Signed-off-by: Bradley Grove <bgrove@xxxxxxxxxxxx> --- drivers/scsi/esas2r/esas2r_main.c | 2032 +++++++++++++++++++++++++++++++++++++ 1 file changed, 2032 insertions(+) create mode 100644 drivers/scsi/esas2r/esas2r_main.c diff --git a/drivers/scsi/esas2r/esas2r_main.c b/drivers/scsi/esas2r/esas2r_main.c new file mode 100644 index 0000000..4abf127 --- /dev/null +++ b/drivers/scsi/esas2r/esas2r_main.c @@ -0,0 +1,2032 @@ +/* + * linux/drivers/scsi/esas2r/esas2r_main.c + * For use with ATTO ExpressSAS R6xx SAS/SATA RAID controllers + * + * Copyright (c) 2001-2013 ATTO Technology, Inc. + * (mailto:linuxdrivers@xxxxxxxxxxxx) + * + * 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 the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * NO WARRANTY + * THE PROGRAM IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OR + * CONDITIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED INCLUDING, WITHOUT + * LIMITATION, ANY WARRANTIES OR CONDITIONS OF TITLE, NON-INFRINGEMENT, + * MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Each Recipient is + * solely responsible for determining the appropriateness of using and + * distributing the Program and assumes all risks associated with its + * exercise of rights under this Agreement, including but not limited to + * the risks and costs of program errors, damage to or loss of data, + * programs or equipment, and unavailability or interruption of operations. + * + * DISCLAIMER OF LIABILITY + * NEITHER RECIPIENT NOR ANY CONTRIBUTORS SHALL HAVE ANY LIABILITY FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING WITHOUT LIMITATION LOST PROFITS), HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR + * TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE + * USE OR DISTRIBUTION OF THE PROGRAM OR THE EXERCISE OF ANY RIGHTS GRANTED + * HEREUNDER, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGES + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include "esas2r.h" + +MODULE_DESCRIPTION(ESAS2R_DRVR_NAME ": " ESAS2R_LONGNAME " driver"); +MODULE_AUTHOR("ATTO Technology, Inc."); +MODULE_LICENSE("GPL"); +MODULE_VERSION(ESAS2R_VERSION_STR); + +/* global definitions */ + +static int found_adapters; +struct esas2r_adapter *esas2r_adapters[MAX_ADAPTERS]; + +#define ESAS2R_VDA_EVENT_PORT1 54414 +#define ESAS2R_VDA_EVENT_PORT2 54415 +#define ESAS2R_VDA_EVENT_SOCK_COUNT 2 + +static struct esas2r_adapter *esas2r_adapter_from_kobj(struct kobject *kobj) +{ + struct device *dev = container_of(kobj, struct device, kobj); + struct Scsi_Host *host = class_to_shost(dev); + + return (struct esas2r_adapter *)host->hostdata; +} + +static ssize_t read_fw(struct file *file, struct kobject *kobj, + struct bin_attribute *attr, + char *buf, loff_t off, size_t count) +{ + struct esas2r_adapter *a = esas2r_adapter_from_kobj(kobj); + + return esas2r_read_fw(a, buf, off, count); +} + +static ssize_t write_fw(struct file *file, struct kobject *kobj, + struct bin_attribute *attr, + char *buf, loff_t off, size_t count) +{ + struct esas2r_adapter *a = esas2r_adapter_from_kobj(kobj); + + return esas2r_write_fw(a, buf, off, count); +} + +static ssize_t read_fs(struct file *file, struct kobject *kobj, + struct bin_attribute *attr, + char *buf, loff_t off, size_t count) +{ + struct esas2r_adapter *a = esas2r_adapter_from_kobj(kobj); + + return esas2r_read_fs(a, buf, off, count); +} + +static ssize_t write_fs(struct file *file, struct kobject *kobj, + struct bin_attribute *attr, + char *buf, loff_t off, size_t count) +{ + struct esas2r_adapter *a = esas2r_adapter_from_kobj(kobj); + int length = min(sizeof(struct esas2r_ioctl_fs), count); + int result = 0; + + result = esas2r_write_fs(a, buf, off, count); + + if (result < 0) + result = 0; + + return length; +} + +static ssize_t read_vda(struct file *file, struct kobject *kobj, + struct bin_attribute *attr, + char *buf, loff_t off, size_t count) +{ + struct esas2r_adapter *a = esas2r_adapter_from_kobj(kobj); + + return esas2r_read_vda(a, buf, off, count); +} + +static ssize_t write_vda(struct file *file, struct kobject *kobj, + struct bin_attribute *attr, + char *buf, loff_t off, size_t count) +{ + struct esas2r_adapter *a = esas2r_adapter_from_kobj(kobj); + + return esas2r_write_vda(a, buf, off, count); +} + +static ssize_t read_live_nvram(struct file *file, struct kobject *kobj, + struct bin_attribute *attr, + char *buf, loff_t off, size_t count) +{ + struct esas2r_adapter *a = esas2r_adapter_from_kobj(kobj); + int length = min_t(size_t, sizeof(struct esas2r_sas_nvram), PAGE_SIZE); + + memcpy(buf, a->nvram, length); + return length; +} + +static ssize_t write_live_nvram(struct file *file, struct kobject *kobj, + struct bin_attribute *attr, + char *buf, loff_t off, size_t count) +{ + struct esas2r_adapter *a = esas2r_adapter_from_kobj(kobj); + struct esas2r_request *rq; + int result = -EFAULT; + + rq = esas2r_alloc_request(a); + if (rq == NULL) + return -ENOMEM; + + if (esas2r_write_params(a, rq, (struct esas2r_sas_nvram *)buf)) + result = count; + + esas2r_free_request(a, rq); + + return result; +} + +static ssize_t read_default_nvram(struct file *file, struct kobject *kobj, + struct bin_attribute *attr, + char *buf, loff_t off, size_t count) +{ + struct esas2r_adapter *a = esas2r_adapter_from_kobj(kobj); + + esas2r_nvram_get_defaults(a, (struct esas2r_sas_nvram *)buf); + + return sizeof(struct esas2r_sas_nvram); +} + +static ssize_t read_hw(struct file *file, struct kobject *kobj, + struct bin_attribute *attr, + char *buf, loff_t off, size_t count) +{ + struct esas2r_adapter *a = esas2r_adapter_from_kobj(kobj); + int length = min_t(size_t, sizeof(struct atto_ioctl), PAGE_SIZE); + + if (!a->local_atto_ioctl) + return -ENOMEM; + + if (handle_hba_ioctl(a, a->local_atto_ioctl) != IOCTL_SUCCESS) + return -ENOMEM; + + memcpy(buf, a->local_atto_ioctl, length); + + return length; +} + +static ssize_t write_hw(struct file *file, struct kobject *kobj, + struct bin_attribute *attr, + char *buf, loff_t off, size_t count) +{ + struct esas2r_adapter *a = esas2r_adapter_from_kobj(kobj); + int length = min(sizeof(struct atto_ioctl), count); + + if (!a->local_atto_ioctl) { + a->local_atto_ioctl = kzalloc(sizeof(struct atto_ioctl), + GFP_KERNEL); + if (a->local_atto_ioctl == NULL) { + esas2r_log(ESAS2R_LOG_WARN, + "write_hw kzalloc failed for %d bytes", + sizeof(struct atto_ioctl)); + return -ENOMEM; + } + } + + memset(a->local_atto_ioctl, 0, sizeof(struct atto_ioctl)); + memcpy(a->local_atto_ioctl, buf, length); + + return length; +} + +#define ESAS2R_RW_BIN_ATTR(_name) \ + struct bin_attribute bin_attr_ ## _name = { \ + .attr = \ + { .name = __stringify(_name), .mode = S_IRUSR | S_IWUSR }, \ + .size = 0, \ + .read = read_ ## _name, \ + .write = write_ ## _name } + +ESAS2R_RW_BIN_ATTR(fw); +ESAS2R_RW_BIN_ATTR(fs); +ESAS2R_RW_BIN_ATTR(vda); +ESAS2R_RW_BIN_ATTR(hw); +ESAS2R_RW_BIN_ATTR(live_nvram); + +struct bin_attribute bin_attr_default_nvram = { + .attr = { .name = "default_nvram", .mode = S_IRUGO }, + .size = 0, + .read = read_default_nvram, + .write = NULL +}; + +static struct scsi_host_template driver_template = { + .module = THIS_MODULE, + .show_info = esas2r_show_info, + .name = ESAS2R_LONGNAME, + .release = esas2r_release, + .info = esas2r_info, + .ioctl = esas2r_ioctl, + .queuecommand = esas2r_queuecommand, + .eh_abort_handler = esas2r_eh_abort, + .eh_device_reset_handler = esas2r_device_reset, + .eh_bus_reset_handler = esas2r_bus_reset, + .eh_host_reset_handler = esas2r_host_reset, + .eh_target_reset_handler = esas2r_target_reset, + .can_queue = 128, + .this_id = -1, + .sg_tablesize = SCSI_MAX_SG_SEGMENTS, + .cmd_per_lun = + ESAS2R_DEFAULT_CMD_PER_LUN, + .present = 0, + .unchecked_isa_dma = 0, + .use_clustering = ENABLE_CLUSTERING, + .emulated = 0, + .proc_name = ESAS2R_DRVR_NAME, + .slave_configure = esas2r_slave_configure, + .slave_alloc = esas2r_slave_alloc, + .slave_destroy = esas2r_slave_destroy, + .change_queue_depth = esas2r_change_queue_depth, + .change_queue_type = esas2r_change_queue_type, + .max_sectors = 0xFFFF, +}; + +int sgl_page_size = 512; +module_param(sgl_page_size, int, 0); +MODULE_PARM_DESC(sgl_page_size, + "Scatter/gather list (SGL) page size in number of S/G " + "entries. If your application is doing a lot of very large " + "transfers, you may want to increase the SGL page size. " + "Default 512."); + +int num_sg_lists = 1024; +module_param(num_sg_lists, int, 0); +MODULE_PARM_DESC(num_sg_lists, + "Number of scatter/gather lists. Default 1024."); + +int sg_tablesize = SCSI_MAX_SG_SEGMENTS; +module_param(sg_tablesize, int, 0); +MODULE_PARM_DESC(sg_tablesize, + "Maximum number of entries in a scatter/gather table."); + +int num_requests = 256; +module_param(num_requests, int, 0); +MODULE_PARM_DESC(num_requests, + "Number of requests. Default 256."); + +int num_ae_requests = 4; +module_param(num_ae_requests, int, 0); +MODULE_PARM_DESC(num_ae_requests, + "Number of VDA asynchromous event requests. Default 4."); + +int cmd_per_lun = ESAS2R_DEFAULT_CMD_PER_LUN; +module_param(cmd_per_lun, int, 0); +MODULE_PARM_DESC(cmd_per_lun, + "Maximum number of commands per LUN. Default " + DEFINED_NUM_TO_STR(ESAS2R_DEFAULT_CMD_PER_LUN) "."); + +int can_queue = 128; +module_param(can_queue, int, 0); +MODULE_PARM_DESC(can_queue, + "Maximum number of commands per adapter. Default 128."); + +int esas2r_max_sectors = 0xFFFF; +module_param(esas2r_max_sectors, int, 0); +MODULE_PARM_DESC(esas2r_max_sectors, + "Maximum number of disk sectors in a single data transfer. " + "Default 65535 (largest possible setting)."); + +int interrupt_mode = 1; +module_param(interrupt_mode, int, 0); +MODULE_PARM_DESC(interrupt_mode, + "Defines the interrupt mode to use. 0 for legacy" + ", 1 for MSI. Default is MSI (1)."); + +static struct pci_device_id + esas2r_pci_table[] = { + { ATTO_VENDOR_ID, 0x0049, ATTO_VENDOR_ID, 0x0049, + 0, + 0, 0 }, + { ATTO_VENDOR_ID, 0x0049, ATTO_VENDOR_ID, 0x004A, + 0, + 0, 0 }, + { ATTO_VENDOR_ID, 0x0049, ATTO_VENDOR_ID, 0x004B, + 0, + 0, 0 }, + { ATTO_VENDOR_ID, 0x0049, ATTO_VENDOR_ID, 0x004C, + 0, + 0, 0 }, + { ATTO_VENDOR_ID, 0x0049, ATTO_VENDOR_ID, 0x004D, + 0, + 0, 0 }, + { ATTO_VENDOR_ID, 0x0049, ATTO_VENDOR_ID, 0x004E, + 0, + 0, 0 }, + { 0, 0, 0, 0, + 0, + 0, 0 } +}; + +MODULE_DEVICE_TABLE(pci, esas2r_pci_table); + +static int +esas2r_probe(struct pci_dev *pcid, const struct pci_device_id *id); + +static void +esas2r_remove(struct pci_dev *pcid); + +static struct pci_driver + esas2r_pci_driver = { + .name = ESAS2R_DRVR_NAME, + .id_table = esas2r_pci_table, + .probe = esas2r_probe, + .remove = esas2r_remove, + .suspend = esas2r_suspend, + .resume = esas2r_resume, +}; + +static int esas2r_probe(struct pci_dev *pcid, + const struct pci_device_id *id) +{ + struct Scsi_Host *host = NULL; + struct esas2r_adapter *a; + int err; + + size_t host_alloc_size = sizeof(struct esas2r_adapter) + + ((num_requests) + + 1) * sizeof(struct esas2r_request); + + esas2r_log_dev(ESAS2R_LOG_DEBG, &(pcid->dev), + "esas2r_probe() 0x%02x 0x%02x 0x%02x 0x%02x", + pcid->vendor, + pcid->device, + pcid->subsystem_vendor, + pcid->subsystem_device); + + esas2r_log_dev(ESAS2R_LOG_INFO, &(pcid->dev), + "before pci_enable_device() " + "enable_cnt: %d", + pcid->enable_cnt.counter); + + err = pci_enable_device(pcid); + if (err != 0) { + esas2r_log_dev(ESAS2R_LOG_CRIT, &(pcid->dev), + "pci_enable_device() FAIL (%d)", + err); + return -ENODEV; + } + + esas2r_log_dev(ESAS2R_LOG_INFO, &(pcid->dev), + "pci_enable_device() OK"); + esas2r_log_dev(ESAS2R_LOG_INFO, &(pcid->dev), + "after pci_device_enable() enable_cnt: %d", + pcid->enable_cnt.counter); + + host = scsi_host_alloc(&driver_template, host_alloc_size); + if (host == NULL) { + esas2r_log(ESAS2R_LOG_CRIT, "scsi_host_alloc() FAIL"); + return -ENODEV; + } + + memset(host->hostdata, 0, host_alloc_size); + + a = (struct esas2r_adapter *)host->hostdata; + + esas2r_log(ESAS2R_LOG_INFO, "scsi_host_alloc() OK host: %p", host); + + /* override max LUN and max target id */ + + host->max_id = ESAS2R_MAX_ID + 1; + host->max_lun = 255; + + /* we can handle 16-byte CDbs */ + + host->max_cmd_len = 16; + + host->can_queue = can_queue; + host->cmd_per_lun = cmd_per_lun; + host->this_id = host->max_id + 1; + host->max_channel = 0; + host->unique_id = found_adapters; + host->sg_tablesize = sg_tablesize; + host->max_sectors = esas2r_max_sectors; + + /* set to bus master for BIOses that don't do it for us */ + + esas2r_log(ESAS2R_LOG_INFO, "pci_set_master() called"); + + pci_set_master(pcid); + + if (!esas2r_init_adapter(host, pcid, found_adapters)) { + esas2r_log(ESAS2R_LOG_CRIT, + "unable to initialize device at PCI bus %x:%x", + pcid->bus->number, + pcid->devfn); + + esas2r_log_dev(ESAS2R_LOG_INFO, &(host->shost_gendev), + "scsi_host_put() called"); + + scsi_host_put(host); + + return 0; + + } + + esas2r_log(ESAS2R_LOG_INFO, "pci_set_drvdata(%p, %p) called", pcid, + host->hostdata); + + pci_set_drvdata(pcid, host); + + esas2r_log(ESAS2R_LOG_INFO, "scsi_add_host() called"); + + err = scsi_add_host(host, &pcid->dev); + + if (err) { + esas2r_log(ESAS2R_LOG_CRIT, "scsi_add_host returned %d", err); + esas2r_log_dev(ESAS2R_LOG_CRIT, &(host->shost_gendev), + "scsi_add_host() FAIL"); + + esas2r_log_dev(ESAS2R_LOG_INFO, &(host->shost_gendev), + "scsi_host_put() called"); + + scsi_host_put(host); + + esas2r_log_dev(ESAS2R_LOG_INFO, &(host->shost_gendev), + "pci_set_drvdata(%p, NULL) called", + pcid); + + pci_set_drvdata(pcid, NULL); + + return -ENODEV; + } + + + esas2r_fw_event_on(a); + + esas2r_log_dev(ESAS2R_LOG_INFO, &(host->shost_gendev), + "scsi_scan_host() called"); + + scsi_scan_host(host); + + /* Add sysfs binary files */ + if (sysfs_create_bin_file(&host->shost_dev.kobj, &bin_attr_fw)) + esas2r_log_dev(ESAS2R_LOG_WARN, &(host->shost_gendev), + "Failed to create sysfs binary file: fw"); + else + a->sysfs_fw_created = 1; + + if (sysfs_create_bin_file(&host->shost_dev.kobj, &bin_attr_fs)) + esas2r_log_dev(ESAS2R_LOG_WARN, &(host->shost_gendev), + "Failed to create sysfs binary file: fs"); + else + a->sysfs_fs_created = 1; + + if (sysfs_create_bin_file(&host->shost_dev.kobj, &bin_attr_vda)) + esas2r_log_dev(ESAS2R_LOG_WARN, &(host->shost_gendev), + "Failed to create sysfs binary file: vda"); + else + a->sysfs_vda_created = 1; + + if (sysfs_create_bin_file(&host->shost_dev.kobj, &bin_attr_hw)) + esas2r_log_dev(ESAS2R_LOG_WARN, &(host->shost_gendev), + "Failed to create sysfs binary file: hw"); + else + a->sysfs_hw_created = 1; + + if (sysfs_create_bin_file(&host->shost_dev.kobj, &bin_attr_live_nvram)) + esas2r_log_dev(ESAS2R_LOG_WARN, &(host->shost_gendev), + "Failed to create sysfs binary file: live_nvram"); + else + a->sysfs_live_nvram_created = 1; + + if (sysfs_create_bin_file(&host->shost_dev.kobj, + &bin_attr_default_nvram)) + esas2r_log_dev(ESAS2R_LOG_WARN, &(host->shost_gendev), + "Failed to create sysfs binary file: default_nvram"); + else + a->sysfs_default_nvram_created = 1; + + found_adapters++; + + return 0; +} + +static void esas2r_remove(struct pci_dev *pdev) +{ + struct Scsi_Host *host; + int index; + + if (pdev == NULL) { + esas2r_log(ESAS2R_LOG_WARN, "esas2r_remove pdev==NULL"); + return; + } + + host = pci_get_drvdata(pdev); + + if (host == NULL) { + /* + * this can happen if pci_set_drvdata was already called + * to clear the host pointer. if this is the case, we + * are okay; this channel has already been cleaned up. + */ + + return; + } + + esas2r_log_dev(ESAS2R_LOG_INFO, &(pdev->dev), + "esas2r_remove(%p) called; " + "host:%p", pdev, + host); + + index = esas2r_cleanup(host); + + if (index < 0) + esas2r_log_dev(ESAS2R_LOG_WARN, &(pdev->dev), + "unknown host in %s", + __func__); + + found_adapters--; + + /* if this was the last adapter, clean up the rest of the driver */ + + if (found_adapters == 0) + esas2r_cleanup(NULL); +} + +static int __init esas2r_init(void) +{ + int i; + + esas2r_log(ESAS2R_LOG_INFO, "%s called", __func__); + + /* verify valid parameters */ + + if (can_queue < 1) { + esas2r_log(ESAS2R_LOG_WARN, + "warning: can_queue must be at least 1, value " + "forced."); + can_queue = 1; + } else if (can_queue > 2048) { + esas2r_log(ESAS2R_LOG_WARN, + "warning: can_queue must be no larger than 2048, " + "value forced."); + can_queue = 2048; + } + + if (cmd_per_lun < 1) { + esas2r_log(ESAS2R_LOG_WARN, + "warning: cmd_per_lun must be at least 1, value " + "forced."); + cmd_per_lun = 1; + } else if (cmd_per_lun > 2048) { + esas2r_log(ESAS2R_LOG_WARN, + "warning: cmd_per_lun must be no larger than " + "2048, value forced."); + cmd_per_lun = 2048; + } + + if (sg_tablesize < 32) { + esas2r_log(ESAS2R_LOG_WARN, + "warning: sg_tablesize must be at least 32, " + "value forced."); + sg_tablesize = 32; + } + + if (esas2r_max_sectors < 1) { + esas2r_log(ESAS2R_LOG_WARN, + "warning: esas2r_max_sectors must be at least " + "1, value forced."); + esas2r_max_sectors = 1; + } else if (esas2r_max_sectors > 0xffff) { + esas2r_log(ESAS2R_LOG_WARN, + "warning: esas2r_max_sectors must be no larger " + "than 0xffff, value forced."); + esas2r_max_sectors = 0xffff; + } + + sgl_page_size &= ~(ESAS2R_SGL_ALIGN - 1); + + if (sgl_page_size < SGL_PG_SZ_MIN) + sgl_page_size = SGL_PG_SZ_MIN; + else if (sgl_page_size > SGL_PG_SZ_MAX) + sgl_page_size = SGL_PG_SZ_MAX; + + if (num_sg_lists < NUM_SGL_MIN) + num_sg_lists = NUM_SGL_MIN; + else if (num_sg_lists > NUM_SGL_MAX) + num_sg_lists = NUM_SGL_MAX; + + if (num_requests < NUM_REQ_MIN) + num_requests = NUM_REQ_MIN; + else if (num_requests > NUM_REQ_MAX) + num_requests = NUM_REQ_MAX; + + if (num_ae_requests < NUM_AE_MIN) + num_ae_requests = NUM_AE_MIN; + else if (num_ae_requests > NUM_AE_MAX) + num_ae_requests = NUM_AE_MAX; + + /* set up other globals */ + + for (i = 0; i < MAX_ADAPTERS; i++) + esas2r_adapters[i] = NULL; + + /* initialize */ + + driver_template.module = THIS_MODULE; + + if (pci_register_driver(&esas2r_pci_driver) != 0) + esas2r_log(ESAS2R_LOG_CRIT, "pci_register_driver FAILED"); + else + esas2r_log(ESAS2R_LOG_INFO, "pci_register_driver() OK"); + + if (!found_adapters) { + pci_unregister_driver(&esas2r_pci_driver); + esas2r_cleanup(NULL); + + esas2r_log(ESAS2R_LOG_CRIT, + "driver will not be loaded because no ATTO " + "%s devices were found", + ESAS2R_DRVR_NAME); + return -1; + } else { + esas2r_log(ESAS2R_LOG_INFO, "found %d adapters", + found_adapters); + } + + return 0; +} + +/* Handle ioctl calls to "/proc/scsi/esas2r/ATTOnode" */ +static const struct file_operations esas2r_proc_fops = { + .compat_ioctl = esas2r_proc_ioctl, + .unlocked_ioctl = esas2r_proc_ioctl, +}; + +static struct Scsi_Host *esas2r_proc_host; +static int esas2r_proc_major; + +long esas2r_proc_ioctl(struct file *fp, unsigned int cmd, unsigned long arg) +{ + return esas2r_ioctl_handler(esas2r_proc_host->hostdata, + (int)cmd, (void __user *)arg); +} + +static void __exit esas2r_exit(void) +{ + esas2r_log(ESAS2R_LOG_INFO, "%s called", __func__); + + if (esas2r_proc_major > 0) { + esas2r_log(ESAS2R_LOG_INFO, "unregister proc"); + + remove_proc_entry(ATTONODE_NAME, + esas2r_proc_host->hostt->proc_dir); + unregister_chrdev(esas2r_proc_major, ESAS2R_DRVR_NAME); + + esas2r_proc_major = 0; + } + + esas2r_log(ESAS2R_LOG_INFO, "pci_unregister_driver() called"); + + pci_unregister_driver(&esas2r_pci_driver); +} + +int esas2r_show_info(struct seq_file *m, struct Scsi_Host *sh) +{ + struct esas2r_adapter *a = (struct esas2r_adapter *)sh->hostdata; + + struct esas2r_target *t; + int dev_count = 0; + + esas2r_log(ESAS2R_LOG_DEBG, "esas2r_show_info (%p,%d)", m, sh->host_no); + + seq_printf(m, ESAS2R_LONGNAME "\n" + "Driver version: "ESAS2R_VERSION_STR "\n" + "Flash version: %s\n" + "Firmware version: %s\n" + "Copyright "ESAS2R_COPYRIGHT_YEARS "\n" + "http://www.attotech.com\n" + "\n", + a->flash_rev, + a->fw_rev[0] ? a->fw_rev : "(none)"); + + + seq_printf(m, "Adapter information:\n" + "--------------------\n" + "Model: %s\n" + "SAS address: %02X%02X%02X%02X:%02X%02X%02X%02X\n", + esas2r_get_model_name(a), + a->nvram->sas_addr[0], + a->nvram->sas_addr[1], + a->nvram->sas_addr[2], + a->nvram->sas_addr[3], + a->nvram->sas_addr[4], + a->nvram->sas_addr[5], + a->nvram->sas_addr[6], + a->nvram->sas_addr[7]); + + seq_puts(m, "\n" + "Discovered devices:\n" + "\n" + " # Target ID\n" + "---------------\n"); + + for (t = a->targetdb; t < a->targetdb_end; t++) + if (t->buffered_target_state == TS_PRESENT) { + seq_printf(m, " %3d %3d\n", + ++dev_count, + (u16)(uintptr_t)(t - a->targetdb)); + } + + if (dev_count == 0) + seq_puts(m, "none\n"); + + seq_puts(m, "\n"); + return 0; + +} + +int esas2r_release(struct Scsi_Host *sh) +{ + esas2r_log_dev(ESAS2R_LOG_INFO, &(sh->shost_gendev), + "esas2r_release() called"); + + esas2r_cleanup(sh); + if (sh->irq) + free_irq(sh->irq, NULL); + scsi_unregister(sh); + return 0; +} + +const char *esas2r_info(struct Scsi_Host *sh) +{ + struct esas2r_adapter *a = (struct esas2r_adapter *)sh->hostdata; + static char esas2r_info_str[512]; + + esas2r_log_dev(ESAS2R_LOG_INFO, &(sh->shost_gendev), + "esas2r_info() called"); + + /* + * if we haven't done so already, register as a char driver + * and stick a node under "/proc/scsi/esas2r/ATTOnode" + */ + + if (esas2r_proc_major <= 0) { + esas2r_proc_host = sh; + + esas2r_proc_major = register_chrdev(0, ESAS2R_DRVR_NAME, + &esas2r_proc_fops); + + esas2r_log_dev(ESAS2R_LOG_DEBG, &(sh->shost_gendev), + "register_chrdev (major %d)", + esas2r_proc_major); + + if (esas2r_proc_major > 0) { + struct proc_dir_entry *pde; + + pde = proc_create(ATTONODE_NAME, 0, + sh->hostt->proc_dir, + &esas2r_proc_fops); + + if (!pde) { + esas2r_log_dev(ESAS2R_LOG_WARN, + &(sh->shost_gendev), + "failed to create_proc_entry"); + esas2r_proc_major = -1; + } + } + } + + sprintf(esas2r_info_str, + ESAS2R_LONGNAME " (bus 0x%02X, device 0x%02X, IRQ 0x%02X)" + " driver version: "ESAS2R_VERSION_STR " firmware version: " + "%s\n", + a->pcid->bus->number, a->pcid->devfn, a->pcid->irq, + a->fw_rev[0] ? a->fw_rev : "(none)"); + + return esas2r_info_str; +} + +/* Callback for building a request scatter/gather list */ +static u32 get_physaddr_from_sgc(struct esas2r_sg_context *sgc, u64 *addr) +{ + u32 len; + + if (likely(sgc->cur_offset == sgc->exp_offset)) { + /* + * the normal case: caller used all bytes from previous call, so + * expected offset is the same as the current offset. + */ + + if (sgc->sgel_count < sgc->num_sgel) { + /* retrieve next segment, except for first time */ + if (sgc->exp_offset > (u8 *)0) { + /* advance current segment */ + sgc->cur_sgel = sg_next(sgc->cur_sgel); + ++(sgc->sgel_count); + } + + + len = sg_dma_len(sgc->cur_sgel); + (*addr) = sg_dma_address(sgc->cur_sgel); + + /* save the total # bytes returned to caller so far */ + sgc->exp_offset += len; + + } else { + len = 0; + } + } else if (sgc->cur_offset < sgc->exp_offset) { + /* + * caller did not use all bytes from previous call. need to + * compute the address based on current segment. + */ + + len = sg_dma_len(sgc->cur_sgel); + (*addr) = sg_dma_address(sgc->cur_sgel); + + sgc->exp_offset -= len; + + /* calculate PA based on prev segment address and offsets */ + *addr = *addr + + (sgc->cur_offset - sgc->exp_offset); + + sgc->exp_offset += len; + + /* re-calculate length based on offset */ + len = lower_32_bits( + sgc->exp_offset - sgc->cur_offset); + } else { /* if ( sgc->cur_offset > sgc->exp_offset ) */ + /* + * we don't expect the caller to skip ahead. + * cur_offset will never exceed the len we return + */ + len = 0; + } + + return len; +} + +int esas2r_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *cmd) +{ + struct esas2r_adapter *a = + (struct esas2r_adapter *)cmd->device->host->hostdata; + struct esas2r_request *rq; + struct esas2r_sg_context sgc; + unsigned bufflen; + + /* Assume success, if it fails we will fix the result later. */ + cmd->result = DID_OK << 16; + + if (unlikely(a->flags & AF_DEGRADED_MODE)) { + cmd->result = DID_NO_CONNECT << 16; + cmd->scsi_done(cmd); + return 0; + } + + rq = esas2r_alloc_request(a); + if (unlikely(rq == NULL)) { + esas2r_debug("esas2r_alloc_request failed"); + return SCSI_MLQUEUE_HOST_BUSY; + } + + rq->cmd = cmd; + bufflen = scsi_bufflen(cmd); + + if (likely(bufflen != 0)) { + if (cmd->sc_data_direction == DMA_TO_DEVICE) + rq->vrq->scsi.flags |= cpu_to_le32(FCP_CMND_WRD); + else if (cmd->sc_data_direction == DMA_FROM_DEVICE) + rq->vrq->scsi.flags |= cpu_to_le32(FCP_CMND_RDD); + } + + memcpy(rq->vrq->scsi.cdb, cmd->cmnd, cmd->cmd_len); + rq->vrq->scsi.length = cpu_to_le32(bufflen); + rq->target_id = cmd->device->id; + rq->vrq->scsi.flags |= cpu_to_le32(cmd->device->lun); + rq->sense_buf = cmd->sense_buffer; + rq->sense_len = SCSI_SENSE_BUFFERSIZE; + + esas2r_sgc_init(&sgc, a, rq, NULL); + + sgc.length = bufflen; + sgc.cur_offset = NULL; + + sgc.cur_sgel = scsi_sglist(cmd); + sgc.exp_offset = NULL; + sgc.num_sgel = scsi_dma_map(cmd); + sgc.sgel_count = 0; + + if (unlikely(sgc.num_sgel < 0)) { + esas2r_free_request(a, rq); + return SCSI_MLQUEUE_HOST_BUSY; + } + + sgc.get_phys_addr = (PGETPHYSADDR)get_physaddr_from_sgc; + + if (unlikely(!esas2r_build_sg_list(a, rq, &sgc))) { + scsi_dma_unmap(cmd); + esas2r_free_request(a, rq); + return SCSI_MLQUEUE_HOST_BUSY; + } + + esas2r_debug("start request %p to %d:%d\n", rq, (int)cmd->device->id, + (int)cmd->device->lun); + + esas2r_start_request(a, rq); + + return 0; +} + +static void complete_task_management_request(struct esas2r_adapter *a, + struct esas2r_request *rq) +{ + (*rq->task_management_status_ptr) = rq->req_stat; + esas2r_free_request(a, rq); +} + +/** + * Searches the specified queue for the specified queue for the command + * to abort. + * + * @param [in] a + * @param [in] abort_request + * @param [in] cmd + * t + * @return 0 on failure, 1 if command was not found, 2 if command was found + */ +static int esas2r_check_active_queue(struct esas2r_adapter *a, + struct esas2r_request **abort_request, + struct scsi_cmnd *cmd, + struct list_head *queue) +{ + bool found = false; + struct esas2r_request *ar = *abort_request; + struct esas2r_request *rq; + struct list_head *element, *next; + + list_for_each_safe(element, next, queue) { + + rq = list_entry(element, struct esas2r_request, req_list); + + if (rq->cmd == cmd) { + + /* Found the request. See what to do with it. */ + if (queue == &a->active_list) { + /* + * We are searching the active queue, which + * means that we need to send an abort request + * to the firmware. + */ + ar = esas2r_alloc_request(a); + if (ar == NULL) { + esas2r_log_dev(ESAS2R_LOG_WARN, + &(a->host->shost_gendev), + "unable to allocate an abort request for cmd %p", + cmd); + return 0; /* Failure */ + } + + /* + * Task management request must be formatted + * with a lock held. + */ + ar->sense_len = 0; + ar->vrq->scsi.length = 0; + ar->target_id = rq->target_id; + ar->vrq->scsi.flags |= cpu_to_le32( + (u8)le32_to_cpu(rq->vrq->scsi.flags)); + + memset(ar->vrq->scsi.cdb, 0, + sizeof(ar->vrq->scsi.cdb)); + + ar->vrq->scsi.flags |= cpu_to_le32( + FCP_CMND_TRM); + ar->vrq->scsi.u.abort_handle = + rq->vrq->scsi.handle; + } else { + /* + * The request is pending but not active on + * the firmware. Just free it now and we'll + * report the successful abort below. + */ + list_del_init(&rq->req_list); + esas2r_free_request(a, rq); + } + + found = true; + break; + } + + } + + if (!found) + return 1; /* Not found */ + + return 2; /* found */ + + +} + +int esas2r_eh_abort(struct scsi_cmnd *cmd) +{ + struct esas2r_adapter *a = + (struct esas2r_adapter *)cmd->device->host->hostdata; + struct esas2r_request *abort_request = NULL; + unsigned long flags; + struct list_head *queue; + int result; + + esas2r_log(ESAS2R_LOG_INFO, "eh_abort (%p)", cmd); + + if (a->flags & AF_DEGRADED_MODE) { + cmd->result = DID_ABORT << 16; + + scsi_set_resid(cmd, 0); + + cmd->scsi_done(cmd); + + return 0; + } + + spin_lock_irqsave(&a->queue_lock, flags); + + /* + * Run through the defer and active queues looking for the request + * to abort. + */ + + queue = &a->defer_list; + +check_active_queue: + + result = esas2r_check_active_queue(a, &abort_request, cmd, queue); + + if (!result) { + spin_unlock_irqrestore(&a->queue_lock, flags); + return FAILED; + } else if (result == 2 && (queue == &a->defer_list)) { + queue = &a->active_list; + goto check_active_queue; + } + + spin_unlock_irqrestore(&a->queue_lock, flags); + + if (abort_request) { + u8 task_management_status = RS_PENDING; + + /* + * the request is already active, so we need to tell + * the firmware to abort it and wait for the response. + */ + + abort_request->comp_cb = complete_task_management_request; + abort_request->task_management_status_ptr = + &task_management_status; + + esas2r_start_request(a, abort_request); + + if (atomic_read(&a->disable_cnt) == 0) + esas2r_do_deferred_processes(a); + + while (task_management_status == RS_PENDING) + msleep(10); + + /* + * Once we get here, the original request will have been + * completed by the firmware and the abort request will have + * been cleaned up. we're done! + */ + + return SUCCESS; + } + + /* + * If we get here, either we found the inactive request and + * freed it, or we didn't find it at all. Either way, success! + */ + + cmd->result = DID_ABORT << 16; + + scsi_set_resid(cmd, 0); + + cmd->scsi_done(cmd); + + return SUCCESS; +} + +static int esas2r_host_bus_reset(struct scsi_cmnd *cmd, bool host_reset) +{ + struct esas2r_adapter *a = + (struct esas2r_adapter *)cmd->device->host->hostdata; + + if (a->flags & AF_DEGRADED_MODE) + return FAILED; + + if (host_reset) + esas2r_reset_adapter(a); + else + esas2r_reset_bus(a); + + /* above call sets the AF_OS_RESET flag. wait for it to clear. */ + + while (a->flags & AF_OS_RESET) { + msleep(10); + + if (a->flags & AF_DEGRADED_MODE) + return FAILED; + } + + if (a->flags & AF_DEGRADED_MODE) + return FAILED; + + return SUCCESS; +} + +int esas2r_host_reset(struct scsi_cmnd *cmd) +{ + esas2r_log(ESAS2R_LOG_INFO, "host_reset (%p)", cmd); + + return esas2r_host_bus_reset(cmd, true); +} + +int esas2r_bus_reset(struct scsi_cmnd *cmd) +{ + esas2r_log(ESAS2R_LOG_INFO, "bus_reset (%p)", cmd); + + return esas2r_host_bus_reset(cmd, false); +} + +static int esas2r_dev_targ_reset(struct scsi_cmnd *cmd, bool target_reset) +{ + struct esas2r_adapter *a = + (struct esas2r_adapter *)cmd->device->host->hostdata; + struct esas2r_request *rq; + u8 task_management_status = RS_PENDING; + bool completed; + + if (a->flags & AF_DEGRADED_MODE) + return FAILED; + +retry: + rq = esas2r_alloc_request(a); + if (rq == NULL) { + if (target_reset) { + esas2r_log(ESAS2R_LOG_CRIT, + "unable to allocate a request for a " + "target reset (%d)!", + cmd->device->id); + } else { + esas2r_log(ESAS2R_LOG_CRIT, + "unable to allocate a request for a " + "device reset (%d:%d)!", + cmd->device->id, + cmd->device->lun); + } + + + return FAILED; + } + + rq->target_id = cmd->device->id; + rq->vrq->scsi.flags |= cpu_to_le32(cmd->device->lun); + rq->req_stat = RS_PENDING; + + rq->comp_cb = complete_task_management_request; + rq->task_management_status_ptr = &task_management_status; + + if (target_reset) { + esas2r_debug("issuing target reset (%p) to id %d", rq, + cmd->device->id); + completed = esas2r_send_task_mgmt(a, rq, 0x20); + } else { + esas2r_debug("issuing device reset (%p) to id %d lun %d", rq, + cmd->device->id, cmd->device->lun); + completed = esas2r_send_task_mgmt(a, rq, 0x10); + } + + if (completed) { + /* Task management cmd completed right away, need to free it. */ + + esas2r_free_request(a, rq); + } else { + /* + * Wait for firmware to complete the request. Completion + * callback will free it. + */ + while (task_management_status == RS_PENDING) + msleep(10); + } + + if (a->flags & AF_DEGRADED_MODE) + return FAILED; + + if (task_management_status == RS_BUSY) { + /* + * Busy, probably because we are flashing. Wait a bit and + * try again. + */ + msleep(100); + goto retry; + } + + return SUCCESS; +} + +int esas2r_device_reset(struct scsi_cmnd *cmd) +{ + esas2r_log(ESAS2R_LOG_INFO, "device_reset (%p)", cmd); + + return esas2r_dev_targ_reset(cmd, false); + +} + +int esas2r_target_reset(struct scsi_cmnd *cmd) +{ + esas2r_log(ESAS2R_LOG_INFO, "target_reset (%p)", cmd); + + return esas2r_dev_targ_reset(cmd, true); +} + +int esas2r_change_queue_depth(struct scsi_device *dev, int depth, int reason) +{ + esas2r_log(ESAS2R_LOG_INFO, "change_queue_depth %p, %d", dev, depth); + + scsi_adjust_queue_depth(dev, scsi_get_tag_type(dev), depth); + + return dev->queue_depth; +} + +int esas2r_change_queue_type(struct scsi_device *dev, int type) +{ + esas2r_log(ESAS2R_LOG_INFO, "change_queue_type %p, %d", dev, type); + + if (dev->tagged_supported) { + scsi_set_tag_type(dev, type); + + if (type) + scsi_activate_tcq(dev, dev->queue_depth); + else + scsi_deactivate_tcq(dev, dev->queue_depth); + } else { + type = 0; + } + + return type; +} + +int esas2r_slave_alloc(struct scsi_device *dev) +{ + return 0; +} + +int esas2r_slave_configure(struct scsi_device *dev) +{ + esas2r_log_dev(ESAS2R_LOG_INFO, &(dev->sdev_gendev), + "esas2r_slave_configure()"); + + if (dev->tagged_supported) { + scsi_set_tag_type(dev, MSG_SIMPLE_TAG); + scsi_activate_tcq(dev, cmd_per_lun); + } else { + scsi_set_tag_type(dev, 0); + scsi_deactivate_tcq(dev, cmd_per_lun); + } + + return 0; +} + +void esas2r_slave_destroy(struct scsi_device *dev) +{ + esas2r_log_dev(ESAS2R_LOG_INFO, &(dev->sdev_gendev), + "esas2r_slave_destroy()"); +} + +void esas2r_log_request_failure(struct esas2r_adapter *a, + struct esas2r_request *rq) +{ + u8 reqstatus = rq->req_stat; + + if (reqstatus == RS_SUCCESS) + return; + + if (rq->vrq->scsi.function == VDA_FUNC_SCSI) { + if (reqstatus == RS_SCSI_ERROR) { + if (rq->func_rsp.scsi_rsp.sense_len >= 13) { + esas2r_log(ESAS2R_LOG_WARN, + "request failure - SCSI error %x ASC:%x ASCQ:%x CDB:%x", + rq->sense_buf[2], rq->sense_buf[12], + rq->sense_buf[13], + rq->vrq->scsi.cdb[0]); + } else { + esas2r_log(ESAS2R_LOG_WARN, + "request failure - SCSI error CDB:%x\n", + rq->vrq->scsi.cdb[0]); + } + } else if ((rq->vrq->scsi.cdb[0] != INQUIRY + && rq->vrq->scsi.cdb[0] != REPORT_LUNS) + || (reqstatus != RS_SEL + && reqstatus != RS_SEL2)) { + if ((reqstatus == RS_UNDERRUN) && + (rq->vrq->scsi.cdb[0] == INQUIRY)) { + /* Don't log inquiry underruns */ + } else { + esas2r_log(ESAS2R_LOG_WARN, + "request failure - cdb:%x reqstatus:%d target:%d", + rq->vrq->scsi.cdb[0], reqstatus, + rq->target_id); + } + } + } +} + +void esas2r_wait_request(struct esas2r_adapter *a, struct esas2r_request *rq) +{ + u32 starttime; + u32 timeout; + + starttime = jiffies_to_msecs(jiffies); + timeout = rq->timeout ? rq->timeout : 5000; + + while (true) { + esas2r_polled_interrupt(a); + + if (rq->req_stat != RS_STARTED) + break; + + schedule_timeout_interruptible(msecs_to_jiffies(100)); + + if ((jiffies_to_msecs(jiffies) - starttime) > timeout) { + esas2r_hdebug("request TMO"); + esas2r_bugon(); + + rq->req_stat = RS_TIMEOUT; + + esas2r_local_reset_adapter(a); + return; + } + } +} + +u32 esas2r_map_data_window(struct esas2r_adapter *a, u32 addr_lo) +{ + u32 offset = addr_lo & (MW_DATA_WINDOW_SIZE - 1); + u32 base = addr_lo & -(signed int)MW_DATA_WINDOW_SIZE; + + if (a->window_base != base) { + esas2r_write_register_dword(a, MVR_PCI_WIN1_REMAP, + base | MVRPW1R_ENABLE); + esas2r_flush_register_dword(a, MVR_PCI_WIN1_REMAP); + a->window_base = base; + } + + return offset; +} + +/* Read a block of data from chip memory */ +bool esas2r_read_mem_block(struct esas2r_adapter *a, + void *to, + u32 from, + u32 size) +{ + u8 *end = (u8 *)to; + + while (size) { + u32 len; + u32 offset; + u32 iatvr; + + iatvr = (from & -(signed int)MW_DATA_WINDOW_SIZE); + + esas2r_map_data_window(a, iatvr); + + offset = from & (MW_DATA_WINDOW_SIZE - 1); + len = size; + + if (len > MW_DATA_WINDOW_SIZE - offset) + len = MW_DATA_WINDOW_SIZE - offset; + + from += len; + size -= len; + + while (len--) { + *end++ = esas2r_read_data_byte(a, offset); + offset++; + } + } + + return true; +} + +void esas2r_nuxi_mgt_data(u8 function, void *data) +{ + struct atto_vda_grp_info *g; + struct atto_vda_devinfo *d; + struct atto_vdapart_info *p; + struct atto_vda_dh_info *h; + struct atto_vda_metrics_info *m; + struct atto_vda_schedule_info *s; + struct atto_vda_buzzer_info *b; + u8 i; + + switch (function) { + case VDAMGT_BUZZER_INFO: + case VDAMGT_BUZZER_SET: + + b = (struct atto_vda_buzzer_info *)data; + + b->duration = le32_to_cpu(b->duration); + break; + + case VDAMGT_SCHEDULE_INFO: + case VDAMGT_SCHEDULE_EVENT: + + s = (struct atto_vda_schedule_info *)data; + + s->id = le32_to_cpu(s->id); + + break; + + case VDAMGT_DEV_INFO: + case VDAMGT_DEV_CLEAN: + case VDAMGT_DEV_PT_INFO: + case VDAMGT_DEV_FEATURES: + case VDAMGT_DEV_PT_FEATURES: + case VDAMGT_DEV_OPERATION: + + d = (struct atto_vda_devinfo *)data; + + d->capacity = le64_to_cpu(d->capacity); + d->block_size = le32_to_cpu(d->block_size); + d->ses_dev_index = le16_to_cpu(d->ses_dev_index); + d->target_id = le16_to_cpu(d->target_id); + d->lun = le16_to_cpu(d->lun); + d->features = le16_to_cpu(d->features); + break; + + case VDAMGT_GRP_INFO: + case VDAMGT_GRP_CREATE: + case VDAMGT_GRP_DELETE: + case VDAMGT_ADD_STORAGE: + case VDAMGT_MEMBER_ADD: + case VDAMGT_GRP_COMMIT: + case VDAMGT_GRP_REBUILD: + case VDAMGT_GRP_COMMIT_INIT: + case VDAMGT_QUICK_RAID: + case VDAMGT_GRP_FEATURES: + case VDAMGT_GRP_COMMIT_INIT_AUTOMAP: + case VDAMGT_QUICK_RAID_INIT_AUTOMAP: + case VDAMGT_SPARE_LIST: + case VDAMGT_SPARE_ADD: + case VDAMGT_SPARE_REMOVE: + case VDAMGT_LOCAL_SPARE_ADD: + case VDAMGT_GRP_OPERATION: + + g = (struct atto_vda_grp_info *)data; + + g->capacity = le64_to_cpu(g->capacity); + g->block_size = le32_to_cpu(g->block_size); + g->interleave = le32_to_cpu(g->interleave); + g->features = le16_to_cpu(g->features); + + for (i = 0; i < 32; i++) + g->members[i] = le16_to_cpu(g->members[i]); + + break; + + case VDAMGT_PART_INFO: + case VDAMGT_PART_MAP: + case VDAMGT_PART_UNMAP: + case VDAMGT_PART_AUTOMAP: + case VDAMGT_PART_SPLIT: + case VDAMGT_PART_MERGE: + + p = (struct atto_vdapart_info *)data; + + p->part_size = le64_to_cpu(p->part_size); + p->start_lba = le32_to_cpu(p->start_lba); + p->block_size = le32_to_cpu(p->block_size); + p->target_id = le16_to_cpu(p->target_id); + break; + + case VDAMGT_DEV_HEALTH_REQ: + + h = (struct atto_vda_dh_info *)data; + + h->med_defect_cnt = le32_to_cpu(h->med_defect_cnt); + h->info_exc_cnt = le32_to_cpu(h->info_exc_cnt); + break; + + case VDAMGT_DEV_METRICS: + + m = (struct atto_vda_metrics_info *)data; + + for (i = 0; i < 32; i++) + m->dev_indexes[i] = le16_to_cpu(m->dev_indexes[i]); + + break; + + default: + break; + } +} + +void esas2r_nuxi_cfg_data(u8 function, void *data) +{ + struct atto_vda_cfg_init *ci; + + switch (function) { + case VDA_CFG_INIT: + case VDA_CFG_GET_INIT: + case VDA_CFG_GET_INIT2: + + ci = (struct atto_vda_cfg_init *)data; + + ci->date_time.year = le16_to_cpu(ci->date_time.year); + ci->sgl_page_size = le32_to_cpu(ci->sgl_page_size); + ci->vda_version = le32_to_cpu(ci->vda_version); + ci->epoch_time = le32_to_cpu(ci->epoch_time); + ci->ioctl_tunnel = le32_to_cpu(ci->ioctl_tunnel); + ci->num_targets_backend = le32_to_cpu(ci->num_targets_backend); + break; + + default: + break; + } +} + +void esas2r_nuxi_ae_data(union atto_vda_ae *ae) +{ + struct atto_vda_ae_raid *r = &ae->raid; + struct atto_vda_ae_lu *l = &ae->lu; + + switch (ae->hdr.bytype) { + case VDAAE_HDR_TYPE_RAID: + + r->dwflags = le32_to_cpu(r->dwflags); + break; + + case VDAAE_HDR_TYPE_LU: + + l->dwevent = le32_to_cpu(l->dwevent); + l->wphys_target_id = le16_to_cpu(l->wphys_target_id); + l->id.tgtlun.wtarget_id = le16_to_cpu(l->id.tgtlun.wtarget_id); + + if (l->hdr.bylength >= offsetof(struct atto_vda_ae_lu, id) + + sizeof(struct atto_vda_ae_lu_tgt_lun_raid)) { + l->id.tgtlun_raid.dwinterleave + = le32_to_cpu(l->id.tgtlun_raid.dwinterleave); + l->id.tgtlun_raid.dwblock_size + = le32_to_cpu(l->id.tgtlun_raid.dwblock_size); + } + + break; + + case VDAAE_HDR_TYPE_DISK: + default: + break; + } +} + +void esas2r_free_request(struct esas2r_adapter *a, struct esas2r_request *rq) +{ + unsigned long flags; + + esas2r_rq_destroy_request(rq, a); + spin_lock_irqsave(&a->request_lock, flags); + list_add(&rq->comp_list, &a->avail_request); + spin_unlock_irqrestore(&a->request_lock, flags); +} + +struct esas2r_request *esas2r_alloc_request(struct esas2r_adapter *a) +{ + struct esas2r_request *rq; + unsigned long flags; + + spin_lock_irqsave(&a->request_lock, flags); + + if (unlikely(list_empty(&a->avail_request))) { + spin_unlock_irqrestore(&a->request_lock, flags); + return NULL; + } + + rq = list_first_entry(&a->avail_request, struct esas2r_request, + comp_list); + list_del(&rq->comp_list); + spin_unlock_irqrestore(&a->request_lock, flags); + esas2r_rq_init_request(rq, a); + + return rq; + +} + +void esas2r_complete_request_cb(struct esas2r_adapter *a, + struct esas2r_request *rq) +{ + esas2r_debug("completing request %p\n", rq); + + scsi_dma_unmap(rq->cmd); + + if (unlikely(rq->req_stat != RS_SUCCESS)) { + esas2r_debug("[%x STATUS %x:%x (%x)]", rq->target_id, + rq->req_stat, + rq->func_rsp.scsi_rsp.scsi_stat, + rq->cmd); + + rq->cmd->result = + ((esas2r_req_status_to_error(rq->req_stat) << 16) + | (rq->func_rsp.scsi_rsp.scsi_stat & STATUS_MASK)); + + if (rq->req_stat == RS_UNDERRUN) + scsi_set_resid(rq->cmd, + le32_to_cpu(rq->func_rsp.scsi_rsp. + residual_length)); + else + scsi_set_resid(rq->cmd, 0); + } + + rq->cmd->scsi_done(rq->cmd); + + esas2r_free_request(a, rq); +} + +/* Run tasklet to handle stuff outside of interrupt context. */ +void esas2r_adapter_tasklet(unsigned long context) +{ + struct esas2r_adapter *a = (struct esas2r_adapter *)context; + + if (unlikely(a->flags2 & AF2_TIMER_TICK)) { + esas2r_lock_clear_flags(&a->flags2, AF2_TIMER_TICK); + esas2r_timer_tick(a); + } + + if (likely(a->flags2 & AF2_INT_PENDING)) { + esas2r_lock_clear_flags(&a->flags2, AF2_INT_PENDING); + esas2r_adapter_interrupt(a); + } + + if (esas2r_is_tasklet_pending(a)) + esas2r_do_tasklet_tasks(a); + + if (esas2r_is_tasklet_pending(a) + || (a->flags2 & AF2_INT_PENDING) + || (a->flags2 & AF2_TIMER_TICK)) { + esas2r_lock_clear_flags(&a->flags, AF_TASKLET_SCHEDULED); + esas2r_schedule_tasklet(a); + } else { + esas2r_lock_clear_flags(&a->flags, AF_TASKLET_SCHEDULED); + } +} + +static void esas2r_timer_callback(unsigned long context); + +void esas2r_kickoff_timer(struct esas2r_adapter *a) +{ + init_timer(&a->timer); + + a->timer.function = esas2r_timer_callback; + a->timer.data = (unsigned long)a; + a->timer.expires = jiffies + + msecs_to_jiffies(100); + + add_timer(&a->timer); +} + +static void esas2r_timer_callback(unsigned long context) +{ + struct esas2r_adapter *a = (struct esas2r_adapter *)context; + + esas2r_lock_set_flags(&a->flags2, AF2_TIMER_TICK); + + esas2r_schedule_tasklet(a); + + esas2r_kickoff_timer(a); +} + +/* + * Firmware events need to be handled outside of interrupt context + * so we schedule a delayed_work to handle them. + */ + +static void +esas2r_free_fw_event(struct esas2r_fw_event_work *fw_event) +{ + unsigned long flags; + struct esas2r_adapter *a = fw_event->a; + + spin_lock_irqsave(&a->fw_event_lock, flags); + list_del(&fw_event->list); + kfree(fw_event); + spin_unlock_irqrestore(&a->fw_event_lock, flags); +} + +void +esas2r_fw_event_off(struct esas2r_adapter *a) +{ + unsigned long flags; + + spin_lock_irqsave(&a->fw_event_lock, flags); + a->fw_events_off = 1; + spin_unlock_irqrestore(&a->fw_event_lock, flags); +} + +void +esas2r_fw_event_on(struct esas2r_adapter *a) +{ + unsigned long flags; + + spin_lock_irqsave(&a->fw_event_lock, flags); + a->fw_events_off = 0; + spin_unlock_irqrestore(&a->fw_event_lock, flags); +} + +static void esas2r_add_device(struct esas2r_adapter *a, u16 target_id) +{ + int ret; + struct scsi_device *scsi_dev; + + scsi_dev = scsi_device_lookup(a->host, 0, target_id, 0); + + if (scsi_dev) { + esas2r_log_dev( + ESAS2R_LOG_WARN, + &(scsi_dev-> + sdev_gendev), + "scsi device already exists at id %d", target_id); + + scsi_device_put(scsi_dev); + } else { + esas2r_log_dev( + ESAS2R_LOG_INFO, + &(a->host-> + shost_gendev), + "scsi_add_device() called for 0:%d:0", + target_id); + + ret = scsi_add_device(a->host, 0, target_id, 0); + if (ret) { + esas2r_log_dev( + ESAS2R_LOG_CRIT, + &(a->host-> + shost_gendev), + "scsi_add_device failed with %d for id %d", + ret, target_id); + } + } +} + +static void esas2r_remove_device(struct esas2r_adapter *a, u16 target_id) +{ + struct scsi_device *scsi_dev; + + scsi_dev = scsi_device_lookup(a->host, 0, target_id, 0); + + if (scsi_dev) { + scsi_device_set_state(scsi_dev, SDEV_OFFLINE); + + esas2r_log_dev( + ESAS2R_LOG_INFO, + &(scsi_dev-> + sdev_gendev), + "scsi_remove_device() called for 0:%d:0", + target_id); + + scsi_remove_device(scsi_dev); + + esas2r_log_dev( + ESAS2R_LOG_INFO, + &(scsi_dev-> + sdev_gendev), + "scsi_device_put() called"); + + scsi_device_put(scsi_dev); + } else { + esas2r_log_dev( + ESAS2R_LOG_WARN, + &(a->host->shost_gendev), + "no target found at id %d", + target_id); + } +} + +/* + * Sends a firmware asynchronous event to anyone who happens to be + * listening on the defined ATTO VDA event ports. + */ +static void esas2r_send_ae_event(struct esas2r_fw_event_work *fw_event) +{ + struct esas2r_vda_ae *ae = (struct esas2r_vda_ae *)fw_event->data; + char *type; + + switch (ae->vda_ae.hdr.bytype) { + case VDAAE_HDR_TYPE_RAID: + type = "RAID group state change"; + break; + + case VDAAE_HDR_TYPE_LU: + type = "Mapped destination LU change"; + break; + + case VDAAE_HDR_TYPE_DISK: + type = "Physical disk inventory change"; + break; + + case VDAAE_HDR_TYPE_RESET: + type = "Firmware reset"; + break; + + case VDAAE_HDR_TYPE_LOG_INFO: + type = "Event Log message (INFO level)"; + break; + + case VDAAE_HDR_TYPE_LOG_WARN: + type = "Event Log message (WARN level)"; + break; + + case VDAAE_HDR_TYPE_LOG_CRIT: + type = "Event Log message (CRIT level)"; + break; + + case VDAAE_HDR_TYPE_LOG_FAIL: + type = "Event Log message (FAIL level)"; + break; + + case VDAAE_HDR_TYPE_NVC: + type = "NVCache change"; + break; + + case VDAAE_HDR_TYPE_TLG_INFO: + type = "Time stamped log message (INFO level)"; + break; + + case VDAAE_HDR_TYPE_TLG_WARN: + type = "Time stamped log message (WARN level)"; + break; + + case VDAAE_HDR_TYPE_TLG_CRIT: + type = "Time stamped log message (CRIT level)"; + break; + + case VDAAE_HDR_TYPE_PWRMGT: + type = "Power management"; + break; + + case VDAAE_HDR_TYPE_MUTE: + type = "Mute button pressed"; + break; + + case VDAAE_HDR_TYPE_DEV: + type = "Device attribute change"; + break; + + default: + type = "Unknown"; + break; + } + + esas2r_log(ESAS2R_LOG_WARN, + "An async event of type \"%s\" was received from the firmware. The event contents are:", + type); + esas2r_log_hexdump(ESAS2R_LOG_WARN, &ae->vda_ae, + ae->vda_ae.hdr.bylength); + +} + +static void +esas2r_firmware_event_work(struct work_struct *work) +{ + struct esas2r_fw_event_work *fw_event = + container_of(work, struct esas2r_fw_event_work, work.work); + + struct esas2r_adapter *a = fw_event->a; + + u16 target_id = *(u16 *)&fw_event->data[0]; + + if (a->fw_events_off) + goto done; + + switch (fw_event->type) { + case fw_event_null: + break; /* do nothing */ + + case fw_event_lun_change: + esas2r_remove_device(a, target_id); + esas2r_add_device(a, target_id); + break; + + case fw_event_present: + esas2r_add_device(a, target_id); + break; + + case fw_event_not_present: + esas2r_remove_device(a, target_id); + break; + + case fw_event_vda_ae: + esas2r_send_ae_event(fw_event); + break; + } + +done: + esas2r_free_fw_event(fw_event); +} + +void esas2r_queue_fw_event(struct esas2r_adapter *a, + enum fw_event_type type, + void *data, + int data_sz) +{ + struct esas2r_fw_event_work *fw_event; + unsigned long flags; + + fw_event = kzalloc(sizeof(struct esas2r_fw_event_work), GFP_ATOMIC); + if (!fw_event) { + esas2r_log(ESAS2R_LOG_WARN, + "esas2r_queue_fw_event failed to alloc"); + return; + } + + if (type == fw_event_vda_ae) { + struct esas2r_vda_ae *ae = + (struct esas2r_vda_ae *)fw_event->data; + + ae->signature = ESAS2R_VDA_EVENT_SIG; + ae->bus_number = a->pcid->bus->number; + ae->devfn = a->pcid->devfn; + memcpy(&ae->vda_ae, data, sizeof(ae->vda_ae)); + } else { + memcpy(fw_event->data, data, data_sz); + } + + fw_event->type = type; + fw_event->a = a; + + spin_lock_irqsave(&a->fw_event_lock, flags); + list_add_tail(&fw_event->list, &a->fw_event_list); + INIT_DELAYED_WORK(&fw_event->work, esas2r_firmware_event_work); + queue_delayed_work_on( + smp_processor_id(), a->fw_event_q, &fw_event->work, + msecs_to_jiffies(1)); + spin_unlock_irqrestore(&a->fw_event_lock, flags); +} + +void esas2r_target_state_changed(struct esas2r_adapter *a, u16 targ_id, + u8 state) +{ + if (state == TS_LUN_CHANGE) + esas2r_queue_fw_event(a, fw_event_lun_change, &targ_id, + sizeof(targ_id)); + else if (state == TS_PRESENT) + esas2r_queue_fw_event(a, fw_event_present, &targ_id, + sizeof(targ_id)); + else if (state == TS_NOT_PRESENT) + esas2r_queue_fw_event(a, fw_event_not_present, &targ_id, + sizeof(targ_id)); +} + +/* Translate status to a Linux SCSI mid-layer error code */ +int esas2r_req_status_to_error(u8 req_stat) +{ + switch (req_stat) { + case RS_OVERRUN: + case RS_UNDERRUN: + case RS_SUCCESS: + /* + * NOTE: SCSI mid-layer wants a good status for a SCSI error, because + * it will check the scsi_stat value in the completion anyway. + */ + case RS_SCSI_ERROR: + return DID_OK; + + case RS_SEL: + case RS_SEL2: + return DID_NO_CONNECT; + + case RS_RESET: + return DID_RESET; + + case RS_ABORTED: + return DID_ABORT; + + case RS_BUSY: + return DID_BUS_BUSY; + } + + /* everything else is just an error. */ + + return DID_ERROR; +} + +module_init(esas2r_init); +module_exit(esas2r_exit); -- 1.8.1.4 -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html