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 "mtip32xx.h"
44 
45 #define HW_CMD_SLOT_SZ		(MTIP_MAX_COMMAND_SLOTS * 32)
46 
47 /* DMA region containing RX Fis, Identify, RLE10, and SMART buffers */
48 #define AHCI_RX_FIS_SZ          0x100
49 #define AHCI_RX_FIS_OFFSET      0x0
50 #define AHCI_IDFY_SZ            ATA_SECT_SIZE
51 #define AHCI_IDFY_OFFSET        0x400
52 #define AHCI_SECTBUF_SZ         ATA_SECT_SIZE
53 #define AHCI_SECTBUF_OFFSET     0x800
54 #define AHCI_SMARTBUF_SZ        ATA_SECT_SIZE
55 #define AHCI_SMARTBUF_OFFSET    0xC00
56 /* 0x100 + 0x200 + 0x200 + 0x200 is smaller than 4k but we pad it out */
57 #define BLOCK_DMA_ALLOC_SZ      4096
58 
59 /* DMA region containing command table (should be 8192 bytes) */
60 #define AHCI_CMD_SLOT_SZ        sizeof(struct mtip_cmd_hdr)
61 #define AHCI_CMD_TBL_SZ         (MTIP_MAX_COMMAND_SLOTS * AHCI_CMD_SLOT_SZ)
62 #define AHCI_CMD_TBL_OFFSET     0x0
63 
64 /* DMA region per command (contains header and SGL) */
65 #define AHCI_CMD_TBL_HDR_SZ     0x80
66 #define AHCI_CMD_TBL_HDR_OFFSET 0x0
67 #define AHCI_CMD_TBL_SGL_SZ     (MTIP_MAX_SG * sizeof(struct mtip_cmd_sg))
68 #define AHCI_CMD_TBL_SGL_OFFSET AHCI_CMD_TBL_HDR_SZ
69 #define CMD_DMA_ALLOC_SZ        (AHCI_CMD_TBL_SGL_SZ + AHCI_CMD_TBL_HDR_SZ)
70 
71 
72 #define HOST_CAP_NZDMA		(1 << 19)
73 #define HOST_HSORG		0xFC
74 #define HSORG_DISABLE_SLOTGRP_INTR (1<<24)
75 #define HSORG_DISABLE_SLOTGRP_PXIS (1<<16)
76 #define HSORG_HWREV		0xFF00
77 #define HSORG_STYLE		0x8
78 #define HSORG_SLOTGROUPS	0x7
79 
80 #define PORT_COMMAND_ISSUE	0x38
81 #define PORT_SDBV		0x7C
82 
83 #define PORT_OFFSET		0x100
84 #define PORT_MEM_SIZE		0x80
85 
86 #define PORT_IRQ_ERR \
87 	(PORT_IRQ_HBUS_ERR | PORT_IRQ_IF_ERR | PORT_IRQ_CONNECT | \
88 	 PORT_IRQ_PHYRDY | PORT_IRQ_UNK_FIS | PORT_IRQ_BAD_PMP | \
89 	 PORT_IRQ_TF_ERR | PORT_IRQ_HBUS_DATA_ERR | PORT_IRQ_IF_NONFATAL | \
90 	 PORT_IRQ_OVERFLOW)
91 #define PORT_IRQ_LEGACY \
92 	(PORT_IRQ_PIOS_FIS | PORT_IRQ_D2H_REG_FIS)
93 #define PORT_IRQ_HANDLED \
94 	(PORT_IRQ_SDB_FIS | PORT_IRQ_LEGACY | \
95 	 PORT_IRQ_TF_ERR | PORT_IRQ_IF_ERR | \
96 	 PORT_IRQ_CONNECT | PORT_IRQ_PHYRDY)
97 #define DEF_PORT_IRQ \
98 	(PORT_IRQ_ERR | PORT_IRQ_LEGACY | PORT_IRQ_SDB_FIS)
99 
100 /* product numbers */
101 #define MTIP_PRODUCT_UNKNOWN	0x00
102 #define MTIP_PRODUCT_ASICFPGA	0x11
103 
104 /* Device instance number, incremented each time a device is probed. */
105 static int instance;
106 
107 static struct list_head online_list;
108 static struct list_head removing_list;
109 static spinlock_t dev_lock;
110 
111 /*
112  * Global variable used to hold the major block device number
113  * allocated in mtip_init().
114  */
115 static int mtip_major;
116 static struct dentry *dfs_parent;
117 static struct dentry *dfs_device_status;
118 
119 static u32 cpu_use[NR_CPUS];
120 
121 static DEFINE_SPINLOCK(rssd_index_lock);
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 			set_bit(QUEUE_FLAG_DEAD, &dd->queue->queue_flags);
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 /* we have to use runtime tag to setup command header */
173 static void mtip_init_cmd_header(struct request *rq)
174 {
175 	struct driver_data *dd = rq->q->queuedata;
176 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
177 	u32 host_cap_64 = readl(dd->mmio + HOST_CAP) & HOST_CAP_64;
178 
179 	/* Point the command headers at the command tables. */
180 	cmd->command_header = dd->port->command_list +
181 				(sizeof(struct mtip_cmd_hdr) * rq->tag);
182 	cmd->command_header_dma = dd->port->command_list_dma +
183 				(sizeof(struct mtip_cmd_hdr) * rq->tag);
184 
185 	if (host_cap_64)
186 		cmd->command_header->ctbau = __force_bit2int cpu_to_le32((cmd->command_dma >> 16) >> 16);
187 
188 	cmd->command_header->ctba = __force_bit2int cpu_to_le32(cmd->command_dma & 0xFFFFFFFF);
189 }
190 
191 static struct mtip_cmd *mtip_get_int_command(struct driver_data *dd)
192 {
193 	struct request *rq;
194 
195 	if (mtip_check_surprise_removal(dd->pdev))
196 		return NULL;
197 
198 	rq = blk_mq_alloc_request(dd->queue, REQ_OP_DRV_IN, BLK_MQ_REQ_RESERVED);
199 	if (IS_ERR(rq))
200 		return NULL;
201 
202 	/* Internal cmd isn't submitted via .queue_rq */
203 	mtip_init_cmd_header(rq);
204 
205 	return blk_mq_rq_to_pdu(rq);
206 }
207 
208 static struct mtip_cmd *mtip_cmd_from_tag(struct driver_data *dd,
209 					  unsigned int tag)
210 {
211 	struct blk_mq_hw_ctx *hctx = dd->queue->queue_hw_ctx[0];
212 
213 	return blk_mq_rq_to_pdu(blk_mq_tag_to_rq(hctx->tags, tag));
214 }
215 
216 /*
217  * Reset the HBA (without sleeping)
218  *
219  * @dd Pointer to the driver data structure.
220  *
221  * return value
222  *	0	The reset was successful.
223  *	-1	The HBA Reset bit did not clear.
224  */
225 static int mtip_hba_reset(struct driver_data *dd)
226 {
227 	unsigned long timeout;
228 
229 	/* Set the reset bit */
230 	writel(HOST_RESET, dd->mmio + HOST_CTL);
231 
232 	/* Flush */
233 	readl(dd->mmio + HOST_CTL);
234 
235 	/*
236 	 * Spin for up to 10 seconds waiting for reset acknowledgement. Spec
237 	 * is 1 sec but in LUN failure conditions, up to 10 secs are required
238 	 */
239 	timeout = jiffies + msecs_to_jiffies(10000);
240 	do {
241 		mdelay(10);
242 		if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag))
243 			return -1;
244 
245 	} while ((readl(dd->mmio + HOST_CTL) & HOST_RESET)
246 		 && time_before(jiffies, timeout));
247 
248 	if (readl(dd->mmio + HOST_CTL) & HOST_RESET)
249 		return -1;
250 
251 	return 0;
252 }
253 
254 /*
255  * Issue a command to the hardware.
256  *
257  * Set the appropriate bit in the s_active and Command Issue hardware
258  * registers, causing hardware command processing to begin.
259  *
260  * @port Pointer to the port structure.
261  * @tag  The tag of the command to be issued.
262  *
263  * return value
264  *      None
265  */
266 static inline void mtip_issue_ncq_command(struct mtip_port *port, int tag)
267 {
268 	int group = tag >> 5;
269 
270 	/* guard SACT and CI registers */
271 	spin_lock(&port->cmd_issue_lock[group]);
272 	writel((1 << MTIP_TAG_BIT(tag)),
273 			port->s_active[MTIP_TAG_INDEX(tag)]);
274 	writel((1 << MTIP_TAG_BIT(tag)),
275 			port->cmd_issue[MTIP_TAG_INDEX(tag)]);
276 	spin_unlock(&port->cmd_issue_lock[group]);
277 }
278 
279 /*
280  * Enable/disable the reception of FIS
281  *
282  * @port   Pointer to the port data structure
283  * @enable 1 to enable, 0 to disable
284  *
285  * return value
286  *	Previous state: 1 enabled, 0 disabled
287  */
288 static int mtip_enable_fis(struct mtip_port *port, int enable)
289 {
290 	u32 tmp;
291 
292 	/* enable FIS reception */
293 	tmp = readl(port->mmio + PORT_CMD);
294 	if (enable)
295 		writel(tmp | PORT_CMD_FIS_RX, port->mmio + PORT_CMD);
296 	else
297 		writel(tmp & ~PORT_CMD_FIS_RX, port->mmio + PORT_CMD);
298 
299 	/* Flush */
300 	readl(port->mmio + PORT_CMD);
301 
302 	return (((tmp & PORT_CMD_FIS_RX) == PORT_CMD_FIS_RX));
303 }
304 
305 /*
306  * Enable/disable the DMA engine
307  *
308  * @port   Pointer to the port data structure
309  * @enable 1 to enable, 0 to disable
310  *
311  * return value
312  *	Previous state: 1 enabled, 0 disabled.
313  */
314 static int mtip_enable_engine(struct mtip_port *port, int enable)
315 {
316 	u32 tmp;
317 
318 	/* enable FIS reception */
319 	tmp = readl(port->mmio + PORT_CMD);
320 	if (enable)
321 		writel(tmp | PORT_CMD_START, port->mmio + PORT_CMD);
322 	else
323 		writel(tmp & ~PORT_CMD_START, port->mmio + PORT_CMD);
324 
325 	readl(port->mmio + PORT_CMD);
326 	return (((tmp & PORT_CMD_START) == PORT_CMD_START));
327 }
328 
329 /*
330  * Enables the port DMA engine and FIS reception.
331  *
332  * return value
333  *	None
334  */
335 static inline void mtip_start_port(struct mtip_port *port)
336 {
337 	/* Enable FIS reception */
338 	mtip_enable_fis(port, 1);
339 
340 	/* Enable the DMA engine */
341 	mtip_enable_engine(port, 1);
342 }
343 
344 /*
345  * Deinitialize a port by disabling port interrupts, the DMA engine,
346  * and FIS reception.
347  *
348  * @port Pointer to the port structure
349  *
350  * return value
351  *	None
352  */
353 static inline void mtip_deinit_port(struct mtip_port *port)
354 {
355 	/* Disable interrupts on this port */
356 	writel(0, port->mmio + PORT_IRQ_MASK);
357 
358 	/* Disable the DMA engine */
359 	mtip_enable_engine(port, 0);
360 
361 	/* Disable FIS reception */
362 	mtip_enable_fis(port, 0);
363 }
364 
365 /*
366  * Initialize a port.
367  *
368  * This function deinitializes the port by calling mtip_deinit_port() and
369  * then initializes it by setting the command header and RX FIS addresses,
370  * clearing the SError register and any pending port interrupts before
371  * re-enabling the default set of port interrupts.
372  *
373  * @port Pointer to the port structure.
374  *
375  * return value
376  *	None
377  */
378 static void mtip_init_port(struct mtip_port *port)
379 {
380 	int i;
381 	mtip_deinit_port(port);
382 
383 	/* Program the command list base and FIS base addresses */
384 	if (readl(port->dd->mmio + HOST_CAP) & HOST_CAP_64) {
385 		writel((port->command_list_dma >> 16) >> 16,
386 			 port->mmio + PORT_LST_ADDR_HI);
387 		writel((port->rxfis_dma >> 16) >> 16,
388 			 port->mmio + PORT_FIS_ADDR_HI);
389 	}
390 
391 	writel(port->command_list_dma & 0xFFFFFFFF,
392 			port->mmio + PORT_LST_ADDR);
393 	writel(port->rxfis_dma & 0xFFFFFFFF, port->mmio + PORT_FIS_ADDR);
394 
395 	/* Clear SError */
396 	writel(readl(port->mmio + PORT_SCR_ERR), port->mmio + PORT_SCR_ERR);
397 
398 	/* reset the completed registers.*/
399 	for (i = 0; i < port->dd->slot_groups; i++)
400 		writel(0xFFFFFFFF, port->completed[i]);
401 
402 	/* Clear any pending interrupts for this port */
403 	writel(readl(port->mmio + PORT_IRQ_STAT), port->mmio + PORT_IRQ_STAT);
404 
405 	/* Clear any pending interrupts on the HBA. */
406 	writel(readl(port->dd->mmio + HOST_IRQ_STAT),
407 					port->dd->mmio + HOST_IRQ_STAT);
408 
409 	/* Enable port interrupts */
410 	writel(DEF_PORT_IRQ, port->mmio + PORT_IRQ_MASK);
411 }
412 
413 /*
414  * Restart a port
415  *
416  * @port Pointer to the port data structure.
417  *
418  * return value
419  *	None
420  */
421 static void mtip_restart_port(struct mtip_port *port)
422 {
423 	unsigned long timeout;
424 
425 	/* Disable the DMA engine */
426 	mtip_enable_engine(port, 0);
427 
428 	/* Chip quirk: wait up to 500ms for PxCMD.CR == 0 */
429 	timeout = jiffies + msecs_to_jiffies(500);
430 	while ((readl(port->mmio + PORT_CMD) & PORT_CMD_LIST_ON)
431 		 && time_before(jiffies, timeout))
432 		;
433 
434 	if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &port->dd->dd_flag))
435 		return;
436 
437 	/*
438 	 * Chip quirk: escalate to hba reset if
439 	 * PxCMD.CR not clear after 500 ms
440 	 */
441 	if (readl(port->mmio + PORT_CMD) & PORT_CMD_LIST_ON) {
442 		dev_warn(&port->dd->pdev->dev,
443 			"PxCMD.CR not clear, escalating reset\n");
444 
445 		if (mtip_hba_reset(port->dd))
446 			dev_err(&port->dd->pdev->dev,
447 				"HBA reset escalation failed.\n");
448 
449 		/* 30 ms delay before com reset to quiesce chip */
450 		mdelay(30);
451 	}
452 
453 	dev_warn(&port->dd->pdev->dev, "Issuing COM reset\n");
454 
455 	/* Set PxSCTL.DET */
456 	writel(readl(port->mmio + PORT_SCR_CTL) |
457 			 1, port->mmio + PORT_SCR_CTL);
458 	readl(port->mmio + PORT_SCR_CTL);
459 
460 	/* Wait 1 ms to quiesce chip function */
461 	timeout = jiffies + msecs_to_jiffies(1);
462 	while (time_before(jiffies, timeout))
463 		;
464 
465 	if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &port->dd->dd_flag))
466 		return;
467 
468 	/* Clear PxSCTL.DET */
469 	writel(readl(port->mmio + PORT_SCR_CTL) & ~1,
470 			 port->mmio + PORT_SCR_CTL);
471 	readl(port->mmio + PORT_SCR_CTL);
472 
473 	/* Wait 500 ms for bit 0 of PORT_SCR_STS to be set */
474 	timeout = jiffies + msecs_to_jiffies(500);
475 	while (((readl(port->mmio + PORT_SCR_STAT) & 0x01) == 0)
476 			 && time_before(jiffies, timeout))
477 		;
478 
479 	if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &port->dd->dd_flag))
480 		return;
481 
482 	if ((readl(port->mmio + PORT_SCR_STAT) & 0x01) == 0)
483 		dev_warn(&port->dd->pdev->dev,
484 			"COM reset failed\n");
485 
486 	mtip_init_port(port);
487 	mtip_start_port(port);
488 
489 }
490 
491 static int mtip_device_reset(struct driver_data *dd)
492 {
493 	int rv = 0;
494 
495 	if (mtip_check_surprise_removal(dd->pdev))
496 		return 0;
497 
498 	if (mtip_hba_reset(dd) < 0)
499 		rv = -EFAULT;
500 
501 	mdelay(1);
502 	mtip_init_port(dd->port);
503 	mtip_start_port(dd->port);
504 
505 	/* Enable interrupts on the HBA. */
506 	writel(readl(dd->mmio + HOST_CTL) | HOST_IRQ_EN,
507 					dd->mmio + HOST_CTL);
508 	return rv;
509 }
510 
511 /*
512  * Helper function for tag logging
513  */
514 static void print_tags(struct driver_data *dd,
515 			char *msg,
516 			unsigned long *tagbits,
517 			int cnt)
518 {
519 	unsigned char tagmap[128];
520 	int group, tagmap_len = 0;
521 
522 	memset(tagmap, 0, sizeof(tagmap));
523 	for (group = SLOTBITS_IN_LONGS; group > 0; group--)
524 		tagmap_len += sprintf(tagmap + tagmap_len, "%016lX ",
525 						tagbits[group-1]);
526 	dev_warn(&dd->pdev->dev,
527 			"%d command(s) %s: tagmap [%s]", cnt, msg, tagmap);
528 }
529 
530 static int mtip_read_log_page(struct mtip_port *port, u8 page, u16 *buffer,
531 				dma_addr_t buffer_dma, unsigned int sectors);
532 static int mtip_get_smart_attr(struct mtip_port *port, unsigned int id,
533 						struct smart_attr *attrib);
534 
535 static void mtip_complete_command(struct mtip_cmd *cmd, int status)
536 {
537 	struct request *req = blk_mq_rq_from_pdu(cmd);
538 
539 	cmd->status = status;
540 	blk_mq_complete_request(req);
541 }
542 
543 /*
544  * Handle an error.
545  *
546  * @dd Pointer to the DRIVER_DATA structure.
547  *
548  * return value
549  *	None
550  */
551 static void mtip_handle_tfe(struct driver_data *dd)
552 {
553 	int group, tag, bit, reissue, rv;
554 	struct mtip_port *port;
555 	struct mtip_cmd  *cmd;
556 	u32 completed;
557 	struct host_to_dev_fis *fis;
558 	unsigned long tagaccum[SLOTBITS_IN_LONGS];
559 	unsigned int cmd_cnt = 0;
560 	unsigned char *buf;
561 	char *fail_reason = NULL;
562 	int fail_all_ncq_write = 0, fail_all_ncq_cmds = 0;
563 
564 	dev_warn(&dd->pdev->dev, "Taskfile error\n");
565 
566 	port = dd->port;
567 
568 	if (test_bit(MTIP_PF_IC_ACTIVE_BIT, &port->flags)) {
569 		cmd = mtip_cmd_from_tag(dd, MTIP_TAG_INTERNAL);
570 		dbg_printk(MTIP_DRV_NAME " TFE for the internal command\n");
571 		mtip_complete_command(cmd, -EIO);
572 		return;
573 	}
574 
575 	/* clear the tag accumulator */
576 	memset(tagaccum, 0, SLOTBITS_IN_LONGS * sizeof(long));
577 
578 	/* Loop through all the groups */
579 	for (group = 0; group < dd->slot_groups; group++) {
580 		completed = readl(port->completed[group]);
581 
582 		dev_warn(&dd->pdev->dev, "g=%u, comp=%x\n", group, completed);
583 
584 		/* clear completed status register in the hardware.*/
585 		writel(completed, port->completed[group]);
586 
587 		/* Process successfully completed commands */
588 		for (bit = 0; bit < 32 && completed; bit++) {
589 			if (!(completed & (1<<bit)))
590 				continue;
591 			tag = (group << 5) + bit;
592 
593 			/* Skip the internal command slot */
594 			if (tag == MTIP_TAG_INTERNAL)
595 				continue;
596 
597 			cmd = mtip_cmd_from_tag(dd, tag);
598 			mtip_complete_command(cmd, 0);
599 			set_bit(tag, tagaccum);
600 			cmd_cnt++;
601 		}
602 	}
603 
604 	print_tags(dd, "completed (TFE)", tagaccum, cmd_cnt);
605 
606 	/* Restart the port */
607 	mdelay(20);
608 	mtip_restart_port(port);
609 
610 	/* Trying to determine the cause of the error */
611 	rv = mtip_read_log_page(dd->port, ATA_LOG_SATA_NCQ,
612 				dd->port->log_buf,
613 				dd->port->log_buf_dma, 1);
614 	if (rv) {
615 		dev_warn(&dd->pdev->dev,
616 			"Error in READ LOG EXT (10h) command\n");
617 		/* non-critical error, don't fail the load */
618 	} else {
619 		buf = (unsigned char *)dd->port->log_buf;
620 		if (buf[259] & 0x1) {
621 			dev_info(&dd->pdev->dev,
622 				"Write protect bit is set.\n");
623 			set_bit(MTIP_DDF_WRITE_PROTECT_BIT, &dd->dd_flag);
624 			fail_all_ncq_write = 1;
625 			fail_reason = "write protect";
626 		}
627 		if (buf[288] == 0xF7) {
628 			dev_info(&dd->pdev->dev,
629 				"Exceeded Tmax, drive in thermal shutdown.\n");
630 			set_bit(MTIP_DDF_OVER_TEMP_BIT, &dd->dd_flag);
631 			fail_all_ncq_cmds = 1;
632 			fail_reason = "thermal shutdown";
633 		}
634 		if (buf[288] == 0xBF) {
635 			set_bit(MTIP_DDF_REBUILD_FAILED_BIT, &dd->dd_flag);
636 			dev_info(&dd->pdev->dev,
637 				"Drive indicates rebuild has failed. Secure erase required.\n");
638 			fail_all_ncq_cmds = 1;
639 			fail_reason = "rebuild failed";
640 		}
641 	}
642 
643 	/* clear the tag accumulator */
644 	memset(tagaccum, 0, SLOTBITS_IN_LONGS * sizeof(long));
645 
646 	/* Loop through all the groups */
647 	for (group = 0; group < dd->slot_groups; group++) {
648 		for (bit = 0; bit < 32; bit++) {
649 			reissue = 1;
650 			tag = (group << 5) + bit;
651 			cmd = mtip_cmd_from_tag(dd, tag);
652 
653 			fis = (struct host_to_dev_fis *)cmd->command;
654 
655 			/* Should re-issue? */
656 			if (tag == MTIP_TAG_INTERNAL ||
657 			    fis->command == ATA_CMD_SET_FEATURES)
658 				reissue = 0;
659 			else {
660 				if (fail_all_ncq_cmds ||
661 					(fail_all_ncq_write &&
662 					fis->command == ATA_CMD_FPDMA_WRITE)) {
663 					dev_warn(&dd->pdev->dev,
664 					"  Fail: %s w/tag %d [%s].\n",
665 					fis->command == ATA_CMD_FPDMA_WRITE ?
666 						"write" : "read",
667 					tag,
668 					fail_reason != NULL ?
669 						fail_reason : "unknown");
670 					mtip_complete_command(cmd, -ENODATA);
671 					continue;
672 				}
673 			}
674 
675 			/*
676 			 * First check if this command has
677 			 *  exceeded its retries.
678 			 */
679 			if (reissue && (cmd->retries-- > 0)) {
680 
681 				set_bit(tag, tagaccum);
682 
683 				/* Re-issue the command. */
684 				mtip_issue_ncq_command(port, tag);
685 
686 				continue;
687 			}
688 
689 			/* Retire a command that will not be reissued */
690 			dev_warn(&port->dd->pdev->dev,
691 				"retiring tag %d\n", tag);
692 
693 			mtip_complete_command(cmd, -EIO);
694 		}
695 	}
696 	print_tags(dd, "reissued (TFE)", tagaccum, cmd_cnt);
697 }
698 
699 /*
700  * Handle a set device bits interrupt
701  */
702 static inline void mtip_workq_sdbfx(struct mtip_port *port, int group,
703 							u32 completed)
704 {
705 	struct driver_data *dd = port->dd;
706 	int tag, bit;
707 	struct mtip_cmd *command;
708 
709 	if (!completed) {
710 		WARN_ON_ONCE(!completed);
711 		return;
712 	}
713 	/* clear completed status register in the hardware.*/
714 	writel(completed, port->completed[group]);
715 
716 	/* Process completed commands. */
717 	for (bit = 0; (bit < 32) && completed; bit++) {
718 		if (completed & 0x01) {
719 			tag = (group << 5) | bit;
720 
721 			/* skip internal command slot. */
722 			if (unlikely(tag == MTIP_TAG_INTERNAL))
723 				continue;
724 
725 			command = mtip_cmd_from_tag(dd, tag);
726 			mtip_complete_command(command, 0);
727 		}
728 		completed >>= 1;
729 	}
730 
731 	/* If last, re-enable interrupts */
732 	if (atomic_dec_return(&dd->irq_workers_active) == 0)
733 		writel(0xffffffff, dd->mmio + HOST_IRQ_STAT);
734 }
735 
736 /*
737  * Process legacy pio and d2h interrupts
738  */
739 static inline void mtip_process_legacy(struct driver_data *dd, u32 port_stat)
740 {
741 	struct mtip_port *port = dd->port;
742 	struct mtip_cmd *cmd = mtip_cmd_from_tag(dd, MTIP_TAG_INTERNAL);
743 
744 	if (test_bit(MTIP_PF_IC_ACTIVE_BIT, &port->flags) && cmd) {
745 		int group = MTIP_TAG_INDEX(MTIP_TAG_INTERNAL);
746 		int status = readl(port->cmd_issue[group]);
747 
748 		if (!(status & (1 << MTIP_TAG_BIT(MTIP_TAG_INTERNAL))))
749 			mtip_complete_command(cmd, 0);
750 	}
751 }
752 
753 /*
754  * Demux and handle errors
755  */
756 static inline void mtip_process_errors(struct driver_data *dd, u32 port_stat)
757 {
758 	if (unlikely(port_stat & PORT_IRQ_CONNECT)) {
759 		dev_warn(&dd->pdev->dev,
760 			"Clearing PxSERR.DIAG.x\n");
761 		writel((1 << 26), dd->port->mmio + PORT_SCR_ERR);
762 	}
763 
764 	if (unlikely(port_stat & PORT_IRQ_PHYRDY)) {
765 		dev_warn(&dd->pdev->dev,
766 			"Clearing PxSERR.DIAG.n\n");
767 		writel((1 << 16), dd->port->mmio + PORT_SCR_ERR);
768 	}
769 
770 	if (unlikely(port_stat & ~PORT_IRQ_HANDLED)) {
771 		dev_warn(&dd->pdev->dev,
772 			"Port stat errors %x unhandled\n",
773 			(port_stat & ~PORT_IRQ_HANDLED));
774 		if (mtip_check_surprise_removal(dd->pdev))
775 			return;
776 	}
777 	if (likely(port_stat & (PORT_IRQ_TF_ERR | PORT_IRQ_IF_ERR))) {
778 		set_bit(MTIP_PF_EH_ACTIVE_BIT, &dd->port->flags);
779 		wake_up_interruptible(&dd->port->svc_wait);
780 	}
781 }
782 
783 static inline irqreturn_t mtip_handle_irq(struct driver_data *data)
784 {
785 	struct driver_data *dd = (struct driver_data *) data;
786 	struct mtip_port *port = dd->port;
787 	u32 hba_stat, port_stat;
788 	int rv = IRQ_NONE;
789 	int do_irq_enable = 1, i, workers;
790 	struct mtip_work *twork;
791 
792 	hba_stat = readl(dd->mmio + HOST_IRQ_STAT);
793 	if (hba_stat) {
794 		rv = IRQ_HANDLED;
795 
796 		/* Acknowledge the interrupt status on the port.*/
797 		port_stat = readl(port->mmio + PORT_IRQ_STAT);
798 		if (unlikely(port_stat == 0xFFFFFFFF)) {
799 			mtip_check_surprise_removal(dd->pdev);
800 			return IRQ_HANDLED;
801 		}
802 		writel(port_stat, port->mmio + PORT_IRQ_STAT);
803 
804 		/* Demux port status */
805 		if (likely(port_stat & PORT_IRQ_SDB_FIS)) {
806 			do_irq_enable = 0;
807 			WARN_ON_ONCE(atomic_read(&dd->irq_workers_active) != 0);
808 
809 			/* Start at 1: group zero is always local? */
810 			for (i = 0, workers = 0; i < MTIP_MAX_SLOT_GROUPS;
811 									i++) {
812 				twork = &dd->work[i];
813 				twork->completed = readl(port->completed[i]);
814 				if (twork->completed)
815 					workers++;
816 			}
817 
818 			atomic_set(&dd->irq_workers_active, workers);
819 			if (workers) {
820 				for (i = 1; i < MTIP_MAX_SLOT_GROUPS; i++) {
821 					twork = &dd->work[i];
822 					if (twork->completed)
823 						queue_work_on(
824 							twork->cpu_binding,
825 							dd->isr_workq,
826 							&twork->work);
827 				}
828 
829 				if (likely(dd->work[0].completed))
830 					mtip_workq_sdbfx(port, 0,
831 							dd->work[0].completed);
832 
833 			} else {
834 				/*
835 				 * Chip quirk: SDB interrupt but nothing
836 				 * to complete
837 				 */
838 				do_irq_enable = 1;
839 			}
840 		}
841 
842 		if (unlikely(port_stat & PORT_IRQ_ERR)) {
843 			if (unlikely(mtip_check_surprise_removal(dd->pdev))) {
844 				/* don't proceed further */
845 				return IRQ_HANDLED;
846 			}
847 			if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
848 							&dd->dd_flag))
849 				return rv;
850 
851 			mtip_process_errors(dd, port_stat & PORT_IRQ_ERR);
852 		}
853 
854 		if (unlikely(port_stat & PORT_IRQ_LEGACY))
855 			mtip_process_legacy(dd, port_stat & PORT_IRQ_LEGACY);
856 	}
857 
858 	/* acknowledge interrupt */
859 	if (unlikely(do_irq_enable))
860 		writel(hba_stat, dd->mmio + HOST_IRQ_STAT);
861 
862 	return rv;
863 }
864 
865 /*
866  * HBA interrupt subroutine.
867  *
868  * @irq		IRQ number.
869  * @instance	Pointer to the driver data structure.
870  *
871  * return value
872  *	IRQ_HANDLED	A HBA interrupt was pending and handled.
873  *	IRQ_NONE	This interrupt was not for the HBA.
874  */
875 static irqreturn_t mtip_irq_handler(int irq, void *instance)
876 {
877 	struct driver_data *dd = instance;
878 
879 	return mtip_handle_irq(dd);
880 }
881 
882 static void mtip_issue_non_ncq_command(struct mtip_port *port, int tag)
883 {
884 	writel(1 << MTIP_TAG_BIT(tag), port->cmd_issue[MTIP_TAG_INDEX(tag)]);
885 }
886 
887 static bool mtip_pause_ncq(struct mtip_port *port,
888 				struct host_to_dev_fis *fis)
889 {
890 	struct host_to_dev_fis *reply;
891 	unsigned long task_file_data;
892 
893 	reply = port->rxfis + RX_FIS_D2H_REG;
894 	task_file_data = readl(port->mmio+PORT_TFDATA);
895 
896 	if ((task_file_data & 1))
897 		return false;
898 
899 	if (fis->command == ATA_CMD_SEC_ERASE_PREP) {
900 		port->ic_pause_timer = jiffies;
901 		return true;
902 	} else if ((fis->command == ATA_CMD_DOWNLOAD_MICRO) &&
903 					(fis->features == 0x03)) {
904 		set_bit(MTIP_PF_DM_ACTIVE_BIT, &port->flags);
905 		port->ic_pause_timer = jiffies;
906 		return true;
907 	} else if ((fis->command == ATA_CMD_SEC_ERASE_UNIT) ||
908 		((fis->command == 0xFC) &&
909 			(fis->features == 0x27 || fis->features == 0x72 ||
910 			 fis->features == 0x62 || fis->features == 0x26))) {
911 		clear_bit(MTIP_DDF_SEC_LOCK_BIT, &port->dd->dd_flag);
912 		clear_bit(MTIP_DDF_REBUILD_FAILED_BIT, &port->dd->dd_flag);
913 		/* Com reset after secure erase or lowlevel format */
914 		mtip_restart_port(port);
915 		clear_bit(MTIP_PF_SE_ACTIVE_BIT, &port->flags);
916 		return false;
917 	}
918 
919 	return false;
920 }
921 
922 static bool mtip_commands_active(struct mtip_port *port)
923 {
924 	unsigned int active;
925 	unsigned int n;
926 
927 	/*
928 	 * Ignore s_active bit 0 of array element 0.
929 	 * This bit will always be set
930 	 */
931 	active = readl(port->s_active[0]) & 0xFFFFFFFE;
932 	for (n = 1; n < port->dd->slot_groups; n++)
933 		active |= readl(port->s_active[n]);
934 
935 	return active != 0;
936 }
937 
938 /*
939  * Wait for port to quiesce
940  *
941  * @port    Pointer to port data structure
942  * @timeout Max duration to wait (ms)
943  *
944  * return value
945  *	0	Success
946  *	-EBUSY  Commands still active
947  */
948 static int mtip_quiesce_io(struct mtip_port *port, unsigned long timeout)
949 {
950 	unsigned long to;
951 	bool active = true;
952 
953 	blk_mq_stop_hw_queues(port->dd->queue);
954 
955 	to = jiffies + msecs_to_jiffies(timeout);
956 	do {
957 		if (test_bit(MTIP_PF_SVC_THD_ACTIVE_BIT, &port->flags) &&
958 			test_bit(MTIP_PF_ISSUE_CMDS_BIT, &port->flags)) {
959 			msleep(20);
960 			continue; /* svc thd is actively issuing commands */
961 		}
962 
963 		msleep(100);
964 
965 		if (mtip_check_surprise_removal(port->dd->pdev))
966 			goto err_fault;
967 
968 		active = mtip_commands_active(port);
969 		if (!active)
970 			break;
971 	} while (time_before(jiffies, to));
972 
973 	blk_mq_start_stopped_hw_queues(port->dd->queue, true);
974 	return active ? -EBUSY : 0;
975 err_fault:
976 	blk_mq_start_stopped_hw_queues(port->dd->queue, true);
977 	return -EFAULT;
978 }
979 
980 struct mtip_int_cmd {
981 	int fis_len;
982 	dma_addr_t buffer;
983 	int buf_len;
984 	u32 opts;
985 };
986 
987 /*
988  * Execute an internal command and wait for the completion.
989  *
990  * @port    Pointer to the port data structure.
991  * @fis     Pointer to the FIS that describes the command.
992  * @fis_len  Length in WORDS of the FIS.
993  * @buffer  DMA accessible for command data.
994  * @buf_len  Length, in bytes, of the data buffer.
995  * @opts    Command header options, excluding the FIS length
996  *             and the number of PRD entries.
997  * @timeout Time in ms to wait for the command to complete.
998  *
999  * return value
1000  *	0	 Command completed successfully.
1001  *	-EFAULT  The buffer address is not correctly aligned.
1002  *	-EBUSY   Internal command or other IO in progress.
1003  *	-EAGAIN  Time out waiting for command to complete.
1004  */
1005 static int mtip_exec_internal_command(struct mtip_port *port,
1006 					struct host_to_dev_fis *fis,
1007 					int fis_len,
1008 					dma_addr_t buffer,
1009 					int buf_len,
1010 					u32 opts,
1011 					unsigned long timeout)
1012 {
1013 	struct mtip_cmd *int_cmd;
1014 	struct driver_data *dd = port->dd;
1015 	struct request *rq;
1016 	struct mtip_int_cmd icmd = {
1017 		.fis_len = fis_len,
1018 		.buffer = buffer,
1019 		.buf_len = buf_len,
1020 		.opts = opts
1021 	};
1022 	int rv = 0;
1023 	unsigned long start;
1024 
1025 	/* Make sure the buffer is 8 byte aligned. This is asic specific. */
1026 	if (buffer & 0x00000007) {
1027 		dev_err(&dd->pdev->dev, "SG buffer is not 8 byte aligned\n");
1028 		return -EFAULT;
1029 	}
1030 
1031 	int_cmd = mtip_get_int_command(dd);
1032 	if (!int_cmd) {
1033 		dbg_printk(MTIP_DRV_NAME "Unable to allocate tag for PIO cmd\n");
1034 		return -EFAULT;
1035 	}
1036 	rq = blk_mq_rq_from_pdu(int_cmd);
1037 	rq->special = &icmd;
1038 
1039 	set_bit(MTIP_PF_IC_ACTIVE_BIT, &port->flags);
1040 
1041 	if (fis->command == ATA_CMD_SEC_ERASE_PREP)
1042 		set_bit(MTIP_PF_SE_ACTIVE_BIT, &port->flags);
1043 
1044 	clear_bit(MTIP_PF_DM_ACTIVE_BIT, &port->flags);
1045 
1046 	if (fis->command != ATA_CMD_STANDBYNOW1) {
1047 		/* wait for io to complete if non atomic */
1048 		if (mtip_quiesce_io(port, MTIP_QUIESCE_IO_TIMEOUT_MS) < 0) {
1049 			dev_warn(&dd->pdev->dev, "Failed to quiesce IO\n");
1050 			blk_mq_free_request(rq);
1051 			clear_bit(MTIP_PF_IC_ACTIVE_BIT, &port->flags);
1052 			wake_up_interruptible(&port->svc_wait);
1053 			return -EBUSY;
1054 		}
1055 	}
1056 
1057 	/* Copy the command to the command table */
1058 	memcpy(int_cmd->command, fis, fis_len*4);
1059 
1060 	start = jiffies;
1061 	rq->timeout = timeout;
1062 
1063 	/* insert request and run queue */
1064 	blk_execute_rq(rq->q, NULL, rq, true);
1065 
1066 	rv = int_cmd->status;
1067 	if (rv < 0) {
1068 		if (rv == -ERESTARTSYS) { /* interrupted */
1069 			dev_err(&dd->pdev->dev,
1070 				"Internal command [%02X] was interrupted after %u ms\n",
1071 				fis->command,
1072 				jiffies_to_msecs(jiffies - start));
1073 			rv = -EINTR;
1074 			goto exec_ic_exit;
1075 		} else if (rv == 0) /* timeout */
1076 			dev_err(&dd->pdev->dev,
1077 				"Internal command did not complete [%02X] within timeout of  %lu ms\n",
1078 				fis->command, timeout);
1079 		else
1080 			dev_err(&dd->pdev->dev,
1081 				"Internal command [%02X] wait returned code [%d] after %lu ms - unhandled\n",
1082 				fis->command, rv, timeout);
1083 
1084 		if (mtip_check_surprise_removal(dd->pdev) ||
1085 			test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
1086 					&dd->dd_flag)) {
1087 			dev_err(&dd->pdev->dev,
1088 				"Internal command [%02X] wait returned due to SR\n",
1089 				fis->command);
1090 			rv = -ENXIO;
1091 			goto exec_ic_exit;
1092 		}
1093 		mtip_device_reset(dd); /* recover from timeout issue */
1094 		rv = -EAGAIN;
1095 		goto exec_ic_exit;
1096 	}
1097 
1098 	if (readl(port->cmd_issue[MTIP_TAG_INDEX(MTIP_TAG_INTERNAL)])
1099 			& (1 << MTIP_TAG_BIT(MTIP_TAG_INTERNAL))) {
1100 		rv = -ENXIO;
1101 		if (!test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)) {
1102 			mtip_device_reset(dd);
1103 			rv = -EAGAIN;
1104 		}
1105 	}
1106 exec_ic_exit:
1107 	/* Clear the allocated and active bits for the internal command. */
1108 	blk_mq_free_request(rq);
1109 	clear_bit(MTIP_PF_IC_ACTIVE_BIT, &port->flags);
1110 	if (rv >= 0 && mtip_pause_ncq(port, fis)) {
1111 		/* NCQ paused */
1112 		return rv;
1113 	}
1114 	wake_up_interruptible(&port->svc_wait);
1115 
1116 	return rv;
1117 }
1118 
1119 /*
1120  * Byte-swap ATA ID strings.
1121  *
1122  * ATA identify data contains strings in byte-swapped 16-bit words.
1123  * They must be swapped (on all architectures) to be usable as C strings.
1124  * This function swaps bytes in-place.
1125  *
1126  * @buf The buffer location of the string
1127  * @len The number of bytes to swap
1128  *
1129  * return value
1130  *	None
1131  */
1132 static inline void ata_swap_string(u16 *buf, unsigned int len)
1133 {
1134 	int i;
1135 	for (i = 0; i < (len/2); i++)
1136 		be16_to_cpus(&buf[i]);
1137 }
1138 
1139 static void mtip_set_timeout(struct driver_data *dd,
1140 					struct host_to_dev_fis *fis,
1141 					unsigned int *timeout, u8 erasemode)
1142 {
1143 	switch (fis->command) {
1144 	case ATA_CMD_DOWNLOAD_MICRO:
1145 		*timeout = 120000; /* 2 minutes */
1146 		break;
1147 	case ATA_CMD_SEC_ERASE_UNIT:
1148 	case 0xFC:
1149 		if (erasemode)
1150 			*timeout = ((*(dd->port->identify + 90) * 2) * 60000);
1151 		else
1152 			*timeout = ((*(dd->port->identify + 89) * 2) * 60000);
1153 		break;
1154 	case ATA_CMD_STANDBYNOW1:
1155 		*timeout = 120000;  /* 2 minutes */
1156 		break;
1157 	case 0xF7:
1158 	case 0xFA:
1159 		*timeout = 60000;  /* 60 seconds */
1160 		break;
1161 	case ATA_CMD_SMART:
1162 		*timeout = 15000;  /* 15 seconds */
1163 		break;
1164 	default:
1165 		*timeout = MTIP_IOCTL_CMD_TIMEOUT_MS;
1166 		break;
1167 	}
1168 }
1169 
1170 /*
1171  * Request the device identity information.
1172  *
1173  * If a user space buffer is not specified, i.e. is NULL, the
1174  * identify information is still read from the drive and placed
1175  * into the identify data buffer (@e port->identify) in the
1176  * port data structure.
1177  * When the identify buffer contains valid identify information @e
1178  * port->identify_valid is non-zero.
1179  *
1180  * @port	 Pointer to the port structure.
1181  * @user_buffer  A user space buffer where the identify data should be
1182  *                    copied.
1183  *
1184  * return value
1185  *	0	Command completed successfully.
1186  *	-EFAULT An error occurred while coping data to the user buffer.
1187  *	-1	Command failed.
1188  */
1189 static int mtip_get_identify(struct mtip_port *port, void __user *user_buffer)
1190 {
1191 	int rv = 0;
1192 	struct host_to_dev_fis fis;
1193 
1194 	if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &port->dd->dd_flag))
1195 		return -EFAULT;
1196 
1197 	/* Build the FIS. */
1198 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1199 	fis.type	= 0x27;
1200 	fis.opts	= 1 << 7;
1201 	fis.command	= ATA_CMD_ID_ATA;
1202 
1203 	/* Set the identify information as invalid. */
1204 	port->identify_valid = 0;
1205 
1206 	/* Clear the identify information. */
1207 	memset(port->identify, 0, sizeof(u16) * ATA_ID_WORDS);
1208 
1209 	/* Execute the command. */
1210 	if (mtip_exec_internal_command(port,
1211 				&fis,
1212 				5,
1213 				port->identify_dma,
1214 				sizeof(u16) * ATA_ID_WORDS,
1215 				0,
1216 				MTIP_INT_CMD_TIMEOUT_MS)
1217 				< 0) {
1218 		rv = -1;
1219 		goto out;
1220 	}
1221 
1222 	/*
1223 	 * Perform any necessary byte-swapping.  Yes, the kernel does in fact
1224 	 * perform field-sensitive swapping on the string fields.
1225 	 * See the kernel use of ata_id_string() for proof of this.
1226 	 */
1227 #ifdef __LITTLE_ENDIAN
1228 	ata_swap_string(port->identify + 27, 40);  /* model string*/
1229 	ata_swap_string(port->identify + 23, 8);   /* firmware string*/
1230 	ata_swap_string(port->identify + 10, 20);  /* serial# string*/
1231 #else
1232 	{
1233 		int i;
1234 		for (i = 0; i < ATA_ID_WORDS; i++)
1235 			port->identify[i] = le16_to_cpu(port->identify[i]);
1236 	}
1237 #endif
1238 
1239 	/* Check security locked state */
1240 	if (port->identify[128] & 0x4)
1241 		set_bit(MTIP_DDF_SEC_LOCK_BIT, &port->dd->dd_flag);
1242 	else
1243 		clear_bit(MTIP_DDF_SEC_LOCK_BIT, &port->dd->dd_flag);
1244 
1245 #ifdef MTIP_TRIM /* Disabling TRIM support temporarily */
1246 	/* Demux ID.DRAT & ID.RZAT to determine trim support */
1247 	if (port->identify[69] & (1 << 14) && port->identify[69] & (1 << 5))
1248 		port->dd->trim_supp = true;
1249 	else
1250 #endif
1251 		port->dd->trim_supp = false;
1252 
1253 	/* Set the identify buffer as valid. */
1254 	port->identify_valid = 1;
1255 
1256 	if (user_buffer) {
1257 		if (copy_to_user(
1258 			user_buffer,
1259 			port->identify,
1260 			ATA_ID_WORDS * sizeof(u16))) {
1261 			rv = -EFAULT;
1262 			goto out;
1263 		}
1264 	}
1265 
1266 out:
1267 	return rv;
1268 }
1269 
1270 /*
1271  * Issue a standby immediate command to the device.
1272  *
1273  * @port Pointer to the port structure.
1274  *
1275  * return value
1276  *	0	Command was executed successfully.
1277  *	-1	An error occurred while executing the command.
1278  */
1279 static int mtip_standby_immediate(struct mtip_port *port)
1280 {
1281 	int rv;
1282 	struct host_to_dev_fis	fis;
1283 	unsigned long start;
1284 	unsigned int timeout;
1285 
1286 	/* Build the FIS. */
1287 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1288 	fis.type	= 0x27;
1289 	fis.opts	= 1 << 7;
1290 	fis.command	= ATA_CMD_STANDBYNOW1;
1291 
1292 	mtip_set_timeout(port->dd, &fis, &timeout, 0);
1293 
1294 	start = jiffies;
1295 	rv = mtip_exec_internal_command(port,
1296 					&fis,
1297 					5,
1298 					0,
1299 					0,
1300 					0,
1301 					timeout);
1302 	dbg_printk(MTIP_DRV_NAME "Time taken to complete standby cmd: %d ms\n",
1303 			jiffies_to_msecs(jiffies - start));
1304 	if (rv)
1305 		dev_warn(&port->dd->pdev->dev,
1306 			"STANDBY IMMEDIATE command failed.\n");
1307 
1308 	return rv;
1309 }
1310 
1311 /*
1312  * Issue a READ LOG EXT command to the device.
1313  *
1314  * @port	pointer to the port structure.
1315  * @page	page number to fetch
1316  * @buffer	pointer to buffer
1317  * @buffer_dma	dma address corresponding to @buffer
1318  * @sectors	page length to fetch, in sectors
1319  *
1320  * return value
1321  *	@rv	return value from mtip_exec_internal_command()
1322  */
1323 static int mtip_read_log_page(struct mtip_port *port, u8 page, u16 *buffer,
1324 				dma_addr_t buffer_dma, unsigned int sectors)
1325 {
1326 	struct host_to_dev_fis fis;
1327 
1328 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1329 	fis.type	= 0x27;
1330 	fis.opts	= 1 << 7;
1331 	fis.command	= ATA_CMD_READ_LOG_EXT;
1332 	fis.sect_count	= sectors & 0xFF;
1333 	fis.sect_cnt_ex	= (sectors >> 8) & 0xFF;
1334 	fis.lba_low	= page;
1335 	fis.lba_mid	= 0;
1336 	fis.device	= ATA_DEVICE_OBS;
1337 
1338 	memset(buffer, 0, sectors * ATA_SECT_SIZE);
1339 
1340 	return mtip_exec_internal_command(port,
1341 					&fis,
1342 					5,
1343 					buffer_dma,
1344 					sectors * ATA_SECT_SIZE,
1345 					0,
1346 					MTIP_INT_CMD_TIMEOUT_MS);
1347 }
1348 
1349 /*
1350  * Issue a SMART READ DATA command to the device.
1351  *
1352  * @port	pointer to the port structure.
1353  * @buffer	pointer to buffer
1354  * @buffer_dma	dma address corresponding to @buffer
1355  *
1356  * return value
1357  *	@rv	return value from mtip_exec_internal_command()
1358  */
1359 static int mtip_get_smart_data(struct mtip_port *port, u8 *buffer,
1360 					dma_addr_t buffer_dma)
1361 {
1362 	struct host_to_dev_fis fis;
1363 
1364 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1365 	fis.type	= 0x27;
1366 	fis.opts	= 1 << 7;
1367 	fis.command	= ATA_CMD_SMART;
1368 	fis.features	= 0xD0;
1369 	fis.sect_count	= 1;
1370 	fis.lba_mid	= 0x4F;
1371 	fis.lba_hi	= 0xC2;
1372 	fis.device	= ATA_DEVICE_OBS;
1373 
1374 	return mtip_exec_internal_command(port,
1375 					&fis,
1376 					5,
1377 					buffer_dma,
1378 					ATA_SECT_SIZE,
1379 					0,
1380 					15000);
1381 }
1382 
1383 /*
1384  * Get the value of a smart attribute
1385  *
1386  * @port	pointer to the port structure
1387  * @id		attribute number
1388  * @attrib	pointer to return attrib information corresponding to @id
1389  *
1390  * return value
1391  *	-EINVAL	NULL buffer passed or unsupported attribute @id.
1392  *	-EPERM	Identify data not valid, SMART not supported or not enabled
1393  */
1394 static int mtip_get_smart_attr(struct mtip_port *port, unsigned int id,
1395 						struct smart_attr *attrib)
1396 {
1397 	int rv, i;
1398 	struct smart_attr *pattr;
1399 
1400 	if (!attrib)
1401 		return -EINVAL;
1402 
1403 	if (!port->identify_valid) {
1404 		dev_warn(&port->dd->pdev->dev, "IDENTIFY DATA not valid\n");
1405 		return -EPERM;
1406 	}
1407 	if (!(port->identify[82] & 0x1)) {
1408 		dev_warn(&port->dd->pdev->dev, "SMART not supported\n");
1409 		return -EPERM;
1410 	}
1411 	if (!(port->identify[85] & 0x1)) {
1412 		dev_warn(&port->dd->pdev->dev, "SMART not enabled\n");
1413 		return -EPERM;
1414 	}
1415 
1416 	memset(port->smart_buf, 0, ATA_SECT_SIZE);
1417 	rv = mtip_get_smart_data(port, port->smart_buf, port->smart_buf_dma);
1418 	if (rv) {
1419 		dev_warn(&port->dd->pdev->dev, "Failed to ge SMART data\n");
1420 		return rv;
1421 	}
1422 
1423 	pattr = (struct smart_attr *)(port->smart_buf + 2);
1424 	for (i = 0; i < 29; i++, pattr++)
1425 		if (pattr->attr_id == id) {
1426 			memcpy(attrib, pattr, sizeof(struct smart_attr));
1427 			break;
1428 		}
1429 
1430 	if (i == 29) {
1431 		dev_warn(&port->dd->pdev->dev,
1432 			"Query for invalid SMART attribute ID\n");
1433 		rv = -EINVAL;
1434 	}
1435 
1436 	return rv;
1437 }
1438 
1439 /*
1440  * Trim unused sectors
1441  *
1442  * @dd		pointer to driver_data structure
1443  * @lba		starting lba
1444  * @len		# of 512b sectors to trim
1445  *
1446  * return value
1447  *      -ENOMEM		Out of dma memory
1448  *      -EINVAL		Invalid parameters passed in, trim not supported
1449  *      -EIO		Error submitting trim request to hw
1450  */
1451 static int mtip_send_trim(struct driver_data *dd, unsigned int lba,
1452 				unsigned int len)
1453 {
1454 	int i, rv = 0;
1455 	u64 tlba, tlen, sect_left;
1456 	struct mtip_trim_entry *buf;
1457 	dma_addr_t dma_addr;
1458 	struct host_to_dev_fis fis;
1459 
1460 	if (!len || dd->trim_supp == false)
1461 		return -EINVAL;
1462 
1463 	/* Trim request too big */
1464 	WARN_ON(len > (MTIP_MAX_TRIM_ENTRY_LEN * MTIP_MAX_TRIM_ENTRIES));
1465 
1466 	/* Trim request not aligned on 4k boundary */
1467 	WARN_ON(len % 8 != 0);
1468 
1469 	/* Warn if vu_trim structure is too big */
1470 	WARN_ON(sizeof(struct mtip_trim) > ATA_SECT_SIZE);
1471 
1472 	/* Allocate a DMA buffer for the trim structure */
1473 	buf = dmam_alloc_coherent(&dd->pdev->dev, ATA_SECT_SIZE, &dma_addr,
1474 								GFP_KERNEL);
1475 	if (!buf)
1476 		return -ENOMEM;
1477 	memset(buf, 0, ATA_SECT_SIZE);
1478 
1479 	for (i = 0, sect_left = len, tlba = lba;
1480 			i < MTIP_MAX_TRIM_ENTRIES && sect_left;
1481 			i++) {
1482 		tlen = (sect_left >= MTIP_MAX_TRIM_ENTRY_LEN ?
1483 					MTIP_MAX_TRIM_ENTRY_LEN :
1484 					sect_left);
1485 		buf[i].lba = __force_bit2int cpu_to_le32(tlba);
1486 		buf[i].range = __force_bit2int cpu_to_le16(tlen);
1487 		tlba += tlen;
1488 		sect_left -= tlen;
1489 	}
1490 	WARN_ON(sect_left != 0);
1491 
1492 	/* Build the fis */
1493 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1494 	fis.type       = 0x27;
1495 	fis.opts       = 1 << 7;
1496 	fis.command    = 0xfb;
1497 	fis.features   = 0x60;
1498 	fis.sect_count = 1;
1499 	fis.device     = ATA_DEVICE_OBS;
1500 
1501 	if (mtip_exec_internal_command(dd->port,
1502 					&fis,
1503 					5,
1504 					dma_addr,
1505 					ATA_SECT_SIZE,
1506 					0,
1507 					MTIP_TRIM_TIMEOUT_MS) < 0)
1508 		rv = -EIO;
1509 
1510 	dmam_free_coherent(&dd->pdev->dev, ATA_SECT_SIZE, buf, dma_addr);
1511 	return rv;
1512 }
1513 
1514 /*
1515  * Get the drive capacity.
1516  *
1517  * @dd      Pointer to the device data structure.
1518  * @sectors Pointer to the variable that will receive the sector count.
1519  *
1520  * return value
1521  *	1 Capacity was returned successfully.
1522  *	0 The identify information is invalid.
1523  */
1524 static bool mtip_hw_get_capacity(struct driver_data *dd, sector_t *sectors)
1525 {
1526 	struct mtip_port *port = dd->port;
1527 	u64 total, raw0, raw1, raw2, raw3;
1528 	raw0 = port->identify[100];
1529 	raw1 = port->identify[101];
1530 	raw2 = port->identify[102];
1531 	raw3 = port->identify[103];
1532 	total = raw0 | raw1<<16 | raw2<<32 | raw3<<48;
1533 	*sectors = total;
1534 	return (bool) !!port->identify_valid;
1535 }
1536 
1537 /*
1538  * Display the identify command data.
1539  *
1540  * @port Pointer to the port data structure.
1541  *
1542  * return value
1543  *	None
1544  */
1545 static void mtip_dump_identify(struct mtip_port *port)
1546 {
1547 	sector_t sectors;
1548 	unsigned short revid;
1549 	char cbuf[42];
1550 
1551 	if (!port->identify_valid)
1552 		return;
1553 
1554 	strlcpy(cbuf, (char *)(port->identify+10), 21);
1555 	dev_info(&port->dd->pdev->dev,
1556 		"Serial No.: %s\n", cbuf);
1557 
1558 	strlcpy(cbuf, (char *)(port->identify+23), 9);
1559 	dev_info(&port->dd->pdev->dev,
1560 		"Firmware Ver.: %s\n", cbuf);
1561 
1562 	strlcpy(cbuf, (char *)(port->identify+27), 41);
1563 	dev_info(&port->dd->pdev->dev, "Model: %s\n", cbuf);
1564 
1565 	dev_info(&port->dd->pdev->dev, "Security: %04x %s\n",
1566 		port->identify[128],
1567 		port->identify[128] & 0x4 ? "(LOCKED)" : "");
1568 
1569 	if (mtip_hw_get_capacity(port->dd, &sectors))
1570 		dev_info(&port->dd->pdev->dev,
1571 			"Capacity: %llu sectors (%llu MB)\n",
1572 			 (u64)sectors,
1573 			 ((u64)sectors) * ATA_SECT_SIZE >> 20);
1574 
1575 	pci_read_config_word(port->dd->pdev, PCI_REVISION_ID, &revid);
1576 	switch (revid & 0xFF) {
1577 	case 0x1:
1578 		strlcpy(cbuf, "A0", 3);
1579 		break;
1580 	case 0x3:
1581 		strlcpy(cbuf, "A2", 3);
1582 		break;
1583 	default:
1584 		strlcpy(cbuf, "?", 2);
1585 		break;
1586 	}
1587 	dev_info(&port->dd->pdev->dev,
1588 		"Card Type: %s\n", cbuf);
1589 }
1590 
1591 /*
1592  * Map the commands scatter list into the command table.
1593  *
1594  * @command Pointer to the command.
1595  * @nents Number of scatter list entries.
1596  *
1597  * return value
1598  *	None
1599  */
1600 static inline void fill_command_sg(struct driver_data *dd,
1601 				struct mtip_cmd *command,
1602 				int nents)
1603 {
1604 	int n;
1605 	unsigned int dma_len;
1606 	struct mtip_cmd_sg *command_sg;
1607 	struct scatterlist *sg = command->sg;
1608 
1609 	command_sg = command->command + AHCI_CMD_TBL_HDR_SZ;
1610 
1611 	for (n = 0; n < nents; n++) {
1612 		dma_len = sg_dma_len(sg);
1613 		if (dma_len > 0x400000)
1614 			dev_err(&dd->pdev->dev,
1615 				"DMA segment length truncated\n");
1616 		command_sg->info = __force_bit2int
1617 			cpu_to_le32((dma_len-1) & 0x3FFFFF);
1618 		command_sg->dba	= __force_bit2int
1619 			cpu_to_le32(sg_dma_address(sg));
1620 		command_sg->dba_upper = __force_bit2int
1621 			cpu_to_le32((sg_dma_address(sg) >> 16) >> 16);
1622 		command_sg++;
1623 		sg++;
1624 	}
1625 }
1626 
1627 /*
1628  * @brief Execute a drive command.
1629  *
1630  * return value 0 The command completed successfully.
1631  * return value -1 An error occurred while executing the command.
1632  */
1633 static int exec_drive_task(struct mtip_port *port, u8 *command)
1634 {
1635 	struct host_to_dev_fis	fis;
1636 	struct host_to_dev_fis *reply = (port->rxfis + RX_FIS_D2H_REG);
1637 	unsigned int to;
1638 
1639 	/* Build the FIS. */
1640 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1641 	fis.type	= 0x27;
1642 	fis.opts	= 1 << 7;
1643 	fis.command	= command[0];
1644 	fis.features	= command[1];
1645 	fis.sect_count	= command[2];
1646 	fis.sector	= command[3];
1647 	fis.cyl_low	= command[4];
1648 	fis.cyl_hi	= command[5];
1649 	fis.device	= command[6] & ~0x10; /* Clear the dev bit*/
1650 
1651 	mtip_set_timeout(port->dd, &fis, &to, 0);
1652 
1653 	dbg_printk(MTIP_DRV_NAME " %s: User Command: cmd %x, feat %x, nsect %x, sect %x, lcyl %x, hcyl %x, sel %x\n",
1654 		__func__,
1655 		command[0],
1656 		command[1],
1657 		command[2],
1658 		command[3],
1659 		command[4],
1660 		command[5],
1661 		command[6]);
1662 
1663 	/* Execute the command. */
1664 	if (mtip_exec_internal_command(port,
1665 				 &fis,
1666 				 5,
1667 				 0,
1668 				 0,
1669 				 0,
1670 				 to) < 0) {
1671 		return -1;
1672 	}
1673 
1674 	command[0] = reply->command; /* Status*/
1675 	command[1] = reply->features; /* Error*/
1676 	command[4] = reply->cyl_low;
1677 	command[5] = reply->cyl_hi;
1678 
1679 	dbg_printk(MTIP_DRV_NAME " %s: Completion Status: stat %x, err %x , cyl_lo %x cyl_hi %x\n",
1680 		__func__,
1681 		command[0],
1682 		command[1],
1683 		command[4],
1684 		command[5]);
1685 
1686 	return 0;
1687 }
1688 
1689 /*
1690  * @brief Execute a drive command.
1691  *
1692  * @param port Pointer to the port data structure.
1693  * @param command Pointer to the user specified command parameters.
1694  * @param user_buffer Pointer to the user space buffer where read sector
1695  *                   data should be copied.
1696  *
1697  * return value 0 The command completed successfully.
1698  * return value -EFAULT An error occurred while copying the completion
1699  *                 data to the user space buffer.
1700  * return value -1 An error occurred while executing the command.
1701  */
1702 static int exec_drive_command(struct mtip_port *port, u8 *command,
1703 				void __user *user_buffer)
1704 {
1705 	struct host_to_dev_fis	fis;
1706 	struct host_to_dev_fis *reply;
1707 	u8 *buf = NULL;
1708 	dma_addr_t dma_addr = 0;
1709 	int rv = 0, xfer_sz = command[3];
1710 	unsigned int to;
1711 
1712 	if (xfer_sz) {
1713 		if (!user_buffer)
1714 			return -EFAULT;
1715 
1716 		buf = dmam_alloc_coherent(&port->dd->pdev->dev,
1717 				ATA_SECT_SIZE * xfer_sz,
1718 				&dma_addr,
1719 				GFP_KERNEL);
1720 		if (!buf) {
1721 			dev_err(&port->dd->pdev->dev,
1722 				"Memory allocation failed (%d bytes)\n",
1723 				ATA_SECT_SIZE * xfer_sz);
1724 			return -ENOMEM;
1725 		}
1726 		memset(buf, 0, ATA_SECT_SIZE * xfer_sz);
1727 	}
1728 
1729 	/* Build the FIS. */
1730 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1731 	fis.type	= 0x27;
1732 	fis.opts	= 1 << 7;
1733 	fis.command	= command[0];
1734 	fis.features	= command[2];
1735 	fis.sect_count	= command[3];
1736 	if (fis.command == ATA_CMD_SMART) {
1737 		fis.sector	= command[1];
1738 		fis.cyl_low	= 0x4F;
1739 		fis.cyl_hi	= 0xC2;
1740 	}
1741 
1742 	mtip_set_timeout(port->dd, &fis, &to, 0);
1743 
1744 	if (xfer_sz)
1745 		reply = (port->rxfis + RX_FIS_PIO_SETUP);
1746 	else
1747 		reply = (port->rxfis + RX_FIS_D2H_REG);
1748 
1749 	dbg_printk(MTIP_DRV_NAME
1750 		" %s: User Command: cmd %x, sect %x, "
1751 		"feat %x, sectcnt %x\n",
1752 		__func__,
1753 		command[0],
1754 		command[1],
1755 		command[2],
1756 		command[3]);
1757 
1758 	/* Execute the command. */
1759 	if (mtip_exec_internal_command(port,
1760 				&fis,
1761 				 5,
1762 				 (xfer_sz ? dma_addr : 0),
1763 				 (xfer_sz ? ATA_SECT_SIZE * xfer_sz : 0),
1764 				 0,
1765 				 to)
1766 				 < 0) {
1767 		rv = -EFAULT;
1768 		goto exit_drive_command;
1769 	}
1770 
1771 	/* Collect the completion status. */
1772 	command[0] = reply->command; /* Status*/
1773 	command[1] = reply->features; /* Error*/
1774 	command[2] = reply->sect_count;
1775 
1776 	dbg_printk(MTIP_DRV_NAME
1777 		" %s: Completion Status: stat %x, "
1778 		"err %x, nsect %x\n",
1779 		__func__,
1780 		command[0],
1781 		command[1],
1782 		command[2]);
1783 
1784 	if (xfer_sz) {
1785 		if (copy_to_user(user_buffer,
1786 				 buf,
1787 				 ATA_SECT_SIZE * command[3])) {
1788 			rv = -EFAULT;
1789 			goto exit_drive_command;
1790 		}
1791 	}
1792 exit_drive_command:
1793 	if (buf)
1794 		dmam_free_coherent(&port->dd->pdev->dev,
1795 				ATA_SECT_SIZE * xfer_sz, buf, dma_addr);
1796 	return rv;
1797 }
1798 
1799 /*
1800  *  Indicates whether a command has a single sector payload.
1801  *
1802  *  @command passed to the device to perform the certain event.
1803  *  @features passed to the device to perform the certain event.
1804  *
1805  *  return value
1806  *	1	command is one that always has a single sector payload,
1807  *		regardless of the value in the Sector Count field.
1808  *      0       otherwise
1809  *
1810  */
1811 static unsigned int implicit_sector(unsigned char command,
1812 				    unsigned char features)
1813 {
1814 	unsigned int rv = 0;
1815 
1816 	/* list of commands that have an implicit sector count of 1 */
1817 	switch (command) {
1818 	case ATA_CMD_SEC_SET_PASS:
1819 	case ATA_CMD_SEC_UNLOCK:
1820 	case ATA_CMD_SEC_ERASE_PREP:
1821 	case ATA_CMD_SEC_ERASE_UNIT:
1822 	case ATA_CMD_SEC_FREEZE_LOCK:
1823 	case ATA_CMD_SEC_DISABLE_PASS:
1824 	case ATA_CMD_PMP_READ:
1825 	case ATA_CMD_PMP_WRITE:
1826 		rv = 1;
1827 		break;
1828 	case ATA_CMD_SET_MAX:
1829 		if (features == ATA_SET_MAX_UNLOCK)
1830 			rv = 1;
1831 		break;
1832 	case ATA_CMD_SMART:
1833 		if ((features == ATA_SMART_READ_VALUES) ||
1834 				(features == ATA_SMART_READ_THRESHOLDS))
1835 			rv = 1;
1836 		break;
1837 	case ATA_CMD_CONF_OVERLAY:
1838 		if ((features == ATA_DCO_IDENTIFY) ||
1839 				(features == ATA_DCO_SET))
1840 			rv = 1;
1841 		break;
1842 	}
1843 	return rv;
1844 }
1845 
1846 /*
1847  * Executes a taskfile
1848  * See ide_taskfile_ioctl() for derivation
1849  */
1850 static int exec_drive_taskfile(struct driver_data *dd,
1851 			       void __user *buf,
1852 			       ide_task_request_t *req_task,
1853 			       int outtotal)
1854 {
1855 	struct host_to_dev_fis	fis;
1856 	struct host_to_dev_fis *reply;
1857 	u8 *outbuf = NULL;
1858 	u8 *inbuf = NULL;
1859 	dma_addr_t outbuf_dma = 0;
1860 	dma_addr_t inbuf_dma = 0;
1861 	dma_addr_t dma_buffer = 0;
1862 	int err = 0;
1863 	unsigned int taskin = 0;
1864 	unsigned int taskout = 0;
1865 	u8 nsect = 0;
1866 	unsigned int timeout;
1867 	unsigned int force_single_sector;
1868 	unsigned int transfer_size;
1869 	unsigned long task_file_data;
1870 	int intotal = outtotal + req_task->out_size;
1871 	int erasemode = 0;
1872 
1873 	taskout = req_task->out_size;
1874 	taskin = req_task->in_size;
1875 	/* 130560 = 512 * 0xFF*/
1876 	if (taskin > 130560 || taskout > 130560)
1877 		return -EINVAL;
1878 
1879 	if (taskout) {
1880 		outbuf = memdup_user(buf + outtotal, taskout);
1881 		if (IS_ERR(outbuf))
1882 			return PTR_ERR(outbuf);
1883 
1884 		outbuf_dma = pci_map_single(dd->pdev,
1885 					 outbuf,
1886 					 taskout,
1887 					 DMA_TO_DEVICE);
1888 		if (pci_dma_mapping_error(dd->pdev, outbuf_dma)) {
1889 			err = -ENOMEM;
1890 			goto abort;
1891 		}
1892 		dma_buffer = outbuf_dma;
1893 	}
1894 
1895 	if (taskin) {
1896 		inbuf = memdup_user(buf + intotal, taskin);
1897 		if (IS_ERR(inbuf)) {
1898 			err = PTR_ERR(inbuf);
1899 			inbuf = NULL;
1900 			goto abort;
1901 		}
1902 		inbuf_dma = pci_map_single(dd->pdev,
1903 					 inbuf,
1904 					 taskin, DMA_FROM_DEVICE);
1905 		if (pci_dma_mapping_error(dd->pdev, inbuf_dma)) {
1906 			err = -ENOMEM;
1907 			goto abort;
1908 		}
1909 		dma_buffer = inbuf_dma;
1910 	}
1911 
1912 	/* only supports PIO and non-data commands from this ioctl. */
1913 	switch (req_task->data_phase) {
1914 	case TASKFILE_OUT:
1915 		nsect = taskout / ATA_SECT_SIZE;
1916 		reply = (dd->port->rxfis + RX_FIS_PIO_SETUP);
1917 		break;
1918 	case TASKFILE_IN:
1919 		reply = (dd->port->rxfis + RX_FIS_PIO_SETUP);
1920 		break;
1921 	case TASKFILE_NO_DATA:
1922 		reply = (dd->port->rxfis + RX_FIS_D2H_REG);
1923 		break;
1924 	default:
1925 		err = -EINVAL;
1926 		goto abort;
1927 	}
1928 
1929 	/* Build the FIS. */
1930 	memset(&fis, 0, sizeof(struct host_to_dev_fis));
1931 
1932 	fis.type	= 0x27;
1933 	fis.opts	= 1 << 7;
1934 	fis.command	= req_task->io_ports[7];
1935 	fis.features	= req_task->io_ports[1];
1936 	fis.sect_count	= req_task->io_ports[2];
1937 	fis.lba_low	= req_task->io_ports[3];
1938 	fis.lba_mid	= req_task->io_ports[4];
1939 	fis.lba_hi	= req_task->io_ports[5];
1940 	 /* Clear the dev bit*/
1941 	fis.device	= req_task->io_ports[6] & ~0x10;
1942 
1943 	if ((req_task->in_flags.all == 0) && (req_task->out_flags.all & 1)) {
1944 		req_task->in_flags.all	=
1945 			IDE_TASKFILE_STD_IN_FLAGS |
1946 			(IDE_HOB_STD_IN_FLAGS << 8);
1947 		fis.lba_low_ex		= req_task->hob_ports[3];
1948 		fis.lba_mid_ex		= req_task->hob_ports[4];
1949 		fis.lba_hi_ex		= req_task->hob_ports[5];
1950 		fis.features_ex		= req_task->hob_ports[1];
1951 		fis.sect_cnt_ex		= req_task->hob_ports[2];
1952 
1953 	} else {
1954 		req_task->in_flags.all = IDE_TASKFILE_STD_IN_FLAGS;
1955 	}
1956 
1957 	force_single_sector = implicit_sector(fis.command, fis.features);
1958 
1959 	if ((taskin || taskout) && (!fis.sect_count)) {
1960 		if (nsect)
1961 			fis.sect_count = nsect;
1962 		else {
1963 			if (!force_single_sector) {
1964 				dev_warn(&dd->pdev->dev,
1965 					"data movement but "
1966 					"sect_count is 0\n");
1967 					err = -EINVAL;
1968 					goto abort;
1969 			}
1970 		}
1971 	}
1972 
1973 	dbg_printk(MTIP_DRV_NAME
1974 		" %s: cmd %x, feat %x, nsect %x,"
1975 		" sect/lbal %x, lcyl/lbam %x, hcyl/lbah %x,"
1976 		" head/dev %x\n",
1977 		__func__,
1978 		fis.command,
1979 		fis.features,
1980 		fis.sect_count,
1981 		fis.lba_low,
1982 		fis.lba_mid,
1983 		fis.lba_hi,
1984 		fis.device);
1985 
1986 	/* check for erase mode support during secure erase.*/
1987 	if ((fis.command == ATA_CMD_SEC_ERASE_UNIT) && outbuf &&
1988 					(outbuf[0] & MTIP_SEC_ERASE_MODE)) {
1989 		erasemode = 1;
1990 	}
1991 
1992 	mtip_set_timeout(dd, &fis, &timeout, erasemode);
1993 
1994 	/* Determine the correct transfer size.*/
1995 	if (force_single_sector)
1996 		transfer_size = ATA_SECT_SIZE;
1997 	else
1998 		transfer_size = ATA_SECT_SIZE * fis.sect_count;
1999 
2000 	/* Execute the command.*/
2001 	if (mtip_exec_internal_command(dd->port,
2002 				 &fis,
2003 				 5,
2004 				 dma_buffer,
2005 				 transfer_size,
2006 				 0,
2007 				 timeout) < 0) {
2008 		err = -EIO;
2009 		goto abort;
2010 	}
2011 
2012 	task_file_data = readl(dd->port->mmio+PORT_TFDATA);
2013 
2014 	if ((req_task->data_phase == TASKFILE_IN) && !(task_file_data & 1)) {
2015 		reply = dd->port->rxfis + RX_FIS_PIO_SETUP;
2016 		req_task->io_ports[7] = reply->control;
2017 	} else {
2018 		reply = dd->port->rxfis + RX_FIS_D2H_REG;
2019 		req_task->io_ports[7] = reply->command;
2020 	}
2021 
2022 	/* reclaim the DMA buffers.*/
2023 	if (inbuf_dma)
2024 		pci_unmap_single(dd->pdev, inbuf_dma,
2025 			taskin, DMA_FROM_DEVICE);
2026 	if (outbuf_dma)
2027 		pci_unmap_single(dd->pdev, outbuf_dma,
2028 			taskout, DMA_TO_DEVICE);
2029 	inbuf_dma  = 0;
2030 	outbuf_dma = 0;
2031 
2032 	/* return the ATA registers to the caller.*/
2033 	req_task->io_ports[1] = reply->features;
2034 	req_task->io_ports[2] = reply->sect_count;
2035 	req_task->io_ports[3] = reply->lba_low;
2036 	req_task->io_ports[4] = reply->lba_mid;
2037 	req_task->io_ports[5] = reply->lba_hi;
2038 	req_task->io_ports[6] = reply->device;
2039 
2040 	if (req_task->out_flags.all & 1)  {
2041 
2042 		req_task->hob_ports[3] = reply->lba_low_ex;
2043 		req_task->hob_ports[4] = reply->lba_mid_ex;
2044 		req_task->hob_ports[5] = reply->lba_hi_ex;
2045 		req_task->hob_ports[1] = reply->features_ex;
2046 		req_task->hob_ports[2] = reply->sect_cnt_ex;
2047 	}
2048 	dbg_printk(MTIP_DRV_NAME
2049 		" %s: Completion: stat %x,"
2050 		"err %x, sect_cnt %x, lbalo %x,"
2051 		"lbamid %x, lbahi %x, dev %x\n",
2052 		__func__,
2053 		req_task->io_ports[7],
2054 		req_task->io_ports[1],
2055 		req_task->io_ports[2],
2056 		req_task->io_ports[3],
2057 		req_task->io_ports[4],
2058 		req_task->io_ports[5],
2059 		req_task->io_ports[6]);
2060 
2061 	if (taskout) {
2062 		if (copy_to_user(buf + outtotal, outbuf, taskout)) {
2063 			err = -EFAULT;
2064 			goto abort;
2065 		}
2066 	}
2067 	if (taskin) {
2068 		if (copy_to_user(buf + intotal, inbuf, taskin)) {
2069 			err = -EFAULT;
2070 			goto abort;
2071 		}
2072 	}
2073 abort:
2074 	if (inbuf_dma)
2075 		pci_unmap_single(dd->pdev, inbuf_dma,
2076 					taskin, DMA_FROM_DEVICE);
2077 	if (outbuf_dma)
2078 		pci_unmap_single(dd->pdev, outbuf_dma,
2079 					taskout, DMA_TO_DEVICE);
2080 	kfree(outbuf);
2081 	kfree(inbuf);
2082 
2083 	return err;
2084 }
2085 
2086 /*
2087  * Handle IOCTL calls from the Block Layer.
2088  *
2089  * This function is called by the Block Layer when it receives an IOCTL
2090  * command that it does not understand. If the IOCTL command is not supported
2091  * this function returns -ENOTTY.
2092  *
2093  * @dd  Pointer to the driver data structure.
2094  * @cmd IOCTL command passed from the Block Layer.
2095  * @arg IOCTL argument passed from the Block Layer.
2096  *
2097  * return value
2098  *	0	The IOCTL completed successfully.
2099  *	-ENOTTY The specified command is not supported.
2100  *	-EFAULT An error occurred copying data to a user space buffer.
2101  *	-EIO	An error occurred while executing the command.
2102  */
2103 static int mtip_hw_ioctl(struct driver_data *dd, unsigned int cmd,
2104 			 unsigned long arg)
2105 {
2106 	switch (cmd) {
2107 	case HDIO_GET_IDENTITY:
2108 	{
2109 		if (copy_to_user((void __user *)arg, dd->port->identify,
2110 						sizeof(u16) * ATA_ID_WORDS))
2111 			return -EFAULT;
2112 		break;
2113 	}
2114 	case HDIO_DRIVE_CMD:
2115 	{
2116 		u8 drive_command[4];
2117 
2118 		/* Copy the user command info to our buffer. */
2119 		if (copy_from_user(drive_command,
2120 					 (void __user *) arg,
2121 					 sizeof(drive_command)))
2122 			return -EFAULT;
2123 
2124 		/* Execute the drive command. */
2125 		if (exec_drive_command(dd->port,
2126 					 drive_command,
2127 					 (void __user *) (arg+4)))
2128 			return -EIO;
2129 
2130 		/* Copy the status back to the users buffer. */
2131 		if (copy_to_user((void __user *) arg,
2132 					 drive_command,
2133 					 sizeof(drive_command)))
2134 			return -EFAULT;
2135 
2136 		break;
2137 	}
2138 	case HDIO_DRIVE_TASK:
2139 	{
2140 		u8 drive_command[7];
2141 
2142 		/* Copy the user command info to our buffer. */
2143 		if (copy_from_user(drive_command,
2144 					 (void __user *) arg,
2145 					 sizeof(drive_command)))
2146 			return -EFAULT;
2147 
2148 		/* Execute the drive command. */
2149 		if (exec_drive_task(dd->port, drive_command))
2150 			return -EIO;
2151 
2152 		/* Copy the status back to the users buffer. */
2153 		if (copy_to_user((void __user *) arg,
2154 					 drive_command,
2155 					 sizeof(drive_command)))
2156 			return -EFAULT;
2157 
2158 		break;
2159 	}
2160 	case HDIO_DRIVE_TASKFILE: {
2161 		ide_task_request_t req_task;
2162 		int ret, outtotal;
2163 
2164 		if (copy_from_user(&req_task, (void __user *) arg,
2165 					sizeof(req_task)))
2166 			return -EFAULT;
2167 
2168 		outtotal = sizeof(req_task);
2169 
2170 		ret = exec_drive_taskfile(dd, (void __user *) arg,
2171 						&req_task, outtotal);
2172 
2173 		if (copy_to_user((void __user *) arg, &req_task,
2174 							sizeof(req_task)))
2175 			return -EFAULT;
2176 
2177 		return ret;
2178 	}
2179 
2180 	default:
2181 		return -EINVAL;
2182 	}
2183 	return 0;
2184 }
2185 
2186 /*
2187  * Submit an IO to the hw
2188  *
2189  * This function is called by the block layer to issue an io
2190  * to the device. Upon completion, the callback function will
2191  * be called with the data parameter passed as the callback data.
2192  *
2193  * @dd       Pointer to the driver data structure.
2194  * @start    First sector to read.
2195  * @nsect    Number of sectors to read.
2196  * @nents    Number of entries in scatter list for the read command.
2197  * @tag      The tag of this read command.
2198  * @callback Pointer to the function that should be called
2199  *	     when the read completes.
2200  * @data     Callback data passed to the callback function
2201  *	     when the read completes.
2202  * @dir      Direction (read or write)
2203  *
2204  * return value
2205  *	None
2206  */
2207 static void mtip_hw_submit_io(struct driver_data *dd, struct request *rq,
2208 			      struct mtip_cmd *command, int nents,
2209 			      struct blk_mq_hw_ctx *hctx)
2210 {
2211 	struct host_to_dev_fis	*fis;
2212 	struct mtip_port *port = dd->port;
2213 	int dma_dir = rq_data_dir(rq) == READ ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
2214 	u64 start = blk_rq_pos(rq);
2215 	unsigned int nsect = blk_rq_sectors(rq);
2216 
2217 	/* Map the scatter list for DMA access */
2218 	nents = dma_map_sg(&dd->pdev->dev, command->sg, nents, dma_dir);
2219 
2220 	prefetch(&port->flags);
2221 
2222 	command->scatter_ents = nents;
2223 
2224 	/*
2225 	 * The number of retries for this command before it is
2226 	 * reported as a failure to the upper layers.
2227 	 */
2228 	command->retries = MTIP_MAX_RETRIES;
2229 
2230 	/* Fill out fis */
2231 	fis = command->command;
2232 	fis->type        = 0x27;
2233 	fis->opts        = 1 << 7;
2234 	if (dma_dir == DMA_FROM_DEVICE)
2235 		fis->command = ATA_CMD_FPDMA_READ;
2236 	else
2237 		fis->command = ATA_CMD_FPDMA_WRITE;
2238 	fis->lba_low     = start & 0xFF;
2239 	fis->lba_mid     = (start >> 8) & 0xFF;
2240 	fis->lba_hi      = (start >> 16) & 0xFF;
2241 	fis->lba_low_ex  = (start >> 24) & 0xFF;
2242 	fis->lba_mid_ex  = (start >> 32) & 0xFF;
2243 	fis->lba_hi_ex   = (start >> 40) & 0xFF;
2244 	fis->device	 = 1 << 6;
2245 	fis->features    = nsect & 0xFF;
2246 	fis->features_ex = (nsect >> 8) & 0xFF;
2247 	fis->sect_count  = ((rq->tag << 3) | (rq->tag >> 5));
2248 	fis->sect_cnt_ex = 0;
2249 	fis->control     = 0;
2250 	fis->res2        = 0;
2251 	fis->res3        = 0;
2252 	fill_command_sg(dd, command, nents);
2253 
2254 	if (unlikely(command->unaligned))
2255 		fis->device |= 1 << 7;
2256 
2257 	/* Populate the command header */
2258 	command->command_header->opts =
2259 			__force_bit2int cpu_to_le32(
2260 				(nents << 16) | 5 | AHCI_CMD_PREFETCH);
2261 	command->command_header->byte_count = 0;
2262 
2263 	command->direction = dma_dir;
2264 
2265 	/*
2266 	 * To prevent this command from being issued
2267 	 * if an internal command is in progress or error handling is active.
2268 	 */
2269 	if (unlikely(port->flags & MTIP_PF_PAUSE_IO)) {
2270 		set_bit(rq->tag, port->cmds_to_issue);
2271 		set_bit(MTIP_PF_ISSUE_CMDS_BIT, &port->flags);
2272 		return;
2273 	}
2274 
2275 	/* Issue the command to the hardware */
2276 	mtip_issue_ncq_command(port, rq->tag);
2277 }
2278 
2279 /*
2280  * Sysfs status dump.
2281  *
2282  * @dev  Pointer to the device structure, passed by the kernrel.
2283  * @attr Pointer to the device_attribute structure passed by the kernel.
2284  * @buf  Pointer to the char buffer that will receive the stats info.
2285  *
2286  * return value
2287  *	The size, in bytes, of the data copied into buf.
2288  */
2289 static ssize_t mtip_hw_show_status(struct device *dev,
2290 				struct device_attribute *attr,
2291 				char *buf)
2292 {
2293 	struct driver_data *dd = dev_to_disk(dev)->private_data;
2294 	int size = 0;
2295 
2296 	if (test_bit(MTIP_DDF_OVER_TEMP_BIT, &dd->dd_flag))
2297 		size += sprintf(buf, "%s", "thermal_shutdown\n");
2298 	else if (test_bit(MTIP_DDF_WRITE_PROTECT_BIT, &dd->dd_flag))
2299 		size += sprintf(buf, "%s", "write_protect\n");
2300 	else
2301 		size += sprintf(buf, "%s", "online\n");
2302 
2303 	return size;
2304 }
2305 
2306 static DEVICE_ATTR(status, S_IRUGO, mtip_hw_show_status, NULL);
2307 
2308 /* debugsfs entries */
2309 
2310 static ssize_t show_device_status(struct device_driver *drv, char *buf)
2311 {
2312 	int size = 0;
2313 	struct driver_data *dd, *tmp;
2314 	unsigned long flags;
2315 	char id_buf[42];
2316 	u16 status = 0;
2317 
2318 	spin_lock_irqsave(&dev_lock, flags);
2319 	size += sprintf(&buf[size], "Devices Present:\n");
2320 	list_for_each_entry_safe(dd, tmp, &online_list, online_list) {
2321 		if (dd->pdev) {
2322 			if (dd->port &&
2323 			    dd->port->identify &&
2324 			    dd->port->identify_valid) {
2325 				strlcpy(id_buf,
2326 					(char *) (dd->port->identify + 10), 21);
2327 				status = *(dd->port->identify + 141);
2328 			} else {
2329 				memset(id_buf, 0, 42);
2330 				status = 0;
2331 			}
2332 
2333 			if (dd->port &&
2334 			    test_bit(MTIP_PF_REBUILD_BIT, &dd->port->flags)) {
2335 				size += sprintf(&buf[size],
2336 					" device %s %s (ftl rebuild %d %%)\n",
2337 					dev_name(&dd->pdev->dev),
2338 					id_buf,
2339 					status);
2340 			} else {
2341 				size += sprintf(&buf[size],
2342 					" device %s %s\n",
2343 					dev_name(&dd->pdev->dev),
2344 					id_buf);
2345 			}
2346 		}
2347 	}
2348 
2349 	size += sprintf(&buf[size], "Devices Being Removed:\n");
2350 	list_for_each_entry_safe(dd, tmp, &removing_list, remove_list) {
2351 		if (dd->pdev) {
2352 			if (dd->port &&
2353 			    dd->port->identify &&
2354 			    dd->port->identify_valid) {
2355 				strlcpy(id_buf,
2356 					(char *) (dd->port->identify+10), 21);
2357 				status = *(dd->port->identify + 141);
2358 			} else {
2359 				memset(id_buf, 0, 42);
2360 				status = 0;
2361 			}
2362 
2363 			if (dd->port &&
2364 			    test_bit(MTIP_PF_REBUILD_BIT, &dd->port->flags)) {
2365 				size += sprintf(&buf[size],
2366 					" device %s %s (ftl rebuild %d %%)\n",
2367 					dev_name(&dd->pdev->dev),
2368 					id_buf,
2369 					status);
2370 			} else {
2371 				size += sprintf(&buf[size],
2372 					" device %s %s\n",
2373 					dev_name(&dd->pdev->dev),
2374 					id_buf);
2375 			}
2376 		}
2377 	}
2378 	spin_unlock_irqrestore(&dev_lock, flags);
2379 
2380 	return size;
2381 }
2382 
2383 static ssize_t mtip_hw_read_device_status(struct file *f, char __user *ubuf,
2384 						size_t len, loff_t *offset)
2385 {
2386 	struct driver_data *dd =  (struct driver_data *)f->private_data;
2387 	int size = *offset;
2388 	char *buf;
2389 	int rv = 0;
2390 
2391 	if (!len || *offset)
2392 		return 0;
2393 
2394 	buf = kzalloc(MTIP_DFS_MAX_BUF_SIZE, GFP_KERNEL);
2395 	if (!buf) {
2396 		dev_err(&dd->pdev->dev,
2397 			"Memory allocation: status buffer\n");
2398 		return -ENOMEM;
2399 	}
2400 
2401 	size += show_device_status(NULL, buf);
2402 
2403 	*offset = size <= len ? size : len;
2404 	size = copy_to_user(ubuf, buf, *offset);
2405 	if (size)
2406 		rv = -EFAULT;
2407 
2408 	kfree(buf);
2409 	return rv ? rv : *offset;
2410 }
2411 
2412 static ssize_t mtip_hw_read_registers(struct file *f, char __user *ubuf,
2413 				  size_t len, loff_t *offset)
2414 {
2415 	struct driver_data *dd =  (struct driver_data *)f->private_data;
2416 	char *buf;
2417 	u32 group_allocated;
2418 	int size = *offset;
2419 	int n, rv = 0;
2420 
2421 	if (!len || size)
2422 		return 0;
2423 
2424 	buf = kzalloc(MTIP_DFS_MAX_BUF_SIZE, GFP_KERNEL);
2425 	if (!buf) {
2426 		dev_err(&dd->pdev->dev,
2427 			"Memory allocation: register buffer\n");
2428 		return -ENOMEM;
2429 	}
2430 
2431 	size += sprintf(&buf[size], "H/ S ACTive      : [ 0x");
2432 
2433 	for (n = dd->slot_groups-1; n >= 0; n--)
2434 		size += sprintf(&buf[size], "%08X ",
2435 					 readl(dd->port->s_active[n]));
2436 
2437 	size += sprintf(&buf[size], "]\n");
2438 	size += sprintf(&buf[size], "H/ Command Issue : [ 0x");
2439 
2440 	for (n = dd->slot_groups-1; n >= 0; n--)
2441 		size += sprintf(&buf[size], "%08X ",
2442 					readl(dd->port->cmd_issue[n]));
2443 
2444 	size += sprintf(&buf[size], "]\n");
2445 	size += sprintf(&buf[size], "H/ Completed     : [ 0x");
2446 
2447 	for (n = dd->slot_groups-1; n >= 0; n--)
2448 		size += sprintf(&buf[size], "%08X ",
2449 				readl(dd->port->completed[n]));
2450 
2451 	size += sprintf(&buf[size], "]\n");
2452 	size += sprintf(&buf[size], "H/ PORT IRQ STAT : [ 0x%08X ]\n",
2453 				readl(dd->port->mmio + PORT_IRQ_STAT));
2454 	size += sprintf(&buf[size], "H/ HOST IRQ STAT : [ 0x%08X ]\n",
2455 				readl(dd->mmio + HOST_IRQ_STAT));
2456 	size += sprintf(&buf[size], "\n");
2457 
2458 	size += sprintf(&buf[size], "L/ Commands in Q : [ 0x");
2459 
2460 	for (n = dd->slot_groups-1; n >= 0; n--) {
2461 		if (sizeof(long) > sizeof(u32))
2462 			group_allocated =
2463 				dd->port->cmds_to_issue[n/2] >> (32*(n&1));
2464 		else
2465 			group_allocated = dd->port->cmds_to_issue[n];
2466 		size += sprintf(&buf[size], "%08X ", group_allocated);
2467 	}
2468 	size += sprintf(&buf[size], "]\n");
2469 
2470 	*offset = size <= len ? size : len;
2471 	size = copy_to_user(ubuf, buf, *offset);
2472 	if (size)
2473 		rv = -EFAULT;
2474 
2475 	kfree(buf);
2476 	return rv ? rv : *offset;
2477 }
2478 
2479 static ssize_t mtip_hw_read_flags(struct file *f, char __user *ubuf,
2480 				  size_t len, loff_t *offset)
2481 {
2482 	struct driver_data *dd =  (struct driver_data *)f->private_data;
2483 	char *buf;
2484 	int size = *offset;
2485 	int rv = 0;
2486 
2487 	if (!len || size)
2488 		return 0;
2489 
2490 	buf = kzalloc(MTIP_DFS_MAX_BUF_SIZE, GFP_KERNEL);
2491 	if (!buf) {
2492 		dev_err(&dd->pdev->dev,
2493 			"Memory allocation: flag buffer\n");
2494 		return -ENOMEM;
2495 	}
2496 
2497 	size += sprintf(&buf[size], "Flag-port : [ %08lX ]\n",
2498 							dd->port->flags);
2499 	size += sprintf(&buf[size], "Flag-dd   : [ %08lX ]\n",
2500 							dd->dd_flag);
2501 
2502 	*offset = size <= len ? size : len;
2503 	size = copy_to_user(ubuf, buf, *offset);
2504 	if (size)
2505 		rv = -EFAULT;
2506 
2507 	kfree(buf);
2508 	return rv ? rv : *offset;
2509 }
2510 
2511 static const struct file_operations mtip_device_status_fops = {
2512 	.owner  = THIS_MODULE,
2513 	.open   = simple_open,
2514 	.read   = mtip_hw_read_device_status,
2515 	.llseek = no_llseek,
2516 };
2517 
2518 static const struct file_operations mtip_regs_fops = {
2519 	.owner  = THIS_MODULE,
2520 	.open   = simple_open,
2521 	.read   = mtip_hw_read_registers,
2522 	.llseek = no_llseek,
2523 };
2524 
2525 static const struct file_operations mtip_flags_fops = {
2526 	.owner  = THIS_MODULE,
2527 	.open   = simple_open,
2528 	.read   = mtip_hw_read_flags,
2529 	.llseek = no_llseek,
2530 };
2531 
2532 /*
2533  * Create the sysfs related attributes.
2534  *
2535  * @dd   Pointer to the driver data structure.
2536  * @kobj Pointer to the kobj for the block device.
2537  *
2538  * return value
2539  *	0	Operation completed successfully.
2540  *	-EINVAL Invalid parameter.
2541  */
2542 static int mtip_hw_sysfs_init(struct driver_data *dd, struct kobject *kobj)
2543 {
2544 	if (!kobj || !dd)
2545 		return -EINVAL;
2546 
2547 	if (sysfs_create_file(kobj, &dev_attr_status.attr))
2548 		dev_warn(&dd->pdev->dev,
2549 			"Error creating 'status' sysfs entry\n");
2550 	return 0;
2551 }
2552 
2553 /*
2554  * Remove the sysfs related attributes.
2555  *
2556  * @dd   Pointer to the driver data structure.
2557  * @kobj Pointer to the kobj for the block device.
2558  *
2559  * return value
2560  *	0	Operation completed successfully.
2561  *	-EINVAL Invalid parameter.
2562  */
2563 static int mtip_hw_sysfs_exit(struct driver_data *dd, struct kobject *kobj)
2564 {
2565 	if (!kobj || !dd)
2566 		return -EINVAL;
2567 
2568 	sysfs_remove_file(kobj, &dev_attr_status.attr);
2569 
2570 	return 0;
2571 }
2572 
2573 static int mtip_hw_debugfs_init(struct driver_data *dd)
2574 {
2575 	if (!dfs_parent)
2576 		return -1;
2577 
2578 	dd->dfs_node = debugfs_create_dir(dd->disk->disk_name, dfs_parent);
2579 	if (IS_ERR_OR_NULL(dd->dfs_node)) {
2580 		dev_warn(&dd->pdev->dev,
2581 			"Error creating node %s under debugfs\n",
2582 						dd->disk->disk_name);
2583 		dd->dfs_node = NULL;
2584 		return -1;
2585 	}
2586 
2587 	debugfs_create_file("flags", S_IRUGO, dd->dfs_node, dd,
2588 							&mtip_flags_fops);
2589 	debugfs_create_file("registers", S_IRUGO, dd->dfs_node, dd,
2590 							&mtip_regs_fops);
2591 
2592 	return 0;
2593 }
2594 
2595 static void mtip_hw_debugfs_exit(struct driver_data *dd)
2596 {
2597 	if (dd->dfs_node)
2598 		debugfs_remove_recursive(dd->dfs_node);
2599 }
2600 
2601 /*
2602  * Perform any init/resume time hardware setup
2603  *
2604  * @dd Pointer to the driver data structure.
2605  *
2606  * return value
2607  *	None
2608  */
2609 static inline void hba_setup(struct driver_data *dd)
2610 {
2611 	u32 hwdata;
2612 	hwdata = readl(dd->mmio + HOST_HSORG);
2613 
2614 	/* interrupt bug workaround: use only 1 IS bit.*/
2615 	writel(hwdata |
2616 		HSORG_DISABLE_SLOTGRP_INTR |
2617 		HSORG_DISABLE_SLOTGRP_PXIS,
2618 		dd->mmio + HOST_HSORG);
2619 }
2620 
2621 static int mtip_device_unaligned_constrained(struct driver_data *dd)
2622 {
2623 	return (dd->pdev->device == P420M_DEVICE_ID ? 1 : 0);
2624 }
2625 
2626 /*
2627  * Detect the details of the product, and store anything needed
2628  * into the driver data structure.  This includes product type and
2629  * version and number of slot groups.
2630  *
2631  * @dd Pointer to the driver data structure.
2632  *
2633  * return value
2634  *	None
2635  */
2636 static void mtip_detect_product(struct driver_data *dd)
2637 {
2638 	u32 hwdata;
2639 	unsigned int rev, slotgroups;
2640 
2641 	/*
2642 	 * HBA base + 0xFC [15:0] - vendor-specific hardware interface
2643 	 * info register:
2644 	 * [15:8] hardware/software interface rev#
2645 	 * [   3] asic-style interface
2646 	 * [ 2:0] number of slot groups, minus 1 (only valid for asic-style).
2647 	 */
2648 	hwdata = readl(dd->mmio + HOST_HSORG);
2649 
2650 	dd->product_type = MTIP_PRODUCT_UNKNOWN;
2651 	dd->slot_groups = 1;
2652 
2653 	if (hwdata & 0x8) {
2654 		dd->product_type = MTIP_PRODUCT_ASICFPGA;
2655 		rev = (hwdata & HSORG_HWREV) >> 8;
2656 		slotgroups = (hwdata & HSORG_SLOTGROUPS) + 1;
2657 		dev_info(&dd->pdev->dev,
2658 			"ASIC-FPGA design, HS rev 0x%x, "
2659 			"%i slot groups [%i slots]\n",
2660 			 rev,
2661 			 slotgroups,
2662 			 slotgroups * 32);
2663 
2664 		if (slotgroups > MTIP_MAX_SLOT_GROUPS) {
2665 			dev_warn(&dd->pdev->dev,
2666 				"Warning: driver only supports "
2667 				"%i slot groups.\n", MTIP_MAX_SLOT_GROUPS);
2668 			slotgroups = MTIP_MAX_SLOT_GROUPS;
2669 		}
2670 		dd->slot_groups = slotgroups;
2671 		return;
2672 	}
2673 
2674 	dev_warn(&dd->pdev->dev, "Unrecognized product id\n");
2675 }
2676 
2677 /*
2678  * Blocking wait for FTL rebuild to complete
2679  *
2680  * @dd Pointer to the DRIVER_DATA structure.
2681  *
2682  * return value
2683  *	0	FTL rebuild completed successfully
2684  *	-EFAULT FTL rebuild error/timeout/interruption
2685  */
2686 static int mtip_ftl_rebuild_poll(struct driver_data *dd)
2687 {
2688 	unsigned long timeout, cnt = 0, start;
2689 
2690 	dev_warn(&dd->pdev->dev,
2691 		"FTL rebuild in progress. Polling for completion.\n");
2692 
2693 	start = jiffies;
2694 	timeout = jiffies + msecs_to_jiffies(MTIP_FTL_REBUILD_TIMEOUT_MS);
2695 
2696 	do {
2697 		if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
2698 				&dd->dd_flag)))
2699 			return -EFAULT;
2700 		if (mtip_check_surprise_removal(dd->pdev))
2701 			return -EFAULT;
2702 
2703 		if (mtip_get_identify(dd->port, NULL) < 0)
2704 			return -EFAULT;
2705 
2706 		if (*(dd->port->identify + MTIP_FTL_REBUILD_OFFSET) ==
2707 			MTIP_FTL_REBUILD_MAGIC) {
2708 			ssleep(1);
2709 			/* Print message every 3 minutes */
2710 			if (cnt++ >= 180) {
2711 				dev_warn(&dd->pdev->dev,
2712 				"FTL rebuild in progress (%d secs).\n",
2713 				jiffies_to_msecs(jiffies - start) / 1000);
2714 				cnt = 0;
2715 			}
2716 		} else {
2717 			dev_warn(&dd->pdev->dev,
2718 				"FTL rebuild complete (%d secs).\n",
2719 			jiffies_to_msecs(jiffies - start) / 1000);
2720 			mtip_block_initialize(dd);
2721 			return 0;
2722 		}
2723 	} while (time_before(jiffies, timeout));
2724 
2725 	/* Check for timeout */
2726 	dev_err(&dd->pdev->dev,
2727 		"Timed out waiting for FTL rebuild to complete (%d secs).\n",
2728 		jiffies_to_msecs(jiffies - start) / 1000);
2729 	return -EFAULT;
2730 }
2731 
2732 static void mtip_softirq_done_fn(struct request *rq)
2733 {
2734 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
2735 	struct driver_data *dd = rq->q->queuedata;
2736 
2737 	/* Unmap the DMA scatter list entries */
2738 	dma_unmap_sg(&dd->pdev->dev, cmd->sg, cmd->scatter_ents,
2739 							cmd->direction);
2740 
2741 	if (unlikely(cmd->unaligned))
2742 		up(&dd->port->cmd_slot_unal);
2743 
2744 	blk_mq_end_request(rq, cmd->status);
2745 }
2746 
2747 static void mtip_abort_cmd(struct request *req, void *data,
2748 							bool reserved)
2749 {
2750 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(req);
2751 	struct driver_data *dd = data;
2752 
2753 	dbg_printk(MTIP_DRV_NAME " Aborting request, tag = %d\n", req->tag);
2754 
2755 	clear_bit(req->tag, dd->port->cmds_to_issue);
2756 	cmd->status = -EIO;
2757 	mtip_softirq_done_fn(req);
2758 }
2759 
2760 static void mtip_queue_cmd(struct request *req, void *data,
2761 							bool reserved)
2762 {
2763 	struct driver_data *dd = data;
2764 
2765 	set_bit(req->tag, dd->port->cmds_to_issue);
2766 	blk_abort_request(req);
2767 }
2768 
2769 /*
2770  * service thread to issue queued commands
2771  *
2772  * @data Pointer to the driver data structure.
2773  *
2774  * return value
2775  *	0
2776  */
2777 
2778 static int mtip_service_thread(void *data)
2779 {
2780 	struct driver_data *dd = (struct driver_data *)data;
2781 	unsigned long slot, slot_start, slot_wrap, to;
2782 	unsigned int num_cmd_slots = dd->slot_groups * 32;
2783 	struct mtip_port *port = dd->port;
2784 
2785 	while (1) {
2786 		if (kthread_should_stop() ||
2787 			test_bit(MTIP_PF_SVC_THD_STOP_BIT, &port->flags))
2788 			goto st_out;
2789 		clear_bit(MTIP_PF_SVC_THD_ACTIVE_BIT, &port->flags);
2790 
2791 		/*
2792 		 * the condition is to check neither an internal command is
2793 		 * is in progress nor error handling is active
2794 		 */
2795 		wait_event_interruptible(port->svc_wait, (port->flags) &&
2796 			(port->flags & MTIP_PF_SVC_THD_WORK));
2797 
2798 		if (kthread_should_stop() ||
2799 			test_bit(MTIP_PF_SVC_THD_STOP_BIT, &port->flags))
2800 			goto st_out;
2801 
2802 		if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
2803 				&dd->dd_flag)))
2804 			goto st_out;
2805 
2806 		set_bit(MTIP_PF_SVC_THD_ACTIVE_BIT, &port->flags);
2807 
2808 restart_eh:
2809 		/* Demux bits: start with error handling */
2810 		if (test_bit(MTIP_PF_EH_ACTIVE_BIT, &port->flags)) {
2811 			mtip_handle_tfe(dd);
2812 			clear_bit(MTIP_PF_EH_ACTIVE_BIT, &port->flags);
2813 		}
2814 
2815 		if (test_bit(MTIP_PF_EH_ACTIVE_BIT, &port->flags))
2816 			goto restart_eh;
2817 
2818 		if (test_bit(MTIP_PF_TO_ACTIVE_BIT, &port->flags)) {
2819 			to = jiffies + msecs_to_jiffies(5000);
2820 
2821 			do {
2822 				mdelay(100);
2823 			} while (atomic_read(&dd->irq_workers_active) != 0 &&
2824 				time_before(jiffies, to));
2825 
2826 			if (atomic_read(&dd->irq_workers_active) != 0)
2827 				dev_warn(&dd->pdev->dev,
2828 					"Completion workers still active!");
2829 
2830 			spin_lock(dd->queue->queue_lock);
2831 			blk_mq_tagset_busy_iter(&dd->tags,
2832 							mtip_queue_cmd, dd);
2833 			spin_unlock(dd->queue->queue_lock);
2834 
2835 			set_bit(MTIP_PF_ISSUE_CMDS_BIT, &dd->port->flags);
2836 
2837 			if (mtip_device_reset(dd))
2838 				blk_mq_tagset_busy_iter(&dd->tags,
2839 							mtip_abort_cmd, dd);
2840 
2841 			clear_bit(MTIP_PF_TO_ACTIVE_BIT, &dd->port->flags);
2842 		}
2843 
2844 		if (test_bit(MTIP_PF_ISSUE_CMDS_BIT, &port->flags)) {
2845 			slot = 1;
2846 			/* used to restrict the loop to one iteration */
2847 			slot_start = num_cmd_slots;
2848 			slot_wrap = 0;
2849 			while (1) {
2850 				slot = find_next_bit(port->cmds_to_issue,
2851 						num_cmd_slots, slot);
2852 				if (slot_wrap == 1) {
2853 					if ((slot_start >= slot) ||
2854 						(slot >= num_cmd_slots))
2855 						break;
2856 				}
2857 				if (unlikely(slot_start == num_cmd_slots))
2858 					slot_start = slot;
2859 
2860 				if (unlikely(slot == num_cmd_slots)) {
2861 					slot = 1;
2862 					slot_wrap = 1;
2863 					continue;
2864 				}
2865 
2866 				/* Issue the command to the hardware */
2867 				mtip_issue_ncq_command(port, slot);
2868 
2869 				clear_bit(slot, port->cmds_to_issue);
2870 			}
2871 
2872 			clear_bit(MTIP_PF_ISSUE_CMDS_BIT, &port->flags);
2873 		}
2874 
2875 		if (test_bit(MTIP_PF_REBUILD_BIT, &port->flags)) {
2876 			if (mtip_ftl_rebuild_poll(dd) == 0)
2877 				clear_bit(MTIP_PF_REBUILD_BIT, &port->flags);
2878 		}
2879 	}
2880 
2881 st_out:
2882 	return 0;
2883 }
2884 
2885 /*
2886  * DMA region teardown
2887  *
2888  * @dd Pointer to driver_data structure
2889  *
2890  * return value
2891  *      None
2892  */
2893 static void mtip_dma_free(struct driver_data *dd)
2894 {
2895 	struct mtip_port *port = dd->port;
2896 
2897 	if (port->block1)
2898 		dmam_free_coherent(&dd->pdev->dev, BLOCK_DMA_ALLOC_SZ,
2899 					port->block1, port->block1_dma);
2900 
2901 	if (port->command_list) {
2902 		dmam_free_coherent(&dd->pdev->dev, AHCI_CMD_TBL_SZ,
2903 				port->command_list, port->command_list_dma);
2904 	}
2905 }
2906 
2907 /*
2908  * DMA region setup
2909  *
2910  * @dd Pointer to driver_data structure
2911  *
2912  * return value
2913  *      -ENOMEM Not enough free DMA region space to initialize driver
2914  */
2915 static int mtip_dma_alloc(struct driver_data *dd)
2916 {
2917 	struct mtip_port *port = dd->port;
2918 
2919 	/* Allocate dma memory for RX Fis, Identify, and Sector Bufffer */
2920 	port->block1 =
2921 		dmam_alloc_coherent(&dd->pdev->dev, BLOCK_DMA_ALLOC_SZ,
2922 					&port->block1_dma, GFP_KERNEL);
2923 	if (!port->block1)
2924 		return -ENOMEM;
2925 	memset(port->block1, 0, BLOCK_DMA_ALLOC_SZ);
2926 
2927 	/* Allocate dma memory for command list */
2928 	port->command_list =
2929 		dmam_alloc_coherent(&dd->pdev->dev, AHCI_CMD_TBL_SZ,
2930 					&port->command_list_dma, GFP_KERNEL);
2931 	if (!port->command_list) {
2932 		dmam_free_coherent(&dd->pdev->dev, BLOCK_DMA_ALLOC_SZ,
2933 					port->block1, port->block1_dma);
2934 		port->block1 = NULL;
2935 		port->block1_dma = 0;
2936 		return -ENOMEM;
2937 	}
2938 	memset(port->command_list, 0, AHCI_CMD_TBL_SZ);
2939 
2940 	/* Setup all pointers into first DMA region */
2941 	port->rxfis         = port->block1 + AHCI_RX_FIS_OFFSET;
2942 	port->rxfis_dma     = port->block1_dma + AHCI_RX_FIS_OFFSET;
2943 	port->identify      = port->block1 + AHCI_IDFY_OFFSET;
2944 	port->identify_dma  = port->block1_dma + AHCI_IDFY_OFFSET;
2945 	port->log_buf       = port->block1 + AHCI_SECTBUF_OFFSET;
2946 	port->log_buf_dma   = port->block1_dma + AHCI_SECTBUF_OFFSET;
2947 	port->smart_buf     = port->block1 + AHCI_SMARTBUF_OFFSET;
2948 	port->smart_buf_dma = port->block1_dma + AHCI_SMARTBUF_OFFSET;
2949 
2950 	return 0;
2951 }
2952 
2953 static int mtip_hw_get_identify(struct driver_data *dd)
2954 {
2955 	struct smart_attr attr242;
2956 	unsigned char *buf;
2957 	int rv;
2958 
2959 	if (mtip_get_identify(dd->port, NULL) < 0)
2960 		return -EFAULT;
2961 
2962 	if (*(dd->port->identify + MTIP_FTL_REBUILD_OFFSET) ==
2963 		MTIP_FTL_REBUILD_MAGIC) {
2964 		set_bit(MTIP_PF_REBUILD_BIT, &dd->port->flags);
2965 		return MTIP_FTL_REBUILD_MAGIC;
2966 	}
2967 	mtip_dump_identify(dd->port);
2968 
2969 	/* check write protect, over temp and rebuild statuses */
2970 	rv = mtip_read_log_page(dd->port, ATA_LOG_SATA_NCQ,
2971 				dd->port->log_buf,
2972 				dd->port->log_buf_dma, 1);
2973 	if (rv) {
2974 		dev_warn(&dd->pdev->dev,
2975 			"Error in READ LOG EXT (10h) command\n");
2976 		/* non-critical error, don't fail the load */
2977 	} else {
2978 		buf = (unsigned char *)dd->port->log_buf;
2979 		if (buf[259] & 0x1) {
2980 			dev_info(&dd->pdev->dev,
2981 				"Write protect bit is set.\n");
2982 			set_bit(MTIP_DDF_WRITE_PROTECT_BIT, &dd->dd_flag);
2983 		}
2984 		if (buf[288] == 0xF7) {
2985 			dev_info(&dd->pdev->dev,
2986 				"Exceeded Tmax, drive in thermal shutdown.\n");
2987 			set_bit(MTIP_DDF_OVER_TEMP_BIT, &dd->dd_flag);
2988 		}
2989 		if (buf[288] == 0xBF) {
2990 			dev_info(&dd->pdev->dev,
2991 				"Drive indicates rebuild has failed.\n");
2992 			set_bit(MTIP_DDF_REBUILD_FAILED_BIT, &dd->dd_flag);
2993 		}
2994 	}
2995 
2996 	/* get write protect progess */
2997 	memset(&attr242, 0, sizeof(struct smart_attr));
2998 	if (mtip_get_smart_attr(dd->port, 242, &attr242))
2999 		dev_warn(&dd->pdev->dev,
3000 				"Unable to check write protect progress\n");
3001 	else
3002 		dev_info(&dd->pdev->dev,
3003 				"Write protect progress: %u%% (%u blocks)\n",
3004 				attr242.cur, le32_to_cpu(attr242.data));
3005 
3006 	return rv;
3007 }
3008 
3009 /*
3010  * Called once for each card.
3011  *
3012  * @dd Pointer to the driver data structure.
3013  *
3014  * return value
3015  *	0 on success, else an error code.
3016  */
3017 static int mtip_hw_init(struct driver_data *dd)
3018 {
3019 	int i;
3020 	int rv;
3021 	unsigned int num_command_slots;
3022 	unsigned long timeout, timetaken;
3023 
3024 	dd->mmio = pcim_iomap_table(dd->pdev)[MTIP_ABAR];
3025 
3026 	mtip_detect_product(dd);
3027 	if (dd->product_type == MTIP_PRODUCT_UNKNOWN) {
3028 		rv = -EIO;
3029 		goto out1;
3030 	}
3031 	num_command_slots = dd->slot_groups * 32;
3032 
3033 	hba_setup(dd);
3034 
3035 	dd->port = kzalloc_node(sizeof(struct mtip_port), GFP_KERNEL,
3036 				dd->numa_node);
3037 	if (!dd->port) {
3038 		dev_err(&dd->pdev->dev,
3039 			"Memory allocation: port structure\n");
3040 		return -ENOMEM;
3041 	}
3042 
3043 	/* Continue workqueue setup */
3044 	for (i = 0; i < MTIP_MAX_SLOT_GROUPS; i++)
3045 		dd->work[i].port = dd->port;
3046 
3047 	/* Enable unaligned IO constraints for some devices */
3048 	if (mtip_device_unaligned_constrained(dd))
3049 		dd->unal_qdepth = MTIP_MAX_UNALIGNED_SLOTS;
3050 	else
3051 		dd->unal_qdepth = 0;
3052 
3053 	sema_init(&dd->port->cmd_slot_unal, dd->unal_qdepth);
3054 
3055 	/* Spinlock to prevent concurrent issue */
3056 	for (i = 0; i < MTIP_MAX_SLOT_GROUPS; i++)
3057 		spin_lock_init(&dd->port->cmd_issue_lock[i]);
3058 
3059 	/* Set the port mmio base address. */
3060 	dd->port->mmio	= dd->mmio + PORT_OFFSET;
3061 	dd->port->dd	= dd;
3062 
3063 	/* DMA allocations */
3064 	rv = mtip_dma_alloc(dd);
3065 	if (rv < 0)
3066 		goto out1;
3067 
3068 	/* Setup the pointers to the extended s_active and CI registers. */
3069 	for (i = 0; i < dd->slot_groups; i++) {
3070 		dd->port->s_active[i] =
3071 			dd->port->mmio + i*0x80 + PORT_SCR_ACT;
3072 		dd->port->cmd_issue[i] =
3073 			dd->port->mmio + i*0x80 + PORT_COMMAND_ISSUE;
3074 		dd->port->completed[i] =
3075 			dd->port->mmio + i*0x80 + PORT_SDBV;
3076 	}
3077 
3078 	timetaken = jiffies;
3079 	timeout = jiffies + msecs_to_jiffies(30000);
3080 	while (((readl(dd->port->mmio + PORT_SCR_STAT) & 0x0F) != 0x03) &&
3081 		 time_before(jiffies, timeout)) {
3082 		mdelay(100);
3083 	}
3084 	if (unlikely(mtip_check_surprise_removal(dd->pdev))) {
3085 		timetaken = jiffies - timetaken;
3086 		dev_warn(&dd->pdev->dev,
3087 			"Surprise removal detected at %u ms\n",
3088 			jiffies_to_msecs(timetaken));
3089 		rv = -ENODEV;
3090 		goto out2 ;
3091 	}
3092 	if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag))) {
3093 		timetaken = jiffies - timetaken;
3094 		dev_warn(&dd->pdev->dev,
3095 			"Removal detected at %u ms\n",
3096 			jiffies_to_msecs(timetaken));
3097 		rv = -EFAULT;
3098 		goto out2;
3099 	}
3100 
3101 	/* Conditionally reset the HBA. */
3102 	if (!(readl(dd->mmio + HOST_CAP) & HOST_CAP_NZDMA)) {
3103 		if (mtip_hba_reset(dd) < 0) {
3104 			dev_err(&dd->pdev->dev,
3105 				"Card did not reset within timeout\n");
3106 			rv = -EIO;
3107 			goto out2;
3108 		}
3109 	} else {
3110 		/* Clear any pending interrupts on the HBA */
3111 		writel(readl(dd->mmio + HOST_IRQ_STAT),
3112 			dd->mmio + HOST_IRQ_STAT);
3113 	}
3114 
3115 	mtip_init_port(dd->port);
3116 	mtip_start_port(dd->port);
3117 
3118 	/* Setup the ISR and enable interrupts. */
3119 	rv = devm_request_irq(&dd->pdev->dev,
3120 				dd->pdev->irq,
3121 				mtip_irq_handler,
3122 				IRQF_SHARED,
3123 				dev_driver_string(&dd->pdev->dev),
3124 				dd);
3125 
3126 	if (rv) {
3127 		dev_err(&dd->pdev->dev,
3128 			"Unable to allocate IRQ %d\n", dd->pdev->irq);
3129 		goto out2;
3130 	}
3131 	irq_set_affinity_hint(dd->pdev->irq, get_cpu_mask(dd->isr_binding));
3132 
3133 	/* Enable interrupts on the HBA. */
3134 	writel(readl(dd->mmio + HOST_CTL) | HOST_IRQ_EN,
3135 					dd->mmio + HOST_CTL);
3136 
3137 	init_waitqueue_head(&dd->port->svc_wait);
3138 
3139 	if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)) {
3140 		rv = -EFAULT;
3141 		goto out3;
3142 	}
3143 
3144 	return rv;
3145 
3146 out3:
3147 	/* Disable interrupts on the HBA. */
3148 	writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
3149 			dd->mmio + HOST_CTL);
3150 
3151 	/* Release the IRQ. */
3152 	irq_set_affinity_hint(dd->pdev->irq, NULL);
3153 	devm_free_irq(&dd->pdev->dev, dd->pdev->irq, dd);
3154 
3155 out2:
3156 	mtip_deinit_port(dd->port);
3157 	mtip_dma_free(dd);
3158 
3159 out1:
3160 	/* Free the memory allocated for the for structure. */
3161 	kfree(dd->port);
3162 
3163 	return rv;
3164 }
3165 
3166 static int mtip_standby_drive(struct driver_data *dd)
3167 {
3168 	int rv = 0;
3169 
3170 	if (dd->sr || !dd->port)
3171 		return -ENODEV;
3172 	/*
3173 	 * Send standby immediate (E0h) to the drive so that it
3174 	 * saves its state.
3175 	 */
3176 	if (!test_bit(MTIP_PF_REBUILD_BIT, &dd->port->flags) &&
3177 	    !test_bit(MTIP_DDF_REBUILD_FAILED_BIT, &dd->dd_flag) &&
3178 	    !test_bit(MTIP_DDF_SEC_LOCK_BIT, &dd->dd_flag)) {
3179 		rv = mtip_standby_immediate(dd->port);
3180 		if (rv)
3181 			dev_warn(&dd->pdev->dev,
3182 				"STANDBY IMMEDIATE failed\n");
3183 	}
3184 	return rv;
3185 }
3186 
3187 /*
3188  * Called to deinitialize an interface.
3189  *
3190  * @dd Pointer to the driver data structure.
3191  *
3192  * return value
3193  *	0
3194  */
3195 static int mtip_hw_exit(struct driver_data *dd)
3196 {
3197 	if (!dd->sr) {
3198 		/* de-initialize the port. */
3199 		mtip_deinit_port(dd->port);
3200 
3201 		/* Disable interrupts on the HBA. */
3202 		writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
3203 				dd->mmio + HOST_CTL);
3204 	}
3205 
3206 	/* Release the IRQ. */
3207 	irq_set_affinity_hint(dd->pdev->irq, NULL);
3208 	devm_free_irq(&dd->pdev->dev, dd->pdev->irq, dd);
3209 	msleep(1000);
3210 
3211 	/* Free dma regions */
3212 	mtip_dma_free(dd);
3213 
3214 	/* Free the memory allocated for the for structure. */
3215 	kfree(dd->port);
3216 	dd->port = NULL;
3217 
3218 	return 0;
3219 }
3220 
3221 /*
3222  * Issue a Standby Immediate command to the device.
3223  *
3224  * This function is called by the Block Layer just before the
3225  * system powers off during a shutdown.
3226  *
3227  * @dd Pointer to the driver data structure.
3228  *
3229  * return value
3230  *	0
3231  */
3232 static int mtip_hw_shutdown(struct driver_data *dd)
3233 {
3234 	/*
3235 	 * Send standby immediate (E0h) to the drive so that it
3236 	 * saves its state.
3237 	 */
3238 	mtip_standby_drive(dd);
3239 
3240 	return 0;
3241 }
3242 
3243 /*
3244  * Suspend function
3245  *
3246  * This function is called by the Block Layer just before the
3247  * system hibernates.
3248  *
3249  * @dd Pointer to the driver data structure.
3250  *
3251  * return value
3252  *	0	Suspend was successful
3253  *	-EFAULT Suspend was not successful
3254  */
3255 static int mtip_hw_suspend(struct driver_data *dd)
3256 {
3257 	/*
3258 	 * Send standby immediate (E0h) to the drive
3259 	 * so that it saves its state.
3260 	 */
3261 	if (mtip_standby_drive(dd) != 0) {
3262 		dev_err(&dd->pdev->dev,
3263 			"Failed standby-immediate command\n");
3264 		return -EFAULT;
3265 	}
3266 
3267 	/* Disable interrupts on the HBA.*/
3268 	writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
3269 			dd->mmio + HOST_CTL);
3270 	mtip_deinit_port(dd->port);
3271 
3272 	return 0;
3273 }
3274 
3275 /*
3276  * Resume function
3277  *
3278  * This function is called by the Block Layer as the
3279  * system resumes.
3280  *
3281  * @dd Pointer to the driver data structure.
3282  *
3283  * return value
3284  *	0	Resume was successful
3285  *      -EFAULT Resume was not successful
3286  */
3287 static int mtip_hw_resume(struct driver_data *dd)
3288 {
3289 	/* Perform any needed hardware setup steps */
3290 	hba_setup(dd);
3291 
3292 	/* Reset the HBA */
3293 	if (mtip_hba_reset(dd) != 0) {
3294 		dev_err(&dd->pdev->dev,
3295 			"Unable to reset the HBA\n");
3296 		return -EFAULT;
3297 	}
3298 
3299 	/*
3300 	 * Enable the port, DMA engine, and FIS reception specific
3301 	 * h/w in controller.
3302 	 */
3303 	mtip_init_port(dd->port);
3304 	mtip_start_port(dd->port);
3305 
3306 	/* Enable interrupts on the HBA.*/
3307 	writel(readl(dd->mmio + HOST_CTL) | HOST_IRQ_EN,
3308 			dd->mmio + HOST_CTL);
3309 
3310 	return 0;
3311 }
3312 
3313 /*
3314  * Helper function for reusing disk name
3315  * upon hot insertion.
3316  */
3317 static int rssd_disk_name_format(char *prefix,
3318 				 int index,
3319 				 char *buf,
3320 				 int buflen)
3321 {
3322 	const int base = 'z' - 'a' + 1;
3323 	char *begin = buf + strlen(prefix);
3324 	char *end = buf + buflen;
3325 	char *p;
3326 	int unit;
3327 
3328 	p = end - 1;
3329 	*p = '\0';
3330 	unit = base;
3331 	do {
3332 		if (p == begin)
3333 			return -EINVAL;
3334 		*--p = 'a' + (index % unit);
3335 		index = (index / unit) - 1;
3336 	} while (index >= 0);
3337 
3338 	memmove(begin, p, end - p);
3339 	memcpy(buf, prefix, strlen(prefix));
3340 
3341 	return 0;
3342 }
3343 
3344 /*
3345  * Block layer IOCTL handler.
3346  *
3347  * @dev Pointer to the block_device structure.
3348  * @mode ignored
3349  * @cmd IOCTL command passed from the user application.
3350  * @arg Argument passed from the user application.
3351  *
3352  * return value
3353  *	0        IOCTL completed successfully.
3354  *	-ENOTTY  IOCTL not supported or invalid driver data
3355  *                 structure pointer.
3356  */
3357 static int mtip_block_ioctl(struct block_device *dev,
3358 			    fmode_t mode,
3359 			    unsigned cmd,
3360 			    unsigned long arg)
3361 {
3362 	struct driver_data *dd = dev->bd_disk->private_data;
3363 
3364 	if (!capable(CAP_SYS_ADMIN))
3365 		return -EACCES;
3366 
3367 	if (!dd)
3368 		return -ENOTTY;
3369 
3370 	if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)))
3371 		return -ENOTTY;
3372 
3373 	switch (cmd) {
3374 	case BLKFLSBUF:
3375 		return -ENOTTY;
3376 	default:
3377 		return mtip_hw_ioctl(dd, cmd, arg);
3378 	}
3379 }
3380 
3381 #ifdef CONFIG_COMPAT
3382 /*
3383  * Block layer compat IOCTL handler.
3384  *
3385  * @dev Pointer to the block_device structure.
3386  * @mode ignored
3387  * @cmd IOCTL command passed from the user application.
3388  * @arg Argument passed from the user application.
3389  *
3390  * return value
3391  *	0        IOCTL completed successfully.
3392  *	-ENOTTY  IOCTL not supported or invalid driver data
3393  *                 structure pointer.
3394  */
3395 static int mtip_block_compat_ioctl(struct block_device *dev,
3396 			    fmode_t mode,
3397 			    unsigned cmd,
3398 			    unsigned long arg)
3399 {
3400 	struct driver_data *dd = dev->bd_disk->private_data;
3401 
3402 	if (!capable(CAP_SYS_ADMIN))
3403 		return -EACCES;
3404 
3405 	if (!dd)
3406 		return -ENOTTY;
3407 
3408 	if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)))
3409 		return -ENOTTY;
3410 
3411 	switch (cmd) {
3412 	case BLKFLSBUF:
3413 		return -ENOTTY;
3414 	case HDIO_DRIVE_TASKFILE: {
3415 		struct mtip_compat_ide_task_request_s __user *compat_req_task;
3416 		ide_task_request_t req_task;
3417 		int compat_tasksize, outtotal, ret;
3418 
3419 		compat_tasksize =
3420 			sizeof(struct mtip_compat_ide_task_request_s);
3421 
3422 		compat_req_task =
3423 			(struct mtip_compat_ide_task_request_s __user *) arg;
3424 
3425 		if (copy_from_user(&req_task, (void __user *) arg,
3426 			compat_tasksize - (2 * sizeof(compat_long_t))))
3427 			return -EFAULT;
3428 
3429 		if (get_user(req_task.out_size, &compat_req_task->out_size))
3430 			return -EFAULT;
3431 
3432 		if (get_user(req_task.in_size, &compat_req_task->in_size))
3433 			return -EFAULT;
3434 
3435 		outtotal = sizeof(struct mtip_compat_ide_task_request_s);
3436 
3437 		ret = exec_drive_taskfile(dd, (void __user *) arg,
3438 						&req_task, outtotal);
3439 
3440 		if (copy_to_user((void __user *) arg, &req_task,
3441 				compat_tasksize -
3442 				(2 * sizeof(compat_long_t))))
3443 			return -EFAULT;
3444 
3445 		if (put_user(req_task.out_size, &compat_req_task->out_size))
3446 			return -EFAULT;
3447 
3448 		if (put_user(req_task.in_size, &compat_req_task->in_size))
3449 			return -EFAULT;
3450 
3451 		return ret;
3452 	}
3453 	default:
3454 		return mtip_hw_ioctl(dd, cmd, arg);
3455 	}
3456 }
3457 #endif
3458 
3459 /*
3460  * Obtain the geometry of the device.
3461  *
3462  * You may think that this function is obsolete, but some applications,
3463  * fdisk for example still used CHS values. This function describes the
3464  * device as having 224 heads and 56 sectors per cylinder. These values are
3465  * chosen so that each cylinder is aligned on a 4KB boundary. Since a
3466  * partition is described in terms of a start and end cylinder this means
3467  * that each partition is also 4KB aligned. Non-aligned partitions adversely
3468  * affects performance.
3469  *
3470  * @dev Pointer to the block_device strucutre.
3471  * @geo Pointer to a hd_geometry structure.
3472  *
3473  * return value
3474  *	0       Operation completed successfully.
3475  *	-ENOTTY An error occurred while reading the drive capacity.
3476  */
3477 static int mtip_block_getgeo(struct block_device *dev,
3478 				struct hd_geometry *geo)
3479 {
3480 	struct driver_data *dd = dev->bd_disk->private_data;
3481 	sector_t capacity;
3482 
3483 	if (!dd)
3484 		return -ENOTTY;
3485 
3486 	if (!(mtip_hw_get_capacity(dd, &capacity))) {
3487 		dev_warn(&dd->pdev->dev,
3488 			"Could not get drive capacity.\n");
3489 		return -ENOTTY;
3490 	}
3491 
3492 	geo->heads = 224;
3493 	geo->sectors = 56;
3494 	sector_div(capacity, (geo->heads * geo->sectors));
3495 	geo->cylinders = capacity;
3496 	return 0;
3497 }
3498 
3499 static int mtip_block_open(struct block_device *dev, fmode_t mode)
3500 {
3501 	struct driver_data *dd;
3502 
3503 	if (dev && dev->bd_disk) {
3504 		dd = (struct driver_data *) dev->bd_disk->private_data;
3505 
3506 		if (dd) {
3507 			if (test_bit(MTIP_DDF_REMOVAL_BIT,
3508 							&dd->dd_flag)) {
3509 				return -ENODEV;
3510 			}
3511 			return 0;
3512 		}
3513 	}
3514 	return -ENODEV;
3515 }
3516 
3517 static void mtip_block_release(struct gendisk *disk, fmode_t mode)
3518 {
3519 }
3520 
3521 /*
3522  * Block device operation function.
3523  *
3524  * This structure contains pointers to the functions required by the block
3525  * layer.
3526  */
3527 static const struct block_device_operations mtip_block_ops = {
3528 	.open		= mtip_block_open,
3529 	.release	= mtip_block_release,
3530 	.ioctl		= mtip_block_ioctl,
3531 #ifdef CONFIG_COMPAT
3532 	.compat_ioctl	= mtip_block_compat_ioctl,
3533 #endif
3534 	.getgeo		= mtip_block_getgeo,
3535 	.owner		= THIS_MODULE
3536 };
3537 
3538 static inline bool is_se_active(struct driver_data *dd)
3539 {
3540 	if (unlikely(test_bit(MTIP_PF_SE_ACTIVE_BIT, &dd->port->flags))) {
3541 		if (dd->port->ic_pause_timer) {
3542 			unsigned long to = dd->port->ic_pause_timer +
3543 							msecs_to_jiffies(1000);
3544 			if (time_after(jiffies, to)) {
3545 				clear_bit(MTIP_PF_SE_ACTIVE_BIT,
3546 							&dd->port->flags);
3547 				clear_bit(MTIP_DDF_SEC_LOCK_BIT, &dd->dd_flag);
3548 				dd->port->ic_pause_timer = 0;
3549 				wake_up_interruptible(&dd->port->svc_wait);
3550 				return false;
3551 			}
3552 		}
3553 		return true;
3554 	}
3555 	return false;
3556 }
3557 
3558 /*
3559  * Block layer make request function.
3560  *
3561  * This function is called by the kernel to process a BIO for
3562  * the P320 device.
3563  *
3564  * @queue Pointer to the request queue. Unused other than to obtain
3565  *              the driver data structure.
3566  * @rq    Pointer to the request.
3567  *
3568  */
3569 static int mtip_submit_request(struct blk_mq_hw_ctx *hctx, struct request *rq)
3570 {
3571 	struct driver_data *dd = hctx->queue->queuedata;
3572 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3573 	unsigned int nents;
3574 
3575 	if (is_se_active(dd))
3576 		return -ENODATA;
3577 
3578 	if (unlikely(dd->dd_flag & MTIP_DDF_STOP_IO)) {
3579 		if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
3580 							&dd->dd_flag))) {
3581 			return -ENXIO;
3582 		}
3583 		if (unlikely(test_bit(MTIP_DDF_OVER_TEMP_BIT, &dd->dd_flag))) {
3584 			return -ENODATA;
3585 		}
3586 		if (unlikely(test_bit(MTIP_DDF_WRITE_PROTECT_BIT,
3587 							&dd->dd_flag) &&
3588 				rq_data_dir(rq))) {
3589 			return -ENODATA;
3590 		}
3591 		if (unlikely(test_bit(MTIP_DDF_SEC_LOCK_BIT, &dd->dd_flag) ||
3592 			test_bit(MTIP_DDF_REBUILD_FAILED_BIT, &dd->dd_flag)))
3593 			return -ENODATA;
3594 	}
3595 
3596 	if (req_op(rq) == REQ_OP_DISCARD) {
3597 		int err;
3598 
3599 		err = mtip_send_trim(dd, blk_rq_pos(rq), blk_rq_sectors(rq));
3600 		blk_mq_end_request(rq, err);
3601 		return 0;
3602 	}
3603 
3604 	/* Create the scatter list for this request. */
3605 	nents = blk_rq_map_sg(hctx->queue, rq, cmd->sg);
3606 
3607 	/* Issue the read/write. */
3608 	mtip_hw_submit_io(dd, rq, cmd, nents, hctx);
3609 	return 0;
3610 }
3611 
3612 static bool mtip_check_unal_depth(struct blk_mq_hw_ctx *hctx,
3613 				  struct request *rq)
3614 {
3615 	struct driver_data *dd = hctx->queue->queuedata;
3616 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3617 
3618 	if (rq_data_dir(rq) == READ || !dd->unal_qdepth)
3619 		return false;
3620 
3621 	/*
3622 	 * If unaligned depth must be limited on this controller, mark it
3623 	 * as unaligned if the IO isn't on a 4k boundary (start of length).
3624 	 */
3625 	if (blk_rq_sectors(rq) <= 64) {
3626 		if ((blk_rq_pos(rq) & 7) || (blk_rq_sectors(rq) & 7))
3627 			cmd->unaligned = 1;
3628 	}
3629 
3630 	if (cmd->unaligned && down_trylock(&dd->port->cmd_slot_unal))
3631 		return true;
3632 
3633 	return false;
3634 }
3635 
3636 static int mtip_issue_reserved_cmd(struct blk_mq_hw_ctx *hctx,
3637 				   struct request *rq)
3638 {
3639 	struct driver_data *dd = hctx->queue->queuedata;
3640 	struct mtip_int_cmd *icmd = rq->special;
3641 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3642 	struct mtip_cmd_sg *command_sg;
3643 
3644 	if (mtip_commands_active(dd->port))
3645 		return BLK_MQ_RQ_QUEUE_BUSY;
3646 
3647 	/* Populate the SG list */
3648 	cmd->command_header->opts =
3649 		 __force_bit2int cpu_to_le32(icmd->opts | icmd->fis_len);
3650 	if (icmd->buf_len) {
3651 		command_sg = cmd->command + AHCI_CMD_TBL_HDR_SZ;
3652 
3653 		command_sg->info =
3654 			__force_bit2int cpu_to_le32((icmd->buf_len-1) & 0x3FFFFF);
3655 		command_sg->dba	=
3656 			__force_bit2int cpu_to_le32(icmd->buffer & 0xFFFFFFFF);
3657 		command_sg->dba_upper =
3658 			__force_bit2int cpu_to_le32((icmd->buffer >> 16) >> 16);
3659 
3660 		cmd->command_header->opts |=
3661 			__force_bit2int cpu_to_le32((1 << 16));
3662 	}
3663 
3664 	/* Populate the command header */
3665 	cmd->command_header->byte_count = 0;
3666 
3667 	blk_mq_start_request(rq);
3668 	mtip_issue_non_ncq_command(dd->port, rq->tag);
3669 	return BLK_MQ_RQ_QUEUE_OK;
3670 }
3671 
3672 static int mtip_queue_rq(struct blk_mq_hw_ctx *hctx,
3673 			 const struct blk_mq_queue_data *bd)
3674 {
3675 	struct request *rq = bd->rq;
3676 	int ret;
3677 
3678 	mtip_init_cmd_header(rq);
3679 
3680 	if (blk_rq_is_passthrough(rq))
3681 		return mtip_issue_reserved_cmd(hctx, rq);
3682 
3683 	if (unlikely(mtip_check_unal_depth(hctx, rq)))
3684 		return BLK_MQ_RQ_QUEUE_BUSY;
3685 
3686 	blk_mq_start_request(rq);
3687 
3688 	ret = mtip_submit_request(hctx, rq);
3689 	if (likely(!ret))
3690 		return BLK_MQ_RQ_QUEUE_OK;
3691 
3692 	return BLK_MQ_RQ_QUEUE_ERROR;
3693 }
3694 
3695 static void mtip_free_cmd(struct blk_mq_tag_set *set, struct request *rq,
3696 			  unsigned int hctx_idx)
3697 {
3698 	struct driver_data *dd = set->driver_data;
3699 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3700 
3701 	if (!cmd->command)
3702 		return;
3703 
3704 	dmam_free_coherent(&dd->pdev->dev, CMD_DMA_ALLOC_SZ,
3705 				cmd->command, cmd->command_dma);
3706 }
3707 
3708 static int mtip_init_cmd(struct blk_mq_tag_set *set, struct request *rq,
3709 			 unsigned int hctx_idx, unsigned int numa_node)
3710 {
3711 	struct driver_data *dd = set->driver_data;
3712 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3713 
3714 	cmd->command = dmam_alloc_coherent(&dd->pdev->dev, CMD_DMA_ALLOC_SZ,
3715 			&cmd->command_dma, GFP_KERNEL);
3716 	if (!cmd->command)
3717 		return -ENOMEM;
3718 
3719 	memset(cmd->command, 0, CMD_DMA_ALLOC_SZ);
3720 
3721 	sg_init_table(cmd->sg, MTIP_MAX_SG);
3722 	return 0;
3723 }
3724 
3725 static enum blk_eh_timer_return mtip_cmd_timeout(struct request *req,
3726 								bool reserved)
3727 {
3728 	struct driver_data *dd = req->q->queuedata;
3729 
3730 	if (reserved) {
3731 		struct mtip_cmd *cmd = blk_mq_rq_to_pdu(req);
3732 
3733 		cmd->status = -ETIME;
3734 		return BLK_EH_HANDLED;
3735 	}
3736 
3737 	if (test_bit(req->tag, dd->port->cmds_to_issue))
3738 		goto exit_handler;
3739 
3740 	if (test_and_set_bit(MTIP_PF_TO_ACTIVE_BIT, &dd->port->flags))
3741 		goto exit_handler;
3742 
3743 	wake_up_interruptible(&dd->port->svc_wait);
3744 exit_handler:
3745 	return BLK_EH_RESET_TIMER;
3746 }
3747 
3748 static const struct blk_mq_ops mtip_mq_ops = {
3749 	.queue_rq	= mtip_queue_rq,
3750 	.init_request	= mtip_init_cmd,
3751 	.exit_request	= mtip_free_cmd,
3752 	.complete	= mtip_softirq_done_fn,
3753 	.timeout        = mtip_cmd_timeout,
3754 };
3755 
3756 /*
3757  * Block layer initialization function.
3758  *
3759  * This function is called once by the PCI layer for each P320
3760  * device that is connected to the system.
3761  *
3762  * @dd Pointer to the driver data structure.
3763  *
3764  * return value
3765  *	0 on success else an error code.
3766  */
3767 static int mtip_block_initialize(struct driver_data *dd)
3768 {
3769 	int rv = 0, wait_for_rebuild = 0;
3770 	sector_t capacity;
3771 	unsigned int index = 0;
3772 	struct kobject *kobj;
3773 
3774 	if (dd->disk)
3775 		goto skip_create_disk; /* hw init done, before rebuild */
3776 
3777 	if (mtip_hw_init(dd)) {
3778 		rv = -EINVAL;
3779 		goto protocol_init_error;
3780 	}
3781 
3782 	dd->disk = alloc_disk_node(MTIP_MAX_MINORS, dd->numa_node);
3783 	if (dd->disk  == NULL) {
3784 		dev_err(&dd->pdev->dev,
3785 			"Unable to allocate gendisk structure\n");
3786 		rv = -EINVAL;
3787 		goto alloc_disk_error;
3788 	}
3789 
3790 	/* Generate the disk name, implemented same as in sd.c */
3791 	do {
3792 		if (!ida_pre_get(&rssd_index_ida, GFP_KERNEL)) {
3793 			rv = -ENOMEM;
3794 			goto ida_get_error;
3795 		}
3796 
3797 		spin_lock(&rssd_index_lock);
3798 		rv = ida_get_new(&rssd_index_ida, &index);
3799 		spin_unlock(&rssd_index_lock);
3800 	} while (rv == -EAGAIN);
3801 
3802 	if (rv)
3803 		goto ida_get_error;
3804 
3805 	rv = rssd_disk_name_format("rssd",
3806 				index,
3807 				dd->disk->disk_name,
3808 				DISK_NAME_LEN);
3809 	if (rv)
3810 		goto disk_index_error;
3811 
3812 	dd->disk->major		= dd->major;
3813 	dd->disk->first_minor	= index * MTIP_MAX_MINORS;
3814 	dd->disk->minors 	= MTIP_MAX_MINORS;
3815 	dd->disk->fops		= &mtip_block_ops;
3816 	dd->disk->private_data	= dd;
3817 	dd->index		= index;
3818 
3819 	mtip_hw_debugfs_init(dd);
3820 
3821 	memset(&dd->tags, 0, sizeof(dd->tags));
3822 	dd->tags.ops = &mtip_mq_ops;
3823 	dd->tags.nr_hw_queues = 1;
3824 	dd->tags.queue_depth = MTIP_MAX_COMMAND_SLOTS;
3825 	dd->tags.reserved_tags = 1;
3826 	dd->tags.cmd_size = sizeof(struct mtip_cmd);
3827 	dd->tags.numa_node = dd->numa_node;
3828 	dd->tags.flags = BLK_MQ_F_SHOULD_MERGE;
3829 	dd->tags.driver_data = dd;
3830 	dd->tags.timeout = MTIP_NCQ_CMD_TIMEOUT_MS;
3831 
3832 	rv = blk_mq_alloc_tag_set(&dd->tags);
3833 	if (rv) {
3834 		dev_err(&dd->pdev->dev,
3835 			"Unable to allocate request queue\n");
3836 		goto block_queue_alloc_tag_error;
3837 	}
3838 
3839 	/* Allocate the request queue. */
3840 	dd->queue = blk_mq_init_queue(&dd->tags);
3841 	if (IS_ERR(dd->queue)) {
3842 		dev_err(&dd->pdev->dev,
3843 			"Unable to allocate request queue\n");
3844 		rv = -ENOMEM;
3845 		goto block_queue_alloc_init_error;
3846 	}
3847 
3848 	dd->disk->queue		= dd->queue;
3849 	dd->queue->queuedata	= dd;
3850 
3851 skip_create_disk:
3852 	/* Initialize the protocol layer. */
3853 	wait_for_rebuild = mtip_hw_get_identify(dd);
3854 	if (wait_for_rebuild < 0) {
3855 		dev_err(&dd->pdev->dev,
3856 			"Protocol layer initialization failed\n");
3857 		rv = -EINVAL;
3858 		goto init_hw_cmds_error;
3859 	}
3860 
3861 	/*
3862 	 * if rebuild pending, start the service thread, and delay the block
3863 	 * queue creation and device_add_disk()
3864 	 */
3865 	if (wait_for_rebuild == MTIP_FTL_REBUILD_MAGIC)
3866 		goto start_service_thread;
3867 
3868 	/* Set device limits. */
3869 	set_bit(QUEUE_FLAG_NONROT, &dd->queue->queue_flags);
3870 	clear_bit(QUEUE_FLAG_ADD_RANDOM, &dd->queue->queue_flags);
3871 	blk_queue_max_segments(dd->queue, MTIP_MAX_SG);
3872 	blk_queue_physical_block_size(dd->queue, 4096);
3873 	blk_queue_max_hw_sectors(dd->queue, 0xffff);
3874 	blk_queue_max_segment_size(dd->queue, 0x400000);
3875 	blk_queue_io_min(dd->queue, 4096);
3876 	blk_queue_bounce_limit(dd->queue, dd->pdev->dma_mask);
3877 
3878 	/* Signal trim support */
3879 	if (dd->trim_supp == true) {
3880 		set_bit(QUEUE_FLAG_DISCARD, &dd->queue->queue_flags);
3881 		dd->queue->limits.discard_granularity = 4096;
3882 		blk_queue_max_discard_sectors(dd->queue,
3883 			MTIP_MAX_TRIM_ENTRY_LEN * MTIP_MAX_TRIM_ENTRIES);
3884 	}
3885 
3886 	/* Set the capacity of the device in 512 byte sectors. */
3887 	if (!(mtip_hw_get_capacity(dd, &capacity))) {
3888 		dev_warn(&dd->pdev->dev,
3889 			"Could not read drive capacity\n");
3890 		rv = -EIO;
3891 		goto read_capacity_error;
3892 	}
3893 	set_capacity(dd->disk, capacity);
3894 
3895 	/* Enable the block device and add it to /dev */
3896 	device_add_disk(&dd->pdev->dev, dd->disk);
3897 
3898 	dd->bdev = bdget_disk(dd->disk, 0);
3899 	/*
3900 	 * Now that the disk is active, initialize any sysfs attributes
3901 	 * managed by the protocol layer.
3902 	 */
3903 	kobj = kobject_get(&disk_to_dev(dd->disk)->kobj);
3904 	if (kobj) {
3905 		mtip_hw_sysfs_init(dd, kobj);
3906 		kobject_put(kobj);
3907 	}
3908 
3909 	if (dd->mtip_svc_handler) {
3910 		set_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag);
3911 		return rv; /* service thread created for handling rebuild */
3912 	}
3913 
3914 start_service_thread:
3915 	dd->mtip_svc_handler = kthread_create_on_node(mtip_service_thread,
3916 						dd, dd->numa_node,
3917 						"mtip_svc_thd_%02d", index);
3918 
3919 	if (IS_ERR(dd->mtip_svc_handler)) {
3920 		dev_err(&dd->pdev->dev, "service thread failed to start\n");
3921 		dd->mtip_svc_handler = NULL;
3922 		rv = -EFAULT;
3923 		goto kthread_run_error;
3924 	}
3925 	wake_up_process(dd->mtip_svc_handler);
3926 	if (wait_for_rebuild == MTIP_FTL_REBUILD_MAGIC)
3927 		rv = wait_for_rebuild;
3928 
3929 	return rv;
3930 
3931 kthread_run_error:
3932 	bdput(dd->bdev);
3933 	dd->bdev = NULL;
3934 
3935 	/* Delete our gendisk. This also removes the device from /dev */
3936 	del_gendisk(dd->disk);
3937 
3938 read_capacity_error:
3939 init_hw_cmds_error:
3940 	blk_cleanup_queue(dd->queue);
3941 block_queue_alloc_init_error:
3942 	blk_mq_free_tag_set(&dd->tags);
3943 block_queue_alloc_tag_error:
3944 	mtip_hw_debugfs_exit(dd);
3945 disk_index_error:
3946 	spin_lock(&rssd_index_lock);
3947 	ida_remove(&rssd_index_ida, index);
3948 	spin_unlock(&rssd_index_lock);
3949 
3950 ida_get_error:
3951 	put_disk(dd->disk);
3952 
3953 alloc_disk_error:
3954 	mtip_hw_exit(dd); /* De-initialize the protocol layer. */
3955 
3956 protocol_init_error:
3957 	return rv;
3958 }
3959 
3960 static void mtip_no_dev_cleanup(struct request *rq, void *data, bool reserv)
3961 {
3962 	struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3963 
3964 	cmd->status = -ENODEV;
3965 	blk_mq_complete_request(rq);
3966 }
3967 
3968 /*
3969  * Block layer deinitialization function.
3970  *
3971  * Called by the PCI layer as each P320 device is removed.
3972  *
3973  * @dd Pointer to the driver data structure.
3974  *
3975  * return value
3976  *	0
3977  */
3978 static int mtip_block_remove(struct driver_data *dd)
3979 {
3980 	struct kobject *kobj;
3981 
3982 	mtip_hw_debugfs_exit(dd);
3983 
3984 	if (dd->mtip_svc_handler) {
3985 		set_bit(MTIP_PF_SVC_THD_STOP_BIT, &dd->port->flags);
3986 		wake_up_interruptible(&dd->port->svc_wait);
3987 		kthread_stop(dd->mtip_svc_handler);
3988 	}
3989 
3990 	/* Clean up the sysfs attributes, if created */
3991 	if (test_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag)) {
3992 		kobj = kobject_get(&disk_to_dev(dd->disk)->kobj);
3993 		if (kobj) {
3994 			mtip_hw_sysfs_exit(dd, kobj);
3995 			kobject_put(kobj);
3996 		}
3997 	}
3998 
3999 	if (!dd->sr) {
4000 		/*
4001 		 * Explicitly wait here for IOs to quiesce,
4002 		 * as mtip_standby_drive usually won't wait for IOs.
4003 		 */
4004 		if (!mtip_quiesce_io(dd->port, MTIP_QUIESCE_IO_TIMEOUT_MS))
4005 			mtip_standby_drive(dd);
4006 	}
4007 	else
4008 		dev_info(&dd->pdev->dev, "device %s surprise removal\n",
4009 						dd->disk->disk_name);
4010 
4011 	blk_freeze_queue_start(dd->queue);
4012 	blk_mq_stop_hw_queues(dd->queue);
4013 	blk_mq_tagset_busy_iter(&dd->tags, mtip_no_dev_cleanup, dd);
4014 
4015 	/*
4016 	 * Delete our gendisk structure. This also removes the device
4017 	 * from /dev
4018 	 */
4019 	if (dd->bdev) {
4020 		bdput(dd->bdev);
4021 		dd->bdev = NULL;
4022 	}
4023 	if (dd->disk) {
4024 		if (test_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag))
4025 			del_gendisk(dd->disk);
4026 		if (dd->disk->queue) {
4027 			blk_cleanup_queue(dd->queue);
4028 			blk_mq_free_tag_set(&dd->tags);
4029 			dd->queue = NULL;
4030 		}
4031 		put_disk(dd->disk);
4032 	}
4033 	dd->disk  = NULL;
4034 
4035 	spin_lock(&rssd_index_lock);
4036 	ida_remove(&rssd_index_ida, dd->index);
4037 	spin_unlock(&rssd_index_lock);
4038 
4039 	/* De-initialize the protocol layer. */
4040 	mtip_hw_exit(dd);
4041 
4042 	return 0;
4043 }
4044 
4045 /*
4046  * Function called by the PCI layer when just before the
4047  * machine shuts down.
4048  *
4049  * If a protocol layer shutdown function is present it will be called
4050  * by this function.
4051  *
4052  * @dd Pointer to the driver data structure.
4053  *
4054  * return value
4055  *	0
4056  */
4057 static int mtip_block_shutdown(struct driver_data *dd)
4058 {
4059 	mtip_hw_shutdown(dd);
4060 
4061 	/* Delete our gendisk structure, and cleanup the blk queue. */
4062 	if (dd->disk) {
4063 		dev_info(&dd->pdev->dev,
4064 			"Shutting down %s ...\n", dd->disk->disk_name);
4065 
4066 		if (test_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag))
4067 			del_gendisk(dd->disk);
4068 		if (dd->disk->queue) {
4069 			blk_cleanup_queue(dd->queue);
4070 			blk_mq_free_tag_set(&dd->tags);
4071 		}
4072 		put_disk(dd->disk);
4073 		dd->disk  = NULL;
4074 		dd->queue = NULL;
4075 	}
4076 
4077 	spin_lock(&rssd_index_lock);
4078 	ida_remove(&rssd_index_ida, dd->index);
4079 	spin_unlock(&rssd_index_lock);
4080 	return 0;
4081 }
4082 
4083 static int mtip_block_suspend(struct driver_data *dd)
4084 {
4085 	dev_info(&dd->pdev->dev,
4086 		"Suspending %s ...\n", dd->disk->disk_name);
4087 	mtip_hw_suspend(dd);
4088 	return 0;
4089 }
4090 
4091 static int mtip_block_resume(struct driver_data *dd)
4092 {
4093 	dev_info(&dd->pdev->dev, "Resuming %s ...\n",
4094 		dd->disk->disk_name);
4095 	mtip_hw_resume(dd);
4096 	return 0;
4097 }
4098 
4099 static void drop_cpu(int cpu)
4100 {
4101 	cpu_use[cpu]--;
4102 }
4103 
4104 static int get_least_used_cpu_on_node(int node)
4105 {
4106 	int cpu, least_used_cpu, least_cnt;
4107 	const struct cpumask *node_mask;
4108 
4109 	node_mask = cpumask_of_node(node);
4110 	least_used_cpu = cpumask_first(node_mask);
4111 	least_cnt = cpu_use[least_used_cpu];
4112 	cpu = least_used_cpu;
4113 
4114 	for_each_cpu(cpu, node_mask) {
4115 		if (cpu_use[cpu] < least_cnt) {
4116 			least_used_cpu = cpu;
4117 			least_cnt = cpu_use[cpu];
4118 		}
4119 	}
4120 	cpu_use[least_used_cpu]++;
4121 	return least_used_cpu;
4122 }
4123 
4124 /* Helper for selecting a node in round robin mode */
4125 static inline int mtip_get_next_rr_node(void)
4126 {
4127 	static int next_node = -1;
4128 
4129 	if (next_node == -1) {
4130 		next_node = first_online_node;
4131 		return next_node;
4132 	}
4133 
4134 	next_node = next_online_node(next_node);
4135 	if (next_node == MAX_NUMNODES)
4136 		next_node = first_online_node;
4137 	return next_node;
4138 }
4139 
4140 static DEFINE_HANDLER(0);
4141 static DEFINE_HANDLER(1);
4142 static DEFINE_HANDLER(2);
4143 static DEFINE_HANDLER(3);
4144 static DEFINE_HANDLER(4);
4145 static DEFINE_HANDLER(5);
4146 static DEFINE_HANDLER(6);
4147 static DEFINE_HANDLER(7);
4148 
4149 static void mtip_disable_link_opts(struct driver_data *dd, struct pci_dev *pdev)
4150 {
4151 	int pos;
4152 	unsigned short pcie_dev_ctrl;
4153 
4154 	pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
4155 	if (pos) {
4156 		pci_read_config_word(pdev,
4157 			pos + PCI_EXP_DEVCTL,
4158 			&pcie_dev_ctrl);
4159 		if (pcie_dev_ctrl & (1 << 11) ||
4160 		    pcie_dev_ctrl & (1 << 4)) {
4161 			dev_info(&dd->pdev->dev,
4162 				"Disabling ERO/No-Snoop on bridge device %04x:%04x\n",
4163 					pdev->vendor, pdev->device);
4164 			pcie_dev_ctrl &= ~(PCI_EXP_DEVCTL_NOSNOOP_EN |
4165 						PCI_EXP_DEVCTL_RELAX_EN);
4166 			pci_write_config_word(pdev,
4167 				pos + PCI_EXP_DEVCTL,
4168 				pcie_dev_ctrl);
4169 		}
4170 	}
4171 }
4172 
4173 static void mtip_fix_ero_nosnoop(struct driver_data *dd, struct pci_dev *pdev)
4174 {
4175 	/*
4176 	 * This workaround is specific to AMD/ATI chipset with a PCI upstream
4177 	 * device with device id 0x5aXX
4178 	 */
4179 	if (pdev->bus && pdev->bus->self) {
4180 		if (pdev->bus->self->vendor == PCI_VENDOR_ID_ATI &&
4181 		    ((pdev->bus->self->device & 0xff00) == 0x5a00)) {
4182 			mtip_disable_link_opts(dd, pdev->bus->self);
4183 		} else {
4184 			/* Check further up the topology */
4185 			struct pci_dev *parent_dev = pdev->bus->self;
4186 			if (parent_dev->bus &&
4187 				parent_dev->bus->parent &&
4188 				parent_dev->bus->parent->self &&
4189 				parent_dev->bus->parent->self->vendor ==
4190 					 PCI_VENDOR_ID_ATI &&
4191 				(parent_dev->bus->parent->self->device &
4192 					0xff00) == 0x5a00) {
4193 				mtip_disable_link_opts(dd,
4194 					parent_dev->bus->parent->self);
4195 			}
4196 		}
4197 	}
4198 }
4199 
4200 /*
4201  * Called for each supported PCI device detected.
4202  *
4203  * This function allocates the private data structure, enables the
4204  * PCI device and then calls the block layer initialization function.
4205  *
4206  * return value
4207  *	0 on success else an error code.
4208  */
4209 static int mtip_pci_probe(struct pci_dev *pdev,
4210 			const struct pci_device_id *ent)
4211 {
4212 	int rv = 0;
4213 	struct driver_data *dd = NULL;
4214 	char cpu_list[256];
4215 	const struct cpumask *node_mask;
4216 	int cpu, i = 0, j = 0;
4217 	int my_node = NUMA_NO_NODE;
4218 	unsigned long flags;
4219 
4220 	/* Allocate memory for this devices private data. */
4221 	my_node = pcibus_to_node(pdev->bus);
4222 	if (my_node != NUMA_NO_NODE) {
4223 		if (!node_online(my_node))
4224 			my_node = mtip_get_next_rr_node();
4225 	} else {
4226 		dev_info(&pdev->dev, "Kernel not reporting proximity, choosing a node\n");
4227 		my_node = mtip_get_next_rr_node();
4228 	}
4229 	dev_info(&pdev->dev, "NUMA node %d (closest: %d,%d, probe on %d:%d)\n",
4230 		my_node, pcibus_to_node(pdev->bus), dev_to_node(&pdev->dev),
4231 		cpu_to_node(raw_smp_processor_id()), raw_smp_processor_id());
4232 
4233 	dd = kzalloc_node(sizeof(struct driver_data), GFP_KERNEL, my_node);
4234 	if (dd == NULL) {
4235 		dev_err(&pdev->dev,
4236 			"Unable to allocate memory for driver data\n");
4237 		return -ENOMEM;
4238 	}
4239 
4240 	/* Attach the private data to this PCI device.  */
4241 	pci_set_drvdata(pdev, dd);
4242 
4243 	rv = pcim_enable_device(pdev);
4244 	if (rv < 0) {
4245 		dev_err(&pdev->dev, "Unable to enable device\n");
4246 		goto iomap_err;
4247 	}
4248 
4249 	/* Map BAR5 to memory. */
4250 	rv = pcim_iomap_regions(pdev, 1 << MTIP_ABAR, MTIP_DRV_NAME);
4251 	if (rv < 0) {
4252 		dev_err(&pdev->dev, "Unable to map regions\n");
4253 		goto iomap_err;
4254 	}
4255 
4256 	if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(64))) {
4257 		rv = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64));
4258 
4259 		if (rv) {
4260 			rv = pci_set_consistent_dma_mask(pdev,
4261 						DMA_BIT_MASK(32));
4262 			if (rv) {
4263 				dev_warn(&pdev->dev,
4264 					"64-bit DMA enable failed\n");
4265 				goto setmask_err;
4266 			}
4267 		}
4268 	}
4269 
4270 	/* Copy the info we may need later into the private data structure. */
4271 	dd->major	= mtip_major;
4272 	dd->instance	= instance;
4273 	dd->pdev	= pdev;
4274 	dd->numa_node	= my_node;
4275 
4276 	INIT_LIST_HEAD(&dd->online_list);
4277 	INIT_LIST_HEAD(&dd->remove_list);
4278 
4279 	memset(dd->workq_name, 0, 32);
4280 	snprintf(dd->workq_name, 31, "mtipq%d", dd->instance);
4281 
4282 	dd->isr_workq = create_workqueue(dd->workq_name);
4283 	if (!dd->isr_workq) {
4284 		dev_warn(&pdev->dev, "Can't create wq %d\n", dd->instance);
4285 		rv = -ENOMEM;
4286 		goto block_initialize_err;
4287 	}
4288 
4289 	memset(cpu_list, 0, sizeof(cpu_list));
4290 
4291 	node_mask = cpumask_of_node(dd->numa_node);
4292 	if (!cpumask_empty(node_mask)) {
4293 		for_each_cpu(cpu, node_mask)
4294 		{
4295 			snprintf(&cpu_list[j], 256 - j, "%d ", cpu);
4296 			j = strlen(cpu_list);
4297 		}
4298 
4299 		dev_info(&pdev->dev, "Node %d on package %d has %d cpu(s): %s\n",
4300 			dd->numa_node,
4301 			topology_physical_package_id(cpumask_first(node_mask)),
4302 			nr_cpus_node(dd->numa_node),
4303 			cpu_list);
4304 	} else
4305 		dev_dbg(&pdev->dev, "mtip32xx: node_mask empty\n");
4306 
4307 	dd->isr_binding = get_least_used_cpu_on_node(dd->numa_node);
4308 	dev_info(&pdev->dev, "Initial IRQ binding node:cpu %d:%d\n",
4309 		cpu_to_node(dd->isr_binding), dd->isr_binding);
4310 
4311 	/* first worker context always runs in ISR */
4312 	dd->work[0].cpu_binding = dd->isr_binding;
4313 	dd->work[1].cpu_binding = get_least_used_cpu_on_node(dd->numa_node);
4314 	dd->work[2].cpu_binding = get_least_used_cpu_on_node(dd->numa_node);
4315 	dd->work[3].cpu_binding = dd->work[0].cpu_binding;
4316 	dd->work[4].cpu_binding = dd->work[1].cpu_binding;
4317 	dd->work[5].cpu_binding = dd->work[2].cpu_binding;
4318 	dd->work[6].cpu_binding = dd->work[2].cpu_binding;
4319 	dd->work[7].cpu_binding = dd->work[1].cpu_binding;
4320 
4321 	/* Log the bindings */
4322 	for_each_present_cpu(cpu) {
4323 		memset(cpu_list, 0, sizeof(cpu_list));
4324 		for (i = 0, j = 0; i < MTIP_MAX_SLOT_GROUPS; i++) {
4325 			if (dd->work[i].cpu_binding == cpu) {
4326 				snprintf(&cpu_list[j], 256 - j, "%d ", i);
4327 				j = strlen(cpu_list);
4328 			}
4329 		}
4330 		if (j)
4331 			dev_info(&pdev->dev, "CPU %d: WQs %s\n", cpu, cpu_list);
4332 	}
4333 
4334 	INIT_WORK(&dd->work[0].work, mtip_workq_sdbf0);
4335 	INIT_WORK(&dd->work[1].work, mtip_workq_sdbf1);
4336 	INIT_WORK(&dd->work[2].work, mtip_workq_sdbf2);
4337 	INIT_WORK(&dd->work[3].work, mtip_workq_sdbf3);
4338 	INIT_WORK(&dd->work[4].work, mtip_workq_sdbf4);
4339 	INIT_WORK(&dd->work[5].work, mtip_workq_sdbf5);
4340 	INIT_WORK(&dd->work[6].work, mtip_workq_sdbf6);
4341 	INIT_WORK(&dd->work[7].work, mtip_workq_sdbf7);
4342 
4343 	pci_set_master(pdev);
4344 	rv = pci_enable_msi(pdev);
4345 	if (rv) {
4346 		dev_warn(&pdev->dev,
4347 			"Unable to enable MSI interrupt.\n");
4348 		goto msi_initialize_err;
4349 	}
4350 
4351 	mtip_fix_ero_nosnoop(dd, pdev);
4352 
4353 	/* Initialize the block layer. */
4354 	rv = mtip_block_initialize(dd);
4355 	if (rv < 0) {
4356 		dev_err(&pdev->dev,
4357 			"Unable to initialize block layer\n");
4358 		goto block_initialize_err;
4359 	}
4360 
4361 	/*
4362 	 * Increment the instance count so that each device has a unique
4363 	 * instance number.
4364 	 */
4365 	instance++;
4366 	if (rv != MTIP_FTL_REBUILD_MAGIC)
4367 		set_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag);
4368 	else
4369 		rv = 0; /* device in rebuild state, return 0 from probe */
4370 
4371 	/* Add to online list even if in ftl rebuild */
4372 	spin_lock_irqsave(&dev_lock, flags);
4373 	list_add(&dd->online_list, &online_list);
4374 	spin_unlock_irqrestore(&dev_lock, flags);
4375 
4376 	goto done;
4377 
4378 block_initialize_err:
4379 	pci_disable_msi(pdev);
4380 
4381 msi_initialize_err:
4382 	if (dd->isr_workq) {
4383 		flush_workqueue(dd->isr_workq);
4384 		destroy_workqueue(dd->isr_workq);
4385 		drop_cpu(dd->work[0].cpu_binding);
4386 		drop_cpu(dd->work[1].cpu_binding);
4387 		drop_cpu(dd->work[2].cpu_binding);
4388 	}
4389 setmask_err:
4390 	pcim_iounmap_regions(pdev, 1 << MTIP_ABAR);
4391 
4392 iomap_err:
4393 	kfree(dd);
4394 	pci_set_drvdata(pdev, NULL);
4395 	return rv;
4396 done:
4397 	return rv;
4398 }
4399 
4400 /*
4401  * Called for each probed device when the device is removed or the
4402  * driver is unloaded.
4403  *
4404  * return value
4405  *	None
4406  */
4407 static void mtip_pci_remove(struct pci_dev *pdev)
4408 {
4409 	struct driver_data *dd = pci_get_drvdata(pdev);
4410 	unsigned long flags, to;
4411 
4412 	set_bit(MTIP_DDF_REMOVAL_BIT, &dd->dd_flag);
4413 
4414 	spin_lock_irqsave(&dev_lock, flags);
4415 	list_del_init(&dd->online_list);
4416 	list_add(&dd->remove_list, &removing_list);
4417 	spin_unlock_irqrestore(&dev_lock, flags);
4418 
4419 	mtip_check_surprise_removal(pdev);
4420 	synchronize_irq(dd->pdev->irq);
4421 
4422 	/* Spin until workers are done */
4423 	to = jiffies + msecs_to_jiffies(4000);
4424 	do {
4425 		msleep(20);
4426 	} while (atomic_read(&dd->irq_workers_active) != 0 &&
4427 		time_before(jiffies, to));
4428 
4429 	if (!dd->sr)
4430 		fsync_bdev(dd->bdev);
4431 
4432 	if (atomic_read(&dd->irq_workers_active) != 0) {
4433 		dev_warn(&dd->pdev->dev,
4434 			"Completion workers still active!\n");
4435 	}
4436 
4437 	blk_set_queue_dying(dd->queue);
4438 	set_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag);
4439 
4440 	/* Clean up the block layer. */
4441 	mtip_block_remove(dd);
4442 
4443 	if (dd->isr_workq) {
4444 		flush_workqueue(dd->isr_workq);
4445 		destroy_workqueue(dd->isr_workq);
4446 		drop_cpu(dd->work[0].cpu_binding);
4447 		drop_cpu(dd->work[1].cpu_binding);
4448 		drop_cpu(dd->work[2].cpu_binding);
4449 	}
4450 
4451 	pci_disable_msi(pdev);
4452 
4453 	spin_lock_irqsave(&dev_lock, flags);
4454 	list_del_init(&dd->remove_list);
4455 	spin_unlock_irqrestore(&dev_lock, flags);
4456 
4457 	kfree(dd);
4458 
4459 	pcim_iounmap_regions(pdev, 1 << MTIP_ABAR);
4460 	pci_set_drvdata(pdev, NULL);
4461 }
4462 
4463 /*
4464  * Called for each probed device when the device is suspended.
4465  *
4466  * return value
4467  *	0  Success
4468  *	<0 Error
4469  */
4470 static int mtip_pci_suspend(struct pci_dev *pdev, pm_message_t mesg)
4471 {
4472 	int rv = 0;
4473 	struct driver_data *dd = pci_get_drvdata(pdev);
4474 
4475 	if (!dd) {
4476 		dev_err(&pdev->dev,
4477 			"Driver private datastructure is NULL\n");
4478 		return -EFAULT;
4479 	}
4480 
4481 	set_bit(MTIP_DDF_RESUME_BIT, &dd->dd_flag);
4482 
4483 	/* Disable ports & interrupts then send standby immediate */
4484 	rv = mtip_block_suspend(dd);
4485 	if (rv < 0) {
4486 		dev_err(&pdev->dev,
4487 			"Failed to suspend controller\n");
4488 		return rv;
4489 	}
4490 
4491 	/*
4492 	 * Save the pci config space to pdev structure &
4493 	 * disable the device
4494 	 */
4495 	pci_save_state(pdev);
4496 	pci_disable_device(pdev);
4497 
4498 	/* Move to Low power state*/
4499 	pci_set_power_state(pdev, PCI_D3hot);
4500 
4501 	return rv;
4502 }
4503 
4504 /*
4505  * Called for each probed device when the device is resumed.
4506  *
4507  * return value
4508  *      0  Success
4509  *      <0 Error
4510  */
4511 static int mtip_pci_resume(struct pci_dev *pdev)
4512 {
4513 	int rv = 0;
4514 	struct driver_data *dd;
4515 
4516 	dd = pci_get_drvdata(pdev);
4517 	if (!dd) {
4518 		dev_err(&pdev->dev,
4519 			"Driver private datastructure is NULL\n");
4520 		return -EFAULT;
4521 	}
4522 
4523 	/* Move the device to active State */
4524 	pci_set_power_state(pdev, PCI_D0);
4525 
4526 	/* Restore PCI configuration space */
4527 	pci_restore_state(pdev);
4528 
4529 	/* Enable the PCI device*/
4530 	rv = pcim_enable_device(pdev);
4531 	if (rv < 0) {
4532 		dev_err(&pdev->dev,
4533 			"Failed to enable card during resume\n");
4534 		goto err;
4535 	}
4536 	pci_set_master(pdev);
4537 
4538 	/*
4539 	 * Calls hbaReset, initPort, & startPort function
4540 	 * then enables interrupts
4541 	 */
4542 	rv = mtip_block_resume(dd);
4543 	if (rv < 0)
4544 		dev_err(&pdev->dev, "Unable to resume\n");
4545 
4546 err:
4547 	clear_bit(MTIP_DDF_RESUME_BIT, &dd->dd_flag);
4548 
4549 	return rv;
4550 }
4551 
4552 /*
4553  * Shutdown routine
4554  *
4555  * return value
4556  *      None
4557  */
4558 static void mtip_pci_shutdown(struct pci_dev *pdev)
4559 {
4560 	struct driver_data *dd = pci_get_drvdata(pdev);
4561 	if (dd)
4562 		mtip_block_shutdown(dd);
4563 }
4564 
4565 /* Table of device ids supported by this driver. */
4566 static const struct pci_device_id mtip_pci_tbl[] = {
4567 	{ PCI_DEVICE(PCI_VENDOR_ID_MICRON, P320H_DEVICE_ID) },
4568 	{ PCI_DEVICE(PCI_VENDOR_ID_MICRON, P320M_DEVICE_ID) },
4569 	{ PCI_DEVICE(PCI_VENDOR_ID_MICRON, P320S_DEVICE_ID) },
4570 	{ PCI_DEVICE(PCI_VENDOR_ID_MICRON, P325M_DEVICE_ID) },
4571 	{ PCI_DEVICE(PCI_VENDOR_ID_MICRON, P420H_DEVICE_ID) },
4572 	{ PCI_DEVICE(PCI_VENDOR_ID_MICRON, P420M_DEVICE_ID) },
4573 	{ PCI_DEVICE(PCI_VENDOR_ID_MICRON, P425M_DEVICE_ID) },
4574 	{ 0 }
4575 };
4576 
4577 /* Structure that describes the PCI driver functions. */
4578 static struct pci_driver mtip_pci_driver = {
4579 	.name			= MTIP_DRV_NAME,
4580 	.id_table		= mtip_pci_tbl,
4581 	.probe			= mtip_pci_probe,
4582 	.remove			= mtip_pci_remove,
4583 	.suspend		= mtip_pci_suspend,
4584 	.resume			= mtip_pci_resume,
4585 	.shutdown		= mtip_pci_shutdown,
4586 };
4587 
4588 MODULE_DEVICE_TABLE(pci, mtip_pci_tbl);
4589 
4590 /*
4591  * Module initialization function.
4592  *
4593  * Called once when the module is loaded. This function allocates a major
4594  * block device number to the Cyclone devices and registers the PCI layer
4595  * of the driver.
4596  *
4597  * Return value
4598  *      0 on success else error code.
4599  */
4600 static int __init mtip_init(void)
4601 {
4602 	int error;
4603 
4604 	pr_info(MTIP_DRV_NAME " Version " MTIP_DRV_VERSION "\n");
4605 
4606 	spin_lock_init(&dev_lock);
4607 
4608 	INIT_LIST_HEAD(&online_list);
4609 	INIT_LIST_HEAD(&removing_list);
4610 
4611 	/* Allocate a major block device number to use with this driver. */
4612 	error = register_blkdev(0, MTIP_DRV_NAME);
4613 	if (error <= 0) {
4614 		pr_err("Unable to register block device (%d)\n",
4615 		error);
4616 		return -EBUSY;
4617 	}
4618 	mtip_major = error;
4619 
4620 	dfs_parent = debugfs_create_dir("rssd", NULL);
4621 	if (IS_ERR_OR_NULL(dfs_parent)) {
4622 		pr_warn("Error creating debugfs parent\n");
4623 		dfs_parent = NULL;
4624 	}
4625 	if (dfs_parent) {
4626 		dfs_device_status = debugfs_create_file("device_status",
4627 					S_IRUGO, dfs_parent, NULL,
4628 					&mtip_device_status_fops);
4629 		if (IS_ERR_OR_NULL(dfs_device_status)) {
4630 			pr_err("Error creating device_status node\n");
4631 			dfs_device_status = NULL;
4632 		}
4633 	}
4634 
4635 	/* Register our PCI operations. */
4636 	error = pci_register_driver(&mtip_pci_driver);
4637 	if (error) {
4638 		debugfs_remove(dfs_parent);
4639 		unregister_blkdev(mtip_major, MTIP_DRV_NAME);
4640 	}
4641 
4642 	return error;
4643 }
4644 
4645 /*
4646  * Module de-initialization function.
4647  *
4648  * Called once when the module is unloaded. This function deallocates
4649  * the major block device number allocated by mtip_init() and
4650  * unregisters the PCI layer of the driver.
4651  *
4652  * Return value
4653  *      none
4654  */
4655 static void __exit mtip_exit(void)
4656 {
4657 	/* Release the allocated major block device number. */
4658 	unregister_blkdev(mtip_major, MTIP_DRV_NAME);
4659 
4660 	/* Unregister the PCI driver. */
4661 	pci_unregister_driver(&mtip_pci_driver);
4662 
4663 	debugfs_remove_recursive(dfs_parent);
4664 }
4665 
4666 MODULE_AUTHOR("Micron Technology, Inc");
4667 MODULE_DESCRIPTION("Micron RealSSD PCIe Block Driver");
4668 MODULE_LICENSE("GPL");
4669 MODULE_VERSION(MTIP_DRV_VERSION);
4670 
4671 module_init(mtip_init);
4672 module_exit(mtip_exit);
4673