2 * Driver for the Micron P320 SSD
3 * Copyright (C) 2011 Micron Technology, Inc.
5 * Portions of this code were derived from works subjected to the
7 * Copyright (C) 2009 Integrated Device Technology, Inc.
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.
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.
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>
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>
45 #define HW_CMD_SLOT_SZ (MTIP_MAX_COMMAND_SLOTS * 32)
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
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
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)
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
80 #define PORT_COMMAND_ISSUE 0x38
81 #define PORT_SDBV 0x7C
83 #define PORT_OFFSET 0x100
84 #define PORT_MEM_SIZE 0x80
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 | \
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)
100 /* product numbers */
101 #define MTIP_PRODUCT_UNKNOWN 0x00
102 #define MTIP_PRODUCT_ASICFPGA 0x11
104 /* Device instance number, incremented each time a device is probed. */
107 static struct list_head online_list;
108 static struct list_head removing_list;
109 static spinlock_t dev_lock;
112 * Global variable used to hold the major block device number
113 * allocated in mtip_init().
115 static int mtip_major;
116 static struct dentry *dfs_parent;
117 static struct dentry *dfs_device_status;
119 static u32 cpu_use[NR_CPUS];
121 static DEFINE_SPINLOCK(rssd_index_lock);
122 static DEFINE_IDA(rssd_index_ida);
124 static int mtip_block_initialize(struct driver_data *dd);
127 struct mtip_compat_ide_task_request_s {
130 ide_reg_valid_t out_flags;
131 ide_reg_valid_t in_flags;
134 compat_ulong_t out_size;
135 compat_ulong_t in_size;
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
144 * @pdev Pointer to the pci_dev structure.
147 * true if device removed, else false
149 static bool mtip_check_surprise_removal(struct pci_dev *pdev)
152 struct driver_data *dd = pci_get_drvdata(pdev);
157 /* Read the vendorID from the configuration space */
158 pci_read_config_word(pdev, 0x00, &vendor_id);
159 if (vendor_id == 0xFFFF) {
162 blk_queue_flag_set(QUEUE_FLAG_DEAD, dd->queue);
164 dev_warn(&dd->pdev->dev,
165 "%s: dd->queue is NULL\n", __func__);
166 return true; /* device removed */
169 return false; /* device present */
172 /* we have to use runtime tag to setup command header */
173 static void mtip_init_cmd_header(struct request *rq)
175 struct driver_data *dd = rq->q->queuedata;
176 struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
178 /* Point the command headers at the command tables. */
179 cmd->command_header = dd->port->command_list +
180 (sizeof(struct mtip_cmd_hdr) * rq->tag);
181 cmd->command_header_dma = dd->port->command_list_dma +
182 (sizeof(struct mtip_cmd_hdr) * rq->tag);
184 if (test_bit(MTIP_PF_HOST_CAP_64, &dd->port->flags))
185 cmd->command_header->ctbau = __force_bit2int cpu_to_le32((cmd->command_dma >> 16) >> 16);
187 cmd->command_header->ctba = __force_bit2int cpu_to_le32(cmd->command_dma & 0xFFFFFFFF);
190 static struct mtip_cmd *mtip_get_int_command(struct driver_data *dd)
194 if (mtip_check_surprise_removal(dd->pdev))
197 rq = blk_mq_alloc_request(dd->queue, REQ_OP_DRV_IN, BLK_MQ_REQ_RESERVED);
201 /* Internal cmd isn't submitted via .queue_rq */
202 mtip_init_cmd_header(rq);
204 return blk_mq_rq_to_pdu(rq);
207 static struct mtip_cmd *mtip_cmd_from_tag(struct driver_data *dd,
210 struct blk_mq_hw_ctx *hctx = dd->queue->queue_hw_ctx[0];
212 return blk_mq_rq_to_pdu(blk_mq_tag_to_rq(hctx->tags, tag));
216 * Reset the HBA (without sleeping)
218 * @dd Pointer to the driver data structure.
221 * 0 The reset was successful.
222 * -1 The HBA Reset bit did not clear.
224 static int mtip_hba_reset(struct driver_data *dd)
226 unsigned long timeout;
228 /* Set the reset bit */
229 writel(HOST_RESET, dd->mmio + HOST_CTL);
232 readl(dd->mmio + HOST_CTL);
235 * Spin for up to 10 seconds waiting for reset acknowledgement. Spec
236 * is 1 sec but in LUN failure conditions, up to 10 secs are required
238 timeout = jiffies + msecs_to_jiffies(10000);
241 if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag))
244 } while ((readl(dd->mmio + HOST_CTL) & HOST_RESET)
245 && time_before(jiffies, timeout));
247 if (readl(dd->mmio + HOST_CTL) & HOST_RESET)
254 * Issue a command to the hardware.
256 * Set the appropriate bit in the s_active and Command Issue hardware
257 * registers, causing hardware command processing to begin.
259 * @port Pointer to the port structure.
260 * @tag The tag of the command to be issued.
265 static inline void mtip_issue_ncq_command(struct mtip_port *port, int tag)
267 int group = tag >> 5;
269 /* guard SACT and CI registers */
270 spin_lock(&port->cmd_issue_lock[group]);
271 writel((1 << MTIP_TAG_BIT(tag)),
272 port->s_active[MTIP_TAG_INDEX(tag)]);
273 writel((1 << MTIP_TAG_BIT(tag)),
274 port->cmd_issue[MTIP_TAG_INDEX(tag)]);
275 spin_unlock(&port->cmd_issue_lock[group]);
279 * Enable/disable the reception of FIS
281 * @port Pointer to the port data structure
282 * @enable 1 to enable, 0 to disable
285 * Previous state: 1 enabled, 0 disabled
287 static int mtip_enable_fis(struct mtip_port *port, int enable)
291 /* enable FIS reception */
292 tmp = readl(port->mmio + PORT_CMD);
294 writel(tmp | PORT_CMD_FIS_RX, port->mmio + PORT_CMD);
296 writel(tmp & ~PORT_CMD_FIS_RX, port->mmio + PORT_CMD);
299 readl(port->mmio + PORT_CMD);
301 return (((tmp & PORT_CMD_FIS_RX) == PORT_CMD_FIS_RX));
305 * Enable/disable the DMA engine
307 * @port Pointer to the port data structure
308 * @enable 1 to enable, 0 to disable
311 * Previous state: 1 enabled, 0 disabled.
313 static int mtip_enable_engine(struct mtip_port *port, int enable)
317 /* enable FIS reception */
318 tmp = readl(port->mmio + PORT_CMD);
320 writel(tmp | PORT_CMD_START, port->mmio + PORT_CMD);
322 writel(tmp & ~PORT_CMD_START, port->mmio + PORT_CMD);
324 readl(port->mmio + PORT_CMD);
325 return (((tmp & PORT_CMD_START) == PORT_CMD_START));
329 * Enables the port DMA engine and FIS reception.
334 static inline void mtip_start_port(struct mtip_port *port)
336 /* Enable FIS reception */
337 mtip_enable_fis(port, 1);
339 /* Enable the DMA engine */
340 mtip_enable_engine(port, 1);
344 * Deinitialize a port by disabling port interrupts, the DMA engine,
347 * @port Pointer to the port structure
352 static inline void mtip_deinit_port(struct mtip_port *port)
354 /* Disable interrupts on this port */
355 writel(0, port->mmio + PORT_IRQ_MASK);
357 /* Disable the DMA engine */
358 mtip_enable_engine(port, 0);
360 /* Disable FIS reception */
361 mtip_enable_fis(port, 0);
367 * This function deinitializes the port by calling mtip_deinit_port() and
368 * then initializes it by setting the command header and RX FIS addresses,
369 * clearing the SError register and any pending port interrupts before
370 * re-enabling the default set of port interrupts.
372 * @port Pointer to the port structure.
377 static void mtip_init_port(struct mtip_port *port)
380 mtip_deinit_port(port);
382 /* Program the command list base and FIS base addresses */
383 if (readl(port->dd->mmio + HOST_CAP) & HOST_CAP_64) {
384 writel((port->command_list_dma >> 16) >> 16,
385 port->mmio + PORT_LST_ADDR_HI);
386 writel((port->rxfis_dma >> 16) >> 16,
387 port->mmio + PORT_FIS_ADDR_HI);
388 set_bit(MTIP_PF_HOST_CAP_64, &port->flags);
391 writel(port->command_list_dma & 0xFFFFFFFF,
392 port->mmio + PORT_LST_ADDR);
393 writel(port->rxfis_dma & 0xFFFFFFFF, port->mmio + PORT_FIS_ADDR);
396 writel(readl(port->mmio + PORT_SCR_ERR), port->mmio + PORT_SCR_ERR);
398 /* reset the completed registers.*/
399 for (i = 0; i < port->dd->slot_groups; i++)
400 writel(0xFFFFFFFF, port->completed[i]);
402 /* Clear any pending interrupts for this port */
403 writel(readl(port->mmio + PORT_IRQ_STAT), port->mmio + PORT_IRQ_STAT);
405 /* Clear any pending interrupts on the HBA. */
406 writel(readl(port->dd->mmio + HOST_IRQ_STAT),
407 port->dd->mmio + HOST_IRQ_STAT);
409 /* Enable port interrupts */
410 writel(DEF_PORT_IRQ, port->mmio + PORT_IRQ_MASK);
416 * @port Pointer to the port data structure.
421 static void mtip_restart_port(struct mtip_port *port)
423 unsigned long timeout;
425 /* Disable the DMA engine */
426 mtip_enable_engine(port, 0);
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))
434 if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &port->dd->dd_flag))
438 * Chip quirk: escalate to hba reset if
439 * PxCMD.CR not clear after 500 ms
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");
445 if (mtip_hba_reset(port->dd))
446 dev_err(&port->dd->pdev->dev,
447 "HBA reset escalation failed.\n");
449 /* 30 ms delay before com reset to quiesce chip */
453 dev_warn(&port->dd->pdev->dev, "Issuing COM reset\n");
456 writel(readl(port->mmio + PORT_SCR_CTL) |
457 1, port->mmio + PORT_SCR_CTL);
458 readl(port->mmio + PORT_SCR_CTL);
460 /* Wait 1 ms to quiesce chip function */
461 timeout = jiffies + msecs_to_jiffies(1);
462 while (time_before(jiffies, timeout))
465 if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &port->dd->dd_flag))
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);
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))
479 if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &port->dd->dd_flag))
482 if ((readl(port->mmio + PORT_SCR_STAT) & 0x01) == 0)
483 dev_warn(&port->dd->pdev->dev,
484 "COM reset failed\n");
486 mtip_init_port(port);
487 mtip_start_port(port);
491 static int mtip_device_reset(struct driver_data *dd)
495 if (mtip_check_surprise_removal(dd->pdev))
498 if (mtip_hba_reset(dd) < 0)
502 mtip_init_port(dd->port);
503 mtip_start_port(dd->port);
505 /* Enable interrupts on the HBA. */
506 writel(readl(dd->mmio + HOST_CTL) | HOST_IRQ_EN,
507 dd->mmio + HOST_CTL);
512 * Helper function for tag logging
514 static void print_tags(struct driver_data *dd,
516 unsigned long *tagbits,
519 unsigned char tagmap[128];
520 int group, tagmap_len = 0;
522 memset(tagmap, 0, sizeof(tagmap));
523 for (group = SLOTBITS_IN_LONGS; group > 0; group--)
524 tagmap_len += sprintf(tagmap + tagmap_len, "%016lX ",
526 dev_warn(&dd->pdev->dev,
527 "%d command(s) %s: tagmap [%s]", cnt, msg, tagmap);
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);
535 static void mtip_complete_command(struct mtip_cmd *cmd, blk_status_t status)
537 struct request *req = blk_mq_rq_from_pdu(cmd);
539 cmd->status = status;
540 blk_mq_complete_request(req);
546 * @dd Pointer to the DRIVER_DATA structure.
551 static void mtip_handle_tfe(struct driver_data *dd)
553 int group, tag, bit, reissue, rv;
554 struct mtip_port *port;
555 struct mtip_cmd *cmd;
557 struct host_to_dev_fis *fis;
558 unsigned long tagaccum[SLOTBITS_IN_LONGS];
559 unsigned int cmd_cnt = 0;
561 char *fail_reason = NULL;
562 int fail_all_ncq_write = 0, fail_all_ncq_cmds = 0;
564 dev_warn(&dd->pdev->dev, "Taskfile error\n");
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, BLK_STS_IOERR);
575 /* clear the tag accumulator */
576 memset(tagaccum, 0, SLOTBITS_IN_LONGS * sizeof(long));
578 /* Loop through all the groups */
579 for (group = 0; group < dd->slot_groups; group++) {
580 completed = readl(port->completed[group]);
582 dev_warn(&dd->pdev->dev, "g=%u, comp=%x\n", group, completed);
584 /* clear completed status register in the hardware.*/
585 writel(completed, port->completed[group]);
587 /* Process successfully completed commands */
588 for (bit = 0; bit < 32 && completed; bit++) {
589 if (!(completed & (1<<bit)))
591 tag = (group << 5) + bit;
593 /* Skip the internal command slot */
594 if (tag == MTIP_TAG_INTERNAL)
597 cmd = mtip_cmd_from_tag(dd, tag);
598 mtip_complete_command(cmd, 0);
599 set_bit(tag, tagaccum);
604 print_tags(dd, "completed (TFE)", tagaccum, cmd_cnt);
606 /* Restart the port */
608 mtip_restart_port(port);
610 /* Trying to determine the cause of the error */
611 rv = mtip_read_log_page(dd->port, ATA_LOG_SATA_NCQ,
613 dd->port->log_buf_dma, 1);
615 dev_warn(&dd->pdev->dev,
616 "Error in READ LOG EXT (10h) command\n");
617 /* non-critical error, don't fail the load */
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";
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";
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";
643 /* clear the tag accumulator */
644 memset(tagaccum, 0, SLOTBITS_IN_LONGS * sizeof(long));
646 /* Loop through all the groups */
647 for (group = 0; group < dd->slot_groups; group++) {
648 for (bit = 0; bit < 32; bit++) {
650 tag = (group << 5) + bit;
651 cmd = mtip_cmd_from_tag(dd, tag);
653 fis = (struct host_to_dev_fis *)cmd->command;
655 /* Should re-issue? */
656 if (tag == MTIP_TAG_INTERNAL ||
657 fis->command == ATA_CMD_SET_FEATURES)
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 ?
668 fail_reason != NULL ?
669 fail_reason : "unknown");
670 mtip_complete_command(cmd, BLK_STS_MEDIUM);
676 * First check if this command has
677 * exceeded its retries.
679 if (reissue && (cmd->retries-- > 0)) {
681 set_bit(tag, tagaccum);
683 /* Re-issue the command. */
684 mtip_issue_ncq_command(port, tag);
689 /* Retire a command that will not be reissued */
690 dev_warn(&port->dd->pdev->dev,
691 "retiring tag %d\n", tag);
693 mtip_complete_command(cmd, BLK_STS_IOERR);
696 print_tags(dd, "reissued (TFE)", tagaccum, cmd_cnt);
700 * Handle a set device bits interrupt
702 static inline void mtip_workq_sdbfx(struct mtip_port *port, int group,
705 struct driver_data *dd = port->dd;
707 struct mtip_cmd *command;
710 WARN_ON_ONCE(!completed);
713 /* clear completed status register in the hardware.*/
714 writel(completed, port->completed[group]);
716 /* Process completed commands. */
717 for (bit = 0; (bit < 32) && completed; bit++) {
718 if (completed & 0x01) {
719 tag = (group << 5) | bit;
721 /* skip internal command slot. */
722 if (unlikely(tag == MTIP_TAG_INTERNAL))
725 command = mtip_cmd_from_tag(dd, tag);
726 mtip_complete_command(command, 0);
731 /* If last, re-enable interrupts */
732 if (atomic_dec_return(&dd->irq_workers_active) == 0)
733 writel(0xffffffff, dd->mmio + HOST_IRQ_STAT);
737 * Process legacy pio and d2h interrupts
739 static inline void mtip_process_legacy(struct driver_data *dd, u32 port_stat)
741 struct mtip_port *port = dd->port;
742 struct mtip_cmd *cmd = mtip_cmd_from_tag(dd, MTIP_TAG_INTERNAL);
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]);
748 if (!(status & (1 << MTIP_TAG_BIT(MTIP_TAG_INTERNAL))))
749 mtip_complete_command(cmd, 0);
754 * Demux and handle errors
756 static inline void mtip_process_errors(struct driver_data *dd, u32 port_stat)
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);
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);
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))
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);
783 static inline irqreturn_t mtip_handle_irq(struct driver_data *data)
785 struct driver_data *dd = (struct driver_data *) data;
786 struct mtip_port *port = dd->port;
787 u32 hba_stat, port_stat;
789 int do_irq_enable = 1, i, workers;
790 struct mtip_work *twork;
792 hba_stat = readl(dd->mmio + HOST_IRQ_STAT);
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);
802 writel(port_stat, port->mmio + PORT_IRQ_STAT);
804 /* Demux port status */
805 if (likely(port_stat & PORT_IRQ_SDB_FIS)) {
807 WARN_ON_ONCE(atomic_read(&dd->irq_workers_active) != 0);
809 /* Start at 1: group zero is always local? */
810 for (i = 0, workers = 0; i < MTIP_MAX_SLOT_GROUPS;
812 twork = &dd->work[i];
813 twork->completed = readl(port->completed[i]);
814 if (twork->completed)
818 atomic_set(&dd->irq_workers_active, workers);
820 for (i = 1; i < MTIP_MAX_SLOT_GROUPS; i++) {
821 twork = &dd->work[i];
822 if (twork->completed)
829 if (likely(dd->work[0].completed))
830 mtip_workq_sdbfx(port, 0,
831 dd->work[0].completed);
835 * Chip quirk: SDB interrupt but nothing
842 if (unlikely(port_stat & PORT_IRQ_ERR)) {
843 if (unlikely(mtip_check_surprise_removal(dd->pdev))) {
844 /* don't proceed further */
847 if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
851 mtip_process_errors(dd, port_stat & PORT_IRQ_ERR);
854 if (unlikely(port_stat & PORT_IRQ_LEGACY))
855 mtip_process_legacy(dd, port_stat & PORT_IRQ_LEGACY);
858 /* acknowledge interrupt */
859 if (unlikely(do_irq_enable))
860 writel(hba_stat, dd->mmio + HOST_IRQ_STAT);
866 * HBA interrupt subroutine.
869 * @instance Pointer to the driver data structure.
872 * IRQ_HANDLED A HBA interrupt was pending and handled.
873 * IRQ_NONE This interrupt was not for the HBA.
875 static irqreturn_t mtip_irq_handler(int irq, void *instance)
877 struct driver_data *dd = instance;
879 return mtip_handle_irq(dd);
882 static void mtip_issue_non_ncq_command(struct mtip_port *port, int tag)
884 writel(1 << MTIP_TAG_BIT(tag), port->cmd_issue[MTIP_TAG_INDEX(tag)]);
887 static bool mtip_pause_ncq(struct mtip_port *port,
888 struct host_to_dev_fis *fis)
890 unsigned long task_file_data;
892 task_file_data = readl(port->mmio+PORT_TFDATA);
893 if ((task_file_data & 1))
896 if (fis->command == ATA_CMD_SEC_ERASE_PREP) {
897 port->ic_pause_timer = jiffies;
899 } else if ((fis->command == ATA_CMD_DOWNLOAD_MICRO) &&
900 (fis->features == 0x03)) {
901 set_bit(MTIP_PF_DM_ACTIVE_BIT, &port->flags);
902 port->ic_pause_timer = jiffies;
904 } else if ((fis->command == ATA_CMD_SEC_ERASE_UNIT) ||
905 ((fis->command == 0xFC) &&
906 (fis->features == 0x27 || fis->features == 0x72 ||
907 fis->features == 0x62 || fis->features == 0x26))) {
908 clear_bit(MTIP_DDF_SEC_LOCK_BIT, &port->dd->dd_flag);
909 clear_bit(MTIP_DDF_REBUILD_FAILED_BIT, &port->dd->dd_flag);
910 /* Com reset after secure erase or lowlevel format */
911 mtip_restart_port(port);
912 clear_bit(MTIP_PF_SE_ACTIVE_BIT, &port->flags);
919 static bool mtip_commands_active(struct mtip_port *port)
925 * Ignore s_active bit 0 of array element 0.
926 * This bit will always be set
928 active = readl(port->s_active[0]) & 0xFFFFFFFE;
929 for (n = 1; n < port->dd->slot_groups; n++)
930 active |= readl(port->s_active[n]);
936 * Wait for port to quiesce
938 * @port Pointer to port data structure
939 * @timeout Max duration to wait (ms)
943 * -EBUSY Commands still active
945 static int mtip_quiesce_io(struct mtip_port *port, unsigned long timeout)
950 blk_mq_quiesce_queue(port->dd->queue);
952 to = jiffies + msecs_to_jiffies(timeout);
954 if (test_bit(MTIP_PF_SVC_THD_ACTIVE_BIT, &port->flags) &&
955 test_bit(MTIP_PF_ISSUE_CMDS_BIT, &port->flags)) {
957 continue; /* svc thd is actively issuing commands */
962 if (mtip_check_surprise_removal(port->dd->pdev))
965 active = mtip_commands_active(port);
968 } while (time_before(jiffies, to));
970 blk_mq_unquiesce_queue(port->dd->queue);
971 return active ? -EBUSY : 0;
973 blk_mq_unquiesce_queue(port->dd->queue);
977 struct mtip_int_cmd {
985 * Execute an internal command and wait for the completion.
987 * @port Pointer to the port data structure.
988 * @fis Pointer to the FIS that describes the command.
989 * @fis_len Length in WORDS of the FIS.
990 * @buffer DMA accessible for command data.
991 * @buf_len Length, in bytes, of the data buffer.
992 * @opts Command header options, excluding the FIS length
993 * and the number of PRD entries.
994 * @timeout Time in ms to wait for the command to complete.
997 * 0 Command completed successfully.
998 * -EFAULT The buffer address is not correctly aligned.
999 * -EBUSY Internal command or other IO in progress.
1000 * -EAGAIN Time out waiting for command to complete.
1002 static int mtip_exec_internal_command(struct mtip_port *port,
1003 struct host_to_dev_fis *fis,
1008 unsigned long timeout)
1010 struct mtip_cmd *int_cmd;
1011 struct driver_data *dd = port->dd;
1013 struct mtip_int_cmd icmd = {
1021 /* Make sure the buffer is 8 byte aligned. This is asic specific. */
1022 if (buffer & 0x00000007) {
1023 dev_err(&dd->pdev->dev, "SG buffer is not 8 byte aligned\n");
1027 int_cmd = mtip_get_int_command(dd);
1029 dbg_printk(MTIP_DRV_NAME "Unable to allocate tag for PIO cmd\n");
1032 rq = blk_mq_rq_from_pdu(int_cmd);
1033 rq->special = &icmd;
1035 set_bit(MTIP_PF_IC_ACTIVE_BIT, &port->flags);
1037 if (fis->command == ATA_CMD_SEC_ERASE_PREP)
1038 set_bit(MTIP_PF_SE_ACTIVE_BIT, &port->flags);
1040 clear_bit(MTIP_PF_DM_ACTIVE_BIT, &port->flags);
1042 if (fis->command != ATA_CMD_STANDBYNOW1) {
1043 /* wait for io to complete if non atomic */
1044 if (mtip_quiesce_io(port, MTIP_QUIESCE_IO_TIMEOUT_MS) < 0) {
1045 dev_warn(&dd->pdev->dev, "Failed to quiesce IO\n");
1046 blk_mq_free_request(rq);
1047 clear_bit(MTIP_PF_IC_ACTIVE_BIT, &port->flags);
1048 wake_up_interruptible(&port->svc_wait);
1053 /* Copy the command to the command table */
1054 memcpy(int_cmd->command, fis, fis_len*4);
1056 rq->timeout = timeout;
1058 /* insert request and run queue */
1059 blk_execute_rq(rq->q, NULL, rq, true);
1061 if (int_cmd->status) {
1062 dev_err(&dd->pdev->dev, "Internal command [%02X] failed %d\n",
1063 fis->command, int_cmd->status);
1066 if (mtip_check_surprise_removal(dd->pdev) ||
1067 test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
1069 dev_err(&dd->pdev->dev,
1070 "Internal command [%02X] wait returned due to SR\n",
1075 mtip_device_reset(dd); /* recover from timeout issue */
1080 if (readl(port->cmd_issue[MTIP_TAG_INDEX(MTIP_TAG_INTERNAL)])
1081 & (1 << MTIP_TAG_BIT(MTIP_TAG_INTERNAL))) {
1083 if (!test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)) {
1084 mtip_device_reset(dd);
1089 /* Clear the allocated and active bits for the internal command. */
1090 blk_mq_free_request(rq);
1091 clear_bit(MTIP_PF_IC_ACTIVE_BIT, &port->flags);
1092 if (rv >= 0 && mtip_pause_ncq(port, fis)) {
1096 wake_up_interruptible(&port->svc_wait);
1102 * Byte-swap ATA ID strings.
1104 * ATA identify data contains strings in byte-swapped 16-bit words.
1105 * They must be swapped (on all architectures) to be usable as C strings.
1106 * This function swaps bytes in-place.
1108 * @buf The buffer location of the string
1109 * @len The number of bytes to swap
1114 static inline void ata_swap_string(u16 *buf, unsigned int len)
1117 for (i = 0; i < (len/2); i++)
1118 be16_to_cpus(&buf[i]);
1121 static void mtip_set_timeout(struct driver_data *dd,
1122 struct host_to_dev_fis *fis,
1123 unsigned int *timeout, u8 erasemode)
1125 switch (fis->command) {
1126 case ATA_CMD_DOWNLOAD_MICRO:
1127 *timeout = 120000; /* 2 minutes */
1129 case ATA_CMD_SEC_ERASE_UNIT:
1132 *timeout = ((*(dd->port->identify + 90) * 2) * 60000);
1134 *timeout = ((*(dd->port->identify + 89) * 2) * 60000);
1136 case ATA_CMD_STANDBYNOW1:
1137 *timeout = 120000; /* 2 minutes */
1141 *timeout = 60000; /* 60 seconds */
1144 *timeout = 15000; /* 15 seconds */
1147 *timeout = MTIP_IOCTL_CMD_TIMEOUT_MS;
1153 * Request the device identity information.
1155 * If a user space buffer is not specified, i.e. is NULL, the
1156 * identify information is still read from the drive and placed
1157 * into the identify data buffer (@e port->identify) in the
1158 * port data structure.
1159 * When the identify buffer contains valid identify information @e
1160 * port->identify_valid is non-zero.
1162 * @port Pointer to the port structure.
1163 * @user_buffer A user space buffer where the identify data should be
1167 * 0 Command completed successfully.
1168 * -EFAULT An error occurred while coping data to the user buffer.
1169 * -1 Command failed.
1171 static int mtip_get_identify(struct mtip_port *port, void __user *user_buffer)
1174 struct host_to_dev_fis fis;
1176 if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &port->dd->dd_flag))
1179 /* Build the FIS. */
1180 memset(&fis, 0, sizeof(struct host_to_dev_fis));
1183 fis.command = ATA_CMD_ID_ATA;
1185 /* Set the identify information as invalid. */
1186 port->identify_valid = 0;
1188 /* Clear the identify information. */
1189 memset(port->identify, 0, sizeof(u16) * ATA_ID_WORDS);
1191 /* Execute the command. */
1192 if (mtip_exec_internal_command(port,
1196 sizeof(u16) * ATA_ID_WORDS,
1198 MTIP_INT_CMD_TIMEOUT_MS)
1205 * Perform any necessary byte-swapping. Yes, the kernel does in fact
1206 * perform field-sensitive swapping on the string fields.
1207 * See the kernel use of ata_id_string() for proof of this.
1209 #ifdef __LITTLE_ENDIAN
1210 ata_swap_string(port->identify + 27, 40); /* model string*/
1211 ata_swap_string(port->identify + 23, 8); /* firmware string*/
1212 ata_swap_string(port->identify + 10, 20); /* serial# string*/
1216 for (i = 0; i < ATA_ID_WORDS; i++)
1217 port->identify[i] = le16_to_cpu(port->identify[i]);
1221 /* Check security locked state */
1222 if (port->identify[128] & 0x4)
1223 set_bit(MTIP_DDF_SEC_LOCK_BIT, &port->dd->dd_flag);
1225 clear_bit(MTIP_DDF_SEC_LOCK_BIT, &port->dd->dd_flag);
1227 #ifdef MTIP_TRIM /* Disabling TRIM support temporarily */
1228 /* Demux ID.DRAT & ID.RZAT to determine trim support */
1229 if (port->identify[69] & (1 << 14) && port->identify[69] & (1 << 5))
1230 port->dd->trim_supp = true;
1233 port->dd->trim_supp = false;
1235 /* Set the identify buffer as valid. */
1236 port->identify_valid = 1;
1242 ATA_ID_WORDS * sizeof(u16))) {
1253 * Issue a standby immediate command to the device.
1255 * @port Pointer to the port structure.
1258 * 0 Command was executed successfully.
1259 * -1 An error occurred while executing the command.
1261 static int mtip_standby_immediate(struct mtip_port *port)
1264 struct host_to_dev_fis fis;
1265 unsigned long start;
1266 unsigned int timeout;
1268 /* Build the FIS. */
1269 memset(&fis, 0, sizeof(struct host_to_dev_fis));
1272 fis.command = ATA_CMD_STANDBYNOW1;
1274 mtip_set_timeout(port->dd, &fis, &timeout, 0);
1277 rv = mtip_exec_internal_command(port,
1284 dbg_printk(MTIP_DRV_NAME "Time taken to complete standby cmd: %d ms\n",
1285 jiffies_to_msecs(jiffies - start));
1287 dev_warn(&port->dd->pdev->dev,
1288 "STANDBY IMMEDIATE command failed.\n");
1294 * Issue a READ LOG EXT command to the device.
1296 * @port pointer to the port structure.
1297 * @page page number to fetch
1298 * @buffer pointer to buffer
1299 * @buffer_dma dma address corresponding to @buffer
1300 * @sectors page length to fetch, in sectors
1303 * @rv return value from mtip_exec_internal_command()
1305 static int mtip_read_log_page(struct mtip_port *port, u8 page, u16 *buffer,
1306 dma_addr_t buffer_dma, unsigned int sectors)
1308 struct host_to_dev_fis fis;
1310 memset(&fis, 0, sizeof(struct host_to_dev_fis));
1313 fis.command = ATA_CMD_READ_LOG_EXT;
1314 fis.sect_count = sectors & 0xFF;
1315 fis.sect_cnt_ex = (sectors >> 8) & 0xFF;
1318 fis.device = ATA_DEVICE_OBS;
1320 memset(buffer, 0, sectors * ATA_SECT_SIZE);
1322 return mtip_exec_internal_command(port,
1326 sectors * ATA_SECT_SIZE,
1328 MTIP_INT_CMD_TIMEOUT_MS);
1332 * Issue a SMART READ DATA command to the device.
1334 * @port pointer to the port structure.
1335 * @buffer pointer to buffer
1336 * @buffer_dma dma address corresponding to @buffer
1339 * @rv return value from mtip_exec_internal_command()
1341 static int mtip_get_smart_data(struct mtip_port *port, u8 *buffer,
1342 dma_addr_t buffer_dma)
1344 struct host_to_dev_fis fis;
1346 memset(&fis, 0, sizeof(struct host_to_dev_fis));
1349 fis.command = ATA_CMD_SMART;
1350 fis.features = 0xD0;
1354 fis.device = ATA_DEVICE_OBS;
1356 return mtip_exec_internal_command(port,
1366 * Get the value of a smart attribute
1368 * @port pointer to the port structure
1369 * @id attribute number
1370 * @attrib pointer to return attrib information corresponding to @id
1373 * -EINVAL NULL buffer passed or unsupported attribute @id.
1374 * -EPERM Identify data not valid, SMART not supported or not enabled
1376 static int mtip_get_smart_attr(struct mtip_port *port, unsigned int id,
1377 struct smart_attr *attrib)
1380 struct smart_attr *pattr;
1385 if (!port->identify_valid) {
1386 dev_warn(&port->dd->pdev->dev, "IDENTIFY DATA not valid\n");
1389 if (!(port->identify[82] & 0x1)) {
1390 dev_warn(&port->dd->pdev->dev, "SMART not supported\n");
1393 if (!(port->identify[85] & 0x1)) {
1394 dev_warn(&port->dd->pdev->dev, "SMART not enabled\n");
1398 memset(port->smart_buf, 0, ATA_SECT_SIZE);
1399 rv = mtip_get_smart_data(port, port->smart_buf, port->smart_buf_dma);
1401 dev_warn(&port->dd->pdev->dev, "Failed to ge SMART data\n");
1405 pattr = (struct smart_attr *)(port->smart_buf + 2);
1406 for (i = 0; i < 29; i++, pattr++)
1407 if (pattr->attr_id == id) {
1408 memcpy(attrib, pattr, sizeof(struct smart_attr));
1413 dev_warn(&port->dd->pdev->dev,
1414 "Query for invalid SMART attribute ID\n");
1422 * Trim unused sectors
1424 * @dd pointer to driver_data structure
1426 * @len # of 512b sectors to trim
1429 * -ENOMEM Out of dma memory
1430 * -EINVAL Invalid parameters passed in, trim not supported
1431 * -EIO Error submitting trim request to hw
1433 static int mtip_send_trim(struct driver_data *dd, unsigned int lba,
1437 u64 tlba, tlen, sect_left;
1438 struct mtip_trim_entry *buf;
1439 dma_addr_t dma_addr;
1440 struct host_to_dev_fis fis;
1442 if (!len || dd->trim_supp == false)
1445 /* Trim request too big */
1446 WARN_ON(len > (MTIP_MAX_TRIM_ENTRY_LEN * MTIP_MAX_TRIM_ENTRIES));
1448 /* Trim request not aligned on 4k boundary */
1449 WARN_ON(len % 8 != 0);
1451 /* Warn if vu_trim structure is too big */
1452 WARN_ON(sizeof(struct mtip_trim) > ATA_SECT_SIZE);
1454 /* Allocate a DMA buffer for the trim structure */
1455 buf = dmam_alloc_coherent(&dd->pdev->dev, ATA_SECT_SIZE, &dma_addr,
1459 memset(buf, 0, ATA_SECT_SIZE);
1461 for (i = 0, sect_left = len, tlba = lba;
1462 i < MTIP_MAX_TRIM_ENTRIES && sect_left;
1464 tlen = (sect_left >= MTIP_MAX_TRIM_ENTRY_LEN ?
1465 MTIP_MAX_TRIM_ENTRY_LEN :
1467 buf[i].lba = __force_bit2int cpu_to_le32(tlba);
1468 buf[i].range = __force_bit2int cpu_to_le16(tlen);
1472 WARN_ON(sect_left != 0);
1475 memset(&fis, 0, sizeof(struct host_to_dev_fis));
1479 fis.features = 0x60;
1481 fis.device = ATA_DEVICE_OBS;
1483 if (mtip_exec_internal_command(dd->port,
1489 MTIP_TRIM_TIMEOUT_MS) < 0)
1492 dmam_free_coherent(&dd->pdev->dev, ATA_SECT_SIZE, buf, dma_addr);
1497 * Get the drive capacity.
1499 * @dd Pointer to the device data structure.
1500 * @sectors Pointer to the variable that will receive the sector count.
1503 * 1 Capacity was returned successfully.
1504 * 0 The identify information is invalid.
1506 static bool mtip_hw_get_capacity(struct driver_data *dd, sector_t *sectors)
1508 struct mtip_port *port = dd->port;
1509 u64 total, raw0, raw1, raw2, raw3;
1510 raw0 = port->identify[100];
1511 raw1 = port->identify[101];
1512 raw2 = port->identify[102];
1513 raw3 = port->identify[103];
1514 total = raw0 | raw1<<16 | raw2<<32 | raw3<<48;
1516 return (bool) !!port->identify_valid;
1520 * Display the identify command data.
1522 * @port Pointer to the port data structure.
1527 static void mtip_dump_identify(struct mtip_port *port)
1530 unsigned short revid;
1533 if (!port->identify_valid)
1536 strlcpy(cbuf, (char *)(port->identify+10), 21);
1537 dev_info(&port->dd->pdev->dev,
1538 "Serial No.: %s\n", cbuf);
1540 strlcpy(cbuf, (char *)(port->identify+23), 9);
1541 dev_info(&port->dd->pdev->dev,
1542 "Firmware Ver.: %s\n", cbuf);
1544 strlcpy(cbuf, (char *)(port->identify+27), 41);
1545 dev_info(&port->dd->pdev->dev, "Model: %s\n", cbuf);
1547 dev_info(&port->dd->pdev->dev, "Security: %04x %s\n",
1548 port->identify[128],
1549 port->identify[128] & 0x4 ? "(LOCKED)" : "");
1551 if (mtip_hw_get_capacity(port->dd, §ors))
1552 dev_info(&port->dd->pdev->dev,
1553 "Capacity: %llu sectors (%llu MB)\n",
1555 ((u64)sectors) * ATA_SECT_SIZE >> 20);
1557 pci_read_config_word(port->dd->pdev, PCI_REVISION_ID, &revid);
1558 switch (revid & 0xFF) {
1560 strlcpy(cbuf, "A0", 3);
1563 strlcpy(cbuf, "A2", 3);
1566 strlcpy(cbuf, "?", 2);
1569 dev_info(&port->dd->pdev->dev,
1570 "Card Type: %s\n", cbuf);
1574 * Map the commands scatter list into the command table.
1576 * @command Pointer to the command.
1577 * @nents Number of scatter list entries.
1582 static inline void fill_command_sg(struct driver_data *dd,
1583 struct mtip_cmd *command,
1587 unsigned int dma_len;
1588 struct mtip_cmd_sg *command_sg;
1589 struct scatterlist *sg = command->sg;
1591 command_sg = command->command + AHCI_CMD_TBL_HDR_SZ;
1593 for (n = 0; n < nents; n++) {
1594 dma_len = sg_dma_len(sg);
1595 if (dma_len > 0x400000)
1596 dev_err(&dd->pdev->dev,
1597 "DMA segment length truncated\n");
1598 command_sg->info = __force_bit2int
1599 cpu_to_le32((dma_len-1) & 0x3FFFFF);
1600 command_sg->dba = __force_bit2int
1601 cpu_to_le32(sg_dma_address(sg));
1602 command_sg->dba_upper = __force_bit2int
1603 cpu_to_le32((sg_dma_address(sg) >> 16) >> 16);
1610 * @brief Execute a drive command.
1612 * return value 0 The command completed successfully.
1613 * return value -1 An error occurred while executing the command.
1615 static int exec_drive_task(struct mtip_port *port, u8 *command)
1617 struct host_to_dev_fis fis;
1618 struct host_to_dev_fis *reply = (port->rxfis + RX_FIS_D2H_REG);
1621 /* Build the FIS. */
1622 memset(&fis, 0, sizeof(struct host_to_dev_fis));
1625 fis.command = command[0];
1626 fis.features = command[1];
1627 fis.sect_count = command[2];
1628 fis.sector = command[3];
1629 fis.cyl_low = command[4];
1630 fis.cyl_hi = command[5];
1631 fis.device = command[6] & ~0x10; /* Clear the dev bit*/
1633 mtip_set_timeout(port->dd, &fis, &to, 0);
1635 dbg_printk(MTIP_DRV_NAME " %s: User Command: cmd %x, feat %x, nsect %x, sect %x, lcyl %x, hcyl %x, sel %x\n",
1645 /* Execute the command. */
1646 if (mtip_exec_internal_command(port,
1656 command[0] = reply->command; /* Status*/
1657 command[1] = reply->features; /* Error*/
1658 command[4] = reply->cyl_low;
1659 command[5] = reply->cyl_hi;
1661 dbg_printk(MTIP_DRV_NAME " %s: Completion Status: stat %x, err %x , cyl_lo %x cyl_hi %x\n",
1672 * @brief Execute a drive command.
1674 * @param port Pointer to the port data structure.
1675 * @param command Pointer to the user specified command parameters.
1676 * @param user_buffer Pointer to the user space buffer where read sector
1677 * data should be copied.
1679 * return value 0 The command completed successfully.
1680 * return value -EFAULT An error occurred while copying the completion
1681 * data to the user space buffer.
1682 * return value -1 An error occurred while executing the command.
1684 static int exec_drive_command(struct mtip_port *port, u8 *command,
1685 void __user *user_buffer)
1687 struct host_to_dev_fis fis;
1688 struct host_to_dev_fis *reply;
1690 dma_addr_t dma_addr = 0;
1691 int rv = 0, xfer_sz = command[3];
1698 buf = dmam_alloc_coherent(&port->dd->pdev->dev,
1699 ATA_SECT_SIZE * xfer_sz,
1703 dev_err(&port->dd->pdev->dev,
1704 "Memory allocation failed (%d bytes)\n",
1705 ATA_SECT_SIZE * xfer_sz);
1708 memset(buf, 0, ATA_SECT_SIZE * xfer_sz);
1711 /* Build the FIS. */
1712 memset(&fis, 0, sizeof(struct host_to_dev_fis));
1715 fis.command = command[0];
1716 fis.features = command[2];
1717 fis.sect_count = command[3];
1718 if (fis.command == ATA_CMD_SMART) {
1719 fis.sector = command[1];
1724 mtip_set_timeout(port->dd, &fis, &to, 0);
1727 reply = (port->rxfis + RX_FIS_PIO_SETUP);
1729 reply = (port->rxfis + RX_FIS_D2H_REG);
1731 dbg_printk(MTIP_DRV_NAME
1732 " %s: User Command: cmd %x, sect %x, "
1733 "feat %x, sectcnt %x\n",
1740 /* Execute the command. */
1741 if (mtip_exec_internal_command(port,
1744 (xfer_sz ? dma_addr : 0),
1745 (xfer_sz ? ATA_SECT_SIZE * xfer_sz : 0),
1750 goto exit_drive_command;
1753 /* Collect the completion status. */
1754 command[0] = reply->command; /* Status*/
1755 command[1] = reply->features; /* Error*/
1756 command[2] = reply->sect_count;
1758 dbg_printk(MTIP_DRV_NAME
1759 " %s: Completion Status: stat %x, "
1760 "err %x, nsect %x\n",
1767 if (copy_to_user(user_buffer,
1769 ATA_SECT_SIZE * command[3])) {
1771 goto exit_drive_command;
1776 dmam_free_coherent(&port->dd->pdev->dev,
1777 ATA_SECT_SIZE * xfer_sz, buf, dma_addr);
1782 * Indicates whether a command has a single sector payload.
1784 * @command passed to the device to perform the certain event.
1785 * @features passed to the device to perform the certain event.
1788 * 1 command is one that always has a single sector payload,
1789 * regardless of the value in the Sector Count field.
1793 static unsigned int implicit_sector(unsigned char command,
1794 unsigned char features)
1796 unsigned int rv = 0;
1798 /* list of commands that have an implicit sector count of 1 */
1800 case ATA_CMD_SEC_SET_PASS:
1801 case ATA_CMD_SEC_UNLOCK:
1802 case ATA_CMD_SEC_ERASE_PREP:
1803 case ATA_CMD_SEC_ERASE_UNIT:
1804 case ATA_CMD_SEC_FREEZE_LOCK:
1805 case ATA_CMD_SEC_DISABLE_PASS:
1806 case ATA_CMD_PMP_READ:
1807 case ATA_CMD_PMP_WRITE:
1810 case ATA_CMD_SET_MAX:
1811 if (features == ATA_SET_MAX_UNLOCK)
1815 if ((features == ATA_SMART_READ_VALUES) ||
1816 (features == ATA_SMART_READ_THRESHOLDS))
1819 case ATA_CMD_CONF_OVERLAY:
1820 if ((features == ATA_DCO_IDENTIFY) ||
1821 (features == ATA_DCO_SET))
1829 * Executes a taskfile
1830 * See ide_taskfile_ioctl() for derivation
1832 static int exec_drive_taskfile(struct driver_data *dd,
1834 ide_task_request_t *req_task,
1837 struct host_to_dev_fis fis;
1838 struct host_to_dev_fis *reply;
1841 dma_addr_t outbuf_dma = 0;
1842 dma_addr_t inbuf_dma = 0;
1843 dma_addr_t dma_buffer = 0;
1845 unsigned int taskin = 0;
1846 unsigned int taskout = 0;
1848 unsigned int timeout;
1849 unsigned int force_single_sector;
1850 unsigned int transfer_size;
1851 unsigned long task_file_data;
1852 int intotal = outtotal + req_task->out_size;
1855 taskout = req_task->out_size;
1856 taskin = req_task->in_size;
1857 /* 130560 = 512 * 0xFF*/
1858 if (taskin > 130560 || taskout > 130560)
1862 outbuf = memdup_user(buf + outtotal, taskout);
1864 return PTR_ERR(outbuf);
1866 outbuf_dma = pci_map_single(dd->pdev,
1870 if (pci_dma_mapping_error(dd->pdev, outbuf_dma)) {
1874 dma_buffer = outbuf_dma;
1878 inbuf = memdup_user(buf + intotal, taskin);
1879 if (IS_ERR(inbuf)) {
1880 err = PTR_ERR(inbuf);
1884 inbuf_dma = pci_map_single(dd->pdev,
1886 taskin, DMA_FROM_DEVICE);
1887 if (pci_dma_mapping_error(dd->pdev, inbuf_dma)) {
1891 dma_buffer = inbuf_dma;
1894 /* only supports PIO and non-data commands from this ioctl. */
1895 switch (req_task->data_phase) {
1897 nsect = taskout / ATA_SECT_SIZE;
1898 reply = (dd->port->rxfis + RX_FIS_PIO_SETUP);
1901 reply = (dd->port->rxfis + RX_FIS_PIO_SETUP);
1903 case TASKFILE_NO_DATA:
1904 reply = (dd->port->rxfis + RX_FIS_D2H_REG);
1911 /* Build the FIS. */
1912 memset(&fis, 0, sizeof(struct host_to_dev_fis));
1916 fis.command = req_task->io_ports[7];
1917 fis.features = req_task->io_ports[1];
1918 fis.sect_count = req_task->io_ports[2];
1919 fis.lba_low = req_task->io_ports[3];
1920 fis.lba_mid = req_task->io_ports[4];
1921 fis.lba_hi = req_task->io_ports[5];
1922 /* Clear the dev bit*/
1923 fis.device = req_task->io_ports[6] & ~0x10;
1925 if ((req_task->in_flags.all == 0) && (req_task->out_flags.all & 1)) {
1926 req_task->in_flags.all =
1927 IDE_TASKFILE_STD_IN_FLAGS |
1928 (IDE_HOB_STD_IN_FLAGS << 8);
1929 fis.lba_low_ex = req_task->hob_ports[3];
1930 fis.lba_mid_ex = req_task->hob_ports[4];
1931 fis.lba_hi_ex = req_task->hob_ports[5];
1932 fis.features_ex = req_task->hob_ports[1];
1933 fis.sect_cnt_ex = req_task->hob_ports[2];
1936 req_task->in_flags.all = IDE_TASKFILE_STD_IN_FLAGS;
1939 force_single_sector = implicit_sector(fis.command, fis.features);
1941 if ((taskin || taskout) && (!fis.sect_count)) {
1943 fis.sect_count = nsect;
1945 if (!force_single_sector) {
1946 dev_warn(&dd->pdev->dev,
1947 "data movement but "
1948 "sect_count is 0\n");
1955 dbg_printk(MTIP_DRV_NAME
1956 " %s: cmd %x, feat %x, nsect %x,"
1957 " sect/lbal %x, lcyl/lbam %x, hcyl/lbah %x,"
1968 /* check for erase mode support during secure erase.*/
1969 if ((fis.command == ATA_CMD_SEC_ERASE_UNIT) && outbuf &&
1970 (outbuf[0] & MTIP_SEC_ERASE_MODE)) {
1974 mtip_set_timeout(dd, &fis, &timeout, erasemode);
1976 /* Determine the correct transfer size.*/
1977 if (force_single_sector)
1978 transfer_size = ATA_SECT_SIZE;
1980 transfer_size = ATA_SECT_SIZE * fis.sect_count;
1982 /* Execute the command.*/
1983 if (mtip_exec_internal_command(dd->port,
1994 task_file_data = readl(dd->port->mmio+PORT_TFDATA);
1996 if ((req_task->data_phase == TASKFILE_IN) && !(task_file_data & 1)) {
1997 reply = dd->port->rxfis + RX_FIS_PIO_SETUP;
1998 req_task->io_ports[7] = reply->control;
2000 reply = dd->port->rxfis + RX_FIS_D2H_REG;
2001 req_task->io_ports[7] = reply->command;
2004 /* reclaim the DMA buffers.*/
2006 pci_unmap_single(dd->pdev, inbuf_dma,
2007 taskin, DMA_FROM_DEVICE);
2009 pci_unmap_single(dd->pdev, outbuf_dma,
2010 taskout, DMA_TO_DEVICE);
2014 /* return the ATA registers to the caller.*/
2015 req_task->io_ports[1] = reply->features;
2016 req_task->io_ports[2] = reply->sect_count;
2017 req_task->io_ports[3] = reply->lba_low;
2018 req_task->io_ports[4] = reply->lba_mid;
2019 req_task->io_ports[5] = reply->lba_hi;
2020 req_task->io_ports[6] = reply->device;
2022 if (req_task->out_flags.all & 1) {
2024 req_task->hob_ports[3] = reply->lba_low_ex;
2025 req_task->hob_ports[4] = reply->lba_mid_ex;
2026 req_task->hob_ports[5] = reply->lba_hi_ex;
2027 req_task->hob_ports[1] = reply->features_ex;
2028 req_task->hob_ports[2] = reply->sect_cnt_ex;
2030 dbg_printk(MTIP_DRV_NAME
2031 " %s: Completion: stat %x,"
2032 "err %x, sect_cnt %x, lbalo %x,"
2033 "lbamid %x, lbahi %x, dev %x\n",
2035 req_task->io_ports[7],
2036 req_task->io_ports[1],
2037 req_task->io_ports[2],
2038 req_task->io_ports[3],
2039 req_task->io_ports[4],
2040 req_task->io_ports[5],
2041 req_task->io_ports[6]);
2044 if (copy_to_user(buf + outtotal, outbuf, taskout)) {
2050 if (copy_to_user(buf + intotal, inbuf, taskin)) {
2057 pci_unmap_single(dd->pdev, inbuf_dma,
2058 taskin, DMA_FROM_DEVICE);
2060 pci_unmap_single(dd->pdev, outbuf_dma,
2061 taskout, DMA_TO_DEVICE);
2069 * Handle IOCTL calls from the Block Layer.
2071 * This function is called by the Block Layer when it receives an IOCTL
2072 * command that it does not understand. If the IOCTL command is not supported
2073 * this function returns -ENOTTY.
2075 * @dd Pointer to the driver data structure.
2076 * @cmd IOCTL command passed from the Block Layer.
2077 * @arg IOCTL argument passed from the Block Layer.
2080 * 0 The IOCTL completed successfully.
2081 * -ENOTTY The specified command is not supported.
2082 * -EFAULT An error occurred copying data to a user space buffer.
2083 * -EIO An error occurred while executing the command.
2085 static int mtip_hw_ioctl(struct driver_data *dd, unsigned int cmd,
2089 case HDIO_GET_IDENTITY:
2091 if (copy_to_user((void __user *)arg, dd->port->identify,
2092 sizeof(u16) * ATA_ID_WORDS))
2096 case HDIO_DRIVE_CMD:
2098 u8 drive_command[4];
2100 /* Copy the user command info to our buffer. */
2101 if (copy_from_user(drive_command,
2102 (void __user *) arg,
2103 sizeof(drive_command)))
2106 /* Execute the drive command. */
2107 if (exec_drive_command(dd->port,
2109 (void __user *) (arg+4)))
2112 /* Copy the status back to the users buffer. */
2113 if (copy_to_user((void __user *) arg,
2115 sizeof(drive_command)))
2120 case HDIO_DRIVE_TASK:
2122 u8 drive_command[7];
2124 /* Copy the user command info to our buffer. */
2125 if (copy_from_user(drive_command,
2126 (void __user *) arg,
2127 sizeof(drive_command)))
2130 /* Execute the drive command. */
2131 if (exec_drive_task(dd->port, drive_command))
2134 /* Copy the status back to the users buffer. */
2135 if (copy_to_user((void __user *) arg,
2137 sizeof(drive_command)))
2142 case HDIO_DRIVE_TASKFILE: {
2143 ide_task_request_t req_task;
2146 if (copy_from_user(&req_task, (void __user *) arg,
2150 outtotal = sizeof(req_task);
2152 ret = exec_drive_taskfile(dd, (void __user *) arg,
2153 &req_task, outtotal);
2155 if (copy_to_user((void __user *) arg, &req_task,
2169 * Submit an IO to the hw
2171 * This function is called by the block layer to issue an io
2172 * to the device. Upon completion, the callback function will
2173 * be called with the data parameter passed as the callback data.
2175 * @dd Pointer to the driver data structure.
2176 * @start First sector to read.
2177 * @nsect Number of sectors to read.
2178 * @nents Number of entries in scatter list for the read command.
2179 * @tag The tag of this read command.
2180 * @callback Pointer to the function that should be called
2181 * when the read completes.
2182 * @data Callback data passed to the callback function
2183 * when the read completes.
2184 * @dir Direction (read or write)
2189 static void mtip_hw_submit_io(struct driver_data *dd, struct request *rq,
2190 struct mtip_cmd *command, int nents,
2191 struct blk_mq_hw_ctx *hctx)
2193 struct host_to_dev_fis *fis;
2194 struct mtip_port *port = dd->port;
2195 int dma_dir = rq_data_dir(rq) == READ ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
2196 u64 start = blk_rq_pos(rq);
2197 unsigned int nsect = blk_rq_sectors(rq);
2199 /* Map the scatter list for DMA access */
2200 nents = dma_map_sg(&dd->pdev->dev, command->sg, nents, dma_dir);
2202 prefetch(&port->flags);
2204 command->scatter_ents = nents;
2207 * The number of retries for this command before it is
2208 * reported as a failure to the upper layers.
2210 command->retries = MTIP_MAX_RETRIES;
2213 fis = command->command;
2216 if (dma_dir == DMA_FROM_DEVICE)
2217 fis->command = ATA_CMD_FPDMA_READ;
2219 fis->command = ATA_CMD_FPDMA_WRITE;
2220 fis->lba_low = start & 0xFF;
2221 fis->lba_mid = (start >> 8) & 0xFF;
2222 fis->lba_hi = (start >> 16) & 0xFF;
2223 fis->lba_low_ex = (start >> 24) & 0xFF;
2224 fis->lba_mid_ex = (start >> 32) & 0xFF;
2225 fis->lba_hi_ex = (start >> 40) & 0xFF;
2226 fis->device = 1 << 6;
2227 fis->features = nsect & 0xFF;
2228 fis->features_ex = (nsect >> 8) & 0xFF;
2229 fis->sect_count = ((rq->tag << 3) | (rq->tag >> 5));
2230 fis->sect_cnt_ex = 0;
2234 fill_command_sg(dd, command, nents);
2236 if (unlikely(command->unaligned))
2237 fis->device |= 1 << 7;
2239 /* Populate the command header */
2240 command->command_header->opts =
2241 __force_bit2int cpu_to_le32(
2242 (nents << 16) | 5 | AHCI_CMD_PREFETCH);
2243 command->command_header->byte_count = 0;
2245 command->direction = dma_dir;
2248 * To prevent this command from being issued
2249 * if an internal command is in progress or error handling is active.
2251 if (unlikely(port->flags & MTIP_PF_PAUSE_IO)) {
2252 set_bit(rq->tag, port->cmds_to_issue);
2253 set_bit(MTIP_PF_ISSUE_CMDS_BIT, &port->flags);
2257 /* Issue the command to the hardware */
2258 mtip_issue_ncq_command(port, rq->tag);
2262 * Sysfs status dump.
2264 * @dev Pointer to the device structure, passed by the kernrel.
2265 * @attr Pointer to the device_attribute structure passed by the kernel.
2266 * @buf Pointer to the char buffer that will receive the stats info.
2269 * The size, in bytes, of the data copied into buf.
2271 static ssize_t mtip_hw_show_status(struct device *dev,
2272 struct device_attribute *attr,
2275 struct driver_data *dd = dev_to_disk(dev)->private_data;
2278 if (test_bit(MTIP_DDF_OVER_TEMP_BIT, &dd->dd_flag))
2279 size += sprintf(buf, "%s", "thermal_shutdown\n");
2280 else if (test_bit(MTIP_DDF_WRITE_PROTECT_BIT, &dd->dd_flag))
2281 size += sprintf(buf, "%s", "write_protect\n");
2283 size += sprintf(buf, "%s", "online\n");
2288 static DEVICE_ATTR(status, 0444, mtip_hw_show_status, NULL);
2290 /* debugsfs entries */
2292 static ssize_t show_device_status(struct device_driver *drv, char *buf)
2295 struct driver_data *dd, *tmp;
2296 unsigned long flags;
2300 spin_lock_irqsave(&dev_lock, flags);
2301 size += sprintf(&buf[size], "Devices Present:\n");
2302 list_for_each_entry_safe(dd, tmp, &online_list, online_list) {
2305 dd->port->identify &&
2306 dd->port->identify_valid) {
2308 (char *) (dd->port->identify + 10), 21);
2309 status = *(dd->port->identify + 141);
2311 memset(id_buf, 0, 42);
2316 test_bit(MTIP_PF_REBUILD_BIT, &dd->port->flags)) {
2317 size += sprintf(&buf[size],
2318 " device %s %s (ftl rebuild %d %%)\n",
2319 dev_name(&dd->pdev->dev),
2323 size += sprintf(&buf[size],
2325 dev_name(&dd->pdev->dev),
2331 size += sprintf(&buf[size], "Devices Being Removed:\n");
2332 list_for_each_entry_safe(dd, tmp, &removing_list, remove_list) {
2335 dd->port->identify &&
2336 dd->port->identify_valid) {
2338 (char *) (dd->port->identify+10), 21);
2339 status = *(dd->port->identify + 141);
2341 memset(id_buf, 0, 42);
2346 test_bit(MTIP_PF_REBUILD_BIT, &dd->port->flags)) {
2347 size += sprintf(&buf[size],
2348 " device %s %s (ftl rebuild %d %%)\n",
2349 dev_name(&dd->pdev->dev),
2353 size += sprintf(&buf[size],
2355 dev_name(&dd->pdev->dev),
2360 spin_unlock_irqrestore(&dev_lock, flags);
2365 static ssize_t mtip_hw_read_device_status(struct file *f, char __user *ubuf,
2366 size_t len, loff_t *offset)
2368 struct driver_data *dd = (struct driver_data *)f->private_data;
2373 if (!len || *offset)
2376 buf = kzalloc(MTIP_DFS_MAX_BUF_SIZE, GFP_KERNEL);
2378 dev_err(&dd->pdev->dev,
2379 "Memory allocation: status buffer\n");
2383 size += show_device_status(NULL, buf);
2385 *offset = size <= len ? size : len;
2386 size = copy_to_user(ubuf, buf, *offset);
2391 return rv ? rv : *offset;
2394 static ssize_t mtip_hw_read_registers(struct file *f, char __user *ubuf,
2395 size_t len, loff_t *offset)
2397 struct driver_data *dd = (struct driver_data *)f->private_data;
2399 u32 group_allocated;
2406 buf = kzalloc(MTIP_DFS_MAX_BUF_SIZE, GFP_KERNEL);
2408 dev_err(&dd->pdev->dev,
2409 "Memory allocation: register buffer\n");
2413 size += sprintf(&buf[size], "H/ S ACTive : [ 0x");
2415 for (n = dd->slot_groups-1; n >= 0; n--)
2416 size += sprintf(&buf[size], "%08X ",
2417 readl(dd->port->s_active[n]));
2419 size += sprintf(&buf[size], "]\n");
2420 size += sprintf(&buf[size], "H/ Command Issue : [ 0x");
2422 for (n = dd->slot_groups-1; n >= 0; n--)
2423 size += sprintf(&buf[size], "%08X ",
2424 readl(dd->port->cmd_issue[n]));
2426 size += sprintf(&buf[size], "]\n");
2427 size += sprintf(&buf[size], "H/ Completed : [ 0x");
2429 for (n = dd->slot_groups-1; n >= 0; n--)
2430 size += sprintf(&buf[size], "%08X ",
2431 readl(dd->port->completed[n]));
2433 size += sprintf(&buf[size], "]\n");
2434 size += sprintf(&buf[size], "H/ PORT IRQ STAT : [ 0x%08X ]\n",
2435 readl(dd->port->mmio + PORT_IRQ_STAT));
2436 size += sprintf(&buf[size], "H/ HOST IRQ STAT : [ 0x%08X ]\n",
2437 readl(dd->mmio + HOST_IRQ_STAT));
2438 size += sprintf(&buf[size], "\n");
2440 size += sprintf(&buf[size], "L/ Commands in Q : [ 0x");
2442 for (n = dd->slot_groups-1; n >= 0; n--) {
2443 if (sizeof(long) > sizeof(u32))
2445 dd->port->cmds_to_issue[n/2] >> (32*(n&1));
2447 group_allocated = dd->port->cmds_to_issue[n];
2448 size += sprintf(&buf[size], "%08X ", group_allocated);
2450 size += sprintf(&buf[size], "]\n");
2452 *offset = size <= len ? size : len;
2453 size = copy_to_user(ubuf, buf, *offset);
2458 return rv ? rv : *offset;
2461 static ssize_t mtip_hw_read_flags(struct file *f, char __user *ubuf,
2462 size_t len, loff_t *offset)
2464 struct driver_data *dd = (struct driver_data *)f->private_data;
2472 buf = kzalloc(MTIP_DFS_MAX_BUF_SIZE, GFP_KERNEL);
2474 dev_err(&dd->pdev->dev,
2475 "Memory allocation: flag buffer\n");
2479 size += sprintf(&buf[size], "Flag-port : [ %08lX ]\n",
2481 size += sprintf(&buf[size], "Flag-dd : [ %08lX ]\n",
2484 *offset = size <= len ? size : len;
2485 size = copy_to_user(ubuf, buf, *offset);
2490 return rv ? rv : *offset;
2493 static const struct file_operations mtip_device_status_fops = {
2494 .owner = THIS_MODULE,
2495 .open = simple_open,
2496 .read = mtip_hw_read_device_status,
2497 .llseek = no_llseek,
2500 static const struct file_operations mtip_regs_fops = {
2501 .owner = THIS_MODULE,
2502 .open = simple_open,
2503 .read = mtip_hw_read_registers,
2504 .llseek = no_llseek,
2507 static const struct file_operations mtip_flags_fops = {
2508 .owner = THIS_MODULE,
2509 .open = simple_open,
2510 .read = mtip_hw_read_flags,
2511 .llseek = no_llseek,
2515 * Create the sysfs related attributes.
2517 * @dd Pointer to the driver data structure.
2518 * @kobj Pointer to the kobj for the block device.
2521 * 0 Operation completed successfully.
2522 * -EINVAL Invalid parameter.
2524 static int mtip_hw_sysfs_init(struct driver_data *dd, struct kobject *kobj)
2529 if (sysfs_create_file(kobj, &dev_attr_status.attr))
2530 dev_warn(&dd->pdev->dev,
2531 "Error creating 'status' sysfs entry\n");
2536 * Remove the sysfs related attributes.
2538 * @dd Pointer to the driver data structure.
2539 * @kobj Pointer to the kobj for the block device.
2542 * 0 Operation completed successfully.
2543 * -EINVAL Invalid parameter.
2545 static int mtip_hw_sysfs_exit(struct driver_data *dd, struct kobject *kobj)
2550 sysfs_remove_file(kobj, &dev_attr_status.attr);
2555 static int mtip_hw_debugfs_init(struct driver_data *dd)
2560 dd->dfs_node = debugfs_create_dir(dd->disk->disk_name, dfs_parent);
2561 if (IS_ERR_OR_NULL(dd->dfs_node)) {
2562 dev_warn(&dd->pdev->dev,
2563 "Error creating node %s under debugfs\n",
2564 dd->disk->disk_name);
2565 dd->dfs_node = NULL;
2569 debugfs_create_file("flags", 0444, dd->dfs_node, dd, &mtip_flags_fops);
2570 debugfs_create_file("registers", 0444, dd->dfs_node, dd,
2576 static void mtip_hw_debugfs_exit(struct driver_data *dd)
2578 debugfs_remove_recursive(dd->dfs_node);
2582 * Perform any init/resume time hardware setup
2584 * @dd Pointer to the driver data structure.
2589 static inline void hba_setup(struct driver_data *dd)
2592 hwdata = readl(dd->mmio + HOST_HSORG);
2594 /* interrupt bug workaround: use only 1 IS bit.*/
2596 HSORG_DISABLE_SLOTGRP_INTR |
2597 HSORG_DISABLE_SLOTGRP_PXIS,
2598 dd->mmio + HOST_HSORG);
2601 static int mtip_device_unaligned_constrained(struct driver_data *dd)
2603 return (dd->pdev->device == P420M_DEVICE_ID ? 1 : 0);
2607 * Detect the details of the product, and store anything needed
2608 * into the driver data structure. This includes product type and
2609 * version and number of slot groups.
2611 * @dd Pointer to the driver data structure.
2616 static void mtip_detect_product(struct driver_data *dd)
2619 unsigned int rev, slotgroups;
2622 * HBA base + 0xFC [15:0] - vendor-specific hardware interface
2624 * [15:8] hardware/software interface rev#
2625 * [ 3] asic-style interface
2626 * [ 2:0] number of slot groups, minus 1 (only valid for asic-style).
2628 hwdata = readl(dd->mmio + HOST_HSORG);
2630 dd->product_type = MTIP_PRODUCT_UNKNOWN;
2631 dd->slot_groups = 1;
2634 dd->product_type = MTIP_PRODUCT_ASICFPGA;
2635 rev = (hwdata & HSORG_HWREV) >> 8;
2636 slotgroups = (hwdata & HSORG_SLOTGROUPS) + 1;
2637 dev_info(&dd->pdev->dev,
2638 "ASIC-FPGA design, HS rev 0x%x, "
2639 "%i slot groups [%i slots]\n",
2644 if (slotgroups > MTIP_MAX_SLOT_GROUPS) {
2645 dev_warn(&dd->pdev->dev,
2646 "Warning: driver only supports "
2647 "%i slot groups.\n", MTIP_MAX_SLOT_GROUPS);
2648 slotgroups = MTIP_MAX_SLOT_GROUPS;
2650 dd->slot_groups = slotgroups;
2654 dev_warn(&dd->pdev->dev, "Unrecognized product id\n");
2658 * Blocking wait for FTL rebuild to complete
2660 * @dd Pointer to the DRIVER_DATA structure.
2663 * 0 FTL rebuild completed successfully
2664 * -EFAULT FTL rebuild error/timeout/interruption
2666 static int mtip_ftl_rebuild_poll(struct driver_data *dd)
2668 unsigned long timeout, cnt = 0, start;
2670 dev_warn(&dd->pdev->dev,
2671 "FTL rebuild in progress. Polling for completion.\n");
2674 timeout = jiffies + msecs_to_jiffies(MTIP_FTL_REBUILD_TIMEOUT_MS);
2677 if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
2680 if (mtip_check_surprise_removal(dd->pdev))
2683 if (mtip_get_identify(dd->port, NULL) < 0)
2686 if (*(dd->port->identify + MTIP_FTL_REBUILD_OFFSET) ==
2687 MTIP_FTL_REBUILD_MAGIC) {
2689 /* Print message every 3 minutes */
2691 dev_warn(&dd->pdev->dev,
2692 "FTL rebuild in progress (%d secs).\n",
2693 jiffies_to_msecs(jiffies - start) / 1000);
2697 dev_warn(&dd->pdev->dev,
2698 "FTL rebuild complete (%d secs).\n",
2699 jiffies_to_msecs(jiffies - start) / 1000);
2700 mtip_block_initialize(dd);
2703 } while (time_before(jiffies, timeout));
2705 /* Check for timeout */
2706 dev_err(&dd->pdev->dev,
2707 "Timed out waiting for FTL rebuild to complete (%d secs).\n",
2708 jiffies_to_msecs(jiffies - start) / 1000);
2712 static void mtip_softirq_done_fn(struct request *rq)
2714 struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
2715 struct driver_data *dd = rq->q->queuedata;
2717 /* Unmap the DMA scatter list entries */
2718 dma_unmap_sg(&dd->pdev->dev, cmd->sg, cmd->scatter_ents,
2721 if (unlikely(cmd->unaligned))
2722 up(&dd->port->cmd_slot_unal);
2724 blk_mq_end_request(rq, cmd->status);
2727 static void mtip_abort_cmd(struct request *req, void *data, bool reserved)
2729 struct mtip_cmd *cmd = blk_mq_rq_to_pdu(req);
2730 struct driver_data *dd = data;
2732 dbg_printk(MTIP_DRV_NAME " Aborting request, tag = %d\n", req->tag);
2734 clear_bit(req->tag, dd->port->cmds_to_issue);
2735 cmd->status = BLK_STS_IOERR;
2736 mtip_softirq_done_fn(req);
2739 static void mtip_queue_cmd(struct request *req, void *data, bool reserved)
2741 struct driver_data *dd = data;
2743 set_bit(req->tag, dd->port->cmds_to_issue);
2744 blk_abort_request(req);
2748 * service thread to issue queued commands
2750 * @data Pointer to the driver data structure.
2756 static int mtip_service_thread(void *data)
2758 struct driver_data *dd = (struct driver_data *)data;
2759 unsigned long slot, slot_start, slot_wrap, to;
2760 unsigned int num_cmd_slots = dd->slot_groups * 32;
2761 struct mtip_port *port = dd->port;
2764 if (kthread_should_stop() ||
2765 test_bit(MTIP_PF_SVC_THD_STOP_BIT, &port->flags))
2767 clear_bit(MTIP_PF_SVC_THD_ACTIVE_BIT, &port->flags);
2770 * the condition is to check neither an internal command is
2771 * is in progress nor error handling is active
2773 wait_event_interruptible(port->svc_wait, (port->flags) &&
2774 (port->flags & MTIP_PF_SVC_THD_WORK));
2776 if (kthread_should_stop() ||
2777 test_bit(MTIP_PF_SVC_THD_STOP_BIT, &port->flags))
2780 if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
2784 set_bit(MTIP_PF_SVC_THD_ACTIVE_BIT, &port->flags);
2787 /* Demux bits: start with error handling */
2788 if (test_bit(MTIP_PF_EH_ACTIVE_BIT, &port->flags)) {
2789 mtip_handle_tfe(dd);
2790 clear_bit(MTIP_PF_EH_ACTIVE_BIT, &port->flags);
2793 if (test_bit(MTIP_PF_EH_ACTIVE_BIT, &port->flags))
2796 if (test_bit(MTIP_PF_TO_ACTIVE_BIT, &port->flags)) {
2797 to = jiffies + msecs_to_jiffies(5000);
2801 } while (atomic_read(&dd->irq_workers_active) != 0 &&
2802 time_before(jiffies, to));
2804 if (atomic_read(&dd->irq_workers_active) != 0)
2805 dev_warn(&dd->pdev->dev,
2806 "Completion workers still active!");
2808 blk_mq_quiesce_queue(dd->queue);
2810 spin_lock(dd->queue->queue_lock);
2811 blk_mq_tagset_busy_iter(&dd->tags,
2812 mtip_queue_cmd, dd);
2813 spin_unlock(dd->queue->queue_lock);
2815 set_bit(MTIP_PF_ISSUE_CMDS_BIT, &dd->port->flags);
2817 if (mtip_device_reset(dd))
2818 blk_mq_tagset_busy_iter(&dd->tags,
2819 mtip_abort_cmd, dd);
2821 clear_bit(MTIP_PF_TO_ACTIVE_BIT, &dd->port->flags);
2823 blk_mq_unquiesce_queue(dd->queue);
2826 if (test_bit(MTIP_PF_ISSUE_CMDS_BIT, &port->flags)) {
2828 /* used to restrict the loop to one iteration */
2829 slot_start = num_cmd_slots;
2832 slot = find_next_bit(port->cmds_to_issue,
2833 num_cmd_slots, slot);
2834 if (slot_wrap == 1) {
2835 if ((slot_start >= slot) ||
2836 (slot >= num_cmd_slots))
2839 if (unlikely(slot_start == num_cmd_slots))
2842 if (unlikely(slot == num_cmd_slots)) {
2848 /* Issue the command to the hardware */
2849 mtip_issue_ncq_command(port, slot);
2851 clear_bit(slot, port->cmds_to_issue);
2854 clear_bit(MTIP_PF_ISSUE_CMDS_BIT, &port->flags);
2857 if (test_bit(MTIP_PF_REBUILD_BIT, &port->flags)) {
2858 if (mtip_ftl_rebuild_poll(dd) == 0)
2859 clear_bit(MTIP_PF_REBUILD_BIT, &port->flags);
2868 * DMA region teardown
2870 * @dd Pointer to driver_data structure
2875 static void mtip_dma_free(struct driver_data *dd)
2877 struct mtip_port *port = dd->port;
2880 dmam_free_coherent(&dd->pdev->dev, BLOCK_DMA_ALLOC_SZ,
2881 port->block1, port->block1_dma);
2883 if (port->command_list) {
2884 dmam_free_coherent(&dd->pdev->dev, AHCI_CMD_TBL_SZ,
2885 port->command_list, port->command_list_dma);
2892 * @dd Pointer to driver_data structure
2895 * -ENOMEM Not enough free DMA region space to initialize driver
2897 static int mtip_dma_alloc(struct driver_data *dd)
2899 struct mtip_port *port = dd->port;
2901 /* Allocate dma memory for RX Fis, Identify, and Sector Bufffer */
2903 dmam_alloc_coherent(&dd->pdev->dev, BLOCK_DMA_ALLOC_SZ,
2904 &port->block1_dma, GFP_KERNEL);
2907 memset(port->block1, 0, BLOCK_DMA_ALLOC_SZ);
2909 /* Allocate dma memory for command list */
2910 port->command_list =
2911 dmam_alloc_coherent(&dd->pdev->dev, AHCI_CMD_TBL_SZ,
2912 &port->command_list_dma, GFP_KERNEL);
2913 if (!port->command_list) {
2914 dmam_free_coherent(&dd->pdev->dev, BLOCK_DMA_ALLOC_SZ,
2915 port->block1, port->block1_dma);
2916 port->block1 = NULL;
2917 port->block1_dma = 0;
2920 memset(port->command_list, 0, AHCI_CMD_TBL_SZ);
2922 /* Setup all pointers into first DMA region */
2923 port->rxfis = port->block1 + AHCI_RX_FIS_OFFSET;
2924 port->rxfis_dma = port->block1_dma + AHCI_RX_FIS_OFFSET;
2925 port->identify = port->block1 + AHCI_IDFY_OFFSET;
2926 port->identify_dma = port->block1_dma + AHCI_IDFY_OFFSET;
2927 port->log_buf = port->block1 + AHCI_SECTBUF_OFFSET;
2928 port->log_buf_dma = port->block1_dma + AHCI_SECTBUF_OFFSET;
2929 port->smart_buf = port->block1 + AHCI_SMARTBUF_OFFSET;
2930 port->smart_buf_dma = port->block1_dma + AHCI_SMARTBUF_OFFSET;
2935 static int mtip_hw_get_identify(struct driver_data *dd)
2937 struct smart_attr attr242;
2941 if (mtip_get_identify(dd->port, NULL) < 0)
2944 if (*(dd->port->identify + MTIP_FTL_REBUILD_OFFSET) ==
2945 MTIP_FTL_REBUILD_MAGIC) {
2946 set_bit(MTIP_PF_REBUILD_BIT, &dd->port->flags);
2947 return MTIP_FTL_REBUILD_MAGIC;
2949 mtip_dump_identify(dd->port);
2951 /* check write protect, over temp and rebuild statuses */
2952 rv = mtip_read_log_page(dd->port, ATA_LOG_SATA_NCQ,
2954 dd->port->log_buf_dma, 1);
2956 dev_warn(&dd->pdev->dev,
2957 "Error in READ LOG EXT (10h) command\n");
2958 /* non-critical error, don't fail the load */
2960 buf = (unsigned char *)dd->port->log_buf;
2961 if (buf[259] & 0x1) {
2962 dev_info(&dd->pdev->dev,
2963 "Write protect bit is set.\n");
2964 set_bit(MTIP_DDF_WRITE_PROTECT_BIT, &dd->dd_flag);
2966 if (buf[288] == 0xF7) {
2967 dev_info(&dd->pdev->dev,
2968 "Exceeded Tmax, drive in thermal shutdown.\n");
2969 set_bit(MTIP_DDF_OVER_TEMP_BIT, &dd->dd_flag);
2971 if (buf[288] == 0xBF) {
2972 dev_info(&dd->pdev->dev,
2973 "Drive indicates rebuild has failed.\n");
2974 set_bit(MTIP_DDF_REBUILD_FAILED_BIT, &dd->dd_flag);
2978 /* get write protect progess */
2979 memset(&attr242, 0, sizeof(struct smart_attr));
2980 if (mtip_get_smart_attr(dd->port, 242, &attr242))
2981 dev_warn(&dd->pdev->dev,
2982 "Unable to check write protect progress\n");
2984 dev_info(&dd->pdev->dev,
2985 "Write protect progress: %u%% (%u blocks)\n",
2986 attr242.cur, le32_to_cpu(attr242.data));
2992 * Called once for each card.
2994 * @dd Pointer to the driver data structure.
2997 * 0 on success, else an error code.
2999 static int mtip_hw_init(struct driver_data *dd)
3003 unsigned long timeout, timetaken;
3005 dd->mmio = pcim_iomap_table(dd->pdev)[MTIP_ABAR];
3007 mtip_detect_product(dd);
3008 if (dd->product_type == MTIP_PRODUCT_UNKNOWN) {
3015 dd->port = kzalloc_node(sizeof(struct mtip_port), GFP_KERNEL,
3018 dev_err(&dd->pdev->dev,
3019 "Memory allocation: port structure\n");
3023 /* Continue workqueue setup */
3024 for (i = 0; i < MTIP_MAX_SLOT_GROUPS; i++)
3025 dd->work[i].port = dd->port;
3027 /* Enable unaligned IO constraints for some devices */
3028 if (mtip_device_unaligned_constrained(dd))
3029 dd->unal_qdepth = MTIP_MAX_UNALIGNED_SLOTS;
3031 dd->unal_qdepth = 0;
3033 sema_init(&dd->port->cmd_slot_unal, dd->unal_qdepth);
3035 /* Spinlock to prevent concurrent issue */
3036 for (i = 0; i < MTIP_MAX_SLOT_GROUPS; i++)
3037 spin_lock_init(&dd->port->cmd_issue_lock[i]);
3039 /* Set the port mmio base address. */
3040 dd->port->mmio = dd->mmio + PORT_OFFSET;
3043 /* DMA allocations */
3044 rv = mtip_dma_alloc(dd);
3048 /* Setup the pointers to the extended s_active and CI registers. */
3049 for (i = 0; i < dd->slot_groups; i++) {
3050 dd->port->s_active[i] =
3051 dd->port->mmio + i*0x80 + PORT_SCR_ACT;
3052 dd->port->cmd_issue[i] =
3053 dd->port->mmio + i*0x80 + PORT_COMMAND_ISSUE;
3054 dd->port->completed[i] =
3055 dd->port->mmio + i*0x80 + PORT_SDBV;
3058 timetaken = jiffies;
3059 timeout = jiffies + msecs_to_jiffies(30000);
3060 while (((readl(dd->port->mmio + PORT_SCR_STAT) & 0x0F) != 0x03) &&
3061 time_before(jiffies, timeout)) {
3064 if (unlikely(mtip_check_surprise_removal(dd->pdev))) {
3065 timetaken = jiffies - timetaken;
3066 dev_warn(&dd->pdev->dev,
3067 "Surprise removal detected at %u ms\n",
3068 jiffies_to_msecs(timetaken));
3072 if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag))) {
3073 timetaken = jiffies - timetaken;
3074 dev_warn(&dd->pdev->dev,
3075 "Removal detected at %u ms\n",
3076 jiffies_to_msecs(timetaken));
3081 /* Conditionally reset the HBA. */
3082 if (!(readl(dd->mmio + HOST_CAP) & HOST_CAP_NZDMA)) {
3083 if (mtip_hba_reset(dd) < 0) {
3084 dev_err(&dd->pdev->dev,
3085 "Card did not reset within timeout\n");
3090 /* Clear any pending interrupts on the HBA */
3091 writel(readl(dd->mmio + HOST_IRQ_STAT),
3092 dd->mmio + HOST_IRQ_STAT);
3095 mtip_init_port(dd->port);
3096 mtip_start_port(dd->port);
3098 /* Setup the ISR and enable interrupts. */
3099 rv = devm_request_irq(&dd->pdev->dev,
3103 dev_driver_string(&dd->pdev->dev),
3107 dev_err(&dd->pdev->dev,
3108 "Unable to allocate IRQ %d\n", dd->pdev->irq);
3111 irq_set_affinity_hint(dd->pdev->irq, get_cpu_mask(dd->isr_binding));
3113 /* Enable interrupts on the HBA. */
3114 writel(readl(dd->mmio + HOST_CTL) | HOST_IRQ_EN,
3115 dd->mmio + HOST_CTL);
3117 init_waitqueue_head(&dd->port->svc_wait);
3119 if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)) {
3127 /* Disable interrupts on the HBA. */
3128 writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
3129 dd->mmio + HOST_CTL);
3131 /* Release the IRQ. */
3132 irq_set_affinity_hint(dd->pdev->irq, NULL);
3133 devm_free_irq(&dd->pdev->dev, dd->pdev->irq, dd);
3136 mtip_deinit_port(dd->port);
3140 /* Free the memory allocated for the for structure. */
3146 static int mtip_standby_drive(struct driver_data *dd)
3150 if (dd->sr || !dd->port)
3153 * Send standby immediate (E0h) to the drive so that it
3156 if (!test_bit(MTIP_PF_REBUILD_BIT, &dd->port->flags) &&
3157 !test_bit(MTIP_DDF_REBUILD_FAILED_BIT, &dd->dd_flag) &&
3158 !test_bit(MTIP_DDF_SEC_LOCK_BIT, &dd->dd_flag)) {
3159 rv = mtip_standby_immediate(dd->port);
3161 dev_warn(&dd->pdev->dev,
3162 "STANDBY IMMEDIATE failed\n");
3168 * Called to deinitialize an interface.
3170 * @dd Pointer to the driver data structure.
3175 static int mtip_hw_exit(struct driver_data *dd)
3178 /* de-initialize the port. */
3179 mtip_deinit_port(dd->port);
3181 /* Disable interrupts on the HBA. */
3182 writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
3183 dd->mmio + HOST_CTL);
3186 /* Release the IRQ. */
3187 irq_set_affinity_hint(dd->pdev->irq, NULL);
3188 devm_free_irq(&dd->pdev->dev, dd->pdev->irq, dd);
3191 /* Free dma regions */
3194 /* Free the memory allocated for the for structure. */
3202 * Issue a Standby Immediate command to the device.
3204 * This function is called by the Block Layer just before the
3205 * system powers off during a shutdown.
3207 * @dd Pointer to the driver data structure.
3212 static int mtip_hw_shutdown(struct driver_data *dd)
3215 * Send standby immediate (E0h) to the drive so that it
3218 mtip_standby_drive(dd);
3226 * This function is called by the Block Layer just before the
3227 * system hibernates.
3229 * @dd Pointer to the driver data structure.
3232 * 0 Suspend was successful
3233 * -EFAULT Suspend was not successful
3235 static int mtip_hw_suspend(struct driver_data *dd)
3238 * Send standby immediate (E0h) to the drive
3239 * so that it saves its state.
3241 if (mtip_standby_drive(dd) != 0) {
3242 dev_err(&dd->pdev->dev,
3243 "Failed standby-immediate command\n");
3247 /* Disable interrupts on the HBA.*/
3248 writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
3249 dd->mmio + HOST_CTL);
3250 mtip_deinit_port(dd->port);
3258 * This function is called by the Block Layer as the
3261 * @dd Pointer to the driver data structure.
3264 * 0 Resume was successful
3265 * -EFAULT Resume was not successful
3267 static int mtip_hw_resume(struct driver_data *dd)
3269 /* Perform any needed hardware setup steps */
3273 if (mtip_hba_reset(dd) != 0) {
3274 dev_err(&dd->pdev->dev,
3275 "Unable to reset the HBA\n");
3280 * Enable the port, DMA engine, and FIS reception specific
3281 * h/w in controller.
3283 mtip_init_port(dd->port);
3284 mtip_start_port(dd->port);
3286 /* Enable interrupts on the HBA.*/
3287 writel(readl(dd->mmio + HOST_CTL) | HOST_IRQ_EN,
3288 dd->mmio + HOST_CTL);
3294 * Helper function for reusing disk name
3295 * upon hot insertion.
3297 static int rssd_disk_name_format(char *prefix,
3302 const int base = 'z' - 'a' + 1;
3303 char *begin = buf + strlen(prefix);
3304 char *end = buf + buflen;
3314 *--p = 'a' + (index % unit);
3315 index = (index / unit) - 1;
3316 } while (index >= 0);
3318 memmove(begin, p, end - p);
3319 memcpy(buf, prefix, strlen(prefix));
3325 * Block layer IOCTL handler.
3327 * @dev Pointer to the block_device structure.
3329 * @cmd IOCTL command passed from the user application.
3330 * @arg Argument passed from the user application.
3333 * 0 IOCTL completed successfully.
3334 * -ENOTTY IOCTL not supported or invalid driver data
3335 * structure pointer.
3337 static int mtip_block_ioctl(struct block_device *dev,
3342 struct driver_data *dd = dev->bd_disk->private_data;
3344 if (!capable(CAP_SYS_ADMIN))
3350 if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)))
3357 return mtip_hw_ioctl(dd, cmd, arg);
3361 #ifdef CONFIG_COMPAT
3363 * Block layer compat IOCTL handler.
3365 * @dev Pointer to the block_device structure.
3367 * @cmd IOCTL command passed from the user application.
3368 * @arg Argument passed from the user application.
3371 * 0 IOCTL completed successfully.
3372 * -ENOTTY IOCTL not supported or invalid driver data
3373 * structure pointer.
3375 static int mtip_block_compat_ioctl(struct block_device *dev,
3380 struct driver_data *dd = dev->bd_disk->private_data;
3382 if (!capable(CAP_SYS_ADMIN))
3388 if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)))
3394 case HDIO_DRIVE_TASKFILE: {
3395 struct mtip_compat_ide_task_request_s __user *compat_req_task;
3396 ide_task_request_t req_task;
3397 int compat_tasksize, outtotal, ret;
3400 sizeof(struct mtip_compat_ide_task_request_s);
3403 (struct mtip_compat_ide_task_request_s __user *) arg;
3405 if (copy_from_user(&req_task, (void __user *) arg,
3406 compat_tasksize - (2 * sizeof(compat_long_t))))
3409 if (get_user(req_task.out_size, &compat_req_task->out_size))
3412 if (get_user(req_task.in_size, &compat_req_task->in_size))
3415 outtotal = sizeof(struct mtip_compat_ide_task_request_s);
3417 ret = exec_drive_taskfile(dd, (void __user *) arg,
3418 &req_task, outtotal);
3420 if (copy_to_user((void __user *) arg, &req_task,
3422 (2 * sizeof(compat_long_t))))
3425 if (put_user(req_task.out_size, &compat_req_task->out_size))
3428 if (put_user(req_task.in_size, &compat_req_task->in_size))
3434 return mtip_hw_ioctl(dd, cmd, arg);
3440 * Obtain the geometry of the device.
3442 * You may think that this function is obsolete, but some applications,
3443 * fdisk for example still used CHS values. This function describes the
3444 * device as having 224 heads and 56 sectors per cylinder. These values are
3445 * chosen so that each cylinder is aligned on a 4KB boundary. Since a
3446 * partition is described in terms of a start and end cylinder this means
3447 * that each partition is also 4KB aligned. Non-aligned partitions adversely
3448 * affects performance.
3450 * @dev Pointer to the block_device strucutre.
3451 * @geo Pointer to a hd_geometry structure.
3454 * 0 Operation completed successfully.
3455 * -ENOTTY An error occurred while reading the drive capacity.
3457 static int mtip_block_getgeo(struct block_device *dev,
3458 struct hd_geometry *geo)
3460 struct driver_data *dd = dev->bd_disk->private_data;
3466 if (!(mtip_hw_get_capacity(dd, &capacity))) {
3467 dev_warn(&dd->pdev->dev,
3468 "Could not get drive capacity.\n");
3474 sector_div(capacity, (geo->heads * geo->sectors));
3475 geo->cylinders = capacity;
3479 static int mtip_block_open(struct block_device *dev, fmode_t mode)
3481 struct driver_data *dd;
3483 if (dev && dev->bd_disk) {
3484 dd = (struct driver_data *) dev->bd_disk->private_data;
3487 if (test_bit(MTIP_DDF_REMOVAL_BIT,
3497 static void mtip_block_release(struct gendisk *disk, fmode_t mode)
3502 * Block device operation function.
3504 * This structure contains pointers to the functions required by the block
3507 static const struct block_device_operations mtip_block_ops = {
3508 .open = mtip_block_open,
3509 .release = mtip_block_release,
3510 .ioctl = mtip_block_ioctl,
3511 #ifdef CONFIG_COMPAT
3512 .compat_ioctl = mtip_block_compat_ioctl,
3514 .getgeo = mtip_block_getgeo,
3515 .owner = THIS_MODULE
3518 static inline bool is_se_active(struct driver_data *dd)
3520 if (unlikely(test_bit(MTIP_PF_SE_ACTIVE_BIT, &dd->port->flags))) {
3521 if (dd->port->ic_pause_timer) {
3522 unsigned long to = dd->port->ic_pause_timer +
3523 msecs_to_jiffies(1000);
3524 if (time_after(jiffies, to)) {
3525 clear_bit(MTIP_PF_SE_ACTIVE_BIT,
3527 clear_bit(MTIP_DDF_SEC_LOCK_BIT, &dd->dd_flag);
3528 dd->port->ic_pause_timer = 0;
3529 wake_up_interruptible(&dd->port->svc_wait);
3539 * Block layer make request function.
3541 * This function is called by the kernel to process a BIO for
3544 * @queue Pointer to the request queue. Unused other than to obtain
3545 * the driver data structure.
3546 * @rq Pointer to the request.
3549 static int mtip_submit_request(struct blk_mq_hw_ctx *hctx, struct request *rq)
3551 struct driver_data *dd = hctx->queue->queuedata;
3552 struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3555 if (is_se_active(dd))
3558 if (unlikely(dd->dd_flag & MTIP_DDF_STOP_IO)) {
3559 if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
3563 if (unlikely(test_bit(MTIP_DDF_OVER_TEMP_BIT, &dd->dd_flag))) {
3566 if (unlikely(test_bit(MTIP_DDF_WRITE_PROTECT_BIT,
3571 if (unlikely(test_bit(MTIP_DDF_SEC_LOCK_BIT, &dd->dd_flag) ||
3572 test_bit(MTIP_DDF_REBUILD_FAILED_BIT, &dd->dd_flag)))
3576 if (req_op(rq) == REQ_OP_DISCARD) {
3579 err = mtip_send_trim(dd, blk_rq_pos(rq), blk_rq_sectors(rq));
3580 blk_mq_end_request(rq, err ? BLK_STS_IOERR : BLK_STS_OK);
3584 /* Create the scatter list for this request. */
3585 nents = blk_rq_map_sg(hctx->queue, rq, cmd->sg);
3587 /* Issue the read/write. */
3588 mtip_hw_submit_io(dd, rq, cmd, nents, hctx);
3592 static bool mtip_check_unal_depth(struct blk_mq_hw_ctx *hctx,
3595 struct driver_data *dd = hctx->queue->queuedata;
3596 struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3598 if (rq_data_dir(rq) == READ || !dd->unal_qdepth)
3602 * If unaligned depth must be limited on this controller, mark it
3603 * as unaligned if the IO isn't on a 4k boundary (start of length).
3605 if (blk_rq_sectors(rq) <= 64) {
3606 if ((blk_rq_pos(rq) & 7) || (blk_rq_sectors(rq) & 7))
3610 if (cmd->unaligned && down_trylock(&dd->port->cmd_slot_unal))
3616 static blk_status_t mtip_issue_reserved_cmd(struct blk_mq_hw_ctx *hctx,
3619 struct driver_data *dd = hctx->queue->queuedata;
3620 struct mtip_int_cmd *icmd = rq->special;
3621 struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3622 struct mtip_cmd_sg *command_sg;
3624 if (mtip_commands_active(dd->port))
3625 return BLK_STS_RESOURCE;
3627 /* Populate the SG list */
3628 cmd->command_header->opts =
3629 __force_bit2int cpu_to_le32(icmd->opts | icmd->fis_len);
3630 if (icmd->buf_len) {
3631 command_sg = cmd->command + AHCI_CMD_TBL_HDR_SZ;
3634 __force_bit2int cpu_to_le32((icmd->buf_len-1) & 0x3FFFFF);
3636 __force_bit2int cpu_to_le32(icmd->buffer & 0xFFFFFFFF);
3637 command_sg->dba_upper =
3638 __force_bit2int cpu_to_le32((icmd->buffer >> 16) >> 16);
3640 cmd->command_header->opts |=
3641 __force_bit2int cpu_to_le32((1 << 16));
3644 /* Populate the command header */
3645 cmd->command_header->byte_count = 0;
3647 blk_mq_start_request(rq);
3648 mtip_issue_non_ncq_command(dd->port, rq->tag);
3652 static blk_status_t mtip_queue_rq(struct blk_mq_hw_ctx *hctx,
3653 const struct blk_mq_queue_data *bd)
3655 struct request *rq = bd->rq;
3658 mtip_init_cmd_header(rq);
3660 if (blk_rq_is_passthrough(rq))
3661 return mtip_issue_reserved_cmd(hctx, rq);
3663 if (unlikely(mtip_check_unal_depth(hctx, rq)))
3664 return BLK_STS_RESOURCE;
3666 blk_mq_start_request(rq);
3668 ret = mtip_submit_request(hctx, rq);
3671 return BLK_STS_IOERR;
3674 static void mtip_free_cmd(struct blk_mq_tag_set *set, struct request *rq,
3675 unsigned int hctx_idx)
3677 struct driver_data *dd = set->driver_data;
3678 struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3683 dmam_free_coherent(&dd->pdev->dev, CMD_DMA_ALLOC_SZ,
3684 cmd->command, cmd->command_dma);
3687 static int mtip_init_cmd(struct blk_mq_tag_set *set, struct request *rq,
3688 unsigned int hctx_idx, unsigned int numa_node)
3690 struct driver_data *dd = set->driver_data;
3691 struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3693 cmd->command = dmam_alloc_coherent(&dd->pdev->dev, CMD_DMA_ALLOC_SZ,
3694 &cmd->command_dma, GFP_KERNEL);
3698 memset(cmd->command, 0, CMD_DMA_ALLOC_SZ);
3700 sg_init_table(cmd->sg, MTIP_MAX_SG);
3704 static enum blk_eh_timer_return mtip_cmd_timeout(struct request *req,
3707 struct driver_data *dd = req->q->queuedata;
3710 struct mtip_cmd *cmd = blk_mq_rq_to_pdu(req);
3712 cmd->status = BLK_STS_TIMEOUT;
3713 blk_mq_complete_request(req);
3717 if (test_bit(req->tag, dd->port->cmds_to_issue))
3720 if (test_and_set_bit(MTIP_PF_TO_ACTIVE_BIT, &dd->port->flags))
3723 wake_up_interruptible(&dd->port->svc_wait);
3725 return BLK_EH_RESET_TIMER;
3728 static const struct blk_mq_ops mtip_mq_ops = {
3729 .queue_rq = mtip_queue_rq,
3730 .init_request = mtip_init_cmd,
3731 .exit_request = mtip_free_cmd,
3732 .complete = mtip_softirq_done_fn,
3733 .timeout = mtip_cmd_timeout,
3737 * Block layer initialization function.
3739 * This function is called once by the PCI layer for each P320
3740 * device that is connected to the system.
3742 * @dd Pointer to the driver data structure.
3745 * 0 on success else an error code.
3747 static int mtip_block_initialize(struct driver_data *dd)
3749 int rv = 0, wait_for_rebuild = 0;
3751 unsigned int index = 0;
3752 struct kobject *kobj;
3755 goto skip_create_disk; /* hw init done, before rebuild */
3757 if (mtip_hw_init(dd)) {
3759 goto protocol_init_error;
3762 dd->disk = alloc_disk_node(MTIP_MAX_MINORS, dd->numa_node);
3763 if (dd->disk == NULL) {
3764 dev_err(&dd->pdev->dev,
3765 "Unable to allocate gendisk structure\n");
3767 goto alloc_disk_error;
3770 /* Generate the disk name, implemented same as in sd.c */
3772 if (!ida_pre_get(&rssd_index_ida, GFP_KERNEL)) {
3777 spin_lock(&rssd_index_lock);
3778 rv = ida_get_new(&rssd_index_ida, &index);
3779 spin_unlock(&rssd_index_lock);
3780 } while (rv == -EAGAIN);
3785 rv = rssd_disk_name_format("rssd",
3787 dd->disk->disk_name,
3790 goto disk_index_error;
3792 dd->disk->major = dd->major;
3793 dd->disk->first_minor = index * MTIP_MAX_MINORS;
3794 dd->disk->minors = MTIP_MAX_MINORS;
3795 dd->disk->fops = &mtip_block_ops;
3796 dd->disk->private_data = dd;
3799 mtip_hw_debugfs_init(dd);
3801 memset(&dd->tags, 0, sizeof(dd->tags));
3802 dd->tags.ops = &mtip_mq_ops;
3803 dd->tags.nr_hw_queues = 1;
3804 dd->tags.queue_depth = MTIP_MAX_COMMAND_SLOTS;
3805 dd->tags.reserved_tags = 1;
3806 dd->tags.cmd_size = sizeof(struct mtip_cmd);
3807 dd->tags.numa_node = dd->numa_node;
3808 dd->tags.flags = BLK_MQ_F_SHOULD_MERGE;
3809 dd->tags.driver_data = dd;
3810 dd->tags.timeout = MTIP_NCQ_CMD_TIMEOUT_MS;
3812 rv = blk_mq_alloc_tag_set(&dd->tags);
3814 dev_err(&dd->pdev->dev,
3815 "Unable to allocate request queue\n");
3816 goto block_queue_alloc_tag_error;
3819 /* Allocate the request queue. */
3820 dd->queue = blk_mq_init_queue(&dd->tags);
3821 if (IS_ERR(dd->queue)) {
3822 dev_err(&dd->pdev->dev,
3823 "Unable to allocate request queue\n");
3825 goto block_queue_alloc_init_error;
3828 dd->disk->queue = dd->queue;
3829 dd->queue->queuedata = dd;
3832 /* Initialize the protocol layer. */
3833 wait_for_rebuild = mtip_hw_get_identify(dd);
3834 if (wait_for_rebuild < 0) {
3835 dev_err(&dd->pdev->dev,
3836 "Protocol layer initialization failed\n");
3838 goto init_hw_cmds_error;
3842 * if rebuild pending, start the service thread, and delay the block
3843 * queue creation and device_add_disk()
3845 if (wait_for_rebuild == MTIP_FTL_REBUILD_MAGIC)
3846 goto start_service_thread;
3848 /* Set device limits. */
3849 blk_queue_flag_set(QUEUE_FLAG_NONROT, dd->queue);
3850 blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, dd->queue);
3851 blk_queue_max_segments(dd->queue, MTIP_MAX_SG);
3852 blk_queue_physical_block_size(dd->queue, 4096);
3853 blk_queue_max_hw_sectors(dd->queue, 0xffff);
3854 blk_queue_max_segment_size(dd->queue, 0x400000);
3855 blk_queue_io_min(dd->queue, 4096);
3857 /* Signal trim support */
3858 if (dd->trim_supp == true) {
3859 blk_queue_flag_set(QUEUE_FLAG_DISCARD, dd->queue);
3860 dd->queue->limits.discard_granularity = 4096;
3861 blk_queue_max_discard_sectors(dd->queue,
3862 MTIP_MAX_TRIM_ENTRY_LEN * MTIP_MAX_TRIM_ENTRIES);
3865 /* Set the capacity of the device in 512 byte sectors. */
3866 if (!(mtip_hw_get_capacity(dd, &capacity))) {
3867 dev_warn(&dd->pdev->dev,
3868 "Could not read drive capacity\n");
3870 goto read_capacity_error;
3872 set_capacity(dd->disk, capacity);
3874 /* Enable the block device and add it to /dev */
3875 device_add_disk(&dd->pdev->dev, dd->disk);
3877 dd->bdev = bdget_disk(dd->disk, 0);
3879 * Now that the disk is active, initialize any sysfs attributes
3880 * managed by the protocol layer.
3882 kobj = kobject_get(&disk_to_dev(dd->disk)->kobj);
3884 mtip_hw_sysfs_init(dd, kobj);
3888 if (dd->mtip_svc_handler) {
3889 set_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag);
3890 return rv; /* service thread created for handling rebuild */
3893 start_service_thread:
3894 dd->mtip_svc_handler = kthread_create_on_node(mtip_service_thread,
3896 "mtip_svc_thd_%02d", index);
3898 if (IS_ERR(dd->mtip_svc_handler)) {
3899 dev_err(&dd->pdev->dev, "service thread failed to start\n");
3900 dd->mtip_svc_handler = NULL;
3902 goto kthread_run_error;
3904 wake_up_process(dd->mtip_svc_handler);
3905 if (wait_for_rebuild == MTIP_FTL_REBUILD_MAGIC)
3906 rv = wait_for_rebuild;
3914 /* Delete our gendisk. This also removes the device from /dev */
3915 del_gendisk(dd->disk);
3917 read_capacity_error:
3919 blk_cleanup_queue(dd->queue);
3920 block_queue_alloc_init_error:
3921 blk_mq_free_tag_set(&dd->tags);
3922 block_queue_alloc_tag_error:
3923 mtip_hw_debugfs_exit(dd);
3925 spin_lock(&rssd_index_lock);
3926 ida_remove(&rssd_index_ida, index);
3927 spin_unlock(&rssd_index_lock);
3933 mtip_hw_exit(dd); /* De-initialize the protocol layer. */
3935 protocol_init_error:
3939 static void mtip_no_dev_cleanup(struct request *rq, void *data, bool reserv)
3941 struct mtip_cmd *cmd = blk_mq_rq_to_pdu(rq);
3943 cmd->status = BLK_STS_IOERR;
3944 blk_mq_complete_request(rq);
3948 * Block layer deinitialization function.
3950 * Called by the PCI layer as each P320 device is removed.
3952 * @dd Pointer to the driver data structure.
3957 static int mtip_block_remove(struct driver_data *dd)
3959 struct kobject *kobj;
3961 mtip_hw_debugfs_exit(dd);
3963 if (dd->mtip_svc_handler) {
3964 set_bit(MTIP_PF_SVC_THD_STOP_BIT, &dd->port->flags);
3965 wake_up_interruptible(&dd->port->svc_wait);
3966 kthread_stop(dd->mtip_svc_handler);
3969 /* Clean up the sysfs attributes, if created */
3970 if (test_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag)) {
3971 kobj = kobject_get(&disk_to_dev(dd->disk)->kobj);
3973 mtip_hw_sysfs_exit(dd, kobj);
3980 * Explicitly wait here for IOs to quiesce,
3981 * as mtip_standby_drive usually won't wait for IOs.
3983 if (!mtip_quiesce_io(dd->port, MTIP_QUIESCE_IO_TIMEOUT_MS))
3984 mtip_standby_drive(dd);
3987 dev_info(&dd->pdev->dev, "device %s surprise removal\n",
3988 dd->disk->disk_name);
3990 blk_freeze_queue_start(dd->queue);
3991 blk_mq_quiesce_queue(dd->queue);
3992 blk_mq_tagset_busy_iter(&dd->tags, mtip_no_dev_cleanup, dd);
3993 blk_mq_unquiesce_queue(dd->queue);
3996 * Delete our gendisk structure. This also removes the device
4004 if (test_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag))
4005 del_gendisk(dd->disk);
4006 if (dd->disk->queue) {
4007 blk_cleanup_queue(dd->queue);
4008 blk_mq_free_tag_set(&dd->tags);
4015 spin_lock(&rssd_index_lock);
4016 ida_remove(&rssd_index_ida, dd->index);
4017 spin_unlock(&rssd_index_lock);
4019 /* De-initialize the protocol layer. */
4026 * Function called by the PCI layer when just before the
4027 * machine shuts down.
4029 * If a protocol layer shutdown function is present it will be called
4032 * @dd Pointer to the driver data structure.
4037 static int mtip_block_shutdown(struct driver_data *dd)
4039 mtip_hw_shutdown(dd);
4041 /* Delete our gendisk structure, and cleanup the blk queue. */
4043 dev_info(&dd->pdev->dev,
4044 "Shutting down %s ...\n", dd->disk->disk_name);
4046 if (test_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag))
4047 del_gendisk(dd->disk);
4048 if (dd->disk->queue) {
4049 blk_cleanup_queue(dd->queue);
4050 blk_mq_free_tag_set(&dd->tags);
4057 spin_lock(&rssd_index_lock);
4058 ida_remove(&rssd_index_ida, dd->index);
4059 spin_unlock(&rssd_index_lock);
4063 static int mtip_block_suspend(struct driver_data *dd)
4065 dev_info(&dd->pdev->dev,
4066 "Suspending %s ...\n", dd->disk->disk_name);
4067 mtip_hw_suspend(dd);
4071 static int mtip_block_resume(struct driver_data *dd)
4073 dev_info(&dd->pdev->dev, "Resuming %s ...\n",
4074 dd->disk->disk_name);
4079 static void drop_cpu(int cpu)
4084 static int get_least_used_cpu_on_node(int node)
4086 int cpu, least_used_cpu, least_cnt;
4087 const struct cpumask *node_mask;
4089 node_mask = cpumask_of_node(node);
4090 least_used_cpu = cpumask_first(node_mask);
4091 least_cnt = cpu_use[least_used_cpu];
4092 cpu = least_used_cpu;
4094 for_each_cpu(cpu, node_mask) {
4095 if (cpu_use[cpu] < least_cnt) {
4096 least_used_cpu = cpu;
4097 least_cnt = cpu_use[cpu];
4100 cpu_use[least_used_cpu]++;
4101 return least_used_cpu;
4104 /* Helper for selecting a node in round robin mode */
4105 static inline int mtip_get_next_rr_node(void)
4107 static int next_node = -1;
4109 if (next_node == -1) {
4110 next_node = first_online_node;
4114 next_node = next_online_node(next_node);
4115 if (next_node == MAX_NUMNODES)
4116 next_node = first_online_node;
4120 static DEFINE_HANDLER(0);
4121 static DEFINE_HANDLER(1);
4122 static DEFINE_HANDLER(2);
4123 static DEFINE_HANDLER(3);
4124 static DEFINE_HANDLER(4);
4125 static DEFINE_HANDLER(5);
4126 static DEFINE_HANDLER(6);
4127 static DEFINE_HANDLER(7);
4129 static void mtip_disable_link_opts(struct driver_data *dd, struct pci_dev *pdev)
4132 unsigned short pcie_dev_ctrl;
4134 pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
4136 pci_read_config_word(pdev,
4137 pos + PCI_EXP_DEVCTL,
4139 if (pcie_dev_ctrl & (1 << 11) ||
4140 pcie_dev_ctrl & (1 << 4)) {
4141 dev_info(&dd->pdev->dev,
4142 "Disabling ERO/No-Snoop on bridge device %04x:%04x\n",
4143 pdev->vendor, pdev->device);
4144 pcie_dev_ctrl &= ~(PCI_EXP_DEVCTL_NOSNOOP_EN |
4145 PCI_EXP_DEVCTL_RELAX_EN);
4146 pci_write_config_word(pdev,
4147 pos + PCI_EXP_DEVCTL,
4153 static void mtip_fix_ero_nosnoop(struct driver_data *dd, struct pci_dev *pdev)
4156 * This workaround is specific to AMD/ATI chipset with a PCI upstream
4157 * device with device id 0x5aXX
4159 if (pdev->bus && pdev->bus->self) {
4160 if (pdev->bus->self->vendor == PCI_VENDOR_ID_ATI &&
4161 ((pdev->bus->self->device & 0xff00) == 0x5a00)) {
4162 mtip_disable_link_opts(dd, pdev->bus->self);
4164 /* Check further up the topology */
4165 struct pci_dev *parent_dev = pdev->bus->self;
4166 if (parent_dev->bus &&
4167 parent_dev->bus->parent &&
4168 parent_dev->bus->parent->self &&
4169 parent_dev->bus->parent->self->vendor ==
4170 PCI_VENDOR_ID_ATI &&
4171 (parent_dev->bus->parent->self->device &
4172 0xff00) == 0x5a00) {
4173 mtip_disable_link_opts(dd,
4174 parent_dev->bus->parent->self);
4181 * Called for each supported PCI device detected.
4183 * This function allocates the private data structure, enables the
4184 * PCI device and then calls the block layer initialization function.
4187 * 0 on success else an error code.
4189 static int mtip_pci_probe(struct pci_dev *pdev,
4190 const struct pci_device_id *ent)
4193 struct driver_data *dd = NULL;
4195 const struct cpumask *node_mask;
4196 int cpu, i = 0, j = 0;
4197 int my_node = NUMA_NO_NODE;
4198 unsigned long flags;
4200 /* Allocate memory for this devices private data. */
4201 my_node = pcibus_to_node(pdev->bus);
4202 if (my_node != NUMA_NO_NODE) {
4203 if (!node_online(my_node))
4204 my_node = mtip_get_next_rr_node();
4206 dev_info(&pdev->dev, "Kernel not reporting proximity, choosing a node\n");
4207 my_node = mtip_get_next_rr_node();
4209 dev_info(&pdev->dev, "NUMA node %d (closest: %d,%d, probe on %d:%d)\n",
4210 my_node, pcibus_to_node(pdev->bus), dev_to_node(&pdev->dev),
4211 cpu_to_node(raw_smp_processor_id()), raw_smp_processor_id());
4213 dd = kzalloc_node(sizeof(struct driver_data), GFP_KERNEL, my_node);
4216 "Unable to allocate memory for driver data\n");
4220 /* Attach the private data to this PCI device. */
4221 pci_set_drvdata(pdev, dd);
4223 rv = pcim_enable_device(pdev);
4225 dev_err(&pdev->dev, "Unable to enable device\n");
4229 /* Map BAR5 to memory. */
4230 rv = pcim_iomap_regions(pdev, 1 << MTIP_ABAR, MTIP_DRV_NAME);
4232 dev_err(&pdev->dev, "Unable to map regions\n");
4236 if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(64))) {
4237 rv = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64));
4240 rv = pci_set_consistent_dma_mask(pdev,
4243 dev_warn(&pdev->dev,
4244 "64-bit DMA enable failed\n");
4250 /* Copy the info we may need later into the private data structure. */
4251 dd->major = mtip_major;
4252 dd->instance = instance;
4254 dd->numa_node = my_node;
4256 INIT_LIST_HEAD(&dd->online_list);
4257 INIT_LIST_HEAD(&dd->remove_list);
4259 memset(dd->workq_name, 0, 32);
4260 snprintf(dd->workq_name, 31, "mtipq%d", dd->instance);
4262 dd->isr_workq = create_workqueue(dd->workq_name);
4263 if (!dd->isr_workq) {
4264 dev_warn(&pdev->dev, "Can't create wq %d\n", dd->instance);
4269 memset(cpu_list, 0, sizeof(cpu_list));
4271 node_mask = cpumask_of_node(dd->numa_node);
4272 if (!cpumask_empty(node_mask)) {
4273 for_each_cpu(cpu, node_mask)
4275 snprintf(&cpu_list[j], 256 - j, "%d ", cpu);
4276 j = strlen(cpu_list);
4279 dev_info(&pdev->dev, "Node %d on package %d has %d cpu(s): %s\n",
4281 topology_physical_package_id(cpumask_first(node_mask)),
4282 nr_cpus_node(dd->numa_node),
4285 dev_dbg(&pdev->dev, "mtip32xx: node_mask empty\n");
4287 dd->isr_binding = get_least_used_cpu_on_node(dd->numa_node);
4288 dev_info(&pdev->dev, "Initial IRQ binding node:cpu %d:%d\n",
4289 cpu_to_node(dd->isr_binding), dd->isr_binding);
4291 /* first worker context always runs in ISR */
4292 dd->work[0].cpu_binding = dd->isr_binding;
4293 dd->work[1].cpu_binding = get_least_used_cpu_on_node(dd->numa_node);
4294 dd->work[2].cpu_binding = get_least_used_cpu_on_node(dd->numa_node);
4295 dd->work[3].cpu_binding = dd->work[0].cpu_binding;
4296 dd->work[4].cpu_binding = dd->work[1].cpu_binding;
4297 dd->work[5].cpu_binding = dd->work[2].cpu_binding;
4298 dd->work[6].cpu_binding = dd->work[2].cpu_binding;
4299 dd->work[7].cpu_binding = dd->work[1].cpu_binding;
4301 /* Log the bindings */
4302 for_each_present_cpu(cpu) {
4303 memset(cpu_list, 0, sizeof(cpu_list));
4304 for (i = 0, j = 0; i < MTIP_MAX_SLOT_GROUPS; i++) {
4305 if (dd->work[i].cpu_binding == cpu) {
4306 snprintf(&cpu_list[j], 256 - j, "%d ", i);
4307 j = strlen(cpu_list);
4311 dev_info(&pdev->dev, "CPU %d: WQs %s\n", cpu, cpu_list);
4314 INIT_WORK(&dd->work[0].work, mtip_workq_sdbf0);
4315 INIT_WORK(&dd->work[1].work, mtip_workq_sdbf1);
4316 INIT_WORK(&dd->work[2].work, mtip_workq_sdbf2);
4317 INIT_WORK(&dd->work[3].work, mtip_workq_sdbf3);
4318 INIT_WORK(&dd->work[4].work, mtip_workq_sdbf4);
4319 INIT_WORK(&dd->work[5].work, mtip_workq_sdbf5);
4320 INIT_WORK(&dd->work[6].work, mtip_workq_sdbf6);
4321 INIT_WORK(&dd->work[7].work, mtip_workq_sdbf7);
4323 pci_set_master(pdev);
4324 rv = pci_enable_msi(pdev);
4326 dev_warn(&pdev->dev,
4327 "Unable to enable MSI interrupt.\n");
4328 goto msi_initialize_err;
4331 mtip_fix_ero_nosnoop(dd, pdev);
4333 /* Initialize the block layer. */
4334 rv = mtip_block_initialize(dd);
4337 "Unable to initialize block layer\n");
4338 goto block_initialize_err;
4342 * Increment the instance count so that each device has a unique
4346 if (rv != MTIP_FTL_REBUILD_MAGIC)
4347 set_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag);
4349 rv = 0; /* device in rebuild state, return 0 from probe */
4351 /* Add to online list even if in ftl rebuild */
4352 spin_lock_irqsave(&dev_lock, flags);
4353 list_add(&dd->online_list, &online_list);
4354 spin_unlock_irqrestore(&dev_lock, flags);
4358 block_initialize_err:
4359 pci_disable_msi(pdev);
4362 if (dd->isr_workq) {
4363 flush_workqueue(dd->isr_workq);
4364 destroy_workqueue(dd->isr_workq);
4365 drop_cpu(dd->work[0].cpu_binding);
4366 drop_cpu(dd->work[1].cpu_binding);
4367 drop_cpu(dd->work[2].cpu_binding);
4370 pcim_iounmap_regions(pdev, 1 << MTIP_ABAR);
4374 pci_set_drvdata(pdev, NULL);
4381 * Called for each probed device when the device is removed or the
4382 * driver is unloaded.
4387 static void mtip_pci_remove(struct pci_dev *pdev)
4389 struct driver_data *dd = pci_get_drvdata(pdev);
4390 unsigned long flags, to;
4392 set_bit(MTIP_DDF_REMOVAL_BIT, &dd->dd_flag);
4394 spin_lock_irqsave(&dev_lock, flags);
4395 list_del_init(&dd->online_list);
4396 list_add(&dd->remove_list, &removing_list);
4397 spin_unlock_irqrestore(&dev_lock, flags);
4399 mtip_check_surprise_removal(pdev);
4400 synchronize_irq(dd->pdev->irq);
4402 /* Spin until workers are done */
4403 to = jiffies + msecs_to_jiffies(4000);
4406 } while (atomic_read(&dd->irq_workers_active) != 0 &&
4407 time_before(jiffies, to));
4410 fsync_bdev(dd->bdev);
4412 if (atomic_read(&dd->irq_workers_active) != 0) {
4413 dev_warn(&dd->pdev->dev,
4414 "Completion workers still active!\n");
4417 blk_set_queue_dying(dd->queue);
4418 set_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag);
4420 /* Clean up the block layer. */
4421 mtip_block_remove(dd);
4423 if (dd->isr_workq) {
4424 flush_workqueue(dd->isr_workq);
4425 destroy_workqueue(dd->isr_workq);
4426 drop_cpu(dd->work[0].cpu_binding);
4427 drop_cpu(dd->work[1].cpu_binding);
4428 drop_cpu(dd->work[2].cpu_binding);
4431 pci_disable_msi(pdev);
4433 spin_lock_irqsave(&dev_lock, flags);
4434 list_del_init(&dd->remove_list);
4435 spin_unlock_irqrestore(&dev_lock, flags);
4439 pcim_iounmap_regions(pdev, 1 << MTIP_ABAR);
4440 pci_set_drvdata(pdev, NULL);
4444 * Called for each probed device when the device is suspended.
4450 static int mtip_pci_suspend(struct pci_dev *pdev, pm_message_t mesg)
4453 struct driver_data *dd = pci_get_drvdata(pdev);
4457 "Driver private datastructure is NULL\n");
4461 set_bit(MTIP_DDF_RESUME_BIT, &dd->dd_flag);
4463 /* Disable ports & interrupts then send standby immediate */
4464 rv = mtip_block_suspend(dd);
4467 "Failed to suspend controller\n");
4472 * Save the pci config space to pdev structure &
4473 * disable the device
4475 pci_save_state(pdev);
4476 pci_disable_device(pdev);
4478 /* Move to Low power state*/
4479 pci_set_power_state(pdev, PCI_D3hot);
4485 * Called for each probed device when the device is resumed.
4491 static int mtip_pci_resume(struct pci_dev *pdev)
4494 struct driver_data *dd;
4496 dd = pci_get_drvdata(pdev);
4499 "Driver private datastructure is NULL\n");
4503 /* Move the device to active State */
4504 pci_set_power_state(pdev, PCI_D0);
4506 /* Restore PCI configuration space */
4507 pci_restore_state(pdev);
4509 /* Enable the PCI device*/
4510 rv = pcim_enable_device(pdev);
4513 "Failed to enable card during resume\n");
4516 pci_set_master(pdev);
4519 * Calls hbaReset, initPort, & startPort function
4520 * then enables interrupts
4522 rv = mtip_block_resume(dd);
4524 dev_err(&pdev->dev, "Unable to resume\n");
4527 clear_bit(MTIP_DDF_RESUME_BIT, &dd->dd_flag);
4538 static void mtip_pci_shutdown(struct pci_dev *pdev)
4540 struct driver_data *dd = pci_get_drvdata(pdev);
4542 mtip_block_shutdown(dd);
4545 /* Table of device ids supported by this driver. */
4546 static const struct pci_device_id mtip_pci_tbl[] = {
4547 { PCI_DEVICE(PCI_VENDOR_ID_MICRON, P320H_DEVICE_ID) },
4548 { PCI_DEVICE(PCI_VENDOR_ID_MICRON, P320M_DEVICE_ID) },
4549 { PCI_DEVICE(PCI_VENDOR_ID_MICRON, P320S_DEVICE_ID) },
4550 { PCI_DEVICE(PCI_VENDOR_ID_MICRON, P325M_DEVICE_ID) },
4551 { PCI_DEVICE(PCI_VENDOR_ID_MICRON, P420H_DEVICE_ID) },
4552 { PCI_DEVICE(PCI_VENDOR_ID_MICRON, P420M_DEVICE_ID) },
4553 { PCI_DEVICE(PCI_VENDOR_ID_MICRON, P425M_DEVICE_ID) },
4557 /* Structure that describes the PCI driver functions. */
4558 static struct pci_driver mtip_pci_driver = {
4559 .name = MTIP_DRV_NAME,
4560 .id_table = mtip_pci_tbl,
4561 .probe = mtip_pci_probe,
4562 .remove = mtip_pci_remove,
4563 .suspend = mtip_pci_suspend,
4564 .resume = mtip_pci_resume,
4565 .shutdown = mtip_pci_shutdown,
4568 MODULE_DEVICE_TABLE(pci, mtip_pci_tbl);
4571 * Module initialization function.
4573 * Called once when the module is loaded. This function allocates a major
4574 * block device number to the Cyclone devices and registers the PCI layer
4578 * 0 on success else error code.
4580 static int __init mtip_init(void)
4584 pr_info(MTIP_DRV_NAME " Version " MTIP_DRV_VERSION "\n");
4586 spin_lock_init(&dev_lock);
4588 INIT_LIST_HEAD(&online_list);
4589 INIT_LIST_HEAD(&removing_list);
4591 /* Allocate a major block device number to use with this driver. */
4592 error = register_blkdev(0, MTIP_DRV_NAME);
4594 pr_err("Unable to register block device (%d)\n",
4600 dfs_parent = debugfs_create_dir("rssd", NULL);
4601 if (IS_ERR_OR_NULL(dfs_parent)) {
4602 pr_warn("Error creating debugfs parent\n");
4606 dfs_device_status = debugfs_create_file("device_status",
4607 0444, dfs_parent, NULL,
4608 &mtip_device_status_fops);
4609 if (IS_ERR_OR_NULL(dfs_device_status)) {
4610 pr_err("Error creating device_status node\n");
4611 dfs_device_status = NULL;
4615 /* Register our PCI operations. */
4616 error = pci_register_driver(&mtip_pci_driver);
4618 debugfs_remove(dfs_parent);
4619 unregister_blkdev(mtip_major, MTIP_DRV_NAME);
4626 * Module de-initialization function.
4628 * Called once when the module is unloaded. This function deallocates
4629 * the major block device number allocated by mtip_init() and
4630 * unregisters the PCI layer of the driver.
4635 static void __exit mtip_exit(void)
4637 /* Release the allocated major block device number. */
4638 unregister_blkdev(mtip_major, MTIP_DRV_NAME);
4640 /* Unregister the PCI driver. */
4641 pci_unregister_driver(&mtip_pci_driver);
4643 debugfs_remove_recursive(dfs_parent);
4646 MODULE_AUTHOR("Micron Technology, Inc");
4647 MODULE_DESCRIPTION("Micron RealSSD PCIe Block Driver");
4648 MODULE_LICENSE("GPL");
4649 MODULE_VERSION(MTIP_DRV_VERSION);
4651 module_init(mtip_init);
4652 module_exit(mtip_exit);