xref: /linux/drivers/spi/spi-cadence-quadspi.c (revision 6d9b262afe0ec1d6e0ef99321ca9d6b921310471)
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/firmware/xlnx-zynqmp.h>
17 #include <linux/interrupt.h>
18 #include <linux/io.h>
19 #include <linux/iopoll.h>
20 #include <linux/jiffies.h>
21 #include <linux/kernel.h>
22 #include <linux/log2.h>
23 #include <linux/module.h>
24 #include <linux/of.h>
25 #include <linux/platform_device.h>
26 #include <linux/pm_runtime.h>
27 #include <linux/reset.h>
28 #include <linux/sched.h>
29 #include <linux/spi/spi.h>
30 #include <linux/spi/spi-mem.h>
31 #include <linux/timer.h>
32 
33 #define CQSPI_NAME			"cadence-qspi"
34 #define CQSPI_MAX_CHIPSELECT		4
35 
36 static_assert(CQSPI_MAX_CHIPSELECT <= SPI_CS_CNT_MAX);
37 
38 /* Quirks */
39 #define CQSPI_NEEDS_WR_DELAY		BIT(0)
40 #define CQSPI_DISABLE_DAC_MODE		BIT(1)
41 #define CQSPI_SUPPORT_EXTERNAL_DMA	BIT(2)
42 #define CQSPI_NO_SUPPORT_WR_COMPLETION	BIT(3)
43 #define CQSPI_SLOW_SRAM		BIT(4)
44 #define CQSPI_NEEDS_APB_AHB_HAZARD_WAR	BIT(5)
45 
46 /* Capabilities */
47 #define CQSPI_SUPPORTS_OCTAL		BIT(0)
48 
49 #define CQSPI_OP_WIDTH(part) ((part).nbytes ? ilog2((part).buswidth) : 0)
50 
51 enum {
52 	CLK_QSPI_APB = 0,
53 	CLK_QSPI_AHB,
54 	CLK_QSPI_NUM,
55 };
56 
57 struct cqspi_st;
58 
59 struct cqspi_flash_pdata {
60 	struct cqspi_st	*cqspi;
61 	u32		clk_rate;
62 	u32		read_delay;
63 	u32		tshsl_ns;
64 	u32		tsd2d_ns;
65 	u32		tchsh_ns;
66 	u32		tslch_ns;
67 	u8		cs;
68 };
69 
70 struct cqspi_st {
71 	struct platform_device	*pdev;
72 	struct spi_controller	*host;
73 	struct clk		*clk;
74 	struct clk		*clks[CLK_QSPI_NUM];
75 	unsigned int		sclk;
76 
77 	void __iomem		*iobase;
78 	void __iomem		*ahb_base;
79 	resource_size_t		ahb_size;
80 	struct completion	transfer_complete;
81 
82 	struct dma_chan		*rx_chan;
83 	struct completion	rx_dma_complete;
84 	dma_addr_t		mmap_phys_base;
85 
86 	int			current_cs;
87 	unsigned long		master_ref_clk_hz;
88 	bool			is_decoded_cs;
89 	u32			fifo_depth;
90 	u32			fifo_width;
91 	u32			num_chipselect;
92 	bool			rclk_en;
93 	u32			trigger_address;
94 	u32			wr_delay;
95 	bool			use_direct_mode;
96 	bool			use_direct_mode_wr;
97 	struct cqspi_flash_pdata f_pdata[CQSPI_MAX_CHIPSELECT];
98 	bool			use_dma_read;
99 	u32			pd_dev_id;
100 	bool			wr_completion;
101 	bool			slow_sram;
102 	bool			apb_ahb_hazard;
103 
104 	bool			is_jh7110; /* Flag for StarFive JH7110 SoC */
105 };
106 
107 struct cqspi_driver_platdata {
108 	u32 hwcaps_mask;
109 	u8 quirks;
110 	int (*indirect_read_dma)(struct cqspi_flash_pdata *f_pdata,
111 				 u_char *rxbuf, loff_t from_addr, size_t n_rx);
112 	u32 (*get_dma_status)(struct cqspi_st *cqspi);
113 	int (*jh7110_clk_init)(struct platform_device *pdev,
114 			       struct cqspi_st *cqspi);
115 };
116 
117 /* Operation timeout value */
118 #define CQSPI_TIMEOUT_MS			500
119 #define CQSPI_READ_TIMEOUT_MS			10
120 
121 /* Runtime_pm autosuspend delay */
122 #define CQSPI_AUTOSUSPEND_TIMEOUT		2000
123 
124 #define CQSPI_DUMMY_CLKS_PER_BYTE		8
125 #define CQSPI_DUMMY_BYTES_MAX			4
126 #define CQSPI_DUMMY_CLKS_MAX			31
127 
128 #define CQSPI_STIG_DATA_LEN_MAX			8
129 
130 /* Register map */
131 #define CQSPI_REG_CONFIG			0x00
132 #define CQSPI_REG_CONFIG_ENABLE_MASK		BIT(0)
133 #define CQSPI_REG_CONFIG_ENB_DIR_ACC_CTRL	BIT(7)
134 #define CQSPI_REG_CONFIG_DECODE_MASK		BIT(9)
135 #define CQSPI_REG_CONFIG_CHIPSELECT_LSB		10
136 #define CQSPI_REG_CONFIG_DMA_MASK		BIT(15)
137 #define CQSPI_REG_CONFIG_BAUD_LSB		19
138 #define CQSPI_REG_CONFIG_DTR_PROTO		BIT(24)
139 #define CQSPI_REG_CONFIG_DUAL_OPCODE		BIT(30)
140 #define CQSPI_REG_CONFIG_IDLE_LSB		31
141 #define CQSPI_REG_CONFIG_CHIPSELECT_MASK	0xF
142 #define CQSPI_REG_CONFIG_BAUD_MASK		0xF
143 
144 #define CQSPI_REG_RD_INSTR			0x04
145 #define CQSPI_REG_RD_INSTR_OPCODE_LSB		0
146 #define CQSPI_REG_RD_INSTR_TYPE_INSTR_LSB	8
147 #define CQSPI_REG_RD_INSTR_TYPE_ADDR_LSB	12
148 #define CQSPI_REG_RD_INSTR_TYPE_DATA_LSB	16
149 #define CQSPI_REG_RD_INSTR_MODE_EN_LSB		20
150 #define CQSPI_REG_RD_INSTR_DUMMY_LSB		24
151 #define CQSPI_REG_RD_INSTR_TYPE_INSTR_MASK	0x3
152 #define CQSPI_REG_RD_INSTR_TYPE_ADDR_MASK	0x3
153 #define CQSPI_REG_RD_INSTR_TYPE_DATA_MASK	0x3
154 #define CQSPI_REG_RD_INSTR_DUMMY_MASK		0x1F
155 
156 #define CQSPI_REG_WR_INSTR			0x08
157 #define CQSPI_REG_WR_INSTR_OPCODE_LSB		0
158 #define CQSPI_REG_WR_INSTR_TYPE_ADDR_LSB	12
159 #define CQSPI_REG_WR_INSTR_TYPE_DATA_LSB	16
160 
161 #define CQSPI_REG_DELAY				0x0C
162 #define CQSPI_REG_DELAY_TSLCH_LSB		0
163 #define CQSPI_REG_DELAY_TCHSH_LSB		8
164 #define CQSPI_REG_DELAY_TSD2D_LSB		16
165 #define CQSPI_REG_DELAY_TSHSL_LSB		24
166 #define CQSPI_REG_DELAY_TSLCH_MASK		0xFF
167 #define CQSPI_REG_DELAY_TCHSH_MASK		0xFF
168 #define CQSPI_REG_DELAY_TSD2D_MASK		0xFF
169 #define CQSPI_REG_DELAY_TSHSL_MASK		0xFF
170 
171 #define CQSPI_REG_READCAPTURE			0x10
172 #define CQSPI_REG_READCAPTURE_BYPASS_LSB	0
173 #define CQSPI_REG_READCAPTURE_DELAY_LSB		1
174 #define CQSPI_REG_READCAPTURE_DELAY_MASK	0xF
175 
176 #define CQSPI_REG_SIZE				0x14
177 #define CQSPI_REG_SIZE_ADDRESS_LSB		0
178 #define CQSPI_REG_SIZE_PAGE_LSB			4
179 #define CQSPI_REG_SIZE_BLOCK_LSB		16
180 #define CQSPI_REG_SIZE_ADDRESS_MASK		0xF
181 #define CQSPI_REG_SIZE_PAGE_MASK		0xFFF
182 #define CQSPI_REG_SIZE_BLOCK_MASK		0x3F
183 
184 #define CQSPI_REG_SRAMPARTITION			0x18
185 #define CQSPI_REG_INDIRECTTRIGGER		0x1C
186 
187 #define CQSPI_REG_DMA				0x20
188 #define CQSPI_REG_DMA_SINGLE_LSB		0
189 #define CQSPI_REG_DMA_BURST_LSB			8
190 #define CQSPI_REG_DMA_SINGLE_MASK		0xFF
191 #define CQSPI_REG_DMA_BURST_MASK		0xFF
192 
193 #define CQSPI_REG_REMAP				0x24
194 #define CQSPI_REG_MODE_BIT			0x28
195 
196 #define CQSPI_REG_SDRAMLEVEL			0x2C
197 #define CQSPI_REG_SDRAMLEVEL_RD_LSB		0
198 #define CQSPI_REG_SDRAMLEVEL_WR_LSB		16
199 #define CQSPI_REG_SDRAMLEVEL_RD_MASK		0xFFFF
200 #define CQSPI_REG_SDRAMLEVEL_WR_MASK		0xFFFF
201 
202 #define CQSPI_REG_WR_COMPLETION_CTRL		0x38
203 #define CQSPI_REG_WR_DISABLE_AUTO_POLL		BIT(14)
204 
205 #define CQSPI_REG_IRQSTATUS			0x40
206 #define CQSPI_REG_IRQMASK			0x44
207 
208 #define CQSPI_REG_INDIRECTRD			0x60
209 #define CQSPI_REG_INDIRECTRD_START_MASK		BIT(0)
210 #define CQSPI_REG_INDIRECTRD_CANCEL_MASK	BIT(1)
211 #define CQSPI_REG_INDIRECTRD_DONE_MASK		BIT(5)
212 
213 #define CQSPI_REG_INDIRECTRDWATERMARK		0x64
214 #define CQSPI_REG_INDIRECTRDSTARTADDR		0x68
215 #define CQSPI_REG_INDIRECTRDBYTES		0x6C
216 
217 #define CQSPI_REG_CMDCTRL			0x90
218 #define CQSPI_REG_CMDCTRL_EXECUTE_MASK		BIT(0)
219 #define CQSPI_REG_CMDCTRL_INPROGRESS_MASK	BIT(1)
220 #define CQSPI_REG_CMDCTRL_DUMMY_LSB		7
221 #define CQSPI_REG_CMDCTRL_WR_BYTES_LSB		12
222 #define CQSPI_REG_CMDCTRL_WR_EN_LSB		15
223 #define CQSPI_REG_CMDCTRL_ADD_BYTES_LSB		16
224 #define CQSPI_REG_CMDCTRL_ADDR_EN_LSB		19
225 #define CQSPI_REG_CMDCTRL_RD_BYTES_LSB		20
226 #define CQSPI_REG_CMDCTRL_RD_EN_LSB		23
227 #define CQSPI_REG_CMDCTRL_OPCODE_LSB		24
228 #define CQSPI_REG_CMDCTRL_WR_BYTES_MASK		0x7
229 #define CQSPI_REG_CMDCTRL_ADD_BYTES_MASK	0x3
230 #define CQSPI_REG_CMDCTRL_RD_BYTES_MASK		0x7
231 #define CQSPI_REG_CMDCTRL_DUMMY_MASK		0x1F
232 
233 #define CQSPI_REG_INDIRECTWR			0x70
234 #define CQSPI_REG_INDIRECTWR_START_MASK		BIT(0)
235 #define CQSPI_REG_INDIRECTWR_CANCEL_MASK	BIT(1)
236 #define CQSPI_REG_INDIRECTWR_DONE_MASK		BIT(5)
237 
238 #define CQSPI_REG_INDIRECTWRWATERMARK		0x74
239 #define CQSPI_REG_INDIRECTWRSTARTADDR		0x78
240 #define CQSPI_REG_INDIRECTWRBYTES		0x7C
241 
242 #define CQSPI_REG_INDTRIG_ADDRRANGE		0x80
243 
244 #define CQSPI_REG_CMDADDRESS			0x94
245 #define CQSPI_REG_CMDREADDATALOWER		0xA0
246 #define CQSPI_REG_CMDREADDATAUPPER		0xA4
247 #define CQSPI_REG_CMDWRITEDATALOWER		0xA8
248 #define CQSPI_REG_CMDWRITEDATAUPPER		0xAC
249 
250 #define CQSPI_REG_POLLING_STATUS		0xB0
251 #define CQSPI_REG_POLLING_STATUS_DUMMY_LSB	16
252 
253 #define CQSPI_REG_OP_EXT_LOWER			0xE0
254 #define CQSPI_REG_OP_EXT_READ_LSB		24
255 #define CQSPI_REG_OP_EXT_WRITE_LSB		16
256 #define CQSPI_REG_OP_EXT_STIG_LSB		0
257 
258 #define CQSPI_REG_VERSAL_DMA_SRC_ADDR		0x1000
259 
260 #define CQSPI_REG_VERSAL_DMA_DST_ADDR		0x1800
261 #define CQSPI_REG_VERSAL_DMA_DST_SIZE		0x1804
262 
263 #define CQSPI_REG_VERSAL_DMA_DST_CTRL		0x180C
264 
265 #define CQSPI_REG_VERSAL_DMA_DST_I_STS		0x1814
266 #define CQSPI_REG_VERSAL_DMA_DST_I_EN		0x1818
267 #define CQSPI_REG_VERSAL_DMA_DST_I_DIS		0x181C
268 #define CQSPI_REG_VERSAL_DMA_DST_DONE_MASK	BIT(1)
269 
270 #define CQSPI_REG_VERSAL_DMA_DST_ADDR_MSB	0x1828
271 
272 #define CQSPI_REG_VERSAL_DMA_DST_CTRL_VAL	0xF43FFA00
273 #define CQSPI_REG_VERSAL_ADDRRANGE_WIDTH_VAL	0x6
274 
275 /* Interrupt status bits */
276 #define CQSPI_REG_IRQ_MODE_ERR			BIT(0)
277 #define CQSPI_REG_IRQ_UNDERFLOW			BIT(1)
278 #define CQSPI_REG_IRQ_IND_COMP			BIT(2)
279 #define CQSPI_REG_IRQ_IND_RD_REJECT		BIT(3)
280 #define CQSPI_REG_IRQ_WR_PROTECTED_ERR		BIT(4)
281 #define CQSPI_REG_IRQ_ILLEGAL_AHB_ERR		BIT(5)
282 #define CQSPI_REG_IRQ_WATERMARK			BIT(6)
283 #define CQSPI_REG_IRQ_IND_SRAM_FULL		BIT(12)
284 
285 #define CQSPI_IRQ_MASK_RD		(CQSPI_REG_IRQ_WATERMARK	| \
286 					 CQSPI_REG_IRQ_IND_SRAM_FULL	| \
287 					 CQSPI_REG_IRQ_IND_COMP)
288 
289 #define CQSPI_IRQ_MASK_WR		(CQSPI_REG_IRQ_IND_COMP		| \
290 					 CQSPI_REG_IRQ_WATERMARK	| \
291 					 CQSPI_REG_IRQ_UNDERFLOW)
292 
293 #define CQSPI_IRQ_STATUS_MASK		0x1FFFF
294 #define CQSPI_DMA_UNALIGN		0x3
295 
296 #define CQSPI_REG_VERSAL_DMA_VAL		0x602
297 
298 static int cqspi_wait_for_bit(void __iomem *reg, const u32 mask, bool clr)
299 {
300 	u32 val;
301 
302 	return readl_relaxed_poll_timeout(reg, val,
303 					  (((clr ? ~val : val) & mask) == mask),
304 					  10, CQSPI_TIMEOUT_MS * 1000);
305 }
306 
307 static bool cqspi_is_idle(struct cqspi_st *cqspi)
308 {
309 	u32 reg = readl(cqspi->iobase + CQSPI_REG_CONFIG);
310 
311 	return reg & (1UL << CQSPI_REG_CONFIG_IDLE_LSB);
312 }
313 
314 static u32 cqspi_get_rd_sram_level(struct cqspi_st *cqspi)
315 {
316 	u32 reg = readl(cqspi->iobase + CQSPI_REG_SDRAMLEVEL);
317 
318 	reg >>= CQSPI_REG_SDRAMLEVEL_RD_LSB;
319 	return reg & CQSPI_REG_SDRAMLEVEL_RD_MASK;
320 }
321 
322 static u32 cqspi_get_versal_dma_status(struct cqspi_st *cqspi)
323 {
324 	u32 dma_status;
325 
326 	dma_status = readl(cqspi->iobase +
327 					   CQSPI_REG_VERSAL_DMA_DST_I_STS);
328 	writel(dma_status, cqspi->iobase +
329 		   CQSPI_REG_VERSAL_DMA_DST_I_STS);
330 
331 	return dma_status & CQSPI_REG_VERSAL_DMA_DST_DONE_MASK;
332 }
333 
334 static irqreturn_t cqspi_irq_handler(int this_irq, void *dev)
335 {
336 	struct cqspi_st *cqspi = dev;
337 	unsigned int irq_status;
338 	struct device *device = &cqspi->pdev->dev;
339 	const struct cqspi_driver_platdata *ddata;
340 
341 	ddata = of_device_get_match_data(device);
342 
343 	/* Read interrupt status */
344 	irq_status = readl(cqspi->iobase + CQSPI_REG_IRQSTATUS);
345 
346 	/* Clear interrupt */
347 	writel(irq_status, cqspi->iobase + CQSPI_REG_IRQSTATUS);
348 
349 	if (cqspi->use_dma_read && ddata && ddata->get_dma_status) {
350 		if (ddata->get_dma_status(cqspi)) {
351 			complete(&cqspi->transfer_complete);
352 			return IRQ_HANDLED;
353 		}
354 	}
355 
356 	else if (!cqspi->slow_sram)
357 		irq_status &= CQSPI_IRQ_MASK_RD | CQSPI_IRQ_MASK_WR;
358 	else
359 		irq_status &= CQSPI_REG_IRQ_WATERMARK | CQSPI_IRQ_MASK_WR;
360 
361 	if (irq_status)
362 		complete(&cqspi->transfer_complete);
363 
364 	return IRQ_HANDLED;
365 }
366 
367 static unsigned int cqspi_calc_rdreg(const struct spi_mem_op *op)
368 {
369 	u32 rdreg = 0;
370 
371 	rdreg |= CQSPI_OP_WIDTH(op->cmd) << CQSPI_REG_RD_INSTR_TYPE_INSTR_LSB;
372 	rdreg |= CQSPI_OP_WIDTH(op->addr) << CQSPI_REG_RD_INSTR_TYPE_ADDR_LSB;
373 	rdreg |= CQSPI_OP_WIDTH(op->data) << CQSPI_REG_RD_INSTR_TYPE_DATA_LSB;
374 
375 	return rdreg;
376 }
377 
378 static unsigned int cqspi_calc_dummy(const struct spi_mem_op *op)
379 {
380 	unsigned int dummy_clk;
381 
382 	if (!op->dummy.nbytes)
383 		return 0;
384 
385 	dummy_clk = op->dummy.nbytes * (8 / op->dummy.buswidth);
386 	if (op->cmd.dtr)
387 		dummy_clk /= 2;
388 
389 	return dummy_clk;
390 }
391 
392 static int cqspi_wait_idle(struct cqspi_st *cqspi)
393 {
394 	const unsigned int poll_idle_retry = 3;
395 	unsigned int count = 0;
396 	unsigned long timeout;
397 
398 	timeout = jiffies + msecs_to_jiffies(CQSPI_TIMEOUT_MS);
399 	while (1) {
400 		/*
401 		 * Read few times in succession to ensure the controller
402 		 * is indeed idle, that is, the bit does not transition
403 		 * low again.
404 		 */
405 		if (cqspi_is_idle(cqspi))
406 			count++;
407 		else
408 			count = 0;
409 
410 		if (count >= poll_idle_retry)
411 			return 0;
412 
413 		if (time_after(jiffies, timeout)) {
414 			/* Timeout, in busy mode. */
415 			dev_err(&cqspi->pdev->dev,
416 				"QSPI is still busy after %dms timeout.\n",
417 				CQSPI_TIMEOUT_MS);
418 			return -ETIMEDOUT;
419 		}
420 
421 		cpu_relax();
422 	}
423 }
424 
425 static int cqspi_exec_flash_cmd(struct cqspi_st *cqspi, unsigned int reg)
426 {
427 	void __iomem *reg_base = cqspi->iobase;
428 	int ret;
429 
430 	/* Write the CMDCTRL without start execution. */
431 	writel(reg, reg_base + CQSPI_REG_CMDCTRL);
432 	/* Start execute */
433 	reg |= CQSPI_REG_CMDCTRL_EXECUTE_MASK;
434 	writel(reg, reg_base + CQSPI_REG_CMDCTRL);
435 
436 	/* Polling for completion. */
437 	ret = cqspi_wait_for_bit(reg_base + CQSPI_REG_CMDCTRL,
438 				 CQSPI_REG_CMDCTRL_INPROGRESS_MASK, 1);
439 	if (ret) {
440 		dev_err(&cqspi->pdev->dev,
441 			"Flash command execution timed out.\n");
442 		return ret;
443 	}
444 
445 	/* Polling QSPI idle status. */
446 	return cqspi_wait_idle(cqspi);
447 }
448 
449 static int cqspi_setup_opcode_ext(struct cqspi_flash_pdata *f_pdata,
450 				  const struct spi_mem_op *op,
451 				  unsigned int shift)
452 {
453 	struct cqspi_st *cqspi = f_pdata->cqspi;
454 	void __iomem *reg_base = cqspi->iobase;
455 	unsigned int reg;
456 	u8 ext;
457 
458 	if (op->cmd.nbytes != 2)
459 		return -EINVAL;
460 
461 	/* Opcode extension is the LSB. */
462 	ext = op->cmd.opcode & 0xff;
463 
464 	reg = readl(reg_base + CQSPI_REG_OP_EXT_LOWER);
465 	reg &= ~(0xff << shift);
466 	reg |= ext << shift;
467 	writel(reg, reg_base + CQSPI_REG_OP_EXT_LOWER);
468 
469 	return 0;
470 }
471 
472 static int cqspi_enable_dtr(struct cqspi_flash_pdata *f_pdata,
473 			    const struct spi_mem_op *op, unsigned int shift)
474 {
475 	struct cqspi_st *cqspi = f_pdata->cqspi;
476 	void __iomem *reg_base = cqspi->iobase;
477 	unsigned int reg;
478 	int ret;
479 
480 	reg = readl(reg_base + CQSPI_REG_CONFIG);
481 
482 	/*
483 	 * We enable dual byte opcode here. The callers have to set up the
484 	 * extension opcode based on which type of operation it is.
485 	 */
486 	if (op->cmd.dtr) {
487 		reg |= CQSPI_REG_CONFIG_DTR_PROTO;
488 		reg |= CQSPI_REG_CONFIG_DUAL_OPCODE;
489 
490 		/* Set up command opcode extension. */
491 		ret = cqspi_setup_opcode_ext(f_pdata, op, shift);
492 		if (ret)
493 			return ret;
494 	} else {
495 		reg &= ~CQSPI_REG_CONFIG_DTR_PROTO;
496 		reg &= ~CQSPI_REG_CONFIG_DUAL_OPCODE;
497 	}
498 
499 	writel(reg, reg_base + CQSPI_REG_CONFIG);
500 
501 	return cqspi_wait_idle(cqspi);
502 }
503 
504 static int cqspi_command_read(struct cqspi_flash_pdata *f_pdata,
505 			      const struct spi_mem_op *op)
506 {
507 	struct cqspi_st *cqspi = f_pdata->cqspi;
508 	void __iomem *reg_base = cqspi->iobase;
509 	u8 *rxbuf = op->data.buf.in;
510 	u8 opcode;
511 	size_t n_rx = op->data.nbytes;
512 	unsigned int rdreg;
513 	unsigned int reg;
514 	unsigned int dummy_clk;
515 	size_t read_len;
516 	int status;
517 
518 	status = cqspi_enable_dtr(f_pdata, op, CQSPI_REG_OP_EXT_STIG_LSB);
519 	if (status)
520 		return status;
521 
522 	if (!n_rx || n_rx > CQSPI_STIG_DATA_LEN_MAX || !rxbuf) {
523 		dev_err(&cqspi->pdev->dev,
524 			"Invalid input argument, len %zu rxbuf 0x%p\n",
525 			n_rx, rxbuf);
526 		return -EINVAL;
527 	}
528 
529 	if (op->cmd.dtr)
530 		opcode = op->cmd.opcode >> 8;
531 	else
532 		opcode = op->cmd.opcode;
533 
534 	reg = opcode << CQSPI_REG_CMDCTRL_OPCODE_LSB;
535 
536 	rdreg = cqspi_calc_rdreg(op);
537 	writel(rdreg, reg_base + CQSPI_REG_RD_INSTR);
538 
539 	dummy_clk = cqspi_calc_dummy(op);
540 	if (dummy_clk > CQSPI_DUMMY_CLKS_MAX)
541 		return -EOPNOTSUPP;
542 
543 	if (dummy_clk)
544 		reg |= (dummy_clk & CQSPI_REG_CMDCTRL_DUMMY_MASK)
545 		     << CQSPI_REG_CMDCTRL_DUMMY_LSB;
546 
547 	reg |= (0x1 << CQSPI_REG_CMDCTRL_RD_EN_LSB);
548 
549 	/* 0 means 1 byte. */
550 	reg |= (((n_rx - 1) & CQSPI_REG_CMDCTRL_RD_BYTES_MASK)
551 		<< CQSPI_REG_CMDCTRL_RD_BYTES_LSB);
552 
553 	/* setup ADDR BIT field */
554 	if (op->addr.nbytes) {
555 		reg |= (0x1 << CQSPI_REG_CMDCTRL_ADDR_EN_LSB);
556 		reg |= ((op->addr.nbytes - 1) &
557 			CQSPI_REG_CMDCTRL_ADD_BYTES_MASK)
558 			<< CQSPI_REG_CMDCTRL_ADD_BYTES_LSB;
559 
560 		writel(op->addr.val, reg_base + CQSPI_REG_CMDADDRESS);
561 	}
562 
563 	status = cqspi_exec_flash_cmd(cqspi, reg);
564 	if (status)
565 		return status;
566 
567 	reg = readl(reg_base + CQSPI_REG_CMDREADDATALOWER);
568 
569 	/* Put the read value into rx_buf */
570 	read_len = (n_rx > 4) ? 4 : n_rx;
571 	memcpy(rxbuf, &reg, read_len);
572 	rxbuf += read_len;
573 
574 	if (n_rx > 4) {
575 		reg = readl(reg_base + CQSPI_REG_CMDREADDATAUPPER);
576 
577 		read_len = n_rx - read_len;
578 		memcpy(rxbuf, &reg, read_len);
579 	}
580 
581 	/* Reset CMD_CTRL Reg once command read completes */
582 	writel(0, reg_base + CQSPI_REG_CMDCTRL);
583 
584 	return 0;
585 }
586 
587 static int cqspi_command_write(struct cqspi_flash_pdata *f_pdata,
588 			       const struct spi_mem_op *op)
589 {
590 	struct cqspi_st *cqspi = f_pdata->cqspi;
591 	void __iomem *reg_base = cqspi->iobase;
592 	u8 opcode;
593 	const u8 *txbuf = op->data.buf.out;
594 	size_t n_tx = op->data.nbytes;
595 	unsigned int reg;
596 	unsigned int data;
597 	size_t write_len;
598 	int ret;
599 
600 	ret = cqspi_enable_dtr(f_pdata, op, CQSPI_REG_OP_EXT_STIG_LSB);
601 	if (ret)
602 		return ret;
603 
604 	if (n_tx > CQSPI_STIG_DATA_LEN_MAX || (n_tx && !txbuf)) {
605 		dev_err(&cqspi->pdev->dev,
606 			"Invalid input argument, cmdlen %zu txbuf 0x%p\n",
607 			n_tx, txbuf);
608 		return -EINVAL;
609 	}
610 
611 	reg = cqspi_calc_rdreg(op);
612 	writel(reg, reg_base + CQSPI_REG_RD_INSTR);
613 
614 	if (op->cmd.dtr)
615 		opcode = op->cmd.opcode >> 8;
616 	else
617 		opcode = op->cmd.opcode;
618 
619 	reg = opcode << CQSPI_REG_CMDCTRL_OPCODE_LSB;
620 
621 	if (op->addr.nbytes) {
622 		reg |= (0x1 << CQSPI_REG_CMDCTRL_ADDR_EN_LSB);
623 		reg |= ((op->addr.nbytes - 1) &
624 			CQSPI_REG_CMDCTRL_ADD_BYTES_MASK)
625 			<< CQSPI_REG_CMDCTRL_ADD_BYTES_LSB;
626 
627 		writel(op->addr.val, reg_base + CQSPI_REG_CMDADDRESS);
628 	}
629 
630 	if (n_tx) {
631 		reg |= (0x1 << CQSPI_REG_CMDCTRL_WR_EN_LSB);
632 		reg |= ((n_tx - 1) & CQSPI_REG_CMDCTRL_WR_BYTES_MASK)
633 			<< CQSPI_REG_CMDCTRL_WR_BYTES_LSB;
634 		data = 0;
635 		write_len = (n_tx > 4) ? 4 : n_tx;
636 		memcpy(&data, txbuf, write_len);
637 		txbuf += write_len;
638 		writel(data, reg_base + CQSPI_REG_CMDWRITEDATALOWER);
639 
640 		if (n_tx > 4) {
641 			data = 0;
642 			write_len = n_tx - 4;
643 			memcpy(&data, txbuf, write_len);
644 			writel(data, reg_base + CQSPI_REG_CMDWRITEDATAUPPER);
645 		}
646 	}
647 
648 	ret = cqspi_exec_flash_cmd(cqspi, reg);
649 
650 	/* Reset CMD_CTRL Reg once command write completes */
651 	writel(0, reg_base + CQSPI_REG_CMDCTRL);
652 
653 	return ret;
654 }
655 
656 static int cqspi_read_setup(struct cqspi_flash_pdata *f_pdata,
657 			    const struct spi_mem_op *op)
658 {
659 	struct cqspi_st *cqspi = f_pdata->cqspi;
660 	void __iomem *reg_base = cqspi->iobase;
661 	unsigned int dummy_clk = 0;
662 	unsigned int reg;
663 	int ret;
664 	u8 opcode;
665 
666 	ret = cqspi_enable_dtr(f_pdata, op, CQSPI_REG_OP_EXT_READ_LSB);
667 	if (ret)
668 		return ret;
669 
670 	if (op->cmd.dtr)
671 		opcode = op->cmd.opcode >> 8;
672 	else
673 		opcode = op->cmd.opcode;
674 
675 	reg = opcode << CQSPI_REG_RD_INSTR_OPCODE_LSB;
676 	reg |= cqspi_calc_rdreg(op);
677 
678 	/* Setup dummy clock cycles */
679 	dummy_clk = cqspi_calc_dummy(op);
680 
681 	if (dummy_clk > CQSPI_DUMMY_CLKS_MAX)
682 		return -EOPNOTSUPP;
683 
684 	if (dummy_clk)
685 		reg |= (dummy_clk & CQSPI_REG_RD_INSTR_DUMMY_MASK)
686 		       << CQSPI_REG_RD_INSTR_DUMMY_LSB;
687 
688 	writel(reg, reg_base + CQSPI_REG_RD_INSTR);
689 
690 	/* Set address width */
691 	reg = readl(reg_base + CQSPI_REG_SIZE);
692 	reg &= ~CQSPI_REG_SIZE_ADDRESS_MASK;
693 	reg |= (op->addr.nbytes - 1);
694 	writel(reg, reg_base + CQSPI_REG_SIZE);
695 	return 0;
696 }
697 
698 static int cqspi_indirect_read_execute(struct cqspi_flash_pdata *f_pdata,
699 				       u8 *rxbuf, loff_t from_addr,
700 				       const size_t n_rx)
701 {
702 	struct cqspi_st *cqspi = f_pdata->cqspi;
703 	struct device *dev = &cqspi->pdev->dev;
704 	void __iomem *reg_base = cqspi->iobase;
705 	void __iomem *ahb_base = cqspi->ahb_base;
706 	unsigned int remaining = n_rx;
707 	unsigned int mod_bytes = n_rx % 4;
708 	unsigned int bytes_to_read = 0;
709 	u8 *rxbuf_end = rxbuf + n_rx;
710 	int ret = 0;
711 
712 	writel(from_addr, reg_base + CQSPI_REG_INDIRECTRDSTARTADDR);
713 	writel(remaining, reg_base + CQSPI_REG_INDIRECTRDBYTES);
714 
715 	/* Clear all interrupts. */
716 	writel(CQSPI_IRQ_STATUS_MASK, reg_base + CQSPI_REG_IRQSTATUS);
717 
718 	/*
719 	 * On SoCFPGA platform reading the SRAM is slow due to
720 	 * hardware limitation and causing read interrupt storm to CPU,
721 	 * so enabling only watermark interrupt to disable all read
722 	 * interrupts later as we want to run "bytes to read" loop with
723 	 * all the read interrupts disabled for max performance.
724 	 */
725 
726 	if (!cqspi->slow_sram)
727 		writel(CQSPI_IRQ_MASK_RD, reg_base + CQSPI_REG_IRQMASK);
728 	else
729 		writel(CQSPI_REG_IRQ_WATERMARK, reg_base + CQSPI_REG_IRQMASK);
730 
731 	reinit_completion(&cqspi->transfer_complete);
732 	writel(CQSPI_REG_INDIRECTRD_START_MASK,
733 	       reg_base + CQSPI_REG_INDIRECTRD);
734 
735 	while (remaining > 0) {
736 		if (!wait_for_completion_timeout(&cqspi->transfer_complete,
737 						 msecs_to_jiffies(CQSPI_READ_TIMEOUT_MS)))
738 			ret = -ETIMEDOUT;
739 
740 		/*
741 		 * Disable all read interrupts until
742 		 * we are out of "bytes to read"
743 		 */
744 		if (cqspi->slow_sram)
745 			writel(0x0, reg_base + CQSPI_REG_IRQMASK);
746 
747 		bytes_to_read = cqspi_get_rd_sram_level(cqspi);
748 
749 		if (ret && bytes_to_read == 0) {
750 			dev_err(dev, "Indirect read timeout, no bytes\n");
751 			goto failrd;
752 		}
753 
754 		while (bytes_to_read != 0) {
755 			unsigned int word_remain = round_down(remaining, 4);
756 
757 			bytes_to_read *= cqspi->fifo_width;
758 			bytes_to_read = bytes_to_read > remaining ?
759 					remaining : bytes_to_read;
760 			bytes_to_read = round_down(bytes_to_read, 4);
761 			/* Read 4 byte word chunks then single bytes */
762 			if (bytes_to_read) {
763 				ioread32_rep(ahb_base, rxbuf,
764 					     (bytes_to_read / 4));
765 			} else if (!word_remain && mod_bytes) {
766 				unsigned int temp = ioread32(ahb_base);
767 
768 				bytes_to_read = mod_bytes;
769 				memcpy(rxbuf, &temp, min((unsigned int)
770 							 (rxbuf_end - rxbuf),
771 							 bytes_to_read));
772 			}
773 			rxbuf += bytes_to_read;
774 			remaining -= bytes_to_read;
775 			bytes_to_read = cqspi_get_rd_sram_level(cqspi);
776 		}
777 
778 		if (remaining > 0) {
779 			reinit_completion(&cqspi->transfer_complete);
780 			if (cqspi->slow_sram)
781 				writel(CQSPI_REG_IRQ_WATERMARK, reg_base + CQSPI_REG_IRQMASK);
782 		}
783 	}
784 
785 	/* Check indirect done status */
786 	ret = cqspi_wait_for_bit(reg_base + CQSPI_REG_INDIRECTRD,
787 				 CQSPI_REG_INDIRECTRD_DONE_MASK, 0);
788 	if (ret) {
789 		dev_err(dev, "Indirect read completion error (%i)\n", ret);
790 		goto failrd;
791 	}
792 
793 	/* Disable interrupt */
794 	writel(0, reg_base + CQSPI_REG_IRQMASK);
795 
796 	/* Clear indirect completion status */
797 	writel(CQSPI_REG_INDIRECTRD_DONE_MASK, reg_base + CQSPI_REG_INDIRECTRD);
798 
799 	return 0;
800 
801 failrd:
802 	/* Disable interrupt */
803 	writel(0, reg_base + CQSPI_REG_IRQMASK);
804 
805 	/* Cancel the indirect read */
806 	writel(CQSPI_REG_INDIRECTRD_CANCEL_MASK,
807 	       reg_base + CQSPI_REG_INDIRECTRD);
808 	return ret;
809 }
810 
811 static void cqspi_controller_enable(struct cqspi_st *cqspi, bool enable)
812 {
813 	void __iomem *reg_base = cqspi->iobase;
814 	unsigned int reg;
815 
816 	reg = readl(reg_base + CQSPI_REG_CONFIG);
817 
818 	if (enable)
819 		reg |= CQSPI_REG_CONFIG_ENABLE_MASK;
820 	else
821 		reg &= ~CQSPI_REG_CONFIG_ENABLE_MASK;
822 
823 	writel(reg, reg_base + CQSPI_REG_CONFIG);
824 }
825 
826 static int cqspi_versal_indirect_read_dma(struct cqspi_flash_pdata *f_pdata,
827 					  u_char *rxbuf, loff_t from_addr,
828 					  size_t n_rx)
829 {
830 	struct cqspi_st *cqspi = f_pdata->cqspi;
831 	struct device *dev = &cqspi->pdev->dev;
832 	void __iomem *reg_base = cqspi->iobase;
833 	u32 reg, bytes_to_dma;
834 	loff_t addr = from_addr;
835 	void *buf = rxbuf;
836 	dma_addr_t dma_addr;
837 	u8 bytes_rem;
838 	int ret = 0;
839 
840 	bytes_rem = n_rx % 4;
841 	bytes_to_dma = (n_rx - bytes_rem);
842 
843 	if (!bytes_to_dma)
844 		goto nondmard;
845 
846 	ret = zynqmp_pm_ospi_mux_select(cqspi->pd_dev_id, PM_OSPI_MUX_SEL_DMA);
847 	if (ret)
848 		return ret;
849 
850 	cqspi_controller_enable(cqspi, 0);
851 
852 	reg = readl(cqspi->iobase + CQSPI_REG_CONFIG);
853 	reg |= CQSPI_REG_CONFIG_DMA_MASK;
854 	writel(reg, cqspi->iobase + CQSPI_REG_CONFIG);
855 
856 	cqspi_controller_enable(cqspi, 1);
857 
858 	dma_addr = dma_map_single(dev, rxbuf, bytes_to_dma, DMA_FROM_DEVICE);
859 	if (dma_mapping_error(dev, dma_addr)) {
860 		dev_err(dev, "dma mapping failed\n");
861 		return -ENOMEM;
862 	}
863 
864 	writel(from_addr, reg_base + CQSPI_REG_INDIRECTRDSTARTADDR);
865 	writel(bytes_to_dma, reg_base + CQSPI_REG_INDIRECTRDBYTES);
866 	writel(CQSPI_REG_VERSAL_ADDRRANGE_WIDTH_VAL,
867 	       reg_base + CQSPI_REG_INDTRIG_ADDRRANGE);
868 
869 	/* Clear all interrupts. */
870 	writel(CQSPI_IRQ_STATUS_MASK, reg_base + CQSPI_REG_IRQSTATUS);
871 
872 	/* Enable DMA done interrupt */
873 	writel(CQSPI_REG_VERSAL_DMA_DST_DONE_MASK,
874 	       reg_base + CQSPI_REG_VERSAL_DMA_DST_I_EN);
875 
876 	/* Default DMA periph configuration */
877 	writel(CQSPI_REG_VERSAL_DMA_VAL, reg_base + CQSPI_REG_DMA);
878 
879 	/* Configure DMA Dst address */
880 	writel(lower_32_bits(dma_addr),
881 	       reg_base + CQSPI_REG_VERSAL_DMA_DST_ADDR);
882 	writel(upper_32_bits(dma_addr),
883 	       reg_base + CQSPI_REG_VERSAL_DMA_DST_ADDR_MSB);
884 
885 	/* Configure DMA Src address */
886 	writel(cqspi->trigger_address, reg_base +
887 	       CQSPI_REG_VERSAL_DMA_SRC_ADDR);
888 
889 	/* Set DMA destination size */
890 	writel(bytes_to_dma, reg_base + CQSPI_REG_VERSAL_DMA_DST_SIZE);
891 
892 	/* Set DMA destination control */
893 	writel(CQSPI_REG_VERSAL_DMA_DST_CTRL_VAL,
894 	       reg_base + CQSPI_REG_VERSAL_DMA_DST_CTRL);
895 
896 	writel(CQSPI_REG_INDIRECTRD_START_MASK,
897 	       reg_base + CQSPI_REG_INDIRECTRD);
898 
899 	reinit_completion(&cqspi->transfer_complete);
900 
901 	if (!wait_for_completion_timeout(&cqspi->transfer_complete,
902 					 msecs_to_jiffies(max_t(size_t, bytes_to_dma, 500)))) {
903 		ret = -ETIMEDOUT;
904 		goto failrd;
905 	}
906 
907 	/* Disable DMA interrupt */
908 	writel(0x0, cqspi->iobase + CQSPI_REG_VERSAL_DMA_DST_I_DIS);
909 
910 	/* Clear indirect completion status */
911 	writel(CQSPI_REG_INDIRECTRD_DONE_MASK,
912 	       cqspi->iobase + CQSPI_REG_INDIRECTRD);
913 	dma_unmap_single(dev, dma_addr, bytes_to_dma, DMA_FROM_DEVICE);
914 
915 	cqspi_controller_enable(cqspi, 0);
916 
917 	reg = readl(cqspi->iobase + CQSPI_REG_CONFIG);
918 	reg &= ~CQSPI_REG_CONFIG_DMA_MASK;
919 	writel(reg, cqspi->iobase + CQSPI_REG_CONFIG);
920 
921 	cqspi_controller_enable(cqspi, 1);
922 
923 	ret = zynqmp_pm_ospi_mux_select(cqspi->pd_dev_id,
924 					PM_OSPI_MUX_SEL_LINEAR);
925 	if (ret)
926 		return ret;
927 
928 nondmard:
929 	if (bytes_rem) {
930 		addr += bytes_to_dma;
931 		buf += bytes_to_dma;
932 		ret = cqspi_indirect_read_execute(f_pdata, buf, addr,
933 						  bytes_rem);
934 		if (ret)
935 			return ret;
936 	}
937 
938 	return 0;
939 
940 failrd:
941 	/* Disable DMA interrupt */
942 	writel(0x0, reg_base + CQSPI_REG_VERSAL_DMA_DST_I_DIS);
943 
944 	/* Cancel the indirect read */
945 	writel(CQSPI_REG_INDIRECTWR_CANCEL_MASK,
946 	       reg_base + CQSPI_REG_INDIRECTRD);
947 
948 	dma_unmap_single(dev, dma_addr, bytes_to_dma, DMA_FROM_DEVICE);
949 
950 	reg = readl(cqspi->iobase + CQSPI_REG_CONFIG);
951 	reg &= ~CQSPI_REG_CONFIG_DMA_MASK;
952 	writel(reg, cqspi->iobase + CQSPI_REG_CONFIG);
953 
954 	zynqmp_pm_ospi_mux_select(cqspi->pd_dev_id, PM_OSPI_MUX_SEL_LINEAR);
955 
956 	return ret;
957 }
958 
959 static int cqspi_write_setup(struct cqspi_flash_pdata *f_pdata,
960 			     const struct spi_mem_op *op)
961 {
962 	unsigned int reg;
963 	int ret;
964 	struct cqspi_st *cqspi = f_pdata->cqspi;
965 	void __iomem *reg_base = cqspi->iobase;
966 	u8 opcode;
967 
968 	ret = cqspi_enable_dtr(f_pdata, op, CQSPI_REG_OP_EXT_WRITE_LSB);
969 	if (ret)
970 		return ret;
971 
972 	if (op->cmd.dtr)
973 		opcode = op->cmd.opcode >> 8;
974 	else
975 		opcode = op->cmd.opcode;
976 
977 	/* Set opcode. */
978 	reg = opcode << CQSPI_REG_WR_INSTR_OPCODE_LSB;
979 	reg |= CQSPI_OP_WIDTH(op->data) << CQSPI_REG_WR_INSTR_TYPE_DATA_LSB;
980 	reg |= CQSPI_OP_WIDTH(op->addr) << CQSPI_REG_WR_INSTR_TYPE_ADDR_LSB;
981 	writel(reg, reg_base + CQSPI_REG_WR_INSTR);
982 	reg = cqspi_calc_rdreg(op);
983 	writel(reg, reg_base + CQSPI_REG_RD_INSTR);
984 
985 	/*
986 	 * SPI NAND flashes require the address of the status register to be
987 	 * passed in the Read SR command. Also, some SPI NOR flashes like the
988 	 * cypress Semper flash expect a 4-byte dummy address in the Read SR
989 	 * command in DTR mode.
990 	 *
991 	 * But this controller does not support address phase in the Read SR
992 	 * command when doing auto-HW polling. So, disable write completion
993 	 * polling on the controller's side. spinand and spi-nor will take
994 	 * care of polling the status register.
995 	 */
996 	if (cqspi->wr_completion) {
997 		reg = readl(reg_base + CQSPI_REG_WR_COMPLETION_CTRL);
998 		reg |= CQSPI_REG_WR_DISABLE_AUTO_POLL;
999 		writel(reg, reg_base + CQSPI_REG_WR_COMPLETION_CTRL);
1000 		/*
1001 		 * DAC mode require auto polling as flash needs to be polled
1002 		 * for write completion in case of bubble in SPI transaction
1003 		 * due to slow CPU/DMA master.
1004 		 */
1005 		cqspi->use_direct_mode_wr = false;
1006 	}
1007 
1008 	reg = readl(reg_base + CQSPI_REG_SIZE);
1009 	reg &= ~CQSPI_REG_SIZE_ADDRESS_MASK;
1010 	reg |= (op->addr.nbytes - 1);
1011 	writel(reg, reg_base + CQSPI_REG_SIZE);
1012 	return 0;
1013 }
1014 
1015 static int cqspi_indirect_write_execute(struct cqspi_flash_pdata *f_pdata,
1016 					loff_t to_addr, const u8 *txbuf,
1017 					const size_t n_tx)
1018 {
1019 	struct cqspi_st *cqspi = f_pdata->cqspi;
1020 	struct device *dev = &cqspi->pdev->dev;
1021 	void __iomem *reg_base = cqspi->iobase;
1022 	unsigned int remaining = n_tx;
1023 	unsigned int write_bytes;
1024 	int ret;
1025 
1026 	writel(to_addr, reg_base + CQSPI_REG_INDIRECTWRSTARTADDR);
1027 	writel(remaining, reg_base + CQSPI_REG_INDIRECTWRBYTES);
1028 
1029 	/* Clear all interrupts. */
1030 	writel(CQSPI_IRQ_STATUS_MASK, reg_base + CQSPI_REG_IRQSTATUS);
1031 
1032 	writel(CQSPI_IRQ_MASK_WR, reg_base + CQSPI_REG_IRQMASK);
1033 
1034 	reinit_completion(&cqspi->transfer_complete);
1035 	writel(CQSPI_REG_INDIRECTWR_START_MASK,
1036 	       reg_base + CQSPI_REG_INDIRECTWR);
1037 	/*
1038 	 * As per 66AK2G02 TRM SPRUHY8F section 11.15.5.3 Indirect Access
1039 	 * Controller programming sequence, couple of cycles of
1040 	 * QSPI_REF_CLK delay is required for the above bit to
1041 	 * be internally synchronized by the QSPI module. Provide 5
1042 	 * cycles of delay.
1043 	 */
1044 	if (cqspi->wr_delay)
1045 		ndelay(cqspi->wr_delay);
1046 
1047 	/*
1048 	 * If a hazard exists between the APB and AHB interfaces, perform a
1049 	 * dummy readback from the controller to ensure synchronization.
1050 	 */
1051 	if (cqspi->apb_ahb_hazard)
1052 		readl(reg_base + CQSPI_REG_INDIRECTWR);
1053 
1054 	while (remaining > 0) {
1055 		size_t write_words, mod_bytes;
1056 
1057 		write_bytes = remaining;
1058 		write_words = write_bytes / 4;
1059 		mod_bytes = write_bytes % 4;
1060 		/* Write 4 bytes at a time then single bytes. */
1061 		if (write_words) {
1062 			iowrite32_rep(cqspi->ahb_base, txbuf, write_words);
1063 			txbuf += (write_words * 4);
1064 		}
1065 		if (mod_bytes) {
1066 			unsigned int temp = 0xFFFFFFFF;
1067 
1068 			memcpy(&temp, txbuf, mod_bytes);
1069 			iowrite32(temp, cqspi->ahb_base);
1070 			txbuf += mod_bytes;
1071 		}
1072 
1073 		if (!wait_for_completion_timeout(&cqspi->transfer_complete,
1074 						 msecs_to_jiffies(CQSPI_TIMEOUT_MS))) {
1075 			dev_err(dev, "Indirect write timeout\n");
1076 			ret = -ETIMEDOUT;
1077 			goto failwr;
1078 		}
1079 
1080 		remaining -= write_bytes;
1081 
1082 		if (remaining > 0)
1083 			reinit_completion(&cqspi->transfer_complete);
1084 	}
1085 
1086 	/* Check indirect done status */
1087 	ret = cqspi_wait_for_bit(reg_base + CQSPI_REG_INDIRECTWR,
1088 				 CQSPI_REG_INDIRECTWR_DONE_MASK, 0);
1089 	if (ret) {
1090 		dev_err(dev, "Indirect write completion error (%i)\n", ret);
1091 		goto failwr;
1092 	}
1093 
1094 	/* Disable interrupt. */
1095 	writel(0, reg_base + CQSPI_REG_IRQMASK);
1096 
1097 	/* Clear indirect completion status */
1098 	writel(CQSPI_REG_INDIRECTWR_DONE_MASK, reg_base + CQSPI_REG_INDIRECTWR);
1099 
1100 	cqspi_wait_idle(cqspi);
1101 
1102 	return 0;
1103 
1104 failwr:
1105 	/* Disable interrupt. */
1106 	writel(0, reg_base + CQSPI_REG_IRQMASK);
1107 
1108 	/* Cancel the indirect write */
1109 	writel(CQSPI_REG_INDIRECTWR_CANCEL_MASK,
1110 	       reg_base + CQSPI_REG_INDIRECTWR);
1111 	return ret;
1112 }
1113 
1114 static void cqspi_chipselect(struct cqspi_flash_pdata *f_pdata)
1115 {
1116 	struct cqspi_st *cqspi = f_pdata->cqspi;
1117 	void __iomem *reg_base = cqspi->iobase;
1118 	unsigned int chip_select = f_pdata->cs;
1119 	unsigned int reg;
1120 
1121 	reg = readl(reg_base + CQSPI_REG_CONFIG);
1122 	if (cqspi->is_decoded_cs) {
1123 		reg |= CQSPI_REG_CONFIG_DECODE_MASK;
1124 	} else {
1125 		reg &= ~CQSPI_REG_CONFIG_DECODE_MASK;
1126 
1127 		/* Convert CS if without decoder.
1128 		 * CS0 to 4b'1110
1129 		 * CS1 to 4b'1101
1130 		 * CS2 to 4b'1011
1131 		 * CS3 to 4b'0111
1132 		 */
1133 		chip_select = 0xF & ~(1 << chip_select);
1134 	}
1135 
1136 	reg &= ~(CQSPI_REG_CONFIG_CHIPSELECT_MASK
1137 		 << CQSPI_REG_CONFIG_CHIPSELECT_LSB);
1138 	reg |= (chip_select & CQSPI_REG_CONFIG_CHIPSELECT_MASK)
1139 	    << CQSPI_REG_CONFIG_CHIPSELECT_LSB;
1140 	writel(reg, reg_base + CQSPI_REG_CONFIG);
1141 }
1142 
1143 static unsigned int calculate_ticks_for_ns(const unsigned int ref_clk_hz,
1144 					   const unsigned int ns_val)
1145 {
1146 	unsigned int ticks;
1147 
1148 	ticks = ref_clk_hz / 1000;	/* kHz */
1149 	ticks = DIV_ROUND_UP(ticks * ns_val, 1000000);
1150 
1151 	return ticks;
1152 }
1153 
1154 static void cqspi_delay(struct cqspi_flash_pdata *f_pdata)
1155 {
1156 	struct cqspi_st *cqspi = f_pdata->cqspi;
1157 	void __iomem *iobase = cqspi->iobase;
1158 	const unsigned int ref_clk_hz = cqspi->master_ref_clk_hz;
1159 	unsigned int tshsl, tchsh, tslch, tsd2d;
1160 	unsigned int reg;
1161 	unsigned int tsclk;
1162 
1163 	/* calculate the number of ref ticks for one sclk tick */
1164 	tsclk = DIV_ROUND_UP(ref_clk_hz, cqspi->sclk);
1165 
1166 	tshsl = calculate_ticks_for_ns(ref_clk_hz, f_pdata->tshsl_ns);
1167 	/* this particular value must be at least one sclk */
1168 	if (tshsl < tsclk)
1169 		tshsl = tsclk;
1170 
1171 	tchsh = calculate_ticks_for_ns(ref_clk_hz, f_pdata->tchsh_ns);
1172 	tslch = calculate_ticks_for_ns(ref_clk_hz, f_pdata->tslch_ns);
1173 	tsd2d = calculate_ticks_for_ns(ref_clk_hz, f_pdata->tsd2d_ns);
1174 
1175 	reg = (tshsl & CQSPI_REG_DELAY_TSHSL_MASK)
1176 	       << CQSPI_REG_DELAY_TSHSL_LSB;
1177 	reg |= (tchsh & CQSPI_REG_DELAY_TCHSH_MASK)
1178 		<< CQSPI_REG_DELAY_TCHSH_LSB;
1179 	reg |= (tslch & CQSPI_REG_DELAY_TSLCH_MASK)
1180 		<< CQSPI_REG_DELAY_TSLCH_LSB;
1181 	reg |= (tsd2d & CQSPI_REG_DELAY_TSD2D_MASK)
1182 		<< CQSPI_REG_DELAY_TSD2D_LSB;
1183 	writel(reg, iobase + CQSPI_REG_DELAY);
1184 }
1185 
1186 static void cqspi_config_baudrate_div(struct cqspi_st *cqspi)
1187 {
1188 	const unsigned int ref_clk_hz = cqspi->master_ref_clk_hz;
1189 	void __iomem *reg_base = cqspi->iobase;
1190 	u32 reg, div;
1191 
1192 	/* Recalculate the baudrate divisor based on QSPI specification. */
1193 	div = DIV_ROUND_UP(ref_clk_hz, 2 * cqspi->sclk) - 1;
1194 
1195 	/* Maximum baud divisor */
1196 	if (div > CQSPI_REG_CONFIG_BAUD_MASK) {
1197 		div = CQSPI_REG_CONFIG_BAUD_MASK;
1198 		dev_warn(&cqspi->pdev->dev,
1199 			"Unable to adjust clock <= %d hz. Reduced to %d hz\n",
1200 			cqspi->sclk, ref_clk_hz/((div+1)*2));
1201 	}
1202 
1203 	reg = readl(reg_base + CQSPI_REG_CONFIG);
1204 	reg &= ~(CQSPI_REG_CONFIG_BAUD_MASK << CQSPI_REG_CONFIG_BAUD_LSB);
1205 	reg |= (div & CQSPI_REG_CONFIG_BAUD_MASK) << CQSPI_REG_CONFIG_BAUD_LSB;
1206 	writel(reg, reg_base + CQSPI_REG_CONFIG);
1207 }
1208 
1209 static void cqspi_readdata_capture(struct cqspi_st *cqspi,
1210 				   const bool bypass,
1211 				   const unsigned int delay)
1212 {
1213 	void __iomem *reg_base = cqspi->iobase;
1214 	unsigned int reg;
1215 
1216 	reg = readl(reg_base + CQSPI_REG_READCAPTURE);
1217 
1218 	if (bypass)
1219 		reg |= (1 << CQSPI_REG_READCAPTURE_BYPASS_LSB);
1220 	else
1221 		reg &= ~(1 << CQSPI_REG_READCAPTURE_BYPASS_LSB);
1222 
1223 	reg &= ~(CQSPI_REG_READCAPTURE_DELAY_MASK
1224 		 << CQSPI_REG_READCAPTURE_DELAY_LSB);
1225 
1226 	reg |= (delay & CQSPI_REG_READCAPTURE_DELAY_MASK)
1227 		<< CQSPI_REG_READCAPTURE_DELAY_LSB;
1228 
1229 	writel(reg, reg_base + CQSPI_REG_READCAPTURE);
1230 }
1231 
1232 static void cqspi_configure(struct cqspi_flash_pdata *f_pdata,
1233 			    unsigned long sclk)
1234 {
1235 	struct cqspi_st *cqspi = f_pdata->cqspi;
1236 	int switch_cs = (cqspi->current_cs != f_pdata->cs);
1237 	int switch_ck = (cqspi->sclk != sclk);
1238 
1239 	if (switch_cs || switch_ck)
1240 		cqspi_controller_enable(cqspi, 0);
1241 
1242 	/* Switch chip select. */
1243 	if (switch_cs) {
1244 		cqspi->current_cs = f_pdata->cs;
1245 		cqspi_chipselect(f_pdata);
1246 	}
1247 
1248 	/* Setup baudrate divisor and delays */
1249 	if (switch_ck) {
1250 		cqspi->sclk = sclk;
1251 		cqspi_config_baudrate_div(cqspi);
1252 		cqspi_delay(f_pdata);
1253 		cqspi_readdata_capture(cqspi, !cqspi->rclk_en,
1254 				       f_pdata->read_delay);
1255 	}
1256 
1257 	if (switch_cs || switch_ck)
1258 		cqspi_controller_enable(cqspi, 1);
1259 }
1260 
1261 static ssize_t cqspi_write(struct cqspi_flash_pdata *f_pdata,
1262 			   const struct spi_mem_op *op)
1263 {
1264 	struct cqspi_st *cqspi = f_pdata->cqspi;
1265 	loff_t to = op->addr.val;
1266 	size_t len = op->data.nbytes;
1267 	const u_char *buf = op->data.buf.out;
1268 	int ret;
1269 
1270 	ret = cqspi_write_setup(f_pdata, op);
1271 	if (ret)
1272 		return ret;
1273 
1274 	/*
1275 	 * Some flashes like the Cypress Semper flash expect a dummy 4-byte
1276 	 * address (all 0s) with the read status register command in DTR mode.
1277 	 * But this controller does not support sending dummy address bytes to
1278 	 * the flash when it is polling the write completion register in DTR
1279 	 * mode. So, we can not use direct mode when in DTR mode for writing
1280 	 * data.
1281 	 */
1282 	if (!op->cmd.dtr && cqspi->use_direct_mode &&
1283 	    cqspi->use_direct_mode_wr && ((to + len) <= cqspi->ahb_size)) {
1284 		memcpy_toio(cqspi->ahb_base + to, buf, len);
1285 		return cqspi_wait_idle(cqspi);
1286 	}
1287 
1288 	return cqspi_indirect_write_execute(f_pdata, to, buf, len);
1289 }
1290 
1291 static void cqspi_rx_dma_callback(void *param)
1292 {
1293 	struct cqspi_st *cqspi = param;
1294 
1295 	complete(&cqspi->rx_dma_complete);
1296 }
1297 
1298 static int cqspi_direct_read_execute(struct cqspi_flash_pdata *f_pdata,
1299 				     u_char *buf, loff_t from, size_t len)
1300 {
1301 	struct cqspi_st *cqspi = f_pdata->cqspi;
1302 	struct device *dev = &cqspi->pdev->dev;
1303 	enum dma_ctrl_flags flags = DMA_CTRL_ACK | DMA_PREP_INTERRUPT;
1304 	dma_addr_t dma_src = (dma_addr_t)cqspi->mmap_phys_base + from;
1305 	int ret = 0;
1306 	struct dma_async_tx_descriptor *tx;
1307 	dma_cookie_t cookie;
1308 	dma_addr_t dma_dst;
1309 	struct device *ddev;
1310 
1311 	if (!cqspi->rx_chan || !virt_addr_valid(buf)) {
1312 		memcpy_fromio(buf, cqspi->ahb_base + from, len);
1313 		return 0;
1314 	}
1315 
1316 	ddev = cqspi->rx_chan->device->dev;
1317 	dma_dst = dma_map_single(ddev, buf, len, DMA_FROM_DEVICE);
1318 	if (dma_mapping_error(ddev, dma_dst)) {
1319 		dev_err(dev, "dma mapping failed\n");
1320 		return -ENOMEM;
1321 	}
1322 	tx = dmaengine_prep_dma_memcpy(cqspi->rx_chan, dma_dst, dma_src,
1323 				       len, flags);
1324 	if (!tx) {
1325 		dev_err(dev, "device_prep_dma_memcpy error\n");
1326 		ret = -EIO;
1327 		goto err_unmap;
1328 	}
1329 
1330 	tx->callback = cqspi_rx_dma_callback;
1331 	tx->callback_param = cqspi;
1332 	cookie = tx->tx_submit(tx);
1333 	reinit_completion(&cqspi->rx_dma_complete);
1334 
1335 	ret = dma_submit_error(cookie);
1336 	if (ret) {
1337 		dev_err(dev, "dma_submit_error %d\n", cookie);
1338 		ret = -EIO;
1339 		goto err_unmap;
1340 	}
1341 
1342 	dma_async_issue_pending(cqspi->rx_chan);
1343 	if (!wait_for_completion_timeout(&cqspi->rx_dma_complete,
1344 					 msecs_to_jiffies(max_t(size_t, len, 500)))) {
1345 		dmaengine_terminate_sync(cqspi->rx_chan);
1346 		dev_err(dev, "DMA wait_for_completion_timeout\n");
1347 		ret = -ETIMEDOUT;
1348 		goto err_unmap;
1349 	}
1350 
1351 err_unmap:
1352 	dma_unmap_single(ddev, dma_dst, len, DMA_FROM_DEVICE);
1353 
1354 	return ret;
1355 }
1356 
1357 static ssize_t cqspi_read(struct cqspi_flash_pdata *f_pdata,
1358 			  const struct spi_mem_op *op)
1359 {
1360 	struct cqspi_st *cqspi = f_pdata->cqspi;
1361 	struct device *dev = &cqspi->pdev->dev;
1362 	const struct cqspi_driver_platdata *ddata;
1363 	loff_t from = op->addr.val;
1364 	size_t len = op->data.nbytes;
1365 	u_char *buf = op->data.buf.in;
1366 	u64 dma_align = (u64)(uintptr_t)buf;
1367 	int ret;
1368 
1369 	ddata = of_device_get_match_data(dev);
1370 
1371 	ret = cqspi_read_setup(f_pdata, op);
1372 	if (ret)
1373 		return ret;
1374 
1375 	if (cqspi->use_direct_mode && ((from + len) <= cqspi->ahb_size))
1376 		return cqspi_direct_read_execute(f_pdata, buf, from, len);
1377 
1378 	if (cqspi->use_dma_read && ddata && ddata->indirect_read_dma &&
1379 	    virt_addr_valid(buf) && ((dma_align & CQSPI_DMA_UNALIGN) == 0))
1380 		return ddata->indirect_read_dma(f_pdata, buf, from, len);
1381 
1382 	return cqspi_indirect_read_execute(f_pdata, buf, from, len);
1383 }
1384 
1385 static int cqspi_mem_process(struct spi_mem *mem, const struct spi_mem_op *op)
1386 {
1387 	struct cqspi_st *cqspi = spi_controller_get_devdata(mem->spi->controller);
1388 	struct cqspi_flash_pdata *f_pdata;
1389 
1390 	f_pdata = &cqspi->f_pdata[spi_get_chipselect(mem->spi, 0)];
1391 	cqspi_configure(f_pdata, mem->spi->max_speed_hz);
1392 
1393 	if (op->data.dir == SPI_MEM_DATA_IN && op->data.buf.in) {
1394 	/*
1395 	 * Performing reads in DAC mode forces to read minimum 4 bytes
1396 	 * which is unsupported on some flash devices during register
1397 	 * reads, prefer STIG mode for such small reads.
1398 	 */
1399 		if (!op->addr.nbytes ||
1400 		    op->data.nbytes <= CQSPI_STIG_DATA_LEN_MAX)
1401 			return cqspi_command_read(f_pdata, op);
1402 
1403 		return cqspi_read(f_pdata, op);
1404 	}
1405 
1406 	if (!op->addr.nbytes || !op->data.buf.out)
1407 		return cqspi_command_write(f_pdata, op);
1408 
1409 	return cqspi_write(f_pdata, op);
1410 }
1411 
1412 static int cqspi_exec_mem_op(struct spi_mem *mem, const struct spi_mem_op *op)
1413 {
1414 	int ret;
1415 	struct cqspi_st *cqspi = spi_controller_get_devdata(mem->spi->controller);
1416 	struct device *dev = &cqspi->pdev->dev;
1417 
1418 	ret = pm_runtime_resume_and_get(dev);
1419 	if (ret) {
1420 		dev_err(&mem->spi->dev, "resume failed with %d\n", ret);
1421 		return ret;
1422 	}
1423 
1424 	ret = cqspi_mem_process(mem, op);
1425 
1426 	pm_runtime_mark_last_busy(dev);
1427 	pm_runtime_put_autosuspend(dev);
1428 
1429 	if (ret)
1430 		dev_err(&mem->spi->dev, "operation failed with %d\n", ret);
1431 
1432 	return ret;
1433 }
1434 
1435 static bool cqspi_supports_mem_op(struct spi_mem *mem,
1436 				  const struct spi_mem_op *op)
1437 {
1438 	bool all_true, all_false;
1439 
1440 	/*
1441 	 * op->dummy.dtr is required for converting nbytes into ncycles.
1442 	 * Also, don't check the dtr field of the op phase having zero nbytes.
1443 	 */
1444 	all_true = op->cmd.dtr &&
1445 		   (!op->addr.nbytes || op->addr.dtr) &&
1446 		   (!op->dummy.nbytes || op->dummy.dtr) &&
1447 		   (!op->data.nbytes || op->data.dtr);
1448 
1449 	all_false = !op->cmd.dtr && !op->addr.dtr && !op->dummy.dtr &&
1450 		    !op->data.dtr;
1451 
1452 	if (all_true) {
1453 		/* Right now we only support 8-8-8 DTR mode. */
1454 		if (op->cmd.nbytes && op->cmd.buswidth != 8)
1455 			return false;
1456 		if (op->addr.nbytes && op->addr.buswidth != 8)
1457 			return false;
1458 		if (op->data.nbytes && op->data.buswidth != 8)
1459 			return false;
1460 	} else if (!all_false) {
1461 		/* Mixed DTR modes are not supported. */
1462 		return false;
1463 	}
1464 
1465 	return spi_mem_default_supports_op(mem, op);
1466 }
1467 
1468 static int cqspi_of_get_flash_pdata(struct platform_device *pdev,
1469 				    struct cqspi_flash_pdata *f_pdata,
1470 				    struct device_node *np)
1471 {
1472 	if (of_property_read_u32(np, "cdns,read-delay", &f_pdata->read_delay)) {
1473 		dev_err(&pdev->dev, "couldn't determine read-delay\n");
1474 		return -ENXIO;
1475 	}
1476 
1477 	if (of_property_read_u32(np, "cdns,tshsl-ns", &f_pdata->tshsl_ns)) {
1478 		dev_err(&pdev->dev, "couldn't determine tshsl-ns\n");
1479 		return -ENXIO;
1480 	}
1481 
1482 	if (of_property_read_u32(np, "cdns,tsd2d-ns", &f_pdata->tsd2d_ns)) {
1483 		dev_err(&pdev->dev, "couldn't determine tsd2d-ns\n");
1484 		return -ENXIO;
1485 	}
1486 
1487 	if (of_property_read_u32(np, "cdns,tchsh-ns", &f_pdata->tchsh_ns)) {
1488 		dev_err(&pdev->dev, "couldn't determine tchsh-ns\n");
1489 		return -ENXIO;
1490 	}
1491 
1492 	if (of_property_read_u32(np, "cdns,tslch-ns", &f_pdata->tslch_ns)) {
1493 		dev_err(&pdev->dev, "couldn't determine tslch-ns\n");
1494 		return -ENXIO;
1495 	}
1496 
1497 	if (of_property_read_u32(np, "spi-max-frequency", &f_pdata->clk_rate)) {
1498 		dev_err(&pdev->dev, "couldn't determine spi-max-frequency\n");
1499 		return -ENXIO;
1500 	}
1501 
1502 	return 0;
1503 }
1504 
1505 static int cqspi_of_get_pdata(struct cqspi_st *cqspi)
1506 {
1507 	struct device *dev = &cqspi->pdev->dev;
1508 	struct device_node *np = dev->of_node;
1509 	u32 id[2];
1510 
1511 	cqspi->is_decoded_cs = of_property_read_bool(np, "cdns,is-decoded-cs");
1512 
1513 	if (of_property_read_u32(np, "cdns,fifo-depth", &cqspi->fifo_depth)) {
1514 		dev_err(dev, "couldn't determine fifo-depth\n");
1515 		return -ENXIO;
1516 	}
1517 
1518 	if (of_property_read_u32(np, "cdns,fifo-width", &cqspi->fifo_width)) {
1519 		dev_err(dev, "couldn't determine fifo-width\n");
1520 		return -ENXIO;
1521 	}
1522 
1523 	if (of_property_read_u32(np, "cdns,trigger-address",
1524 				 &cqspi->trigger_address)) {
1525 		dev_err(dev, "couldn't determine trigger-address\n");
1526 		return -ENXIO;
1527 	}
1528 
1529 	if (of_property_read_u32(np, "num-cs", &cqspi->num_chipselect))
1530 		cqspi->num_chipselect = CQSPI_MAX_CHIPSELECT;
1531 
1532 	cqspi->rclk_en = of_property_read_bool(np, "cdns,rclk-en");
1533 
1534 	if (!of_property_read_u32_array(np, "power-domains", id,
1535 					ARRAY_SIZE(id)))
1536 		cqspi->pd_dev_id = id[1];
1537 
1538 	return 0;
1539 }
1540 
1541 static void cqspi_controller_init(struct cqspi_st *cqspi)
1542 {
1543 	u32 reg;
1544 
1545 	cqspi_controller_enable(cqspi, 0);
1546 
1547 	/* Configure the remap address register, no remap */
1548 	writel(0, cqspi->iobase + CQSPI_REG_REMAP);
1549 
1550 	/* Disable all interrupts. */
1551 	writel(0, cqspi->iobase + CQSPI_REG_IRQMASK);
1552 
1553 	/* Configure the SRAM split to 1:1 . */
1554 	writel(cqspi->fifo_depth / 2, cqspi->iobase + CQSPI_REG_SRAMPARTITION);
1555 
1556 	/* Load indirect trigger address. */
1557 	writel(cqspi->trigger_address,
1558 	       cqspi->iobase + CQSPI_REG_INDIRECTTRIGGER);
1559 
1560 	/* Program read watermark -- 1/2 of the FIFO. */
1561 	writel(cqspi->fifo_depth * cqspi->fifo_width / 2,
1562 	       cqspi->iobase + CQSPI_REG_INDIRECTRDWATERMARK);
1563 	/* Program write watermark -- 1/8 of the FIFO. */
1564 	writel(cqspi->fifo_depth * cqspi->fifo_width / 8,
1565 	       cqspi->iobase + CQSPI_REG_INDIRECTWRWATERMARK);
1566 
1567 	/* Disable direct access controller */
1568 	if (!cqspi->use_direct_mode) {
1569 		reg = readl(cqspi->iobase + CQSPI_REG_CONFIG);
1570 		reg &= ~CQSPI_REG_CONFIG_ENB_DIR_ACC_CTRL;
1571 		writel(reg, cqspi->iobase + CQSPI_REG_CONFIG);
1572 	}
1573 
1574 	/* Enable DMA interface */
1575 	if (cqspi->use_dma_read) {
1576 		reg = readl(cqspi->iobase + CQSPI_REG_CONFIG);
1577 		reg |= CQSPI_REG_CONFIG_DMA_MASK;
1578 		writel(reg, cqspi->iobase + CQSPI_REG_CONFIG);
1579 	}
1580 
1581 	cqspi_controller_enable(cqspi, 1);
1582 }
1583 
1584 static int cqspi_request_mmap_dma(struct cqspi_st *cqspi)
1585 {
1586 	dma_cap_mask_t mask;
1587 
1588 	dma_cap_zero(mask);
1589 	dma_cap_set(DMA_MEMCPY, mask);
1590 
1591 	cqspi->rx_chan = dma_request_chan_by_mask(&mask);
1592 	if (IS_ERR(cqspi->rx_chan)) {
1593 		int ret = PTR_ERR(cqspi->rx_chan);
1594 
1595 		cqspi->rx_chan = NULL;
1596 		return dev_err_probe(&cqspi->pdev->dev, ret, "No Rx DMA available\n");
1597 	}
1598 	init_completion(&cqspi->rx_dma_complete);
1599 
1600 	return 0;
1601 }
1602 
1603 static const char *cqspi_get_name(struct spi_mem *mem)
1604 {
1605 	struct cqspi_st *cqspi = spi_controller_get_devdata(mem->spi->controller);
1606 	struct device *dev = &cqspi->pdev->dev;
1607 
1608 	return devm_kasprintf(dev, GFP_KERNEL, "%s.%d", dev_name(dev),
1609 			      spi_get_chipselect(mem->spi, 0));
1610 }
1611 
1612 static const struct spi_controller_mem_ops cqspi_mem_ops = {
1613 	.exec_op = cqspi_exec_mem_op,
1614 	.get_name = cqspi_get_name,
1615 	.supports_op = cqspi_supports_mem_op,
1616 };
1617 
1618 static const struct spi_controller_mem_caps cqspi_mem_caps = {
1619 	.dtr = true,
1620 };
1621 
1622 static int cqspi_setup_flash(struct cqspi_st *cqspi)
1623 {
1624 	unsigned int max_cs = cqspi->num_chipselect - 1;
1625 	struct platform_device *pdev = cqspi->pdev;
1626 	struct device *dev = &pdev->dev;
1627 	struct device_node *np = dev->of_node;
1628 	struct cqspi_flash_pdata *f_pdata;
1629 	unsigned int cs;
1630 	int ret;
1631 
1632 	/* Get flash device data */
1633 	for_each_available_child_of_node(dev->of_node, np) {
1634 		ret = of_property_read_u32(np, "reg", &cs);
1635 		if (ret) {
1636 			dev_err(dev, "Couldn't determine chip select.\n");
1637 			of_node_put(np);
1638 			return ret;
1639 		}
1640 
1641 		if (cs >= cqspi->num_chipselect) {
1642 			dev_err(dev, "Chip select %d out of range.\n", cs);
1643 			of_node_put(np);
1644 			return -EINVAL;
1645 		} else if (cs < max_cs) {
1646 			max_cs = cs;
1647 		}
1648 
1649 		f_pdata = &cqspi->f_pdata[cs];
1650 		f_pdata->cqspi = cqspi;
1651 		f_pdata->cs = cs;
1652 
1653 		ret = cqspi_of_get_flash_pdata(pdev, f_pdata, np);
1654 		if (ret) {
1655 			of_node_put(np);
1656 			return ret;
1657 		}
1658 	}
1659 
1660 	cqspi->num_chipselect = max_cs + 1;
1661 	return 0;
1662 }
1663 
1664 static int cqspi_jh7110_clk_init(struct platform_device *pdev, struct cqspi_st *cqspi)
1665 {
1666 	static struct clk_bulk_data qspiclk[] = {
1667 		{ .id = "apb" },
1668 		{ .id = "ahb" },
1669 	};
1670 
1671 	int ret = 0;
1672 
1673 	ret = devm_clk_bulk_get(&pdev->dev, ARRAY_SIZE(qspiclk), qspiclk);
1674 	if (ret) {
1675 		dev_err(&pdev->dev, "%s: failed to get qspi clocks\n", __func__);
1676 		return ret;
1677 	}
1678 
1679 	cqspi->clks[CLK_QSPI_APB] = qspiclk[0].clk;
1680 	cqspi->clks[CLK_QSPI_AHB] = qspiclk[1].clk;
1681 
1682 	ret = clk_prepare_enable(cqspi->clks[CLK_QSPI_APB]);
1683 	if (ret) {
1684 		dev_err(&pdev->dev, "%s: failed to enable CLK_QSPI_APB\n", __func__);
1685 		return ret;
1686 	}
1687 
1688 	ret = clk_prepare_enable(cqspi->clks[CLK_QSPI_AHB]);
1689 	if (ret) {
1690 		dev_err(&pdev->dev, "%s: failed to enable CLK_QSPI_AHB\n", __func__);
1691 		goto disable_apb_clk;
1692 	}
1693 
1694 	cqspi->is_jh7110 = true;
1695 
1696 	return 0;
1697 
1698 disable_apb_clk:
1699 	clk_disable_unprepare(cqspi->clks[CLK_QSPI_APB]);
1700 
1701 	return ret;
1702 }
1703 
1704 static void cqspi_jh7110_disable_clk(struct platform_device *pdev, struct cqspi_st *cqspi)
1705 {
1706 	clk_disable_unprepare(cqspi->clks[CLK_QSPI_AHB]);
1707 	clk_disable_unprepare(cqspi->clks[CLK_QSPI_APB]);
1708 }
1709 static int cqspi_probe(struct platform_device *pdev)
1710 {
1711 	const struct cqspi_driver_platdata *ddata;
1712 	struct reset_control *rstc, *rstc_ocp, *rstc_ref;
1713 	struct device *dev = &pdev->dev;
1714 	struct spi_controller *host;
1715 	struct resource *res_ahb;
1716 	struct cqspi_st *cqspi;
1717 	int ret;
1718 	int irq;
1719 
1720 	host = devm_spi_alloc_host(&pdev->dev, sizeof(*cqspi));
1721 	if (!host)
1722 		return -ENOMEM;
1723 
1724 	host->mode_bits = SPI_RX_QUAD | SPI_RX_DUAL;
1725 	host->mem_ops = &cqspi_mem_ops;
1726 	host->mem_caps = &cqspi_mem_caps;
1727 	host->dev.of_node = pdev->dev.of_node;
1728 
1729 	cqspi = spi_controller_get_devdata(host);
1730 
1731 	cqspi->pdev = pdev;
1732 	cqspi->host = host;
1733 	cqspi->is_jh7110 = false;
1734 	platform_set_drvdata(pdev, cqspi);
1735 
1736 	/* Obtain configuration from OF. */
1737 	ret = cqspi_of_get_pdata(cqspi);
1738 	if (ret) {
1739 		dev_err(dev, "Cannot get mandatory OF data.\n");
1740 		return -ENODEV;
1741 	}
1742 
1743 	/* Obtain QSPI clock. */
1744 	cqspi->clk = devm_clk_get(dev, NULL);
1745 	if (IS_ERR(cqspi->clk)) {
1746 		dev_err(dev, "Cannot claim QSPI clock.\n");
1747 		ret = PTR_ERR(cqspi->clk);
1748 		return ret;
1749 	}
1750 
1751 	/* Obtain and remap controller address. */
1752 	cqspi->iobase = devm_platform_ioremap_resource(pdev, 0);
1753 	if (IS_ERR(cqspi->iobase)) {
1754 		dev_err(dev, "Cannot remap controller address.\n");
1755 		ret = PTR_ERR(cqspi->iobase);
1756 		return ret;
1757 	}
1758 
1759 	/* Obtain and remap AHB address. */
1760 	cqspi->ahb_base = devm_platform_get_and_ioremap_resource(pdev, 1, &res_ahb);
1761 	if (IS_ERR(cqspi->ahb_base)) {
1762 		dev_err(dev, "Cannot remap AHB address.\n");
1763 		ret = PTR_ERR(cqspi->ahb_base);
1764 		return ret;
1765 	}
1766 	cqspi->mmap_phys_base = (dma_addr_t)res_ahb->start;
1767 	cqspi->ahb_size = resource_size(res_ahb);
1768 
1769 	init_completion(&cqspi->transfer_complete);
1770 
1771 	/* Obtain IRQ line. */
1772 	irq = platform_get_irq(pdev, 0);
1773 	if (irq < 0)
1774 		return -ENXIO;
1775 
1776 	ret = pm_runtime_set_active(dev);
1777 	if (ret)
1778 		return ret;
1779 
1780 
1781 	ret = clk_prepare_enable(cqspi->clk);
1782 	if (ret) {
1783 		dev_err(dev, "Cannot enable QSPI clock.\n");
1784 		goto probe_clk_failed;
1785 	}
1786 
1787 	/* Obtain QSPI reset control */
1788 	rstc = devm_reset_control_get_optional_exclusive(dev, "qspi");
1789 	if (IS_ERR(rstc)) {
1790 		ret = PTR_ERR(rstc);
1791 		dev_err(dev, "Cannot get QSPI reset.\n");
1792 		goto probe_reset_failed;
1793 	}
1794 
1795 	rstc_ocp = devm_reset_control_get_optional_exclusive(dev, "qspi-ocp");
1796 	if (IS_ERR(rstc_ocp)) {
1797 		ret = PTR_ERR(rstc_ocp);
1798 		dev_err(dev, "Cannot get QSPI OCP reset.\n");
1799 		goto probe_reset_failed;
1800 	}
1801 
1802 	if (of_device_is_compatible(pdev->dev.of_node, "starfive,jh7110-qspi")) {
1803 		rstc_ref = devm_reset_control_get_optional_exclusive(dev, "rstc_ref");
1804 		if (IS_ERR(rstc_ref)) {
1805 			ret = PTR_ERR(rstc_ref);
1806 			dev_err(dev, "Cannot get QSPI REF reset.\n");
1807 			goto probe_reset_failed;
1808 		}
1809 		reset_control_assert(rstc_ref);
1810 		reset_control_deassert(rstc_ref);
1811 	}
1812 
1813 	reset_control_assert(rstc);
1814 	reset_control_deassert(rstc);
1815 
1816 	reset_control_assert(rstc_ocp);
1817 	reset_control_deassert(rstc_ocp);
1818 
1819 	cqspi->master_ref_clk_hz = clk_get_rate(cqspi->clk);
1820 	host->max_speed_hz = cqspi->master_ref_clk_hz;
1821 
1822 	/* write completion is supported by default */
1823 	cqspi->wr_completion = true;
1824 
1825 	ddata  = of_device_get_match_data(dev);
1826 	if (ddata) {
1827 		if (ddata->quirks & CQSPI_NEEDS_WR_DELAY)
1828 			cqspi->wr_delay = 50 * DIV_ROUND_UP(NSEC_PER_SEC,
1829 						cqspi->master_ref_clk_hz);
1830 		if (ddata->hwcaps_mask & CQSPI_SUPPORTS_OCTAL)
1831 			host->mode_bits |= SPI_RX_OCTAL | SPI_TX_OCTAL;
1832 		if (!(ddata->quirks & CQSPI_DISABLE_DAC_MODE)) {
1833 			cqspi->use_direct_mode = true;
1834 			cqspi->use_direct_mode_wr = true;
1835 		}
1836 		if (ddata->quirks & CQSPI_SUPPORT_EXTERNAL_DMA)
1837 			cqspi->use_dma_read = true;
1838 		if (ddata->quirks & CQSPI_NO_SUPPORT_WR_COMPLETION)
1839 			cqspi->wr_completion = false;
1840 		if (ddata->quirks & CQSPI_SLOW_SRAM)
1841 			cqspi->slow_sram = true;
1842 		if (ddata->quirks & CQSPI_NEEDS_APB_AHB_HAZARD_WAR)
1843 			cqspi->apb_ahb_hazard = true;
1844 
1845 		if (ddata->jh7110_clk_init) {
1846 			ret = cqspi_jh7110_clk_init(pdev, cqspi);
1847 			if (ret)
1848 				goto probe_reset_failed;
1849 		}
1850 
1851 		if (of_device_is_compatible(pdev->dev.of_node,
1852 					    "xlnx,versal-ospi-1.0")) {
1853 			ret = dma_set_mask(&pdev->dev, DMA_BIT_MASK(64));
1854 			if (ret)
1855 				goto probe_reset_failed;
1856 		}
1857 	}
1858 
1859 	ret = devm_request_irq(dev, irq, cqspi_irq_handler, 0,
1860 			       pdev->name, cqspi);
1861 	if (ret) {
1862 		dev_err(dev, "Cannot request IRQ.\n");
1863 		goto probe_reset_failed;
1864 	}
1865 
1866 	cqspi_wait_idle(cqspi);
1867 	cqspi_controller_init(cqspi);
1868 	cqspi->current_cs = -1;
1869 	cqspi->sclk = 0;
1870 
1871 	ret = cqspi_setup_flash(cqspi);
1872 	if (ret) {
1873 		dev_err(dev, "failed to setup flash parameters %d\n", ret);
1874 		goto probe_setup_failed;
1875 	}
1876 
1877 	host->num_chipselect = cqspi->num_chipselect;
1878 
1879 	if (cqspi->use_direct_mode) {
1880 		ret = cqspi_request_mmap_dma(cqspi);
1881 		if (ret == -EPROBE_DEFER)
1882 			goto probe_setup_failed;
1883 	}
1884 
1885 	ret = devm_pm_runtime_enable(dev);
1886 	if (ret) {
1887 		if (cqspi->rx_chan)
1888 			dma_release_channel(cqspi->rx_chan);
1889 		goto probe_setup_failed;
1890 	}
1891 
1892 	pm_runtime_set_autosuspend_delay(dev, CQSPI_AUTOSUSPEND_TIMEOUT);
1893 	pm_runtime_use_autosuspend(dev);
1894 	pm_runtime_get_noresume(dev);
1895 
1896 	ret = spi_register_controller(host);
1897 	if (ret) {
1898 		dev_err(&pdev->dev, "failed to register SPI ctlr %d\n", ret);
1899 		goto probe_setup_failed;
1900 	}
1901 
1902 	pm_runtime_mark_last_busy(dev);
1903 	pm_runtime_put_autosuspend(dev);
1904 
1905 	return 0;
1906 probe_setup_failed:
1907 	cqspi_controller_enable(cqspi, 0);
1908 probe_reset_failed:
1909 	if (cqspi->is_jh7110)
1910 		cqspi_jh7110_disable_clk(pdev, cqspi);
1911 	clk_disable_unprepare(cqspi->clk);
1912 probe_clk_failed:
1913 	return ret;
1914 }
1915 
1916 static void cqspi_remove(struct platform_device *pdev)
1917 {
1918 	struct cqspi_st *cqspi = platform_get_drvdata(pdev);
1919 
1920 	spi_unregister_controller(cqspi->host);
1921 	cqspi_controller_enable(cqspi, 0);
1922 
1923 	if (cqspi->rx_chan)
1924 		dma_release_channel(cqspi->rx_chan);
1925 
1926 	clk_disable_unprepare(cqspi->clk);
1927 
1928 	if (cqspi->is_jh7110)
1929 		cqspi_jh7110_disable_clk(pdev, cqspi);
1930 
1931 	pm_runtime_put_sync(&pdev->dev);
1932 	pm_runtime_disable(&pdev->dev);
1933 }
1934 
1935 static int cqspi_runtime_suspend(struct device *dev)
1936 {
1937 	struct cqspi_st *cqspi = dev_get_drvdata(dev);
1938 
1939 	cqspi_controller_enable(cqspi, 0);
1940 	clk_disable_unprepare(cqspi->clk);
1941 	return 0;
1942 }
1943 
1944 static int cqspi_runtime_resume(struct device *dev)
1945 {
1946 	struct cqspi_st *cqspi = dev_get_drvdata(dev);
1947 
1948 	clk_prepare_enable(cqspi->clk);
1949 	cqspi_wait_idle(cqspi);
1950 	cqspi_controller_init(cqspi);
1951 
1952 	cqspi->current_cs = -1;
1953 	cqspi->sclk = 0;
1954 	return 0;
1955 }
1956 
1957 static int cqspi_suspend(struct device *dev)
1958 {
1959 	struct cqspi_st *cqspi = dev_get_drvdata(dev);
1960 
1961 	return spi_controller_suspend(cqspi->host);
1962 }
1963 
1964 static int cqspi_resume(struct device *dev)
1965 {
1966 	struct cqspi_st *cqspi = dev_get_drvdata(dev);
1967 
1968 	return spi_controller_resume(cqspi->host);
1969 }
1970 
1971 static const struct dev_pm_ops cqspi_dev_pm_ops = {
1972 	RUNTIME_PM_OPS(cqspi_runtime_suspend, cqspi_runtime_resume, NULL)
1973 	SYSTEM_SLEEP_PM_OPS(cqspi_suspend, cqspi_resume)
1974 };
1975 
1976 static const struct cqspi_driver_platdata cdns_qspi = {
1977 	.quirks = CQSPI_DISABLE_DAC_MODE,
1978 };
1979 
1980 static const struct cqspi_driver_platdata k2g_qspi = {
1981 	.quirks = CQSPI_NEEDS_WR_DELAY,
1982 };
1983 
1984 static const struct cqspi_driver_platdata am654_ospi = {
1985 	.hwcaps_mask = CQSPI_SUPPORTS_OCTAL,
1986 	.quirks = CQSPI_NEEDS_WR_DELAY,
1987 };
1988 
1989 static const struct cqspi_driver_platdata intel_lgm_qspi = {
1990 	.quirks = CQSPI_DISABLE_DAC_MODE,
1991 };
1992 
1993 static const struct cqspi_driver_platdata socfpga_qspi = {
1994 	.quirks = CQSPI_DISABLE_DAC_MODE
1995 			| CQSPI_NO_SUPPORT_WR_COMPLETION
1996 			| CQSPI_SLOW_SRAM,
1997 };
1998 
1999 static const struct cqspi_driver_platdata versal_ospi = {
2000 	.hwcaps_mask = CQSPI_SUPPORTS_OCTAL,
2001 	.quirks = CQSPI_DISABLE_DAC_MODE | CQSPI_SUPPORT_EXTERNAL_DMA,
2002 	.indirect_read_dma = cqspi_versal_indirect_read_dma,
2003 	.get_dma_status = cqspi_get_versal_dma_status,
2004 };
2005 
2006 static const struct cqspi_driver_platdata jh7110_qspi = {
2007 	.quirks = CQSPI_DISABLE_DAC_MODE,
2008 	.jh7110_clk_init = cqspi_jh7110_clk_init,
2009 };
2010 
2011 static const struct cqspi_driver_platdata pensando_cdns_qspi = {
2012 	.quirks = CQSPI_NEEDS_APB_AHB_HAZARD_WAR | CQSPI_DISABLE_DAC_MODE,
2013 };
2014 
2015 static const struct of_device_id cqspi_dt_ids[] = {
2016 	{
2017 		.compatible = "cdns,qspi-nor",
2018 		.data = &cdns_qspi,
2019 	},
2020 	{
2021 		.compatible = "ti,k2g-qspi",
2022 		.data = &k2g_qspi,
2023 	},
2024 	{
2025 		.compatible = "ti,am654-ospi",
2026 		.data = &am654_ospi,
2027 	},
2028 	{
2029 		.compatible = "intel,lgm-qspi",
2030 		.data = &intel_lgm_qspi,
2031 	},
2032 	{
2033 		.compatible = "xlnx,versal-ospi-1.0",
2034 		.data = &versal_ospi,
2035 	},
2036 	{
2037 		.compatible = "intel,socfpga-qspi",
2038 		.data = &socfpga_qspi,
2039 	},
2040 	{
2041 		.compatible = "starfive,jh7110-qspi",
2042 		.data = &jh7110_qspi,
2043 	},
2044 	{
2045 		.compatible = "amd,pensando-elba-qspi",
2046 		.data = &pensando_cdns_qspi,
2047 	},
2048 	{ /* end of table */ }
2049 };
2050 
2051 MODULE_DEVICE_TABLE(of, cqspi_dt_ids);
2052 
2053 static struct platform_driver cqspi_platform_driver = {
2054 	.probe = cqspi_probe,
2055 	.remove_new = cqspi_remove,
2056 	.driver = {
2057 		.name = CQSPI_NAME,
2058 		.pm = pm_ptr(&cqspi_dev_pm_ops),
2059 		.of_match_table = cqspi_dt_ids,
2060 	},
2061 };
2062 
2063 module_platform_driver(cqspi_platform_driver);
2064 
2065 MODULE_DESCRIPTION("Cadence QSPI Controller Driver");
2066 MODULE_LICENSE("GPL v2");
2067 MODULE_ALIAS("platform:" CQSPI_NAME);
2068 MODULE_AUTHOR("Ley Foon Tan <lftan@altera.com>");
2069 MODULE_AUTHOR("Graham Moore <grmoore@opensource.altera.com>");
2070 MODULE_AUTHOR("Vadivel Murugan R <vadivel.muruganx.ramuthevar@intel.com>");
2071 MODULE_AUTHOR("Vignesh Raghavendra <vigneshr@ti.com>");
2072 MODULE_AUTHOR("Pratyush Yadav <p.yadav@ti.com>");
2073