1 // SPDX-License-Identifier: GPL-2.0-only
2 //
3 // Copyright (C) 2020 NVIDIA CORPORATION.
4 
5 #include <linux/clk.h>
6 #include <linux/completion.h>
7 #include <linux/delay.h>
8 #include <linux/dmaengine.h>
9 #include <linux/dma-mapping.h>
10 #include <linux/dmapool.h>
11 #include <linux/err.h>
12 #include <linux/interrupt.h>
13 #include <linux/io.h>
14 #include <linux/iopoll.h>
15 #include <linux/kernel.h>
16 #include <linux/kthread.h>
17 #include <linux/module.h>
18 #include <linux/platform_device.h>
19 #include <linux/pm_runtime.h>
20 #include <linux/of.h>
21 #include <linux/of_device.h>
22 #include <linux/reset.h>
23 #include <linux/spi/spi.h>
24 #include <linux/acpi.h>
25 #include <linux/property.h>
26 
27 #define QSPI_COMMAND1				0x000
28 #define QSPI_BIT_LENGTH(x)			(((x) & 0x1f) << 0)
29 #define QSPI_PACKED				BIT(5)
30 #define QSPI_INTERFACE_WIDTH_MASK		(0x03 << 7)
31 #define QSPI_INTERFACE_WIDTH(x)			(((x) & 0x03) << 7)
32 #define QSPI_INTERFACE_WIDTH_SINGLE		QSPI_INTERFACE_WIDTH(0)
33 #define QSPI_INTERFACE_WIDTH_DUAL		QSPI_INTERFACE_WIDTH(1)
34 #define QSPI_INTERFACE_WIDTH_QUAD		QSPI_INTERFACE_WIDTH(2)
35 #define QSPI_SDR_DDR_SEL			BIT(9)
36 #define QSPI_TX_EN				BIT(11)
37 #define QSPI_RX_EN				BIT(12)
38 #define QSPI_CS_SW_VAL				BIT(20)
39 #define QSPI_CS_SW_HW				BIT(21)
40 
41 #define QSPI_CS_POL_INACTIVE(n)			(1 << (22 + (n)))
42 #define QSPI_CS_POL_INACTIVE_MASK		(0xF << 22)
43 #define QSPI_CS_SEL_0				(0 << 26)
44 #define QSPI_CS_SEL_1				(1 << 26)
45 #define QSPI_CS_SEL_2				(2 << 26)
46 #define QSPI_CS_SEL_3				(3 << 26)
47 #define QSPI_CS_SEL_MASK			(3 << 26)
48 #define QSPI_CS_SEL(x)				(((x) & 0x3) << 26)
49 
50 #define QSPI_CONTROL_MODE_0			(0 << 28)
51 #define QSPI_CONTROL_MODE_3			(3 << 28)
52 #define QSPI_CONTROL_MODE_MASK			(3 << 28)
53 #define QSPI_M_S				BIT(30)
54 #define QSPI_PIO				BIT(31)
55 
56 #define QSPI_COMMAND2				0x004
57 #define QSPI_TX_TAP_DELAY(x)			(((x) & 0x3f) << 10)
58 #define QSPI_RX_TAP_DELAY(x)			(((x) & 0xff) << 0)
59 
60 #define QSPI_CS_TIMING1				0x008
61 #define QSPI_SETUP_HOLD(setup, hold)		(((setup) << 4) | (hold))
62 
63 #define QSPI_CS_TIMING2				0x00c
64 #define CYCLES_BETWEEN_PACKETS_0(x)		(((x) & 0x1f) << 0)
65 #define CS_ACTIVE_BETWEEN_PACKETS_0		BIT(5)
66 
67 #define QSPI_TRANS_STATUS			0x010
68 #define QSPI_BLK_CNT(val)			(((val) >> 0) & 0xffff)
69 #define QSPI_RDY				BIT(30)
70 
71 #define QSPI_FIFO_STATUS			0x014
72 #define QSPI_RX_FIFO_EMPTY			BIT(0)
73 #define QSPI_RX_FIFO_FULL			BIT(1)
74 #define QSPI_TX_FIFO_EMPTY			BIT(2)
75 #define QSPI_TX_FIFO_FULL			BIT(3)
76 #define QSPI_RX_FIFO_UNF			BIT(4)
77 #define QSPI_RX_FIFO_OVF			BIT(5)
78 #define QSPI_TX_FIFO_UNF			BIT(6)
79 #define QSPI_TX_FIFO_OVF			BIT(7)
80 #define QSPI_ERR				BIT(8)
81 #define QSPI_TX_FIFO_FLUSH			BIT(14)
82 #define QSPI_RX_FIFO_FLUSH			BIT(15)
83 #define QSPI_TX_FIFO_EMPTY_COUNT(val)		(((val) >> 16) & 0x7f)
84 #define QSPI_RX_FIFO_FULL_COUNT(val)		(((val) >> 23) & 0x7f)
85 
86 #define QSPI_FIFO_ERROR				(QSPI_RX_FIFO_UNF | \
87 						 QSPI_RX_FIFO_OVF | \
88 						 QSPI_TX_FIFO_UNF | \
89 						 QSPI_TX_FIFO_OVF)
90 #define QSPI_FIFO_EMPTY				(QSPI_RX_FIFO_EMPTY | \
91 						 QSPI_TX_FIFO_EMPTY)
92 
93 #define QSPI_TX_DATA				0x018
94 #define QSPI_RX_DATA				0x01c
95 
96 #define QSPI_DMA_CTL				0x020
97 #define QSPI_TX_TRIG(n)				(((n) & 0x3) << 15)
98 #define QSPI_TX_TRIG_1				QSPI_TX_TRIG(0)
99 #define QSPI_TX_TRIG_4				QSPI_TX_TRIG(1)
100 #define QSPI_TX_TRIG_8				QSPI_TX_TRIG(2)
101 #define QSPI_TX_TRIG_16				QSPI_TX_TRIG(3)
102 
103 #define QSPI_RX_TRIG(n)				(((n) & 0x3) << 19)
104 #define QSPI_RX_TRIG_1				QSPI_RX_TRIG(0)
105 #define QSPI_RX_TRIG_4				QSPI_RX_TRIG(1)
106 #define QSPI_RX_TRIG_8				QSPI_RX_TRIG(2)
107 #define QSPI_RX_TRIG_16				QSPI_RX_TRIG(3)
108 
109 #define QSPI_DMA_EN				BIT(31)
110 
111 #define QSPI_DMA_BLK				0x024
112 #define QSPI_DMA_BLK_SET(x)			(((x) & 0xffff) << 0)
113 
114 #define QSPI_TX_FIFO				0x108
115 #define QSPI_RX_FIFO				0x188
116 
117 #define QSPI_FIFO_DEPTH				64
118 
119 #define QSPI_INTR_MASK				0x18c
120 #define QSPI_INTR_RX_FIFO_UNF_MASK		BIT(25)
121 #define QSPI_INTR_RX_FIFO_OVF_MASK		BIT(26)
122 #define QSPI_INTR_TX_FIFO_UNF_MASK		BIT(27)
123 #define QSPI_INTR_TX_FIFO_OVF_MASK		BIT(28)
124 #define QSPI_INTR_RDY_MASK			BIT(29)
125 #define QSPI_INTR_RX_TX_FIFO_ERR		(QSPI_INTR_RX_FIFO_UNF_MASK | \
126 						 QSPI_INTR_RX_FIFO_OVF_MASK | \
127 						 QSPI_INTR_TX_FIFO_UNF_MASK | \
128 						 QSPI_INTR_TX_FIFO_OVF_MASK)
129 
130 #define QSPI_MISC_REG                           0x194
131 #define QSPI_NUM_DUMMY_CYCLE(x)			(((x) & 0xff) << 0)
132 #define QSPI_DUMMY_CYCLES_MAX			0xff
133 
134 #define QSPI_CMB_SEQ_CMD			0x19c
135 #define QSPI_COMMAND_VALUE_SET(X)		(((x) & 0xFF) << 0)
136 
137 #define QSPI_CMB_SEQ_CMD_CFG			0x1a0
138 #define QSPI_COMMAND_X1_X2_X4(x)		(((x) & 0x3) << 13)
139 #define QSPI_COMMAND_X1_X2_X4_MASK		(0x03 << 13)
140 #define QSPI_COMMAND_SDR_DDR			BIT(12)
141 #define QSPI_COMMAND_SIZE_SET(x)		(((x) & 0xFF) << 0)
142 
143 #define QSPI_GLOBAL_CONFIG			0X1a4
144 #define QSPI_CMB_SEQ_EN				BIT(0)
145 
146 #define QSPI_CMB_SEQ_ADDR			0x1a8
147 #define QSPI_ADDRESS_VALUE_SET(X)		(((x) & 0xFFFF) << 0)
148 
149 #define QSPI_CMB_SEQ_ADDR_CFG			0x1ac
150 #define QSPI_ADDRESS_X1_X2_X4(x)		(((x) & 0x3) << 13)
151 #define QSPI_ADDRESS_X1_X2_X4_MASK		(0x03 << 13)
152 #define QSPI_ADDRESS_SDR_DDR			BIT(12)
153 #define QSPI_ADDRESS_SIZE_SET(x)		(((x) & 0xFF) << 0)
154 
155 #define DATA_DIR_TX				BIT(0)
156 #define DATA_DIR_RX				BIT(1)
157 
158 #define QSPI_DMA_TIMEOUT			(msecs_to_jiffies(1000))
159 #define DEFAULT_QSPI_DMA_BUF_LEN		(64 * 1024)
160 #define CMD_TRANSFER				0
161 #define ADDR_TRANSFER				1
162 #define DATA_TRANSFER				2
163 
164 struct tegra_qspi_soc_data {
165 	bool has_dma;
166 	bool cmb_xfer_capable;
167 	unsigned int cs_count;
168 };
169 
170 struct tegra_qspi_client_data {
171 	int tx_clk_tap_delay;
172 	int rx_clk_tap_delay;
173 };
174 
175 struct tegra_qspi {
176 	struct device				*dev;
177 	struct spi_master			*master;
178 	/* lock to protect data accessed by irq */
179 	spinlock_t				lock;
180 
181 	struct clk				*clk;
182 	void __iomem				*base;
183 	phys_addr_t				phys;
184 	unsigned int				irq;
185 
186 	u32					cur_speed;
187 	unsigned int				cur_pos;
188 	unsigned int				words_per_32bit;
189 	unsigned int				bytes_per_word;
190 	unsigned int				curr_dma_words;
191 	unsigned int				cur_direction;
192 
193 	unsigned int				cur_rx_pos;
194 	unsigned int				cur_tx_pos;
195 
196 	unsigned int				dma_buf_size;
197 	unsigned int				max_buf_size;
198 	bool					is_curr_dma_xfer;
199 
200 	struct completion			rx_dma_complete;
201 	struct completion			tx_dma_complete;
202 
203 	u32					tx_status;
204 	u32					rx_status;
205 	u32					status_reg;
206 	bool					is_packed;
207 	bool					use_dma;
208 
209 	u32					command1_reg;
210 	u32					dma_control_reg;
211 	u32					def_command1_reg;
212 	u32					def_command2_reg;
213 	u32					spi_cs_timing1;
214 	u32					spi_cs_timing2;
215 	u8					dummy_cycles;
216 
217 	struct completion			xfer_completion;
218 	struct spi_transfer			*curr_xfer;
219 
220 	struct dma_chan				*rx_dma_chan;
221 	u32					*rx_dma_buf;
222 	dma_addr_t				rx_dma_phys;
223 	struct dma_async_tx_descriptor		*rx_dma_desc;
224 
225 	struct dma_chan				*tx_dma_chan;
226 	u32					*tx_dma_buf;
227 	dma_addr_t				tx_dma_phys;
228 	struct dma_async_tx_descriptor		*tx_dma_desc;
229 	const struct tegra_qspi_soc_data	*soc_data;
230 };
231 
232 static inline u32 tegra_qspi_readl(struct tegra_qspi *tqspi, unsigned long offset)
233 {
234 	return readl(tqspi->base + offset);
235 }
236 
237 static inline void tegra_qspi_writel(struct tegra_qspi *tqspi, u32 value, unsigned long offset)
238 {
239 	writel(value, tqspi->base + offset);
240 
241 	/* read back register to make sure that register writes completed */
242 	if (offset != QSPI_TX_FIFO)
243 		readl(tqspi->base + QSPI_COMMAND1);
244 }
245 
246 static void tegra_qspi_mask_clear_irq(struct tegra_qspi *tqspi)
247 {
248 	u32 value;
249 
250 	/* write 1 to clear status register */
251 	value = tegra_qspi_readl(tqspi, QSPI_TRANS_STATUS);
252 	tegra_qspi_writel(tqspi, value, QSPI_TRANS_STATUS);
253 
254 	value = tegra_qspi_readl(tqspi, QSPI_INTR_MASK);
255 	if (!(value & QSPI_INTR_RDY_MASK)) {
256 		value |= (QSPI_INTR_RDY_MASK | QSPI_INTR_RX_TX_FIFO_ERR);
257 		tegra_qspi_writel(tqspi, value, QSPI_INTR_MASK);
258 	}
259 
260 	/* clear fifo status error if any */
261 	value = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
262 	if (value & QSPI_ERR)
263 		tegra_qspi_writel(tqspi, QSPI_ERR | QSPI_FIFO_ERROR, QSPI_FIFO_STATUS);
264 }
265 
266 static unsigned int
267 tegra_qspi_calculate_curr_xfer_param(struct tegra_qspi *tqspi, struct spi_transfer *t)
268 {
269 	unsigned int max_word, max_len, total_fifo_words;
270 	unsigned int remain_len = t->len - tqspi->cur_pos;
271 	unsigned int bits_per_word = t->bits_per_word;
272 
273 	tqspi->bytes_per_word = DIV_ROUND_UP(bits_per_word, 8);
274 
275 	/*
276 	 * Tegra QSPI controller supports packed or unpacked mode transfers.
277 	 * Packed mode is used for data transfers using 8, 16, or 32 bits per
278 	 * word with a minimum transfer of 1 word and for all other transfers
279 	 * unpacked mode will be used.
280 	 */
281 
282 	if ((bits_per_word == 8 || bits_per_word == 16 ||
283 	     bits_per_word == 32) && t->len > 3) {
284 		tqspi->is_packed = true;
285 		tqspi->words_per_32bit = 32 / bits_per_word;
286 	} else {
287 		tqspi->is_packed = false;
288 		tqspi->words_per_32bit = 1;
289 	}
290 
291 	if (tqspi->is_packed) {
292 		max_len = min(remain_len, tqspi->max_buf_size);
293 		tqspi->curr_dma_words = max_len / tqspi->bytes_per_word;
294 		total_fifo_words = (max_len + 3) / 4;
295 	} else {
296 		max_word = (remain_len - 1) / tqspi->bytes_per_word + 1;
297 		max_word = min(max_word, tqspi->max_buf_size / 4);
298 		tqspi->curr_dma_words = max_word;
299 		total_fifo_words = max_word;
300 	}
301 
302 	return total_fifo_words;
303 }
304 
305 static unsigned int
306 tegra_qspi_fill_tx_fifo_from_client_txbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
307 {
308 	unsigned int written_words, fifo_words_left, count;
309 	unsigned int len, tx_empty_count, max_n_32bit, i;
310 	u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
311 	u32 fifo_status;
312 
313 	fifo_status = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
314 	tx_empty_count = QSPI_TX_FIFO_EMPTY_COUNT(fifo_status);
315 
316 	if (tqspi->is_packed) {
317 		fifo_words_left = tx_empty_count * tqspi->words_per_32bit;
318 		written_words = min(fifo_words_left, tqspi->curr_dma_words);
319 		len = written_words * tqspi->bytes_per_word;
320 		max_n_32bit = DIV_ROUND_UP(len, 4);
321 		for (count = 0; count < max_n_32bit; count++) {
322 			u32 x = 0;
323 
324 			for (i = 0; (i < 4) && len; i++, len--)
325 				x |= (u32)(*tx_buf++) << (i * 8);
326 			tegra_qspi_writel(tqspi, x, QSPI_TX_FIFO);
327 		}
328 
329 		tqspi->cur_tx_pos += written_words * tqspi->bytes_per_word;
330 	} else {
331 		unsigned int write_bytes;
332 		u8 bytes_per_word = tqspi->bytes_per_word;
333 
334 		max_n_32bit = min(tqspi->curr_dma_words, tx_empty_count);
335 		written_words = max_n_32bit;
336 		len = written_words * tqspi->bytes_per_word;
337 		if (len > t->len - tqspi->cur_pos)
338 			len = t->len - tqspi->cur_pos;
339 		write_bytes = len;
340 		for (count = 0; count < max_n_32bit; count++) {
341 			u32 x = 0;
342 
343 			for (i = 0; len && (i < bytes_per_word); i++, len--)
344 				x |= (u32)(*tx_buf++) << (i * 8);
345 			tegra_qspi_writel(tqspi, x, QSPI_TX_FIFO);
346 		}
347 
348 		tqspi->cur_tx_pos += write_bytes;
349 	}
350 
351 	return written_words;
352 }
353 
354 static unsigned int
355 tegra_qspi_read_rx_fifo_to_client_rxbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
356 {
357 	u8 *rx_buf = (u8 *)t->rx_buf + tqspi->cur_rx_pos;
358 	unsigned int len, rx_full_count, count, i;
359 	unsigned int read_words = 0;
360 	u32 fifo_status, x;
361 
362 	fifo_status = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
363 	rx_full_count = QSPI_RX_FIFO_FULL_COUNT(fifo_status);
364 	if (tqspi->is_packed) {
365 		len = tqspi->curr_dma_words * tqspi->bytes_per_word;
366 		for (count = 0; count < rx_full_count; count++) {
367 			x = tegra_qspi_readl(tqspi, QSPI_RX_FIFO);
368 
369 			for (i = 0; len && (i < 4); i++, len--)
370 				*rx_buf++ = (x >> i * 8) & 0xff;
371 		}
372 
373 		read_words += tqspi->curr_dma_words;
374 		tqspi->cur_rx_pos += tqspi->curr_dma_words * tqspi->bytes_per_word;
375 	} else {
376 		u32 rx_mask = ((u32)1 << t->bits_per_word) - 1;
377 		u8 bytes_per_word = tqspi->bytes_per_word;
378 		unsigned int read_bytes;
379 
380 		len = rx_full_count * bytes_per_word;
381 		if (len > t->len - tqspi->cur_pos)
382 			len = t->len - tqspi->cur_pos;
383 		read_bytes = len;
384 		for (count = 0; count < rx_full_count; count++) {
385 			x = tegra_qspi_readl(tqspi, QSPI_RX_FIFO) & rx_mask;
386 
387 			for (i = 0; len && (i < bytes_per_word); i++, len--)
388 				*rx_buf++ = (x >> (i * 8)) & 0xff;
389 		}
390 
391 		read_words += rx_full_count;
392 		tqspi->cur_rx_pos += read_bytes;
393 	}
394 
395 	return read_words;
396 }
397 
398 static void
399 tegra_qspi_copy_client_txbuf_to_qspi_txbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
400 {
401 	dma_sync_single_for_cpu(tqspi->dev, tqspi->tx_dma_phys,
402 				tqspi->dma_buf_size, DMA_TO_DEVICE);
403 
404 	/*
405 	 * In packed mode, each word in FIFO may contain multiple packets
406 	 * based on bits per word. So all bytes in each FIFO word are valid.
407 	 *
408 	 * In unpacked mode, each word in FIFO contains single packet and
409 	 * based on bits per word any remaining bits in FIFO word will be
410 	 * ignored by the hardware and are invalid bits.
411 	 */
412 	if (tqspi->is_packed) {
413 		tqspi->cur_tx_pos += tqspi->curr_dma_words * tqspi->bytes_per_word;
414 	} else {
415 		u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
416 		unsigned int i, count, consume, write_bytes;
417 
418 		/*
419 		 * Fill tx_dma_buf to contain single packet in each word based
420 		 * on bits per word from SPI core tx_buf.
421 		 */
422 		consume = tqspi->curr_dma_words * tqspi->bytes_per_word;
423 		if (consume > t->len - tqspi->cur_pos)
424 			consume = t->len - tqspi->cur_pos;
425 		write_bytes = consume;
426 		for (count = 0; count < tqspi->curr_dma_words; count++) {
427 			u32 x = 0;
428 
429 			for (i = 0; consume && (i < tqspi->bytes_per_word); i++, consume--)
430 				x |= (u32)(*tx_buf++) << (i * 8);
431 			tqspi->tx_dma_buf[count] = x;
432 		}
433 
434 		tqspi->cur_tx_pos += write_bytes;
435 	}
436 
437 	dma_sync_single_for_device(tqspi->dev, tqspi->tx_dma_phys,
438 				   tqspi->dma_buf_size, DMA_TO_DEVICE);
439 }
440 
441 static void
442 tegra_qspi_copy_qspi_rxbuf_to_client_rxbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
443 {
444 	dma_sync_single_for_cpu(tqspi->dev, tqspi->rx_dma_phys,
445 				tqspi->dma_buf_size, DMA_FROM_DEVICE);
446 
447 	if (tqspi->is_packed) {
448 		tqspi->cur_rx_pos += tqspi->curr_dma_words * tqspi->bytes_per_word;
449 	} else {
450 		unsigned char *rx_buf = t->rx_buf + tqspi->cur_rx_pos;
451 		u32 rx_mask = ((u32)1 << t->bits_per_word) - 1;
452 		unsigned int i, count, consume, read_bytes;
453 
454 		/*
455 		 * Each FIFO word contains single data packet.
456 		 * Skip invalid bits in each FIFO word based on bits per word
457 		 * and align bytes while filling in SPI core rx_buf.
458 		 */
459 		consume = tqspi->curr_dma_words * tqspi->bytes_per_word;
460 		if (consume > t->len - tqspi->cur_pos)
461 			consume = t->len - tqspi->cur_pos;
462 		read_bytes = consume;
463 		for (count = 0; count < tqspi->curr_dma_words; count++) {
464 			u32 x = tqspi->rx_dma_buf[count] & rx_mask;
465 
466 			for (i = 0; consume && (i < tqspi->bytes_per_word); i++, consume--)
467 				*rx_buf++ = (x >> (i * 8)) & 0xff;
468 		}
469 
470 		tqspi->cur_rx_pos += read_bytes;
471 	}
472 
473 	dma_sync_single_for_device(tqspi->dev, tqspi->rx_dma_phys,
474 				   tqspi->dma_buf_size, DMA_FROM_DEVICE);
475 }
476 
477 static void tegra_qspi_dma_complete(void *args)
478 {
479 	struct completion *dma_complete = args;
480 
481 	complete(dma_complete);
482 }
483 
484 static int tegra_qspi_start_tx_dma(struct tegra_qspi *tqspi, struct spi_transfer *t, int len)
485 {
486 	dma_addr_t tx_dma_phys;
487 
488 	reinit_completion(&tqspi->tx_dma_complete);
489 
490 	if (tqspi->is_packed)
491 		tx_dma_phys = t->tx_dma;
492 	else
493 		tx_dma_phys = tqspi->tx_dma_phys;
494 
495 	tqspi->tx_dma_desc = dmaengine_prep_slave_single(tqspi->tx_dma_chan, tx_dma_phys,
496 							 len, DMA_MEM_TO_DEV,
497 							 DMA_PREP_INTERRUPT |  DMA_CTRL_ACK);
498 
499 	if (!tqspi->tx_dma_desc) {
500 		dev_err(tqspi->dev, "Unable to get TX descriptor\n");
501 		return -EIO;
502 	}
503 
504 	tqspi->tx_dma_desc->callback = tegra_qspi_dma_complete;
505 	tqspi->tx_dma_desc->callback_param = &tqspi->tx_dma_complete;
506 	dmaengine_submit(tqspi->tx_dma_desc);
507 	dma_async_issue_pending(tqspi->tx_dma_chan);
508 
509 	return 0;
510 }
511 
512 static int tegra_qspi_start_rx_dma(struct tegra_qspi *tqspi, struct spi_transfer *t, int len)
513 {
514 	dma_addr_t rx_dma_phys;
515 
516 	reinit_completion(&tqspi->rx_dma_complete);
517 
518 	if (tqspi->is_packed)
519 		rx_dma_phys = t->rx_dma;
520 	else
521 		rx_dma_phys = tqspi->rx_dma_phys;
522 
523 	tqspi->rx_dma_desc = dmaengine_prep_slave_single(tqspi->rx_dma_chan, rx_dma_phys,
524 							 len, DMA_DEV_TO_MEM,
525 							 DMA_PREP_INTERRUPT |  DMA_CTRL_ACK);
526 
527 	if (!tqspi->rx_dma_desc) {
528 		dev_err(tqspi->dev, "Unable to get RX descriptor\n");
529 		return -EIO;
530 	}
531 
532 	tqspi->rx_dma_desc->callback = tegra_qspi_dma_complete;
533 	tqspi->rx_dma_desc->callback_param = &tqspi->rx_dma_complete;
534 	dmaengine_submit(tqspi->rx_dma_desc);
535 	dma_async_issue_pending(tqspi->rx_dma_chan);
536 
537 	return 0;
538 }
539 
540 static int tegra_qspi_flush_fifos(struct tegra_qspi *tqspi, bool atomic)
541 {
542 	void __iomem *addr = tqspi->base + QSPI_FIFO_STATUS;
543 	u32 val;
544 
545 	val = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
546 	if ((val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY)
547 		return 0;
548 
549 	val |= QSPI_RX_FIFO_FLUSH | QSPI_TX_FIFO_FLUSH;
550 	tegra_qspi_writel(tqspi, val, QSPI_FIFO_STATUS);
551 
552 	if (!atomic)
553 		return readl_relaxed_poll_timeout(addr, val,
554 						  (val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY,
555 						  1000, 1000000);
556 
557 	return readl_relaxed_poll_timeout_atomic(addr, val,
558 						 (val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY,
559 						 1000, 1000000);
560 }
561 
562 static void tegra_qspi_unmask_irq(struct tegra_qspi *tqspi)
563 {
564 	u32 intr_mask;
565 
566 	intr_mask = tegra_qspi_readl(tqspi, QSPI_INTR_MASK);
567 	intr_mask &= ~(QSPI_INTR_RDY_MASK | QSPI_INTR_RX_TX_FIFO_ERR);
568 	tegra_qspi_writel(tqspi, intr_mask, QSPI_INTR_MASK);
569 }
570 
571 static int tegra_qspi_dma_map_xfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
572 {
573 	u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
574 	u8 *rx_buf = (u8 *)t->rx_buf + tqspi->cur_rx_pos;
575 	unsigned int len;
576 
577 	len = DIV_ROUND_UP(tqspi->curr_dma_words * tqspi->bytes_per_word, 4) * 4;
578 
579 	if (t->tx_buf) {
580 		t->tx_dma = dma_map_single(tqspi->dev, (void *)tx_buf, len, DMA_TO_DEVICE);
581 		if (dma_mapping_error(tqspi->dev, t->tx_dma))
582 			return -ENOMEM;
583 	}
584 
585 	if (t->rx_buf) {
586 		t->rx_dma = dma_map_single(tqspi->dev, (void *)rx_buf, len, DMA_FROM_DEVICE);
587 		if (dma_mapping_error(tqspi->dev, t->rx_dma)) {
588 			dma_unmap_single(tqspi->dev, t->tx_dma, len, DMA_TO_DEVICE);
589 			return -ENOMEM;
590 		}
591 	}
592 
593 	return 0;
594 }
595 
596 static void tegra_qspi_dma_unmap_xfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
597 {
598 	unsigned int len;
599 
600 	len = DIV_ROUND_UP(tqspi->curr_dma_words * tqspi->bytes_per_word, 4) * 4;
601 
602 	dma_unmap_single(tqspi->dev, t->tx_dma, len, DMA_TO_DEVICE);
603 	dma_unmap_single(tqspi->dev, t->rx_dma, len, DMA_FROM_DEVICE);
604 }
605 
606 static int tegra_qspi_start_dma_based_transfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
607 {
608 	struct dma_slave_config dma_sconfig = { 0 };
609 	unsigned int len;
610 	u8 dma_burst;
611 	int ret = 0;
612 	u32 val;
613 
614 	if (tqspi->is_packed) {
615 		ret = tegra_qspi_dma_map_xfer(tqspi, t);
616 		if (ret < 0)
617 			return ret;
618 	}
619 
620 	val = QSPI_DMA_BLK_SET(tqspi->curr_dma_words - 1);
621 	tegra_qspi_writel(tqspi, val, QSPI_DMA_BLK);
622 
623 	tegra_qspi_unmask_irq(tqspi);
624 
625 	if (tqspi->is_packed)
626 		len = DIV_ROUND_UP(tqspi->curr_dma_words * tqspi->bytes_per_word, 4) * 4;
627 	else
628 		len = tqspi->curr_dma_words * 4;
629 
630 	/* set attention level based on length of transfer */
631 	val = 0;
632 	if (len & 0xf) {
633 		val |= QSPI_TX_TRIG_1 | QSPI_RX_TRIG_1;
634 		dma_burst = 1;
635 	} else if (((len) >> 4) & 0x1) {
636 		val |= QSPI_TX_TRIG_4 | QSPI_RX_TRIG_4;
637 		dma_burst = 4;
638 	} else {
639 		val |= QSPI_TX_TRIG_8 | QSPI_RX_TRIG_8;
640 		dma_burst = 8;
641 	}
642 
643 	tegra_qspi_writel(tqspi, val, QSPI_DMA_CTL);
644 	tqspi->dma_control_reg = val;
645 
646 	dma_sconfig.device_fc = true;
647 	if (tqspi->cur_direction & DATA_DIR_TX) {
648 		dma_sconfig.dst_addr = tqspi->phys + QSPI_TX_FIFO;
649 		dma_sconfig.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
650 		dma_sconfig.dst_maxburst = dma_burst;
651 		ret = dmaengine_slave_config(tqspi->tx_dma_chan, &dma_sconfig);
652 		if (ret < 0) {
653 			dev_err(tqspi->dev, "failed DMA slave config: %d\n", ret);
654 			return ret;
655 		}
656 
657 		tegra_qspi_copy_client_txbuf_to_qspi_txbuf(tqspi, t);
658 		ret = tegra_qspi_start_tx_dma(tqspi, t, len);
659 		if (ret < 0) {
660 			dev_err(tqspi->dev, "failed to starting TX DMA: %d\n", ret);
661 			return ret;
662 		}
663 	}
664 
665 	if (tqspi->cur_direction & DATA_DIR_RX) {
666 		dma_sconfig.src_addr = tqspi->phys + QSPI_RX_FIFO;
667 		dma_sconfig.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
668 		dma_sconfig.src_maxburst = dma_burst;
669 		ret = dmaengine_slave_config(tqspi->rx_dma_chan, &dma_sconfig);
670 		if (ret < 0) {
671 			dev_err(tqspi->dev, "failed DMA slave config: %d\n", ret);
672 			return ret;
673 		}
674 
675 		dma_sync_single_for_device(tqspi->dev, tqspi->rx_dma_phys,
676 					   tqspi->dma_buf_size,
677 					   DMA_FROM_DEVICE);
678 
679 		ret = tegra_qspi_start_rx_dma(tqspi, t, len);
680 		if (ret < 0) {
681 			dev_err(tqspi->dev, "failed to start RX DMA: %d\n", ret);
682 			if (tqspi->cur_direction & DATA_DIR_TX)
683 				dmaengine_terminate_all(tqspi->tx_dma_chan);
684 			return ret;
685 		}
686 	}
687 
688 	tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
689 
690 	tqspi->is_curr_dma_xfer = true;
691 	tqspi->dma_control_reg = val;
692 	val |= QSPI_DMA_EN;
693 	tegra_qspi_writel(tqspi, val, QSPI_DMA_CTL);
694 
695 	return ret;
696 }
697 
698 static int tegra_qspi_start_cpu_based_transfer(struct tegra_qspi *qspi, struct spi_transfer *t)
699 {
700 	u32 val;
701 	unsigned int cur_words;
702 
703 	if (qspi->cur_direction & DATA_DIR_TX)
704 		cur_words = tegra_qspi_fill_tx_fifo_from_client_txbuf(qspi, t);
705 	else
706 		cur_words = qspi->curr_dma_words;
707 
708 	val = QSPI_DMA_BLK_SET(cur_words - 1);
709 	tegra_qspi_writel(qspi, val, QSPI_DMA_BLK);
710 
711 	tegra_qspi_unmask_irq(qspi);
712 
713 	qspi->is_curr_dma_xfer = false;
714 	val = qspi->command1_reg;
715 	val |= QSPI_PIO;
716 	tegra_qspi_writel(qspi, val, QSPI_COMMAND1);
717 
718 	return 0;
719 }
720 
721 static void tegra_qspi_deinit_dma(struct tegra_qspi *tqspi)
722 {
723 	if (!tqspi->soc_data->has_dma)
724 		return;
725 
726 	if (tqspi->tx_dma_buf) {
727 		dma_free_coherent(tqspi->dev, tqspi->dma_buf_size,
728 				  tqspi->tx_dma_buf, tqspi->tx_dma_phys);
729 		tqspi->tx_dma_buf = NULL;
730 	}
731 
732 	if (tqspi->tx_dma_chan) {
733 		dma_release_channel(tqspi->tx_dma_chan);
734 		tqspi->tx_dma_chan = NULL;
735 	}
736 
737 	if (tqspi->rx_dma_buf) {
738 		dma_free_coherent(tqspi->dev, tqspi->dma_buf_size,
739 				  tqspi->rx_dma_buf, tqspi->rx_dma_phys);
740 		tqspi->rx_dma_buf = NULL;
741 	}
742 
743 	if (tqspi->rx_dma_chan) {
744 		dma_release_channel(tqspi->rx_dma_chan);
745 		tqspi->rx_dma_chan = NULL;
746 	}
747 }
748 
749 static int tegra_qspi_init_dma(struct tegra_qspi *tqspi)
750 {
751 	struct dma_chan *dma_chan;
752 	dma_addr_t dma_phys;
753 	u32 *dma_buf;
754 	int err;
755 
756 	if (!tqspi->soc_data->has_dma)
757 		return 0;
758 
759 	dma_chan = dma_request_chan(tqspi->dev, "rx");
760 	if (IS_ERR(dma_chan)) {
761 		err = PTR_ERR(dma_chan);
762 		goto err_out;
763 	}
764 
765 	tqspi->rx_dma_chan = dma_chan;
766 
767 	dma_buf = dma_alloc_coherent(tqspi->dev, tqspi->dma_buf_size, &dma_phys, GFP_KERNEL);
768 	if (!dma_buf) {
769 		err = -ENOMEM;
770 		goto err_out;
771 	}
772 
773 	tqspi->rx_dma_buf = dma_buf;
774 	tqspi->rx_dma_phys = dma_phys;
775 
776 	dma_chan = dma_request_chan(tqspi->dev, "tx");
777 	if (IS_ERR(dma_chan)) {
778 		err = PTR_ERR(dma_chan);
779 		goto err_out;
780 	}
781 
782 	tqspi->tx_dma_chan = dma_chan;
783 
784 	dma_buf = dma_alloc_coherent(tqspi->dev, tqspi->dma_buf_size, &dma_phys, GFP_KERNEL);
785 	if (!dma_buf) {
786 		err = -ENOMEM;
787 		goto err_out;
788 	}
789 
790 	tqspi->tx_dma_buf = dma_buf;
791 	tqspi->tx_dma_phys = dma_phys;
792 	tqspi->use_dma = true;
793 
794 	return 0;
795 
796 err_out:
797 	tegra_qspi_deinit_dma(tqspi);
798 
799 	if (err != -EPROBE_DEFER) {
800 		dev_err(tqspi->dev, "cannot use DMA: %d\n", err);
801 		dev_err(tqspi->dev, "falling back to PIO\n");
802 		return 0;
803 	}
804 
805 	return err;
806 }
807 
808 static u32 tegra_qspi_setup_transfer_one(struct spi_device *spi, struct spi_transfer *t,
809 					 bool is_first_of_msg)
810 {
811 	struct tegra_qspi *tqspi = spi_master_get_devdata(spi->master);
812 	struct tegra_qspi_client_data *cdata = spi->controller_data;
813 	u32 command1, command2, speed = t->speed_hz;
814 	u8 bits_per_word = t->bits_per_word;
815 	u32 tx_tap = 0, rx_tap = 0;
816 	int req_mode;
817 
818 	if (!has_acpi_companion(tqspi->dev) && speed != tqspi->cur_speed) {
819 		clk_set_rate(tqspi->clk, speed);
820 		tqspi->cur_speed = speed;
821 	}
822 
823 	tqspi->cur_pos = 0;
824 	tqspi->cur_rx_pos = 0;
825 	tqspi->cur_tx_pos = 0;
826 	tqspi->curr_xfer = t;
827 
828 	if (is_first_of_msg) {
829 		tegra_qspi_mask_clear_irq(tqspi);
830 
831 		command1 = tqspi->def_command1_reg;
832 		command1 |= QSPI_CS_SEL(spi->chip_select);
833 		command1 |= QSPI_BIT_LENGTH(bits_per_word - 1);
834 
835 		command1 &= ~QSPI_CONTROL_MODE_MASK;
836 		req_mode = spi->mode & 0x3;
837 		if (req_mode == SPI_MODE_3)
838 			command1 |= QSPI_CONTROL_MODE_3;
839 		else
840 			command1 |= QSPI_CONTROL_MODE_0;
841 
842 		if (spi->mode & SPI_CS_HIGH)
843 			command1 |= QSPI_CS_SW_VAL;
844 		else
845 			command1 &= ~QSPI_CS_SW_VAL;
846 		tegra_qspi_writel(tqspi, command1, QSPI_COMMAND1);
847 
848 		if (cdata && cdata->tx_clk_tap_delay)
849 			tx_tap = cdata->tx_clk_tap_delay;
850 
851 		if (cdata && cdata->rx_clk_tap_delay)
852 			rx_tap = cdata->rx_clk_tap_delay;
853 
854 		command2 = QSPI_TX_TAP_DELAY(tx_tap) | QSPI_RX_TAP_DELAY(rx_tap);
855 		if (command2 != tqspi->def_command2_reg)
856 			tegra_qspi_writel(tqspi, command2, QSPI_COMMAND2);
857 
858 	} else {
859 		command1 = tqspi->command1_reg;
860 		command1 &= ~QSPI_BIT_LENGTH(~0);
861 		command1 |= QSPI_BIT_LENGTH(bits_per_word - 1);
862 	}
863 
864 	command1 &= ~QSPI_SDR_DDR_SEL;
865 
866 	return command1;
867 }
868 
869 static int tegra_qspi_start_transfer_one(struct spi_device *spi,
870 					 struct spi_transfer *t, u32 command1)
871 {
872 	struct tegra_qspi *tqspi = spi_master_get_devdata(spi->master);
873 	unsigned int total_fifo_words;
874 	u8 bus_width = 0;
875 	int ret;
876 
877 	total_fifo_words = tegra_qspi_calculate_curr_xfer_param(tqspi, t);
878 
879 	command1 &= ~QSPI_PACKED;
880 	if (tqspi->is_packed)
881 		command1 |= QSPI_PACKED;
882 	tegra_qspi_writel(tqspi, command1, QSPI_COMMAND1);
883 
884 	tqspi->cur_direction = 0;
885 
886 	command1 &= ~(QSPI_TX_EN | QSPI_RX_EN);
887 	if (t->rx_buf) {
888 		command1 |= QSPI_RX_EN;
889 		tqspi->cur_direction |= DATA_DIR_RX;
890 		bus_width = t->rx_nbits;
891 	}
892 
893 	if (t->tx_buf) {
894 		command1 |= QSPI_TX_EN;
895 		tqspi->cur_direction |= DATA_DIR_TX;
896 		bus_width = t->tx_nbits;
897 	}
898 
899 	command1 &= ~QSPI_INTERFACE_WIDTH_MASK;
900 
901 	if (bus_width == SPI_NBITS_QUAD)
902 		command1 |= QSPI_INTERFACE_WIDTH_QUAD;
903 	else if (bus_width == SPI_NBITS_DUAL)
904 		command1 |= QSPI_INTERFACE_WIDTH_DUAL;
905 	else
906 		command1 |= QSPI_INTERFACE_WIDTH_SINGLE;
907 
908 	tqspi->command1_reg = command1;
909 
910 	tegra_qspi_writel(tqspi, QSPI_NUM_DUMMY_CYCLE(tqspi->dummy_cycles), QSPI_MISC_REG);
911 
912 	ret = tegra_qspi_flush_fifos(tqspi, false);
913 	if (ret < 0)
914 		return ret;
915 
916 	if (tqspi->use_dma && total_fifo_words > QSPI_FIFO_DEPTH)
917 		ret = tegra_qspi_start_dma_based_transfer(tqspi, t);
918 	else
919 		ret = tegra_qspi_start_cpu_based_transfer(tqspi, t);
920 
921 	return ret;
922 }
923 
924 static struct tegra_qspi_client_data *tegra_qspi_parse_cdata_dt(struct spi_device *spi)
925 {
926 	struct tegra_qspi_client_data *cdata;
927 	struct tegra_qspi *tqspi = spi_master_get_devdata(spi->master);
928 
929 	cdata = devm_kzalloc(tqspi->dev, sizeof(*cdata), GFP_KERNEL);
930 	if (!cdata)
931 		return NULL;
932 
933 	device_property_read_u32(&spi->dev, "nvidia,tx-clk-tap-delay",
934 				 &cdata->tx_clk_tap_delay);
935 	device_property_read_u32(&spi->dev, "nvidia,rx-clk-tap-delay",
936 				 &cdata->rx_clk_tap_delay);
937 
938 	return cdata;
939 }
940 
941 static int tegra_qspi_setup(struct spi_device *spi)
942 {
943 	struct tegra_qspi *tqspi = spi_master_get_devdata(spi->master);
944 	struct tegra_qspi_client_data *cdata = spi->controller_data;
945 	unsigned long flags;
946 	u32 val;
947 	int ret;
948 
949 	ret = pm_runtime_resume_and_get(tqspi->dev);
950 	if (ret < 0) {
951 		dev_err(tqspi->dev, "failed to get runtime PM: %d\n", ret);
952 		return ret;
953 	}
954 
955 	if (!cdata) {
956 		cdata = tegra_qspi_parse_cdata_dt(spi);
957 		spi->controller_data = cdata;
958 	}
959 	spin_lock_irqsave(&tqspi->lock, flags);
960 
961 	/* keep default cs state to inactive */
962 	val = tqspi->def_command1_reg;
963 	val |= QSPI_CS_SEL(spi->chip_select);
964 	if (spi->mode & SPI_CS_HIGH)
965 		val &= ~QSPI_CS_POL_INACTIVE(spi->chip_select);
966 	else
967 		val |= QSPI_CS_POL_INACTIVE(spi->chip_select);
968 
969 	tqspi->def_command1_reg = val;
970 	tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
971 
972 	spin_unlock_irqrestore(&tqspi->lock, flags);
973 
974 	pm_runtime_put(tqspi->dev);
975 
976 	return 0;
977 }
978 
979 static void tegra_qspi_dump_regs(struct tegra_qspi *tqspi)
980 {
981 	dev_dbg(tqspi->dev, "============ QSPI REGISTER DUMP ============\n");
982 	dev_dbg(tqspi->dev, "Command1:    0x%08x | Command2:    0x%08x\n",
983 		tegra_qspi_readl(tqspi, QSPI_COMMAND1),
984 		tegra_qspi_readl(tqspi, QSPI_COMMAND2));
985 	dev_dbg(tqspi->dev, "DMA_CTL:     0x%08x | DMA_BLK:     0x%08x\n",
986 		tegra_qspi_readl(tqspi, QSPI_DMA_CTL),
987 		tegra_qspi_readl(tqspi, QSPI_DMA_BLK));
988 	dev_dbg(tqspi->dev, "INTR_MASK:  0x%08x | MISC: 0x%08x\n",
989 		tegra_qspi_readl(tqspi, QSPI_INTR_MASK),
990 		tegra_qspi_readl(tqspi, QSPI_MISC_REG));
991 	dev_dbg(tqspi->dev, "TRANS_STAT:  0x%08x | FIFO_STATUS: 0x%08x\n",
992 		tegra_qspi_readl(tqspi, QSPI_TRANS_STATUS),
993 		tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS));
994 }
995 
996 static void tegra_qspi_handle_error(struct tegra_qspi *tqspi)
997 {
998 	dev_err(tqspi->dev, "error in transfer, fifo status 0x%08x\n", tqspi->status_reg);
999 	tegra_qspi_dump_regs(tqspi);
1000 	tegra_qspi_flush_fifos(tqspi, true);
1001 	if (device_reset(tqspi->dev) < 0)
1002 		dev_warn_once(tqspi->dev, "device reset failed\n");
1003 }
1004 
1005 static void tegra_qspi_transfer_end(struct spi_device *spi)
1006 {
1007 	struct tegra_qspi *tqspi = spi_master_get_devdata(spi->master);
1008 	int cs_val = (spi->mode & SPI_CS_HIGH) ? 0 : 1;
1009 
1010 	if (cs_val)
1011 		tqspi->command1_reg |= QSPI_CS_SW_VAL;
1012 	else
1013 		tqspi->command1_reg &= ~QSPI_CS_SW_VAL;
1014 	tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
1015 	tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
1016 }
1017 
1018 static u32 tegra_qspi_cmd_config(bool is_ddr, u8 bus_width, u8 len)
1019 {
1020 	u32 cmd_config = 0;
1021 
1022 	/* Extract Command configuration and value */
1023 	if (is_ddr)
1024 		cmd_config |= QSPI_COMMAND_SDR_DDR;
1025 	else
1026 		cmd_config &= ~QSPI_COMMAND_SDR_DDR;
1027 
1028 	cmd_config |= QSPI_COMMAND_X1_X2_X4(bus_width);
1029 	cmd_config |= QSPI_COMMAND_SIZE_SET((len * 8) - 1);
1030 
1031 	return cmd_config;
1032 }
1033 
1034 static u32 tegra_qspi_addr_config(bool is_ddr, u8 bus_width, u8 len)
1035 {
1036 	u32 addr_config = 0;
1037 
1038 	/* Extract Address configuration and value */
1039 	is_ddr = 0; //Only SDR mode supported
1040 	bus_width = 0; //X1 mode
1041 
1042 	if (is_ddr)
1043 		addr_config |= QSPI_ADDRESS_SDR_DDR;
1044 	else
1045 		addr_config &= ~QSPI_ADDRESS_SDR_DDR;
1046 
1047 	addr_config |= QSPI_ADDRESS_X1_X2_X4(bus_width);
1048 	addr_config |= QSPI_ADDRESS_SIZE_SET((len * 8) - 1);
1049 
1050 	return addr_config;
1051 }
1052 
1053 static int tegra_qspi_combined_seq_xfer(struct tegra_qspi *tqspi,
1054 					struct spi_message *msg)
1055 {
1056 	bool is_first_msg = true;
1057 	struct spi_transfer *xfer;
1058 	struct spi_device *spi = msg->spi;
1059 	u8 transfer_phase = 0;
1060 	u32 cmd1 = 0, dma_ctl = 0;
1061 	int ret = 0;
1062 	u32 address_value = 0;
1063 	u32 cmd_config = 0, addr_config = 0;
1064 	u8 cmd_value = 0, val = 0;
1065 
1066 	/* Enable Combined sequence mode */
1067 	val = tegra_qspi_readl(tqspi, QSPI_GLOBAL_CONFIG);
1068 	val |= QSPI_CMB_SEQ_EN;
1069 	tegra_qspi_writel(tqspi, val, QSPI_GLOBAL_CONFIG);
1070 	/* Process individual transfer list */
1071 	list_for_each_entry(xfer, &msg->transfers, transfer_list) {
1072 		switch (transfer_phase) {
1073 		case CMD_TRANSFER:
1074 			/* X1 SDR mode */
1075 			cmd_config = tegra_qspi_cmd_config(false, 0,
1076 							   xfer->len);
1077 			cmd_value = *((const u8 *)(xfer->tx_buf));
1078 			break;
1079 		case ADDR_TRANSFER:
1080 			/* X1 SDR mode */
1081 			addr_config = tegra_qspi_addr_config(false, 0,
1082 							     xfer->len);
1083 			address_value = *((const u32 *)(xfer->tx_buf));
1084 			break;
1085 		case DATA_TRANSFER:
1086 			/* Program Command, Address value in register */
1087 			tegra_qspi_writel(tqspi, cmd_value, QSPI_CMB_SEQ_CMD);
1088 			tegra_qspi_writel(tqspi, address_value,
1089 					  QSPI_CMB_SEQ_ADDR);
1090 			/* Program Command and Address config in register */
1091 			tegra_qspi_writel(tqspi, cmd_config,
1092 					  QSPI_CMB_SEQ_CMD_CFG);
1093 			tegra_qspi_writel(tqspi, addr_config,
1094 					  QSPI_CMB_SEQ_ADDR_CFG);
1095 
1096 			reinit_completion(&tqspi->xfer_completion);
1097 			cmd1 = tegra_qspi_setup_transfer_one(spi, xfer,
1098 							     is_first_msg);
1099 			ret = tegra_qspi_start_transfer_one(spi, xfer,
1100 							    cmd1);
1101 
1102 			if (ret < 0) {
1103 				dev_err(tqspi->dev, "Failed to start transfer-one: %d\n",
1104 					ret);
1105 				return ret;
1106 			}
1107 
1108 			is_first_msg = false;
1109 			ret = wait_for_completion_timeout
1110 					(&tqspi->xfer_completion,
1111 					QSPI_DMA_TIMEOUT);
1112 
1113 			if (WARN_ON(ret == 0)) {
1114 				dev_err(tqspi->dev, "QSPI Transfer failed with timeout: %d\n",
1115 					ret);
1116 				if (tqspi->is_curr_dma_xfer &&
1117 				    (tqspi->cur_direction & DATA_DIR_TX))
1118 					dmaengine_terminate_all
1119 						(tqspi->tx_dma_chan);
1120 
1121 				if (tqspi->is_curr_dma_xfer &&
1122 				    (tqspi->cur_direction & DATA_DIR_RX))
1123 					dmaengine_terminate_all
1124 						(tqspi->rx_dma_chan);
1125 
1126 				/* Abort transfer by resetting pio/dma bit */
1127 				if (!tqspi->is_curr_dma_xfer) {
1128 					cmd1 = tegra_qspi_readl
1129 							(tqspi,
1130 							 QSPI_COMMAND1);
1131 					cmd1 &= ~QSPI_PIO;
1132 					tegra_qspi_writel
1133 							(tqspi, cmd1,
1134 							 QSPI_COMMAND1);
1135 				} else {
1136 					dma_ctl = tegra_qspi_readl
1137 							(tqspi,
1138 							 QSPI_DMA_CTL);
1139 					dma_ctl &= ~QSPI_DMA_EN;
1140 					tegra_qspi_writel(tqspi, dma_ctl,
1141 							  QSPI_DMA_CTL);
1142 				}
1143 
1144 				/* Reset controller if timeout happens */
1145 				if (device_reset(tqspi->dev) < 0)
1146 					dev_warn_once(tqspi->dev,
1147 						      "device reset failed\n");
1148 				ret = -EIO;
1149 				goto exit;
1150 			}
1151 
1152 			if (tqspi->tx_status ||  tqspi->rx_status) {
1153 				dev_err(tqspi->dev, "QSPI Transfer failed\n");
1154 				tqspi->tx_status = 0;
1155 				tqspi->rx_status = 0;
1156 				ret = -EIO;
1157 				goto exit;
1158 			}
1159 			if (!xfer->cs_change) {
1160 				tegra_qspi_transfer_end(spi);
1161 				spi_transfer_delay_exec(xfer);
1162 			}
1163 			break;
1164 		default:
1165 			ret = -EINVAL;
1166 			goto exit;
1167 		}
1168 		msg->actual_length += xfer->len;
1169 		transfer_phase++;
1170 	}
1171 	ret = 0;
1172 
1173 exit:
1174 	msg->status = ret;
1175 	if (ret < 0) {
1176 		tegra_qspi_transfer_end(spi);
1177 		spi_transfer_delay_exec(xfer);
1178 	}
1179 
1180 	return ret;
1181 }
1182 
1183 static int tegra_qspi_non_combined_seq_xfer(struct tegra_qspi *tqspi,
1184 					    struct spi_message *msg)
1185 {
1186 	struct spi_device *spi = msg->spi;
1187 	struct spi_transfer *transfer;
1188 	bool is_first_msg = true;
1189 	int ret = 0, val = 0;
1190 
1191 	msg->status = 0;
1192 	msg->actual_length = 0;
1193 	tqspi->tx_status = 0;
1194 	tqspi->rx_status = 0;
1195 
1196 	/* Disable Combined sequence mode */
1197 	val = tegra_qspi_readl(tqspi, QSPI_GLOBAL_CONFIG);
1198 	val &= ~QSPI_CMB_SEQ_EN;
1199 	tegra_qspi_writel(tqspi, val, QSPI_GLOBAL_CONFIG);
1200 	list_for_each_entry(transfer, &msg->transfers, transfer_list) {
1201 		struct spi_transfer *xfer = transfer;
1202 		u8 dummy_bytes = 0;
1203 		u32 cmd1;
1204 
1205 		tqspi->dummy_cycles = 0;
1206 		/*
1207 		 * Tegra QSPI hardware supports dummy bytes transfer after actual transfer
1208 		 * bytes based on programmed dummy clock cycles in the QSPI_MISC register.
1209 		 * So, check if the next transfer is dummy data transfer and program dummy
1210 		 * clock cycles along with the current transfer and skip next transfer.
1211 		 */
1212 		if (!list_is_last(&xfer->transfer_list, &msg->transfers)) {
1213 			struct spi_transfer *next_xfer;
1214 
1215 			next_xfer = list_next_entry(xfer, transfer_list);
1216 			if (next_xfer->dummy_data) {
1217 				u32 dummy_cycles = next_xfer->len * 8 / next_xfer->tx_nbits;
1218 
1219 				if (dummy_cycles <= QSPI_DUMMY_CYCLES_MAX) {
1220 					tqspi->dummy_cycles = dummy_cycles;
1221 					dummy_bytes = next_xfer->len;
1222 					transfer = next_xfer;
1223 				}
1224 			}
1225 		}
1226 
1227 		reinit_completion(&tqspi->xfer_completion);
1228 
1229 		cmd1 = tegra_qspi_setup_transfer_one(spi, xfer, is_first_msg);
1230 
1231 		ret = tegra_qspi_start_transfer_one(spi, xfer, cmd1);
1232 		if (ret < 0) {
1233 			dev_err(tqspi->dev, "failed to start transfer: %d\n", ret);
1234 			goto complete_xfer;
1235 		}
1236 
1237 		ret = wait_for_completion_timeout(&tqspi->xfer_completion,
1238 						  QSPI_DMA_TIMEOUT);
1239 		if (WARN_ON(ret == 0)) {
1240 			dev_err(tqspi->dev, "transfer timeout\n");
1241 			if (tqspi->is_curr_dma_xfer && (tqspi->cur_direction & DATA_DIR_TX))
1242 				dmaengine_terminate_all(tqspi->tx_dma_chan);
1243 			if (tqspi->is_curr_dma_xfer && (tqspi->cur_direction & DATA_DIR_RX))
1244 				dmaengine_terminate_all(tqspi->rx_dma_chan);
1245 			tegra_qspi_handle_error(tqspi);
1246 			ret = -EIO;
1247 			goto complete_xfer;
1248 		}
1249 
1250 		if (tqspi->tx_status ||  tqspi->rx_status) {
1251 			tegra_qspi_handle_error(tqspi);
1252 			ret = -EIO;
1253 			goto complete_xfer;
1254 		}
1255 
1256 		msg->actual_length += xfer->len + dummy_bytes;
1257 
1258 complete_xfer:
1259 		if (ret < 0) {
1260 			tegra_qspi_transfer_end(spi);
1261 			spi_transfer_delay_exec(xfer);
1262 			goto exit;
1263 		}
1264 
1265 		if (list_is_last(&xfer->transfer_list, &msg->transfers)) {
1266 			/* de-activate CS after last transfer only when cs_change is not set */
1267 			if (!xfer->cs_change) {
1268 				tegra_qspi_transfer_end(spi);
1269 				spi_transfer_delay_exec(xfer);
1270 			}
1271 		} else if (xfer->cs_change) {
1272 			 /* de-activated CS between the transfers only when cs_change is set */
1273 			tegra_qspi_transfer_end(spi);
1274 			spi_transfer_delay_exec(xfer);
1275 		}
1276 	}
1277 
1278 	ret = 0;
1279 exit:
1280 	msg->status = ret;
1281 
1282 	return ret;
1283 }
1284 
1285 static bool tegra_qspi_validate_cmb_seq(struct tegra_qspi *tqspi,
1286 					struct spi_message *msg)
1287 {
1288 	int transfer_count = 0;
1289 	struct spi_transfer *xfer;
1290 
1291 	list_for_each_entry(xfer, &msg->transfers, transfer_list) {
1292 		transfer_count++;
1293 	}
1294 	if (!tqspi->soc_data->cmb_xfer_capable || transfer_count != 3)
1295 		return false;
1296 	xfer = list_first_entry(&msg->transfers, typeof(*xfer),
1297 				transfer_list);
1298 	if (xfer->len > 2)
1299 		return false;
1300 	xfer = list_next_entry(xfer, transfer_list);
1301 	if (xfer->len > 4 || xfer->len < 3)
1302 		return false;
1303 	xfer = list_next_entry(xfer, transfer_list);
1304 	if (!tqspi->soc_data->has_dma && xfer->len > (QSPI_FIFO_DEPTH << 2))
1305 		return false;
1306 
1307 	return true;
1308 }
1309 
1310 static int tegra_qspi_transfer_one_message(struct spi_master *master,
1311 					   struct spi_message *msg)
1312 {
1313 	struct tegra_qspi *tqspi = spi_master_get_devdata(master);
1314 	int ret;
1315 
1316 	if (tegra_qspi_validate_cmb_seq(tqspi, msg))
1317 		ret = tegra_qspi_combined_seq_xfer(tqspi, msg);
1318 	else
1319 		ret = tegra_qspi_non_combined_seq_xfer(tqspi, msg);
1320 
1321 	spi_finalize_current_message(master);
1322 
1323 	return ret;
1324 }
1325 
1326 static irqreturn_t handle_cpu_based_xfer(struct tegra_qspi *tqspi)
1327 {
1328 	struct spi_transfer *t = tqspi->curr_xfer;
1329 	unsigned long flags;
1330 
1331 	spin_lock_irqsave(&tqspi->lock, flags);
1332 
1333 	if (tqspi->tx_status ||  tqspi->rx_status) {
1334 		tegra_qspi_handle_error(tqspi);
1335 		complete(&tqspi->xfer_completion);
1336 		goto exit;
1337 	}
1338 
1339 	if (tqspi->cur_direction & DATA_DIR_RX)
1340 		tegra_qspi_read_rx_fifo_to_client_rxbuf(tqspi, t);
1341 
1342 	if (tqspi->cur_direction & DATA_DIR_TX)
1343 		tqspi->cur_pos = tqspi->cur_tx_pos;
1344 	else
1345 		tqspi->cur_pos = tqspi->cur_rx_pos;
1346 
1347 	if (tqspi->cur_pos == t->len) {
1348 		complete(&tqspi->xfer_completion);
1349 		goto exit;
1350 	}
1351 
1352 	tegra_qspi_calculate_curr_xfer_param(tqspi, t);
1353 	tegra_qspi_start_cpu_based_transfer(tqspi, t);
1354 exit:
1355 	spin_unlock_irqrestore(&tqspi->lock, flags);
1356 	return IRQ_HANDLED;
1357 }
1358 
1359 static irqreturn_t handle_dma_based_xfer(struct tegra_qspi *tqspi)
1360 {
1361 	struct spi_transfer *t = tqspi->curr_xfer;
1362 	unsigned int total_fifo_words;
1363 	unsigned long flags;
1364 	long wait_status;
1365 	int err = 0;
1366 
1367 	if (tqspi->cur_direction & DATA_DIR_TX) {
1368 		if (tqspi->tx_status) {
1369 			dmaengine_terminate_all(tqspi->tx_dma_chan);
1370 			err += 1;
1371 		} else {
1372 			wait_status = wait_for_completion_interruptible_timeout(
1373 				&tqspi->tx_dma_complete, QSPI_DMA_TIMEOUT);
1374 			if (wait_status <= 0) {
1375 				dmaengine_terminate_all(tqspi->tx_dma_chan);
1376 				dev_err(tqspi->dev, "failed TX DMA transfer\n");
1377 				err += 1;
1378 			}
1379 		}
1380 	}
1381 
1382 	if (tqspi->cur_direction & DATA_DIR_RX) {
1383 		if (tqspi->rx_status) {
1384 			dmaengine_terminate_all(tqspi->rx_dma_chan);
1385 			err += 2;
1386 		} else {
1387 			wait_status = wait_for_completion_interruptible_timeout(
1388 				&tqspi->rx_dma_complete, QSPI_DMA_TIMEOUT);
1389 			if (wait_status <= 0) {
1390 				dmaengine_terminate_all(tqspi->rx_dma_chan);
1391 				dev_err(tqspi->dev, "failed RX DMA transfer\n");
1392 				err += 2;
1393 			}
1394 		}
1395 	}
1396 
1397 	spin_lock_irqsave(&tqspi->lock, flags);
1398 
1399 	if (err) {
1400 		tegra_qspi_dma_unmap_xfer(tqspi, t);
1401 		tegra_qspi_handle_error(tqspi);
1402 		complete(&tqspi->xfer_completion);
1403 		goto exit;
1404 	}
1405 
1406 	if (tqspi->cur_direction & DATA_DIR_RX)
1407 		tegra_qspi_copy_qspi_rxbuf_to_client_rxbuf(tqspi, t);
1408 
1409 	if (tqspi->cur_direction & DATA_DIR_TX)
1410 		tqspi->cur_pos = tqspi->cur_tx_pos;
1411 	else
1412 		tqspi->cur_pos = tqspi->cur_rx_pos;
1413 
1414 	if (tqspi->cur_pos == t->len) {
1415 		tegra_qspi_dma_unmap_xfer(tqspi, t);
1416 		complete(&tqspi->xfer_completion);
1417 		goto exit;
1418 	}
1419 
1420 	tegra_qspi_dma_unmap_xfer(tqspi, t);
1421 
1422 	/* continue transfer in current message */
1423 	total_fifo_words = tegra_qspi_calculate_curr_xfer_param(tqspi, t);
1424 	if (total_fifo_words > QSPI_FIFO_DEPTH)
1425 		err = tegra_qspi_start_dma_based_transfer(tqspi, t);
1426 	else
1427 		err = tegra_qspi_start_cpu_based_transfer(tqspi, t);
1428 
1429 exit:
1430 	spin_unlock_irqrestore(&tqspi->lock, flags);
1431 	return IRQ_HANDLED;
1432 }
1433 
1434 static irqreturn_t tegra_qspi_isr_thread(int irq, void *context_data)
1435 {
1436 	struct tegra_qspi *tqspi = context_data;
1437 
1438 	tqspi->status_reg = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
1439 
1440 	if (tqspi->cur_direction & DATA_DIR_TX)
1441 		tqspi->tx_status = tqspi->status_reg & (QSPI_TX_FIFO_UNF | QSPI_TX_FIFO_OVF);
1442 
1443 	if (tqspi->cur_direction & DATA_DIR_RX)
1444 		tqspi->rx_status = tqspi->status_reg & (QSPI_RX_FIFO_OVF | QSPI_RX_FIFO_UNF);
1445 
1446 	tegra_qspi_mask_clear_irq(tqspi);
1447 
1448 	if (!tqspi->is_curr_dma_xfer)
1449 		return handle_cpu_based_xfer(tqspi);
1450 
1451 	return handle_dma_based_xfer(tqspi);
1452 }
1453 
1454 static struct tegra_qspi_soc_data tegra210_qspi_soc_data = {
1455 	.has_dma = true,
1456 	.cmb_xfer_capable = false,
1457 	.cs_count = 1,
1458 };
1459 
1460 static struct tegra_qspi_soc_data tegra186_qspi_soc_data = {
1461 	.has_dma = true,
1462 	.cmb_xfer_capable = true,
1463 	.cs_count = 1,
1464 };
1465 
1466 static struct tegra_qspi_soc_data tegra234_qspi_soc_data = {
1467 	.has_dma = false,
1468 	.cmb_xfer_capable = true,
1469 	.cs_count = 1,
1470 };
1471 
1472 static struct tegra_qspi_soc_data tegra241_qspi_soc_data = {
1473 	.has_dma = false,
1474 	.cmb_xfer_capable = true,
1475 	.cs_count = 4,
1476 };
1477 
1478 static const struct of_device_id tegra_qspi_of_match[] = {
1479 	{
1480 		.compatible = "nvidia,tegra210-qspi",
1481 		.data	    = &tegra210_qspi_soc_data,
1482 	}, {
1483 		.compatible = "nvidia,tegra186-qspi",
1484 		.data	    = &tegra186_qspi_soc_data,
1485 	}, {
1486 		.compatible = "nvidia,tegra194-qspi",
1487 		.data	    = &tegra186_qspi_soc_data,
1488 	}, {
1489 		.compatible = "nvidia,tegra234-qspi",
1490 		.data	    = &tegra234_qspi_soc_data,
1491 	}, {
1492 		.compatible = "nvidia,tegra241-qspi",
1493 		.data	    = &tegra241_qspi_soc_data,
1494 	},
1495 	{}
1496 };
1497 
1498 MODULE_DEVICE_TABLE(of, tegra_qspi_of_match);
1499 
1500 #ifdef CONFIG_ACPI
1501 static const struct acpi_device_id tegra_qspi_acpi_match[] = {
1502 	{
1503 		.id = "NVDA1213",
1504 		.driver_data = (kernel_ulong_t)&tegra210_qspi_soc_data,
1505 	}, {
1506 		.id = "NVDA1313",
1507 		.driver_data = (kernel_ulong_t)&tegra186_qspi_soc_data,
1508 	}, {
1509 		.id = "NVDA1413",
1510 		.driver_data = (kernel_ulong_t)&tegra234_qspi_soc_data,
1511 	}, {
1512 		.id = "NVDA1513",
1513 		.driver_data = (kernel_ulong_t)&tegra241_qspi_soc_data,
1514 	},
1515 	{}
1516 };
1517 
1518 MODULE_DEVICE_TABLE(acpi, tegra_qspi_acpi_match);
1519 #endif
1520 
1521 static int tegra_qspi_probe(struct platform_device *pdev)
1522 {
1523 	struct spi_master	*master;
1524 	struct tegra_qspi	*tqspi;
1525 	struct resource		*r;
1526 	int ret, qspi_irq;
1527 	int bus_num;
1528 
1529 	master = devm_spi_alloc_master(&pdev->dev, sizeof(*tqspi));
1530 	if (!master)
1531 		return -ENOMEM;
1532 
1533 	platform_set_drvdata(pdev, master);
1534 	tqspi = spi_master_get_devdata(master);
1535 
1536 	master->mode_bits = SPI_MODE_0 | SPI_MODE_3 | SPI_CS_HIGH |
1537 			    SPI_TX_DUAL | SPI_RX_DUAL | SPI_TX_QUAD | SPI_RX_QUAD;
1538 	master->bits_per_word_mask = SPI_BPW_MASK(32) | SPI_BPW_MASK(16) | SPI_BPW_MASK(8);
1539 	master->flags = SPI_CONTROLLER_HALF_DUPLEX;
1540 	master->setup = tegra_qspi_setup;
1541 	master->transfer_one_message = tegra_qspi_transfer_one_message;
1542 	master->num_chipselect = 1;
1543 	master->auto_runtime_pm = true;
1544 
1545 	bus_num = of_alias_get_id(pdev->dev.of_node, "spi");
1546 	if (bus_num >= 0)
1547 		master->bus_num = bus_num;
1548 
1549 	tqspi->master = master;
1550 	tqspi->dev = &pdev->dev;
1551 	spin_lock_init(&tqspi->lock);
1552 
1553 	tqspi->soc_data = device_get_match_data(&pdev->dev);
1554 	master->num_chipselect = tqspi->soc_data->cs_count;
1555 	r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
1556 	tqspi->base = devm_ioremap_resource(&pdev->dev, r);
1557 	if (IS_ERR(tqspi->base))
1558 		return PTR_ERR(tqspi->base);
1559 
1560 	tqspi->phys = r->start;
1561 	qspi_irq = platform_get_irq(pdev, 0);
1562 	if (qspi_irq < 0)
1563 		return qspi_irq;
1564 	tqspi->irq = qspi_irq;
1565 
1566 	if (!has_acpi_companion(tqspi->dev)) {
1567 		tqspi->clk = devm_clk_get(&pdev->dev, "qspi");
1568 		if (IS_ERR(tqspi->clk)) {
1569 			ret = PTR_ERR(tqspi->clk);
1570 			dev_err(&pdev->dev, "failed to get clock: %d\n", ret);
1571 			return ret;
1572 		}
1573 
1574 	}
1575 
1576 	tqspi->max_buf_size = QSPI_FIFO_DEPTH << 2;
1577 	tqspi->dma_buf_size = DEFAULT_QSPI_DMA_BUF_LEN;
1578 
1579 	ret = tegra_qspi_init_dma(tqspi);
1580 	if (ret < 0)
1581 		return ret;
1582 
1583 	if (tqspi->use_dma)
1584 		tqspi->max_buf_size = tqspi->dma_buf_size;
1585 
1586 	init_completion(&tqspi->tx_dma_complete);
1587 	init_completion(&tqspi->rx_dma_complete);
1588 	init_completion(&tqspi->xfer_completion);
1589 
1590 	pm_runtime_enable(&pdev->dev);
1591 	ret = pm_runtime_resume_and_get(&pdev->dev);
1592 	if (ret < 0) {
1593 		dev_err(&pdev->dev, "failed to get runtime PM: %d\n", ret);
1594 		goto exit_pm_disable;
1595 	}
1596 
1597 	if (device_reset(tqspi->dev) < 0)
1598 		dev_warn_once(tqspi->dev, "device reset failed\n");
1599 
1600 	tqspi->def_command1_reg = QSPI_M_S | QSPI_CS_SW_HW |  QSPI_CS_SW_VAL;
1601 	tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
1602 	tqspi->spi_cs_timing1 = tegra_qspi_readl(tqspi, QSPI_CS_TIMING1);
1603 	tqspi->spi_cs_timing2 = tegra_qspi_readl(tqspi, QSPI_CS_TIMING2);
1604 	tqspi->def_command2_reg = tegra_qspi_readl(tqspi, QSPI_COMMAND2);
1605 
1606 	pm_runtime_put(&pdev->dev);
1607 
1608 	ret = request_threaded_irq(tqspi->irq, NULL,
1609 				   tegra_qspi_isr_thread, IRQF_ONESHOT,
1610 				   dev_name(&pdev->dev), tqspi);
1611 	if (ret < 0) {
1612 		dev_err(&pdev->dev, "failed to request IRQ#%u: %d\n", tqspi->irq, ret);
1613 		goto exit_pm_disable;
1614 	}
1615 
1616 	master->dev.of_node = pdev->dev.of_node;
1617 	ret = spi_register_master(master);
1618 	if (ret < 0) {
1619 		dev_err(&pdev->dev, "failed to register master: %d\n", ret);
1620 		goto exit_free_irq;
1621 	}
1622 
1623 	return 0;
1624 
1625 exit_free_irq:
1626 	free_irq(qspi_irq, tqspi);
1627 exit_pm_disable:
1628 	pm_runtime_force_suspend(&pdev->dev);
1629 	tegra_qspi_deinit_dma(tqspi);
1630 	return ret;
1631 }
1632 
1633 static int tegra_qspi_remove(struct platform_device *pdev)
1634 {
1635 	struct spi_master *master = platform_get_drvdata(pdev);
1636 	struct tegra_qspi *tqspi = spi_master_get_devdata(master);
1637 
1638 	spi_unregister_master(master);
1639 	free_irq(tqspi->irq, tqspi);
1640 	pm_runtime_force_suspend(&pdev->dev);
1641 	tegra_qspi_deinit_dma(tqspi);
1642 
1643 	return 0;
1644 }
1645 
1646 static int __maybe_unused tegra_qspi_suspend(struct device *dev)
1647 {
1648 	struct spi_master *master = dev_get_drvdata(dev);
1649 
1650 	return spi_master_suspend(master);
1651 }
1652 
1653 static int __maybe_unused tegra_qspi_resume(struct device *dev)
1654 {
1655 	struct spi_master *master = dev_get_drvdata(dev);
1656 	struct tegra_qspi *tqspi = spi_master_get_devdata(master);
1657 	int ret;
1658 
1659 	ret = pm_runtime_resume_and_get(dev);
1660 	if (ret < 0) {
1661 		dev_err(dev, "failed to get runtime PM: %d\n", ret);
1662 		return ret;
1663 	}
1664 
1665 	tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
1666 	tegra_qspi_writel(tqspi, tqspi->def_command2_reg, QSPI_COMMAND2);
1667 	pm_runtime_put(dev);
1668 
1669 	return spi_master_resume(master);
1670 }
1671 
1672 static int __maybe_unused tegra_qspi_runtime_suspend(struct device *dev)
1673 {
1674 	struct spi_master *master = dev_get_drvdata(dev);
1675 	struct tegra_qspi *tqspi = spi_master_get_devdata(master);
1676 
1677 	/* Runtime pm disabled with ACPI */
1678 	if (has_acpi_companion(tqspi->dev))
1679 		return 0;
1680 	/* flush all write which are in PPSB queue by reading back */
1681 	tegra_qspi_readl(tqspi, QSPI_COMMAND1);
1682 
1683 	clk_disable_unprepare(tqspi->clk);
1684 
1685 	return 0;
1686 }
1687 
1688 static int __maybe_unused tegra_qspi_runtime_resume(struct device *dev)
1689 {
1690 	struct spi_master *master = dev_get_drvdata(dev);
1691 	struct tegra_qspi *tqspi = spi_master_get_devdata(master);
1692 	int ret;
1693 
1694 	/* Runtime pm disabled with ACPI */
1695 	if (has_acpi_companion(tqspi->dev))
1696 		return 0;
1697 	ret = clk_prepare_enable(tqspi->clk);
1698 	if (ret < 0)
1699 		dev_err(tqspi->dev, "failed to enable clock: %d\n", ret);
1700 
1701 	return ret;
1702 }
1703 
1704 static const struct dev_pm_ops tegra_qspi_pm_ops = {
1705 	SET_RUNTIME_PM_OPS(tegra_qspi_runtime_suspend, tegra_qspi_runtime_resume, NULL)
1706 	SET_SYSTEM_SLEEP_PM_OPS(tegra_qspi_suspend, tegra_qspi_resume)
1707 };
1708 
1709 static struct platform_driver tegra_qspi_driver = {
1710 	.driver = {
1711 		.name		= "tegra-qspi",
1712 		.pm		= &tegra_qspi_pm_ops,
1713 		.of_match_table	= tegra_qspi_of_match,
1714 		.acpi_match_table = ACPI_PTR(tegra_qspi_acpi_match),
1715 	},
1716 	.probe =	tegra_qspi_probe,
1717 	.remove =	tegra_qspi_remove,
1718 };
1719 module_platform_driver(tegra_qspi_driver);
1720 
1721 MODULE_ALIAS("platform:qspi-tegra");
1722 MODULE_DESCRIPTION("NVIDIA Tegra QSPI Controller Driver");
1723 MODULE_AUTHOR("Sowjanya Komatineni <skomatineni@nvidia.com>");
1724 MODULE_LICENSE("GPL v2");
1725