Hi Narsimhulu,
On 03/11/2015 06:01 PM, Narsimhulu Musini wrote:
> snic_scsi.c contains scsi handling, includes queuing io, abort, lun reset,
> and host reset. Also it handles asynchronous event notifications from FW.
>
> v2
> Changed queuecommand to lock-free version.
> Converted custom error codes to standard error codes.
>
> Signed-off-by: Narsimhulu Musini <nmus...@cisco.com>
> Signed-off-by: Sesidhar Baddela <sebad...@cisco.com>
> ---
> drivers/scsi/snic/snic_scsi.c | 2631
> +++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 2631 insertions(+)
> create mode 100644 drivers/scsi/snic/snic_scsi.c
>
> diff --git a/drivers/scsi/snic/snic_scsi.c b/drivers/scsi/snic/snic_scsi.c
> new file mode 100644
> index 0000000..f82ff02
> --- /dev/null
> +++ b/drivers/scsi/snic/snic_scsi.c
> @@ -0,0 +1,2631 @@
> +/*
> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
> + *
> + * This program is free software; you may redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; version 2 of the License.
> + *
> + * 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/mempool.h>
> +#include <linux/errno.h>
> +#include <linux/init.h>
> +#include <linux/workqueue.h>
> +#include <linux/pci.h>
> +#include <linux/spinlock.h>
> +#include <linux/delay.h>
> +#include <linux/gfp.h>
> +#include <scsi/scsi.h>
> +#include <scsi/scsi_host.h>
> +#include <scsi/scsi_device.h>
> +#include <scsi/scsi_cmnd.h>
> +#include <scsi/scsi_tcq.h>
> +#include <scsi/scsi_dbg.h>
> +
> +#include "snic_io.h"
> +#include "snic.h"
> +#include "snic_os.h"
> +
> +const char *snic_state_str[] = {
> + [SNIC_INIT] = "SNIC_INIT",
> + [SNIC_ERROR] = "SNIC_ERROR",
> + [SNIC_ONLINE] = "SNIC_ONLINE",
> + [SNIC_OFFLINE] = "SNIC_OFFLINE",
> + [SNIC_FWRESET] = "SNIC_FWRESET",
> +};
> +
> +static const char * const snic_req_state_str[] = {
> + [SNIC_IOREQ_NOT_INITED] = "SNIC_IOREQ_NOT_INITED",
> + [SNIC_IOREQ_PENDING] = "SNIC_IOREQ_PENDING",
> + [SNIC_IOREQ_ABTS_PENDING] = "SNIC_IOREQ_ABTS_PENDING",
> + [SNIC_IOREQ_ABTS_COMPLETE] = "SNIC_IOREQ_ABTS_COMPELTE",
> + [SNIC_IOREQ_LR_PENDING] = "SNIC_IOREQ_LR_PENDING",
> + [SNIC_IOREQ_LR_COMPLETE] = "SNIC_IOREQ_LR_COMPELTE",
> + [SNIC_IOREQ_COMPLETE] = "SNIC_IOREQ_CMD_COMPELTE",
> +};
> +
> +/* snic cmd status strings */
> +static const char * const snic_io_status_str[] = {
> + [SNIC_STAT_IO_SUCCESS] = "SNIC_STAT_IO_SUCCESS", /* 0x0 */
> + [SNIC_STAT_INVALID_HDR] = "SNIC_STAT_INVALID_HDR",
> + [SNIC_STAT_OUT_OF_RES] = "SNIC_STAT_OUT_OF_RES",
> + [SNIC_STAT_INVALID_PARM] = "SNIC_STAT_INVALID_PARM",
> + [SNIC_STAT_REQ_NOT_SUP] = "SNIC_STAT_REQ_NOT_SUP",
> + [SNIC_STAT_IO_NOT_FOUND] = "SNIC_STAT_IO_NOT_FOUND",
> + [SNIC_STAT_ABORTED] = "SNIC_STAT_ABORTED",
> + [SNIC_STAT_TIMEOUT] = "SNIC_STAT_TIMEOUT",
> + [SNIC_STAT_SGL_INVALID] = "SNIC_STAT_SGL_INVALID",
> + [SNIC_STAT_DATA_CNT_MISMATCH] = "SNIC_STAT_DATA_CNT_MISMATCH",
> + [SNIC_STAT_FW_ERR] = "SNIC_STAT_FW_ERR",
> + [SNIC_STAT_ITMF_REJECT] = "SNIC_STAT_ITMF_REJECT",
> + [SNIC_STAT_ITMF_FAIL] = "SNIC_STAT_ITMF_FAIL",
> + [SNIC_STAT_ITMF_INCORRECT_LUN] = "SNIC_STAT_ITMF_INCORRECT_LUN",
> + [SNIC_STAT_CMND_REJECT] = "SNIC_STAT_CMND_REJECT",
> + [SNIC_STAT_DEV_OFFLINE] = "SNIC_STAT_DEV_OFFLINE",
> + [SNIC_STAT_NO_BOOTLUN] = "SNIC_STAT_NO_BOOTLUN",
> + [SNIC_STAT_SCSI_ERR] = "SNIC_STAT_SCSI_ERR",
> + [SNIC_STAT_NOT_READY] = "SNIC_STAT_NOT_READY",
> + [SNIC_STAT_FATAL_ERROR] = "SNIC_STAT_FATAL_ERROR",
> +};
> +
> +static void snic_scsi_cleanup(struct snic *, int);
> +
> +const char *
> +snic_state_to_str(unsigned int state)
> +{
> + if (state >= ARRAY_SIZE(snic_state_str) || !snic_state_str[state])
> + return "Unknown";
> +
> + return snic_state_str[state];
> +}
> +
> +static const char *
> +snic_io_status_to_str(unsigned int state)
> +{
> + if ((state >= ARRAY_SIZE(snic_io_status_str)) ||
> + (!snic_io_status_str[state]))
> + return "Unknown";
> +
> + return snic_io_status_str[state];
> +}
> +
> +static const char *
> +snic_ioreq_state_to_str(unsigned int state)
> +{
> + if (state >= ARRAY_SIZE(snic_req_state_str) ||
> + !snic_req_state_str[state])
> + return "Unknown";
> +
> + return snic_req_state_str[state];
> +}
> +
> +static inline spinlock_t *
> +snic_io_lock_hash(struct snic *snic, struct scsi_cmnd *sc)
> +{
> + u32 hash = snic_cmd_tag(sc) & (SNIC_IO_LOCKS - 1);
> +
> + return &snic->io_req_lock[hash];
> +}
> +
> +static inline spinlock_t *
> +snic_io_lock_tag(struct snic *snic, int tag)
> +{
> + return &snic->io_req_lock[tag & (SNIC_IO_LOCKS - 1)];
> +}
> +
> +/* snic_release_req_buf : Releases snic_req_info */
> +static void
> +snic_release_req_buf(struct snic *snic,
> + struct snic_req_info *rqi,
> + struct scsi_cmnd *sc)
> +{
> + struct snic_host_req *req = rqi_to_req(rqi);
> +
> + /* Freeing cmd without marking completion, not okay */
> + SNIC_BUG_ON(!((CMD_STATE(sc) == SNIC_IOREQ_COMPLETE) ||
> + (CMD_STATE(sc) == SNIC_IOREQ_ABTS_COMPLETE) ||
> + (CMD_FLAGS(sc) & SNIC_DEV_RST_NOTSUP) ||
> + (CMD_FLAGS(sc) & SNIC_IO_INTERNAL_TERM_ISSUED) ||
> + (CMD_FLAGS(sc) & SNIC_DEV_RST_TERM_ISSUED) ||
> + (CMD_FLAGS(sc) & SNIC_SCSI_CLEANUP) ||
> + (CMD_STATE(sc) == SNIC_IOREQ_LR_COMPLETE)));
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "Rel_req:sc %p:tag %x:rqi %p:ioreq %p:abt %p:dr %p: state
> %s:flags 0x%x\n",
> + sc, snic_cmd_tag(sc), rqi, rqi->req, rqi->abort_req,
> + rqi->dr_req, snic_ioreq_state_to_str(CMD_STATE(sc)),
> + CMD_FLAGS(sc));
> +
> + if (req->u.icmnd.sense_addr)
> + pci_unmap_single(snic->pdev,
> + req->u.icmnd.sense_addr,
> + SCSI_SENSE_BUFFERSIZE,
> + PCI_DMA_FROMDEVICE);
> +
> + scsi_dma_unmap(sc);
> +
> + snic_req_free(snic, rqi);
> +} /* end of snic_release_req_buf */
> +
> +/*
> + * snic_queue_icmnd_req : Queues snic_icmnd request
> + */
> +static int
> +snic_queue_icmnd_req(struct snic *snic,
> + struct snic_req_info *rqi,
> + struct scsi_cmnd *sc,
> + int sg_cnt)
> +{
> + struct scatterlist *sg;
> + struct snic_sg_desc *sgd;
> + dma_addr_t pa = 0;
> + struct scsi_lun lun;
> + int flags = 0;
> + int ret = 0;
> + unsigned int i;
> +
> + if (sg_cnt) {
> + flags = SNIC_ICMND_ESGL;
> + sgd = (struct snic_sg_desc *) req_to_sgl(rqi->req);
> +
> + for_each_sg(scsi_sglist(sc), sg, sg_cnt, i) {
> + sgd->addr = cpu_to_le64(sg_dma_address(sg));
> + sgd->len = cpu_to_le32(sg_dma_len(sg));
> + sgd->_resvd = 0;
> + sgd++;
> + }
> + }
> +
> + pa = pci_map_single(snic->pdev,
> + sc->sense_buffer,
> + SCSI_SENSE_BUFFERSIZE,
> + PCI_DMA_FROMDEVICE);
> +
> + if (pci_dma_mapping_error(snic->pdev, pa)) {
> + SNIC_HOST_ERR(snic->shost,
> + "QIcmnd:PCI Map Failed for sns buf %p tag %x\n",
> + sc->sense_buffer, snic_cmd_tag(sc));
> + ret = -ENOMEM;
> +
> + return ret;
> + }
> +
> + int_to_scsilun(sc->device->lun, &lun);
> + if (sc->sc_data_direction == DMA_FROM_DEVICE)
> + flags |= SNIC_ICMND_RD;
> + if (sc->sc_data_direction == DMA_TO_DEVICE)
> + flags |= SNIC_ICMND_WR;
> +
> + /* Initialize icmnd */
> + snic_icmnd_init(rqi->req,
> + snic_cmd_tag(sc),
> + snic->config.hid, /* hid */
> + (u64)rqi,
> + flags, /* command flags */
> + rqi->tgt_id,
> + lun.scsi_lun,
> + sc->cmnd,
> + sc->cmd_len,
> + scsi_bufflen(sc),
> + sg_cnt,
> + (u64)req_to_sgl(rqi->req),
> + pa, /* sense buffer pa */
> + SCSI_SENSE_BUFFERSIZE);
> +
> + ret = snic_queue_wq_desc(snic, rqi->req, rqi->req_len);
> + if (ret)
> + SNIC_HOST_ERR(snic->shost,
> + "QIcmnd: Queuing Icmnd Failed. ret = %d\n",
> + ret);
> +
> + return ret;
> +} /* end of snic_queue_icmnd_req */
> +
> +/*
> + * snic_issue_scsi_req : Prepares IO request and Issues to FW.
> + */
> +static int
> +snic_issue_scsi_req(struct snic *snic,
> + struct snic_tgt *tgt,
> + struct scsi_cmnd *sc)
> +{
> + struct snic_req_info *rqi = NULL;
> + int sg_cnt = 0;
> + int ret = 0;
> + u32 tag = snic_cmd_tag(sc);
> + u64 cmd_trc = 0, cmd_st_flags = 0;
> + spinlock_t *io_lock = NULL;
> + unsigned long flags;
> +
> + CMD_STATE(sc) = SNIC_IOREQ_NOT_INITED;
> + CMD_FLAGS(sc) = SNIC_NO_FLAGS;
> + sg_cnt = scsi_dma_map(sc);
> + if (sg_cnt < 0) {
> + SNIC_TRC((u16)snic->shost->host_no, tag, (u64)sc, 0,
> + sc->cmnd[0], sg_cnt, CMD_STATE(sc));
> +
> + SNIC_HOST_ERR(snic->shost, "issue_sc:Failed to map SG List.\n");
> + ret = -ENOMEM;
> +
> + goto issue_sc_end;
> + }
> +
> + rqi = snic_req_init(snic, sg_cnt);
> + if (!rqi) {
> + scsi_dma_unmap(sc);
> + ret = -ENOMEM;
> +
> + goto issue_sc_end;
> + }
> +
> + rqi->tgt_id = tgt->id;
> + rqi->sc = sc;
> +
> + CMD_STATE(sc) = SNIC_IOREQ_PENDING;
> + CMD_SP(sc) = (char *) rqi;
> + cmd_trc = SNIC_TRC_CMD(sc);
> + CMD_FLAGS(sc) |= (SNIC_IO_INITIALIZED | SNIC_IO_ISSUED);
> + cmd_st_flags = SNIC_TRC_CMD_STATE_FLAGS(sc);
> + io_lock = snic_io_lock_hash(snic, sc);
> +
> + /* create wq desc and enqueue it */
> + ret = snic_queue_icmnd_req(snic, rqi, sc, sg_cnt);
> + if (ret) {
> + SNIC_HOST_ERR(snic->shost,
> + "issue_sc: icmnd qing Failed for sc %p, err %d\n",
> + sc, ret);
> +
> + spin_lock_irqsave(io_lock, flags);
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + CMD_SP(sc) = NULL;
> + CMD_STATE(sc) = SNIC_IOREQ_COMPLETE;
> + CMD_FLAGS(sc) &= ~SNIC_IO_ISSUED; /* turn off the flag */
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + if (rqi)
> + snic_release_req_buf(snic, rqi, sc);
> +
> + SNIC_TRC(snic->shost->host_no, tag, (u64)sc, 0, 0, 0,
> + SNIC_TRC_CMD_STATE_FLAGS(sc));
> + } else {
> + u32 io_sz = scsi_bufflen(sc) >> 9;
> + u32 qtime = jiffies - rqi->start_time;
> + struct snic_io_stats *iostats = &snic->s_stats.io;
> +
> + if (io_sz > atomic64_read(&iostats->max_io_sz))
> + atomic64_set(&iostats->max_io_sz, io_sz);
> +
> + if (qtime > atomic64_read(&iostats->max_qtime))
> + atomic64_set(&iostats->max_qtime, qtime);
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "issue_sc:sc %p, tag %d queued to WQ.\n",
> + sc, tag);
> +
> + SNIC_TRC(snic->shost->host_no, tag, (u64)sc, (u64)rqi, sg_cnt,
> + cmd_trc, cmd_st_flags);
> + }
> +
> +issue_sc_end:
> +
> + return ret;
> +} /* end of snic_issue_scsi_req */
> +
> +
> +/*
> + * snic_queuecommand
> + * Routine to send a scsi cdb to LLD
> + * Called with host_lock held and interrupts disabled
> + */
> +int
> +snic_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *sc)
> +{
> + struct snic_tgt *tgt = NULL;
> + struct snic *snic = shost_priv(shost);
> + int ret;
> +
> + tgt = starget_to_tgt(scsi_target(sc->device));
> + ret = snic_tgt_chkready(tgt);
> + if (ret) {
> + SNIC_HOST_ERR(shost, "Tgt %p id %d Not Ready.\n", tgt, tgt->id);
> + atomic64_inc(&snic->s_stats.misc.tgt_not_rdy);
> + sc->result = ret;
> + sc->scsi_done(sc);
> +
> + return 0;
> + }
> +
> + if (snic_get_state(snic) != SNIC_ONLINE) {
> + SNIC_HOST_ERR(shost, "snic state is %s\n",
> + snic_state_str[snic_get_state(snic)]);
> +
> + return SCSI_MLQUEUE_HOST_BUSY;
> + }
> + atomic_inc(&snic->ios_inflight);
> +
> + SNIC_SCSI_DBG(shost, "sc %p Tag %d (sc %0x) lun %lld in snic_qcmd\n",
> + sc, snic_cmd_tag(sc), sc->cmnd[0], sc->device->lun);
> +
> +#ifdef SNIC_DEBUG
> + scsi_print_command(sc);
> +#endif
> +
> + ret = snic_issue_scsi_req(snic, tgt, sc);
> + if (ret) {
> + SNIC_HOST_ERR(shost, "Failed to Q, Scsi Req w/ err %d.\n", ret);
> + ret = SCSI_MLQUEUE_HOST_BUSY;
> + } else
> + snic_stats_update_active_ios(&snic->s_stats);
> +
> + atomic_dec(&snic->ios_inflight);
> +
> + return ret;
> +} /* end of snic_queuecommand */
> +
> +/*
> + * snic_process_abts_pending_state:
> + * caller should hold IO lock
> + */
> +static void
> +snic_proc_tmreq_pending_state(struct snic *snic,
> + struct scsi_cmnd *sc,
> + u8 cmpl_status)
> +{
> + int state = CMD_STATE(sc);
> +
> + if (state == SNIC_IOREQ_ABTS_PENDING)
> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_PENDING;
> + else if (state == SNIC_IOREQ_LR_PENDING)
> + CMD_FLAGS(sc) |= SNIC_DEV_RST_PENDING;
> + else
> + SNIC_BUG_ON(1);
> +
> + switch (cmpl_status) {
> + case SNIC_STAT_IO_SUCCESS:
> + CMD_FLAGS(sc) |= SNIC_IO_DONE;
> + break;
> +
> + case SNIC_STAT_ABORTED:
> + CMD_FLAGS(sc) |= SNIC_IO_ABORTED;
> + break;
> +
> + default:
> + SNIC_BUG_ON(1);
> + }
> +}
> +
> +/*
> + * snic_process_io_failed_state:
> + * Processes IO's error states
> + */
> +static void
> +snic_process_io_failed_state(struct snic *snic,
> + struct snic_icmnd_cmpl *icmnd_cmpl,
> + struct scsi_cmnd *sc,
> + u8 cmpl_stat)
> +{
> + int res = 0;
> +
> + switch (cmpl_stat) {
> + case SNIC_STAT_TIMEOUT: /* Req was timedout */
> + atomic64_inc(&snic->s_stats.misc.io_tmo);
> + res = DID_TIME_OUT;
> + break;
> +
> + case SNIC_STAT_ABORTED: /* Req was aborted */
> + atomic64_inc(&snic->s_stats.misc.io_aborted);
> + res = DID_ERROR;
> + break;
> +
> + case SNIC_STAT_DATA_CNT_MISMATCH:/* Recv/Sent more/less data than exp */
> + atomic64_inc(&snic->s_stats.misc.data_cnt_mismat);
> + scsi_set_resid(sc, icmnd_cmpl->resid);
> + res = DID_ERROR;
> + break;
> +
> + case SNIC_STAT_OUT_OF_RES: /* Out of resources to complete request */
> + atomic64_inc(&snic->s_stats.fw.out_of_res);
> + res = DID_REQUEUE;
> + break;
> +
> + case SNIC_STAT_IO_NOT_FOUND: /* Requested I/O was not found */
> + atomic64_inc(&snic->s_stats.io.io_not_found);
> + res = DID_ERROR;
> + break;
> +
> + case SNIC_STAT_SGL_INVALID: /* Req was aborted to due to sgl error*/
> + atomic64_inc(&snic->s_stats.misc.sgl_inval);
> + res = DID_ERROR;
> + break;
> +
> + case SNIC_STAT_FW_ERR: /* Req terminated due to FW Error */
> + atomic64_inc(&snic->s_stats.fw.io_errs);
> + res = DID_ERROR;
> + break;
> +
> + case SNIC_STAT_SCSI_ERR: /* FW hits SCSI Error */
> + atomic64_inc(&snic->s_stats.fw.scsi_errs);
> + res = DID_ERROR;
> + break;
> +
> + case SNIC_STAT_INVALID_HDR: /* Hdr contains invalid data */
> + case SNIC_STAT_INVALID_PARM: /* Some param in req is invalid */
> + case SNIC_STAT_REQ_NOT_SUP: /* Req type is not supported */
> + case SNIC_STAT_CMND_REJECT: /* Req rejected */
> + case SNIC_STAT_DEV_OFFLINE: /* Device offline */
> + case SNIC_STAT_NOT_READY: /* XPT yet to initialize */
> + case SNIC_STAT_FATAL_ERROR: /* XPT Error */
> + default:
> + SNIC_SCSI_DBG(snic->shost,
> + "Invalid Hdr/Param or Req Not Supported or Cmnd
> Rejected or Device Offline. or Unknown\n");
> + res = DID_ERROR;
> + break;
> + }
> +
> + SNIC_HOST_ERR(snic->shost, "fw returns failed status %s flags 0x%x\n",
> + snic_io_status_to_str(cmpl_stat), CMD_FLAGS(sc));
> +
> + /* Set sc->result */
> + sc->result = (res << 16) | icmnd_cmpl->scsi_status;
> +} /* end of snic_process_io_failed_state */
> +
> +/*
> + * snic_tmreq_pending : is task management in progress.
> + */
> +static int
> +snic_tmreq_pending(struct scsi_cmnd *sc)
> +{
> + int state = CMD_STATE(sc);
> +
> + return ((state == SNIC_IOREQ_ABTS_PENDING) ||
> + (state == SNIC_IOREQ_LR_PENDING));
> +}
> +
> +/*
> + * snic_process_icmnd_cmpl_status:
> + * Caller should hold io_lock
> + */
> +static int
> +snic_process_icmnd_cmpl_status(struct snic *snic,
> + struct snic_icmnd_cmpl *icmnd_cmpl,
> + u8 cmpl_stat,
> + struct scsi_cmnd *sc)
> +{
> + u8 scsi_stat = icmnd_cmpl->scsi_status;
> + u64 xfer_len = 0;
> + int ret = 0;
> +
> + /* Mark the IO as complete */
> + CMD_STATE(sc) = SNIC_IOREQ_COMPLETE;
> +
> + if (likely(cmpl_stat == SNIC_STAT_IO_SUCCESS)) {
> + sc->result = (DID_OK << 16) | scsi_stat;
> +
> + xfer_len = scsi_bufflen(sc);
> +
> + /* Update SCSI Cmd with resid value */
> + scsi_set_resid(sc, icmnd_cmpl->resid);
> +
> + if (icmnd_cmpl->flags & SNIC_ICMND_CMPL_UNDR_RUN) {
> + xfer_len -= icmnd_cmpl->resid;
> + atomic64_inc(&snic->s_stats.misc.io_under_run);
> + }
> +
> + if (icmnd_cmpl->scsi_status == SAM_STAT_TASK_SET_FULL)
> + atomic64_inc(&snic->s_stats.misc.qfull);
> +
> + ret = 0;
> + } else {
> + snic_process_io_failed_state(snic, icmnd_cmpl, sc, cmpl_stat);
> + atomic64_inc(&snic->s_stats.io.fail);
> + SNIC_HOST_ERR(snic->shost,
> + "icmnd_cmpl: IO Failed : Hdr Status %s flags
> 0x%x\n",
> + snic_io_status_to_str(cmpl_stat), CMD_FLAGS(sc));
> + ret = 1;
> + }
> +
> + return ret;
> +} /* end of snic_process_icmnd_cmpl_status */
> +
> +
> +/*
> + * snic_icmnd_cmpl_handler
> + * Routine to handle icmnd completions
> + */
> +static void
> +snic_icmnd_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq)
> +{
> + u8 typ, hdr_stat;
> + u32 cmnd_id, hid;
> + u64 ctx;
> + struct scsi_cmnd *sc = NULL;
> + struct snic_icmnd_cmpl *icmnd_cmpl = NULL;
> + struct snic_host_req *req = NULL;
> + struct snic_req_info *rqi = NULL;
> + unsigned long flags, start_time;
> + spinlock_t *io_lock;
> + u8 sc_stat = 0;
> +
> + snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
> + icmnd_cmpl = &fwreq->u.icmnd_cmpl;
> + sc_stat = icmnd_cmpl->scsi_status;
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "Icmnd_cmpl: type = %x, hdr_stat = %x, cmnd_id = %x, hid
> = %x,i ctx = %llx\n",
> + typ, hdr_stat, cmnd_id, hid, ctx);
> +
> + if (cmnd_id >= snic->max_tag_id) {
> + SNIC_HOST_ERR(snic->shost,
> + "Icmnd_cmpl:Tag Error:Out of Range Tag %d, hdr
> status = %s\n",
> + cmnd_id, snic_io_status_to_str(hdr_stat));
> + return;
> + }
> +
> + sc = scsi_host_find_tag(snic->shost, cmnd_id);
> + WARN_ON_ONCE(!sc);
> +
> + if (!sc) {
> + atomic64_inc(&snic->s_stats.io.sc_null);
> + SNIC_HOST_ERR(snic->shost,
> + "Icmnd_cmpl: Scsi Cmnd Not found, sc = NULL Hdr
> Status = %s tag = 0x%x fwreq = 0x%p\n",
> + snic_io_status_to_str(hdr_stat),
> + cmnd_id,
> + fwreq);
> +
> + SNIC_TRC(snic->shost->host_no, cmnd_id, 0,
> + ((u64)hdr_stat << 16 |
> + (u64)sc_stat << 8 | (u64)icmnd_cmpl->flags),
> + fwreq, icmnd_cmpl->resid, ctx);
> +
> + return;
> + }
> +
> +
> +#ifdef SNIC_DEBUG
> + scsi_print_result(sc);
> +#endif
> +
> + io_lock = snic_io_lock_hash(snic, sc);
> +
> + spin_lock_irqsave(io_lock, flags);
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + SNIC_SCSI_DBG(snic->shost,
> + "Icmnd_cmpl:lun %lld sc %p cmd %xtag %d flags %xrqi %p\n",
> + sc->device->lun, sc, sc->cmnd[0], snic_cmd_tag(sc),
> + CMD_FLAGS(sc), rqi);
> +
> + SNIC_BUG_ON(rqi != (struct snic_req_info *)ctx);
> + WARN_ON_ONCE(req);
Now that is curious.
'ctx' already points to the 'snic_req_info', so why do we need to
have this double lookup here?
'scsi_host_find_tag' above already will tell us if the command is
valid; the tag map is managed by the driver, after all, so the
driver will guarantee that a busy tag will not be reused.
Why can't we do away with the 'CMD_SP' reference and get the
scsi command via tag lookup?
That would simplify the code by quite a bit ...
> + if (!rqi) {
> + atomic64_inc(&snic->s_stats.io.req_null);
> + CMD_FLAGS(sc) |= SNIC_IO_REQ_NULL;
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + SNIC_HOST_ERR(snic->shost,
> + "Icmnd_cmpl:Host Req Not Found(null), Hdr Status
> %s, Tag 0x%x, sc 0x%p flags 0x%x\n",
> + snic_io_status_to_str(hdr_stat),
> + cmnd_id, sc, CMD_FLAGS(sc));
> + return;
> + }
> +
> + rqi = (struct snic_req_info *) ctx;
> + start_time = rqi->start_time;
> +
> + /* firmware completed the io */
> + rqi->io_cmpl = 1;
> +
> + /*
> + * if SCSI-ML has already issued abort on this command,
> + * ignore completion of the IO. The abts path will clean it up
> + */
> + if (unlikely(snic_tmreq_pending(sc))) {
> + snic_proc_tmreq_pending_state(snic, sc, hdr_stat);
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + snic_stats_update_io_cmpl(&snic->s_stats);
> +
> + /* Expected value is SNIC_STAT_ABORTED */
> + if (likely(hdr_stat == SNIC_STAT_ABORTED))
> + return;
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "icmnd_cmpl:TM Req Pending(%s), Hdr Status %s sc
> 0x%p scsi status %x resid %d flags 0x%x\n",
> + snic_ioreq_state_to_str(CMD_STATE(sc)),
> + snic_io_status_to_str(hdr_stat),
> + sc, sc_stat, icmnd_cmpl->resid, CMD_FLAGS(sc));
> +
> + SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
> + jiffies_to_msecs(jiffies - start_time), fwreq,
> + SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
> +
> + return;
> + }
> +
> + if (snic_process_icmnd_cmpl_status(snic, icmnd_cmpl, hdr_stat, sc)) {
> +#ifdef SNIC_DEBUG
> + scsi_print_command(sc);
> +#endif
> + SNIC_HOST_ERR(snic->shost,
> + "icmnd_cmpl:IO Failed, sc 0x%p Tag %d Cmd %x Hdr
> Status %s flags 0x%x\n",
> + sc, sc->cmnd[0], cmnd_id,
> + snic_io_status_to_str(hdr_stat), CMD_FLAGS(sc));
> + }
> +
> + /* Break link with the SCSI Command */
> + CMD_SP(sc) = NULL;
> + CMD_FLAGS(sc) |= SNIC_IO_DONE;
> +
And then we wouldn't have to modify the CMD_SP() here ...
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + /* For now, consider only successful IO. */
> + snic_calc_io_process_time(snic, rqi);
> +
> + snic_release_req_buf(snic, rqi, sc);
> +
> + SNIC_TRC(snic->shost->host_no, cmnd_id, (u64)sc,
> + jiffies_to_msecs(jiffies - start_time), (u64)fwreq,
> + SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
> +
> +
> + if (sc->scsi_done)
> + sc->scsi_done(sc);
> +
> + snic_stats_update_io_cmpl(&snic->s_stats);
> +} /* end of snic_icmnd_cmpl_handler */
> +
> +static void
> +snic_proc_dr_cmpl_locked(struct snic *snic,
> + struct snic_fw_req *fwreq,
> + u8 cmpl_stat,
> + u32 cmnd_id,
> + struct scsi_cmnd *sc)
> +{
> + struct snic_req_info *rqi = (struct snic_req_info *) CMD_SP(sc);
> + u32 start_time = rqi->start_time;
> +
> + CMD_LR_STATUS(sc) = cmpl_stat;
> +
> + SNIC_SCSI_DBG(snic->shost, "itmf_cmpl: Cmd State = %s\n",
> + snic_ioreq_state_to_str(CMD_STATE(sc)));
> +
> + if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING) {
> + CMD_FLAGS(sc) |= SNIC_DEV_RST_ABTS_PENDING;
> +
> + SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
> + jiffies_to_msecs(jiffies - start_time),
> + fwreq, 0, SNIC_TRC_CMD_STATE_FLAGS(sc));
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "itmf_cmpl: Terminate Pending Dev Reset Cmpl
> Recvd.id %x, status %s flags 0x%x\n",
> + (int)(cmnd_id & SNIC_TAG_MASK),
> + snic_io_status_to_str(cmpl_stat),
> + CMD_FLAGS(sc));
> +
> + return;
> + }
> +
> +
> + if (CMD_FLAGS(sc) & SNIC_DEV_RST_TIMEDOUT) {
> + SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
> + jiffies_to_msecs(jiffies - start_time),
> + fwreq, 0, SNIC_TRC_CMD_STATE_FLAGS(sc));
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "itmf_cmpl:Dev Reset Completion Received after
> timeout. id %d cmpl status %s flags 0x%x\n",
> + (int)(cmnd_id & SNIC_TAG_MASK),
> + snic_io_status_to_str(cmpl_stat),
> + CMD_FLAGS(sc));
> +
> + return;
> + }
> +
> + CMD_STATE(sc) = SNIC_IOREQ_LR_COMPLETE;
> + CMD_FLAGS(sc) |= SNIC_DEV_RST_DONE;
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "itmf_cmpl:Dev Reset Cmpl Recvd id %d cmpl status %s
> flags 0x%x\n",
> + (int)(cmnd_id & SNIC_TAG_MASK),
> + snic_io_status_to_str(cmpl_stat),
> + CMD_FLAGS(sc));
> +
> + if (rqi->dr_done)
> + complete(rqi->dr_done);
> +} /* end of snic_proc_dr_cmpl_locked */
> +
> +/*
> + * snic_update_abort_stats : Updates abort stats based on completion status.
> + */
> +static void
> +snic_update_abort_stats(struct snic *snic, u8 cmpl_stat)
> +{
> + struct snic_abort_stats *abt_stats = &snic->s_stats.abts;
> +
> + SNIC_SCSI_DBG(snic->shost, "Updating Abort stats.\n");
> +
> + switch (cmpl_stat) {
> + case SNIC_STAT_IO_SUCCESS:
> + break;
> +
> + case SNIC_STAT_TIMEOUT:
> + atomic64_inc(&abt_stats->fw_tmo);
> + break;
> +
> + case SNIC_STAT_IO_NOT_FOUND:
> + atomic64_inc(&abt_stats->io_not_found);
> + break;
> +
> + default:
> + atomic64_inc(&abt_stats->fail);
> + break;
> + }
> +}
> +
> +static int
> +snic_process_itmf_cmpl(struct snic *snic,
> + struct snic_fw_req *fwreq,
> + u32 cmnd_id,
> + u8 cmpl_stat,
> + struct scsi_cmnd *sc)
> +{
> + struct snic_req_info *rqi = NULL;
> + u32 tm_tags = 0;
> + spinlock_t *io_lock = NULL;
> + unsigned long flags;
> + u32 start_time = 0;
> + int ret = 0;
> +
> + io_lock = snic_io_lock_hash(snic, sc);
> + spin_lock_irqsave(io_lock, flags);
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + WARN_ON_ONCE(!rqi);
> +
> + if (!rqi) {
> + atomic64_inc(&snic->s_stats.io.req_null);
> + spin_unlock_irqrestore(io_lock, flags);
> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_REQ_NULL;
> + SNIC_HOST_ERR(snic->shost,
> + "itmf_cmpl: rqi is null,Hdr stat = %s Tag = 0x%x
> sc = 0x%p flags 0x%x\n",
> + snic_io_status_to_str(cmpl_stat), cmnd_id, sc,
> + CMD_FLAGS(sc));
> +
> + return ret;
> + }
> +
> + /* Extract task management flags */
> + tm_tags = cmnd_id & ~(SNIC_TAG_MASK);
> +
> + start_time = rqi->start_time;
> + cmnd_id &= (SNIC_TAG_MASK);
> +
> + switch (tm_tags) {
> + case SNIC_TAG_ABORT:
> + /* Abort only issued on cmd */
> + snic_update_abort_stats(snic, cmpl_stat);
> +
> + if (CMD_STATE(sc) != SNIC_IOREQ_ABTS_PENDING) {
> + /* This is a late completion. Ignore it. */
> + ret = -1;
> + spin_unlock_irqrestore(io_lock, flags);
> + break;
> + }
> +
> + CMD_STATE(sc) = SNIC_IOREQ_ABTS_COMPLETE;
> + CMD_ABTS_STATUS(sc) = cmpl_stat;
> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_DONE;
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "itmf_cmpl:Abort Cmpl Recvd.Tag 0x%x Status %s
> flags 0x%x\n",
> + cmnd_id,
> + snic_io_status_to_str(cmpl_stat),
> + CMD_FLAGS(sc));
> +
> + /*
> + * If scsi_eh thread is blocked waiting for abts complete,
> + * signal completion to it. IO will be cleaned in the thread,
> + * else clean it in this context.
> + */
> + if (rqi->abts_done) {
> + complete(rqi->abts_done);
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + break; /* jump out */
> + }
> +
> + CMD_SP(sc) = NULL;
> + sc->result = (DID_ERROR << 16);
> + SNIC_SCSI_DBG(snic->shost,
> + "itmf_cmpl: Completing IO. sc %p flags 0x%x\n",
> + sc, CMD_FLAGS(sc));
> +
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + snic_release_req_buf(snic, rqi, sc);
> +
> + if (sc->scsi_done) {
> + SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
> + jiffies_to_msecs(jiffies - start_time),
> + fwreq,
> + SNIC_TRC_CMD(sc),
> + SNIC_TRC_CMD_STATE_FLAGS(sc));
> +
> + sc->scsi_done(sc);
> + }
> +
> + break;
> +
> + case SNIC_TAG_DEV_RST:
> + case SNIC_TAG_DEV_RST | SNIC_TAG_IOCTL_DEV_RST:
> + snic_proc_dr_cmpl_locked(snic, fwreq, cmpl_stat, cmnd_id, sc);
> + spin_unlock_irqrestore(io_lock, flags);
> + ret = 0;
> +
> + break;
> +
> + case SNIC_TAG_ABORT | SNIC_TAG_DEV_RST:
> + /* Abort and terminate completion of device reset req */
> +
> + CMD_STATE(sc) = SNIC_IOREQ_ABTS_COMPLETE;
> + CMD_ABTS_STATUS(sc) = cmpl_stat;
> + CMD_FLAGS(sc) |= SNIC_DEV_RST_DONE;
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "itmf_cmpl:dev reset abts cmpl recvd. id %d
> status %s flags 0x%x\n",
> + cmnd_id, snic_io_status_to_str(cmpl_stat),
> + CMD_FLAGS(sc));
> +
> + if (rqi->abts_done)
> + complete(rqi->abts_done);
> +
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + break;
> +
> + default:
> + spin_unlock_irqrestore(io_lock, flags);
> + SNIC_HOST_ERR(snic->shost,
> + "itmf_cmpl: Unknown TM tag bit 0x%x\n", tm_tags);
> +
> + SNIC_HOST_ERR(snic->shost,
> + "itmf_cmpl:Unexpected itmf io stat %s Tag = 0x%x
> flags 0x%x\n",
> + snic_ioreq_state_to_str(CMD_STATE(sc)),
> + cmnd_id,
> + CMD_FLAGS(sc));
> + ret = -1;
> + SNIC_BUG_ON(1);
> +
> + break;
> + }
> +
> + return ret;
> +} /* end of snic_process_itmf_cmpl_status */
> +
> +/*
> + * snic_itmf_cmpl_handler.
> + * Routine to handle itmf completions.
> + */
> +static void
> +snic_itmf_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq)
> +{
> + struct scsi_cmnd *sc = NULL;
> + struct snic_req_info *rqi = NULL;
> + struct snic_itmf_cmpl *itmf_cmpl = NULL;
> + u64 ctx;
> + u32 cmnd_id;
> + u32 hid;
> + u8 typ;
> + u8 hdr_stat;
> +
> + snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
> + SNIC_SCSI_DBG(snic->shost,
> + "Itmf_cmpl: %s: type = %x, hdr_stat = %x, cmnd_id = %x,
> hid = %x,ctx = %llx\n",
> + __func__, typ, hdr_stat, cmnd_id, hid, ctx);
> +
> + itmf_cmpl = &fwreq->u.itmf_cmpl;
> + SNIC_SCSI_DBG(snic->shost,
> + "Itmf_cmpl: nterm %u , flags 0x%x\n",
> + itmf_cmpl->nterminated, itmf_cmpl->flags);
> +
> + /* spl case, dev reset issued through ioctl */
> + if (cmnd_id & SNIC_TAG_IOCTL_DEV_RST) {
> + rqi = (struct snic_req_info *) ctx;
> + sc = rqi->sc;
> +
> + goto ioctl_dev_rst;
> + }
> +
> + if ((cmnd_id & SNIC_TAG_MASK) >= snic->max_tag_id) {
> + SNIC_HOST_ERR(snic->shost,
> + "Itmf_cmpl: Tag 0x%x out of Range,HdrStat %s\n",
> + cmnd_id, snic_io_status_to_str(hdr_stat));
> + SNIC_BUG_ON(1);
> +
> + return;
> + }
> +
> + sc = scsi_host_find_tag(snic->shost, cmnd_id & SNIC_TAG_MASK);
> + WARN_ON_ONCE(!sc);
> +
Same issue here.
Why can't you rely on the tag to get the command?
Then you wouldn't have to to the triple indirection
rqi->sc->rpi
_and_ you would be insulated against command timeouts, where
'sc' might become invalid ...
> +ioctl_dev_rst:
> + if (!sc) {
> + atomic64_inc(&snic->s_stats.io.sc_null);
> + SNIC_HOST_ERR(snic->shost,
> + "Itmf_cmpl: sc is NULL - Hdr Stat %s Tag 0x%x\n",
> + snic_io_status_to_str(hdr_stat), cmnd_id);
> +
> + return;
> + }
> +
> + snic_process_itmf_cmpl(snic, fwreq, cmnd_id, hdr_stat, sc);
> +} /* end of snic_itmf_cmpl_handler */
> +
> +
> +
> +static void
> +snic_hba_reset_scsi_cleanup(struct snic *snic, struct scsi_cmnd *sc)
> +{
> + struct snic_stats *st = &snic->s_stats;
> + long act_ios = 0, act_fwreqs = 0;
> +
> + SNIC_SCSI_DBG(snic->shost, "HBA Reset scsi cleanup.\n");
> + snic_scsi_cleanup(snic, snic_cmd_tag(sc));
> +
> + /* Update stats on pending IOs */
> + act_ios = atomic64_read(&st->io.active);
> + atomic64_add(act_ios, &st->io.compl);
> + atomic64_sub(act_ios, &st->io.active);
> +
> + act_fwreqs = atomic64_read(&st->fw.actv_reqs);
> + atomic64_sub(act_fwreqs, &st->fw.actv_reqs);
> +}
> +
> +/*
> + * snic_hba_reset_cmpl_handler :
> + *
> + * Notes :
> + * 1. Cleanup all the scsi cmds, release all snic specific cmds
> + * 2. Issue Report Targets in case of SAN targets
> + */
> +static int
> +snic_hba_reset_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq)
> +{
> + u64 ctx;
> + u32 cmnd_id;
> + u32 hid;
> + u8 typ;
> + u8 hdr_stat;
> + struct scsi_cmnd *sc = NULL;
> + struct snic_req_info *rqi = NULL;
> + spinlock_t *io_lock = NULL;
> + unsigned long flags, gflags;
> + int ret = 0;
> +
> + SNIC_HOST_INFO(snic->shost,
> + "reset_cmpl:HBA Reset Completion received.\n");
> +
> + snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
> + SNIC_SCSI_DBG(snic->shost,
> + "reset_cmpl: type = %x, hdr_stat = %x, cmnd_id = %x, hid
> = %x, ctx = %llx\n",
> + typ, hdr_stat, cmnd_id, hid, ctx);
> +
> + /* spl case, host reset issued through ioctl */
> + if (cmnd_id == SCSI_NO_TAG) {
> + rqi = (struct snic_req_info *) ctx;
> + sc = rqi->sc;
> +
> + goto ioctl_hba_rst;
> + }
> +
> + if (cmnd_id >= snic->max_tag_id) {
> + SNIC_HOST_ERR(snic->shost,
> + "reset_cmpl: Tag 0x%x out of Range,HdrStat %s\n",
> + cmnd_id, snic_io_status_to_str(hdr_stat));
> + SNIC_BUG_ON(1);
> +
> + return 1;
> + }
> +
> + sc = scsi_host_find_tag(snic->shost, cmnd_id);
> +ioctl_hba_rst:
> + if (!sc) {
> + atomic64_inc(&snic->s_stats.io.sc_null);
> + SNIC_HOST_ERR(snic->shost,
> + "reset_cmpl: sc is NULL - Hdr Stat %s Tag 0x%x\n",
> + snic_io_status_to_str(hdr_stat), cmnd_id);
> + ret = 1;
> +
> + return ret;
> + }
> +
> + io_lock = snic_io_lock_hash(snic, sc);
> + spin_lock_irqsave(io_lock, flags);
> +
> + if (!snic->remove_wait) {
> + spin_unlock_irqrestore(io_lock, flags);
> + SNIC_HOST_ERR(snic->shost,
> + "reset_cmpl:host reset completed after timout\n");
> + ret = 1;
> +
> + return ret;
> + }
> +
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + WARN_ON_ONCE(!rqi);
> +
See above.
If possible I would avoid keeping direct references to the scsi
command in the firmware request structure; the scsi command might
timeout while the firmware request is in flight, leaving the
firmware request with a stale pointer.
I would rather use the tag in the firmware request and use
the tag lookup to get to the command.
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
h...@suse.de +49 911 74053 688
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: F. Imendörffer, J. Smithard, J. Guild, D. Upmanyu, G. Norton
HRB 21284 (AG Nürnberg)
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
