This patch contains code to implement the interrupt handling and the fast
path I/O functionality. The interrupt handling includes allocation of
MSIX vectors, registering and implemeting the interrupt service routines.
The fast path I/O functionality includes posting the I/O request to firmware
via Work Requests, tracking/completing them, and handling task management
requests. SCSI midlayer host template implementation is also covered by
this patch.
Signed-off-by: Naresh Kumar Inna <naresh@xxxxxxxxxxx>
---
drivers/scsi/csiostor/csio_isr.c | 631 ++++++++++
drivers/scsi/csiostor/csio_scsi.c | 2498 +++++++++++++++++++++++++++++++++++++
2 files changed, 3129 insertions(+), 0 deletions(-)
create mode 100644 drivers/scsi/csiostor/csio_isr.c
create mode 100644 drivers/scsi/csiostor/csio_scsi.c
diff --git a/drivers/scsi/csiostor/csio_isr.c b/drivers/scsi/csiostor/csio_isr.c
new file mode 100644
index 0000000..96633e9
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_isr.c
@@ -0,0 +1,631 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * 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.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/pci.h>
+#include <linux/interrupt.h>
+#include <linux/cpumask.h>
+#include <linux/string.h>
+
+#include "csio_init.h"
+#include "csio_hw.h"
+
+static irqreturn_t
+csio_nondata_isr(int irq, void *dev_id)
+{
+ struct csio_hw *hw = (struct csio_hw *) dev_id;
+ csio_retval_t rv;
+ unsigned long flags;
+
+ if (unlikely(!hw))
+ return IRQ_NONE;
+
+ if (unlikely(pci_channel_offline(hw->pdev))) {
+ csio_inc_stats(hw, n_pcich_offline);
+ return IRQ_NONE;
+ }
+
+ spin_lock_irqsave(&hw->lock, flags);
+ csio_hw_slow_intr_handler(hw);
+ rv = csio_mb_isr_handler(hw);
+
+ if (rv == CSIO_SUCCESS && !(hw->flags & CSIO_HWF_FWEVT_PENDING)) {
+ hw->flags |= CSIO_HWF_FWEVT_PENDING;
+ spin_unlock_irqrestore(&hw->lock, flags);
+ schedule_work(&hw->evtq_work);
+ return IRQ_HANDLED;
+ }
+ spin_unlock_irqrestore(&hw->lock, flags);
+ return IRQ_HANDLED;
+}
+
+/*
+ * csio_fwevt_handler - Common FW event handler routine.
+ * @hw: HW module.
+ *
+ * This is the ISR for FW events. It is shared b/w MSIX
+ * and INTx handlers.
+ */
+static void
+csio_fwevt_handler(struct csio_hw *hw)
+{
+ csio_retval_t rv;
+ unsigned long flags;
+
+ rv = csio_fwevtq_handler(hw);
+
+ spin_lock_irqsave(&hw->lock, flags);
+ if (rv == CSIO_SUCCESS && !(hw->flags & CSIO_HWF_FWEVT_PENDING)) {
+ hw->flags |= CSIO_HWF_FWEVT_PENDING;
+ spin_unlock_irqrestore(&hw->lock, flags);
+ schedule_work(&hw->evtq_work);
+ return;
+ }
+ spin_unlock_irqrestore(&hw->lock, flags);
+
+} /* csio_fwevt_handler */
+
+/*
+ * csio_fwevt_isr() - FW events MSIX ISR
+ * @irq:
+ * @dev_id:
+ *
+ * Process WRs on the FW event queue.
+ *
+ */
+static irqreturn_t
+csio_fwevt_isr(int irq, void *dev_id)
+{
+ struct csio_hw *hw = (struct csio_hw *) dev_id;
+
+ if (unlikely(!hw))
+ return IRQ_NONE;
+
+ if (unlikely(pci_channel_offline(hw->pdev))) {
+ csio_inc_stats(hw, n_pcich_offline);
+ return IRQ_NONE;
+ }
+
+ csio_fwevt_handler(hw);
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * csio_fwevt_isr() - INTx wrapper for handling FW events.
+ * @irq:
+ * @dev_id:
+ */
+void
+csio_fwevt_intx_handler(struct csio_hw *hw, void *wr, uint32_t len,
+ struct csio_fl_dma_buf *flb, void *priv)
+{
+ csio_fwevt_handler(hw);
+} /* csio_fwevt_intx_handler */
+
+/*
+ * csio_process_scsi_cmpl - Process a SCSI WR completion.
+ * @hw: HW module.
+ * @wr: The completed WR from the ingress queue.
+ * @len: Length of the WR.
+ * @flb: Freelist buffer array.
+ *
+ */
+static void
+csio_process_scsi_cmpl(struct csio_hw *hw, void *wr, uint32_t len,
+ struct csio_fl_dma_buf *flb, void *cbfn_q)
+{
+ struct csio_ioreq *ioreq;
+ uint8_t *scsiwr;
+ uint8_t subop;
+ void *cmnd;
+ unsigned long flags;
+
+ ioreq = csio_scsi_cmpl_handler(hw, wr, len, flb, NULL, &scsiwr);
+ if (likely(ioreq)) {
+ if (unlikely(*scsiwr == FW_SCSI_ABRT_CLS_WR)) {
+ subop = FW_SCSI_ABRT_CLS_WR_SUB_OPCODE_GET(
+ ((struct fw_scsi_abrt_cls_wr *)
+ scsiwr)->sub_opcode_to_chk_all_io);
+
+ csio_dbg(hw, "%s cmpl recvd ioreq:%p status:%d\n",
+ subop ? "Close" : "Abort",
+ ioreq, ioreq->wr_status);
+
+ spin_lock_irqsave(&hw->lock, flags);
+ if (subop)
+ csio_scsi_closed(ioreq,
+ (struct list_head *)cbfn_q);
+ else
+ csio_scsi_aborted(ioreq,
+ (struct list_head *)cbfn_q);
+ /*
+ * We call scsi_done for I/Os that driver thinks aborts
+ * have timed out. If there is a race caused by FW
+ * completing abort at the exact same time that the
+ * driver has deteced the abort timeout, the following
+ * check prevents calling of scsi_done twice for the
+ * same command: once from the eh_abort_handler, another
+ * from csio_scsi_isr_handler(). This also avoids the
+ * need to check if csio_scsi_cmnd(req) is NULL in the
+ * fast path.
+ */
+ cmnd = csio_scsi_cmnd(ioreq);
+ if (unlikely(cmnd == NULL))
+ list_del_init(&ioreq->sm.sm_list);
+
+ spin_unlock_irqrestore(&hw->lock, flags);
+
+ if (unlikely(cmnd == NULL))
+ csio_put_scsi_ioreq_lock(hw,
+ csio_hw_to_scsim(hw), ioreq);
+ } else {
+ spin_lock_irqsave(&hw->lock, flags);
+ csio_scsi_completed(ioreq, (struct list_head *)cbfn_q);
+ spin_unlock_irqrestore(&hw->lock, flags);
+ }
+ }
+}
+
+/*
+ * csio_scsi_isr_handler() - Common SCSI ISR handler.
+ * @iq: Ingress queue pointer.
+ *
+ * Processes SCSI completions on the SCSI IQ indicated by scm->iq_idx
+ * by calling csio_wr_process_iq_idx. If there are completions on the
+ * isr_cbfn_q, yank them out into a local queue and call their io_cbfns.
+ * Once done, add these completions onto the freelist.
+ * This routine is shared b/w MSIX and INTx.
+ */
+static inline irqreturn_t
+csio_scsi_isr_handler(struct csio_q *iq)
+{
+ struct csio_hw *hw = (struct csio_hw *)iq->owner;
+ LIST_HEAD(cbfn_q);
+ struct list_head *tmp;
+ struct csio_scsim *scm;
+ struct csio_ioreq *ioreq;
+ int isr_completions = 0;
+
+ scm = csio_hw_to_scsim(hw);
+
+ if (unlikely(csio_wr_process_iq(hw, iq, csio_process_scsi_cmpl,
+ &cbfn_q) != CSIO_SUCCESS))
+ return IRQ_NONE;
+
+ /* Call back the completion routines */
+ list_for_each(tmp, &cbfn_q) {
+ ioreq = (struct csio_ioreq *)tmp;
+ isr_completions++;
+ ioreq->io_cbfn(hw, ioreq);
+#ifdef __CSIO_DDP_SUPPORT__
+ /* Release ddp buffer if used for this req */
+ if (unlikely(ioreq->dcopy))
+ csio_put_scsi_ddp_list_lock(hw, scm, &ioreq->gen_list,
+ ioreq->nsge);
+#endif
+ }
+
+ if (isr_completions) {
+ /* Return the ioreqs back to ioreq->freelist */
+ csio_put_scsi_ioreq_list_lock(hw, scm, &cbfn_q,
+ isr_completions);
+ }
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * csio_scsi_isr() - SCSI MSIX handler
+ * @irq:
+ * @dev_id:
+ *
+ * This is the top level SCSI MSIX handler. Calls csio_scsi_isr_handler()
+ * for handling SCSI completions.
+ */
+static irqreturn_t
+csio_scsi_isr(int irq, void *dev_id)
+{
+ struct csio_q *iq = (struct csio_q *) dev_id;
+ struct csio_hw *hw;
+
+ if (unlikely(!iq))
+ return IRQ_NONE;
+
+ hw = (struct csio_hw *)iq->owner;
+
+ if (unlikely(pci_channel_offline(hw->pdev))) {
+ csio_inc_stats(hw, n_pcich_offline);
+ return IRQ_NONE;
+ }
+
+ csio_scsi_isr_handler(iq);
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * csio_scsi_intx_handler() - SCSI INTx handler
+ * @irq:
+ * @dev_id:
+ *
+ * This is the top level SCSI INTx handler. Calls csio_scsi_isr_handler()
+ * for handling SCSI completions.
+ */
+void
+csio_scsi_intx_handler(struct csio_hw *hw, void *wr, uint32_t len,
+ struct csio_fl_dma_buf *flb, void *priv)
+{
+ struct csio_q *iq = priv;
+
+ csio_scsi_isr_handler(iq);
+
+} /* csio_scsi_intx_handler */
+
+/*
+ * csio_fcoe_isr() - INTx/MSI interrupt service routine for FCoE.
+ * @irq:
+ * @dev_id:
+ *
+ *
+ */
+static irqreturn_t
+csio_fcoe_isr(int irq, void *dev_id)
+{
+ struct csio_hw *hw = (struct csio_hw *) dev_id;
+ struct csio_q *intx_q = NULL;
+ csio_retval_t rv;
+ irqreturn_t ret = IRQ_NONE;
+ unsigned long flags;
+
+ if (unlikely(!hw))
+ return IRQ_NONE;
+
+ if (unlikely(pci_channel_offline(hw->pdev))) {
+ csio_inc_stats(hw, n_pcich_offline);
+ return IRQ_NONE;
+ }
+
+ /* Disable the interrupt for this PCI function. */
+ if (hw->intr_mode == CSIO_IM_INTX)
+ csio_wr_reg32(hw, 0, MYPF_REG(PCIE_PF_CLI));
+
+ /*
+ * The read in the following function will flush the
+ * above write.
+ */
+ if (csio_hw_slow_intr_handler(hw))
+ ret = IRQ_HANDLED;
+
+ /* Get the INTx Forward interrupt IQ. */
+ intx_q = csio_get_q(hw, hw->intr_iq_idx);
+
+ CSIO_DB_ASSERT(intx_q);
+
+ /* IQ handler is not possible for intx_q, hence pass in NULL */
+ if (likely(csio_wr_process_iq(hw, intx_q, NULL, NULL) == CSIO_SUCCESS))
+ ret = IRQ_HANDLED;
+
+ spin_lock_irqsave(&hw->lock, flags);
+ rv = csio_mb_isr_handler(hw);
+ if (rv == CSIO_SUCCESS && !(hw->flags & CSIO_HWF_FWEVT_PENDING)) {
+ hw->flags |= CSIO_HWF_FWEVT_PENDING;
+ spin_unlock_irqrestore(&hw->lock, flags);
+ schedule_work(&hw->evtq_work);
+ return IRQ_HANDLED;
+ }
+ spin_unlock_irqrestore(&hw->lock, flags);
+
+ return ret;
+}
+
+#define csio_extra_msix_desc(_desc, _len, _str, _arg1, _arg2, _arg3) \
+do { \
+ memset((_desc), 0, (_len) + 1); \
+ snprintf((_desc), (_len), (_str), (_arg1), (_arg2), (_arg3)); \
+} while (0)
+
+static void
+csio_add_msix_desc(struct csio_hw *hw)
+{
+ int i;
+ struct csio_msix_entries *entryp = &hw->msix_entries[0];
+ int k = CSIO_EXTRA_VECS;
+ int len = sizeof(entryp->desc) - 1;
+ int cnt = hw->num_sqsets + k;
+
+ /* Non-data vector */
+ csio_extra_msix_desc(entryp->desc, len, "csio-%02x:%02x:%x-nondata",
+ CSIO_PCI_BUS(hw), CSIO_PCI_DEV(hw),
+ CSIO_PCI_FUNC(hw));
+ entryp++;
+ csio_extra_msix_desc(entryp->desc, len, "csio-%02x:%02x:%x-fwevt",
+ CSIO_PCI_BUS(hw), CSIO_PCI_DEV(hw),
+ CSIO_PCI_FUNC(hw));
+ entryp++;
+
+ /* Name SCSI vecs */
+ for (i = k; i < cnt; i++, entryp++) {
+ memset(entryp->desc, 0, len + 1);
+ snprintf(entryp->desc, len, "csio-%02x:%02x:%x-scsi%d",
+ CSIO_PCI_BUS(hw), CSIO_PCI_DEV(hw),
+ CSIO_PCI_FUNC(hw), i - CSIO_EXTRA_VECS);
+ }
+}
+
+csio_retval_t
+csio_request_irqs(struct csio_hw *hw)
+{
+ int rv, i, j, k = 0;
+ struct csio_msix_entries *entryp = &hw->msix_entries[0];
+ struct csio_scsi_cpu_info *info;
+
+ if (hw->intr_mode != CSIO_IM_MSIX) {
+ rv = request_irq(hw->pdev->irq, csio_fcoe_isr,
+ (hw->intr_mode == CSIO_IM_MSI) ?
+ 0 : IRQF_SHARED,
+ KBUILD_MODNAME, hw);
+ if (rv) {
+ if (hw->intr_mode == CSIO_IM_MSI)
+ pci_disable_msi(hw->pdev);
+ csio_err(hw, "Failed to allocate interrupt line.\n");
+ return CSIO_INVAL;
+ }
+
+ goto out;
+ }
+
+ /* Add the MSIX vector descriptions */
+ csio_add_msix_desc(hw);
+
+ rv = request_irq(entryp[k].vector, csio_nondata_isr, 0,
+ entryp[k].desc, hw);
+ if (rv) {
+ csio_err(hw, "IRQ request failed for vec %d err:%d\n",
+ entryp[k].vector, rv);
+ goto err;
+ }
+
+ entryp[k++].dev_id = (void *)hw;
+
+ rv = request_irq(entryp[k].vector, csio_fwevt_isr, 0,
+ entryp[k].desc, hw);
+ if (rv) {
+ csio_err(hw, "IRQ request failed for vec %d err:%d\n",
+ entryp[k].vector, rv);
+ goto err;
+ }
+
+ entryp[k++].dev_id = (void *)hw;
+
+ /* Allocate IRQs for SCSI */
+ for (i = 0; i < hw->num_pports; i++) {
+ info = &hw->scsi_cpu_info[i];
+ for (j = 0; j < info->max_cpus; j++, k++) {
+ struct csio_scsi_qset *sqset = &hw->sqset[i][j];
+ struct csio_q *q = hw->wrm.q_arr[sqset->iq_idx];
+
+ rv = request_irq(entryp[k].vector, csio_scsi_isr, 0,
+ entryp[k].desc, q);
+ if (rv) {
+ csio_err(hw,
+ "IRQ request failed for vec %d err:%d\n",
+ entryp[k].vector, rv);
+ goto err;
+ }
+
+ entryp[k].dev_id = (void *)q;
+
+ } /* for all scsi cpus */
+ } /* for all ports */
+
+out:
+ hw->flags |= CSIO_HWF_HOST_INTR_ENABLED;
+
+ return CSIO_SUCCESS;
+
+err:
+ for (i = 0; i < k; i++) {
+ entryp = &hw->msix_entries[i];
+ free_irq(entryp->vector, entryp->dev_id);
+ }
+ pci_disable_msix(hw->pdev);
+
+ return CSIO_INVAL;
+}
+
+static void
+csio_disable_msix(struct csio_hw *hw, bool free)
+{
+ int i;
+ struct csio_msix_entries *entryp;
+ int cnt = hw->num_sqsets + CSIO_EXTRA_VECS;
+
+ if (free) {
+ for (i = 0; i < cnt; i++) {
+ entryp = &hw->msix_entries[i];
+ free_irq(entryp->vector, entryp->dev_id);
+ }
+ }
+ pci_disable_msix(hw->pdev);
+}
+
+/* Reduce per-port max possible CPUs */
+static void
+csio_reduce_sqsets(struct csio_hw *hw, int cnt)
+{
+ int i;
+ struct csio_scsi_cpu_info *info;
+
+ while (cnt < hw->num_sqsets) {
+ for (i = 0; i < hw->num_pports; i++) {
+ info = &hw->scsi_cpu_info[i];
+ if (info->max_cpus > 1) {
+ info->max_cpus--;
+ hw->num_sqsets--;
+ if (hw->num_sqsets <= cnt)
+ break;
+ }
+ }
+ }
+
+ csio_dbg(hw, "Reduced sqsets to %d\n", hw->num_sqsets);
+}
+
+static csio_retval_t
+csio_enable_msix(struct csio_hw *hw)
+{
+ int rv, i, j, k, n, min, cnt;
+ struct csio_msix_entries *entryp;
+ struct msix_entry *entries;
+ int extra = CSIO_EXTRA_VECS;
+ struct csio_scsi_cpu_info *info;
+
+ min = hw->num_pports + extra;
+ cnt = hw->num_sqsets + extra;
+
+ /* Max vectors required based on #niqs configured in fw */
+ if (hw->flags & CSIO_HWF_USING_SOFT_PARAMS || !csio_is_hw_master(hw))
+ cnt = min_t(uint8_t, hw->cfg_niq, cnt);
+
+ entries = kzalloc(sizeof(struct msix_entry) * cnt, GFP_KERNEL);
+ if (!entries)
+ return CSIO_NOMEM;
+
+ for (i = 0; i < cnt; i++)
+ entries[i].entry = (uint16_t)i;
+
+ csio_dbg(hw, "FW supp #niq:%d, trying %d msix's\n", hw->cfg_niq, cnt);
+
+ while ((rv = pci_enable_msix(hw->pdev, entries, cnt)) >= min)
+ cnt = rv;
+ if (!rv) {
+ if (cnt < (hw->num_sqsets + extra)) {
+ csio_dbg(hw, "Reducing sqsets to %d\n", cnt - extra);
+ csio_reduce_sqsets(hw, cnt - extra);
+ }
+ } else {
+ if (rv > 0) {
+ pci_disable_msix(hw->pdev);
+ csio_info(hw, "Not using MSI-X, remainder:%d\n", rv);
+ }
+
+ kfree(entries);
+ return CSIO_NOMEM;
+ }
+
+ /* Save off vectors */
+ for (i = 0; i < cnt; i++) {
+ entryp = &hw->msix_entries[i];
+ entryp->vector = entries[i].vector;
+ }
+
+ /* Distribute vectors */
+ k = 0;
+ csio_set_nondata_intr_idx(hw, entries[k].entry);
+ csio_set_mb_intr_idx(csio_hw_to_mbm(hw), entries[k++].entry);
+ csio_set_fwevt_intr_idx(hw, entries[k++].entry);
+
+ for (i = 0; i < hw->num_pports; i++) {
+ info = &hw->scsi_cpu_info[i];
+
+ for (j = 0; j < hw->num_scsi_msix_cpus; j++) {
+ n = (j % info->max_cpus) + k;
+ hw->sqset[i][j].intr_idx = entries[n].entry;
+ }
+
+ k += info->max_cpus;
+ }
+
+ kfree(entries);
+ return CSIO_SUCCESS;
+}
+
+void
+csio_intr_enable(struct csio_hw *hw)
+{
+ hw->intr_mode = CSIO_IM_NONE;
+ hw->flags &= ~CSIO_HWF_HOST_INTR_ENABLED;
+
+ /* Try MSIX, then MSI or fall back to INTx */
+ if ((csio_msi == 2) && !csio_enable_msix(hw))
+ hw->intr_mode = CSIO_IM_MSIX;
+ else {
+ /* Max iqs required based on #niqs configured in fw */
+ if (hw->flags & CSIO_HWF_USING_SOFT_PARAMS ||
+ !csio_is_hw_master(hw)) {
+ int extra = CSIO_EXTRA_MSI_IQS;
+
+ if (hw->cfg_niq < (hw->num_sqsets + extra)) {
+ csio_dbg(hw, "Reducing sqsets to %d\n",
+ hw->cfg_niq - extra);
+ csio_reduce_sqsets(hw, hw->cfg_niq - extra);
+ }
+ }
+
+ if ((csio_msi == 1) && !pci_enable_msi(hw->pdev))
+ hw->intr_mode = CSIO_IM_MSI;
+ else
+ hw->intr_mode = CSIO_IM_INTX;
+ }
+
+ csio_dbg(hw, "Using %s interrupt mode.\n",
+ (hw->intr_mode == CSIO_IM_MSIX) ? "MSIX" :
+ ((hw->intr_mode == CSIO_IM_MSI) ? "MSI" : "INTx"));
+}
+
+void
+csio_intr_disable(struct csio_hw *hw, bool free)
+{
+ csio_hw_intr_disable(hw);
+
+ switch (hw->intr_mode) {
+ case CSIO_IM_MSIX:
+ csio_disable_msix(hw, free);
+ break;
+ case CSIO_IM_MSI:
+ if (free)
+ free_irq(hw->pdev->irq, hw);
+ pci_disable_msi(hw->pdev);
+ break;
+ case CSIO_IM_INTX:
+ if (free)
+ free_irq(hw->pdev->irq, hw);
+ break;
+ default:
+ break;
+ }
+ hw->intr_mode = CSIO_IM_NONE;
+ hw->flags &= ~CSIO_HWF_HOST_INTR_ENABLED;
+}
diff --git a/drivers/scsi/csiostor/csio_scsi.c b/drivers/scsi/csiostor/csio_scsi.c
new file mode 100644
index 0000000..0f87b00
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_scsi.c
@@ -0,0 +1,2498 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * 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.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/moduleparam.h>
+#include <linux/device.h>
+#include <linux/delay.h>
+#include <linux/ctype.h>
+#include <linux/version.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+#include <linux/compiler.h>
+#include <linux/export.h>
+#include <linux/module.h>
+#include <asm/unaligned.h>
+#include <asm/page.h>
+#include <scsi/scsi.h>
+#include <scsi/scsi_transport_fc.h>
+
+#include "csio_hw.h"
+#include "csio_lnode.h"
+#include "csio_rnode.h"
+#include "csio_scsi.h"
+#include "csio_init.h"
+
+int csio_scsi_eqsize = 65536;
+int csio_scsi_iqlen = 128;
+int csio_scsi_ioreqs = 2048;
+uint32_t csio_max_scan_tmo;
+uint32_t csio_delta_scan_tmo = 5;
+int csio_lun_qdepth = 32;
+
+static int csio_ddp_descs = 128;
+
+static csio_retval_t csio_do_abrt_cls(struct csio_hw *,
+ struct csio_ioreq *, bool);
+
+static void csio_scsis_uninit(struct csio_ioreq *, enum csio_scsi_ev);
+static void csio_scsis_io_active(struct csio_ioreq *, enum csio_scsi_ev);
+static void csio_scsis_tm_active(struct csio_ioreq *, enum csio_scsi_ev);
+static void csio_scsis_aborting(struct csio_ioreq *, enum csio_scsi_ev);
+static void csio_scsis_closing(struct csio_ioreq *, enum csio_scsi_ev);
+static void csio_scsis_shost_cmpl_await(struct csio_ioreq *, enum csio_scsi_ev);
+
+/*
+ * csio_scsi_match_io - Match an ioreq with the given SCSI level data.
+ * @ioreq: The I/O request
+ * @sld: Level information
+ *
+ * Should be called with lock held.
+ *
+ */
+static bool
+csio_scsi_match_io(struct csio_ioreq *ioreq, struct csio_scsi_level_data *sld)
+{
+ struct scsi_cmnd *scmnd = csio_scsi_cmnd(ioreq);
+
+ switch (sld->level) {
+ case CSIO_LEV_LUN:
+ if (scmnd == NULL)
+ return CSIO_FALSE;
+
+ return ((ioreq->lnode == sld->lnode) &&
+ (ioreq->rnode == sld->rnode) &&
+ ((uint64_t)scmnd->device->lun == sld->oslun));
+
+ case CSIO_LEV_RNODE:
+ return ((ioreq->lnode == sld->lnode) &&
+ (ioreq->rnode == sld->rnode));
+ case CSIO_LEV_LNODE:
+ return (ioreq->lnode == sld->lnode);
+ case CSIO_LEV_ALL:
+ return CSIO_TRUE;
+ default:
+ return CSIO_FALSE;
+ }
+}
+
+/*
+ * csio_scsi_gather_active_ios - Gather active I/Os based on level
+ * @scm: SCSI module
+ * @sld: Level information
+ * @dest: The queue where these I/Os have to be gathered.
+ *
+ * Should be called with lock held.
+ */
+static void
+csio_scsi_gather_active_ios(struct csio_scsim *scm,
+ struct csio_scsi_level_data *sld,
+ struct list_head *dest)
+{
+ struct list_head *tmp, *next;
+
+ if (list_empty(&scm->active_q))
+ return;
+
+ /* Just splice the entire active_q into dest */
+ if (sld->level == CSIO_LEV_ALL) {
+ list_splice_tail_init(&scm->active_q, dest);
+ return;
+ }
+
+ list_for_each_safe(tmp, next, &scm->active_q) {
+ if (csio_scsi_match_io((struct csio_ioreq *)tmp, sld)) {
+ list_del_init(tmp);
+ list_add_tail(tmp, dest);
+ }
+ }
+}
+
+static inline bool
+csio_scsi_itnexus_loss_error(uint16_t error)
+{
+ switch (error) {
+ case FW_ERR_LINK_DOWN:
+ case FW_RDEV_NOT_READY:
+ case FW_ERR_RDEV_LOST:
+ case FW_ERR_RDEV_LOGO:
+ case FW_ERR_RDEV_IMPL_LOGO:
+ return 1;
+ }
+ return 0;
+}
+
+static inline void
+csio_scsi_tag(struct scsi_cmnd *scmnd, uint8_t *tag, uint8_t hq,
+ uint8_t oq, uint8_t sq)
+{
+ char stag[2];
+
+ if (scsi_populate_tag_msg(scmnd, stag)) {
+ switch (stag[0]) {
+ case HEAD_OF_QUEUE_TAG:
+ *tag = hq;
+ break;
+ case ORDERED_QUEUE_TAG:
+ *tag = oq;
+ break;
+ default:
+ *tag = sq;
+ break;
+ }
+ } else
+ *tag = 0;
+}
+
+/*
+ * csio_scsi_fcp_cmnd - Frame the SCSI FCP command paylod.
+ * @req: IO req structure.
+ * @addr: DMA location to place the payload.
+ *
+ * This routine is shared between FCP_WRITE, FCP_READ and FCP_CMD requests.
+ */
+static inline void
+csio_scsi_fcp_cmnd(struct csio_ioreq *req, void *addr)
+{
+ struct csio_fcp_cmnd *fcp_cmnd = (struct csio_fcp_cmnd *)addr;
+ struct scsi_cmnd *scmnd = csio_scsi_cmnd(req);
+
+ /* Check for Task Management */
+ if (likely(scmnd->SCp.Message == 0)) {
+ int_to_scsilun(scmnd->device->lun,
+ (struct scsi_lun *)fcp_cmnd->lun);
+ fcp_cmnd->tm_flags = 0;
+ fcp_cmnd->cmdref = 0;
+ fcp_cmnd->pri_ta = 0;
+
+ memcpy(fcp_cmnd->cdb, scmnd->cmnd, 16);
+ csio_scsi_tag(scmnd, &fcp_cmnd->pri_ta,
+ FCP_PTA_HEADQ, FCP_PTA_ORDERED, FCP_PTA_SIMPLE);
+ fcp_cmnd->dl = cpu_to_be32(scsi_bufflen(scmnd));
+
+ if (req->nsge)
+ if (req->datadir == CSIO_IOREQF_DMA_WRITE)
+ fcp_cmnd->flags = FCP_CFL_WRDATA;
+ else
+ fcp_cmnd->flags = FCP_CFL_RDDATA;
+ else
+ fcp_cmnd->flags = 0;
+ } else {
+ memset(fcp_cmnd, 0, sizeof(*fcp_cmnd));
+ int_to_scsilun(scmnd->device->lun,
+ (struct scsi_lun *)fcp_cmnd->lun);
+ fcp_cmnd->tm_flags = (uint8_t)scmnd->SCp.Message;
+ }
+}
+
+/*
+ * csio_scsi_init_cmd_wr - Initialize the SCSI CMD WR.
+ * @req: IO req structure.
+ * @addr: DMA location to place the payload.
+ * @size: Size of WR (including FW WR + immed data + rsp SG entry
+ *
+ * Wrapper for populating fw_scsi_cmd_wr.
+ */
+static inline void
+csio_scsi_init_cmd_wr(struct csio_ioreq *req, void *addr, uint32_t size)
+{
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_rnode *rn = req->rnode;
+ struct fw_scsi_cmd_wr *wr = (struct fw_scsi_cmd_wr *)addr;
+ struct csio_dma_buf *dma_buf;
+ uint8_t imm = csio_hw_to_scsim(hw)->proto_cmd_len;
+
+ wr->op_immdlen = cpu_to_be32(FW_WR_OP(FW_SCSI_CMD_WR) |
+ FW_SCSI_CMD_WR_IMMDLEN(imm));
+ wr->flowid_len16 = cpu_to_be32(FW_WR_FLOWID(rn->flowid) |
+ FW_WR_LEN16(
+ CSIO_ROUNDUP(size, 16)));
+
+ wr->cookie = (uintptr_t) req;
+ wr->iqid = (uint16_t)cpu_to_be16(csio_q_physiqid(hw, req->iq_idx));
+ wr->tmo_val = (uint8_t) req->tmo;
+ wr->r3 = 0;
+ memset(&wr->r5, 0, 8);
+
+ /* Get RSP DMA buffer */
+ dma_buf = &req->dma_buf;
+
+ /* Prepare RSP SGL */
+ wr->rsp_dmalen = cpu_to_be32(dma_buf->len);
+ wr->rsp_dmaaddr = cpu_to_be64(dma_buf->paddr);
+
+ wr->r6 = 0;
+
+ wr->u.fcoe.ctl_pri = 0;
+ wr->u.fcoe.cp_en_class = 0;
+ wr->u.fcoe.r4_lo[0] = 0;
+ wr->u.fcoe.r4_lo[1] = 0;
+
+ /* Frame a FCP command */
+ csio_scsi_fcp_cmnd(req, (void *)((uintptr_t)addr +
+ sizeof(struct fw_scsi_cmd_wr)));
+}
+
+#define CSIO_SCSI_CMD_WR_SZ(_imm) \
+ (sizeof(struct fw_scsi_cmd_wr) + /* WR size */ \
+ ALIGN((_imm), 16)) /* Immed data */
+
+#define CSIO_SCSI_CMD_WR_SZ_16(_imm) \
+ (ALIGN(CSIO_SCSI_CMD_WR_SZ((_imm)), 16))
+
+/*
+ * csio_scsi_cmd - Create a SCSI CMD WR.
+ * @req: IO req structure.
+ *
+ * Gets a WR slot in the ingress queue and initializes it with SCSI CMD WR.
+ *
+ */
+static inline void
+csio_scsi_cmd(struct csio_ioreq *req)
+{
+ struct csio_wr_pair wrp;
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_scsim *scsim = csio_hw_to_scsim(hw);
+ uint32_t size = CSIO_SCSI_CMD_WR_SZ_16(scsim->proto_cmd_len);
+
+ req->drv_status = csio_wr_get(hw, req->eq_idx, size, &wrp);
+ if (unlikely(req->drv_status != CSIO_SUCCESS))
+ return;
+
+ if (wrp.size1 >= size) {
+ /* Initialize WR in one shot */
+ csio_scsi_init_cmd_wr(req, wrp.addr1, size);
+ } else {
+ uint8_t tmpwr[512];
+ /*
+ * Make a temporary copy of the WR and write back
+ * the copy into the WR pair.
+ */
+ csio_scsi_init_cmd_wr(req, (void *)tmpwr, size);
+ memcpy(wrp.addr1, tmpwr, wrp.size1);
+ memcpy(wrp.addr2, tmpwr + wrp.size1, size - wrp.size1);
+ }
+}
+
+/*
+ * The following is fast path code. Therefore it is inlined with multi-line
+ * macros using name substitution, thus avoiding if-else switches for
+ * operation (read/write), as well as serving the purpose of code re-use.
+ */
+/*
+ * csio_scsi_init_ulptx_dsgl - Fill in a ULP_TX_SC_DSGL
+ * @hw: HW module
+ * @req: IO request
+ * @sgl: ULP TX SGL pointer.
+ *
+ */
+#define csio_scsi_init_ultptx_dsgl(hw, req, sgl) \
+do { \
+ struct ulptx_sge_pair *_sge_pair = NULL; \
+ struct scatterlist *_sgel; \
+ uint32_t _i = 0; \
+ uint32_t _xfer_len; \
+ struct list_head *_tmp; \
+ struct csio_dma_buf *_dma_buf; \
+ struct scsi_cmnd *scmnd = csio_scsi_cmnd((req)); \
+ \
+ (sgl)->cmd_nsge = htonl(ULPTX_CMD(ULP_TX_SC_DSGL) | ULPTX_MORE | \
+ ULPTX_NSGE((req)->nsge)); \
+ /* Now add the data SGLs */ \
+ if (likely(!(req)->dcopy)) { \
+ scsi_for_each_sg(scmnd, _sgel, (req)->nsge, _i) { \
+ if (_i == 0) { \
+ (sgl)->addr0 = cpu_to_be64( \
+ sg_dma_address(_sgel)); \
+ (sgl)->len0 = cpu_to_be32( \
+ sg_dma_len(_sgel)); \
+ _sge_pair = \
+ (struct ulptx_sge_pair *)((sgl) + 1); \
+ continue; \
+ } \
+ if ((_i - 1) & 0x1) { \
+ _sge_pair->addr[1] = cpu_to_be64( \
+ sg_dma_address(_sgel)); \
+ _sge_pair->len[1] = cpu_to_be32( \
+ sg_dma_len(_sgel)); \
+ _sge_pair++; \
+ } else { \
+ _sge_pair->addr[0] = cpu_to_be64( \
+ sg_dma_address(_sgel)); \
+ _sge_pair->len[0] = cpu_to_be32( \
+ sg_dma_len(_sgel)); \
+ } \
+ } \
+ } else { \
+ /* Program sg elements with driver's DDP buffer */ \
+ _xfer_len = scsi_bufflen(scmnd); \
+ list_for_each(_tmp, &(req)->gen_list) { \
+ _dma_buf = (struct csio_dma_buf *)_tmp; \
+ if (_i == 0) { \
+ (sgl)->addr0 = cpu_to_be64(_dma_buf->paddr); \
+ (sgl)->len0 = cpu_to_be32( \
+ min(_xfer_len, _dma_buf->len)); \
+ _sge_pair = \
+ (struct ulptx_sge_pair *)((sgl) + 1); \
+ } \
+ else if ((_i - 1) & 0x1) { \
+ _sge_pair->addr[1] = cpu_to_be64( \
+ _dma_buf->paddr); \
+ _sge_pair->len[1] = cpu_to_be32( \
+ min(_xfer_len, _dma_buf->len)); \
+ _sge_pair++; \
+ } else { \
+ _sge_pair->addr[0] = cpu_to_be64( \
+ _dma_buf->paddr); \
+ _sge_pair->len[0] = cpu_to_be32( \
+ min(_xfer_len, _dma_buf->len)); \
+ } \
+ _xfer_len -= min(_xfer_len, _dma_buf->len); \
+ _i++; \
+ } \
+ } \
+} while (0)
+
+/*
+ * csio_scsi_init_data_wr - Initialize the READ/WRITE SCSI WR.
+ * @req: IO req structure.
+ * @oper: read/write
+ * @wrp: DMA location to place the payload.
+ * @size: Size of WR (including FW WR + immed data + rsp SG entry + data SGL
+ * @wrop: _READ_/_WRITE_
+ *
+ * Wrapper for populating fw_scsi_read_wr/fw_scsi_write_wr.
+ */
+#define csio_scsi_init_data_wr(req, oper, wrp, size, wrop) \
+do { \
+ struct csio_hw *_hw = (req)->lnode->hwp; \
+ struct csio_rnode *_rn = (req)->rnode; \
+ struct fw_scsi_##oper##_wr *__wr = (struct fw_scsi_##oper##_wr *)(wrp);\
+ struct ulptx_sgl *_sgl; \
+ struct csio_dma_buf *_dma_buf; \
+ uint8_t _imm = csio_hw_to_scsim(_hw)->proto_cmd_len; \
+ struct scsi_cmnd *scmnd = csio_scsi_cmnd((req)); \
+ \
+ __wr->op_immdlen = cpu_to_be32(FW_WR_OP(FW_SCSI##wrop##WR) | \
+ FW_SCSI##wrop##WR_IMMDLEN(_imm)); \
+ __wr->flowid_len16 = cpu_to_be32(FW_WR_FLOWID(_rn->flowid) | \
+ FW_WR_LEN16( \
+ CSIO_ROUNDUP((size), 16))); \
+ __wr->cookie = (uintptr_t) (req); \
+ __wr->iqid = (uint16_t)cpu_to_be16(csio_q_physiqid(_hw, \
+ (req)->iq_idx));\
+ __wr->tmo_val = (uint8_t)((req)->tmo); \
+ __wr->use_xfer_cnt = 1; \
+ __wr->xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
+ __wr->ini_xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
+ /* Get RSP DMA buffer */ \
+ _dma_buf = &(req)->dma_buf; \
+ \
+ /* Prepare RSP SGL */ \
+ __wr->rsp_dmalen = cpu_to_be32(_dma_buf->len); \
+ __wr->rsp_dmaaddr = cpu_to_be64(_dma_buf->paddr); \
+ \
+ __wr->r4 = 0; \
+ \
+ __wr->u.fcoe.ctl_pri = 0; \
+ __wr->u.fcoe.cp_en_class = 0; \
+ __wr->u.fcoe.r3_lo[0] = 0; \
+ __wr->u.fcoe.r3_lo[1] = 0; \
+ csio_scsi_fcp_cmnd((req), (void *)((uintptr_t)(wrp) + \
+ sizeof(struct fw_scsi_##oper##_wr))); \
+ \
+ /* Move WR pointer past command and immediate data */ \
+ _sgl = (struct ulptx_sgl *) ((uintptr_t)(wrp) + \
+ sizeof(struct fw_scsi_##oper##_wr) + \
+ ALIGN(_imm, 16)); \
+ \
+ /* Fill in the DSGL */ \
+ csio_scsi_init_ultptx_dsgl(_hw, (req), _sgl); \
+ \
+} while (0)
+
+/* Calculate WR size needed for fw_scsi_read_wr/fw_scsi_write_wr */
+#define csio_scsi_data_wrsz(req, oper, sz, imm) \
+do { \
+ (sz) = sizeof(struct fw_scsi_##oper##_wr) + /* WR size */ \
+ ALIGN((imm), 16) + /* Immed data */ \
+ sizeof(struct ulptx_sgl); /* ulptx_sgl */ \
+ \
+ if (unlikely((req)->nsge > 1)) \
+ (sz) += (sizeof(struct ulptx_sge_pair) * \
+ (ALIGN(((req)->nsge - 1), 2) / 2)); \
+ /* Data SGE */ \
+} while (0)
+
+/*
+ * csio_scsi_data - Create a SCSI WRITE/READ WR.
+ * @req: IO req structure.
+ * @oper: read/write
+ * @wrop: _READ_/_WRITE_ (string subsitutions to use with the FW bit field
+ * macros).
+ *
+ * Gets a WR slot in the ingress queue and initializes it with
+ * SCSI CMD READ/WRITE WR.
+ *
+ */
+#define csio_scsi_data(req, oper, wrop) \
+do { \
+ struct csio_wr_pair _wrp; \
+ uint32_t _size; \
+ struct csio_hw *_hw = (req)->lnode->hwp; \
+ struct csio_scsim *_scsim = csio_hw_to_scsim(_hw); \
+ \
+ csio_scsi_data_wrsz((req), oper, _size, _scsim->proto_cmd_len); \
+ _size = ALIGN(_size, 16); \
+ \
+ (req)->drv_status = csio_wr_get(_hw, (req)->eq_idx, _size, &_wrp); \
+ if (likely((req)->drv_status == CSIO_SUCCESS)) { \
+ if (likely(_wrp.size1 >= _size)) { \
+ /* Initialize WR in one shot */ \
+ csio_scsi_init_data_wr((req), oper, _wrp.addr1, \
+ _size, wrop); \
+ } else { \
+ uint8_t tmpwr[512]; \
+ /* \
+ * Make a temporary copy of the WR and write back \
+ * the copy into the WR pair. \
+ */ \
+ csio_scsi_init_data_wr((req), oper, (void *)tmpwr, \
+ _size, wrop); \
+ memcpy(_wrp.addr1, tmpwr, _wrp.size1); \
+ memcpy(_wrp.addr2, tmpwr + _wrp.size1, \
+ _size - _wrp.size1); \
+ } \
+ } \
+} while (0)
+
+/*
+ * csio_setup_ddp - Setup DDP buffers for Read request.
+ * @req: IO req structure.
+ *
+ * Checks SGLs/Data buffers are virtually contiguous required for DDP.
+ * If contiguous,driver posts SGLs in the WR otherwise post internal
+ * buffers for such request for DDP.
+ */
+static inline void
+csio_setup_ddp(struct csio_scsim *scsim, struct csio_ioreq *req)
+{
+#ifdef __CSIO_DEBUG__
+ struct csio_hw *hw = req->lnode->hwp;
+#endif
+ struct scatterlist *sgel = NULL;
+ struct scsi_cmnd *scmnd = csio_scsi_cmnd(req);
+ uint64_t sg_addr = 0;
+ uint32_t ddp_pagesz = 4096;
+ uint32_t buf_off;
+ struct csio_dma_buf *dma_buf = NULL;
+ uint32_t alloc_len = 0;
+ uint32_t xfer_len = 0;
+ uint32_t sg_len = 0;
+ uint32_t i;
+
+ scsi_for_each_sg(scmnd, sgel, req->nsge, i) {
+ sg_addr = sg_dma_address(sgel);
+ sg_len = sg_dma_len(sgel);
+
+ buf_off = sg_addr & (ddp_pagesz - 1);
+
+ /* Except 1st buffer,all buffer addr have to be Page aligned */
+ if (i != 0 && buf_off) {
+ csio_dbg(hw, "SGL addr not DDP aligned (%llx:%d)\n",
+ sg_addr, sg_len);
+ goto unaligned;
+ }
+
+ /* Except last buffer,all buffer must end on page boundary */
+ if ((i != (req->nsge - 1)) &&
+ ((buf_off + sg_len) & (ddp_pagesz - 1))) {
+ csio_dbg(hw,
+ "SGL addr not ending on page boundary"
+ "(%llx:%d)\n", sg_addr, sg_len);
+ goto unaligned;
+ }
+ }
+
+ /* SGL's are virtually contiguous. HW will DDP to SGLs */
+ req->dcopy = 0;
+ csio_scsi_data(req, read, _READ_);
+
+ return;
+
+unaligned:
+ csio_inc_stats(scsim, n_unaligned);
+ /*
+ * For unaligned SGLs, driver will allocate internal DDP buffer.
+ * Once command is completed data from DDP buffer copied to SGLs
+ */
+ req->dcopy = 1;
+
+ /* Use gen_list to store the DDP buffers */
+ INIT_LIST_HEAD(&req->gen_list);
+ xfer_len = scsi_bufflen(scmnd);
+
+ i = 0;
+ /* Allocate ddp buffers for this request */
+ while (alloc_len < xfer_len) {
+ dma_buf = csio_get_scsi_ddp(scsim);
+ if (dma_buf == NULL || i > scsim->max_sge) {
+ req->drv_status = CSIO_BUSY;
+ break;
+ }
+ alloc_len += dma_buf->len;
+ /* Added to IO req */
+ list_add_tail(&dma_buf->list, &req->gen_list);
+ i++;
+ }
+
+ if (!req->drv_status) {
+ /* set number of ddp bufs used */
+ req->nsge = i;
+ csio_scsi_data(req, read, _READ_);
+ return;
+ }
+
+ /* release dma descs */
+ if (i > 0)
+ csio_put_scsi_ddp_list(scsim, &req->gen_list, i);
+}
+
+/*
+ * csio_scsi_init_abrt_cls_wr - Initialize an ABORT/CLOSE WR.
+ * @req: IO req structure.
+ * @addr: DMA location to place the payload.
+ * @size: Size of WR
+ * @abort: abort OR close
+ *
+ * Wrapper for populating fw_scsi_cmd_wr.
+ */
+static inline void
+csio_scsi_init_abrt_cls_wr(struct csio_ioreq *req, void *addr, uint32_t size,
+ bool abort)
+{
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_rnode *rn = req->rnode;
+ struct fw_scsi_abrt_cls_wr *wr = (struct fw_scsi_abrt_cls_wr *)addr;
+
+ wr->op_immdlen = cpu_to_be32(FW_WR_OP(FW_SCSI_ABRT_CLS_WR));
+ wr->flowid_len16 = cpu_to_be32(FW_WR_FLOWID(rn->flowid) |
+ FW_WR_LEN16(
+ CSIO_ROUNDUP(size, 16)));
+
+ wr->cookie = (uintptr_t) req;
+ wr->iqid = (uint16_t)cpu_to_be16(csio_q_physiqid(hw, req->iq_idx));
+ wr->tmo_val = (uint8_t) req->tmo;
+ /* 0 for CHK_ALL_IO tells FW to look up t_cookie */
+ wr->sub_opcode_to_chk_all_io =
+ (FW_SCSI_ABRT_CLS_WR_SUB_OPCODE(abort) |
+ FW_SCSI_ABRT_CLS_WR_CHK_ALL_IO(0));
+ wr->r3[0] = 0;
+ wr->r3[1] = 0;
+ wr->r3[2] = 0;
+ wr->r3[3] = 0;
+ /* Since we re-use the same ioreq for abort as well */
+ wr->t_cookie = (uintptr_t) req;
+}
+
+static inline void
+csio_scsi_abrt_cls(struct csio_ioreq *req, bool abort)
+{
+ struct csio_wr_pair wrp;
+ struct csio_hw *hw = req->lnode->hwp;
+ uint32_t size = ALIGN(sizeof(struct fw_scsi_abrt_cls_wr), 16);
+
+ req->drv_status = csio_wr_get(hw, req->eq_idx, size, &wrp);
+ if (req->drv_status != CSIO_SUCCESS)
+ return;
+
+ if (wrp.size1 >= size) {
+ /* Initialize WR in one shot */
+ csio_scsi_init_abrt_cls_wr(req, wrp.addr1, size, abort);
+ } else {
+ uint8_t tmpwr[512];
+ /*
+ * Make a temporary copy of the WR and write back
+ * the copy into the WR pair.
+ */
+ csio_scsi_init_abrt_cls_wr(req, (void *)tmpwr, size, abort);
+ memcpy(wrp.addr1, tmpwr, wrp.size1);
+ memcpy(wrp.addr2, tmpwr + wrp.size1, size - wrp.size1);
+ }
+}
+
+/*****************************************************************************/
+/* START: SCSI SM */
+/*****************************************************************************/
+static void
+csio_scsis_uninit(struct csio_ioreq *req, enum csio_scsi_ev evt)
+{
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_scsim *scsim = csio_hw_to_scsim(hw);
+
+ switch (evt) {
+
+ case CSIO_SCSIE_START_IO:
+
+ /* There is data */
+ if (req->nsge) {
+ if (req->datadir == CSIO_IOREQF_DMA_WRITE) {
+ req->dcopy = 0;
+ csio_scsi_data(req, write, _WRITE_);
+ } else
+ csio_setup_ddp(scsim, req);
+ } else {
+ csio_scsi_cmd(req);
+ }
+
+ if (likely(req->drv_status == CSIO_SUCCESS)) {
+ /* change state and enqueue on active_q */
+ csio_set_state(&req->sm, csio_scsis_io_active);
+ list_add_tail(&req->sm.sm_list, &scsim->active_q);
+ csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
+ csio_inc_stats(scsim, n_active);
+
+ return;
+ }
+ break;
+
+ case CSIO_SCSIE_START_TM:
+ csio_scsi_cmd(req);
+ if (req->drv_status == CSIO_SUCCESS) {
+ /*
+ * NOTE: We collect the affected I/Os prior to issuing
+ * LUN reset, and not after it. This is to prevent
+ * aborting I/Os that get issued after the LUN reset,
+ * but prior to LUN reset completion (in the event that
+ * the host stack has not blocked I/Os to a LUN that is
+ * being reset.
+ */
+ csio_set_state(&req->sm, csio_scsis_tm_active);
+ list_add_tail(&req->sm.sm_list, &scsim->active_q);
+ csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
+ csio_inc_stats(scsim, n_tm_active);
+ }
+ return;
+
+ case CSIO_SCSIE_ABORT:
+ case CSIO_SCSIE_CLOSE:
+ /*
+ * NOTE:
+ * We could get here due to :
+ * - a window in the cleanup path of the SCSI module
+ * (csio_scsi_abort_io()). Please see NOTE in this function.
+ * - a window in the time we tried to issue an abort/close
+ * of a request to FW, and the FW completed the request
+ * itself.
+ * Print a message for now, and return INVAL either way.
+ */
+ req->drv_status = CSIO_INVAL;
+ csio_warn(hw, "Trying to abort/close completed IO:%p!\n", req);
+ break;
+
+ default:
+ csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
+ CSIO_DB_ASSERT(0);
+ }
+}
+
+static void
+csio_scsis_io_active(struct csio_ioreq *req, enum csio_scsi_ev evt)
+{
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_scsim *scm = csio_hw_to_scsim(hw);
+ struct csio_rnode *rn;
+
+ switch (evt) {
+
+ case CSIO_SCSIE_COMPLETED:
+ csio_dec_stats(scm, n_active);
+ list_del_init(&req->sm.sm_list);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ /*
+ * In MSIX mode, with multiple queues, the SCSI compeltions
+ * could reach us sooner than the FW events sent to indicate
+ * I-T nexus loss (link down, remote device logo etc). We
+ * dont want to be returning such I/Os to the upper layer
+ * immediately, since we wouldnt have reported the I-T nexus
+ * loss itself. This forces us to serialize such completions
+ * with the reporting of the I-T nexus loss. Therefore, we
+ * internally queue up such up such completions in the rnode.
+ * The reporting of I-T nexus loss to the upper layer is then
+ * followed by the returning of I/Os in this internal queue.
+ * Having another state alongwith another queue helps us take
+ * actions for events such as ABORT received while we are
+ * in this rnode queue.
+ */
+ if (unlikely(req->wr_status != FW_SUCCESS)) {
+ rn = req->rnode;
+ /*
+ * FW says remote device is lost, but rnode
+ * doesnt reflect it.
+ */
+ if (csio_scsi_itnexus_loss_error(req->wr_status) &&
+ csio_is_rnode_ready(rn)) {
+ csio_set_state(&req->sm,
+ csio_scsis_shost_cmpl_await);
+ list_add_tail(&req->sm.sm_list,
+ &rn->host_cmpl_q);
+ }
+ }
+
+ break;
+
+ case CSIO_SCSIE_ABORT:
+ csio_scsi_abrt_cls(req, SCSI_ABORT);
+ if (req->drv_status == CSIO_SUCCESS) {
+ csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
+ csio_set_state(&req->sm, csio_scsis_aborting);
+ }
+ break;
+
+ case CSIO_SCSIE_CLOSE:
+ csio_scsi_abrt_cls(req, SCSI_CLOSE);
+ if (req->drv_status == CSIO_SUCCESS) {
+ csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
+ csio_set_state(&req->sm, csio_scsis_closing);
+ }
+ break;
+
+ case CSIO_SCSIE_DRVCLEANUP:
+ req->wr_status = FW_HOSTERROR;
+ csio_dec_stats(scm, n_active);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ break;
+
+ default:
+ csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
+ CSIO_DB_ASSERT(0);
+ }
+}
+
+static void
+csio_scsis_tm_active(struct csio_ioreq *req, enum csio_scsi_ev evt)
+{
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_scsim *scm = csio_hw_to_scsim(hw);
+
+ switch (evt) {
+
+ case CSIO_SCSIE_COMPLETED:
+ csio_dec_stats(scm, n_tm_active);
+ list_del_init(&req->sm.sm_list);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+
+ break;
+
+ case CSIO_SCSIE_ABORT:
+ csio_scsi_abrt_cls(req, SCSI_ABORT);
+ if (req->drv_status == CSIO_SUCCESS) {
+ csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
+ csio_set_state(&req->sm, csio_scsis_aborting);
+ }
+ break;
+
+
+ case CSIO_SCSIE_CLOSE:
+ csio_scsi_abrt_cls(req, SCSI_CLOSE);
+ if (req->drv_status == CSIO_SUCCESS) {
+ csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
+ csio_set_state(&req->sm, csio_scsis_closing);
+ }
+ break;
+
+ case CSIO_SCSIE_DRVCLEANUP:
+ req->wr_status = FW_HOSTERROR;
+ csio_dec_stats(scm, n_tm_active);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ break;
+
+ default:
+ csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
+ CSIO_DB_ASSERT(0);
+ }
+}
+
+static void
+csio_scsis_aborting(struct csio_ioreq *req, enum csio_scsi_ev evt)
+{
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_scsim *scm = csio_hw_to_scsim(hw);
+
+ switch (evt) {
+
+ case CSIO_SCSIE_COMPLETED:
+ csio_dbg(hw,
+ "ioreq %p recvd cmpltd (wr_status:%d) "
+ "in aborting st\n", req, req->wr_status);
+ /*
+ * Use CSIO_CANCELLED to explicitly tell the ABORTED event that
+ * the original I/O was returned to driver by FW.
+ * We dont really care if the I/O was returned with success by
+ * FW (because the ABORT and completion of the I/O crossed each
+ * other), or any other return value. Once we are in aborting
+ * state, the success or failure of the I/O is unimportant to
+ * us.
+ */
+ req->drv_status = CSIO_CANCELLED;
+ break;
+
+ case CSIO_SCSIE_ABORT:
+ csio_inc_stats(scm, n_abrt_dups);
+ break;
+
+ case CSIO_SCSIE_ABORTED:
+
+ csio_dbg(hw, "abort of %p return status:0x%x drv_status:%x\n",
+ req, req->wr_status, req->drv_status);
+ /*
+ * Check if original I/O WR completed before the Abort
+ * completion.
+ */
+ if (req->drv_status != CSIO_CANCELLED) {
+ csio_fatal(hw,
+ "Abort completed before original I/O,"
+ " req:%p\n", req);
+ CSIO_DB_ASSERT(0);
+ }
+
+ /*
+ * There are the following possible scenarios:
+ * 1. The abort completed successfully, FW returned FW_SUCCESS.
+ * 2. The completion of an I/O and the receipt of
+ * abort for that I/O by the FW crossed each other.
+ * The FW returned FW_EINVAL. The original I/O would have
+ * returned with FW_SUCCESS or any other SCSI error.
+ * 3. The FW couldnt sent the abort out on the wire, as there
+ * was an I-T nexus loss (link down, remote device logged
+ * out etc). FW sent back an appropriate IT nexus loss status
+ * for the abort.
+ * 4. FW sent an abort, but abort timed out (remote device
+ * didnt respond). FW replied back with
+ * FW_SCSI_ABORT_TIMEDOUT.
+ * 5. FW couldnt genuinely abort the request for some reason,
+ * and sent us an error.
+ *
+ * The first 3 scenarios are treated as succesful abort
+ * operations by the host, while the last 2 are failed attempts
+ * to abort. Manipulate the return value of the request
+ * appropriately, so that host can convey these results
+ * back to the upper layer.
+ */
+ if ((req->wr_status == FW_SUCCESS) ||
+ (req->wr_status == FW_EINVAL) ||
+ csio_scsi_itnexus_loss_error(req->wr_status))
+ req->wr_status = FW_SCSI_ABORT_REQUESTED;
+
+ csio_dec_stats(scm, n_active);
+ list_del_init(&req->sm.sm_list);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ break;
+
+ case CSIO_SCSIE_DRVCLEANUP:
+ req->wr_status = FW_HOSTERROR;
+ csio_dec_stats(scm, n_active);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ break;
+
+ case CSIO_SCSIE_CLOSE:
+ /*
+ * We can receive this event from the module
+ * cleanup paths, if the FW forgot to reply to the ABORT WR
+ * and left this ioreq in this state. For now, just ignore
+ * the event. The CLOSE event is sent to this state, as
+ * the LINK may have already gone down.
+ */
+ break;
+
+ default:
+ csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
+ CSIO_DB_ASSERT(0);
+ }
+}
+
+static void
+csio_scsis_closing(struct csio_ioreq *req, enum csio_scsi_ev evt)
+{
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_scsim *scm = csio_hw_to_scsim(hw);
+
+ switch (evt) {
+
+ case CSIO_SCSIE_COMPLETED:
+ csio_dbg(hw,
+ "ioreq %p recvd cmpltd (wr_status:%d) "
+ "in closing st\n", req, req->wr_status);
+ /*
+ * Use CSIO_CANCELLED to explicitly tell the CLOSED event that
+ * the original I/O was returned to driver by FW.
+ * We dont really care if the I/O was returned with success by
+ * FW (because the CLOSE and completion of the I/O crossed each
+ * other), or any other return value. Once we are in aborting
+ * state, the success or failure of the I/O is unimportant to
+ * us.
+ */
+ req->drv_status = CSIO_CANCELLED;
+ break;
+
+ case CSIO_SCSIE_CLOSED:
+ /*
+ * Check if original I/O WR completed before the Close
+ * completion.
+ */
+ if (req->drv_status != CSIO_CANCELLED) {
+ csio_fatal(hw,
+ "Close completed before original I/O,"
+ " req:%p\n", req);
+ CSIO_DB_ASSERT(0);
+ }
+
+ /*
+ * Either close succeeded, or we issued close to FW at the
+ * same time FW compelted it to us. Either way, the I/O
+ * is closed.
+ */
+ CSIO_DB_ASSERT((req->wr_status == FW_SUCCESS) ||
+ (req->wr_status == FW_EINVAL));
+ req->wr_status = FW_SCSI_CLOSE_REQUESTED;
+
+ csio_dec_stats(scm, n_active);
+ list_del_init(&req->sm.sm_list);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ break;
+
+ case CSIO_SCSIE_CLOSE:
+ break;
+
+ case CSIO_SCSIE_DRVCLEANUP:
+ req->wr_status = FW_HOSTERROR;
+ csio_dec_stats(scm, n_active);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ break;
+
+ default:
+ csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
+ CSIO_DB_ASSERT(0);
+ }
+}
+
+static void
+csio_scsis_shost_cmpl_await(struct csio_ioreq *req, enum csio_scsi_ev evt)
+{
+ switch (evt) {
+ case CSIO_SCSIE_ABORT:
+ case CSIO_SCSIE_CLOSE:
+ /*
+ * Just succeed the abort request, and hope that
+ * the remote device unregister path will cleanup
+ * this I/O to the upper layer within a sane
+ * amount of time.
+ */
+ /*
+ * A close can come in during a LINK DOWN. The FW would have
+ * returned us the I/O back, but not the remote device lost
+ * FW event. In this interval, if the I/O times out at the upper
+ * layer, a close can come in. Take the same action as abort:
+ * return success, and hope that the remote device unregister
+ * path will cleanup this I/O. If the FW still doesnt send
+ * the msg, the close times out, and the upper layer resorts
+ * to the next level of error recovery.
+ */
+ req->drv_status = CSIO_SUCCESS;
+ break;
+ case CSIO_SCSIE_DRVCLEANUP:
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ break;
+ default:
+ csio_dbg(req->lnode->hwp, "Unhandled event:%d sent to req:%p\n",
+ evt, req);
+ CSIO_DB_ASSERT(0);
+ }
+}
+
+/*
+ * csio_scsi_cmpl_handler - WR completion handler for SCSI.
+ * @hw: HW module.
+ * @wr: The completed WR from the ingress queue.
+ * @len: Length of the WR.
+ * @flb: Freelist buffer array.
+ * @priv: Private object
+ * @scsiwr: Pointer to SCSI WR.
+ *
+ * This is the WR completion handler called per completion from the
+ * ISR. It is called with lock held. It walks past the RSS and CPL message
+ * header where the actual WR is present.
+ * It then gets the status, WR handle (ioreq pointer) and the len of
+ * the WR, based on WR opcode. Only on a non-good status is the entire
+ * WR copied into the WR cache (ioreq->fw_wr).
+ * The ioreq corresponding to the WR is returned to the caller.
+ * NOTE: The SCSI queue doesnt allocate a freelist today, hence
+ * no freelist buffer is expected.
+ */
+struct csio_ioreq *
+csio_scsi_cmpl_handler(struct csio_hw *hw, void *wr, uint32_t len,
+ struct csio_fl_dma_buf *flb, void *priv, uint8_t **scsiwr)
+{
+ struct csio_ioreq *ioreq = NULL;
+ struct cpl_fw6_msg *cpl;
+ uint8_t *tempwr;
+ uint8_t status;
+ struct csio_scsim *scm = csio_hw_to_scsim(hw);
+
+ /* skip RSS header */
+ cpl = (struct cpl_fw6_msg *)((uintptr_t)wr + sizeof(__be64));
+
+ if (unlikely(cpl->opcode != CPL_FW6_MSG)) {
+ csio_warn(hw, "Error: Invalid CPL msg %x recvd on SCSI q\n",
+ cpl->opcode);
+ csio_inc_stats(scm, n_inval_cplop);
+ return NULL;
+ }
+
+ tempwr = (uint8_t *)(cpl->data);
+ status = csio_wr_status(tempwr);
+ *scsiwr = tempwr;
+
+ if (likely((*tempwr == FW_SCSI_READ_WR) ||
+ (*tempwr == FW_SCSI_WRITE_WR) ||
+ (*tempwr == FW_SCSI_CMD_WR))) {
+ ioreq = (struct csio_ioreq *)((uintptr_t)
+ (((struct fw_scsi_read_wr *)tempwr)->cookie));
+ CSIO_DB_ASSERT(virt_addr_valid(ioreq));
+
+ ioreq->wr_status = status;
+
+ return ioreq;
+ }
+
+ if (*tempwr == FW_SCSI_ABRT_CLS_WR) {
+ ioreq = (struct csio_ioreq *)((uintptr_t)
+ (((struct fw_scsi_abrt_cls_wr *)tempwr)->cookie));
+ CSIO_DB_ASSERT(virt_addr_valid(ioreq));
+
+ ioreq->wr_status = status;
+ return ioreq;
+ }
+
+ csio_warn(hw, "WR with invalid opcode in SCSI IQ: %x\n", *tempwr);
+ csio_inc_stats(scm, n_inval_scsiop);
+ return NULL;
+}
+
+/*
+ * csio_scsi_cleanup_io_q - Cleanup the given queue.
+ * @scm: SCSI module.
+ * @q: Queue to be cleaned up.
+ *
+ * Called with lock held. Has to exit with lock held.
+ */
+void
+csio_scsi_cleanup_io_q(struct csio_scsim *scm, struct list_head *q)
+{
+ struct csio_hw *hw = scm->hw;
+ struct csio_ioreq *ioreq;
+ struct list_head *tmp, *next;
+ struct scsi_cmnd *scmnd;
+
+ /* Call back the completion routines of the active_q */
+ list_for_each_safe(tmp, next, q) {
+ ioreq = (struct csio_ioreq *)tmp;
+ csio_scsi_drvcleanup(ioreq);
+ list_del_init(&ioreq->sm.sm_list);
+ scmnd = csio_scsi_cmnd(ioreq);
+ spin_unlock_irq(&hw->lock);
+
+ /*
+ * Upper layers may have cleared this command, hence this
+ * check to avoid accessing stale references.
+ */
+ if (scmnd != NULL)
+ ioreq->io_cbfn(hw, ioreq);
+
+ spin_lock_irq(&scm->freelist_lock);
+ csio_put_scsi_ioreq(scm, ioreq);
+ spin_unlock_irq(&scm->freelist_lock);
+
+ spin_lock_irq(&hw->lock);
+ }
+}
+
+#define CSIO_SCSI_ABORT_Q_POLL_MS 2000
+
+static void
+csio_abrt_cls(struct csio_ioreq *ioreq, struct scsi_cmnd *scmnd)
+{
+ struct csio_lnode *ln = ioreq->lnode;
+ struct csio_hw *hw = ln->hwp;
+ int ready = 0;
+ struct csio_scsim *scsim = csio_hw_to_scsim(hw);
+ csio_retval_t rv;
+
+ if (csio_scsi_cmnd(ioreq) != scmnd) {
+ csio_inc_stats(scsim, n_abrt_race_comp);
+ return;
+ }
+
+ ready = csio_is_lnode_ready(ln);
+
+ rv = csio_do_abrt_cls(hw, ioreq, (ready ? SCSI_ABORT : SCSI_CLOSE));
+ if (rv != CSIO_SUCCESS) {
+ if (ready)
+ csio_inc_stats(scsim, n_abrt_busy_error);
+ else
+ csio_inc_stats(scsim, n_cls_busy_error);
+ }
+}
+
+/*
+ * csio_scsi_abort_io_q - Abort all I/Os on given queue
+ * @scm: SCSI module.
+ * @q: Queue to abort.
+ * @tmo: Timeout in ms
+ *
+ * Attempt to abort all I/Os on given queue, and wait for a max
+ * of tmo milliseconds for them to complete. Returns success
+ * if all I/Os are aborted. Else returns CSIO_TIMEOUT.
+ * Should be entered with lock held. Exits with lock held.
+ * NOTE:
+ * Lock has to be held across the loop that aborts I/Os, since dropping the lock
+ * in between can cause the list to be corrupted. As a result, the caller
+ * of this function has to ensure that the number of I/os to be aborted
+ * is finite enough to not cause lock-held-for-too-long issues.
+ */
+static csio_retval_t
+csio_scsi_abort_io_q(struct csio_scsim *scm, struct list_head *q, uint32_t tmo)
+{
+ struct csio_hw *hw = scm->hw;
+ struct list_head *tmp, *next;
+ int count = CSIO_ROUNDUP(tmo, CSIO_SCSI_ABORT_Q_POLL_MS);
+ struct scsi_cmnd *scmnd;
+
+ if (list_empty(q))
+ return CSIO_SUCCESS;
+
+ csio_dbg(hw, "Aborting SCSI I/Os\n");
+
+ /* Now abort/close I/Os in the queue passed */
+ list_for_each_safe(tmp, next, q) {
+ scmnd = csio_scsi_cmnd((struct csio_ioreq *)tmp);
+ csio_abrt_cls((struct csio_ioreq *)tmp, scmnd);
+ }
+
+ /* Wait till all active I/Os are completed/aborted/closed */
+ while (!list_empty(q) && count--) {
+ spin_unlock_irq(&hw->lock);
+ msleep(CSIO_SCSI_ABORT_Q_POLL_MS);
+ spin_lock_irq(&hw->lock);
+ }
+
+ /* all aborts completed */
+ if (list_empty(q))
+ return CSIO_SUCCESS;
+
+ return CSIO_TIMEOUT;
+}
+
+/*
+ * csio_scsim_cleanup_io - Cleanup all I/Os in SCSI module.
+ * @scm: SCSI module.
+ * @abort: abort required.
+ * Called with lock held, should exit with lock held.
+ * Can sleep when waiting for I/Os to complete.
+ */
+csio_retval_t
+csio_scsim_cleanup_io(struct csio_scsim *scm, bool abort)
+{
+ struct csio_hw *hw = scm->hw;
+ csio_retval_t rv = CSIO_SUCCESS;
+ int count = CSIO_ROUNDUP(60 * 1000, CSIO_SCSI_ABORT_Q_POLL_MS);
+
+ /* No I/Os pending */
+ if (list_empty(&scm->active_q))
+ return CSIO_SUCCESS;
+
+ /* Wait until all active I/Os are completed */
+ while (!list_empty(&scm->active_q) && count--) {
+ spin_unlock_irq(&hw->lock);
+ msleep(CSIO_SCSI_ABORT_Q_POLL_MS);
+ spin_lock_irq(&hw->lock);
+ }
+
+ /* all I/Os completed */
+ if (list_empty(&scm->active_q))
+ return CSIO_SUCCESS;
+
+ /* Else abort */
+ if (abort) {
+ rv = csio_scsi_abort_io_q(scm, &scm->active_q, 30000);
+ if (rv == CSIO_SUCCESS)
+ return rv;
+ csio_dbg(hw, "Some I/O aborts timed out, cleaning up..\n");
+ }
+
+ csio_scsi_cleanup_io_q(scm, &scm->active_q);
+
+ CSIO_DB_ASSERT(list_empty(&scm->active_q));
+
+ return rv;
+}
+
+/*
+ * csio_scsim_cleanup_io_lnode - Cleanup all I/Os of given lnode.
+ * @scm: SCSI module.
+ * @lnode: lnode
+ *
+ * Called with lock held, should exit with lock held.
+ * Can sleep (with dropped lock) when waiting for I/Os to complete.
+ */
+csio_retval_t
+csio_scsim_cleanup_io_lnode(struct csio_scsim *scm, struct csio_lnode *ln)
+{
+ struct csio_hw *hw = scm->hw;
+ struct csio_scsi_level_data sld;
+ csio_retval_t rv;
+ int count = CSIO_ROUNDUP(60 * 1000, CSIO_SCSI_ABORT_Q_POLL_MS);
+
+ csio_dbg(hw, "Gathering all SCSI I/Os on lnode %p\n", ln);
+
+ sld.level = CSIO_LEV_LNODE;
+ sld.lnode = ln;
+ INIT_LIST_HEAD(&ln->cmpl_q);
+ csio_scsi_gather_active_ios(scm, &sld, &ln->cmpl_q);
+
+ /* No I/Os pending on this lnode */
+ if (list_empty(&ln->cmpl_q))
+ return CSIO_SUCCESS;
+
+ /* Wait until all active I/Os on this lnode are completed */
+ while (!list_empty(&ln->cmpl_q) && count--) {
+ spin_unlock_irq(&hw->lock);
+ msleep(CSIO_SCSI_ABORT_Q_POLL_MS);
+ spin_lock_irq(&hw->lock);
+ }
+
+ /* all I/Os completed */
+ if (list_empty(&ln->cmpl_q))
+ return CSIO_SUCCESS;
+
+ csio_dbg(hw, "Some I/Os pending on ln:%p, aborting them..\n", ln);
+
+ /* I/Os are pending, abort them */
+ rv = csio_scsi_abort_io_q(scm, &ln->cmpl_q, 30000);
+ if (rv != CSIO_SUCCESS) {
+ csio_dbg(hw, "Some I/O aborts timed out, cleaning up..\n");
+ csio_scsi_cleanup_io_q(scm, &ln->cmpl_q);
+ }
+
+ CSIO_DB_ASSERT(list_empty(&ln->cmpl_q));
+
+ return rv;
+}
+
+static ssize_t
+csio_show_hw_state(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ struct csio_lnode *ln = shost_priv(class_to_shost(dev));
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ if (csio_is_hw_ready(hw))
+ return snprintf(buf, PAGE_SIZE, "ready\n");
+ else
+ return snprintf(buf, PAGE_SIZE, "not ready\n");
+}
+
+/* Device reset */
+static ssize_t
+csio_device_reset(struct device *dev,
+ struct device_attribute *attr, const char *buf, size_t count)
+{
+ struct csio_lnode *ln = shost_priv(class_to_shost(dev));
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ if (*buf != '1')
+ return -EINVAL;
+
+ /* Delete NPIV lnodes */
+ csio_lnodes_exit(hw, 1);
+
+ /* Block upper IOs */
+ csio_lnodes_block_request(hw);
+
+ spin_lock_irq(&hw->lock);
+ csio_hw_reset(hw);
+ spin_unlock_irq(&hw->lock);
+
+ /* Unblock upper IOs */
+ csio_lnodes_unblock_request(hw);
+ return count;
+}
+
+/* disable port */
+static ssize_t
+csio_disable_port(struct device *dev,
+ struct device_attribute *attr, const char *buf, size_t count)
+{
+ struct csio_lnode *ln = shost_priv(class_to_shost(dev));
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ bool disable;
+
+ if (*buf == '1' || *buf == '0')
+ disable = (*buf == '1') ? CSIO_TRUE : CSIO_FALSE;
+ else
+ return -EINVAL;
+
+ /* Block upper IOs */
+ csio_lnodes_block_by_port(hw, ln->portid);
+
+ spin_lock_irq(&hw->lock);
+ csio_disable_lnodes(hw, ln->portid, disable);
+ spin_unlock_irq(&hw->lock);
+
+ /* Unblock upper IOs */
+ csio_lnodes_unblock_by_port(hw, ln->portid);
+ return count;
+}
+
+/* Show debug level */
+static ssize_t
+csio_show_dbg_level(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ struct csio_lnode *ln = shost_priv(class_to_shost(dev));
+
+ return snprintf(buf, PAGE_SIZE, "%x\n", ln->params.log_level);
+}
+
+/* Store debug level */
+static ssize_t
+csio_store_dbg_level(struct device *dev,
+ struct device_attribute *attr, const char *buf, size_t count)
+{
+ struct csio_lnode *ln = shost_priv(class_to_shost(dev));
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ uint32_t dbg_level = 0;
+
+ if (!isdigit(buf[0]))
+ return -EINVAL;
+
+ if (sscanf(buf, "%i", &dbg_level))
+ return -EINVAL;
+
+ ln->params.log_level = dbg_level;
+ hw->params.log_level = dbg_level;
+
+ return 0;
+}
+
+static DEVICE_ATTR(hw_state, S_IRUGO, csio_show_hw_state, NULL);
+static DEVICE_ATTR(device_reset, S_IRUGO | S_IWUSR, NULL, csio_device_reset);
+static DEVICE_ATTR(disable_port, S_IRUGO | S_IWUSR, NULL, csio_disable_port);
+static DEVICE_ATTR(dbg_level, S_IRUGO | S_IWUSR, csio_show_dbg_level,
+ csio_store_dbg_level);
+
+static struct device_attribute *csio_fcoe_lport_attrs[] = {
+ &dev_attr_hw_state,
+ &dev_attr_device_reset,
+ &dev_attr_disable_port,
+ &dev_attr_dbg_level,
+ NULL,
+};
+
+static ssize_t
+csio_show_num_reg_rnodes(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ struct csio_lnode *ln = shost_priv(class_to_shost(dev));
+
+ return snprintf(buf, PAGE_SIZE, "%d\n", ln->num_reg_rnodes);
+}
+
+static DEVICE_ATTR(num_reg_rnodes, S_IRUGO, csio_show_num_reg_rnodes, NULL);
+
+static struct device_attribute *csio_fcoe_vport_attrs[] = {
+ &dev_attr_num_reg_rnodes,
+ &dev_attr_dbg_level,
+ NULL,
+};
+
+static inline uint32_t
+csio_scsi_copy_to_sgl(struct csio_hw *hw, struct csio_ioreq *req)
+{
+ struct scsi_cmnd *scmnd = (struct scsi_cmnd *)csio_scsi_cmnd(req);
+ struct scatterlist *sg;
+ uint32_t bytes_left;
+ uint32_t bytes_copy;
+ uint32_t buf_off = 0;
+ uint32_t start_off = 0;
+ uint32_t sg_off = 0;
+ void *sg_addr;
+ void *buf_addr;
+ struct csio_dma_buf *dma_buf;
+
+ bytes_left = scsi_bufflen(scmnd);
+ sg = scsi_sglist(scmnd);
+ dma_buf = (struct csio_dma_buf *)csio_list_next(&req->gen_list);
+
+ /* Copy data from driver buffer to SGs of SCSI CMD */
+ while (bytes_left > 0 && sg && dma_buf) {
+ if (buf_off >= dma_buf->len) {
+ buf_off = 0;
+ dma_buf = (struct csio_dma_buf *)
+ csio_list_next(dma_buf);
+ continue;
+ }
+
+ if (start_off >= sg->length) {
+ start_off -= sg->length;
+ sg = sg_next(sg);
+ continue;
+ }
+
+ buf_addr = dma_buf->vaddr + buf_off;
+ sg_off = sg->offset + start_off;
+ bytes_copy = min((dma_buf->len - buf_off),
+ sg->length - start_off);
+ bytes_copy = min((uint32_t)(PAGE_SIZE - (sg_off & ~PAGE_MASK)),
+ bytes_copy);
+
+ sg_addr = kmap_atomic(sg_page(sg) + (sg_off >> PAGE_SHIFT));
+ if (!sg_addr) {
+ csio_err(hw, "failed to kmap sg:%p of ioreq:%p\n",
+ sg, req);
+ break;
+ }
+
+ csio_dbg(hw, "copy_to_sgl:sg_addr %p sg_off %d buf %p len %d\n",
+ sg_addr, sg_off, buf_addr, bytes_copy);
+ memcpy(sg_addr + (sg_off & ~PAGE_MASK), buf_addr, bytes_copy);
+ kunmap_atomic(sg_addr);
+
+ start_off += bytes_copy;
+ buf_off += bytes_copy;
+ bytes_left -= bytes_copy;
+ }
+
+ if (bytes_left > 0)
+ return DID_ERROR;
+ else
+ return DID_OK;
+}
+
+/*
+ * csio_scsi_err_handler - SCSI error handler.
+ * @hw: HW module.
+ * @req: IO request.
+ *
+ */
+static inline void
+csio_scsi_err_handler(struct csio_hw *hw, struct csio_ioreq *req)
+{
+ struct scsi_cmnd *cmnd = (struct scsi_cmnd *)csio_scsi_cmnd(req);
+ struct csio_scsim *scm = csio_hw_to_scsim(hw);
+ struct csio_fcp_resp *fcp_resp;
+ struct csio_dma_buf *dma_buf;
+ uint8_t flags, scsi_status = 0;
+ uint32_t host_status = DID_OK;
+ uint32_t rsp_len = 0, sns_len = 0;
+ struct csio_rnode *rn = (struct csio_rnode *)(cmnd->device->hostdata);
+
+
+ switch (req->wr_status) {
+ case FW_HOSTERROR:
+ if (unlikely(!csio_is_hw_ready(hw)))
+ return;
+
+ host_status = DID_ERROR;
+ csio_inc_stats(scm, n_hosterror);
+
+ break;
+ case FW_SCSI_RSP_ERR:
+ dma_buf = &req->dma_buf;
+ fcp_resp = (struct csio_fcp_resp *)dma_buf->vaddr;
+ flags = fcp_resp->flags;
+ scsi_status = fcp_resp->scsi_status;
+
+ if (flags & FCP_RSP_LEN_VAL) {
+ rsp_len = be32_to_cpu(fcp_resp->rsp_len);
+ if ((rsp_len != 0 && rsp_len != 4 && rsp_len != 8) ||
+ (fcp_resp->rsp_code != FCP_TMF_CMPL)) {
+ host_status = DID_ERROR;
+ goto out;
+ }
+ }
+
+ if ((flags & FCP_SNS_LEN_VAL) && fcp_resp->sns_len) {
+ sns_len = be32_to_cpu(fcp_resp->sns_len);
+ if (sns_len > SCSI_SENSE_BUFFERSIZE)
+ sns_len = SCSI_SENSE_BUFFERSIZE;
+
+ memcpy(cmnd->sense_buffer, &fcp_resp->rsvd1 + rsp_len,
+ sns_len);
+ csio_inc_stats(scm, n_autosense);
+ }
+
+ scsi_set_resid(cmnd, 0);
+
+ /* Under run */
+ if (flags & FCP_RESID_UNDER) {
+ scsi_set_resid(cmnd, be32_to_cpu(fcp_resp->resid));
+
+ if (!(flags & FCP_SNS_LEN_VAL) &&
+ (scsi_status == SAM_STAT_GOOD) &&
+ ((scsi_bufflen(cmnd) - scsi_get_resid(cmnd))
+ < cmnd->underflow))
+ host_status = DID_ERROR;
+ } else if (flags & FCP_RESID_OVER)
+ host_status = DID_ERROR;
+
+ csio_inc_stats(scm, n_rsperror);
+ break;
+
+ case FW_SCSI_OVER_FLOW_ERR:
+ csio_warn(hw,
+ "Over-flow error,cmnd:0x%x expected len:0x%x"
+ " resid:0x%x\n", cmnd->cmnd[0],
+ scsi_bufflen(cmnd), scsi_get_resid(cmnd));
+ host_status = DID_ERROR;
+ csio_inc_stats(scm, n_ovflerror);
+ break;
+
+ case FW_SCSI_UNDER_FLOW_ERR:
+ csio_warn(hw,
+ "Under-flow error,cmnd:0x%x expected"
+ " len:0x%x resid:0x%x lun:0x%x ssn:0x%x\n",
+ cmnd->cmnd[0], scsi_bufflen(cmnd),
+ scsi_get_resid(cmnd), cmnd->device->lun,
+ rn->flowid);
+ host_status = DID_ERROR;
+ csio_inc_stats(scm, n_unflerror);
+ break;
+
+ case FW_SCSI_ABORT_REQUESTED:
+ case FW_SCSI_ABORTED:
+ case FW_SCSI_CLOSE_REQUESTED:
+ csio_dbg(hw, "Req %p cmd:%p op:%x %s\n", req, cmnd,
+ cmnd->cmnd[0],
+ (req->wr_status == FW_SCSI_CLOSE_REQUESTED) ?
+ "closed" : "aborted");
+ /*
+ * csio_eh_abort_handler checks this value to
+ * succeed or fail the abort request.
+ */
+ host_status = DID_REQUEUE;
+ if (req->wr_status == FW_SCSI_CLOSE_REQUESTED)
+ csio_inc_stats(scm, n_closed);
+ else
+ csio_inc_stats(scm, n_aborted);
+ break;
+
+ case FW_SCSI_ABORT_TIMEDOUT:
+ /* FW timed out the abort itself */
+ csio_dbg(hw, "FW timed out abort req:%p cmnd:%p status:%x\n",
+ req, cmnd, req->wr_status);
+ host_status = DID_ERROR;
+ csio_inc_stats(scm, n_abrt_timedout);
+ break;
+
+ case FW_RDEV_NOT_READY:
+ /*
+ * In firmware, a RDEV can get into this state
+ * temporarily, before moving into dissapeared/lost
+ * state. So, the driver should complete the request equivalent
+ * to device-disappeared!
+ */
+ csio_inc_stats(scm, n_rdev_nr_error);
+ host_status = DID_ERROR;
+ break;
+
+ case FW_ERR_RDEV_LOST:
+ csio_inc_stats(scm, n_rdev_lost_error);
+ host_status = DID_ERROR;
+ break;
+
+ case FW_ERR_RDEV_LOGO:
+ csio_inc_stats(scm, n_rdev_logo_error);
+ host_status = DID_ERROR;
+ break;
+
+ case FW_ERR_RDEV_IMPL_LOGO:
+ host_status = DID_ERROR;
+ break;
+
+ case FW_ERR_LINK_DOWN:
+ csio_inc_stats(scm, n_link_down_error);
+ host_status = DID_ERROR;
+ break;
+
+ case FW_FCOE_NO_XCHG:
+ csio_inc_stats(scm, n_no_xchg_error);
+ host_status = DID_ERROR;
+ break;
+
+ default:
+ csio_err(hw, "Unknown SCSI FW WR status:%d req:%p cmnd:%p\n",
+ req->wr_status, req, cmnd);
+ CSIO_DB_ASSERT(0);
+
+ csio_inc_stats(scm, n_unknown_error);
+ host_status = DID_ERROR;
+ break;
+ }
+
+out:
+ if (req->nsge > 0)
+ scsi_dma_unmap(cmnd);
+
+ cmnd->result = (((host_status) << 16) | scsi_status);
+ cmnd->scsi_done(cmnd);
+
+ /* Wake up waiting threads */
+ csio_scsi_cmnd(req) = NULL;
+ complete_all(&req->cmplobj);
+}
+
+/*
+ * csio_scsi_cbfn - SCSI callback function.
+ * @hw: HW module.
+ * @req: IO request.
+ *
+ */
+static void
+csio_scsi_cbfn(struct csio_hw *hw, struct csio_ioreq *req)
+{
+ struct scsi_cmnd *cmnd = (struct scsi_cmnd *)csio_scsi_cmnd(req);
+ uint8_t scsi_status = SAM_STAT_GOOD;
+ uint32_t host_status = DID_OK;
+
+ if (likely(req->wr_status == FW_SUCCESS)) {
+ if (req->nsge > 0) {
+ scsi_dma_unmap(cmnd);
+ if (req->dcopy)
+ host_status = csio_scsi_copy_to_sgl(hw, req);
+ }
+
+ cmnd->result = (((host_status) << 16) | scsi_status);
+ cmnd->scsi_done(cmnd);
+ csio_scsi_cmnd(req) = NULL;
+ csio_inc_stats(csio_hw_to_scsim(hw), n_tot_success);
+ } else {
+ /* Error handling */
+ csio_scsi_err_handler(hw, req);
+ }
+}
+
+/**
+ * csio_queuecommand_lck - Entry point to kickstart an I/O request.
+ * @cmnd: The I/O request from ML.
+ * @done: The ML callback routine.
+ *
+ * This routine does the following:
+ * - Checks for HW and Rnode module readiness.
+ * - Gets a free ioreq structure (which is already initialized
+ * to uninit during its allocation).
+ * - Maps SG elements.
+ * - Initializes ioreq members.
+ * - Kicks off the SCSI state machine for this IO.
+ * - Returns busy status on error.
+ */
+static int
+csio_queuecommand_lck(struct scsi_cmnd *cmnd, void (*done)(struct scsi_cmnd *))
+{
+ struct csio_lnode *ln = shost_priv(cmnd->device->host);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ struct csio_scsim *scsim = csio_hw_to_scsim(hw);
+ struct csio_rnode *rn = (struct csio_rnode *)(cmnd->device->hostdata);
+ struct csio_ioreq *ioreq = NULL;
+ unsigned long flags;
+ int nsge = 0;
+ int rv = SCSI_MLQUEUE_HOST_BUSY, nr;
+ csio_retval_t retval;
+ int cpu;
+ struct csio_scsi_qset *sqset;
+ struct fc_rport *rport = starget_to_rport(scsi_target(cmnd->device));
+
+ if (!blk_rq_cpu_valid(cmnd->request))
+ cpu = smp_processor_id();
+ else
+ cpu = cmnd->request->cpu;
+
+ sqset = &hw->sqset[ln->portid][cpu];
+
+ nr = fc_remote_port_chkready(rport);
+ if (nr) {
+ cmnd->result = nr;
+ csio_inc_stats(scsim, n_rn_nr_error);
+ goto err_done;
+ }
+
+ if (unlikely(!csio_is_hw_ready(hw))) {
+ cmnd->result = (DID_REQUEUE << 16);
+ csio_inc_stats(scsim, n_hw_nr_error);
+ goto err_done;
+ }
+
+ /* Get req->nsge, if there are SG elements to be mapped */
+ nsge = scsi_dma_map(cmnd);
+ if (unlikely(nsge < 0)) {
+ csio_inc_stats(scsim, n_dmamap_error);
+ goto err;
+ }
+
+ /* Do we support so many mappings? */
+ if (unlikely(nsge > scsim->max_sge)) {
+ csio_warn(hw,
+ "More SGEs than can be supported."
+ " SGEs: %d, Max SGEs: %d\n", nsge, scsim->max_sge);
+ csio_inc_stats(scsim, n_unsupp_sge_error);
+ goto err_dma_unmap;
+ }
+
+ /* Get a free ioreq structure - SM is already set to uninit */
+ ioreq = csio_get_scsi_ioreq_lock(hw, scsim);
+ if (!ioreq) {
+ csio_err(hw, "Out of I/O request elements. Active #:%d\n",
+ scsim->stats.n_active);
+ csio_inc_stats(scsim, n_no_req_error);
+ goto err_dma_unmap;
+ }
+
+ ioreq->nsge = nsge;
+ ioreq->lnode = ln;
+ ioreq->rnode = rn;
+ ioreq->iq_idx = sqset->iq_idx;
+ ioreq->eq_idx = sqset->eq_idx;
+ ioreq->wr_status = 0;
+ ioreq->drv_status = CSIO_SUCCESS;
+ csio_scsi_cmnd(ioreq) = (void *)cmnd;
+ ioreq->tmo = 0;
+
+ switch (cmnd->sc_data_direction) {
+ case DMA_BIDIRECTIONAL:
+ ioreq->datadir = CSIO_IOREQF_DMA_BIDI;
+ csio_inc_stats(ln, n_control_requests);
+ break;
+ case DMA_TO_DEVICE:
+ ioreq->datadir = CSIO_IOREQF_DMA_WRITE;
+ csio_inc_stats(ln, n_output_requests);
+ ln->stats.n_output_bytes += scsi_bufflen(cmnd);
+ break;
+ case DMA_FROM_DEVICE:
+ ioreq->datadir = CSIO_IOREQF_DMA_READ;
+ csio_inc_stats(ln, n_input_requests);
+ ln->stats.n_input_bytes += scsi_bufflen(cmnd);
+ break;
+ case DMA_NONE:
+ ioreq->datadir = CSIO_IOREQF_DMA_NONE;
+ csio_inc_stats(ln, n_control_requests);
+ break;
+ default:
+ CSIO_DB_ASSERT(0);
+ break;
+ }
+
+ /* Set cbfn */
+ ioreq->io_cbfn = csio_scsi_cbfn;
+
+ /* Needed during abort */
+ cmnd->host_scribble = (unsigned char *)ioreq;
+ cmnd->scsi_done = done;
+ cmnd->SCp.Message = 0;
+
+ /* Kick off SCSI IO SM on the ioreq */
+ spin_lock_irqsave(&hw->lock, flags);
+ retval = csio_scsi_start_io(ioreq);
+ spin_unlock_irqrestore(&hw->lock, flags);
+
+ if (retval != CSIO_SUCCESS) {
+ csio_err(hw, "ioreq: %p couldnt be started, status:%d\n",
+ ioreq, retval);
+ csio_inc_stats(scsim, n_busy_error);
+ goto err_put_req;
+ }
+
+ return 0;
+
+err_put_req:
+ csio_put_scsi_ioreq_lock(hw, scsim, ioreq);
+err_dma_unmap:
+ if (nsge > 0)
+ scsi_dma_unmap(cmnd);
+err:
+ return rv;
+
+err_done:
+ done(cmnd);
+ return 0;
+}
+
+static DEF_SCSI_QCMD(csio_queuecommand);
+
+static csio_retval_t
+csio_do_abrt_cls(struct csio_hw *hw, struct csio_ioreq *ioreq, bool abort)
+{
+ csio_retval_t rv;
+ int cpu = smp_processor_id();
+ struct csio_lnode *ln = ioreq->lnode;
+ struct csio_scsi_qset *sqset = &hw->sqset[ln->portid][cpu];
+
+ ioreq->tmo = CSIO_SCSI_ABRT_TMO_MS;
+ /*
+ * Use current processor queue for posting the abort/close, but retain
+ * the ingress queue ID of the original I/O being aborted/closed - we
+ * need the abort/close completion to be received on the same queue
+ * as the original I/O.
+ */
+ ioreq->eq_idx = sqset->eq_idx;
+
+ if (abort == SCSI_ABORT)
+ rv = csio_scsi_abort(ioreq);
+ else /* close */
+ rv = csio_scsi_close(ioreq);
+
+ return rv;
+}
+
+static int
+csio_eh_abort_handler(struct scsi_cmnd *cmnd)
+{
+ struct csio_ioreq *ioreq;
+ struct csio_lnode *ln = shost_priv(cmnd->device->host);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ struct csio_scsim *scsim = csio_hw_to_scsim(hw);
+ int ready = 0, ret;
+ unsigned long tmo = 0;
+ csio_retval_t rv;
+ struct csio_rnode *rn = (struct csio_rnode *)(cmnd->device->hostdata);
+
+ ret = fc_block_scsi_eh(cmnd);
+ if (ret)
+ return ret;
+
+ ioreq = (struct csio_ioreq *)cmnd->host_scribble;
+ if (!ioreq)
+ return SUCCESS;
+
+ if (!rn)
+ return FAILED;
+
+ csio_dbg(hw,
+ "Request to abort ioreq:%p cmd:%p cdb:%08llx"
+ " ssni:0x%x lun:%d iq:0x%x\n",
+ ioreq, cmnd, *((uint64_t *)cmnd->cmnd), rn->flowid,
+ cmnd->device->lun, csio_q_physiqid(hw, ioreq->iq_idx));
+
+ if (((struct scsi_cmnd *)csio_scsi_cmnd(ioreq)) != cmnd) {
+ csio_inc_stats(scsim, n_abrt_race_comp);
+ return SUCCESS;
+ }
+
+ ready = csio_is_lnode_ready(ln);
+ tmo = CSIO_SCSI_ABRT_TMO_MS;
+
+ spin_lock_irq(&hw->lock);
+ rv = csio_do_abrt_cls(hw, ioreq, (ready ? SCSI_ABORT : SCSI_CLOSE));
+ spin_unlock_irq(&hw->lock);
+
+ if (rv != CSIO_SUCCESS) {
+ if (rv == CSIO_INVAL) {
+ /* Return success, if abort/close request issued on
+ * already completed IO
+ */
+ return SUCCESS;
+ }
+ if (ready)
+ csio_inc_stats(scsim, n_abrt_busy_error);
+ else
+ csio_inc_stats(scsim, n_cls_busy_error);
+
+ goto inval_scmnd;
+ }
+
+ /* Wait for completion */
+ init_completion(&ioreq->cmplobj);
+ wait_for_completion_timeout(&ioreq->cmplobj, msecs_to_jiffies(tmo));
+
+ /* FW didnt respond to abort within our timeout */
+ if (((struct scsi_cmnd *)csio_scsi_cmnd(ioreq)) == cmnd) {
+
+ csio_err(hw, "Abort timed out -- req: %p\n", ioreq);
+ csio_inc_stats(scsim, n_abrt_timedout);
+
+inval_scmnd:
+ if (ioreq->nsge > 0)
+ scsi_dma_unmap(cmnd);
+
+ spin_lock_irq(&hw->lock);
+ csio_scsi_cmnd(ioreq) = NULL;
+ spin_unlock_irq(&hw->lock);
+
+ cmnd->result = (DID_ERROR << 16);
+ cmnd->scsi_done(cmnd);
+
+ return FAILED;
+ }
+
+ /* FW successfully aborted the request */
+ if (host_byte(cmnd->result) == DID_REQUEUE) {
+ csio_info(hw,
+ "Aborted SCSI command to (%d:%d) serial#:0x%lx\n",
+ cmnd->device->id, cmnd->device->lun,
+ cmnd->serial_number);
+ return SUCCESS;
+ } else {
+ csio_info(hw,
+ "Failed to abort SCSI command, (%d:%d) serial#:0x%lx\n",
+ cmnd->device->id, cmnd->device->lun,
+ cmnd->serial_number);
+ return FAILED;
+ }
+}
+
+/*
+ * csio_tm_cbfn - TM callback function.
+ * @hw: HW module.
+ * @req: IO request.
+ *
+ * Cache the result in 'cmnd', since ioreq will be freed soon
+ * after we return from here, and the waiting thread shouldnt trust
+ * the ioreq contents.
+ */
+static void
+csio_tm_cbfn(struct csio_hw *hw, struct csio_ioreq *req)
+{
+ struct scsi_cmnd *cmnd = (struct scsi_cmnd *)csio_scsi_cmnd(req);
+ struct csio_dma_buf *dma_buf;
+ uint8_t flags = 0;
+ struct csio_fcp_resp *fcp_resp;
+
+ csio_dbg(hw, "req: %p in csio_tm_cbfn status: %d\n",
+ req, req->wr_status);
+
+ /* Cache FW return status */
+ cmnd->SCp.Status = req->wr_status;
+
+ /* Special handling based on FCP response */
+
+ /*
+ * FW returns us this error, if flags were set. FCP4 says
+ * FCP_RSP_LEN_VAL in flags shall be set for TM completions.
+ * So if a target were to set this bit, we expect that the
+ * rsp_code is set to FCP_TMF_CMPL for a successful TM
+ * completion. Any other rsp_code means TM operation failed.
+ * If a target were to just ignore setting flags, we treat
+ * the TM operation as success, and FW returns FW_SUCCESS.
+ */
+ if (req->wr_status == FW_SCSI_RSP_ERR) {
+ dma_buf = &req->dma_buf;
+ fcp_resp = (struct csio_fcp_resp *)dma_buf->vaddr;
+ flags = fcp_resp->flags;
+
+ /* Modify return status if flags indicate success */
+ if (flags & FCP_RSP_LEN_VAL)
+ if (fcp_resp->rsp_code == FCP_TMF_CMPL)
+ cmnd->SCp.Status = FW_SUCCESS;
+
+ csio_dbg(hw, "TM FCP rsp code: %d\n", fcp_resp->rsp_code);
+ }
+
+ /* Wake up the TM handler thread */
+ csio_scsi_cmnd(req) = NULL;
+}
+
+static int
+csio_eh_lun_reset_handler(struct scsi_cmnd *cmnd)
+{
+ struct csio_lnode *ln = shost_priv(cmnd->device->host);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ struct csio_scsim *scsim = csio_hw_to_scsim(hw);
+ struct csio_rnode *rn = (struct csio_rnode *)(cmnd->device->hostdata);
+ struct csio_ioreq *ioreq = NULL;
+ struct csio_scsi_qset *sqset;
+ unsigned long flags;
+ csio_retval_t retval;
+ int count, ret;
+ LIST_HEAD(local_q);
+ struct csio_scsi_level_data sld;
+
+ if (!rn)
+ goto fail;
+
+ csio_dbg(hw, "Request to reset LUN:%d (ssni:0x%x tgtid:%d)\n",
+ cmnd->device->lun, rn->flowid, rn->scsi_id);
+
+ if (!csio_is_lnode_ready(ln)) {
+ csio_err(hw,
+ "LUN reset cannot be issued on non-ready"
+ " local node vnpi:0x%x (LUN:%d)\n",
+ ln->vnp_flowid, cmnd->device->lun);
+ goto fail;
+ }
+
+ /* Lnode is ready, now wait on rport node readiness */
+ ret = fc_block_scsi_eh(cmnd);
+ if (ret)
+ return ret;
+
+ /*
+ * If we have blocked in the previous call, at this point, either the
+ * remote node has come back online, or device loss timer has fired
+ * and the remote node is destroyed. Allow the LUN reset only for
+ * the former case, since LUN reset is a TMF I/O on the wire, and we
+ * need a valid session to issue it.
+ */
+ if (fc_remote_port_chkready(rn->rport)) {
+ csio_err(hw,
+ "LUN reset cannot be issued on non-ready"
+ " remote node ssni:0x%x (LUN:%d)\n",
+ rn->flowid, cmnd->device->lun);
+ goto fail;
+ }
+
+ /* Get a free ioreq structure - SM is already set to uninit */
+ ioreq = csio_get_scsi_ioreq_lock(hw, scsim);
+
+ if (!ioreq) {
+ csio_err(hw, "Out of IO request elements. Active # :%d\n",
+ scsim->stats.n_active);
+ goto fail;
+ }
+
+ sqset = &hw->sqset[ln->portid][smp_processor_id()];
+ ioreq->nsge = 0;
+ ioreq->lnode = ln;
+ ioreq->rnode = rn;
+ ioreq->iq_idx = sqset->iq_idx;
+ ioreq->eq_idx = sqset->eq_idx;
+
+ csio_scsi_cmnd(ioreq) = cmnd;
+ cmnd->host_scribble = (unsigned char *)ioreq;
+ cmnd->SCp.Status = 0;
+
+ cmnd->SCp.Message = FCP_TMF_LUN_RESET;
+ ioreq->tmo = CSIO_SCSI_LUNRST_TMO_MS / 1000;
+
+ /*
+ * FW times the LUN reset for ioreq->tmo, so we got to wait a little
+ * longer (10s for now) than that to allow FW to return the timed
+ * out command.
+ */
+ count = CSIO_ROUNDUP((ioreq->tmo + 10) * 1000, CSIO_SCSI_TM_POLL_MS);
+
+ /* Set cbfn */
+ ioreq->io_cbfn = csio_tm_cbfn;
+
+ /* Save of the ioreq info for later use */
+ sld.level = CSIO_LEV_LUN;
+ sld.lnode = ioreq->lnode;
+ sld.rnode = ioreq->rnode;
+ sld.oslun = (uint64_t)cmnd->device->lun;
+
+ spin_lock_irqsave(&hw->lock, flags);
+ /* Kick off TM SM on the ioreq */
+ retval = csio_scsi_start_tm(ioreq);
+ spin_unlock_irqrestore(&hw->lock, flags);
+
+ if (retval != CSIO_SUCCESS) {
+ csio_err(hw, "Failed to issue LUN reset, req:%p, status:%d\n",
+ ioreq, retval);
+ goto fail_ret_ioreq;
+ }
+
+ csio_dbg(hw, "Waiting max %d secs for LUN reset completion\n",
+ count * (CSIO_SCSI_TM_POLL_MS / 1000));
+ /* Wait for completion */
+ while ((((struct scsi_cmnd *)csio_scsi_cmnd(ioreq)) == cmnd)
+ && count--)
+ msleep(CSIO_SCSI_TM_POLL_MS);
+
+ /* LUN reset timed-out */
+ if (((struct scsi_cmnd *)csio_scsi_cmnd(ioreq)) == cmnd) {
+ csio_err(hw, "LUN reset (%d:%d) timed out\n",
+ cmnd->device->id, cmnd->device->lun);
+
+ spin_lock_irq(&hw->lock);
+ csio_scsi_drvcleanup(ioreq);
+ list_del_init(&ioreq->sm.sm_list);
+ spin_unlock_irq(&hw->lock);
+
+ goto fail_ret_ioreq;
+ }
+
+ /* LUN reset returned, check cached status */
+ if (cmnd->SCp.Status != FW_SUCCESS) {
+ csio_err(hw, "LUN reset failed (%d:%d), status: %d\n",
+ cmnd->device->id, cmnd->device->lun, cmnd->SCp.Status);
+ goto fail;
+ }
+
+ /* LUN reset succeeded, Start aborting affected I/Os */
+ /*
+ * Since the host guarantees during LUN reset that there
+ * will not be any more I/Os to that LUN, until the LUN reset
+ * completes, we gather pending I/Os after the LUN reset.
+ */
+ spin_lock_irq(&hw->lock);
+ csio_scsi_gather_active_ios(scsim, &sld, &local_q);
+
+ retval = csio_scsi_abort_io_q(scsim, &local_q, 30000);
+ spin_unlock_irq(&hw->lock);
+
+ /* Aborts may have timed out */
+ if (retval != CSIO_SUCCESS) {
+ csio_err(hw,
+ "Attempt to abort I/Os during LUN reset of %d"
+ " returned %d\n", cmnd->device->lun, retval);
+ /* Return I/Os back to active_q */
+ spin_lock_irq(&hw->lock);
+ list_splice_tail_init(&local_q, &scsim->active_q);
+ spin_unlock_irq(&hw->lock);
+ goto fail;
+ }
+
+ csio_inc_stats(rn, n_lun_rst);
+
+ csio_info(hw, "LUN reset occurred (%d:%d)\n",
+ cmnd->device->id, cmnd->device->lun);
+
+ return SUCCESS;
+
+fail_ret_ioreq:
+ csio_put_scsi_ioreq_lock(hw, scsim, ioreq);
+fail:
+ csio_inc_stats(rn, n_lun_rst_fail);
+ return FAILED;
+}
+
+static int
+csio_slave_alloc(struct scsi_device *sdev)
+{
+ struct fc_rport *rport = starget_to_rport(scsi_target(sdev));
+
+ if (!rport || fc_remote_port_chkready(rport))
+ return -ENXIO;
+
+ sdev->hostdata = *((struct csio_lnode **)(rport->dd_data));
+
+ return 0;
+}
+
+static int
+csio_slave_configure(struct scsi_device *sdev)
+{
+ if (sdev->tagged_supported)
+ scsi_activate_tcq(sdev, csio_lun_qdepth);
+ else
+ scsi_deactivate_tcq(sdev, csio_lun_qdepth);
+
+ return 0;
+}
+
+static void
+csio_slave_destroy(struct scsi_device *sdev)
+{
+ sdev->hostdata = NULL;
+}
+
+static int
+csio_scan_finished(struct Scsi_Host *shost, unsigned long time)
+{
+ struct csio_lnode *ln = shost_priv(shost);
+ int rv = 0;
+
+ spin_lock_irq(shost->host_lock);
+ if (!ln->hwp || csio_list_deleted(&ln->sm.sm_list)) {
+ spin_unlock_irq(shost->host_lock);
+ return 1;
+ }
+
+ rv = csio_scan_done(ln, jiffies, time, csio_max_scan_tmo * HZ,
+ csio_delta_scan_tmo * HZ);
+
+ spin_unlock_irq(shost->host_lock);
+
+ return rv;
+}
+
+struct scsi_host_template csio_fcoe_shost_template = {
+ .module = THIS_MODULE,
+ .name = CSIO_DRV_DESC,
+ .proc_name = KBUILD_MODNAME,
+ .queuecommand = csio_queuecommand,
+ .eh_abort_handler = csio_eh_abort_handler,
+ .eh_device_reset_handler = csio_eh_lun_reset_handler,
+ .slave_alloc = csio_slave_alloc,
+ .slave_configure = csio_slave_configure,
+ .slave_destroy = csio_slave_destroy,
+ .scan_finished = csio_scan_finished,
+ .this_id = -1,
+ .sg_tablesize = CSIO_SCSI_MAX_SGE,
+ .cmd_per_lun = CSIO_MAX_CMD_PER_LUN,
+ .use_clustering = ENABLE_CLUSTERING,
+ .shost_attrs = csio_fcoe_lport_attrs,
+ .max_sectors = CSIO_MAX_SECTOR_SIZE,
+};
+
+struct scsi_host_template csio_fcoe_shost_vport_template = {
+ .module = THIS_MODULE,
+ .name = CSIO_DRV_DESC,
+ .proc_name = KBUILD_MODNAME,
+ .queuecommand = csio_queuecommand,
+ .eh_abort_handler = csio_eh_abort_handler,
+ .eh_device_reset_handler = csio_eh_lun_reset_handler,
+ .slave_alloc = csio_slave_alloc,
+ .slave_configure = csio_slave_configure,
+ .slave_destroy = csio_slave_destroy,
+ .scan_finished = csio_scan_finished,
+ .this_id = -1,
+ .sg_tablesize = CSIO_SCSI_MAX_SGE,
+ .cmd_per_lun = CSIO_MAX_CMD_PER_LUN,
+ .use_clustering = ENABLE_CLUSTERING,
+ .shost_attrs = csio_fcoe_vport_attrs,
+ .max_sectors = CSIO_MAX_SECTOR_SIZE,
+};
+
+/*
+ * csio_scsi_alloc_ddp_bufs - Allocate buffers for DDP of unaligned SGLs.
+ * @scm: SCSI Module
+ * @hw: HW device.
+ * @buf_size: buffer size
+ * @num_buf : Number of buffers.
+ *
+ * This routine allocates DMA buffers required for SCSI Data xfer, if
+ * each SGL buffer for a SCSI Read request posted by SCSI midlayer are
+ * not virtually contiguous.
+ */
+static csio_retval_t
+csio_scsi_alloc_ddp_bufs(struct csio_scsim *scm, struct csio_hw *hw,
+ int buf_size, int num_buf)
+{
+ int n = 0;
+ struct list_head *tmp;
+ struct csio_dma_buf *ddp_desc = NULL;
+ uint32_t unit_size = 0;
+
+ if (!num_buf)
+ return CSIO_SUCCESS;
+
+ if (!buf_size)
+ return CSIO_INVAL;
+
+ INIT_LIST_HEAD(&scm->ddp_freelist);
+
+ /* Align buf size to page size */
+ buf_size = (buf_size + PAGE_SIZE - 1) & PAGE_MASK;
+ /* Initialize dma descriptors */
+ for (n = 0; n < num_buf; n++) {
+ /* Set unit size to request size */
+ unit_size = buf_size;
+ ddp_desc = kzalloc(sizeof(struct csio_dma_buf), GFP_KERNEL);
+ if (!ddp_desc) {
+ csio_err(hw,
+ "Failed to allocate ddp descriptors,"
+ " Num allocated = %d.\n",
+ scm->stats.n_free_ddp);
+ goto no_mem;
+ }
+
+ /* Allocate Dma buffers for DDP */
+ ddp_desc->vaddr = pci_alloc_consistent(hw->pdev, unit_size,
+ &ddp_desc->paddr);
+ if (!ddp_desc->vaddr) {
+ csio_err(hw,
+ "SCSI response DMA buffer (ddp) allocation"
+ " failed!\n");
+ kfree(ddp_desc);
+ goto no_mem;
+ }
+
+ ddp_desc->len = unit_size;
+
+ /* Added it to scsi ddp freelist */
+ list_add_tail(&ddp_desc->list, &scm->ddp_freelist);
+ csio_inc_stats(scm, n_free_ddp);
+ }
+
+ return CSIO_SUCCESS;
+no_mem:
+ /* release dma descs back to freelist and free dma memory */
+ list_for_each(tmp, &scm->ddp_freelist) {
+ ddp_desc = (struct csio_dma_buf *) tmp;
+ tmp = csio_list_prev(tmp);
+ pci_free_consistent(hw->pdev, ddp_desc->len, ddp_desc->vaddr,
+ ddp_desc->paddr);
+ list_del_init(&ddp_desc->list);
+ kfree(ddp_desc);
+ }
+ scm->stats.n_free_ddp = 0;
+
+ return CSIO_NOMEM;
+}
+
+/*
+ * csio_scsi_free_ddp_bufs - free DDP buffers of unaligned SGLs.
+ * @scm: SCSI Module
+ * @hw: HW device.
+ *
+ * This routine frees ddp buffers.
+ */
+static csio_retval_t
+csio_scsi_free_ddp_bufs(struct csio_scsim *scm, struct csio_hw *hw)
+{
+ struct list_head *tmp;
+ struct csio_dma_buf *ddp_desc;
+
+ /* release dma descs back to freelist and free dma memory */
+ list_for_each(tmp, &scm->ddp_freelist) {
+ ddp_desc = (struct csio_dma_buf *) tmp;
+ tmp = csio_list_prev(tmp);
+ pci_free_consistent(hw->pdev, ddp_desc->len, ddp_desc->vaddr,
+ ddp_desc->paddr);
+ list_del_init(&ddp_desc->list);
+ kfree(ddp_desc);
+ }
+ scm->stats.n_free_ddp = 0;
+
+ return CSIO_NOMEM;
+}
+
+/**
+ * csio_scsim_init - Initialize SCSI Module
+ * @scm: SCSI Module
+ * @hw: HW module
+ *
+ */
+csio_retval_t
+csio_scsim_init(struct csio_scsim *scm, struct csio_hw *hw)
+{
+ int i;
+ struct csio_ioreq *ioreq;
+ struct csio_dma_buf *dma_buf;
+
+ INIT_LIST_HEAD(&scm->active_q);
+ scm->hw = hw;
+
+ scm->proto_cmd_len = sizeof(struct csio_fcp_cmnd);
+ scm->proto_rsp_len = sizeof(struct csio_fcp_resp);
+ scm->max_sge = CSIO_SCSI_MAX_SGE;
+
+ spin_lock_init(&scm->freelist_lock);
+
+ /* Pre-allocate ioreqs and initialize them */
+ INIT_LIST_HEAD(&scm->ioreq_freelist);
+ for (i = 0; i < csio_scsi_ioreqs; i++) {
+
+ ioreq = kzalloc(sizeof(struct csio_ioreq), GFP_KERNEL);
+ if (!ioreq) {
+ csio_err(hw,
+ "I/O request element allocation failed, "
+ " Num allocated = %d.\n",
+ scm->stats.n_free_ioreq);
+
+ goto free_ioreq;
+ }
+
+ /* Allocate Dma buffers for Response Payload */
+ dma_buf = &ioreq->dma_buf;
+ dma_buf->vaddr = pci_pool_alloc(hw->scsi_pci_pool, GFP_KERNEL,
+ &dma_buf->paddr);
+ if (!dma_buf->vaddr) {
+ csio_err(hw,
+ "SCSI response DMA buffer allocation"
+ " failed!\n");
+ kfree(ioreq);
+ goto free_ioreq;
+ }
+
+ dma_buf->len = scm->proto_rsp_len;
+
+ /* Set state to uninit */
+ csio_init_state(&ioreq->sm, csio_scsis_uninit);
+ INIT_LIST_HEAD(&ioreq->gen_list);
+ init_completion(&ioreq->cmplobj);
+
+ list_add_tail(&ioreq->sm.sm_list, &scm->ioreq_freelist);
+ csio_inc_stats(scm, n_free_ioreq);
+ }
+
+ if (csio_scsi_alloc_ddp_bufs(scm, hw, PAGE_SIZE, csio_ddp_descs))
+ goto free_ioreq;
+
+ return CSIO_SUCCESS;
+
+free_ioreq:
+ /*
+ * Free up existing allocations, since an error
+ * from here means we are returning for good
+ */
+ while (!list_empty(&scm->ioreq_freelist)) {
+ csio_deq_from_head(&scm->ioreq_freelist, &ioreq);
+
+ dma_buf = &ioreq->dma_buf;
+ pci_pool_free(hw->scsi_pci_pool, dma_buf->vaddr,
+ dma_buf->paddr);
+
+ kfree(ioreq);
+ }
+
+ scm->stats.n_free_ioreq = 0;
+
+ return CSIO_NOMEM;
+}
+
+/**
+ * csio_scsim_exit: Uninitialize SCSI Module
+ * @scm: SCSI Module
+ *
+ */
+void
+csio_scsim_exit(struct csio_scsim *scm)
+{
+ struct csio_ioreq *ioreq;
+ struct csio_dma_buf *dma_buf;
+
+ while (!list_empty(&scm->ioreq_freelist)) {
+ csio_deq_from_head(&scm->ioreq_freelist, &ioreq);
+
+ dma_buf = &ioreq->dma_buf;
+ pci_pool_free(scm->hw->scsi_pci_pool, dma_buf->vaddr,
+ dma_buf->paddr);
+
+ kfree(ioreq);
+ }
+
+ scm->stats.n_free_ioreq = 0;
+
+ csio_scsi_free_ddp_bufs(scm, scm->hw);
+}
--
1.7.1
--
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
