1 /*
2  * Driver for the Micron P320 SSD
3  *   Copyright (C) 2011 Micron Technology, Inc.
4  *
5  * Portions of this code were derived from works subjected to the
6  * following copyright:
7  *    Copyright (C) 2009 Integrated Device Technology, Inc.
8  *
9  * This program is free software; you can redistribute it and/or modify
10  * it under the terms of the GNU General Public License as published by
11  * the Free Software Foundation; either version 2 of the License, or
12  * (at your option) any later version.
13  *
14  * This program is distributed in the hope that it will be useful,
15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
17  * GNU General Public License for more details.
18  *
19  */
20 
21 #include <linux/pci.h>
22 #include <linux/interrupt.h>
23 #include <linux/ata.h>
24 #include <linux/delay.h>
25 #include <linux/hdreg.h>
26 #include <linux/uaccess.h>
27 #include <linux/random.h>
28 #include <linux/smp.h>
29 #include <linux/compat.h>
30 #include <linux/fs.h>
31 #include <linux/module.h>
32 #include <linux/genhd.h>
33 #include <linux/blkdev.h>
34 #include <linux/blk-mq.h>
35 #include <linux/bio.h>
36 #include <linux/dma-mapping.h>
37 #include <linux/idr.h>
38 #include <linux/kthread.h>
39 #include <../drivers/ata/ahci.h>
40 #include <linux/export.h>
41 #include <linux/debugfs.h>
42 #include <linux/prefetch.h>
43 #include <linux/numa.h>
44 #include "mtip32xx.h"
45 
46 #define HW_CMD_SLOT_SZ		(MTIP_MAX_COMMAND_SLOTS * 32)
47 
48 /* DMA region containing RX Fis, Identify, RLE10, and SMART buffers */
49 #define AHCI_RX_FIS_SZ          0x100
50 #define AHCI_RX_FIS_OFFSET      0x0
51 #define AHCI_IDFY_SZ            ATA_SECT_SIZE
52 #define AHCI_IDFY_OFFSET        0x400
53 #define AHCI_SECTBUF_SZ         ATA_SECT_SIZE
54 #define AHCI_SECTBUF_OFFSET     0x800
55 #define AHCI_SMARTBUF_SZ        ATA_SECT_SIZE
56 #define AHCI_SMARTBUF_OFFSET    0xC00
57 /* 0x100 + 0x200 + 0x200 + 0x200 is smaller than 4k but we pad it out */
58 #define BLOCK_DMA_ALLOC_SZ      4096
59 
60 /* DMA region containing command table (should be 8192 bytes) */
61 #define AHCI_CMD_SLOT_SZ        sizeof(struct mtip_cmd_hdr)
62 #define AHCI_CMD_TBL_SZ         (MTIP_MAX_COMMAND_SLOTS * AHCI_CMD_SLOT_SZ)
63 #define AHCI_CMD_TBL_OFFSET     0x0
64 
65 /* DMA region per command (contains header and SGL) */
66 #define AHCI_CMD_TBL_HDR_SZ     0x80
67 #define AHCI_CMD_TBL_HDR_OFFSET 0x0
68 #define AHCI_CMD_TBL_SGL_SZ     (MTIP_MAX_SG * sizeof(struct mtip_cmd_sg))
69 #define AHCI_CMD_TBL_SGL_OFFSET AHCI_CMD_TBL_HDR_SZ
70 #define CMD_DMA_ALLOC_SZ        (AHCI_CMD_TBL_SGL_SZ + AHCI_CMD_TBL_HDR_SZ)
71 
72 
73 #define HOST_CAP_NZDMA		(1 << 19)
74 #define HOST_HSORG		0xFC
75 #define HSORG_DISABLE_SLOTGRP_INTR (1<<24)
76 #define HSORG_DISABLE_SLOTGRP_PXIS (1<<16)
77 #define HSORG_HWREV		0xFF00
78 #define HSORG_STYLE		0x8
79 #define HSORG_SLOTGROUPS	0x7
80 
81 #define PORT_COMMAND_ISSUE	0x38
82 #define PORT_SDBV		0x7C
83 
84 #define PORT_OFFSET		0x100
85 #define PORT_MEM_SIZE		0x80
86 
87 #define PORT_IRQ_ERR \
88 	(PORT_IRQ_HBUS_ERR | PORT_IRQ_IF_ERR | PORT_IRQ_CONNECT | \
89 	 PORT_IRQ_PHYRDY | PORT_IRQ_UNK_FIS | PORT_IRQ_BAD_PMP | \
90 	 PORT_IRQ_TF_ERR | PORT_IRQ_HBUS_DATA_ERR | PORT_IRQ_IF_NONFATAL | \
91 	 PORT_IRQ_OVERFLOW)
92 #define PORT_IRQ_LEGACY \
93 	(PORT_IRQ_PIOS_FIS | PORT_IRQ_D2H_REG_FIS)
94 #define PORT_IRQ_HANDLED \
95 	(PORT_IRQ_SDB_FIS | PORT_IRQ_LEGACY | \
96 	 PORT_IRQ_TF_ERR | PORT_IRQ_IF_ERR | \
97 	 PORT_IRQ_CONNECT | PORT_IRQ_PHYRDY)
98 #define DEF_PORT_IRQ \
99 	(PORT_IRQ_ERR | PORT_IRQ_LEGACY | PORT_IRQ_SDB_FIS)
100 
101 /* product numbers */
102 #define MTIP_PRODUCT_UNKNOWN	0x00
103 #define MTIP_PRODUCT_ASICFPGA	0x11
104 
105 /* Device instance number, incremented each time a device is probed. */
106 static int instance;
107 
108 static struct list_head online_list;
109 static struct list_head removing_list;
110 static spinlock_t dev_lock;
111 
112 /*
113  * Global variable used to hold the major block device number
114  * allocated in mtip_init().
115  */
116 static int mtip_major;
117 static struct dentry *dfs_parent;
118 static struct dentry *dfs_device_status;
119 
120 static u32 cpu_use[NR_CPUS];
121 
122 static DEFINE_IDA(rssd_index_ida);
123 
124 static int mtip_block_initialize(struct driver_data *dd);
125 
126 #ifdef CONFIG_COMPAT
127 struct mtip_compat_ide_task_request_s {
128 	__u8		io_ports[8];
129 	__u8		hob_ports[8];
130 	ide_reg_valid_t	out_flags;
131 	ide_reg_valid_t	in_flags;
132 	int		data_phase;
133 	int		req_cmd;
134 	compat_ulong_t	out_size;
135 	compat_ulong_t	in_size;
136 };
137 #endif
138 
139 /*
140  * This function check_for_surprise_removal is called
141  * while card is removed from the system and it will
142  * read the vendor id from the configration space
143  *
144  * @pdev Pointer to the pci_dev structure.
145  *
146  * return value
147  *	 true if device removed, else false
148  */
149 static bool mtip_check_surprise_removal(struct pci_dev *pdev)
150 {
151 	u16 vendor_id = 0;
152 	struct driver_data *dd = pci_get_drvdata(pdev);
153 
154 	if (dd->sr)
155 		return true;
156 
157        /* Read the vendorID from the configuration space */
158 	pci_read_config_word(pdev, 0x00, &vendor_id);
159 	if (vendor_id == 0xFFFF) {
160 		dd->sr = true;
161 		if (dd->queue)
162 			blk_queue_flag_set(QUEUE_FLAG_DEAD, dd->queue);
163 		else
164 			dev_warn(&dd->pdev->dev,
165 				"%s: dd->queue is NULL\n", __func__);
166 		return true; /* device removed */
167 	}
168 
169 	return false; /* device present */
170 }
171 
172 static struct mtip_cmd *mtip_cmd_from_tag(struct driver_data *dd,
173 					  unsigned int tag)
174 {
175 	struct blk_mq_hw_ctx *hctx = dd->queue->queue_hw_ctx[0];
176 
177 	return blk_mq_rq_to_pdu(blk_mq_tag_to_rq(hctx->tags, tag));
178 }
179 
180 /*
181  * Reset the HBA (without sleeping)
182  *
183  * @dd Pointer to the driver data structure.
184  *
185  * return value
186  *	0	The reset was successful.
187  *	-1	The HBA Reset bit did not clear.
188  */
189 static int mtip_hba_reset(struct driver_data *dd)
190 {
191 	unsigned long timeout;
192 
193 	/* Set the reset bit */
194 	writel(HOST_RESET, dd->mmio + HOST_CTL);
195 
196 	/* Flush */
197 	readl(dd->mmio + HOST_CTL);
198 
199 	/*
200 	 * Spin for up to 10 seconds waiting for reset acknowledgement. Spec
201 	 * is 1 sec but in LUN failure conditions, up to 10 secs are required
202 	 */
203 	timeout = jiffies + msecs_to_jiffies(10000);
204 	do {
205 		mdelay(10);
206 		if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag))
207 			return -1;
208 
209 	} while ((readl(dd->mmio + HOST_CTL) & HOST_RESET)
210 		 && time_before(jiffies, timeout));
211 
212 	if (readl(dd->mmio + HOST_CTL) & HOST_RESET)
213 		return -1;
214 
215 	return 0;
216 }
217 
218 /*
219  * Issue a command to the hardware.
220  *
221  * Set the appropriate bit in the s_active and Command Issue hardware
222  * registers, causing hardware command processing to begin.
223  *
224  * @port Pointer to the port structure.
225  * @tag  The tag of the command to be issued.
226  *
227  * return value
228  *      None
229  */
230 static inline void mtip_issue_ncq_command(struct mtip_port *port, int tag)
231 {
232 	int group = tag >> 5;
233 
234 	/* guard SACT and CI registers */
235 	spin_lock(&port->cmd_issue_lock[group]);
236 	writel((1 << MTIP_TAG_BIT(tag)),
237 			port->s_active[MTIP_TAG_INDEX(tag)]);
238 	writel((1 << MTIP_TAG_BIT(tag)),
239 			port->cmd_issue[MTIP_TAG_INDEX(tag)]);
240 	spin_unlock(&port->cmd_issue_lock[group]);
241 }
242 
243 /*
244  * Enable/disable the reception of FIS
245  *
246  * @port   Pointer to the port data structure
247  * @enable 1 to enable, 0 to disable
248  *
249  * return value
250  *	Previous state: 1 enabled, 0 disabled
251  */
252 static int mtip_enable_fis(struct mtip_port *port, int enable)
253 {
254 	u32 tmp;
255 
256 	/* enable FIS reception */
257 	tmp = readl(port->mmio + PORT_CMD);
258 	if (enable)
259 		writel(tmp | PORT_CMD_FIS_RX, port->mmio + PORT_CMD);
260 	else
261 		writel(tmp & ~PORT_CMD_FIS_RX, port->mmio + PORT_CMD);
262 
263 	/* Flush */
264 	readl(port->mmio + PORT_CMD);
265 
266 	return (((tmp & PORT_CMD_FIS_RX) == PORT_CMD_FIS_RX));
267 }
268 
269 /*
270  * Enable/disable the DMA engine
271  *
272  * @port   Pointer to the port data structure
273  * @enable 1 to enable, 0 to disable
274  *
275  * return value
276  *	Previous state: 1 enabled, 0 disabled.
277  */
278 static int mtip_enable_engine(struct mtip_port *port, int enable)
279 {
280 	u32 tmp;
281 
282 	/* enable FIS reception */
283 	tmp = readl(port->mmio + PORT_CMD);
284 	if (enable)
285 		writel(tmp | PORT_CMD_START, port->mmio + PORT_CMD);
286 	else
287 		writel(tmp & ~PORT_CMD_START, port->mmio + PORT_CMD);
288 
289 	readl(port->mmio + PORT_CMD);
290 	return (((tmp & PORT_CMD_START) == PORT_CMD_START));
291 }
292 
293 /*
294  * Enables the port DMA engine and FIS reception.
295  *
296  * return value
297  *	None
298  */
299 static inline void mtip_start_port(struct mtip_port *port)
300 {
301 	/* Enable FIS reception */
302 	mtip_enable_fis(port, 1);
303 
304 	/* Enable the DMA engine */
305 	mtip_enable_engine(port, 1);
306 }
307 
308 /*
309  * Deinitialize a port by disabling port interrupts, the DMA engine,
310  * and FIS reception.
311  *
312  * @port Pointer to the port structure
313  *
314  * return value
315  *	None
316  */
317 static inline void mtip_deinit_port(struct mtip_port *port)
318 {
319 	/* Disable interrupts on this port */
320 	writel(0, port->mmio + PORT_IRQ_MASK);
321 
322 	/* Disable the DMA engine */
323 	mtip_enable_engine(port, 0);
324 
325 	/* Disable FIS reception */
326 	mtip_enable_fis(port, 0);
327 }
328 
329 /*
330  * Initialize a port.
331  *
332  * This function deinitializes the port by calling mtip_deinit_port() and
333  * then initializes it by setting the command header and RX FIS addresses,
334  * clearing the SError register and any pending port interrupts before
335  * re-enabling the default set of port interrupts.
336  *
337  * @port Pointer to the port structure.
338  *
339  * return value
340  *	None
341  */
342 static void mtip_init_port(struct mtip_port *port)
343 {
344 	int i;
345 	mtip_deinit_port(port);
346 
347 	/* Program the command list base and FIS base addresses */
348 	if (readl(port->dd->mmio + HOST_CAP) & HOST_CAP_64) {
349 		writel((port->command_list_dma >> 16) >> 16,
350 			 port->mmio + PORT_LST_ADDR_HI);
351 		writel((port->rxfis_dma >> 16) >> 16,
352 			 port->mmio + PORT_FIS_ADDR_HI);
353 		set_bit(MTIP_PF_HOST_CAP_64, &port->flags);
354 	}
355 
356 	writel(port->command_list_dma & 0xFFFFFFFF,
357 			port->mmio + PORT_LST_ADDR);
358 	writel(port->rxfis_dma & 0xFFFFFFFF, port->mmio + PORT_FIS_ADDR);
359 
360 	/* Clear SError */
361 	writel(readl(port->mmio + PORT_SCR_ERR), port->mmio + PORT_SCR_ERR);
362 
363 	/* reset the completed registers.*/
364 	for (i = 0; i < port->dd->slot_groups; i++)
365 		writel(0xFFFFFFFF, port->completed[i]);
366 
367 	/* Clear any pending interrupts for this port */
368 	writel(readl(port->mmio + PORT_IRQ_STAT), port->mmio + PORT_IRQ_STAT);
369 
370 	/* Clear any pending interrupts on the HBA. */
371 	writel(readl(port->dd->mmio + HOST_IRQ_STAT),
372 					port->dd->mmio + HOST_IRQ_STAT);
373 
374 	/* Enable port interrupts */
375 	writel(DEF_PORT_IRQ, port->mmio + PORT_IRQ_MASK);
376 }
377 
378 /*
379  * Restart a port
380  *
381  * @port Pointer to the port data structure.
382  *
383  * return value
384  *	None
385  */
386 static void mtip_restart_port(struct mtip_port *port)
387 {
388 	unsigned long timeout;
389 
390 	/* Disable the DMA engine */
391 	mtip_enable_engine(port, 0);
392 
393 	/* Chip quirk: wait up to 500ms for PxCMD.CR == 0 */
394 	timeout = jiffies + msecs_to_jiffies(500);
395 	while ((readl(port->mmio + PORT_CMD) & PORT_CMD_LIST_ON)
396 		 && time_before(jiffies, timeout))
397 		;
398 
399 	if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &port->dd->dd_flag))
400 		return;
401 
402 	/*
403 	 * Chip quirk: escalate to hba reset if
404 	 * PxCMD.CR not clear after 500 ms
405 	 */
406 	if (readl(port->mmio + PORT_CMD) & PORT_CMD_LIST_ON) {
407 		dev_warn(&port->dd->pdev->dev,
408 			"PxCMD.CR not clear, escalating reset\n");
409 
410 		if (mtip_hba_reset(port->dd))
411 			dev_err(&port->dd->pdev->dev,
412 				"HBA reset escalation failed.\n");
413 
414 		/* 30 ms delay before com reset to quiesce chip */
415 		mdelay(30);
416 	}
417 
418 	dev_warn(&port->dd->pdev->dev, "Issuing COM reset\n");
419 
420 	/* Set PxSCTL.DET */
421 	writel(readl(port->mmio + PORT_SCR_CTL) |
422 			 1, port->mmio + PORT_SCR_CTL);
423 	readl(port->mmio + PORT_SCR_CTL);
424 
425 	/* Wait 1 ms to quiesce chip function */
426 	timeout = jiffies + msecs_to_jiffies(1);
427 	while (time_before(jiffies, timeout))
428 		;
429 
430 	if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &port->dd->dd_flag))
431 		return;
432 
433 	/* Clear PxSCTL.DET */
434 	writel(readl(port->mmio + PORT_SCR_CTL) & ~1,
435 			 port->mmio + PORT_SCR_CTL);
436 	readl(port->mmio + PORT_SCR_CTL);
437 
438 	/* Wait 500 ms for bit 0 of PORT_SCR_STS to be set */
439 	timeout = jiffies + msecs_to_jiffies(500);
440 	while (((readl(port->mmio + PORT_SCR_STAT) & 0x01) == 0)
441 			 && time_before(jiffies, timeout))
442 		;
443 
444 	if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &port->dd->dd_flag))
445 		return;
446 
447 	if ((readl(port->mmio + PORT_SCR_STAT) & 0x01) == 0)
448 		dev_warn(&port->dd->pdev->dev,
449 			"COM reset failed\n");
450 
451 	mtip_init_port(port);
452 	mtip_start_port(port);
453 
454 }
455 
456 static int mtip_device_reset(struct driver_data *dd)
457 {
458 	int rv = 0;
459 
460 	if (mtip_check_surprise_removal(dd->pdev))
461 		return 0;
462 
463 	if (mtip_hba_reset(dd) < 0)
464 		rv = -EFAULT;
465 
466 	mdelay(1);
467 	mtip_init_port(dd->port);
468 	mtip_start_port(dd->port);
469 
470 	/* Enable interrupts on the HBA. */
471 	writel(readl(dd->mmio + HOST_CTL) | HOST_IRQ_EN,
472 					dd->mmio + HOST_CTL);
473 	return rv;
474 }
475 
476 /*
477  * Helper function for tag logging
478  */
479 static void print_tags(struct driver_data *dd,
480 			char *msg,
481 			unsigned long *tagbits,
482 			int cnt)
483 {
484 	unsigned char tagmap[128];
485 	int group, tagmap_len = 0;
486 
487 	memset(tagmap, 0, sizeof(tagmap));
488 	for (group = SLOTBITS_IN_LONGS; group > 0; group--)
489 		tagmap_len += sprintf(tagmap + tagmap_len, "%016lX ",
490 						tagbits[group-1]);
491 	dev_warn(&dd->pdev->dev,
492 			"%d command(s) %s: tagmap [%s]", cnt, msg, tagmap);
493 }
494 
495 static int mtip_read_log_page(struct mtip_port *port, u8 page, u16 *buffer,
496 				dma_addr_t buffer_dma, unsigned int sectors);
497 static int mtip_get_smart_attr(struct mtip_port *port, unsigned int id,
498 						struct smart_attr *attrib);
499 
500 static void mtip_complete_command(struct mtip_cmd *cmd, blk_status_t status)
501 {
502 	struct request *req = blk_mq_rq_from_pdu(cmd);
503 
504 	cmd->status = status;
505 	blk_mq_complete_request(req);
506 }
507 
508 /*
509  * Handle an error.
510  *
511  * @dd Pointer to the DRIVER_DATA structure.
512  *
513  * return value
514  *	None
515  */
516 static void mtip_handle_tfe(struct driver_data *dd)
517 {
518 	int group, tag, bit, reissue, rv;
519 	struct mtip_port *port;
520 	struct mtip_cmd  *cmd;
521 	u32 completed;
522 	struct host_to_dev_fis *fis;
523 	unsigned long tagaccum[SLOTBITS_IN_LONGS];
524 	unsigned int cmd_cnt = 0;
525 	unsigned char *buf;
526 	char *fail_reason = NULL;
527 	int fail_all_ncq_write = 0, fail_all_ncq_cmds = 0;
528 
529 	dev_warn(&dd->pdev->dev, "Taskfile error\n");
530 
531 	port = dd->port;
532 
533 	if (test_bit(MTIP_PF_IC_ACTIVE_BIT, &port->flags)) {
534 		cmd = mtip_cmd_from_tag(dd, MTIP_TAG_INTERNAL);
535 		dbg_printk(MTIP_DRV_NAME " TFE for the internal command\n");
536 		mtip_complete_command(cmd, BLK_STS_IOERR);
537 		return;
538 	}
539 
540 	/* clear the tag accumulator */
541 	memset(tagaccum, 0, SLOTBITS_IN_LONGS * sizeof(long));
542 
543 	/* Loop through all the groups */
544 	for (group = 0; group < dd->slot_groups; group++) {
545 		completed = readl(port->completed[group]);
546 
547 		dev_warn(&dd->pdev->dev, "g=%u, comp=%x\n", group, completed);
548 
549 		/* clear completed status register in the hardware.*/
550 		writel(completed, port->completed[group]);
551 
552 		/* Process successfully completed commands */
553 		for (bit = 0; bit < 32 && completed; bit++) {
554 			if (!(completed & (1<<bit)))
555 				continue;
556 			tag = (group << 5) + bit;
557 
558 			/* Skip the internal command slot */
559 			if (tag == MTIP_TAG_INTERNAL)
560 				continue;
561 
562 			cmd = mtip_cmd_from_tag(dd, tag);
563 			mtip_complete_command(cmd, 0);
564 			set_bit(tag, tagaccum);
565 			cmd_cnt++;
566 		}
567 	}
568 
569 	print_tags(dd, "completed (TFE)", tagaccum, cmd_cnt);
570 
571 	/* Restart the port */
572 	mdelay(20);
573 	mtip_restart_port(port);
574 
575 	/* Trying to determine the cause of the error */
576 	rv = mtip_read_log_page(dd->port, ATA_LOG_SATA_NCQ,
577 				dd->port->log_buf,
578 				dd->port->log_buf_dma, 1);
579 	if (rv) {
580 		dev_warn(&dd->pdev->dev,
581 			"Error in READ LOG EXT (10h) command\n");
582 		/* non-critical error, don't fail the load */
583 	} else {
584 		buf = (unsigned char *)dd->port->log_buf;
585 		if (buf[259] & 0x1) {
586 			dev_info(&dd->pdev->dev,
587 				"Write protect bit is set.\n");
588 			set_bit(MTIP_DDF_WRITE_PROTECT_BIT, &dd->dd_flag);
589 			fail_all_ncq_write = 1;
590 			fail_reason = "write protect";
591 		}
592 		if (buf[288] == 0xF7) {
593 			dev_info(&dd->pdev->dev,
594 				"Exceeded Tmax, drive in thermal shutdown.\n");
595 			set_bit(MTIP_DDF_OVER_TEMP_BIT, &dd->dd_flag);
596 			fail_all_ncq_cmds = 1;
597 			fail_reason = "thermal shutdown";
598 		}
599 		if (buf[288] == 0xBF) {
600 			set_bit(MTIP_DDF_REBUILD_FAILED_BIT, &dd->dd_flag);
601 			dev_info(&dd->pdev->dev,
602 				"Drive indicates rebuild has failed. Secure erase required.\n");
603 			fail_all_ncq_cmds = 1;
604 			fail_reason = "rebuild failed";
605 		}
606 	}
607 
608 	/* clear the tag accumulator */
609 	memset(tagaccum, 0, SLOTBITS_IN_LONGS * sizeof(long));
610 
611 	/* Loop through all the groups */
612 	for (group = 0; group < dd->slot_groups; group++) {
613 		for (bit = 0; bit < 32; bit++) {
614 			reissue = 1;
615 			tag = (group << 5) + bit;
616 			cmd = mtip_cmd_from_tag(dd, tag);
617 
618 			fis = (struct host_to_dev_fis *)cmd->command;
619 
620 			/* Should re-issue? */
621 			if (tag == MTIP_TAG_INTERNAL ||
622 			    fis->command == ATA_CMD_SET_FEATURES)
623 				reissue = 0;
624 			else {
625 				if (fail_all_ncq_cmds ||
626 					(fail_all_ncq_write &&
627 					fis->command == ATA_CMD_FPDMA_WRITE)) {
628 					dev_warn(&dd->pdev->dev,
629 					"  Fail: %s w/tag %d [%s].\n",
630 					fis->command == ATA_CMD_FPDMA_WRITE ?
631 						"write" : "read",
632 					tag,
633 					fail_reason != NULL ?
634 						fail_reason : "unknown");
635 					mtip_complete_command(cmd, BLK_STS_MEDIUM);
636 					continue;
637 				}
638 			}
639 
640 			/*
641 			 * First check if this command has
642 			 *  exceeded its retries.
643 			 */
644 			if (reissue && (cmd->retries-- > 0)) {
645 
646 				set_bit(tag, tagaccum);
647 
648 				/* Re-issue the command. */
649 				mtip_issue_ncq_command(port, tag);
650 
651 				continue;
652 			}
653 
654 			/* Retire a command that will not be reissued */
655 			dev_warn(&port->dd->pdev->dev,
656 				"retiring tag %d\n", tag);
657 
658 			mtip_complete_command(cmd, BLK_STS_IOERR);
659 		}
660 	}
661 	print_tags(dd, "reissued (TFE)", tagaccum, cmd_cnt);
662 }
663 
664 /*
665  * Handle a set device bits interrupt
666  */
667 static inline void mtip_workq_sdbfx(struct mtip_port *port, int group,
668 							u32 completed)
669 {
670 	struct driver_data *dd = port->dd;
671 	int tag, bit;
672 	struct mtip_cmd *command;
673 
674 	if (!completed) {
675 		WARN_ON_ONCE(!completed);
676 		return;
677 	}
678 	/* clear completed status register in the hardware.*/
679 	writel(completed, port->completed[group]);
680 
681 	/* Process completed commands. */
682 	for (bit = 0; (bit < 32) && completed; bit++) {
683 		if (completed & 0x01) {
684 			tag = (group << 5) | bit;
685 
686 			/* skip internal command slot. */
687 			if (unlikely(tag == MTIP_TAG_INTERNAL))
688 				continue;
689 
690 			command = mtip_cmd_from_tag(dd, tag);
691 			mtip_complete_command(command, 0);
692 		}
693 		completed >>= 1;
694 	}
695 
696 	/* If last, re-enable interrupts */
697 	if (atomic_dec_return(&dd->irq_workers_active) == 0)
698 		writel(0xffffffff, dd->mmio + HOST_IRQ_STAT);
699 }
700 
701 /*
702  * Process legacy pio and d2h interrupts
703  */
704 static inline void mtip_process_legacy(struct driver_data *dd, u32 port_stat)
705 {
706 	struct mtip_port *port = dd->port;
707 	struct mtip_cmd *cmd = mtip_cmd_from_tag(dd, MTIP_TAG_INTERNAL);
708 
709 	if (test_bit(MTIP_PF_IC_ACTIVE_BIT, &port->flags) && cmd) {
710 		int group = MTIP_TAG_INDEX(MTIP_TAG_INTERNAL);
711 		int status = readl(port->cmd_issue[group]);
712 
713 		if (!(status & (1 << MTIP_TAG_BIT(MTIP_TAG_INTERNAL))))
714 			mtip_complete_command(cmd, 0);
715 	}
716 }
717 
718 /*
719  * Demux and handle errors
720  */
721 static inline void mtip_process_errors(struct driver_data *dd, u32 port_stat)
722 {
723 	if (unlikely(port_stat & PORT_IRQ_CONNECT)) {
724 		dev_warn(&dd->pdev->dev,
725 			"Clearing PxSERR.DIAG.x\n");
726 		writel((1 << 26), dd->port->mmio + PORT_SCR_ERR);
727 	}
728 
729 	if (unlikely(port_stat & PORT_IRQ_PHYRDY)) {
730 		dev_warn(&dd->pdev->dev,
731 			"Clearing PxSERR.DIAG.n\n");
732 		writel((1 << 16), dd->port->mmio + PORT_SCR_ERR);
733 	}
734 
735 	if (unlikely(port_stat & ~PORT_IRQ_HANDLED)) {
736 		dev_warn(&dd->pdev->dev,
737 			"Port stat errors %x unhandled\n",
738 			(port_stat & ~PORT_IRQ_HANDLED));
739 		if (mtip_check_surprise_removal(dd->pdev))
740 			return;
741 	}
742 	if (likely(port_stat & (PORT_IRQ_TF_ERR | PORT_IRQ_IF_ERR))) {
743 		set_bit(MTIP_PF_EH_ACTIVE_BIT, &dd->port->flags);
744 		wake_up_interruptible(&dd->port->svc_wait);
745 	}
746 }
747 
748 static inline irqreturn_t mtip_handle_irq(struct driver_data *data)
749 {
750 	struct driver_data *dd = (struct driver_data *) data;
751 	struct mtip_port *port = dd->port;
752 	u32 hba_stat, port_stat;
753 	int rv = IRQ_NONE;
754 	int do_irq_enable = 1, i, workers;
755 	struct mtip_work *twork;
756 
757 	hba_stat = readl(dd->mmio + HOST_IRQ_STAT);
758 	if (hba_stat) {
759 		rv = IRQ_HANDLED;
760 
761 		/* Acknowledge the interrupt status on the port.*/
762 		port_stat = readl(port->mmio + PORT_IRQ_STAT);
763 		if (unlikely(port_stat == 0xFFFFFFFF)) {
764 			mtip_check_surprise_removal(dd->pdev);
765 			return IRQ_HANDLED;
766 		}
767 		writel(port_stat, port->mmio + PORT_IRQ_STAT);
768 
769 		/* Demux port status */
770 		if (likely(port_stat & PORT_IRQ_SDB_FIS)) {
771 			do_irq_enable = 0;
772 			WARN_ON_ONCE(atomic_read(&dd->irq_workers_active) != 0);
773 
774 			/* Start at 1: group zero is always local? */
775 			for (i = 0, workers = 0; i < MTIP_MAX_SLOT_GROUPS;
776 									i++) {
777 				twork = &dd->work[i];
778 				twork->completed = readl(port->completed[i]);
779 				if (twork->completed)
780 					workers++;
781 			}
782 
783 			atomic_set(&dd->irq_workers_active, workers);
784 			if (workers) {
785 				for (i = 1; i < MTIP_MAX_SLOT_GROUPS; i++) {
786 					twork = &dd->work[i];
787 					if (twork->completed)
788 						queue_work_on(
789 							twork->cpu_binding,
790 							dd->isr_workq,
791 							&twork->work);
792 				}
793 
794 				if (likely(dd->work[0].completed))
795 					mtip_workq_sdbfx(port, 0,
796 							dd->work[0].completed);
797 
798 			} else {
799 				/*
800 				 * Chip quirk: SDB interrupt but nothing
801 				 * to complete
802 				 */
803 				do_irq_enable = 1;
804 			}
805 		}
806 
807 		if (unlikely(port_stat & PORT_IRQ_ERR)) {
808 			if (unlikely(mtip_check_surprise_removal(dd->pdev))) {
809 				/* don't proceed further */
810 				return IRQ_HANDLED;
811 			}
812 			if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
813 							&dd->dd_flag))
814 				return rv;
815 
816 			mtip_process_errors(dd, port_stat & PORT_IRQ_ERR);
817 		}
818 
819 		if (unlikely(port_stat & PORT_IRQ_LEGACY))
820 			mtip_process_legacy(dd, port_stat & PORT_IRQ_LEGACY);
821 	}
822 
823 	/* acknowledge interrupt */
824 	if (unlikely(do_irq_enable))
825 		writel(hba_stat, dd->mmio + HOST_IRQ_STAT);
826 
827 	return rv;
828 }
829 
830 /*
831  * HBA interrupt subroutine.
832  *
833  * @irq		IRQ number.
834  * @instance	Pointer to the driver data structure.
835  *
836  * return value
837  *	IRQ_HANDLED	A HBA interrupt was pending and handled.
838  *	IRQ_NONE	This interrupt was not for the HBA.
839  */
840 static irqreturn_t mtip_irq_handler(int irq, void *instance)
841 {
842 	struct driver_data *dd = instance;
843 
844 	return mtip_handle_irq(dd);
845 }
846 
847 static void mtip_issue_non_ncq_command(struct mtip_port *port, int tag)
848 {
849 	writel(1 << MTIP_TAG_BIT(tag), port->cmd_issue[MTIP_TAG_INDEX(tag)]);
850 }
851 
852 static bool mtip_pause_ncq(struct mtip_port *port,
853 				struct host_to_dev_fis *fis)
854 {
855 	unsigned long task_file_data;
856 
857 	task_file_data = readl(port->mmio+PORT_TFDATA);
858 	if ((task_file_data & 1))
859 		return false;
860 
861 	if (fis->command == ATA_CMD_SEC_ERASE_PREP) {
862 		port->ic_pause_timer = jiffies;
863 		return true;
864 	} else if ((fis->command == ATA_CMD_DOWNLOAD_MICRO) &&
865 					(fis->features == 0x03)) {
866 		set_bit(MTIP_PF_DM_ACTIVE_BIT, &port->flags);
867 		port->ic_pause_timer = jiffies;
868 		return true;
869 	} else if ((fis->command == ATA_CMD_SEC_ERASE_UNIT) ||
870 		((fis->command == 0xFC) &&
871 			(fis->features == 0x27 || fis->features == 0x72 ||
872 			 fis->features == 0x62 || fis->features == 0x26))) {
873 		clear_bit(MTIP_DDF_SEC_LOCK_BIT, &port->dd->dd_flag);
874 		clear_bit(MTIP_DDF_REBUILD_FAILED_BIT, &port->dd->dd_flag);
875 		/* Com reset after secure erase or lowlevel format */
876 		mtip_restart_port(port);
877 		clear_bit(MTIP_PF_SE_ACTIVE_BIT, &port->flags);
878 		return false;
879 	}
880 
881 	return false;
882 }
883 
884 static bool mtip_commands_active(struct mtip_port *port)
885 {
886 	unsigned int active;
887 	unsigned int n;
888 
889 	/*
890 	 * Ignore s_active bit 0 of array element 0.
891 	 * This bit will always be set
892 	 */
893 	active = readl(port->s_active[0]) & 0xFFFFFFFE;
894 	for (n = 1; n < port->dd->slot_groups; n++)
895 		active |= readl(port->s_active[n]);
896 
897 	return active != 0;
898 }
899 
900 /*
901  * Wait for port to quiesce
902  *
903  * @port    Pointer to port data structure
904  * @timeout Max duration to wait (ms)
905  *
906  * return value
907  *	0	Success
908  *	-EBUSY  Commands still active
909  */
910 static int mtip_quiesce_io(struct mtip_port *port, unsigned long timeout)
911 {
912 	unsigned long to;
913 	bool active = true;
914 
915 	blk_mq_quiesce_queue(port->dd->queue);
916 
917 	to = jiffies + msecs_to_jiffies(timeout);
918 	do {
919 		if (test_bit(MTIP_PF_SVC_THD_ACTIVE_BIT, &port->flags) &&
920 			test_bit(MTIP_PF_ISSUE_CMDS_BIT, &port->flags)) {
921 			msleep(20);
922 			continue; /* svc thd is actively issuing commands */
923 		}
924 
925 		msleep(100);
926 
927 		if (mtip_check_surprise_removal(port->dd->pdev))
928 			goto err_fault;
929 
930 		active = mtip_commands_active(port);
931 		if (!active)
932 			break;
933 	} while (time_before(jiffies, to));
934 
935 	blk_mq_unquiesce_queue(port->dd->queue);
936 	return active ? -EBUSY : 0;
937 err_fault:
938 	blk_mq_unquiesce_queue(port->dd->queue);
939 	return -EFAULT;
940 }
941 
942 struct mtip_int_cmd {
943 	int fis_len;
944 	dma_addr_t buffer;
945 	int buf_len;
946 	u32 opts;
947 };
948 
949 /*
950  * Execute an internal command and wait for the completion.
951  *
952  * @port    Pointer to the port data structure.
953  * @fis     Pointer to the FIS that describes the command.
954  * @fis_len  Length in WORDS of the FIS.
955  * @buffer  DMA accessible for command data.
956  * @buf_len  Length, in bytes, of the data buffer.
957  * @opts    Command header options, excluding the FIS length
958  *             and the number of PRD entries.
959  * @timeout Time in ms to wait for the command to complete.
960  *
961  * return value
962  *	0	 Command completed successfully.
963  *	-EFAULT  The buffer address is not correctly aligned.
964  *	-EBUSY   Internal command or other IO in progress.
965  *	-EAGAIN  Time out waiting for command to complete.
966  */
967 static int mtip_exec_internal_command(struct mtip_port *port,
968 					struct host_to_dev_fis *fis,
969 					int fis_len,
970 					dma_addr_t buffer,
971 					int buf_len,
972 					u32 opts,
973 					unsigned long timeout)
974 {
975 	struct mtip_cmd *int_cmd;
976 	struct driver_data *dd = port->dd;
977 	struct request *rq;
978 	struct mtip_int_cmd icmd = {
979 		.fis_len = fis_len,
980 		.buffer = buffer,
981 		.buf_len = buf_len,
982 		.opts = opts
983 	};
984 	int rv = 0;
985 
986 	/* Make sure the buffer is 8 byte aligned. This is asic specific. */
987 	if (buffer & 0x00000007) {
988 		dev_err(&dd->pdev->dev, "SG buffer is not 8 byte aligned\n");
989 		return -EFAULT;
990 	}
991 
992 	if (mtip_check_surprise_removal(dd->pdev))
993 		return -EFAULT;
994 
995 	rq = blk_mq_alloc_request(dd->queue, REQ_OP_DRV_IN, BLK_MQ_REQ_RESERVED);
996 	if (IS_ERR(rq)) {
997 		dbg_printk(MTIP_DRV_NAME "Unable to allocate tag for PIO cmd\n");
998 		return -EFAULT;
999 	}
1000 
1001 	set_bit(MTIP_PF_IC_ACTIVE_BIT, &port->flags);
1002 
1003 	if (fis->command == ATA_CMD_SEC_ERASE_PREP)
1004 		set_bit(MTIP_PF_SE_ACTIVE_BIT, &port->flags);
1005 
1006 	clear_bit(MTIP_PF_DM_ACTIVE_BIT, &port->flags);
1007 
1008 	if (fis->command != ATA_CMD_STANDBYNOW1) {
1009 		/* wait for io to complete if non atomic */
1010 		if (mtip_quiesce_io(port, MTIP_QUIESCE_IO_TIMEOUT_MS) < 0) {
1011 			dev_warn(&dd->pdev->dev, "Failed to quiesce IO\n");
1012 			blk_mq_free_request(rq);
1013 			clear_bit(MTIP_PF_IC_ACTIVE_BIT, &port->flags);
1014 			wake_up_interruptible(&port->svc_wait);
1015 			return -EBUSY;
1016 		}
1017 	}
1018 
1019 	/* Copy the command to the command table */
1020 	int_cmd = blk_mq_rq_to_pdu(rq);
1021 	int_cmd->icmd = &icmd;
1022 	memcpy(int_cmd->command, fis, fis_len*4);
1023 
1024 	rq->timeout = timeout;
1025 
1026 	/* insert request and run queue */
1027 	blk_execute_rq(rq->q, NULL, rq, true);
1028 
1029 	if (int_cmd->status) {
1030 		dev_err(&dd->pdev->dev, "Internal command [%02X] failed %d\n",
1031 				fis->command, int_cmd->status);
1032 		rv = -EIO;
1033 
1034 		if (mtip_check_surprise_removal(dd->pdev) ||
1035 			test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
1036 					&dd->dd_flag)) {
1037 			dev_err(&dd->pdev->dev,
1038 				"Internal command [%02X] wait returned due to SR\n",
1039 				fis->command);
1040 			rv = -ENXIO;
1041 			goto exec_ic_exit;
1042 		}
1043 		mtip_device_reset(dd); /* recover from timeout issue */
1044 		rv = -EAGAIN;
1045 		goto exec_ic_exit;
1046 	}
1047 
1048 	if (readl(port->cmd_issue[MTIP_TAG_INDEX(MTIP_TAG_INTERNAL)])
1049 			& (1 << MTIP_TAG_BIT(MTIP_TAG_INTERNAL))) {
1050 		rv = -ENXIO;
1051 		if (!test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)) {
1052 			mtip_device_reset(dd);
1053 			rv = -EAGAIN;
1054 		}
1055 	}
1056 exec_ic_exit:
1057 	/* Clear the allocated and active bits for the internal command. */
1058 	blk_mq_free_request(rq);
1059 	clear_bit(MTIP_PF_IC_ACTIVE_BIT, &port->flags);
1060 	if (rv >= 0 && mtip_pause_ncq(port, fis)) {
1061 		/* NCQ paused */
1062 		return rv;
1063 	}
1064 	wake_up_interruptible(&port->svc_wait);
1065 
1066 	return rv;
1067 }
1068 
1069 /*
1070  * Byte-swap ATA ID strings.
1071  *
1072  * ATA identify data contains strings in byte-swapped 16-bit words.
1073  * They must be swapped (on all architectures) to be usable as C strings.
1074  * This function swaps bytes in-place.
1075  *
1076  * @buf The buffer location of the string
1077  * @len The number of bytes to swap
1078  *
1079  * return value
1080  *	None
1081  */
1082 static inline void ata_swap_string(u16 *buf, unsigned int len)
1083 {
1084 	int i;
1085 	for (i = 0; i < (len/2); i++)
1086 		be16_to_cpus(&buf[i]);
1087 }
1088 
1089 static void mtip_set_timeout(struct driver_data *dd,
1090 					struct host_to_dev_fis *fis,
1091 					unsigned int *timeout, u8 erasemode)
1092 {
1093 	switch (fis->command) {
1094 	case ATA_CMD_DOWNLOAD_MICRO:
1095 		*timeout = 120000; /* 2 minutes */
1096 		break;
1097 	case ATA_CMD_SEC_ERASE_UNIT:
1098 	case 0xFC:
1099 		if (erasemode)
1100 			*timeout = ((*(dd->port->identify + 90) * 2) * 60000);
1101 		else
1102 			*timeout = ((*(dd->port->identify + 89) * 2) * 60000);
1103 		break;
1104 	case ATA_CMD_STANDBYNOW1:
1105 		*timeout = 120000;  /* 2 minutes */
1106 		break;
1107 	case 0xF7:
1108 	case 0xFA:
1109 		*timeout = 60000;  /* 60 seconds */
1110 		break;
1111 	case ATA_CMD_SMART:
1112 		*timeout = 15000;  /* 15 seconds */
1113 		break;
1114 	default:
1115 		*timeout = MTIP_IOCTL_CMD_TIMEOUT_MS;
1116 		break;
1117 	}
1118 }
1119 
1120 /*
1121  * Request the device identity information.
1122  *
1123  * If a user space buffer is not specified, i.e. is NULL, the
1124  * identify information is still read from the drive and placed
1125  * into the identify data buffer (@e port->identify) in the
1126  * port data structure.
1127  * When the identify buffer contains valid identify information @e
1128  * port->identify_valid is non-zero.
1129  *
1130  * @port	 Pointer to the port structure.
1131  * @user_buffer  A user space buffer where the identify data should be
1132  *                    copied.
1133  *
1134  * return value
1135  *	0	Command completed successfully.
1136  *	-EFAULT An error occurred while coping data to the user buffer.
1137  *	-1	Command failed.
1138  */
1139 static int mtip_get_identify(struct mtip_port *port, void __user *user_buffer)
1140 {
1141 	int rv = 0;
1142 	struct host_to_dev_fis fis;
1143 
1144 	if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &port->dd->dd_flag))
1145 		return -EFAULT;
1146 
1147 	/* Build the FIS. */
1148 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1149 	fis.type	= 0x27;
1150 	fis.opts	= 1 << 7;
1151 	fis.command	= ATA_CMD_ID_ATA;
1152 
1153 	/* Set the identify information as invalid. */
1154 	port->identify_valid = 0;
1155 
1156 	/* Clear the identify information. */
1157 	memset(port->identify, 0, sizeof(u16) * ATA_ID_WORDS);
1158 
1159 	/* Execute the command. */
1160 	if (mtip_exec_internal_command(port,
1161 				&fis,
1162 				5,
1163 				port->identify_dma,
1164 				sizeof(u16) * ATA_ID_WORDS,
1165 				0,
1166 				MTIP_INT_CMD_TIMEOUT_MS)
1167 				< 0) {
1168 		rv = -1;
1169 		goto out;
1170 	}
1171 
1172 	/*
1173 	 * Perform any necessary byte-swapping.  Yes, the kernel does in fact
1174 	 * perform field-sensitive swapping on the string fields.
1175 	 * See the kernel use of ata_id_string() for proof of this.
1176 	 */
1177 #ifdef __LITTLE_ENDIAN
1178 	ata_swap_string(port->identify + 27, 40);  /* model string*/
1179 	ata_swap_string(port->identify + 23, 8);   /* firmware string*/
1180 	ata_swap_string(port->identify + 10, 20);  /* serial# string*/
1181 #else
1182 	{
1183 		int i;
1184 		for (i = 0; i < ATA_ID_WORDS; i++)
1185 			port->identify[i] = le16_to_cpu(port->identify[i]);
1186 	}
1187 #endif
1188 
1189 	/* Check security locked state */
1190 	if (port->identify[128] & 0x4)
1191 		set_bit(MTIP_DDF_SEC_LOCK_BIT, &port->dd->dd_flag);
1192 	else
1193 		clear_bit(MTIP_DDF_SEC_LOCK_BIT, &port->dd->dd_flag);
1194 
1195 	/* Set the identify buffer as valid. */
1196 	port->identify_valid = 1;
1197 
1198 	if (user_buffer) {
1199 		if (copy_to_user(
1200 			user_buffer,
1201 			port->identify,
1202 			ATA_ID_WORDS * sizeof(u16))) {
1203 			rv = -EFAULT;
1204 			goto out;
1205 		}
1206 	}
1207 
1208 out:
1209 	return rv;
1210 }
1211 
1212 /*
1213  * Issue a standby immediate command to the device.
1214  *
1215  * @port Pointer to the port structure.
1216  *
1217  * return value
1218  *	0	Command was executed successfully.
1219  *	-1	An error occurred while executing the command.
1220  */
1221 static int mtip_standby_immediate(struct mtip_port *port)
1222 {
1223 	int rv;
1224 	struct host_to_dev_fis	fis;
1225 	unsigned long start;
1226 	unsigned int timeout;
1227 
1228 	/* Build the FIS. */
1229 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1230 	fis.type	= 0x27;
1231 	fis.opts	= 1 << 7;
1232 	fis.command	= ATA_CMD_STANDBYNOW1;
1233 
1234 	mtip_set_timeout(port->dd, &fis, &timeout, 0);
1235 
1236 	start = jiffies;
1237 	rv = mtip_exec_internal_command(port,
1238 					&fis,
1239 					5,
1240 					0,
1241 					0,
1242 					0,
1243 					timeout);
1244 	dbg_printk(MTIP_DRV_NAME "Time taken to complete standby cmd: %d ms\n",
1245 			jiffies_to_msecs(jiffies - start));
1246 	if (rv)
1247 		dev_warn(&port->dd->pdev->dev,
1248 			"STANDBY IMMEDIATE command failed.\n");
1249 
1250 	return rv;
1251 }
1252 
1253 /*
1254  * Issue a READ LOG EXT command to the device.
1255  *
1256  * @port	pointer to the port structure.
1257  * @page	page number to fetch
1258  * @buffer	pointer to buffer
1259  * @buffer_dma	dma address corresponding to @buffer
1260  * @sectors	page length to fetch, in sectors
1261  *
1262  * return value
1263  *	@rv	return value from mtip_exec_internal_command()
1264  */
1265 static int mtip_read_log_page(struct mtip_port *port, u8 page, u16 *buffer,
1266 				dma_addr_t buffer_dma, unsigned int sectors)
1267 {
1268 	struct host_to_dev_fis fis;
1269 
1270 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1271 	fis.type	= 0x27;
1272 	fis.opts	= 1 << 7;
1273 	fis.command	= ATA_CMD_READ_LOG_EXT;
1274 	fis.sect_count	= sectors & 0xFF;
1275 	fis.sect_cnt_ex	= (sectors >> 8) & 0xFF;
1276 	fis.lba_low	= page;
1277 	fis.lba_mid	= 0;
1278 	fis.device	= ATA_DEVICE_OBS;
1279 
1280 	memset(buffer, 0, sectors * ATA_SECT_SIZE);
1281 
1282 	return mtip_exec_internal_command(port,
1283 					&fis,
1284 					5,
1285 					buffer_dma,
1286 					sectors * ATA_SECT_SIZE,
1287 					0,
1288 					MTIP_INT_CMD_TIMEOUT_MS);
1289 }
1290 
1291 /*
1292  * Issue a SMART READ DATA command to the device.
1293  *
1294  * @port	pointer to the port structure.
1295  * @buffer	pointer to buffer
1296  * @buffer_dma	dma address corresponding to @buffer
1297  *
1298  * return value
1299  *	@rv	return value from mtip_exec_internal_command()
1300  */
1301 static int mtip_get_smart_data(struct mtip_port *port, u8 *buffer,
1302 					dma_addr_t buffer_dma)
1303 {
1304 	struct host_to_dev_fis fis;
1305 
1306 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1307 	fis.type	= 0x27;
1308 	fis.opts	= 1 << 7;
1309 	fis.command	= ATA_CMD_SMART;
1310 	fis.features	= 0xD0;
1311 	fis.sect_count	= 1;
1312 	fis.lba_mid	= 0x4F;
1313 	fis.lba_hi	= 0xC2;
1314 	fis.device	= ATA_DEVICE_OBS;
1315 
1316 	return mtip_exec_internal_command(port,
1317 					&fis,
1318 					5,
1319 					buffer_dma,
1320 					ATA_SECT_SIZE,
1321 					0,
1322 					15000);
1323 }
1324 
1325 /*
1326  * Get the value of a smart attribute
1327  *
1328  * @port	pointer to the port structure
1329  * @id		attribute number
1330  * @attrib	pointer to return attrib information corresponding to @id
1331  *
1332  * return value
1333  *	-EINVAL	NULL buffer passed or unsupported attribute @id.
1334  *	-EPERM	Identify data not valid, SMART not supported or not enabled
1335  */
1336 static int mtip_get_smart_attr(struct mtip_port *port, unsigned int id,
1337 						struct smart_attr *attrib)
1338 {
1339 	int rv, i;
1340 	struct smart_attr *pattr;
1341 
1342 	if (!attrib)
1343 		return -EINVAL;
1344 
1345 	if (!port->identify_valid) {
1346 		dev_warn(&port->dd->pdev->dev, "IDENTIFY DATA not valid\n");
1347 		return -EPERM;
1348 	}
1349 	if (!(port->identify[82] & 0x1)) {
1350 		dev_warn(&port->dd->pdev->dev, "SMART not supported\n");
1351 		return -EPERM;
1352 	}
1353 	if (!(port->identify[85] & 0x1)) {
1354 		dev_warn(&port->dd->pdev->dev, "SMART not enabled\n");
1355 		return -EPERM;
1356 	}
1357 
1358 	memset(port->smart_buf, 0, ATA_SECT_SIZE);
1359 	rv = mtip_get_smart_data(port, port->smart_buf, port->smart_buf_dma);
1360 	if (rv) {
1361 		dev_warn(&port->dd->pdev->dev, "Failed to ge SMART data\n");
1362 		return rv;
1363 	}
1364 
1365 	pattr = (struct smart_attr *)(port->smart_buf + 2);
1366 	for (i = 0; i < 29; i++, pattr++)
1367 		if (pattr->attr_id == id) {
1368 			memcpy(attrib, pattr, sizeof(struct smart_attr));
1369 			break;
1370 		}
1371 
1372 	if (i == 29) {
1373 		dev_warn(&port->dd->pdev->dev,
1374 			"Query for invalid SMART attribute ID\n");
1375 		rv = -EINVAL;
1376 	}
1377 
1378 	return rv;
1379 }
1380 
1381 /*
1382  * Get the drive capacity.
1383  *
1384  * @dd      Pointer to the device data structure.
1385  * @sectors Pointer to the variable that will receive the sector count.
1386  *
1387  * return value
1388  *	1 Capacity was returned successfully.
1389  *	0 The identify information is invalid.
1390  */
1391 static bool mtip_hw_get_capacity(struct driver_data *dd, sector_t *sectors)
1392 {
1393 	struct mtip_port *port = dd->port;
1394 	u64 total, raw0, raw1, raw2, raw3;
1395 	raw0 = port->identify[100];
1396 	raw1 = port->identify[101];
1397 	raw2 = port->identify[102];
1398 	raw3 = port->identify[103];
1399 	total = raw0 | raw1<<16 | raw2<<32 | raw3<<48;
1400 	*sectors = total;
1401 	return (bool) !!port->identify_valid;
1402 }
1403 
1404 /*
1405  * Display the identify command data.
1406  *
1407  * @port Pointer to the port data structure.
1408  *
1409  * return value
1410  *	None
1411  */
1412 static void mtip_dump_identify(struct mtip_port *port)
1413 {
1414 	sector_t sectors;
1415 	unsigned short revid;
1416 	char cbuf[42];
1417 
1418 	if (!port->identify_valid)
1419 		return;
1420 
1421 	strlcpy(cbuf, (char *)(port->identify+10), 21);
1422 	dev_info(&port->dd->pdev->dev,
1423 		"Serial No.: %s\n", cbuf);
1424 
1425 	strlcpy(cbuf, (char *)(port->identify+23), 9);
1426 	dev_info(&port->dd->pdev->dev,
1427 		"Firmware Ver.: %s\n", cbuf);
1428 
1429 	strlcpy(cbuf, (char *)(port->identify+27), 41);
1430 	dev_info(&port->dd->pdev->dev, "Model: %s\n", cbuf);
1431 
1432 	dev_info(&port->dd->pdev->dev, "Security: %04x %s\n",
1433 		port->identify[128],
1434 		port->identify[128] & 0x4 ? "(LOCKED)" : "");
1435 
1436 	if (mtip_hw_get_capacity(port->dd, &sectors))
1437 		dev_info(&port->dd->pdev->dev,
1438 			"Capacity: %llu sectors (%llu MB)\n",
1439 			 (u64)sectors,
1440 			 ((u64)sectors) * ATA_SECT_SIZE >> 20);
1441 
1442 	pci_read_config_word(port->dd->pdev, PCI_REVISION_ID, &revid);
1443 	switch (revid & 0xFF) {
1444 	case 0x1:
1445 		strlcpy(cbuf, "A0", 3);
1446 		break;
1447 	case 0x3:
1448 		strlcpy(cbuf, "A2", 3);
1449 		break;
1450 	default:
1451 		strlcpy(cbuf, "?", 2);
1452 		break;
1453 	}
1454 	dev_info(&port->dd->pdev->dev,
1455 		"Card Type: %s\n", cbuf);
1456 }
1457 
1458 /*
1459  * Map the commands scatter list into the command table.
1460  *
1461  * @command Pointer to the command.
1462  * @nents Number of scatter list entries.
1463  *
1464  * return value
1465  *	None
1466  */
1467 static inline void fill_command_sg(struct driver_data *dd,
1468 				struct mtip_cmd *command,
1469 				int nents)
1470 {
1471 	int n;
1472 	unsigned int dma_len;
1473 	struct mtip_cmd_sg *command_sg;
1474 	struct scatterlist *sg;
1475 
1476 	command_sg = command->command + AHCI_CMD_TBL_HDR_SZ;
1477 
1478 	for_each_sg(command->sg, sg, nents, n) {
1479 		dma_len = sg_dma_len(sg);
1480 		if (dma_len > 0x400000)
1481 			dev_err(&dd->pdev->dev,
1482 				"DMA segment length truncated\n");
1483 		command_sg->info = cpu_to_le32((dma_len-1) & 0x3FFFFF);
1484 		command_sg->dba	=  cpu_to_le32(sg_dma_address(sg));
1485 		command_sg->dba_upper =
1486 			cpu_to_le32((sg_dma_address(sg) >> 16) >> 16);
1487 		command_sg++;
1488 	}
1489 }
1490 
1491 /*
1492  * @brief Execute a drive command.
1493  *
1494  * return value 0 The command completed successfully.
1495  * return value -1 An error occurred while executing the command.
1496  */
1497 static int exec_drive_task(struct mtip_port *port, u8 *command)
1498 {
1499 	struct host_to_dev_fis	fis;
1500 	struct host_to_dev_fis *reply = (port->rxfis + RX_FIS_D2H_REG);
1501 	unsigned int to;
1502 
1503 	/* Build the FIS. */
1504 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1505 	fis.type	= 0x27;
1506 	fis.opts	= 1 << 7;
1507 	fis.command	= command[0];
1508 	fis.features	= command[1];
1509 	fis.sect_count	= command[2];
1510 	fis.sector	= command[3];
1511 	fis.cyl_low	= command[4];
1512 	fis.cyl_hi	= command[5];
1513 	fis.device	= command[6] & ~0x10; /* Clear the dev bit*/
1514 
1515 	mtip_set_timeout(port->dd, &fis, &to, 0);
1516 
1517 	dbg_printk(MTIP_DRV_NAME " %s: User Command: cmd %x, feat %x, nsect %x, sect %x, lcyl %x, hcyl %x, sel %x\n",
1518 		__func__,
1519 		command[0],
1520 		command[1],
1521 		command[2],
1522 		command[3],
1523 		command[4],
1524 		command[5],
1525 		command[6]);
1526 
1527 	/* Execute the command. */
1528 	if (mtip_exec_internal_command(port,
1529 				 &fis,
1530 				 5,
1531 				 0,
1532 				 0,
1533 				 0,
1534 				 to) < 0) {
1535 		return -1;
1536 	}
1537 
1538 	command[0] = reply->command; /* Status*/
1539 	command[1] = reply->features; /* Error*/
1540 	command[4] = reply->cyl_low;
1541 	command[5] = reply->cyl_hi;
1542 
1543 	dbg_printk(MTIP_DRV_NAME " %s: Completion Status: stat %x, err %x , cyl_lo %x cyl_hi %x\n",
1544 		__func__,
1545 		command[0],
1546 		command[1],
1547 		command[4],
1548 		command[5]);
1549 
1550 	return 0;
1551 }
1552 
1553 /*
1554  * @brief Execute a drive command.
1555  *
1556  * @param port Pointer to the port data structure.
1557  * @param command Pointer to the user specified command parameters.
1558  * @param user_buffer Pointer to the user space buffer where read sector
1559  *                   data should be copied.
1560  *
1561  * return value 0 The command completed successfully.
1562  * return value -EFAULT An error occurred while copying the completion
1563  *                 data to the user space buffer.
1564  * return value -1 An error occurred while executing the command.
1565  */
1566 static int exec_drive_command(struct mtip_port *port, u8 *command,
1567 				void __user *user_buffer)
1568 {
1569 	struct host_to_dev_fis	fis;
1570 	struct host_to_dev_fis *reply;
1571 	u8 *buf = NULL;
1572 	dma_addr_t dma_addr = 0;
1573 	int rv = 0, xfer_sz = command[3];
1574 	unsigned int to;
1575 
1576 	if (xfer_sz) {
1577 		if (!user_buffer)
1578 			return -EFAULT;
1579 
1580 		buf = dma_alloc_coherent(&port->dd->pdev->dev,
1581 				ATA_SECT_SIZE * xfer_sz,
1582 				&dma_addr,
1583 				GFP_KERNEL);
1584 		if (!buf) {
1585 			dev_err(&port->dd->pdev->dev,
1586 				"Memory allocation failed (%d bytes)\n",
1587 				ATA_SECT_SIZE * xfer_sz);
1588 			return -ENOMEM;
1589 		}
1590 		memset(buf, 0, ATA_SECT_SIZE * xfer_sz);
1591 	}
1592 
1593 	/* Build the FIS. */
1594 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1595 	fis.type	= 0x27;
1596 	fis.opts	= 1 << 7;
1597 	fis.command	= command[0];
1598 	fis.features	= command[2];
1599 	fis.sect_count	= command[3];
1600 	if (fis.command == ATA_CMD_SMART) {
1601 		fis.sector	= command[1];
1602 		fis.cyl_low	= 0x4F;
1603 		fis.cyl_hi	= 0xC2;
1604 	}
1605 
1606 	mtip_set_timeout(port->dd, &fis, &to, 0);
1607 
1608 	if (xfer_sz)
1609 		reply = (port->rxfis + RX_FIS_PIO_SETUP);
1610 	else
1611 		reply = (port->rxfis + RX_FIS_D2H_REG);
1612 
1613 	dbg_printk(MTIP_DRV_NAME
1614 		" %s: User Command: cmd %x, sect %x, "
1615 		"feat %x, sectcnt %x\n",
1616 		__func__,
1617 		command[0],
1618 		command[1],
1619 		command[2],
1620 		command[3]);
1621 
1622 	/* Execute the command. */
1623 	if (mtip_exec_internal_command(port,
1624 				&fis,
1625 				 5,
1626 				 (xfer_sz ? dma_addr : 0),
1627 				 (xfer_sz ? ATA_SECT_SIZE * xfer_sz : 0),
1628 				 0,
1629 				 to)
1630 				 < 0) {
1631 		rv = -EFAULT;
1632 		goto exit_drive_command;
1633 	}
1634 
1635 	/* Collect the completion status. */
1636 	command[0] = reply->command; /* Status*/
1637 	command[1] = reply->features; /* Error*/
1638 	command[2] = reply->sect_count;
1639 
1640 	dbg_printk(MTIP_DRV_NAME
1641 		" %s: Completion Status: stat %x, "
1642 		"err %x, nsect %x\n",
1643 		__func__,
1644 		command[0],
1645 		command[1],
1646 		command[2]);
1647 
1648 	if (xfer_sz) {
1649 		if (copy_to_user(user_buffer,
1650 				 buf,
1651 				 ATA_SECT_SIZE * command[3])) {
1652 			rv = -EFAULT;
1653 			goto exit_drive_command;
1654 		}
1655 	}
1656 exit_drive_command:
1657 	if (buf)
1658 		dma_free_coherent(&port->dd->pdev->dev,
1659 				ATA_SECT_SIZE * xfer_sz, buf, dma_addr);
1660 	return rv;
1661 }
1662 
1663 /*
1664  *  Indicates whether a command has a single sector payload.
1665  *
1666  *  @command passed to the device to perform the certain event.
1667  *  @features passed to the device to perform the certain event.
1668  *
1669  *  return value
1670  *	1	command is one that always has a single sector payload,
1671  *		regardless of the value in the Sector Count field.
1672  *      0       otherwise
1673  *
1674  */
1675 static unsigned int implicit_sector(unsigned char command,
1676 				    unsigned char features)
1677 {
1678 	unsigned int rv = 0;
1679 
1680 	/* list of commands that have an implicit sector count of 1 */
1681 	switch (command) {
1682 	case ATA_CMD_SEC_SET_PASS:
1683 	case ATA_CMD_SEC_UNLOCK:
1684 	case ATA_CMD_SEC_ERASE_PREP:
1685 	case ATA_CMD_SEC_ERASE_UNIT:
1686 	case ATA_CMD_SEC_FREEZE_LOCK:
1687 	case ATA_CMD_SEC_DISABLE_PASS:
1688 	case ATA_CMD_PMP_READ:
1689 	case ATA_CMD_PMP_WRITE:
1690 		rv = 1;
1691 		break;
1692 	case ATA_CMD_SET_MAX:
1693 		if (features == ATA_SET_MAX_UNLOCK)
1694 			rv = 1;
1695 		break;
1696 	case ATA_CMD_SMART:
1697 		if ((features == ATA_SMART_READ_VALUES) ||
1698 				(features == ATA_SMART_READ_THRESHOLDS))
1699 			rv = 1;
1700 		break;
1701 	case ATA_CMD_CONF_OVERLAY:
1702 		if ((features == ATA_DCO_IDENTIFY) ||
1703 				(features == ATA_DCO_SET))
1704 			rv = 1;
1705 		break;
1706 	}
1707 	return rv;
1708 }
1709 
1710 /*
1711  * Executes a taskfile
1712  * See ide_taskfile_ioctl() for derivation
1713  */
1714 static int exec_drive_taskfile(struct driver_data *dd,
1715 			       void __user *buf,
1716 			       ide_task_request_t *req_task,
1717 			       int outtotal)
1718 {
1719 	struct host_to_dev_fis	fis;
1720 	struct host_to_dev_fis *reply;
1721 	u8 *outbuf = NULL;
1722 	u8 *inbuf = NULL;
1723 	dma_addr_t outbuf_dma = 0;
1724 	dma_addr_t inbuf_dma = 0;
1725 	dma_addr_t dma_buffer = 0;
1726 	int err = 0;
1727 	unsigned int taskin = 0;
1728 	unsigned int taskout = 0;
1729 	u8 nsect = 0;
1730 	unsigned int timeout;
1731 	unsigned int force_single_sector;
1732 	unsigned int transfer_size;
1733 	unsigned long task_file_data;
1734 	int intotal = outtotal + req_task->out_size;
1735 	int erasemode = 0;
1736 
1737 	taskout = req_task->out_size;
1738 	taskin = req_task->in_size;
1739 	/* 130560 = 512 * 0xFF*/
1740 	if (taskin > 130560 || taskout > 130560)
1741 		return -EINVAL;
1742 
1743 	if (taskout) {
1744 		outbuf = memdup_user(buf + outtotal, taskout);
1745 		if (IS_ERR(outbuf))
1746 			return PTR_ERR(outbuf);
1747 
1748 		outbuf_dma = dma_map_single(&dd->pdev->dev, outbuf,
1749 					    taskout, DMA_TO_DEVICE);
1750 		if (dma_mapping_error(&dd->pdev->dev, outbuf_dma)) {
1751 			err = -ENOMEM;
1752 			goto abort;
1753 		}
1754 		dma_buffer = outbuf_dma;
1755 	}
1756 
1757 	if (taskin) {
1758 		inbuf = memdup_user(buf + intotal, taskin);
1759 		if (IS_ERR(inbuf)) {
1760 			err = PTR_ERR(inbuf);
1761 			inbuf = NULL;
1762 			goto abort;
1763 		}
1764 		inbuf_dma = dma_map_single(&dd->pdev->dev, inbuf,
1765 					   taskin, DMA_FROM_DEVICE);
1766 		if (dma_mapping_error(&dd->pdev->dev, inbuf_dma)) {
1767 			err = -ENOMEM;
1768 			goto abort;
1769 		}
1770 		dma_buffer = inbuf_dma;
1771 	}
1772 
1773 	/* only supports PIO and non-data commands from this ioctl. */
1774 	switch (req_task->data_phase) {
1775 	case TASKFILE_OUT:
1776 		nsect = taskout / ATA_SECT_SIZE;
1777 		reply = (dd->port->rxfis + RX_FIS_PIO_SETUP);
1778 		break;
1779 	case TASKFILE_IN:
1780 		reply = (dd->port->rxfis + RX_FIS_PIO_SETUP);
1781 		break;
1782 	case TASKFILE_NO_DATA:
1783 		reply = (dd->port->rxfis + RX_FIS_D2H_REG);
1784 		break;
1785 	default:
1786 		err = -EINVAL;
1787 		goto abort;
1788 	}
1789 
1790 	/* Build the FIS. */
1791 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1792 
1793 	fis.type	= 0x27;
1794 	fis.opts	= 1 << 7;
1795 	fis.command	= req_task->io_ports[7];
1796 	fis.features	= req_task->io_ports[1];
1797 	fis.sect_count	= req_task->io_ports[2];
1798 	fis.lba_low	= req_task->io_ports[3];
1799 	fis.lba_mid	= req_task->io_ports[4];
1800 	fis.lba_hi	= req_task->io_ports[5];
1801 	 /* Clear the dev bit*/
1802 	fis.device	= req_task->io_ports[6] & ~0x10;
1803 
1804 	if ((req_task->in_flags.all == 0) && (req_task->out_flags.all & 1)) {
1805 		req_task->in_flags.all	=
1806 			IDE_TASKFILE_STD_IN_FLAGS |
1807 			(IDE_HOB_STD_IN_FLAGS << 8);
1808 		fis.lba_low_ex		= req_task->hob_ports[3];
1809 		fis.lba_mid_ex		= req_task->hob_ports[4];
1810 		fis.lba_hi_ex		= req_task->hob_ports[5];
1811 		fis.features_ex		= req_task->hob_ports[1];
1812 		fis.sect_cnt_ex		= req_task->hob_ports[2];
1813 
1814 	} else {
1815 		req_task->in_flags.all = IDE_TASKFILE_STD_IN_FLAGS;
1816 	}
1817 
1818 	force_single_sector = implicit_sector(fis.command, fis.features);
1819 
1820 	if ((taskin || taskout) && (!fis.sect_count)) {
1821 		if (nsect)
1822 			fis.sect_count = nsect;
1823 		else {
1824 			if (!force_single_sector) {
1825 				dev_warn(&dd->pdev->dev,
1826 					"data movement but "
1827 					"sect_count is 0\n");
1828 				err = -EINVAL;
1829 				goto abort;
1830 			}
1831 		}
1832 	}
1833 
1834 	dbg_printk(MTIP_DRV_NAME
1835 		" %s: cmd %x, feat %x, nsect %x,"
1836 		" sect/lbal %x, lcyl/lbam %x, hcyl/lbah %x,"
1837 		" head/dev %x\n",
1838 		__func__,
1839 		fis.command,
1840 		fis.features,
1841 		fis.sect_count,
1842 		fis.lba_low,
1843 		fis.lba_mid,
1844 		fis.lba_hi,
1845 		fis.device);
1846 
1847 	/* check for erase mode support during secure erase.*/
1848 	if ((fis.command == ATA_CMD_SEC_ERASE_UNIT) && outbuf &&
1849 					(outbuf[0] & MTIP_SEC_ERASE_MODE)) {
1850 		erasemode = 1;
1851 	}
1852 
1853 	mtip_set_timeout(dd, &fis, &timeout, erasemode);
1854 
1855 	/* Determine the correct transfer size.*/
1856 	if (force_single_sector)
1857 		transfer_size = ATA_SECT_SIZE;
1858 	else
1859 		transfer_size = ATA_SECT_SIZE * fis.sect_count;
1860 
1861 	/* Execute the command.*/
1862 	if (mtip_exec_internal_command(dd->port,
1863 				 &fis,
1864 				 5,
1865 				 dma_buffer,
1866 				 transfer_size,
1867 				 0,
1868 				 timeout) < 0) {
1869 		err = -EIO;
1870 		goto abort;
1871 	}
1872 
1873 	task_file_data = readl(dd->port->mmio+PORT_TFDATA);
1874 
1875 	if ((req_task->data_phase == TASKFILE_IN) && !(task_file_data & 1)) {
1876 		reply = dd->port->rxfis + RX_FIS_PIO_SETUP;
1877 		req_task->io_ports[7] = reply->control;
1878 	} else {
1879 		reply = dd->port->rxfis + RX_FIS_D2H_REG;
1880 		req_task->io_ports[7] = reply->command;
1881 	}
1882 
1883 	/* reclaim the DMA buffers.*/
1884 	if (inbuf_dma)
1885 		dma_unmap_single(&dd->pdev->dev, inbuf_dma, taskin,
1886 				 DMA_FROM_DEVICE);
1887 	if (outbuf_dma)
1888 		dma_unmap_single(&dd->pdev->dev, outbuf_dma, taskout,
1889 				 DMA_TO_DEVICE);
1890 	inbuf_dma  = 0;
1891 	outbuf_dma = 0;
1892 
1893 	/* return the ATA registers to the caller.*/
1894 	req_task->io_ports[1] = reply->features;
1895 	req_task->io_ports[2] = reply->sect_count;
1896 	req_task->io_ports[3] = reply->lba_low;
1897 	req_task->io_ports[4] = reply->lba_mid;
1898 	req_task->io_ports[5] = reply->lba_hi;
1899 	req_task->io_ports[6] = reply->device;
1900 
1901 	if (req_task->out_flags.all & 1)  {
1902 
1903 		req_task->hob_ports[3] = reply->lba_low_ex;
1904 		req_task->hob_ports[4] = reply->lba_mid_ex;
1905 		req_task->hob_ports[5] = reply->lba_hi_ex;
1906 		req_task->hob_ports[1] = reply->features_ex;
1907 		req_task->hob_ports[2] = reply->sect_cnt_ex;
1908 	}
1909 	dbg_printk(MTIP_DRV_NAME
1910 		" %s: Completion: stat %x,"
1911 		"err %x, sect_cnt %x, lbalo %x,"
1912 		"lbamid %x, lbahi %x, dev %x\n",
1913 		__func__,
1914 		req_task->io_ports[7],
1915 		req_task->io_ports[1],
1916 		req_task->io_ports[2],
1917 		req_task->io_ports[3],
1918 		req_task->io_ports[4],
1919 		req_task->io_ports[5],
1920 		req_task->io_ports[6]);
1921 
1922 	if (taskout) {
1923 		if (copy_to_user(buf + outtotal, outbuf, taskout)) {
1924 			err = -EFAULT;
1925 			goto abort;
1926 		}
1927 	}
1928 	if (taskin) {
1929 		if (copy_to_user(buf + intotal, inbuf, taskin)) {
1930 			err = -EFAULT;
1931 			goto abort;
1932 		}
1933 	}
1934 abort:
1935 	if (inbuf_dma)
1936 		dma_unmap_single(&dd->pdev->dev, inbuf_dma, taskin,
1937 				 DMA_FROM_DEVICE);
1938 	if (outbuf_dma)
1939 		dma_unmap_single(&dd->pdev->dev, outbuf_dma, taskout,
1940 				 DMA_TO_DEVICE);
1941 	kfree(outbuf);
1942 	kfree(inbuf);
1943 
1944 	return err;
1945 }
1946 
1947 /*
1948  * Handle IOCTL calls from the Block Layer.
1949  *
1950  * This function is called by the Block Layer when it receives an IOCTL
1951  * command that it does not understand. If the IOCTL command is not supported
1952  * this function returns -ENOTTY.
1953  *
1954  * @dd  Pointer to the driver data structure.
1955  * @cmd IOCTL command passed from the Block Layer.
1956  * @arg IOCTL argument passed from the Block Layer.
1957  *
1958  * return value
1959  *	0	The IOCTL completed successfully.
1960  *	-ENOTTY The specified command is not supported.
1961  *	-EFAULT An error occurred copying data to a user space buffer.
1962  *	-EIO	An error occurred while executing the command.
1963  */
1964 static int mtip_hw_ioctl(struct driver_data *dd, unsigned int cmd,
1965 			 unsigned long arg)
1966 {
1967 	switch (cmd) {
1968 	case HDIO_GET_IDENTITY:
1969 	{
1970 		if (copy_to_user((void __user *)arg, dd->port->identify,
1971 						sizeof(u16) * ATA_ID_WORDS))
1972 			return -EFAULT;
1973 		break;
1974 	}
1975 	case HDIO_DRIVE_CMD:
1976 	{
1977 		u8 drive_command[4];
1978 
1979 		/* Copy the user command info to our buffer. */
1980 		if (copy_from_user(drive_command,
1981 					 (void __user *) arg,
1982 					 sizeof(drive_command)))
1983 			return -EFAULT;
1984 
1985 		/* Execute the drive command. */
1986 		if (exec_drive_command(dd->port,
1987 					 drive_command,
1988 					 (void __user *) (arg+4)))
1989 			return -EIO;
1990 
1991 		/* Copy the status back to the users buffer. */
1992 		if (copy_to_user((void __user *) arg,
1993 					 drive_command,
1994 					 sizeof(drive_command)))
1995 			return -EFAULT;
1996 
1997 		break;
1998 	}
1999 	case HDIO_DRIVE_TASK:
2000 	{
2001 		u8 drive_command[7];
2002 
2003 		/* Copy the user command info to our buffer. */
2004 		if (copy_from_user(drive_command,
2005 					 (void __user *) arg,
2006 					 sizeof(drive_command)))
2007 			return -EFAULT;
2008 
2009 		/* Execute the drive command. */
2010 		if (exec_drive_task(dd->port, drive_command))
2011 			return -EIO;
2012 
2013 		/* Copy the status back to the users buffer. */
2014 		if (copy_to_user((void __user *) arg,
2015 					 drive_command,
2016 					 sizeof(drive_command)))
2017 			return -EFAULT;
2018 
2019 		break;
2020 	}
2021 	case HDIO_DRIVE_TASKFILE: {
2022 		ide_task_request_t req_task;
2023 		int ret, outtotal;
2024 
2025 		if (copy_from_user(&req_task, (void __user *) arg,
2026 					sizeof(req_task)))
2027 			return -EFAULT;
2028 
2029 		outtotal = sizeof(req_task);
2030 
2031 		ret = exec_drive_taskfile(dd, (void __user *) arg,
2032 						&req_task, outtotal);
2033 
2034 		if (copy_to_user((void __user *) arg, &req_task,
2035 							sizeof(req_task)))
2036 			return -EFAULT;
2037 
2038 		return ret;
2039 	}
2040 
2041 	default:
2042 		return -EINVAL;
2043 	}
2044 	return 0;
2045 }
2046 
2047 /*
2048  * Submit an IO to the hw
2049  *
2050  * This function is called by the block layer to issue an io
2051  * to the device. Upon completion, the callback function will
2052  * be called with the data parameter passed as the callback data.
2053  *
2054  * @dd       Pointer to the driver data structure.
2055  * @start    First sector to read.
2056  * @nsect    Number of sectors to read.
2057  * @tag      The tag of this read command.
2058  * @callback Pointer to the function that should be called
2059  *	     when the read completes.
2060  * @data     Callback data passed to the callback function
2061  *	     when the read completes.
2062  * @dir      Direction (read or write)
2063  *
2064  * return value
2065  *	None
2066  */
2067 static void mtip_hw_submit_io(struct driver_data *dd, struct request *rq,
2068 			      struct mtip_cmd *command,
2069 			      struct blk_mq_hw_ctx *hctx)
2070 {
2071 	struct mtip_cmd_hdr *hdr =
2072 		dd->port->command_list + sizeof(struct mtip_cmd_hdr) * rq->tag;
2073 	struct host_to_dev_fis	*fis;
2074 	struct mtip_port *port = dd->port;
2075 	int dma_dir = rq_data_dir(rq) == READ ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
2076 	u64 start = blk_rq_pos(rq);
2077 	unsigned int nsect = blk_rq_sectors(rq);
2078 	unsigned int nents;
2079 
2080 	/* Map the scatter list for DMA access */
2081 	nents = blk_rq_map_sg(hctx->queue, rq, command->sg);
2082 	nents = dma_map_sg(&dd->pdev->dev, command->sg, nents, dma_dir);
2083 
2084 	prefetch(&port->flags);
2085 
2086 	command->scatter_ents = nents;
2087 
2088 	/*
2089 	 * The number of retries for this command before it is
2090 	 * reported as a failure to the upper layers.
2091 	 */
2092 	command->retries = MTIP_MAX_RETRIES;
2093 
2094 	/* Fill out fis */
2095 	fis = command->command;
2096 	fis->type        = 0x27;
2097 	fis->opts        = 1 << 7;
2098 	if (dma_dir == DMA_FROM_DEVICE)
2099 		fis->command = ATA_CMD_FPDMA_READ;
2100 	else
2101 		fis->command = ATA_CMD_FPDMA_WRITE;
2102 	fis->lba_low     = start & 0xFF;
2103 	fis->lba_mid     = (start >> 8) & 0xFF;
2104 	fis->lba_hi      = (start >> 16) & 0xFF;
2105 	fis->lba_low_ex  = (start >> 24) & 0xFF;
2106 	fis->lba_mid_ex  = (start >> 32) & 0xFF;
2107 	fis->lba_hi_ex   = (start >> 40) & 0xFF;
2108 	fis->device	 = 1 << 6;
2109 	fis->features    = nsect & 0xFF;
2110 	fis->features_ex = (nsect >> 8) & 0xFF;
2111 	fis->sect_count  = ((rq->tag << 3) | (rq->tag >> 5));
2112 	fis->sect_cnt_ex = 0;
2113 	fis->control     = 0;
2114 	fis->res2        = 0;
2115 	fis->res3        = 0;
2116 	fill_command_sg(dd, command, nents);
2117 
2118 	if (unlikely(command->unaligned))
2119 		fis->device |= 1 << 7;
2120 
2121 	/* Populate the command header */
2122 	hdr->ctba = cpu_to_le32(command->command_dma & 0xFFFFFFFF);
2123 	if (test_bit(MTIP_PF_HOST_CAP_64, &dd->port->flags))
2124 		hdr->ctbau = cpu_to_le32((command->command_dma >> 16) >> 16);
2125 	hdr->opts = cpu_to_le32((nents << 16) | 5 | AHCI_CMD_PREFETCH);
2126 	hdr->byte_count = 0;
2127 
2128 	command->direction = dma_dir;
2129 
2130 	/*
2131 	 * To prevent this command from being issued
2132 	 * if an internal command is in progress or error handling is active.
2133 	 */
2134 	if (unlikely(port->flags & MTIP_PF_PAUSE_IO)) {
2135 		set_bit(rq->tag, port->cmds_to_issue);
2136 		set_bit(MTIP_PF_ISSUE_CMDS_BIT, &port->flags);
2137 		return;
2138 	}
2139 
2140 	/* Issue the command to the hardware */
2141 	mtip_issue_ncq_command(port, rq->tag);
2142 }
2143 
2144 /*
2145  * Sysfs status dump.
2146  *
2147  * @dev  Pointer to the device structure, passed by the kernrel.
2148  * @attr Pointer to the device_attribute structure passed by the kernel.
2149  * @buf  Pointer to the char buffer that will receive the stats info.
2150  *
2151  * return value
2152  *	The size, in bytes, of the data copied into buf.
2153  */
2154 static ssize_t mtip_hw_show_status(struct device *dev,
2155 				struct device_attribute *attr,
2156 				char *buf)
2157 {
2158 	struct driver_data *dd = dev_to_disk(dev)->private_data;
2159 	int size = 0;
2160 
2161 	if (test_bit(MTIP_DDF_OVER_TEMP_BIT, &dd->dd_flag))
2162 		size += sprintf(buf, "%s", "thermal_shutdown\n");
2163 	else if (test_bit(MTIP_DDF_WRITE_PROTECT_BIT, &dd->dd_flag))
2164 		size += sprintf(buf, "%s", "write_protect\n");
2165 	else
2166 		size += sprintf(buf, "%s", "online\n");
2167 
2168 	return size;
2169 }
2170 
2171 static DEVICE_ATTR(status, 0444, mtip_hw_show_status, NULL);
2172 
2173 /* debugsfs entries */
2174 
2175 static ssize_t show_device_status(struct device_driver *drv, char *buf)
2176 {
2177 	int size = 0;
2178 	struct driver_data *dd, *tmp;
2179 	unsigned long flags;
2180 	char id_buf[42];
2181 	u16 status = 0;
2182 
2183 	spin_lock_irqsave(&dev_lock, flags);
2184 	size += sprintf(&buf[size], "Devices Present:\n");
2185 	list_for_each_entry_safe(dd, tmp, &online_list, online_list) {
2186 		if (dd->pdev) {
2187 			if (dd->port &&
2188 			    dd->port->identify &&
2189 			    dd->port->identify_valid) {
2190 				strlcpy(id_buf,
2191 					(char *) (dd->port->identify + 10), 21);
2192 				status = *(dd->port->identify + 141);
2193 			} else {
2194 				memset(id_buf, 0, 42);
2195 				status = 0;
2196 			}
2197 
2198 			if (dd->port &&
2199 			    test_bit(MTIP_PF_REBUILD_BIT, &dd->port->flags)) {
2200 				size += sprintf(&buf[size],
2201 					" device %s %s (ftl rebuild %d %%)\n",
2202 					dev_name(&dd->pdev->dev),
2203 					id_buf,
2204 					status);
2205 			} else {
2206 				size += sprintf(&buf[size],
2207 					" device %s %s\n",
2208 					dev_name(&dd->pdev->dev),
2209 					id_buf);
2210 			}
2211 		}
2212 	}
2213 
2214 	size += sprintf(&buf[size], "Devices Being Removed:\n");
2215 	list_for_each_entry_safe(dd, tmp, &removing_list, remove_list) {
2216 		if (dd->pdev) {
2217 			if (dd->port &&
2218 			    dd->port->identify &&
2219 			    dd->port->identify_valid) {
2220 				strlcpy(id_buf,
2221 					(char *) (dd->port->identify+10), 21);
2222 				status = *(dd->port->identify + 141);
2223 			} else {
2224 				memset(id_buf, 0, 42);
2225 				status = 0;
2226 			}
2227 
2228 			if (dd->port &&
2229 			    test_bit(MTIP_PF_REBUILD_BIT, &dd->port->flags)) {
2230 				size += sprintf(&buf[size],
2231 					" device %s %s (ftl rebuild %d %%)\n",
2232 					dev_name(&dd->pdev->dev),
2233 					id_buf,
2234 					status);
2235 			} else {
2236 				size += sprintf(&buf[size],
2237 					" device %s %s\n",
2238 					dev_name(&dd->pdev->dev),
2239 					id_buf);
2240 			}
2241 		}
2242 	}
2243 	spin_unlock_irqrestore(&dev_lock, flags);
2244 
2245 	return size;
2246 }
2247 
2248 static ssize_t mtip_hw_read_device_status(struct file *f, char __user *ubuf,
2249 						size_t len, loff_t *offset)
2250 {
2251 	struct driver_data *dd =  (struct driver_data *)f->private_data;
2252 	int size = *offset;
2253 	char *buf;
2254 	int rv = 0;
2255 
2256 	if (!len || *offset)
2257 		return 0;
2258 
2259 	buf = kzalloc(MTIP_DFS_MAX_BUF_SIZE, GFP_KERNEL);
2260 	if (!buf) {
2261 		dev_err(&dd->pdev->dev,
2262 			"Memory allocation: status buffer\n");
2263 		return -ENOMEM;
2264 	}
2265 
2266 	size += show_device_status(NULL, buf);
2267 
2268 	*offset = size <= len ? size : len;
2269 	size = copy_to_user(ubuf, buf, *offset);
2270 	if (size)
2271 		rv = -EFAULT;
2272 
2273 	kfree(buf);
2274 	return rv ? rv : *offset;
2275 }
2276 
2277 static ssize_t mtip_hw_read_registers(struct file *f, char __user *ubuf,
2278 				  size_t len, loff_t *offset)
2279 {
2280 	struct driver_data *dd =  (struct driver_data *)f->private_data;
2281 	char *buf;
2282 	u32 group_allocated;
2283 	int size = *offset;
2284 	int n, rv = 0;
2285 
2286 	if (!len || size)
2287 		return 0;
2288 
2289 	buf = kzalloc(MTIP_DFS_MAX_BUF_SIZE, GFP_KERNEL);
2290 	if (!buf) {
2291 		dev_err(&dd->pdev->dev,
2292 			"Memory allocation: register buffer\n");
2293 		return -ENOMEM;
2294 	}
2295 
2296 	size += sprintf(&buf[size], "H/ S ACTive      : [ 0x");
2297 
2298 	for (n = dd->slot_groups-1; n >= 0; n--)
2299 		size += sprintf(&buf[size], "%08X ",
2300 					 readl(dd->port->s_active[n]));
2301 
2302 	size += sprintf(&buf[size], "]\n");
2303 	size += sprintf(&buf[size], "H/ Command Issue : [ 0x");
2304 
2305 	for (n = dd->slot_groups-1; n >= 0; n--)
2306 		size += sprintf(&buf[size], "%08X ",
2307 					readl(dd->port->cmd_issue[n]));
2308 
2309 	size += sprintf(&buf[size], "]\n");
2310 	size += sprintf(&buf[size], "H/ Completed     : [ 0x");
2311 
2312 	for (n = dd->slot_groups-1; n >= 0; n--)
2313 		size += sprintf(&buf[size], "%08X ",
2314 				readl(dd->port->completed[n]));
2315 
2316 	size += sprintf(&buf[size], "]\n");
2317 	size += sprintf(&buf[size], "H/ PORT IRQ STAT : [ 0x%08X ]\n",
2318 				readl(dd->port->mmio + PORT_IRQ_STAT));
2319 	size += sprintf(&buf[size], "H/ HOST IRQ STAT : [ 0x%08X ]\n",
2320 				readl(dd->mmio + HOST_IRQ_STAT));
2321 	size += sprintf(&buf[size], "\n");
2322 
2323 	size += sprintf(&buf[size], "L/ Commands in Q : [ 0x");
2324 
2325 	for (n = dd->slot_groups-1; n >= 0; n--) {
2326 		if (sizeof(long) > sizeof(u32))
2327 			group_allocated =
2328 				dd->port->cmds_to_issue[n/2] >> (32*(n&1));
2329 		else
2330 			group_allocated = dd->port->cmds_to_issue[n];
2331 		size += sprintf(&buf[size], "%08X ", group_allocated);
2332 	}
2333 	size += sprintf(&buf[size], "]\n");
2334 
2335 	*offset = size <= len ? size : len;
2336 	size = copy_to_user(ubuf, buf, *offset);
2337 	if (size)
2338 		rv = -EFAULT;
2339 
2340 	kfree(buf);
2341 	return rv ? rv : *offset;
2342 }
2343 
2344 static ssize_t mtip_hw_read_flags(struct file *f, char __user *ubuf,
2345 				  size_t len, loff_t *offset)
2346 {
2347 	struct driver_data *dd =  (struct driver_data *)f->private_data;
2348 	char *buf;
2349 	int size = *offset;
2350 	int rv = 0;
2351 
2352 	if (!len || size)
2353 		return 0;
2354 
2355 	buf = kzalloc(MTIP_DFS_MAX_BUF_SIZE, GFP_KERNEL);
2356 	if (!buf) {
2357 		dev_err(&dd->pdev->dev,
2358 			"Memory allocation: flag buffer\n");
2359 		return -ENOMEM;
2360 	}
2361 
2362 	size += sprintf(&buf[size], "Flag-port : [ %08lX ]\n",
2363 							dd->port->flags);
2364 	size += sprintf(&buf[size], "Flag-dd   : [ %08lX ]\n",
2365 							dd->dd_flag);
2366 
2367 	*offset = size <= len ? size : len;
2368 	size = copy_to_user(ubuf, buf, *offset);
2369 	if (size)
2370 		rv = -EFAULT;
2371 
2372 	kfree(buf);
2373 	return rv ? rv : *offset;
2374 }
2375 
2376 static const struct file_operations mtip_device_status_fops = {
2377 	.owner  = THIS_MODULE,
2378 	.open   = simple_open,
2379 	.read   = mtip_hw_read_device_status,
2380 	.llseek = no_llseek,
2381 };
2382 
2383 static const struct file_operations mtip_regs_fops = {
2384 	.owner  = THIS_MODULE,
2385 	.open   = simple_open,
2386 	.read   = mtip_hw_read_registers,
2387 	.llseek = no_llseek,
2388 };
2389 
2390 static const struct file_operations mtip_flags_fops = {
2391 	.owner  = THIS_MODULE,
2392 	.open   = simple_open,
2393 	.read   = mtip_hw_read_flags,
2394 	.llseek = no_llseek,
2395 };
2396 
2397 /*
2398  * Create the sysfs related attributes.
2399  *
2400  * @dd   Pointer to the driver data structure.
2401  * @kobj Pointer to the kobj for the block device.
2402  *
2403  * return value
2404  *	0	Operation completed successfully.
2405  *	-EINVAL Invalid parameter.
2406  */
2407 static int mtip_hw_sysfs_init(struct driver_data *dd, struct kobject *kobj)
2408 {
2409 	if (!kobj || !dd)
2410 		return -EINVAL;
2411 
2412 	if (sysfs_create_file(kobj, &dev_attr_status.attr))
2413 		dev_warn(&dd->pdev->dev,
2414 			"Error creating 'status' sysfs entry\n");
2415 	return 0;
2416 }
2417 
2418 /*
2419  * Remove the sysfs related attributes.
2420  *
2421  * @dd   Pointer to the driver data structure.
2422  * @kobj Pointer to the kobj for the block device.
2423  *
2424  * return value
2425  *	0	Operation completed successfully.
2426  *	-EINVAL Invalid parameter.
2427  */
2428 static int mtip_hw_sysfs_exit(struct driver_data *dd, struct kobject *kobj)
2429 {
2430 	if (!kobj || !dd)
2431 		return -EINVAL;
2432 
2433 	sysfs_remove_file(kobj, &dev_attr_status.attr);
2434 
2435 	return 0;
2436 }
2437 
2438 static int mtip_hw_debugfs_init(struct driver_data *dd)
2439 {
2440 	if (!dfs_parent)
2441 		return -1;
2442 
2443 	dd->dfs_node = debugfs_create_dir(dd->disk->disk_name, dfs_parent);
2444 	if (IS_ERR_OR_NULL(dd->dfs_node)) {
2445 		dev_warn(&dd->pdev->dev,
2446 			"Error creating node %s under debugfs\n",
2447 						dd->disk->disk_name);
2448 		dd->dfs_node = NULL;
2449 		return -1;
2450 	}
2451 
2452 	debugfs_create_file("flags", 0444, dd->dfs_node, dd, &mtip_flags_fops);
2453 	debugfs_create_file("registers", 0444, dd->dfs_node, dd,
2454 			    &mtip_regs_fops);
2455 
2456 	return 0;
2457 }
2458 
2459 static void mtip_hw_debugfs_exit(struct driver_data *dd)
2460 {
2461 	debugfs_remove_recursive(dd->dfs_node);
2462 }
2463 
2464 /*
2465  * Perform any init/resume time hardware setup
2466  *
2467  * @dd Pointer to the driver data structure.
2468  *
2469  * return value
2470  *	None
2471  */
2472 static inline void hba_setup(struct driver_data *dd)
2473 {
2474 	u32 hwdata;
2475 	hwdata = readl(dd->mmio + HOST_HSORG);
2476 
2477 	/* interrupt bug workaround: use only 1 IS bit.*/
2478 	writel(hwdata |
2479 		HSORG_DISABLE_SLOTGRP_INTR |
2480 		HSORG_DISABLE_SLOTGRP_PXIS,
2481 		dd->mmio + HOST_HSORG);
2482 }
2483 
2484 static int mtip_device_unaligned_constrained(struct driver_data *dd)
2485 {
2486 	return (dd->pdev->device == P420M_DEVICE_ID ? 1 : 0);
2487 }
2488 
2489 /*
2490  * Detect the details of the product, and store anything needed
2491  * into the driver data structure.  This includes product type and
2492  * version and number of slot groups.
2493  *
2494  * @dd Pointer to the driver data structure.
2495  *
2496  * return value
2497  *	None
2498  */
2499 static void mtip_detect_product(struct driver_data *dd)
2500 {
2501 	u32 hwdata;
2502 	unsigned int rev, slotgroups;
2503 
2504 	/*
2505 	 * HBA base + 0xFC [15:0] - vendor-specific hardware interface
2506 	 * info register:
2507 	 * [15:8] hardware/software interface rev#
2508 	 * [   3] asic-style interface
2509 	 * [ 2:0] number of slot groups, minus 1 (only valid for asic-style).
2510 	 */
2511 	hwdata = readl(dd->mmio + HOST_HSORG);
2512 
2513 	dd->product_type = MTIP_PRODUCT_UNKNOWN;
2514 	dd->slot_groups = 1;
2515 
2516 	if (hwdata & 0x8) {
2517 		dd->product_type = MTIP_PRODUCT_ASICFPGA;
2518 		rev = (hwdata & HSORG_HWREV) >> 8;
2519 		slotgroups = (hwdata & HSORG_SLOTGROUPS) + 1;
2520 		dev_info(&dd->pdev->dev,
2521 			"ASIC-FPGA design, HS rev 0x%x, "
2522 			"%i slot groups [%i slots]\n",
2523 			 rev,
2524 			 slotgroups,
2525 			 slotgroups * 32);
2526 
2527 		if (slotgroups > MTIP_MAX_SLOT_GROUPS) {
2528 			dev_warn(&dd->pdev->dev,
2529 				"Warning: driver only supports "
2530 				"%i slot groups.\n", MTIP_MAX_SLOT_GROUPS);
2531 			slotgroups = MTIP_MAX_SLOT_GROUPS;
2532 		}
2533 		dd->slot_groups = slotgroups;
2534 		return;
2535 	}
2536 
2537 	dev_warn(&dd->pdev->dev, "Unrecognized product id\n");
2538 }
2539 
2540 /*
2541  * Blocking wait for FTL rebuild to complete
2542  *
2543  * @dd Pointer to the DRIVER_DATA structure.
2544  *
2545  * return value
2546  *	0	FTL rebuild completed successfully
2547  *	-EFAULT FTL rebuild error/timeout/interruption
2548  */
2549 static int mtip_ftl_rebuild_poll(struct driver_data *dd)
2550 {
2551 	unsigned long timeout, cnt = 0, start;
2552 
2553 	dev_warn(&dd->pdev->dev,
2554 		"FTL rebuild in progress. Polling for completion.\n");
2555 
2556 	start = jiffies;
2557 	timeout = jiffies + msecs_to_jiffies(MTIP_FTL_REBUILD_TIMEOUT_MS);
2558 
2559 	do {
2560 		if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
2561 				&dd->dd_flag)))
2562 			return -EFAULT;
2563 		if (mtip_check_surprise_removal(dd->pdev))
2564 			return -EFAULT;
2565 
2566 		if (mtip_get_identify(dd->port, NULL) < 0)
2567 			return -EFAULT;
2568 
2569 		if (*(dd->port->identify + MTIP_FTL_REBUILD_OFFSET) ==
2570 			MTIP_FTL_REBUILD_MAGIC) {
2571 			ssleep(1);
2572 			/* Print message every 3 minutes */
2573 			if (cnt++ >= 180) {
2574 				dev_warn(&dd->pdev->dev,
2575 				"FTL rebuild in progress (%d secs).\n",
2576 				jiffies_to_msecs(jiffies - start) / 1000);
2577 				cnt = 0;
2578 			}
2579 		} else {
2580 			dev_warn(&dd->pdev->dev,
2581 				"FTL rebuild complete (%d secs).\n",
2582 			jiffies_to_msecs(jiffies - start) / 1000);
2583 			mtip_block_initialize(dd);
2584 			return 0;
2585 		}
2586 	} while (time_before(jiffies, timeout));
2587 
2588 	/* Check for timeout */
2589 	dev_err(&dd->pdev->dev,
2590 		"Timed out waiting for FTL rebuild to complete (%d secs).\n",
2591 		jiffies_to_msecs(jiffies - start) / 1000);
2592 	return -EFAULT;
2593 }
2594 
2595 static void mtip_softirq_done_fn(struct request *rq)
2596 {
2597 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
2598 	struct driver_data *dd = rq->q->queuedata;
2599 
2600 	/* Unmap the DMA scatter list entries */
2601 	dma_unmap_sg(&dd->pdev->dev, cmd->sg, cmd->scatter_ents,
2602 							cmd->direction);
2603 
2604 	if (unlikely(cmd->unaligned))
2605 		atomic_inc(&dd->port->cmd_slot_unal);
2606 
2607 	blk_mq_end_request(rq, cmd->status);
2608 }
2609 
2610 static bool mtip_abort_cmd(struct request *req, void *data, bool reserved)
2611 {
2612 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(req);
2613 	struct driver_data *dd = data;
2614 
2615 	dbg_printk(MTIP_DRV_NAME " Aborting request, tag = %d\n", req->tag);
2616 
2617 	clear_bit(req->tag, dd->port->cmds_to_issue);
2618 	cmd->status = BLK_STS_IOERR;
2619 	mtip_softirq_done_fn(req);
2620 	return true;
2621 }
2622 
2623 static bool mtip_queue_cmd(struct request *req, void *data, bool reserved)
2624 {
2625 	struct driver_data *dd = data;
2626 
2627 	set_bit(req->tag, dd->port->cmds_to_issue);
2628 	blk_abort_request(req);
2629 	return true;
2630 }
2631 
2632 /*
2633  * service thread to issue queued commands
2634  *
2635  * @data Pointer to the driver data structure.
2636  *
2637  * return value
2638  *	0
2639  */
2640 
2641 static int mtip_service_thread(void *data)
2642 {
2643 	struct driver_data *dd = (struct driver_data *)data;
2644 	unsigned long slot, slot_start, slot_wrap, to;
2645 	unsigned int num_cmd_slots = dd->slot_groups * 32;
2646 	struct mtip_port *port = dd->port;
2647 
2648 	while (1) {
2649 		if (kthread_should_stop() ||
2650 			test_bit(MTIP_PF_SVC_THD_STOP_BIT, &port->flags))
2651 			goto st_out;
2652 		clear_bit(MTIP_PF_SVC_THD_ACTIVE_BIT, &port->flags);
2653 
2654 		/*
2655 		 * the condition is to check neither an internal command is
2656 		 * is in progress nor error handling is active
2657 		 */
2658 		wait_event_interruptible(port->svc_wait, (port->flags) &&
2659 			(port->flags & MTIP_PF_SVC_THD_WORK));
2660 
2661 		if (kthread_should_stop() ||
2662 			test_bit(MTIP_PF_SVC_THD_STOP_BIT, &port->flags))
2663 			goto st_out;
2664 
2665 		if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
2666 				&dd->dd_flag)))
2667 			goto st_out;
2668 
2669 		set_bit(MTIP_PF_SVC_THD_ACTIVE_BIT, &port->flags);
2670 
2671 restart_eh:
2672 		/* Demux bits: start with error handling */
2673 		if (test_bit(MTIP_PF_EH_ACTIVE_BIT, &port->flags)) {
2674 			mtip_handle_tfe(dd);
2675 			clear_bit(MTIP_PF_EH_ACTIVE_BIT, &port->flags);
2676 		}
2677 
2678 		if (test_bit(MTIP_PF_EH_ACTIVE_BIT, &port->flags))
2679 			goto restart_eh;
2680 
2681 		if (test_bit(MTIP_PF_TO_ACTIVE_BIT, &port->flags)) {
2682 			to = jiffies + msecs_to_jiffies(5000);
2683 
2684 			do {
2685 				mdelay(100);
2686 			} while (atomic_read(&dd->irq_workers_active) != 0 &&
2687 				time_before(jiffies, to));
2688 
2689 			if (atomic_read(&dd->irq_workers_active) != 0)
2690 				dev_warn(&dd->pdev->dev,
2691 					"Completion workers still active!");
2692 
2693 			blk_mq_quiesce_queue(dd->queue);
2694 
2695 			blk_mq_tagset_busy_iter(&dd->tags, mtip_queue_cmd, dd);
2696 
2697 			set_bit(MTIP_PF_ISSUE_CMDS_BIT, &dd->port->flags);
2698 
2699 			if (mtip_device_reset(dd))
2700 				blk_mq_tagset_busy_iter(&dd->tags,
2701 							mtip_abort_cmd, dd);
2702 
2703 			clear_bit(MTIP_PF_TO_ACTIVE_BIT, &dd->port->flags);
2704 
2705 			blk_mq_unquiesce_queue(dd->queue);
2706 		}
2707 
2708 		if (test_bit(MTIP_PF_ISSUE_CMDS_BIT, &port->flags)) {
2709 			slot = 1;
2710 			/* used to restrict the loop to one iteration */
2711 			slot_start = num_cmd_slots;
2712 			slot_wrap = 0;
2713 			while (1) {
2714 				slot = find_next_bit(port->cmds_to_issue,
2715 						num_cmd_slots, slot);
2716 				if (slot_wrap == 1) {
2717 					if ((slot_start >= slot) ||
2718 						(slot >= num_cmd_slots))
2719 						break;
2720 				}
2721 				if (unlikely(slot_start == num_cmd_slots))
2722 					slot_start = slot;
2723 
2724 				if (unlikely(slot == num_cmd_slots)) {
2725 					slot = 1;
2726 					slot_wrap = 1;
2727 					continue;
2728 				}
2729 
2730 				/* Issue the command to the hardware */
2731 				mtip_issue_ncq_command(port, slot);
2732 
2733 				clear_bit(slot, port->cmds_to_issue);
2734 			}
2735 
2736 			clear_bit(MTIP_PF_ISSUE_CMDS_BIT, &port->flags);
2737 		}
2738 
2739 		if (test_bit(MTIP_PF_REBUILD_BIT, &port->flags)) {
2740 			if (mtip_ftl_rebuild_poll(dd) == 0)
2741 				clear_bit(MTIP_PF_REBUILD_BIT, &port->flags);
2742 		}
2743 	}
2744 
2745 st_out:
2746 	return 0;
2747 }
2748 
2749 /*
2750  * DMA region teardown
2751  *
2752  * @dd Pointer to driver_data structure
2753  *
2754  * return value
2755  *      None
2756  */
2757 static void mtip_dma_free(struct driver_data *dd)
2758 {
2759 	struct mtip_port *port = dd->port;
2760 
2761 	if (port->block1)
2762 		dma_free_coherent(&dd->pdev->dev, BLOCK_DMA_ALLOC_SZ,
2763 					port->block1, port->block1_dma);
2764 
2765 	if (port->command_list) {
2766 		dma_free_coherent(&dd->pdev->dev, AHCI_CMD_TBL_SZ,
2767 				port->command_list, port->command_list_dma);
2768 	}
2769 }
2770 
2771 /*
2772  * DMA region setup
2773  *
2774  * @dd Pointer to driver_data structure
2775  *
2776  * return value
2777  *      -ENOMEM Not enough free DMA region space to initialize driver
2778  */
2779 static int mtip_dma_alloc(struct driver_data *dd)
2780 {
2781 	struct mtip_port *port = dd->port;
2782 
2783 	/* Allocate dma memory for RX Fis, Identify, and Sector Bufffer */
2784 	port->block1 =
2785 		dma_alloc_coherent(&dd->pdev->dev, BLOCK_DMA_ALLOC_SZ,
2786 					&port->block1_dma, GFP_KERNEL);
2787 	if (!port->block1)
2788 		return -ENOMEM;
2789 	memset(port->block1, 0, BLOCK_DMA_ALLOC_SZ);
2790 
2791 	/* Allocate dma memory for command list */
2792 	port->command_list =
2793 		dma_alloc_coherent(&dd->pdev->dev, AHCI_CMD_TBL_SZ,
2794 					&port->command_list_dma, GFP_KERNEL);
2795 	if (!port->command_list) {
2796 		dma_free_coherent(&dd->pdev->dev, BLOCK_DMA_ALLOC_SZ,
2797 					port->block1, port->block1_dma);
2798 		port->block1 = NULL;
2799 		port->block1_dma = 0;
2800 		return -ENOMEM;
2801 	}
2802 	memset(port->command_list, 0, AHCI_CMD_TBL_SZ);
2803 
2804 	/* Setup all pointers into first DMA region */
2805 	port->rxfis         = port->block1 + AHCI_RX_FIS_OFFSET;
2806 	port->rxfis_dma     = port->block1_dma + AHCI_RX_FIS_OFFSET;
2807 	port->identify      = port->block1 + AHCI_IDFY_OFFSET;
2808 	port->identify_dma  = port->block1_dma + AHCI_IDFY_OFFSET;
2809 	port->log_buf       = port->block1 + AHCI_SECTBUF_OFFSET;
2810 	port->log_buf_dma   = port->block1_dma + AHCI_SECTBUF_OFFSET;
2811 	port->smart_buf     = port->block1 + AHCI_SMARTBUF_OFFSET;
2812 	port->smart_buf_dma = port->block1_dma + AHCI_SMARTBUF_OFFSET;
2813 
2814 	return 0;
2815 }
2816 
2817 static int mtip_hw_get_identify(struct driver_data *dd)
2818 {
2819 	struct smart_attr attr242;
2820 	unsigned char *buf;
2821 	int rv;
2822 
2823 	if (mtip_get_identify(dd->port, NULL) < 0)
2824 		return -EFAULT;
2825 
2826 	if (*(dd->port->identify + MTIP_FTL_REBUILD_OFFSET) ==
2827 		MTIP_FTL_REBUILD_MAGIC) {
2828 		set_bit(MTIP_PF_REBUILD_BIT, &dd->port->flags);
2829 		return MTIP_FTL_REBUILD_MAGIC;
2830 	}
2831 	mtip_dump_identify(dd->port);
2832 
2833 	/* check write protect, over temp and rebuild statuses */
2834 	rv = mtip_read_log_page(dd->port, ATA_LOG_SATA_NCQ,
2835 				dd->port->log_buf,
2836 				dd->port->log_buf_dma, 1);
2837 	if (rv) {
2838 		dev_warn(&dd->pdev->dev,
2839 			"Error in READ LOG EXT (10h) command\n");
2840 		/* non-critical error, don't fail the load */
2841 	} else {
2842 		buf = (unsigned char *)dd->port->log_buf;
2843 		if (buf[259] & 0x1) {
2844 			dev_info(&dd->pdev->dev,
2845 				"Write protect bit is set.\n");
2846 			set_bit(MTIP_DDF_WRITE_PROTECT_BIT, &dd->dd_flag);
2847 		}
2848 		if (buf[288] == 0xF7) {
2849 			dev_info(&dd->pdev->dev,
2850 				"Exceeded Tmax, drive in thermal shutdown.\n");
2851 			set_bit(MTIP_DDF_OVER_TEMP_BIT, &dd->dd_flag);
2852 		}
2853 		if (buf[288] == 0xBF) {
2854 			dev_info(&dd->pdev->dev,
2855 				"Drive indicates rebuild has failed.\n");
2856 			set_bit(MTIP_DDF_REBUILD_FAILED_BIT, &dd->dd_flag);
2857 		}
2858 	}
2859 
2860 	/* get write protect progess */
2861 	memset(&attr242, 0, sizeof(struct smart_attr));
2862 	if (mtip_get_smart_attr(dd->port, 242, &attr242))
2863 		dev_warn(&dd->pdev->dev,
2864 				"Unable to check write protect progress\n");
2865 	else
2866 		dev_info(&dd->pdev->dev,
2867 				"Write protect progress: %u%% (%u blocks)\n",
2868 				attr242.cur, le32_to_cpu(attr242.data));
2869 
2870 	return rv;
2871 }
2872 
2873 /*
2874  * Called once for each card.
2875  *
2876  * @dd Pointer to the driver data structure.
2877  *
2878  * return value
2879  *	0 on success, else an error code.
2880  */
2881 static int mtip_hw_init(struct driver_data *dd)
2882 {
2883 	int i;
2884 	int rv;
2885 	unsigned long timeout, timetaken;
2886 
2887 	dd->mmio = pcim_iomap_table(dd->pdev)[MTIP_ABAR];
2888 
2889 	mtip_detect_product(dd);
2890 	if (dd->product_type == MTIP_PRODUCT_UNKNOWN) {
2891 		rv = -EIO;
2892 		goto out1;
2893 	}
2894 
2895 	hba_setup(dd);
2896 
2897 	dd->port = kzalloc_node(sizeof(struct mtip_port), GFP_KERNEL,
2898 				dd->numa_node);
2899 	if (!dd->port) {
2900 		dev_err(&dd->pdev->dev,
2901 			"Memory allocation: port structure\n");
2902 		return -ENOMEM;
2903 	}
2904 
2905 	/* Continue workqueue setup */
2906 	for (i = 0; i < MTIP_MAX_SLOT_GROUPS; i++)
2907 		dd->work[i].port = dd->port;
2908 
2909 	/* Enable unaligned IO constraints for some devices */
2910 	if (mtip_device_unaligned_constrained(dd))
2911 		dd->unal_qdepth = MTIP_MAX_UNALIGNED_SLOTS;
2912 	else
2913 		dd->unal_qdepth = 0;
2914 
2915 	atomic_set(&dd->port->cmd_slot_unal, dd->unal_qdepth);
2916 
2917 	/* Spinlock to prevent concurrent issue */
2918 	for (i = 0; i < MTIP_MAX_SLOT_GROUPS; i++)
2919 		spin_lock_init(&dd->port->cmd_issue_lock[i]);
2920 
2921 	/* Set the port mmio base address. */
2922 	dd->port->mmio	= dd->mmio + PORT_OFFSET;
2923 	dd->port->dd	= dd;
2924 
2925 	/* DMA allocations */
2926 	rv = mtip_dma_alloc(dd);
2927 	if (rv < 0)
2928 		goto out1;
2929 
2930 	/* Setup the pointers to the extended s_active and CI registers. */
2931 	for (i = 0; i < dd->slot_groups; i++) {
2932 		dd->port->s_active[i] =
2933 			dd->port->mmio + i*0x80 + PORT_SCR_ACT;
2934 		dd->port->cmd_issue[i] =
2935 			dd->port->mmio + i*0x80 + PORT_COMMAND_ISSUE;
2936 		dd->port->completed[i] =
2937 			dd->port->mmio + i*0x80 + PORT_SDBV;
2938 	}
2939 
2940 	timetaken = jiffies;
2941 	timeout = jiffies + msecs_to_jiffies(30000);
2942 	while (((readl(dd->port->mmio + PORT_SCR_STAT) & 0x0F) != 0x03) &&
2943 		 time_before(jiffies, timeout)) {
2944 		mdelay(100);
2945 	}
2946 	if (unlikely(mtip_check_surprise_removal(dd->pdev))) {
2947 		timetaken = jiffies - timetaken;
2948 		dev_warn(&dd->pdev->dev,
2949 			"Surprise removal detected at %u ms\n",
2950 			jiffies_to_msecs(timetaken));
2951 		rv = -ENODEV;
2952 		goto out2 ;
2953 	}
2954 	if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag))) {
2955 		timetaken = jiffies - timetaken;
2956 		dev_warn(&dd->pdev->dev,
2957 			"Removal detected at %u ms\n",
2958 			jiffies_to_msecs(timetaken));
2959 		rv = -EFAULT;
2960 		goto out2;
2961 	}
2962 
2963 	/* Conditionally reset the HBA. */
2964 	if (!(readl(dd->mmio + HOST_CAP) & HOST_CAP_NZDMA)) {
2965 		if (mtip_hba_reset(dd) < 0) {
2966 			dev_err(&dd->pdev->dev,
2967 				"Card did not reset within timeout\n");
2968 			rv = -EIO;
2969 			goto out2;
2970 		}
2971 	} else {
2972 		/* Clear any pending interrupts on the HBA */
2973 		writel(readl(dd->mmio + HOST_IRQ_STAT),
2974 			dd->mmio + HOST_IRQ_STAT);
2975 	}
2976 
2977 	mtip_init_port(dd->port);
2978 	mtip_start_port(dd->port);
2979 
2980 	/* Setup the ISR and enable interrupts. */
2981 	rv = request_irq(dd->pdev->irq, mtip_irq_handler, IRQF_SHARED,
2982 			 dev_driver_string(&dd->pdev->dev), dd);
2983 	if (rv) {
2984 		dev_err(&dd->pdev->dev,
2985 			"Unable to allocate IRQ %d\n", dd->pdev->irq);
2986 		goto out2;
2987 	}
2988 	irq_set_affinity_hint(dd->pdev->irq, get_cpu_mask(dd->isr_binding));
2989 
2990 	/* Enable interrupts on the HBA. */
2991 	writel(readl(dd->mmio + HOST_CTL) | HOST_IRQ_EN,
2992 					dd->mmio + HOST_CTL);
2993 
2994 	init_waitqueue_head(&dd->port->svc_wait);
2995 
2996 	if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)) {
2997 		rv = -EFAULT;
2998 		goto out3;
2999 	}
3000 
3001 	return rv;
3002 
3003 out3:
3004 	/* Disable interrupts on the HBA. */
3005 	writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
3006 			dd->mmio + HOST_CTL);
3007 
3008 	/* Release the IRQ. */
3009 	irq_set_affinity_hint(dd->pdev->irq, NULL);
3010 	free_irq(dd->pdev->irq, dd);
3011 
3012 out2:
3013 	mtip_deinit_port(dd->port);
3014 	mtip_dma_free(dd);
3015 
3016 out1:
3017 	/* Free the memory allocated for the for structure. */
3018 	kfree(dd->port);
3019 
3020 	return rv;
3021 }
3022 
3023 static int mtip_standby_drive(struct driver_data *dd)
3024 {
3025 	int rv = 0;
3026 
3027 	if (dd->sr || !dd->port)
3028 		return -ENODEV;
3029 	/*
3030 	 * Send standby immediate (E0h) to the drive so that it
3031 	 * saves its state.
3032 	 */
3033 	if (!test_bit(MTIP_PF_REBUILD_BIT, &dd->port->flags) &&
3034 	    !test_bit(MTIP_DDF_REBUILD_FAILED_BIT, &dd->dd_flag) &&
3035 	    !test_bit(MTIP_DDF_SEC_LOCK_BIT, &dd->dd_flag)) {
3036 		rv = mtip_standby_immediate(dd->port);
3037 		if (rv)
3038 			dev_warn(&dd->pdev->dev,
3039 				"STANDBY IMMEDIATE failed\n");
3040 	}
3041 	return rv;
3042 }
3043 
3044 /*
3045  * Called to deinitialize an interface.
3046  *
3047  * @dd Pointer to the driver data structure.
3048  *
3049  * return value
3050  *	0
3051  */
3052 static int mtip_hw_exit(struct driver_data *dd)
3053 {
3054 	if (!dd->sr) {
3055 		/* de-initialize the port. */
3056 		mtip_deinit_port(dd->port);
3057 
3058 		/* Disable interrupts on the HBA. */
3059 		writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
3060 				dd->mmio + HOST_CTL);
3061 	}
3062 
3063 	/* Release the IRQ. */
3064 	irq_set_affinity_hint(dd->pdev->irq, NULL);
3065 	free_irq(dd->pdev->irq, dd);
3066 	msleep(1000);
3067 
3068 	/* Free dma regions */
3069 	mtip_dma_free(dd);
3070 
3071 	/* Free the memory allocated for the for structure. */
3072 	kfree(dd->port);
3073 	dd->port = NULL;
3074 
3075 	return 0;
3076 }
3077 
3078 /*
3079  * Issue a Standby Immediate command to the device.
3080  *
3081  * This function is called by the Block Layer just before the
3082  * system powers off during a shutdown.
3083  *
3084  * @dd Pointer to the driver data structure.
3085  *
3086  * return value
3087  *	0
3088  */
3089 static int mtip_hw_shutdown(struct driver_data *dd)
3090 {
3091 	/*
3092 	 * Send standby immediate (E0h) to the drive so that it
3093 	 * saves its state.
3094 	 */
3095 	mtip_standby_drive(dd);
3096 
3097 	return 0;
3098 }
3099 
3100 /*
3101  * Suspend function
3102  *
3103  * This function is called by the Block Layer just before the
3104  * system hibernates.
3105  *
3106  * @dd Pointer to the driver data structure.
3107  *
3108  * return value
3109  *	0	Suspend was successful
3110  *	-EFAULT Suspend was not successful
3111  */
3112 static int mtip_hw_suspend(struct driver_data *dd)
3113 {
3114 	/*
3115 	 * Send standby immediate (E0h) to the drive
3116 	 * so that it saves its state.
3117 	 */
3118 	if (mtip_standby_drive(dd) != 0) {
3119 		dev_err(&dd->pdev->dev,
3120 			"Failed standby-immediate command\n");
3121 		return -EFAULT;
3122 	}
3123 
3124 	/* Disable interrupts on the HBA.*/
3125 	writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
3126 			dd->mmio + HOST_CTL);
3127 	mtip_deinit_port(dd->port);
3128 
3129 	return 0;
3130 }
3131 
3132 /*
3133  * Resume function
3134  *
3135  * This function is called by the Block Layer as the
3136  * system resumes.
3137  *
3138  * @dd Pointer to the driver data structure.
3139  *
3140  * return value
3141  *	0	Resume was successful
3142  *      -EFAULT Resume was not successful
3143  */
3144 static int mtip_hw_resume(struct driver_data *dd)
3145 {
3146 	/* Perform any needed hardware setup steps */
3147 	hba_setup(dd);
3148 
3149 	/* Reset the HBA */
3150 	if (mtip_hba_reset(dd) != 0) {
3151 		dev_err(&dd->pdev->dev,
3152 			"Unable to reset the HBA\n");
3153 		return -EFAULT;
3154 	}
3155 
3156 	/*
3157 	 * Enable the port, DMA engine, and FIS reception specific
3158 	 * h/w in controller.
3159 	 */
3160 	mtip_init_port(dd->port);
3161 	mtip_start_port(dd->port);
3162 
3163 	/* Enable interrupts on the HBA.*/
3164 	writel(readl(dd->mmio + HOST_CTL) | HOST_IRQ_EN,
3165 			dd->mmio + HOST_CTL);
3166 
3167 	return 0;
3168 }
3169 
3170 /*
3171  * Helper function for reusing disk name
3172  * upon hot insertion.
3173  */
3174 static int rssd_disk_name_format(char *prefix,
3175 				 int index,
3176 				 char *buf,
3177 				 int buflen)
3178 {
3179 	const int base = 'z' - 'a' + 1;
3180 	char *begin = buf + strlen(prefix);
3181 	char *end = buf + buflen;
3182 	char *p;
3183 	int unit;
3184 
3185 	p = end - 1;
3186 	*p = '\0';
3187 	unit = base;
3188 	do {
3189 		if (p == begin)
3190 			return -EINVAL;
3191 		*--p = 'a' + (index % unit);
3192 		index = (index / unit) - 1;
3193 	} while (index >= 0);
3194 
3195 	memmove(begin, p, end - p);
3196 	memcpy(buf, prefix, strlen(prefix));
3197 
3198 	return 0;
3199 }
3200 
3201 /*
3202  * Block layer IOCTL handler.
3203  *
3204  * @dev Pointer to the block_device structure.
3205  * @mode ignored
3206  * @cmd IOCTL command passed from the user application.
3207  * @arg Argument passed from the user application.
3208  *
3209  * return value
3210  *	0        IOCTL completed successfully.
3211  *	-ENOTTY  IOCTL not supported or invalid driver data
3212  *                 structure pointer.
3213  */
3214 static int mtip_block_ioctl(struct block_device *dev,
3215 			    fmode_t mode,
3216 			    unsigned cmd,
3217 			    unsigned long arg)
3218 {
3219 	struct driver_data *dd = dev->bd_disk->private_data;
3220 
3221 	if (!capable(CAP_SYS_ADMIN))
3222 		return -EACCES;
3223 
3224 	if (!dd)
3225 		return -ENOTTY;
3226 
3227 	if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)))
3228 		return -ENOTTY;
3229 
3230 	switch (cmd) {
3231 	case BLKFLSBUF:
3232 		return -ENOTTY;
3233 	default:
3234 		return mtip_hw_ioctl(dd, cmd, arg);
3235 	}
3236 }
3237 
3238 #ifdef CONFIG_COMPAT
3239 /*
3240  * Block layer compat IOCTL handler.
3241  *
3242  * @dev Pointer to the block_device structure.
3243  * @mode ignored
3244  * @cmd IOCTL command passed from the user application.
3245  * @arg Argument passed from the user application.
3246  *
3247  * return value
3248  *	0        IOCTL completed successfully.
3249  *	-ENOTTY  IOCTL not supported or invalid driver data
3250  *                 structure pointer.
3251  */
3252 static int mtip_block_compat_ioctl(struct block_device *dev,
3253 			    fmode_t mode,
3254 			    unsigned cmd,
3255 			    unsigned long arg)
3256 {
3257 	struct driver_data *dd = dev->bd_disk->private_data;
3258 
3259 	if (!capable(CAP_SYS_ADMIN))
3260 		return -EACCES;
3261 
3262 	if (!dd)
3263 		return -ENOTTY;
3264 
3265 	if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)))
3266 		return -ENOTTY;
3267 
3268 	switch (cmd) {
3269 	case BLKFLSBUF:
3270 		return -ENOTTY;
3271 	case HDIO_DRIVE_TASKFILE: {
3272 		struct mtip_compat_ide_task_request_s __user *compat_req_task;
3273 		ide_task_request_t req_task;
3274 		int compat_tasksize, outtotal, ret;
3275 
3276 		compat_tasksize =
3277 			sizeof(struct mtip_compat_ide_task_request_s);
3278 
3279 		compat_req_task =
3280 			(struct mtip_compat_ide_task_request_s __user *) arg;
3281 
3282 		if (copy_from_user(&req_task, (void __user *) arg,
3283 			compat_tasksize - (2 * sizeof(compat_long_t))))
3284 			return -EFAULT;
3285 
3286 		if (get_user(req_task.out_size, &compat_req_task->out_size))
3287 			return -EFAULT;
3288 
3289 		if (get_user(req_task.in_size, &compat_req_task->in_size))
3290 			return -EFAULT;
3291 
3292 		outtotal = sizeof(struct mtip_compat_ide_task_request_s);
3293 
3294 		ret = exec_drive_taskfile(dd, (void __user *) arg,
3295 						&req_task, outtotal);
3296 
3297 		if (copy_to_user((void __user *) arg, &req_task,
3298 				compat_tasksize -
3299 				(2 * sizeof(compat_long_t))))
3300 			return -EFAULT;
3301 
3302 		if (put_user(req_task.out_size, &compat_req_task->out_size))
3303 			return -EFAULT;
3304 
3305 		if (put_user(req_task.in_size, &compat_req_task->in_size))
3306 			return -EFAULT;
3307 
3308 		return ret;
3309 	}
3310 	default:
3311 		return mtip_hw_ioctl(dd, cmd, arg);
3312 	}
3313 }
3314 #endif
3315 
3316 /*
3317  * Obtain the geometry of the device.
3318  *
3319  * You may think that this function is obsolete, but some applications,
3320  * fdisk for example still used CHS values. This function describes the
3321  * device as having 224 heads and 56 sectors per cylinder. These values are
3322  * chosen so that each cylinder is aligned on a 4KB boundary. Since a
3323  * partition is described in terms of a start and end cylinder this means
3324  * that each partition is also 4KB aligned. Non-aligned partitions adversely
3325  * affects performance.
3326  *
3327  * @dev Pointer to the block_device strucutre.
3328  * @geo Pointer to a hd_geometry structure.
3329  *
3330  * return value
3331  *	0       Operation completed successfully.
3332  *	-ENOTTY An error occurred while reading the drive capacity.
3333  */
3334 static int mtip_block_getgeo(struct block_device *dev,
3335 				struct hd_geometry *geo)
3336 {
3337 	struct driver_data *dd = dev->bd_disk->private_data;
3338 	sector_t capacity;
3339 
3340 	if (!dd)
3341 		return -ENOTTY;
3342 
3343 	if (!(mtip_hw_get_capacity(dd, &capacity))) {
3344 		dev_warn(&dd->pdev->dev,
3345 			"Could not get drive capacity.\n");
3346 		return -ENOTTY;
3347 	}
3348 
3349 	geo->heads = 224;
3350 	geo->sectors = 56;
3351 	sector_div(capacity, (geo->heads * geo->sectors));
3352 	geo->cylinders = capacity;
3353 	return 0;
3354 }
3355 
3356 static int mtip_block_open(struct block_device *dev, fmode_t mode)
3357 {
3358 	struct driver_data *dd;
3359 
3360 	if (dev && dev->bd_disk) {
3361 		dd = (struct driver_data *) dev->bd_disk->private_data;
3362 
3363 		if (dd) {
3364 			if (test_bit(MTIP_DDF_REMOVAL_BIT,
3365 							&dd->dd_flag)) {
3366 				return -ENODEV;
3367 			}
3368 			return 0;
3369 		}
3370 	}
3371 	return -ENODEV;
3372 }
3373 
3374 static void mtip_block_release(struct gendisk *disk, fmode_t mode)
3375 {
3376 }
3377 
3378 /*
3379  * Block device operation function.
3380  *
3381  * This structure contains pointers to the functions required by the block
3382  * layer.
3383  */
3384 static const struct block_device_operations mtip_block_ops = {
3385 	.open		= mtip_block_open,
3386 	.release	= mtip_block_release,
3387 	.ioctl		= mtip_block_ioctl,
3388 #ifdef CONFIG_COMPAT
3389 	.compat_ioctl	= mtip_block_compat_ioctl,
3390 #endif
3391 	.getgeo		= mtip_block_getgeo,
3392 	.owner		= THIS_MODULE
3393 };
3394 
3395 static inline bool is_se_active(struct driver_data *dd)
3396 {
3397 	if (unlikely(test_bit(MTIP_PF_SE_ACTIVE_BIT, &dd->port->flags))) {
3398 		if (dd->port->ic_pause_timer) {
3399 			unsigned long to = dd->port->ic_pause_timer +
3400 							msecs_to_jiffies(1000);
3401 			if (time_after(jiffies, to)) {
3402 				clear_bit(MTIP_PF_SE_ACTIVE_BIT,
3403 							&dd->port->flags);
3404 				clear_bit(MTIP_DDF_SEC_LOCK_BIT, &dd->dd_flag);
3405 				dd->port->ic_pause_timer = 0;
3406 				wake_up_interruptible(&dd->port->svc_wait);
3407 				return false;
3408 			}
3409 		}
3410 		return true;
3411 	}
3412 	return false;
3413 }
3414 
3415 static inline bool is_stopped(struct driver_data *dd, struct request *rq)
3416 {
3417 	if (likely(!(dd->dd_flag & MTIP_DDF_STOP_IO)))
3418 		return false;
3419 
3420 	if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag))
3421 		return true;
3422 	if (test_bit(MTIP_DDF_OVER_TEMP_BIT, &dd->dd_flag))
3423 		return true;
3424 	if (test_bit(MTIP_DDF_WRITE_PROTECT_BIT, &dd->dd_flag) &&
3425 	    rq_data_dir(rq))
3426 		return true;
3427 	if (test_bit(MTIP_DDF_SEC_LOCK_BIT, &dd->dd_flag))
3428 		return true;
3429 	if (test_bit(MTIP_DDF_REBUILD_FAILED_BIT, &dd->dd_flag))
3430 		return true;
3431 
3432 	return false;
3433 }
3434 
3435 static bool mtip_check_unal_depth(struct blk_mq_hw_ctx *hctx,
3436 				  struct request *rq)
3437 {
3438 	struct driver_data *dd = hctx->queue->queuedata;
3439 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3440 
3441 	if (rq_data_dir(rq) == READ || !dd->unal_qdepth)
3442 		return false;
3443 
3444 	/*
3445 	 * If unaligned depth must be limited on this controller, mark it
3446 	 * as unaligned if the IO isn't on a 4k boundary (start of length).
3447 	 */
3448 	if (blk_rq_sectors(rq) <= 64) {
3449 		if ((blk_rq_pos(rq) & 7) || (blk_rq_sectors(rq) & 7))
3450 			cmd->unaligned = 1;
3451 	}
3452 
3453 	if (cmd->unaligned && atomic_dec_if_positive(&dd->port->cmd_slot_unal) >= 0)
3454 		return true;
3455 
3456 	return false;
3457 }
3458 
3459 static blk_status_t mtip_issue_reserved_cmd(struct blk_mq_hw_ctx *hctx,
3460 		struct request *rq)
3461 {
3462 	struct driver_data *dd = hctx->queue->queuedata;
3463 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3464 	struct mtip_int_cmd *icmd = cmd->icmd;
3465 	struct mtip_cmd_hdr *hdr =
3466 		dd->port->command_list + sizeof(struct mtip_cmd_hdr) * rq->tag;
3467 	struct mtip_cmd_sg *command_sg;
3468 
3469 	if (mtip_commands_active(dd->port))
3470 		return BLK_STS_DEV_RESOURCE;
3471 
3472 	hdr->ctba = cpu_to_le32(cmd->command_dma & 0xFFFFFFFF);
3473 	if (test_bit(MTIP_PF_HOST_CAP_64, &dd->port->flags))
3474 		hdr->ctbau = cpu_to_le32((cmd->command_dma >> 16) >> 16);
3475 	/* Populate the SG list */
3476 	hdr->opts = cpu_to_le32(icmd->opts | icmd->fis_len);
3477 	if (icmd->buf_len) {
3478 		command_sg = cmd->command + AHCI_CMD_TBL_HDR_SZ;
3479 
3480 		command_sg->info = cpu_to_le32((icmd->buf_len-1) & 0x3FFFFF);
3481 		command_sg->dba	= cpu_to_le32(icmd->buffer & 0xFFFFFFFF);
3482 		command_sg->dba_upper =
3483 			cpu_to_le32((icmd->buffer >> 16) >> 16);
3484 
3485 		hdr->opts |= cpu_to_le32((1 << 16));
3486 	}
3487 
3488 	/* Populate the command header */
3489 	hdr->byte_count = 0;
3490 
3491 	blk_mq_start_request(rq);
3492 	mtip_issue_non_ncq_command(dd->port, rq->tag);
3493 	return 0;
3494 }
3495 
3496 static blk_status_t mtip_queue_rq(struct blk_mq_hw_ctx *hctx,
3497 			 const struct blk_mq_queue_data *bd)
3498 {
3499 	struct driver_data *dd = hctx->queue->queuedata;
3500 	struct request *rq = bd->rq;
3501 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3502 
3503 	if (blk_rq_is_passthrough(rq))
3504 		return mtip_issue_reserved_cmd(hctx, rq);
3505 
3506 	if (unlikely(mtip_check_unal_depth(hctx, rq)))
3507 		return BLK_STS_DEV_RESOURCE;
3508 
3509 	if (is_se_active(dd) || is_stopped(dd, rq))
3510 		return BLK_STS_IOERR;
3511 
3512 	blk_mq_start_request(rq);
3513 
3514 	mtip_hw_submit_io(dd, rq, cmd, hctx);
3515 	return BLK_STS_OK;
3516 }
3517 
3518 static void mtip_free_cmd(struct blk_mq_tag_set *set, struct request *rq,
3519 			  unsigned int hctx_idx)
3520 {
3521 	struct driver_data *dd = set->driver_data;
3522 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3523 
3524 	if (!cmd->command)
3525 		return;
3526 
3527 	dma_free_coherent(&dd->pdev->dev, CMD_DMA_ALLOC_SZ, cmd->command,
3528 			  cmd->command_dma);
3529 }
3530 
3531 static int mtip_init_cmd(struct blk_mq_tag_set *set, struct request *rq,
3532 			 unsigned int hctx_idx, unsigned int numa_node)
3533 {
3534 	struct driver_data *dd = set->driver_data;
3535 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3536 
3537 	cmd->command = dma_alloc_coherent(&dd->pdev->dev, CMD_DMA_ALLOC_SZ,
3538 			&cmd->command_dma, GFP_KERNEL);
3539 	if (!cmd->command)
3540 		return -ENOMEM;
3541 
3542 	memset(cmd->command, 0, CMD_DMA_ALLOC_SZ);
3543 
3544 	sg_init_table(cmd->sg, MTIP_MAX_SG);
3545 	return 0;
3546 }
3547 
3548 static enum blk_eh_timer_return mtip_cmd_timeout(struct request *req,
3549 								bool reserved)
3550 {
3551 	struct driver_data *dd = req->q->queuedata;
3552 
3553 	if (reserved) {
3554 		struct mtip_cmd *cmd = blk_mq_rq_to_pdu(req);
3555 
3556 		cmd->status = BLK_STS_TIMEOUT;
3557 		blk_mq_complete_request(req);
3558 		return BLK_EH_DONE;
3559 	}
3560 
3561 	if (test_bit(req->tag, dd->port->cmds_to_issue))
3562 		goto exit_handler;
3563 
3564 	if (test_and_set_bit(MTIP_PF_TO_ACTIVE_BIT, &dd->port->flags))
3565 		goto exit_handler;
3566 
3567 	wake_up_interruptible(&dd->port->svc_wait);
3568 exit_handler:
3569 	return BLK_EH_RESET_TIMER;
3570 }
3571 
3572 static const struct blk_mq_ops mtip_mq_ops = {
3573 	.queue_rq	= mtip_queue_rq,
3574 	.init_request	= mtip_init_cmd,
3575 	.exit_request	= mtip_free_cmd,
3576 	.complete	= mtip_softirq_done_fn,
3577 	.timeout        = mtip_cmd_timeout,
3578 };
3579 
3580 /*
3581  * Block layer initialization function.
3582  *
3583  * This function is called once by the PCI layer for each P320
3584  * device that is connected to the system.
3585  *
3586  * @dd Pointer to the driver data structure.
3587  *
3588  * return value
3589  *	0 on success else an error code.
3590  */
3591 static int mtip_block_initialize(struct driver_data *dd)
3592 {
3593 	int rv = 0, wait_for_rebuild = 0;
3594 	sector_t capacity;
3595 	unsigned int index = 0;
3596 	struct kobject *kobj;
3597 
3598 	if (dd->disk)
3599 		goto skip_create_disk; /* hw init done, before rebuild */
3600 
3601 	if (mtip_hw_init(dd)) {
3602 		rv = -EINVAL;
3603 		goto protocol_init_error;
3604 	}
3605 
3606 	dd->disk = alloc_disk_node(MTIP_MAX_MINORS, dd->numa_node);
3607 	if (dd->disk  == NULL) {
3608 		dev_err(&dd->pdev->dev,
3609 			"Unable to allocate gendisk structure\n");
3610 		rv = -EINVAL;
3611 		goto alloc_disk_error;
3612 	}
3613 
3614 	rv = ida_alloc(&rssd_index_ida, GFP_KERNEL);
3615 	if (rv < 0)
3616 		goto ida_get_error;
3617 	index = rv;
3618 
3619 	rv = rssd_disk_name_format("rssd",
3620 				index,
3621 				dd->disk->disk_name,
3622 				DISK_NAME_LEN);
3623 	if (rv)
3624 		goto disk_index_error;
3625 
3626 	dd->disk->major		= dd->major;
3627 	dd->disk->first_minor	= index * MTIP_MAX_MINORS;
3628 	dd->disk->minors 	= MTIP_MAX_MINORS;
3629 	dd->disk->fops		= &mtip_block_ops;
3630 	dd->disk->private_data	= dd;
3631 	dd->index		= index;
3632 
3633 	mtip_hw_debugfs_init(dd);
3634 
3635 	memset(&dd->tags, 0, sizeof(dd->tags));
3636 	dd->tags.ops = &mtip_mq_ops;
3637 	dd->tags.nr_hw_queues = 1;
3638 	dd->tags.queue_depth = MTIP_MAX_COMMAND_SLOTS;
3639 	dd->tags.reserved_tags = 1;
3640 	dd->tags.cmd_size = sizeof(struct mtip_cmd);
3641 	dd->tags.numa_node = dd->numa_node;
3642 	dd->tags.flags = BLK_MQ_F_SHOULD_MERGE;
3643 	dd->tags.driver_data = dd;
3644 	dd->tags.timeout = MTIP_NCQ_CMD_TIMEOUT_MS;
3645 
3646 	rv = blk_mq_alloc_tag_set(&dd->tags);
3647 	if (rv) {
3648 		dev_err(&dd->pdev->dev,
3649 			"Unable to allocate request queue\n");
3650 		goto block_queue_alloc_tag_error;
3651 	}
3652 
3653 	/* Allocate the request queue. */
3654 	dd->queue = blk_mq_init_queue(&dd->tags);
3655 	if (IS_ERR(dd->queue)) {
3656 		dev_err(&dd->pdev->dev,
3657 			"Unable to allocate request queue\n");
3658 		rv = -ENOMEM;
3659 		goto block_queue_alloc_init_error;
3660 	}
3661 
3662 	dd->disk->queue		= dd->queue;
3663 	dd->queue->queuedata	= dd;
3664 
3665 skip_create_disk:
3666 	/* Initialize the protocol layer. */
3667 	wait_for_rebuild = mtip_hw_get_identify(dd);
3668 	if (wait_for_rebuild < 0) {
3669 		dev_err(&dd->pdev->dev,
3670 			"Protocol layer initialization failed\n");
3671 		rv = -EINVAL;
3672 		goto init_hw_cmds_error;
3673 	}
3674 
3675 	/*
3676 	 * if rebuild pending, start the service thread, and delay the block
3677 	 * queue creation and device_add_disk()
3678 	 */
3679 	if (wait_for_rebuild == MTIP_FTL_REBUILD_MAGIC)
3680 		goto start_service_thread;
3681 
3682 	/* Set device limits. */
3683 	blk_queue_flag_set(QUEUE_FLAG_NONROT, dd->queue);
3684 	blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, dd->queue);
3685 	blk_queue_max_segments(dd->queue, MTIP_MAX_SG);
3686 	blk_queue_physical_block_size(dd->queue, 4096);
3687 	blk_queue_max_hw_sectors(dd->queue, 0xffff);
3688 	blk_queue_max_segment_size(dd->queue, 0x400000);
3689 	blk_queue_io_min(dd->queue, 4096);
3690 
3691 	/* Set the capacity of the device in 512 byte sectors. */
3692 	if (!(mtip_hw_get_capacity(dd, &capacity))) {
3693 		dev_warn(&dd->pdev->dev,
3694 			"Could not read drive capacity\n");
3695 		rv = -EIO;
3696 		goto read_capacity_error;
3697 	}
3698 	set_capacity(dd->disk, capacity);
3699 
3700 	/* Enable the block device and add it to /dev */
3701 	device_add_disk(&dd->pdev->dev, dd->disk, NULL);
3702 
3703 	dd->bdev = bdget_disk(dd->disk, 0);
3704 	/*
3705 	 * Now that the disk is active, initialize any sysfs attributes
3706 	 * managed by the protocol layer.
3707 	 */
3708 	kobj = kobject_get(&disk_to_dev(dd->disk)->kobj);
3709 	if (kobj) {
3710 		mtip_hw_sysfs_init(dd, kobj);
3711 		kobject_put(kobj);
3712 	}
3713 
3714 	if (dd->mtip_svc_handler) {
3715 		set_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag);
3716 		return rv; /* service thread created for handling rebuild */
3717 	}
3718 
3719 start_service_thread:
3720 	dd->mtip_svc_handler = kthread_create_on_node(mtip_service_thread,
3721 						dd, dd->numa_node,
3722 						"mtip_svc_thd_%02d", index);
3723 
3724 	if (IS_ERR(dd->mtip_svc_handler)) {
3725 		dev_err(&dd->pdev->dev, "service thread failed to start\n");
3726 		dd->mtip_svc_handler = NULL;
3727 		rv = -EFAULT;
3728 		goto kthread_run_error;
3729 	}
3730 	wake_up_process(dd->mtip_svc_handler);
3731 	if (wait_for_rebuild == MTIP_FTL_REBUILD_MAGIC)
3732 		rv = wait_for_rebuild;
3733 
3734 	return rv;
3735 
3736 kthread_run_error:
3737 	bdput(dd->bdev);
3738 	dd->bdev = NULL;
3739 
3740 	/* Delete our gendisk. This also removes the device from /dev */
3741 	del_gendisk(dd->disk);
3742 
3743 read_capacity_error:
3744 init_hw_cmds_error:
3745 	blk_cleanup_queue(dd->queue);
3746 block_queue_alloc_init_error:
3747 	blk_mq_free_tag_set(&dd->tags);
3748 block_queue_alloc_tag_error:
3749 	mtip_hw_debugfs_exit(dd);
3750 disk_index_error:
3751 	ida_free(&rssd_index_ida, index);
3752 
3753 ida_get_error:
3754 	put_disk(dd->disk);
3755 
3756 alloc_disk_error:
3757 	mtip_hw_exit(dd); /* De-initialize the protocol layer. */
3758 
3759 protocol_init_error:
3760 	return rv;
3761 }
3762 
3763 static bool mtip_no_dev_cleanup(struct request *rq, void *data, bool reserv)
3764 {
3765 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3766 
3767 	cmd->status = BLK_STS_IOERR;
3768 	blk_mq_complete_request(rq);
3769 	return true;
3770 }
3771 
3772 /*
3773  * Block layer deinitialization function.
3774  *
3775  * Called by the PCI layer as each P320 device is removed.
3776  *
3777  * @dd Pointer to the driver data structure.
3778  *
3779  * return value
3780  *	0
3781  */
3782 static int mtip_block_remove(struct driver_data *dd)
3783 {
3784 	struct kobject *kobj;
3785 
3786 	mtip_hw_debugfs_exit(dd);
3787 
3788 	if (dd->mtip_svc_handler) {
3789 		set_bit(MTIP_PF_SVC_THD_STOP_BIT, &dd->port->flags);
3790 		wake_up_interruptible(&dd->port->svc_wait);
3791 		kthread_stop(dd->mtip_svc_handler);
3792 	}
3793 
3794 	/* Clean up the sysfs attributes, if created */
3795 	if (test_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag)) {
3796 		kobj = kobject_get(&disk_to_dev(dd->disk)->kobj);
3797 		if (kobj) {
3798 			mtip_hw_sysfs_exit(dd, kobj);
3799 			kobject_put(kobj);
3800 		}
3801 	}
3802 
3803 	if (!dd->sr) {
3804 		/*
3805 		 * Explicitly wait here for IOs to quiesce,
3806 		 * as mtip_standby_drive usually won't wait for IOs.
3807 		 */
3808 		if (!mtip_quiesce_io(dd->port, MTIP_QUIESCE_IO_TIMEOUT_MS))
3809 			mtip_standby_drive(dd);
3810 	}
3811 	else
3812 		dev_info(&dd->pdev->dev, "device %s surprise removal\n",
3813 						dd->disk->disk_name);
3814 
3815 	blk_freeze_queue_start(dd->queue);
3816 	blk_mq_quiesce_queue(dd->queue);
3817 	blk_mq_tagset_busy_iter(&dd->tags, mtip_no_dev_cleanup, dd);
3818 	blk_mq_unquiesce_queue(dd->queue);
3819 
3820 	/*
3821 	 * Delete our gendisk structure. This also removes the device
3822 	 * from /dev
3823 	 */
3824 	if (dd->bdev) {
3825 		bdput(dd->bdev);
3826 		dd->bdev = NULL;
3827 	}
3828 	if (dd->disk) {
3829 		if (test_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag))
3830 			del_gendisk(dd->disk);
3831 		if (dd->disk->queue) {
3832 			blk_cleanup_queue(dd->queue);
3833 			blk_mq_free_tag_set(&dd->tags);
3834 			dd->queue = NULL;
3835 		}
3836 		put_disk(dd->disk);
3837 	}
3838 	dd->disk  = NULL;
3839 
3840 	ida_free(&rssd_index_ida, dd->index);
3841 
3842 	/* De-initialize the protocol layer. */
3843 	mtip_hw_exit(dd);
3844 
3845 	return 0;
3846 }
3847 
3848 /*
3849  * Function called by the PCI layer when just before the
3850  * machine shuts down.
3851  *
3852  * If a protocol layer shutdown function is present it will be called
3853  * by this function.
3854  *
3855  * @dd Pointer to the driver data structure.
3856  *
3857  * return value
3858  *	0
3859  */
3860 static int mtip_block_shutdown(struct driver_data *dd)
3861 {
3862 	mtip_hw_shutdown(dd);
3863 
3864 	/* Delete our gendisk structure, and cleanup the blk queue. */
3865 	if (dd->disk) {
3866 		dev_info(&dd->pdev->dev,
3867 			"Shutting down %s ...\n", dd->disk->disk_name);
3868 
3869 		if (test_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag))
3870 			del_gendisk(dd->disk);
3871 		if (dd->disk->queue) {
3872 			blk_cleanup_queue(dd->queue);
3873 			blk_mq_free_tag_set(&dd->tags);
3874 		}
3875 		put_disk(dd->disk);
3876 		dd->disk  = NULL;
3877 		dd->queue = NULL;
3878 	}
3879 
3880 	ida_free(&rssd_index_ida, dd->index);
3881 	return 0;
3882 }
3883 
3884 static int mtip_block_suspend(struct driver_data *dd)
3885 {
3886 	dev_info(&dd->pdev->dev,
3887 		"Suspending %s ...\n", dd->disk->disk_name);
3888 	mtip_hw_suspend(dd);
3889 	return 0;
3890 }
3891 
3892 static int mtip_block_resume(struct driver_data *dd)
3893 {
3894 	dev_info(&dd->pdev->dev, "Resuming %s ...\n",
3895 		dd->disk->disk_name);
3896 	mtip_hw_resume(dd);
3897 	return 0;
3898 }
3899 
3900 static void drop_cpu(int cpu)
3901 {
3902 	cpu_use[cpu]--;
3903 }
3904 
3905 static int get_least_used_cpu_on_node(int node)
3906 {
3907 	int cpu, least_used_cpu, least_cnt;
3908 	const struct cpumask *node_mask;
3909 
3910 	node_mask = cpumask_of_node(node);
3911 	least_used_cpu = cpumask_first(node_mask);
3912 	least_cnt = cpu_use[least_used_cpu];
3913 	cpu = least_used_cpu;
3914 
3915 	for_each_cpu(cpu, node_mask) {
3916 		if (cpu_use[cpu] < least_cnt) {
3917 			least_used_cpu = cpu;
3918 			least_cnt = cpu_use[cpu];
3919 		}
3920 	}
3921 	cpu_use[least_used_cpu]++;
3922 	return least_used_cpu;
3923 }
3924 
3925 /* Helper for selecting a node in round robin mode */
3926 static inline int mtip_get_next_rr_node(void)
3927 {
3928 	static int next_node = NUMA_NO_NODE;
3929 
3930 	if (next_node == NUMA_NO_NODE) {
3931 		next_node = first_online_node;
3932 		return next_node;
3933 	}
3934 
3935 	next_node = next_online_node(next_node);
3936 	if (next_node == MAX_NUMNODES)
3937 		next_node = first_online_node;
3938 	return next_node;
3939 }
3940 
3941 static DEFINE_HANDLER(0);
3942 static DEFINE_HANDLER(1);
3943 static DEFINE_HANDLER(2);
3944 static DEFINE_HANDLER(3);
3945 static DEFINE_HANDLER(4);
3946 static DEFINE_HANDLER(5);
3947 static DEFINE_HANDLER(6);
3948 static DEFINE_HANDLER(7);
3949 
3950 static void mtip_disable_link_opts(struct driver_data *dd, struct pci_dev *pdev)
3951 {
3952 	int pos;
3953 	unsigned short pcie_dev_ctrl;
3954 
3955 	pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
3956 	if (pos) {
3957 		pci_read_config_word(pdev,
3958 			pos + PCI_EXP_DEVCTL,
3959 			&pcie_dev_ctrl);
3960 		if (pcie_dev_ctrl & (1 << 11) ||
3961 		    pcie_dev_ctrl & (1 << 4)) {
3962 			dev_info(&dd->pdev->dev,
3963 				"Disabling ERO/No-Snoop on bridge device %04x:%04x\n",
3964 					pdev->vendor, pdev->device);
3965 			pcie_dev_ctrl &= ~(PCI_EXP_DEVCTL_NOSNOOP_EN |
3966 						PCI_EXP_DEVCTL_RELAX_EN);
3967 			pci_write_config_word(pdev,
3968 				pos + PCI_EXP_DEVCTL,
3969 				pcie_dev_ctrl);
3970 		}
3971 	}
3972 }
3973 
3974 static void mtip_fix_ero_nosnoop(struct driver_data *dd, struct pci_dev *pdev)
3975 {
3976 	/*
3977 	 * This workaround is specific to AMD/ATI chipset with a PCI upstream
3978 	 * device with device id 0x5aXX
3979 	 */
3980 	if (pdev->bus && pdev->bus->self) {
3981 		if (pdev->bus->self->vendor == PCI_VENDOR_ID_ATI &&
3982 		    ((pdev->bus->self->device & 0xff00) == 0x5a00)) {
3983 			mtip_disable_link_opts(dd, pdev->bus->self);
3984 		} else {
3985 			/* Check further up the topology */
3986 			struct pci_dev *parent_dev = pdev->bus->self;
3987 			if (parent_dev->bus &&
3988 				parent_dev->bus->parent &&
3989 				parent_dev->bus->parent->self &&
3990 				parent_dev->bus->parent->self->vendor ==
3991 					 PCI_VENDOR_ID_ATI &&
3992 				(parent_dev->bus->parent->self->device &
3993 					0xff00) == 0x5a00) {
3994 				mtip_disable_link_opts(dd,
3995 					parent_dev->bus->parent->self);
3996 			}
3997 		}
3998 	}
3999 }
4000 
4001 /*
4002  * Called for each supported PCI device detected.
4003  *
4004  * This function allocates the private data structure, enables the
4005  * PCI device and then calls the block layer initialization function.
4006  *
4007  * return value
4008  *	0 on success else an error code.
4009  */
4010 static int mtip_pci_probe(struct pci_dev *pdev,
4011 			const struct pci_device_id *ent)
4012 {
4013 	int rv = 0;
4014 	struct driver_data *dd = NULL;
4015 	char cpu_list[256];
4016 	const struct cpumask *node_mask;
4017 	int cpu, i = 0, j = 0;
4018 	int my_node = NUMA_NO_NODE;
4019 	unsigned long flags;
4020 
4021 	/* Allocate memory for this devices private data. */
4022 	my_node = pcibus_to_node(pdev->bus);
4023 	if (my_node != NUMA_NO_NODE) {
4024 		if (!node_online(my_node))
4025 			my_node = mtip_get_next_rr_node();
4026 	} else {
4027 		dev_info(&pdev->dev, "Kernel not reporting proximity, choosing a node\n");
4028 		my_node = mtip_get_next_rr_node();
4029 	}
4030 	dev_info(&pdev->dev, "NUMA node %d (closest: %d,%d, probe on %d:%d)\n",
4031 		my_node, pcibus_to_node(pdev->bus), dev_to_node(&pdev->dev),
4032 		cpu_to_node(raw_smp_processor_id()), raw_smp_processor_id());
4033 
4034 	dd = kzalloc_node(sizeof(struct driver_data), GFP_KERNEL, my_node);
4035 	if (dd == NULL) {
4036 		dev_err(&pdev->dev,
4037 			"Unable to allocate memory for driver data\n");
4038 		return -ENOMEM;
4039 	}
4040 
4041 	/* Attach the private data to this PCI device.  */
4042 	pci_set_drvdata(pdev, dd);
4043 
4044 	rv = pcim_enable_device(pdev);
4045 	if (rv < 0) {
4046 		dev_err(&pdev->dev, "Unable to enable device\n");
4047 		goto iomap_err;
4048 	}
4049 
4050 	/* Map BAR5 to memory. */
4051 	rv = pcim_iomap_regions(pdev, 1 << MTIP_ABAR, MTIP_DRV_NAME);
4052 	if (rv < 0) {
4053 		dev_err(&pdev->dev, "Unable to map regions\n");
4054 		goto iomap_err;
4055 	}
4056 
4057 	rv = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
4058 	if (rv) {
4059 		dev_warn(&pdev->dev, "64-bit DMA enable failed\n");
4060 		goto setmask_err;
4061 	}
4062 
4063 	/* Copy the info we may need later into the private data structure. */
4064 	dd->major	= mtip_major;
4065 	dd->instance	= instance;
4066 	dd->pdev	= pdev;
4067 	dd->numa_node	= my_node;
4068 
4069 	INIT_LIST_HEAD(&dd->online_list);
4070 	INIT_LIST_HEAD(&dd->remove_list);
4071 
4072 	memset(dd->workq_name, 0, 32);
4073 	snprintf(dd->workq_name, 31, "mtipq%d", dd->instance);
4074 
4075 	dd->isr_workq = create_workqueue(dd->workq_name);
4076 	if (!dd->isr_workq) {
4077 		dev_warn(&pdev->dev, "Can't create wq %d\n", dd->instance);
4078 		rv = -ENOMEM;
4079 		goto setmask_err;
4080 	}
4081 
4082 	memset(cpu_list, 0, sizeof(cpu_list));
4083 
4084 	node_mask = cpumask_of_node(dd->numa_node);
4085 	if (!cpumask_empty(node_mask)) {
4086 		for_each_cpu(cpu, node_mask)
4087 		{
4088 			snprintf(&cpu_list[j], 256 - j, "%d ", cpu);
4089 			j = strlen(cpu_list);
4090 		}
4091 
4092 		dev_info(&pdev->dev, "Node %d on package %d has %d cpu(s): %s\n",
4093 			dd->numa_node,
4094 			topology_physical_package_id(cpumask_first(node_mask)),
4095 			nr_cpus_node(dd->numa_node),
4096 			cpu_list);
4097 	} else
4098 		dev_dbg(&pdev->dev, "mtip32xx: node_mask empty\n");
4099 
4100 	dd->isr_binding = get_least_used_cpu_on_node(dd->numa_node);
4101 	dev_info(&pdev->dev, "Initial IRQ binding node:cpu %d:%d\n",
4102 		cpu_to_node(dd->isr_binding), dd->isr_binding);
4103 
4104 	/* first worker context always runs in ISR */
4105 	dd->work[0].cpu_binding = dd->isr_binding;
4106 	dd->work[1].cpu_binding = get_least_used_cpu_on_node(dd->numa_node);
4107 	dd->work[2].cpu_binding = get_least_used_cpu_on_node(dd->numa_node);
4108 	dd->work[3].cpu_binding = dd->work[0].cpu_binding;
4109 	dd->work[4].cpu_binding = dd->work[1].cpu_binding;
4110 	dd->work[5].cpu_binding = dd->work[2].cpu_binding;
4111 	dd->work[6].cpu_binding = dd->work[2].cpu_binding;
4112 	dd->work[7].cpu_binding = dd->work[1].cpu_binding;
4113 
4114 	/* Log the bindings */
4115 	for_each_present_cpu(cpu) {
4116 		memset(cpu_list, 0, sizeof(cpu_list));
4117 		for (i = 0, j = 0; i < MTIP_MAX_SLOT_GROUPS; i++) {
4118 			if (dd->work[i].cpu_binding == cpu) {
4119 				snprintf(&cpu_list[j], 256 - j, "%d ", i);
4120 				j = strlen(cpu_list);
4121 			}
4122 		}
4123 		if (j)
4124 			dev_info(&pdev->dev, "CPU %d: WQs %s\n", cpu, cpu_list);
4125 	}
4126 
4127 	INIT_WORK(&dd->work[0].work, mtip_workq_sdbf0);
4128 	INIT_WORK(&dd->work[1].work, mtip_workq_sdbf1);
4129 	INIT_WORK(&dd->work[2].work, mtip_workq_sdbf2);
4130 	INIT_WORK(&dd->work[3].work, mtip_workq_sdbf3);
4131 	INIT_WORK(&dd->work[4].work, mtip_workq_sdbf4);
4132 	INIT_WORK(&dd->work[5].work, mtip_workq_sdbf5);
4133 	INIT_WORK(&dd->work[6].work, mtip_workq_sdbf6);
4134 	INIT_WORK(&dd->work[7].work, mtip_workq_sdbf7);
4135 
4136 	pci_set_master(pdev);
4137 	rv = pci_enable_msi(pdev);
4138 	if (rv) {
4139 		dev_warn(&pdev->dev,
4140 			"Unable to enable MSI interrupt.\n");
4141 		goto msi_initialize_err;
4142 	}
4143 
4144 	mtip_fix_ero_nosnoop(dd, pdev);
4145 
4146 	/* Initialize the block layer. */
4147 	rv = mtip_block_initialize(dd);
4148 	if (rv < 0) {
4149 		dev_err(&pdev->dev,
4150 			"Unable to initialize block layer\n");
4151 		goto block_initialize_err;
4152 	}
4153 
4154 	/*
4155 	 * Increment the instance count so that each device has a unique
4156 	 * instance number.
4157 	 */
4158 	instance++;
4159 	if (rv != MTIP_FTL_REBUILD_MAGIC)
4160 		set_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag);
4161 	else
4162 		rv = 0; /* device in rebuild state, return 0 from probe */
4163 
4164 	/* Add to online list even if in ftl rebuild */
4165 	spin_lock_irqsave(&dev_lock, flags);
4166 	list_add(&dd->online_list, &online_list);
4167 	spin_unlock_irqrestore(&dev_lock, flags);
4168 
4169 	goto done;
4170 
4171 block_initialize_err:
4172 	pci_disable_msi(pdev);
4173 
4174 msi_initialize_err:
4175 	if (dd->isr_workq) {
4176 		flush_workqueue(dd->isr_workq);
4177 		destroy_workqueue(dd->isr_workq);
4178 		drop_cpu(dd->work[0].cpu_binding);
4179 		drop_cpu(dd->work[1].cpu_binding);
4180 		drop_cpu(dd->work[2].cpu_binding);
4181 	}
4182 setmask_err:
4183 	pcim_iounmap_regions(pdev, 1 << MTIP_ABAR);
4184 
4185 iomap_err:
4186 	kfree(dd);
4187 	pci_set_drvdata(pdev, NULL);
4188 	return rv;
4189 done:
4190 	return rv;
4191 }
4192 
4193 /*
4194  * Called for each probed device when the device is removed or the
4195  * driver is unloaded.
4196  *
4197  * return value
4198  *	None
4199  */
4200 static void mtip_pci_remove(struct pci_dev *pdev)
4201 {
4202 	struct driver_data *dd = pci_get_drvdata(pdev);
4203 	unsigned long flags, to;
4204 
4205 	set_bit(MTIP_DDF_REMOVAL_BIT, &dd->dd_flag);
4206 
4207 	spin_lock_irqsave(&dev_lock, flags);
4208 	list_del_init(&dd->online_list);
4209 	list_add(&dd->remove_list, &removing_list);
4210 	spin_unlock_irqrestore(&dev_lock, flags);
4211 
4212 	mtip_check_surprise_removal(pdev);
4213 	synchronize_irq(dd->pdev->irq);
4214 
4215 	/* Spin until workers are done */
4216 	to = jiffies + msecs_to_jiffies(4000);
4217 	do {
4218 		msleep(20);
4219 	} while (atomic_read(&dd->irq_workers_active) != 0 &&
4220 		time_before(jiffies, to));
4221 
4222 	if (!dd->sr)
4223 		fsync_bdev(dd->bdev);
4224 
4225 	if (atomic_read(&dd->irq_workers_active) != 0) {
4226 		dev_warn(&dd->pdev->dev,
4227 			"Completion workers still active!\n");
4228 	}
4229 
4230 	blk_set_queue_dying(dd->queue);
4231 	set_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag);
4232 
4233 	/* Clean up the block layer. */
4234 	mtip_block_remove(dd);
4235 
4236 	if (dd->isr_workq) {
4237 		flush_workqueue(dd->isr_workq);
4238 		destroy_workqueue(dd->isr_workq);
4239 		drop_cpu(dd->work[0].cpu_binding);
4240 		drop_cpu(dd->work[1].cpu_binding);
4241 		drop_cpu(dd->work[2].cpu_binding);
4242 	}
4243 
4244 	pci_disable_msi(pdev);
4245 
4246 	spin_lock_irqsave(&dev_lock, flags);
4247 	list_del_init(&dd->remove_list);
4248 	spin_unlock_irqrestore(&dev_lock, flags);
4249 
4250 	kfree(dd);
4251 
4252 	pcim_iounmap_regions(pdev, 1 << MTIP_ABAR);
4253 	pci_set_drvdata(pdev, NULL);
4254 }
4255 
4256 /*
4257  * Called for each probed device when the device is suspended.
4258  *
4259  * return value
4260  *	0  Success
4261  *	<0 Error
4262  */
4263 static int mtip_pci_suspend(struct pci_dev *pdev, pm_message_t mesg)
4264 {
4265 	int rv = 0;
4266 	struct driver_data *dd = pci_get_drvdata(pdev);
4267 
4268 	if (!dd) {
4269 		dev_err(&pdev->dev,
4270 			"Driver private datastructure is NULL\n");
4271 		return -EFAULT;
4272 	}
4273 
4274 	set_bit(MTIP_DDF_RESUME_BIT, &dd->dd_flag);
4275 
4276 	/* Disable ports & interrupts then send standby immediate */
4277 	rv = mtip_block_suspend(dd);
4278 	if (rv < 0) {
4279 		dev_err(&pdev->dev,
4280 			"Failed to suspend controller\n");
4281 		return rv;
4282 	}
4283 
4284 	/*
4285 	 * Save the pci config space to pdev structure &
4286 	 * disable the device
4287 	 */
4288 	pci_save_state(pdev);
4289 	pci_disable_device(pdev);
4290 
4291 	/* Move to Low power state*/
4292 	pci_set_power_state(pdev, PCI_D3hot);
4293 
4294 	return rv;
4295 }
4296 
4297 /*
4298  * Called for each probed device when the device is resumed.
4299  *
4300  * return value
4301  *      0  Success
4302  *      <0 Error
4303  */
4304 static int mtip_pci_resume(struct pci_dev *pdev)
4305 {
4306 	int rv = 0;
4307 	struct driver_data *dd;
4308 
4309 	dd = pci_get_drvdata(pdev);
4310 	if (!dd) {
4311 		dev_err(&pdev->dev,
4312 			"Driver private datastructure is NULL\n");
4313 		return -EFAULT;
4314 	}
4315 
4316 	/* Move the device to active State */
4317 	pci_set_power_state(pdev, PCI_D0);
4318 
4319 	/* Restore PCI configuration space */
4320 	pci_restore_state(pdev);
4321 
4322 	/* Enable the PCI device*/
4323 	rv = pcim_enable_device(pdev);
4324 	if (rv < 0) {
4325 		dev_err(&pdev->dev,
4326 			"Failed to enable card during resume\n");
4327 		goto err;
4328 	}
4329 	pci_set_master(pdev);
4330 
4331 	/*
4332 	 * Calls hbaReset, initPort, & startPort function
4333 	 * then enables interrupts
4334 	 */
4335 	rv = mtip_block_resume(dd);
4336 	if (rv < 0)
4337 		dev_err(&pdev->dev, "Unable to resume\n");
4338 
4339 err:
4340 	clear_bit(MTIP_DDF_RESUME_BIT, &dd->dd_flag);
4341 
4342 	return rv;
4343 }
4344 
4345 /*
4346  * Shutdown routine
4347  *
4348  * return value
4349  *      None
4350  */
4351 static void mtip_pci_shutdown(struct pci_dev *pdev)
4352 {
4353 	struct driver_data *dd = pci_get_drvdata(pdev);
4354 	if (dd)
4355 		mtip_block_shutdown(dd);
4356 }
4357 
4358 /* Table of device ids supported by this driver. */
4359 static const struct pci_device_id mtip_pci_tbl[] = {
4360 	{ PCI_DEVICE(PCI_VENDOR_ID_MICRON, P320H_DEVICE_ID) },
4361 	{ PCI_DEVICE(PCI_VENDOR_ID_MICRON, P320M_DEVICE_ID) },
4362 	{ PCI_DEVICE(PCI_VENDOR_ID_MICRON, P320S_DEVICE_ID) },
4363 	{ PCI_DEVICE(PCI_VENDOR_ID_MICRON, P325M_DEVICE_ID) },
4364 	{ PCI_DEVICE(PCI_VENDOR_ID_MICRON, P420H_DEVICE_ID) },
4365 	{ PCI_DEVICE(PCI_VENDOR_ID_MICRON, P420M_DEVICE_ID) },
4366 	{ PCI_DEVICE(PCI_VENDOR_ID_MICRON, P425M_DEVICE_ID) },
4367 	{ 0 }
4368 };
4369 
4370 /* Structure that describes the PCI driver functions. */
4371 static struct pci_driver mtip_pci_driver = {
4372 	.name			= MTIP_DRV_NAME,
4373 	.id_table		= mtip_pci_tbl,
4374 	.probe			= mtip_pci_probe,
4375 	.remove			= mtip_pci_remove,
4376 	.suspend		= mtip_pci_suspend,
4377 	.resume			= mtip_pci_resume,
4378 	.shutdown		= mtip_pci_shutdown,
4379 };
4380 
4381 MODULE_DEVICE_TABLE(pci, mtip_pci_tbl);
4382 
4383 /*
4384  * Module initialization function.
4385  *
4386  * Called once when the module is loaded. This function allocates a major
4387  * block device number to the Cyclone devices and registers the PCI layer
4388  * of the driver.
4389  *
4390  * Return value
4391  *      0 on success else error code.
4392  */
4393 static int __init mtip_init(void)
4394 {
4395 	int error;
4396 
4397 	pr_info(MTIP_DRV_NAME " Version " MTIP_DRV_VERSION "\n");
4398 
4399 	spin_lock_init(&dev_lock);
4400 
4401 	INIT_LIST_HEAD(&online_list);
4402 	INIT_LIST_HEAD(&removing_list);
4403 
4404 	/* Allocate a major block device number to use with this driver. */
4405 	error = register_blkdev(0, MTIP_DRV_NAME);
4406 	if (error <= 0) {
4407 		pr_err("Unable to register block device (%d)\n",
4408 		error);
4409 		return -EBUSY;
4410 	}
4411 	mtip_major = error;
4412 
4413 	dfs_parent = debugfs_create_dir("rssd", NULL);
4414 	if (IS_ERR_OR_NULL(dfs_parent)) {
4415 		pr_warn("Error creating debugfs parent\n");
4416 		dfs_parent = NULL;
4417 	}
4418 	if (dfs_parent) {
4419 		dfs_device_status = debugfs_create_file("device_status",
4420 					0444, dfs_parent, NULL,
4421 					&mtip_device_status_fops);
4422 		if (IS_ERR_OR_NULL(dfs_device_status)) {
4423 			pr_err("Error creating device_status node\n");
4424 			dfs_device_status = NULL;
4425 		}
4426 	}
4427 
4428 	/* Register our PCI operations. */
4429 	error = pci_register_driver(&mtip_pci_driver);
4430 	if (error) {
4431 		debugfs_remove(dfs_parent);
4432 		unregister_blkdev(mtip_major, MTIP_DRV_NAME);
4433 	}
4434 
4435 	return error;
4436 }
4437 
4438 /*
4439  * Module de-initialization function.
4440  *
4441  * Called once when the module is unloaded. This function deallocates
4442  * the major block device number allocated by mtip_init() and
4443  * unregisters the PCI layer of the driver.
4444  *
4445  * Return value
4446  *      none
4447  */
4448 static void __exit mtip_exit(void)
4449 {
4450 	/* Release the allocated major block device number. */
4451 	unregister_blkdev(mtip_major, MTIP_DRV_NAME);
4452 
4453 	/* Unregister the PCI driver. */
4454 	pci_unregister_driver(&mtip_pci_driver);
4455 
4456 	debugfs_remove_recursive(dfs_parent);
4457 }
4458 
4459 MODULE_AUTHOR("Micron Technology, Inc");
4460 MODULE_DESCRIPTION("Micron RealSSD PCIe Block Driver");
4461 MODULE_LICENSE("GPL");
4462 MODULE_VERSION(MTIP_DRV_VERSION);
4463 
4464 module_init(mtip_init);
4465 module_exit(mtip_exit);
4466