summaryrefslogtreecommitdiffstats
path: root/drivers/scsi/esas2r/esas2r_io.c
diff options
context:
space:
mode:
Diffstat (limited to '')
-rw-r--r--drivers/scsi/esas2r/esas2r_io.c877
1 files changed, 877 insertions, 0 deletions
diff --git a/drivers/scsi/esas2r/esas2r_io.c b/drivers/scsi/esas2r/esas2r_io.c
new file mode 100644
index 000000000..a8df916cd
--- /dev/null
+++ b/drivers/scsi/esas2r/esas2r_io.c
@@ -0,0 +1,877 @@
+/*
+ * linux/drivers/scsi/esas2r/esas2r_io.c
+ * For use with ATTO ExpressSAS R6xx SAS/SATA RAID controllers
+ *
+ * Copyright (c) 2001-2013 ATTO Technology, Inc.
+ * (mailto:linuxdrivers@attotech.com)mpt3sas/mpt3sas_trigger_diag.
+ *
+ * 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"
+
+void esas2r_start_request(struct esas2r_adapter *a, struct esas2r_request *rq)
+{
+ struct esas2r_target *t = NULL;
+ struct esas2r_request *startrq = rq;
+ unsigned long flags;
+
+ if (unlikely(test_bit(AF_DEGRADED_MODE, &a->flags) ||
+ test_bit(AF_POWER_DOWN, &a->flags))) {
+ if (rq->vrq->scsi.function == VDA_FUNC_SCSI)
+ rq->req_stat = RS_SEL2;
+ else
+ rq->req_stat = RS_DEGRADED;
+ } else if (likely(rq->vrq->scsi.function == VDA_FUNC_SCSI)) {
+ t = a->targetdb + rq->target_id;
+
+ if (unlikely(t >= a->targetdb_end
+ || !(t->flags & TF_USED))) {
+ rq->req_stat = RS_SEL;
+ } else {
+ /* copy in the target ID. */
+ rq->vrq->scsi.target_id = cpu_to_le16(t->virt_targ_id);
+
+ /*
+ * Test if we want to report RS_SEL for missing target.
+ * Note that if AF_DISC_PENDING is set than this will
+ * go on the defer queue.
+ */
+ if (unlikely(t->target_state != TS_PRESENT &&
+ !test_bit(AF_DISC_PENDING, &a->flags)))
+ rq->req_stat = RS_SEL;
+ }
+ }
+
+ if (unlikely(rq->req_stat != RS_PENDING)) {
+ esas2r_complete_request(a, rq);
+ return;
+ }
+
+ esas2r_trace("rq=%p", rq);
+ esas2r_trace("rq->vrq->scsi.handle=%x", rq->vrq->scsi.handle);
+
+ if (rq->vrq->scsi.function == VDA_FUNC_SCSI) {
+ esas2r_trace("rq->target_id=%d", rq->target_id);
+ esas2r_trace("rq->vrq->scsi.flags=%x", rq->vrq->scsi.flags);
+ }
+
+ spin_lock_irqsave(&a->queue_lock, flags);
+
+ if (likely(list_empty(&a->defer_list) &&
+ !test_bit(AF_CHPRST_PENDING, &a->flags) &&
+ !test_bit(AF_FLASHING, &a->flags) &&
+ !test_bit(AF_DISC_PENDING, &a->flags)))
+ esas2r_local_start_request(a, startrq);
+ else
+ list_add_tail(&startrq->req_list, &a->defer_list);
+
+ spin_unlock_irqrestore(&a->queue_lock, flags);
+}
+
+/*
+ * Starts the specified request. all requests have RS_PENDING set when this
+ * routine is called. The caller is usually esas2r_start_request, but
+ * esas2r_do_deferred_processes will start request that are deferred.
+ *
+ * The caller must ensure that requests can be started.
+ *
+ * esas2r_start_request will defer a request if there are already requests
+ * waiting or there is a chip reset pending. once the reset condition clears,
+ * esas2r_do_deferred_processes will call this function to start the request.
+ *
+ * When a request is started, it is placed on the active list and queued to
+ * the controller.
+ */
+void esas2r_local_start_request(struct esas2r_adapter *a,
+ struct esas2r_request *rq)
+{
+ esas2r_trace_enter();
+ esas2r_trace("rq=%p", rq);
+ esas2r_trace("rq->vrq:%p", rq->vrq);
+ esas2r_trace("rq->vrq_md->phys_addr:%x", rq->vrq_md->phys_addr);
+
+ if (unlikely(rq->vrq->scsi.function == VDA_FUNC_FLASH
+ && rq->vrq->flash.sub_func == VDA_FLASH_COMMIT))
+ set_bit(AF_FLASHING, &a->flags);
+
+ list_add_tail(&rq->req_list, &a->active_list);
+ esas2r_start_vda_request(a, rq);
+ esas2r_trace_exit();
+ return;
+}
+
+void esas2r_start_vda_request(struct esas2r_adapter *a,
+ struct esas2r_request *rq)
+{
+ struct esas2r_inbound_list_source_entry *element;
+ u32 dw;
+
+ rq->req_stat = RS_STARTED;
+ /*
+ * Calculate the inbound list entry location and the current state of
+ * toggle bit.
+ */
+ a->last_write++;
+ if (a->last_write >= a->list_size) {
+ a->last_write = 0;
+ /* update the toggle bit */
+ if (test_bit(AF_COMM_LIST_TOGGLE, &a->flags))
+ clear_bit(AF_COMM_LIST_TOGGLE, &a->flags);
+ else
+ set_bit(AF_COMM_LIST_TOGGLE, &a->flags);
+ }
+
+ element =
+ (struct esas2r_inbound_list_source_entry *)a->inbound_list_md.
+ virt_addr
+ + a->last_write;
+
+ /* Set the VDA request size if it was never modified */
+ if (rq->vda_req_sz == RQ_SIZE_DEFAULT)
+ rq->vda_req_sz = (u16)(a->max_vdareq_size / sizeof(u32));
+
+ element->address = cpu_to_le64(rq->vrq_md->phys_addr);
+ element->length = cpu_to_le32(rq->vda_req_sz);
+
+ /* Update the write pointer */
+ dw = a->last_write;
+
+ if (test_bit(AF_COMM_LIST_TOGGLE, &a->flags))
+ dw |= MU_ILW_TOGGLE;
+
+ esas2r_trace("rq->vrq->scsi.handle:%x", rq->vrq->scsi.handle);
+ esas2r_trace("dw:%x", dw);
+ esas2r_trace("rq->vda_req_sz:%x", rq->vda_req_sz);
+ esas2r_write_register_dword(a, MU_IN_LIST_WRITE, dw);
+}
+
+/*
+ * Build the scatter/gather list for an I/O request according to the
+ * specifications placed in the s/g context. The caller must initialize
+ * context prior to the initial call by calling esas2r_sgc_init().
+ */
+bool esas2r_build_sg_list_sge(struct esas2r_adapter *a,
+ struct esas2r_sg_context *sgc)
+{
+ struct esas2r_request *rq = sgc->first_req;
+ union atto_vda_req *vrq = rq->vrq;
+
+ while (sgc->length) {
+ u32 rem = 0;
+ u64 addr;
+ u32 len;
+
+ len = (*sgc->get_phys_addr)(sgc, &addr);
+
+ if (unlikely(len == 0))
+ return false;
+
+ /* if current length is more than what's left, stop there */
+ if (unlikely(len > sgc->length))
+ len = sgc->length;
+
+another_entry:
+ /* limit to a round number less than the maximum length */
+ if (len > SGE_LEN_MAX) {
+ /*
+ * Save the remainder of the split. Whenever we limit
+ * an entry we come back around to build entries out
+ * of the leftover. We do this to prevent multiple
+ * calls to the get_phys_addr() function for an SGE
+ * that is too large.
+ */
+ rem = len - SGE_LEN_MAX;
+ len = SGE_LEN_MAX;
+ }
+
+ /* See if we need to allocate a new SGL */
+ if (unlikely(sgc->sge.a64.curr > sgc->sge.a64.limit)) {
+ u8 sgelen;
+ struct esas2r_mem_desc *sgl;
+
+ /*
+ * If no SGls are available, return failure. The
+ * caller can call us later with the current context
+ * to pick up here.
+ */
+ sgl = esas2r_alloc_sgl(a);
+
+ if (unlikely(sgl == NULL))
+ return false;
+
+ /* Calculate the length of the last SGE filled in */
+ sgelen = (u8)((u8 *)sgc->sge.a64.curr
+ - (u8 *)sgc->sge.a64.last);
+
+ /*
+ * Copy the last SGE filled in to the first entry of
+ * the new SGL to make room for the chain entry.
+ */
+ memcpy(sgl->virt_addr, sgc->sge.a64.last, sgelen);
+
+ /* Figure out the new curr pointer in the new segment */
+ sgc->sge.a64.curr =
+ (struct atto_vda_sge *)((u8 *)sgl->virt_addr +
+ sgelen);
+
+ /* Set the limit pointer and build the chain entry */
+ sgc->sge.a64.limit =
+ (struct atto_vda_sge *)((u8 *)sgl->virt_addr
+ + sgl_page_size
+ - sizeof(struct
+ atto_vda_sge));
+ sgc->sge.a64.last->length = cpu_to_le32(
+ SGE_CHAIN | SGE_ADDR_64);
+ sgc->sge.a64.last->address =
+ cpu_to_le64(sgl->phys_addr);
+
+ /*
+ * Now, if there was a previous chain entry, then
+ * update it to contain the length of this segment
+ * and size of this chain. otherwise this is the
+ * first SGL, so set the chain_offset in the request.
+ */
+ if (sgc->sge.a64.chain) {
+ sgc->sge.a64.chain->length |=
+ cpu_to_le32(
+ ((u8 *)(sgc->sge.a64.
+ last + 1)
+ - (u8 *)rq->sg_table->
+ virt_addr)
+ + sizeof(struct atto_vda_sge) *
+ LOBIT(SGE_CHAIN_SZ));
+ } else {
+ vrq->scsi.chain_offset = (u8)
+ ((u8 *)sgc->
+ sge.a64.last -
+ (u8 *)vrq);
+
+ /*
+ * This is the first SGL, so set the
+ * chain_offset and the VDA request size in
+ * the request.
+ */
+ rq->vda_req_sz =
+ (vrq->scsi.chain_offset +
+ sizeof(struct atto_vda_sge) +
+ 3)
+ / sizeof(u32);
+ }
+
+ /*
+ * Remember this so when we get a new SGL filled in we
+ * can update the length of this chain entry.
+ */
+ sgc->sge.a64.chain = sgc->sge.a64.last;
+
+ /* Now link the new SGL onto the primary request. */
+ list_add(&sgl->next_desc, &rq->sg_table_head);
+ }
+
+ /* Update last one filled in */
+ sgc->sge.a64.last = sgc->sge.a64.curr;
+
+ /* Build the new SGE and update the S/G context */
+ sgc->sge.a64.curr->length = cpu_to_le32(SGE_ADDR_64 | len);
+ sgc->sge.a64.curr->address = cpu_to_le32(addr);
+ sgc->sge.a64.curr++;
+ sgc->cur_offset += len;
+ sgc->length -= len;
+
+ /*
+ * Check if we previously split an entry. If so we have to
+ * pick up where we left off.
+ */
+ if (rem) {
+ addr += len;
+ len = rem;
+ rem = 0;
+ goto another_entry;
+ }
+ }
+
+ /* Mark the end of the SGL */
+ sgc->sge.a64.last->length |= cpu_to_le32(SGE_LAST);
+
+ /*
+ * If there was a previous chain entry, update the length to indicate
+ * the length of this last segment.
+ */
+ if (sgc->sge.a64.chain) {
+ sgc->sge.a64.chain->length |= cpu_to_le32(
+ ((u8 *)(sgc->sge.a64.curr) -
+ (u8 *)rq->sg_table->virt_addr));
+ } else {
+ u16 reqsize;
+
+ /*
+ * The entire VDA request was not used so lets
+ * set the size of the VDA request to be DMA'd
+ */
+ reqsize =
+ ((u16)((u8 *)sgc->sge.a64.last - (u8 *)vrq)
+ + sizeof(struct atto_vda_sge) + 3) / sizeof(u32);
+
+ /*
+ * Only update the request size if it is bigger than what is
+ * already there. We can come in here twice for some management
+ * commands.
+ */
+ if (reqsize > rq->vda_req_sz)
+ rq->vda_req_sz = reqsize;
+ }
+ return true;
+}
+
+
+/*
+ * Create PRD list for each I-block consumed by the command. This routine
+ * determines how much data is required from each I-block being consumed
+ * by the command. The first and last I-blocks can be partials and all of
+ * the I-blocks in between are for a full I-block of data.
+ *
+ * The interleave size is used to determine the number of bytes in the 1st
+ * I-block and the remaining I-blocks are what remeains.
+ */
+static bool esas2r_build_prd_iblk(struct esas2r_adapter *a,
+ struct esas2r_sg_context *sgc)
+{
+ struct esas2r_request *rq = sgc->first_req;
+ u64 addr;
+ u32 len;
+ struct esas2r_mem_desc *sgl;
+ u32 numchain = 1;
+ u32 rem = 0;
+
+ while (sgc->length) {
+ /* Get the next address/length pair */
+
+ len = (*sgc->get_phys_addr)(sgc, &addr);
+
+ if (unlikely(len == 0))
+ return false;
+
+ /* If current length is more than what's left, stop there */
+
+ if (unlikely(len > sgc->length))
+ len = sgc->length;
+
+another_entry:
+ /* Limit to a round number less than the maximum length */
+
+ if (len > PRD_LEN_MAX) {
+ /*
+ * Save the remainder of the split. whenever we limit
+ * an entry we come back around to build entries out
+ * of the leftover. We do this to prevent multiple
+ * calls to the get_phys_addr() function for an SGE
+ * that is too large.
+ */
+ rem = len - PRD_LEN_MAX;
+ len = PRD_LEN_MAX;
+ }
+
+ /* See if we need to allocate a new SGL */
+ if (sgc->sge.prd.sge_cnt == 0) {
+ if (len == sgc->length) {
+ /*
+ * We only have 1 PRD entry left.
+ * It can be placed where the chain
+ * entry would have gone
+ */
+
+ /* Build the simple SGE */
+ sgc->sge.prd.curr->ctl_len = cpu_to_le32(
+ PRD_DATA | len);
+ sgc->sge.prd.curr->address = cpu_to_le64(addr);
+
+ /* Adjust length related fields */
+ sgc->cur_offset += len;
+ sgc->length -= len;
+
+ /* We use the reserved chain entry for data */
+ numchain = 0;
+
+ break;
+ }
+
+ if (sgc->sge.prd.chain) {
+ /*
+ * Fill # of entries of current SGL in previous
+ * chain the length of this current SGL may not
+ * full.
+ */
+
+ sgc->sge.prd.chain->ctl_len |= cpu_to_le32(
+ sgc->sge.prd.sgl_max_cnt);
+ }
+
+ /*
+ * If no SGls are available, return failure. The
+ * caller can call us later with the current context
+ * to pick up here.
+ */
+
+ sgl = esas2r_alloc_sgl(a);
+
+ if (unlikely(sgl == NULL))
+ return false;
+
+ /*
+ * Link the new SGL onto the chain
+ * They are in reverse order
+ */
+ list_add(&sgl->next_desc, &rq->sg_table_head);
+
+ /*
+ * An SGL was just filled in and we are starting
+ * a new SGL. Prime the chain of the ending SGL with
+ * info that points to the new SGL. The length gets
+ * filled in when the new SGL is filled or ended
+ */
+
+ sgc->sge.prd.chain = sgc->sge.prd.curr;
+
+ sgc->sge.prd.chain->ctl_len = cpu_to_le32(PRD_CHAIN);
+ sgc->sge.prd.chain->address =
+ cpu_to_le64(sgl->phys_addr);
+
+ /*
+ * Start a new segment.
+ * Take one away and save for chain SGE
+ */
+
+ sgc->sge.prd.curr =
+ (struct atto_physical_region_description *)sgl
+ ->
+ virt_addr;
+ sgc->sge.prd.sge_cnt = sgc->sge.prd.sgl_max_cnt - 1;
+ }
+
+ sgc->sge.prd.sge_cnt--;
+ /* Build the simple SGE */
+ sgc->sge.prd.curr->ctl_len = cpu_to_le32(PRD_DATA | len);
+ sgc->sge.prd.curr->address = cpu_to_le64(addr);
+
+ /* Used another element. Point to the next one */
+
+ sgc->sge.prd.curr++;
+
+ /* Adjust length related fields */
+
+ sgc->cur_offset += len;
+ sgc->length -= len;
+
+ /*
+ * Check if we previously split an entry. If so we have to
+ * pick up where we left off.
+ */
+
+ if (rem) {
+ addr += len;
+ len = rem;
+ rem = 0;
+ goto another_entry;
+ }
+ }
+
+ if (!list_empty(&rq->sg_table_head)) {
+ if (sgc->sge.prd.chain) {
+ sgc->sge.prd.chain->ctl_len |=
+ cpu_to_le32(sgc->sge.prd.sgl_max_cnt
+ - sgc->sge.prd.sge_cnt
+ - numchain);
+ }
+ }
+
+ return true;
+}
+
+bool esas2r_build_sg_list_prd(struct esas2r_adapter *a,
+ struct esas2r_sg_context *sgc)
+{
+ struct esas2r_request *rq = sgc->first_req;
+ u32 len = sgc->length;
+ struct esas2r_target *t = a->targetdb + rq->target_id;
+ u8 is_i_o = 0;
+ u16 reqsize;
+ struct atto_physical_region_description *curr_iblk_chn;
+ u8 *cdb = (u8 *)&rq->vrq->scsi.cdb[0];
+
+ /*
+ * extract LBA from command so we can determine
+ * the I-Block boundary
+ */
+
+ if (rq->vrq->scsi.function == VDA_FUNC_SCSI
+ && t->target_state == TS_PRESENT
+ && !(t->flags & TF_PASS_THRU)) {
+ u32 lbalo = 0;
+
+ switch (rq->vrq->scsi.cdb[0]) {
+ case READ_16:
+ case WRITE_16:
+ {
+ lbalo =
+ MAKEDWORD(MAKEWORD(cdb[9],
+ cdb[8]),
+ MAKEWORD(cdb[7],
+ cdb[6]));
+ is_i_o = 1;
+ break;
+ }
+
+ case READ_12:
+ case WRITE_12:
+ case READ_10:
+ case WRITE_10:
+ {
+ lbalo =
+ MAKEDWORD(MAKEWORD(cdb[5],
+ cdb[4]),
+ MAKEWORD(cdb[3],
+ cdb[2]));
+ is_i_o = 1;
+ break;
+ }
+
+ case READ_6:
+ case WRITE_6:
+ {
+ lbalo =
+ MAKEDWORD(MAKEWORD(cdb[3],
+ cdb[2]),
+ MAKEWORD(cdb[1] & 0x1F,
+ 0));
+ is_i_o = 1;
+ break;
+ }
+
+ default:
+ break;
+ }
+
+ if (is_i_o) {
+ u32 startlba;
+
+ rq->vrq->scsi.iblk_cnt_prd = 0;
+
+ /* Determine size of 1st I-block PRD list */
+ startlba = t->inter_block - (lbalo & (t->inter_block -
+ 1));
+ sgc->length = startlba * t->block_size;
+
+ /* Chk if the 1st iblk chain starts at base of Iblock */
+ if ((lbalo & (t->inter_block - 1)) == 0)
+ rq->flags |= RF_1ST_IBLK_BASE;
+
+ if (sgc->length > len)
+ sgc->length = len;
+ } else {
+ sgc->length = len;
+ }
+ } else {
+ sgc->length = len;
+ }
+
+ /* get our starting chain address */
+
+ curr_iblk_chn =
+ (struct atto_physical_region_description *)sgc->sge.a64.curr;
+
+ sgc->sge.prd.sgl_max_cnt = sgl_page_size /
+ sizeof(struct
+ atto_physical_region_description);
+
+ /* create all of the I-block PRD lists */
+
+ while (len) {
+ sgc->sge.prd.sge_cnt = 0;
+ sgc->sge.prd.chain = NULL;
+ sgc->sge.prd.curr = curr_iblk_chn;
+
+ /* increment to next I-Block */
+
+ len -= sgc->length;
+
+ /* go build the next I-Block PRD list */
+
+ if (unlikely(!esas2r_build_prd_iblk(a, sgc)))
+ return false;
+
+ curr_iblk_chn++;
+
+ if (is_i_o) {
+ rq->vrq->scsi.iblk_cnt_prd++;
+
+ if (len > t->inter_byte)
+ sgc->length = t->inter_byte;
+ else
+ sgc->length = len;
+ }
+ }
+
+ /* figure out the size used of the VDA request */
+
+ reqsize = ((u16)((u8 *)curr_iblk_chn - (u8 *)rq->vrq))
+ / sizeof(u32);
+
+ /*
+ * only update the request size if it is bigger than what is
+ * already there. we can come in here twice for some management
+ * commands.
+ */
+
+ if (reqsize > rq->vda_req_sz)
+ rq->vda_req_sz = reqsize;
+
+ return true;
+}
+
+static void esas2r_handle_pending_reset(struct esas2r_adapter *a, u32 currtime)
+{
+ u32 delta = currtime - a->chip_init_time;
+
+ if (delta <= ESAS2R_CHPRST_WAIT_TIME) {
+ /* Wait before accessing registers */
+ } else if (delta >= ESAS2R_CHPRST_TIME) {
+ /*
+ * The last reset failed so try again. Reset
+ * processing will give up after three tries.
+ */
+ esas2r_local_reset_adapter(a);
+ } else {
+ /* We can now see if the firmware is ready */
+ u32 doorbell;
+
+ doorbell = esas2r_read_register_dword(a, MU_DOORBELL_OUT);
+ if (doorbell == 0xFFFFFFFF || !(doorbell & DRBL_FORCE_INT)) {
+ esas2r_force_interrupt(a);
+ } else {
+ u32 ver = (doorbell & DRBL_FW_VER_MSK);
+
+ /* Driver supports API version 0 and 1 */
+ esas2r_write_register_dword(a, MU_DOORBELL_OUT,
+ doorbell);
+ if (ver == DRBL_FW_VER_0) {
+ set_bit(AF_CHPRST_DETECTED, &a->flags);
+ set_bit(AF_LEGACY_SGE_MODE, &a->flags);
+
+ a->max_vdareq_size = 128;
+ a->build_sgl = esas2r_build_sg_list_sge;
+ } else if (ver == DRBL_FW_VER_1) {
+ set_bit(AF_CHPRST_DETECTED, &a->flags);
+ clear_bit(AF_LEGACY_SGE_MODE, &a->flags);
+
+ a->max_vdareq_size = 1024;
+ a->build_sgl = esas2r_build_sg_list_prd;
+ } else {
+ esas2r_local_reset_adapter(a);
+ }
+ }
+ }
+}
+
+
+/* This function must be called once per timer tick */
+void esas2r_timer_tick(struct esas2r_adapter *a)
+{
+ u32 currtime = jiffies_to_msecs(jiffies);
+ u32 deltatime = currtime - a->last_tick_time;
+
+ a->last_tick_time = currtime;
+
+ /* count down the uptime */
+ if (a->chip_uptime &&
+ !test_bit(AF_CHPRST_PENDING, &a->flags) &&
+ !test_bit(AF_DISC_PENDING, &a->flags)) {
+ if (deltatime >= a->chip_uptime)
+ a->chip_uptime = 0;
+ else
+ a->chip_uptime -= deltatime;
+ }
+
+ if (test_bit(AF_CHPRST_PENDING, &a->flags)) {
+ if (!test_bit(AF_CHPRST_NEEDED, &a->flags) &&
+ !test_bit(AF_CHPRST_DETECTED, &a->flags))
+ esas2r_handle_pending_reset(a, currtime);
+ } else {
+ if (test_bit(AF_DISC_PENDING, &a->flags))
+ esas2r_disc_check_complete(a);
+ if (test_bit(AF_HEARTBEAT_ENB, &a->flags)) {
+ if (test_bit(AF_HEARTBEAT, &a->flags)) {
+ if ((currtime - a->heartbeat_time) >=
+ ESAS2R_HEARTBEAT_TIME) {
+ clear_bit(AF_HEARTBEAT, &a->flags);
+ esas2r_hdebug("heartbeat failed");
+ esas2r_log(ESAS2R_LOG_CRIT,
+ "heartbeat failed");
+ esas2r_bugon();
+ esas2r_local_reset_adapter(a);
+ }
+ } else {
+ set_bit(AF_HEARTBEAT, &a->flags);
+ a->heartbeat_time = currtime;
+ esas2r_force_interrupt(a);
+ }
+ }
+ }
+
+ if (atomic_read(&a->disable_cnt) == 0)
+ esas2r_do_deferred_processes(a);
+}
+
+/*
+ * Send the specified task management function to the target and LUN
+ * specified in rqaux. in addition, immediately abort any commands that
+ * are queued but not sent to the device according to the rules specified
+ * by the task management function.
+ */
+bool esas2r_send_task_mgmt(struct esas2r_adapter *a,
+ struct esas2r_request *rqaux, u8 task_mgt_func)
+{
+ u16 targetid = rqaux->target_id;
+ u8 lun = (u8)le32_to_cpu(rqaux->vrq->scsi.flags);
+ bool ret = false;
+ struct esas2r_request *rq;
+ struct list_head *next, *element;
+ unsigned long flags;
+
+ LIST_HEAD(comp_list);
+
+ esas2r_trace_enter();
+ esas2r_trace("rqaux:%p", rqaux);
+ esas2r_trace("task_mgt_func:%x", task_mgt_func);
+ spin_lock_irqsave(&a->queue_lock, flags);
+
+ /* search the defer queue looking for requests for the device */
+ list_for_each_safe(element, next, &a->defer_list) {
+ rq = list_entry(element, struct esas2r_request, req_list);
+
+ if (rq->vrq->scsi.function == VDA_FUNC_SCSI
+ && rq->target_id == targetid
+ && (((u8)le32_to_cpu(rq->vrq->scsi.flags)) == lun
+ || task_mgt_func == 0x20)) { /* target reset */
+ /* Found a request affected by the task management */
+ if (rq->req_stat == RS_PENDING) {
+ /*
+ * The request is pending or waiting. We can
+ * safelycomplete the request now.
+ */
+ if (esas2r_ioreq_aborted(a, rq, RS_ABORTED))
+ list_add_tail(&rq->comp_list,
+ &comp_list);
+ }
+ }
+ }
+
+ /* Send the task management request to the firmware */
+ rqaux->sense_len = 0;
+ rqaux->vrq->scsi.length = 0;
+ rqaux->target_id = targetid;
+ rqaux->vrq->scsi.flags |= cpu_to_le32(lun);
+ memset(rqaux->vrq->scsi.cdb, 0, sizeof(rqaux->vrq->scsi.cdb));
+ rqaux->vrq->scsi.flags |=
+ cpu_to_le16(task_mgt_func * LOBIT(FCP_CMND_TM_MASK));
+
+ if (test_bit(AF_FLASHING, &a->flags)) {
+ /* Assume success. if there are active requests, return busy */
+ rqaux->req_stat = RS_SUCCESS;
+
+ list_for_each_safe(element, next, &a->active_list) {
+ rq = list_entry(element, struct esas2r_request,
+ req_list);
+ if (rq->vrq->scsi.function == VDA_FUNC_SCSI
+ && rq->target_id == targetid
+ && (((u8)le32_to_cpu(rq->vrq->scsi.flags)) == lun
+ || task_mgt_func == 0x20)) /* target reset */
+ rqaux->req_stat = RS_BUSY;
+ }
+
+ ret = true;
+ }
+
+ spin_unlock_irqrestore(&a->queue_lock, flags);
+
+ if (!test_bit(AF_FLASHING, &a->flags))
+ esas2r_start_request(a, rqaux);
+
+ esas2r_comp_list_drain(a, &comp_list);
+
+ if (atomic_read(&a->disable_cnt) == 0)
+ esas2r_do_deferred_processes(a);
+
+ esas2r_trace_exit();
+
+ return ret;
+}
+
+void esas2r_reset_bus(struct esas2r_adapter *a)
+{
+ esas2r_log(ESAS2R_LOG_INFO, "performing a bus reset");
+
+ if (!test_bit(AF_DEGRADED_MODE, &a->flags) &&
+ !test_bit(AF_CHPRST_PENDING, &a->flags) &&
+ !test_bit(AF_DISC_PENDING, &a->flags)) {
+ set_bit(AF_BUSRST_NEEDED, &a->flags);
+ set_bit(AF_BUSRST_PENDING, &a->flags);
+ set_bit(AF_OS_RESET, &a->flags);
+
+ esas2r_schedule_tasklet(a);
+ }
+}
+
+bool esas2r_ioreq_aborted(struct esas2r_adapter *a, struct esas2r_request *rq,
+ u8 status)
+{
+ esas2r_trace_enter();
+ esas2r_trace("rq:%p", rq);
+ list_del_init(&rq->req_list);
+ if (rq->timeout > RQ_MAX_TIMEOUT) {
+ /*
+ * The request timed out, but we could not abort it because a
+ * chip reset occurred. Return busy status.
+ */
+ rq->req_stat = RS_BUSY;
+ esas2r_trace_exit();
+ return true;
+ }
+
+ rq->req_stat = status;
+ esas2r_trace_exit();
+ return true;
+}