diff options
Diffstat (limited to 'drivers/scsi/hpsa.c')
| -rw-r--r-- | drivers/scsi/hpsa.c | 4227 |
1 files changed, 3619 insertions, 608 deletions
diff --git a/drivers/scsi/hpsa.c b/drivers/scsi/hpsa.c index b96962c3944..31184b35370 100644 --- a/drivers/scsi/hpsa.c +++ b/drivers/scsi/hpsa.c @@ -1,6 +1,6 @@ /* * Disk Array driver for HP Smart Array SAS controllers - * Copyright 2000, 2009 Hewlett-Packard Development Company, L.P. + * Copyright 2000, 2014 Hewlett-Packard Development Company, L.P. * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -29,7 +29,6 @@ #include <linux/delay.h> #include <linux/fs.h> #include <linux/timer.h> -#include <linux/seq_file.h> #include <linux/init.h> #include <linux/spinlock.h> #include <linux/compat.h> @@ -48,14 +47,16 @@ #include <linux/string.h> #include <linux/bitmap.h> #include <linux/atomic.h> -#include <linux/kthread.h> #include <linux/jiffies.h> +#include <linux/percpu.h> +#include <asm/div64.h> #include "hpsa_cmd.h" #include "hpsa.h" /* HPSA_DRIVER_VERSION must be 3 byte values (0-255) separated by '.' */ -#define HPSA_DRIVER_VERSION "2.0.2-1" +#define HPSA_DRIVER_VERSION "3.4.4-1" #define DRIVER_NAME "HP HPSA Driver (v " HPSA_DRIVER_VERSION ")" +#define HPSA "hpsa" /* How long to wait (in milliseconds) for board to go into simple mode */ #define MAX_CONFIG_WAIT 30000 @@ -88,8 +89,8 @@ static const struct pci_device_id hpsa_pci_device_id[] = { {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3245}, {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3247}, {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3249}, - {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x324a}, - {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x324b}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x324A}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x324B}, {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3233}, {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSF, 0x103C, 0x3350}, {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSF, 0x103C, 0x3351}, @@ -98,6 +99,37 @@ static const struct pci_device_id hpsa_pci_device_id[] = { {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSF, 0x103C, 0x3354}, {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSF, 0x103C, 0x3355}, {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSF, 0x103C, 0x3356}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1921}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1922}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1923}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1924}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1925}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1926}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1928}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1929}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21BD}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21BE}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21BF}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21C0}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21C1}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21C2}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21C3}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21C4}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21C5}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21C6}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21C7}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21C8}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21C9}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21CA}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21CB}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21CC}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21CD}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSI, 0x103C, 0x21CE}, + {PCI_VENDOR_ID_HP_3PAR, 0x0075, 0x1590, 0x0076}, + {PCI_VENDOR_ID_HP_3PAR, 0x0075, 0x1590, 0x0087}, + {PCI_VENDOR_ID_HP_3PAR, 0x0075, 0x1590, 0x007D}, + {PCI_VENDOR_ID_HP_3PAR, 0x0075, 0x1590, 0x0088}, + {PCI_VENDOR_ID_HP, 0x333f, 0x103c, 0x333f}, {PCI_VENDOR_ID_HP, PCI_ANY_ID, PCI_ANY_ID, PCI_ANY_ID, PCI_CLASS_STORAGE_RAID << 8, 0xffff << 8, 0}, {0,} @@ -115,28 +147,55 @@ static struct board_type products[] = { {0x3245103C, "Smart Array P410i", &SA5_access}, {0x3247103C, "Smart Array P411", &SA5_access}, {0x3249103C, "Smart Array P812", &SA5_access}, - {0x324a103C, "Smart Array P712m", &SA5_access}, - {0x324b103C, "Smart Array P711m", &SA5_access}, - {0x3350103C, "Smart Array", &SA5_access}, - {0x3351103C, "Smart Array", &SA5_access}, - {0x3352103C, "Smart Array", &SA5_access}, - {0x3353103C, "Smart Array", &SA5_access}, - {0x3354103C, "Smart Array", &SA5_access}, - {0x3355103C, "Smart Array", &SA5_access}, - {0x3356103C, "Smart Array", &SA5_access}, + {0x324A103C, "Smart Array P712m", &SA5_access}, + {0x324B103C, "Smart Array P711m", &SA5_access}, + {0x3350103C, "Smart Array P222", &SA5_access}, + {0x3351103C, "Smart Array P420", &SA5_access}, + {0x3352103C, "Smart Array P421", &SA5_access}, + {0x3353103C, "Smart Array P822", &SA5_access}, + {0x3354103C, "Smart Array P420i", &SA5_access}, + {0x3355103C, "Smart Array P220i", &SA5_access}, + {0x3356103C, "Smart Array P721m", &SA5_access}, + {0x1921103C, "Smart Array P830i", &SA5_access}, + {0x1922103C, "Smart Array P430", &SA5_access}, + {0x1923103C, "Smart Array P431", &SA5_access}, + {0x1924103C, "Smart Array P830", &SA5_access}, + {0x1926103C, "Smart Array P731m", &SA5_access}, + {0x1928103C, "Smart Array P230i", &SA5_access}, + {0x1929103C, "Smart Array P530", &SA5_access}, + {0x21BD103C, "Smart Array", &SA5_access}, + {0x21BE103C, "Smart Array", &SA5_access}, + {0x21BF103C, "Smart Array", &SA5_access}, + {0x21C0103C, "Smart Array", &SA5_access}, + {0x21C1103C, "Smart Array", &SA5_access}, + {0x21C2103C, "Smart Array", &SA5_access}, + {0x21C3103C, "Smart Array", &SA5_access}, + {0x21C4103C, "Smart Array", &SA5_access}, + {0x21C5103C, "Smart Array", &SA5_access}, + {0x21C6103C, "Smart Array", &SA5_access}, + {0x21C7103C, "Smart Array", &SA5_access}, + {0x21C8103C, "Smart Array", &SA5_access}, + {0x21C9103C, "Smart Array", &SA5_access}, + {0x21CA103C, "Smart Array", &SA5_access}, + {0x21CB103C, "Smart Array", &SA5_access}, + {0x21CC103C, "Smart Array", &SA5_access}, + {0x21CD103C, "Smart Array", &SA5_access}, + {0x21CE103C, "Smart Array", &SA5_access}, + {0x00761590, "HP Storage P1224 Array Controller", &SA5_access}, + {0x00871590, "HP Storage P1224e Array Controller", &SA5_access}, + {0x007D1590, "HP Storage P1228 Array Controller", &SA5_access}, + {0x00881590, "HP Storage P1228e Array Controller", &SA5_access}, + {0x333f103c, "HP StorageWorks 1210m Array Controller", &SA5_access}, {0xFFFF103C, "Unknown Smart Array", &SA5_access}, }; static int number_of_controllers; -static struct list_head hpsa_ctlr_list = LIST_HEAD_INIT(hpsa_ctlr_list); -static spinlock_t lockup_detector_lock; -static struct task_struct *hpsa_lockup_detector; - static irqreturn_t do_hpsa_intr_intx(int irq, void *dev_id); static irqreturn_t do_hpsa_intr_msi(int irq, void *dev_id); static int hpsa_ioctl(struct scsi_device *dev, int cmd, void *arg); -static void start_io(struct ctlr_info *h); +static void lock_and_start_io(struct ctlr_info *h); +static void start_io(struct ctlr_info *h, unsigned long *flags); #ifdef CONFIG_COMPAT static int hpsa_compat_ioctl(struct scsi_device *dev, int cmd, void *arg); @@ -146,9 +205,10 @@ static void cmd_free(struct ctlr_info *h, struct CommandList *c); static void cmd_special_free(struct ctlr_info *h, struct CommandList *c); static struct CommandList *cmd_alloc(struct ctlr_info *h); static struct CommandList *cmd_special_alloc(struct ctlr_info *h); -static void fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, - void *buff, size_t size, u8 page_code, unsigned char *scsi3addr, +static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, + void *buff, size_t size, u16 page_code, unsigned char *scsi3addr, int cmd_type); +#define VPD_PAGE (1 << 8) static int hpsa_scsi_queue_command(struct Scsi_Host *h, struct scsi_cmnd *cmd); static void hpsa_scan_start(struct Scsi_Host *); @@ -158,6 +218,7 @@ static int hpsa_change_queue_depth(struct scsi_device *sdev, int qdepth, int reason); static int hpsa_eh_device_reset_handler(struct scsi_cmnd *scsicmd); +static int hpsa_eh_abort_handler(struct scsi_cmnd *scsicmd); static int hpsa_slave_alloc(struct scsi_device *sdev); static void hpsa_slave_destroy(struct scsi_device *sdev); @@ -168,19 +229,26 @@ static void check_ioctl_unit_attention(struct ctlr_info *h, struct CommandList *c); /* performant mode helper functions */ static void calc_bucket_map(int *bucket, int num_buckets, - int nsgs, int *bucket_map); -static __devinit void hpsa_put_ctlr_into_performant_mode(struct ctlr_info *h); -static inline u32 next_command(struct ctlr_info *h); -static int __devinit hpsa_find_cfg_addrs(struct pci_dev *pdev, - void __iomem *vaddr, u32 *cfg_base_addr, u64 *cfg_base_addr_index, - u64 *cfg_offset); -static int __devinit hpsa_pci_find_memory_BAR(struct pci_dev *pdev, - unsigned long *memory_bar); -static int __devinit hpsa_lookup_board_id(struct pci_dev *pdev, u32 *board_id); -static int __devinit hpsa_wait_for_board_state(struct pci_dev *pdev, - void __iomem *vaddr, int wait_for_ready); + int nsgs, int min_blocks, int *bucket_map); +static void hpsa_put_ctlr_into_performant_mode(struct ctlr_info *h); +static inline u32 next_command(struct ctlr_info *h, u8 q); +static int hpsa_find_cfg_addrs(struct pci_dev *pdev, void __iomem *vaddr, + u32 *cfg_base_addr, u64 *cfg_base_addr_index, + u64 *cfg_offset); +static int hpsa_pci_find_memory_BAR(struct pci_dev *pdev, + unsigned long *memory_bar); +static int hpsa_lookup_board_id(struct pci_dev *pdev, u32 *board_id); +static int hpsa_wait_for_board_state(struct pci_dev *pdev, void __iomem *vaddr, + int wait_for_ready); +static inline void finish_cmd(struct CommandList *c); +static void hpsa_wait_for_mode_change_ack(struct ctlr_info *h); #define BOARD_NOT_READY 0 #define BOARD_READY 1 +static void hpsa_drain_accel_commands(struct ctlr_info *h); +static void hpsa_flush_cache(struct ctlr_info *h); +static int hpsa_scsi_ioaccel_queue_command(struct ctlr_info *h, + struct CommandList *c, u32 ioaccel_handle, u8 *cdb, int cdb_len, + u8 *scsi3addr); static inline struct ctlr_info *sdev_to_hba(struct scsi_device *sdev) { @@ -202,36 +270,96 @@ static int check_for_unit_attention(struct ctlr_info *h, switch (c->err_info->SenseInfo[12]) { case STATE_CHANGED: - dev_warn(&h->pdev->dev, "hpsa%d: a state change " + dev_warn(&h->pdev->dev, HPSA "%d: a state change " "detected, command retried\n", h->ctlr); break; case LUN_FAILED: - dev_warn(&h->pdev->dev, "hpsa%d: LUN failure " + dev_warn(&h->pdev->dev, HPSA "%d: LUN failure " "detected, action required\n", h->ctlr); break; case REPORT_LUNS_CHANGED: - dev_warn(&h->pdev->dev, "hpsa%d: report LUN data " + dev_warn(&h->pdev->dev, HPSA "%d: report LUN data " "changed, action required\n", h->ctlr); /* - * Note: this REPORT_LUNS_CHANGED condition only occurs on the MSA2012. + * Note: this REPORT_LUNS_CHANGED condition only occurs on the external + * target (array) devices. */ break; case POWER_OR_RESET: - dev_warn(&h->pdev->dev, "hpsa%d: a power on " + dev_warn(&h->pdev->dev, HPSA "%d: a power on " "or device reset detected\n", h->ctlr); break; case UNIT_ATTENTION_CLEARED: - dev_warn(&h->pdev->dev, "hpsa%d: unit attention " + dev_warn(&h->pdev->dev, HPSA "%d: unit attention " "cleared by another initiator\n", h->ctlr); break; default: - dev_warn(&h->pdev->dev, "hpsa%d: unknown " + dev_warn(&h->pdev->dev, HPSA "%d: unknown " "unit attention detected\n", h->ctlr); break; } return 1; } +static int check_for_busy(struct ctlr_info *h, struct CommandList *c) +{ + if (c->err_info->CommandStatus != CMD_TARGET_STATUS || + (c->err_info->ScsiStatus != SAM_STAT_BUSY && + c->err_info->ScsiStatus != SAM_STAT_TASK_SET_FULL)) + return 0; + dev_warn(&h->pdev->dev, HPSA "device busy"); + return 1; +} + +static ssize_t host_store_hp_ssd_smart_path_status(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + int status, len; + struct ctlr_info *h; + struct Scsi_Host *shost = class_to_shost(dev); + char tmpbuf[10]; + + if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SYS_RAWIO)) + return -EACCES; + len = count > sizeof(tmpbuf) - 1 ? sizeof(tmpbuf) - 1 : count; + strncpy(tmpbuf, buf, len); + tmpbuf[len] = '\0'; + if (sscanf(tmpbuf, "%d", &status) != 1) + return -EINVAL; + h = shost_to_hba(shost); + h->acciopath_status = !!status; + dev_warn(&h->pdev->dev, + "hpsa: HP SSD Smart Path %s via sysfs update.\n", + h->acciopath_status ? "enabled" : "disabled"); + return count; +} + +static ssize_t host_store_raid_offload_debug(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + int debug_level, len; + struct ctlr_info *h; + struct Scsi_Host *shost = class_to_shost(dev); + char tmpbuf[10]; + + if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SYS_RAWIO)) + return -EACCES; + len = count > sizeof(tmpbuf) - 1 ? sizeof(tmpbuf) - 1 : count; + strncpy(tmpbuf, buf, len); + tmpbuf[len] = '\0'; + if (sscanf(tmpbuf, "%d", &debug_level) != 1) + return -EINVAL; + if (debug_level < 0) + debug_level = 0; + h = shost_to_hba(shost); + h->raid_offload_debug = debug_level; + dev_warn(&h->pdev->dev, "hpsa: Set raid_offload_debug level = %d\n", + h->raid_offload_debug); + return count; +} + static ssize_t host_store_rescan(struct device *dev, struct device_attribute *attr, const char *buf, size_t count) @@ -279,6 +407,17 @@ static ssize_t host_show_transport_mode(struct device *dev, "performant" : "simple"); } +static ssize_t host_show_hp_ssd_smart_path_status(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct ctlr_info *h; + struct Scsi_Host *shost = class_to_shost(dev); + + h = shost_to_hba(shost); + return snprintf(buf, 30, "HP SSD Smart Path %s\n", + (h->acciopath_status == 1) ? "enabled" : "disabled"); +} + /* List of controllers which cannot be hard reset on kexec with reset_devices */ static u32 unresettable_controller[] = { 0x324a103C, /* Smart Array P712m */ @@ -296,11 +435,23 @@ static u32 unresettable_controller[] = { 0x40800E11, /* Smart Array 5i */ 0x409C0E11, /* Smart Array 6400 */ 0x409D0E11, /* Smart Array 6400 EM */ + 0x40700E11, /* Smart Array 5300 */ + 0x40820E11, /* Smart Array 532 */ + 0x40830E11, /* Smart Array 5312 */ + 0x409A0E11, /* Smart Array 641 */ + 0x409B0E11, /* Smart Array 642 */ + 0x40910E11, /* Smart Array 6i */ }; /* List of controllers which cannot even be soft reset */ static u32 soft_unresettable_controller[] = { 0x40800E11, /* Smart Array 5i */ + 0x40700E11, /* Smart Array 5300 */ + 0x40820E11, /* Smart Array 532 */ + 0x40830E11, /* Smart Array 5312 */ + 0x409A0E11, /* Smart Array 641 */ + 0x409B0E11, /* Smart Array 642 */ + 0x40910E11, /* Smart Array 6i */ /* Exclude 640x boards. These are two pci devices in one slot * which share a battery backed cache module. One controls the * cache, the other accesses the cache through the one that controls @@ -354,8 +505,15 @@ static inline int is_logical_dev_addr_mode(unsigned char scsi3addr[]) } static const char *raid_label[] = { "0", "4", "1(1+0)", "5", "5+1", "ADG", - "UNKNOWN" + "1(ADM)", "UNKNOWN" }; +#define HPSA_RAID_0 0 +#define HPSA_RAID_4 1 +#define HPSA_RAID_1 2 /* also used for RAID 10 */ +#define HPSA_RAID_5 3 /* also used for RAID 50 */ +#define HPSA_RAID_51 4 +#define HPSA_RAID_6 5 /* also used for RAID 60 */ +#define HPSA_RAID_ADM 6 /* also used for RAID 1+0 ADM */ #define RAID_UNKNOWN (ARRAY_SIZE(raid_label) - 1) static ssize_t raid_level_show(struct device *dev, @@ -444,10 +602,39 @@ static ssize_t unique_id_show(struct device *dev, sn[12], sn[13], sn[14], sn[15]); } +static ssize_t host_show_hp_ssd_smart_path_enabled(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct ctlr_info *h; + struct scsi_device *sdev; + struct hpsa_scsi_dev_t *hdev; + unsigned long flags; + int offload_enabled; + + sdev = to_scsi_device(dev); + h = sdev_to_hba(sdev); + spin_lock_irqsave(&h->lock, flags); + hdev = sdev->hostdata; + if (!hdev) { + spin_unlock_irqrestore(&h->lock, flags); + return -ENODEV; + } + offload_enabled = hdev->offload_enabled; + spin_unlock_irqrestore(&h->lock, flags); + return snprintf(buf, 20, "%d\n", offload_enabled); +} + static DEVICE_ATTR(raid_level, S_IRUGO, raid_level_show, NULL); static DEVICE_ATTR(lunid, S_IRUGO, lunid_show, NULL); static DEVICE_ATTR(unique_id, S_IRUGO, unique_id_show, NULL); static DEVICE_ATTR(rescan, S_IWUSR, NULL, host_store_rescan); +static DEVICE_ATTR(hp_ssd_smart_path_enabled, S_IRUGO, + host_show_hp_ssd_smart_path_enabled, NULL); +static DEVICE_ATTR(hp_ssd_smart_path_status, S_IWUSR|S_IRUGO|S_IROTH, + host_show_hp_ssd_smart_path_status, + host_store_hp_ssd_smart_path_status); +static DEVICE_ATTR(raid_offload_debug, S_IWUSR, NULL, + host_store_raid_offload_debug); static DEVICE_ATTR(firmware_revision, S_IRUGO, host_show_firmware_revision, NULL); static DEVICE_ATTR(commands_outstanding, S_IRUGO, @@ -461,6 +648,7 @@ static struct device_attribute *hpsa_sdev_attrs[] = { &dev_attr_raid_level, &dev_attr_lunid, &dev_attr_unique_id, + &dev_attr_hp_ssd_smart_path_enabled, NULL, }; @@ -470,19 +658,22 @@ static struct device_attribute *hpsa_shost_attrs[] = { &dev_attr_commands_outstanding, &dev_attr_transport_mode, &dev_attr_resettable, + &dev_attr_hp_ssd_smart_path_status, + &dev_attr_raid_offload_debug, NULL, }; static struct scsi_host_template hpsa_driver_template = { .module = THIS_MODULE, - .name = "hpsa", - .proc_name = "hpsa", + .name = HPSA, + .proc_name = HPSA, .queuecommand = hpsa_scsi_queue_command, .scan_start = hpsa_scan_start, .scan_finished = hpsa_scan_finished, .change_queue_depth = hpsa_change_queue_depth, .this_id = -1, .use_clustering = ENABLE_CLUSTERING, + .eh_abort_handler = hpsa_eh_abort_handler, .eh_device_reset_handler = hpsa_eh_device_reset_handler, .ioctl = hpsa_ioctl, .slave_alloc = hpsa_slave_alloc, @@ -493,6 +684,7 @@ static struct scsi_host_template hpsa_driver_template = { .sdev_attrs = hpsa_sdev_attrs, .shost_attrs = hpsa_shost_attrs, .max_sectors = 8192, + .no_write_same = 1, }; @@ -502,36 +694,137 @@ static inline void addQ(struct list_head *list, struct CommandList *c) list_add_tail(&c->list, list); } -static inline u32 next_command(struct ctlr_info *h) +static inline u32 next_command(struct ctlr_info *h, u8 q) { u32 a; + struct reply_queue_buffer *rq = &h->reply_queue[q]; + unsigned long flags; + + if (h->transMethod & CFGTBL_Trans_io_accel1) + return h->access.command_completed(h, q); if (unlikely(!(h->transMethod & CFGTBL_Trans_Performant))) - return h->access.command_completed(h); + return h->access.command_completed(h, q); - if ((*(h->reply_pool_head) & 1) == (h->reply_pool_wraparound)) { - a = *(h->reply_pool_head); /* Next cmd in ring buffer */ - (h->reply_pool_head)++; + if ((rq->head[rq->current_entry] & 1) == rq->wraparound) { + a = rq->head[rq->current_entry]; + rq->current_entry++; + spin_lock_irqsave(&h->lock, flags); h->commands_outstanding--; + spin_unlock_irqrestore(&h->lock, flags); } else { a = FIFO_EMPTY; } /* Check for wraparound */ - if (h->reply_pool_head == (h->reply_pool + h->max_commands)) { - h->reply_pool_head = h->reply_pool; - h->reply_pool_wraparound ^= 1; + if (rq->current_entry == h->max_commands) { + rq->current_entry = 0; + rq->wraparound ^= 1; } return a; } +/* + * There are some special bits in the bus address of the + * command that we have to set for the controller to know + * how to process the command: + * + * Normal performant mode: + * bit 0: 1 means performant mode, 0 means simple mode. + * bits 1-3 = block fetch table entry + * bits 4-6 = command type (== 0) + * + * ioaccel1 mode: + * bit 0 = "performant mode" bit. + * bits 1-3 = block fetch table entry + * bits 4-6 = command type (== 110) + * (command type is needed because ioaccel1 mode + * commands are submitted through the same register as normal + * mode commands, so this is how the controller knows whether + * the command is normal mode or ioaccel1 mode.) + * + * ioaccel2 mode: + * bit 0 = "performant mode" bit. + * bits 1-4 = block fetch table entry (note extra bit) + * bits 4-6 = not needed, because ioaccel2 mode has + * a separate special register for submitting commands. + */ + /* set_performant_mode: Modify the tag for cciss performant * set bit 0 for pull model, bits 3-1 for block fetch * register number */ static void set_performant_mode(struct ctlr_info *h, struct CommandList *c) { - if (likely(h->transMethod & CFGTBL_Trans_Performant)) + if (likely(h->transMethod & CFGTBL_Trans_Performant)) { c->busaddr |= 1 | (h->blockFetchTable[c->Header.SGList] << 1); + if (likely(h->msix_vector > 0)) + c->Header.ReplyQueue = + raw_smp_processor_id() % h->nreply_queues; + } +} + +static void set_ioaccel1_performant_mode(struct ctlr_info *h, + struct CommandList *c) +{ + struct io_accel1_cmd *cp = &h->ioaccel_cmd_pool[c->cmdindex]; + + /* Tell the controller to post the reply to the queue for this + * processor. This seems to give the best I/O throughput. + */ + cp->ReplyQueue = smp_processor_id() % h->nreply_queues; + /* Set the bits in the address sent down to include: + * - performant mode bit (bit 0) + * - pull count (bits 1-3) + * - command type (bits 4-6) + */ + c->busaddr |= 1 | (h->ioaccel1_blockFetchTable[c->Header.SGList] << 1) | + IOACCEL1_BUSADDR_CMDTYPE; +} + +static void set_ioaccel2_performant_mode(struct ctlr_info *h, + struct CommandList *c) +{ + struct io_accel2_cmd *cp = &h->ioaccel2_cmd_pool[c->cmdindex]; + + /* Tell the controller to post the reply to the queue for this + * processor. This seems to give the best I/O throughput. + */ + cp->reply_queue = smp_processor_id() % h->nreply_queues; + /* Set the bits in the address sent down to include: + * - performant mode bit not used in ioaccel mode 2 + * - pull count (bits 0-3) + * - command type isn't needed for ioaccel2 + */ + c->busaddr |= (h->ioaccel2_blockFetchTable[cp->sg_count]); +} + +static int is_firmware_flash_cmd(u8 *cdb) +{ + return cdb[0] == BMIC_WRITE && cdb[6] == BMIC_FLASH_FIRMWARE; +} + +/* + * During firmware flash, the heartbeat register may not update as frequently + * as it should. So we dial down lockup detection during firmware flash. and + * dial it back up when firmware flash completes. + */ +#define HEARTBEAT_SAMPLE_INTERVAL_DURING_FLASH (240 * HZ) +#define HEARTBEAT_SAMPLE_INTERVAL (30 * HZ) +static void dial_down_lockup_detection_during_fw_flash(struct ctlr_info *h, + struct CommandList *c) +{ + if (!is_firmware_flash_cmd(c->Request.CDB)) + return; + atomic_inc(&h->firmware_flash_in_progress); + h->heartbeat_sample_interval = HEARTBEAT_SAMPLE_INTERVAL_DURING_FLASH; +} + +static void dial_up_lockup_detection_on_fw_flash_complete(struct ctlr_info *h, + struct CommandList *c) +{ + if (is_firmware_flash_cmd(c->Request.CDB) && + atomic_dec_and_test(&h->firmware_flash_in_progress)) + h->heartbeat_sample_interval = HEARTBEAT_SAMPLE_INTERVAL; } static void enqueue_cmd_and_start_io(struct ctlr_info *h, @@ -539,11 +832,21 @@ static void enqueue_cmd_and_start_io(struct ctlr_info *h, { unsigned long flags; - set_performant_mode(h, c); + switch (c->cmd_type) { + case CMD_IOACCEL1: + set_ioaccel1_performant_mode(h, c); + break; + case CMD_IOACCEL2: + set_ioaccel2_performant_mode(h, c); + break; + default: + set_performant_mode(h, c); + } + dial_down_lockup_detection_during_fw_flash(h, c); spin_lock_irqsave(&h->lock, flags); addQ(&h->reqQ, c); h->Qdepth++; - start_io(h); + start_io(h, &flags); spin_unlock_irqrestore(&h->lock, flags); } @@ -577,21 +880,19 @@ static int hpsa_find_target_lun(struct ctlr_info *h, int i, found = 0; DECLARE_BITMAP(lun_taken, HPSA_MAX_DEVICES); - memset(&lun_taken[0], 0, HPSA_MAX_DEVICES >> 3); + bitmap_zero(lun_taken, HPSA_MAX_DEVICES); for (i = 0; i < h->ndevices; i++) { if (h->dev[i]->bus == bus && h->dev[i]->target != -1) - set_bit(h->dev[i]->target, lun_taken); + __set_bit(h->dev[i]->target, lun_taken); } - for (i = 0; i < HPSA_MAX_DEVICES; i++) { - if (!test_bit(i, lun_taken)) { - /* *bus = 1; */ - *target = i; - *lun = 0; - found = 1; - break; - } + i = find_first_zero_bit(lun_taken, HPSA_MAX_DEVICES); + if (i < HPSA_MAX_DEVICES) { + /* *bus = 1; */ + *target = i; + *lun = 0; + found = 1; } return !found; } @@ -675,6 +976,28 @@ lun_assigned: return 0; } +/* Update an entry in h->dev[] array. */ +static void hpsa_scsi_update_entry(struct ctlr_info *h, int hostno, + int entry, struct hpsa_scsi_dev_t *new_entry) +{ + /* assumes h->devlock is held */ + BUG_ON(entry < 0 || entry >= HPSA_MAX_DEVICES); + + /* Raid level changed. */ + h->dev[entry]->raid_level = new_entry->raid_level; + + /* Raid offload parameters changed. */ + h->dev[entry]->offload_config = new_entry->offload_config; + h->dev[entry]->offload_enabled = new_entry->offload_enabled; + h->dev[entry]->ioaccel_handle = new_entry->ioaccel_handle; + h->dev[entry]->offload_to_mirror = new_entry->offload_to_mirror; + h->dev[entry]->raid_map = new_entry->raid_map; + + dev_info(&h->pdev->dev, "%s device c%db%dt%dl%d updated.\n", + scsi_device_type(new_entry->devtype), hostno, new_entry->bus, + new_entry->target, new_entry->lun); +} + /* Replace an entry from h->dev[] array. */ static void hpsa_scsi_replace_entry(struct ctlr_info *h, int hostno, int entry, struct hpsa_scsi_dev_t *new_entry, @@ -781,10 +1104,29 @@ static inline int device_is_the_same(struct hpsa_scsi_dev_t *dev1, return 1; } +static inline int device_updated(struct hpsa_scsi_dev_t *dev1, + struct hpsa_scsi_dev_t *dev2) +{ + /* Device attributes that can change, but don't mean + * that the device is a different device, nor that the OS + * needs to be told anything about the change. + */ + if (dev1->raid_level != dev2->raid_level) + return 1; + if (dev1->offload_config != dev2->offload_config) + return 1; + if (dev1->offload_enabled != dev2->offload_enabled) + return 1; + return 0; +} + /* Find needle in haystack. If exact match found, return DEVICE_SAME, * and return needle location in *index. If scsi3addr matches, but not * vendor, model, serial num, etc. return DEVICE_CHANGED, and return needle - * location in *index. If needle not found, return DEVICE_NOT_FOUND. + * location in *index. + * In the case of a minor device attribute change, such as RAID level, just + * return DEVICE_UPDATED, along with the updated device's location in index. + * If needle not found, return DEVICE_NOT_FOUND. */ static int hpsa_scsi_find_entry(struct hpsa_scsi_dev_t *needle, struct hpsa_scsi_dev_t *haystack[], int haystack_size, @@ -794,21 +1136,132 @@ static int hpsa_scsi_find_entry(struct hpsa_scsi_dev_t *needle, #define DEVICE_NOT_FOUND 0 #define DEVICE_CHANGED 1 #define DEVICE_SAME 2 +#define DEVICE_UPDATED 3 for (i = 0; i < haystack_size; i++) { if (haystack[i] == NULL) /* previously removed. */ continue; if (SCSI3ADDR_EQ(needle->scsi3addr, haystack[i]->scsi3addr)) { *index = i; - if (device_is_the_same(needle, haystack[i])) + if (device_is_the_same(needle, haystack[i])) { + if (device_updated(needle, haystack[i])) + return DEVICE_UPDATED; return DEVICE_SAME; - else + } else { + /* Keep offline devices offline */ + if (needle->volume_offline) + return DEVICE_NOT_FOUND; return DEVICE_CHANGED; + } } } *index = -1; return DEVICE_NOT_FOUND; } +static void hpsa_monitor_offline_device(struct ctlr_info *h, + unsigned char scsi3addr[]) +{ + struct offline_device_entry *device; + unsigned long flags; + + /* Check to see if device is already on the list */ + spin_lock_irqsave(&h->offline_device_lock, flags); + list_for_each_entry(device, &h->offline_device_list, offline_list) { + if (memcmp(device->scsi3addr, scsi3addr, + sizeof(device->scsi3addr)) == 0) { + spin_unlock_irqrestore(&h->offline_device_lock, flags); + return; + } + } + spin_unlock_irqrestore(&h->offline_device_lock, flags); + + /* Device is not on the list, add it. */ + device = kmalloc(sizeof(*device), GFP_KERNEL); + if (!device) { + dev_warn(&h->pdev->dev, "out of memory in %s\n", __func__); + return; + } + memcpy(device->scsi3addr, scsi3addr, sizeof(device->scsi3addr)); + spin_lock_irqsave(&h->offline_device_lock, flags); + list_add_tail(&device->offline_list, &h->offline_device_list); + spin_unlock_irqrestore(&h->offline_device_lock, flags); +} + +/* Print a message explaining various offline volume states */ +static void hpsa_show_volume_status(struct ctlr_info *h, + struct hpsa_scsi_dev_t *sd) +{ + if (sd->volume_offline == HPSA_VPD_LV_STATUS_UNSUPPORTED) + dev_info(&h->pdev->dev, + "C%d:B%d:T%d:L%d Volume status is not available through vital product data pages.\n", + h->scsi_host->host_no, + sd->bus, sd->target, sd->lun); + switch (sd->volume_offline) { + case HPSA_LV_OK: + break; + case HPSA_LV_UNDERGOING_ERASE: + dev_info(&h->pdev->dev, + "C%d:B%d:T%d:L%d Volume is undergoing background erase process.\n", + h->scsi_host->host_no, + sd->bus, sd->target, sd->lun); + break; + case HPSA_LV_UNDERGOING_RPI: + dev_info(&h->pdev->dev, + "C%d:B%d:T%d:L%d Volume is undergoing rapid parity initialization process.\n", + h->scsi_host->host_no, + sd->bus, sd->target, sd->lun); + break; + case HPSA_LV_PENDING_RPI: + dev_info(&h->pdev->dev, + "C%d:B%d:T%d:L%d Volume is queued for rapid parity initialization process.\n", + h->scsi_host->host_no, + sd->bus, sd->target, sd->lun); + break; + case HPSA_LV_ENCRYPTED_NO_KEY: + dev_info(&h->pdev->dev, + "C%d:B%d:T%d:L%d Volume is encrypted and cannot be accessed because key is not present.\n", + h->scsi_host->host_no, + sd->bus, sd->target, sd->lun); + break; + case HPSA_LV_PLAINTEXT_IN_ENCRYPT_ONLY_CONTROLLER: + dev_info(&h->pdev->dev, + "C%d:B%d:T%d:L%d Volume is not encrypted and cannot be accessed because controller is in encryption-only mode.\n", + h->scsi_host->host_no, + sd->bus, sd->target, sd->lun); + break; + case HPSA_LV_UNDERGOING_ENCRYPTION: + dev_info(&h->pdev->dev, + "C%d:B%d:T%d:L%d Volume is undergoing encryption process.\n", + h->scsi_host->host_no, + sd->bus, sd->target, sd->lun); + break; + case HPSA_LV_UNDERGOING_ENCRYPTION_REKEYING: + dev_info(&h->pdev->dev, + "C%d:B%d:T%d:L%d Volume is undergoing encryption re-keying process.\n", + h->scsi_host->host_no, + sd->bus, sd->target, sd->lun); + break; + case HPSA_LV_ENCRYPTED_IN_NON_ENCRYPTED_CONTROLLER: + dev_info(&h->pdev->dev, + "C%d:B%d:T%d:L%d Volume is encrypted and cannot be accessed because controller does not have encryption enabled.\n", + h->scsi_host->host_no, + sd->bus, sd->target, sd->lun); + break; + case HPSA_LV_PENDING_ENCRYPTION: + dev_info(&h->pdev->dev, + "C%d:B%d:T%d:L%d Volume is pending migration to encrypted state, but process has not started.\n", + h->scsi_host->host_no, + sd->bus, sd->target, sd->lun); + break; + case HPSA_LV_PENDING_ENCRYPTION_REKEYING: + dev_info(&h->pdev->dev, + "C%d:B%d:T%d:L%d Volume is encrypted and is pending encryption rekeying.\n", + h->scsi_host->host_no, + sd->bus, sd->target, sd->lun); + break; + } +} + static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno, struct hpsa_scsi_dev_t *sd[], int nsds) { @@ -838,6 +1291,8 @@ static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno, * sd[] and remove them from h->dev[], and for any * devices which have changed, remove the old device * info and add the new device info. + * If minor device attributes change, just update + * the existing device structure. */ i = 0; nremoved = 0; @@ -858,6 +1313,8 @@ static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno, * at the bottom of hpsa_update_scsi_devices() */ sd[entry] = NULL; + } else if (device_change == DEVICE_UPDATED) { + hpsa_scsi_update_entry(h, hostno, i, sd[entry]); } i++; } @@ -869,6 +1326,20 @@ static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno, for (i = 0; i < nsds; i++) { if (!sd[i]) /* if already added above. */ continue; + + /* Don't add devices which are NOT READY, FORMAT IN PROGRESS + * as the SCSI mid-layer does not handle such devices well. + * It relentlessly loops sending TUR at 3Hz, then READ(10) + * at 160Hz, and prevents the system from coming up. + */ + if (sd[i]->volume_offline) { + hpsa_show_volume_status(h, sd[i]); + dev_info(&h->pdev->dev, "c%db%dt%dl%d: temporarily offline\n", + h->scsi_host->host_no, + sd[i]->bus, sd[i]->target, sd[i]->lun); + continue; + } + device_change = hpsa_scsi_find_entry(sd[i], h->dev, h->ndevices, &entry); if (device_change == DEVICE_NOT_FOUND) { @@ -887,6 +1358,17 @@ static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno, } spin_unlock_irqrestore(&h->devlock, flags); + /* Monitor devices which are in one of several NOT READY states to be + * brought online later. This must be done without holding h->devlock, + * so don't touch h->dev[] + */ + for (i = 0; i < nsds; i++) { + if (!sd[i]) /* if already added above. */ + continue; + if (sd[i]->volume_offline) + hpsa_monitor_offline_device(h, sd[i]->scsi3addr); + } + /* Don't notify scsi mid layer of any changes the first time through * (or if there are no changes) scsi_scan_host will do it later the * first time through. @@ -936,7 +1418,7 @@ free_and_out: } /* - * Lookup bus/target/lun and retrun corresponding struct hpsa_scsi_dev_t * + * Lookup bus/target/lun and return corresponding struct hpsa_scsi_dev_t * * Assume's h->devlock is held. */ static struct hpsa_scsi_dev_t *lookup_hpsa_scsi_dev(struct ctlr_info *h, @@ -1013,7 +1495,7 @@ clean: return -ENOMEM; } -static void hpsa_map_sg_chain_block(struct ctlr_info *h, +static int hpsa_map_sg_chain_block(struct ctlr_info *h, struct CommandList *c) { struct SGDescriptor *chain_sg, *chain_block; @@ -1026,8 +1508,15 @@ static void hpsa_map_sg_chain_block(struct ctlr_info *h, (c->Header.SGTotal - h->max_cmd_sg_entries); temp64 = pci_map_single(h->pdev, chain_block, chain_sg->Len, PCI_DMA_TODEVICE); + if (dma_mapping_error(&h->pdev->dev, temp64)) { + /* prevent subsequent unmapping */ + chain_sg->Addr.lower = 0; + chain_sg->Addr.upper = 0; + return -1; + } chain_sg->Addr.lower = (u32) (temp64 & 0x0FFFFFFFFULL); chain_sg->Addr.upper = (u32) ((temp64 >> 32) & 0x0FFFFFFFFULL); + return 0; } static void hpsa_unmap_sg_chain_block(struct ctlr_info *h, @@ -1045,11 +1534,156 @@ static void hpsa_unmap_sg_chain_block(struct ctlr_info *h, pci_unmap_single(h->pdev, temp64.val, chain_sg->Len, PCI_DMA_TODEVICE); } + +/* Decode the various types of errors on ioaccel2 path. + * Return 1 for any error that should generate a RAID path retry. + * Return 0 for errors that don't require a RAID path retry. + */ +static int handle_ioaccel_mode2_error(struct ctlr_info *h, + struct CommandList *c, + struct scsi_cmnd *cmd, + struct io_accel2_cmd *c2) +{ + int data_len; + int retry = 0; + + switch (c2->error_data.serv_response) { + case IOACCEL2_SERV_RESPONSE_COMPLETE: + switch (c2->error_data.status) { + case IOACCEL2_STATUS_SR_TASK_COMP_GOOD: + break; + case IOACCEL2_STATUS_SR_TASK_COMP_CHK_COND: + dev_warn(&h->pdev->dev, + "%s: task complete with check condition.\n", + "HP SSD Smart Path"); + cmd->result |= SAM_STAT_CHECK_CONDITION; + if (c2->error_data.data_present != + IOACCEL2_SENSE_DATA_PRESENT) { + memset(cmd->sense_buffer, 0, + SCSI_SENSE_BUFFERSIZE); + break; + } + /* copy the sense data */ + data_len = c2->error_data.sense_data_len; + if (data_len > SCSI_SENSE_BUFFERSIZE) + data_len = SCSI_SENSE_BUFFERSIZE; + if (data_len > sizeof(c2->error_data.sense_data_buff)) + data_len = + sizeof(c2->error_data.sense_data_buff); + memcpy(cmd->sense_buffer, + c2->error_data.sense_data_buff, data_len); + retry = 1; + break; + case IOACCEL2_STATUS_SR_TASK_COMP_BUSY: + dev_warn(&h->pdev->dev, + "%s: task complete with BUSY status.\n", + "HP SSD Smart Path"); + retry = 1; + break; + case IOACCEL2_STATUS_SR_TASK_COMP_RES_CON: + dev_warn(&h->pdev->dev, + "%s: task complete with reservation conflict.\n", + "HP SSD Smart Path"); + retry = 1; + break; + case IOACCEL2_STATUS_SR_TASK_COMP_SET_FULL: + /* Make scsi midlayer do unlimited retries */ + cmd->result = DID_IMM_RETRY << 16; + break; + case IOACCEL2_STATUS_SR_TASK_COMP_ABORTED: + dev_warn(&h->pdev->dev, + "%s: task complete with aborted status.\n", + "HP SSD Smart Path"); + retry = 1; + break; + default: + dev_warn(&h->pdev->dev, + "%s: task complete with unrecognized status: 0x%02x\n", + "HP SSD Smart Path", c2->error_data.status); + retry = 1; + break; + } + break; + case IOACCEL2_SERV_RESPONSE_FAILURE: + /* don't expect to get here. */ + dev_warn(&h->pdev->dev, + "unexpected delivery or target failure, status = 0x%02x\n", + c2->error_data.status); + retry = 1; + break; + case IOACCEL2_SERV_RESPONSE_TMF_COMPLETE: + break; + case IOACCEL2_SERV_RESPONSE_TMF_SUCCESS: + break; + case IOACCEL2_SERV_RESPONSE_TMF_REJECTED: + dev_warn(&h->pdev->dev, "task management function rejected.\n"); + retry = 1; + break; + case IOACCEL2_SERV_RESPONSE_TMF_WRONG_LUN: + dev_warn(&h->pdev->dev, "task management function invalid LUN\n"); + break; + default: + dev_warn(&h->pdev->dev, + "%s: Unrecognized server response: 0x%02x\n", + "HP SSD Smart Path", + c2->error_data.serv_response); + retry = 1; + break; + } + + return retry; /* retry on raid path? */ +} + +static void process_ioaccel2_completion(struct ctlr_info *h, + struct CommandList *c, struct scsi_cmnd *cmd, + struct hpsa_scsi_dev_t *dev) +{ + struct io_accel2_cmd *c2 = &h->ioaccel2_cmd_pool[c->cmdindex]; + int raid_retry = 0; + + /* check for good status */ + if (likely(c2->error_data.serv_response == 0 && + c2->error_data.status == 0)) { + cmd_free(h, c); + cmd->scsi_done(cmd); + return; + } + + /* Any RAID offload error results in retry which will use + * the normal I/O path so the controller can handle whatever's + * wrong. + */ + if (is_logical_dev_addr_mode(dev->scsi3addr) && + c2->error_data.serv_response == + IOACCEL2_SERV_RESPONSE_FAILURE) { + dev->offload_enabled = 0; + h->drv_req_rescan = 1; /* schedule controller for a rescan */ + cmd->result = DID_SOFT_ERROR << 16; + cmd_free(h, c); + cmd->scsi_done(cmd); + return; + } + raid_retry = handle_ioaccel_mode2_error(h, c, cmd, c2); + /* If error found, disable Smart Path, schedule a rescan, + * and force a retry on the standard path. + */ + if (raid_retry) { + dev_warn(&h->pdev->dev, "%s: Retrying on standard path.\n", + "HP SSD Smart Path"); + dev->offload_enabled = 0; /* Disable Smart Path */ + h->drv_req_rescan = 1; /* schedule controller rescan */ + cmd->result = DID_SOFT_ERROR << 16; + } + cmd_free(h, c); + cmd->scsi_done(cmd); +} + static void complete_scsi_command(struct CommandList *cp) { struct scsi_cmnd *cmd; struct ctlr_info *h; struct ErrorInfo *ei; + struct hpsa_scsi_dev_t *dev; unsigned char sense_key; unsigned char asc; /* additional sense code */ @@ -1059,13 +1693,19 @@ static void complete_scsi_command(struct CommandList *cp) ei = cp->err_info; cmd = (struct scsi_cmnd *) cp->scsi_cmd; h = cp->h; + dev = cmd->device->hostdata; scsi_dma_unmap(cmd); /* undo the DMA mappings */ - if (cp->Header.SGTotal > h->max_cmd_sg_entries) + if ((cp->cmd_type == CMD_SCSI) && + (cp->Header.SGTotal > h->max_cmd_sg_entries)) hpsa_unmap_sg_chain_block(h, cp); cmd->result = (DID_OK << 16); /* host byte */ cmd->result |= (COMMAND_COMPLETE << 8); /* msg byte */ + + if (cp->cmd_type == CMD_IOACCEL2) + return process_ioaccel2_completion(h, cp, cmd, dev); + cmd->result |= ei->ScsiStatus; /* copy the sense data whether we need to or not. */ @@ -1080,11 +1720,37 @@ static void complete_scsi_command(struct CommandList *cp) scsi_set_resid(cmd, ei->ResidualCnt); if (ei->CommandStatus == 0) { - cmd->scsi_done(cmd); cmd_free(h, cp); + cmd->scsi_done(cmd); return; } + /* For I/O accelerator commands, copy over some fields to the normal + * CISS header used below for error handling. + */ + if (cp->cmd_type == CMD_IOACCEL1) { + struct io_accel1_cmd *c = &h->ioaccel_cmd_pool[cp->cmdindex]; + cp->Header.SGList = cp->Header.SGTotal = scsi_sg_count(cmd); + cp->Request.CDBLen = c->io_flags & IOACCEL1_IOFLAGS_CDBLEN_MASK; + cp->Header.Tag.lower = c->Tag.lower; + cp->Header.Tag.upper = c->Tag.upper; + memcpy(cp->Header.LUN.LunAddrBytes, c->CISS_LUN, 8); + memcpy(cp->Request.CDB, c->CDB, cp->Request.CDBLen); + + /* Any RAID offload error results in retry which will use + * the normal I/O path so the controller can handle whatever's + * wrong. + */ + if (is_logical_dev_addr_mode(dev->scsi3addr)) { + if (ei->CommandStatus == CMD_IOACCEL_DISABLED) + dev->offload_enabled = 0; + cmd->result = DID_SOFT_ERROR << 16; + cmd_free(h, cp); + cmd->scsi_done(cmd); + return; + } + } + /* an error has occurred */ switch (ei->CommandStatus) { @@ -1099,10 +1765,8 @@ static void complete_scsi_command(struct CommandList *cp) } if (ei->ScsiStatus == SAM_STAT_CHECK_CONDITION) { - if (check_for_unit_attention(h, cp)) { - cmd->result = DID_SOFT_ERROR << 16; + if (check_for_unit_attention(h, cp)) break; - } if (sense_key == ILLEGAL_REQUEST) { /* * SCSI REPORT_LUNS is commonly unsupported on @@ -1140,11 +1804,11 @@ static void complete_scsi_command(struct CommandList *cp) "has check condition: aborted command: " "ASC: 0x%x, ASCQ: 0x%x\n", cp, asc, ascq); - cmd->result = DID_SOFT_ERROR << 16; + cmd->result |= DID_SOFT_ERROR << 16; break; } /* Must be some other type of check condition */ - dev_warn(&h->pdev->dev, "cp %p has check condition: " + dev_dbg(&h->pdev->dev, "cp %p has check condition: " "unknown type: " "Sense: 0x%x, ASC: 0x%x, ASCQ: 0x%x, " "Returning result: 0x%x, " @@ -1215,8 +1879,9 @@ static void complete_scsi_command(struct CommandList *cp) } break; case CMD_PROTOCOL_ERR: + cmd->result = DID_ERROR << 16; dev_warn(&h->pdev->dev, "cp %p has " - "protocol error \n", cp); + "protocol error\n", cp); break; case CMD_HARDWARE_ERR: cmd->result = DID_ERROR << 16; @@ -1248,53 +1913,21 @@ static void complete_scsi_command(struct CommandList *cp) cmd->result = DID_ERROR << 16; dev_warn(&h->pdev->dev, "Command unabortable\n"); break; + case CMD_IOACCEL_DISABLED: + /* This only handles the direct pass-through case since RAID + * offload is handled above. Just attempt a retry. + */ + cmd->result = DID_SOFT_ERROR << 16; + dev_warn(&h->pdev->dev, + "cp %p had HP SSD Smart Path error\n", cp); + break; default: cmd->result = DID_ERROR << 16; dev_warn(&h->pdev->dev, "cp %p returned unknown status %x\n", cp, ei->CommandStatus); } - cmd->scsi_done(cmd); cmd_free(h, cp); -} - -static int hpsa_scsi_detect(struct ctlr_info *h) -{ - struct Scsi_Host *sh; - int error; - - sh = scsi_host_alloc(&hpsa_driver_template, sizeof(h)); - if (sh == NULL) - goto fail; - - sh->io_port = 0; - sh->n_io_port = 0; - sh->this_id = -1; - sh->max_channel = 3; - sh->max_cmd_len = MAX_COMMAND_SIZE; - sh->max_lun = HPSA_MAX_LUN; - sh->max_id = HPSA_MAX_LUN; - sh->can_queue = h->nr_cmds; - sh->cmd_per_lun = h->nr_cmds; - sh->sg_tablesize = h->maxsgentries; - h->scsi_host = sh; - sh->hostdata[0] = (unsigned long) h; - sh->irq = h->intr[h->intr_mode]; - sh->unique_id = sh->irq; - error = scsi_add_host(sh, &h->pdev->dev); - if (error) - goto fail_host_put; - scsi_scan_host(sh); - return 0; - - fail_host_put: - dev_err(&h->pdev->dev, "hpsa_scsi_detect: scsi_add_host" - " failed for controller %d\n", h->ctlr); - scsi_host_put(sh); - return error; - fail: - dev_err(&h->pdev->dev, "hpsa_scsi_detect: scsi_host_alloc" - " failed for controller %d\n", h->ctlr); - return -ENOMEM; + cmd->scsi_done(cmd); } static void hpsa_pci_unmap(struct pci_dev *pdev, @@ -1311,7 +1944,7 @@ static void hpsa_pci_unmap(struct pci_dev *pdev, } } -static void hpsa_map_one(struct pci_dev *pdev, +static int hpsa_map_one(struct pci_dev *pdev, struct CommandList *cp, unsigned char *buf, size_t buflen, @@ -1322,17 +1955,25 @@ static void hpsa_map_one(struct pci_dev *pdev, if (buflen == 0 || data_direction == PCI_DMA_NONE) { cp->Header.SGList = 0; cp->Header.SGTotal = 0; - return; + return 0; } addr64 = (u64) pci_map_single(pdev, buf, buflen, data_direction); + if (dma_mapping_error(&pdev->dev, addr64)) { + /* Prevent subsequent unmap of something never mapped */ + cp->Header.SGList = 0; + cp->Header.SGTotal = 0; + return -1; + } cp->SG[0].Addr.lower = (u32) (addr64 & (u64) 0x00000000FFFFFFFF); cp->SG[0].Addr.upper = (u32) ((addr64 >> 32) & (u64) 0x00000000FFFFFFFF); cp->SG[0].Len = buflen; + cp->SG[0].Ext = HPSA_SG_LAST; /* we are not chaining */ cp->Header.SGList = (u8) 1; /* no. SGs contig in this cmd */ cp->Header.SGTotal = (u16) 1; /* total sgs in this cmd list */ + return 0; } static inline void hpsa_scsi_do_simple_cmd_core(struct ctlr_info *h, @@ -1345,46 +1986,80 @@ static inline void hpsa_scsi_do_simple_cmd_core(struct ctlr_info *h, wait_for_completion(&wait); } +static u32 lockup_detected(struct ctlr_info *h) +{ + int cpu; + u32 rc, *lockup_detected; + + cpu = get_cpu(); + lockup_detected = per_cpu_ptr(h->lockup_detected, cpu); + rc = *lockup_detected; + put_cpu(); + return rc; +} + static void hpsa_scsi_do_simple_cmd_core_if_no_lockup(struct ctlr_info *h, struct CommandList *c) { - unsigned long flags; - /* If controller lockup detected, fake a hardware error. */ - spin_lock_irqsave(&h->lock, flags); - if (unlikely(h->lockup_detected)) { - spin_unlock_irqrestore(&h->lock, flags); + if (unlikely(lockup_detected(h))) c->err_info->CommandStatus = CMD_HARDWARE_ERR; - } else { - spin_unlock_irqrestore(&h->lock, flags); + else hpsa_scsi_do_simple_cmd_core(h, c); - } } +#define MAX_DRIVER_CMD_RETRIES 25 static void hpsa_scsi_do_simple_cmd_with_retry(struct ctlr_info *h, struct CommandList *c, int data_direction) { - int retry_count = 0; + int backoff_time = 10, retry_count = 0; do { memset(c->err_info, 0, sizeof(*c->err_info)); hpsa_scsi_do_simple_cmd_core(h, c); retry_count++; - } while (check_for_unit_attention(h, c) && retry_count <= 3); + if (retry_count > 3) { + msleep(backoff_time); + if (backoff_time < 1000) + backoff_time *= 2; + } + } while ((check_for_unit_attention(h, c) || + check_for_busy(h, c)) && + retry_count <= MAX_DRIVER_CMD_RETRIES); hpsa_pci_unmap(h->pdev, c, 1, data_direction); } -static void hpsa_scsi_interpret_error(struct CommandList *cp) +static void hpsa_print_cmd(struct ctlr_info *h, char *txt, + struct CommandList *c) { - struct ErrorInfo *ei; + const u8 *cdb = c->Request.CDB; + const u8 *lun = c->Header.LUN.LunAddrBytes; + + dev_warn(&h->pdev->dev, "%s: LUN:%02x%02x%02x%02x%02x%02x%02x%02x" + " CDB:%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x\n", + txt, lun[0], lun[1], lun[2], lun[3], + lun[4], lun[5], lun[6], lun[7], + cdb[0], cdb[1], cdb[2], cdb[3], + cdb[4], cdb[5], cdb[6], cdb[7], + cdb[8], cdb[9], cdb[10], cdb[11], + cdb[12], cdb[13], cdb[14], cdb[15]); +} + +static void hpsa_scsi_interpret_error(struct ctlr_info *h, + struct CommandList *cp) +{ + const struct ErrorInfo *ei = cp->err_info; struct device *d = &cp->h->pdev->dev; + const u8 *sd = ei->SenseInfo; - ei = cp->err_info; switch (ei->CommandStatus) { case CMD_TARGET_STATUS: - dev_warn(d, "cmd %p has completed with errors\n", cp); - dev_warn(d, "cmd %p has SCSI Status = %x\n", cp, - ei->ScsiStatus); + hpsa_print_cmd(h, "SCSI status", cp); + if (ei->ScsiStatus == SAM_STAT_CHECK_CONDITION) + dev_warn(d, "SCSI Status = 02, Sense key = %02x, ASC = %02x, ASCQ = %02x\n", + sd[2] & 0x0f, sd[12], sd[13]); + else + dev_warn(d, "SCSI Status = %02x\n", ei->ScsiStatus); if (ei->ScsiStatus == 0) dev_warn(d, "SCSI status is abnormally zero. " "(probably indicates selection timeout " @@ -1392,54 +2067,51 @@ static void hpsa_scsi_interpret_error(struct CommandList *cp) "firmware bug, circa July, 2001.)\n"); break; case CMD_DATA_UNDERRUN: /* let mid layer handle it. */ - dev_info(d, "UNDERRUN\n"); break; case CMD_DATA_OVERRUN: - dev_warn(d, "cp %p has completed with data overrun\n", cp); + hpsa_print_cmd(h, "overrun condition", cp); break; case CMD_INVALID: { /* controller unfortunately reports SCSI passthru's * to non-existent targets as invalid commands. */ - dev_warn(d, "cp %p is reported invalid (probably means " - "target device no longer present)\n", cp); - /* print_bytes((unsigned char *) cp, sizeof(*cp), 1, 0); - print_cmd(cp); */ + hpsa_print_cmd(h, "invalid command", cp); + dev_warn(d, "probably means device no longer present\n"); } break; case CMD_PROTOCOL_ERR: - dev_warn(d, "cp %p has protocol error \n", cp); + hpsa_print_cmd(h, "protocol error", cp); break; case CMD_HARDWARE_ERR: - /* cmd->result = DID_ERROR << 16; */ - dev_warn(d, "cp %p had hardware error\n", cp); + hpsa_print_cmd(h, "hardware error", cp); break; case CMD_CONNECTION_LOST: - dev_warn(d, "cp %p had connection lost\n", cp); + hpsa_print_cmd(h, "connection lost", cp); break; case CMD_ABORTED: - dev_warn(d, "cp %p was aborted\n", cp); + hpsa_print_cmd(h, "aborted", cp); break; case CMD_ABORT_FAILED: - dev_warn(d, "cp %p reports abort failed\n", cp); + hpsa_print_cmd(h, "abort failed", cp); break; case CMD_UNSOLICITED_ABORT: - dev_warn(d, "cp %p aborted due to an unsolicited abort\n", cp); + hpsa_print_cmd(h, "unsolicited abort", cp); break; case CMD_TIMEOUT: - dev_warn(d, "cp %p timed out\n", cp); + hpsa_print_cmd(h, "timed out", cp); break; case CMD_UNABORTABLE: - dev_warn(d, "Command unabortable\n"); + hpsa_print_cmd(h, "unabortable", cp); break; default: - dev_warn(d, "cp %p returned unknown status %x\n", cp, + hpsa_print_cmd(h, "unknown status", cp); + dev_warn(d, "Unknown command status %x\n", ei->CommandStatus); } } static int hpsa_scsi_do_inquiry(struct ctlr_info *h, unsigned char *scsi3addr, - unsigned char page, unsigned char *buf, + u16 page, unsigned char *buf, unsigned char bufsize) { int rc = IO_OK; @@ -1453,18 +2125,55 @@ static int hpsa_scsi_do_inquiry(struct ctlr_info *h, unsigned char *scsi3addr, return -ENOMEM; } - fill_cmd(c, HPSA_INQUIRY, h, buf, bufsize, page, scsi3addr, TYPE_CMD); + if (fill_cmd(c, HPSA_INQUIRY, h, buf, bufsize, + page, scsi3addr, TYPE_CMD)) { + rc = -1; + goto out; + } hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_FROMDEVICE); ei = c->err_info; if (ei->CommandStatus != 0 && ei->CommandStatus != CMD_DATA_UNDERRUN) { - hpsa_scsi_interpret_error(c); + hpsa_scsi_interpret_error(h, c); rc = -1; } +out: cmd_special_free(h, c); return rc; } -static int hpsa_send_reset(struct ctlr_info *h, unsigned char *scsi3addr) +static int hpsa_bmic_ctrl_mode_sense(struct ctlr_info *h, + unsigned char *scsi3addr, unsigned char page, + struct bmic_controller_parameters *buf, size_t bufsize) +{ + int rc = IO_OK; + struct CommandList *c; + struct ErrorInfo *ei; + + c = cmd_special_alloc(h); + + if (c == NULL) { /* trouble... */ + dev_warn(&h->pdev->dev, "cmd_special_alloc returned NULL!\n"); + return -ENOMEM; + } + + if (fill_cmd(c, BMIC_SENSE_CONTROLLER_PARAMETERS, h, buf, bufsize, + page, scsi3addr, TYPE_CMD)) { + rc = -1; + goto out; + } + hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_FROMDEVICE); + ei = c->err_info; + if (ei->CommandStatus != 0 && ei->CommandStatus != CMD_DATA_UNDERRUN) { + hpsa_scsi_interpret_error(h, c); + rc = -1; + } +out: + cmd_special_free(h, c); + return rc; + } + +static int hpsa_send_reset(struct ctlr_info *h, unsigned char *scsi3addr, + u8 reset_type) { int rc = IO_OK; struct CommandList *c; @@ -1477,13 +2186,16 @@ static int hpsa_send_reset(struct ctlr_info *h, unsigned char *scsi3addr) return -ENOMEM; } - fill_cmd(c, HPSA_DEVICE_RESET_MSG, h, NULL, 0, 0, scsi3addr, TYPE_MSG); + /* fill_cmd can't fail here, no data buffer to map. */ + (void) fill_cmd(c, HPSA_DEVICE_RESET_MSG, h, NULL, 0, 0, + scsi3addr, TYPE_MSG); + c->Request.CDB[1] = reset_type; /* fill_cmd defaults to LUN reset */ hpsa_scsi_do_simple_cmd_core(h, c); /* no unmap needed here because no data xfer. */ ei = c->err_info; if (ei->CommandStatus != 0) { - hpsa_scsi_interpret_error(c); + hpsa_scsi_interpret_error(h, c); rc = -1; } cmd_special_free(h, c); @@ -1500,7 +2212,7 @@ static void hpsa_get_raid_level(struct ctlr_info *h, buf = kzalloc(64, GFP_KERNEL); if (!buf) return; - rc = hpsa_scsi_do_inquiry(h, scsi3addr, 0xC1, buf, 64); + rc = hpsa_scsi_do_inquiry(h, scsi3addr, VPD_PAGE | 0xC1, buf, 64); if (rc == 0) *raid_level = buf[8]; if (*raid_level > RAID_UNKNOWN) @@ -1509,6 +2221,204 @@ static void hpsa_get_raid_level(struct ctlr_info *h, return; } +#define HPSA_MAP_DEBUG +#ifdef HPSA_MAP_DEBUG +static void hpsa_debug_map_buff(struct ctlr_info *h, int rc, + struct raid_map_data *map_buff) +{ + struct raid_map_disk_data *dd = &map_buff->data[0]; + int map, row, col; + u16 map_cnt, row_cnt, disks_per_row; + + if (rc != 0) + return; + + /* Show details only if debugging has been activated. */ + if (h->raid_offload_debug < 2) + return; + + dev_info(&h->pdev->dev, "structure_size = %u\n", + le32_to_cpu(map_buff->structure_size)); + dev_info(&h->pdev->dev, "volume_blk_size = %u\n", + le32_to_cpu(map_buff->volume_blk_size)); + dev_info(&h->pdev->dev, "volume_blk_cnt = 0x%llx\n", + le64_to_cpu(map_buff->volume_blk_cnt)); + dev_info(&h->pdev->dev, "physicalBlockShift = %u\n", + map_buff->phys_blk_shift); + dev_info(&h->pdev->dev, "parity_rotation_shift = %u\n", + map_buff->parity_rotation_shift); + dev_info(&h->pdev->dev, "strip_size = %u\n", + le16_to_cpu(map_buff->strip_size)); + dev_info(&h->pdev->dev, "disk_starting_blk = 0x%llx\n", + le64_to_cpu(map_buff->disk_starting_blk)); + dev_info(&h->pdev->dev, "disk_blk_cnt = 0x%llx\n", + le64_to_cpu(map_buff->disk_blk_cnt)); + dev_info(&h->pdev->dev, "data_disks_per_row = %u\n", + le16_to_cpu(map_buff->data_disks_per_row)); + dev_info(&h->pdev->dev, "metadata_disks_per_row = %u\n", + le16_to_cpu(map_buff->metadata_disks_per_row)); + dev_info(&h->pdev->dev, "row_cnt = %u\n", + le16_to_cpu(map_buff->row_cnt)); + dev_info(&h->pdev->dev, "layout_map_count = %u\n", + le16_to_cpu(map_buff->layout_map_count)); + dev_info(&h->pdev->dev, "flags = %u\n", + le16_to_cpu(map_buff->flags)); + if (map_buff->flags & RAID_MAP_FLAG_ENCRYPT_ON) + dev_info(&h->pdev->dev, "encrypytion = ON\n"); + else + dev_info(&h->pdev->dev, "encrypytion = OFF\n"); + dev_info(&h->pdev->dev, "dekindex = %u\n", + le16_to_cpu(map_buff->dekindex)); + + map_cnt = le16_to_cpu(map_buff->layout_map_count); + for (map = 0; map < map_cnt; map++) { + dev_info(&h->pdev->dev, "Map%u:\n", map); + row_cnt = le16_to_cpu(map_buff->row_cnt); + for (row = 0; row < row_cnt; row++) { + dev_info(&h->pdev->dev, " Row%u:\n", row); + disks_per_row = + le16_to_cpu(map_buff->data_disks_per_row); + for (col = 0; col < disks_per_row; col++, dd++) + dev_info(&h->pdev->dev, + " D%02u: h=0x%04x xor=%u,%u\n", + col, dd->ioaccel_handle, + dd->xor_mult[0], dd->xor_mult[1]); + disks_per_row = + le16_to_cpu(map_buff->metadata_disks_per_row); + for (col = 0; col < disks_per_row; col++, dd++) + dev_info(&h->pdev->dev, + " M%02u: h=0x%04x xor=%u,%u\n", + col, dd->ioaccel_handle, + dd->xor_mult[0], dd->xor_mult[1]); + } + } +} +#else +static void hpsa_debug_map_buff(__attribute__((unused)) struct ctlr_info *h, + __attribute__((unused)) int rc, + __attribute__((unused)) struct raid_map_data *map_buff) +{ +} +#endif + +static int hpsa_get_raid_map(struct ctlr_info *h, + unsigned char *scsi3addr, struct hpsa_scsi_dev_t *this_device) +{ + int rc = 0; + struct CommandList *c; + struct ErrorInfo *ei; + + c = cmd_special_alloc(h); + if (c == NULL) { + dev_warn(&h->pdev->dev, "cmd_special_alloc returned NULL!\n"); + return -ENOMEM; + } + if (fill_cmd(c, HPSA_GET_RAID_MAP, h, &this_device->raid_map, + sizeof(this_device->raid_map), 0, + scsi3addr, TYPE_CMD)) { + dev_warn(&h->pdev->dev, "Out of memory in hpsa_get_raid_map()\n"); + cmd_special_free(h, c); + return -ENOMEM; + } + hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_FROMDEVICE); + ei = c->err_info; + if (ei->CommandStatus != 0 && ei->CommandStatus != CMD_DATA_UNDERRUN) { + hpsa_scsi_interpret_error(h, c); + cmd_special_free(h, c); + return -1; + } + cmd_special_free(h, c); + + /* @todo in the future, dynamically allocate RAID map memory */ + if (le32_to_cpu(this_device->raid_map.structure_size) > + sizeof(this_device->raid_map)) { + dev_warn(&h->pdev->dev, "RAID map size is too large!\n"); + rc = -1; + } + hpsa_debug_map_buff(h, rc, &this_device->raid_map); + return rc; +} + +static int hpsa_vpd_page_supported(struct ctlr_info *h, + unsigned char scsi3addr[], u8 page) +{ + int rc; + int i; + int pages; + unsigned char *buf, bufsize; + + buf = kzalloc(256, GFP_KERNEL); + if (!buf) + return 0; + + /* Get the size of the page list first */ + rc = hpsa_scsi_do_inquiry(h, scsi3addr, + VPD_PAGE | HPSA_VPD_SUPPORTED_PAGES, + buf, HPSA_VPD_HEADER_SZ); + if (rc != 0) + goto exit_unsupported; + pages = buf[3]; + if ((pages + HPSA_VPD_HEADER_SZ) <= 255) + bufsize = pages + HPSA_VPD_HEADER_SZ; + else + bufsize = 255; + + /* Get the whole VPD page list */ + rc = hpsa_scsi_do_inquiry(h, scsi3addr, + VPD_PAGE | HPSA_VPD_SUPPORTED_PAGES, + buf, bufsize); + if (rc != 0) + goto exit_unsupported; + + pages = buf[3]; + for (i = 1; i <= pages; i++) + if (buf[3 + i] == page) + goto exit_supported; +exit_unsupported: + kfree(buf); + return 0; +exit_supported: + kfree(buf); + return 1; +} + +static void hpsa_get_ioaccel_status(struct ctlr_info *h, + unsigned char *scsi3addr, struct hpsa_scsi_dev_t *this_device) +{ + int rc; + unsigned char *buf; + u8 ioaccel_status; + + this_device->offload_config = 0; + this_device->offload_enabled = 0; + + buf = kzalloc(64, GFP_KERNEL); + if (!buf) + return; + if (!hpsa_vpd_page_supported(h, scsi3addr, HPSA_VPD_LV_IOACCEL_STATUS)) + goto out; + rc = hpsa_scsi_do_inquiry(h, scsi3addr, + VPD_PAGE | HPSA_VPD_LV_IOACCEL_STATUS, buf, 64); + if (rc != 0) + goto out; + +#define IOACCEL_STATUS_BYTE 4 +#define OFFLOAD_CONFIGURED_BIT 0x01 +#define OFFLOAD_ENABLED_BIT 0x02 + ioaccel_status = buf[IOACCEL_STATUS_BYTE]; + this_device->offload_config = + !!(ioaccel_status & OFFLOAD_CONFIGURED_BIT); + if (this_device->offload_config) { + this_device->offload_enabled = + !!(ioaccel_status & OFFLOAD_ENABLED_BIT); + if (hpsa_get_raid_map(h, scsi3addr, this_device)) + this_device->offload_enabled = 0; + } +out: + kfree(buf); + return; +} + /* Get the device id from inquiry page 0x83 */ static int hpsa_get_device_id(struct ctlr_info *h, unsigned char *scsi3addr, unsigned char *device_id, int buflen) @@ -1520,8 +2430,8 @@ static int hpsa_get_device_id(struct ctlr_info *h, unsigned char *scsi3addr, buflen = 16; buf = kzalloc(64, GFP_KERNEL); if (!buf) - return -1; - rc = hpsa_scsi_do_inquiry(h, scsi3addr, 0x83, buf, 64); + return -ENOMEM; + rc = hpsa_scsi_do_inquiry(h, scsi3addr, VPD_PAGE | 0x83, buf, 64); if (rc == 0) memcpy(device_id, &buf[8], buflen); kfree(buf); @@ -1544,17 +2454,29 @@ static int hpsa_scsi_do_report_luns(struct ctlr_info *h, int logical, } /* address the controller */ memset(scsi3addr, 0, sizeof(scsi3addr)); - fill_cmd(c, logical ? HPSA_REPORT_LOG : HPSA_REPORT_PHYS, h, - buf, bufsize, 0, scsi3addr, TYPE_CMD); + if (fill_cmd(c, logical ? HPSA_REPORT_LOG : HPSA_REPORT_PHYS, h, + buf, bufsize, 0, scsi3addr, TYPE_CMD)) { + rc = -1; + goto out; + } if (extended_response) c->Request.CDB[1] = extended_response; hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_FROMDEVICE); ei = c->err_info; if (ei->CommandStatus != 0 && ei->CommandStatus != CMD_DATA_UNDERRUN) { - hpsa_scsi_interpret_error(c); + hpsa_scsi_interpret_error(h, c); rc = -1; + } else { + if (buf->extended_response_flag != extended_response) { + dev_err(&h->pdev->dev, + "report luns requested format %u, got %u\n", + extended_response, + buf->extended_response_flag); + rc = -1; + } } +out: cmd_special_free(h, c); return rc; } @@ -1580,6 +2502,111 @@ static inline void hpsa_set_bus_target_lun(struct hpsa_scsi_dev_t *device, device->lun = lun; } +/* Use VPD inquiry to get details of volume status */ +static int hpsa_get_volume_status(struct ctlr_info *h, + unsigned char scsi3addr[]) +{ + int rc; + int status; + int size; + unsigned char *buf; + + buf = kzalloc(64, GFP_KERNEL); + if (!buf) + return HPSA_VPD_LV_STATUS_UNSUPPORTED; + + /* Does controller have VPD for logical volume status? */ + if (!hpsa_vpd_page_supported(h, scsi3addr, HPSA_VPD_LV_STATUS)) + goto exit_failed; + + /* Get the size of the VPD return buffer */ + rc = hpsa_scsi_do_inquiry(h, scsi3addr, VPD_PAGE | HPSA_VPD_LV_STATUS, + buf, HPSA_VPD_HEADER_SZ); + if (rc != 0) + goto exit_failed; + size = buf[3]; + + /* Now get the whole VPD buffer */ + rc = hpsa_scsi_do_inquiry(h, scsi3addr, VPD_PAGE | HPSA_VPD_LV_STATUS, + buf, size + HPSA_VPD_HEADER_SZ); + if (rc != 0) + goto exit_failed; + status = buf[4]; /* status byte */ + + kfree(buf); + return status; +exit_failed: + kfree(buf); + return HPSA_VPD_LV_STATUS_UNSUPPORTED; +} + +/* Determine offline status of a volume. + * Return either: + * 0 (not offline) + * 0xff (offline for unknown reasons) + * # (integer code indicating one of several NOT READY states + * describing why a volume is to be kept offline) + */ +static int hpsa_volume_offline(struct ctlr_info *h, + unsigned char scsi3addr[]) +{ + struct CommandList *c; + unsigned char *sense, sense_key, asc, ascq; + int ldstat = 0; + u16 cmd_status; + u8 scsi_status; +#define ASC_LUN_NOT_READY 0x04 +#define ASCQ_LUN_NOT_READY_FORMAT_IN_PROGRESS 0x04 +#define ASCQ_LUN_NOT_READY_INITIALIZING_CMD_REQ 0x02 + + c = cmd_alloc(h); + if (!c) + return 0; + (void) fill_cmd(c, TEST_UNIT_READY, h, NULL, 0, 0, scsi3addr, TYPE_CMD); + hpsa_scsi_do_simple_cmd_core(h, c); + sense = c->err_info->SenseInfo; + sense_key = sense[2]; + asc = sense[12]; + ascq = sense[13]; + cmd_status = c->err_info->CommandStatus; + scsi_status = c->err_info->ScsiStatus; + cmd_free(h, c); + /* Is the volume 'not ready'? */ + if (cmd_status != CMD_TARGET_STATUS || + scsi_status != SAM_STAT_CHECK_CONDITION || + sense_key != NOT_READY || + asc != ASC_LUN_NOT_READY) { + return 0; + } + + /* Determine the reason for not ready state */ + ldstat = hpsa_get_volume_status(h, scsi3addr); + + /* Keep volume offline in certain cases: */ + switch (ldstat) { + case HPSA_LV_UNDERGOING_ERASE: + case HPSA_LV_UNDERGOING_RPI: + case HPSA_LV_PENDING_RPI: + case HPSA_LV_ENCRYPTED_NO_KEY: + case HPSA_LV_PLAINTEXT_IN_ENCRYPT_ONLY_CONTROLLER: + case HPSA_LV_UNDERGOING_ENCRYPTION: + case HPSA_LV_UNDERGOING_ENCRYPTION_REKEYING: + case HPSA_LV_ENCRYPTED_IN_NON_ENCRYPTED_CONTROLLER: + return ldstat; + case HPSA_VPD_LV_STATUS_UNSUPPORTED: + /* If VPD status page isn't available, + * use ASC/ASCQ to determine state + */ + if ((ascq == ASCQ_LUN_NOT_READY_FORMAT_IN_PROGRESS) || + (ascq == ASCQ_LUN_NOT_READY_INITIALIZING_CMD_REQ)) + return ldstat; + break; + default: + break; + } + return 0; +} + static int hpsa_update_device_info(struct ctlr_info *h, unsigned char scsi3addr[], struct hpsa_scsi_dev_t *this_device, unsigned char *is_OBDR_device) @@ -1618,10 +2645,22 @@ static int hpsa_update_device_info(struct ctlr_info *h, sizeof(this_device->device_id)); if (this_device->devtype == TYPE_DISK && - is_logical_dev_addr_mode(scsi3addr)) + is_logical_dev_addr_mode(scsi3addr)) { + int volume_offline; + hpsa_get_raid_level(h, scsi3addr, &this_device->raid_level); - else + if (h->fw_support & MISC_FW_RAID_OFFLOAD_BASIC) + hpsa_get_ioaccel_status(h, scsi3addr, this_device); + volume_offline = hpsa_volume_offline(h, scsi3addr); + if (volume_offline < 0 || volume_offline > 0xff) + volume_offline = HPSA_VPD_LV_STATUS_UNSUPPORTED; + this_device->volume_offline = volume_offline & 0xff; + } else { this_device->raid_level = RAID_UNKNOWN; + this_device->offload_config = 0; + this_device->offload_enabled = 0; + this_device->volume_offline = 0; + } if (is_OBDR_device) { /* See if this is a One-Button-Disaster-Recovery device @@ -1641,87 +2680,64 @@ bail_out: return 1; } -static unsigned char *msa2xxx_model[] = { +static unsigned char *ext_target_model[] = { "MSA2012", "MSA2024", "MSA2312", "MSA2324", "P2000 G3 SAS", + "MSA 2040 SAS", NULL, }; -static int is_msa2xxx(struct ctlr_info *h, struct hpsa_scsi_dev_t *device) +static int is_ext_target(struct ctlr_info *h, struct hpsa_scsi_dev_t *device) { int i; - for (i = 0; msa2xxx_model[i]; i++) - if (strncmp(device->model, msa2xxx_model[i], - strlen(msa2xxx_model[i])) == 0) + for (i = 0; ext_target_model[i]; i++) + if (strncmp(device->model, ext_target_model[i], + strlen(ext_target_model[i])) == 0) return 1; return 0; } /* Helper function to assign bus, target, lun mapping of devices. - * Puts non-msa2xxx logical volumes on bus 0, msa2xxx logical + * Puts non-external target logical volumes on bus 0, external target logical * volumes on bus 1, physical devices on bus 2. and the hba on bus 3. * Logical drive target and lun are assigned at this time, but * physical device lun and target assignment are deferred (assigned * in hpsa_find_target_lun, called by hpsa_scsi_add_entry.) */ static void figure_bus_target_lun(struct ctlr_info *h, - u8 *lunaddrbytes, int *bus, int *target, int *lun, - struct hpsa_scsi_dev_t *device) + u8 *lunaddrbytes, struct hpsa_scsi_dev_t *device) { - u32 lunid; + u32 lunid = le32_to_cpu(*((__le32 *) lunaddrbytes)); - if (is_logical_dev_addr_mode(lunaddrbytes)) { - /* logical device */ - if (unlikely(is_scsi_rev_5(h))) { - /* p1210m, logical drives lun assignments - * match SCSI REPORT LUNS data. - */ - lunid = le32_to_cpu(*((__le32 *) lunaddrbytes)); - *bus = 0; - *target = 0; - *lun = (lunid & 0x3fff) + 1; - } else { - /* not p1210m... */ - lunid = le32_to_cpu(*((__le32 *) lunaddrbytes)); - if (is_msa2xxx(h, device)) { - /* msa2xxx way, put logicals on bus 1 - * and match target/lun numbers box - * reports. - */ - *bus = 1; - *target = (lunid >> 16) & 0x3fff; - *lun = lunid & 0x00ff; - } else { - /* Traditional smart array way. */ - *bus = 0; - *lun = 0; - *target = lunid & 0x3fff; - } - } - } else { - /* physical device */ + if (!is_logical_dev_addr_mode(lunaddrbytes)) { + /* physical device, target and lun filled in later */ if (is_hba_lunid(lunaddrbytes)) - if (unlikely(is_scsi_rev_5(h))) { - *bus = 0; /* put p1210m ctlr at 0,0,0 */ - *target = 0; - *lun = 0; - return; - } else - *bus = 3; /* traditional smartarray */ + hpsa_set_bus_target_lun(device, 3, 0, lunid & 0x3fff); else - *bus = 2; /* physical disk */ - *target = -1; - *lun = -1; /* we will fill these in later. */ + /* defer target, lun assignment for physical devices */ + hpsa_set_bus_target_lun(device, 2, -1, -1); + return; + } + /* It's a logical device */ + if (is_ext_target(h, device)) { + /* external target way, put logicals on bus 1 + * and match target/lun numbers box + * reports, other smart array, bus 0, target 0, match lunid + */ + hpsa_set_bus_target_lun(device, + 1, (lunid >> 16) & 0x3fff, lunid & 0x00ff); + return; } + hpsa_set_bus_target_lun(device, 0, 0, lunid & 0x3fff); } /* * If there is no lun 0 on a target, linux won't find any devices. - * For the MSA2xxx boxes, we have to manually detect the enclosure + * For the external targets (arrays), we have to manually detect the enclosure * which is at lun zero, as CCISS_REPORT_PHYSICAL_LUNS doesn't report * it for some reason. *tmpdevice is the target we're adding, * this_device is a pointer into the current element of currentsd[] @@ -1730,50 +2746,145 @@ static void figure_bus_target_lun(struct ctlr_info *h, * lun 0 assigned. * Returns 1 if an enclosure was added, 0 if not. */ -static int add_msa2xxx_enclosure_device(struct ctlr_info *h, +static int add_ext_target_dev(struct ctlr_info *h, struct hpsa_scsi_dev_t *tmpdevice, struct hpsa_scsi_dev_t *this_device, u8 *lunaddrbytes, - int bus, int target, int lun, unsigned long lunzerobits[], - int *nmsa2xxx_enclosures) + unsigned long lunzerobits[], int *n_ext_target_devs) { unsigned char scsi3addr[8]; - if (test_bit(target, lunzerobits)) + if (test_bit(tmpdevice->target, lunzerobits)) return 0; /* There is already a lun 0 on this target. */ if (!is_logical_dev_addr_mode(lunaddrbytes)) return 0; /* It's the logical targets that may lack lun 0. */ - if (!is_msa2xxx(h, tmpdevice)) - return 0; /* It's only the MSA2xxx that have this problem. */ + if (!is_ext_target(h, tmpdevice)) + return 0; /* Only external target devices have this problem. */ - if (lun == 0) /* if lun is 0, then obviously we have a lun 0. */ + if (tmpdevice->lun == 0) /* if lun is 0, then we have a lun 0. */ return 0; memset(scsi3addr, 0, 8); - scsi3addr[3] = target; + scsi3addr[3] = tmpdevice->target; if (is_hba_lunid(scsi3addr)) return 0; /* Don't add the RAID controller here. */ if (is_scsi_rev_5(h)) return 0; /* p1210m doesn't need to do this. */ - if (*nmsa2xxx_enclosures >= MAX_MSA2XXX_ENCLOSURES) { - dev_warn(&h->pdev->dev, "Maximum number of MSA2XXX " - "enclosures exceeded. Check your hardware " + if (*n_ext_target_devs >= MAX_EXT_TARGETS) { + dev_warn(&h->pdev->dev, "Maximum number of external " + "target devices exceeded. Check your hardware " "configuration."); return 0; } if (hpsa_update_device_info(h, scsi3addr, this_device, NULL)) return 0; - (*nmsa2xxx_enclosures)++; - hpsa_set_bus_target_lun(this_device, bus, target, 0); - set_bit(target, lunzerobits); + (*n_ext_target_devs)++; + hpsa_set_bus_target_lun(this_device, + tmpdevice->bus, tmpdevice->target, 0); + set_bit(tmpdevice->target, lunzerobits); return 1; } /* + * Get address of physical disk used for an ioaccel2 mode command: + * 1. Extract ioaccel2 handle from the command. + * 2. Find a matching ioaccel2 handle from list of physical disks. + * 3. Return: + * 1 and set scsi3addr to address of matching physical + * 0 if no matching physical disk was found. + */ +static int hpsa_get_pdisk_of_ioaccel2(struct ctlr_info *h, + struct CommandList *ioaccel2_cmd_to_abort, unsigned char *scsi3addr) +{ + struct ReportExtendedLUNdata *physicals = NULL; + int responsesize = 24; /* size of physical extended response */ + int extended = 2; /* flag forces reporting 'other dev info'. */ + int reportsize = sizeof(*physicals) + HPSA_MAX_PHYS_LUN * responsesize; + u32 nphysicals = 0; /* number of reported physical devs */ + int found = 0; /* found match (1) or not (0) */ + u32 find; /* handle we need to match */ + int i; + struct scsi_cmnd *scmd; /* scsi command within request being aborted */ + struct hpsa_scsi_dev_t *d; /* device of request being aborted */ + struct io_accel2_cmd *c2a; /* ioaccel2 command to abort */ + u32 it_nexus; /* 4 byte device handle for the ioaccel2 cmd */ + u32 scsi_nexus; /* 4 byte device handle for the ioaccel2 cmd */ + + if (ioaccel2_cmd_to_abort->cmd_type != CMD_IOACCEL2) + return 0; /* no match */ + + /* point to the ioaccel2 device handle */ + c2a = &h->ioaccel2_cmd_pool[ioaccel2_cmd_to_abort->cmdindex]; + if (c2a == NULL) + return 0; /* no match */ + + scmd = (struct scsi_cmnd *) ioaccel2_cmd_to_abort->scsi_cmd; + if (scmd == NULL) + return 0; /* no match */ + + d = scmd->device->hostdata; + if (d == NULL) + return 0; /* no match */ + + it_nexus = cpu_to_le32((u32) d->ioaccel_handle); + scsi_nexus = cpu_to_le32((u32) c2a->scsi_nexus); + find = c2a->scsi_nexus; + + if (h->raid_offload_debug > 0) + dev_info(&h->pdev->dev, + "%s: scsi_nexus:0x%08x device id: 0x%02x%02x%02x%02x %02x%02x%02x%02x %02x%02x%02x%02x %02x%02x%02x%02x\n", + __func__, scsi_nexus, + d->device_id[0], d->device_id[1], d->device_id[2], + d->device_id[3], d->device_id[4], d->device_id[5], + d->device_id[6], d->device_id[7], d->device_id[8], + d->device_id[9], d->device_id[10], d->device_id[11], + d->device_id[12], d->device_id[13], d->device_id[14], + d->device_id[15]); + + /* Get the list of physical devices */ + physicals = kzalloc(reportsize, GFP_KERNEL); + if (physicals == NULL) + return 0; + if (hpsa_scsi_do_report_phys_luns(h, (struct ReportLUNdata *) physicals, + reportsize, extended)) { + dev_err(&h->pdev->dev, + "Can't lookup %s device handle: report physical LUNs failed.\n", + "HP SSD Smart Path"); + kfree(physicals); + return 0; + } + nphysicals = be32_to_cpu(*((__be32 *)physicals->LUNListLength)) / + responsesize; + + /* find ioaccel2 handle in list of physicals: */ + for (i = 0; i < nphysicals; i++) { + struct ext_report_lun_entry *entry = &physicals->LUN[i]; + + /* handle is in bytes 28-31 of each lun */ + if (entry->ioaccel_handle != find) + continue; /* didn't match */ + found = 1; + memcpy(scsi3addr, entry->lunid, 8); + if (h->raid_offload_debug > 0) + dev_info(&h->pdev->dev, + "%s: Searched h=0x%08x, Found h=0x%08x, scsiaddr 0x%8phN\n", + __func__, find, + entry->ioaccel_handle, scsi3addr); + break; /* found it */ + } + + kfree(physicals); + if (found) + return 1; + else + return 0; + +} +/* * Do CISS_REPORT_PHYS and CISS_REPORT_LOG. Data is returned in physdev, * logdev. The number of luns in physdev and logdev are returned in * *nphysicals and *nlogicals, respectively. @@ -1781,14 +2892,26 @@ static int add_msa2xxx_enclosure_device(struct ctlr_info *h, */ static int hpsa_gather_lun_info(struct ctlr_info *h, int reportlunsize, - struct ReportLUNdata *physdev, u32 *nphysicals, + struct ReportLUNdata *physdev, u32 *nphysicals, int *physical_mode, struct ReportLUNdata *logdev, u32 *nlogicals) { - if (hpsa_scsi_do_report_phys_luns(h, physdev, reportlunsize, 0)) { + int physical_entry_size = 8; + + *physical_mode = 0; + + /* For I/O accelerator mode we need to read physical device handles */ + if (h->transMethod & CFGTBL_Trans_io_accel1 || + h->transMethod & CFGTBL_Trans_io_accel2) { + *physical_mode = HPSA_REPORT_PHYS_EXTENDED; + physical_entry_size = 24; + } + if (hpsa_scsi_do_report_phys_luns(h, physdev, reportlunsize, + *physical_mode)) { dev_err(&h->pdev->dev, "report physical LUNs failed.\n"); return -1; } - *nphysicals = be32_to_cpu(*((__be32 *)physdev->LUNListLength)) / 8; + *nphysicals = be32_to_cpu(*((__be32 *)physdev->LUNListLength)) / + physical_entry_size; if (*nphysicals > HPSA_MAX_PHYS_LUN) { dev_warn(&h->pdev->dev, "maximum physical LUNs (%d) exceeded." " %d LUNs ignored.\n", HPSA_MAX_PHYS_LUN, @@ -1819,7 +2942,8 @@ static int hpsa_gather_lun_info(struct ctlr_info *h, } u8 *figure_lunaddrbytes(struct ctlr_info *h, int raid_ctlr_position, int i, - int nphysicals, int nlogicals, struct ReportLUNdata *physdev_list, + int nphysicals, int nlogicals, + struct ReportExtendedLUNdata *physdev_list, struct ReportLUNdata *logdev_list) { /* Helper function, figure out where the LUN ID info is coming from @@ -1834,7 +2958,8 @@ u8 *figure_lunaddrbytes(struct ctlr_info *h, int raid_ctlr_position, int i, return RAID_CTLR_LUNID; if (i < logicals_start) - return &physdev_list->LUN[i - (raid_ctlr_position == 0)][0]; + return &physdev_list->LUN[i - + (raid_ctlr_position == 0)].lunid[0]; if (i < last_device) return &logdev_list->LUN[i - nphysicals - @@ -1843,6 +2968,29 @@ u8 *figure_lunaddrbytes(struct ctlr_info *h, int raid_ctlr_position, int i, return NULL; } +static int hpsa_hba_mode_enabled(struct ctlr_info *h) +{ + int rc; + int hba_mode_enabled; + struct bmic_controller_parameters *ctlr_params; + ctlr_params = kzalloc(sizeof(struct bmic_controller_parameters), + GFP_KERNEL); + + if (!ctlr_params) + return -ENOMEM; + rc = hpsa_bmic_ctrl_mode_sense(h, RAID_CTLR_LUNID, 0, ctlr_params, + sizeof(struct bmic_controller_parameters)); + if (rc) { + kfree(ctlr_params); + return rc; + } + + hba_mode_enabled = + ((ctlr_params->nvram_flags & HBA_MODE_ENABLED_FLAG) != 0); + kfree(ctlr_params); + return hba_mode_enabled; +} + static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) { /* the idea here is we could get notified @@ -1855,18 +3003,19 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) * tell which devices we already know about, vs. new * devices, vs. disappearing devices. */ - struct ReportLUNdata *physdev_list = NULL; + struct ReportExtendedLUNdata *physdev_list = NULL; struct ReportLUNdata *logdev_list = NULL; u32 nphysicals = 0; u32 nlogicals = 0; + int physical_mode = 0; u32 ndev_allocated = 0; struct hpsa_scsi_dev_t **currentsd, *this_device, *tmpdevice; int ncurrent = 0; - int reportlunsize = sizeof(*physdev_list) + HPSA_MAX_PHYS_LUN * 8; - int i, nmsa2xxx_enclosures, ndevs_to_allocate; - int bus, target, lun; + int reportlunsize = sizeof(*physdev_list) + HPSA_MAX_PHYS_LUN * 24; + int i, n_ext_target_devs, ndevs_to_allocate; int raid_ctlr_position; - DECLARE_BITMAP(lunzerobits, HPSA_MAX_TARGETS_PER_CTLR); + int rescan_hba_mode; + DECLARE_BITMAP(lunzerobits, MAX_EXT_TARGETS); currentsd = kzalloc(sizeof(*currentsd) * HPSA_MAX_DEVICES, GFP_KERNEL); physdev_list = kzalloc(reportlunsize, GFP_KERNEL); @@ -1879,15 +3028,27 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) } memset(lunzerobits, 0, sizeof(lunzerobits)); - if (hpsa_gather_lun_info(h, reportlunsize, physdev_list, &nphysicals, - logdev_list, &nlogicals)) + rescan_hba_mode = hpsa_hba_mode_enabled(h); + if (rescan_hba_mode < 0) goto out; - /* We might see up to 32 MSA2xxx enclosures, actually 8 of them - * but each of them 4 times through different paths. The plus 1 - * is for the RAID controller. + if (!h->hba_mode_enabled && rescan_hba_mode) + dev_warn(&h->pdev->dev, "HBA mode enabled\n"); + else if (h->hba_mode_enabled && !rescan_hba_mode) + dev_warn(&h->pdev->dev, "HBA mode disabled\n"); + + h->hba_mode_enabled = rescan_hba_mode; + + if (hpsa_gather_lun_info(h, reportlunsize, + (struct ReportLUNdata *) physdev_list, &nphysicals, + &physical_mode, logdev_list, &nlogicals)) + goto out; + + /* We might see up to the maximum number of logical and physical disks + * plus external target devices, and a device for the local RAID + * controller. */ - ndevs_to_allocate = nphysicals + nlogicals + MAX_MSA2XXX_ENCLOSURES + 1; + ndevs_to_allocate = nphysicals + nlogicals + MAX_EXT_TARGETS + 1; /* Allocate the per device structures */ for (i = 0; i < ndevs_to_allocate; i++) { @@ -1907,13 +3068,13 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) ndev_allocated++; } - if (unlikely(is_scsi_rev_5(h))) + if (is_scsi_rev_5(h)) raid_ctlr_position = 0; else raid_ctlr_position = nphysicals + nlogicals; /* adjust our table of devices */ - nmsa2xxx_enclosures = 0; + n_ext_target_devs = 0; for (i = 0; i < nphysicals + nlogicals + 1; i++) { u8 *lunaddrbytes, is_OBDR = 0; @@ -1929,26 +3090,24 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) if (hpsa_update_device_info(h, lunaddrbytes, tmpdevice, &is_OBDR)) continue; /* skip it if we can't talk to it. */ - figure_bus_target_lun(h, lunaddrbytes, &bus, &target, &lun, - tmpdevice); + figure_bus_target_lun(h, lunaddrbytes, tmpdevice); this_device = currentsd[ncurrent]; /* - * For the msa2xxx boxes, we have to insert a LUN 0 which + * For external target devices, we have to insert a LUN 0 which * doesn't show up in CCISS_REPORT_PHYSICAL data, but there * is nonetheless an enclosure device there. We have to * present that otherwise linux won't find anything if * there is no lun 0. */ - if (add_msa2xxx_enclosure_device(h, tmpdevice, this_device, - lunaddrbytes, bus, target, lun, lunzerobits, - &nmsa2xxx_enclosures)) { + if (add_ext_target_dev(h, tmpdevice, this_device, + lunaddrbytes, lunzerobits, + &n_ext_target_devs)) { ncurrent++; this_device = currentsd[ncurrent]; } *this_device = *tmpdevice; - hpsa_set_bus_target_lun(this_device, bus, target, lun); switch (this_device->devtype) { case TYPE_ROM: @@ -1963,9 +3122,28 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) ncurrent++; break; case TYPE_DISK: - if (i < nphysicals) + if (h->hba_mode_enabled) { + /* never use raid mapper in HBA mode */ + this_device->offload_enabled = 0; + ncurrent++; break; - ncurrent++; + } else if (h->acciopath_status) { + if (i >= nphysicals) { + ncurrent++; + break; + } + } else { + if (i < nphysicals) + break; + ncurrent++; + break; + } + if (physical_mode == HPSA_REPORT_PHYS_EXTENDED) { + memcpy(&this_device->ioaccel_handle, + &lunaddrbytes[20], + sizeof(this_device->ioaccel_handle)); + ncurrent++; + } break; case TYPE_TAPE: case TYPE_MEDIUM_CHANGER: @@ -2035,7 +3213,7 @@ static int hpsa_scatter_gather(struct ctlr_info *h, curr_sg->Addr.lower = (u32) (addr64 & 0x0FFFFFFFFULL); curr_sg->Addr.upper = (u32) ((addr64 >> 32) & 0x0FFFFFFFFULL); curr_sg->Len = len; - curr_sg->Ext = 0; /* we are not chaining */ + curr_sg->Ext = (i < scsi_sg_count(cmd) - 1) ? 0 : HPSA_SG_LAST; curr_sg++; } @@ -2045,7 +3223,10 @@ static int hpsa_scatter_gather(struct ctlr_info *h, if (chained) { cp->Header.SGList = h->max_cmd_sg_entries; cp->Header.SGTotal = (u16) (use_sg + 1); - hpsa_map_sg_chain_block(h, cp); + if (hpsa_map_sg_chain_block(h, cp)) { + scsi_dma_unmap(cmd); + return -1; + } return 0; } @@ -2056,6 +3237,726 @@ sglist_finished: return 0; } +#define IO_ACCEL_INELIGIBLE (1) +static int fixup_ioaccel_cdb(u8 *cdb, int *cdb_len) +{ + int is_write = 0; + u32 block; + u32 block_cnt; + + /* Perform some CDB fixups if needed using 10 byte reads/writes only */ + switch (cdb[0]) { + case WRITE_6: + case WRITE_12: + is_write = 1; + case READ_6: + case READ_12: + if (*cdb_len == 6) { + block = (((u32) cdb[2]) << 8) | cdb[3]; + block_cnt = cdb[4]; + } else { + BUG_ON(*cdb_len != 12); + block = (((u32) cdb[2]) << 24) | + (((u32) cdb[3]) << 16) | + (((u32) cdb[4]) << 8) | + cdb[5]; + block_cnt = + (((u32) cdb[6]) << 24) | + (((u32) cdb[7]) << 16) | + (((u32) cdb[8]) << 8) | + cdb[9]; + } + if (block_cnt > 0xffff) + return IO_ACCEL_INELIGIBLE; + + cdb[0] = is_write ? WRITE_10 : READ_10; + cdb[1] = 0; + cdb[2] = (u8) (block >> 24); + cdb[3] = (u8) (block >> 16); + cdb[4] = (u8) (block >> 8); + cdb[5] = (u8) (block); + cdb[6] = 0; + cdb[7] = (u8) (block_cnt >> 8); + cdb[8] = (u8) (block_cnt); + cdb[9] = 0; + *cdb_len = 10; + break; + } + return 0; +} + +static int hpsa_scsi_ioaccel1_queue_command(struct ctlr_info *h, + struct CommandList *c, u32 ioaccel_handle, u8 *cdb, int cdb_len, + u8 *scsi3addr) +{ + struct scsi_cmnd *cmd = c->scsi_cmd; + struct io_accel1_cmd *cp = &h->ioaccel_cmd_pool[c->cmdindex]; + unsigned int len; + unsigned int total_len = 0; + struct scatterlist *sg; + u64 addr64; + int use_sg, i; + struct SGDescriptor *curr_sg; + u32 control = IOACCEL1_CONTROL_SIMPLEQUEUE; + + /* TODO: implement chaining support */ + if (scsi_sg_count(cmd) > h->ioaccel_maxsg) + return IO_ACCEL_INELIGIBLE; + + BUG_ON(cmd->cmd_len > IOACCEL1_IOFLAGS_CDBLEN_MAX); + + if (fixup_ioaccel_cdb(cdb, &cdb_len)) + return IO_ACCEL_INELIGIBLE; + + c->cmd_type = CMD_IOACCEL1; + + /* Adjust the DMA address to point to the accelerated command buffer */ + c->busaddr = (u32) h->ioaccel_cmd_pool_dhandle + + (c->cmdindex * sizeof(*cp)); + BUG_ON(c->busaddr & 0x0000007F); + + use_sg = scsi_dma_map(cmd); + if (use_sg < 0) + return use_sg; + + if (use_sg) { + curr_sg = cp->SG; + scsi_for_each_sg(cmd, sg, use_sg, i) { + addr64 = (u64) sg_dma_address(sg); + len = sg_dma_len(sg); + total_len += len; + curr_sg->Addr.lower = (u32) (addr64 & 0x0FFFFFFFFULL); + curr_sg->Addr.upper = + (u32) ((addr64 >> 32) & 0x0FFFFFFFFULL); + curr_sg->Len = len; + + if (i == (scsi_sg_count(cmd) - 1)) + curr_sg->Ext = HPSA_SG_LAST; + else + curr_sg->Ext = 0; /* we are not chaining */ + curr_sg++; + } + + switch (cmd->sc_data_direction) { + case DMA_TO_DEVICE: + control |= IOACCEL1_CONTROL_DATA_OUT; + break; + case DMA_FROM_DEVICE: + control |= IOACCEL1_CONTROL_DATA_IN; + break; + case DMA_NONE: + control |= IOACCEL1_CONTROL_NODATAXFER; + break; + default: + dev_err(&h->pdev->dev, "unknown data direction: %d\n", + cmd->sc_data_direction); + BUG(); + break; + } + } else { + control |= IOACCEL1_CONTROL_NODATAXFER; + } + + c->Header.SGList = use_sg; + /* Fill out the command structure to submit */ + cp->dev_handle = ioaccel_handle & 0xFFFF; + cp->transfer_len = total_len; + cp->io_flags = IOACCEL1_IOFLAGS_IO_REQ | + (cdb_len & IOACCEL1_IOFLAGS_CDBLEN_MASK); + cp->control = control; + memcpy(cp->CDB, cdb, cdb_len); + memcpy(cp->CISS_LUN, scsi3addr, 8); + /* Tag was already set at init time. */ + enqueue_cmd_and_start_io(h, c); + return 0; +} + +/* + * Queue a command directly to a device behind the controller using the + * I/O accelerator path. + */ +static int hpsa_scsi_ioaccel_direct_map(struct ctlr_info *h, + struct CommandList *c) +{ + struct scsi_cmnd *cmd = c->scsi_cmd; + struct hpsa_scsi_dev_t *dev = cmd->device->hostdata; + + return hpsa_scsi_ioaccel_queue_command(h, c, dev->ioaccel_handle, + cmd->cmnd, cmd->cmd_len, dev->scsi3addr); +} + +/* + * Set encryption parameters for the ioaccel2 request + */ +static void set_encrypt_ioaccel2(struct ctlr_info *h, + struct CommandList *c, struct io_accel2_cmd *cp) +{ + struct scsi_cmnd *cmd = c->scsi_cmd; + struct hpsa_scsi_dev_t *dev = cmd->device->hostdata; + struct raid_map_data *map = &dev->raid_map; + u64 first_block; + + BUG_ON(!(dev->offload_config && dev->offload_enabled)); + + /* Are we doing encryption on this device */ + if (!(map->flags & RAID_MAP_FLAG_ENCRYPT_ON)) + return; + /* Set the data encryption key index. */ + cp->dekindex = map->dekindex; + + /* Set the encryption enable flag, encoded into direction field. */ + cp->direction |= IOACCEL2_DIRECTION_ENCRYPT_MASK; + + /* Set encryption tweak values based on logical block address + * If block size is 512, tweak value is LBA. + * For other block sizes, tweak is (LBA * block size)/ 512) + */ + switch (cmd->cmnd[0]) { + /* Required? 6-byte cdbs eliminated by fixup_ioaccel_cdb */ + case WRITE_6: + case READ_6: + if (map->volume_blk_size == 512) { + cp->tweak_lower = + (((u32) cmd->cmnd[2]) << 8) | + cmd->cmnd[3]; + cp->tweak_upper = 0; + } else { + first_block = + (((u64) cmd->cmnd[2]) << 8) | + cmd->cmnd[3]; + first_block = (first_block * map->volume_blk_size)/512; + cp->tweak_lower = (u32)first_block; + cp->tweak_upper = (u32)(first_block >> 32); + } + break; + case WRITE_10: + case READ_10: + if (map->volume_blk_size == 512) { + cp->tweak_lower = + (((u32) cmd->cmnd[2]) << 24) | + (((u32) cmd->cmnd[3]) << 16) | + (((u32) cmd->cmnd[4]) << 8) | + cmd->cmnd[5]; + cp->tweak_upper = 0; + } else { + first_block = + (((u64) cmd->cmnd[2]) << 24) | + (((u64) cmd->cmnd[3]) << 16) | + (((u64) cmd->cmnd[4]) << 8) | + cmd->cmnd[5]; + first_block = (first_block * map->volume_blk_size)/512; + cp->tweak_lower = (u32)first_block; + cp->tweak_upper = (u32)(first_block >> 32); + } + break; + /* Required? 12-byte cdbs eliminated by fixup_ioaccel_cdb */ + case WRITE_12: + case READ_12: + if (map->volume_blk_size == 512) { + cp->tweak_lower = + (((u32) cmd->cmnd[2]) << 24) | + (((u32) cmd->cmnd[3]) << 16) | + (((u32) cmd->cmnd[4]) << 8) | + cmd->cmnd[5]; + cp->tweak_upper = 0; + } else { + first_block = + (((u64) cmd->cmnd[2]) << 24) | + (((u64) cmd->cmnd[3]) << 16) | + (((u64) cmd->cmnd[4]) << 8) | + cmd->cmnd[5]; + first_block = (first_block * map->volume_blk_size)/512; + cp->tweak_lower = (u32)first_block; + cp->tweak_upper = (u32)(first_block >> 32); + } + break; + case WRITE_16: + case READ_16: + if (map->volume_blk_size == 512) { + cp->tweak_lower = + (((u32) cmd->cmnd[6]) << 24) | + (((u32) cmd->cmnd[7]) << 16) | + (((u32) cmd->cmnd[8]) << 8) | + cmd->cmnd[9]; + cp->tweak_upper = + (((u32) cmd->cmnd[2]) << 24) | + (((u32) cmd->cmnd[3]) << 16) | + (((u32) cmd->cmnd[4]) << 8) | + cmd->cmnd[5]; + } else { + first_block = + (((u64) cmd->cmnd[2]) << 56) | + (((u64) cmd->cmnd[3]) << 48) | + (((u64) cmd->cmnd[4]) << 40) | + (((u64) cmd->cmnd[5]) << 32) | + (((u64) cmd->cmnd[6]) << 24) | + (((u64) cmd->cmnd[7]) << 16) | + (((u64) cmd->cmnd[8]) << 8) | + cmd->cmnd[9]; + first_block = (first_block * map->volume_blk_size)/512; + cp->tweak_lower = (u32)first_block; + cp->tweak_upper = (u32)(first_block >> 32); + } + break; + default: + dev_err(&h->pdev->dev, + "ERROR: %s: IOACCEL request CDB size not supported for encryption\n", + __func__); + BUG(); + break; + } +} + +static int hpsa_scsi_ioaccel2_queue_command(struct ctlr_info *h, + struct CommandList *c, u32 ioaccel_handle, u8 *cdb, int cdb_len, + u8 *scsi3addr) +{ + struct scsi_cmnd *cmd = c->scsi_cmd; + struct io_accel2_cmd *cp = &h->ioaccel2_cmd_pool[c->cmdindex]; + struct ioaccel2_sg_element *curr_sg; + int use_sg, i; + struct scatterlist *sg; + u64 addr64; + u32 len; + u32 total_len = 0; + + if (scsi_sg_count(cmd) > h->ioaccel_maxsg) + return IO_ACCEL_INELIGIBLE; + + if (fixup_ioaccel_cdb(cdb, &cdb_len)) + return IO_ACCEL_INELIGIBLE; + c->cmd_type = CMD_IOACCEL2; + /* Adjust the DMA address to point to the accelerated command buffer */ + c->busaddr = (u32) h->ioaccel2_cmd_pool_dhandle + + (c->cmdindex * sizeof(*cp)); + BUG_ON(c->busaddr & 0x0000007F); + + memset(cp, 0, sizeof(*cp)); + cp->IU_type = IOACCEL2_IU_TYPE; + + use_sg = scsi_dma_map(cmd); + if (use_sg < 0) + return use_sg; + + if (use_sg) { + BUG_ON(use_sg > IOACCEL2_MAXSGENTRIES); + curr_sg = cp->sg; + scsi_for_each_sg(cmd, sg, use_sg, i) { + addr64 = (u64) sg_dma_address(sg); + len = sg_dma_len(sg); + total_len += len; + curr_sg->address = cpu_to_le64(addr64); + curr_sg->length = cpu_to_le32(len); + curr_sg->reserved[0] = 0; + curr_sg->reserved[1] = 0; + curr_sg->reserved[2] = 0; + curr_sg->chain_indicator = 0; + curr_sg++; + } + + switch (cmd->sc_data_direction) { + case DMA_TO_DEVICE: + cp->direction &= ~IOACCEL2_DIRECTION_MASK; + cp->direction |= IOACCEL2_DIR_DATA_OUT; + break; + case DMA_FROM_DEVICE: + cp->direction &= ~IOACCEL2_DIRECTION_MASK; + cp->direction |= IOACCEL2_DIR_DATA_IN; + break; + case DMA_NONE: + cp->direction &= ~IOACCEL2_DIRECTION_MASK; + cp->direction |= IOACCEL2_DIR_NO_DATA; + break; + default: + dev_err(&h->pdev->dev, "unknown data direction: %d\n", + cmd->sc_data_direction); + BUG(); + break; + } + } else { + cp->direction &= ~IOACCEL2_DIRECTION_MASK; + cp->direction |= IOACCEL2_DIR_NO_DATA; + } + + /* Set encryption parameters, if necessary */ + set_encrypt_ioaccel2(h, c, cp); + + cp->scsi_nexus = ioaccel_handle; + cp->Tag = (c->cmdindex << DIRECT_LOOKUP_SHIFT) | + DIRECT_LOOKUP_BIT; + memcpy(cp->cdb, cdb, sizeof(cp->cdb)); + + /* fill in sg elements */ + cp->sg_count = (u8) use_sg; + + cp->data_len = cpu_to_le32(total_len); + cp->err_ptr = cpu_to_le64(c->busaddr + + offsetof(struct io_accel2_cmd, error_data)); + cp->err_len = cpu_to_le32((u32) sizeof(cp->error_data)); + + enqueue_cmd_and_start_io(h, c); + return 0; +} + +/* + * Queue a command to the correct I/O accelerator path. + */ +static int hpsa_scsi_ioaccel_queue_command(struct ctlr_info *h, + struct CommandList *c, u32 ioaccel_handle, u8 *cdb, int cdb_len, + u8 *scsi3addr) +{ + if (h->transMethod & CFGTBL_Trans_io_accel1) + return hpsa_scsi_ioaccel1_queue_command(h, c, ioaccel_handle, + cdb, cdb_len, scsi3addr); + else + return hpsa_scsi_ioaccel2_queue_command(h, c, ioaccel_handle, + cdb, cdb_len, scsi3addr); +} + +static void raid_map_helper(struct raid_map_data *map, + int offload_to_mirror, u32 *map_index, u32 *current_group) +{ + if (offload_to_mirror == 0) { + /* use physical disk in the first mirrored group. */ + *map_index %= map->data_disks_per_row; + return; + } + do { + /* determine mirror group that *map_index indicates */ + *current_group = *map_index / map->data_disks_per_row; + if (offload_to_mirror == *current_group) + continue; + if (*current_group < (map->layout_map_count - 1)) { + /* select map index from next group */ + *map_index += map->data_disks_per_row; + (*current_group)++; + } else { + /* select map index from first group */ + *map_index %= map->data_disks_per_row; + *current_group = 0; + } + } while (offload_to_mirror != *current_group); +} + +/* + * Attempt to perform offload RAID mapping for a logical volume I/O. + */ +static int hpsa_scsi_ioaccel_raid_map(struct ctlr_info *h, + struct CommandList *c) +{ + struct scsi_cmnd *cmd = c->scsi_cmd; + struct hpsa_scsi_dev_t *dev = cmd->device->hostdata; + struct raid_map_data *map = &dev->raid_map; + struct raid_map_disk_data *dd = &map->data[0]; + int is_write = 0; + u32 map_index; + u64 first_block, last_block; + u32 block_cnt; + u32 blocks_per_row; + u64 first_row, last_row; + u32 first_row_offset, last_row_offset; + u32 first_column, last_column; + u64 r0_first_row, r0_last_row; + u32 r5or6_blocks_per_row; + u64 r5or6_first_row, r5or6_last_row; + u32 r5or6_first_row_offset, r5or6_last_row_offset; + u32 r5or6_first_column, r5or6_last_column; + u32 total_disks_per_row; + u32 stripesize; + u32 first_group, last_group, current_group; + u32 map_row; + u32 disk_handle; + u64 disk_block; + u32 disk_block_cnt; + u8 cdb[16]; + u8 cdb_len; +#if BITS_PER_LONG == 32 + u64 tmpdiv; +#endif + int offload_to_mirror; + + BUG_ON(!(dev->offload_config && dev->offload_enabled)); + + /* check for valid opcode, get LBA and block count */ + switch (cmd->cmnd[0]) { + case WRITE_6: + is_write = 1; + case READ_6: + first_block = + (((u64) cmd->cmnd[2]) << 8) | + cmd->cmnd[3]; + block_cnt = cmd->cmnd[4]; + break; + case WRITE_10: + is_write = 1; + case READ_10: + first_block = + (((u64) cmd->cmnd[2]) << 24) | + (((u64) cmd->cmnd[3]) << 16) | + (((u64) cmd->cmnd[4]) << 8) | + cmd->cmnd[5]; + block_cnt = + (((u32) cmd->cmnd[7]) << 8) | + cmd->cmnd[8]; + break; + case WRITE_12: + is_write = 1; + case READ_12: + first_block = + (((u64) cmd->cmnd[2]) << 24) | + (((u64) cmd->cmnd[3]) << 16) | + (((u64) cmd->cmnd[4]) << 8) | + cmd->cmnd[5]; + block_cnt = + (((u32) cmd->cmnd[6]) << 24) | + (((u32) cmd->cmnd[7]) << 16) | + (((u32) cmd->cmnd[8]) << 8) | + cmd->cmnd[9]; + break; + case WRITE_16: + is_write = 1; + case READ_16: + first_block = + (((u64) cmd->cmnd[2]) << 56) | + (((u64) cmd->cmnd[3]) << 48) | + (((u64) cmd->cmnd[4]) << 40) | + (((u64) cmd->cmnd[5]) << 32) | + (((u64) cmd->cmnd[6]) << 24) | + (((u64) cmd->cmnd[7]) << 16) | + (((u64) cmd->cmnd[8]) << 8) | + cmd->cmnd[9]; + block_cnt = + (((u32) cmd->cmnd[10]) << 24) | + (((u32) cmd->cmnd[11]) << 16) | + (((u32) cmd->cmnd[12]) << 8) | + cmd->cmnd[13]; + break; + default: + return IO_ACCEL_INELIGIBLE; /* process via normal I/O path */ + } + BUG_ON(block_cnt == 0); + last_block = first_block + block_cnt - 1; + + /* check for write to non-RAID-0 */ + if (is_write && dev->raid_level != 0) + return IO_ACCEL_INELIGIBLE; + + /* check for invalid block or wraparound */ + if (last_block >= map->volume_blk_cnt || last_block < first_block) + return IO_ACCEL_INELIGIBLE; + + /* calculate stripe information for the request */ + blocks_per_row = map->data_disks_per_row * map->strip_size; +#if BITS_PER_LONG == 32 + tmpdiv = first_block; + (void) do_div(tmpdiv, blocks_per_row); + first_row = tmpdiv; + tmpdiv = last_block; + (void) do_div(tmpdiv, blocks_per_row); + last_row = tmpdiv; + first_row_offset = (u32) (first_block - (first_row * blocks_per_row)); + last_row_offset = (u32) (last_block - (last_row * blocks_per_row)); + tmpdiv = first_row_offset; + (void) do_div(tmpdiv, map->strip_size); + first_column = tmpdiv; + tmpdiv = last_row_offset; + (void) do_div(tmpdiv, map->strip_size); + last_column = tmpdiv; +#else + first_row = first_block / blocks_per_row; + last_row = last_block / blocks_per_row; + first_row_offset = (u32) (first_block - (first_row * blocks_per_row)); + last_row_offset = (u32) (last_block - (last_row * blocks_per_row)); + first_column = first_row_offset / map->strip_size; + last_column = last_row_offset / map->strip_size; +#endif + + /* if this isn't a single row/column then give to the controller */ + if ((first_row != last_row) || (first_column != last_column)) + return IO_ACCEL_INELIGIBLE; + + /* proceeding with driver mapping */ + total_disks_per_row = map->data_disks_per_row + + map->metadata_disks_per_row; + map_row = ((u32)(first_row >> map->parity_rotation_shift)) % + map->row_cnt; + map_index = (map_row * total_disks_per_row) + first_column; + + switch (dev->raid_level) { + case HPSA_RAID_0: + break; /* nothing special to do */ + case HPSA_RAID_1: + /* Handles load balance across RAID 1 members. + * (2-drive R1 and R10 with even # of drives.) + * Appropriate for SSDs, not optimal for HDDs + */ + BUG_ON(map->layout_map_count != 2); + if (dev->offload_to_mirror) + map_index += map->data_disks_per_row; + dev->offload_to_mirror = !dev->offload_to_mirror; + break; + case HPSA_RAID_ADM: + /* Handles N-way mirrors (R1-ADM) + * and R10 with # of drives divisible by 3.) + */ + BUG_ON(map->layout_map_count != 3); + + offload_to_mirror = dev->offload_to_mirror; + raid_map_helper(map, offload_to_mirror, + &map_index, ¤t_group); + /* set mirror group to use next time */ + offload_to_mirror = + (offload_to_mirror >= map->layout_map_count - 1) + ? 0 : offload_to_mirror + 1; + /* FIXME: remove after debug/dev */ + BUG_ON(offload_to_mirror >= map->layout_map_count); + dev_warn(&h->pdev->dev, + "DEBUG: Using physical disk map index %d from mirror group %d\n", + map_index, offload_to_mirror); + dev->offload_to_mirror = offload_to_mirror; + /* Avoid direct use of dev->offload_to_mirror within this + * function since multiple threads might simultaneously + * increment it beyond the range of dev->layout_map_count -1. + */ + break; + case HPSA_RAID_5: + case HPSA_RAID_6: + if (map->layout_map_count <= 1) + break; + + /* Verify first and last block are in same RAID group */ + r5or6_blocks_per_row = + map->strip_size * map->data_disks_per_row; + BUG_ON(r5or6_blocks_per_row == 0); + stripesize = r5or6_blocks_per_row * map->layout_map_count; +#if BITS_PER_LONG == 32 + tmpdiv = first_block; + first_group = do_div(tmpdiv, stripesize); + tmpdiv = first_group; + (void) do_div(tmpdiv, r5or6_blocks_per_row); + first_group = tmpdiv; + tmpdiv = last_block; + last_group = do_div(tmpdiv, stripesize); + tmpdiv = last_group; + (void) do_div(tmpdiv, r5or6_blocks_per_row); + last_group = tmpdiv; +#else + first_group = (first_block % stripesize) / r5or6_blocks_per_row; + last_group = (last_block % stripesize) / r5or6_blocks_per_row; +#endif + if (first_group != last_group) + return IO_ACCEL_INELIGIBLE; + + /* Verify request is in a single row of RAID 5/6 */ +#if BITS_PER_LONG == 32 + tmpdiv = first_block; + (void) do_div(tmpdiv, stripesize); + first_row = r5or6_first_row = r0_first_row = tmpdiv; + tmpdiv = last_block; + (void) do_div(tmpdiv, stripesize); + r5or6_last_row = r0_last_row = tmpdiv; +#else + first_row = r5or6_first_row = r0_first_row = + first_block / stripesize; + r5or6_last_row = r0_last_row = last_block / stripesize; +#endif + if (r5or6_first_row != r5or6_last_row) + return IO_ACCEL_INELIGIBLE; + + + /* Verify request is in a single column */ +#if BITS_PER_LONG == 32 + tmpdiv = first_block; + first_row_offset = do_div(tmpdiv, stripesize); + tmpdiv = first_row_offset; + first_row_offset = (u32) do_div(tmpdiv, r5or6_blocks_per_row); + r5or6_first_row_offset = first_row_offset; + tmpdiv = last_block; + r5or6_last_row_offset = do_div(tmpdiv, stripesize); + tmpdiv = r5or6_last_row_offset; + r5or6_last_row_offset = do_div(tmpdiv, r5or6_blocks_per_row); + tmpdiv = r5or6_first_row_offset; + (void) do_div(tmpdiv, map->strip_size); + first_column = r5or6_first_column = tmpdiv; + tmpdiv = r5or6_last_row_offset; + (void) do_div(tmpdiv, map->strip_size); + r5or6_last_column = tmpdiv; +#else + first_row_offset = r5or6_first_row_offset = + (u32)((first_block % stripesize) % + r5or6_blocks_per_row); + + r5or6_last_row_offset = + (u32)((last_block % stripesize) % + r5or6_blocks_per_row); + + first_column = r5or6_first_column = + r5or6_first_row_offset / map->strip_size; + r5or6_last_column = + r5or6_last_row_offset / map->strip_size; +#endif + if (r5or6_first_column != r5or6_last_column) + return IO_ACCEL_INELIGIBLE; + + /* Request is eligible */ + map_row = ((u32)(first_row >> map->parity_rotation_shift)) % + map->row_cnt; + + map_index = (first_group * + (map->row_cnt * total_disks_per_row)) + + (map_row * total_disks_per_row) + first_column; + break; + default: + return IO_ACCEL_INELIGIBLE; + } + + disk_handle = dd[map_index].ioaccel_handle; + disk_block = map->disk_starting_blk + (first_row * map->strip_size) + + (first_row_offset - (first_column * map->strip_size)); + disk_block_cnt = block_cnt; + + /* handle differing logical/physical block sizes */ + if (map->phys_blk_shift) { + disk_block <<= map->phys_blk_shift; + disk_block_cnt <<= map->phys_blk_shift; + } + BUG_ON(disk_block_cnt > 0xffff); + + /* build the new CDB for the physical disk I/O */ + if (disk_block > 0xffffffff) { + cdb[0] = is_write ? WRITE_16 : READ_16; + cdb[1] = 0; + cdb[2] = (u8) (disk_block >> 56); + cdb[3] = (u8) (disk_block >> 48); + cdb[4] = (u8) (disk_block >> 40); + cdb[5] = (u8) (disk_block >> 32); + cdb[6] = (u8) (disk_block >> 24); + cdb[7] = (u8) (disk_block >> 16); + cdb[8] = (u8) (disk_block >> 8); + cdb[9] = (u8) (disk_block); + cdb[10] = (u8) (disk_block_cnt >> 24); + cdb[11] = (u8) (disk_block_cnt >> 16); + cdb[12] = (u8) (disk_block_cnt >> 8); + cdb[13] = (u8) (disk_block_cnt); + cdb[14] = 0; + cdb[15] = 0; + cdb_len = 16; + } else { + cdb[0] = is_write ? WRITE_10 : READ_10; + cdb[1] = 0; + cdb[2] = (u8) (disk_block >> 24); + cdb[3] = (u8) (disk_block >> 16); + cdb[4] = (u8) (disk_block >> 8); + cdb[5] = (u8) (disk_block); + cdb[6] = 0; + cdb[7] = (u8) (disk_block_cnt >> 8); + cdb[8] = (u8) (disk_block_cnt); + cdb[9] = 0; + cdb_len = 10; + } + return hpsa_scsi_ioaccel_queue_command(h, c, disk_handle, cdb, cdb_len, + dev->scsi3addr); +} static int hpsa_scsi_queue_command_lck(struct scsi_cmnd *cmd, void (*done)(struct scsi_cmnd *)) @@ -2064,7 +3965,7 @@ static int hpsa_scsi_queue_command_lck(struct scsi_cmnd *cmd, struct hpsa_scsi_dev_t *dev; unsigned char scsi3addr[8]; struct CommandList *c; - unsigned long flags; + int rc = 0; /* Get the ptr to our adapter structure out of cmd->host. */ h = sdev_to_hba(cmd->device); @@ -2076,16 +3977,12 @@ static int hpsa_scsi_queue_command_lck(struct scsi_cmnd *cmd, } memcpy(scsi3addr, dev->scsi3addr, sizeof(scsi3addr)); - spin_lock_irqsave(&h->lock, flags); - if (unlikely(h->lockup_detected)) { - spin_unlock_irqrestore(&h->lock, flags); + if (unlikely(lockup_detected(h))) { cmd->result = DID_ERROR << 16; done(cmd); return 0; } - /* Need a lock as this is being allocated from the pool */ c = cmd_alloc(h); - spin_unlock_irqrestore(&h->lock, flags); if (c == NULL) { /* trouble... */ dev_err(&h->pdev->dev, "cmd_alloc returned NULL!\n"); return SCSI_MLQUEUE_HOST_BUSY; @@ -2100,6 +3997,32 @@ static int hpsa_scsi_queue_command_lck(struct scsi_cmnd *cmd, c->cmd_type = CMD_SCSI; c->scsi_cmd = cmd; + + /* Call alternate submit routine for I/O accelerated commands. + * Retries always go down the normal I/O path. + */ + if (likely(cmd->retries == 0 && + cmd->request->cmd_type == REQ_TYPE_FS && + h->acciopath_status)) { + if (dev->offload_enabled) { + rc = hpsa_scsi_ioaccel_raid_map(h, c); + if (rc == 0) + return 0; /* Sent on ioaccel path */ + if (rc < 0) { /* scsi_dma_map failed. */ + cmd_free(h, c); + return SCSI_MLQUEUE_HOST_BUSY; + } + } else if (dev->ioaccel_handle) { + rc = hpsa_scsi_ioaccel_direct_map(h, c); + if (rc == 0) + return 0; /* Sent on direct map path */ + if (rc < 0) { /* scsi_dma_map failed. */ + cmd_free(h, c); + return SCSI_MLQUEUE_HOST_BUSY; + } + } + } + c->Header.ReplyQueue = 0; /* unused in simple mode */ memcpy(&c->Header.LUN.LunAddrBytes[0], &scsi3addr[0], 8); c->Header.Tag.lower = (c->cmdindex << DIRECT_LOOKUP_SHIFT); @@ -2159,11 +4082,35 @@ static int hpsa_scsi_queue_command_lck(struct scsi_cmnd *cmd, static DEF_SCSI_QCMD(hpsa_scsi_queue_command) +static int do_not_scan_if_controller_locked_up(struct ctlr_info *h) +{ + unsigned long flags; + + /* + * Don't let rescans be initiated on a controller known + * to be locked up. If the controller locks up *during* + * a rescan, that thread is probably hosed, but at least + * we can prevent new rescan threads from piling up on a + * locked up controller. + */ + if (unlikely(lockup_detected(h))) { + spin_lock_irqsave(&h->scan_lock, flags); + h->scan_finished = 1; + wake_up_all(&h->scan_wait_queue); + spin_unlock_irqrestore(&h->scan_lock, flags); + return 1; + } + return 0; +} + static void hpsa_scan_start(struct Scsi_Host *sh) { struct ctlr_info *h = shost_to_hba(sh); unsigned long flags; + if (do_not_scan_if_controller_locked_up(h)) + return; + /* wait until any scan already in progress is finished. */ while (1) { spin_lock_irqsave(&h->scan_lock, flags); @@ -2180,6 +4127,9 @@ static void hpsa_scan_start(struct Scsi_Host *sh) h->scan_finished = 0; /* mark scan as in progress */ spin_unlock_irqrestore(&h->scan_lock, flags); + if (do_not_scan_if_controller_locked_up(h)) + return; + hpsa_update_scsi_devices(h, h->scsi_host->host_no); spin_lock_irqsave(&h->scan_lock, flags); @@ -2228,19 +4178,51 @@ static void hpsa_unregister_scsi(struct ctlr_info *h) static int hpsa_register_scsi(struct ctlr_info *h) { - int rc; + struct Scsi_Host *sh; + int error; - rc = hpsa_scsi_detect(h); - if (rc != 0) - dev_err(&h->pdev->dev, "hpsa_register_scsi: failed" - " hpsa_scsi_detect(), rc is %d\n", rc); - return rc; + sh = scsi_host_alloc(&hpsa_driver_template, sizeof(h)); + if (sh == NULL) + goto fail; + + sh->io_port = 0; + sh->n_io_port = 0; + sh->this_id = -1; + sh->max_channel = 3; + sh->max_cmd_len = MAX_COMMAND_SIZE; + sh->max_lun = HPSA_MAX_LUN; + sh->max_id = HPSA_MAX_LUN; + sh->can_queue = h->nr_cmds; + if (h->hba_mode_enabled) + sh->cmd_per_lun = 7; + else + sh->cmd_per_lun = h->nr_cmds; + sh->sg_tablesize = h->maxsgentries; + h->scsi_host = sh; + sh->hostdata[0] = (unsigned long) h; + sh->irq = h->intr[h->intr_mode]; + sh->unique_id = sh->irq; + error = scsi_add_host(sh, &h->pdev->dev); + if (error) + goto fail_host_put; + scsi_scan_host(sh); + return 0; + + fail_host_put: + dev_err(&h->pdev->dev, "%s: scsi_add_host" + " failed for controller %d\n", __func__, h->ctlr); + scsi_host_put(sh); + return error; + fail: + dev_err(&h->pdev->dev, "%s: scsi_host_alloc" + " failed for controller %d\n", __func__, h->ctlr); + return -ENOMEM; } static int wait_for_device_to_become_ready(struct ctlr_info *h, unsigned char lunaddr[]) { - int rc = 0; + int rc; int count = 0; int waittime = 1; /* seconds */ struct CommandList *c; @@ -2260,13 +4242,15 @@ static int wait_for_device_to_become_ready(struct ctlr_info *h, */ msleep(1000 * waittime); count++; + rc = 0; /* Device ready. */ /* Increase wait time with each try, up to a point. */ if (waittime < HPSA_MAX_WAIT_INTERVAL_SECS) waittime = waittime * 2; - /* Send the Test Unit Ready */ - fill_cmd(c, TEST_UNIT_READY, h, NULL, 0, 0, lunaddr, TYPE_CMD); + /* Send the Test Unit Ready, fill_cmd can't fail, no mapping */ + (void) fill_cmd(c, TEST_UNIT_READY, h, + NULL, 0, 0, lunaddr, TYPE_CMD); hpsa_scsi_do_simple_cmd_core(h, c); /* no unmap needed here because no data xfer. */ @@ -2315,7 +4299,7 @@ static int hpsa_eh_device_reset_handler(struct scsi_cmnd *scsicmd) dev_warn(&h->pdev->dev, "resetting device %d:%d:%d:%d\n", h->scsi_host->host_no, dev->bus, dev->target, dev->lun); /* send a reset to the SCSI LUN which the command was sent to */ - rc = hpsa_send_reset(h, dev->scsi3addr); + rc = hpsa_send_reset(h, dev->scsi3addr, HPSA_RESET_TYPE_LUN); if (rc == 0 && wait_for_device_to_become_ready(h, dev->scsi3addr) == 0) return SUCCESS; @@ -2323,6 +4307,372 @@ static int hpsa_eh_device_reset_handler(struct scsi_cmnd *scsicmd) return FAILED; } +static void swizzle_abort_tag(u8 *tag) +{ + u8 original_tag[8]; + + memcpy(original_tag, tag, 8); + tag[0] = original_tag[3]; + tag[1] = original_tag[2]; + tag[2] = original_tag[1]; + tag[3] = original_tag[0]; + tag[4] = original_tag[7]; + tag[5] = original_tag[6]; + tag[6] = original_tag[5]; + tag[7] = original_tag[4]; +} + +static void hpsa_get_tag(struct ctlr_info *h, + struct CommandList *c, u32 *taglower, u32 *tagupper) +{ + if (c->cmd_type == CMD_IOACCEL1) { + struct io_accel1_cmd *cm1 = (struct io_accel1_cmd *) + &h->ioaccel_cmd_pool[c->cmdindex]; + *tagupper = cm1->Tag.upper; + *taglower = cm1->Tag.lower; + return; + } + if (c->cmd_type == CMD_IOACCEL2) { + struct io_accel2_cmd *cm2 = (struct io_accel2_cmd *) + &h->ioaccel2_cmd_pool[c->cmdindex]; + /* upper tag not used in ioaccel2 mode */ + memset(tagupper, 0, sizeof(*tagupper)); + *taglower = cm2->Tag; + return; + } + *tagupper = c->Header.Tag.upper; + *taglower = c->Header.Tag.lower; +} + + +static int hpsa_send_abort(struct ctlr_info *h, unsigned char *scsi3addr, + struct CommandList *abort, int swizzle) +{ + int rc = IO_OK; + struct CommandList *c; + struct ErrorInfo *ei; + u32 tagupper, taglower; + + c = cmd_special_alloc(h); + if (c == NULL) { /* trouble... */ + dev_warn(&h->pdev->dev, "cmd_special_alloc returned NULL!\n"); + return -ENOMEM; + } + + /* fill_cmd can't fail here, no buffer to map */ + (void) fill_cmd(c, HPSA_ABORT_MSG, h, abort, + 0, 0, scsi3addr, TYPE_MSG); + if (swizzle) + swizzle_abort_tag(&c->Request.CDB[4]); + hpsa_scsi_do_simple_cmd_core(h, c); + hpsa_get_tag(h, abort, &taglower, &tagupper); + dev_dbg(&h->pdev->dev, "%s: Tag:0x%08x:%08x: do_simple_cmd_core completed.\n", + __func__, tagupper, taglower); + /* no unmap needed here because no data xfer. */ + + ei = c->err_info; + switch (ei->CommandStatus) { + case CMD_SUCCESS: + break; + case CMD_UNABORTABLE: /* Very common, don't make noise. */ + rc = -1; + break; + default: + dev_dbg(&h->pdev->dev, "%s: Tag:0x%08x:%08x: interpreting error.\n", + __func__, tagupper, taglower); + hpsa_scsi_interpret_error(h, c); + rc = -1; + break; + } + cmd_special_free(h, c); + dev_dbg(&h->pdev->dev, "%s: Tag:0x%08x:%08x: Finished.\n", + __func__, tagupper, taglower); + return rc; +} + +/* + * hpsa_find_cmd_in_queue + * + * Used to determine whether a command (find) is still present + * in queue_head. Optionally excludes the last element of queue_head. + * + * This is used to avoid unnecessary aborts. Commands in h->reqQ have + * not yet been submitted, and so can be aborted by the driver without + * sending an abort to the hardware. + * + * Returns pointer to command if found in queue, NULL otherwise. + */ +static struct CommandList *hpsa_find_cmd_in_queue(struct ctlr_info *h, + struct scsi_cmnd *find, struct list_head *queue_head) +{ + unsigned long flags; + struct CommandList *c = NULL; /* ptr into cmpQ */ + + if (!find) + return 0; + spin_lock_irqsave(&h->lock, flags); + list_for_each_entry(c, queue_head, list) { + if (c->scsi_cmd == NULL) /* e.g.: passthru ioctl */ + continue; + if (c->scsi_cmd == find) { + spin_unlock_irqrestore(&h->lock, flags); + return c; + } + } + spin_unlock_irqrestore(&h->lock, flags); + return NULL; +} + +static struct CommandList *hpsa_find_cmd_in_queue_by_tag(struct ctlr_info *h, + u8 *tag, struct list_head *queue_head) +{ + unsigned long flags; + struct CommandList *c; + + spin_lock_irqsave(&h->lock, flags); + list_for_each_entry(c, queue_head, list) { + if (memcmp(&c->Header.Tag, tag, 8) != 0) + continue; + spin_unlock_irqrestore(&h->lock, flags); + return c; + } + spin_unlock_irqrestore(&h->lock, flags); + return NULL; +} + +/* ioaccel2 path firmware cannot handle abort task requests. + * Change abort requests to physical target reset, and send to the + * address of the physical disk used for the ioaccel 2 command. + * Return 0 on success (IO_OK) + * -1 on failure + */ + +static int hpsa_send_reset_as_abort_ioaccel2(struct ctlr_info *h, + unsigned char *scsi3addr, struct CommandList *abort) +{ + int rc = IO_OK; + struct scsi_cmnd *scmd; /* scsi command within request being aborted */ + struct hpsa_scsi_dev_t *dev; /* device to which scsi cmd was sent */ + unsigned char phys_scsi3addr[8]; /* addr of phys disk with volume */ + unsigned char *psa = &phys_scsi3addr[0]; + + /* Get a pointer to the hpsa logical device. */ + scmd = (struct scsi_cmnd *) abort->scsi_cmd; + dev = (struct hpsa_scsi_dev_t *)(scmd->device->hostdata); + if (dev == NULL) { + dev_warn(&h->pdev->dev, + "Cannot abort: no device pointer for command.\n"); + return -1; /* not abortable */ + } + + if (h->raid_offload_debug > 0) + dev_info(&h->pdev->dev, + "Reset as abort: Abort requested on C%d:B%d:T%d:L%d scsi3addr 0x%02x%02x%02x%02x%02x%02x%02x%02x\n", + h->scsi_host->host_no, dev->bus, dev->target, dev->lun, + scsi3addr[0], scsi3addr[1], scsi3addr[2], scsi3addr[3], + scsi3addr[4], scsi3addr[5], scsi3addr[6], scsi3addr[7]); + + if (!dev->offload_enabled) { + dev_warn(&h->pdev->dev, + "Can't abort: device is not operating in HP SSD Smart Path mode.\n"); + return -1; /* not abortable */ + } + + /* Incoming scsi3addr is logical addr. We need physical disk addr. */ + if (!hpsa_get_pdisk_of_ioaccel2(h, abort, psa)) { + dev_warn(&h->pdev->dev, "Can't abort: Failed lookup of physical address.\n"); + return -1; /* not abortable */ + } + + /* send the reset */ + if (h->raid_offload_debug > 0) + dev_info(&h->pdev->dev, + "Reset as abort: Resetting physical device at scsi3addr 0x%02x%02x%02x%02x%02x%02x%02x%02x\n", + psa[0], psa[1], psa[2], psa[3], + psa[4], psa[5], psa[6], psa[7]); + rc = hpsa_send_reset(h, psa, HPSA_RESET_TYPE_TARGET); + if (rc != 0) { + dev_warn(&h->pdev->dev, + "Reset as abort: Failed on physical device at scsi3addr 0x%02x%02x%02x%02x%02x%02x%02x%02x\n", + psa[0], psa[1], psa[2], psa[3], + psa[4], psa[5], psa[6], psa[7]); + return rc; /* failed to reset */ + } + + /* wait for device to recover */ + if (wait_for_device_to_become_ready(h, psa) != 0) { + dev_warn(&h->pdev->dev, + "Reset as abort: Failed: Device never recovered from reset: 0x%02x%02x%02x%02x%02x%02x%02x%02x\n", + psa[0], psa[1], psa[2], psa[3], + psa[4], psa[5], psa[6], psa[7]); + return -1; /* failed to recover */ + } + + /* device recovered */ + dev_info(&h->pdev->dev, + "Reset as abort: Device recovered from reset: scsi3addr 0x%02x%02x%02x%02x%02x%02x%02x%02x\n", + psa[0], psa[1], psa[2], psa[3], + psa[4], psa[5], psa[6], psa[7]); + + return rc; /* success */ +} + +/* Some Smart Arrays need the abort tag swizzled, and some don't. It's hard to + * tell which kind we're dealing with, so we send the abort both ways. There + * shouldn't be any collisions between swizzled and unswizzled tags due to the + * way we construct our tags but we check anyway in case the assumptions which + * make this true someday become false. + */ +static int hpsa_send_abort_both_ways(struct ctlr_info *h, + unsigned char *scsi3addr, struct CommandList *abort) +{ + u8 swizzled_tag[8]; + struct CommandList *c; + int rc = 0, rc2 = 0; + + /* ioccelerator mode 2 commands should be aborted via the + * accelerated path, since RAID path is unaware of these commands, + * but underlying firmware can't handle abort TMF. + * Change abort to physical device reset. + */ + if (abort->cmd_type == CMD_IOACCEL2) + return hpsa_send_reset_as_abort_ioaccel2(h, scsi3addr, abort); + + /* we do not expect to find the swizzled tag in our queue, but + * check anyway just to be sure the assumptions which make this + * the case haven't become wrong. + */ + memcpy(swizzled_tag, &abort->Request.CDB[4], 8); + swizzle_abort_tag(swizzled_tag); + c = hpsa_find_cmd_in_queue_by_tag(h, swizzled_tag, &h->cmpQ); + if (c != NULL) { + dev_warn(&h->pdev->dev, "Unexpectedly found byte-swapped tag in completion queue.\n"); + return hpsa_send_abort(h, scsi3addr, abort, 0); + } + rc = hpsa_send_abort(h, scsi3addr, abort, 0); + + /* if the command is still in our queue, we can't conclude that it was + * aborted (it might have just completed normally) but in any case + * we don't need to try to abort it another way. + */ + c = hpsa_find_cmd_in_queue(h, abort->scsi_cmd, &h->cmpQ); + if (c) + rc2 = hpsa_send_abort(h, scsi3addr, abort, 1); + return rc && rc2; +} + +/* Send an abort for the specified command. + * If the device and controller support it, + * send a task abort request. + */ +static int hpsa_eh_abort_handler(struct scsi_cmnd *sc) +{ + + int i, rc; + struct ctlr_info *h; + struct hpsa_scsi_dev_t *dev; + struct CommandList *abort; /* pointer to command to be aborted */ + struct CommandList *found; + struct scsi_cmnd *as; /* ptr to scsi cmd inside aborted command. */ + char msg[256]; /* For debug messaging. */ + int ml = 0; + u32 tagupper, taglower; + + /* Find the controller of the command to be aborted */ + h = sdev_to_hba(sc->device); + if (WARN(h == NULL, + "ABORT REQUEST FAILED, Controller lookup failed.\n")) + return FAILED; + + /* Check that controller supports some kind of task abort */ + if (!(HPSATMF_PHYS_TASK_ABORT & h->TMFSupportFlags) && + !(HPSATMF_LOG_TASK_ABORT & h->TMFSupportFlags)) + return FAILED; + + memset(msg, 0, sizeof(msg)); + ml += sprintf(msg+ml, "ABORT REQUEST on C%d:B%d:T%d:L%d ", + h->scsi_host->host_no, sc->device->channel, + sc->device->id, sc->device->lun); + + /* Find the device of the command to be aborted */ + dev = sc->device->hostdata; + if (!dev) { + dev_err(&h->pdev->dev, "%s FAILED, Device lookup failed.\n", + msg); + return FAILED; + } + + /* Get SCSI command to be aborted */ + abort = (struct CommandList *) sc->host_scribble; + if (abort == NULL) { + dev_err(&h->pdev->dev, "%s FAILED, Command to abort is NULL.\n", + msg); + return FAILED; + } + hpsa_get_tag(h, abort, &taglower, &tagupper); + ml += sprintf(msg+ml, "Tag:0x%08x:%08x ", tagupper, taglower); + as = (struct scsi_cmnd *) abort->scsi_cmd; + if (as != NULL) + ml += sprintf(msg+ml, "Command:0x%x SN:0x%lx ", + as->cmnd[0], as->serial_number); + dev_dbg(&h->pdev->dev, "%s\n", msg); + dev_warn(&h->pdev->dev, "Abort request on C%d:B%d:T%d:L%d\n", + h->scsi_host->host_no, dev->bus, dev->target, dev->lun); + + /* Search reqQ to See if command is queued but not submitted, + * if so, complete the command with aborted status and remove + * it from the reqQ. + */ + found = hpsa_find_cmd_in_queue(h, sc, &h->reqQ); + if (found) { + found->err_info->CommandStatus = CMD_ABORTED; + finish_cmd(found); + dev_info(&h->pdev->dev, "%s Request SUCCEEDED (driver queue).\n", + msg); + return SUCCESS; + } + + /* not in reqQ, if also not in cmpQ, must have already completed */ + found = hpsa_find_cmd_in_queue(h, sc, &h->cmpQ); + if (!found) { + dev_dbg(&h->pdev->dev, "%s Request SUCCEEDED (not known to driver).\n", + msg); + return SUCCESS; + } + + /* + * Command is in flight, or possibly already completed + * by the firmware (but not to the scsi mid layer) but we can't + * distinguish which. Send the abort down. + */ + rc = hpsa_send_abort_both_ways(h, dev->scsi3addr, abort); + if (rc != 0) { + dev_dbg(&h->pdev->dev, "%s Request FAILED.\n", msg); + dev_warn(&h->pdev->dev, "FAILED abort on device C%d:B%d:T%d:L%d\n", + h->scsi_host->host_no, + dev->bus, dev->target, dev->lun); + return FAILED; + } + dev_info(&h->pdev->dev, "%s REQUEST SUCCEEDED.\n", msg); + + /* If the abort(s) above completed and actually aborted the + * command, then the command to be aborted should already be + * completed. If not, wait around a bit more to see if they + * manage to complete normally. + */ +#define ABORT_COMPLETE_WAIT_SECS 30 + for (i = 0; i < ABORT_COMPLETE_WAIT_SECS * 10; i++) { + found = hpsa_find_cmd_in_queue(h, sc, &h->cmpQ); + if (!found) + return SUCCESS; + msleep(100); + } + dev_warn(&h->pdev->dev, "%s FAILED. Aborted command has not completed after %d seconds.\n", + msg, ABORT_COMPLETE_WAIT_SECS); + return FAILED; +} + + /* * For operations that cannot sleep, a command block is allocated at init, * and managed by cmd_alloc() and cmd_free() using a simple bitmap to track @@ -2335,14 +4685,20 @@ static struct CommandList *cmd_alloc(struct ctlr_info *h) int i; union u64bit temp64; dma_addr_t cmd_dma_handle, err_dma_handle; + unsigned long flags; + spin_lock_irqsave(&h->lock, flags); do { i = find_first_zero_bit(h->cmd_pool_bits, h->nr_cmds); - if (i == h->nr_cmds) + if (i == h->nr_cmds) { + spin_unlock_irqrestore(&h->lock, flags); return NULL; + } } while (test_and_set_bit (i & (BITS_PER_LONG - 1), h->cmd_pool_bits + (i / BITS_PER_LONG)) != 0); + spin_unlock_irqrestore(&h->lock, flags); + c = h->cmd_pool + i; memset(c, 0, sizeof(*c)); cmd_dma_handle = h->cmd_pool_dhandle @@ -2351,7 +4707,6 @@ static struct CommandList *cmd_alloc(struct ctlr_info *h) memset(c->err_info, 0, sizeof(*c->err_info)); err_dma_handle = h->errinfo_pool_dhandle + i * sizeof(*c->err_info); - h->nr_allocs++; c->cmdindex = i; @@ -2381,6 +4736,7 @@ static struct CommandList *cmd_special_alloc(struct ctlr_info *h) return NULL; memset(c, 0, sizeof(*c)); + c->cmd_type = CMD_SCSI; c->cmdindex = -1; c->err_info = pci_alloc_consistent(h->pdev, sizeof(*c->err_info), @@ -2407,11 +4763,13 @@ static struct CommandList *cmd_special_alloc(struct ctlr_info *h) static void cmd_free(struct ctlr_info *h, struct CommandList *c) { int i; + unsigned long flags; i = c - h->cmd_pool; + spin_lock_irqsave(&h->lock, flags); clear_bit(i & (BITS_PER_LONG - 1), h->cmd_pool_bits + (i / BITS_PER_LONG)); - h->nr_frees++; + spin_unlock_irqrestore(&h->lock, flags); } static void cmd_special_free(struct ctlr_info *h, struct CommandList *c) @@ -2576,6 +4934,7 @@ static int hpsa_passthru_ioctl(struct ctlr_info *h, void __user *argp) struct CommandList *c; char *buff = NULL; union u64bit temp64; + int rc = 0; if (!argp) return -EINVAL; @@ -2591,12 +4950,12 @@ static int hpsa_passthru_ioctl(struct ctlr_info *h, void __user *argp) buff = kmalloc(iocommand.buf_size, GFP_KERNEL); if (buff == NULL) return -EFAULT; - if (iocommand.Request.Type.Direction == XFER_WRITE) { + if (iocommand.Request.Type.Direction & XFER_WRITE) { /* Copy the data into the buffer we created */ if (copy_from_user(buff, iocommand.buf, iocommand.buf_size)) { - kfree(buff); - return -EFAULT; + rc = -EFAULT; + goto out_kfree; } } else { memset(buff, 0, iocommand.buf_size); @@ -2604,8 +4963,8 @@ static int hpsa_passthru_ioctl(struct ctlr_info *h, void __user *argp) } c = cmd_special_alloc(h); if (c == NULL) { - kfree(buff); - return -ENOMEM; + rc = -ENOMEM; + goto out_kfree; } /* Fill in the command type */ c->cmd_type = CMD_IOCTL_PEND; @@ -2630,10 +4989,17 @@ static int hpsa_passthru_ioctl(struct ctlr_info *h, void __user *argp) if (iocommand.buf_size > 0) { temp64.val = pci_map_single(h->pdev, buff, iocommand.buf_size, PCI_DMA_BIDIRECTIONAL); + if (dma_mapping_error(&h->pdev->dev, temp64.val)) { + c->SG[0].Addr.lower = 0; + c->SG[0].Addr.upper = 0; + c->SG[0].Len = 0; + rc = -ENOMEM; + goto out; + } c->SG[0].Addr.lower = temp64.val32.lower; c->SG[0].Addr.upper = temp64.val32.upper; c->SG[0].Len = iocommand.buf_size; - c->SG[0].Ext = 0; /* we are not chaining*/ + c->SG[0].Ext = HPSA_SG_LAST; /* we are not chaining*/ } hpsa_scsi_do_simple_cmd_core_if_no_lockup(h, c); if (iocommand.buf_size > 0) @@ -2644,22 +5010,22 @@ static int hpsa_passthru_ioctl(struct ctlr_info *h, void __user *argp) memcpy(&iocommand.error_info, c->err_info, sizeof(iocommand.error_info)); if (copy_to_user(argp, &iocommand, sizeof(iocommand))) { - kfree(buff); - cmd_special_free(h, c); - return -EFAULT; + rc = -EFAULT; + goto out; } - if (iocommand.Request.Type.Direction == XFER_READ && + if ((iocommand.Request.Type.Direction & XFER_READ) && iocommand.buf_size > 0) { /* Copy the data out of the buffer we created */ if (copy_to_user(iocommand.buf, buff, iocommand.buf_size)) { - kfree(buff); - cmd_special_free(h, c); - return -EFAULT; + rc = -EFAULT; + goto out; } } - kfree(buff); +out: cmd_special_free(h, c); - return 0; +out_kfree: + kfree(buff); + return rc; } static int hpsa_big_passthru_ioctl(struct ctlr_info *h, void __user *argp) @@ -2700,16 +5066,16 @@ static int hpsa_big_passthru_ioctl(struct ctlr_info *h, void __user *argp) status = -EINVAL; goto cleanup1; } - if (ioc->buf_size > ioc->malloc_size * MAXSGENTRIES) { + if (ioc->buf_size > ioc->malloc_size * SG_ENTRIES_IN_CMD) { status = -EINVAL; goto cleanup1; } - buff = kzalloc(MAXSGENTRIES * sizeof(char *), GFP_KERNEL); + buff = kzalloc(SG_ENTRIES_IN_CMD * sizeof(char *), GFP_KERNEL); if (!buff) { status = -ENOMEM; goto cleanup1; } - buff_size = kmalloc(MAXSGENTRIES * sizeof(int), GFP_KERNEL); + buff_size = kmalloc(SG_ENTRIES_IN_CMD * sizeof(int), GFP_KERNEL); if (!buff_size) { status = -ENOMEM; goto cleanup1; @@ -2724,7 +5090,7 @@ static int hpsa_big_passthru_ioctl(struct ctlr_info *h, void __user *argp) status = -ENOMEM; goto cleanup1; } - if (ioc->Request.Type.Direction == XFER_WRITE) { + if (ioc->Request.Type.Direction & XFER_WRITE) { if (copy_from_user(buff[sg_used], data_ptr, sz)) { status = -ENOMEM; goto cleanup1; @@ -2751,11 +5117,19 @@ static int hpsa_big_passthru_ioctl(struct ctlr_info *h, void __user *argp) for (i = 0; i < sg_used; i++) { temp64.val = pci_map_single(h->pdev, buff[i], buff_size[i], PCI_DMA_BIDIRECTIONAL); + if (dma_mapping_error(&h->pdev->dev, temp64.val)) { + c->SG[i].Addr.lower = 0; + c->SG[i].Addr.upper = 0; + c->SG[i].Len = 0; + hpsa_pci_unmap(h->pdev, c, i, + PCI_DMA_BIDIRECTIONAL); + status = -ENOMEM; + goto cleanup0; + } c->SG[i].Addr.lower = temp64.val32.lower; c->SG[i].Addr.upper = temp64.val32.upper; c->SG[i].Len = buff_size[i]; - /* we are not chaining */ - c->SG[i].Ext = 0; + c->SG[i].Ext = i < sg_used - 1 ? 0 : HPSA_SG_LAST; } } hpsa_scsi_do_simple_cmd_core_if_no_lockup(h, c); @@ -2765,24 +5139,23 @@ static int hpsa_big_passthru_ioctl(struct ctlr_info *h, void __user *argp) /* Copy the error information out */ memcpy(&ioc->error_info, c->err_info, sizeof(ioc->error_info)); if (copy_to_user(argp, ioc, sizeof(*ioc))) { - cmd_special_free(h, c); status = -EFAULT; - goto cleanup1; + goto cleanup0; } - if (ioc->Request.Type.Direction == XFER_READ && ioc->buf_size > 0) { + if ((ioc->Request.Type.Direction & XFER_READ) && ioc->buf_size > 0) { /* Copy the data out of the buffer we created */ BYTE __user *ptr = ioc->buf; for (i = 0; i < sg_used; i++) { if (copy_to_user(ptr, buff[i], buff_size[i])) { - cmd_special_free(h, c); status = -EFAULT; - goto cleanup1; + goto cleanup0; } ptr += buff_size[i]; } } - cmd_special_free(h, c); status = 0; +cleanup0: + cmd_special_free(h, c); cleanup1: if (buff) { for (i = 0; i < sg_used; i++) @@ -2801,6 +5174,36 @@ static void check_ioctl_unit_attention(struct ctlr_info *h, c->err_info->ScsiStatus != SAM_STAT_CHECK_CONDITION) (void) check_for_unit_attention(h, c); } + +static int increment_passthru_count(struct ctlr_info *h) +{ + unsigned long flags; + + spin_lock_irqsave(&h->passthru_count_lock, flags); + if (h->passthru_count >= HPSA_MAX_CONCURRENT_PASSTHRUS) { + spin_unlock_irqrestore(&h->passthru_count_lock, flags); + return -1; + } + h->passthru_count++; + spin_unlock_irqrestore(&h->passthru_count_lock, flags); + return 0; +} + +static void decrement_passthru_count(struct ctlr_info *h) +{ + unsigned long flags; + + spin_lock_irqsave(&h->passthru_count_lock, flags); + if (h->passthru_count <= 0) { + spin_unlock_irqrestore(&h->passthru_count_lock, flags); + /* not expecting to get here. */ + dev_warn(&h->pdev->dev, "Bug detected, passthru_count seems to be incorrect.\n"); + return; + } + h->passthru_count--; + spin_unlock_irqrestore(&h->passthru_count_lock, flags); +} + /* * ioctl */ @@ -2808,6 +5211,7 @@ static int hpsa_ioctl(struct scsi_device *dev, int cmd, void *arg) { struct ctlr_info *h; void __user *argp = (void __user *)arg; + int rc; h = sdev_to_hba(dev); @@ -2822,23 +5226,32 @@ static int hpsa_ioctl(struct scsi_device *dev, int cmd, void *arg) case CCISS_GETDRIVVER: return hpsa_getdrivver_ioctl(h, argp); case CCISS_PASSTHRU: - return hpsa_passthru_ioctl(h, argp); + if (increment_passthru_count(h)) + return -EAGAIN; + rc = hpsa_passthru_ioctl(h, argp); + decrement_passthru_count(h); + return rc; case CCISS_BIG_PASSTHRU: - return hpsa_big_passthru_ioctl(h, argp); + if (increment_passthru_count(h)) + return -EAGAIN; + rc = hpsa_big_passthru_ioctl(h, argp); + decrement_passthru_count(h); + return rc; default: return -ENOTTY; } } -static int __devinit hpsa_send_host_reset(struct ctlr_info *h, - unsigned char *scsi3addr, u8 reset_type) +static int hpsa_send_host_reset(struct ctlr_info *h, unsigned char *scsi3addr, + u8 reset_type) { struct CommandList *c; c = cmd_alloc(h); if (!c) return -ENOMEM; - fill_cmd(c, HPSA_DEVICE_RESET_MSG, h, NULL, 0, 0, + /* fill_cmd can't fail here, no data buffer to map */ + (void) fill_cmd(c, HPSA_DEVICE_RESET_MSG, h, NULL, 0, 0, RAID_CTLR_LUNID, TYPE_MSG); c->Request.CDB[1] = reset_type; /* fill_cmd defaults to target reset */ c->waiting = NULL; @@ -2850,11 +5263,12 @@ static int __devinit hpsa_send_host_reset(struct ctlr_info *h, return 0; } -static void fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, - void *buff, size_t size, u8 page_code, unsigned char *scsi3addr, +static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, + void *buff, size_t size, u16 page_code, unsigned char *scsi3addr, int cmd_type) { int pci_dir = XFER_NONE; + struct CommandList *a; /* for commands to be aborted */ c->cmd_type = CMD_IOCTL_PEND; c->Header.ReplyQueue = 0; @@ -2873,9 +5287,9 @@ static void fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, switch (cmd) { case HPSA_INQUIRY: /* are we trying to read a vital product page */ - if (page_code != 0) { + if (page_code & VPD_PAGE) { c->Request.CDB[1] = 0x01; - c->Request.CDB[2] = page_code; + c->Request.CDB[2] = (page_code & 0xff); } c->Request.CDBLen = 6; c->Request.Type.Attribute = ATTR_SIMPLE; @@ -2915,10 +5329,32 @@ static void fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, c->Request.Type.Direction = XFER_NONE; c->Request.Timeout = 0; break; + case HPSA_GET_RAID_MAP: + c->Request.CDBLen = 12; + c->Request.Type.Attribute = ATTR_SIMPLE; + c->Request.Type.Direction = XFER_READ; + c->Request.Timeout = 0; + c->Request.CDB[0] = HPSA_CISS_READ; + c->Request.CDB[1] = cmd; + c->Request.CDB[6] = (size >> 24) & 0xFF; /* MSB */ + c->Request.CDB[7] = (size >> 16) & 0xFF; + c->Request.CDB[8] = (size >> 8) & 0xFF; + c->Request.CDB[9] = size & 0xFF; + break; + case BMIC_SENSE_CONTROLLER_PARAMETERS: + c->Request.CDBLen = 10; + c->Request.Type.Attribute = ATTR_SIMPLE; + c->Request.Type.Direction = XFER_READ; + c->Request.Timeout = 0; + c->Request.CDB[0] = BMIC_READ; + c->Request.CDB[6] = BMIC_SENSE_CONTROLLER_PARAMETERS; + c->Request.CDB[7] = (size >> 16) & 0xFF; + c->Request.CDB[8] = (size >> 8) & 0xFF; + break; default: dev_warn(&h->pdev->dev, "unknown command 0x%c\n", cmd); BUG(); - return; + return -1; } } else if (cmd_type == TYPE_MSG) { switch (cmd) { @@ -2931,15 +5367,42 @@ static void fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, c->Request.Timeout = 0; /* Don't time out */ memset(&c->Request.CDB[0], 0, sizeof(c->Request.CDB)); c->Request.CDB[0] = cmd; - c->Request.CDB[1] = 0x03; /* Reset target above */ + c->Request.CDB[1] = HPSA_RESET_TYPE_LUN; /* If bytes 4-7 are zero, it means reset the */ /* LunID device */ c->Request.CDB[4] = 0x00; c->Request.CDB[5] = 0x00; c->Request.CDB[6] = 0x00; c->Request.CDB[7] = 0x00; + break; + case HPSA_ABORT_MSG: + a = buff; /* point to command to be aborted */ + dev_dbg(&h->pdev->dev, "Abort Tag:0x%08x:%08x using request Tag:0x%08x:%08x\n", + a->Header.Tag.upper, a->Header.Tag.lower, + c->Header.Tag.upper, c->Header.Tag.lower); + c->Request.CDBLen = 16; + c->Request.Type.Type = TYPE_MSG; + c->Request.Type.Attribute = ATTR_SIMPLE; + c->Request.Type.Direction = XFER_WRITE; + c->Request.Timeout = 0; /* Don't time out */ + c->Request.CDB[0] = HPSA_TASK_MANAGEMENT; + c->Request.CDB[1] = HPSA_TMF_ABORT_TASK; + c->Request.CDB[2] = 0x00; /* reserved */ + c->Request.CDB[3] = 0x00; /* reserved */ + /* Tag to abort goes in CDB[4]-CDB[11] */ + c->Request.CDB[4] = a->Header.Tag.lower & 0xFF; + c->Request.CDB[5] = (a->Header.Tag.lower >> 8) & 0xFF; + c->Request.CDB[6] = (a->Header.Tag.lower >> 16) & 0xFF; + c->Request.CDB[7] = (a->Header.Tag.lower >> 24) & 0xFF; + c->Request.CDB[8] = a->Header.Tag.upper & 0xFF; + c->Request.CDB[9] = (a->Header.Tag.upper >> 8) & 0xFF; + c->Request.CDB[10] = (a->Header.Tag.upper >> 16) & 0xFF; + c->Request.CDB[11] = (a->Header.Tag.upper >> 24) & 0xFF; + c->Request.CDB[12] = 0x00; /* reserved */ + c->Request.CDB[13] = 0x00; /* reserved */ + c->Request.CDB[14] = 0x00; /* reserved */ + c->Request.CDB[15] = 0x00; /* reserved */ break; - default: dev_warn(&h->pdev->dev, "unknown message type %d\n", cmd); @@ -2963,10 +5426,9 @@ static void fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, default: pci_dir = PCI_DMA_BIDIRECTIONAL; } - - hpsa_map_one(h->pdev, c, buff, size, pci_dir); - - return; + if (hpsa_map_one(h->pdev, c, buff, size, pci_dir)) + return -1; + return 0; } /* @@ -2976,15 +5438,17 @@ static void __iomem *remap_pci_mem(ulong base, ulong size) { ulong page_base = ((ulong) base) & PAGE_MASK; ulong page_offs = ((ulong) base) - page_base; - void __iomem *page_remapped = ioremap(page_base, page_offs + size); + void __iomem *page_remapped = ioremap_nocache(page_base, + page_offs + size); return page_remapped ? (page_remapped + page_offs) : NULL; } /* Takes cmds off the submission queue and sends them to the hardware, * then puts them on the queue of cmds waiting for completion. + * Assumes h->lock is held */ -static void start_io(struct ctlr_info *h) +static void start_io(struct ctlr_info *h, unsigned long *flags) { struct CommandList *c; @@ -2992,25 +5456,44 @@ static void start_io(struct ctlr_info *h) c = list_entry(h->reqQ.next, struct CommandList, list); /* can't do anything if fifo is full */ if ((h->access.fifo_full(h))) { + h->fifo_recently_full = 1; dev_warn(&h->pdev->dev, "fifo full\n"); break; } + h->fifo_recently_full = 0; /* Get the first entry from the Request Q */ removeQ(c); h->Qdepth--; - /* Tell the controller execute command */ - h->access.submit_command(h, c); - /* Put job onto the completed Q */ addQ(&h->cmpQ, c); + + /* Must increment commands_outstanding before unlocking + * and submitting to avoid race checking for fifo full + * condition. + */ + h->commands_outstanding++; + + /* Tell the controller execute command */ + spin_unlock_irqrestore(&h->lock, *flags); + h->access.submit_command(h, c); + spin_lock_irqsave(&h->lock, *flags); } } -static inline unsigned long get_next_completion(struct ctlr_info *h) +static void lock_and_start_io(struct ctlr_info *h) +{ + unsigned long flags; + + spin_lock_irqsave(&h->lock, flags); + start_io(h, &flags); + spin_unlock_irqrestore(&h->lock, flags); +} + +static inline unsigned long get_next_completion(struct ctlr_info *h, u8 q) { - return h->access.command_completed(h); + return h->access.command_completed(h, q); } static inline bool interrupt_pending(struct ctlr_info *h) @@ -3034,13 +5517,45 @@ static inline int bad_tag(struct ctlr_info *h, u32 tag_index, return 0; } -static inline void finish_cmd(struct CommandList *c, u32 raw_tag) +static inline void finish_cmd(struct CommandList *c) { + unsigned long flags; + int io_may_be_stalled = 0; + struct ctlr_info *h = c->h; + + spin_lock_irqsave(&h->lock, flags); removeQ(c); - if (likely(c->cmd_type == CMD_SCSI)) + + /* + * Check for possibly stalled i/o. + * + * If a fifo_full condition is encountered, requests will back up + * in h->reqQ. This queue is only emptied out by start_io which is + * only called when a new i/o request comes in. If no i/o's are + * forthcoming, the i/o's in h->reqQ can get stuck. So we call + * start_io from here if we detect such a danger. + * + * Normally, we shouldn't hit this case, but pounding on the + * CCISS_PASSTHRU ioctl can provoke it. Only call start_io if + * commands_outstanding is low. We want to avoid calling + * start_io from in here as much as possible, and esp. don't + * want to get in a cycle where we call start_io every time + * through here. + */ + if (unlikely(h->fifo_recently_full) && + h->commands_outstanding < 5) + io_may_be_stalled = 1; + + spin_unlock_irqrestore(&h->lock, flags); + + dial_up_lockup_detection_on_fw_flash_complete(c->h, c); + if (likely(c->cmd_type == CMD_IOACCEL1 || c->cmd_type == CMD_SCSI + || c->cmd_type == CMD_IOACCEL2)) complete_scsi_command(c); else if (c->cmd_type == CMD_IOCTL_PEND) complete(c->waiting); + if (unlikely(io_may_be_stalled)) + lock_and_start_io(h); } static inline u32 hpsa_tag_contains_index(u32 tag) @@ -3064,36 +5579,38 @@ static inline u32 hpsa_tag_discard_error_bits(struct ctlr_info *h, u32 tag) } /* process completion of an indexed ("direct lookup") command */ -static inline u32 process_indexed_cmd(struct ctlr_info *h, +static inline void process_indexed_cmd(struct ctlr_info *h, u32 raw_tag) { u32 tag_index; struct CommandList *c; tag_index = hpsa_tag_to_index(raw_tag); - if (bad_tag(h, tag_index, raw_tag)) - return next_command(h); - c = h->cmd_pool + tag_index; - finish_cmd(c, raw_tag); - return next_command(h); + if (!bad_tag(h, tag_index, raw_tag)) { + c = h->cmd_pool + tag_index; + finish_cmd(c); + } } /* process completion of a non-indexed command */ -static inline u32 process_nonindexed_cmd(struct ctlr_info *h, +static inline void process_nonindexed_cmd(struct ctlr_info *h, u32 raw_tag) { u32 tag; struct CommandList *c = NULL; + unsigned long flags; tag = hpsa_tag_discard_error_bits(h, raw_tag); + spin_lock_irqsave(&h->lock, flags); list_for_each_entry(c, &h->cmpQ, list) { if ((c->busaddr & 0xFFFFFFE0) == (tag & 0xFFFFFFE0)) { - finish_cmd(c, raw_tag); - return next_command(h); + spin_unlock_irqrestore(&h->lock, flags); + finish_cmd(c); + return; } } + spin_unlock_irqrestore(&h->lock, flags); bad_tag(h, h->nr_cmds + 1, raw_tag); - return next_command(h); } /* Some controllers, like p400, will give us one interrupt @@ -3115,10 +5632,20 @@ static int ignore_bogus_interrupt(struct ctlr_info *h) return 1; } -static irqreturn_t hpsa_intx_discard_completions(int irq, void *dev_id) +/* + * Convert &h->q[x] (passed to interrupt handlers) back to h. + * Relies on (h-q[x] == x) being true for x such that + * 0 <= x < MAX_REPLY_QUEUES. + */ +static struct ctlr_info *queue_to_hba(u8 *queue) { - struct ctlr_info *h = dev_id; - unsigned long flags; + return container_of((queue - *queue), struct ctlr_info, q[0]); +} + +static irqreturn_t hpsa_intx_discard_completions(int irq, void *queue) +{ + struct ctlr_info *h = queue_to_hba(queue); + u8 q = *(u8 *) queue; u32 raw_tag; if (ignore_bogus_interrupt(h)) @@ -3126,74 +5653,68 @@ static irqreturn_t hpsa_intx_discard_completions(int irq, void *dev_id) if (interrupt_not_for_us(h)) return IRQ_NONE; - spin_lock_irqsave(&h->lock, flags); h->last_intr_timestamp = get_jiffies_64(); while (interrupt_pending(h)) { - raw_tag = get_next_completion(h); + raw_tag = get_next_completion(h, q); while (raw_tag != FIFO_EMPTY) - raw_tag = next_command(h); + raw_tag = next_command(h, q); } - spin_unlock_irqrestore(&h->lock, flags); return IRQ_HANDLED; } -static irqreturn_t hpsa_msix_discard_completions(int irq, void *dev_id) +static irqreturn_t hpsa_msix_discard_completions(int irq, void *queue) { - struct ctlr_info *h = dev_id; - unsigned long flags; + struct ctlr_info *h = queue_to_hba(queue); u32 raw_tag; + u8 q = *(u8 *) queue; if (ignore_bogus_interrupt(h)) return IRQ_NONE; - spin_lock_irqsave(&h->lock, flags); h->last_intr_timestamp = get_jiffies_64(); - raw_tag = get_next_completion(h); + raw_tag = get_next_completion(h, q); while (raw_tag != FIFO_EMPTY) - raw_tag = next_command(h); - spin_unlock_irqrestore(&h->lock, flags); + raw_tag = next_command(h, q); return IRQ_HANDLED; } -static irqreturn_t do_hpsa_intr_intx(int irq, void *dev_id) +static irqreturn_t do_hpsa_intr_intx(int irq, void *queue) { - struct ctlr_info *h = dev_id; - unsigned long flags; + struct ctlr_info *h = queue_to_hba((u8 *) queue); u32 raw_tag; + u8 q = *(u8 *) queue; if (interrupt_not_for_us(h)) return IRQ_NONE; - spin_lock_irqsave(&h->lock, flags); h->last_intr_timestamp = get_jiffies_64(); while (interrupt_pending(h)) { - raw_tag = get_next_completion(h); + raw_tag = get_next_completion(h, q); while (raw_tag != FIFO_EMPTY) { - if (hpsa_tag_contains_index(raw_tag)) - raw_tag = process_indexed_cmd(h, raw_tag); + if (likely(hpsa_tag_contains_index(raw_tag))) + process_indexed_cmd(h, raw_tag); else - raw_tag = process_nonindexed_cmd(h, raw_tag); + process_nonindexed_cmd(h, raw_tag); + raw_tag = next_command(h, q); } } - spin_unlock_irqrestore(&h->lock, flags); return IRQ_HANDLED; } -static irqreturn_t do_hpsa_intr_msi(int irq, void *dev_id) +static irqreturn_t do_hpsa_intr_msi(int irq, void *queue) { - struct ctlr_info *h = dev_id; - unsigned long flags; + struct ctlr_info *h = queue_to_hba(queue); u32 raw_tag; + u8 q = *(u8 *) queue; - spin_lock_irqsave(&h->lock, flags); h->last_intr_timestamp = get_jiffies_64(); - raw_tag = get_next_completion(h); + raw_tag = get_next_completion(h, q); while (raw_tag != FIFO_EMPTY) { - if (hpsa_tag_contains_index(raw_tag)) - raw_tag = process_indexed_cmd(h, raw_tag); + if (likely(hpsa_tag_contains_index(raw_tag))) + process_indexed_cmd(h, raw_tag); else - raw_tag = process_nonindexed_cmd(h, raw_tag); + process_nonindexed_cmd(h, raw_tag); + raw_tag = next_command(h, q); } - spin_unlock_irqrestore(&h->lock, flags); return IRQ_HANDLED; } @@ -3201,8 +5722,8 @@ static irqreturn_t do_hpsa_intr_msi(int irq, void *dev_id) * in simple mode, not performant mode due to the tag lookup. * We only ever use this immediately after a controller reset. */ -static __devinit int hpsa_message(struct pci_dev *pdev, unsigned char opcode, - unsigned char type) +static int hpsa_message(struct pci_dev *pdev, unsigned char opcode, + unsigned char type) { struct Command { struct CommandListHeader CommandHeader; @@ -3310,6 +5831,13 @@ static int hpsa_controller_hard_reset(struct pci_dev *pdev, */ dev_info(&pdev->dev, "using doorbell to reset controller\n"); writel(use_doorbell, vaddr + SA5_DOORBELL); + + /* PMC hardware guys tell us we need a 10 second delay after + * doorbell reset and before any attempt to talk to the board + * at all to ensure that this actually works and doesn't fall + * over in some weird corner cases. + */ + msleep(10000); } else { /* Try to do it the PCI power state way */ /* Quoting from the Open CISS Specification: "The Power @@ -3351,14 +5879,13 @@ static int hpsa_controller_hard_reset(struct pci_dev *pdev, return 0; } -static __devinit void init_driver_version(char *driver_version, int len) +static void init_driver_version(char *driver_version, int len) { memset(driver_version, 0, len); - strncpy(driver_version, "hpsa " HPSA_DRIVER_VERSION, len - 1); + strncpy(driver_version, HPSA " " HPSA_DRIVER_VERSION, len - 1); } -static __devinit int write_driver_ver_to_cfgtable( - struct CfgTable __iomem *cfgtable) +static int write_driver_ver_to_cfgtable(struct CfgTable __iomem *cfgtable) { char *driver_version; int i, size = sizeof(cfgtable->driver_version); @@ -3374,8 +5901,8 @@ static __devinit int write_driver_ver_to_cfgtable( return 0; } -static __devinit void read_driver_ver_from_cfgtable( - struct CfgTable __iomem *cfgtable, unsigned char *driver_ver) +static void read_driver_ver_from_cfgtable(struct CfgTable __iomem *cfgtable, + unsigned char *driver_ver) { int i; @@ -3383,8 +5910,7 @@ static __devinit void read_driver_ver_from_cfgtable( driver_ver[i] = readb(&cfgtable->driver_version[i]); } -static __devinit int controller_reset_failed( - struct CfgTable __iomem *cfgtable) +static int controller_reset_failed(struct CfgTable __iomem *cfgtable) { char *driver_ver, *old_driver_ver; @@ -3407,7 +5933,7 @@ static __devinit int controller_reset_failed( /* This does a hard reset of the controller using PCI power management * states or the using the doorbell register. */ -static __devinit int hpsa_kdump_hard_reset_controller(struct pci_dev *pdev) +static int hpsa_kdump_hard_reset_controller(struct pci_dev *pdev) { u64 cfg_offset; u32 cfg_base_addr; @@ -3508,16 +6034,6 @@ static __devinit int hpsa_kdump_hard_reset_controller(struct pci_dev *pdev) need a little pause here */ msleep(HPSA_POST_RESET_PAUSE_MSECS); - /* Wait for board to become not ready, then ready. */ - dev_info(&pdev->dev, "Waiting for board to reset.\n"); - rc = hpsa_wait_for_board_state(pdev, vaddr, BOARD_NOT_READY); - if (rc) { - dev_warn(&pdev->dev, - "failed waiting for board to reset." - " Will try soft reset.\n"); - rc = -ENOTSUPP; /* Not expected, but try soft reset later */ - goto unmap_cfgtable; - } rc = hpsa_wait_for_board_state(pdev, vaddr, BOARD_READY); if (rc) { dev_warn(&pdev->dev, @@ -3624,13 +6140,16 @@ static int find_PCI_BAR_index(struct pci_dev *pdev, unsigned long pci_bar_addr) * controllers that are capable. If not, we use IO-APIC mode. */ -static void __devinit hpsa_interrupt_mode(struct ctlr_info *h) +static void hpsa_interrupt_mode(struct ctlr_info *h) { #ifdef CONFIG_PCI_MSI - int err; - struct msix_entry hpsa_msix_entries[4] = { {0, 0}, {0, 1}, - {0, 2}, {0, 3} - }; + int err, i; + struct msix_entry hpsa_msix_entries[MAX_REPLY_QUEUES]; + + for (i = 0; i < MAX_REPLY_QUEUES; i++) { + hpsa_msix_entries[i].vector = 0; + hpsa_msix_entries[i].entry = i; + } /* Some boards advertise MSI but don't really support it */ if ((h->board_id == 0x40700E11) || (h->board_id == 0x40800E11) || @@ -3638,22 +6157,26 @@ static void __devinit hpsa_interrupt_mode(struct ctlr_info *h) goto default_int_mode; if (pci_find_capability(h->pdev, PCI_CAP_ID_MSIX)) { dev_info(&h->pdev->dev, "MSIX\n"); - err = pci_enable_msix(h->pdev, hpsa_msix_entries, 4); - if (!err) { - h->intr[0] = hpsa_msix_entries[0].vector; - h->intr[1] = hpsa_msix_entries[1].vector; - h->intr[2] = hpsa_msix_entries[2].vector; - h->intr[3] = hpsa_msix_entries[3].vector; - h->msix_vector = 1; - return; - } + h->msix_vector = MAX_REPLY_QUEUES; + if (h->msix_vector > num_online_cpus()) + h->msix_vector = num_online_cpus(); + err = pci_enable_msix(h->pdev, hpsa_msix_entries, + h->msix_vector); if (err > 0) { dev_warn(&h->pdev->dev, "only %d MSI-X vectors " "available\n", err); - goto default_int_mode; + h->msix_vector = err; + err = pci_enable_msix(h->pdev, hpsa_msix_entries, + h->msix_vector); + } + if (!err) { + for (i = 0; i < h->msix_vector; i++) + h->intr[i] = hpsa_msix_entries[i].vector; + return; } else { dev_warn(&h->pdev->dev, "MSI-X init failed %d\n", err); + h->msix_vector = 0; goto default_int_mode; } } @@ -3670,7 +6193,7 @@ default_int_mode: h->intr[h->intr_mode] = h->pdev->irq; } -static int __devinit hpsa_lookup_board_id(struct pci_dev *pdev, u32 *board_id) +static int hpsa_lookup_board_id(struct pci_dev *pdev, u32 *board_id) { int i; u32 subsystem_vendor_id, subsystem_device_id; @@ -3694,16 +6217,8 @@ static int __devinit hpsa_lookup_board_id(struct pci_dev *pdev, u32 *board_id) return ARRAY_SIZE(products) - 1; /* generic unknown smart array */ } -static inline bool hpsa_board_disabled(struct pci_dev *pdev) -{ - u16 command; - - (void) pci_read_config_word(pdev, PCI_COMMAND, &command); - return ((command & PCI_COMMAND_MEMORY) == 0); -} - -static int __devinit hpsa_pci_find_memory_BAR(struct pci_dev *pdev, - unsigned long *memory_bar) +static int hpsa_pci_find_memory_BAR(struct pci_dev *pdev, + unsigned long *memory_bar) { int i; @@ -3719,8 +6234,8 @@ static int __devinit hpsa_pci_find_memory_BAR(struct pci_dev *pdev, return -ENODEV; } -static int __devinit hpsa_wait_for_board_state(struct pci_dev *pdev, - void __iomem *vaddr, int wait_for_ready) +static int hpsa_wait_for_board_state(struct pci_dev *pdev, void __iomem *vaddr, + int wait_for_ready) { int i, iterations; u32 scratchpad; @@ -3744,9 +6259,9 @@ static int __devinit hpsa_wait_for_board_state(struct pci_dev *pdev, return -ENODEV; } -static int __devinit hpsa_find_cfg_addrs(struct pci_dev *pdev, - void __iomem *vaddr, u32 *cfg_base_addr, u64 *cfg_base_addr_index, - u64 *cfg_offset) +static int hpsa_find_cfg_addrs(struct pci_dev *pdev, void __iomem *vaddr, + u32 *cfg_base_addr, u64 *cfg_base_addr_index, + u64 *cfg_offset) { *cfg_base_addr = readl(vaddr + SA5_CTCFG_OFFSET); *cfg_offset = readl(vaddr + SA5_CTMEM_OFFSET); @@ -3759,7 +6274,7 @@ static int __devinit hpsa_find_cfg_addrs(struct pci_dev *pdev, return 0; } -static int __devinit hpsa_find_cfgtables(struct ctlr_info *h) +static int hpsa_find_cfgtables(struct ctlr_info *h) { u64 cfg_offset; u32 cfg_base_addr; @@ -3788,7 +6303,7 @@ static int __devinit hpsa_find_cfgtables(struct ctlr_info *h) return 0; } -static void __devinit hpsa_get_max_perf_mode_cmds(struct ctlr_info *h) +static void hpsa_get_max_perf_mode_cmds(struct ctlr_info *h) { h->max_commands = readl(&(h->cfgtable->MaxPerformantModeCommands)); @@ -3809,11 +6324,12 @@ static void __devinit hpsa_get_max_perf_mode_cmds(struct ctlr_info *h) * max commands, max SG elements without chaining, and with chaining, * SG chain block size, etc. */ -static void __devinit hpsa_find_board_params(struct ctlr_info *h) +static void hpsa_find_board_params(struct ctlr_info *h) { hpsa_get_max_perf_mode_cmds(h); h->nr_cmds = h->max_commands - 4; /* Allow room for some ioctls */ h->maxsgentries = readl(&(h->cfgtable->MaxScatterGatherElements)); + h->fw_support = readl(&(h->cfgtable->misc_fw_support)); /* * Limit in-command s/g elements to 32 save dma'able memory. * Howvever spec says if 0, use 31 @@ -3827,30 +6343,35 @@ static void __devinit hpsa_find_board_params(struct ctlr_info *h) h->maxsgentries = 31; /* default to traditional values */ h->chainsize = 0; } + + /* Find out what task management functions are supported and cache */ + h->TMFSupportFlags = readl(&(h->cfgtable->TMFSupportFlags)); + if (!(HPSATMF_PHYS_TASK_ABORT & h->TMFSupportFlags)) + dev_warn(&h->pdev->dev, "Physical aborts not supported\n"); + if (!(HPSATMF_LOG_TASK_ABORT & h->TMFSupportFlags)) + dev_warn(&h->pdev->dev, "Logical aborts not supported\n"); } static inline bool hpsa_CISS_signature_present(struct ctlr_info *h) { - if ((readb(&h->cfgtable->Signature[0]) != 'C') || - (readb(&h->cfgtable->Signature[1]) != 'I') || - (readb(&h->cfgtable->Signature[2]) != 'S') || - (readb(&h->cfgtable->Signature[3]) != 'S')) { + if (!check_signature(h->cfgtable->Signature, "CISS", 4)) { dev_warn(&h->pdev->dev, "not a valid CISS config table\n"); return false; } return true; } -/* Need to enable prefetch in the SCSI core for 6400 in x86 */ -static inline void hpsa_enable_scsi_prefetch(struct ctlr_info *h) +static inline void hpsa_set_driver_support_bits(struct ctlr_info *h) { -#ifdef CONFIG_X86 - u32 prefetch; + u32 driver_support; - prefetch = readl(&(h->cfgtable->SCSI_Prefetch)); - prefetch |= 0x100; - writel(prefetch, &(h->cfgtable->SCSI_Prefetch)); +#ifdef CONFIG_X86 + /* Need to enable prefetch in the SCSI core for 6400 in x86 */ + driver_support = readl(&(h->cfgtable->driver_support)); + driver_support |= ENABLE_SCSI_PREFETCH; #endif + driver_support |= ENABLE_UNIT_ATTN; + writel(driver_support, &(h->cfgtable->driver_support)); } /* Disable DMA prefetch for the P600. Otherwise an ASIC bug may result @@ -3867,7 +6388,24 @@ static inline void hpsa_p600_dma_prefetch_quirk(struct ctlr_info *h) writel(dma_prefetch, h->vaddr + I2O_DMA1_CFG); } -static void __devinit hpsa_wait_for_mode_change_ack(struct ctlr_info *h) +static void hpsa_wait_for_clear_event_notify_ack(struct ctlr_info *h) +{ + int i; + u32 doorbell_value; + unsigned long flags; + /* wait until the clear_event_notify bit 6 is cleared by controller. */ + for (i = 0; i < MAX_CONFIG_WAIT; i++) { + spin_lock_irqsave(&h->lock, flags); + doorbell_value = readl(h->vaddr + SA5_DOORBELL); + spin_unlock_irqrestore(&h->lock, flags); + if (!(doorbell_value & DOORBELL_CLEAR_EVENTS)) + break; + /* delay and try again */ + msleep(20); + } +} + +static void hpsa_wait_for_mode_change_ack(struct ctlr_info *h) { int i; u32 doorbell_value; @@ -3888,7 +6426,7 @@ static void __devinit hpsa_wait_for_mode_change_ack(struct ctlr_info *h) } } -static int __devinit hpsa_enter_simple_mode(struct ctlr_info *h) +static int hpsa_enter_simple_mode(struct ctlr_info *h) { u32 trans_support; @@ -3897,21 +6435,23 @@ static int __devinit hpsa_enter_simple_mode(struct ctlr_info *h) return -ENOTSUPP; h->max_commands = readl(&(h->cfgtable->CmdsOutMax)); + /* Update the field, and then ring the doorbell */ writel(CFGTBL_Trans_Simple, &(h->cfgtable->HostWrite.TransportRequest)); + writel(0, &h->cfgtable->HostWrite.command_pool_addr_hi); writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL); hpsa_wait_for_mode_change_ack(h); print_cfg_table(&h->pdev->dev, h->cfgtable); - if (!(readl(&(h->cfgtable->TransportActive)) & CFGTBL_Trans_Simple)) { - dev_warn(&h->pdev->dev, - "unable to get board into simple mode\n"); - return -ENODEV; - } + if (!(readl(&(h->cfgtable->TransportActive)) & CFGTBL_Trans_Simple)) + goto error; h->transMethod = CFGTBL_Trans_Simple; return 0; +error: + dev_warn(&h->pdev->dev, "unable to get board into simple mode\n"); + return -ENODEV; } -static int __devinit hpsa_pci_init(struct ctlr_info *h) +static int hpsa_pci_init(struct ctlr_info *h) { int prod_index, err; @@ -3921,11 +6461,6 @@ static int __devinit hpsa_pci_init(struct ctlr_info *h) h->product_name = products[prod_index].product_name; h->access = *(products[prod_index].access); - if (hpsa_board_disabled(h->pdev)) { - dev_warn(&h->pdev->dev, "controller appears to be disabled\n"); - return -ENODEV; - } - pci_disable_link_state(h->pdev, PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1 | PCIE_LINK_STATE_CLKPM); @@ -3935,7 +6470,10 @@ static int __devinit hpsa_pci_init(struct ctlr_info *h) return err; } - err = pci_request_regions(h->pdev, "hpsa"); + /* Enable bus mastering (pci_disable_device may disable this) */ + pci_set_master(h->pdev); + + err = pci_request_regions(h->pdev, HPSA); if (err) { dev_err(&h->pdev->dev, "cannot obtain PCI resources, aborting\n"); @@ -3962,7 +6500,7 @@ static int __devinit hpsa_pci_init(struct ctlr_info *h) err = -ENODEV; goto err_out_free_res; } - hpsa_enable_scsi_prefetch(h); + hpsa_set_driver_support_bits(h); hpsa_p600_dma_prefetch_quirk(h); err = hpsa_enter_simple_mode(h); if (err) @@ -3976,15 +6514,12 @@ err_out_free_res: iounmap(h->cfgtable); if (h->vaddr) iounmap(h->vaddr); - /* - * Deliberately omit pci_disable_device(): it does something nasty to - * Smart Array controllers that pci_enable_device does not undo - */ + pci_disable_device(h->pdev); pci_release_regions(h->pdev); return err; } -static void __devinit hpsa_hba_inquiry(struct ctlr_info *h) +static void hpsa_hba_inquiry(struct ctlr_info *h) { int rc; @@ -4000,7 +6535,7 @@ static void __devinit hpsa_hba_inquiry(struct ctlr_info *h) } } -static __devinit int hpsa_init_reset_devices(struct pci_dev *pdev) +static int hpsa_init_reset_devices(struct pci_dev *pdev) { int rc, i; @@ -4032,7 +6567,7 @@ static __devinit int hpsa_init_reset_devices(struct pci_dev *pdev) return 0; } -static __devinit int hpsa_allocate_cmd_pool(struct ctlr_info *h) +static int hpsa_allocate_cmd_pool(struct ctlr_info *h) { h->cmd_pool_bits = kzalloc( DIV_ROUND_UP(h->nr_cmds, BITS_PER_LONG) * @@ -4059,25 +6594,64 @@ static void hpsa_free_cmd_pool(struct ctlr_info *h) pci_free_consistent(h->pdev, h->nr_cmds * sizeof(struct CommandList), h->cmd_pool, h->cmd_pool_dhandle); + if (h->ioaccel2_cmd_pool) + pci_free_consistent(h->pdev, + h->nr_cmds * sizeof(*h->ioaccel2_cmd_pool), + h->ioaccel2_cmd_pool, h->ioaccel2_cmd_pool_dhandle); if (h->errinfo_pool) pci_free_consistent(h->pdev, h->nr_cmds * sizeof(struct ErrorInfo), h->errinfo_pool, h->errinfo_pool_dhandle); + if (h->ioaccel_cmd_pool) + pci_free_consistent(h->pdev, + h->nr_cmds * sizeof(struct io_accel1_cmd), + h->ioaccel_cmd_pool, h->ioaccel_cmd_pool_dhandle); +} + +static void hpsa_irq_affinity_hints(struct ctlr_info *h) +{ + int i, cpu, rc; + + cpu = cpumask_first(cpu_online_mask); + for (i = 0; i < h->msix_vector; i++) { + rc = irq_set_affinity_hint(h->intr[i], get_cpu_mask(cpu)); + cpu = cpumask_next(cpu, cpu_online_mask); + } } static int hpsa_request_irq(struct ctlr_info *h, irqreturn_t (*msixhandler)(int, void *), irqreturn_t (*intxhandler)(int, void *)) { - int rc; + int rc, i; - if (h->msix_vector || h->msi_vector) - rc = request_irq(h->intr[h->intr_mode], msixhandler, - 0, h->devname, h); - else - rc = request_irq(h->intr[h->intr_mode], intxhandler, - IRQF_SHARED, h->devname, h); + /* + * initialize h->q[x] = x so that interrupt handlers know which + * queue to process. + */ + for (i = 0; i < MAX_REPLY_QUEUES; i++) + h->q[i] = (u8) i; + + if (h->intr_mode == PERF_MODE_INT && h->msix_vector > 0) { + /* If performant mode and MSI-X, use multiple reply queues */ + for (i = 0; i < h->msix_vector; i++) + rc = request_irq(h->intr[i], msixhandler, + 0, h->devname, + &h->q[i]); + hpsa_irq_affinity_hints(h); + } else { + /* Use single reply pool */ + if (h->msix_vector > 0 || h->msi_vector) { + rc = request_irq(h->intr[h->intr_mode], + msixhandler, 0, h->devname, + &h->q[h->intr_mode]); + } else { + rc = request_irq(h->intr[h->intr_mode], + intxhandler, IRQF_SHARED, h->devname, + &h->q[h->intr_mode]); + } + } if (rc) { dev_err(&h->pdev->dev, "unable to get irq %d for %s\n", h->intr[h->intr_mode], h->devname); @@ -4086,7 +6660,7 @@ static int hpsa_request_irq(struct ctlr_info *h, return 0; } -static int __devinit hpsa_kdump_soft_reset(struct ctlr_info *h) +static int hpsa_kdump_soft_reset(struct ctlr_info *h) { if (hpsa_send_host_reset(h, RAID_CTLR_LUNID, HPSA_RESET_TYPE_CONTROLLER)) { @@ -4110,20 +6684,60 @@ static int __devinit hpsa_kdump_soft_reset(struct ctlr_info *h) return 0; } -static void hpsa_undo_allocations_after_kdump_soft_reset(struct ctlr_info *h) +static void free_irqs(struct ctlr_info *h) { - free_irq(h->intr[h->intr_mode], h); + int i; + + if (!h->msix_vector || h->intr_mode != PERF_MODE_INT) { + /* Single reply queue, only one irq to free */ + i = h->intr_mode; + irq_set_affinity_hint(h->intr[i], NULL); + free_irq(h->intr[i], &h->q[i]); + return; + } + + for (i = 0; i < h->msix_vector; i++) { + irq_set_affinity_hint(h->intr[i], NULL); + free_irq(h->intr[i], &h->q[i]); + } +} + +static void hpsa_free_irqs_and_disable_msix(struct ctlr_info *h) +{ + free_irqs(h); #ifdef CONFIG_PCI_MSI - if (h->msix_vector) - pci_disable_msix(h->pdev); - else if (h->msi_vector) - pci_disable_msi(h->pdev); + if (h->msix_vector) { + if (h->pdev->msix_enabled) + pci_disable_msix(h->pdev); + } else if (h->msi_vector) { + if (h->pdev->msi_enabled) + pci_disable_msi(h->pdev); + } #endif /* CONFIG_PCI_MSI */ +} + +static void hpsa_free_reply_queues(struct ctlr_info *h) +{ + int i; + + for (i = 0; i < h->nreply_queues; i++) { + if (!h->reply_queue[i].head) + continue; + pci_free_consistent(h->pdev, h->reply_queue_size, + h->reply_queue[i].head, h->reply_queue[i].busaddr); + h->reply_queue[i].head = NULL; + h->reply_queue[i].busaddr = 0; + } +} + +static void hpsa_undo_allocations_after_kdump_soft_reset(struct ctlr_info *h) +{ + hpsa_free_irqs_and_disable_msix(h); hpsa_free_sg_chain_blocks(h); hpsa_free_cmd_pool(h); + kfree(h->ioaccel1_blockFetchTable); kfree(h->blockFetchTable); - pci_free_consistent(h->pdev, h->reply_pool_size, - h->reply_pool, h->reply_pool_dhandle); + hpsa_free_reply_queues(h); if (h->vaddr) iounmap(h->vaddr); if (h->transtable) @@ -4134,16 +6748,6 @@ static void hpsa_undo_allocations_after_kdump_soft_reset(struct ctlr_info *h) kfree(h); } -static void remove_ctlr_from_lockup_detector_list(struct ctlr_info *h) -{ - assert_spin_locked(&lockup_detector_lock); - if (!hpsa_lockup_detector) - return; - if (h->lockup_detected) - return; /* already stopped the lockup detector */ - list_del(&h->lockup_list); -} - /* Called when controller lockup detected. */ static void fail_all_cmds_on_list(struct ctlr_info *h, struct list_head *list) { @@ -4154,22 +6758,42 @@ static void fail_all_cmds_on_list(struct ctlr_info *h, struct list_head *list) while (!list_empty(list)) { c = list_entry(list->next, struct CommandList, list); c->err_info->CommandStatus = CMD_HARDWARE_ERR; - finish_cmd(c, c->Header.Tag.lower); + finish_cmd(c); + } +} + +static void set_lockup_detected_for_all_cpus(struct ctlr_info *h, u32 value) +{ + int i, cpu; + + cpu = cpumask_first(cpu_online_mask); + for (i = 0; i < num_online_cpus(); i++) { + u32 *lockup_detected; + lockup_detected = per_cpu_ptr(h->lockup_detected, cpu); + *lockup_detected = value; + cpu = cpumask_next(cpu, cpu_online_mask); } + wmb(); /* be sure the per-cpu variables are out to memory */ } static void controller_lockup_detected(struct ctlr_info *h) { unsigned long flags; + u32 lockup_detected; - assert_spin_locked(&lockup_detector_lock); - remove_ctlr_from_lockup_detector_list(h); h->access.set_intr_mask(h, HPSA_INTR_OFF); spin_lock_irqsave(&h->lock, flags); - h->lockup_detected = readl(h->vaddr + SA5_SCRATCHPAD_OFFSET); + lockup_detected = readl(h->vaddr + SA5_SCRATCHPAD_OFFSET); + if (!lockup_detected) { + /* no heartbeat, but controller gave us a zero. */ + dev_warn(&h->pdev->dev, + "lockup detected but scratchpad register is zero\n"); + lockup_detected = 0xffffffff; + } + set_lockup_detected_for_all_cpus(h, lockup_detected); spin_unlock_irqrestore(&h->lock, flags); dev_warn(&h->pdev->dev, "Controller lockup detected: 0x%08x\n", - h->lockup_detected); + lockup_detected); pci_disable_device(h->pdev); spin_lock_irqsave(&h->lock, flags); fail_all_cmds_on_list(h, &h->cmpQ); @@ -4177,20 +6801,16 @@ static void controller_lockup_detected(struct ctlr_info *h) spin_unlock_irqrestore(&h->lock, flags); } -#define HEARTBEAT_SAMPLE_INTERVAL (10 * HZ) -#define HEARTBEAT_CHECK_MINIMUM_INTERVAL (HEARTBEAT_SAMPLE_INTERVAL / 2) - static void detect_controller_lockup(struct ctlr_info *h) { u64 now; u32 heartbeat; unsigned long flags; - assert_spin_locked(&lockup_detector_lock); now = get_jiffies_64(); /* If we've received an interrupt recently, we're ok. */ if (time_after64(h->last_intr_timestamp + - (HEARTBEAT_CHECK_MINIMUM_INTERVAL), now)) + (h->heartbeat_sample_interval), now)) return; /* @@ -4199,7 +6819,7 @@ static void detect_controller_lockup(struct ctlr_info *h) * otherwise don't care about signals in this thread. */ if (time_after64(h->last_heartbeat_timestamp + - (HEARTBEAT_CHECK_MINIMUM_INTERVAL), now)) + (h->heartbeat_sample_interval), now)) return; /* If heartbeat has not changed since we last looked, we're not ok. */ @@ -4216,71 +6836,120 @@ static void detect_controller_lockup(struct ctlr_info *h) h->last_heartbeat_timestamp = now; } -static int detect_controller_lockup_thread(void *notused) +static void hpsa_ack_ctlr_events(struct ctlr_info *h) { - struct ctlr_info *h; - unsigned long flags; - - while (1) { - struct list_head *this, *tmp; - - schedule_timeout_interruptible(HEARTBEAT_SAMPLE_INTERVAL); - if (kthread_should_stop()) - break; - spin_lock_irqsave(&lockup_detector_lock, flags); - list_for_each_safe(this, tmp, &hpsa_ctlr_list) { - h = list_entry(this, struct ctlr_info, lockup_list); - detect_controller_lockup(h); - } - spin_unlock_irqrestore(&lockup_detector_lock, flags); + int i; + char *event_type; + + /* Clear the driver-requested rescan flag */ + h->drv_req_rescan = 0; + + /* Ask the controller to clear the events we're handling. */ + if ((h->transMethod & (CFGTBL_Trans_io_accel1 + | CFGTBL_Trans_io_accel2)) && + (h->events & HPSA_EVENT_NOTIFY_ACCEL_IO_PATH_STATE_CHANGE || + h->events & HPSA_EVENT_NOTIFY_ACCEL_IO_PATH_CONFIG_CHANGE)) { + + if (h->events & HPSA_EVENT_NOTIFY_ACCEL_IO_PATH_STATE_CHANGE) + event_type = "state change"; + if (h->events & HPSA_EVENT_NOTIFY_ACCEL_IO_PATH_CONFIG_CHANGE) + event_type = "configuration change"; + /* Stop sending new RAID offload reqs via the IO accelerator */ + scsi_block_requests(h->scsi_host); + for (i = 0; i < h->ndevices; i++) + h->dev[i]->offload_enabled = 0; + hpsa_drain_accel_commands(h); + /* Set 'accelerator path config change' bit */ + dev_warn(&h->pdev->dev, + "Acknowledging event: 0x%08x (HP SSD Smart Path %s)\n", + h->events, event_type); + writel(h->events, &(h->cfgtable->clear_event_notify)); + /* Set the "clear event notify field update" bit 6 */ + writel(DOORBELL_CLEAR_EVENTS, h->vaddr + SA5_DOORBELL); + /* Wait until ctlr clears 'clear event notify field', bit 6 */ + hpsa_wait_for_clear_event_notify_ack(h); + scsi_unblock_requests(h->scsi_host); + } else { + /* Acknowledge controller notification events. */ + writel(h->events, &(h->cfgtable->clear_event_notify)); + writel(DOORBELL_CLEAR_EVENTS, h->vaddr + SA5_DOORBELL); + hpsa_wait_for_clear_event_notify_ack(h); +#if 0 + writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL); + hpsa_wait_for_mode_change_ack(h); +#endif } - return 0; + return; } -static void add_ctlr_to_lockup_detector_list(struct ctlr_info *h) +/* Check a register on the controller to see if there are configuration + * changes (added/changed/removed logical drives, etc.) which mean that + * we should rescan the controller for devices. + * Also check flag for driver-initiated rescan. + */ +static int hpsa_ctlr_needs_rescan(struct ctlr_info *h) { - unsigned long flags; + if (h->drv_req_rescan) + return 1; + + if (!(h->fw_support & MISC_FW_EVENT_NOTIFY)) + return 0; - spin_lock_irqsave(&lockup_detector_lock, flags); - list_add_tail(&h->lockup_list, &hpsa_ctlr_list); - spin_unlock_irqrestore(&lockup_detector_lock, flags); + h->events = readl(&(h->cfgtable->event_notify)); + return h->events & RESCAN_REQUIRED_EVENT_BITS; } -static void start_controller_lockup_detector(struct ctlr_info *h) +/* + * Check if any of the offline devices have become ready + */ +static int hpsa_offline_devices_ready(struct ctlr_info *h) { - /* Start the lockup detector thread if not already started */ - if (!hpsa_lockup_detector) { - spin_lock_init(&lockup_detector_lock); - hpsa_lockup_detector = - kthread_run(detect_controller_lockup_thread, - NULL, "hpsa"); - } - if (!hpsa_lockup_detector) { - dev_warn(&h->pdev->dev, - "Could not start lockup detector thread\n"); - return; + unsigned long flags; + struct offline_device_entry *d; + struct list_head *this, *tmp; + + spin_lock_irqsave(&h->offline_device_lock, flags); + list_for_each_safe(this, tmp, &h->offline_device_list) { + d = list_entry(this, struct offline_device_entry, + offline_list); + spin_unlock_irqrestore(&h->offline_device_lock, flags); + if (!hpsa_volume_offline(h, d->scsi3addr)) + return 1; + spin_lock_irqsave(&h->offline_device_lock, flags); } - add_ctlr_to_lockup_detector_list(h); + spin_unlock_irqrestore(&h->offline_device_lock, flags); + return 0; } -static void stop_controller_lockup_detector(struct ctlr_info *h) + +static void hpsa_monitor_ctlr_worker(struct work_struct *work) { unsigned long flags; + struct ctlr_info *h = container_of(to_delayed_work(work), + struct ctlr_info, monitor_ctlr_work); + detect_controller_lockup(h); + if (lockup_detected(h)) + return; - spin_lock_irqsave(&lockup_detector_lock, flags); - remove_ctlr_from_lockup_detector_list(h); - /* If the list of ctlr's to monitor is empty, stop the thread */ - if (list_empty(&hpsa_ctlr_list)) { - spin_unlock_irqrestore(&lockup_detector_lock, flags); - kthread_stop(hpsa_lockup_detector); - spin_lock_irqsave(&lockup_detector_lock, flags); - hpsa_lockup_detector = NULL; + if (hpsa_ctlr_needs_rescan(h) || hpsa_offline_devices_ready(h)) { + scsi_host_get(h->scsi_host); + h->drv_req_rescan = 0; + hpsa_ack_ctlr_events(h); + hpsa_scan_start(h->scsi_host); + scsi_host_put(h->scsi_host); } - spin_unlock_irqrestore(&lockup_detector_lock, flags); + + spin_lock_irqsave(&h->lock, flags); + if (h->remove_in_progress) { + spin_unlock_irqrestore(&h->lock, flags); + return; + } + schedule_delayed_work(&h->monitor_ctlr_work, + h->heartbeat_sample_interval); + spin_unlock_irqrestore(&h->lock, flags); } -static int __devinit hpsa_init_one(struct pci_dev *pdev, - const struct pci_device_id *ent) +static int hpsa_init_one(struct pci_dev *pdev, const struct pci_device_id *ent) { int dac, rc; struct ctlr_info *h; @@ -4309,7 +6978,6 @@ reinit_after_soft_reset: * the 5 lower bits of the address are used by the hardware. and by * the driver. See comments in hpsa.h for more info. */ -#define COMMANDLIST_ALIGNMENT 32 BUILD_BUG_ON(sizeof(struct CommandList) % COMMANDLIST_ALIGNMENT); h = kzalloc(sizeof(*h), GFP_KERNEL); if (!h) @@ -4319,13 +6987,23 @@ reinit_after_soft_reset: h->intr_mode = hpsa_simple_mode ? SIMPLE_MODE_INT : PERF_MODE_INT; INIT_LIST_HEAD(&h->cmpQ); INIT_LIST_HEAD(&h->reqQ); + INIT_LIST_HEAD(&h->offline_device_list); spin_lock_init(&h->lock); + spin_lock_init(&h->offline_device_lock); spin_lock_init(&h->scan_lock); + spin_lock_init(&h->passthru_count_lock); + + /* Allocate and clear per-cpu variable lockup_detected */ + h->lockup_detected = alloc_percpu(u32); + if (!h->lockup_detected) + goto clean1; + set_lockup_detected_for_all_cpus(h, 0); + rc = hpsa_pci_init(h); if (rc != 0) goto clean1; - sprintf(h->devname, "hpsa%d", number_of_controllers); + sprintf(h->devname, HPSA "%d", number_of_controllers); h->ctlr = number_of_controllers; number_of_controllers++; @@ -4360,6 +7038,7 @@ reinit_after_soft_reset: pci_set_drvdata(pdev, h); h->ndevices = 0; + h->hba_mode_enabled = 0; h->scsi_host = NULL; spin_lock_init(&h->devlock); hpsa_put_ctlr_into_performant_mode(h); @@ -4380,7 +7059,7 @@ reinit_after_soft_reset: spin_lock_irqsave(&h->lock, flags); h->access.set_intr_mask(h, HPSA_INTR_OFF); spin_unlock_irqrestore(&h->lock, flags); - free_irq(h->intr[h->intr_mode], h); + free_irqs(h); rc = hpsa_request_irq(h, hpsa_msix_discard_completions, hpsa_intx_discard_completions); if (rc) { @@ -4419,20 +7098,32 @@ reinit_after_soft_reset: goto reinit_after_soft_reset; } + /* Enable Accelerated IO path at driver layer */ + h->acciopath_status = 1; + + h->drv_req_rescan = 0; + /* Turn the interrupts on so we can service requests */ h->access.set_intr_mask(h, HPSA_INTR_ON); hpsa_hba_inquiry(h); hpsa_register_scsi(h); /* hook ourselves into SCSI subsystem */ - start_controller_lockup_detector(h); - return 1; + + /* Monitor the controller for firmware lockups */ + h->heartbeat_sample_interval = HEARTBEAT_SAMPLE_INTERVAL; + INIT_DELAYED_WORK(&h->monitor_ctlr_work, hpsa_monitor_ctlr_worker); + schedule_delayed_work(&h->monitor_ctlr_work, + h->heartbeat_sample_interval); + return 0; clean4: hpsa_free_sg_chain_blocks(h); hpsa_free_cmd_pool(h); - free_irq(h->intr[h->intr_mode], h); + free_irqs(h); clean2: clean1: + if (h->lockup_detected) + free_percpu(h->lockup_detected); kfree(h); return rc; } @@ -4442,6 +7133,9 @@ static void hpsa_flush_cache(struct ctlr_info *h) char *flush_buf; struct CommandList *c; + /* Don't bother trying to flush the cache if locked up */ + if (unlikely(lockup_detected(h))) + return; flush_buf = kzalloc(4, GFP_KERNEL); if (!flush_buf) return; @@ -4451,10 +7145,13 @@ static void hpsa_flush_cache(struct ctlr_info *h) dev_warn(&h->pdev->dev, "cmd_special_alloc returned NULL!\n"); goto out_of_memory; } - fill_cmd(c, HPSA_CACHE_FLUSH, h, flush_buf, 4, 0, - RAID_CTLR_LUNID, TYPE_CMD); + if (fill_cmd(c, HPSA_CACHE_FLUSH, h, flush_buf, 4, 0, + RAID_CTLR_LUNID, TYPE_CMD)) { + goto out; + } hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_TODEVICE); if (c->err_info->CommandStatus != 0) +out: dev_warn(&h->pdev->dev, "error flushing cache on controller\n"); cmd_special_free(h, c); @@ -4473,30 +7170,40 @@ static void hpsa_shutdown(struct pci_dev *pdev) */ hpsa_flush_cache(h); h->access.set_intr_mask(h, HPSA_INTR_OFF); - free_irq(h->intr[h->intr_mode], h); -#ifdef CONFIG_PCI_MSI - if (h->msix_vector) - pci_disable_msix(h->pdev); - else if (h->msi_vector) - pci_disable_msi(h->pdev); -#endif /* CONFIG_PCI_MSI */ + hpsa_free_irqs_and_disable_msix(h); } -static void __devexit hpsa_remove_one(struct pci_dev *pdev) +static void hpsa_free_device_info(struct ctlr_info *h) +{ + int i; + + for (i = 0; i < h->ndevices; i++) + kfree(h->dev[i]); +} + +static void hpsa_remove_one(struct pci_dev *pdev) { struct ctlr_info *h; + unsigned long flags; if (pci_get_drvdata(pdev) == NULL) { dev_err(&pdev->dev, "unable to remove device\n"); return; } h = pci_get_drvdata(pdev); - stop_controller_lockup_detector(h); + + /* Get rid of any controller monitoring work items */ + spin_lock_irqsave(&h->lock, flags); + h->remove_in_progress = 1; + cancel_delayed_work(&h->monitor_ctlr_work); + spin_unlock_irqrestore(&h->lock, flags); + hpsa_unregister_scsi(h); /* unhook from SCSI subsystem */ hpsa_shutdown(pdev); iounmap(h->vaddr); iounmap(h->transtable); iounmap(h->cfgtable); + hpsa_free_device_info(h); hpsa_free_sg_chain_blocks(h); pci_free_consistent(h->pdev, h->nr_cmds * sizeof(struct CommandList), @@ -4504,17 +7211,15 @@ static void __devexit hpsa_remove_one(struct pci_dev *pdev) pci_free_consistent(h->pdev, h->nr_cmds * sizeof(struct ErrorInfo), h->errinfo_pool, h->errinfo_pool_dhandle); - pci_free_consistent(h->pdev, h->reply_pool_size, - h->reply_pool, h->reply_pool_dhandle); + hpsa_free_reply_queues(h); kfree(h->cmd_pool_bits); kfree(h->blockFetchTable); + kfree(h->ioaccel1_blockFetchTable); + kfree(h->ioaccel2_blockFetchTable); kfree(h->hba_inquiry_data); - /* - * Deliberately omit pci_disable_device(): it does something nasty to - * Smart Array controllers that pci_enable_device does not undo - */ + pci_disable_device(pdev); pci_release_regions(pdev); - pci_set_drvdata(pdev, NULL); + free_percpu(h->lockup_detected); kfree(h); } @@ -4530,9 +7235,9 @@ static int hpsa_resume(__attribute__((unused)) struct pci_dev *pdev) } static struct pci_driver hpsa_pci_driver = { - .name = "hpsa", + .name = HPSA, .probe = hpsa_init_one, - .remove = __devexit_p(hpsa_remove_one), + .remove = hpsa_remove_one, .id_table = hpsa_pci_device_id, /* id_table */ .shutdown = hpsa_shutdown, .suspend = hpsa_suspend, @@ -4552,20 +7257,17 @@ static struct pci_driver hpsa_pci_driver = { * bits of the command address. */ static void calc_bucket_map(int bucket[], int num_buckets, - int nsgs, int *bucket_map) + int nsgs, int min_blocks, int *bucket_map) { int i, j, b, size; - /* even a command with 0 SGs requires 4 blocks */ -#define MINIMUM_TRANSFER_BLOCKS 4 -#define NUM_BUCKETS 8 /* Note, bucket_map must have nsgs+1 entries. */ for (i = 0; i <= nsgs; i++) { /* Compute size of a command with i SG entries */ - size = i + MINIMUM_TRANSFER_BLOCKS; + size = i + min_blocks; b = num_buckets; /* Assume the biggest bucket */ /* Find the bucket that is just big enough */ - for (j = 0; j < 8; j++) { + for (j = 0; j < num_buckets; j++) { if (bucket[j] >= size) { b = j; break; @@ -4576,11 +7278,16 @@ static void calc_bucket_map(int bucket[], int num_buckets, } } -static __devinit void hpsa_enter_performant_mode(struct ctlr_info *h, - u32 use_short_tags) +static void hpsa_enter_performant_mode(struct ctlr_info *h, u32 trans_support) { int i; unsigned long register_value; + unsigned long transMethod = CFGTBL_Trans_Performant | + (trans_support & CFGTBL_Trans_use_short_tags) | + CFGTBL_Trans_enable_directed_msix | + (trans_support & (CFGTBL_Trans_io_accel1 | + CFGTBL_Trans_io_accel2)); + struct access_method access = SA5_performant_access; /* This is a bit complicated. There are 8 registers on * the controller which we write to to tell it 8 different @@ -4592,41 +7299,76 @@ static __devinit void hpsa_enter_performant_mode(struct ctlr_info *h, * Each SG entry requires 16 bytes. The eight registers are programmed * with the number of 16-byte blocks a command of that size requires. * The smallest command possible requires 5 such 16 byte blocks. - * the largest command possible requires MAXSGENTRIES + 4 16-byte + * the largest command possible requires SG_ENTRIES_IN_CMD + 4 16-byte * blocks. Note, this only extends to the SG entries contained * within the command block, and does not extend to chained blocks * of SG elements. bft[] contains the eight values we write to * the registers. They are not evenly distributed, but have more * sizes for small commands, and fewer sizes for larger commands. */ - int bft[8] = {5, 6, 8, 10, 12, 20, 28, MAXSGENTRIES + 4}; - BUILD_BUG_ON(28 > MAXSGENTRIES + 4); + int bft[8] = {5, 6, 8, 10, 12, 20, 28, SG_ENTRIES_IN_CMD + 4}; +#define MIN_IOACCEL2_BFT_ENTRY 5 +#define HPSA_IOACCEL2_HEADER_SZ 4 + int bft2[16] = {MIN_IOACCEL2_BFT_ENTRY, 6, 7, 8, 9, 10, 11, 12, + 13, 14, 15, 16, 17, 18, 19, + HPSA_IOACCEL2_HEADER_SZ + IOACCEL2_MAXSGENTRIES}; + BUILD_BUG_ON(ARRAY_SIZE(bft2) != 16); + BUILD_BUG_ON(ARRAY_SIZE(bft) != 8); + BUILD_BUG_ON(offsetof(struct io_accel2_cmd, sg) > + 16 * MIN_IOACCEL2_BFT_ENTRY); + BUILD_BUG_ON(sizeof(struct ioaccel2_sg_element) != 16); + BUILD_BUG_ON(28 > SG_ENTRIES_IN_CMD + 4); /* 5 = 1 s/g entry or 4k * 6 = 2 s/g entry or 8k * 8 = 4 s/g entry or 16k * 10 = 6 s/g entry or 24k */ - h->reply_pool_wraparound = 1; /* spec: init to 1 */ + /* If the controller supports either ioaccel method then + * we can also use the RAID stack submit path that does not + * perform the superfluous readl() after each command submission. + */ + if (trans_support & (CFGTBL_Trans_io_accel1 | CFGTBL_Trans_io_accel2)) + access = SA5_performant_access_no_read; /* Controller spec: zero out this buffer. */ - memset(h->reply_pool, 0, h->reply_pool_size); - h->reply_pool_head = h->reply_pool; + for (i = 0; i < h->nreply_queues; i++) + memset(h->reply_queue[i].head, 0, h->reply_queue_size); - bft[7] = h->max_sg_entries + 4; - calc_bucket_map(bft, ARRAY_SIZE(bft), 32, h->blockFetchTable); + bft[7] = SG_ENTRIES_IN_CMD + 4; + calc_bucket_map(bft, ARRAY_SIZE(bft), + SG_ENTRIES_IN_CMD, 4, h->blockFetchTable); for (i = 0; i < 8; i++) writel(bft[i], &h->transtable->BlockFetch[i]); /* size of controller ring buffer */ writel(h->max_commands, &h->transtable->RepQSize); - writel(1, &h->transtable->RepQCount); + writel(h->nreply_queues, &h->transtable->RepQCount); writel(0, &h->transtable->RepQCtrAddrLow32); writel(0, &h->transtable->RepQCtrAddrHigh32); - writel(h->reply_pool_dhandle, &h->transtable->RepQAddr0Low32); - writel(0, &h->transtable->RepQAddr0High32); - writel(CFGTBL_Trans_Performant | use_short_tags, - &(h->cfgtable->HostWrite.TransportRequest)); + + for (i = 0; i < h->nreply_queues; i++) { + writel(0, &h->transtable->RepQAddr[i].upper); + writel(h->reply_queue[i].busaddr, + &h->transtable->RepQAddr[i].lower); + } + + writel(0, &h->cfgtable->HostWrite.command_pool_addr_hi); + writel(transMethod, &(h->cfgtable->HostWrite.TransportRequest)); + /* + * enable outbound interrupt coalescing in accelerator mode; + */ + if (trans_support & CFGTBL_Trans_io_accel1) { + access = SA5_ioaccel_mode1_access; + writel(10, &h->cfgtable->HostWrite.CoalIntDelay); + writel(4, &h->cfgtable->HostWrite.CoalIntCount); + } else { + if (trans_support & CFGTBL_Trans_io_accel2) { + access = SA5_ioaccel_mode2_access; + writel(10, &h->cfgtable->HostWrite.CoalIntDelay); + writel(4, &h->cfgtable->HostWrite.CoalIntCount); + } + } writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL); hpsa_wait_for_mode_change_ack(h); register_value = readl(&(h->cfgtable->TransportActive)); @@ -4636,13 +7378,161 @@ static __devinit void hpsa_enter_performant_mode(struct ctlr_info *h, return; } /* Change the access methods to the performant access methods */ - h->access = SA5_performant_access; - h->transMethod = CFGTBL_Trans_Performant; + h->access = access; + h->transMethod = transMethod; + + if (!((trans_support & CFGTBL_Trans_io_accel1) || + (trans_support & CFGTBL_Trans_io_accel2))) + return; + + if (trans_support & CFGTBL_Trans_io_accel1) { + /* Set up I/O accelerator mode */ + for (i = 0; i < h->nreply_queues; i++) { + writel(i, h->vaddr + IOACCEL_MODE1_REPLY_QUEUE_INDEX); + h->reply_queue[i].current_entry = + readl(h->vaddr + IOACCEL_MODE1_PRODUCER_INDEX); + } + bft[7] = h->ioaccel_maxsg + 8; + calc_bucket_map(bft, ARRAY_SIZE(bft), h->ioaccel_maxsg, 8, + h->ioaccel1_blockFetchTable); + + /* initialize all reply queue entries to unused */ + for (i = 0; i < h->nreply_queues; i++) + memset(h->reply_queue[i].head, + (u8) IOACCEL_MODE1_REPLY_UNUSED, + h->reply_queue_size); + + /* set all the constant fields in the accelerator command + * frames once at init time to save CPU cycles later. + */ + for (i = 0; i < h->nr_cmds; i++) { + struct io_accel1_cmd *cp = &h->ioaccel_cmd_pool[i]; + + cp->function = IOACCEL1_FUNCTION_SCSIIO; + cp->err_info = (u32) (h->errinfo_pool_dhandle + + (i * sizeof(struct ErrorInfo))); + cp->err_info_len = sizeof(struct ErrorInfo); + cp->sgl_offset = IOACCEL1_SGLOFFSET; + cp->host_context_flags = IOACCEL1_HCFLAGS_CISS_FORMAT; + cp->timeout_sec = 0; + cp->ReplyQueue = 0; + cp->Tag.lower = (i << DIRECT_LOOKUP_SHIFT) | + DIRECT_LOOKUP_BIT; + cp->Tag.upper = 0; + cp->host_addr.lower = + (u32) (h->ioaccel_cmd_pool_dhandle + + (i * sizeof(struct io_accel1_cmd))); + cp->host_addr.upper = 0; + } + } else if (trans_support & CFGTBL_Trans_io_accel2) { + u64 cfg_offset, cfg_base_addr_index; + u32 bft2_offset, cfg_base_addr; + int rc; + + rc = hpsa_find_cfg_addrs(h->pdev, h->vaddr, &cfg_base_addr, + &cfg_base_addr_index, &cfg_offset); + BUILD_BUG_ON(offsetof(struct io_accel2_cmd, sg) != 64); + bft2[15] = h->ioaccel_maxsg + HPSA_IOACCEL2_HEADER_SZ; + calc_bucket_map(bft2, ARRAY_SIZE(bft2), h->ioaccel_maxsg, + 4, h->ioaccel2_blockFetchTable); + bft2_offset = readl(&h->cfgtable->io_accel_request_size_offset); + BUILD_BUG_ON(offsetof(struct CfgTable, + io_accel_request_size_offset) != 0xb8); + h->ioaccel2_bft2_regs = + remap_pci_mem(pci_resource_start(h->pdev, + cfg_base_addr_index) + + cfg_offset + bft2_offset, + ARRAY_SIZE(bft2) * + sizeof(*h->ioaccel2_bft2_regs)); + for (i = 0; i < ARRAY_SIZE(bft2); i++) + writel(bft2[i], &h->ioaccel2_bft2_regs[i]); + } + writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL); + hpsa_wait_for_mode_change_ack(h); +} + +static int hpsa_alloc_ioaccel_cmd_and_bft(struct ctlr_info *h) +{ + h->ioaccel_maxsg = + readl(&(h->cfgtable->io_accel_max_embedded_sg_count)); + if (h->ioaccel_maxsg > IOACCEL1_MAXSGENTRIES) + h->ioaccel_maxsg = IOACCEL1_MAXSGENTRIES; + + /* Command structures must be aligned on a 128-byte boundary + * because the 7 lower bits of the address are used by the + * hardware. + */ + BUILD_BUG_ON(sizeof(struct io_accel1_cmd) % + IOACCEL1_COMMANDLIST_ALIGNMENT); + h->ioaccel_cmd_pool = + pci_alloc_consistent(h->pdev, + h->nr_cmds * sizeof(*h->ioaccel_cmd_pool), + &(h->ioaccel_cmd_pool_dhandle)); + + h->ioaccel1_blockFetchTable = + kmalloc(((h->ioaccel_maxsg + 1) * + sizeof(u32)), GFP_KERNEL); + + if ((h->ioaccel_cmd_pool == NULL) || + (h->ioaccel1_blockFetchTable == NULL)) + goto clean_up; + + memset(h->ioaccel_cmd_pool, 0, + h->nr_cmds * sizeof(*h->ioaccel_cmd_pool)); + return 0; + +clean_up: + if (h->ioaccel_cmd_pool) + pci_free_consistent(h->pdev, + h->nr_cmds * sizeof(*h->ioaccel_cmd_pool), + h->ioaccel_cmd_pool, h->ioaccel_cmd_pool_dhandle); + kfree(h->ioaccel1_blockFetchTable); + return 1; } -static __devinit void hpsa_put_ctlr_into_performant_mode(struct ctlr_info *h) +static int ioaccel2_alloc_cmds_and_bft(struct ctlr_info *h) +{ + /* Allocate ioaccel2 mode command blocks and block fetch table */ + + h->ioaccel_maxsg = + readl(&(h->cfgtable->io_accel_max_embedded_sg_count)); + if (h->ioaccel_maxsg > IOACCEL2_MAXSGENTRIES) + h->ioaccel_maxsg = IOACCEL2_MAXSGENTRIES; + + BUILD_BUG_ON(sizeof(struct io_accel2_cmd) % + IOACCEL2_COMMANDLIST_ALIGNMENT); + h->ioaccel2_cmd_pool = + pci_alloc_consistent(h->pdev, + h->nr_cmds * sizeof(*h->ioaccel2_cmd_pool), + &(h->ioaccel2_cmd_pool_dhandle)); + + h->ioaccel2_blockFetchTable = + kmalloc(((h->ioaccel_maxsg + 1) * + sizeof(u32)), GFP_KERNEL); + + if ((h->ioaccel2_cmd_pool == NULL) || + (h->ioaccel2_blockFetchTable == NULL)) + goto clean_up; + + memset(h->ioaccel2_cmd_pool, 0, + h->nr_cmds * sizeof(*h->ioaccel2_cmd_pool)); + return 0; + +clean_up: + if (h->ioaccel2_cmd_pool) + pci_free_consistent(h->pdev, + h->nr_cmds * sizeof(*h->ioaccel2_cmd_pool), + h->ioaccel2_cmd_pool, h->ioaccel2_cmd_pool_dhandle); + kfree(h->ioaccel2_blockFetchTable); + return 1; +} + +static void hpsa_put_ctlr_into_performant_mode(struct ctlr_info *h) { u32 trans_support; + unsigned long transMethod = CFGTBL_Trans_Performant | + CFGTBL_Trans_use_short_tags; + int i; if (hpsa_simple_mode) return; @@ -4651,33 +7541,76 @@ static __devinit void hpsa_put_ctlr_into_performant_mode(struct ctlr_info *h) if (!(trans_support & PERFORMANT_MODE)) return; + /* Check for I/O accelerator mode support */ + if (trans_support & CFGTBL_Trans_io_accel1) { + transMethod |= CFGTBL_Trans_io_accel1 | + CFGTBL_Trans_enable_directed_msix; + if (hpsa_alloc_ioaccel_cmd_and_bft(h)) + goto clean_up; + } else { + if (trans_support & CFGTBL_Trans_io_accel2) { + transMethod |= CFGTBL_Trans_io_accel2 | + CFGTBL_Trans_enable_directed_msix; + if (ioaccel2_alloc_cmds_and_bft(h)) + goto clean_up; + } + } + + h->nreply_queues = h->msix_vector > 0 ? h->msix_vector : 1; hpsa_get_max_perf_mode_cmds(h); - h->max_sg_entries = 32; /* Performant mode ring buffer and supporting data structures */ - h->reply_pool_size = h->max_commands * sizeof(u64); - h->reply_pool = pci_alloc_consistent(h->pdev, h->reply_pool_size, - &(h->reply_pool_dhandle)); + h->reply_queue_size = h->max_commands * sizeof(u64); + + for (i = 0; i < h->nreply_queues; i++) { + h->reply_queue[i].head = pci_alloc_consistent(h->pdev, + h->reply_queue_size, + &(h->reply_queue[i].busaddr)); + if (!h->reply_queue[i].head) + goto clean_up; + h->reply_queue[i].size = h->max_commands; + h->reply_queue[i].wraparound = 1; /* spec: init to 1 */ + h->reply_queue[i].current_entry = 0; + } /* Need a block fetch table for performant mode */ - h->blockFetchTable = kmalloc(((h->max_sg_entries+1) * + h->blockFetchTable = kmalloc(((SG_ENTRIES_IN_CMD + 1) * sizeof(u32)), GFP_KERNEL); - - if ((h->reply_pool == NULL) - || (h->blockFetchTable == NULL)) + if (!h->blockFetchTable) goto clean_up; - hpsa_enter_performant_mode(h, - trans_support & CFGTBL_Trans_use_short_tags); - + hpsa_enter_performant_mode(h, trans_support); return; clean_up: - if (h->reply_pool) - pci_free_consistent(h->pdev, h->reply_pool_size, - h->reply_pool, h->reply_pool_dhandle); + hpsa_free_reply_queues(h); kfree(h->blockFetchTable); } +static int is_accelerated_cmd(struct CommandList *c) +{ + return c->cmd_type == CMD_IOACCEL1 || c->cmd_type == CMD_IOACCEL2; +} + +static void hpsa_drain_accel_commands(struct ctlr_info *h) +{ + struct CommandList *c = NULL; + unsigned long flags; + int accel_cmds_out; + + do { /* wait for all outstanding commands to drain out */ + accel_cmds_out = 0; + spin_lock_irqsave(&h->lock, flags); + list_for_each_entry(c, &h->cmpQ, list) + accel_cmds_out += is_accelerated_cmd(c); + list_for_each_entry(c, &h->reqQ, list) + accel_cmds_out += is_accelerated_cmd(c); + spin_unlock_irqrestore(&h->lock, flags); + if (accel_cmds_out <= 0) + break; + msleep(100); + } while (1); +} + /* * This is it. Register the PCI driver information for the cards we control * the OS will call our registered routines when it finds one of our cards. @@ -4692,5 +7625,83 @@ static void __exit hpsa_cleanup(void) pci_unregister_driver(&hpsa_pci_driver); } +static void __attribute__((unused)) verify_offsets(void) +{ +#define VERIFY_OFFSET(member, offset) \ + BUILD_BUG_ON(offsetof(struct raid_map_data, member) != offset) + + VERIFY_OFFSET(structure_size, 0); + VERIFY_OFFSET(volume_blk_size, 4); + VERIFY_OFFSET(volume_blk_cnt, 8); + VERIFY_OFFSET(phys_blk_shift, 16); + VERIFY_OFFSET(parity_rotation_shift, 17); + VERIFY_OFFSET(strip_size, 18); + VERIFY_OFFSET(disk_starting_blk, 20); + VERIFY_OFFSET(disk_blk_cnt, 28); + VERIFY_OFFSET(data_disks_per_row, 36); + VERIFY_OFFSET(metadata_disks_per_row, 38); + VERIFY_OFFSET(row_cnt, 40); + VERIFY_OFFSET(layout_map_count, 42); + VERIFY_OFFSET(flags, 44); + VERIFY_OFFSET(dekindex, 46); + /* VERIFY_OFFSET(reserved, 48 */ + VERIFY_OFFSET(data, 64); + +#undef VERIFY_OFFSET + +#define VERIFY_OFFSET(member, offset) \ + BUILD_BUG_ON(offsetof(struct io_accel2_cmd, member) != offset) + + VERIFY_OFFSET(IU_type, 0); + VERIFY_OFFSET(direction, 1); + VERIFY_OFFSET(reply_queue, 2); + /* VERIFY_OFFSET(reserved1, 3); */ + VERIFY_OFFSET(scsi_nexus, 4); + VERIFY_OFFSET(Tag, 8); + VERIFY_OFFSET(cdb, 16); + VERIFY_OFFSET(cciss_lun, 32); + VERIFY_OFFSET(data_len, 40); + VERIFY_OFFSET(cmd_priority_task_attr, 44); + VERIFY_OFFSET(sg_count, 45); + /* VERIFY_OFFSET(reserved3 */ + VERIFY_OFFSET(err_ptr, 48); + VERIFY_OFFSET(err_len, 56); + /* VERIFY_OFFSET(reserved4 */ + VERIFY_OFFSET(sg, 64); + +#undef VERIFY_OFFSET + +#define VERIFY_OFFSET(member, offset) \ + BUILD_BUG_ON(offsetof(struct io_accel1_cmd, member) != offset) + + VERIFY_OFFSET(dev_handle, 0x00); + VERIFY_OFFSET(reserved1, 0x02); + VERIFY_OFFSET(function, 0x03); + VERIFY_OFFSET(reserved2, 0x04); + VERIFY_OFFSET(err_info, 0x0C); + VERIFY_OFFSET(reserved3, 0x10); + VERIFY_OFFSET(err_info_len, 0x12); + VERIFY_OFFSET(reserved4, 0x13); + VERIFY_OFFSET(sgl_offset, 0x14); + VERIFY_OFFSET(reserved5, 0x15); + VERIFY_OFFSET(transfer_len, 0x1C); + VERIFY_OFFSET(reserved6, 0x20); + VERIFY_OFFSET(io_flags, 0x24); + VERIFY_OFFSET(reserved7, 0x26); + VERIFY_OFFSET(LUN, 0x34); + VERIFY_OFFSET(control, 0x3C); + VERIFY_OFFSET(CDB, 0x40); + VERIFY_OFFSET(reserved8, 0x50); + VERIFY_OFFSET(host_context_flags, 0x60); + VERIFY_OFFSET(timeout_sec, 0x62); + VERIFY_OFFSET(ReplyQueue, 0x64); + VERIFY_OFFSET(reserved9, 0x65); + VERIFY_OFFSET(Tag, 0x68); + VERIFY_OFFSET(host_addr, 0x70); + VERIFY_OFFSET(CISS_LUN, 0x78); + VERIFY_OFFSET(SG, 0x78 + 8); +#undef VERIFY_OFFSET +} + module_init(hpsa_init); module_exit(hpsa_cleanup); |