Loading...
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
164struct tegra_qspi_soc_data {
165 bool has_dma;
166 bool cmb_xfer_capable;
167 unsigned int cs_count;
168};
169
170struct tegra_qspi_client_data {
171 int tx_clk_tap_delay;
172 int rx_clk_tap_delay;
173};
174
175struct 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
232static inline u32 tegra_qspi_readl(struct tegra_qspi *tqspi, unsigned long offset)
233{
234 return readl(tqspi->base + offset);
235}
236
237static 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
246static 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
266static unsigned int
267tegra_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
305static unsigned int
306tegra_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
354static unsigned int
355tegra_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
398static void
399tegra_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
441static void
442tegra_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
477static void tegra_qspi_dma_complete(void *args)
478{
479 struct completion *dma_complete = args;
480
481 complete(dma_complete);
482}
483
484static 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
512static 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
540static 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
562static 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
571static 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
596static 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
606static 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
698static 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
721static 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
749static 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
796err_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
808static 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
869static 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
924static 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
941static 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
979static 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
996static 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
1005static 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
1018static 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
1034static 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
1053static 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 break;
1160 default:
1161 ret = -EINVAL;
1162 goto exit;
1163 }
1164 msg->actual_length += xfer->len;
1165 transfer_phase++;
1166 }
1167 if (!xfer->cs_change) {
1168 tegra_qspi_transfer_end(spi);
1169 spi_transfer_delay_exec(xfer);
1170 }
1171 ret = 0;
1172
1173exit:
1174 msg->status = ret;
1175
1176 return ret;
1177}
1178
1179static int tegra_qspi_non_combined_seq_xfer(struct tegra_qspi *tqspi,
1180 struct spi_message *msg)
1181{
1182 struct spi_device *spi = msg->spi;
1183 struct spi_transfer *transfer;
1184 bool is_first_msg = true;
1185 int ret = 0, val = 0;
1186
1187 msg->status = 0;
1188 msg->actual_length = 0;
1189 tqspi->tx_status = 0;
1190 tqspi->rx_status = 0;
1191
1192 /* Disable Combined sequence mode */
1193 val = tegra_qspi_readl(tqspi, QSPI_GLOBAL_CONFIG);
1194 val &= ~QSPI_CMB_SEQ_EN;
1195 tegra_qspi_writel(tqspi, val, QSPI_GLOBAL_CONFIG);
1196 list_for_each_entry(transfer, &msg->transfers, transfer_list) {
1197 struct spi_transfer *xfer = transfer;
1198 u8 dummy_bytes = 0;
1199 u32 cmd1;
1200
1201 tqspi->dummy_cycles = 0;
1202 /*
1203 * Tegra QSPI hardware supports dummy bytes transfer after actual transfer
1204 * bytes based on programmed dummy clock cycles in the QSPI_MISC register.
1205 * So, check if the next transfer is dummy data transfer and program dummy
1206 * clock cycles along with the current transfer and skip next transfer.
1207 */
1208 if (!list_is_last(&xfer->transfer_list, &msg->transfers)) {
1209 struct spi_transfer *next_xfer;
1210
1211 next_xfer = list_next_entry(xfer, transfer_list);
1212 if (next_xfer->dummy_data) {
1213 u32 dummy_cycles = next_xfer->len * 8 / next_xfer->tx_nbits;
1214
1215 if (dummy_cycles <= QSPI_DUMMY_CYCLES_MAX) {
1216 tqspi->dummy_cycles = dummy_cycles;
1217 dummy_bytes = next_xfer->len;
1218 transfer = next_xfer;
1219 }
1220 }
1221 }
1222
1223 reinit_completion(&tqspi->xfer_completion);
1224
1225 cmd1 = tegra_qspi_setup_transfer_one(spi, xfer, is_first_msg);
1226
1227 ret = tegra_qspi_start_transfer_one(spi, xfer, cmd1);
1228 if (ret < 0) {
1229 dev_err(tqspi->dev, "failed to start transfer: %d\n", ret);
1230 goto complete_xfer;
1231 }
1232
1233 ret = wait_for_completion_timeout(&tqspi->xfer_completion,
1234 QSPI_DMA_TIMEOUT);
1235 if (WARN_ON(ret == 0)) {
1236 dev_err(tqspi->dev, "transfer timeout\n");
1237 if (tqspi->is_curr_dma_xfer && (tqspi->cur_direction & DATA_DIR_TX))
1238 dmaengine_terminate_all(tqspi->tx_dma_chan);
1239 if (tqspi->is_curr_dma_xfer && (tqspi->cur_direction & DATA_DIR_RX))
1240 dmaengine_terminate_all(tqspi->rx_dma_chan);
1241 tegra_qspi_handle_error(tqspi);
1242 ret = -EIO;
1243 goto complete_xfer;
1244 }
1245
1246 if (tqspi->tx_status || tqspi->rx_status) {
1247 tegra_qspi_handle_error(tqspi);
1248 ret = -EIO;
1249 goto complete_xfer;
1250 }
1251
1252 msg->actual_length += xfer->len + dummy_bytes;
1253
1254complete_xfer:
1255 if (ret < 0) {
1256 tegra_qspi_transfer_end(spi);
1257 spi_transfer_delay_exec(xfer);
1258 goto exit;
1259 }
1260
1261 if (list_is_last(&xfer->transfer_list, &msg->transfers)) {
1262 /* de-activate CS after last transfer only when cs_change is not set */
1263 if (!xfer->cs_change) {
1264 tegra_qspi_transfer_end(spi);
1265 spi_transfer_delay_exec(xfer);
1266 }
1267 } else if (xfer->cs_change) {
1268 /* de-activated CS between the transfers only when cs_change is set */
1269 tegra_qspi_transfer_end(spi);
1270 spi_transfer_delay_exec(xfer);
1271 }
1272 }
1273
1274 ret = 0;
1275exit:
1276 msg->status = ret;
1277
1278 return ret;
1279}
1280
1281static bool tegra_qspi_validate_cmb_seq(struct tegra_qspi *tqspi,
1282 struct spi_message *msg)
1283{
1284 int transfer_count = 0;
1285 struct spi_transfer *xfer;
1286
1287 list_for_each_entry(xfer, &msg->transfers, transfer_list) {
1288 transfer_count++;
1289 }
1290 if (!tqspi->soc_data->cmb_xfer_capable || transfer_count != 3)
1291 return false;
1292 xfer = list_first_entry(&msg->transfers, typeof(*xfer),
1293 transfer_list);
1294 if (xfer->len > 2)
1295 return false;
1296 xfer = list_next_entry(xfer, transfer_list);
1297 if (xfer->len > 4 || xfer->len < 3)
1298 return false;
1299 xfer = list_next_entry(xfer, transfer_list);
1300 if (!tqspi->soc_data->has_dma || xfer->len > (QSPI_FIFO_DEPTH << 2))
1301 return false;
1302
1303 return true;
1304}
1305
1306static int tegra_qspi_transfer_one_message(struct spi_master *master,
1307 struct spi_message *msg)
1308{
1309 struct tegra_qspi *tqspi = spi_master_get_devdata(master);
1310 int ret;
1311
1312 if (tegra_qspi_validate_cmb_seq(tqspi, msg))
1313 ret = tegra_qspi_combined_seq_xfer(tqspi, msg);
1314 else
1315 ret = tegra_qspi_non_combined_seq_xfer(tqspi, msg);
1316
1317 spi_finalize_current_message(master);
1318
1319 return ret;
1320}
1321
1322static irqreturn_t handle_cpu_based_xfer(struct tegra_qspi *tqspi)
1323{
1324 struct spi_transfer *t = tqspi->curr_xfer;
1325 unsigned long flags;
1326
1327 spin_lock_irqsave(&tqspi->lock, flags);
1328
1329 if (tqspi->tx_status || tqspi->rx_status) {
1330 tegra_qspi_handle_error(tqspi);
1331 complete(&tqspi->xfer_completion);
1332 goto exit;
1333 }
1334
1335 if (tqspi->cur_direction & DATA_DIR_RX)
1336 tegra_qspi_read_rx_fifo_to_client_rxbuf(tqspi, t);
1337
1338 if (tqspi->cur_direction & DATA_DIR_TX)
1339 tqspi->cur_pos = tqspi->cur_tx_pos;
1340 else
1341 tqspi->cur_pos = tqspi->cur_rx_pos;
1342
1343 if (tqspi->cur_pos == t->len) {
1344 complete(&tqspi->xfer_completion);
1345 goto exit;
1346 }
1347
1348 tegra_qspi_calculate_curr_xfer_param(tqspi, t);
1349 tegra_qspi_start_cpu_based_transfer(tqspi, t);
1350exit:
1351 spin_unlock_irqrestore(&tqspi->lock, flags);
1352 return IRQ_HANDLED;
1353}
1354
1355static irqreturn_t handle_dma_based_xfer(struct tegra_qspi *tqspi)
1356{
1357 struct spi_transfer *t = tqspi->curr_xfer;
1358 unsigned int total_fifo_words;
1359 unsigned long flags;
1360 long wait_status;
1361 int err = 0;
1362
1363 if (tqspi->cur_direction & DATA_DIR_TX) {
1364 if (tqspi->tx_status) {
1365 dmaengine_terminate_all(tqspi->tx_dma_chan);
1366 err += 1;
1367 } else {
1368 wait_status = wait_for_completion_interruptible_timeout(
1369 &tqspi->tx_dma_complete, QSPI_DMA_TIMEOUT);
1370 if (wait_status <= 0) {
1371 dmaengine_terminate_all(tqspi->tx_dma_chan);
1372 dev_err(tqspi->dev, "failed TX DMA transfer\n");
1373 err += 1;
1374 }
1375 }
1376 }
1377
1378 if (tqspi->cur_direction & DATA_DIR_RX) {
1379 if (tqspi->rx_status) {
1380 dmaengine_terminate_all(tqspi->rx_dma_chan);
1381 err += 2;
1382 } else {
1383 wait_status = wait_for_completion_interruptible_timeout(
1384 &tqspi->rx_dma_complete, QSPI_DMA_TIMEOUT);
1385 if (wait_status <= 0) {
1386 dmaengine_terminate_all(tqspi->rx_dma_chan);
1387 dev_err(tqspi->dev, "failed RX DMA transfer\n");
1388 err += 2;
1389 }
1390 }
1391 }
1392
1393 spin_lock_irqsave(&tqspi->lock, flags);
1394
1395 if (err) {
1396 tegra_qspi_dma_unmap_xfer(tqspi, t);
1397 tegra_qspi_handle_error(tqspi);
1398 complete(&tqspi->xfer_completion);
1399 goto exit;
1400 }
1401
1402 if (tqspi->cur_direction & DATA_DIR_RX)
1403 tegra_qspi_copy_qspi_rxbuf_to_client_rxbuf(tqspi, t);
1404
1405 if (tqspi->cur_direction & DATA_DIR_TX)
1406 tqspi->cur_pos = tqspi->cur_tx_pos;
1407 else
1408 tqspi->cur_pos = tqspi->cur_rx_pos;
1409
1410 if (tqspi->cur_pos == t->len) {
1411 tegra_qspi_dma_unmap_xfer(tqspi, t);
1412 complete(&tqspi->xfer_completion);
1413 goto exit;
1414 }
1415
1416 tegra_qspi_dma_unmap_xfer(tqspi, t);
1417
1418 /* continue transfer in current message */
1419 total_fifo_words = tegra_qspi_calculate_curr_xfer_param(tqspi, t);
1420 if (total_fifo_words > QSPI_FIFO_DEPTH)
1421 err = tegra_qspi_start_dma_based_transfer(tqspi, t);
1422 else
1423 err = tegra_qspi_start_cpu_based_transfer(tqspi, t);
1424
1425exit:
1426 spin_unlock_irqrestore(&tqspi->lock, flags);
1427 return IRQ_HANDLED;
1428}
1429
1430static irqreturn_t tegra_qspi_isr_thread(int irq, void *context_data)
1431{
1432 struct tegra_qspi *tqspi = context_data;
1433
1434 tqspi->status_reg = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
1435
1436 if (tqspi->cur_direction & DATA_DIR_TX)
1437 tqspi->tx_status = tqspi->status_reg & (QSPI_TX_FIFO_UNF | QSPI_TX_FIFO_OVF);
1438
1439 if (tqspi->cur_direction & DATA_DIR_RX)
1440 tqspi->rx_status = tqspi->status_reg & (QSPI_RX_FIFO_OVF | QSPI_RX_FIFO_UNF);
1441
1442 tegra_qspi_mask_clear_irq(tqspi);
1443
1444 if (!tqspi->is_curr_dma_xfer)
1445 return handle_cpu_based_xfer(tqspi);
1446
1447 return handle_dma_based_xfer(tqspi);
1448}
1449
1450static struct tegra_qspi_soc_data tegra210_qspi_soc_data = {
1451 .has_dma = true,
1452 .cmb_xfer_capable = false,
1453 .cs_count = 1,
1454};
1455
1456static struct tegra_qspi_soc_data tegra186_qspi_soc_data = {
1457 .has_dma = true,
1458 .cmb_xfer_capable = true,
1459 .cs_count = 1,
1460};
1461
1462static struct tegra_qspi_soc_data tegra234_qspi_soc_data = {
1463 .has_dma = false,
1464 .cmb_xfer_capable = true,
1465 .cs_count = 1,
1466};
1467
1468static struct tegra_qspi_soc_data tegra241_qspi_soc_data = {
1469 .has_dma = false,
1470 .cmb_xfer_capable = true,
1471 .cs_count = 4,
1472};
1473
1474static const struct of_device_id tegra_qspi_of_match[] = {
1475 {
1476 .compatible = "nvidia,tegra210-qspi",
1477 .data = &tegra210_qspi_soc_data,
1478 }, {
1479 .compatible = "nvidia,tegra186-qspi",
1480 .data = &tegra186_qspi_soc_data,
1481 }, {
1482 .compatible = "nvidia,tegra194-qspi",
1483 .data = &tegra186_qspi_soc_data,
1484 }, {
1485 .compatible = "nvidia,tegra234-qspi",
1486 .data = &tegra234_qspi_soc_data,
1487 }, {
1488 .compatible = "nvidia,tegra241-qspi",
1489 .data = &tegra241_qspi_soc_data,
1490 },
1491 {}
1492};
1493
1494MODULE_DEVICE_TABLE(of, tegra_qspi_of_match);
1495
1496#ifdef CONFIG_ACPI
1497static const struct acpi_device_id tegra_qspi_acpi_match[] = {
1498 {
1499 .id = "NVDA1213",
1500 .driver_data = (kernel_ulong_t)&tegra210_qspi_soc_data,
1501 }, {
1502 .id = "NVDA1313",
1503 .driver_data = (kernel_ulong_t)&tegra186_qspi_soc_data,
1504 }, {
1505 .id = "NVDA1413",
1506 .driver_data = (kernel_ulong_t)&tegra234_qspi_soc_data,
1507 }, {
1508 .id = "NVDA1513",
1509 .driver_data = (kernel_ulong_t)&tegra241_qspi_soc_data,
1510 },
1511 {}
1512};
1513
1514MODULE_DEVICE_TABLE(acpi, tegra_qspi_acpi_match);
1515#endif
1516
1517static int tegra_qspi_probe(struct platform_device *pdev)
1518{
1519 struct spi_master *master;
1520 struct tegra_qspi *tqspi;
1521 struct resource *r;
1522 int ret, qspi_irq;
1523 int bus_num;
1524
1525 master = devm_spi_alloc_master(&pdev->dev, sizeof(*tqspi));
1526 if (!master)
1527 return -ENOMEM;
1528
1529 platform_set_drvdata(pdev, master);
1530 tqspi = spi_master_get_devdata(master);
1531
1532 master->mode_bits = SPI_MODE_0 | SPI_MODE_3 | SPI_CS_HIGH |
1533 SPI_TX_DUAL | SPI_RX_DUAL | SPI_TX_QUAD | SPI_RX_QUAD;
1534 master->bits_per_word_mask = SPI_BPW_MASK(32) | SPI_BPW_MASK(16) | SPI_BPW_MASK(8);
1535 master->setup = tegra_qspi_setup;
1536 master->transfer_one_message = tegra_qspi_transfer_one_message;
1537 master->num_chipselect = 1;
1538 master->auto_runtime_pm = true;
1539
1540 bus_num = of_alias_get_id(pdev->dev.of_node, "spi");
1541 if (bus_num >= 0)
1542 master->bus_num = bus_num;
1543
1544 tqspi->master = master;
1545 tqspi->dev = &pdev->dev;
1546 spin_lock_init(&tqspi->lock);
1547
1548 tqspi->soc_data = device_get_match_data(&pdev->dev);
1549 master->num_chipselect = tqspi->soc_data->cs_count;
1550 r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
1551 tqspi->base = devm_ioremap_resource(&pdev->dev, r);
1552 if (IS_ERR(tqspi->base))
1553 return PTR_ERR(tqspi->base);
1554
1555 tqspi->phys = r->start;
1556 qspi_irq = platform_get_irq(pdev, 0);
1557 if (qspi_irq < 0)
1558 return qspi_irq;
1559 tqspi->irq = qspi_irq;
1560
1561 if (!has_acpi_companion(tqspi->dev)) {
1562 tqspi->clk = devm_clk_get(&pdev->dev, "qspi");
1563 if (IS_ERR(tqspi->clk)) {
1564 ret = PTR_ERR(tqspi->clk);
1565 dev_err(&pdev->dev, "failed to get clock: %d\n", ret);
1566 return ret;
1567 }
1568
1569 }
1570
1571 tqspi->max_buf_size = QSPI_FIFO_DEPTH << 2;
1572 tqspi->dma_buf_size = DEFAULT_QSPI_DMA_BUF_LEN;
1573
1574 ret = tegra_qspi_init_dma(tqspi);
1575 if (ret < 0)
1576 return ret;
1577
1578 if (tqspi->use_dma)
1579 tqspi->max_buf_size = tqspi->dma_buf_size;
1580
1581 init_completion(&tqspi->tx_dma_complete);
1582 init_completion(&tqspi->rx_dma_complete);
1583 init_completion(&tqspi->xfer_completion);
1584
1585 pm_runtime_enable(&pdev->dev);
1586 ret = pm_runtime_resume_and_get(&pdev->dev);
1587 if (ret < 0) {
1588 dev_err(&pdev->dev, "failed to get runtime PM: %d\n", ret);
1589 goto exit_pm_disable;
1590 }
1591
1592 if (device_reset(tqspi->dev) < 0)
1593 dev_warn_once(tqspi->dev, "device reset failed\n");
1594
1595 tqspi->def_command1_reg = QSPI_M_S | QSPI_CS_SW_HW | QSPI_CS_SW_VAL;
1596 tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
1597 tqspi->spi_cs_timing1 = tegra_qspi_readl(tqspi, QSPI_CS_TIMING1);
1598 tqspi->spi_cs_timing2 = tegra_qspi_readl(tqspi, QSPI_CS_TIMING2);
1599 tqspi->def_command2_reg = tegra_qspi_readl(tqspi, QSPI_COMMAND2);
1600
1601 pm_runtime_put(&pdev->dev);
1602
1603 ret = request_threaded_irq(tqspi->irq, NULL,
1604 tegra_qspi_isr_thread, IRQF_ONESHOT,
1605 dev_name(&pdev->dev), tqspi);
1606 if (ret < 0) {
1607 dev_err(&pdev->dev, "failed to request IRQ#%u: %d\n", tqspi->irq, ret);
1608 goto exit_pm_disable;
1609 }
1610
1611 master->dev.of_node = pdev->dev.of_node;
1612 ret = spi_register_master(master);
1613 if (ret < 0) {
1614 dev_err(&pdev->dev, "failed to register master: %d\n", ret);
1615 goto exit_free_irq;
1616 }
1617
1618 return 0;
1619
1620exit_free_irq:
1621 free_irq(qspi_irq, tqspi);
1622exit_pm_disable:
1623 pm_runtime_force_suspend(&pdev->dev);
1624 tegra_qspi_deinit_dma(tqspi);
1625 return ret;
1626}
1627
1628static int tegra_qspi_remove(struct platform_device *pdev)
1629{
1630 struct spi_master *master = platform_get_drvdata(pdev);
1631 struct tegra_qspi *tqspi = spi_master_get_devdata(master);
1632
1633 spi_unregister_master(master);
1634 free_irq(tqspi->irq, tqspi);
1635 pm_runtime_force_suspend(&pdev->dev);
1636 tegra_qspi_deinit_dma(tqspi);
1637
1638 return 0;
1639}
1640
1641static int __maybe_unused tegra_qspi_suspend(struct device *dev)
1642{
1643 struct spi_master *master = dev_get_drvdata(dev);
1644
1645 return spi_master_suspend(master);
1646}
1647
1648static int __maybe_unused tegra_qspi_resume(struct device *dev)
1649{
1650 struct spi_master *master = dev_get_drvdata(dev);
1651 struct tegra_qspi *tqspi = spi_master_get_devdata(master);
1652 int ret;
1653
1654 ret = pm_runtime_resume_and_get(dev);
1655 if (ret < 0) {
1656 dev_err(dev, "failed to get runtime PM: %d\n", ret);
1657 return ret;
1658 }
1659
1660 tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
1661 tegra_qspi_writel(tqspi, tqspi->def_command2_reg, QSPI_COMMAND2);
1662 pm_runtime_put(dev);
1663
1664 return spi_master_resume(master);
1665}
1666
1667static int __maybe_unused tegra_qspi_runtime_suspend(struct device *dev)
1668{
1669 struct spi_master *master = dev_get_drvdata(dev);
1670 struct tegra_qspi *tqspi = spi_master_get_devdata(master);
1671
1672 /* Runtime pm disabled with ACPI */
1673 if (has_acpi_companion(tqspi->dev))
1674 return 0;
1675 /* flush all write which are in PPSB queue by reading back */
1676 tegra_qspi_readl(tqspi, QSPI_COMMAND1);
1677
1678 clk_disable_unprepare(tqspi->clk);
1679
1680 return 0;
1681}
1682
1683static int __maybe_unused tegra_qspi_runtime_resume(struct device *dev)
1684{
1685 struct spi_master *master = dev_get_drvdata(dev);
1686 struct tegra_qspi *tqspi = spi_master_get_devdata(master);
1687 int ret;
1688
1689 /* Runtime pm disabled with ACPI */
1690 if (has_acpi_companion(tqspi->dev))
1691 return 0;
1692 ret = clk_prepare_enable(tqspi->clk);
1693 if (ret < 0)
1694 dev_err(tqspi->dev, "failed to enable clock: %d\n", ret);
1695
1696 return ret;
1697}
1698
1699static const struct dev_pm_ops tegra_qspi_pm_ops = {
1700 SET_RUNTIME_PM_OPS(tegra_qspi_runtime_suspend, tegra_qspi_runtime_resume, NULL)
1701 SET_SYSTEM_SLEEP_PM_OPS(tegra_qspi_suspend, tegra_qspi_resume)
1702};
1703
1704static struct platform_driver tegra_qspi_driver = {
1705 .driver = {
1706 .name = "tegra-qspi",
1707 .pm = &tegra_qspi_pm_ops,
1708 .of_match_table = tegra_qspi_of_match,
1709 .acpi_match_table = ACPI_PTR(tegra_qspi_acpi_match),
1710 },
1711 .probe = tegra_qspi_probe,
1712 .remove = tegra_qspi_remove,
1713};
1714module_platform_driver(tegra_qspi_driver);
1715
1716MODULE_ALIAS("platform:qspi-tegra");
1717MODULE_DESCRIPTION("NVIDIA Tegra QSPI Controller Driver");
1718MODULE_AUTHOR("Sowjanya Komatineni <skomatineni@nvidia.com>");
1719MODULE_LICENSE("GPL v2");
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
26#define QSPI_COMMAND1 0x000
27#define QSPI_BIT_LENGTH(x) (((x) & 0x1f) << 0)
28#define QSPI_PACKED BIT(5)
29#define QSPI_INTERFACE_WIDTH_MASK (0x03 << 7)
30#define QSPI_INTERFACE_WIDTH(x) (((x) & 0x03) << 7)
31#define QSPI_INTERFACE_WIDTH_SINGLE QSPI_INTERFACE_WIDTH(0)
32#define QSPI_INTERFACE_WIDTH_DUAL QSPI_INTERFACE_WIDTH(1)
33#define QSPI_INTERFACE_WIDTH_QUAD QSPI_INTERFACE_WIDTH(2)
34#define QSPI_SDR_DDR_SEL BIT(9)
35#define QSPI_TX_EN BIT(11)
36#define QSPI_RX_EN BIT(12)
37#define QSPI_CS_SW_VAL BIT(20)
38#define QSPI_CS_SW_HW BIT(21)
39
40#define QSPI_CS_POL_INACTIVE(n) (1 << (22 + (n)))
41#define QSPI_CS_POL_INACTIVE_MASK (0xF << 22)
42#define QSPI_CS_SEL_0 (0 << 26)
43#define QSPI_CS_SEL_1 (1 << 26)
44#define QSPI_CS_SEL_2 (2 << 26)
45#define QSPI_CS_SEL_3 (3 << 26)
46#define QSPI_CS_SEL_MASK (3 << 26)
47#define QSPI_CS_SEL(x) (((x) & 0x3) << 26)
48
49#define QSPI_CONTROL_MODE_0 (0 << 28)
50#define QSPI_CONTROL_MODE_3 (3 << 28)
51#define QSPI_CONTROL_MODE_MASK (3 << 28)
52#define QSPI_M_S BIT(30)
53#define QSPI_PIO BIT(31)
54
55#define QSPI_COMMAND2 0x004
56#define QSPI_TX_TAP_DELAY(x) (((x) & 0x3f) << 10)
57#define QSPI_RX_TAP_DELAY(x) (((x) & 0xff) << 0)
58
59#define QSPI_CS_TIMING1 0x008
60#define QSPI_SETUP_HOLD(setup, hold) (((setup) << 4) | (hold))
61
62#define QSPI_CS_TIMING2 0x00c
63#define CYCLES_BETWEEN_PACKETS_0(x) (((x) & 0x1f) << 0)
64#define CS_ACTIVE_BETWEEN_PACKETS_0 BIT(5)
65
66#define QSPI_TRANS_STATUS 0x010
67#define QSPI_BLK_CNT(val) (((val) >> 0) & 0xffff)
68#define QSPI_RDY BIT(30)
69
70#define QSPI_FIFO_STATUS 0x014
71#define QSPI_RX_FIFO_EMPTY BIT(0)
72#define QSPI_RX_FIFO_FULL BIT(1)
73#define QSPI_TX_FIFO_EMPTY BIT(2)
74#define QSPI_TX_FIFO_FULL BIT(3)
75#define QSPI_RX_FIFO_UNF BIT(4)
76#define QSPI_RX_FIFO_OVF BIT(5)
77#define QSPI_TX_FIFO_UNF BIT(6)
78#define QSPI_TX_FIFO_OVF BIT(7)
79#define QSPI_ERR BIT(8)
80#define QSPI_TX_FIFO_FLUSH BIT(14)
81#define QSPI_RX_FIFO_FLUSH BIT(15)
82#define QSPI_TX_FIFO_EMPTY_COUNT(val) (((val) >> 16) & 0x7f)
83#define QSPI_RX_FIFO_FULL_COUNT(val) (((val) >> 23) & 0x7f)
84
85#define QSPI_FIFO_ERROR (QSPI_RX_FIFO_UNF | \
86 QSPI_RX_FIFO_OVF | \
87 QSPI_TX_FIFO_UNF | \
88 QSPI_TX_FIFO_OVF)
89#define QSPI_FIFO_EMPTY (QSPI_RX_FIFO_EMPTY | \
90 QSPI_TX_FIFO_EMPTY)
91
92#define QSPI_TX_DATA 0x018
93#define QSPI_RX_DATA 0x01c
94
95#define QSPI_DMA_CTL 0x020
96#define QSPI_TX_TRIG(n) (((n) & 0x3) << 15)
97#define QSPI_TX_TRIG_1 QSPI_TX_TRIG(0)
98#define QSPI_TX_TRIG_4 QSPI_TX_TRIG(1)
99#define QSPI_TX_TRIG_8 QSPI_TX_TRIG(2)
100#define QSPI_TX_TRIG_16 QSPI_TX_TRIG(3)
101
102#define QSPI_RX_TRIG(n) (((n) & 0x3) << 19)
103#define QSPI_RX_TRIG_1 QSPI_RX_TRIG(0)
104#define QSPI_RX_TRIG_4 QSPI_RX_TRIG(1)
105#define QSPI_RX_TRIG_8 QSPI_RX_TRIG(2)
106#define QSPI_RX_TRIG_16 QSPI_RX_TRIG(3)
107
108#define QSPI_DMA_EN BIT(31)
109
110#define QSPI_DMA_BLK 0x024
111#define QSPI_DMA_BLK_SET(x) (((x) & 0xffff) << 0)
112
113#define QSPI_TX_FIFO 0x108
114#define QSPI_RX_FIFO 0x188
115
116#define QSPI_FIFO_DEPTH 64
117
118#define QSPI_INTR_MASK 0x18c
119#define QSPI_INTR_RX_FIFO_UNF_MASK BIT(25)
120#define QSPI_INTR_RX_FIFO_OVF_MASK BIT(26)
121#define QSPI_INTR_TX_FIFO_UNF_MASK BIT(27)
122#define QSPI_INTR_TX_FIFO_OVF_MASK BIT(28)
123#define QSPI_INTR_RDY_MASK BIT(29)
124#define QSPI_INTR_RX_TX_FIFO_ERR (QSPI_INTR_RX_FIFO_UNF_MASK | \
125 QSPI_INTR_RX_FIFO_OVF_MASK | \
126 QSPI_INTR_TX_FIFO_UNF_MASK | \
127 QSPI_INTR_TX_FIFO_OVF_MASK)
128
129#define QSPI_MISC_REG 0x194
130#define QSPI_NUM_DUMMY_CYCLE(x) (((x) & 0xff) << 0)
131#define QSPI_DUMMY_CYCLES_MAX 0xff
132
133#define QSPI_CMB_SEQ_CMD 0x19c
134#define QSPI_COMMAND_VALUE_SET(X) (((x) & 0xFF) << 0)
135
136#define QSPI_CMB_SEQ_CMD_CFG 0x1a0
137#define QSPI_COMMAND_X1_X2_X4(x) (((x) & 0x3) << 13)
138#define QSPI_COMMAND_X1_X2_X4_MASK (0x03 << 13)
139#define QSPI_COMMAND_SDR_DDR BIT(12)
140#define QSPI_COMMAND_SIZE_SET(x) (((x) & 0xFF) << 0)
141
142#define QSPI_GLOBAL_CONFIG 0X1a4
143#define QSPI_CMB_SEQ_EN BIT(0)
144#define QSPI_TPM_WAIT_POLL_EN BIT(1)
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
164struct tegra_qspi_soc_data {
165 bool has_dma;
166 bool cmb_xfer_capable;
167 bool supports_tpm;
168 unsigned int cs_count;
169};
170
171struct tegra_qspi_client_data {
172 int tx_clk_tap_delay;
173 int rx_clk_tap_delay;
174};
175
176struct tegra_qspi {
177 struct device *dev;
178 struct spi_controller *host;
179 /* lock to protect data accessed by irq */
180 spinlock_t lock;
181
182 struct clk *clk;
183 void __iomem *base;
184 phys_addr_t phys;
185 unsigned int irq;
186
187 u32 cur_speed;
188 unsigned int cur_pos;
189 unsigned int words_per_32bit;
190 unsigned int bytes_per_word;
191 unsigned int curr_dma_words;
192 unsigned int cur_direction;
193
194 unsigned int cur_rx_pos;
195 unsigned int cur_tx_pos;
196
197 unsigned int dma_buf_size;
198 unsigned int max_buf_size;
199 bool is_curr_dma_xfer;
200
201 struct completion rx_dma_complete;
202 struct completion tx_dma_complete;
203
204 u32 tx_status;
205 u32 rx_status;
206 u32 status_reg;
207 bool is_packed;
208 bool use_dma;
209
210 u32 command1_reg;
211 u32 dma_control_reg;
212 u32 def_command1_reg;
213 u32 def_command2_reg;
214 u32 spi_cs_timing1;
215 u32 spi_cs_timing2;
216 u8 dummy_cycles;
217
218 struct completion xfer_completion;
219 struct spi_transfer *curr_xfer;
220
221 struct dma_chan *rx_dma_chan;
222 u32 *rx_dma_buf;
223 dma_addr_t rx_dma_phys;
224 struct dma_async_tx_descriptor *rx_dma_desc;
225
226 struct dma_chan *tx_dma_chan;
227 u32 *tx_dma_buf;
228 dma_addr_t tx_dma_phys;
229 struct dma_async_tx_descriptor *tx_dma_desc;
230 const struct tegra_qspi_soc_data *soc_data;
231};
232
233static inline u32 tegra_qspi_readl(struct tegra_qspi *tqspi, unsigned long offset)
234{
235 return readl(tqspi->base + offset);
236}
237
238static inline void tegra_qspi_writel(struct tegra_qspi *tqspi, u32 value, unsigned long offset)
239{
240 writel(value, tqspi->base + offset);
241
242 /* read back register to make sure that register writes completed */
243 if (offset != QSPI_TX_FIFO)
244 readl(tqspi->base + QSPI_COMMAND1);
245}
246
247static void tegra_qspi_mask_clear_irq(struct tegra_qspi *tqspi)
248{
249 u32 value;
250
251 /* write 1 to clear status register */
252 value = tegra_qspi_readl(tqspi, QSPI_TRANS_STATUS);
253 tegra_qspi_writel(tqspi, value, QSPI_TRANS_STATUS);
254
255 value = tegra_qspi_readl(tqspi, QSPI_INTR_MASK);
256 if (!(value & QSPI_INTR_RDY_MASK)) {
257 value |= (QSPI_INTR_RDY_MASK | QSPI_INTR_RX_TX_FIFO_ERR);
258 tegra_qspi_writel(tqspi, value, QSPI_INTR_MASK);
259 }
260
261 /* clear fifo status error if any */
262 value = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
263 if (value & QSPI_ERR)
264 tegra_qspi_writel(tqspi, QSPI_ERR | QSPI_FIFO_ERROR, QSPI_FIFO_STATUS);
265}
266
267static unsigned int
268tegra_qspi_calculate_curr_xfer_param(struct tegra_qspi *tqspi, struct spi_transfer *t)
269{
270 unsigned int max_word, max_len, total_fifo_words;
271 unsigned int remain_len = t->len - tqspi->cur_pos;
272 unsigned int bits_per_word = t->bits_per_word;
273
274 tqspi->bytes_per_word = DIV_ROUND_UP(bits_per_word, 8);
275
276 /*
277 * Tegra QSPI controller supports packed or unpacked mode transfers.
278 * Packed mode is used for data transfers using 8, 16, or 32 bits per
279 * word with a minimum transfer of 1 word and for all other transfers
280 * unpacked mode will be used.
281 */
282
283 if ((bits_per_word == 8 || bits_per_word == 16 ||
284 bits_per_word == 32) && t->len > 3) {
285 tqspi->is_packed = true;
286 tqspi->words_per_32bit = 32 / bits_per_word;
287 } else {
288 tqspi->is_packed = false;
289 tqspi->words_per_32bit = 1;
290 }
291
292 if (tqspi->is_packed) {
293 max_len = min(remain_len, tqspi->max_buf_size);
294 tqspi->curr_dma_words = max_len / tqspi->bytes_per_word;
295 total_fifo_words = (max_len + 3) / 4;
296 } else {
297 max_word = (remain_len - 1) / tqspi->bytes_per_word + 1;
298 max_word = min(max_word, tqspi->max_buf_size / 4);
299 tqspi->curr_dma_words = max_word;
300 total_fifo_words = max_word;
301 }
302
303 return total_fifo_words;
304}
305
306static unsigned int
307tegra_qspi_fill_tx_fifo_from_client_txbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
308{
309 unsigned int written_words, fifo_words_left, count;
310 unsigned int len, tx_empty_count, max_n_32bit, i;
311 u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
312 u32 fifo_status;
313
314 fifo_status = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
315 tx_empty_count = QSPI_TX_FIFO_EMPTY_COUNT(fifo_status);
316
317 if (tqspi->is_packed) {
318 fifo_words_left = tx_empty_count * tqspi->words_per_32bit;
319 written_words = min(fifo_words_left, tqspi->curr_dma_words);
320 len = written_words * tqspi->bytes_per_word;
321 max_n_32bit = DIV_ROUND_UP(len, 4);
322 for (count = 0; count < max_n_32bit; count++) {
323 u32 x = 0;
324
325 for (i = 0; (i < 4) && len; i++, len--)
326 x |= (u32)(*tx_buf++) << (i * 8);
327 tegra_qspi_writel(tqspi, x, QSPI_TX_FIFO);
328 }
329
330 tqspi->cur_tx_pos += written_words * tqspi->bytes_per_word;
331 } else {
332 unsigned int write_bytes;
333 u8 bytes_per_word = tqspi->bytes_per_word;
334
335 max_n_32bit = min(tqspi->curr_dma_words, tx_empty_count);
336 written_words = max_n_32bit;
337 len = written_words * tqspi->bytes_per_word;
338 if (len > t->len - tqspi->cur_pos)
339 len = t->len - tqspi->cur_pos;
340 write_bytes = len;
341 for (count = 0; count < max_n_32bit; count++) {
342 u32 x = 0;
343
344 for (i = 0; len && (i < bytes_per_word); i++, len--)
345 x |= (u32)(*tx_buf++) << (i * 8);
346 tegra_qspi_writel(tqspi, x, QSPI_TX_FIFO);
347 }
348
349 tqspi->cur_tx_pos += write_bytes;
350 }
351
352 return written_words;
353}
354
355static unsigned int
356tegra_qspi_read_rx_fifo_to_client_rxbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
357{
358 u8 *rx_buf = (u8 *)t->rx_buf + tqspi->cur_rx_pos;
359 unsigned int len, rx_full_count, count, i;
360 unsigned int read_words = 0;
361 u32 fifo_status, x;
362
363 fifo_status = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
364 rx_full_count = QSPI_RX_FIFO_FULL_COUNT(fifo_status);
365 if (tqspi->is_packed) {
366 len = tqspi->curr_dma_words * tqspi->bytes_per_word;
367 for (count = 0; count < rx_full_count; count++) {
368 x = tegra_qspi_readl(tqspi, QSPI_RX_FIFO);
369
370 for (i = 0; len && (i < 4); i++, len--)
371 *rx_buf++ = (x >> i * 8) & 0xff;
372 }
373
374 read_words += tqspi->curr_dma_words;
375 tqspi->cur_rx_pos += tqspi->curr_dma_words * tqspi->bytes_per_word;
376 } else {
377 u32 rx_mask = ((u32)1 << t->bits_per_word) - 1;
378 u8 bytes_per_word = tqspi->bytes_per_word;
379 unsigned int read_bytes;
380
381 len = rx_full_count * bytes_per_word;
382 if (len > t->len - tqspi->cur_pos)
383 len = t->len - tqspi->cur_pos;
384 read_bytes = len;
385 for (count = 0; count < rx_full_count; count++) {
386 x = tegra_qspi_readl(tqspi, QSPI_RX_FIFO) & rx_mask;
387
388 for (i = 0; len && (i < bytes_per_word); i++, len--)
389 *rx_buf++ = (x >> (i * 8)) & 0xff;
390 }
391
392 read_words += rx_full_count;
393 tqspi->cur_rx_pos += read_bytes;
394 }
395
396 return read_words;
397}
398
399static void
400tegra_qspi_copy_client_txbuf_to_qspi_txbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
401{
402 dma_sync_single_for_cpu(tqspi->dev, tqspi->tx_dma_phys,
403 tqspi->dma_buf_size, DMA_TO_DEVICE);
404
405 /*
406 * In packed mode, each word in FIFO may contain multiple packets
407 * based on bits per word. So all bytes in each FIFO word are valid.
408 *
409 * In unpacked mode, each word in FIFO contains single packet and
410 * based on bits per word any remaining bits in FIFO word will be
411 * ignored by the hardware and are invalid bits.
412 */
413 if (tqspi->is_packed) {
414 tqspi->cur_tx_pos += tqspi->curr_dma_words * tqspi->bytes_per_word;
415 } else {
416 u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
417 unsigned int i, count, consume, write_bytes;
418
419 /*
420 * Fill tx_dma_buf to contain single packet in each word based
421 * on bits per word from SPI core tx_buf.
422 */
423 consume = tqspi->curr_dma_words * tqspi->bytes_per_word;
424 if (consume > t->len - tqspi->cur_pos)
425 consume = t->len - tqspi->cur_pos;
426 write_bytes = consume;
427 for (count = 0; count < tqspi->curr_dma_words; count++) {
428 u32 x = 0;
429
430 for (i = 0; consume && (i < tqspi->bytes_per_word); i++, consume--)
431 x |= (u32)(*tx_buf++) << (i * 8);
432 tqspi->tx_dma_buf[count] = x;
433 }
434
435 tqspi->cur_tx_pos += write_bytes;
436 }
437
438 dma_sync_single_for_device(tqspi->dev, tqspi->tx_dma_phys,
439 tqspi->dma_buf_size, DMA_TO_DEVICE);
440}
441
442static void
443tegra_qspi_copy_qspi_rxbuf_to_client_rxbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
444{
445 dma_sync_single_for_cpu(tqspi->dev, tqspi->rx_dma_phys,
446 tqspi->dma_buf_size, DMA_FROM_DEVICE);
447
448 if (tqspi->is_packed) {
449 tqspi->cur_rx_pos += tqspi->curr_dma_words * tqspi->bytes_per_word;
450 } else {
451 unsigned char *rx_buf = t->rx_buf + tqspi->cur_rx_pos;
452 u32 rx_mask = ((u32)1 << t->bits_per_word) - 1;
453 unsigned int i, count, consume, read_bytes;
454
455 /*
456 * Each FIFO word contains single data packet.
457 * Skip invalid bits in each FIFO word based on bits per word
458 * and align bytes while filling in SPI core rx_buf.
459 */
460 consume = tqspi->curr_dma_words * tqspi->bytes_per_word;
461 if (consume > t->len - tqspi->cur_pos)
462 consume = t->len - tqspi->cur_pos;
463 read_bytes = consume;
464 for (count = 0; count < tqspi->curr_dma_words; count++) {
465 u32 x = tqspi->rx_dma_buf[count] & rx_mask;
466
467 for (i = 0; consume && (i < tqspi->bytes_per_word); i++, consume--)
468 *rx_buf++ = (x >> (i * 8)) & 0xff;
469 }
470
471 tqspi->cur_rx_pos += read_bytes;
472 }
473
474 dma_sync_single_for_device(tqspi->dev, tqspi->rx_dma_phys,
475 tqspi->dma_buf_size, DMA_FROM_DEVICE);
476}
477
478static void tegra_qspi_dma_complete(void *args)
479{
480 struct completion *dma_complete = args;
481
482 complete(dma_complete);
483}
484
485static int tegra_qspi_start_tx_dma(struct tegra_qspi *tqspi, struct spi_transfer *t, int len)
486{
487 dma_addr_t tx_dma_phys;
488
489 reinit_completion(&tqspi->tx_dma_complete);
490
491 if (tqspi->is_packed)
492 tx_dma_phys = t->tx_dma;
493 else
494 tx_dma_phys = tqspi->tx_dma_phys;
495
496 tqspi->tx_dma_desc = dmaengine_prep_slave_single(tqspi->tx_dma_chan, tx_dma_phys,
497 len, DMA_MEM_TO_DEV,
498 DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
499
500 if (!tqspi->tx_dma_desc) {
501 dev_err(tqspi->dev, "Unable to get TX descriptor\n");
502 return -EIO;
503 }
504
505 tqspi->tx_dma_desc->callback = tegra_qspi_dma_complete;
506 tqspi->tx_dma_desc->callback_param = &tqspi->tx_dma_complete;
507 dmaengine_submit(tqspi->tx_dma_desc);
508 dma_async_issue_pending(tqspi->tx_dma_chan);
509
510 return 0;
511}
512
513static int tegra_qspi_start_rx_dma(struct tegra_qspi *tqspi, struct spi_transfer *t, int len)
514{
515 dma_addr_t rx_dma_phys;
516
517 reinit_completion(&tqspi->rx_dma_complete);
518
519 if (tqspi->is_packed)
520 rx_dma_phys = t->rx_dma;
521 else
522 rx_dma_phys = tqspi->rx_dma_phys;
523
524 tqspi->rx_dma_desc = dmaengine_prep_slave_single(tqspi->rx_dma_chan, rx_dma_phys,
525 len, DMA_DEV_TO_MEM,
526 DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
527
528 if (!tqspi->rx_dma_desc) {
529 dev_err(tqspi->dev, "Unable to get RX descriptor\n");
530 return -EIO;
531 }
532
533 tqspi->rx_dma_desc->callback = tegra_qspi_dma_complete;
534 tqspi->rx_dma_desc->callback_param = &tqspi->rx_dma_complete;
535 dmaengine_submit(tqspi->rx_dma_desc);
536 dma_async_issue_pending(tqspi->rx_dma_chan);
537
538 return 0;
539}
540
541static int tegra_qspi_flush_fifos(struct tegra_qspi *tqspi, bool atomic)
542{
543 void __iomem *addr = tqspi->base + QSPI_FIFO_STATUS;
544 u32 val;
545
546 val = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
547 if ((val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY)
548 return 0;
549
550 val |= QSPI_RX_FIFO_FLUSH | QSPI_TX_FIFO_FLUSH;
551 tegra_qspi_writel(tqspi, val, QSPI_FIFO_STATUS);
552
553 if (!atomic)
554 return readl_relaxed_poll_timeout(addr, val,
555 (val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY,
556 1000, 1000000);
557
558 return readl_relaxed_poll_timeout_atomic(addr, val,
559 (val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY,
560 1000, 1000000);
561}
562
563static void tegra_qspi_unmask_irq(struct tegra_qspi *tqspi)
564{
565 u32 intr_mask;
566
567 intr_mask = tegra_qspi_readl(tqspi, QSPI_INTR_MASK);
568 intr_mask &= ~(QSPI_INTR_RDY_MASK | QSPI_INTR_RX_TX_FIFO_ERR);
569 tegra_qspi_writel(tqspi, intr_mask, QSPI_INTR_MASK);
570}
571
572static int tegra_qspi_dma_map_xfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
573{
574 u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
575 u8 *rx_buf = (u8 *)t->rx_buf + tqspi->cur_rx_pos;
576 unsigned int len;
577
578 len = DIV_ROUND_UP(tqspi->curr_dma_words * tqspi->bytes_per_word, 4) * 4;
579
580 if (t->tx_buf) {
581 t->tx_dma = dma_map_single(tqspi->dev, (void *)tx_buf, len, DMA_TO_DEVICE);
582 if (dma_mapping_error(tqspi->dev, t->tx_dma))
583 return -ENOMEM;
584 }
585
586 if (t->rx_buf) {
587 t->rx_dma = dma_map_single(tqspi->dev, (void *)rx_buf, len, DMA_FROM_DEVICE);
588 if (dma_mapping_error(tqspi->dev, t->rx_dma)) {
589 dma_unmap_single(tqspi->dev, t->tx_dma, len, DMA_TO_DEVICE);
590 return -ENOMEM;
591 }
592 }
593
594 return 0;
595}
596
597static void tegra_qspi_dma_unmap_xfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
598{
599 unsigned int len;
600
601 len = DIV_ROUND_UP(tqspi->curr_dma_words * tqspi->bytes_per_word, 4) * 4;
602
603 dma_unmap_single(tqspi->dev, t->tx_dma, len, DMA_TO_DEVICE);
604 dma_unmap_single(tqspi->dev, t->rx_dma, len, DMA_FROM_DEVICE);
605}
606
607static int tegra_qspi_start_dma_based_transfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
608{
609 struct dma_slave_config dma_sconfig = { 0 };
610 unsigned int len;
611 u8 dma_burst;
612 int ret = 0;
613 u32 val;
614
615 if (tqspi->is_packed) {
616 ret = tegra_qspi_dma_map_xfer(tqspi, t);
617 if (ret < 0)
618 return ret;
619 }
620
621 val = QSPI_DMA_BLK_SET(tqspi->curr_dma_words - 1);
622 tegra_qspi_writel(tqspi, val, QSPI_DMA_BLK);
623
624 tegra_qspi_unmask_irq(tqspi);
625
626 if (tqspi->is_packed)
627 len = DIV_ROUND_UP(tqspi->curr_dma_words * tqspi->bytes_per_word, 4) * 4;
628 else
629 len = tqspi->curr_dma_words * 4;
630
631 /* set attention level based on length of transfer */
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 tqspi->dma_control_reg = val;
646
647 dma_sconfig.device_fc = true;
648 if (tqspi->cur_direction & DATA_DIR_TX) {
649 dma_sconfig.dst_addr = tqspi->phys + QSPI_TX_FIFO;
650 dma_sconfig.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
651 dma_sconfig.dst_maxburst = dma_burst;
652 ret = dmaengine_slave_config(tqspi->tx_dma_chan, &dma_sconfig);
653 if (ret < 0) {
654 dev_err(tqspi->dev, "failed DMA slave config: %d\n", ret);
655 return ret;
656 }
657
658 tegra_qspi_copy_client_txbuf_to_qspi_txbuf(tqspi, t);
659 ret = tegra_qspi_start_tx_dma(tqspi, t, len);
660 if (ret < 0) {
661 dev_err(tqspi->dev, "failed to starting TX DMA: %d\n", ret);
662 return ret;
663 }
664 }
665
666 if (tqspi->cur_direction & DATA_DIR_RX) {
667 dma_sconfig.src_addr = tqspi->phys + QSPI_RX_FIFO;
668 dma_sconfig.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
669 dma_sconfig.src_maxburst = dma_burst;
670 ret = dmaengine_slave_config(tqspi->rx_dma_chan, &dma_sconfig);
671 if (ret < 0) {
672 dev_err(tqspi->dev, "failed DMA slave config: %d\n", ret);
673 return ret;
674 }
675
676 dma_sync_single_for_device(tqspi->dev, tqspi->rx_dma_phys,
677 tqspi->dma_buf_size,
678 DMA_FROM_DEVICE);
679
680 ret = tegra_qspi_start_rx_dma(tqspi, t, len);
681 if (ret < 0) {
682 dev_err(tqspi->dev, "failed to start RX DMA: %d\n", ret);
683 if (tqspi->cur_direction & DATA_DIR_TX)
684 dmaengine_terminate_all(tqspi->tx_dma_chan);
685 return ret;
686 }
687 }
688
689 tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
690
691 tqspi->is_curr_dma_xfer = true;
692 tqspi->dma_control_reg = val;
693 val |= QSPI_DMA_EN;
694 tegra_qspi_writel(tqspi, val, QSPI_DMA_CTL);
695
696 return ret;
697}
698
699static int tegra_qspi_start_cpu_based_transfer(struct tegra_qspi *qspi, struct spi_transfer *t)
700{
701 u32 val;
702 unsigned int cur_words;
703
704 if (qspi->cur_direction & DATA_DIR_TX)
705 cur_words = tegra_qspi_fill_tx_fifo_from_client_txbuf(qspi, t);
706 else
707 cur_words = qspi->curr_dma_words;
708
709 val = QSPI_DMA_BLK_SET(cur_words - 1);
710 tegra_qspi_writel(qspi, val, QSPI_DMA_BLK);
711
712 tegra_qspi_unmask_irq(qspi);
713
714 qspi->is_curr_dma_xfer = false;
715 val = qspi->command1_reg;
716 val |= QSPI_PIO;
717 tegra_qspi_writel(qspi, val, QSPI_COMMAND1);
718
719 return 0;
720}
721
722static void tegra_qspi_deinit_dma(struct tegra_qspi *tqspi)
723{
724 if (!tqspi->soc_data->has_dma)
725 return;
726
727 if (tqspi->tx_dma_buf) {
728 dma_free_coherent(tqspi->dev, tqspi->dma_buf_size,
729 tqspi->tx_dma_buf, tqspi->tx_dma_phys);
730 tqspi->tx_dma_buf = NULL;
731 }
732
733 if (tqspi->tx_dma_chan) {
734 dma_release_channel(tqspi->tx_dma_chan);
735 tqspi->tx_dma_chan = NULL;
736 }
737
738 if (tqspi->rx_dma_buf) {
739 dma_free_coherent(tqspi->dev, tqspi->dma_buf_size,
740 tqspi->rx_dma_buf, tqspi->rx_dma_phys);
741 tqspi->rx_dma_buf = NULL;
742 }
743
744 if (tqspi->rx_dma_chan) {
745 dma_release_channel(tqspi->rx_dma_chan);
746 tqspi->rx_dma_chan = NULL;
747 }
748}
749
750static int tegra_qspi_init_dma(struct tegra_qspi *tqspi)
751{
752 struct dma_chan *dma_chan;
753 dma_addr_t dma_phys;
754 u32 *dma_buf;
755 int err;
756
757 if (!tqspi->soc_data->has_dma)
758 return 0;
759
760 dma_chan = dma_request_chan(tqspi->dev, "rx");
761 if (IS_ERR(dma_chan)) {
762 err = PTR_ERR(dma_chan);
763 goto err_out;
764 }
765
766 tqspi->rx_dma_chan = dma_chan;
767
768 dma_buf = dma_alloc_coherent(tqspi->dev, tqspi->dma_buf_size, &dma_phys, GFP_KERNEL);
769 if (!dma_buf) {
770 err = -ENOMEM;
771 goto err_out;
772 }
773
774 tqspi->rx_dma_buf = dma_buf;
775 tqspi->rx_dma_phys = dma_phys;
776
777 dma_chan = dma_request_chan(tqspi->dev, "tx");
778 if (IS_ERR(dma_chan)) {
779 err = PTR_ERR(dma_chan);
780 goto err_out;
781 }
782
783 tqspi->tx_dma_chan = dma_chan;
784
785 dma_buf = dma_alloc_coherent(tqspi->dev, tqspi->dma_buf_size, &dma_phys, GFP_KERNEL);
786 if (!dma_buf) {
787 err = -ENOMEM;
788 goto err_out;
789 }
790
791 tqspi->tx_dma_buf = dma_buf;
792 tqspi->tx_dma_phys = dma_phys;
793 tqspi->use_dma = true;
794
795 return 0;
796
797err_out:
798 tegra_qspi_deinit_dma(tqspi);
799
800 if (err != -EPROBE_DEFER) {
801 dev_err(tqspi->dev, "cannot use DMA: %d\n", err);
802 dev_err(tqspi->dev, "falling back to PIO\n");
803 return 0;
804 }
805
806 return err;
807}
808
809static u32 tegra_qspi_setup_transfer_one(struct spi_device *spi, struct spi_transfer *t,
810 bool is_first_of_msg)
811{
812 struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);
813 struct tegra_qspi_client_data *cdata = spi->controller_data;
814 u32 command1, command2, speed = t->speed_hz;
815 u8 bits_per_word = t->bits_per_word;
816 u32 tx_tap = 0, rx_tap = 0;
817 int req_mode;
818
819 if (!has_acpi_companion(tqspi->dev) && speed != tqspi->cur_speed) {
820 clk_set_rate(tqspi->clk, speed);
821 tqspi->cur_speed = speed;
822 }
823
824 tqspi->cur_pos = 0;
825 tqspi->cur_rx_pos = 0;
826 tqspi->cur_tx_pos = 0;
827 tqspi->curr_xfer = t;
828
829 if (is_first_of_msg) {
830 tegra_qspi_mask_clear_irq(tqspi);
831
832 command1 = tqspi->def_command1_reg;
833 command1 |= QSPI_CS_SEL(spi_get_chipselect(spi, 0));
834 command1 |= QSPI_BIT_LENGTH(bits_per_word - 1);
835
836 command1 &= ~QSPI_CONTROL_MODE_MASK;
837 req_mode = spi->mode & 0x3;
838 if (req_mode == SPI_MODE_3)
839 command1 |= QSPI_CONTROL_MODE_3;
840 else
841 command1 |= QSPI_CONTROL_MODE_0;
842
843 if (spi->mode & SPI_CS_HIGH)
844 command1 |= QSPI_CS_SW_VAL;
845 else
846 command1 &= ~QSPI_CS_SW_VAL;
847 tegra_qspi_writel(tqspi, command1, QSPI_COMMAND1);
848
849 if (cdata && cdata->tx_clk_tap_delay)
850 tx_tap = cdata->tx_clk_tap_delay;
851
852 if (cdata && cdata->rx_clk_tap_delay)
853 rx_tap = cdata->rx_clk_tap_delay;
854
855 command2 = QSPI_TX_TAP_DELAY(tx_tap) | QSPI_RX_TAP_DELAY(rx_tap);
856 if (command2 != tqspi->def_command2_reg)
857 tegra_qspi_writel(tqspi, command2, QSPI_COMMAND2);
858
859 } else {
860 command1 = tqspi->command1_reg;
861 command1 &= ~QSPI_BIT_LENGTH(~0);
862 command1 |= QSPI_BIT_LENGTH(bits_per_word - 1);
863 }
864
865 command1 &= ~QSPI_SDR_DDR_SEL;
866
867 return command1;
868}
869
870static int tegra_qspi_start_transfer_one(struct spi_device *spi,
871 struct spi_transfer *t, u32 command1)
872{
873 struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);
874 unsigned int total_fifo_words;
875 u8 bus_width = 0;
876 int ret;
877
878 total_fifo_words = tegra_qspi_calculate_curr_xfer_param(tqspi, t);
879
880 command1 &= ~QSPI_PACKED;
881 if (tqspi->is_packed)
882 command1 |= QSPI_PACKED;
883 tegra_qspi_writel(tqspi, command1, QSPI_COMMAND1);
884
885 tqspi->cur_direction = 0;
886
887 command1 &= ~(QSPI_TX_EN | QSPI_RX_EN);
888 if (t->rx_buf) {
889 command1 |= QSPI_RX_EN;
890 tqspi->cur_direction |= DATA_DIR_RX;
891 bus_width = t->rx_nbits;
892 }
893
894 if (t->tx_buf) {
895 command1 |= QSPI_TX_EN;
896 tqspi->cur_direction |= DATA_DIR_TX;
897 bus_width = t->tx_nbits;
898 }
899
900 command1 &= ~QSPI_INTERFACE_WIDTH_MASK;
901
902 if (bus_width == SPI_NBITS_QUAD)
903 command1 |= QSPI_INTERFACE_WIDTH_QUAD;
904 else if (bus_width == SPI_NBITS_DUAL)
905 command1 |= QSPI_INTERFACE_WIDTH_DUAL;
906 else
907 command1 |= QSPI_INTERFACE_WIDTH_SINGLE;
908
909 tqspi->command1_reg = command1;
910
911 tegra_qspi_writel(tqspi, QSPI_NUM_DUMMY_CYCLE(tqspi->dummy_cycles), QSPI_MISC_REG);
912
913 ret = tegra_qspi_flush_fifos(tqspi, false);
914 if (ret < 0)
915 return ret;
916
917 if (tqspi->use_dma && total_fifo_words > QSPI_FIFO_DEPTH)
918 ret = tegra_qspi_start_dma_based_transfer(tqspi, t);
919 else
920 ret = tegra_qspi_start_cpu_based_transfer(tqspi, t);
921
922 return ret;
923}
924
925static struct tegra_qspi_client_data *tegra_qspi_parse_cdata_dt(struct spi_device *spi)
926{
927 struct tegra_qspi_client_data *cdata;
928 struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);
929
930 cdata = devm_kzalloc(tqspi->dev, sizeof(*cdata), GFP_KERNEL);
931 if (!cdata)
932 return NULL;
933
934 device_property_read_u32(&spi->dev, "nvidia,tx-clk-tap-delay",
935 &cdata->tx_clk_tap_delay);
936 device_property_read_u32(&spi->dev, "nvidia,rx-clk-tap-delay",
937 &cdata->rx_clk_tap_delay);
938
939 return cdata;
940}
941
942static int tegra_qspi_setup(struct spi_device *spi)
943{
944 struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);
945 struct tegra_qspi_client_data *cdata = spi->controller_data;
946 unsigned long flags;
947 u32 val;
948 int ret;
949
950 ret = pm_runtime_resume_and_get(tqspi->dev);
951 if (ret < 0) {
952 dev_err(tqspi->dev, "failed to get runtime PM: %d\n", ret);
953 return ret;
954 }
955
956 if (!cdata) {
957 cdata = tegra_qspi_parse_cdata_dt(spi);
958 spi->controller_data = cdata;
959 }
960 spin_lock_irqsave(&tqspi->lock, flags);
961
962 /* keep default cs state to inactive */
963 val = tqspi->def_command1_reg;
964 val |= QSPI_CS_SEL(spi_get_chipselect(spi, 0));
965 if (spi->mode & SPI_CS_HIGH)
966 val &= ~QSPI_CS_POL_INACTIVE(spi_get_chipselect(spi, 0));
967 else
968 val |= QSPI_CS_POL_INACTIVE(spi_get_chipselect(spi, 0));
969
970 tqspi->def_command1_reg = val;
971 tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
972
973 spin_unlock_irqrestore(&tqspi->lock, flags);
974
975 pm_runtime_put(tqspi->dev);
976
977 return 0;
978}
979
980static void tegra_qspi_dump_regs(struct tegra_qspi *tqspi)
981{
982 dev_dbg(tqspi->dev, "============ QSPI REGISTER DUMP ============\n");
983 dev_dbg(tqspi->dev, "Command1: 0x%08x | Command2: 0x%08x\n",
984 tegra_qspi_readl(tqspi, QSPI_COMMAND1),
985 tegra_qspi_readl(tqspi, QSPI_COMMAND2));
986 dev_dbg(tqspi->dev, "DMA_CTL: 0x%08x | DMA_BLK: 0x%08x\n",
987 tegra_qspi_readl(tqspi, QSPI_DMA_CTL),
988 tegra_qspi_readl(tqspi, QSPI_DMA_BLK));
989 dev_dbg(tqspi->dev, "INTR_MASK: 0x%08x | MISC: 0x%08x\n",
990 tegra_qspi_readl(tqspi, QSPI_INTR_MASK),
991 tegra_qspi_readl(tqspi, QSPI_MISC_REG));
992 dev_dbg(tqspi->dev, "TRANS_STAT: 0x%08x | FIFO_STATUS: 0x%08x\n",
993 tegra_qspi_readl(tqspi, QSPI_TRANS_STATUS),
994 tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS));
995}
996
997static void tegra_qspi_handle_error(struct tegra_qspi *tqspi)
998{
999 dev_err(tqspi->dev, "error in transfer, fifo status 0x%08x\n", tqspi->status_reg);
1000 tegra_qspi_dump_regs(tqspi);
1001 tegra_qspi_flush_fifos(tqspi, true);
1002 if (device_reset(tqspi->dev) < 0)
1003 dev_warn_once(tqspi->dev, "device reset failed\n");
1004}
1005
1006static void tegra_qspi_transfer_end(struct spi_device *spi)
1007{
1008 struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);
1009 int cs_val = (spi->mode & SPI_CS_HIGH) ? 0 : 1;
1010
1011 if (cs_val)
1012 tqspi->command1_reg |= QSPI_CS_SW_VAL;
1013 else
1014 tqspi->command1_reg &= ~QSPI_CS_SW_VAL;
1015 tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
1016 tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
1017}
1018
1019static u32 tegra_qspi_cmd_config(bool is_ddr, u8 bus_width, u8 len)
1020{
1021 u32 cmd_config = 0;
1022
1023 /* Extract Command configuration and value */
1024 if (is_ddr)
1025 cmd_config |= QSPI_COMMAND_SDR_DDR;
1026 else
1027 cmd_config &= ~QSPI_COMMAND_SDR_DDR;
1028
1029 cmd_config |= QSPI_COMMAND_X1_X2_X4(bus_width);
1030 cmd_config |= QSPI_COMMAND_SIZE_SET((len * 8) - 1);
1031
1032 return cmd_config;
1033}
1034
1035static u32 tegra_qspi_addr_config(bool is_ddr, u8 bus_width, u8 len)
1036{
1037 u32 addr_config = 0;
1038
1039 /* Extract Address configuration and value */
1040 is_ddr = 0; //Only SDR mode supported
1041 bus_width = 0; //X1 mode
1042
1043 if (is_ddr)
1044 addr_config |= QSPI_ADDRESS_SDR_DDR;
1045 else
1046 addr_config &= ~QSPI_ADDRESS_SDR_DDR;
1047
1048 addr_config |= QSPI_ADDRESS_X1_X2_X4(bus_width);
1049 addr_config |= QSPI_ADDRESS_SIZE_SET((len * 8) - 1);
1050
1051 return addr_config;
1052}
1053
1054static int tegra_qspi_combined_seq_xfer(struct tegra_qspi *tqspi,
1055 struct spi_message *msg)
1056{
1057 bool is_first_msg = true;
1058 struct spi_transfer *xfer;
1059 struct spi_device *spi = msg->spi;
1060 u8 transfer_phase = 0;
1061 u32 cmd1 = 0, dma_ctl = 0;
1062 int ret = 0;
1063 u32 address_value = 0;
1064 u32 cmd_config = 0, addr_config = 0;
1065 u8 cmd_value = 0, val = 0;
1066
1067 /* Enable Combined sequence mode */
1068 val = tegra_qspi_readl(tqspi, QSPI_GLOBAL_CONFIG);
1069 if (spi->mode & SPI_TPM_HW_FLOW) {
1070 if (tqspi->soc_data->supports_tpm)
1071 val |= QSPI_TPM_WAIT_POLL_EN;
1072 else
1073 return -EIO;
1074 }
1075 val |= QSPI_CMB_SEQ_EN;
1076 tegra_qspi_writel(tqspi, val, QSPI_GLOBAL_CONFIG);
1077 /* Process individual transfer list */
1078 list_for_each_entry(xfer, &msg->transfers, transfer_list) {
1079 switch (transfer_phase) {
1080 case CMD_TRANSFER:
1081 /* X1 SDR mode */
1082 cmd_config = tegra_qspi_cmd_config(false, 0,
1083 xfer->len);
1084 cmd_value = *((const u8 *)(xfer->tx_buf));
1085 break;
1086 case ADDR_TRANSFER:
1087 /* X1 SDR mode */
1088 addr_config = tegra_qspi_addr_config(false, 0,
1089 xfer->len);
1090 address_value = *((const u32 *)(xfer->tx_buf));
1091 break;
1092 case DATA_TRANSFER:
1093 /* Program Command, Address value in register */
1094 tegra_qspi_writel(tqspi, cmd_value, QSPI_CMB_SEQ_CMD);
1095 tegra_qspi_writel(tqspi, address_value,
1096 QSPI_CMB_SEQ_ADDR);
1097 /* Program Command and Address config in register */
1098 tegra_qspi_writel(tqspi, cmd_config,
1099 QSPI_CMB_SEQ_CMD_CFG);
1100 tegra_qspi_writel(tqspi, addr_config,
1101 QSPI_CMB_SEQ_ADDR_CFG);
1102
1103 reinit_completion(&tqspi->xfer_completion);
1104 cmd1 = tegra_qspi_setup_transfer_one(spi, xfer,
1105 is_first_msg);
1106 ret = tegra_qspi_start_transfer_one(spi, xfer,
1107 cmd1);
1108
1109 if (ret < 0) {
1110 dev_err(tqspi->dev, "Failed to start transfer-one: %d\n",
1111 ret);
1112 return ret;
1113 }
1114
1115 is_first_msg = false;
1116 ret = wait_for_completion_timeout
1117 (&tqspi->xfer_completion,
1118 QSPI_DMA_TIMEOUT);
1119
1120 if (WARN_ON(ret == 0)) {
1121 dev_err(tqspi->dev, "QSPI Transfer failed with timeout: %d\n",
1122 ret);
1123 if (tqspi->is_curr_dma_xfer &&
1124 (tqspi->cur_direction & DATA_DIR_TX))
1125 dmaengine_terminate_all
1126 (tqspi->tx_dma_chan);
1127
1128 if (tqspi->is_curr_dma_xfer &&
1129 (tqspi->cur_direction & DATA_DIR_RX))
1130 dmaengine_terminate_all
1131 (tqspi->rx_dma_chan);
1132
1133 /* Abort transfer by resetting pio/dma bit */
1134 if (!tqspi->is_curr_dma_xfer) {
1135 cmd1 = tegra_qspi_readl
1136 (tqspi,
1137 QSPI_COMMAND1);
1138 cmd1 &= ~QSPI_PIO;
1139 tegra_qspi_writel
1140 (tqspi, cmd1,
1141 QSPI_COMMAND1);
1142 } else {
1143 dma_ctl = tegra_qspi_readl
1144 (tqspi,
1145 QSPI_DMA_CTL);
1146 dma_ctl &= ~QSPI_DMA_EN;
1147 tegra_qspi_writel(tqspi, dma_ctl,
1148 QSPI_DMA_CTL);
1149 }
1150
1151 /* Reset controller if timeout happens */
1152 if (device_reset(tqspi->dev) < 0)
1153 dev_warn_once(tqspi->dev,
1154 "device reset failed\n");
1155 ret = -EIO;
1156 goto exit;
1157 }
1158
1159 if (tqspi->tx_status || tqspi->rx_status) {
1160 dev_err(tqspi->dev, "QSPI Transfer failed\n");
1161 tqspi->tx_status = 0;
1162 tqspi->rx_status = 0;
1163 ret = -EIO;
1164 goto exit;
1165 }
1166 if (!xfer->cs_change) {
1167 tegra_qspi_transfer_end(spi);
1168 spi_transfer_delay_exec(xfer);
1169 }
1170 break;
1171 default:
1172 ret = -EINVAL;
1173 goto exit;
1174 }
1175 msg->actual_length += xfer->len;
1176 transfer_phase++;
1177 }
1178 ret = 0;
1179
1180exit:
1181 msg->status = ret;
1182 if (ret < 0) {
1183 tegra_qspi_transfer_end(spi);
1184 spi_transfer_delay_exec(xfer);
1185 }
1186
1187 return ret;
1188}
1189
1190static int tegra_qspi_non_combined_seq_xfer(struct tegra_qspi *tqspi,
1191 struct spi_message *msg)
1192{
1193 struct spi_device *spi = msg->spi;
1194 struct spi_transfer *transfer;
1195 bool is_first_msg = true;
1196 int ret = 0, val = 0;
1197
1198 msg->status = 0;
1199 msg->actual_length = 0;
1200 tqspi->tx_status = 0;
1201 tqspi->rx_status = 0;
1202
1203 /* Disable Combined sequence mode */
1204 val = tegra_qspi_readl(tqspi, QSPI_GLOBAL_CONFIG);
1205 val &= ~QSPI_CMB_SEQ_EN;
1206 if (tqspi->soc_data->supports_tpm)
1207 val &= ~QSPI_TPM_WAIT_POLL_EN;
1208 tegra_qspi_writel(tqspi, val, QSPI_GLOBAL_CONFIG);
1209 list_for_each_entry(transfer, &msg->transfers, transfer_list) {
1210 struct spi_transfer *xfer = transfer;
1211 u8 dummy_bytes = 0;
1212 u32 cmd1;
1213
1214 tqspi->dummy_cycles = 0;
1215 /*
1216 * Tegra QSPI hardware supports dummy bytes transfer after actual transfer
1217 * bytes based on programmed dummy clock cycles in the QSPI_MISC register.
1218 * So, check if the next transfer is dummy data transfer and program dummy
1219 * clock cycles along with the current transfer and skip next transfer.
1220 */
1221 if (!list_is_last(&xfer->transfer_list, &msg->transfers)) {
1222 struct spi_transfer *next_xfer;
1223
1224 next_xfer = list_next_entry(xfer, transfer_list);
1225 if (next_xfer->dummy_data) {
1226 u32 dummy_cycles = next_xfer->len * 8 / next_xfer->tx_nbits;
1227
1228 if (dummy_cycles <= QSPI_DUMMY_CYCLES_MAX) {
1229 tqspi->dummy_cycles = dummy_cycles;
1230 dummy_bytes = next_xfer->len;
1231 transfer = next_xfer;
1232 }
1233 }
1234 }
1235
1236 reinit_completion(&tqspi->xfer_completion);
1237
1238 cmd1 = tegra_qspi_setup_transfer_one(spi, xfer, is_first_msg);
1239
1240 ret = tegra_qspi_start_transfer_one(spi, xfer, cmd1);
1241 if (ret < 0) {
1242 dev_err(tqspi->dev, "failed to start transfer: %d\n", ret);
1243 goto complete_xfer;
1244 }
1245
1246 ret = wait_for_completion_timeout(&tqspi->xfer_completion,
1247 QSPI_DMA_TIMEOUT);
1248 if (WARN_ON(ret == 0)) {
1249 dev_err(tqspi->dev, "transfer timeout\n");
1250 if (tqspi->is_curr_dma_xfer && (tqspi->cur_direction & DATA_DIR_TX))
1251 dmaengine_terminate_all(tqspi->tx_dma_chan);
1252 if (tqspi->is_curr_dma_xfer && (tqspi->cur_direction & DATA_DIR_RX))
1253 dmaengine_terminate_all(tqspi->rx_dma_chan);
1254 tegra_qspi_handle_error(tqspi);
1255 ret = -EIO;
1256 goto complete_xfer;
1257 }
1258
1259 if (tqspi->tx_status || tqspi->rx_status) {
1260 tegra_qspi_handle_error(tqspi);
1261 ret = -EIO;
1262 goto complete_xfer;
1263 }
1264
1265 msg->actual_length += xfer->len + dummy_bytes;
1266
1267complete_xfer:
1268 if (ret < 0) {
1269 tegra_qspi_transfer_end(spi);
1270 spi_transfer_delay_exec(xfer);
1271 goto exit;
1272 }
1273
1274 if (list_is_last(&xfer->transfer_list, &msg->transfers)) {
1275 /* de-activate CS after last transfer only when cs_change is not set */
1276 if (!xfer->cs_change) {
1277 tegra_qspi_transfer_end(spi);
1278 spi_transfer_delay_exec(xfer);
1279 }
1280 } else if (xfer->cs_change) {
1281 /* de-activated CS between the transfers only when cs_change is set */
1282 tegra_qspi_transfer_end(spi);
1283 spi_transfer_delay_exec(xfer);
1284 }
1285 }
1286
1287 ret = 0;
1288exit:
1289 msg->status = ret;
1290
1291 return ret;
1292}
1293
1294static bool tegra_qspi_validate_cmb_seq(struct tegra_qspi *tqspi,
1295 struct spi_message *msg)
1296{
1297 int transfer_count = 0;
1298 struct spi_transfer *xfer;
1299
1300 list_for_each_entry(xfer, &msg->transfers, transfer_list) {
1301 transfer_count++;
1302 }
1303 if (!tqspi->soc_data->cmb_xfer_capable || transfer_count != 3)
1304 return false;
1305 xfer = list_first_entry(&msg->transfers, typeof(*xfer),
1306 transfer_list);
1307 if (xfer->len > 2)
1308 return false;
1309 xfer = list_next_entry(xfer, transfer_list);
1310 if (xfer->len > 4 || xfer->len < 3)
1311 return false;
1312 xfer = list_next_entry(xfer, transfer_list);
1313 if (!tqspi->soc_data->has_dma && xfer->len > (QSPI_FIFO_DEPTH << 2))
1314 return false;
1315
1316 return true;
1317}
1318
1319static int tegra_qspi_transfer_one_message(struct spi_controller *host,
1320 struct spi_message *msg)
1321{
1322 struct tegra_qspi *tqspi = spi_controller_get_devdata(host);
1323 int ret;
1324
1325 if (tegra_qspi_validate_cmb_seq(tqspi, msg))
1326 ret = tegra_qspi_combined_seq_xfer(tqspi, msg);
1327 else
1328 ret = tegra_qspi_non_combined_seq_xfer(tqspi, msg);
1329
1330 spi_finalize_current_message(host);
1331
1332 return ret;
1333}
1334
1335static irqreturn_t handle_cpu_based_xfer(struct tegra_qspi *tqspi)
1336{
1337 struct spi_transfer *t = tqspi->curr_xfer;
1338 unsigned long flags;
1339
1340 spin_lock_irqsave(&tqspi->lock, flags);
1341
1342 if (tqspi->tx_status || tqspi->rx_status) {
1343 tegra_qspi_handle_error(tqspi);
1344 complete(&tqspi->xfer_completion);
1345 goto exit;
1346 }
1347
1348 if (tqspi->cur_direction & DATA_DIR_RX)
1349 tegra_qspi_read_rx_fifo_to_client_rxbuf(tqspi, t);
1350
1351 if (tqspi->cur_direction & DATA_DIR_TX)
1352 tqspi->cur_pos = tqspi->cur_tx_pos;
1353 else
1354 tqspi->cur_pos = tqspi->cur_rx_pos;
1355
1356 if (tqspi->cur_pos == t->len) {
1357 complete(&tqspi->xfer_completion);
1358 goto exit;
1359 }
1360
1361 tegra_qspi_calculate_curr_xfer_param(tqspi, t);
1362 tegra_qspi_start_cpu_based_transfer(tqspi, t);
1363exit:
1364 spin_unlock_irqrestore(&tqspi->lock, flags);
1365 return IRQ_HANDLED;
1366}
1367
1368static irqreturn_t handle_dma_based_xfer(struct tegra_qspi *tqspi)
1369{
1370 struct spi_transfer *t = tqspi->curr_xfer;
1371 unsigned int total_fifo_words;
1372 unsigned long flags;
1373 long wait_status;
1374 int err = 0;
1375
1376 if (tqspi->cur_direction & DATA_DIR_TX) {
1377 if (tqspi->tx_status) {
1378 dmaengine_terminate_all(tqspi->tx_dma_chan);
1379 err += 1;
1380 } else {
1381 wait_status = wait_for_completion_interruptible_timeout(
1382 &tqspi->tx_dma_complete, QSPI_DMA_TIMEOUT);
1383 if (wait_status <= 0) {
1384 dmaengine_terminate_all(tqspi->tx_dma_chan);
1385 dev_err(tqspi->dev, "failed TX DMA transfer\n");
1386 err += 1;
1387 }
1388 }
1389 }
1390
1391 if (tqspi->cur_direction & DATA_DIR_RX) {
1392 if (tqspi->rx_status) {
1393 dmaengine_terminate_all(tqspi->rx_dma_chan);
1394 err += 2;
1395 } else {
1396 wait_status = wait_for_completion_interruptible_timeout(
1397 &tqspi->rx_dma_complete, QSPI_DMA_TIMEOUT);
1398 if (wait_status <= 0) {
1399 dmaengine_terminate_all(tqspi->rx_dma_chan);
1400 dev_err(tqspi->dev, "failed RX DMA transfer\n");
1401 err += 2;
1402 }
1403 }
1404 }
1405
1406 spin_lock_irqsave(&tqspi->lock, flags);
1407
1408 if (err) {
1409 tegra_qspi_dma_unmap_xfer(tqspi, t);
1410 tegra_qspi_handle_error(tqspi);
1411 complete(&tqspi->xfer_completion);
1412 goto exit;
1413 }
1414
1415 if (tqspi->cur_direction & DATA_DIR_RX)
1416 tegra_qspi_copy_qspi_rxbuf_to_client_rxbuf(tqspi, t);
1417
1418 if (tqspi->cur_direction & DATA_DIR_TX)
1419 tqspi->cur_pos = tqspi->cur_tx_pos;
1420 else
1421 tqspi->cur_pos = tqspi->cur_rx_pos;
1422
1423 if (tqspi->cur_pos == t->len) {
1424 tegra_qspi_dma_unmap_xfer(tqspi, t);
1425 complete(&tqspi->xfer_completion);
1426 goto exit;
1427 }
1428
1429 tegra_qspi_dma_unmap_xfer(tqspi, t);
1430
1431 /* continue transfer in current message */
1432 total_fifo_words = tegra_qspi_calculate_curr_xfer_param(tqspi, t);
1433 if (total_fifo_words > QSPI_FIFO_DEPTH)
1434 err = tegra_qspi_start_dma_based_transfer(tqspi, t);
1435 else
1436 err = tegra_qspi_start_cpu_based_transfer(tqspi, t);
1437
1438exit:
1439 spin_unlock_irqrestore(&tqspi->lock, flags);
1440 return IRQ_HANDLED;
1441}
1442
1443static irqreturn_t tegra_qspi_isr_thread(int irq, void *context_data)
1444{
1445 struct tegra_qspi *tqspi = context_data;
1446
1447 tqspi->status_reg = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
1448
1449 if (tqspi->cur_direction & DATA_DIR_TX)
1450 tqspi->tx_status = tqspi->status_reg & (QSPI_TX_FIFO_UNF | QSPI_TX_FIFO_OVF);
1451
1452 if (tqspi->cur_direction & DATA_DIR_RX)
1453 tqspi->rx_status = tqspi->status_reg & (QSPI_RX_FIFO_OVF | QSPI_RX_FIFO_UNF);
1454
1455 tegra_qspi_mask_clear_irq(tqspi);
1456
1457 if (!tqspi->is_curr_dma_xfer)
1458 return handle_cpu_based_xfer(tqspi);
1459
1460 return handle_dma_based_xfer(tqspi);
1461}
1462
1463static struct tegra_qspi_soc_data tegra210_qspi_soc_data = {
1464 .has_dma = true,
1465 .cmb_xfer_capable = false,
1466 .supports_tpm = false,
1467 .cs_count = 1,
1468};
1469
1470static struct tegra_qspi_soc_data tegra186_qspi_soc_data = {
1471 .has_dma = true,
1472 .cmb_xfer_capable = true,
1473 .supports_tpm = false,
1474 .cs_count = 1,
1475};
1476
1477static struct tegra_qspi_soc_data tegra234_qspi_soc_data = {
1478 .has_dma = false,
1479 .cmb_xfer_capable = true,
1480 .supports_tpm = true,
1481 .cs_count = 1,
1482};
1483
1484static struct tegra_qspi_soc_data tegra241_qspi_soc_data = {
1485 .has_dma = false,
1486 .cmb_xfer_capable = true,
1487 .supports_tpm = true,
1488 .cs_count = 4,
1489};
1490
1491static const struct of_device_id tegra_qspi_of_match[] = {
1492 {
1493 .compatible = "nvidia,tegra210-qspi",
1494 .data = &tegra210_qspi_soc_data,
1495 }, {
1496 .compatible = "nvidia,tegra186-qspi",
1497 .data = &tegra186_qspi_soc_data,
1498 }, {
1499 .compatible = "nvidia,tegra194-qspi",
1500 .data = &tegra186_qspi_soc_data,
1501 }, {
1502 .compatible = "nvidia,tegra234-qspi",
1503 .data = &tegra234_qspi_soc_data,
1504 }, {
1505 .compatible = "nvidia,tegra241-qspi",
1506 .data = &tegra241_qspi_soc_data,
1507 },
1508 {}
1509};
1510
1511MODULE_DEVICE_TABLE(of, tegra_qspi_of_match);
1512
1513#ifdef CONFIG_ACPI
1514static const struct acpi_device_id tegra_qspi_acpi_match[] = {
1515 {
1516 .id = "NVDA1213",
1517 .driver_data = (kernel_ulong_t)&tegra210_qspi_soc_data,
1518 }, {
1519 .id = "NVDA1313",
1520 .driver_data = (kernel_ulong_t)&tegra186_qspi_soc_data,
1521 }, {
1522 .id = "NVDA1413",
1523 .driver_data = (kernel_ulong_t)&tegra234_qspi_soc_data,
1524 }, {
1525 .id = "NVDA1513",
1526 .driver_data = (kernel_ulong_t)&tegra241_qspi_soc_data,
1527 },
1528 {}
1529};
1530
1531MODULE_DEVICE_TABLE(acpi, tegra_qspi_acpi_match);
1532#endif
1533
1534static int tegra_qspi_probe(struct platform_device *pdev)
1535{
1536 struct spi_controller *host;
1537 struct tegra_qspi *tqspi;
1538 struct resource *r;
1539 int ret, qspi_irq;
1540 int bus_num;
1541
1542 host = devm_spi_alloc_host(&pdev->dev, sizeof(*tqspi));
1543 if (!host)
1544 return -ENOMEM;
1545
1546 platform_set_drvdata(pdev, host);
1547 tqspi = spi_controller_get_devdata(host);
1548
1549 host->mode_bits = SPI_MODE_0 | SPI_MODE_3 | SPI_CS_HIGH |
1550 SPI_TX_DUAL | SPI_RX_DUAL | SPI_TX_QUAD | SPI_RX_QUAD;
1551 host->bits_per_word_mask = SPI_BPW_MASK(32) | SPI_BPW_MASK(16) | SPI_BPW_MASK(8);
1552 host->flags = SPI_CONTROLLER_HALF_DUPLEX;
1553 host->setup = tegra_qspi_setup;
1554 host->transfer_one_message = tegra_qspi_transfer_one_message;
1555 host->num_chipselect = 1;
1556 host->auto_runtime_pm = true;
1557
1558 bus_num = of_alias_get_id(pdev->dev.of_node, "spi");
1559 if (bus_num >= 0)
1560 host->bus_num = bus_num;
1561
1562 tqspi->host = host;
1563 tqspi->dev = &pdev->dev;
1564 spin_lock_init(&tqspi->lock);
1565
1566 tqspi->soc_data = device_get_match_data(&pdev->dev);
1567 host->num_chipselect = tqspi->soc_data->cs_count;
1568 tqspi->base = devm_platform_get_and_ioremap_resource(pdev, 0, &r);
1569 if (IS_ERR(tqspi->base))
1570 return PTR_ERR(tqspi->base);
1571
1572 tqspi->phys = r->start;
1573 qspi_irq = platform_get_irq(pdev, 0);
1574 if (qspi_irq < 0)
1575 return qspi_irq;
1576 tqspi->irq = qspi_irq;
1577
1578 if (!has_acpi_companion(tqspi->dev)) {
1579 tqspi->clk = devm_clk_get(&pdev->dev, "qspi");
1580 if (IS_ERR(tqspi->clk)) {
1581 ret = PTR_ERR(tqspi->clk);
1582 dev_err(&pdev->dev, "failed to get clock: %d\n", ret);
1583 return ret;
1584 }
1585
1586 }
1587
1588 tqspi->max_buf_size = QSPI_FIFO_DEPTH << 2;
1589 tqspi->dma_buf_size = DEFAULT_QSPI_DMA_BUF_LEN;
1590
1591 ret = tegra_qspi_init_dma(tqspi);
1592 if (ret < 0)
1593 return ret;
1594
1595 if (tqspi->use_dma)
1596 tqspi->max_buf_size = tqspi->dma_buf_size;
1597
1598 init_completion(&tqspi->tx_dma_complete);
1599 init_completion(&tqspi->rx_dma_complete);
1600 init_completion(&tqspi->xfer_completion);
1601
1602 pm_runtime_enable(&pdev->dev);
1603 ret = pm_runtime_resume_and_get(&pdev->dev);
1604 if (ret < 0) {
1605 dev_err(&pdev->dev, "failed to get runtime PM: %d\n", ret);
1606 goto exit_pm_disable;
1607 }
1608
1609 if (device_reset(tqspi->dev) < 0)
1610 dev_warn_once(tqspi->dev, "device reset failed\n");
1611
1612 tqspi->def_command1_reg = QSPI_M_S | QSPI_CS_SW_HW | QSPI_CS_SW_VAL;
1613 tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
1614 tqspi->spi_cs_timing1 = tegra_qspi_readl(tqspi, QSPI_CS_TIMING1);
1615 tqspi->spi_cs_timing2 = tegra_qspi_readl(tqspi, QSPI_CS_TIMING2);
1616 tqspi->def_command2_reg = tegra_qspi_readl(tqspi, QSPI_COMMAND2);
1617
1618 pm_runtime_put(&pdev->dev);
1619
1620 ret = request_threaded_irq(tqspi->irq, NULL,
1621 tegra_qspi_isr_thread, IRQF_ONESHOT,
1622 dev_name(&pdev->dev), tqspi);
1623 if (ret < 0) {
1624 dev_err(&pdev->dev, "failed to request IRQ#%u: %d\n", tqspi->irq, ret);
1625 goto exit_pm_disable;
1626 }
1627
1628 host->dev.of_node = pdev->dev.of_node;
1629 ret = spi_register_controller(host);
1630 if (ret < 0) {
1631 dev_err(&pdev->dev, "failed to register host: %d\n", ret);
1632 goto exit_free_irq;
1633 }
1634
1635 return 0;
1636
1637exit_free_irq:
1638 free_irq(qspi_irq, tqspi);
1639exit_pm_disable:
1640 pm_runtime_force_suspend(&pdev->dev);
1641 tegra_qspi_deinit_dma(tqspi);
1642 return ret;
1643}
1644
1645static void tegra_qspi_remove(struct platform_device *pdev)
1646{
1647 struct spi_controller *host = platform_get_drvdata(pdev);
1648 struct tegra_qspi *tqspi = spi_controller_get_devdata(host);
1649
1650 spi_unregister_controller(host);
1651 free_irq(tqspi->irq, tqspi);
1652 pm_runtime_force_suspend(&pdev->dev);
1653 tegra_qspi_deinit_dma(tqspi);
1654}
1655
1656static int __maybe_unused tegra_qspi_suspend(struct device *dev)
1657{
1658 struct spi_controller *host = dev_get_drvdata(dev);
1659
1660 return spi_controller_suspend(host);
1661}
1662
1663static int __maybe_unused tegra_qspi_resume(struct device *dev)
1664{
1665 struct spi_controller *host = dev_get_drvdata(dev);
1666 struct tegra_qspi *tqspi = spi_controller_get_devdata(host);
1667 int ret;
1668
1669 ret = pm_runtime_resume_and_get(dev);
1670 if (ret < 0) {
1671 dev_err(dev, "failed to get runtime PM: %d\n", ret);
1672 return ret;
1673 }
1674
1675 tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
1676 tegra_qspi_writel(tqspi, tqspi->def_command2_reg, QSPI_COMMAND2);
1677 pm_runtime_put(dev);
1678
1679 return spi_controller_resume(host);
1680}
1681
1682static int __maybe_unused tegra_qspi_runtime_suspend(struct device *dev)
1683{
1684 struct spi_controller *host = dev_get_drvdata(dev);
1685 struct tegra_qspi *tqspi = spi_controller_get_devdata(host);
1686
1687 /* Runtime pm disabled with ACPI */
1688 if (has_acpi_companion(tqspi->dev))
1689 return 0;
1690 /* flush all write which are in PPSB queue by reading back */
1691 tegra_qspi_readl(tqspi, QSPI_COMMAND1);
1692
1693 clk_disable_unprepare(tqspi->clk);
1694
1695 return 0;
1696}
1697
1698static int __maybe_unused tegra_qspi_runtime_resume(struct device *dev)
1699{
1700 struct spi_controller *host = dev_get_drvdata(dev);
1701 struct tegra_qspi *tqspi = spi_controller_get_devdata(host);
1702 int ret;
1703
1704 /* Runtime pm disabled with ACPI */
1705 if (has_acpi_companion(tqspi->dev))
1706 return 0;
1707 ret = clk_prepare_enable(tqspi->clk);
1708 if (ret < 0)
1709 dev_err(tqspi->dev, "failed to enable clock: %d\n", ret);
1710
1711 return ret;
1712}
1713
1714static const struct dev_pm_ops tegra_qspi_pm_ops = {
1715 SET_RUNTIME_PM_OPS(tegra_qspi_runtime_suspend, tegra_qspi_runtime_resume, NULL)
1716 SET_SYSTEM_SLEEP_PM_OPS(tegra_qspi_suspend, tegra_qspi_resume)
1717};
1718
1719static struct platform_driver tegra_qspi_driver = {
1720 .driver = {
1721 .name = "tegra-qspi",
1722 .pm = &tegra_qspi_pm_ops,
1723 .of_match_table = tegra_qspi_of_match,
1724 .acpi_match_table = ACPI_PTR(tegra_qspi_acpi_match),
1725 },
1726 .probe = tegra_qspi_probe,
1727 .remove_new = tegra_qspi_remove,
1728};
1729module_platform_driver(tegra_qspi_driver);
1730
1731MODULE_ALIAS("platform:qspi-tegra");
1732MODULE_DESCRIPTION("NVIDIA Tegra QSPI Controller Driver");
1733MODULE_AUTHOR("Sowjanya Komatineni <skomatineni@nvidia.com>");
1734MODULE_LICENSE("GPL v2");