freebsd-skq/sys/dev/isci/isci_controller.c
Alexander Motin e5dfa058da MFprojects/camlock r248982:
Stop abusing xpt_periph in random plases that really have no periph related
to CCB, for example, bus scanning.  NULL value is fine in such cases and it
is correctly logged in debug messages as "noperiph".  If at some point we
need some real XPT periphs (alike to pmpX now), quite likely they will be
per-bus, and not a single global instance as xpt_periph now.
2013-04-14 09:55:48 +00:00

827 lines
26 KiB
C

/*-
* BSD LICENSE
*
* Copyright(c) 2008 - 2011 Intel Corporation. All rights reserved.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in
* the documentation and/or other materials provided with the
* distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 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
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <dev/isci/isci.h>
#include <sys/conf.h>
#include <sys/malloc.h>
#include <cam/cam_periph.h>
#include <cam/cam_xpt_periph.h>
#include <dev/isci/scil/sci_memory_descriptor_list.h>
#include <dev/isci/scil/sci_memory_descriptor_list_decorator.h>
#include <dev/isci/scil/scif_controller.h>
#include <dev/isci/scil/scif_library.h>
#include <dev/isci/scil/scif_io_request.h>
#include <dev/isci/scil/scif_task_request.h>
#include <dev/isci/scil/scif_remote_device.h>
#include <dev/isci/scil/scif_domain.h>
#include <dev/isci/scil/scif_user_callback.h>
#include <dev/isci/scil/scic_sgpio.h>
#include <dev/led/led.h>
void isci_action(struct cam_sim *sim, union ccb *ccb);
void isci_poll(struct cam_sim *sim);
#define ccb_sim_ptr sim_priv.entries[0].ptr
/**
* @brief This user callback will inform the user that the controller has
* had a serious unexpected error. The user should not the error,
* disable interrupts, and wait for current ongoing processing to
* complete. Subsequently, the user should reset the controller.
*
* @param[in] controller This parameter specifies the controller that had
* an error.
*
* @return none
*/
void scif_cb_controller_error(SCI_CONTROLLER_HANDLE_T controller,
SCI_CONTROLLER_ERROR error)
{
isci_log_message(0, "ISCI", "scif_cb_controller_error: 0x%x\n",
error);
}
/**
* @brief This user callback will inform the user that the controller has
* finished the start process.
*
* @param[in] controller This parameter specifies the controller that was
* started.
* @param[in] completion_status This parameter specifies the results of
* the start operation. SCI_SUCCESS indicates successful
* completion.
*
* @return none
*/
void scif_cb_controller_start_complete(SCI_CONTROLLER_HANDLE_T controller,
SCI_STATUS completion_status)
{
uint32_t index;
struct ISCI_CONTROLLER *isci_controller = (struct ISCI_CONTROLLER *)
sci_object_get_association(controller);
isci_controller->is_started = TRUE;
/* Set bits for all domains. We will clear them one-by-one once
* the domains complete discovery, or return error when calling
* scif_domain_discover. Once all bits are clear, we will register
* the controller with CAM.
*/
isci_controller->initial_discovery_mask = (1 << SCI_MAX_DOMAINS) - 1;
for(index = 0; index < SCI_MAX_DOMAINS; index++) {
SCI_STATUS status;
SCI_DOMAIN_HANDLE_T domain =
isci_controller->domain[index].sci_object;
status = scif_domain_discover(
domain,
scif_domain_get_suggested_discover_timeout(domain),
DEVICE_TIMEOUT
);
if (status != SCI_SUCCESS)
{
isci_controller_domain_discovery_complete(
isci_controller, &isci_controller->domain[index]);
}
}
}
/**
* @brief This user callback will inform the user that the controller has
* finished the stop process. Note, after user calls
* scif_controller_stop(), before user receives this controller stop
* complete callback, user should not expect any callback from
* framework, such like scif_cb_domain_change_notification().
*
* @param[in] controller This parameter specifies the controller that was
* stopped.
* @param[in] completion_status This parameter specifies the results of
* the stop operation. SCI_SUCCESS indicates successful
* completion.
*
* @return none
*/
void scif_cb_controller_stop_complete(SCI_CONTROLLER_HANDLE_T controller,
SCI_STATUS completion_status)
{
struct ISCI_CONTROLLER *isci_controller = (struct ISCI_CONTROLLER *)
sci_object_get_association(controller);
isci_controller->is_started = FALSE;
}
static void
isci_single_map(void *arg, bus_dma_segment_t *seg, int nseg, int error)
{
SCI_PHYSICAL_ADDRESS *phys_addr = arg;
*phys_addr = seg[0].ds_addr;
}
/**
* @brief This method will be invoked to allocate memory dynamically.
*
* @param[in] controller This parameter represents the controller
* object for which to allocate memory.
* @param[out] mde This parameter represents the memory descriptor to
* be filled in by the user that will reference the newly
* allocated memory.
*
* @return none
*/
void scif_cb_controller_allocate_memory(SCI_CONTROLLER_HANDLE_T controller,
SCI_PHYSICAL_MEMORY_DESCRIPTOR_T *mde)
{
struct ISCI_CONTROLLER *isci_controller = (struct ISCI_CONTROLLER *)
sci_object_get_association(controller);
/*
* Note this routine is only used for buffers needed to translate
* SCSI UNMAP commands to ATA DSM commands for SATA disks.
*
* We first try to pull a buffer from the controller's pool, and only
* call contigmalloc if one isn't there.
*/
if (!sci_pool_empty(isci_controller->unmap_buffer_pool)) {
sci_pool_get(isci_controller->unmap_buffer_pool,
mde->virtual_address);
} else
mde->virtual_address = contigmalloc(PAGE_SIZE,
M_ISCI, M_NOWAIT, 0, BUS_SPACE_MAXADDR,
mde->constant_memory_alignment, 0);
if (mde->virtual_address != NULL)
bus_dmamap_load(isci_controller->buffer_dma_tag,
NULL, mde->virtual_address, PAGE_SIZE,
isci_single_map, &mde->physical_address,
BUS_DMA_NOWAIT);
}
/**
* @brief This method will be invoked to allocate memory dynamically.
*
* @param[in] controller This parameter represents the controller
* object for which to allocate memory.
* @param[out] mde This parameter represents the memory descriptor to
* be filled in by the user that will reference the newly
* allocated memory.
*
* @return none
*/
void scif_cb_controller_free_memory(SCI_CONTROLLER_HANDLE_T controller,
SCI_PHYSICAL_MEMORY_DESCRIPTOR_T * mde)
{
struct ISCI_CONTROLLER *isci_controller = (struct ISCI_CONTROLLER *)
sci_object_get_association(controller);
/*
* Put the buffer back into the controller's buffer pool, rather
* than invoking configfree. This helps reduce chance we won't
* have buffers available when system is under memory pressure.
*/
sci_pool_put(isci_controller->unmap_buffer_pool,
mde->virtual_address);
}
void isci_controller_construct(struct ISCI_CONTROLLER *controller,
struct isci_softc *isci)
{
SCI_CONTROLLER_HANDLE_T scif_controller_handle;
scif_library_allocate_controller(isci->sci_library_handle,
&scif_controller_handle);
scif_controller_construct(isci->sci_library_handle,
scif_controller_handle, NULL);
controller->isci = isci;
controller->scif_controller_handle = scif_controller_handle;
/* This allows us to later use
* sci_object_get_association(scif_controller_handle)
* inside of a callback routine to get our struct ISCI_CONTROLLER object
*/
sci_object_set_association(scif_controller_handle, (void *)controller);
controller->is_started = FALSE;
controller->is_frozen = FALSE;
controller->release_queued_ccbs = FALSE;
controller->sim = NULL;
controller->initial_discovery_mask = 0;
sci_fast_list_init(&controller->pending_device_reset_list);
mtx_init(&controller->lock, "isci", NULL, MTX_DEF);
uint32_t domain_index;
for(domain_index = 0; domain_index < SCI_MAX_DOMAINS; domain_index++) {
isci_domain_construct( &controller->domain[domain_index],
domain_index, controller);
}
controller->timer_memory = malloc(
sizeof(struct ISCI_TIMER) * SCI_MAX_TIMERS, M_ISCI,
M_NOWAIT | M_ZERO);
sci_pool_initialize(controller->timer_pool);
struct ISCI_TIMER *timer = (struct ISCI_TIMER *)
controller->timer_memory;
for ( int i = 0; i < SCI_MAX_TIMERS; i++ ) {
sci_pool_put(controller->timer_pool, timer++);
}
sci_pool_initialize(controller->unmap_buffer_pool);
}
static void isci_led_fault_func(void *priv, int onoff)
{
struct ISCI_PHY *phy = priv;
/* map onoff to the fault LED */
phy->led_fault = onoff;
scic_sgpio_update_led_state(phy->handle, 1 << phy->index,
phy->led_fault, phy->led_locate, 0);
}
static void isci_led_locate_func(void *priv, int onoff)
{
struct ISCI_PHY *phy = priv;
/* map onoff to the locate LED */
phy->led_locate = onoff;
scic_sgpio_update_led_state(phy->handle, 1 << phy->index,
phy->led_fault, phy->led_locate, 0);
}
SCI_STATUS isci_controller_initialize(struct ISCI_CONTROLLER *controller)
{
SCIC_USER_PARAMETERS_T scic_user_parameters;
SCI_CONTROLLER_HANDLE_T scic_controller_handle;
char led_name[64];
unsigned long tunable;
int i;
scic_controller_handle =
scif_controller_get_scic_handle(controller->scif_controller_handle);
if (controller->isci->oem_parameters_found == TRUE)
{
scic_oem_parameters_set(
scic_controller_handle,
&controller->oem_parameters,
(uint8_t)(controller->oem_parameters_version));
}
scic_user_parameters_get(scic_controller_handle, &scic_user_parameters);
if (TUNABLE_ULONG_FETCH("hw.isci.no_outbound_task_timeout", &tunable))
scic_user_parameters.sds1.no_outbound_task_timeout =
(uint8_t)tunable;
if (TUNABLE_ULONG_FETCH("hw.isci.ssp_max_occupancy_timeout", &tunable))
scic_user_parameters.sds1.ssp_max_occupancy_timeout =
(uint16_t)tunable;
if (TUNABLE_ULONG_FETCH("hw.isci.stp_max_occupancy_timeout", &tunable))
scic_user_parameters.sds1.stp_max_occupancy_timeout =
(uint16_t)tunable;
if (TUNABLE_ULONG_FETCH("hw.isci.ssp_inactivity_timeout", &tunable))
scic_user_parameters.sds1.ssp_inactivity_timeout =
(uint16_t)tunable;
if (TUNABLE_ULONG_FETCH("hw.isci.stp_inactivity_timeout", &tunable))
scic_user_parameters.sds1.stp_inactivity_timeout =
(uint16_t)tunable;
if (TUNABLE_ULONG_FETCH("hw.isci.max_speed_generation", &tunable))
for (i = 0; i < SCI_MAX_PHYS; i++)
scic_user_parameters.sds1.phys[i].max_speed_generation =
(uint8_t)tunable;
scic_user_parameters_set(scic_controller_handle, &scic_user_parameters);
/* Scheduler bug in SCU requires SCIL to reserve some task contexts as a
* a workaround - one per domain.
*/
controller->queue_depth = SCI_MAX_IO_REQUESTS - SCI_MAX_DOMAINS;
if (TUNABLE_INT_FETCH("hw.isci.controller_queue_depth",
&controller->queue_depth)) {
controller->queue_depth = max(1, min(controller->queue_depth,
SCI_MAX_IO_REQUESTS - SCI_MAX_DOMAINS));
}
/* Reserve one request so that we can ensure we have one available TC
* to do internal device resets.
*/
controller->sim_queue_depth = controller->queue_depth - 1;
/* Although we save one TC to do internal device resets, it is possible
* we could end up using several TCs for simultaneous device resets
* while at the same time having CAM fill our controller queue. To
* simulate this condition, and how our driver handles it, we can set
* this io_shortage parameter, which will tell CAM that we have a
* large queue depth than we really do.
*/
uint32_t io_shortage = 0;
TUNABLE_INT_FETCH("hw.isci.io_shortage", &io_shortage);
controller->sim_queue_depth += io_shortage;
/* Attach to CAM using xpt_bus_register now, then immediately freeze
* the simq. It will get released later when initial domain discovery
* is complete.
*/
controller->has_been_scanned = FALSE;
mtx_lock(&controller->lock);
isci_controller_attach_to_cam(controller);
xpt_freeze_simq(controller->sim, 1);
mtx_unlock(&controller->lock);
for (i = 0; i < SCI_MAX_PHYS; i++) {
controller->phys[i].handle = scic_controller_handle;
controller->phys[i].index = i;
/* fault */
controller->phys[i].led_fault = 0;
sprintf(led_name, "isci.bus%d.port%d.fault", controller->index, i);
controller->phys[i].cdev_fault = led_create(isci_led_fault_func,
&controller->phys[i], led_name);
/* locate */
controller->phys[i].led_locate = 0;
sprintf(led_name, "isci.bus%d.port%d.locate", controller->index, i);
controller->phys[i].cdev_locate = led_create(isci_led_locate_func,
&controller->phys[i], led_name);
}
return (scif_controller_initialize(controller->scif_controller_handle));
}
int isci_controller_allocate_memory(struct ISCI_CONTROLLER *controller)
{
int error;
device_t device = controller->isci->device;
uint32_t max_segment_size = isci_io_request_get_max_io_size();
uint32_t status = 0;
struct ISCI_MEMORY *uncached_controller_memory =
&controller->uncached_controller_memory;
struct ISCI_MEMORY *cached_controller_memory =
&controller->cached_controller_memory;
struct ISCI_MEMORY *request_memory =
&controller->request_memory;
POINTER_UINT virtual_address;
bus_addr_t physical_address;
controller->mdl = sci_controller_get_memory_descriptor_list_handle(
controller->scif_controller_handle);
uncached_controller_memory->size = sci_mdl_decorator_get_memory_size(
controller->mdl, SCI_MDE_ATTRIBUTE_PHYSICALLY_CONTIGUOUS);
error = isci_allocate_dma_buffer(device, uncached_controller_memory);
if (error != 0)
return (error);
sci_mdl_decorator_assign_memory( controller->mdl,
SCI_MDE_ATTRIBUTE_PHYSICALLY_CONTIGUOUS,
uncached_controller_memory->virtual_address,
uncached_controller_memory->physical_address);
cached_controller_memory->size = sci_mdl_decorator_get_memory_size(
controller->mdl,
SCI_MDE_ATTRIBUTE_CACHEABLE | SCI_MDE_ATTRIBUTE_PHYSICALLY_CONTIGUOUS
);
error = isci_allocate_dma_buffer(device, cached_controller_memory);
if (error != 0)
return (error);
sci_mdl_decorator_assign_memory(controller->mdl,
SCI_MDE_ATTRIBUTE_CACHEABLE | SCI_MDE_ATTRIBUTE_PHYSICALLY_CONTIGUOUS,
cached_controller_memory->virtual_address,
cached_controller_memory->physical_address);
request_memory->size =
controller->queue_depth * isci_io_request_get_object_size();
error = isci_allocate_dma_buffer(device, request_memory);
if (error != 0)
return (error);
/* For STP PIO testing, we want to ensure we can force multiple SGLs
* since this has been a problem area in SCIL. This tunable parameter
* will allow us to force DMA segments to a smaller size, ensuring
* that even if a physically contiguous buffer is attached to this
* I/O, the DMA subsystem will pass us multiple segments in our DMA
* load callback.
*/
TUNABLE_INT_FETCH("hw.isci.max_segment_size", &max_segment_size);
/* Create DMA tag for our I/O requests. Then we can create DMA maps based off
* of this tag and store them in each of our ISCI_IO_REQUEST objects. This
* will enable better performance than creating the DMA maps everytime we get
* an I/O.
*/
status = bus_dma_tag_create(bus_get_dma_tag(device), 0x1, 0x0,
BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL,
isci_io_request_get_max_io_size(),
SCI_MAX_SCATTER_GATHER_ELEMENTS, max_segment_size, 0, NULL, NULL,
&controller->buffer_dma_tag);
sci_pool_initialize(controller->request_pool);
virtual_address = request_memory->virtual_address;
physical_address = request_memory->physical_address;
for (int i = 0; i < controller->queue_depth; i++) {
struct ISCI_REQUEST *request =
(struct ISCI_REQUEST *)virtual_address;
isci_request_construct(request,
controller->scif_controller_handle,
controller->buffer_dma_tag, physical_address);
sci_pool_put(controller->request_pool, request);
virtual_address += isci_request_get_object_size();
physical_address += isci_request_get_object_size();
}
uint32_t remote_device_size = sizeof(struct ISCI_REMOTE_DEVICE) +
scif_remote_device_get_object_size();
controller->remote_device_memory = (uint8_t *) malloc(
remote_device_size * SCI_MAX_REMOTE_DEVICES, M_ISCI,
M_NOWAIT | M_ZERO);
sci_pool_initialize(controller->remote_device_pool);
uint8_t *remote_device_memory_ptr = controller->remote_device_memory;
for (int i = 0; i < SCI_MAX_REMOTE_DEVICES; i++) {
struct ISCI_REMOTE_DEVICE *remote_device =
(struct ISCI_REMOTE_DEVICE *)remote_device_memory_ptr;
controller->remote_device[i] = NULL;
remote_device->index = i;
remote_device->is_resetting = FALSE;
remote_device->frozen_lun_mask = 0;
sci_fast_list_element_init(remote_device,
&remote_device->pending_device_reset_element);
TAILQ_INIT(&remote_device->queued_ccbs);
remote_device->release_queued_ccb = FALSE;
remote_device->queued_ccb_in_progress = NULL;
/*
* For the first SCI_MAX_DOMAINS device objects, do not put
* them in the pool, rather assign them to each domain. This
* ensures that any device attached directly to port "i" will
* always get CAM target id "i".
*/
if (i < SCI_MAX_DOMAINS)
controller->domain[i].da_remote_device = remote_device;
else
sci_pool_put(controller->remote_device_pool,
remote_device);
remote_device_memory_ptr += remote_device_size;
}
return (0);
}
void isci_controller_start(void *controller_handle)
{
struct ISCI_CONTROLLER *controller =
(struct ISCI_CONTROLLER *)controller_handle;
SCI_CONTROLLER_HANDLE_T scif_controller_handle =
controller->scif_controller_handle;
scif_controller_start(scif_controller_handle,
scif_controller_get_suggested_start_timeout(scif_controller_handle));
scic_controller_enable_interrupts(
scif_controller_get_scic_handle(controller->scif_controller_handle));
}
void isci_controller_domain_discovery_complete(
struct ISCI_CONTROLLER *isci_controller, struct ISCI_DOMAIN *isci_domain)
{
if (!isci_controller->has_been_scanned)
{
/* Controller has not been scanned yet. We'll clear
* the discovery bit for this domain, then check if all bits
* are now clear. That would indicate that all domains are
* done with discovery and we can then proceed with initial
* scan.
*/
isci_controller->initial_discovery_mask &=
~(1 << isci_domain->index);
if (isci_controller->initial_discovery_mask == 0) {
struct isci_softc *driver = isci_controller->isci;
uint8_t next_index = isci_controller->index + 1;
isci_controller->has_been_scanned = TRUE;
/* Unfreeze simq to allow initial scan to proceed. */
xpt_release_simq(isci_controller->sim, TRUE);
#if __FreeBSD_version < 800000
/* When driver is loaded after boot, we need to
* explicitly rescan here for versions <8.0, because
* CAM only automatically scans new buses at boot
* time.
*/
union ccb *ccb = xpt_alloc_ccb_nowait();
xpt_create_path(&ccb->ccb_h.path, NULL,
cam_sim_path(isci_controller->sim),
CAM_TARGET_WILDCARD, CAM_LUN_WILDCARD);
xpt_rescan(ccb);
#endif
if (next_index < driver->controller_count) {
/* There are more controllers that need to
* start. So start the next one.
*/
isci_controller_start(
&driver->controllers[next_index]);
}
else
{
/* All controllers have been started and completed discovery.
* Disestablish the config hook while will signal to the
* kernel during boot that it is safe to try to find and
* mount the root partition.
*/
config_intrhook_disestablish(
&driver->config_hook);
}
}
}
}
int isci_controller_attach_to_cam(struct ISCI_CONTROLLER *controller)
{
struct isci_softc *isci = controller->isci;
device_t parent = device_get_parent(isci->device);
int unit = device_get_unit(isci->device);
struct cam_devq *isci_devq = cam_simq_alloc(controller->sim_queue_depth);
if(isci_devq == NULL) {
isci_log_message(0, "ISCI", "isci_devq is NULL \n");
return (-1);
}
controller->sim = cam_sim_alloc(isci_action, isci_poll, "isci",
controller, unit, &controller->lock, controller->sim_queue_depth,
controller->sim_queue_depth, isci_devq);
if(controller->sim == NULL) {
isci_log_message(0, "ISCI", "cam_sim_alloc... fails\n");
cam_simq_free(isci_devq);
return (-1);
}
if(xpt_bus_register(controller->sim, parent, controller->index)
!= CAM_SUCCESS) {
isci_log_message(0, "ISCI", "xpt_bus_register...fails \n");
cam_sim_free(controller->sim, TRUE);
mtx_unlock(&controller->lock);
return (-1);
}
if(xpt_create_path(&controller->path, NULL,
cam_sim_path(controller->sim), CAM_TARGET_WILDCARD,
CAM_LUN_WILDCARD) != CAM_REQ_CMP) {
isci_log_message(0, "ISCI", "xpt_create_path....fails\n");
xpt_bus_deregister(cam_sim_path(controller->sim));
cam_sim_free(controller->sim, TRUE);
mtx_unlock(&controller->lock);
return (-1);
}
return (0);
}
void isci_poll(struct cam_sim *sim)
{
struct ISCI_CONTROLLER *controller =
(struct ISCI_CONTROLLER *)cam_sim_softc(sim);
isci_interrupt_poll_handler(controller);
}
void isci_action(struct cam_sim *sim, union ccb *ccb)
{
struct ISCI_CONTROLLER *controller =
(struct ISCI_CONTROLLER *)cam_sim_softc(sim);
switch ( ccb->ccb_h.func_code ) {
case XPT_PATH_INQ:
{
struct ccb_pathinq *cpi = &ccb->cpi;
int bus = cam_sim_bus(sim);
ccb->ccb_h.ccb_sim_ptr = sim;
cpi->version_num = 1;
cpi->hba_inquiry = PI_TAG_ABLE;
cpi->target_sprt = 0;
cpi->hba_misc = PIM_NOBUSRESET | PIM_SEQSCAN |
PIM_UNMAPPED;
cpi->hba_eng_cnt = 0;
cpi->max_target = SCI_MAX_REMOTE_DEVICES - 1;
cpi->max_lun = ISCI_MAX_LUN;
#if __FreeBSD_version >= 800102
cpi->maxio = isci_io_request_get_max_io_size();
#endif
cpi->unit_number = cam_sim_unit(sim);
cpi->bus_id = bus;
cpi->initiator_id = SCI_MAX_REMOTE_DEVICES;
cpi->base_transfer_speed = 300000;
strncpy(cpi->sim_vid, "FreeBSD", SIM_IDLEN);
strncpy(cpi->hba_vid, "Intel Corp.", HBA_IDLEN);
strncpy(cpi->dev_name, cam_sim_name(sim), DEV_IDLEN);
cpi->transport = XPORT_SAS;
cpi->transport_version = 0;
cpi->protocol = PROTO_SCSI;
cpi->protocol_version = SCSI_REV_SPC2;
cpi->ccb_h.status = CAM_REQ_CMP;
xpt_done(ccb);
}
break;
case XPT_GET_TRAN_SETTINGS:
{
struct ccb_trans_settings *general_settings = &ccb->cts;
struct ccb_trans_settings_sas *sas_settings =
&general_settings->xport_specific.sas;
struct ccb_trans_settings_scsi *scsi_settings =
&general_settings->proto_specific.scsi;
struct ISCI_REMOTE_DEVICE *remote_device;
remote_device = controller->remote_device[ccb->ccb_h.target_id];
if (remote_device == NULL) {
ccb->ccb_h.status &= ~CAM_SIM_QUEUED;
ccb->ccb_h.status &= ~CAM_STATUS_MASK;
ccb->ccb_h.status |= CAM_DEV_NOT_THERE;
xpt_done(ccb);
break;
}
general_settings->protocol = PROTO_SCSI;
general_settings->transport = XPORT_SAS;
general_settings->protocol_version = SCSI_REV_SPC2;
general_settings->transport_version = 0;
scsi_settings->valid = CTS_SCSI_VALID_TQ;
scsi_settings->flags = CTS_SCSI_FLAGS_TAG_ENB;
ccb->ccb_h.status &= ~CAM_STATUS_MASK;
ccb->ccb_h.status |= CAM_REQ_CMP;
sas_settings->bitrate =
isci_remote_device_get_bitrate(remote_device);
if (sas_settings->bitrate != 0)
sas_settings->valid = CTS_SAS_VALID_SPEED;
xpt_done(ccb);
}
break;
case XPT_SCSI_IO:
isci_io_request_execute_scsi_io(ccb, controller);
break;
#if __FreeBSD_version >= 900026
case XPT_SMP_IO:
isci_io_request_execute_smp_io(ccb, controller);
break;
#endif
case XPT_SET_TRAN_SETTINGS:
ccb->ccb_h.status &= ~CAM_STATUS_MASK;
ccb->ccb_h.status |= CAM_REQ_CMP;
xpt_done(ccb);
break;
case XPT_CALC_GEOMETRY:
cam_calc_geometry(&ccb->ccg, /*extended*/1);
xpt_done(ccb);
break;
case XPT_RESET_DEV:
{
struct ISCI_REMOTE_DEVICE *remote_device =
controller->remote_device[ccb->ccb_h.target_id];
if (remote_device != NULL)
isci_remote_device_reset(remote_device, ccb);
else {
ccb->ccb_h.status &= ~CAM_SIM_QUEUED;
ccb->ccb_h.status &= ~CAM_STATUS_MASK;
ccb->ccb_h.status |= CAM_DEV_NOT_THERE;
xpt_done(ccb);
}
}
break;
case XPT_RESET_BUS:
ccb->ccb_h.status = CAM_REQ_CMP;
xpt_done(ccb);
break;
default:
isci_log_message(0, "ISCI", "Unhandled func_code 0x%x\n",
ccb->ccb_h.func_code);
ccb->ccb_h.status &= ~CAM_SIM_QUEUED;
ccb->ccb_h.status &= ~CAM_STATUS_MASK;
ccb->ccb_h.status |= CAM_REQ_INVALID;
xpt_done(ccb);
break;
}
}
/*
* Unfortunately, SCIL doesn't cleanly handle retry conditions.
* CAM_REQUEUE_REQ works only when no one is using the pass(4) interface. So
* when SCIL denotes an I/O needs to be retried (typically because of mixing
* tagged/non-tagged ATA commands, or running out of NCQ slots), we queue
* these I/O internally. Once SCIL completes an I/O to this device, or we get
* a ready notification, we will retry the first I/O on the queue.
* Unfortunately, SCIL also doesn't cleanly handle starting the new I/O within
* the context of the completion handler, so we need to retry these I/O after
* the completion handler is done executing.
*/
void
isci_controller_release_queued_ccbs(struct ISCI_CONTROLLER *controller)
{
struct ISCI_REMOTE_DEVICE *dev;
struct ccb_hdr *ccb_h;
int dev_idx;
KASSERT(mtx_owned(&controller->lock), ("controller lock not owned"));
controller->release_queued_ccbs = FALSE;
for (dev_idx = 0;
dev_idx < SCI_MAX_REMOTE_DEVICES;
dev_idx++) {
dev = controller->remote_device[dev_idx];
if (dev != NULL &&
dev->release_queued_ccb == TRUE &&
dev->queued_ccb_in_progress == NULL) {
dev->release_queued_ccb = FALSE;
ccb_h = TAILQ_FIRST(&dev->queued_ccbs);
if (ccb_h == NULL)
continue;
isci_log_message(1, "ISCI", "release %p %x\n", ccb_h,
((union ccb *)ccb_h)->csio.cdb_io.cdb_bytes[0]);
dev->queued_ccb_in_progress = (union ccb *)ccb_h;
isci_io_request_execute_scsi_io(
(union ccb *)ccb_h, controller);
}
}
}