. */
427 	ret = cqspi_wait_for_bit(reg_base + CQSPI_REG_CMDCTRL,
428 				 CQSPI_REG_CMDCTRL_INPROGRESS_MASK, 1);
429 	if (ret) {
430 		dev_err(&cqspi->pdev->dev,
431 			"Flash command execution timed out.\n");
432 		return ret;
433 	}
434 
435 	/* Polling QSPI idle status. */
436 	return cqspi_wait_idle(cqspi);
437 }
438 
439 static int cqspi_setup_opcode_ext(struct cqspi_flash_pdata *f_pdata,
440 				  const struct spi_mem_op *op,
441 				  unsigned int shift)
442 {
443 	struct cqspi_st *cqspi = f_pdata->cqspi;
444 	void __iomem *reg_base = cqspi->iobase;
445 	unsigned int reg;
446 	u8 ext;
447 
448 	if (op->cmd.nbytes != 2)
449 		return -EINVAL;
450 
451 	/* Opcode extension is the LSB. */
452 	ext = op->cmd.opcode & 0xff;
453 
454 	reg = readl(reg_base + CQSPI_REG_OP_EXT_LOWER);
455 	reg &= ~(0xff << shift);
456 	reg |= ext << shift;
457 	writel(reg, reg_base + CQSPI_REG_OP_EXT_LOWER);
458 
459 	return 0;
460 }
461 
462 static int cqspi_enable_dtr(struct cqspi_flash_pdata *f_pdata,
463 			    const struct spi_mem_op *op, unsigned int shift,
464 			    bool enable)
465 {
466 	struct cqspi_st *cqspi = f_pdata->cqspi;
467 	void __iomem *reg_base = cqspi->iobase;
468 	unsigned int reg;
469 	int ret;
470 
471 	reg = readl(reg_base + CQSPI_REG_CONFIG);
472 
473 	/*
474 	 * We enable dual byte opcode here. The callers have to set up the
475 	 * extension opcode based on which type of operation it is.
476 	 */
477 	if (enable) {
478 		reg |= CQSPI_REG_CONFIG_DTR_PROTO;
479 		reg |= CQSPI_REG_CONFIG_DUAL_OPCODE;
480 
481 		/* Set up command opcode extension. */
482 		ret = cqspi_setup_opcode_ext(f_pdata, op, shift);
483 		if (ret)
484 			return ret;
485 	} else {
486 		reg &= ~CQSPI_REG_CONFIG_DTR_PROTO;
487 		reg &= ~CQSPI_REG_CONFIG_DUAL_OPCODE;
488 	}
489 
490 	writel(reg, reg_base + CQSPI_REG_CONFIG);
491 
492 	return cqspi_wait_idle(cqspi);
493 }
494 
495 static int cqspi_command_read(struct cqspi_flash_pdata *f_pdata,
496 			      const struct spi_mem_op *op)
497 {
498 	struct cqspi_st *cqspi = f_pdata->cqspi;
499 	void __iomem *reg_base = cqspi->iobase;
500 	u8 *rxbuf = op->data.buf.in;
501 	u8 opcode;
502 	size_t n_rx = op->data.nbytes;
503 	unsigned int rdreg;
504 	unsigned int reg;
505 	unsigned int dummy_clk;
506 	size_t read_len;
507 	int status;
508 
509 	status = cqspi_set_protocol(f_pdata, op);
510 	if (status)
511 		return status;
512 
513 	status = cqspi_enable_dtr(f_pdata, op, CQSPI_REG_OP_EXT_STIG_LSB,
514 				  f_pdata->dtr);
515 	if (status)
516 		return status;
517 
518 	if (!n_rx || n_rx > CQSPI_STIG_DATA_LEN_MAX || !rxbuf) {
519 		dev_err(&cqspi->pdev->dev,
520 			"Invalid input argument, len %zu rxbuf 0x%p\n",
521 			n_rx, rxbuf);
522 		return -EINVAL;
523 	}
524 
525 	if (f_pdata->dtr)
526 		opcode = op->cmd.opcode >> 8;
527 	else
528 		opcode = op->cmd.opcode;
529 
530 	reg = opcode << CQSPI_REG_CMDCTRL_OPCODE_LSB;
531 
532 	rdreg = cqspi_calc_rdreg(f_pdata);
533 	writel(rdreg, reg_base + CQSPI_REG_RD_INSTR);
534 
535 	dummy_clk = cqspi_calc_dummy(op, f_pdata->dtr);
536 	if (dummy_clk > CQSPI_DUMMY_CLKS_MAX)
537 		return -EOPNOTSUPP;
538 
539 	if (dummy_clk)
540 		reg |= (dummy_clk & CQSPI_REG_CMDCTRL_DUMMY_MASK)
541 		     << CQSPI_REG_CMDCTRL_DUMMY_LSB;
542 
543 	reg |= (0x1 << CQSPI_REG_CMDCTRL_RD_EN_LSB);
544 
545 	/* 0 means 1 byte. */
546 	reg |= (((n_rx - 1) & CQSPI_REG_CMDCTRL_RD_BYTES_MASK)
547 		<< CQSPI_REG_CMDCTRL_RD_BYTES_LSB);
548 	status = cqspi_exec_flash_cmd(cqspi, reg);
549 	if (status)
550 		return status;
551 
552 	reg = readl(reg_base + CQSPI_REG_CMDREADDATALOWER);
553 
554 	/* Put the read value into rx_buf */
555 	read_len = (n_rx > 4) ? 4 : n_rx;
556 	memcpy(rxbuf, &reg, read_len);
557 	rxbuf += read_len;
558 
559 	if (n_rx > 4) {
560 		reg = readl(reg_base + CQSPI_REG_CMDREADDATAUPPER);
561 
562 		read_len = n_rx - read_len;
563 		memcpy(rxbuf, &reg, read_len);
564 	}
565 
566 	return 0;
567 }
568 
569 static int cqspi_command_write(struct cqspi_flash_pdata *f_pdata,
570 			       const struct spi_mem_op *op)
571 {
572 	struct cqspi_st *cqspi = f_pdata->cqspi;
573 	void __iomem *reg_base = cqspi->iobase;
574 	u8 opcode;
575 	const u8 *txbuf = op->data.buf.out;
576 	size_t n_tx = op->data.nbytes;
577 	unsigned int reg;
578 	unsigned int data;
579 	size_t write_len;
580 	int ret;
581 
582 	ret = cqspi_set_protocol(f_pdata, op);
583 	if (ret)
584 		return ret;
585 
586 	ret = cqspi_enable_dtr(f_pdata, op, CQSPI_REG_OP_EXT_STIG_LSB,
587 			       f_pdata->dtr);
588 	if (ret)
589 		return ret;
590 
591 	if (n_tx > CQSPI_STIG_DATA_LEN_MAX || (n_tx && !txbuf)) {
592 		dev_err(&cqspi->pdev->dev,
593 			"Invalid input argument, cmdlen %zu txbuf 0x%p\n",
594 			n_tx, txbuf);
595 		return -EINVAL;
596 	}
597 
598 	reg = cqspi_calc_rdreg(f_pdata);
599 	writel(reg, reg_base + CQSPI_REG_RD_INSTR);
600 
601 	if (f_pdata->dtr)
602 		opcode = op->cmd.opcode >> 8;
603 	else
604 		opcode = op->cmd.opcode;
605 
606 	reg = opcode << CQSPI_REG_CMDCTRL_OPCODE_LSB;
607 
608 	if (op->addr.nbytes) {
609 		reg |= (0x1 << CQSPI_REG_CMDCTRL_ADDR_EN_LSB);
610 		reg |= ((op->addr.nbytes - 1) &
611 			CQSPI_REG_CMDCTRL_ADD_BYTES_MASK)
612 			<< CQSPI_REG_CMDCTRL_ADD_BYTES_LSB;
613 
614 		writel(op->addr.val, reg_base + CQSPI_REG_CMDADDRESS);
615 	}
616 
617 	if (n_tx) {
618 		reg |= (0x1 << CQSPI_REG_CMDCTRL_WR_EN_LSB);
619 		reg |= ((n_tx - 1) & CQSPI_REG_CMDCTRL_WR_BYTES_MASK)
620 			<< CQSPI_REG_CMDCTRL_WR_BYTES_LSB;
621 		data = 0;
622 		write_len = (n_tx > 4) ? 4 : n_tx;
623 		memcpy(&data, txbuf, write_len);
624 		txbuf += write_len;
625 		writel(data, reg_base + CQSPI_REG_CMDWRITEDATALOWER);
626 
627 		if (n_tx > 4) {
628 			data = 0;
629 			write_len = n_tx - 4;
630 			memcpy(&data, txbuf, write_len);
631 			writel(data, reg_base + CQSPI_REG_CMDWRITEDATAUPPER);
632 		}
633 	}
634 
635 	return cqspi_exec_flash_cmd(cqspi, reg);
636 }
637 
638 static int cqspi_read_setup(struct cqspi_flash_pdata *f_pdata,
639 			    const struct spi_mem_op *op)
640 {
641 	struct cqspi_st *cqspi = f_pdata->cqspi;
642 	void __iomem *reg_base = cqspi->iobase;
643 	unsigned int dummy_clk = 0;
644 	unsigned int reg;
645 	int ret;
646 	u8 opcode;
647 
648 	ret = cqspi_enable_dtr(f_pdata, op, CQSPI_REG_OP_EXT_READ_LSB,
649 			       f_pdata->dtr);
650 	if (ret)
651 		return ret;
652 
653 	if (f_pdata->dtr)
654 		opcode = op->cmd.opcode >> 8;
655 	else
656 		opcode = op->cmd.opcode;
657 
658 	reg = opcode << CQSPI_REG_RD_INSTR_OPCODE_LSB;
659 	reg |= cqspi_calc_rdreg(f_pdata);
660 
661 	/* Setup dummy clock cycles */
662 	dummy_clk = cqspi_calc_dummy(op, f_pdata->dtr);
663 
664 	if (dummy_clk > CQSPI_DUMMY_CLKS_MAX)
665 		return -EOPNOTSUPP;
666 
667 	if (dummy_clk)
668 		reg |= (dummy_clk & CQSPI_REG_RD_INSTR_DUMMY_MASK)
669 		       << CQSPI_REG_RD_INSTR_DUMMY_LSB;
670 
671 	writel(reg, reg_base + CQSPI_REG_RD_INSTR);
672 
673 	/* Set address width */
674 	reg = readl(reg_base + CQSPI_REG_SIZE);
675 	reg &= ~CQSPI_REG_SIZE_ADDRESS_MASK;
676 	reg |= (op->addr.nbytes - 1);
677 	writel(reg, reg_base + CQSPI_REG_SIZE);
678 	return 0;
679 }
680 
681 static int cqspi_indirect_read_execute(struct cqspi_flash_pdata *f_pdata,
682 				       u8 *rxbuf, loff_t from_addr,
683 				       const size_t n_rx)
684 {
685 	struct cqspi_st *cqspi = f_pdata->cqspi;
686 	struct device *dev = &cqspi->pdev->dev;
687 	void __iomem *reg_base = cqspi->iobase;
688 	void __iomem *ahb_base = cqspi->ahb_base;
689 	unsigned int remaining = n_rx;
690 	unsigned int mod_bytes = n_rx % 4;
691 	unsigned int bytes_to_read = 0;
692 	u8 *rxbuf_end = rxbuf + n_rx;
693 	int ret = 0;
694 
695 	writel(from_addr, reg_base + CQSPI_REG_INDIRECTRDSTARTADDR);
696 	writel(remaining, reg_base + CQSPI_REG_INDIRECTRDBYTES);
697 
698 	/* Clear all interrupts. */
699 	writel(CQSPI_IRQ_STATUS_MASK, reg_base + CQSPI_REG_IRQSTATUS);
700 
701 	writel(CQSPI_IRQ_MASK_RD, reg_base + CQSPI_REG_IRQMASK);
702 
703 	reinit_completion(&cqspi->transfer_complete);
704 	writel(CQSPI_REG_INDIRECTRD_START_MASK,
705 	       reg_base + CQSPI_REG_INDIRECTRD);
706 
707 	while (remaining > 0) {
708 		if (!wait_for_completion_timeout(&cqspi->transfer_complete,
709 						 msecs_to_jiffies(CQSPI_READ_TIMEOUT_MS)))
710 			ret = -ETIMEDOUT;
711 
712 		bytes_to_read = cqspi_get_rd_sram_level(cqspi);
713 
714 		if (ret && bytes_to_read == 0) {
715 			dev_err(dev, "Indirect read timeout, no bytes\n");
716 			goto failrd;
717 		}
718 
719 		while (bytes_to_read != 0) {
720 			unsigned int word_remain = round_down(remaining, 4);
721 
722 			bytes_to_read *= cqspi->fifo_width;
723 			bytes_to_read = bytes_to_read > remaining ?
724 					remaining : bytes_to_read;
725 			bytes_to_read = round_down(bytes_to_read, 4);
726 			/* Read 4 byte word chunks then single bytes */
727 			if (bytes_to_read) {
728 				ioread32_rep(ahb_base, rxbuf,
729 					     (bytes_to_read / 4));
730 			} else if (!word_remain && mod_bytes) {
731 				unsigned int temp = ioread32(ahb_base);
732 
733 				bytes_to_read = mod_bytes;
734 				memcpy(rxbuf, &temp, min((unsigned int)
735 							 (rxbuf_end - rxbuf),
736 							 bytes_to_read));
737 			}
738 			rxbuf += bytes_to_read;
739 			remaining -= bytes_to_read;
740 			bytes_to_read = cqspi_get_rd_sram_level(cqspi);
741 		}
742 
743 		if (remaining > 0)
744 			reinit_completion(&cqspi->transfer_complete);
745 	}
746 
747 	/* Check indirect done status */
748 	ret = cqspi_wait_for_bit(reg_base + CQSPI_REG_INDIRECTRD,
749 				 CQSPI_REG_INDIRECTRD_DONE_MASK, 0);
750 	if (ret) {
751 		dev_err(dev, "Indirect read completion error (%i)\n", ret);
752 		goto failrd;
753 	}
754 
755 	/* Disable interrupt */
756 	writel(0, reg_base + CQSPI_REG_IRQMASK);
757 
758 	/* Clear indirect completion status */
759 	writel(CQSPI_REG_INDIRECTRD_DONE_MASK, reg_base + CQSPI_REG_INDIRECTRD);
760 
761 	return 0;
762 
763 failrd:
764 	/* Disable interrupt */
765 	writel(0, reg_base + CQSPI_REG_IRQMASK);
766 
767 	/* Cancel the indirect read */
768 	writel(CQSPI_REG_INDIRECTWR_CANCEL_MASK,
769 	       reg_base + CQSPI_REG_INDIRECTRD);
770 	return ret;
771 }
772 
773 static int cqspi_write_setup(struct cqspi_flash_pdata *f_pdata,
774 			     const struct spi_mem_op *op)
775 {
776 	unsigned int reg;
777 	int ret;
778 	struct cqspi_st *cqspi = f_pdata->cqspi;
779 	void __iomem *reg_base = cqspi->iobase;
780 	u8 opcode;
781 
782 	ret = cqspi_enable_dtr(f_pdata, op, CQSPI_REG_OP_EXT_WRITE_LSB,
783 			       f_pdata->dtr);
784 	if (ret)
785 		return ret;
786 
787 	if (f_pdata->dtr)
788 		opcode = op->cmd.opcode >> 8;
789 	else
790 		opcode = op->cmd.opcode;
791 
792 	/* Set opcode. */
793 	reg = opcode << CQSPI_REG_WR_INSTR_OPCODE_LSB;
794 	reg |= f_pdata->data_width << CQSPI_REG_WR_INSTR_TYPE_DATA_LSB;
795 	reg |= f_pdata->addr_width << CQSPI_REG_WR_INSTR_TYPE_ADDR_LSB;
796 	writel(reg, reg_base + CQSPI_REG_WR_INSTR);
797 	reg = cqspi_calc_rdreg(f_pdata);
798 	writel(reg, reg_base + CQSPI_REG_RD_INSTR);
799 
800 	if (f_pdata->dtr) {
801 		/*
802 		 * Some flashes like the cypress Semper flash expect a 4-byte
803 		 * dummy address with the Read SR command in DTR mode, but this
804 		 * controller does not support sending address with the Read SR
805 		 * command. So, disable write completion polling on the
806 		 * controller's side. spi-nor will take care of polling the
807 		 * status register.
808 		 */
809 		reg = readl(reg_base + CQSPI_REG_WR_COMPLETION_CTRL);
810 		reg |= CQSPI_REG_WR_DISABLE_AUTO_POLL;
811 		writel(reg, reg_base + CQSPI_REG_WR_COMPLETION_CTRL);
812 	}
813 
814 	reg = readl(reg_base + CQSPI_REG_SIZE);
815 	reg &= ~CQSPI_REG_SIZE_ADDRESS_MASK;
816 	reg |= (op->addr.nbytes - 1);
817 	writel(reg, reg_base + CQSPI_REG_SIZE);
818 	return 0;
819 }
820 
821 static int cqspi_indirect_write_execute(struct cqspi_flash_pdata *f_pdata,
822 					loff_t to_addr, const u8 *txbuf,
823 					const size_t n_tx)
824 {
825 	struct cqspi_st *cqspi = f_pdata->cqspi;
826 	struct device *dev = &cqspi->pdev->dev;
827 	void __iomem *reg_base = cqspi->iobase;
828 	unsigned int remaining = n_tx;
829 	unsigned int write_bytes;
830 	int ret;
831 
832 	writel(to_addr, reg_base + CQSPI_REG_INDIRECTWRSTARTADDR);
833 	writel(remaining, reg_base + CQSPI_REG_INDIRECTWRBYTES);
834 
835 	/* Clear all interrupts. */
836 	writel(CQSPI_IRQ_STATUS_MASK, reg_base + CQSPI_REG_IRQSTATUS);
837 
838 	writel(CQSPI_IRQ_MASK_WR, reg_base + CQSPI_REG_IRQMASK);
839 
840 	reinit_completion(&cqspi->transfer_complete);
841 	writel(CQSPI_REG_INDIRECTWR_START_MASK,
842 	       reg_base + CQSPI_REG_INDIRECTWR);
843 	/*
844 	 * As per 66AK2G02 TRM SPRUHY8F section 11.15.5.3 Indirect Access
845 	 * Controller programming sequence, couple of cycles of
846 	 * QSPI_REF_CLK delay is required for the above bit to
847 	 * be internally synchronized by the QSPI module. Provide 5
848 	 * cycles of delay.
849 	 */
850 	if (cqspi->wr_delay)
851 		ndelay(cqspi->wr_delay);
852 
853 	while (remaining > 0) {
854 		size_t write_words, mod_bytes;
855 
856 		write_bytes = remaining;
857 		write_words = write_bytes / 4;
858 		mod_bytes = write_bytes % 4;
859 		/* Write 4 bytes at a time then single bytes. */
860 		if (write_words) {
861 			iowrite32_rep(cqspi->ahb_base, txbuf, write_words);
862 			txbuf += (write_words * 4);
863 		}
864 		if (mod_bytes) {
865 			unsigned int temp = 0xFFFFFFFF;
866 
867 			memcpy(&temp, txbuf, mod_bytes);
868 			iowrite32(temp, cqspi->ahb_base);
869 			txbuf += mod_bytes;
870 		}
871 
872 		if (!wait_for_completion_timeout(&cqspi->transfer_complete,
873 						 msecs_to_jiffies(CQSPI_TIMEOUT_MS))) {
874 			dev_err(dev, "Indirect write timeout\n");
875 			ret = -ETIMEDOUT;
876 			goto failwr;
877 		}
878 
879 		remaining -= write_bytes;
880 
881 		if (remaining > 0)
882 			reinit_completion(&cqspi->transfer_complete);
883 	}
884 
885 	/* Check indirect done status */
886 	ret = cqspi_wait_for_bit(reg_base + CQSPI_REG_INDIRECTWR,
887 				 CQSPI_REG_INDIRECTWR_DONE_MASK, 0);
888 	if (ret) {
889 		dev_err(dev, "Indirect write completion error (%i)\n", ret);
890 		goto failwr;
891 	}
892 
893 	/* Disable interrupt. */
894 	writel(0, reg_base + CQSPI_REG_IRQMASK);
895 
896 	/* Clear indirect completion status */
897 	writel(CQSPI_REG_INDIRECTWR_DONE_MASK, reg_base + CQSPI_REG_INDIRECTWR);
898 
899 	cqspi_wait_idle(cqspi);
900 
901 	return 0;
902 
903 failwr:
904 	/* Disable interrupt. */
905 	writel(0, reg_base + CQSPI_REG_IRQMASK);
906 
907 	/* Cancel the indirect write */
908 	writel(CQSPI_REG_INDIRECTWR_CANCEL_MASK,
909 	       reg_base + CQSPI_REG_INDIRECTWR);
910 	return ret;
911 }
912 
913 static void cqspi_chipselect(struct cqspi_flash_pdata *f_pdata)
914 {
915 	struct cqspi_st *cqspi = f_pdata->cqspi;
916 	void __iomem *reg_base = cqspi->iobase;
917 	unsigned int chip_select = f_pdata->cs;
918 	unsigned int reg;
919 
920 	reg = readl(reg_base + CQSPI_REG_CONFIG);
921 	if (cqspi->is_decoded_cs) {
922 		reg |= CQSPI_REG_CONFIG_DECODE_MASK;
923 	} else {
924 		reg &= ~CQSPI_REG_CONFIG_DECODE_MASK;
925 
926 		/* Convert CS if without decoder.
927 		 * CS0 to 4b'1110
928 		 * CS1 to 4b'1101
929 		 * CS2 to 4b'1011
930 		 * CS3 to 4b'0111
931 		 */
932 		chip_select = 0xF & ~(1 << chip_select);
933 	}
934 
935 	reg &= ~(CQSPI_REG_CONFIG_CHIPSELECT_MASK
936 		 << CQSPI_REG_CONFIG_CHIPSELECT_LSB);
937 	reg |= (chip_select & CQSPI_REG_CONFIG_CHIPSELECT_MASK)
938 	    << CQSPI_REG_CONFIG_CHIPSELECT_LSB;
939 	writel(reg, reg_base + CQSPI_REG_CONFIG);
940 }
941 
942 static unsigned int calculate_ticks_for_ns(const unsigned int ref_clk_hz,
943 					   const unsigned int ns_val)
944 {
945 	unsigned int ticks;
946 
947 	ticks = ref_clk_hz / 1000;	/* kHz */
948 	ticks = DIV_ROUND_UP(ticks * ns_val, 1000000);
949 
950 	return ticks;
951 }
952 
953 static void cqspi_delay(struct cqspi_flash_pdata *f_pdata)
954 {
955 	struct cqspi_st *cqspi = f_pdata->cqspi;
956 	void __iomem *iobase = cqspi->iobase;
957 	const unsigned int ref_clk_hz = cqspi->master_ref_clk_hz;
958 	unsigned int tshsl, tchsh, tslch, tsd2d;
959 	unsigned int reg;
960 	unsigned int tsclk;
961 
962 	/* calculate the number of ref ticks for one sclk tick */
963 	tsclk = DIV_ROUND_UP(ref_clk_hz, cqspi->sclk);
964 
965 	tshsl = calculate_ticks_for_ns(ref_clk_hz, f_pdata->tshsl_ns);
966 	/* this particular value must be at least one sclk */
967 	if (tshsl < tsclk)
968 		tshsl = tsclk;
969 
970 	tchsh = calculate_ticks_for_ns(ref_clk_hz, f_pdata->tchsh_ns);
971 	tslch = calculate_ticks_for_ns(ref_clk_hz, f_pdata->tslch_ns);
972 	tsd2d = calculate_ticks_for_ns(ref_clk_hz, f_pdata->tsd2d_ns);
973 
974 	reg = (tshsl & CQSPI_REG_DELAY_TSHSL_MASK)
975 	       << CQSPI_REG_DELAY_TSHSL_LSB;
976 	reg |= (tchsh & CQSPI_REG_DELAY_TCHSH_MASK)
977 		<< CQSPI_REG_DELAY_TCHSH_LSB;
978 	reg |= (tslch & CQSPI_REG_DELAY_TSLCH_MASK)
979 		<< CQSPI_REG_DELAY_TSLCH_LSB;
980 	reg |= (tsd2d & CQSPI_REG_DELAY_TSD2D_MASK)
981 		<< CQSPI_REG_DELAY_TSD2D_LSB;
982 	writel(reg, iobase + CQSPI_REG_DELAY);
983 }
984 
985 static void cqspi_config_baudrate_div(struct cqspi_st *cqspi)
986 {
987 	const unsigned int ref_clk_hz = cqspi->master_ref_clk_hz;
988 	void __iomem *reg_base = cqspi->iobase;
989 	u32 reg, div;
990 
991 	/* Recalculate the baudrate divisor based on QSPI specification. */
992 	div = DIV_ROUND_UP(ref_clk_hz, 2 * cqspi->sclk) - 1;
993 
994 	reg = readl(reg_base + CQSPI_REG_CONFIG);
995 	reg &= ~(CQSPI_REG_CONFIG_BAUD_MASK << CQSPI_REG_CONFIG_BAUD_LSB);
996 	reg |= (div & CQSPI_REG_CONFIG_BAUD_MASK) << CQSPI_REG_CONFIG_BAUD_LSB;
997 	writel(reg, reg_base + CQSPI_REG_CONFIG);
998 }
999 
1000 static void cqspi_readdata_capture(struct cqspi_st *cqspi,
1001 				   const bool bypass,
1002 				   const unsigned int delay)
1003 {
1004 	void __iomem *reg_base = cqspi->iobase;
1005 	unsigned int reg;
1006 
1007 	reg = readl(reg_base + CQSPI_REG_READCAPTURE);
1008 
1009 	if (bypass)
1010 		reg |= (1 << CQSPI_REG_READCAPTURE_BYPASS_LSB);
1011 	else
1012 		reg &= ~(1 << CQSPI_REG_READCAPTURE_BYPASS_LSB);
1013 
1014 	reg &= ~(CQSPI_REG_READCAPTURE_DELAY_MASK
1015 		 << CQSPI_REG_READCAPTURE_DELAY_LSB);
1016 
1017 	reg |= (delay & CQSPI_REG_READCAPTURE_DELAY_MASK)
1018 		<< CQSPI_REG_READCAPTURE_DELAY_LSB;
1019 
1020 	writel(reg, reg_base + CQSPI_REG_READCAPTURE);
1021 }
1022 
1023 static void cqspi_controller_enable(struct cqspi_st *cqspi, bool enable)
1024 {
1025 	void __iomem *reg_base = cqspi->iobase;
1026 	unsigned int reg;
1027 
1028 	reg = readl(reg_base + CQSPI_REG_CONFIG);
1029 
1030 	if (enable)
1031 		reg |= CQSPI_REG_CONFIG_ENABLE_MASK;
1032 	else
1033 		reg &= ~CQSPI_REG_CONFIG_ENABLE_MASK;
1034 
1035 	writel(reg, reg_base + CQSPI_REG_CONFIG);
1036 }
1037 
1038 static void cqspi_configure(struct cqspi_flash_pdata *f_pdata,
1039 			    unsigned long sclk)
1040 {
1041 	struct cqspi_st *cqspi = f_pdata->cqspi;
1042 	int switch_cs = (cqspi->current_cs != f_pdata->cs);
1043 	int switch_ck = (cqspi->sclk != sclk);
1044 
1045 	if (switch_cs || switch_ck)
1046 		cqspi_controller_enable(cqspi, 0);
1047 
1048 	/* Switch chip select. */
1049 	if (switch_cs) {
1050 		cqspi->current_cs = f_pdata->cs;
1051 		cqspi_chipselect(f_pdata);
1052 	}
1053 
1054 	/* Setup baudrate divisor and delays */
1055 	if (switch_ck) {
1056 		cqspi->sclk = sclk;
1057 		cqspi_config_baudrate_div(cqspi);
1058 		cqspi_delay(f_pdata);
1059 		cqspi_readdata_capture(cqspi, !cqspi->rclk_en,
1060 				       f_pdata->read_delay);
1061 	}
1062 
1063 	if (switch_cs || switch_ck)
1064 		cqspi_controller_enable(cqspi, 1);
1065 }
1066 
1067 static ssize_t cqspi_write(struct cqspi_flash_pdata *f_pdata,
1068 			   const struct spi_mem_op *op)
1069 {
1070 	struct cqspi_st *cqspi = f_pdata->cqspi;
1071 	loff_t to = op->addr.val;
1072 	size_t len = op->data.nbytes;
1073 	const u_char *buf = op->data.buf.out;
1074 	int ret;
1075 
1076 	ret = cqspi_set_protocol(f_pdata, op);
1077 	if (ret)
1078 		return ret;
1079 
1080 	ret = cqspi_write_setup(f_pdata, op);
1081 	if (ret)
1082 		return ret;
1083 
1084 	/*
1085 	 * Some flashes like the Cypress Semper flash expect a dummy 4-byte
1086 	 * address (all 0s) with the read status register command in DTR mode.
1087 	 * But this controller does not support sending dummy address bytes to
1088 	 * the flash when it is polling the write completion register in DTR
1089 	 * mode. So, we can not use direct mode when in DTR mode for writing
1090 	 * data.
1091 	 */
1092 	if (!f_pdata->dtr && cqspi->use_direct_mode &&
1093 	    ((to + len) <= cqspi->ahb_size)) {
1094 		memcpy_toio(cqspi->ahb_base + to, buf, len);
1095 		return cqspi_wait_idle(cqspi);
1096 	}
1097 
1098 	return cqspi_indirect_write_execute(f_pdata, to, buf, len);
1099 }
1100 
1101 static void cqspi_rx_dma_callback(void *param)
1102 {
1103 	struct cqspi_st *cqspi = param;
1104 
1105 	complete(&cqspi->rx_dma_complete);
1106 }
1107 
1108 static int cqspi_direct_read_execute(struct cqspi_flash_pdata *f_pdata,
1109 				     u_char *buf, loff_t from, size_t len)
1110 {
1111 	struct cqspi_st *cqspi = f_pdata->cqspi;
1112 	struct device *dev = &cqspi->pdev->dev;
1113 	enum dma_ctrl_flags flags = DMA_CTRL_ACK | DMA_PREP_INTERRUPT;
1114 	dma_addr_t dma_src = (dma_addr_t)cqspi->mmap_phys_base + from;
1115 	int ret = 0;
1116 	struct dma_async_tx_descriptor *tx;
1117 	dma_cookie_t cookie;
1118 	dma_addr_t dma_dst;
1119 	struct device *ddev;
1120 
1121 	if (!cqspi->rx_chan || !virt_addr_valid(buf)) {
1122 		memcpy_fromio(buf, cqspi->ahb_base + from, len);
1123 		return 0;
1124 	}
1125 
1126 	ddev = cqspi->rx_chan->device->dev;
1127 	dma_dst = dma_map_single(ddev, buf, len, DMA_FROM_DEVICE);
1128 	if (dma_mapping_error(ddev, dma_dst)) {
1129 		dev_err(dev, "dma mapping failed\n");
1130 		return -ENOMEM;
1131 	}
1132 	tx = dmaengine_prep_dma_memcpy(cqspi->rx_chan, dma_dst, dma_src,
1133 				       len, flags);
1134 	if (!tx) {
1135 		dev_err(dev, "device_prep_dma_memcpy error\n");
1136 		ret = -EIO;
1137 		goto err_unmap;
1138 	}
1139 
1140 	tx->callback = cqspi_rx_dma_callback;
1141 	tx->callback_param = cqspi;
1142 	cookie = tx->tx_submit(tx);
1143 	reinit_completion(&cqspi->rx_dma_complete);
1144 
1145 	ret = dma_submit_error(cookie);
1146 	if (ret) {
1147 		dev_err(dev, "dma_submit_error %d\n", cookie);
1148 		ret = -EIO;
1149 		goto err_unmap;
1150 	}
1151 
1152 	dma_async_issue_pending(cqspi->rx_chan);
1153 	if (!wait_for_completion_timeout(&cqspi->rx_dma_complete,
1154 					 msecs_to_jiffies(max_t(size_t, len, 500)))) {
1155 		dmaengine_terminate_sync(cqspi->rx_chan);
1156 		dev_err(dev, "DMA wait_for_completion_timeout\n");
1157 		ret = -ETIMEDOUT;
1158 		goto err_unmap;
1159 	}
1160 
1161 err_unmap:
1162 	dma_unmap_single(ddev, dma_dst, len, DMA_FROM_DEVICE);
1163 
1164 	return ret;
1165 }
1166 
1167 static ssize_t cqspi_read(struct cqspi_flash_pdata *f_pdata,
1168 			  const struct spi_mem_op *op)
1169 {
1170 	struct cqspi_st *cqspi = f_pdata->cqspi;
1171 	loff_t from = op->addr.val;
1172 	size_t len = op->data.nbytes;
1173 	u_char *buf = op->data.buf.in;
1174 	int ret;
1175 
1176 	ret = cqspi_set_protocol(f_pdata, op);
1177 	if (ret)
1178 		return ret;
1179 
1180 	ret = cqspi_read_setup(f_pdata, op);
1181 	if (ret)
1182 		return ret;
1183 
1184 	if (cqspi->use_direct_mode && ((from + len) <= cqspi->ahb_size))
1185 		return cqspi_direct_read_execute(f_pdata, buf, from, len);
1186 
1187 	return cqspi_indirect_read_execute(f_pdata, buf, from, len);
1188 }
1189 
1190 static int cqspi_mem_process(struct spi_mem *mem, const struct spi_mem_op *op)
1191 {
1192 	struct cqspi_st *cqspi = spi_master_get_devdata(mem->spi->master);
1193 	struct cqspi_flash_pdata *f_pdata;
1194 
1195 	f_pdata = &cqspi->f_pdata[mem->spi->chip_select];
1196 	cqspi_configure(f_pdata, mem->spi->max_speed_hz);
1197 
1198 	if (op->data.dir == SPI_MEM_DATA_IN && op->data.buf.in) {
1199 		if (!op->addr.nbytes)
1200 			return cqspi_command_read(f_pdata, op);
1201 
1202 		return cqspi_read(f_pdata, op);
1203 	}
1204 
1205 	if (!op->addr.nbytes || !op->data.buf.out)
1206 		return cqspi_command_write(f_pdata, op);
1207 
1208 	return cqspi_write(f_pdata, op);
1209 }
1210 
1211 static int cqspi_exec_mem_op(struct spi_mem *mem, const struct spi_mem_op *op)
1212 {
1213 	int ret;
1214 
1215 	ret = cqspi_mem_process(mem, op);
1216 	if (ret)
1217 		dev_err(&mem->spi->dev, "operation failed with %d\n", ret);
1218 
1219 	return ret;
1220 }
1221 
1222 static bool cqspi_supports_mem_op(struct spi_mem *mem,
1223 				  const struct spi_mem_op *op)
1224 {
1225 	bool all_true, all_false;
1226 
1227 	all_true = op->cmd.dtr && op->addr.dtr && op->dummy.dtr &&
1228 		   op->data.dtr;
1229 	all_false = !op->cmd.dtr && !op->addr.dtr && !op->dummy.dtr &&
1230 		    !op->data.dtr;
1231 
1232 	/* Mixed DTR modes not supported. */
1233 	if (!(all_true || all_false))
1234 		return false;
1235 
1236 	if (all_true)
1237 		return spi_mem_dtr_supports_op(mem, op);
1238 	else
1239 		return spi_mem_default_supports_op(mem, op);
1240 }
1241 
1242 static int cqspi_of_get_flash_pdata(struct platform_device *pdev,
1243 				    struct cqspi_flash_pdata *f_pdata,
1244 				    struct device_node *np)
1245 {
1246 	if (of_property_read_u32(np, "cdns,read-delay", &f_pdata->read_delay)) {
1247 		dev_err(&pdev->dev, "couldn't determine read-delay\n");
1248 		return -ENXIO;
1249 	}
1250 
1251 	if (of_property_read_u32(np, "cdns,tshsl-ns", &f_pdata->tshsl_ns)) {
1252 		dev_err(&pdev->dev, "couldn't determine tshsl-ns\n");
1253 		return -ENXIO;
1254 	}
1255 
1256 	if (of_property_read_u32(np, "cdns,tsd2d-ns", &f_pdata->tsd2d_ns)) {
1257 		dev_err(&pdev->dev, "couldn't determine tsd2d-ns\n");
1258 		return -ENXIO;
1259 	}
1260 
1261 	if (of_property_read_u32(np, "cdns,tchsh-ns", &f_pdata->tchsh_ns)) {
1262 		dev_err(&pdev->dev, "couldn't determine tchsh-ns\n");
1263 		return -ENXIO;
1264 	}
1265 
1266 	if (of_property_read_u32(np, "cdns,tslch-ns", &f_pdata->tslch_ns)) {
1267 		dev_err(&pdev->dev, "couldn't determine tslch-ns\n");
1268 		return -ENXIO;
1269 	}
1270 
1271 	if (of_property_read_u32(np, "spi-max-frequency", &f_pdata->clk_rate)) {
1272 		dev_err(&pdev->dev, "couldn't determine spi-max-frequency\n");
1273 		return -ENXIO;
1274 	}
1275 
1276 	return 0;
1277 }
1278 
1279 static int cqspi_of_get_pdata(struct cqspi_st *cqspi)
1280 {
1281 	struct device *dev = &cqspi->pdev->dev;
1282 	struct device_node *np = dev->of_node;
1283 
1284 	cqspi->is_decoded_cs = of_property_read_bool(np, "cdns,is-decoded-cs");
1285 
1286 	if (of_property_read_u32(np, "cdns,fifo-depth", &cqspi->fifo_depth)) {
1287 		dev_err(dev, "couldn't determine fifo-depth\n");
1288 		return -ENXIO;
1289 	}
1290 
1291 	if (of_property_read_u32(np, "cdns,fifo-width", &cqspi->fifo_width)) {
1292 		dev_err(dev, "couldn't determine fifo-width\n");
1293 		return -ENXIO;
1294 	}
1295 
1296 	if (of_property_read_u32(np, "cdns,trigger-address",
1297 				 &cqspi->trigger_address)) {
1298 		dev_err(dev, "couldn't determine trigger-address\n");
1299 		return -ENXIO;
1300 	}
1301 
1302 	if (of_property_read_u32(np, "num-cs", &cqspi->num_chipselect))
1303 		cqspi->num_chipselect = CQSPI_MAX_CHIPSELECT;
1304 
1305 	cqspi->rclk_en = of_property_read_bool(np, "cdns,rclk-en");
1306 
1307 	return 0;
1308 }
1309 
1310 static void cqspi_controller_init(struct cqspi_st *cqspi)
1311 {
1312 	u32 reg;
1313 
1314 	cqspi_controller_enable(cqspi, 0);
1315 
1316 	/* Configure the remap address register, no remap */
1317 	writel(0, cqspi->iobase + CQSPI_REG_REMAP);
1318 
1319 	/* Disable all interrupts. */
1320 	writel(0, cqspi->iobase + CQSPI_REG_IRQMASK);
1321 
1322 	/* Configure the SRAM split to 1:1 . */
1323 	writel(cqspi->fifo_depth / 2, cqspi->iobase + CQSPI_REG_SRAMPARTITION);
1324 
1325 	/* Load indirect trigger address. */
1326 	writel(cqspi->trigger_address,
1327 	       cqspi->iobase + CQSPI_REG_INDIRECTTRIGGER);
1328 
1329 	/* Program read watermark -- 1/2 of the FIFO. */
1330 	writel(cqspi->fifo_depth * cqspi->fifo_width / 2,
1331 	       cqspi->iobase + CQSPI_REG_INDIRECTRDWATERMARK);
1332 	/* Program write watermark -- 1/8 of the FIFO. */
1333 	writel(cqspi->fifo_depth * cqspi->fifo_width / 8,
1334 	       cqspi->iobase + CQSPI_REG_INDIRECTWRWATERMARK);
1335 
1336 	/* Disable direct access controller */
1337 	if (!cqspi->use_direct_mode) {
1338 		reg = readl(cqspi->iobase + CQSPI_REG_CONFIG);
1339 		reg &= ~CQSPI_REG_CONFIG_ENB_DIR_ACC_CTRL;
1340 		writel(reg, cqspi->iobase + CQSPI_REG_CONFIG);
1341 	}
1342 
1343 	cqspi_controller_enable(cqspi, 1);
1344 }
1345 
1346 static int cqspi_request_mmap_dma(struct cqspi_st *cqspi)
1347 {
1348 	dma_cap_mask_t mask;
1349 
1350 	dma_cap_zero(mask);
1351 	dma_cap_set(DMA_MEMCPY, mask);
1352 
1353 	cqspi->rx_chan = dma_request_chan_by_mask(&mask);
1354 	if (IS_ERR(cqspi->rx_chan)) {
1355 		int ret = PTR_ERR(cqspi->rx_chan);
1356 		cqspi->rx_chan = NULL;
1357 		return dev_err_probe(&cqspi->pdev->dev, ret, "No Rx DMA available\n");
1358 	}
1359 	init_completion(&cqspi->rx_dma_complete);
1360 
1361 	return 0;
1362 }
1363 
1364 static const char *cqspi_get_name(struct spi_mem *mem)
1365 {
1366 	struct cqspi_st *cqspi = spi_master_get_devdata(mem->spi->master);
1367 	struct device *dev = &cqspi->pdev->dev;
1368 
1369 	return devm_kasprintf(dev, GFP_KERNEL, "%s.%d", dev_name(dev), mem->spi->chip_select);
1370 }
1371 
1372 static const struct spi_controller_mem_ops cqspi_mem_ops = {
1373 	.exec_op = cqspi_exec_mem_op,
1374 	.get_name = cqspi_get_name,
1375 	.supports_op = cqspi_supports_mem_op,
1376 };
1377 
1378 static int cqspi_setup_flash(struct cqspi_st *cqspi)
1379 {
1380 	struct platform_device *pdev = cqspi->pdev;
1381 	struct device *dev = &pdev->dev;
1382 	struct device_node *np = dev->of_node;
1383 	struct cqspi_flash_pdata *f_pdata;
1384 	unsigned int cs;
1385 	int ret;
1386 
1387 	/* Get flash device data */
1388 	for_each_available_child_of_node(dev->of_node, np) {
1389 		ret = of_property_read_u32(np, "reg", &cs);
1390 		if (ret) {
1391 			dev_err(dev, "Couldn't determine chip select.\n");
1392 			return ret;
1393 		}
1394 
1395 		if (cs >= CQSPI_MAX_CHIPSELECT) {
1396 			dev_err(dev, "Chip select %d out of range.\n", cs);
1397 			return -EINVAL;
1398 		}
1399 
1400 		f_pdata = &cqspi->f_pdata[cs];
1401 		f_pdata->cqspi = cqspi;
1402 		f_pdata->cs = cs;
1403 
1404 		ret = cqspi_of_get_flash_pdata(pdev, f_pdata, np);
1405 		if (ret)
1406 			return ret;
1407 	}
1408 
1409 	return 0;
1410 }
1411 
1412 static int cqspi_probe(struct platform_device *pdev)
1413 {
1414 	const struct cqspi_driver_platdata *ddata;
1415 	struct reset_control *rstc, *rstc_ocp;
1416 	struct device *dev = &pdev->dev;
1417 	struct spi_master *master;
1418 	struct resource *res_ahb;
1419 	struct cqspi_st *cqspi;
1420 	struct resource *res;
1421 	int ret;
1422 	int irq;
1423 
1424 	master = spi_alloc_master(&pdev->dev, sizeof(*cqspi));
1425 	if (!master) {
1426 		dev_err(&pdev->dev, "spi_alloc_master failed\n");
1427 		return -ENOMEM;
1428 	}
1429 	master->mode_bits = SPI_RX_QUAD | SPI_RX_DUAL;
1430 	master->mem_ops = &cqspi_mem_ops;
1431 	master->dev.of_node = pdev->dev.of_node;
1432 
1433 	cqspi = spi_master_get_devdata(master);
1434 
1435 	cqspi->pdev = pdev;
1436 	platform_set_drvdata(pdev, cqspi);
1437 
1438 	/* Obtain configuration from OF. */
1439 	ret = cqspi_of_get_pdata(cqspi);
1440 	if (ret) {
1441 		dev_err(dev, "Cannot get mandatory OF data.\n");
1442 		ret = -ENODEV;
1443 		goto probe_master_put;
1444 	}
1445 
1446 	/* Obtain QSPI clock. */
1447 	cqspi->clk = devm_clk_get(dev, NULL);
1448 	if (IS_ERR(cqspi->clk)) {
1449 		dev_err(dev, "Cannot claim QSPI clock.\n");
1450 		ret = PTR_ERR(cqspi->clk);
1451 		goto probe_master_put;
1452 	}
1453 
1454 	/* Obtain and remap controller address. */
1455 	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
1456 	cqspi->iobase = devm_ioremap_resource(dev, res);
1457 	if (IS_ERR(cqspi->iobase)) {
1458 		dev_err(dev, "Cannot remap controller address.\n");
1459 		ret = PTR_ERR(cqspi->iobase);
1460 		goto probe_master_put;
1461 	}
1462 
1463 	/* Obtain and remap AHB address. */
1464 	res_ahb = platform_get_resource(pdev, IORESOURCE_MEM, 1);
1465 	cqspi->ahb_base = devm_ioremap_resource(dev, res_ahb);
1466 	if (IS_ERR(cqspi->ahb_base)) {
1467 		dev_err(dev, "Cannot remap AHB address.\n");
1468 		ret = PTR_ERR(cqspi->ahb_base);
1469 		goto probe_master_put;
1470 	}
1471 	cqspi->mmap_phys_base = (dma_addr_t)res_ahb->start;
1472 	cqspi->ahb_size = resource_size(res_ahb);
1473 
1474 	init_completion(&cqspi->transfer_complete);
1475 
1476 	/* Obtain IRQ line. */
1477 	irq = platform_get_irq(pdev, 0);
1478 	if (irq < 0) {
1479 		ret = -ENXIO;
1480 		goto probe_master_put;
1481 	}
1482 
1483 	pm_runtime_enable(dev);
1484 	ret = pm_runtime_get_sync(dev);
1485 	if (ret < 0) {
1486 		pm_runtime_put_noidle(dev);
1487 		goto probe_master_put;
1488 	}
1489 
1490 	ret = clk_prepare_enable(cqspi->clk);
1491 	if (ret) {
1492 		dev_err(dev, "Cannot enable QSPI clock.\n");
1493 		goto probe_clk_failed;
1494 	}
1495 
1496 	/* Obtain QSPI reset control */
1497 	rstc = devm_reset_control_get_optional_exclusive(dev, "qspi");
1498 	if (IS_ERR(rstc)) {
1499 		ret = PTR_ERR(rstc);
1500 		dev_err(dev, "Cannot get QSPI reset.\n");
1501 		goto probe_reset_failed;
1502 	}
1503 
1504 	rstc_ocp = devm_reset_control_get_optional_exclusive(dev, "qspi-ocp");
1505 	if (IS_ERR(rstc_ocp)) {
1506 		ret = PTR_ERR(rstc_ocp);
1507 		dev_err(dev, "Cannot get QSPI OCP reset.\n");
1508 		goto probe_reset_failed;
1509 	}
1510 
1511 	reset_control_assert(rstc);
1512 	reset_control_deassert(rstc);
1513 
1514 	reset_control_assert(rstc_ocp);
1515 	reset_control_deassert(rstc_ocp);
1516 
1517 	cqspi->master_ref_clk_hz = clk_get_rate(cqspi->clk);
1518 	master->max_speed_hz = cqspi->master_ref_clk_hz;
1519 	ddata  = of_device_get_match_data(dev);
1520 	if (ddata) {
1521 		if (ddata->quirks & CQSPI_NEEDS_WR_DELAY)
1522 			cqspi->wr_delay = 50 * DIV_ROUND_UP(NSEC_PER_SEC,
1523 						cqspi->master_ref_clk_hz);
1524 		if (ddata->hwcaps_mask & CQSPI_SUPPORTS_OCTAL)
1525 			master->mode_bits |= SPI_RX_OCTAL | SPI_TX_OCTAL;
1526 		if (!(ddata->quirks & CQSPI_DISABLE_DAC_MODE))
1527 			cqspi->use_direct_mode = true;
1528 	}
1529 
1530 	ret = devm_request_irq(dev, irq, cqspi_irq_handler, 0,
1531 			       pdev->name, cqspi);
1532 	if (ret) {
1533 		dev_err(dev, "Cannot request IRQ.\n");
1534 		goto probe_reset_failed;
1535 	}
1536 
1537 	cqspi_wait_idle(cqspi);
1538 	cqspi_controller_init(cqspi);
1539 	cqspi->current_cs = -1;
1540 	cqspi->sclk = 0;
1541 
1542 	master->num_chipselect = cqspi->num_chipselect;
1543 
1544 	ret = cqspi_setup_flash(cqspi);
1545 	if (ret) {
1546 		dev_err(dev, "failed to setup flash parameters %d\n", ret);
1547 		goto probe_setup_failed;
1548 	}
1549 
1550 	if (cqspi->use_direct_mode) {
1551 		ret = cqspi_request_mmap_dma(cqspi);
1552 		if (ret == -EPROBE_DEFER)
1553 			goto probe_setup_failed;
1554 	}
1555 
1556 	ret = devm_spi_register_master(dev, master);
1557 	if (ret) {
1558 		dev_err(&pdev->dev, "failed to register SPI ctlr %d\n", ret);
1559 		goto probe_setup_failed;
1560 	}
1561 
1562 	return 0;
1563 probe_setup_failed:
1564 	cqspi_controller_enable(cqspi, 0);
1565 probe_reset_failed:
1566 	clk_disable_unprepare(cqspi->clk);
1567 probe_clk_failed:
1568 	pm_runtime_put_sync(dev);
1569 	pm_runtime_disable(dev);
1570 probe_master_put:
1571 	spi_master_put(master);
1572 	return ret;
1573 }
1574 
1575 static int cqspi_remove(struct platform_device *pdev)
1576 {
1577 	struct cqspi_st *cqspi = platform_get_drvdata(pdev);
1578 
1579 	cqspi_controller_enable(cqspi, 0);
1580 
1581 	if (cqspi->rx_chan)
1582 		dma_release_channel(cqspi->rx_chan);
1583 
1584 	clk_disable_unprepare(cqspi->clk);
1585 
1586 	pm_runtime_put_sync(&pdev->dev);
1587 	pm_runtime_disable(&pdev->dev);
1588 
1589 	return 0;
1590 }
1591 
1592 #ifdef CONFIG_PM_SLEEP
1593 static int cqspi_suspend(struct device *dev)
1594 {
1595 	struct cqspi_st *cqspi = dev_get_drvdata(dev);
1596 
1597 	cqspi_controller_enable(cqspi, 0);
1598 	return 0;
1599 }
1600 
1601 static int cqspi_resume(struct device *dev)
1602 {
1603 	struct cqspi_st *cqspi = dev_get_drvdata(dev);
1604 
1605 	cqspi_controller_enable(cqspi, 1);
1606 	return 0;
1607 }
1608 
1609 static const struct dev_pm_ops cqspi__dev_pm_ops = {
1610 	.suspend = cqspi_suspend,
1611 	.resume = cqspi_resume,
1612 };
1613 
1614 #define CQSPI_DEV_PM_OPS	(&cqspi__dev_pm_ops)
1615 #else
1616 #define CQSPI_DEV_PM_OPS	NULL
1617 #endif
1618 
1619 static const struct cqspi_driver_platdata cdns_qspi = {
1620 	.quirks = CQSPI_DISABLE_DAC_MODE,
1621 };
1622 
1623 static const struct cqspi_driver_platdata k2g_qspi = {
1624 	.quirks = CQSPI_NEEDS_WR_DELAY,
1625 };
1626 
1627 static const struct cqspi_driver_platdata am654_ospi = {
1628 	.hwcaps_mask = CQSPI_SUPPORTS_OCTAL,
1629 	.quirks = CQSPI_NEEDS_WR_DELAY,
1630 };
1631 
1632 static const struct cqspi_driver_platdata intel_lgm_qspi = {
1633 	.quirks = CQSPI_DISABLE_DAC_MODE,
1634 };
1635 
1636 static const struct of_device_id cqspi_dt_ids[] = {
1637 	{
1638 		.compatible = "cdns,qspi-nor",
1639 		.data = &cdns_qspi,
1640 	},
1641 	{
1642 		.compatible = "ti,k2g-qspi",
1643 		.data = &k2g_qspi,
1644 	},
1645 	{
1646 		.compatible = "ti,am654-ospi",
1647 		.data = &am654_ospi,
1648 	},
1649 	{
1650 		.compatible = "intel,lgm-qspi",
1651 		.data = &intel_lgm_qspi,
1652 	},
1653 	{ /* end of table */ }
1654 };
1655 
1656 MODULE_DEVICE_TABLE(of, cqspi_dt_ids);
1657 
1658 static struct platform_driver cqspi_platform_driver = {
1659 	.probe = cqspi_probe,
1660 	.remove = cqspi_remove,
1661 	.driver = {
1662 		.name = CQSPI_NAME,
1663 		.pm = CQSPI_DEV_PM_OPS,
1664 		.of_match_table = cqspi_dt_ids,
1665 	},
1666 };
1667 
1668 module_platform_driver(cqspi_platform_driver);
1669 
1670 MODULE_DESCRIPTION("Cadence QSPI Controller Driver");
1671 MODULE_LICENSE("GPL v2");
1672 MODULE_ALIAS("platform:" CQSPI_NAME);
1673 MODULE_AUTHOR("Ley Foon Tan <lftan@altera.com>");
1674 MODULE_AUTHOR("Graham Moore <grmoore@opensource.altera.com>");
1675 MODULE_AUTHOR("Vadivel Murugan R <vadivel.muruganx.ramuthevar@intel.com>");
1676 MODULE_AUTHOR("Vignesh Raghavendra <vigneshr@ti.com>");
1677 MODULE_AUTHOR("Pratyush Yadav <p.yadav@ti.com>");
1678