mirror of
https://github.com/torvalds/linux.git
synced 2024-12-30 14:52:05 +00:00
9588d24e36
Previously the code embedded the kernel's test_bit/clear_bit functions in wrappers that accepted u32 parameters. The wrapper cast these parameters to longs before passing them to the kernel's bit functions. This did not work properly on platforms with 64-bit longs. Signed-off-by: Bradley Grove <bgrove@attotech.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
943 lines
24 KiB
C
943 lines
24 KiB
C
/*
|
|
* linux/drivers/scsi/esas2r/esas2r_int.c
|
|
* esas2r interrupt handling
|
|
*
|
|
* Copyright (c) 2001-2013 ATTO Technology, Inc.
|
|
* (mailto:linuxdrivers@attotech.com)
|
|
*/
|
|
/*=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=*/
|
|
/*
|
|
* 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; version 2 of the License.
|
|
*
|
|
* 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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
|
|
*/
|
|
/*=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=*/
|
|
|
|
#include "esas2r.h"
|
|
|
|
/* Local function prototypes */
|
|
static void esas2r_doorbell_interrupt(struct esas2r_adapter *a, u32 doorbell);
|
|
static void esas2r_get_outbound_responses(struct esas2r_adapter *a);
|
|
static void esas2r_process_bus_reset(struct esas2r_adapter *a);
|
|
|
|
/*
|
|
* Poll the adapter for interrupts and service them.
|
|
* This function handles both legacy interrupts and MSI.
|
|
*/
|
|
void esas2r_polled_interrupt(struct esas2r_adapter *a)
|
|
{
|
|
u32 intstat;
|
|
u32 doorbell;
|
|
|
|
esas2r_disable_chip_interrupts(a);
|
|
|
|
intstat = esas2r_read_register_dword(a, MU_INT_STATUS_OUT);
|
|
|
|
if (intstat & MU_INTSTAT_POST_OUT) {
|
|
/* clear the interrupt */
|
|
|
|
esas2r_write_register_dword(a, MU_OUT_LIST_INT_STAT,
|
|
MU_OLIS_INT);
|
|
esas2r_flush_register_dword(a, MU_OUT_LIST_INT_STAT);
|
|
|
|
esas2r_get_outbound_responses(a);
|
|
}
|
|
|
|
if (intstat & MU_INTSTAT_DRBL) {
|
|
doorbell = esas2r_read_register_dword(a, MU_DOORBELL_OUT);
|
|
if (doorbell != 0)
|
|
esas2r_doorbell_interrupt(a, doorbell);
|
|
}
|
|
|
|
esas2r_enable_chip_interrupts(a);
|
|
|
|
if (atomic_read(&a->disable_cnt) == 0)
|
|
esas2r_do_deferred_processes(a);
|
|
}
|
|
|
|
/*
|
|
* Legacy and MSI interrupt handlers. Note that the legacy interrupt handler
|
|
* schedules a TASKLET to process events, whereas the MSI handler just
|
|
* processes interrupt events directly.
|
|
*/
|
|
irqreturn_t esas2r_interrupt(int irq, void *dev_id)
|
|
{
|
|
struct esas2r_adapter *a = (struct esas2r_adapter *)dev_id;
|
|
|
|
if (!esas2r_adapter_interrupt_pending(a))
|
|
return IRQ_NONE;
|
|
|
|
set_bit(AF2_INT_PENDING, &a->flags2);
|
|
esas2r_schedule_tasklet(a);
|
|
|
|
return IRQ_HANDLED;
|
|
}
|
|
|
|
void esas2r_adapter_interrupt(struct esas2r_adapter *a)
|
|
{
|
|
u32 doorbell;
|
|
|
|
if (likely(a->int_stat & MU_INTSTAT_POST_OUT)) {
|
|
/* clear the interrupt */
|
|
esas2r_write_register_dword(a, MU_OUT_LIST_INT_STAT,
|
|
MU_OLIS_INT);
|
|
esas2r_flush_register_dword(a, MU_OUT_LIST_INT_STAT);
|
|
esas2r_get_outbound_responses(a);
|
|
}
|
|
|
|
if (unlikely(a->int_stat & MU_INTSTAT_DRBL)) {
|
|
doorbell = esas2r_read_register_dword(a, MU_DOORBELL_OUT);
|
|
if (doorbell != 0)
|
|
esas2r_doorbell_interrupt(a, doorbell);
|
|
}
|
|
|
|
a->int_mask = ESAS2R_INT_STS_MASK;
|
|
|
|
esas2r_enable_chip_interrupts(a);
|
|
|
|
if (likely(atomic_read(&a->disable_cnt) == 0))
|
|
esas2r_do_deferred_processes(a);
|
|
}
|
|
|
|
irqreturn_t esas2r_msi_interrupt(int irq, void *dev_id)
|
|
{
|
|
struct esas2r_adapter *a = (struct esas2r_adapter *)dev_id;
|
|
u32 intstat;
|
|
u32 doorbell;
|
|
|
|
intstat = esas2r_read_register_dword(a, MU_INT_STATUS_OUT);
|
|
|
|
if (likely(intstat & MU_INTSTAT_POST_OUT)) {
|
|
/* clear the interrupt */
|
|
|
|
esas2r_write_register_dword(a, MU_OUT_LIST_INT_STAT,
|
|
MU_OLIS_INT);
|
|
esas2r_flush_register_dword(a, MU_OUT_LIST_INT_STAT);
|
|
|
|
esas2r_get_outbound_responses(a);
|
|
}
|
|
|
|
if (unlikely(intstat & MU_INTSTAT_DRBL)) {
|
|
doorbell = esas2r_read_register_dword(a, MU_DOORBELL_OUT);
|
|
if (doorbell != 0)
|
|
esas2r_doorbell_interrupt(a, doorbell);
|
|
}
|
|
|
|
/*
|
|
* Work around a chip bug and force a new MSI to be sent if one is
|
|
* still pending.
|
|
*/
|
|
esas2r_disable_chip_interrupts(a);
|
|
esas2r_enable_chip_interrupts(a);
|
|
|
|
if (likely(atomic_read(&a->disable_cnt) == 0))
|
|
esas2r_do_deferred_processes(a);
|
|
|
|
esas2r_do_tasklet_tasks(a);
|
|
|
|
return 1;
|
|
}
|
|
|
|
|
|
|
|
static void esas2r_handle_outbound_rsp_err(struct esas2r_adapter *a,
|
|
struct esas2r_request *rq,
|
|
struct atto_vda_ob_rsp *rsp)
|
|
{
|
|
|
|
/*
|
|
* For I/O requests, only copy the response if an error
|
|
* occurred and setup a callback to do error processing.
|
|
*/
|
|
if (unlikely(rq->req_stat != RS_SUCCESS)) {
|
|
memcpy(&rq->func_rsp, &rsp->func_rsp, sizeof(rsp->func_rsp));
|
|
|
|
if (rq->req_stat == RS_ABORTED) {
|
|
if (rq->timeout > RQ_MAX_TIMEOUT)
|
|
rq->req_stat = RS_TIMEOUT;
|
|
} else if (rq->req_stat == RS_SCSI_ERROR) {
|
|
u8 scsistatus = rq->func_rsp.scsi_rsp.scsi_stat;
|
|
|
|
esas2r_trace("scsistatus: %x", scsistatus);
|
|
|
|
/* Any of these are a good result. */
|
|
if (scsistatus == SAM_STAT_GOOD || scsistatus ==
|
|
SAM_STAT_CONDITION_MET || scsistatus ==
|
|
SAM_STAT_INTERMEDIATE || scsistatus ==
|
|
SAM_STAT_INTERMEDIATE_CONDITION_MET) {
|
|
rq->req_stat = RS_SUCCESS;
|
|
rq->func_rsp.scsi_rsp.scsi_stat =
|
|
SAM_STAT_GOOD;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
static void esas2r_get_outbound_responses(struct esas2r_adapter *a)
|
|
{
|
|
struct atto_vda_ob_rsp *rsp;
|
|
u32 rspput_ptr;
|
|
u32 rspget_ptr;
|
|
struct esas2r_request *rq;
|
|
u32 handle;
|
|
unsigned long flags;
|
|
|
|
LIST_HEAD(comp_list);
|
|
|
|
esas2r_trace_enter();
|
|
|
|
spin_lock_irqsave(&a->queue_lock, flags);
|
|
|
|
/* Get the outbound limit and pointers */
|
|
rspput_ptr = le32_to_cpu(*a->outbound_copy) & MU_OLC_WRT_PTR;
|
|
rspget_ptr = a->last_read;
|
|
|
|
esas2r_trace("rspput_ptr: %x, rspget_ptr: %x", rspput_ptr, rspget_ptr);
|
|
|
|
/* If we don't have anything to process, get out */
|
|
if (unlikely(rspget_ptr == rspput_ptr)) {
|
|
spin_unlock_irqrestore(&a->queue_lock, flags);
|
|
esas2r_trace_exit();
|
|
return;
|
|
}
|
|
|
|
/* Make sure the firmware is healthy */
|
|
if (unlikely(rspput_ptr >= a->list_size)) {
|
|
spin_unlock_irqrestore(&a->queue_lock, flags);
|
|
esas2r_bugon();
|
|
esas2r_local_reset_adapter(a);
|
|
esas2r_trace_exit();
|
|
return;
|
|
}
|
|
|
|
do {
|
|
rspget_ptr++;
|
|
|
|
if (rspget_ptr >= a->list_size)
|
|
rspget_ptr = 0;
|
|
|
|
rsp = (struct atto_vda_ob_rsp *)a->outbound_list_md.virt_addr
|
|
+ rspget_ptr;
|
|
|
|
handle = rsp->handle;
|
|
|
|
/* Verify the handle range */
|
|
if (unlikely(LOWORD(handle) == 0
|
|
|| LOWORD(handle) > num_requests +
|
|
num_ae_requests + 1)) {
|
|
esas2r_bugon();
|
|
continue;
|
|
}
|
|
|
|
/* Get the request for this handle */
|
|
rq = a->req_table[LOWORD(handle)];
|
|
|
|
if (unlikely(rq == NULL || rq->vrq->scsi.handle != handle)) {
|
|
esas2r_bugon();
|
|
continue;
|
|
}
|
|
|
|
list_del(&rq->req_list);
|
|
|
|
/* Get the completion status */
|
|
rq->req_stat = rsp->req_stat;
|
|
|
|
esas2r_trace("handle: %x", handle);
|
|
esas2r_trace("rq: %p", rq);
|
|
esas2r_trace("req_status: %x", rq->req_stat);
|
|
|
|
if (likely(rq->vrq->scsi.function == VDA_FUNC_SCSI)) {
|
|
esas2r_handle_outbound_rsp_err(a, rq, rsp);
|
|
} else {
|
|
/*
|
|
* Copy the outbound completion struct for non-I/O
|
|
* requests.
|
|
*/
|
|
memcpy(&rq->func_rsp, &rsp->func_rsp,
|
|
sizeof(rsp->func_rsp));
|
|
}
|
|
|
|
/* Queue the request for completion. */
|
|
list_add_tail(&rq->comp_list, &comp_list);
|
|
|
|
} while (rspget_ptr != rspput_ptr);
|
|
|
|
a->last_read = rspget_ptr;
|
|
spin_unlock_irqrestore(&a->queue_lock, flags);
|
|
|
|
esas2r_comp_list_drain(a, &comp_list);
|
|
esas2r_trace_exit();
|
|
}
|
|
|
|
/*
|
|
* Perform all deferred processes for the adapter. Deferred
|
|
* processes can only be done while the current interrupt
|
|
* disable_cnt for the adapter is zero.
|
|
*/
|
|
void esas2r_do_deferred_processes(struct esas2r_adapter *a)
|
|
{
|
|
int startreqs = 2;
|
|
struct esas2r_request *rq;
|
|
unsigned long flags;
|
|
|
|
/*
|
|
* startreqs is used to control starting requests
|
|
* that are on the deferred queue
|
|
* = 0 - do not start any requests
|
|
* = 1 - can start discovery requests
|
|
* = 2 - can start any request
|
|
*/
|
|
|
|
if (test_bit(AF_CHPRST_PENDING, &a->flags) ||
|
|
test_bit(AF_FLASHING, &a->flags))
|
|
startreqs = 0;
|
|
else if (test_bit(AF_DISC_PENDING, &a->flags))
|
|
startreqs = 1;
|
|
|
|
atomic_inc(&a->disable_cnt);
|
|
|
|
/* Clear off the completed list to be processed later. */
|
|
|
|
if (esas2r_is_tasklet_pending(a)) {
|
|
esas2r_schedule_tasklet(a);
|
|
|
|
startreqs = 0;
|
|
}
|
|
|
|
/*
|
|
* If we can start requests then traverse the defer queue
|
|
* looking for requests to start or complete
|
|
*/
|
|
if (startreqs && !list_empty(&a->defer_list)) {
|
|
LIST_HEAD(comp_list);
|
|
struct list_head *element, *next;
|
|
|
|
spin_lock_irqsave(&a->queue_lock, flags);
|
|
|
|
list_for_each_safe(element, next, &a->defer_list) {
|
|
rq = list_entry(element, struct esas2r_request,
|
|
req_list);
|
|
|
|
if (rq->req_stat != RS_PENDING) {
|
|
list_del(element);
|
|
list_add_tail(&rq->comp_list, &comp_list);
|
|
}
|
|
/*
|
|
* Process discovery and OS requests separately. We
|
|
* can't hold up discovery requests when discovery is
|
|
* pending. In general, there may be different sets of
|
|
* conditions for starting different types of requests.
|
|
*/
|
|
else if (rq->req_type == RT_DISC_REQ) {
|
|
list_del(element);
|
|
esas2r_disc_local_start_request(a, rq);
|
|
} else if (startreqs == 2) {
|
|
list_del(element);
|
|
esas2r_local_start_request(a, rq);
|
|
|
|
/*
|
|
* Flashing could have been set by last local
|
|
* start
|
|
*/
|
|
if (test_bit(AF_FLASHING, &a->flags))
|
|
break;
|
|
}
|
|
}
|
|
|
|
spin_unlock_irqrestore(&a->queue_lock, flags);
|
|
esas2r_comp_list_drain(a, &comp_list);
|
|
}
|
|
|
|
atomic_dec(&a->disable_cnt);
|
|
}
|
|
|
|
/*
|
|
* Process an adapter reset (or one that is about to happen)
|
|
* by making sure all outstanding requests are completed that
|
|
* haven't been already.
|
|
*/
|
|
void esas2r_process_adapter_reset(struct esas2r_adapter *a)
|
|
{
|
|
struct esas2r_request *rq = &a->general_req;
|
|
unsigned long flags;
|
|
struct esas2r_disc_context *dc;
|
|
|
|
LIST_HEAD(comp_list);
|
|
struct list_head *element;
|
|
|
|
esas2r_trace_enter();
|
|
|
|
spin_lock_irqsave(&a->queue_lock, flags);
|
|
|
|
/* abort the active discovery, if any. */
|
|
|
|
if (rq->interrupt_cx) {
|
|
dc = (struct esas2r_disc_context *)rq->interrupt_cx;
|
|
|
|
dc->disc_evt = 0;
|
|
|
|
clear_bit(AF_DISC_IN_PROG, &a->flags);
|
|
}
|
|
|
|
/*
|
|
* just clear the interrupt callback for now. it will be dequeued if
|
|
* and when we find it on the active queue and we don't want the
|
|
* callback called. also set the dummy completion callback in case we
|
|
* were doing an I/O request.
|
|
*/
|
|
|
|
rq->interrupt_cx = NULL;
|
|
rq->interrupt_cb = NULL;
|
|
|
|
rq->comp_cb = esas2r_dummy_complete;
|
|
|
|
/* Reset the read and write pointers */
|
|
|
|
*a->outbound_copy =
|
|
a->last_write =
|
|
a->last_read = a->list_size - 1;
|
|
|
|
set_bit(AF_COMM_LIST_TOGGLE, &a->flags);
|
|
|
|
/* Kill all the requests on the active list */
|
|
list_for_each(element, &a->defer_list) {
|
|
rq = list_entry(element, struct esas2r_request, req_list);
|
|
|
|
if (rq->req_stat == RS_STARTED)
|
|
if (esas2r_ioreq_aborted(a, rq, RS_ABORTED))
|
|
list_add_tail(&rq->comp_list, &comp_list);
|
|
}
|
|
|
|
spin_unlock_irqrestore(&a->queue_lock, flags);
|
|
esas2r_comp_list_drain(a, &comp_list);
|
|
esas2r_process_bus_reset(a);
|
|
esas2r_trace_exit();
|
|
}
|
|
|
|
static void esas2r_process_bus_reset(struct esas2r_adapter *a)
|
|
{
|
|
struct esas2r_request *rq;
|
|
struct list_head *element;
|
|
unsigned long flags;
|
|
|
|
LIST_HEAD(comp_list);
|
|
|
|
esas2r_trace_enter();
|
|
|
|
esas2r_hdebug("reset detected");
|
|
|
|
spin_lock_irqsave(&a->queue_lock, flags);
|
|
|
|
/* kill all the requests on the deferred queue */
|
|
list_for_each(element, &a->defer_list) {
|
|
rq = list_entry(element, struct esas2r_request, req_list);
|
|
if (esas2r_ioreq_aborted(a, rq, RS_ABORTED))
|
|
list_add_tail(&rq->comp_list, &comp_list);
|
|
}
|
|
|
|
spin_unlock_irqrestore(&a->queue_lock, flags);
|
|
|
|
esas2r_comp_list_drain(a, &comp_list);
|
|
|
|
if (atomic_read(&a->disable_cnt) == 0)
|
|
esas2r_do_deferred_processes(a);
|
|
|
|
clear_bit(AF_OS_RESET, &a->flags);
|
|
|
|
esas2r_trace_exit();
|
|
}
|
|
|
|
static void esas2r_chip_rst_needed_during_tasklet(struct esas2r_adapter *a)
|
|
{
|
|
|
|
clear_bit(AF_CHPRST_NEEDED, &a->flags);
|
|
clear_bit(AF_BUSRST_NEEDED, &a->flags);
|
|
clear_bit(AF_BUSRST_DETECTED, &a->flags);
|
|
clear_bit(AF_BUSRST_PENDING, &a->flags);
|
|
/*
|
|
* Make sure we don't get attempt more than 3 resets
|
|
* when the uptime between resets does not exceed one
|
|
* minute. This will stop any situation where there is
|
|
* really something wrong with the hardware. The way
|
|
* this works is that we start with uptime ticks at 0.
|
|
* Each time we do a reset, we add 20 seconds worth to
|
|
* the count. Each time a timer tick occurs, as long
|
|
* as a chip reset is not pending, we decrement the
|
|
* tick count. If the uptime ticks ever gets to 60
|
|
* seconds worth, we disable the adapter from that
|
|
* point forward. Three strikes, you're out.
|
|
*/
|
|
if (!esas2r_is_adapter_present(a) || (a->chip_uptime >=
|
|
ESAS2R_CHP_UPTIME_MAX)) {
|
|
esas2r_hdebug("*** adapter disabled ***");
|
|
|
|
/*
|
|
* Ok, some kind of hard failure. Make sure we
|
|
* exit this loop with chip interrupts
|
|
* permanently disabled so we don't lock up the
|
|
* entire system. Also flag degraded mode to
|
|
* prevent the heartbeat from trying to recover.
|
|
*/
|
|
|
|
set_bit(AF_DEGRADED_MODE, &a->flags);
|
|
set_bit(AF_DISABLED, &a->flags);
|
|
clear_bit(AF_CHPRST_PENDING, &a->flags);
|
|
clear_bit(AF_DISC_PENDING, &a->flags);
|
|
|
|
esas2r_disable_chip_interrupts(a);
|
|
a->int_mask = 0;
|
|
esas2r_process_adapter_reset(a);
|
|
|
|
esas2r_log(ESAS2R_LOG_CRIT,
|
|
"Adapter disabled because of hardware failure");
|
|
} else {
|
|
bool alrdyrst = test_and_set_bit(AF_CHPRST_STARTED, &a->flags);
|
|
|
|
if (!alrdyrst)
|
|
/*
|
|
* Only disable interrupts if this is
|
|
* the first reset attempt.
|
|
*/
|
|
esas2r_disable_chip_interrupts(a);
|
|
|
|
if ((test_bit(AF_POWER_MGT, &a->flags)) &&
|
|
!test_bit(AF_FIRST_INIT, &a->flags) && !alrdyrst) {
|
|
/*
|
|
* Don't reset the chip on the first
|
|
* deferred power up attempt.
|
|
*/
|
|
} else {
|
|
esas2r_hdebug("*** resetting chip ***");
|
|
esas2r_reset_chip(a);
|
|
}
|
|
|
|
/* Kick off the reinitialization */
|
|
a->chip_uptime += ESAS2R_CHP_UPTIME_CNT;
|
|
a->chip_init_time = jiffies_to_msecs(jiffies);
|
|
if (!test_bit(AF_POWER_MGT, &a->flags)) {
|
|
esas2r_process_adapter_reset(a);
|
|
|
|
if (!alrdyrst) {
|
|
/* Remove devices now that I/O is cleaned up. */
|
|
a->prev_dev_cnt =
|
|
esas2r_targ_db_get_tgt_cnt(a);
|
|
esas2r_targ_db_remove_all(a, false);
|
|
}
|
|
}
|
|
|
|
a->int_mask = 0;
|
|
}
|
|
}
|
|
|
|
static void esas2r_handle_chip_rst_during_tasklet(struct esas2r_adapter *a)
|
|
{
|
|
while (test_bit(AF_CHPRST_DETECTED, &a->flags)) {
|
|
/*
|
|
* Balance the enable in esas2r_initadapter_hw.
|
|
* Esas2r_power_down already took care of it for power
|
|
* management.
|
|
*/
|
|
if (!test_bit(AF_DEGRADED_MODE, &a->flags) &&
|
|
!test_bit(AF_POWER_MGT, &a->flags))
|
|
esas2r_disable_chip_interrupts(a);
|
|
|
|
/* Reinitialize the chip. */
|
|
esas2r_check_adapter(a);
|
|
esas2r_init_adapter_hw(a, 0);
|
|
|
|
if (test_bit(AF_CHPRST_NEEDED, &a->flags))
|
|
break;
|
|
|
|
if (test_bit(AF_POWER_MGT, &a->flags)) {
|
|
/* Recovery from power management. */
|
|
if (test_bit(AF_FIRST_INIT, &a->flags)) {
|
|
/* Chip reset during normal power up */
|
|
esas2r_log(ESAS2R_LOG_CRIT,
|
|
"The firmware was reset during a normal power-up sequence");
|
|
} else {
|
|
/* Deferred power up complete. */
|
|
clear_bit(AF_POWER_MGT, &a->flags);
|
|
esas2r_send_reset_ae(a, true);
|
|
}
|
|
} else {
|
|
/* Recovery from online chip reset. */
|
|
if (test_bit(AF_FIRST_INIT, &a->flags)) {
|
|
/* Chip reset during driver load */
|
|
} else {
|
|
/* Chip reset after driver load */
|
|
esas2r_send_reset_ae(a, false);
|
|
}
|
|
|
|
esas2r_log(ESAS2R_LOG_CRIT,
|
|
"Recovering from a chip reset while the chip was online");
|
|
}
|
|
|
|
clear_bit(AF_CHPRST_STARTED, &a->flags);
|
|
esas2r_enable_chip_interrupts(a);
|
|
|
|
/*
|
|
* Clear this flag last! this indicates that the chip has been
|
|
* reset already during initialization.
|
|
*/
|
|
clear_bit(AF_CHPRST_DETECTED, &a->flags);
|
|
}
|
|
}
|
|
|
|
|
|
/* Perform deferred tasks when chip interrupts are disabled */
|
|
void esas2r_do_tasklet_tasks(struct esas2r_adapter *a)
|
|
{
|
|
|
|
if (test_bit(AF_CHPRST_NEEDED, &a->flags) ||
|
|
test_bit(AF_CHPRST_DETECTED, &a->flags)) {
|
|
if (test_bit(AF_CHPRST_NEEDED, &a->flags))
|
|
esas2r_chip_rst_needed_during_tasklet(a);
|
|
|
|
esas2r_handle_chip_rst_during_tasklet(a);
|
|
}
|
|
|
|
if (test_bit(AF_BUSRST_NEEDED, &a->flags)) {
|
|
esas2r_hdebug("hard resetting bus");
|
|
|
|
clear_bit(AF_BUSRST_NEEDED, &a->flags);
|
|
|
|
if (test_bit(AF_FLASHING, &a->flags))
|
|
set_bit(AF_BUSRST_DETECTED, &a->flags);
|
|
else
|
|
esas2r_write_register_dword(a, MU_DOORBELL_IN,
|
|
DRBL_RESET_BUS);
|
|
}
|
|
|
|
if (test_bit(AF_BUSRST_DETECTED, &a->flags)) {
|
|
esas2r_process_bus_reset(a);
|
|
|
|
esas2r_log_dev(ESAS2R_LOG_WARN,
|
|
&(a->host->shost_gendev),
|
|
"scsi_report_bus_reset() called");
|
|
|
|
scsi_report_bus_reset(a->host, 0);
|
|
|
|
clear_bit(AF_BUSRST_DETECTED, &a->flags);
|
|
clear_bit(AF_BUSRST_PENDING, &a->flags);
|
|
|
|
esas2r_log(ESAS2R_LOG_WARN, "Bus reset complete");
|
|
}
|
|
|
|
if (test_bit(AF_PORT_CHANGE, &a->flags)) {
|
|
clear_bit(AF_PORT_CHANGE, &a->flags);
|
|
|
|
esas2r_targ_db_report_changes(a);
|
|
}
|
|
|
|
if (atomic_read(&a->disable_cnt) == 0)
|
|
esas2r_do_deferred_processes(a);
|
|
}
|
|
|
|
static void esas2r_doorbell_interrupt(struct esas2r_adapter *a, u32 doorbell)
|
|
{
|
|
if (!(doorbell & DRBL_FORCE_INT)) {
|
|
esas2r_trace_enter();
|
|
esas2r_trace("doorbell: %x", doorbell);
|
|
}
|
|
|
|
/* First clear the doorbell bits */
|
|
esas2r_write_register_dword(a, MU_DOORBELL_OUT, doorbell);
|
|
|
|
if (doorbell & DRBL_RESET_BUS)
|
|
set_bit(AF_BUSRST_DETECTED, &a->flags);
|
|
|
|
if (doorbell & DRBL_FORCE_INT)
|
|
clear_bit(AF_HEARTBEAT, &a->flags);
|
|
|
|
if (doorbell & DRBL_PANIC_REASON_MASK) {
|
|
esas2r_hdebug("*** Firmware Panic ***");
|
|
esas2r_log(ESAS2R_LOG_CRIT, "The firmware has panicked");
|
|
}
|
|
|
|
if (doorbell & DRBL_FW_RESET) {
|
|
set_bit(AF2_COREDUMP_AVAIL, &a->flags2);
|
|
esas2r_local_reset_adapter(a);
|
|
}
|
|
|
|
if (!(doorbell & DRBL_FORCE_INT))
|
|
esas2r_trace_exit();
|
|
}
|
|
|
|
void esas2r_force_interrupt(struct esas2r_adapter *a)
|
|
{
|
|
esas2r_write_register_dword(a, MU_DOORBELL_IN, DRBL_FORCE_INT |
|
|
DRBL_DRV_VER);
|
|
}
|
|
|
|
|
|
static void esas2r_lun_event(struct esas2r_adapter *a, union atto_vda_ae *ae,
|
|
u16 target, u32 length)
|
|
{
|
|
struct esas2r_target *t = a->targetdb + target;
|
|
u32 cplen = length;
|
|
unsigned long flags;
|
|
|
|
if (cplen > sizeof(t->lu_event))
|
|
cplen = sizeof(t->lu_event);
|
|
|
|
esas2r_trace("ae->lu.dwevent: %x", ae->lu.dwevent);
|
|
esas2r_trace("ae->lu.bystate: %x", ae->lu.bystate);
|
|
|
|
spin_lock_irqsave(&a->mem_lock, flags);
|
|
|
|
t->new_target_state = TS_INVALID;
|
|
|
|
if (ae->lu.dwevent & VDAAE_LU_LOST) {
|
|
t->new_target_state = TS_NOT_PRESENT;
|
|
} else {
|
|
switch (ae->lu.bystate) {
|
|
case VDAAE_LU_NOT_PRESENT:
|
|
case VDAAE_LU_OFFLINE:
|
|
case VDAAE_LU_DELETED:
|
|
case VDAAE_LU_FACTORY_DISABLED:
|
|
t->new_target_state = TS_NOT_PRESENT;
|
|
break;
|
|
|
|
case VDAAE_LU_ONLINE:
|
|
case VDAAE_LU_DEGRADED:
|
|
t->new_target_state = TS_PRESENT;
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (t->new_target_state != TS_INVALID) {
|
|
memcpy(&t->lu_event, &ae->lu, cplen);
|
|
|
|
esas2r_disc_queue_event(a, DCDE_DEV_CHANGE);
|
|
}
|
|
|
|
spin_unlock_irqrestore(&a->mem_lock, flags);
|
|
}
|
|
|
|
|
|
|
|
void esas2r_ae_complete(struct esas2r_adapter *a, struct esas2r_request *rq)
|
|
{
|
|
union atto_vda_ae *ae =
|
|
(union atto_vda_ae *)rq->vda_rsp_data->ae_data.event_data;
|
|
u32 length = le32_to_cpu(rq->func_rsp.ae_rsp.length);
|
|
union atto_vda_ae *last =
|
|
(union atto_vda_ae *)(rq->vda_rsp_data->ae_data.event_data
|
|
+ length);
|
|
|
|
esas2r_trace_enter();
|
|
esas2r_trace("length: %d", length);
|
|
|
|
if (length > sizeof(struct atto_vda_ae_data)
|
|
|| (length & 3) != 0
|
|
|| length == 0) {
|
|
esas2r_log(ESAS2R_LOG_WARN,
|
|
"The AE request response length (%p) is too long: %d",
|
|
rq, length);
|
|
|
|
esas2r_hdebug("aereq->length (0x%x) too long", length);
|
|
esas2r_bugon();
|
|
|
|
last = ae;
|
|
}
|
|
|
|
while (ae < last) {
|
|
u16 target;
|
|
|
|
esas2r_trace("ae: %p", ae);
|
|
esas2r_trace("ae->hdr: %p", &(ae->hdr));
|
|
|
|
length = ae->hdr.bylength;
|
|
|
|
if (length > (u32)((u8 *)last - (u8 *)ae)
|
|
|| (length & 3) != 0
|
|
|| length == 0) {
|
|
esas2r_log(ESAS2R_LOG_CRIT,
|
|
"the async event length is invalid (%p): %d",
|
|
ae, length);
|
|
|
|
esas2r_hdebug("ae->hdr.length (0x%x) invalid", length);
|
|
esas2r_bugon();
|
|
|
|
break;
|
|
}
|
|
|
|
esas2r_nuxi_ae_data(ae);
|
|
|
|
esas2r_queue_fw_event(a, fw_event_vda_ae, ae,
|
|
sizeof(union atto_vda_ae));
|
|
|
|
switch (ae->hdr.bytype) {
|
|
case VDAAE_HDR_TYPE_RAID:
|
|
|
|
if (ae->raid.dwflags & (VDAAE_GROUP_STATE
|
|
| VDAAE_RBLD_STATE
|
|
| VDAAE_MEMBER_CHG
|
|
| VDAAE_PART_CHG)) {
|
|
esas2r_log(ESAS2R_LOG_INFO,
|
|
"RAID event received - name:%s rebuild_state:%d group_state:%d",
|
|
ae->raid.acname,
|
|
ae->raid.byrebuild_state,
|
|
ae->raid.bygroup_state);
|
|
}
|
|
|
|
break;
|
|
|
|
case VDAAE_HDR_TYPE_LU:
|
|
esas2r_log(ESAS2R_LOG_INFO,
|
|
"LUN event received: event:%d target_id:%d LUN:%d state:%d",
|
|
ae->lu.dwevent,
|
|
ae->lu.id.tgtlun.wtarget_id,
|
|
ae->lu.id.tgtlun.bylun,
|
|
ae->lu.bystate);
|
|
|
|
target = ae->lu.id.tgtlun.wtarget_id;
|
|
|
|
if (target < ESAS2R_MAX_TARGETS)
|
|
esas2r_lun_event(a, ae, target, length);
|
|
|
|
break;
|
|
|
|
case VDAAE_HDR_TYPE_DISK:
|
|
esas2r_log(ESAS2R_LOG_INFO, "Disk event received");
|
|
break;
|
|
|
|
default:
|
|
|
|
/* Silently ignore the rest and let the apps deal with
|
|
* them.
|
|
*/
|
|
|
|
break;
|
|
}
|
|
|
|
ae = (union atto_vda_ae *)((u8 *)ae + length);
|
|
}
|
|
|
|
/* Now requeue it. */
|
|
esas2r_start_ae_request(a, rq);
|
|
esas2r_trace_exit();
|
|
}
|
|
|
|
/* Send an asynchronous event for a chip reset or power management. */
|
|
void esas2r_send_reset_ae(struct esas2r_adapter *a, bool pwr_mgt)
|
|
{
|
|
struct atto_vda_ae_hdr ae;
|
|
|
|
if (pwr_mgt)
|
|
ae.bytype = VDAAE_HDR_TYPE_PWRMGT;
|
|
else
|
|
ae.bytype = VDAAE_HDR_TYPE_RESET;
|
|
|
|
ae.byversion = VDAAE_HDR_VER_0;
|
|
ae.byflags = 0;
|
|
ae.bylength = (u8)sizeof(struct atto_vda_ae_hdr);
|
|
|
|
if (pwr_mgt)
|
|
esas2r_hdebug("*** sending power management AE ***");
|
|
else
|
|
esas2r_hdebug("*** sending reset AE ***");
|
|
|
|
esas2r_queue_fw_event(a, fw_event_vda_ae, &ae,
|
|
sizeof(union atto_vda_ae));
|
|
}
|
|
|
|
void esas2r_dummy_complete(struct esas2r_adapter *a, struct esas2r_request *rq)
|
|
{}
|
|
|
|
static void esas2r_check_req_rsp_sense(struct esas2r_adapter *a,
|
|
struct esas2r_request *rq)
|
|
{
|
|
u8 snslen, snslen2;
|
|
|
|
snslen = snslen2 = rq->func_rsp.scsi_rsp.sense_len;
|
|
|
|
if (snslen > rq->sense_len)
|
|
snslen = rq->sense_len;
|
|
|
|
if (snslen) {
|
|
if (rq->sense_buf)
|
|
memcpy(rq->sense_buf, rq->data_buf, snslen);
|
|
else
|
|
rq->sense_buf = (u8 *)rq->data_buf;
|
|
|
|
/* See about possible sense data */
|
|
if (snslen2 > 0x0c) {
|
|
u8 *s = (u8 *)rq->data_buf;
|
|
|
|
esas2r_trace_enter();
|
|
|
|
/* Report LUNS data has changed */
|
|
if (s[0x0c] == 0x3f && s[0x0d] == 0x0E) {
|
|
esas2r_trace("rq->target_id: %d",
|
|
rq->target_id);
|
|
esas2r_target_state_changed(a, rq->target_id,
|
|
TS_LUN_CHANGE);
|
|
}
|
|
|
|
esas2r_trace("add_sense_key=%x", s[0x0c]);
|
|
esas2r_trace("add_sense_qual=%x", s[0x0d]);
|
|
esas2r_trace_exit();
|
|
}
|
|
}
|
|
|
|
rq->sense_len = snslen;
|
|
}
|
|
|
|
|
|
void esas2r_complete_request(struct esas2r_adapter *a,
|
|
struct esas2r_request *rq)
|
|
{
|
|
if (rq->vrq->scsi.function == VDA_FUNC_FLASH
|
|
&& rq->vrq->flash.sub_func == VDA_FLASH_COMMIT)
|
|
clear_bit(AF_FLASHING, &a->flags);
|
|
|
|
/* See if we setup a callback to do special processing */
|
|
|
|
if (rq->interrupt_cb) {
|
|
(*rq->interrupt_cb)(a, rq);
|
|
|
|
if (rq->req_stat == RS_PENDING) {
|
|
esas2r_start_request(a, rq);
|
|
return;
|
|
}
|
|
}
|
|
|
|
if (likely(rq->vrq->scsi.function == VDA_FUNC_SCSI)
|
|
&& unlikely(rq->req_stat != RS_SUCCESS)) {
|
|
esas2r_check_req_rsp_sense(a, rq);
|
|
esas2r_log_request_failure(a, rq);
|
|
}
|
|
|
|
(*rq->comp_cb)(a, rq);
|
|
}
|