xref: /linux/drivers/spi/spi-tegra210-quad.c (revision fa5ef105618ae9b5aaa51b3f09e41d88d4514207)
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/reset.h>
22 #include <linux/spi/spi.h>
23 #include <linux/acpi.h>
24 #include <linux/property.h>
25 #include <linux/sizes.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_DMA_MEM_ADDRESS			0x028
115 #define QSPI_DMA_HI_ADDRESS			0x02c
116 
117 #define QSPI_TX_FIFO				0x108
118 #define QSPI_RX_FIFO				0x188
119 
120 #define QSPI_FIFO_DEPTH				64
121 
122 #define QSPI_INTR_MASK				0x18c
123 #define QSPI_INTR_RX_FIFO_UNF_MASK		BIT(25)
124 #define QSPI_INTR_RX_FIFO_OVF_MASK		BIT(26)
125 #define QSPI_INTR_TX_FIFO_UNF_MASK		BIT(27)
126 #define QSPI_INTR_TX_FIFO_OVF_MASK		BIT(28)
127 #define QSPI_INTR_RDY_MASK			BIT(29)
128 #define QSPI_INTR_RX_TX_FIFO_ERR		(QSPI_INTR_RX_FIFO_UNF_MASK | \
129 						 QSPI_INTR_RX_FIFO_OVF_MASK | \
130 						 QSPI_INTR_TX_FIFO_UNF_MASK | \
131 						 QSPI_INTR_TX_FIFO_OVF_MASK)
132 
133 #define QSPI_MISC_REG                           0x194
134 #define QSPI_NUM_DUMMY_CYCLE(x)			(((x) & 0xff) << 0)
135 #define QSPI_DUMMY_CYCLES_MAX			0xff
136 
137 #define QSPI_CMB_SEQ_CMD			0x19c
138 #define QSPI_COMMAND_VALUE_SET(X)		(((x) & 0xFF) << 0)
139 
140 #define QSPI_CMB_SEQ_CMD_CFG			0x1a0
141 #define QSPI_COMMAND_X1_X2_X4(x)		((((x) >> 1) & 0x3) << 13)
142 #define QSPI_COMMAND_X1_X2_X4_MASK		(0x03 << 13)
143 #define QSPI_COMMAND_SDR_DDR			BIT(12)
144 #define QSPI_COMMAND_SIZE_SET(x)		(((x) & 0xFF) << 0)
145 
146 #define QSPI_GLOBAL_CONFIG			0X1a4
147 #define QSPI_CMB_SEQ_EN				BIT(0)
148 #define QSPI_TPM_WAIT_POLL_EN			BIT(1)
149 
150 #define QSPI_CMB_SEQ_ADDR			0x1a8
151 #define QSPI_ADDRESS_VALUE_SET(X)		(((x) & 0xFFFF) << 0)
152 
153 #define QSPI_CMB_SEQ_ADDR_CFG			0x1ac
154 #define QSPI_ADDRESS_X1_X2_X4(x)		((((x) >> 1) & 0x3) << 13)
155 #define QSPI_ADDRESS_X1_X2_X4_MASK		(0x03 << 13)
156 #define QSPI_ADDRESS_SDR_DDR			BIT(12)
157 #define QSPI_ADDRESS_SIZE_SET(x)		(((x) & 0xFF) << 0)
158 
159 #define DATA_DIR_TX				BIT(0)
160 #define DATA_DIR_RX				BIT(1)
161 
162 #define QSPI_DMA_TIMEOUT			(msecs_to_jiffies(1000))
163 #define DEFAULT_QSPI_DMA_BUF_LEN		SZ_64K
164 
165 enum tegra_qspi_transfer_type {
166 	CMD_TRANSFER   = 0,
167 	ADDR_TRANSFER  = 1,
168 	DUMMY_TRANSFER = 2,
169 	DATA_TRANSFER  = 3
170 };
171 
172 struct tegra_qspi_soc_data {
173 	bool cmb_xfer_capable;
174 	bool supports_tpm;
175 	bool has_ext_dma;
176 	unsigned int cs_count;
177 };
178 
179 struct tegra_qspi_client_data {
180 	int tx_clk_tap_delay;
181 	int rx_clk_tap_delay;
182 };
183 
184 struct tegra_qspi {
185 	struct device				*dev;
186 	struct spi_controller			*host;
187 	/* lock to protect data accessed by irq */
188 	spinlock_t				lock;
189 
190 	struct clk				*clk;
191 	void __iomem				*base;
192 	phys_addr_t				phys;
193 	unsigned int				irq;
194 
195 	u32					cur_speed;
196 	unsigned int				cur_pos;
197 	unsigned int				words_per_32bit;
198 	unsigned int				bytes_per_word;
199 	unsigned int				curr_dma_words;
200 	unsigned int				cur_direction;
201 
202 	unsigned int				cur_rx_pos;
203 	unsigned int				cur_tx_pos;
204 
205 	unsigned int				dma_buf_size;
206 	unsigned int				max_buf_size;
207 	bool					is_curr_dma_xfer;
208 
209 	struct completion			rx_dma_complete;
210 	struct completion			tx_dma_complete;
211 
212 	u32					tx_status;
213 	u32					rx_status;
214 	u32					status_reg;
215 	bool					is_packed;
216 	bool					use_dma;
217 
218 	u32					command1_reg;
219 	u32					dma_control_reg;
220 	u32					def_command1_reg;
221 	u32					def_command2_reg;
222 	u32					spi_cs_timing1;
223 	u32					spi_cs_timing2;
224 	u8					dummy_cycles;
225 
226 	struct completion			xfer_completion;
227 	struct spi_transfer			*curr_xfer;
228 
229 	struct dma_chan				*rx_dma_chan;
230 	u32					*rx_dma_buf;
231 	dma_addr_t				rx_dma_phys;
232 	struct dma_async_tx_descriptor		*rx_dma_desc;
233 
234 	struct dma_chan				*tx_dma_chan;
235 	u32					*tx_dma_buf;
236 	dma_addr_t				tx_dma_phys;
237 	struct dma_async_tx_descriptor		*tx_dma_desc;
238 	const struct tegra_qspi_soc_data	*soc_data;
239 };
240 
241 static inline u32 tegra_qspi_readl(struct tegra_qspi *tqspi, unsigned long offset)
242 {
243 	return readl(tqspi->base + offset);
244 }
245 
246 static inline void tegra_qspi_writel(struct tegra_qspi *tqspi, u32 value, unsigned long offset)
247 {
248 	writel(value, tqspi->base + offset);
249 
250 	/* read back register to make sure that register writes completed */
251 	if (offset != QSPI_TX_FIFO)
252 		readl(tqspi->base + QSPI_COMMAND1);
253 }
254 
255 static void tegra_qspi_mask_clear_irq(struct tegra_qspi *tqspi)
256 {
257 	u32 value;
258 
259 	/* write 1 to clear status register */
260 	value = tegra_qspi_readl(tqspi, QSPI_TRANS_STATUS);
261 	tegra_qspi_writel(tqspi, value, QSPI_TRANS_STATUS);
262 
263 	value = tegra_qspi_readl(tqspi, QSPI_INTR_MASK);
264 	if (!(value & QSPI_INTR_RDY_MASK)) {
265 		value |= (QSPI_INTR_RDY_MASK | QSPI_INTR_RX_TX_FIFO_ERR);
266 		tegra_qspi_writel(tqspi, value, QSPI_INTR_MASK);
267 	}
268 
269 	/* clear fifo status error if any */
270 	value = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
271 	if (value & QSPI_ERR)
272 		tegra_qspi_writel(tqspi, QSPI_ERR | QSPI_FIFO_ERROR, QSPI_FIFO_STATUS);
273 }
274 
275 static unsigned int
276 tegra_qspi_calculate_curr_xfer_param(struct tegra_qspi *tqspi, struct spi_transfer *t)
277 {
278 	unsigned int max_word, max_len, total_fifo_words;
279 	unsigned int remain_len = t->len - tqspi->cur_pos;
280 	unsigned int bits_per_word = t->bits_per_word;
281 
282 	tqspi->bytes_per_word = DIV_ROUND_UP(bits_per_word, 8);
283 
284 	/*
285 	 * Tegra QSPI controller supports packed or unpacked mode transfers.
286 	 * Packed mode is used for data transfers using 8, 16, or 32 bits per
287 	 * word with a minimum transfer of 1 word and for all other transfers
288 	 * unpacked mode will be used.
289 	 */
290 
291 	if ((bits_per_word == 8 || bits_per_word == 16 ||
292 	     bits_per_word == 32) && t->len > 3) {
293 		tqspi->is_packed = true;
294 		tqspi->words_per_32bit = 32 / bits_per_word;
295 	} else {
296 		tqspi->is_packed = false;
297 		tqspi->words_per_32bit = 1;
298 	}
299 
300 	if (tqspi->is_packed) {
301 		max_len = min(remain_len, tqspi->max_buf_size);
302 		tqspi->curr_dma_words = max_len / tqspi->bytes_per_word;
303 		total_fifo_words = (max_len + 3) / 4;
304 	} else {
305 		max_word = (remain_len - 1) / tqspi->bytes_per_word + 1;
306 		max_word = min(max_word, tqspi->max_buf_size / 4);
307 		tqspi->curr_dma_words = max_word;
308 		total_fifo_words = max_word;
309 	}
310 
311 	return total_fifo_words;
312 }
313 
314 static unsigned int
315 tegra_qspi_fill_tx_fifo_from_client_txbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
316 {
317 	unsigned int written_words, fifo_words_left, count;
318 	unsigned int len, tx_empty_count, max_n_32bit, i;
319 	u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
320 	u32 fifo_status;
321 
322 	fifo_status = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
323 	tx_empty_count = QSPI_TX_FIFO_EMPTY_COUNT(fifo_status);
324 
325 	if (tqspi->is_packed) {
326 		fifo_words_left = tx_empty_count * tqspi->words_per_32bit;
327 		written_words = min(fifo_words_left, tqspi->curr_dma_words);
328 		len = written_words * tqspi->bytes_per_word;
329 		max_n_32bit = DIV_ROUND_UP(len, 4);
330 		for (count = 0; count < max_n_32bit; count++) {
331 			u32 x = 0;
332 
333 			for (i = 0; (i < 4) && len; i++, len--)
334 				x |= (u32)(*tx_buf++) << (i * 8);
335 			tegra_qspi_writel(tqspi, x, QSPI_TX_FIFO);
336 		}
337 
338 		tqspi->cur_tx_pos += written_words * tqspi->bytes_per_word;
339 	} else {
340 		unsigned int write_bytes;
341 		u8 bytes_per_word = tqspi->bytes_per_word;
342 
343 		max_n_32bit = min(tqspi->curr_dma_words, tx_empty_count);
344 		written_words = max_n_32bit;
345 		len = written_words * tqspi->bytes_per_word;
346 		if (len > t->len - tqspi->cur_pos)
347 			len = t->len - tqspi->cur_pos;
348 		write_bytes = len;
349 		for (count = 0; count < max_n_32bit; count++) {
350 			u32 x = 0;
351 
352 			for (i = 0; len && (i < min(4, bytes_per_word)); i++, len--)
353 				x |= (u32)(*tx_buf++) << (i * 8);
354 			tegra_qspi_writel(tqspi, x, QSPI_TX_FIFO);
355 		}
356 
357 		tqspi->cur_tx_pos += write_bytes;
358 	}
359 
360 	return written_words;
361 }
362 
363 static unsigned int
364 tegra_qspi_read_rx_fifo_to_client_rxbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
365 {
366 	u8 *rx_buf = (u8 *)t->rx_buf + tqspi->cur_rx_pos;
367 	unsigned int len, rx_full_count, count, i;
368 	unsigned int read_words = 0;
369 	u32 fifo_status, x;
370 
371 	fifo_status = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
372 	rx_full_count = QSPI_RX_FIFO_FULL_COUNT(fifo_status);
373 	if (tqspi->is_packed) {
374 		len = tqspi->curr_dma_words * tqspi->bytes_per_word;
375 		for (count = 0; count < rx_full_count; count++) {
376 			x = tegra_qspi_readl(tqspi, QSPI_RX_FIFO);
377 
378 			for (i = 0; len && (i < 4); i++, len--)
379 				*rx_buf++ = (x >> i * 8) & 0xff;
380 		}
381 
382 		read_words += tqspi->curr_dma_words;
383 		tqspi->cur_rx_pos += tqspi->curr_dma_words * tqspi->bytes_per_word;
384 	} else {
385 		u32 rx_mask = ((u32)1 << t->bits_per_word) - 1;
386 		u8 bytes_per_word = tqspi->bytes_per_word;
387 		unsigned int read_bytes;
388 
389 		len = rx_full_count * bytes_per_word;
390 		if (len > t->len - tqspi->cur_pos)
391 			len = t->len - tqspi->cur_pos;
392 		read_bytes = len;
393 		for (count = 0; count < rx_full_count; count++) {
394 			x = tegra_qspi_readl(tqspi, QSPI_RX_FIFO) & rx_mask;
395 
396 			for (i = 0; len && (i < bytes_per_word); i++, len--)
397 				*rx_buf++ = (x >> (i * 8)) & 0xff;
398 		}
399 
400 		read_words += rx_full_count;
401 		tqspi->cur_rx_pos += read_bytes;
402 	}
403 
404 	return read_words;
405 }
406 
407 static void
408 tegra_qspi_copy_client_txbuf_to_qspi_txbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
409 {
410 	/*
411 	 * In packed mode, each word in FIFO may contain multiple packets
412 	 * based on bits per word. So all bytes in each FIFO word are valid.
413 	 *
414 	 * In unpacked mode, each word in FIFO contains single packet and
415 	 * based on bits per word any remaining bits in FIFO word will be
416 	 * ignored by the hardware and are invalid bits.
417 	 */
418 	if (tqspi->is_packed) {
419 		tqspi->cur_tx_pos += tqspi->curr_dma_words * tqspi->bytes_per_word;
420 	} else {
421 		u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
422 		unsigned int i, count, consume, write_bytes;
423 
424 		/*
425 		 * Fill tx_dma_buf to contain single packet in each word based
426 		 * on bits per word from SPI core tx_buf.
427 		 */
428 		consume = tqspi->curr_dma_words * tqspi->bytes_per_word;
429 		if (consume > t->len - tqspi->cur_pos)
430 			consume = t->len - tqspi->cur_pos;
431 		write_bytes = consume;
432 		for (count = 0; count < tqspi->curr_dma_words; count++) {
433 			u32 x = 0;
434 
435 			for (i = 0; consume && (i < tqspi->bytes_per_word); i++, consume--)
436 				x |= (u32)(*tx_buf++) << (i * 8);
437 			tqspi->tx_dma_buf[count] = x;
438 		}
439 
440 		tqspi->cur_tx_pos += write_bytes;
441 	}
442 }
443 
444 static void
445 tegra_qspi_copy_qspi_rxbuf_to_client_rxbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
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 
474 static void tegra_qspi_dma_complete(void *args)
475 {
476 	struct completion *dma_complete = args;
477 
478 	complete(dma_complete);
479 }
480 
481 static int tegra_qspi_start_tx_dma(struct tegra_qspi *tqspi, struct spi_transfer *t, int len)
482 {
483 	dma_addr_t tx_dma_phys;
484 
485 	reinit_completion(&tqspi->tx_dma_complete);
486 
487 	if (tqspi->is_packed)
488 		tx_dma_phys = t->tx_dma;
489 	else
490 		tx_dma_phys = tqspi->tx_dma_phys;
491 
492 	tqspi->tx_dma_desc = dmaengine_prep_slave_single(tqspi->tx_dma_chan, tx_dma_phys,
493 							 len, DMA_MEM_TO_DEV,
494 							 DMA_PREP_INTERRUPT |  DMA_CTRL_ACK);
495 
496 	if (!tqspi->tx_dma_desc) {
497 		dev_err(tqspi->dev, "Unable to get TX descriptor\n");
498 		return -EIO;
499 	}
500 
501 	tqspi->tx_dma_desc->callback = tegra_qspi_dma_complete;
502 	tqspi->tx_dma_desc->callback_param = &tqspi->tx_dma_complete;
503 	dmaengine_submit(tqspi->tx_dma_desc);
504 	dma_async_issue_pending(tqspi->tx_dma_chan);
505 
506 	return 0;
507 }
508 
509 static int tegra_qspi_start_rx_dma(struct tegra_qspi *tqspi, struct spi_transfer *t, int len)
510 {
511 	dma_addr_t rx_dma_phys;
512 
513 	reinit_completion(&tqspi->rx_dma_complete);
514 
515 	if (tqspi->is_packed)
516 		rx_dma_phys = t->rx_dma;
517 	else
518 		rx_dma_phys = tqspi->rx_dma_phys;
519 
520 	tqspi->rx_dma_desc = dmaengine_prep_slave_single(tqspi->rx_dma_chan, rx_dma_phys,
521 							 len, DMA_DEV_TO_MEM,
522 							 DMA_PREP_INTERRUPT |  DMA_CTRL_ACK);
523 
524 	if (!tqspi->rx_dma_desc) {
525 		dev_err(tqspi->dev, "Unable to get RX descriptor\n");
526 		return -EIO;
527 	}
528 
529 	tqspi->rx_dma_desc->callback = tegra_qspi_dma_complete;
530 	tqspi->rx_dma_desc->callback_param = &tqspi->rx_dma_complete;
531 	dmaengine_submit(tqspi->rx_dma_desc);
532 	dma_async_issue_pending(tqspi->rx_dma_chan);
533 
534 	return 0;
535 }
536 
537 static int tegra_qspi_flush_fifos(struct tegra_qspi *tqspi, bool atomic)
538 {
539 	void __iomem *addr = tqspi->base + QSPI_FIFO_STATUS;
540 	u32 val;
541 
542 	val = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
543 	if ((val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY)
544 		return 0;
545 
546 	val |= QSPI_RX_FIFO_FLUSH | QSPI_TX_FIFO_FLUSH;
547 	tegra_qspi_writel(tqspi, val, QSPI_FIFO_STATUS);
548 
549 	if (!atomic)
550 		return readl_relaxed_poll_timeout(addr, val,
551 						  (val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY,
552 						  1000, 1000000);
553 
554 	return readl_relaxed_poll_timeout_atomic(addr, val,
555 						 (val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY,
556 						 1000, 1000000);
557 }
558 
559 static void tegra_qspi_unmask_irq(struct tegra_qspi *tqspi)
560 {
561 	u32 intr_mask;
562 
563 	intr_mask = tegra_qspi_readl(tqspi, QSPI_INTR_MASK);
564 	intr_mask &= ~(QSPI_INTR_RDY_MASK | QSPI_INTR_RX_TX_FIFO_ERR);
565 	tegra_qspi_writel(tqspi, intr_mask, QSPI_INTR_MASK);
566 }
567 
568 static int tegra_qspi_dma_map_xfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
569 {
570 	u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
571 	u8 *rx_buf = (u8 *)t->rx_buf + tqspi->cur_rx_pos;
572 	unsigned int len;
573 
574 	len = DIV_ROUND_UP(tqspi->curr_dma_words * tqspi->bytes_per_word, 4) * 4;
575 
576 	if (t->tx_buf) {
577 		t->tx_dma = dma_map_single(tqspi->dev, (void *)tx_buf, len, DMA_TO_DEVICE);
578 		if (dma_mapping_error(tqspi->dev, t->tx_dma))
579 			return -ENOMEM;
580 	}
581 
582 	if (t->rx_buf) {
583 		t->rx_dma = dma_map_single(tqspi->dev, (void *)rx_buf, len, DMA_FROM_DEVICE);
584 		if (dma_mapping_error(tqspi->dev, t->rx_dma)) {
585 			dma_unmap_single(tqspi->dev, t->tx_dma, len, DMA_TO_DEVICE);
586 			return -ENOMEM;
587 		}
588 	}
589 
590 	return 0;
591 }
592 
593 static void tegra_qspi_dma_unmap_xfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
594 {
595 	unsigned int len;
596 
597 	len = DIV_ROUND_UP(tqspi->curr_dma_words * tqspi->bytes_per_word, 4) * 4;
598 
599 	if (t->tx_buf)
600 		dma_unmap_single(tqspi->dev, t->tx_dma, len, DMA_TO_DEVICE);
601 	if (t->rx_buf)
602 		dma_unmap_single(tqspi->dev, t->rx_dma, len, DMA_FROM_DEVICE);
603 }
604 
605 static int tegra_qspi_start_dma_based_transfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
606 {
607 	struct dma_slave_config dma_sconfig = { 0 };
608 	dma_addr_t rx_dma_phys, tx_dma_phys;
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 	if (tqspi->soc_data->has_ext_dma) {
632 		val = 0;
633 		if (len & 0xf) {
634 			val |= QSPI_TX_TRIG_1 | QSPI_RX_TRIG_1;
635 			dma_burst = 1;
636 		} else if (((len) >> 4) & 0x1) {
637 			val |= QSPI_TX_TRIG_4 | QSPI_RX_TRIG_4;
638 			dma_burst = 4;
639 		} else {
640 			val |= QSPI_TX_TRIG_8 | QSPI_RX_TRIG_8;
641 			dma_burst = 8;
642 		}
643 
644 		tegra_qspi_writel(tqspi, val, QSPI_DMA_CTL);
645 	}
646 
647 	tqspi->dma_control_reg = val;
648 
649 	dma_sconfig.device_fc = true;
650 
651 	if (tqspi->cur_direction & DATA_DIR_TX) {
652 		if (tqspi->tx_dma_chan) {
653 			dma_sconfig.dst_addr = tqspi->phys + QSPI_TX_FIFO;
654 			dma_sconfig.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
655 			dma_sconfig.dst_maxburst = dma_burst;
656 			ret = dmaengine_slave_config(tqspi->tx_dma_chan, &dma_sconfig);
657 			if (ret < 0) {
658 				dev_err(tqspi->dev, "failed DMA slave config: %d\n", ret);
659 				return ret;
660 			}
661 
662 			tegra_qspi_copy_client_txbuf_to_qspi_txbuf(tqspi, t);
663 			ret = tegra_qspi_start_tx_dma(tqspi, t, len);
664 			if (ret < 0) {
665 				dev_err(tqspi->dev, "failed to starting TX DMA: %d\n", ret);
666 				return ret;
667 			}
668 		} else {
669 			if (tqspi->is_packed)
670 				tx_dma_phys = t->tx_dma;
671 			else
672 				tx_dma_phys = tqspi->tx_dma_phys;
673 			tegra_qspi_copy_client_txbuf_to_qspi_txbuf(tqspi, t);
674 			tegra_qspi_writel(tqspi, lower_32_bits(tx_dma_phys),
675 					  QSPI_DMA_MEM_ADDRESS);
676 			tegra_qspi_writel(tqspi, (upper_32_bits(tx_dma_phys) & 0xff),
677 					  QSPI_DMA_HI_ADDRESS);
678 		}
679 	}
680 
681 	if (tqspi->cur_direction & DATA_DIR_RX) {
682 		if (tqspi->rx_dma_chan) {
683 			dma_sconfig.src_addr = tqspi->phys + QSPI_RX_FIFO;
684 			dma_sconfig.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
685 			dma_sconfig.src_maxburst = dma_burst;
686 			ret = dmaengine_slave_config(tqspi->rx_dma_chan, &dma_sconfig);
687 			if (ret < 0) {
688 				dev_err(tqspi->dev, "failed DMA slave config: %d\n", ret);
689 				return ret;
690 			}
691 
692 			ret = tegra_qspi_start_rx_dma(tqspi, t, len);
693 			if (ret < 0) {
694 				dev_err(tqspi->dev, "failed to start RX DMA: %d\n", ret);
695 				if (tqspi->cur_direction & DATA_DIR_TX)
696 					dmaengine_terminate_all(tqspi->tx_dma_chan);
697 				return ret;
698 			}
699 		} else {
700 			if (tqspi->is_packed)
701 				rx_dma_phys = t->rx_dma;
702 			else
703 				rx_dma_phys = tqspi->rx_dma_phys;
704 
705 			tegra_qspi_writel(tqspi, lower_32_bits(rx_dma_phys),
706 					  QSPI_DMA_MEM_ADDRESS);
707 			tegra_qspi_writel(tqspi, (upper_32_bits(rx_dma_phys) & 0xff),
708 					  QSPI_DMA_HI_ADDRESS);
709 		}
710 	}
711 
712 	tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
713 
714 	tqspi->is_curr_dma_xfer = true;
715 	tqspi->dma_control_reg = val;
716 	val |= QSPI_DMA_EN;
717 	tegra_qspi_writel(tqspi, val, QSPI_DMA_CTL);
718 
719 	return ret;
720 }
721 
722 static int tegra_qspi_start_cpu_based_transfer(struct tegra_qspi *qspi, struct spi_transfer *t)
723 {
724 	u32 val;
725 	unsigned int cur_words;
726 
727 	if (qspi->cur_direction & DATA_DIR_TX)
728 		cur_words = tegra_qspi_fill_tx_fifo_from_client_txbuf(qspi, t);
729 	else
730 		cur_words = qspi->curr_dma_words;
731 
732 	val = QSPI_DMA_BLK_SET(cur_words - 1);
733 	tegra_qspi_writel(qspi, val, QSPI_DMA_BLK);
734 
735 	tegra_qspi_unmask_irq(qspi);
736 
737 	qspi->is_curr_dma_xfer = false;
738 	val = qspi->command1_reg;
739 	val |= QSPI_PIO;
740 	tegra_qspi_writel(qspi, val, QSPI_COMMAND1);
741 
742 	return 0;
743 }
744 
745 static void tegra_qspi_deinit_dma(struct tegra_qspi *tqspi)
746 {
747 	if (tqspi->tx_dma_buf) {
748 		dma_free_coherent(tqspi->dev, tqspi->dma_buf_size,
749 				  tqspi->tx_dma_buf, tqspi->tx_dma_phys);
750 		tqspi->tx_dma_buf = NULL;
751 	}
752 
753 	if (tqspi->tx_dma_chan) {
754 		dma_release_channel(tqspi->tx_dma_chan);
755 		tqspi->tx_dma_chan = NULL;
756 	}
757 
758 	if (tqspi->rx_dma_buf) {
759 		dma_free_coherent(tqspi->dev, tqspi->dma_buf_size,
760 				  tqspi->rx_dma_buf, tqspi->rx_dma_phys);
761 		tqspi->rx_dma_buf = NULL;
762 	}
763 
764 	if (tqspi->rx_dma_chan) {
765 		dma_release_channel(tqspi->rx_dma_chan);
766 		tqspi->rx_dma_chan = NULL;
767 	}
768 }
769 
770 static int tegra_qspi_init_dma(struct tegra_qspi *tqspi)
771 {
772 	struct dma_chan *dma_chan;
773 	dma_addr_t dma_phys;
774 	u32 *dma_buf;
775 	int err;
776 
777 	if (tqspi->soc_data->has_ext_dma) {
778 		dma_chan = dma_request_chan(tqspi->dev, "rx");
779 		if (IS_ERR(dma_chan)) {
780 			err = PTR_ERR(dma_chan);
781 			goto err_out;
782 		}
783 
784 		tqspi->rx_dma_chan = dma_chan;
785 
786 		dma_chan = dma_request_chan(tqspi->dev, "tx");
787 		if (IS_ERR(dma_chan)) {
788 			err = PTR_ERR(dma_chan);
789 			goto err_out;
790 		}
791 
792 		tqspi->tx_dma_chan = dma_chan;
793 	} else {
794 		if (!device_iommu_mapped(tqspi->dev)) {
795 			dev_warn(tqspi->dev,
796 				 "IOMMU not enabled in device-tree, falling back to PIO mode\n");
797 			return 0;
798 		}
799 	}
800 
801 	dma_buf = dma_alloc_coherent(tqspi->dev, tqspi->dma_buf_size, &dma_phys, GFP_KERNEL);
802 	if (!dma_buf) {
803 		err = -ENOMEM;
804 		goto err_out;
805 	}
806 
807 	tqspi->rx_dma_buf = dma_buf;
808 	tqspi->rx_dma_phys = dma_phys;
809 
810 	dma_buf = dma_alloc_coherent(tqspi->dev, tqspi->dma_buf_size, &dma_phys, GFP_KERNEL);
811 	if (!dma_buf) {
812 		err = -ENOMEM;
813 		goto err_out;
814 	}
815 
816 	tqspi->tx_dma_buf = dma_buf;
817 	tqspi->tx_dma_phys = dma_phys;
818 	tqspi->use_dma = true;
819 
820 	return 0;
821 
822 err_out:
823 	tegra_qspi_deinit_dma(tqspi);
824 
825 	if (err != -EPROBE_DEFER) {
826 		dev_err(tqspi->dev, "cannot use DMA: %d\n", err);
827 		dev_err(tqspi->dev, "falling back to PIO\n");
828 		return 0;
829 	}
830 
831 	return err;
832 }
833 
834 static u32 tegra_qspi_setup_transfer_one(struct spi_device *spi, struct spi_transfer *t,
835 					 bool is_first_of_msg)
836 {
837 	struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);
838 	struct tegra_qspi_client_data *cdata = spi->controller_data;
839 	u32 command1, command2, speed = t->speed_hz;
840 	u8 bits_per_word = t->bits_per_word;
841 	u32 tx_tap = 0, rx_tap = 0;
842 	int req_mode;
843 
844 	if (!has_acpi_companion(tqspi->dev) && speed != tqspi->cur_speed) {
845 		clk_set_rate(tqspi->clk, speed);
846 		tqspi->cur_speed = speed;
847 	}
848 
849 	tqspi->cur_pos = 0;
850 	tqspi->cur_rx_pos = 0;
851 	tqspi->cur_tx_pos = 0;
852 	tqspi->curr_xfer = t;
853 
854 	if (is_first_of_msg) {
855 		tegra_qspi_mask_clear_irq(tqspi);
856 
857 		command1 = tqspi->def_command1_reg;
858 		command1 |= QSPI_CS_SEL(spi_get_chipselect(spi, 0));
859 		command1 |= QSPI_BIT_LENGTH(bits_per_word - 1);
860 
861 		command1 &= ~QSPI_CONTROL_MODE_MASK;
862 		req_mode = spi->mode & 0x3;
863 		if (req_mode == SPI_MODE_3)
864 			command1 |= QSPI_CONTROL_MODE_3;
865 		else
866 			command1 |= QSPI_CONTROL_MODE_0;
867 
868 		if (spi->mode & SPI_CS_HIGH)
869 			command1 |= QSPI_CS_SW_VAL;
870 		else
871 			command1 &= ~QSPI_CS_SW_VAL;
872 		tegra_qspi_writel(tqspi, command1, QSPI_COMMAND1);
873 
874 		if (cdata && cdata->tx_clk_tap_delay)
875 			tx_tap = cdata->tx_clk_tap_delay;
876 
877 		if (cdata && cdata->rx_clk_tap_delay)
878 			rx_tap = cdata->rx_clk_tap_delay;
879 
880 		command2 = QSPI_TX_TAP_DELAY(tx_tap) | QSPI_RX_TAP_DELAY(rx_tap);
881 		if (command2 != tqspi->def_command2_reg)
882 			tegra_qspi_writel(tqspi, command2, QSPI_COMMAND2);
883 
884 	} else {
885 		command1 = tqspi->command1_reg;
886 		command1 &= ~QSPI_BIT_LENGTH(~0);
887 		command1 |= QSPI_BIT_LENGTH(bits_per_word - 1);
888 	}
889 
890 	command1 &= ~QSPI_SDR_DDR_SEL;
891 
892 	return command1;
893 }
894 
895 static int tegra_qspi_start_transfer_one(struct spi_device *spi,
896 					 struct spi_transfer *t, u32 command1)
897 {
898 	struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);
899 	unsigned int total_fifo_words;
900 	u8 bus_width = 0;
901 	int ret;
902 
903 	total_fifo_words = tegra_qspi_calculate_curr_xfer_param(tqspi, t);
904 
905 	command1 &= ~QSPI_PACKED;
906 	if (tqspi->is_packed)
907 		command1 |= QSPI_PACKED;
908 	tegra_qspi_writel(tqspi, command1, QSPI_COMMAND1);
909 
910 	tqspi->cur_direction = 0;
911 
912 	command1 &= ~(QSPI_TX_EN | QSPI_RX_EN);
913 	if (t->rx_buf) {
914 		command1 |= QSPI_RX_EN;
915 		tqspi->cur_direction |= DATA_DIR_RX;
916 		bus_width = t->rx_nbits;
917 	}
918 
919 	if (t->tx_buf) {
920 		command1 |= QSPI_TX_EN;
921 		tqspi->cur_direction |= DATA_DIR_TX;
922 		bus_width = t->tx_nbits;
923 	}
924 
925 	command1 &= ~QSPI_INTERFACE_WIDTH_MASK;
926 
927 	if (bus_width == SPI_NBITS_QUAD)
928 		command1 |= QSPI_INTERFACE_WIDTH_QUAD;
929 	else if (bus_width == SPI_NBITS_DUAL)
930 		command1 |= QSPI_INTERFACE_WIDTH_DUAL;
931 	else
932 		command1 |= QSPI_INTERFACE_WIDTH_SINGLE;
933 
934 	tqspi->command1_reg = command1;
935 
936 	tegra_qspi_writel(tqspi, QSPI_NUM_DUMMY_CYCLE(tqspi->dummy_cycles), QSPI_MISC_REG);
937 
938 	ret = tegra_qspi_flush_fifos(tqspi, false);
939 	if (ret < 0)
940 		return ret;
941 
942 	if (tqspi->use_dma && total_fifo_words > QSPI_FIFO_DEPTH)
943 		ret = tegra_qspi_start_dma_based_transfer(tqspi, t);
944 	else
945 		ret = tegra_qspi_start_cpu_based_transfer(tqspi, t);
946 
947 	return ret;
948 }
949 
950 static struct tegra_qspi_client_data *tegra_qspi_parse_cdata_dt(struct spi_device *spi)
951 {
952 	struct tegra_qspi_client_data *cdata;
953 	struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);
954 
955 	cdata = devm_kzalloc(tqspi->dev, sizeof(*cdata), GFP_KERNEL);
956 	if (!cdata)
957 		return NULL;
958 
959 	device_property_read_u32(&spi->dev, "nvidia,tx-clk-tap-delay",
960 				 &cdata->tx_clk_tap_delay);
961 	device_property_read_u32(&spi->dev, "nvidia,rx-clk-tap-delay",
962 				 &cdata->rx_clk_tap_delay);
963 
964 	return cdata;
965 }
966 
967 static int tegra_qspi_setup(struct spi_device *spi)
968 {
969 	struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);
970 	struct tegra_qspi_client_data *cdata = spi->controller_data;
971 	unsigned long flags;
972 	u32 val;
973 	int ret;
974 
975 	ret = pm_runtime_resume_and_get(tqspi->dev);
976 	if (ret < 0) {
977 		dev_err(tqspi->dev, "failed to get runtime PM: %d\n", ret);
978 		return ret;
979 	}
980 
981 	if (!cdata) {
982 		cdata = tegra_qspi_parse_cdata_dt(spi);
983 		spi->controller_data = cdata;
984 	}
985 	spin_lock_irqsave(&tqspi->lock, flags);
986 
987 	/* keep default cs state to inactive */
988 	val = tqspi->def_command1_reg;
989 	val |= QSPI_CS_SEL(spi_get_chipselect(spi, 0));
990 	if (spi->mode & SPI_CS_HIGH)
991 		val &= ~QSPI_CS_POL_INACTIVE(spi_get_chipselect(spi, 0));
992 	else
993 		val |= QSPI_CS_POL_INACTIVE(spi_get_chipselect(spi, 0));
994 
995 	tqspi->def_command1_reg = val;
996 	tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
997 
998 	spin_unlock_irqrestore(&tqspi->lock, flags);
999 
1000 	pm_runtime_put(tqspi->dev);
1001 
1002 	return 0;
1003 }
1004 
1005 static void tegra_qspi_dump_regs(struct tegra_qspi *tqspi)
1006 {
1007 	dev_dbg(tqspi->dev, "============ QSPI REGISTER DUMP ============\n");
1008 	dev_dbg(tqspi->dev, "Command1:    0x%08x | Command2:    0x%08x\n",
1009 		tegra_qspi_readl(tqspi, QSPI_COMMAND1),
1010 		tegra_qspi_readl(tqspi, QSPI_COMMAND2));
1011 	dev_dbg(tqspi->dev, "DMA_CTL:     0x%08x | DMA_BLK:     0x%08x\n",
1012 		tegra_qspi_readl(tqspi, QSPI_DMA_CTL),
1013 		tegra_qspi_readl(tqspi, QSPI_DMA_BLK));
1014 	dev_dbg(tqspi->dev, "INTR_MASK:  0x%08x | MISC: 0x%08x\n",
1015 		tegra_qspi_readl(tqspi, QSPI_INTR_MASK),
1016 		tegra_qspi_readl(tqspi, QSPI_MISC_REG));
1017 	dev_dbg(tqspi->dev, "TRANS_STAT:  0x%08x | FIFO_STATUS: 0x%08x\n",
1018 		tegra_qspi_readl(tqspi, QSPI_TRANS_STATUS),
1019 		tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS));
1020 }
1021 
1022 static void tegra_qspi_reset(struct tegra_qspi *tqspi)
1023 {
1024 	if (device_reset(tqspi->dev) < 0) {
1025 		dev_warn_once(tqspi->dev, "device reset failed\n");
1026 		tegra_qspi_mask_clear_irq(tqspi);
1027 	}
1028 }
1029 
1030 static void tegra_qspi_handle_error(struct tegra_qspi *tqspi)
1031 {
1032 	dev_err(tqspi->dev, "error in transfer, fifo status 0x%08x\n", tqspi->status_reg);
1033 	tegra_qspi_dump_regs(tqspi);
1034 	tegra_qspi_flush_fifos(tqspi, true);
1035 	tegra_qspi_reset(tqspi);
1036 }
1037 
1038 static void tegra_qspi_transfer_end(struct spi_device *spi)
1039 {
1040 	struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);
1041 	int cs_val = (spi->mode & SPI_CS_HIGH) ? 0 : 1;
1042 
1043 	if (cs_val)
1044 		tqspi->command1_reg |= QSPI_CS_SW_VAL;
1045 	else
1046 		tqspi->command1_reg &= ~QSPI_CS_SW_VAL;
1047 	tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
1048 	tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
1049 }
1050 
1051 static irqreturn_t handle_cpu_based_xfer(struct tegra_qspi *tqspi);
1052 static irqreturn_t handle_dma_based_xfer(struct tegra_qspi *tqspi);
1053 
1054 /**
1055  * tegra_qspi_handle_timeout - Handle transfer timeout with hardware check
1056  * @tqspi: QSPI controller instance
1057  *
1058  * When a timeout occurs but hardware has completed the transfer (interrupt
1059  * was lost or delayed), manually trigger transfer completion processing.
1060  * This avoids failing transfers that actually succeeded.
1061  *
1062  * Returns: 0 if transfer was completed, -ETIMEDOUT if real timeout
1063  */
1064 static int tegra_qspi_handle_timeout(struct tegra_qspi *tqspi)
1065 {
1066 	irqreturn_t ret;
1067 	u32 status;
1068 
1069 	/* Check if hardware actually completed the transfer */
1070 	status = tegra_qspi_readl(tqspi, QSPI_TRANS_STATUS);
1071 	if (!(status & QSPI_RDY))
1072 		return -ETIMEDOUT;
1073 
1074 	/*
1075 	 * Hardware completed but interrupt was lost/delayed. Manually
1076 	 * process the completion by calling the appropriate handler.
1077 	 */
1078 	dev_warn_ratelimited(tqspi->dev,
1079 			     "QSPI interrupt timeout, but transfer complete\n");
1080 
1081 	/* Clear the transfer status */
1082 	status = tegra_qspi_readl(tqspi, QSPI_TRANS_STATUS);
1083 	tegra_qspi_writel(tqspi, status, QSPI_TRANS_STATUS);
1084 
1085 	/* Manually trigger completion handler */
1086 	if (!tqspi->is_curr_dma_xfer)
1087 		ret = handle_cpu_based_xfer(tqspi);
1088 	else
1089 		ret = handle_dma_based_xfer(tqspi);
1090 
1091 	return (ret == IRQ_HANDLED) ? 0 : -EIO;
1092 }
1093 
1094 static u32 tegra_qspi_cmd_config(bool is_ddr, u8 bus_width, u8 len)
1095 {
1096 	u32 cmd_config = 0;
1097 
1098 	/* Extract Command configuration and value */
1099 	if (is_ddr)
1100 		cmd_config |= QSPI_COMMAND_SDR_DDR;
1101 	else
1102 		cmd_config &= ~QSPI_COMMAND_SDR_DDR;
1103 
1104 	cmd_config |= QSPI_COMMAND_X1_X2_X4(bus_width);
1105 	cmd_config |= QSPI_COMMAND_SIZE_SET((len * 8) - 1);
1106 
1107 	return cmd_config;
1108 }
1109 
1110 static u32 tegra_qspi_addr_config(bool is_ddr, u8 bus_width, u8 len)
1111 {
1112 	u32 addr_config = 0;
1113 
1114 	if (is_ddr)
1115 		addr_config |= QSPI_ADDRESS_SDR_DDR;
1116 	else
1117 		addr_config &= ~QSPI_ADDRESS_SDR_DDR;
1118 
1119 	addr_config |= QSPI_ADDRESS_X1_X2_X4(bus_width);
1120 	addr_config |= QSPI_ADDRESS_SIZE_SET((len * 8) - 1);
1121 
1122 	return addr_config;
1123 }
1124 
1125 static void tegra_qspi_dma_stop(struct tegra_qspi *tqspi)
1126 {
1127 	u32 value;
1128 
1129 	if ((tqspi->cur_direction & DATA_DIR_TX) && tqspi->tx_dma_chan)
1130 		dmaengine_terminate_all(tqspi->tx_dma_chan);
1131 
1132 	if ((tqspi->cur_direction & DATA_DIR_RX) && tqspi->rx_dma_chan)
1133 		dmaengine_terminate_all(tqspi->rx_dma_chan);
1134 
1135 	value = tegra_qspi_readl(tqspi, QSPI_DMA_CTL);
1136 	value &= ~QSPI_DMA_EN;
1137 	tegra_qspi_writel(tqspi, value, QSPI_DMA_CTL);
1138 }
1139 
1140 static void tegra_qspi_pio_stop(struct tegra_qspi *tqspi)
1141 {
1142 	u32 value;
1143 
1144 	value = tegra_qspi_readl(tqspi, QSPI_COMMAND1);
1145 	value &= ~QSPI_PIO;
1146 	tegra_qspi_writel(tqspi, value, QSPI_COMMAND1);
1147 }
1148 
1149 static int tegra_qspi_combined_seq_xfer(struct tegra_qspi *tqspi,
1150 					struct spi_message *msg)
1151 {
1152 	bool is_first_msg = true;
1153 	struct spi_transfer *xfer;
1154 	struct spi_device *spi = msg->spi;
1155 	u8 transfer_phase = 0;
1156 	u32 cmd1 = 0;
1157 	int ret = 0;
1158 	u32 address_value = 0;
1159 	u32 cmd_config = 0, addr_config = 0;
1160 	u8 cmd_value = 0, val = 0;
1161 
1162 	/* Enable Combined sequence mode */
1163 	val = tegra_qspi_readl(tqspi, QSPI_GLOBAL_CONFIG);
1164 	if (spi->mode & SPI_TPM_HW_FLOW) {
1165 		if (tqspi->soc_data->supports_tpm)
1166 			val |= QSPI_TPM_WAIT_POLL_EN;
1167 		else
1168 			return -EIO;
1169 	}
1170 	val |= QSPI_CMB_SEQ_EN;
1171 	tegra_qspi_writel(tqspi, val, QSPI_GLOBAL_CONFIG);
1172 	/* Process individual transfer list */
1173 	list_for_each_entry(xfer, &msg->transfers, transfer_list) {
1174 		switch (transfer_phase) {
1175 		case CMD_TRANSFER:
1176 			/* X1 SDR mode */
1177 			cmd_config = tegra_qspi_cmd_config(false, xfer->tx_nbits,
1178 							   xfer->len);
1179 			cmd_value = *((const u8 *)(xfer->tx_buf));
1180 			break;
1181 		case ADDR_TRANSFER:
1182 			/* X1 SDR mode */
1183 			addr_config = tegra_qspi_addr_config(false, xfer->tx_nbits,
1184 							     xfer->len);
1185 			address_value = *((const u32 *)(xfer->tx_buf));
1186 			break;
1187 		case DUMMY_TRANSFER:
1188 			if (xfer->dummy_data) {
1189 				tqspi->dummy_cycles = xfer->len * 8 / xfer->tx_nbits;
1190 				break;
1191 			}
1192 			transfer_phase++;
1193 			fallthrough;
1194 		case DATA_TRANSFER:
1195 			/* Program Command, Address value in register */
1196 			tegra_qspi_writel(tqspi, cmd_value, QSPI_CMB_SEQ_CMD);
1197 			tegra_qspi_writel(tqspi, address_value,
1198 					  QSPI_CMB_SEQ_ADDR);
1199 			/* Program Command and Address config in register */
1200 			tegra_qspi_writel(tqspi, cmd_config,
1201 					  QSPI_CMB_SEQ_CMD_CFG);
1202 			tegra_qspi_writel(tqspi, addr_config,
1203 					  QSPI_CMB_SEQ_ADDR_CFG);
1204 
1205 			reinit_completion(&tqspi->xfer_completion);
1206 			cmd1 = tegra_qspi_setup_transfer_one(spi, xfer,
1207 							     is_first_msg);
1208 			ret = tegra_qspi_start_transfer_one(spi, xfer,
1209 							    cmd1);
1210 
1211 			if (ret < 0) {
1212 				dev_err(tqspi->dev, "Failed to start transfer-one: %d\n",
1213 					ret);
1214 				return ret;
1215 			}
1216 
1217 			is_first_msg = false;
1218 			ret = wait_for_completion_timeout
1219 					(&tqspi->xfer_completion,
1220 					QSPI_DMA_TIMEOUT);
1221 
1222 			if (WARN_ON_ONCE(ret == 0)) {
1223 				/*
1224 				 * Check if hardware completed the transfer
1225 				 * even though interrupt was lost or delayed.
1226 				 * If so, process the completion and continue.
1227 				 */
1228 				ret = tegra_qspi_handle_timeout(tqspi);
1229 				if (ret < 0) {
1230 					/* Real timeout - clean up and fail */
1231 					dev_err(tqspi->dev, "transfer timeout\n");
1232 
1233 					/* Abort transfer by resetting pio/dma bit */
1234 					if (tqspi->is_curr_dma_xfer)
1235 						tegra_qspi_dma_stop(tqspi);
1236 					else
1237 						tegra_qspi_pio_stop(tqspi);
1238 
1239 					/* Reset controller if timeout happens */
1240 					tegra_qspi_reset(tqspi);
1241 
1242 					ret = -EIO;
1243 					goto exit;
1244 				}
1245 			}
1246 
1247 			if (tqspi->tx_status ||  tqspi->rx_status) {
1248 				dev_err(tqspi->dev, "QSPI Transfer failed\n");
1249 				tqspi->tx_status = 0;
1250 				tqspi->rx_status = 0;
1251 				ret = -EIO;
1252 				goto exit;
1253 			}
1254 			break;
1255 		default:
1256 			ret = -EINVAL;
1257 			goto exit;
1258 		}
1259 		msg->actual_length += xfer->len;
1260 		if (!xfer->cs_change && transfer_phase == DATA_TRANSFER) {
1261 			tegra_qspi_transfer_end(spi);
1262 			spi_transfer_delay_exec(xfer);
1263 		}
1264 		tqspi->curr_xfer = NULL;
1265 		transfer_phase++;
1266 	}
1267 	ret = 0;
1268 
1269 exit:
1270 	tqspi->curr_xfer = NULL;
1271 	msg->status = ret;
1272 
1273 	return ret;
1274 }
1275 
1276 static int tegra_qspi_non_combined_seq_xfer(struct tegra_qspi *tqspi,
1277 					    struct spi_message *msg)
1278 {
1279 	struct spi_device *spi = msg->spi;
1280 	struct spi_transfer *transfer;
1281 	bool is_first_msg = true;
1282 	int ret = 0, val = 0;
1283 
1284 	msg->status = 0;
1285 	msg->actual_length = 0;
1286 	tqspi->tx_status = 0;
1287 	tqspi->rx_status = 0;
1288 
1289 	/* Disable Combined sequence mode */
1290 	val = tegra_qspi_readl(tqspi, QSPI_GLOBAL_CONFIG);
1291 	val &= ~QSPI_CMB_SEQ_EN;
1292 	if (tqspi->soc_data->supports_tpm)
1293 		val &= ~QSPI_TPM_WAIT_POLL_EN;
1294 	tegra_qspi_writel(tqspi, val, QSPI_GLOBAL_CONFIG);
1295 	list_for_each_entry(transfer, &msg->transfers, transfer_list) {
1296 		struct spi_transfer *xfer = transfer;
1297 		u8 dummy_bytes = 0;
1298 		u32 cmd1;
1299 
1300 		tqspi->dummy_cycles = 0;
1301 		/*
1302 		 * Tegra QSPI hardware supports dummy bytes transfer after actual transfer
1303 		 * bytes based on programmed dummy clock cycles in the QSPI_MISC register.
1304 		 * So, check if the next transfer is dummy data transfer and program dummy
1305 		 * clock cycles along with the current transfer and skip next transfer.
1306 		 */
1307 		if (!list_is_last(&xfer->transfer_list, &msg->transfers)) {
1308 			struct spi_transfer *next_xfer;
1309 
1310 			next_xfer = list_next_entry(xfer, transfer_list);
1311 			if (next_xfer->dummy_data) {
1312 				u32 dummy_cycles = next_xfer->len * 8 / next_xfer->tx_nbits;
1313 
1314 				if (dummy_cycles <= QSPI_DUMMY_CYCLES_MAX) {
1315 					tqspi->dummy_cycles = dummy_cycles;
1316 					dummy_bytes = next_xfer->len;
1317 					transfer = next_xfer;
1318 				}
1319 			}
1320 		}
1321 
1322 		reinit_completion(&tqspi->xfer_completion);
1323 
1324 		cmd1 = tegra_qspi_setup_transfer_one(spi, xfer, is_first_msg);
1325 
1326 		ret = tegra_qspi_start_transfer_one(spi, xfer, cmd1);
1327 		if (ret < 0) {
1328 			dev_err(tqspi->dev, "failed to start transfer: %d\n", ret);
1329 			goto complete_xfer;
1330 		}
1331 
1332 		ret = wait_for_completion_timeout(&tqspi->xfer_completion,
1333 						  QSPI_DMA_TIMEOUT);
1334 		if (WARN_ON(ret == 0)) {
1335 			/*
1336 			 * Check if hardware completed the transfer even though
1337 			 * interrupt was lost or delayed. If so, process the
1338 			 * completion and continue.
1339 			 */
1340 			ret = tegra_qspi_handle_timeout(tqspi);
1341 			if (ret < 0) {
1342 				/* Real timeout - clean up and fail */
1343 				dev_err(tqspi->dev, "transfer timeout\n");
1344 
1345 				if (tqspi->is_curr_dma_xfer)
1346 					tegra_qspi_dma_stop(tqspi);
1347 
1348 				tegra_qspi_handle_error(tqspi);
1349 				ret = -EIO;
1350 				goto complete_xfer;
1351 			}
1352 		}
1353 
1354 		if (tqspi->tx_status ||  tqspi->rx_status) {
1355 			tegra_qspi_handle_error(tqspi);
1356 			ret = -EIO;
1357 			goto complete_xfer;
1358 		}
1359 
1360 		msg->actual_length += xfer->len + dummy_bytes;
1361 
1362 complete_xfer:
1363 		tqspi->curr_xfer = NULL;
1364 
1365 		if (ret < 0) {
1366 			tegra_qspi_transfer_end(spi);
1367 			spi_transfer_delay_exec(xfer);
1368 			goto exit;
1369 		}
1370 
1371 		if (list_is_last(&xfer->transfer_list, &msg->transfers)) {
1372 			/* de-activate CS after last transfer only when cs_change is not set */
1373 			if (!xfer->cs_change) {
1374 				tegra_qspi_transfer_end(spi);
1375 				spi_transfer_delay_exec(xfer);
1376 			}
1377 		} else if (xfer->cs_change) {
1378 			 /* de-activated CS between the transfers only when cs_change is set */
1379 			tegra_qspi_transfer_end(spi);
1380 			spi_transfer_delay_exec(xfer);
1381 		}
1382 	}
1383 
1384 	ret = 0;
1385 exit:
1386 	msg->status = ret;
1387 
1388 	return ret;
1389 }
1390 
1391 static bool tegra_qspi_validate_cmb_seq(struct tegra_qspi *tqspi,
1392 					struct spi_message *msg)
1393 {
1394 	int transfer_count = 0;
1395 	struct spi_transfer *xfer;
1396 
1397 	list_for_each_entry(xfer, &msg->transfers, transfer_list) {
1398 		transfer_count++;
1399 	}
1400 	if (!tqspi->soc_data->cmb_xfer_capable)
1401 		return false;
1402 	if (transfer_count > 4 || transfer_count < 3)
1403 		return false;
1404 	xfer = list_first_entry(&msg->transfers, typeof(*xfer),
1405 				transfer_list);
1406 	if (xfer->len > 2)
1407 		return false;
1408 	xfer = list_next_entry(xfer, transfer_list);
1409 	if (xfer->len > 4 || xfer->len < 3)
1410 		return false;
1411 	xfer = list_next_entry(xfer, transfer_list);
1412 	if (transfer_count == 4) {
1413 		if (xfer->dummy_data != 1)
1414 			return false;
1415 		if ((xfer->len * 8 / xfer->tx_nbits) > QSPI_DUMMY_CYCLES_MAX)
1416 			return false;
1417 		xfer = list_next_entry(xfer, transfer_list);
1418 	}
1419 	if (!tqspi->soc_data->has_ext_dma && xfer->len > (QSPI_FIFO_DEPTH << 2))
1420 		return false;
1421 
1422 	return true;
1423 }
1424 
1425 static int tegra_qspi_transfer_one_message(struct spi_controller *host,
1426 					   struct spi_message *msg)
1427 {
1428 	struct tegra_qspi *tqspi = spi_controller_get_devdata(host);
1429 	int ret;
1430 
1431 	if (tegra_qspi_validate_cmb_seq(tqspi, msg))
1432 		ret = tegra_qspi_combined_seq_xfer(tqspi, msg);
1433 	else
1434 		ret = tegra_qspi_non_combined_seq_xfer(tqspi, msg);
1435 
1436 	spi_finalize_current_message(host);
1437 
1438 	return ret;
1439 }
1440 
1441 static irqreturn_t handle_cpu_based_xfer(struct tegra_qspi *tqspi)
1442 {
1443 	struct spi_transfer *t = tqspi->curr_xfer;
1444 	unsigned long flags;
1445 
1446 	spin_lock_irqsave(&tqspi->lock, flags);
1447 
1448 	if (tqspi->tx_status ||  tqspi->rx_status) {
1449 		tegra_qspi_handle_error(tqspi);
1450 		complete(&tqspi->xfer_completion);
1451 		goto exit;
1452 	}
1453 
1454 	if (tqspi->cur_direction & DATA_DIR_RX)
1455 		tegra_qspi_read_rx_fifo_to_client_rxbuf(tqspi, t);
1456 
1457 	if (tqspi->cur_direction & DATA_DIR_TX)
1458 		tqspi->cur_pos = tqspi->cur_tx_pos;
1459 	else
1460 		tqspi->cur_pos = tqspi->cur_rx_pos;
1461 
1462 	if (tqspi->cur_pos == t->len) {
1463 		complete(&tqspi->xfer_completion);
1464 		goto exit;
1465 	}
1466 
1467 	tegra_qspi_calculate_curr_xfer_param(tqspi, t);
1468 	tegra_qspi_start_cpu_based_transfer(tqspi, t);
1469 exit:
1470 	tqspi->curr_xfer = NULL;
1471 	spin_unlock_irqrestore(&tqspi->lock, flags);
1472 	return IRQ_HANDLED;
1473 }
1474 
1475 static irqreturn_t handle_dma_based_xfer(struct tegra_qspi *tqspi)
1476 {
1477 	struct spi_transfer *t = tqspi->curr_xfer;
1478 	unsigned int total_fifo_words;
1479 	unsigned long flags;
1480 	long wait_status;
1481 	int num_errors = 0;
1482 
1483 	if (tqspi->cur_direction & DATA_DIR_TX) {
1484 		if (tqspi->tx_status) {
1485 			if (tqspi->tx_dma_chan)
1486 				dmaengine_terminate_all(tqspi->tx_dma_chan);
1487 			num_errors++;
1488 		} else if (tqspi->tx_dma_chan) {
1489 			wait_status = wait_for_completion_interruptible_timeout(
1490 				&tqspi->tx_dma_complete, QSPI_DMA_TIMEOUT);
1491 			if (wait_status <= 0) {
1492 				dmaengine_terminate_all(tqspi->tx_dma_chan);
1493 				dev_err(tqspi->dev, "failed TX DMA transfer\n");
1494 				num_errors++;
1495 			}
1496 		}
1497 	}
1498 
1499 	if (tqspi->cur_direction & DATA_DIR_RX) {
1500 		if (tqspi->rx_status) {
1501 			if (tqspi->rx_dma_chan)
1502 				dmaengine_terminate_all(tqspi->rx_dma_chan);
1503 			num_errors++;
1504 		} else if (tqspi->rx_dma_chan) {
1505 			wait_status = wait_for_completion_interruptible_timeout(
1506 				&tqspi->rx_dma_complete, QSPI_DMA_TIMEOUT);
1507 			if (wait_status <= 0) {
1508 				dmaengine_terminate_all(tqspi->rx_dma_chan);
1509 				dev_err(tqspi->dev, "failed RX DMA transfer\n");
1510 				num_errors++;
1511 			}
1512 		}
1513 	}
1514 
1515 	spin_lock_irqsave(&tqspi->lock, flags);
1516 
1517 	if (num_errors) {
1518 		tegra_qspi_dma_unmap_xfer(tqspi, t);
1519 		tegra_qspi_handle_error(tqspi);
1520 		complete(&tqspi->xfer_completion);
1521 		goto exit;
1522 	}
1523 
1524 	if (tqspi->cur_direction & DATA_DIR_RX)
1525 		tegra_qspi_copy_qspi_rxbuf_to_client_rxbuf(tqspi, t);
1526 
1527 	if (tqspi->cur_direction & DATA_DIR_TX)
1528 		tqspi->cur_pos = tqspi->cur_tx_pos;
1529 	else
1530 		tqspi->cur_pos = tqspi->cur_rx_pos;
1531 
1532 	if (tqspi->cur_pos == t->len) {
1533 		tegra_qspi_dma_unmap_xfer(tqspi, t);
1534 		complete(&tqspi->xfer_completion);
1535 		goto exit;
1536 	}
1537 
1538 	tegra_qspi_dma_unmap_xfer(tqspi, t);
1539 
1540 	/* continue transfer in current message */
1541 	total_fifo_words = tegra_qspi_calculate_curr_xfer_param(tqspi, t);
1542 	if (total_fifo_words > QSPI_FIFO_DEPTH)
1543 		num_errors = tegra_qspi_start_dma_based_transfer(tqspi, t);
1544 	else
1545 		num_errors = tegra_qspi_start_cpu_based_transfer(tqspi, t);
1546 
1547 exit:
1548 	spin_unlock_irqrestore(&tqspi->lock, flags);
1549 	return IRQ_HANDLED;
1550 }
1551 
1552 static irqreturn_t tegra_qspi_isr_thread(int irq, void *context_data)
1553 {
1554 	struct tegra_qspi *tqspi = context_data;
1555 
1556 	/*
1557 	 * Occasionally the IRQ thread takes a long time to wake up (usually
1558 	 * when the CPU that it's running on is excessively busy) and we have
1559 	 * already reached the timeout before and cleaned up the timed out
1560 	 * transfer. Avoid any processing in that case and bail out early.
1561 	 */
1562 	if (!tqspi->curr_xfer)
1563 		return IRQ_NONE;
1564 
1565 	tqspi->status_reg = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
1566 
1567 	if (tqspi->cur_direction & DATA_DIR_TX)
1568 		tqspi->tx_status = tqspi->status_reg & (QSPI_TX_FIFO_UNF | QSPI_TX_FIFO_OVF);
1569 
1570 	if (tqspi->cur_direction & DATA_DIR_RX)
1571 		tqspi->rx_status = tqspi->status_reg & (QSPI_RX_FIFO_OVF | QSPI_RX_FIFO_UNF);
1572 
1573 	tegra_qspi_mask_clear_irq(tqspi);
1574 
1575 	if (!tqspi->is_curr_dma_xfer)
1576 		return handle_cpu_based_xfer(tqspi);
1577 
1578 	return handle_dma_based_xfer(tqspi);
1579 }
1580 
1581 static struct tegra_qspi_soc_data tegra210_qspi_soc_data = {
1582 	.has_ext_dma = true,
1583 	.cmb_xfer_capable = false,
1584 	.supports_tpm = false,
1585 	.cs_count = 1,
1586 };
1587 
1588 static struct tegra_qspi_soc_data tegra186_qspi_soc_data = {
1589 	.has_ext_dma = true,
1590 	.cmb_xfer_capable = true,
1591 	.supports_tpm = false,
1592 	.cs_count = 1,
1593 };
1594 
1595 static struct tegra_qspi_soc_data tegra234_qspi_soc_data = {
1596 	.has_ext_dma = false,
1597 	.cmb_xfer_capable = true,
1598 	.supports_tpm = true,
1599 	.cs_count = 1,
1600 };
1601 
1602 static struct tegra_qspi_soc_data tegra241_qspi_soc_data = {
1603 	.has_ext_dma = true,
1604 	.cmb_xfer_capable = true,
1605 	.supports_tpm = true,
1606 	.cs_count = 4,
1607 };
1608 
1609 static const struct of_device_id tegra_qspi_of_match[] = {
1610 	{
1611 		.compatible = "nvidia,tegra210-qspi",
1612 		.data	    = &tegra210_qspi_soc_data,
1613 	}, {
1614 		.compatible = "nvidia,tegra186-qspi",
1615 		.data	    = &tegra186_qspi_soc_data,
1616 	}, {
1617 		.compatible = "nvidia,tegra194-qspi",
1618 		.data	    = &tegra186_qspi_soc_data,
1619 	}, {
1620 		.compatible = "nvidia,tegra234-qspi",
1621 		.data	    = &tegra234_qspi_soc_data,
1622 	}, {
1623 		.compatible = "nvidia,tegra241-qspi",
1624 		.data	    = &tegra241_qspi_soc_data,
1625 	},
1626 	{}
1627 };
1628 
1629 MODULE_DEVICE_TABLE(of, tegra_qspi_of_match);
1630 
1631 #ifdef CONFIG_ACPI
1632 static const struct acpi_device_id tegra_qspi_acpi_match[] = {
1633 	{
1634 		.id = "NVDA1213",
1635 		.driver_data = (kernel_ulong_t)&tegra210_qspi_soc_data,
1636 	}, {
1637 		.id = "NVDA1313",
1638 		.driver_data = (kernel_ulong_t)&tegra186_qspi_soc_data,
1639 	}, {
1640 		.id = "NVDA1413",
1641 		.driver_data = (kernel_ulong_t)&tegra234_qspi_soc_data,
1642 	}, {
1643 		.id = "NVDA1513",
1644 		.driver_data = (kernel_ulong_t)&tegra241_qspi_soc_data,
1645 	},
1646 	{}
1647 };
1648 
1649 MODULE_DEVICE_TABLE(acpi, tegra_qspi_acpi_match);
1650 #endif
1651 
1652 static int tegra_qspi_probe(struct platform_device *pdev)
1653 {
1654 	struct spi_controller	*host;
1655 	struct tegra_qspi	*tqspi;
1656 	struct resource		*r;
1657 	int ret, qspi_irq;
1658 	int bus_num;
1659 
1660 	host = devm_spi_alloc_host(&pdev->dev, sizeof(*tqspi));
1661 	if (!host)
1662 		return -ENOMEM;
1663 
1664 	platform_set_drvdata(pdev, host);
1665 	tqspi = spi_controller_get_devdata(host);
1666 
1667 	host->mode_bits = SPI_MODE_0 | SPI_MODE_3 | SPI_CS_HIGH |
1668 			  SPI_TX_DUAL | SPI_RX_DUAL | SPI_TX_QUAD | SPI_RX_QUAD;
1669 	host->bits_per_word_mask = SPI_BPW_MASK(32) | SPI_BPW_MASK(16) | SPI_BPW_MASK(8);
1670 	host->flags = SPI_CONTROLLER_HALF_DUPLEX;
1671 	host->setup = tegra_qspi_setup;
1672 	host->transfer_one_message = tegra_qspi_transfer_one_message;
1673 	host->num_chipselect = 1;
1674 	host->auto_runtime_pm = true;
1675 
1676 	bus_num = of_alias_get_id(pdev->dev.of_node, "spi");
1677 	if (bus_num >= 0)
1678 		host->bus_num = bus_num;
1679 
1680 	tqspi->host = host;
1681 	tqspi->dev = &pdev->dev;
1682 	spin_lock_init(&tqspi->lock);
1683 
1684 	tqspi->soc_data = device_get_match_data(&pdev->dev);
1685 	host->num_chipselect = tqspi->soc_data->cs_count;
1686 	tqspi->base = devm_platform_get_and_ioremap_resource(pdev, 0, &r);
1687 	if (IS_ERR(tqspi->base))
1688 		return PTR_ERR(tqspi->base);
1689 
1690 	tqspi->phys = r->start;
1691 	qspi_irq = platform_get_irq(pdev, 0);
1692 	if (qspi_irq < 0)
1693 		return qspi_irq;
1694 	tqspi->irq = qspi_irq;
1695 
1696 	if (!has_acpi_companion(tqspi->dev)) {
1697 		tqspi->clk = devm_clk_get(&pdev->dev, "qspi");
1698 		if (IS_ERR(tqspi->clk)) {
1699 			ret = PTR_ERR(tqspi->clk);
1700 			dev_err(&pdev->dev, "failed to get clock: %d\n", ret);
1701 			return ret;
1702 		}
1703 
1704 	}
1705 
1706 	tqspi->max_buf_size = QSPI_FIFO_DEPTH << 2;
1707 	tqspi->dma_buf_size = DEFAULT_QSPI_DMA_BUF_LEN;
1708 
1709 	ret = tegra_qspi_init_dma(tqspi);
1710 	if (ret < 0)
1711 		return ret;
1712 
1713 	if (tqspi->use_dma)
1714 		tqspi->max_buf_size = tqspi->dma_buf_size;
1715 
1716 	init_completion(&tqspi->tx_dma_complete);
1717 	init_completion(&tqspi->rx_dma_complete);
1718 	init_completion(&tqspi->xfer_completion);
1719 
1720 	pm_runtime_enable(&pdev->dev);
1721 	ret = pm_runtime_resume_and_get(&pdev->dev);
1722 	if (ret < 0) {
1723 		dev_err(&pdev->dev, "failed to get runtime PM: %d\n", ret);
1724 		goto exit_pm_disable;
1725 	}
1726 
1727 	if (device_reset(tqspi->dev) < 0)
1728 		dev_warn_once(tqspi->dev, "device reset failed\n");
1729 
1730 	tqspi->def_command1_reg = QSPI_M_S | QSPI_CS_SW_HW |  QSPI_CS_SW_VAL;
1731 	tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
1732 	tqspi->spi_cs_timing1 = tegra_qspi_readl(tqspi, QSPI_CS_TIMING1);
1733 	tqspi->spi_cs_timing2 = tegra_qspi_readl(tqspi, QSPI_CS_TIMING2);
1734 	tqspi->def_command2_reg = tegra_qspi_readl(tqspi, QSPI_COMMAND2);
1735 
1736 	pm_runtime_put(&pdev->dev);
1737 
1738 	ret = request_threaded_irq(tqspi->irq, NULL,
1739 				   tegra_qspi_isr_thread, IRQF_ONESHOT,
1740 				   dev_name(&pdev->dev), tqspi);
1741 	if (ret < 0) {
1742 		dev_err(&pdev->dev, "failed to request IRQ#%u: %d\n", tqspi->irq, ret);
1743 		goto exit_pm_disable;
1744 	}
1745 
1746 	host->dev.of_node = pdev->dev.of_node;
1747 	ret = spi_register_controller(host);
1748 	if (ret < 0) {
1749 		dev_err(&pdev->dev, "failed to register host: %d\n", ret);
1750 		goto exit_free_irq;
1751 	}
1752 
1753 	return 0;
1754 
1755 exit_free_irq:
1756 	free_irq(qspi_irq, tqspi);
1757 exit_pm_disable:
1758 	pm_runtime_force_suspend(&pdev->dev);
1759 	tegra_qspi_deinit_dma(tqspi);
1760 	return ret;
1761 }
1762 
1763 static void tegra_qspi_remove(struct platform_device *pdev)
1764 {
1765 	struct spi_controller *host = platform_get_drvdata(pdev);
1766 	struct tegra_qspi *tqspi = spi_controller_get_devdata(host);
1767 
1768 	spi_unregister_controller(host);
1769 	free_irq(tqspi->irq, tqspi);
1770 	pm_runtime_force_suspend(&pdev->dev);
1771 	tegra_qspi_deinit_dma(tqspi);
1772 }
1773 
1774 static int __maybe_unused tegra_qspi_suspend(struct device *dev)
1775 {
1776 	struct spi_controller *host = dev_get_drvdata(dev);
1777 
1778 	return spi_controller_suspend(host);
1779 }
1780 
1781 static int __maybe_unused tegra_qspi_resume(struct device *dev)
1782 {
1783 	struct spi_controller *host = dev_get_drvdata(dev);
1784 	struct tegra_qspi *tqspi = spi_controller_get_devdata(host);
1785 	int ret;
1786 
1787 	ret = pm_runtime_resume_and_get(dev);
1788 	if (ret < 0) {
1789 		dev_err(dev, "failed to get runtime PM: %d\n", ret);
1790 		return ret;
1791 	}
1792 
1793 	tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
1794 	tegra_qspi_writel(tqspi, tqspi->def_command2_reg, QSPI_COMMAND2);
1795 	pm_runtime_put(dev);
1796 
1797 	return spi_controller_resume(host);
1798 }
1799 
1800 static int __maybe_unused tegra_qspi_runtime_suspend(struct device *dev)
1801 {
1802 	struct spi_controller *host = dev_get_drvdata(dev);
1803 	struct tegra_qspi *tqspi = spi_controller_get_devdata(host);
1804 
1805 	/* Runtime pm disabled with ACPI */
1806 	if (has_acpi_companion(tqspi->dev))
1807 		return 0;
1808 	/* flush all write which are in PPSB queue by reading back */
1809 	tegra_qspi_readl(tqspi, QSPI_COMMAND1);
1810 
1811 	clk_disable_unprepare(tqspi->clk);
1812 
1813 	return 0;
1814 }
1815 
1816 static int __maybe_unused tegra_qspi_runtime_resume(struct device *dev)
1817 {
1818 	struct spi_controller *host = dev_get_drvdata(dev);
1819 	struct tegra_qspi *tqspi = spi_controller_get_devdata(host);
1820 	int ret;
1821 
1822 	/* Runtime pm disabled with ACPI */
1823 	if (has_acpi_companion(tqspi->dev))
1824 		return 0;
1825 	ret = clk_prepare_enable(tqspi->clk);
1826 	if (ret < 0)
1827 		dev_err(tqspi->dev, "failed to enable clock: %d\n", ret);
1828 
1829 	return ret;
1830 }
1831 
1832 static const struct dev_pm_ops tegra_qspi_pm_ops = {
1833 	SET_RUNTIME_PM_OPS(tegra_qspi_runtime_suspend, tegra_qspi_runtime_resume, NULL)
1834 	SET_SYSTEM_SLEEP_PM_OPS(tegra_qspi_suspend, tegra_qspi_resume)
1835 };
1836 
1837 static struct platform_driver tegra_qspi_driver = {
1838 	.driver = {
1839 		.name		= "tegra-qspi",
1840 		.pm		= &tegra_qspi_pm_ops,
1841 		.of_match_table	= tegra_qspi_of_match,
1842 		.acpi_match_table = ACPI_PTR(tegra_qspi_acpi_match),
1843 	},
1844 	.probe =	tegra_qspi_probe,
1845 	.remove =	tegra_qspi_remove,
1846 };
1847 module_platform_driver(tegra_qspi_driver);
1848 
1849 MODULE_ALIAS("platform:qspi-tegra");
1850 MODULE_DESCRIPTION("NVIDIA Tegra QSPI Controller Driver");
1851 MODULE_AUTHOR("Sowjanya Komatineni <skomatineni@nvidia.com>");
1852 MODULE_LICENSE("GPL v2");
1853