The patch below does not apply to the 5.15-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable@xxxxxxxxxxxxxxx>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-5.15.y git checkout FETCH_HEAD git cherry-pick -x 9ae615c5bfd37bd091772969b1153de5335ea986 # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable@xxxxxxxxxxxxxxx>' --in-reply-to '2023072148-easiness-jawed-2659@gregkh' --subject-prefix 'PATCH 5.15.y' HEAD^.. Possible dependencies: 9ae615c5bfd3 ("scsi: qla2xxx: Fix hang in task management") 6a87679626b5 ("scsi: qla2xxx: Fix task management cmd fail due to unavailable resource") 9803fb5d2759 ("scsi: qla2xxx: Fix task management cmd failure") d90171dd0da5 ("scsi: qla2xxx: Multi-que support for TMF") 5f63a163ed2f ("scsi: qla2xxx: Fix exchange oversubscription for management commands") 41e5afe51f75 ("scsi: qla2xxx: Fix exchange oversubscription") 68ad83188d78 ("scsi: qla2xxx: Fix crash when I/O abort times out") 1b80addaae09 ("scsi: qla2xxx: Remove unused declarations for qla2xxx") f12d2d130efc ("scsi: qla2xxx: Add debug prints in the device remove path") 118b0c863c8f ("scsi: qla2xxx: Fix losing target when it reappears during delete") 63ab6cb582fa ("scsi: qla2xxx: edif: Fix I/O timeout due to over-subscription") c02aada06d19 ("scsi: qla2xxx: Fix hang due to session stuck") 31e6cdbe0eae ("scsi: qla2xxx: Implement ref count for SRB") d4523bd6fd5d ("scsi: qla2xxx: Refactor asynchronous command initialization") 91f6f5fbe87b ("scsi: qla2xxx: edif: Reduce connection thrash") bb2ca6b3f09a ("scsi: qla2xxx: Relogin during fabric disturbance") thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From 9ae615c5bfd37bd091772969b1153de5335ea986 Mon Sep 17 00:00:00 2001 From: Quinn Tran <qutran@xxxxxxxxxxx> Date: Fri, 28 Apr 2023 00:53:36 -0700 Subject: [PATCH] scsi: qla2xxx: Fix hang in task management Task management command hangs where a side band chip reset failed to nudge the TMF from it's current send path. Add additional error check to block TMF from entering during chip reset and along the TMF path to cause it to bail out, skip over abort of marker. Cc: stable@xxxxxxxxxxxxxxx Signed-off-by: Quinn Tran <qutran@xxxxxxxxxxx> Signed-off-by: Nilesh Javali <njavali@xxxxxxxxxxx> Link: https://lore.kernel.org/r/20230428075339.32551-5-njavali@xxxxxxxxxxx Reviewed-by: Himanshu Madhani <himanshu.madhani@xxxxxxxxxx> Signed-off-by: Martin K. Petersen <martin.petersen@xxxxxxxxxx> diff --git a/drivers/scsi/qla2xxx/qla_def.h b/drivers/scsi/qla2xxx/qla_def.h index e345ccbff807..dfee3b41bdf1 100644 --- a/drivers/scsi/qla2xxx/qla_def.h +++ b/drivers/scsi/qla2xxx/qla_def.h @@ -5515,4 +5515,8 @@ struct ql_vnd_tgt_stats_resp { _fp->disc_state, _fp->scan_state, _fp->loop_id, _fp->deleted, \ _fp->flags +#define TMF_NOT_READY(_fcport) \ + (!_fcport || IS_SESSION_DELETED(_fcport) || atomic_read(&_fcport->state) != FCS_ONLINE || \ + !_fcport->vha->hw->flags.fw_started) + #endif diff --git a/drivers/scsi/qla2xxx/qla_init.c b/drivers/scsi/qla2xxx/qla_init.c index 84841edcd1b5..0df6eae7324e 100644 --- a/drivers/scsi/qla2xxx/qla_init.c +++ b/drivers/scsi/qla2xxx/qla_init.c @@ -1996,6 +1996,11 @@ qla2x00_tmf_iocb_timeout(void *data) int rc, h; unsigned long flags; + if (sp->type == SRB_MARKER) { + complete(&tmf->u.tmf.comp); + return; + } + rc = qla24xx_async_abort_cmd(sp, false); if (rc) { spin_lock_irqsave(sp->qpair->qp_lock_ptr, flags); @@ -2023,6 +2028,7 @@ static void qla_marker_sp_done(srb_t *sp, int res) sp->handle, sp->fcport->d_id.b24, sp->u.iocb_cmd.u.tmf.flags, sp->u.iocb_cmd.u.tmf.lun, sp->qpair->id); + sp->u.iocb_cmd.u.tmf.data = res; complete(&tmf->u.tmf.comp); } @@ -2039,6 +2045,11 @@ static void qla_marker_sp_done(srb_t *sp, int res) } while (cnt); \ } +/** + * qla26xx_marker: send marker IOCB and wait for the completion of it. + * @arg: pointer to argument list. + * It is assume caller will provide an fcport pointer and modifier + */ static int qla26xx_marker(struct tmf_arg *arg) { @@ -2048,6 +2059,14 @@ qla26xx_marker(struct tmf_arg *arg) int rval = QLA_FUNCTION_FAILED; fc_port_t *fcport = arg->fcport; + if (TMF_NOT_READY(arg->fcport)) { + ql_dbg(ql_dbg_taskm, vha, 0x8039, + "FC port not ready for marker loop-id=%x portid=%06x modifier=%x lun=%lld qp=%d.\n", + fcport->loop_id, fcport->d_id.b24, + arg->modifier, arg->lun, arg->qpair->id); + return QLA_SUSPENDED; + } + /* ref: INIT */ sp = qla2xxx_get_qpair_sp(vha, arg->qpair, fcport, GFP_KERNEL); if (!sp) @@ -2074,11 +2093,19 @@ qla26xx_marker(struct tmf_arg *arg) if (rval != QLA_SUCCESS) { ql_log(ql_log_warn, vha, 0x8031, - "Marker IOCB failed (%x).\n", rval); + "Marker IOCB send failure (%x).\n", rval); goto done_free_sp; } wait_for_completion(&tm_iocb->u.tmf.comp); + rval = tm_iocb->u.tmf.data; + + if (rval != QLA_SUCCESS) { + ql_log(ql_log_warn, vha, 0x8019, + "Marker failed hdl=%x loop-id=%x portid=%06x modifier=%x lun=%lld qp=%d rval %d.\n", + sp->handle, fcport->loop_id, fcport->d_id.b24, + arg->modifier, arg->lun, sp->qpair->id, rval); + } done_free_sp: /* ref: INIT */ @@ -2091,6 +2118,8 @@ static void qla2x00_tmf_sp_done(srb_t *sp, int res) { struct srb_iocb *tmf = &sp->u.iocb_cmd; + if (res) + tmf->u.tmf.data = res; complete(&tmf->u.tmf.comp); } @@ -2104,6 +2133,14 @@ __qla2x00_async_tm_cmd(struct tmf_arg *arg) fc_port_t *fcport = arg->fcport; + if (TMF_NOT_READY(arg->fcport)) { + ql_dbg(ql_dbg_taskm, vha, 0x8032, + "FC port not ready for TM command loop-id=%x portid=%06x modifier=%x lun=%lld qp=%d.\n", + fcport->loop_id, fcport->d_id.b24, + arg->modifier, arg->lun, arg->qpair->id); + return QLA_SUSPENDED; + } + /* ref: INIT */ sp = qla2xxx_get_qpair_sp(vha, arg->qpair, fcport, GFP_KERNEL); if (!sp) @@ -2178,7 +2215,9 @@ int qla_get_tmf(fc_port_t *fcport) msleep(1); spin_lock_irqsave(&ha->tgt.sess_lock, flags); - if (fcport->deleted) { + if (TMF_NOT_READY(fcport)) { + ql_log(ql_log_warn, vha, 0x802c, + "Unable to acquire TM resource due to disruption.\n"); rc = EIO; break; } @@ -2204,7 +2243,10 @@ qla2x00_async_tm_cmd(fc_port_t *fcport, uint32_t flags, uint64_t lun, struct scsi_qla_host *vha = fcport->vha; struct qla_qpair *qpair; struct tmf_arg a; - int i, rval; + int i, rval = QLA_SUCCESS; + + if (TMF_NOT_READY(fcport)) + return QLA_SUSPENDED; a.vha = fcport->vha; a.fcport = fcport; @@ -2223,6 +2265,14 @@ qla2x00_async_tm_cmd(fc_port_t *fcport, uint32_t flags, uint64_t lun, qpair = vha->hw->queue_pair_map[i]; if (!qpair) continue; + + if (TMF_NOT_READY(fcport)) { + ql_log(ql_log_warn, vha, 0x8026, + "Unable to send TM due to disruption.\n"); + rval = QLA_SUSPENDED; + break; + } + a.qpair = qpair; a.flags = flags|TCF_NOTMCMD_TO_TARGET; rval = __qla2x00_async_tm_cmd(&a); @@ -2231,10 +2281,14 @@ qla2x00_async_tm_cmd(fc_port_t *fcport, uint32_t flags, uint64_t lun, } } + if (rval) + goto bailout; + a.qpair = vha->hw->base_qpair; a.flags = flags; rval = __qla2x00_async_tm_cmd(&a); +bailout: if (a.modifier == MK_SYNC_ID_LUN) qla_put_tmf(fcport);