freebsd-nq/sys/dev/mps/mps_sas.c

3571 lines
101 KiB
C
Raw Normal View History

/*-
* Copyright (c) 2009 Yahoo! Inc.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
/*-
* Copyright (c) 2011 LSI Corp.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* LSI MPT-Fusion Host Adapter FreeBSD
*
* $FreeBSD$
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
/* Communications core for LSI MPT2 */
/* TODO Move headers to mpsvar */
#include <sys/types.h>
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/selinfo.h>
#include <sys/module.h>
#include <sys/bus.h>
#include <sys/conf.h>
#include <sys/bio.h>
#include <sys/malloc.h>
#include <sys/uio.h>
#include <sys/sysctl.h>
Add Serial Management Protocol (SMP) passthrough support to CAM. This includes support in the kernel, camcontrol(8), libcam and the mps(4) driver for SMP passthrough. The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00 to determine supported pages, and will now fetch page 0x83 in addition to page 0x80 if supported. Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is intended for SMP requests and responses. The ADVINFO is currently used to fetch cached VPD page 0x83 data from the transport layer, but is intended to be extensible to fetch other types of device-specific data. SMP-only devices are not currently represented in the CAM topology, and so the current semantics are that the SIM will route SMP CCBs to either the addressed device, if it contains an SMP target, or its parent, if it contains an SMP target. (This is noted in cam_ccb.h, since it will change later once we have the ability to have SMP-only devices in CAM's topology.) smp_all.c, smp_all.h: New helper routines for SMP. This includes SMP request building routines, response parsing routines, error decoding routines, and structure definitions for a number of SMP commands. libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality is available to userland applications. camcontrol.8, camcontrol.c: Add smp passthrough support to camcontrol. Several new subcommands are now available: 'smpcmd' functions much like 'cmd', except that it allows the user to send generic SMP commands. 'smprg' sends the SMP report general command, and displays the decoded output. It will automatically fetch extended output if it is available. 'smppc' sends the SMP phy control command, with any number of potential options. Among other things, this allows the user to reset a phy on a SAS expander, or disable a phy on an expander. 'smpmaninfo' sends the SMP report manufacturer information and displays the decoded output. 'smpphylist' displays a list of phys on an expander, and the CAM devices attached to those phys, if any. cam.h, cam.c: Add a status value for SMP errors (CAM_SMP_STATUS_ERROR). Add a missing description for CAM_SCSI_IT_NEXUS_LOST. Add support for SMP commands to cam_error_string(). cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP commands are by nature bi-directional, and we may need to support bi-directional SCSI commands later. Add the XPT_SMP_IO CCB. Since SMP commands are bi-directional, there are pointers for both the request and response. Add a fill routine for SMP CCBs. Add the XPT_GDEV_ADVINFO CCB. This is currently used to fetch cached page 0x83 data from the transport later, but is extensible to fetch many other types of data. cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. cam_xpt.c: Add support for executing XPT_SMP_IO CCBs. cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct cam_ed. scsi_all.c: Add scsi_get_sas_addr(), a function that parses VPD page 0x83 data and pulls out a SAS address. scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a prototype for scsi_get_sas_addr(). scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. scsi_xpt.c: In the SCSI probe code, first ask the device for VPD page 0x00. If any VPD pages are supported, that page is required to be implemented. Based on the response, we may probe for the serial number (page 0x80) or device id (page 0x83). Add support for the XPT_GDEV_ADVINFO CCB. sys/conf/files: Add smp_all.c. mps.c: Add support for passing in a uio in mps_map_command(), so we can map a S/G list at once. Add support for SMP passthrough commands in mps_data_cb(). SMP is a special case, because the first buffer in the S/G list is outbound and the second buffer is inbound. Add support for warning the user if the busdma code comes back with more buffers than will work for the command. This will, for example, help the user determine why an SMP command failed if busdma comes back with three buffers. mps_pci.c: Add sys/uio.h. mps_sas.c: Add the SAS address and the parent handle to the list of fields we pull from device page 0 and cache in struct mpssas_target. These are needed for SMP passthrough. Add support for the XPT_SMP_IO CCB. For now, this CCB is routed to the addressed device if it supports SMP, or to its parent if it does not and the parent does. This is necessary because CAM does not currently support SMP-only nodes in the topology. Make SMP passthrough support conditional on __FreeBSD_version >= 900026. This will make it easier to MFC this change to the driver without MFCing the CAM changes as well. mps_user.c: Un-staticize mpi_init_sge() so we can use it for the SMP passthrough code. mpsvar.h: Add a uio and iovecs into struct mps_command for SMP passthrough commands. Add a cm_max_segs field to struct mps_command so that we can warn the user if busdma comes back with too many segments. Clear the cm_reply when a command gets freed. If it is not cleared, reply frames will eventually get freed into the pool multiple times and corrupt the pool. (This fix is from scottl.) Add a prototype for mpi_init_sge(). sys/param.h: Bump __FreeBSD_version to 900026 for the for the inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO CAM CCBs.
2010-11-30 22:39:46 +00:00
#include <sys/endian.h>
#include <sys/queue.h>
#include <sys/kthread.h>
#include <sys/taskqueue.h>
#include <sys/sbuf.h>
#include <machine/bus.h>
#include <machine/resource.h>
#include <sys/rman.h>
#include <machine/stdarg.h>
#include <cam/cam.h>
#include <cam/cam_ccb.h>
#include <cam/cam_xpt.h>
#include <cam/cam_debug.h>
#include <cam/cam_sim.h>
#include <cam/cam_xpt_sim.h>
#include <cam/cam_xpt_periph.h>
#include <cam/cam_periph.h>
#include <cam/scsi/scsi_all.h>
#include <cam/scsi/scsi_message.h>
Add Serial Management Protocol (SMP) passthrough support to CAM. This includes support in the kernel, camcontrol(8), libcam and the mps(4) driver for SMP passthrough. The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00 to determine supported pages, and will now fetch page 0x83 in addition to page 0x80 if supported. Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is intended for SMP requests and responses. The ADVINFO is currently used to fetch cached VPD page 0x83 data from the transport layer, but is intended to be extensible to fetch other types of device-specific data. SMP-only devices are not currently represented in the CAM topology, and so the current semantics are that the SIM will route SMP CCBs to either the addressed device, if it contains an SMP target, or its parent, if it contains an SMP target. (This is noted in cam_ccb.h, since it will change later once we have the ability to have SMP-only devices in CAM's topology.) smp_all.c, smp_all.h: New helper routines for SMP. This includes SMP request building routines, response parsing routines, error decoding routines, and structure definitions for a number of SMP commands. libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality is available to userland applications. camcontrol.8, camcontrol.c: Add smp passthrough support to camcontrol. Several new subcommands are now available: 'smpcmd' functions much like 'cmd', except that it allows the user to send generic SMP commands. 'smprg' sends the SMP report general command, and displays the decoded output. It will automatically fetch extended output if it is available. 'smppc' sends the SMP phy control command, with any number of potential options. Among other things, this allows the user to reset a phy on a SAS expander, or disable a phy on an expander. 'smpmaninfo' sends the SMP report manufacturer information and displays the decoded output. 'smpphylist' displays a list of phys on an expander, and the CAM devices attached to those phys, if any. cam.h, cam.c: Add a status value for SMP errors (CAM_SMP_STATUS_ERROR). Add a missing description for CAM_SCSI_IT_NEXUS_LOST. Add support for SMP commands to cam_error_string(). cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP commands are by nature bi-directional, and we may need to support bi-directional SCSI commands later. Add the XPT_SMP_IO CCB. Since SMP commands are bi-directional, there are pointers for both the request and response. Add a fill routine for SMP CCBs. Add the XPT_GDEV_ADVINFO CCB. This is currently used to fetch cached page 0x83 data from the transport later, but is extensible to fetch many other types of data. cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. cam_xpt.c: Add support for executing XPT_SMP_IO CCBs. cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct cam_ed. scsi_all.c: Add scsi_get_sas_addr(), a function that parses VPD page 0x83 data and pulls out a SAS address. scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a prototype for scsi_get_sas_addr(). scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. scsi_xpt.c: In the SCSI probe code, first ask the device for VPD page 0x00. If any VPD pages are supported, that page is required to be implemented. Based on the response, we may probe for the serial number (page 0x80) or device id (page 0x83). Add support for the XPT_GDEV_ADVINFO CCB. sys/conf/files: Add smp_all.c. mps.c: Add support for passing in a uio in mps_map_command(), so we can map a S/G list at once. Add support for SMP passthrough commands in mps_data_cb(). SMP is a special case, because the first buffer in the S/G list is outbound and the second buffer is inbound. Add support for warning the user if the busdma code comes back with more buffers than will work for the command. This will, for example, help the user determine why an SMP command failed if busdma comes back with three buffers. mps_pci.c: Add sys/uio.h. mps_sas.c: Add the SAS address and the parent handle to the list of fields we pull from device page 0 and cache in struct mpssas_target. These are needed for SMP passthrough. Add support for the XPT_SMP_IO CCB. For now, this CCB is routed to the addressed device if it supports SMP, or to its parent if it does not and the parent does. This is necessary because CAM does not currently support SMP-only nodes in the topology. Make SMP passthrough support conditional on __FreeBSD_version >= 900026. This will make it easier to MFC this change to the driver without MFCing the CAM changes as well. mps_user.c: Un-staticize mpi_init_sge() so we can use it for the SMP passthrough code. mpsvar.h: Add a uio and iovecs into struct mps_command for SMP passthrough commands. Add a cm_max_segs field to struct mps_command so that we can warn the user if busdma comes back with too many segments. Clear the cm_reply when a command gets freed. If it is not cleared, reply frames will eventually get freed into the pool multiple times and corrupt the pool. (This fix is from scottl.) Add a prototype for mpi_init_sge(). sys/param.h: Bump __FreeBSD_version to 900026 for the for the inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO CAM CCBs.
2010-11-30 22:39:46 +00:00
#if __FreeBSD_version >= 900026
#include <cam/scsi/smp_all.h>
#endif
#include <dev/mps/mpi/mpi2_type.h>
#include <dev/mps/mpi/mpi2.h>
#include <dev/mps/mpi/mpi2_ioc.h>
#include <dev/mps/mpi/mpi2_sas.h>
#include <dev/mps/mpi/mpi2_cnfg.h>
#include <dev/mps/mpi/mpi2_init.h>
#include <dev/mps/mpi/mpi2_tool.h>
#include <dev/mps/mps_ioctl.h>
#include <dev/mps/mpsvar.h>
#include <dev/mps/mps_table.h>
#include <dev/mps/mps_sas.h>
#define MPSSAS_DISCOVERY_TIMEOUT 20
#define MPSSAS_MAX_DISCOVERY_TIMEOUTS 10 /* 200 seconds */
/*
* static array to check SCSI OpCode for EEDP protection bits
*/
#define PRO_R MPI2_SCSIIO_EEDPFLAGS_CHECK_REMOVE_OP
#define PRO_W MPI2_SCSIIO_EEDPFLAGS_INSERT_OP
#define PRO_V MPI2_SCSIIO_EEDPFLAGS_INSERT_OP
static uint8_t op_code_prot[256] = {
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, PRO_R, 0, PRO_W, 0, 0, 0, PRO_W, PRO_V,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, PRO_W, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, PRO_R, 0, PRO_W, 0, 0, 0, PRO_W, PRO_V,
0, 0, 0, PRO_W, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, PRO_R, 0, PRO_W, 0, 0, 0, PRO_W, PRO_V,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
};
MALLOC_DEFINE(M_MPSSAS, "MPSSAS", "MPS SAS memory");
static void mpssas_discovery_timeout(void *data);
static void mpssas_remove_device(struct mps_softc *, struct mps_command *);
static void mpssas_remove_complete(struct mps_softc *, struct mps_command *);
static void mpssas_action(struct cam_sim *sim, union ccb *ccb);
static void mpssas_poll(struct cam_sim *sim);
static void mpssas_scsiio_timeout(void *data);
static void mpssas_abort_complete(struct mps_softc *sc, struct mps_command *cm);
static void mpssas_direct_drive_io(struct mpssas_softc *sassc,
struct mps_command *cm, union ccb *ccb);
static void mpssas_action_scsiio(struct mpssas_softc *, union ccb *);
static void mpssas_scsiio_complete(struct mps_softc *, struct mps_command *);
static void mpssas_action_resetdev(struct mpssas_softc *, union ccb *);
Add Serial Management Protocol (SMP) passthrough support to CAM. This includes support in the kernel, camcontrol(8), libcam and the mps(4) driver for SMP passthrough. The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00 to determine supported pages, and will now fetch page 0x83 in addition to page 0x80 if supported. Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is intended for SMP requests and responses. The ADVINFO is currently used to fetch cached VPD page 0x83 data from the transport layer, but is intended to be extensible to fetch other types of device-specific data. SMP-only devices are not currently represented in the CAM topology, and so the current semantics are that the SIM will route SMP CCBs to either the addressed device, if it contains an SMP target, or its parent, if it contains an SMP target. (This is noted in cam_ccb.h, since it will change later once we have the ability to have SMP-only devices in CAM's topology.) smp_all.c, smp_all.h: New helper routines for SMP. This includes SMP request building routines, response parsing routines, error decoding routines, and structure definitions for a number of SMP commands. libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality is available to userland applications. camcontrol.8, camcontrol.c: Add smp passthrough support to camcontrol. Several new subcommands are now available: 'smpcmd' functions much like 'cmd', except that it allows the user to send generic SMP commands. 'smprg' sends the SMP report general command, and displays the decoded output. It will automatically fetch extended output if it is available. 'smppc' sends the SMP phy control command, with any number of potential options. Among other things, this allows the user to reset a phy on a SAS expander, or disable a phy on an expander. 'smpmaninfo' sends the SMP report manufacturer information and displays the decoded output. 'smpphylist' displays a list of phys on an expander, and the CAM devices attached to those phys, if any. cam.h, cam.c: Add a status value for SMP errors (CAM_SMP_STATUS_ERROR). Add a missing description for CAM_SCSI_IT_NEXUS_LOST. Add support for SMP commands to cam_error_string(). cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP commands are by nature bi-directional, and we may need to support bi-directional SCSI commands later. Add the XPT_SMP_IO CCB. Since SMP commands are bi-directional, there are pointers for both the request and response. Add a fill routine for SMP CCBs. Add the XPT_GDEV_ADVINFO CCB. This is currently used to fetch cached page 0x83 data from the transport later, but is extensible to fetch many other types of data. cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. cam_xpt.c: Add support for executing XPT_SMP_IO CCBs. cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct cam_ed. scsi_all.c: Add scsi_get_sas_addr(), a function that parses VPD page 0x83 data and pulls out a SAS address. scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a prototype for scsi_get_sas_addr(). scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. scsi_xpt.c: In the SCSI probe code, first ask the device for VPD page 0x00. If any VPD pages are supported, that page is required to be implemented. Based on the response, we may probe for the serial number (page 0x80) or device id (page 0x83). Add support for the XPT_GDEV_ADVINFO CCB. sys/conf/files: Add smp_all.c. mps.c: Add support for passing in a uio in mps_map_command(), so we can map a S/G list at once. Add support for SMP passthrough commands in mps_data_cb(). SMP is a special case, because the first buffer in the S/G list is outbound and the second buffer is inbound. Add support for warning the user if the busdma code comes back with more buffers than will work for the command. This will, for example, help the user determine why an SMP command failed if busdma comes back with three buffers. mps_pci.c: Add sys/uio.h. mps_sas.c: Add the SAS address and the parent handle to the list of fields we pull from device page 0 and cache in struct mpssas_target. These are needed for SMP passthrough. Add support for the XPT_SMP_IO CCB. For now, this CCB is routed to the addressed device if it supports SMP, or to its parent if it does not and the parent does. This is necessary because CAM does not currently support SMP-only nodes in the topology. Make SMP passthrough support conditional on __FreeBSD_version >= 900026. This will make it easier to MFC this change to the driver without MFCing the CAM changes as well. mps_user.c: Un-staticize mpi_init_sge() so we can use it for the SMP passthrough code. mpsvar.h: Add a uio and iovecs into struct mps_command for SMP passthrough commands. Add a cm_max_segs field to struct mps_command so that we can warn the user if busdma comes back with too many segments. Clear the cm_reply when a command gets freed. If it is not cleared, reply frames will eventually get freed into the pool multiple times and corrupt the pool. (This fix is from scottl.) Add a prototype for mpi_init_sge(). sys/param.h: Bump __FreeBSD_version to 900026 for the for the inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO CAM CCBs.
2010-11-30 22:39:46 +00:00
#if __FreeBSD_version >= 900026
static void mpssas_smpio_complete(struct mps_softc *sc, struct mps_command *cm);
static void mpssas_send_smpcmd(struct mpssas_softc *sassc, union ccb *ccb,
uint64_t sasaddr);
static void mpssas_action_smpio(struct mpssas_softc *sassc, union ccb *ccb);
#endif //FreeBSD_version >= 900026
static void mpssas_resetdev_complete(struct mps_softc *, struct mps_command *);
static int mpssas_send_abort(struct mps_softc *sc, struct mps_command *tm, struct mps_command *cm);
static int mpssas_send_reset(struct mps_softc *sc, struct mps_command *tm, uint8_t type);
static void mpssas_rescan(struct mpssas_softc *sassc, union ccb *ccb);
static void mpssas_rescan_done(struct cam_periph *periph, union ccb *done_ccb);
static void mpssas_scanner_thread(void *arg);
#if __FreeBSD_version >= 1000006
static void mpssas_async(void *callback_arg, uint32_t code,
struct cam_path *path, void *arg);
#else
static void mpssas_check_eedp(struct mpssas_softc *sassc);
static void mpssas_read_cap_done(struct cam_periph *periph, union ccb *done_ccb);
#endif
static int mpssas_send_portenable(struct mps_softc *sc);
static void mpssas_portenable_complete(struct mps_softc *sc,
struct mps_command *cm);
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
struct mpssas_target *
mpssas_find_target_by_handle(struct mpssas_softc *sassc, int start, uint16_t handle)
{
struct mpssas_target *target;
int i;
for (i = start; i < sassc->sc->facts->MaxTargets; i++) {
target = &sassc->targets[i];
if (target->handle == handle)
return (target);
}
return (NULL);
}
/* we need to freeze the simq during attach and diag reset, to avoid failing
* commands before device handles have been found by discovery. Since
* discovery involves reading config pages and possibly sending commands,
* discovery actions may continue even after we receive the end of discovery
* event, so refcount discovery actions instead of assuming we can unfreeze
* the simq when we get the event.
*/
void
mpssas_startup_increment(struct mpssas_softc *sassc)
{
if ((sassc->flags & MPSSAS_IN_STARTUP) != 0) {
if (sassc->startup_refcount++ == 0) {
/* just starting, freeze the simq */
mps_dprint(sassc->sc, MPS_INFO,
"%s freezing simq\n", __func__);
xpt_freeze_simq(sassc->sim, 1);
}
mps_dprint(sassc->sc, MPS_TRACE, "%s refcount %u\n", __func__,
sassc->startup_refcount);
}
}
void
mpssas_startup_decrement(struct mpssas_softc *sassc)
{
if ((sassc->flags & MPSSAS_IN_STARTUP) != 0) {
if (--sassc->startup_refcount == 0) {
/* finished all discovery-related actions, release
* the simq and rescan for the latest topology.
*/
mps_dprint(sassc->sc, MPS_INFO,
"%s releasing simq\n", __func__);
sassc->flags &= ~MPSSAS_IN_STARTUP;
xpt_release_simq(sassc->sim, 1);
mpssas_rescan_target(sassc->sc, NULL);
}
mps_dprint(sassc->sc, MPS_TRACE, "%s refcount %u\n", __func__,
sassc->startup_refcount);
}
}
/* LSI's firmware requires us to stop sending commands when we're doing task
* management, so refcount the TMs and keep the simq frozen when any are in
* use.
*/
struct mps_command *
mpssas_alloc_tm(struct mps_softc *sc)
{
struct mps_command *tm;
tm = mps_alloc_high_priority_command(sc);
if (tm != NULL) {
if (sc->sassc->tm_count++ == 0) {
mps_printf(sc, "%s freezing simq\n", __func__);
xpt_freeze_simq(sc->sassc->sim, 1);
}
mps_dprint(sc, MPS_TRACE, "%s tm_count %u\n", __func__,
sc->sassc->tm_count);
}
return tm;
}
void
mpssas_free_tm(struct mps_softc *sc, struct mps_command *tm)
{
if (tm == NULL)
return;
/* if there are no TMs in use, we can release the simq. We use our
* own refcount so that it's easier for a diag reset to cleanup and
* release the simq.
*/
if (--sc->sassc->tm_count == 0) {
mps_printf(sc, "%s releasing simq\n", __func__);
xpt_release_simq(sc->sassc->sim, 1);
}
mps_dprint(sc, MPS_TRACE, "%s tm_count %u\n", __func__,
sc->sassc->tm_count);
mps_free_high_priority_command(sc, tm);
}
void
mpssas_rescan_target(struct mps_softc *sc, struct mpssas_target *targ)
{
struct mpssas_softc *sassc = sc->sassc;
path_id_t pathid;
target_id_t targetid;
union ccb *ccb;
pathid = cam_sim_path(sassc->sim);
if (targ == NULL)
targetid = CAM_TARGET_WILDCARD;
else
targetid = targ - sassc->targets;
/*
* Allocate a CCB and schedule a rescan.
*/
ccb = xpt_alloc_ccb_nowait();
if (ccb == NULL) {
mps_dprint(sc, MPS_FAULT, "unable to alloc CCB for rescan\n");
return;
}
if (xpt_create_path(&ccb->ccb_h.path, xpt_periph, pathid,
targetid, CAM_LUN_WILDCARD) != CAM_REQ_CMP) {
mps_dprint(sc, MPS_FAULT, "unable to create path for rescan\n");
xpt_free_ccb(ccb);
return;
}
/* XXX Hardwired to scan the bus for now */
ccb->ccb_h.func_code = XPT_SCAN_BUS;
mps_dprint(sc, MPS_TRACE, "%s targetid %u\n", __func__, targetid);
mpssas_rescan(sassc, ccb);
}
static void
mpssas_log_command(struct mps_command *cm, const char *fmt, ...)
{
struct sbuf sb;
va_list ap;
char str[192];
char path_str[64];
if (cm == NULL)
return;
sbuf_new(&sb, str, sizeof(str), 0);
va_start(ap, fmt);
if (cm->cm_ccb != NULL) {
xpt_path_string(cm->cm_ccb->csio.ccb_h.path, path_str,
sizeof(path_str));
sbuf_cat(&sb, path_str);
if (cm->cm_ccb->ccb_h.func_code == XPT_SCSI_IO) {
scsi_command_string(&cm->cm_ccb->csio, &sb);
sbuf_printf(&sb, "length %d ",
cm->cm_ccb->csio.dxfer_len);
}
}
else {
sbuf_printf(&sb, "(noperiph:%s%d:%u:%u:%u): ",
cam_sim_name(cm->cm_sc->sassc->sim),
cam_sim_unit(cm->cm_sc->sassc->sim),
cam_sim_bus(cm->cm_sc->sassc->sim),
cm->cm_targ ? cm->cm_targ->tid : 0xFFFFFFFF,
cm->cm_lun);
}
sbuf_printf(&sb, "SMID %u ", cm->cm_desc.Default.SMID);
sbuf_vprintf(&sb, fmt, ap);
sbuf_finish(&sb);
printf("%s", sbuf_data(&sb));
va_end(ap);
}
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
static void
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
mpssas_remove_volume(struct mps_softc *sc, struct mps_command *tm)
{
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
MPI2_SCSI_TASK_MANAGE_REPLY *reply;
struct mpssas_target *targ;
uint16_t handle;
mps_dprint(sc, MPS_INFO, "%s\n", __func__);
reply = (MPI2_SCSI_TASK_MANAGE_REPLY *)tm->cm_reply;
handle = (uint16_t)(uintptr_t)tm->cm_complete_data;
targ = tm->cm_targ;
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
if (reply == NULL) {
/* XXX retry the remove after the diag reset completes? */
mps_printf(sc, "%s NULL reply reseting device 0x%04x\n",
__func__, handle);
mpssas_free_tm(sc, tm);
return;
}
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
if (reply->IOCStatus != MPI2_IOCSTATUS_SUCCESS) {
mps_printf(sc, "IOCStatus = 0x%x while resetting device 0x%x\n",
reply->IOCStatus, handle);
mpssas_free_tm(sc, tm);
return;
}
mps_printf(sc, "Reset aborted %u commands\n", reply->TerminationCount);
mps_free_reply(sc, tm->cm_reply_data);
tm->cm_reply = NULL; /* Ensures the the reply won't get re-freed */
mps_printf(sc, "clearing target %u handle 0x%04x\n", targ->tid, handle);
/*
* Don't clear target if remove fails because things will get confusing.
* Leave the devname and sasaddr intact so that we know to avoid reusing
* this target id if possible, and so we can assign the same target id
* to this device if it comes back in the future.
*/
if (reply->IOCStatus == MPI2_IOCSTATUS_SUCCESS) {
targ = tm->cm_targ;
targ->handle = 0x0;
targ->encl_handle = 0x0;
targ->encl_slot = 0x0;
targ->exp_dev_handle = 0x0;
targ->phy_num = 0x0;
targ->linkrate = 0x0;
targ->devinfo = 0x0;
targ->flags = 0x0;
}
mpssas_free_tm(sc, tm);
}
/*
* No Need to call "MPI2_SAS_OP_REMOVE_DEVICE" For Volume removal.
* Otherwise Volume Delete is same as Bare Drive Removal.
*/
void
mpssas_prepare_volume_remove(struct mpssas_softc *sassc, uint16_t handle)
{
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
struct mps_softc *sc;
struct mps_command *cm;
struct mpssas_target *targ = NULL;
mps_dprint(sassc->sc, MPS_INFO, "%s\n", __func__);
sc = sassc->sc;
#ifdef WD_SUPPORT
/*
* If this is a WD controller, determine if the disk should be exposed
* to the OS or not. If disk should be exposed, return from this
* function without doing anything.
*/
if (sc->WD_available && (sc->WD_hide_expose ==
MPS_WD_EXPOSE_ALWAYS)) {
return;
}
#endif //WD_SUPPORT
targ = mpssas_find_target_by_handle(sassc, 0, handle);
if (targ == NULL) {
/* FIXME: what is the action? */
/* We don't know about this device? */
printf("%s %d : invalid handle 0x%x \n", __func__,__LINE__, handle);
return;
}
targ->flags |= MPSSAS_TARGET_INREMOVAL;
cm = mpssas_alloc_tm(sc);
if (cm == NULL) {
mps_printf(sc, "%s: command alloc failure\n", __func__);
return;
}
mpssas_rescan_target(sc, targ);
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)cm->cm_req;
req->DevHandle = targ->handle;
req->Function = MPI2_FUNCTION_SCSI_TASK_MGMT;
req->TaskType = MPI2_SCSITASKMGMT_TASKTYPE_TARGET_RESET;
/* SAS Hard Link Reset / SATA Link Reset */
req->MsgFlags = MPI2_SCSITASKMGMT_MSGFLAGS_LINK_RESET;
cm->cm_targ = targ;
cm->cm_data = NULL;
cm->cm_desc.HighPriority.RequestFlags =
MPI2_REQ_DESCRIPT_FLAGS_HIGH_PRIORITY;
cm->cm_complete = mpssas_remove_volume;
cm->cm_complete_data = (void *)(uintptr_t)handle;
mps_map_command(sc, cm);
}
/*
* The MPT2 firmware performs debounce on the link to avoid transient link
* errors and false removals. When it does decide that link has been lost
* and a device need to go away, it expects that the host will perform a
* target reset and then an op remove. The reset has the side-effect of
* aborting any outstanding requests for the device, which is required for
* the op-remove to succeed. It's not clear if the host should check for
* the device coming back alive after the reset.
*/
void
mpssas_prepare_remove(struct mpssas_softc *sassc, uint16_t handle)
{
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
struct mps_softc *sc;
struct mps_command *cm;
struct mpssas_target *targ = NULL;
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
mps_dprint(sassc->sc, MPS_INFO, "%s\n", __func__);
/*
* If this is a WD controller, determine if the disk should be exposed
* to the OS or not. If disk should be exposed, return from this
* function without doing anything.
*/
sc = sassc->sc;
if ((sc->mps_flags & MPS_FLAGS_WD_AVAILABLE) && (sc->WD_hide_expose ==
MPS_WD_EXPOSE_ALWAYS)) {
return;
}
targ = mpssas_find_target_by_handle(sassc, 0, handle);
if (targ == NULL) {
/* FIXME: what is the action? */
/* We don't know about this device? */
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
printf("%s %d : invalid handle 0x%x \n", __func__,__LINE__, handle);
return;
}
targ->flags |= MPSSAS_TARGET_INREMOVAL;
cm = mpssas_alloc_tm(sc);
if (cm == NULL) {
mps_printf(sc, "%s: command alloc failure\n", __func__);
return;
}
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
mpssas_rescan_target(sc, targ);
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)cm->cm_req;
memset(req, 0, sizeof(*req));
req->DevHandle = targ->handle;
req->Function = MPI2_FUNCTION_SCSI_TASK_MGMT;
req->TaskType = MPI2_SCSITASKMGMT_TASKTYPE_TARGET_RESET;
/* SAS Hard Link Reset / SATA Link Reset */
req->MsgFlags = MPI2_SCSITASKMGMT_MSGFLAGS_LINK_RESET;
cm->cm_targ = targ;
cm->cm_data = NULL;
cm->cm_desc.HighPriority.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_HIGH_PRIORITY;
cm->cm_complete = mpssas_remove_device;
cm->cm_complete_data = (void *)(uintptr_t)handle;
mps_map_command(sc, cm);
}
static void
mpssas_remove_device(struct mps_softc *sc, struct mps_command *tm)
{
MPI2_SCSI_TASK_MANAGE_REPLY *reply;
MPI2_SAS_IOUNIT_CONTROL_REQUEST *req;
struct mpssas_target *targ;
struct mps_command *next_cm;
uint16_t handle;
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
mps_dprint(sc, MPS_INFO, "%s\n", __func__);
reply = (MPI2_SCSI_TASK_MANAGE_REPLY *)tm->cm_reply;
handle = (uint16_t)(uintptr_t)tm->cm_complete_data;
targ = tm->cm_targ;
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
/*
* Currently there should be no way we can hit this case. It only
* happens when we have a failure to allocate chain frames, and
* task management commands don't have S/G lists.
*/
if ((tm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
mps_printf(sc, "%s: cm_flags = %#x for remove of handle %#04x! "
"This should not happen!\n", __func__, tm->cm_flags,
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
handle);
mpssas_free_tm(sc, tm);
return;
}
if (reply == NULL) {
/* XXX retry the remove after the diag reset completes? */
mps_printf(sc, "%s NULL reply reseting device 0x%04x\n",
__func__, handle);
mpssas_free_tm(sc, tm);
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
return;
}
if (reply->IOCStatus != MPI2_IOCSTATUS_SUCCESS) {
mps_printf(sc, "IOCStatus = 0x%x while resetting device 0x%x\n",
reply->IOCStatus, handle);
mpssas_free_tm(sc, tm);
return;
}
mps_dprint(sc, MPS_INFO, "Reset aborted %u commands\n",
reply->TerminationCount);
mps_free_reply(sc, tm->cm_reply_data);
tm->cm_reply = NULL; /* Ensures the the reply won't get re-freed */
/* Reuse the existing command */
req = (MPI2_SAS_IOUNIT_CONTROL_REQUEST *)tm->cm_req;
memset(req, 0, sizeof(*req));
req->Function = MPI2_FUNCTION_SAS_IO_UNIT_CONTROL;
req->Operation = MPI2_SAS_OP_REMOVE_DEVICE;
req->DevHandle = handle;
tm->cm_data = NULL;
tm->cm_desc.Default.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_DEFAULT_TYPE;
tm->cm_complete = mpssas_remove_complete;
tm->cm_complete_data = (void *)(uintptr_t)handle;
mps_map_command(sc, tm);
mps_dprint(sc, MPS_INFO, "clearing target %u handle 0x%04x\n",
targ->tid, handle);
TAILQ_FOREACH_SAFE(tm, &targ->commands, cm_link, next_cm) {
union ccb *ccb;
mps_dprint(sc, MPS_INFO, "Completing missed command %p\n", tm);
ccb = tm->cm_complete_data;
ccb->ccb_h.status = CAM_DEV_NOT_THERE;
mpssas_scsiio_complete(sc, tm);
}
}
static void
mpssas_remove_complete(struct mps_softc *sc, struct mps_command *tm)
{
MPI2_SAS_IOUNIT_CONTROL_REPLY *reply;
uint16_t handle;
struct mpssas_target *targ;
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
mps_dprint(sc, MPS_INFO, "%s\n", __func__);
reply = (MPI2_SAS_IOUNIT_CONTROL_REPLY *)tm->cm_reply;
handle = (uint16_t)(uintptr_t)tm->cm_complete_data;
/*
* Currently there should be no way we can hit this case. It only
* happens when we have a failure to allocate chain frames, and
* task management commands don't have S/G lists.
*/
if ((tm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
mps_printf(sc, "%s: cm_flags = %#x for remove of handle %#04x! "
"This should not happen!\n", __func__, tm->cm_flags,
handle);
mpssas_free_tm(sc, tm);
return;
}
if (reply == NULL) {
/* most likely a chip reset */
mps_printf(sc, "%s NULL reply removing device 0x%04x\n",
__func__, handle);
mpssas_free_tm(sc, tm);
return;
}
mps_printf(sc, "%s on handle 0x%04x, IOCStatus= 0x%x\n", __func__,
handle, reply->IOCStatus);
/*
* Don't clear target if remove fails because things will get confusing.
* Leave the devname and sasaddr intact so that we know to avoid reusing
* this target id if possible, and so we can assign the same target id
* to this device if it comes back in the future.
*/
if (reply->IOCStatus == MPI2_IOCSTATUS_SUCCESS) {
targ = tm->cm_targ;
targ->handle = 0x0;
targ->encl_handle = 0x0;
targ->encl_slot = 0x0;
targ->exp_dev_handle = 0x0;
targ->phy_num = 0x0;
targ->linkrate = 0x0;
targ->devinfo = 0x0;
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
targ->flags = 0x0;
}
mpssas_free_tm(sc, tm);
}
static int
mpssas_register_events(struct mps_softc *sc)
{
uint8_t events[16];
bzero(events, 16);
setbit(events, MPI2_EVENT_SAS_DEVICE_STATUS_CHANGE);
setbit(events, MPI2_EVENT_SAS_DISCOVERY);
setbit(events, MPI2_EVENT_SAS_BROADCAST_PRIMITIVE);
setbit(events, MPI2_EVENT_SAS_INIT_DEVICE_STATUS_CHANGE);
setbit(events, MPI2_EVENT_SAS_INIT_TABLE_OVERFLOW);
setbit(events, MPI2_EVENT_SAS_TOPOLOGY_CHANGE_LIST);
setbit(events, MPI2_EVENT_SAS_ENCL_DEVICE_STATUS_CHANGE);
setbit(events, MPI2_EVENT_IR_CONFIGURATION_CHANGE_LIST);
setbit(events, MPI2_EVENT_IR_VOLUME);
setbit(events, MPI2_EVENT_IR_PHYSICAL_DISK);
setbit(events, MPI2_EVENT_IR_OPERATION_STATUS);
setbit(events, MPI2_EVENT_LOG_ENTRY_ADDED);
mps_register_events(sc, events, mpssas_evt_handler, NULL,
&sc->sassc->mpssas_eh);
return (0);
}
int
mps_attach_sas(struct mps_softc *sc)
{
struct mpssas_softc *sassc;
#if __FreeBSD_version >= 1000006
cam_status status;
#endif
int unit, error = 0;
mps_dprint(sc, MPS_TRACE, "%s\n", __func__);
sassc = malloc(sizeof(struct mpssas_softc), M_MPT2, M_WAITOK|M_ZERO);
sassc->targets = malloc(sizeof(struct mpssas_target) *
sc->facts->MaxTargets, M_MPT2, M_WAITOK|M_ZERO);
sc->sassc = sassc;
sassc->sc = sc;
if ((sassc->devq = cam_simq_alloc(sc->num_reqs)) == NULL) {
mps_dprint(sc, MPS_FAULT, "Cannot allocate SIMQ\n");
error = ENOMEM;
goto out;
}
unit = device_get_unit(sc->mps_dev);
sassc->sim = cam_sim_alloc(mpssas_action, mpssas_poll, "mps", sassc,
unit, &sc->mps_mtx, sc->num_reqs, sc->num_reqs, sassc->devq);
if (sassc->sim == NULL) {
mps_dprint(sc, MPS_FAULT, "Cannot allocate SIM\n");
error = EINVAL;
goto out;
}
TAILQ_INIT(&sassc->ev_queue);
/* Initialize taskqueue for Event Handling */
TASK_INIT(&sassc->ev_task, 0, mpssas_firmware_event_work, sc);
sassc->ev_tq = taskqueue_create("mps_taskq", M_NOWAIT | M_ZERO,
taskqueue_thread_enqueue, &sassc->ev_tq);
/* Run the task queue with lowest priority */
taskqueue_start_threads(&sassc->ev_tq, 1, 255, "%s taskq",
device_get_nameunit(sc->mps_dev));
TAILQ_INIT(&sassc->ccb_scanq);
error = mps_kproc_create(mpssas_scanner_thread, sassc,
&sassc->rescan_thread, 0, 0, "mps_scan%d", unit);
if (error) {
mps_printf(sc, "Error %d starting rescan thread\n", error);
goto out;
}
mps_lock(sc);
sassc->flags |= MPSSAS_SCANTHREAD;
/*
* XXX There should be a bus for every port on the adapter, but since
* we're just going to fake the topology for now, we'll pretend that
* everything is just a target on a single bus.
*/
if ((error = xpt_bus_register(sassc->sim, sc->mps_dev, 0)) != 0) {
mps_dprint(sc, MPS_FAULT, "Error %d registering SCSI bus\n",
error);
mps_unlock(sc);
goto out;
}
/*
* Assume that discovery events will start right away. Freezing
* the simq will prevent the CAM boottime scanner from running
* before discovery is complete.
*/
sassc->flags |= MPSSAS_IN_STARTUP | MPSSAS_IN_DISCOVERY;
xpt_freeze_simq(sassc->sim, 1);
sc->sassc->startup_refcount = 0;
callout_init(&sassc->discovery_callout, 1 /*mpsafe*/);
sassc->discovery_timeouts = 0;
sassc->tm_count = 0;
#if __FreeBSD_version >= 1000006
status = xpt_register_async(AC_ADVINFO_CHANGED, mpssas_async, sc, NULL);
if (status != CAM_REQ_CMP) {
mps_printf(sc, "Error %#x registering async handler for "
"AC_ADVINFO_CHANGED events\n", status);
}
#endif
mps_unlock(sc);
mpssas_register_events(sc);
out:
if (error)
mps_detach_sas(sc);
return (error);
}
int
mps_detach_sas(struct mps_softc *sc)
{
struct mpssas_softc *sassc;
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
mps_dprint(sc, MPS_INFO, "%s\n", __func__);
if (sc->sassc == NULL)
return (0);
sassc = sc->sassc;
mps_deregister_events(sc, sassc->mpssas_eh);
/*
* Drain and free the event handling taskqueue with the lock
* unheld so that any parallel processing tasks drain properly
* without deadlocking.
*/
if (sassc->ev_tq != NULL)
taskqueue_free(sassc->ev_tq);
/* Make sure CAM doesn't wedge if we had to bail out early. */
mps_lock(sc);
/* Deregister our async handler */
#if __FreeBSD_version >= 1000006
xpt_register_async(0, mpssas_async, sc, NULL);
#endif
if (sassc->flags & MPSSAS_IN_STARTUP)
xpt_release_simq(sassc->sim, 1);
if (sassc->sim != NULL) {
xpt_bus_deregister(cam_sim_path(sassc->sim));
cam_sim_free(sassc->sim, FALSE);
}
if (sassc->flags & MPSSAS_SCANTHREAD) {
sassc->flags |= MPSSAS_SHUTDOWN;
wakeup(&sassc->ccb_scanq);
if (sassc->flags & MPSSAS_SCANTHREAD) {
msleep(&sassc->flags, &sc->mps_mtx, PRIBIO,
"mps_shutdown", 30 * hz);
}
}
mps_unlock(sc);
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
mps_dprint(sc, MPS_INFO, "%s:%d\n", __func__,__LINE__);
if (sassc->devq != NULL)
cam_simq_free(sassc->devq);
free(sassc->targets, M_MPT2);
free(sassc, M_MPT2);
sc->sassc = NULL;
return (0);
}
void
mpssas_discovery_end(struct mpssas_softc *sassc)
{
struct mps_softc *sc = sassc->sc;
mps_dprint(sc, MPS_TRACE, "%s\n", __func__);
if (sassc->flags & MPSSAS_DISCOVERY_TIMEOUT_PENDING)
callout_stop(&sassc->discovery_callout);
}
static void
mpssas_discovery_timeout(void *data)
{
struct mpssas_softc *sassc = data;
struct mps_softc *sc;
sc = sassc->sc;
mps_dprint(sc, MPS_TRACE, "%s\n", __func__);
mps_lock(sc);
mps_printf(sc,
"Timeout waiting for discovery, interrupts may not be working!\n");
sassc->flags &= ~MPSSAS_DISCOVERY_TIMEOUT_PENDING;
/* Poll the hardware for events in case interrupts aren't working */
mps_intr_locked(sc);
mps_printf(sassc->sc,
"Finished polling after discovery timeout at %d\n", ticks);
if ((sassc->flags & MPSSAS_IN_DISCOVERY) == 0) {
mpssas_discovery_end(sassc);
} else {
if (sassc->discovery_timeouts < MPSSAS_MAX_DISCOVERY_TIMEOUTS) {
sassc->flags |= MPSSAS_DISCOVERY_TIMEOUT_PENDING;
callout_reset(&sassc->discovery_callout,
MPSSAS_DISCOVERY_TIMEOUT * hz,
mpssas_discovery_timeout, sassc);
sassc->discovery_timeouts++;
} else {
mps_dprint(sassc->sc, MPS_FAULT,
"Discovery timed out, continuing.\n");
sassc->flags &= ~MPSSAS_IN_DISCOVERY;
mpssas_discovery_end(sassc);
}
}
mps_unlock(sc);
}
static void
mpssas_action(struct cam_sim *sim, union ccb *ccb)
{
struct mpssas_softc *sassc;
sassc = cam_sim_softc(sim);
mps_dprint(sassc->sc, MPS_TRACE, "%s func 0x%x\n", __func__,
ccb->ccb_h.func_code);
mtx_assert(&sassc->sc->mps_mtx, MA_OWNED);
switch (ccb->ccb_h.func_code) {
case XPT_PATH_INQ:
{
struct ccb_pathinq *cpi = &ccb->cpi;
cpi->version_num = 1;
cpi->hba_inquiry = PI_SDTR_ABLE|PI_TAG_ABLE|PI_WIDE_16;
cpi->target_sprt = 0;
cpi->hba_misc = PIM_NOBUSRESET;
cpi->hba_eng_cnt = 0;
cpi->max_target = sassc->sc->facts->MaxTargets - 1;
cpi->max_lun = 8;
cpi->initiator_id = 255;
strncpy(cpi->sim_vid, "FreeBSD", SIM_IDLEN);
strncpy(cpi->hba_vid, "LSILogic", HBA_IDLEN);
strncpy(cpi->dev_name, cam_sim_name(sim), DEV_IDLEN);
cpi->unit_number = cam_sim_unit(sim);
cpi->bus_id = cam_sim_bus(sim);
cpi->base_transfer_speed = 150000;
cpi->transport = XPORT_SAS;
cpi->transport_version = 0;
cpi->protocol = PROTO_SCSI;
cpi->protocol_version = SCSI_REV_SPC;
#if __FreeBSD_version >= 800001
/*
* XXX KDM where does this number come from?
*/
cpi->maxio = 256 * 1024;
#endif
cpi->ccb_h.status = CAM_REQ_CMP;
break;
}
case XPT_GET_TRAN_SETTINGS:
{
struct ccb_trans_settings *cts;
struct ccb_trans_settings_sas *sas;
struct ccb_trans_settings_scsi *scsi;
struct mpssas_target *targ;
cts = &ccb->cts;
sas = &cts->xport_specific.sas;
scsi = &cts->proto_specific.scsi;
targ = &sassc->targets[cts->ccb_h.target_id];
if (targ->handle == 0x0) {
cts->ccb_h.status = CAM_TID_INVALID;
break;
}
cts->protocol_version = SCSI_REV_SPC2;
cts->transport = XPORT_SAS;
cts->transport_version = 0;
sas->valid = CTS_SAS_VALID_SPEED;
switch (targ->linkrate) {
case 0x08:
sas->bitrate = 150000;
break;
case 0x09:
sas->bitrate = 300000;
break;
case 0x0a:
sas->bitrate = 600000;
break;
default:
sas->valid = 0;
}
cts->protocol = PROTO_SCSI;
scsi->valid = CTS_SCSI_VALID_TQ;
scsi->flags = CTS_SCSI_FLAGS_TAG_ENB;
cts->ccb_h.status = CAM_REQ_CMP;
break;
}
case XPT_CALC_GEOMETRY:
cam_calc_geometry(&ccb->ccg, /*extended*/1);
ccb->ccb_h.status = CAM_REQ_CMP;
break;
case XPT_RESET_DEV:
mps_printf(sassc->sc, "mpssas_action XPT_RESET_DEV\n");
mpssas_action_resetdev(sassc, ccb);
return;
case XPT_RESET_BUS:
case XPT_ABORT:
case XPT_TERM_IO:
mps_printf(sassc->sc, "mpssas_action faking success for "
"abort or reset\n");
ccb->ccb_h.status = CAM_REQ_CMP;
break;
case XPT_SCSI_IO:
mpssas_action_scsiio(sassc, ccb);
return;
Add Serial Management Protocol (SMP) passthrough support to CAM. This includes support in the kernel, camcontrol(8), libcam and the mps(4) driver for SMP passthrough. The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00 to determine supported pages, and will now fetch page 0x83 in addition to page 0x80 if supported. Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is intended for SMP requests and responses. The ADVINFO is currently used to fetch cached VPD page 0x83 data from the transport layer, but is intended to be extensible to fetch other types of device-specific data. SMP-only devices are not currently represented in the CAM topology, and so the current semantics are that the SIM will route SMP CCBs to either the addressed device, if it contains an SMP target, or its parent, if it contains an SMP target. (This is noted in cam_ccb.h, since it will change later once we have the ability to have SMP-only devices in CAM's topology.) smp_all.c, smp_all.h: New helper routines for SMP. This includes SMP request building routines, response parsing routines, error decoding routines, and structure definitions for a number of SMP commands. libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality is available to userland applications. camcontrol.8, camcontrol.c: Add smp passthrough support to camcontrol. Several new subcommands are now available: 'smpcmd' functions much like 'cmd', except that it allows the user to send generic SMP commands. 'smprg' sends the SMP report general command, and displays the decoded output. It will automatically fetch extended output if it is available. 'smppc' sends the SMP phy control command, with any number of potential options. Among other things, this allows the user to reset a phy on a SAS expander, or disable a phy on an expander. 'smpmaninfo' sends the SMP report manufacturer information and displays the decoded output. 'smpphylist' displays a list of phys on an expander, and the CAM devices attached to those phys, if any. cam.h, cam.c: Add a status value for SMP errors (CAM_SMP_STATUS_ERROR). Add a missing description for CAM_SCSI_IT_NEXUS_LOST. Add support for SMP commands to cam_error_string(). cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP commands are by nature bi-directional, and we may need to support bi-directional SCSI commands later. Add the XPT_SMP_IO CCB. Since SMP commands are bi-directional, there are pointers for both the request and response. Add a fill routine for SMP CCBs. Add the XPT_GDEV_ADVINFO CCB. This is currently used to fetch cached page 0x83 data from the transport later, but is extensible to fetch many other types of data. cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. cam_xpt.c: Add support for executing XPT_SMP_IO CCBs. cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct cam_ed. scsi_all.c: Add scsi_get_sas_addr(), a function that parses VPD page 0x83 data and pulls out a SAS address. scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a prototype for scsi_get_sas_addr(). scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. scsi_xpt.c: In the SCSI probe code, first ask the device for VPD page 0x00. If any VPD pages are supported, that page is required to be implemented. Based on the response, we may probe for the serial number (page 0x80) or device id (page 0x83). Add support for the XPT_GDEV_ADVINFO CCB. sys/conf/files: Add smp_all.c. mps.c: Add support for passing in a uio in mps_map_command(), so we can map a S/G list at once. Add support for SMP passthrough commands in mps_data_cb(). SMP is a special case, because the first buffer in the S/G list is outbound and the second buffer is inbound. Add support for warning the user if the busdma code comes back with more buffers than will work for the command. This will, for example, help the user determine why an SMP command failed if busdma comes back with three buffers. mps_pci.c: Add sys/uio.h. mps_sas.c: Add the SAS address and the parent handle to the list of fields we pull from device page 0 and cache in struct mpssas_target. These are needed for SMP passthrough. Add support for the XPT_SMP_IO CCB. For now, this CCB is routed to the addressed device if it supports SMP, or to its parent if it does not and the parent does. This is necessary because CAM does not currently support SMP-only nodes in the topology. Make SMP passthrough support conditional on __FreeBSD_version >= 900026. This will make it easier to MFC this change to the driver without MFCing the CAM changes as well. mps_user.c: Un-staticize mpi_init_sge() so we can use it for the SMP passthrough code. mpsvar.h: Add a uio and iovecs into struct mps_command for SMP passthrough commands. Add a cm_max_segs field to struct mps_command so that we can warn the user if busdma comes back with too many segments. Clear the cm_reply when a command gets freed. If it is not cleared, reply frames will eventually get freed into the pool multiple times and corrupt the pool. (This fix is from scottl.) Add a prototype for mpi_init_sge(). sys/param.h: Bump __FreeBSD_version to 900026 for the for the inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO CAM CCBs.
2010-11-30 22:39:46 +00:00
#if __FreeBSD_version >= 900026
case XPT_SMP_IO:
mpssas_action_smpio(sassc, ccb);
return;
#endif
default:
ccb->ccb_h.status = CAM_FUNC_NOTAVAIL;
break;
}
xpt_done(ccb);
}
static void
mpssas_announce_reset(struct mps_softc *sc, uint32_t ac_code,
target_id_t target_id, lun_id_t lun_id)
{
path_id_t path_id = cam_sim_path(sc->sassc->sim);
struct cam_path *path;
mps_printf(sc, "%s code %x target %d lun %d\n", __func__,
ac_code, target_id, lun_id);
if (xpt_create_path(&path, NULL,
path_id, target_id, lun_id) != CAM_REQ_CMP) {
mps_printf(sc, "unable to create path for reset "
"notification\n");
return;
}
xpt_async(ac_code, path, NULL);
xpt_free_path(path);
}
static void
mpssas_complete_all_commands(struct mps_softc *sc)
{
struct mps_command *cm;
int i;
int completed;
mps_printf(sc, "%s\n", __func__);
mtx_assert(&sc->mps_mtx, MA_OWNED);
/* complete all commands with a NULL reply */
for (i = 1; i < sc->num_reqs; i++) {
cm = &sc->commands[i];
cm->cm_reply = NULL;
completed = 0;
if (cm->cm_flags & MPS_CM_FLAGS_POLLED)
cm->cm_flags |= MPS_CM_FLAGS_COMPLETE;
if (cm->cm_complete != NULL) {
mpssas_log_command(cm,
"completing cm %p state %x ccb %p for diag reset\n",
cm, cm->cm_state, cm->cm_ccb);
cm->cm_complete(sc, cm);
completed = 1;
}
if (cm->cm_flags & MPS_CM_FLAGS_WAKEUP) {
mpssas_log_command(cm,
"waking up cm %p state %x ccb %p for diag reset\n",
cm, cm->cm_state, cm->cm_ccb);
wakeup(cm);
completed = 1;
}
if ((completed == 0) && (cm->cm_state != MPS_CM_STATE_FREE)) {
/* this should never happen, but if it does, log */
mpssas_log_command(cm,
"cm %p state %x flags 0x%x ccb %p during diag "
"reset\n", cm, cm->cm_state, cm->cm_flags,
cm->cm_ccb);
}
}
}
void
mpssas_handle_reinit(struct mps_softc *sc)
{
int i;
/* Go back into startup mode and freeze the simq, so that CAM
* doesn't send any commands until after we've rediscovered all
* targets and found the proper device handles for them.
*
* After the reset, portenable will trigger discovery, and after all
* discovery-related activities have finished, the simq will be
* released.
*/
mps_printf(sc, "%s startup\n", __func__);
sc->sassc->flags |= MPSSAS_IN_STARTUP;
sc->sassc->flags |= MPSSAS_IN_DISCOVERY;
xpt_freeze_simq(sc->sassc->sim, 1);
/* notify CAM of a bus reset */
mpssas_announce_reset(sc, AC_BUS_RESET, CAM_TARGET_WILDCARD,
CAM_LUN_WILDCARD);
/* complete and cleanup after all outstanding commands */
mpssas_complete_all_commands(sc);
mps_printf(sc, "%s startup %u tm %u after command completion\n",
__func__, sc->sassc->startup_refcount, sc->sassc->tm_count);
/*
* The simq was explicitly frozen above, so set the refcount to 0.
* The simq will be explicitly released after port enable completes.
*/
sc->sassc->startup_refcount = 0;
/* zero all the target handles, since they may change after the
* reset, and we have to rediscover all the targets and use the new
* handles.
*/
for (i = 0; i < sc->facts->MaxTargets; i++) {
if (sc->sassc->targets[i].outstanding != 0)
mps_printf(sc, "target %u outstanding %u\n",
i, sc->sassc->targets[i].outstanding);
sc->sassc->targets[i].handle = 0x0;
sc->sassc->targets[i].exp_dev_handle = 0x0;
sc->sassc->targets[i].outstanding = 0;
sc->sassc->targets[i].flags = MPSSAS_TARGET_INDIAGRESET;
}
}
static void
mpssas_tm_timeout(void *data)
{
struct mps_command *tm = data;
struct mps_softc *sc = tm->cm_sc;
mtx_assert(&sc->mps_mtx, MA_OWNED);
mpssas_log_command(tm, "task mgmt %p timed out\n", tm);
mps_reinit(sc);
}
static void
mpssas_logical_unit_reset_complete(struct mps_softc *sc, struct mps_command *tm)
{
MPI2_SCSI_TASK_MANAGE_REPLY *reply;
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
unsigned int cm_count = 0;
struct mps_command *cm;
struct mpssas_target *targ;
callout_stop(&tm->cm_callout);
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)tm->cm_req;
reply = (MPI2_SCSI_TASK_MANAGE_REPLY *)tm->cm_reply;
targ = tm->cm_targ;
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
/*
* Currently there should be no way we can hit this case. It only
* happens when we have a failure to allocate chain frames, and
* task management commands don't have S/G lists.
*/
if ((tm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
mps_printf(sc, "%s: cm_flags = %#x for LUN reset! "
"This should not happen!\n", __func__, tm->cm_flags);
mpssas_free_tm(sc, tm);
return;
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
}
if (reply == NULL) {
mpssas_log_command(tm, "NULL reset reply for tm %p\n", tm);
if ((sc->mps_flags & MPS_FLAGS_DIAGRESET) != 0) {
/* this completion was due to a reset, just cleanup */
targ->flags &= ~MPSSAS_TARGET_INRESET;
targ->tm = NULL;
mpssas_free_tm(sc, tm);
}
else {
/* we should have gotten a reply. */
mps_reinit(sc);
}
return;
}
mpssas_log_command(tm,
"logical unit reset status 0x%x code 0x%x count %u\n",
reply->IOCStatus, reply->ResponseCode,
reply->TerminationCount);
/* See if there are any outstanding commands for this LUN.
* This could be made more efficient by using a per-LU data
* structure of some sort.
*/
TAILQ_FOREACH(cm, &targ->commands, cm_link) {
if (cm->cm_lun == tm->cm_lun)
cm_count++;
}
if (cm_count == 0) {
mpssas_log_command(tm,
"logical unit %u finished recovery after reset\n",
tm->cm_lun, tm);
mpssas_announce_reset(sc, AC_SENT_BDR, tm->cm_targ->tid,
tm->cm_lun);
/* we've finished recovery for this logical unit. check and
* see if some other logical unit has a timedout command
* that needs to be processed.
*/
cm = TAILQ_FIRST(&targ->timedout_commands);
if (cm) {
mpssas_send_abort(sc, tm, cm);
}
else {
targ->tm = NULL;
mpssas_free_tm(sc, tm);
}
}
else {
/* if we still have commands for this LUN, the reset
* effectively failed, regardless of the status reported.
* Escalate to a target reset.
*/
mpssas_log_command(tm,
"logical unit reset complete for tm %p, but still have %u command(s)\n",
tm, cm_count);
mpssas_send_reset(sc, tm,
MPI2_SCSITASKMGMT_TASKTYPE_TARGET_RESET);
}
}
static void
mpssas_target_reset_complete(struct mps_softc *sc, struct mps_command *tm)
{
MPI2_SCSI_TASK_MANAGE_REPLY *reply;
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
struct mpssas_target *targ;
callout_stop(&tm->cm_callout);
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)tm->cm_req;
reply = (MPI2_SCSI_TASK_MANAGE_REPLY *)tm->cm_reply;
targ = tm->cm_targ;
/*
* Currently there should be no way we can hit this case. It only
* happens when we have a failure to allocate chain frames, and
* task management commands don't have S/G lists.
*/
if ((tm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
mps_printf(sc, "%s: cm_flags = %#x for target reset! "
"This should not happen!\n", __func__, tm->cm_flags);
mpssas_free_tm(sc, tm);
return;
}
if (reply == NULL) {
mpssas_log_command(tm, "NULL reset reply for tm %p\n", tm);
if ((sc->mps_flags & MPS_FLAGS_DIAGRESET) != 0) {
/* this completion was due to a reset, just cleanup */
targ->flags &= ~MPSSAS_TARGET_INRESET;
targ->tm = NULL;
mpssas_free_tm(sc, tm);
}
else {
/* we should have gotten a reply. */
mps_reinit(sc);
}
return;
}
mpssas_log_command(tm,
"target reset status 0x%x code 0x%x count %u\n",
reply->IOCStatus, reply->ResponseCode,
reply->TerminationCount);
targ->flags &= ~MPSSAS_TARGET_INRESET;
if (targ->outstanding == 0) {
/* we've finished recovery for this target and all
* of its logical units.
*/
mpssas_log_command(tm,
"recovery finished after target reset\n");
mpssas_announce_reset(sc, AC_SENT_BDR, tm->cm_targ->tid,
CAM_LUN_WILDCARD);
targ->tm = NULL;
mpssas_free_tm(sc, tm);
}
else {
/* after a target reset, if this target still has
* outstanding commands, the reset effectively failed,
* regardless of the status reported. escalate.
*/
mpssas_log_command(tm,
"target reset complete for tm %p, but still have %u command(s)\n",
tm, targ->outstanding);
mps_reinit(sc);
}
}
#define MPS_RESET_TIMEOUT 30
static int
mpssas_send_reset(struct mps_softc *sc, struct mps_command *tm, uint8_t type)
{
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
struct mpssas_target *target;
int err;
target = tm->cm_targ;
if (target->handle == 0) {
mps_printf(sc, "%s null devhandle for target_id %d\n",
__func__, target->tid);
return -1;
}
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)tm->cm_req;
req->DevHandle = target->handle;
req->Function = MPI2_FUNCTION_SCSI_TASK_MGMT;
req->TaskType = type;
if (type == MPI2_SCSITASKMGMT_TASKTYPE_LOGICAL_UNIT_RESET) {
/* XXX Need to handle invalid LUNs */
MPS_SET_LUN(req->LUN, tm->cm_lun);
tm->cm_targ->logical_unit_resets++;
mpssas_log_command(tm, "sending logical unit reset\n");
tm->cm_complete = mpssas_logical_unit_reset_complete;
}
else if (type == MPI2_SCSITASKMGMT_TASKTYPE_TARGET_RESET) {
/* Target reset method = SAS Hard Link Reset / SATA Link Reset */
req->MsgFlags = MPI2_SCSITASKMGMT_MSGFLAGS_LINK_RESET;
tm->cm_targ->target_resets++;
tm->cm_targ->flags |= MPSSAS_TARGET_INRESET;
mpssas_log_command(tm, "sending target reset\n");
tm->cm_complete = mpssas_target_reset_complete;
}
else {
mps_printf(sc, "unexpected reset type 0x%x\n", type);
return -1;
}
tm->cm_data = NULL;
tm->cm_desc.HighPriority.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_HIGH_PRIORITY;
tm->cm_complete_data = (void *)tm;
callout_reset(&tm->cm_callout, MPS_RESET_TIMEOUT * hz,
mpssas_tm_timeout, tm);
err = mps_map_command(sc, tm);
if (err)
mpssas_log_command(tm,
"error %d sending reset type %u\n",
err, type);
return err;
}
static void
mpssas_abort_complete(struct mps_softc *sc, struct mps_command *tm)
{
struct mps_command *cm;
MPI2_SCSI_TASK_MANAGE_REPLY *reply;
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
struct mpssas_target *targ;
callout_stop(&tm->cm_callout);
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)tm->cm_req;
reply = (MPI2_SCSI_TASK_MANAGE_REPLY *)tm->cm_reply;
targ = tm->cm_targ;
/*
* Currently there should be no way we can hit this case. It only
* happens when we have a failure to allocate chain frames, and
* task management commands don't have S/G lists.
*/
if ((tm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
mpssas_log_command(tm,
"cm_flags = %#x for abort %p TaskMID %u!\n",
tm->cm_flags, tm, req->TaskMID);
mpssas_free_tm(sc, tm);
return;
}
if (reply == NULL) {
mpssas_log_command(tm,
"NULL abort reply for tm %p TaskMID %u\n",
tm, req->TaskMID);
if ((sc->mps_flags & MPS_FLAGS_DIAGRESET) != 0) {
/* this completion was due to a reset, just cleanup */
targ->tm = NULL;
mpssas_free_tm(sc, tm);
}
else {
/* we should have gotten a reply. */
mps_reinit(sc);
}
return;
}
mpssas_log_command(tm,
"abort TaskMID %u status 0x%x code 0x%x count %u\n",
req->TaskMID,
reply->IOCStatus, reply->ResponseCode,
reply->TerminationCount);
cm = TAILQ_FIRST(&tm->cm_targ->timedout_commands);
if (cm == NULL) {
/* if there are no more timedout commands, we're done with
* error recovery for this target.
*/
mpssas_log_command(tm,
"finished recovery after aborting TaskMID %u\n",
req->TaskMID);
targ->tm = NULL;
mpssas_free_tm(sc, tm);
}
else if (req->TaskMID != cm->cm_desc.Default.SMID) {
/* abort success, but we have more timedout commands to abort */
mpssas_log_command(tm,
"continuing recovery after aborting TaskMID %u\n",
req->TaskMID);
mpssas_send_abort(sc, tm, cm);
}
else {
/* we didn't get a command completion, so the abort
* failed as far as we're concerned. escalate.
*/
mpssas_log_command(tm,
"abort failed for TaskMID %u tm %p\n",
req->TaskMID, tm);
mpssas_send_reset(sc, tm,
MPI2_SCSITASKMGMT_TASKTYPE_LOGICAL_UNIT_RESET);
}
}
#define MPS_ABORT_TIMEOUT 5
static int
mpssas_send_abort(struct mps_softc *sc, struct mps_command *tm, struct mps_command *cm)
{
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
struct mpssas_target *targ;
int err;
targ = cm->cm_targ;
if (targ->handle == 0) {
mps_printf(sc, "%s null devhandle for target_id %d\n",
__func__, cm->cm_ccb->ccb_h.target_id);
return -1;
}
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)tm->cm_req;
req->DevHandle = targ->handle;
req->Function = MPI2_FUNCTION_SCSI_TASK_MGMT;
req->TaskType = MPI2_SCSITASKMGMT_TASKTYPE_ABORT_TASK;
/* XXX Need to handle invalid LUNs */
MPS_SET_LUN(req->LUN, cm->cm_ccb->ccb_h.target_lun);
req->TaskMID = cm->cm_desc.Default.SMID;
tm->cm_data = NULL;
tm->cm_desc.HighPriority.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_HIGH_PRIORITY;
tm->cm_complete = mpssas_abort_complete;
tm->cm_complete_data = (void *)tm;
tm->cm_targ = cm->cm_targ;
tm->cm_lun = cm->cm_lun;
callout_reset(&tm->cm_callout, MPS_ABORT_TIMEOUT * hz,
mpssas_tm_timeout, tm);
targ->aborts++;
err = mps_map_command(sc, tm);
if (err)
mpssas_log_command(tm,
"error %d sending abort for cm %p SMID %u\n",
err, cm, req->TaskMID);
return err;
}
static void
mpssas_scsiio_timeout(void *data)
{
struct mps_softc *sc;
struct mps_command *cm;
struct mpssas_target *targ;
cm = (struct mps_command *)data;
sc = cm->cm_sc;
mtx_assert(&sc->mps_mtx, MA_OWNED);
mps_printf(sc, "%s checking sc %p cm %p\n", __func__, sc, cm);
/*
* Run the interrupt handler to make sure it's not pending. This
* isn't perfect because the command could have already completed
* and been re-used, though this is unlikely.
*/
mps_intr_locked(sc);
if (cm->cm_state == MPS_CM_STATE_FREE) {
mps_printf(sc, "SCSI command %p sc %p almost timed out\n", cm, sc);
return;
}
if (cm->cm_ccb == NULL) {
mps_printf(sc, "command timeout with NULL ccb\n");
return;
}
mpssas_log_command(cm, "command timeout cm %p ccb %p\n",
cm, cm->cm_ccb);
targ = cm->cm_targ;
targ->timeouts++;
/* XXX first, check the firmware state, to see if it's still
* operational. if not, do a diag reset.
*/
cm->cm_ccb->ccb_h.status = CAM_CMD_TIMEOUT;
cm->cm_state = MPS_CM_STATE_TIMEDOUT;
TAILQ_INSERT_TAIL(&targ->timedout_commands, cm, cm_recovery);
if (targ->tm != NULL) {
/* target already in recovery, just queue up another
* timedout command to be processed later.
*/
mps_printf(sc, "queued timedout cm %p for processing by tm %p\n",
cm, targ->tm);
}
else if ((targ->tm = mpssas_alloc_tm(sc)) != NULL) {
mps_printf(sc, "timedout cm %p allocated tm %p\n",
cm, targ->tm);
/* start recovery by aborting the first timedout command */
mpssas_send_abort(sc, targ->tm, cm);
}
else {
/* XXX queue this target up for recovery once a TM becomes
* available. The firmware only has a limited number of
* HighPriority credits for the high priority requests used
* for task management, and we ran out.
*
* Isilon: don't worry about this for now, since we have
* more credits than disks in an enclosure, and limit
* ourselves to one TM per target for recovery.
*/
mps_printf(sc, "timedout cm %p failed to allocate a tm\n",
cm);
}
}
static void
mpssas_action_scsiio(struct mpssas_softc *sassc, union ccb *ccb)
{
MPI2_SCSI_IO_REQUEST *req;
struct ccb_scsiio *csio;
struct mps_softc *sc;
struct mpssas_target *targ;
struct mpssas_lun *lun;
struct mps_command *cm;
uint8_t i, lba_byte, *ref_tag_addr;
uint16_t eedp_flags;
sc = sassc->sc;
mtx_assert(&sc->mps_mtx, MA_OWNED);
csio = &ccb->csio;
targ = &sassc->targets[csio->ccb_h.target_id];
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
mps_dprint(sc, MPS_TRACE, "%s ccb %p target flag %x\n", __func__, ccb, targ->flags);
if (targ->handle == 0x0) {
mps_dprint(sc, MPS_TRACE, "%s NULL handle for target %u\n",
__func__, csio->ccb_h.target_id);
csio->ccb_h.status = CAM_TID_INVALID;
xpt_done(ccb);
return;
}
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
if (targ->flags & MPS_TARGET_FLAGS_RAID_COMPONENT) {
mps_dprint(sc, MPS_TRACE, "%s Raid component no SCSI IO supported %u\n",
__func__, csio->ccb_h.target_id);
csio->ccb_h.status = CAM_TID_INVALID;
xpt_done(ccb);
return;
}
/*
* If devinfo is 0 this will be a volume. In that case don't tell CAM
* that the volume has timed out. We want volumes to be enumerated
* until they are deleted/removed, not just failed.
*/
if (targ->flags & MPSSAS_TARGET_INREMOVAL) {
if (targ->devinfo == 0)
csio->ccb_h.status = CAM_REQ_CMP;
else
csio->ccb_h.status = CAM_SEL_TIMEOUT;
xpt_done(ccb);
return;
}
if ((sc->mps_flags & MPS_FLAGS_SHUTDOWN) != 0) {
mps_dprint(sc, MPS_TRACE, "%s shutting down\n", __func__);
csio->ccb_h.status = CAM_TID_INVALID;
xpt_done(ccb);
return;
}
cm = mps_alloc_command(sc);
if (cm == NULL) {
if ((sassc->flags & MPSSAS_QUEUE_FROZEN) == 0) {
xpt_freeze_simq(sassc->sim, 1);
sassc->flags |= MPSSAS_QUEUE_FROZEN;
}
ccb->ccb_h.status &= ~CAM_SIM_QUEUED;
ccb->ccb_h.status |= CAM_REQUEUE_REQ;
xpt_done(ccb);
return;
}
req = (MPI2_SCSI_IO_REQUEST *)cm->cm_req;
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
bzero(req, sizeof(*req));
req->DevHandle = targ->handle;
req->Function = MPI2_FUNCTION_SCSI_IO_REQUEST;
req->MsgFlags = 0;
req->SenseBufferLowAddress = cm->cm_sense_busaddr;
req->SenseBufferLength = MPS_SENSE_LEN;
req->SGLFlags = 0;
req->ChainOffset = 0;
req->SGLOffset0 = 24; /* 32bit word offset to the SGL */
req->SGLOffset1= 0;
req->SGLOffset2= 0;
req->SGLOffset3= 0;
req->SkipCount = 0;
req->DataLength = csio->dxfer_len;
req->BidirectionalDataLength = 0;
req->IoFlags = csio->cdb_len;
req->EEDPFlags = 0;
/* Note: BiDirectional transfers are not supported */
switch (csio->ccb_h.flags & CAM_DIR_MASK) {
case CAM_DIR_IN:
req->Control = MPI2_SCSIIO_CONTROL_READ;
cm->cm_flags |= MPS_CM_FLAGS_DATAIN;
break;
case CAM_DIR_OUT:
req->Control = MPI2_SCSIIO_CONTROL_WRITE;
cm->cm_flags |= MPS_CM_FLAGS_DATAOUT;
break;
case CAM_DIR_NONE:
default:
req->Control = MPI2_SCSIIO_CONTROL_NODATATRANSFER;
break;
}
/*
* It looks like the hardware doesn't require an explicit tag
* number for each transaction. SAM Task Management not supported
* at the moment.
*/
switch (csio->tag_action) {
case MSG_HEAD_OF_Q_TAG:
req->Control |= MPI2_SCSIIO_CONTROL_HEADOFQ;
break;
case MSG_ORDERED_Q_TAG:
req->Control |= MPI2_SCSIIO_CONTROL_ORDEREDQ;
break;
case MSG_ACA_TASK:
req->Control |= MPI2_SCSIIO_CONTROL_ACAQ;
break;
case CAM_TAG_ACTION_NONE:
case MSG_SIMPLE_Q_TAG:
default:
req->Control |= MPI2_SCSIIO_CONTROL_SIMPLEQ;
break;
}
req->Control |= sc->mapping_table[csio->ccb_h.target_id].TLR_bits;
if (MPS_SET_LUN(req->LUN, csio->ccb_h.target_lun) != 0) {
mps_free_command(sc, cm);
ccb->ccb_h.status = CAM_LUN_INVALID;
xpt_done(ccb);
return;
}
if (csio->ccb_h.flags & CAM_CDB_POINTER)
bcopy(csio->cdb_io.cdb_ptr, &req->CDB.CDB32[0], csio->cdb_len);
else
bcopy(csio->cdb_io.cdb_bytes, &req->CDB.CDB32[0],csio->cdb_len);
req->IoFlags = csio->cdb_len;
Add Serial Management Protocol (SMP) passthrough support to CAM. This includes support in the kernel, camcontrol(8), libcam and the mps(4) driver for SMP passthrough. The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00 to determine supported pages, and will now fetch page 0x83 in addition to page 0x80 if supported. Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is intended for SMP requests and responses. The ADVINFO is currently used to fetch cached VPD page 0x83 data from the transport layer, but is intended to be extensible to fetch other types of device-specific data. SMP-only devices are not currently represented in the CAM topology, and so the current semantics are that the SIM will route SMP CCBs to either the addressed device, if it contains an SMP target, or its parent, if it contains an SMP target. (This is noted in cam_ccb.h, since it will change later once we have the ability to have SMP-only devices in CAM's topology.) smp_all.c, smp_all.h: New helper routines for SMP. This includes SMP request building routines, response parsing routines, error decoding routines, and structure definitions for a number of SMP commands. libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality is available to userland applications. camcontrol.8, camcontrol.c: Add smp passthrough support to camcontrol. Several new subcommands are now available: 'smpcmd' functions much like 'cmd', except that it allows the user to send generic SMP commands. 'smprg' sends the SMP report general command, and displays the decoded output. It will automatically fetch extended output if it is available. 'smppc' sends the SMP phy control command, with any number of potential options. Among other things, this allows the user to reset a phy on a SAS expander, or disable a phy on an expander. 'smpmaninfo' sends the SMP report manufacturer information and displays the decoded output. 'smpphylist' displays a list of phys on an expander, and the CAM devices attached to those phys, if any. cam.h, cam.c: Add a status value for SMP errors (CAM_SMP_STATUS_ERROR). Add a missing description for CAM_SCSI_IT_NEXUS_LOST. Add support for SMP commands to cam_error_string(). cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP commands are by nature bi-directional, and we may need to support bi-directional SCSI commands later. Add the XPT_SMP_IO CCB. Since SMP commands are bi-directional, there are pointers for both the request and response. Add a fill routine for SMP CCBs. Add the XPT_GDEV_ADVINFO CCB. This is currently used to fetch cached page 0x83 data from the transport later, but is extensible to fetch many other types of data. cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. cam_xpt.c: Add support for executing XPT_SMP_IO CCBs. cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct cam_ed. scsi_all.c: Add scsi_get_sas_addr(), a function that parses VPD page 0x83 data and pulls out a SAS address. scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a prototype for scsi_get_sas_addr(). scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. scsi_xpt.c: In the SCSI probe code, first ask the device for VPD page 0x00. If any VPD pages are supported, that page is required to be implemented. Based on the response, we may probe for the serial number (page 0x80) or device id (page 0x83). Add support for the XPT_GDEV_ADVINFO CCB. sys/conf/files: Add smp_all.c. mps.c: Add support for passing in a uio in mps_map_command(), so we can map a S/G list at once. Add support for SMP passthrough commands in mps_data_cb(). SMP is a special case, because the first buffer in the S/G list is outbound and the second buffer is inbound. Add support for warning the user if the busdma code comes back with more buffers than will work for the command. This will, for example, help the user determine why an SMP command failed if busdma comes back with three buffers. mps_pci.c: Add sys/uio.h. mps_sas.c: Add the SAS address and the parent handle to the list of fields we pull from device page 0 and cache in struct mpssas_target. These are needed for SMP passthrough. Add support for the XPT_SMP_IO CCB. For now, this CCB is routed to the addressed device if it supports SMP, or to its parent if it does not and the parent does. This is necessary because CAM does not currently support SMP-only nodes in the topology. Make SMP passthrough support conditional on __FreeBSD_version >= 900026. This will make it easier to MFC this change to the driver without MFCing the CAM changes as well. mps_user.c: Un-staticize mpi_init_sge() so we can use it for the SMP passthrough code. mpsvar.h: Add a uio and iovecs into struct mps_command for SMP passthrough commands. Add a cm_max_segs field to struct mps_command so that we can warn the user if busdma comes back with too many segments. Clear the cm_reply when a command gets freed. If it is not cleared, reply frames will eventually get freed into the pool multiple times and corrupt the pool. (This fix is from scottl.) Add a prototype for mpi_init_sge(). sys/param.h: Bump __FreeBSD_version to 900026 for the for the inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO CAM CCBs.
2010-11-30 22:39:46 +00:00
/*
* Check if EEDP is supported and enabled. If it is then check if the
* SCSI opcode could be using EEDP. If so, make sure the LUN exists and
* is formatted for EEDP support. If all of this is true, set CDB up
* for EEDP transfer.
Add Serial Management Protocol (SMP) passthrough support to CAM. This includes support in the kernel, camcontrol(8), libcam and the mps(4) driver for SMP passthrough. The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00 to determine supported pages, and will now fetch page 0x83 in addition to page 0x80 if supported. Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is intended for SMP requests and responses. The ADVINFO is currently used to fetch cached VPD page 0x83 data from the transport layer, but is intended to be extensible to fetch other types of device-specific data. SMP-only devices are not currently represented in the CAM topology, and so the current semantics are that the SIM will route SMP CCBs to either the addressed device, if it contains an SMP target, or its parent, if it contains an SMP target. (This is noted in cam_ccb.h, since it will change later once we have the ability to have SMP-only devices in CAM's topology.) smp_all.c, smp_all.h: New helper routines for SMP. This includes SMP request building routines, response parsing routines, error decoding routines, and structure definitions for a number of SMP commands. libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality is available to userland applications. camcontrol.8, camcontrol.c: Add smp passthrough support to camcontrol. Several new subcommands are now available: 'smpcmd' functions much like 'cmd', except that it allows the user to send generic SMP commands. 'smprg' sends the SMP report general command, and displays the decoded output. It will automatically fetch extended output if it is available. 'smppc' sends the SMP phy control command, with any number of potential options. Among other things, this allows the user to reset a phy on a SAS expander, or disable a phy on an expander. 'smpmaninfo' sends the SMP report manufacturer information and displays the decoded output. 'smpphylist' displays a list of phys on an expander, and the CAM devices attached to those phys, if any. cam.h, cam.c: Add a status value for SMP errors (CAM_SMP_STATUS_ERROR). Add a missing description for CAM_SCSI_IT_NEXUS_LOST. Add support for SMP commands to cam_error_string(). cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP commands are by nature bi-directional, and we may need to support bi-directional SCSI commands later. Add the XPT_SMP_IO CCB. Since SMP commands are bi-directional, there are pointers for both the request and response. Add a fill routine for SMP CCBs. Add the XPT_GDEV_ADVINFO CCB. This is currently used to fetch cached page 0x83 data from the transport later, but is extensible to fetch many other types of data. cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. cam_xpt.c: Add support for executing XPT_SMP_IO CCBs. cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct cam_ed. scsi_all.c: Add scsi_get_sas_addr(), a function that parses VPD page 0x83 data and pulls out a SAS address. scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a prototype for scsi_get_sas_addr(). scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. scsi_xpt.c: In the SCSI probe code, first ask the device for VPD page 0x00. If any VPD pages are supported, that page is required to be implemented. Based on the response, we may probe for the serial number (page 0x80) or device id (page 0x83). Add support for the XPT_GDEV_ADVINFO CCB. sys/conf/files: Add smp_all.c. mps.c: Add support for passing in a uio in mps_map_command(), so we can map a S/G list at once. Add support for SMP passthrough commands in mps_data_cb(). SMP is a special case, because the first buffer in the S/G list is outbound and the second buffer is inbound. Add support for warning the user if the busdma code comes back with more buffers than will work for the command. This will, for example, help the user determine why an SMP command failed if busdma comes back with three buffers. mps_pci.c: Add sys/uio.h. mps_sas.c: Add the SAS address and the parent handle to the list of fields we pull from device page 0 and cache in struct mpssas_target. These are needed for SMP passthrough. Add support for the XPT_SMP_IO CCB. For now, this CCB is routed to the addressed device if it supports SMP, or to its parent if it does not and the parent does. This is necessary because CAM does not currently support SMP-only nodes in the topology. Make SMP passthrough support conditional on __FreeBSD_version >= 900026. This will make it easier to MFC this change to the driver without MFCing the CAM changes as well. mps_user.c: Un-staticize mpi_init_sge() so we can use it for the SMP passthrough code. mpsvar.h: Add a uio and iovecs into struct mps_command for SMP passthrough commands. Add a cm_max_segs field to struct mps_command so that we can warn the user if busdma comes back with too many segments. Clear the cm_reply when a command gets freed. If it is not cleared, reply frames will eventually get freed into the pool multiple times and corrupt the pool. (This fix is from scottl.) Add a prototype for mpi_init_sge(). sys/param.h: Bump __FreeBSD_version to 900026 for the for the inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO CAM CCBs.
2010-11-30 22:39:46 +00:00
*/
eedp_flags = op_code_prot[req->CDB.CDB32[0]];
if (sc->eedp_enabled && eedp_flags) {
SLIST_FOREACH(lun, &targ->luns, lun_link) {
if (lun->lun_id == csio->ccb_h.target_lun) {
break;
}
}
if ((lun != NULL) && (lun->eedp_formatted)) {
req->EEDPBlockSize = lun->eedp_block_size;
eedp_flags |= (MPI2_SCSIIO_EEDPFLAGS_INC_PRI_REFTAG |
MPI2_SCSIIO_EEDPFLAGS_CHECK_REFTAG |
MPI2_SCSIIO_EEDPFLAGS_CHECK_GUARD);
req->EEDPFlags = eedp_flags;
/*
* If CDB less than 32, fill in Primary Ref Tag with
* low 4 bytes of LBA. If CDB is 32, tag stuff is
* already there. Also, set protection bit. FreeBSD
* currently does not support CDBs bigger than 16, but
* the code doesn't hurt, and will be here for the
* future.
*/
if (csio->cdb_len != 32) {
lba_byte = (csio->cdb_len == 16) ? 6 : 2;
ref_tag_addr = (uint8_t *)&req->CDB.EEDP32.
PrimaryReferenceTag;
for (i = 0; i < 4; i++) {
*ref_tag_addr =
req->CDB.CDB32[lba_byte + i];
ref_tag_addr++;
}
req->CDB.EEDP32.PrimaryApplicationTagMask =
0xFFFF;
req->CDB.CDB32[1] = (req->CDB.CDB32[1] & 0x1F) |
0x20;
} else {
eedp_flags |=
MPI2_SCSIIO_EEDPFLAGS_INC_PRI_APPTAG;
req->EEDPFlags = eedp_flags;
req->CDB.CDB32[10] = (req->CDB.CDB32[10] &
0x1F) | 0x20;
}
}
}
cm->cm_data = csio->data_ptr;
cm->cm_length = csio->dxfer_len;
cm->cm_sge = &req->SGL;
cm->cm_sglsize = (32 - 24) * 4;
cm->cm_desc.SCSIIO.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_SCSI_IO;
cm->cm_desc.SCSIIO.DevHandle = targ->handle;
cm->cm_complete = mpssas_scsiio_complete;
cm->cm_complete_data = ccb;
cm->cm_targ = targ;
cm->cm_lun = csio->ccb_h.target_lun;
cm->cm_ccb = ccb;
/*
* If HBA is a WD and the command is not for a retry, try to build a
* direct I/O message. If failed, or the command is for a retry, send
* the I/O to the IR volume itself.
*/
if (sc->WD_valid_config) {
if (ccb->ccb_h.status != MPS_WD_RETRY) {
mpssas_direct_drive_io(sassc, cm, ccb);
} else {
ccb->ccb_h.status = CAM_REQ_INPROG;
}
}
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
callout_reset(&cm->cm_callout, (ccb->ccb_h.timeout * hz) / 1000,
mpssas_scsiio_timeout, cm);
targ->issued++;
targ->outstanding++;
TAILQ_INSERT_TAIL(&targ->commands, cm, cm_link);
if ((sc->mps_debug & MPS_TRACE) != 0)
mpssas_log_command(cm, "%s cm %p ccb %p outstanding %u\n",
__func__, cm, ccb, targ->outstanding);
mps_map_command(sc, cm);
return;
}
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
static void
mps_response_code(struct mps_softc *sc, u8 response_code)
{
char *desc;
switch (response_code) {
case MPI2_SCSITASKMGMT_RSP_TM_COMPLETE:
desc = "task management request completed";
break;
case MPI2_SCSITASKMGMT_RSP_INVALID_FRAME:
desc = "invalid frame";
break;
case MPI2_SCSITASKMGMT_RSP_TM_NOT_SUPPORTED:
desc = "task management request not supported";
break;
case MPI2_SCSITASKMGMT_RSP_TM_FAILED:
desc = "task management request failed";
break;
case MPI2_SCSITASKMGMT_RSP_TM_SUCCEEDED:
desc = "task management request succeeded";
break;
case MPI2_SCSITASKMGMT_RSP_TM_INVALID_LUN:
desc = "invalid lun";
break;
case 0xA:
desc = "overlapped tag attempted";
break;
case MPI2_SCSITASKMGMT_RSP_IO_QUEUED_ON_IOC:
desc = "task queued, however not sent to target";
break;
default:
desc = "unknown";
break;
}
mps_dprint(sc, MPS_INFO, "response_code(0x%01x): %s\n",
response_code, desc);
}
/**
* mps_sc_failed_io_info - translated non-succesfull SCSI_IO request
*/
static void
mps_sc_failed_io_info(struct mps_softc *sc, struct ccb_scsiio *csio,
Mpi2SCSIIOReply_t *mpi_reply)
{
u32 response_info;
u8 *response_bytes;
u16 ioc_status = le16toh(mpi_reply->IOCStatus) &
MPI2_IOCSTATUS_MASK;
u8 scsi_state = mpi_reply->SCSIState;
u8 scsi_status = mpi_reply->SCSIStatus;
char *desc_ioc_state = NULL;
char *desc_scsi_status = NULL;
char *desc_scsi_state = sc->tmp_string;
u32 log_info = le32toh(mpi_reply->IOCLogInfo);
if (log_info == 0x31170000)
return;
switch (ioc_status) {
case MPI2_IOCSTATUS_SUCCESS:
desc_ioc_state = "success";
break;
case MPI2_IOCSTATUS_INVALID_FUNCTION:
desc_ioc_state = "invalid function";
break;
case MPI2_IOCSTATUS_SCSI_RECOVERED_ERROR:
desc_ioc_state = "scsi recovered error";
break;
case MPI2_IOCSTATUS_SCSI_INVALID_DEVHANDLE:
desc_ioc_state = "scsi invalid dev handle";
break;
case MPI2_IOCSTATUS_SCSI_DEVICE_NOT_THERE:
desc_ioc_state = "scsi device not there";
break;
case MPI2_IOCSTATUS_SCSI_DATA_OVERRUN:
desc_ioc_state = "scsi data overrun";
break;
case MPI2_IOCSTATUS_SCSI_DATA_UNDERRUN:
desc_ioc_state = "scsi data underrun";
break;
case MPI2_IOCSTATUS_SCSI_IO_DATA_ERROR:
desc_ioc_state = "scsi io data error";
break;
case MPI2_IOCSTATUS_SCSI_PROTOCOL_ERROR:
desc_ioc_state = "scsi protocol error";
break;
case MPI2_IOCSTATUS_SCSI_TASK_TERMINATED:
desc_ioc_state = "scsi task terminated";
break;
case MPI2_IOCSTATUS_SCSI_RESIDUAL_MISMATCH:
desc_ioc_state = "scsi residual mismatch";
break;
case MPI2_IOCSTATUS_SCSI_TASK_MGMT_FAILED:
desc_ioc_state = "scsi task mgmt failed";
break;
case MPI2_IOCSTATUS_SCSI_IOC_TERMINATED:
desc_ioc_state = "scsi ioc terminated";
break;
case MPI2_IOCSTATUS_SCSI_EXT_TERMINATED:
desc_ioc_state = "scsi ext terminated";
break;
case MPI2_IOCSTATUS_EEDP_GUARD_ERROR:
desc_ioc_state = "eedp guard error";
break;
case MPI2_IOCSTATUS_EEDP_REF_TAG_ERROR:
desc_ioc_state = "eedp ref tag error";
break;
case MPI2_IOCSTATUS_EEDP_APP_TAG_ERROR:
desc_ioc_state = "eedp app tag error";
break;
default:
desc_ioc_state = "unknown";
break;
}
switch (scsi_status) {
case MPI2_SCSI_STATUS_GOOD:
desc_scsi_status = "good";
break;
case MPI2_SCSI_STATUS_CHECK_CONDITION:
desc_scsi_status = "check condition";
break;
case MPI2_SCSI_STATUS_CONDITION_MET:
desc_scsi_status = "condition met";
break;
case MPI2_SCSI_STATUS_BUSY:
desc_scsi_status = "busy";
break;
case MPI2_SCSI_STATUS_INTERMEDIATE:
desc_scsi_status = "intermediate";
break;
case MPI2_SCSI_STATUS_INTERMEDIATE_CONDMET:
desc_scsi_status = "intermediate condmet";
break;
case MPI2_SCSI_STATUS_RESERVATION_CONFLICT:
desc_scsi_status = "reservation conflict";
break;
case MPI2_SCSI_STATUS_COMMAND_TERMINATED:
desc_scsi_status = "command terminated";
break;
case MPI2_SCSI_STATUS_TASK_SET_FULL:
desc_scsi_status = "task set full";
break;
case MPI2_SCSI_STATUS_ACA_ACTIVE:
desc_scsi_status = "aca active";
break;
case MPI2_SCSI_STATUS_TASK_ABORTED:
desc_scsi_status = "task aborted";
break;
default:
desc_scsi_status = "unknown";
break;
}
desc_scsi_state[0] = '\0';
if (!scsi_state)
desc_scsi_state = " ";
if (scsi_state & MPI2_SCSI_STATE_RESPONSE_INFO_VALID)
strcat(desc_scsi_state, "response info ");
if (scsi_state & MPI2_SCSI_STATE_TERMINATED)
strcat(desc_scsi_state, "state terminated ");
if (scsi_state & MPI2_SCSI_STATE_NO_SCSI_STATUS)
strcat(desc_scsi_state, "no status ");
if (scsi_state & MPI2_SCSI_STATE_AUTOSENSE_FAILED)
strcat(desc_scsi_state, "autosense failed ");
if (scsi_state & MPI2_SCSI_STATE_AUTOSENSE_VALID)
strcat(desc_scsi_state, "autosense valid ");
mps_dprint(sc, MPS_INFO, "\thandle(0x%04x), ioc_status(%s)(0x%04x), \n",
le16toh(mpi_reply->DevHandle),
desc_ioc_state, ioc_status);
/* We can add more detail about underflow data here
* TO-DO
* */
mps_dprint(sc, MPS_INFO, "\tscsi_status(%s)(0x%02x), "
"scsi_state(%s)(0x%02x)\n", desc_scsi_status,
scsi_status, desc_scsi_state, scsi_state);
if (sc->mps_debug & MPS_INFO &&
scsi_state & MPI2_SCSI_STATE_AUTOSENSE_VALID) {
mps_dprint(sc, MPS_INFO, "-> Sense Buffer Data : Start :\n");
scsi_sense_print(csio);
mps_dprint(sc, MPS_INFO, "-> Sense Buffer Data : End :\n");
}
if (scsi_state & MPI2_SCSI_STATE_RESPONSE_INFO_VALID) {
response_info = le32toh(mpi_reply->ResponseInfo);
response_bytes = (u8 *)&response_info;
mps_response_code(sc,response_bytes[0]);
}
}
static void
mpssas_scsiio_complete(struct mps_softc *sc, struct mps_command *cm)
{
MPI2_SCSI_IO_REPLY *rep;
union ccb *ccb;
struct ccb_scsiio *csio;
struct mpssas_softc *sassc;
struct scsi_vpd_supported_page_list *vpd_list = NULL;
u8 *TLR_bits, TLR_on;
int dir = 0, i;
u16 alloc_len;
mps_dprint(sc, MPS_TRACE,
"%s cm %p SMID %u ccb %p reply %p outstanding %u\n",
__func__, cm, cm->cm_desc.Default.SMID, cm->cm_ccb, cm->cm_reply,
cm->cm_targ->outstanding);
callout_stop(&cm->cm_callout);
mtx_assert(&sc->mps_mtx, MA_OWNED);
sassc = sc->sassc;
ccb = cm->cm_complete_data;
csio = &ccb->csio;
rep = (MPI2_SCSI_IO_REPLY *)cm->cm_reply;
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
/*
* XXX KDM if the chain allocation fails, does it matter if we do
* the sync and unload here? It is simpler to do it in every case,
* assuming it doesn't cause problems.
*/
if (cm->cm_data != NULL) {
if (cm->cm_flags & MPS_CM_FLAGS_DATAIN)
dir = BUS_DMASYNC_POSTREAD;
else if (cm->cm_flags & MPS_CM_FLAGS_DATAOUT)
dir = BUS_DMASYNC_POSTWRITE;;
bus_dmamap_sync(sc->buffer_dmat, cm->cm_dmamap, dir);
bus_dmamap_unload(sc->buffer_dmat, cm->cm_dmamap);
}
cm->cm_targ->completed++;
cm->cm_targ->outstanding--;
TAILQ_REMOVE(&cm->cm_targ->commands, cm, cm_link);
if (cm->cm_state == MPS_CM_STATE_TIMEDOUT) {
TAILQ_REMOVE(&cm->cm_targ->timedout_commands, cm, cm_recovery);
if (cm->cm_reply != NULL)
mpssas_log_command(cm,
"completed timedout cm %p ccb %p during recovery "
"ioc %x scsi %x state %x xfer %u\n",
cm, cm->cm_ccb,
rep->IOCStatus, rep->SCSIStatus, rep->SCSIState,
rep->TransferCount);
else
mpssas_log_command(cm,
"completed timedout cm %p ccb %p during recovery\n",
cm, cm->cm_ccb);
} else if (cm->cm_targ->tm != NULL) {
if (cm->cm_reply != NULL)
mpssas_log_command(cm,
"completed cm %p ccb %p during recovery "
"ioc %x scsi %x state %x xfer %u\n",
cm, cm->cm_ccb,
rep->IOCStatus, rep->SCSIStatus, rep->SCSIState,
rep->TransferCount);
else
mpssas_log_command(cm,
"completed cm %p ccb %p during recovery\n",
cm, cm->cm_ccb);
} else if ((sc->mps_flags & MPS_FLAGS_DIAGRESET) != 0) {
mpssas_log_command(cm,
"reset completed cm %p ccb %p\n",
cm, cm->cm_ccb);
}
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
if ((cm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
/*
* We ran into an error after we tried to map the command,
* so we're getting a callback without queueing the command
* to the hardware. So we set the status here, and it will
* be retained below. We'll go through the "fast path",
* because there can be no reply when we haven't actually
* gone out to the hardware.
*/
ccb->ccb_h.status |= CAM_REQUEUE_REQ;
/*
* Currently the only error included in the mask is
* MPS_CM_FLAGS_CHAIN_FAILED, which means we're out of
* chain frames. We need to freeze the queue until we get
* a command that completed without this error, which will
* hopefully have some chain frames attached that we can
* use. If we wanted to get smarter about it, we would
* only unfreeze the queue in this condition when we're
* sure that we're getting some chain frames back. That's
* probably unnecessary.
*/
if ((sassc->flags & MPSSAS_QUEUE_FROZEN) == 0) {
xpt_freeze_simq(sassc->sim, 1);
sassc->flags |= MPSSAS_QUEUE_FROZEN;
mps_dprint(sc, MPS_INFO, "Error sending command, "
"freezing SIM queue\n");
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
}
}
/* Take the fast path to completion */
if (cm->cm_reply == NULL) {
if ((ccb->ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_INPROG) {
if ((sc->mps_flags & MPS_FLAGS_DIAGRESET) != 0)
ccb->ccb_h.status = CAM_SCSI_BUS_RESET;
else {
ccb->ccb_h.status = CAM_REQ_CMP;
ccb->csio.scsi_status = SCSI_STATUS_OK;
}
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
if (sassc->flags & MPSSAS_QUEUE_FROZEN) {
ccb->ccb_h.status |= CAM_RELEASE_SIMQ;
sassc->flags &= ~MPSSAS_QUEUE_FROZEN;
mps_dprint(sc, MPS_INFO,
"Unfreezing SIM queue\n");
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
}
}
/*
* There are two scenarios where the status won't be
* CAM_REQ_CMP. The first is if MPS_CM_FLAGS_ERROR_MASK is
* set, the second is in the MPS_FLAGS_DIAGRESET above.
*/
if ((ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) {
/*
* Freeze the dev queue so that commands are
* executed in the correct order with after error
* recovery.
*/
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
ccb->ccb_h.status |= CAM_DEV_QFRZN;
xpt_freeze_devq(ccb->ccb_h.path, /*count*/ 1);
}
mps_free_command(sc, cm);
xpt_done(ccb);
return;
}
if (sc->mps_debug & MPS_TRACE)
mpssas_log_command(cm,
"ioc %x scsi %x state %x xfer %u\n",
rep->IOCStatus, rep->SCSIStatus,
rep->SCSIState, rep->TransferCount);
/*
* If this is a Direct Drive I/O, reissue the I/O to the original IR
* Volume if an error occurred (normal I/O retry). Use the original
* CCB, but set a flag that this will be a retry so that it's sent to
* the original volume. Free the command but reuse the CCB.
*/
if (cm->cm_flags & MPS_CM_FLAGS_DD_IO) {
mps_free_command(sc, cm);
ccb->ccb_h.status = MPS_WD_RETRY;
mpssas_action_scsiio(sassc, ccb);
return;
}
switch (rep->IOCStatus & MPI2_IOCSTATUS_MASK) {
case MPI2_IOCSTATUS_SCSI_DATA_UNDERRUN:
csio->resid = cm->cm_length - rep->TransferCount;
/* FALLTHROUGH */
case MPI2_IOCSTATUS_SUCCESS:
case MPI2_IOCSTATUS_SCSI_RECOVERED_ERROR:
if ((rep->IOCStatus & MPI2_IOCSTATUS_MASK) ==
MPI2_IOCSTATUS_SCSI_RECOVERED_ERROR)
mpssas_log_command(cm, "recovered error\n");
/* Completion failed at the transport level. */
if (rep->SCSIState & (MPI2_SCSI_STATE_NO_SCSI_STATUS |
MPI2_SCSI_STATE_TERMINATED)) {
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
break;
}
/* In a modern packetized environment, an autosense failure
* implies that there's not much else that can be done to
* recover the command.
*/
if (rep->SCSIState & MPI2_SCSI_STATE_AUTOSENSE_FAILED) {
ccb->ccb_h.status = CAM_AUTOSENSE_FAIL;
break;
}
/*
* CAM doesn't care about SAS Response Info data, but if this is
* the state check if TLR should be done. If not, clear the
* TLR_bits for the target.
*/
if ((rep->SCSIState & MPI2_SCSI_STATE_RESPONSE_INFO_VALID) &&
((rep->ResponseInfo & MPI2_SCSI_RI_MASK_REASONCODE) ==
MPS_SCSI_RI_INVALID_FRAME)) {
sc->mapping_table[csio->ccb_h.target_id].TLR_bits =
(u8)MPI2_SCSIIO_CONTROL_NO_TLR;
}
/*
* Intentionally override the normal SCSI status reporting
* for these two cases. These are likely to happen in a
* multi-initiator environment, and we want to make sure that
* CAM retries these commands rather than fail them.
*/
if ((rep->SCSIStatus == MPI2_SCSI_STATUS_COMMAND_TERMINATED) ||
(rep->SCSIStatus == MPI2_SCSI_STATUS_TASK_ABORTED)) {
ccb->ccb_h.status = CAM_REQ_ABORTED;
break;
}
/* Handle normal status and sense */
csio->scsi_status = rep->SCSIStatus;
if (rep->SCSIStatus == MPI2_SCSI_STATUS_GOOD)
ccb->ccb_h.status = CAM_REQ_CMP;
else
ccb->ccb_h.status = CAM_SCSI_STATUS_ERROR;
if (rep->SCSIState & MPI2_SCSI_STATE_AUTOSENSE_VALID) {
int sense_len, returned_sense_len;
returned_sense_len = min(rep->SenseCount,
sizeof(struct scsi_sense_data));
if (returned_sense_len < ccb->csio.sense_len)
ccb->csio.sense_resid = ccb->csio.sense_len -
returned_sense_len;
else
ccb->csio.sense_resid = 0;
sense_len = min(returned_sense_len,
ccb->csio.sense_len - ccb->csio.sense_resid);
bzero(&ccb->csio.sense_data,
sizeof(&ccb->csio.sense_data));
bcopy(cm->cm_sense, &ccb->csio.sense_data, sense_len);
ccb->ccb_h.status |= CAM_AUTOSNS_VALID;
}
/*
* Check if this is an INQUIRY command. If it's a VPD inquiry,
* and it's page code 0 (Supported Page List), and there is
* inquiry data, and this is for a sequential access device, and
* the device is an SSP target, and TLR is supported by the
* controller, turn the TLR_bits value ON if page 0x90 is
* supported.
*/
if ((csio->cdb_io.cdb_bytes[0] == INQUIRY) &&
(csio->cdb_io.cdb_bytes[1] & SI_EVPD) &&
(csio->cdb_io.cdb_bytes[2] == SVPD_SUPPORTED_PAGE_LIST) &&
(csio->data_ptr != NULL) && (((uint8_t *)cm->cm_data)[0] ==
T_SEQUENTIAL) && (sc->control_TLR) &&
(sc->mapping_table[csio->ccb_h.target_id].device_info &
MPI2_SAS_DEVICE_INFO_SSP_TARGET)) {
vpd_list = (struct scsi_vpd_supported_page_list *)
csio->data_ptr;
TLR_bits = &sc->mapping_table[csio->ccb_h.target_id].
TLR_bits;
*TLR_bits = (u8)MPI2_SCSIIO_CONTROL_NO_TLR;
TLR_on = (u8)MPI2_SCSIIO_CONTROL_TLR_ON;
alloc_len = ((u16)csio->cdb_io.cdb_bytes[3] << 8) +
csio->cdb_io.cdb_bytes[4];
for (i = 0; i < MIN(vpd_list->length, alloc_len); i++) {
if (vpd_list->list[i] == 0x90) {
*TLR_bits = TLR_on;
break;
}
}
}
break;
case MPI2_IOCSTATUS_SCSI_INVALID_DEVHANDLE:
case MPI2_IOCSTATUS_SCSI_DEVICE_NOT_THERE:
/*
* If devinfo is 0 this will be a volume. In that case don't
* tell CAM that the volume is not there. We want volumes to
* be enumerated until they are deleted/removed, not just
* failed.
*/
if (cm->cm_targ->devinfo == 0)
ccb->ccb_h.status = CAM_REQ_CMP;
else
ccb->ccb_h.status = CAM_DEV_NOT_THERE;
break;
case MPI2_IOCSTATUS_INVALID_SGL:
mps_print_scsiio_cmd(sc, cm);
ccb->ccb_h.status = CAM_UNREC_HBA_ERROR;
break;
case MPI2_IOCSTATUS_SCSI_TASK_TERMINATED:
/*
* This is one of the responses that comes back when an I/O
* has been aborted. If it is because of a timeout that we
* initiated, just set the status to CAM_CMD_TIMEOUT.
* Otherwise set it to CAM_REQ_ABORTED. The effect on the
* command is the same (it gets retried, subject to the
* retry counter), the only difference is what gets printed
* on the console.
*/
if (cm->cm_state == MPS_CM_STATE_TIMEDOUT)
ccb->ccb_h.status = CAM_CMD_TIMEOUT;
else
ccb->ccb_h.status = CAM_REQ_ABORTED;
break;
case MPI2_IOCSTATUS_SCSI_DATA_OVERRUN:
/* resid is ignored for this condition */
csio->resid = 0;
ccb->ccb_h.status = CAM_DATA_RUN_ERR;
break;
case MPI2_IOCSTATUS_SCSI_IOC_TERMINATED:
case MPI2_IOCSTATUS_SCSI_EXT_TERMINATED:
/*
* Since these are generally external (i.e. hopefully
* transient transport-related) errors, retry these without
* decrementing the retry count.
*/
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
ccb->ccb_h.status = CAM_REQUEUE_REQ;
mpssas_log_command(cm,
"terminated ioc %x scsi %x state %x xfer %u\n",
rep->IOCStatus, rep->SCSIStatus, rep->SCSIState,
rep->TransferCount);
break;
case MPI2_IOCSTATUS_INVALID_FUNCTION:
case MPI2_IOCSTATUS_INTERNAL_ERROR:
case MPI2_IOCSTATUS_INVALID_VPID:
case MPI2_IOCSTATUS_INVALID_FIELD:
case MPI2_IOCSTATUS_INVALID_STATE:
case MPI2_IOCSTATUS_OP_STATE_NOT_SUPPORTED:
case MPI2_IOCSTATUS_SCSI_IO_DATA_ERROR:
case MPI2_IOCSTATUS_SCSI_PROTOCOL_ERROR:
case MPI2_IOCSTATUS_SCSI_RESIDUAL_MISMATCH:
case MPI2_IOCSTATUS_SCSI_TASK_MGMT_FAILED:
default:
mpssas_log_command(cm,
"completed ioc %x scsi %x state %x xfer %u\n",
rep->IOCStatus, rep->SCSIStatus, rep->SCSIState,
rep->TransferCount);
csio->resid = cm->cm_length;
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
break;
}
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
mps_sc_failed_io_info(sc,csio,rep);
if (sassc->flags & MPSSAS_QUEUE_FROZEN) {
ccb->ccb_h.status |= CAM_RELEASE_SIMQ;
sassc->flags &= ~MPSSAS_QUEUE_FROZEN;
mps_dprint(sc, MPS_INFO, "Command completed, "
"unfreezing SIM queue\n");
}
if ((ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) {
ccb->ccb_h.status |= CAM_DEV_QFRZN;
xpt_freeze_devq(ccb->ccb_h.path, /*count*/ 1);
}
mps_free_command(sc, cm);
xpt_done(ccb);
}
Add descriptor sense support to CAM, and honor sense residuals properly in CAM. Desriptor sense is a new sense data format that originated in SPC-3. Among other things, it allows for an 8-byte info field, which is necessary to pass back block numbers larger than 4 bytes. This change adds a number of new functions to scsi_all.c (and therefore libcam) that abstract out most access to sense data. This includes a bump of CAM_VERSION, because the CCB ABI has changed. Userland programs that use the CAM pass(4) driver will need to be recompiled. camcontrol.c: Change uses of scsi_extract_sense() to use scsi_extract_sense_len(). Use scsi_get_sks() instead of accessing sense key specific data directly. scsi_modes: Update the control mode page to the latest version (SPC-4). scsi_cmds.c, scsi_target.c: Change references to struct scsi_sense_data to struct scsi_sense_data_fixed. This should be changed to allow the user to specify fixed or descriptor sense, and then use scsi_set_sense_data() to build the sense data. ps3cdrom.c: Use scsi_set_sense_data() instead of setting sense data manually. cam_periph.c: Use scsi_extract_sense_len() instead of using scsi_extract_sense() or accessing sense data directly. cam_ccb.h: Bump the CAM_VERSION from 0x15 to 0x16. The change of struct scsi_sense_data from 32 to 252 bytes changes the size of struct ccb_scsiio, but not the size of union ccb. So the version must be bumped to prevent structure mis-matches. scsi_all.h: Lots of updated SCSI sense data and other structures. Add function prototypes for the new sense data functions. Take out the inline implementation of scsi_extract_sense(). It is now too large to put in a header file. Add macros to calculate whether fields are present and filled in fixed and descriptor sense data scsi_all.c: In scsi_op_desc(), allow the user to pass in NULL inquiry data, and we'll assume a direct access device in that case. Changed the SCSI RESERVED sense key name and description to COMPLETED, as it is now defined in the spec. Change the error recovery action for a number of read errors to prevent lots of retries when the drive has said that the block isn't accessible. This speeds up reconstruction of the block by any RAID software running on top of the drive (e.g. ZFS). In scsi_sense_desc(), allow for invalid sense key numbers. This allows calling this routine without checking the input values first. Change scsi_error_action() to use scsi_extract_sense_len(), and handle things when invalid asc/ascq values are encountered. Add a new routine, scsi_desc_iterate(), that will call the supplied function for every descriptor in descriptor format sense data. Add scsi_set_sense_data(), and scsi_set_sense_data_va(), which build descriptor and fixed format sense data. They currently default to fixed format sense data. Add a number of scsi_get_*() functions, which get different types of sense data fields from either fixed or descriptor format sense data, if the data is present. Add a number of scsi_*_sbuf() functions, which print formatted versions of various sense data fields. These functions work for either fixed or descriptor sense. Add a number of scsi_sense_*_sbuf() functions, which have a standard calling interface and print the indicated field. These functions take descriptors only. Add scsi_sense_desc_sbuf(), which will print a formatted version of the given sense descriptor. Pull out a majority of the scsi_sense_sbuf() function and put it into scsi_sense_only_sbuf(). This allows callers that don't use struct ccb_scsiio to easily utilize the printing routines. Revamp that function to handle descriptor sense and use the new sense fetching and printing routines. Move scsi_extract_sense() into scsi_all.c, and implement it in terms of the new function, scsi_extract_sense_len(). The _len() version takes a length (which should be the sense length - residual) and can indicate which fields are present and valid in the sense data. Add a couple of new scsi_get_*() routines to get the sense key, asc, and ascq only. mly.c: Rename struct scsi_sense_data to struct scsi_sense_data_fixed. sbp_targ.c: Use the new sense fetching routines to get sense data instead of accessing it directly. sbp.c: Change the firewire/SCSI sense data transformation code to use struct scsi_sense_data_fixed instead of struct scsi_sense_data. This should be changed later to use scsi_set_sense_data(). ciss.c: Calculate the sense residual properly. Use scsi_get_sense_key() to fetch the sense key. mps_sas.c, mpt_cam.c: Set the sense residual properly. iir.c: Use scsi_set_sense_data() instead of building sense data by hand. iscsi_subr.c: Use scsi_extract_sense_len() instead of grabbing sense data directly. umass.c: Use scsi_set_sense_data() to build sense data. Grab the sense key using scsi_get_sense_key(). Calculate the sense residual properly. isp_freebsd.h: Use scsi_get_*() routines to grab asc, ascq, and sense key values. Calculate and set the sense residual. MFC after: 3 days Sponsored by: Spectra Logic Corporation
2011-10-03 20:32:55 +00:00
static void
mpssas_direct_drive_io(struct mpssas_softc *sassc, struct mps_command *cm,
union ccb *ccb) {
pMpi2SCSIIORequest_t pIO_req;
struct mps_softc *sc = sassc->sc;
uint64_t virtLBA;
uint32_t physLBA, stripe_offset, stripe_unit;
uint32_t io_size, column;
uint8_t *ptrLBA, lba_idx, physLBA_byte, *CDB;
Add descriptor sense support to CAM, and honor sense residuals properly in CAM. Desriptor sense is a new sense data format that originated in SPC-3. Among other things, it allows for an 8-byte info field, which is necessary to pass back block numbers larger than 4 bytes. This change adds a number of new functions to scsi_all.c (and therefore libcam) that abstract out most access to sense data. This includes a bump of CAM_VERSION, because the CCB ABI has changed. Userland programs that use the CAM pass(4) driver will need to be recompiled. camcontrol.c: Change uses of scsi_extract_sense() to use scsi_extract_sense_len(). Use scsi_get_sks() instead of accessing sense key specific data directly. scsi_modes: Update the control mode page to the latest version (SPC-4). scsi_cmds.c, scsi_target.c: Change references to struct scsi_sense_data to struct scsi_sense_data_fixed. This should be changed to allow the user to specify fixed or descriptor sense, and then use scsi_set_sense_data() to build the sense data. ps3cdrom.c: Use scsi_set_sense_data() instead of setting sense data manually. cam_periph.c: Use scsi_extract_sense_len() instead of using scsi_extract_sense() or accessing sense data directly. cam_ccb.h: Bump the CAM_VERSION from 0x15 to 0x16. The change of struct scsi_sense_data from 32 to 252 bytes changes the size of struct ccb_scsiio, but not the size of union ccb. So the version must be bumped to prevent structure mis-matches. scsi_all.h: Lots of updated SCSI sense data and other structures. Add function prototypes for the new sense data functions. Take out the inline implementation of scsi_extract_sense(). It is now too large to put in a header file. Add macros to calculate whether fields are present and filled in fixed and descriptor sense data scsi_all.c: In scsi_op_desc(), allow the user to pass in NULL inquiry data, and we'll assume a direct access device in that case. Changed the SCSI RESERVED sense key name and description to COMPLETED, as it is now defined in the spec. Change the error recovery action for a number of read errors to prevent lots of retries when the drive has said that the block isn't accessible. This speeds up reconstruction of the block by any RAID software running on top of the drive (e.g. ZFS). In scsi_sense_desc(), allow for invalid sense key numbers. This allows calling this routine without checking the input values first. Change scsi_error_action() to use scsi_extract_sense_len(), and handle things when invalid asc/ascq values are encountered. Add a new routine, scsi_desc_iterate(), that will call the supplied function for every descriptor in descriptor format sense data. Add scsi_set_sense_data(), and scsi_set_sense_data_va(), which build descriptor and fixed format sense data. They currently default to fixed format sense data. Add a number of scsi_get_*() functions, which get different types of sense data fields from either fixed or descriptor format sense data, if the data is present. Add a number of scsi_*_sbuf() functions, which print formatted versions of various sense data fields. These functions work for either fixed or descriptor sense. Add a number of scsi_sense_*_sbuf() functions, which have a standard calling interface and print the indicated field. These functions take descriptors only. Add scsi_sense_desc_sbuf(), which will print a formatted version of the given sense descriptor. Pull out a majority of the scsi_sense_sbuf() function and put it into scsi_sense_only_sbuf(). This allows callers that don't use struct ccb_scsiio to easily utilize the printing routines. Revamp that function to handle descriptor sense and use the new sense fetching and printing routines. Move scsi_extract_sense() into scsi_all.c, and implement it in terms of the new function, scsi_extract_sense_len(). The _len() version takes a length (which should be the sense length - residual) and can indicate which fields are present and valid in the sense data. Add a couple of new scsi_get_*() routines to get the sense key, asc, and ascq only. mly.c: Rename struct scsi_sense_data to struct scsi_sense_data_fixed. sbp_targ.c: Use the new sense fetching routines to get sense data instead of accessing it directly. sbp.c: Change the firewire/SCSI sense data transformation code to use struct scsi_sense_data_fixed instead of struct scsi_sense_data. This should be changed later to use scsi_set_sense_data(). ciss.c: Calculate the sense residual properly. Use scsi_get_sense_key() to fetch the sense key. mps_sas.c, mpt_cam.c: Set the sense residual properly. iir.c: Use scsi_set_sense_data() instead of building sense data by hand. iscsi_subr.c: Use scsi_extract_sense_len() instead of grabbing sense data directly. umass.c: Use scsi_set_sense_data() to build sense data. Grab the sense key using scsi_get_sense_key(). Calculate the sense residual properly. isp_freebsd.h: Use scsi_get_*() routines to grab asc, ascq, and sense key values. Calculate and set the sense residual. MFC after: 3 days Sponsored by: Spectra Logic Corporation
2011-10-03 20:32:55 +00:00
/*
* If this is a valid SCSI command (Read6, Read10, Read16, Write6,
* Write10, or Write16), build a direct I/O message. Otherwise, the I/O
* will be sent to the IR volume itself. Since Read6 and Write6 are a
* bit different than the 10/16 CDBs, handle them separately.
*/
pIO_req = (pMpi2SCSIIORequest_t)cm->cm_req;
CDB = pIO_req->CDB.CDB32;
/*
* Handle 6 byte CDBs.
*/
if ((pIO_req->DevHandle == sc->DD_dev_handle) && ((CDB[0] == READ_6) ||
(CDB[0] == WRITE_6))) {
/*
* Get the transfer size in blocks.
*/
io_size = (cm->cm_length >> sc->DD_block_exponent);
/*
* Get virtual LBA given in the CDB.
*/
virtLBA = ((uint64_t)(CDB[1] & 0x1F) << 16) |
((uint64_t)CDB[2] << 8) | (uint64_t)CDB[3];
/*
* Check that LBA range for I/O does not exceed volume's
* MaxLBA.
*/
if ((virtLBA + (uint64_t)io_size - 1) <=
sc->DD_max_lba) {
/*
* Check if the I/O crosses a stripe boundary. If not,
* translate the virtual LBA to a physical LBA and set
* the DevHandle for the PhysDisk to be used. If it
* does cross a boundry, do normal I/O. To get the
* right DevHandle to use, get the map number for the
* column, then use that map number to look up the
* DevHandle of the PhysDisk.
*/
stripe_offset = (uint32_t)virtLBA &
(sc->DD_stripe_size - 1);
if ((stripe_offset + io_size) <= sc->DD_stripe_size) {
physLBA = (uint32_t)virtLBA >>
sc->DD_stripe_exponent;
stripe_unit = physLBA / sc->DD_num_phys_disks;
column = physLBA % sc->DD_num_phys_disks;
pIO_req->DevHandle =
sc->DD_column_map[column].dev_handle;
cm->cm_desc.SCSIIO.DevHandle =
pIO_req->DevHandle;
physLBA = (stripe_unit <<
sc->DD_stripe_exponent) + stripe_offset;
ptrLBA = &pIO_req->CDB.CDB32[1];
physLBA_byte = (uint8_t)(physLBA >> 16);
*ptrLBA = physLBA_byte;
ptrLBA = &pIO_req->CDB.CDB32[2];
physLBA_byte = (uint8_t)(physLBA >> 8);
*ptrLBA = physLBA_byte;
ptrLBA = &pIO_req->CDB.CDB32[3];
physLBA_byte = (uint8_t)physLBA;
*ptrLBA = physLBA_byte;
/*
* Set flag that Direct Drive I/O is
* being done.
*/
cm->cm_flags |= MPS_CM_FLAGS_DD_IO;
}
}
return;
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
}
/*
* Handle 10 or 16 byte CDBs.
*/
if ((pIO_req->DevHandle == sc->DD_dev_handle) && ((CDB[0] == READ_10) ||
(CDB[0] == WRITE_10) || (CDB[0] == READ_16) ||
(CDB[0] == WRITE_16))) {
/*
* For 16-byte CDB's, verify that the upper 4 bytes of the CDB
* are 0. If not, this is accessing beyond 2TB so handle it in
* the else section. 10-byte CDB's are OK.
*/
if ((CDB[0] < READ_16) ||
!(CDB[2] | CDB[3] | CDB[4] | CDB[5])) {
/*
* Get the transfer size in blocks.
*/
io_size = (cm->cm_length >> sc->DD_block_exponent);
/*
* Get virtual LBA. Point to correct lower 4 bytes of
* LBA in the CDB depending on command.
*/
lba_idx = (CDB[0] < READ_16) ? 2 : 6;
virtLBA = ((uint64_t)CDB[lba_idx] << 24) |
((uint64_t)CDB[lba_idx + 1] << 16) |
((uint64_t)CDB[lba_idx + 2] << 8) |
(uint64_t)CDB[lba_idx + 3];
/*
* Check that LBA range for I/O does not exceed volume's
* MaxLBA.
*/
if ((virtLBA + (uint64_t)io_size - 1) <=
sc->DD_max_lba) {
/*
* Check if the I/O crosses a stripe boundary.
* If not, translate the virtual LBA to a
* physical LBA and set the DevHandle for the
* PhysDisk to be used. If it does cross a
* boundry, do normal I/O. To get the right
* DevHandle to use, get the map number for the
* column, then use that map number to look up
* the DevHandle of the PhysDisk.
*/
stripe_offset = (uint32_t)virtLBA &
(sc->DD_stripe_size - 1);
if ((stripe_offset + io_size) <=
sc->DD_stripe_size) {
physLBA = (uint32_t)virtLBA >>
sc->DD_stripe_exponent;
stripe_unit = physLBA /
sc->DD_num_phys_disks;
column = physLBA %
sc->DD_num_phys_disks;
pIO_req->DevHandle =
sc->DD_column_map[column].
dev_handle;
cm->cm_desc.SCSIIO.DevHandle =
pIO_req->DevHandle;
physLBA = (stripe_unit <<
sc->DD_stripe_exponent) +
stripe_offset;
ptrLBA =
&pIO_req->CDB.CDB32[lba_idx];
physLBA_byte = (uint8_t)(physLBA >> 24);
*ptrLBA = physLBA_byte;
ptrLBA =
&pIO_req->CDB.CDB32[lba_idx + 1];
physLBA_byte = (uint8_t)(physLBA >> 16);
*ptrLBA = physLBA_byte;
ptrLBA =
&pIO_req->CDB.CDB32[lba_idx + 2];
physLBA_byte = (uint8_t)(physLBA >> 8);
*ptrLBA = physLBA_byte;
ptrLBA =
&pIO_req->CDB.CDB32[lba_idx + 3];
physLBA_byte = (uint8_t)physLBA;
*ptrLBA = physLBA_byte;
/*
* Set flag that Direct Drive I/O is
* being done.
*/
cm->cm_flags |= MPS_CM_FLAGS_DD_IO;
}
}
} else {
/*
* 16-byte CDB and the upper 4 bytes of the CDB are not
* 0. Get the transfer size in blocks.
*/
io_size = (cm->cm_length >> sc->DD_block_exponent);
/*
* Get virtual LBA.
*/
virtLBA = ((uint64_t)CDB[2] << 54) |
((uint64_t)CDB[3] << 48) |
((uint64_t)CDB[4] << 40) |
((uint64_t)CDB[5] << 32) |
((uint64_t)CDB[6] << 24) |
((uint64_t)CDB[7] << 16) |
((uint64_t)CDB[8] << 8) |
(uint64_t)CDB[9];
/*
* Check that LBA range for I/O does not exceed volume's
* MaxLBA.
*/
if ((virtLBA + (uint64_t)io_size - 1) <=
sc->DD_max_lba) {
/*
* Check if the I/O crosses a stripe boundary.
* If not, translate the virtual LBA to a
* physical LBA and set the DevHandle for the
* PhysDisk to be used. If it does cross a
* boundry, do normal I/O. To get the right
* DevHandle to use, get the map number for the
* column, then use that map number to look up
* the DevHandle of the PhysDisk.
*/
stripe_offset = (uint32_t)virtLBA &
(sc->DD_stripe_size - 1);
if ((stripe_offset + io_size) <=
sc->DD_stripe_size) {
physLBA = (uint32_t)(virtLBA >>
sc->DD_stripe_exponent);
stripe_unit = physLBA /
sc->DD_num_phys_disks;
column = physLBA %
sc->DD_num_phys_disks;
pIO_req->DevHandle =
sc->DD_column_map[column].
dev_handle;
cm->cm_desc.SCSIIO.DevHandle =
pIO_req->DevHandle;
physLBA = (stripe_unit <<
sc->DD_stripe_exponent) +
stripe_offset;
/*
* Set upper 4 bytes of LBA to 0. We
* assume that the phys disks are less
* than 2 TB's in size. Then, set the
* lower 4 bytes.
*/
pIO_req->CDB.CDB32[2] = 0;
pIO_req->CDB.CDB32[3] = 0;
pIO_req->CDB.CDB32[4] = 0;
pIO_req->CDB.CDB32[5] = 0;
ptrLBA = &pIO_req->CDB.CDB32[6];
physLBA_byte = (uint8_t)(physLBA >> 24);
*ptrLBA = physLBA_byte;
ptrLBA = &pIO_req->CDB.CDB32[7];
physLBA_byte = (uint8_t)(physLBA >> 16);
*ptrLBA = physLBA_byte;
ptrLBA = &pIO_req->CDB.CDB32[8];
physLBA_byte = (uint8_t)(physLBA >> 8);
*ptrLBA = physLBA_byte;
ptrLBA = &pIO_req->CDB.CDB32[9];
physLBA_byte = (uint8_t)physLBA;
*ptrLBA = physLBA_byte;
/*
* Set flag that Direct Drive I/O is
* being done.
*/
cm->cm_flags |= MPS_CM_FLAGS_DD_IO;
}
}
}
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
}
}
Add Serial Management Protocol (SMP) passthrough support to CAM. This includes support in the kernel, camcontrol(8), libcam and the mps(4) driver for SMP passthrough. The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00 to determine supported pages, and will now fetch page 0x83 in addition to page 0x80 if supported. Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is intended for SMP requests and responses. The ADVINFO is currently used to fetch cached VPD page 0x83 data from the transport layer, but is intended to be extensible to fetch other types of device-specific data. SMP-only devices are not currently represented in the CAM topology, and so the current semantics are that the SIM will route SMP CCBs to either the addressed device, if it contains an SMP target, or its parent, if it contains an SMP target. (This is noted in cam_ccb.h, since it will change later once we have the ability to have SMP-only devices in CAM's topology.) smp_all.c, smp_all.h: New helper routines for SMP. This includes SMP request building routines, response parsing routines, error decoding routines, and structure definitions for a number of SMP commands. libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality is available to userland applications. camcontrol.8, camcontrol.c: Add smp passthrough support to camcontrol. Several new subcommands are now available: 'smpcmd' functions much like 'cmd', except that it allows the user to send generic SMP commands. 'smprg' sends the SMP report general command, and displays the decoded output. It will automatically fetch extended output if it is available. 'smppc' sends the SMP phy control command, with any number of potential options. Among other things, this allows the user to reset a phy on a SAS expander, or disable a phy on an expander. 'smpmaninfo' sends the SMP report manufacturer information and displays the decoded output. 'smpphylist' displays a list of phys on an expander, and the CAM devices attached to those phys, if any. cam.h, cam.c: Add a status value for SMP errors (CAM_SMP_STATUS_ERROR). Add a missing description for CAM_SCSI_IT_NEXUS_LOST. Add support for SMP commands to cam_error_string(). cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP commands are by nature bi-directional, and we may need to support bi-directional SCSI commands later. Add the XPT_SMP_IO CCB. Since SMP commands are bi-directional, there are pointers for both the request and response. Add a fill routine for SMP CCBs. Add the XPT_GDEV_ADVINFO CCB. This is currently used to fetch cached page 0x83 data from the transport later, but is extensible to fetch many other types of data. cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. cam_xpt.c: Add support for executing XPT_SMP_IO CCBs. cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct cam_ed. scsi_all.c: Add scsi_get_sas_addr(), a function that parses VPD page 0x83 data and pulls out a SAS address. scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a prototype for scsi_get_sas_addr(). scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. scsi_xpt.c: In the SCSI probe code, first ask the device for VPD page 0x00. If any VPD pages are supported, that page is required to be implemented. Based on the response, we may probe for the serial number (page 0x80) or device id (page 0x83). Add support for the XPT_GDEV_ADVINFO CCB. sys/conf/files: Add smp_all.c. mps.c: Add support for passing in a uio in mps_map_command(), so we can map a S/G list at once. Add support for SMP passthrough commands in mps_data_cb(). SMP is a special case, because the first buffer in the S/G list is outbound and the second buffer is inbound. Add support for warning the user if the busdma code comes back with more buffers than will work for the command. This will, for example, help the user determine why an SMP command failed if busdma comes back with three buffers. mps_pci.c: Add sys/uio.h. mps_sas.c: Add the SAS address and the parent handle to the list of fields we pull from device page 0 and cache in struct mpssas_target. These are needed for SMP passthrough. Add support for the XPT_SMP_IO CCB. For now, this CCB is routed to the addressed device if it supports SMP, or to its parent if it does not and the parent does. This is necessary because CAM does not currently support SMP-only nodes in the topology. Make SMP passthrough support conditional on __FreeBSD_version >= 900026. This will make it easier to MFC this change to the driver without MFCing the CAM changes as well. mps_user.c: Un-staticize mpi_init_sge() so we can use it for the SMP passthrough code. mpsvar.h: Add a uio and iovecs into struct mps_command for SMP passthrough commands. Add a cm_max_segs field to struct mps_command so that we can warn the user if busdma comes back with too many segments. Clear the cm_reply when a command gets freed. If it is not cleared, reply frames will eventually get freed into the pool multiple times and corrupt the pool. (This fix is from scottl.) Add a prototype for mpi_init_sge(). sys/param.h: Bump __FreeBSD_version to 900026 for the for the inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO CAM CCBs.
2010-11-30 22:39:46 +00:00
#if __FreeBSD_version >= 900026
static void
mpssas_smpio_complete(struct mps_softc *sc, struct mps_command *cm)
{
MPI2_SMP_PASSTHROUGH_REPLY *rpl;
MPI2_SMP_PASSTHROUGH_REQUEST *req;
uint64_t sasaddr;
union ccb *ccb;
ccb = cm->cm_complete_data;
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
/*
* Currently there should be no way we can hit this case. It only
* happens when we have a failure to allocate chain frames, and SMP
* commands require two S/G elements only. That should be handled
* in the standard request size.
*/
if ((cm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
mps_printf(sc, "%s: cm_flags = %#x on SMP request!\n",
__func__, cm->cm_flags);
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
goto bailout;
}
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
Add Serial Management Protocol (SMP) passthrough support to CAM. This includes support in the kernel, camcontrol(8), libcam and the mps(4) driver for SMP passthrough. The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00 to determine supported pages, and will now fetch page 0x83 in addition to page 0x80 if supported. Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is intended for SMP requests and responses. The ADVINFO is currently used to fetch cached VPD page 0x83 data from the transport layer, but is intended to be extensible to fetch other types of device-specific data. SMP-only devices are not currently represented in the CAM topology, and so the current semantics are that the SIM will route SMP CCBs to either the addressed device, if it contains an SMP target, or its parent, if it contains an SMP target. (This is noted in cam_ccb.h, since it will change later once we have the ability to have SMP-only devices in CAM's topology.) smp_all.c, smp_all.h: New helper routines for SMP. This includes SMP request building routines, response parsing routines, error decoding routines, and structure definitions for a number of SMP commands. libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality is available to userland applications. camcontrol.8, camcontrol.c: Add smp passthrough support to camcontrol. Several new subcommands are now available: 'smpcmd' functions much like 'cmd', except that it allows the user to send generic SMP commands. 'smprg' sends the SMP report general command, and displays the decoded output. It will automatically fetch extended output if it is available. 'smppc' sends the SMP phy control command, with any number of potential options. Among other things, this allows the user to reset a phy on a SAS expander, or disable a phy on an expander. 'smpmaninfo' sends the SMP report manufacturer information and displays the decoded output. 'smpphylist' displays a list of phys on an expander, and the CAM devices attached to those phys, if any. cam.h, cam.c: Add a status value for SMP errors (CAM_SMP_STATUS_ERROR). Add a missing description for CAM_SCSI_IT_NEXUS_LOST. Add support for SMP commands to cam_error_string(). cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP commands are by nature bi-directional, and we may need to support bi-directional SCSI commands later. Add the XPT_SMP_IO CCB. Since SMP commands are bi-directional, there are pointers for both the request and response. Add a fill routine for SMP CCBs. Add the XPT_GDEV_ADVINFO CCB. This is currently used to fetch cached page 0x83 data from the transport later, but is extensible to fetch many other types of data. cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. cam_xpt.c: Add support for executing XPT_SMP_IO CCBs. cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct cam_ed. scsi_all.c: Add scsi_get_sas_addr(), a function that parses VPD page 0x83 data and pulls out a SAS address. scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a prototype for scsi_get_sas_addr(). scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. scsi_xpt.c: In the SCSI probe code, first ask the device for VPD page 0x00. If any VPD pages are supported, that page is required to be implemented. Based on the response, we may probe for the serial number (page 0x80) or device id (page 0x83). Add support for the XPT_GDEV_ADVINFO CCB. sys/conf/files: Add smp_all.c. mps.c: Add support for passing in a uio in mps_map_command(), so we can map a S/G list at once. Add support for SMP passthrough commands in mps_data_cb(). SMP is a special case, because the first buffer in the S/G list is outbound and the second buffer is inbound. Add support for warning the user if the busdma code comes back with more buffers than will work for the command. This will, for example, help the user determine why an SMP command failed if busdma comes back with three buffers. mps_pci.c: Add sys/uio.h. mps_sas.c: Add the SAS address and the parent handle to the list of fields we pull from device page 0 and cache in struct mpssas_target. These are needed for SMP passthrough. Add support for the XPT_SMP_IO CCB. For now, this CCB is routed to the addressed device if it supports SMP, or to its parent if it does not and the parent does. This is necessary because CAM does not currently support SMP-only nodes in the topology. Make SMP passthrough support conditional on __FreeBSD_version >= 900026. This will make it easier to MFC this change to the driver without MFCing the CAM changes as well. mps_user.c: Un-staticize mpi_init_sge() so we can use it for the SMP passthrough code. mpsvar.h: Add a uio and iovecs into struct mps_command for SMP passthrough commands. Add a cm_max_segs field to struct mps_command so that we can warn the user if busdma comes back with too many segments. Clear the cm_reply when a command gets freed. If it is not cleared, reply frames will eventually get freed into the pool multiple times and corrupt the pool. (This fix is from scottl.) Add a prototype for mpi_init_sge(). sys/param.h: Bump __FreeBSD_version to 900026 for the for the inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO CAM CCBs.
2010-11-30 22:39:46 +00:00
rpl = (MPI2_SMP_PASSTHROUGH_REPLY *)cm->cm_reply;
if (rpl == NULL) {
mps_dprint(sc, MPS_INFO, "%s: NULL cm_reply!\n", __func__);
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
goto bailout;
}
req = (MPI2_SMP_PASSTHROUGH_REQUEST *)cm->cm_req;
sasaddr = le32toh(req->SASAddress.Low);
sasaddr |= ((uint64_t)(le32toh(req->SASAddress.High))) << 32;
if ((rpl->IOCStatus & MPI2_IOCSTATUS_MASK) != MPI2_IOCSTATUS_SUCCESS ||
rpl->SASStatus != MPI2_SASSTATUS_SUCCESS) {
mps_dprint(sc, MPS_INFO, "%s: IOCStatus %04x SASStatus %02x\n",
__func__, rpl->IOCStatus, rpl->SASStatus);
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
goto bailout;
}
mps_dprint(sc, MPS_INFO, "%s: SMP request to SAS address "
"%#jx completed successfully\n", __func__,
(uintmax_t)sasaddr);
if (ccb->smpio.smp_response[2] == SMP_FR_ACCEPTED)
ccb->ccb_h.status = CAM_REQ_CMP;
else
ccb->ccb_h.status = CAM_SMP_STATUS_ERROR;
bailout:
/*
* We sync in both directions because we had DMAs in the S/G list
* in both directions.
*/
bus_dmamap_sync(sc->buffer_dmat, cm->cm_dmamap,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(sc->buffer_dmat, cm->cm_dmamap);
mps_free_command(sc, cm);
xpt_done(ccb);
}
static void
mpssas_send_smpcmd(struct mpssas_softc *sassc, union ccb *ccb, uint64_t sasaddr)
{
struct mps_command *cm;
uint8_t *request, *response;
MPI2_SMP_PASSTHROUGH_REQUEST *req;
struct mps_softc *sc;
struct sglist *sg;
int error;
sc = sassc->sc;
sg = NULL;
error = 0;
/*
* XXX We don't yet support physical addresses here.
*/
if (ccb->ccb_h.flags & (CAM_DATA_PHYS|CAM_SG_LIST_PHYS)) {
mps_printf(sc, "%s: physical addresses not supported\n",
__func__);
ccb->ccb_h.status = CAM_REQ_INVALID;
xpt_done(ccb);
return;
}
/*
* If the user wants to send an S/G list, check to make sure they
* have single buffers.
*/
if (ccb->ccb_h.flags & CAM_SCATTER_VALID) {
/*
* The chip does not support more than one buffer for the
* request or response.
*/
if ((ccb->smpio.smp_request_sglist_cnt > 1)
|| (ccb->smpio.smp_response_sglist_cnt > 1)) {
mps_printf(sc, "%s: multiple request or response "
"buffer segments not supported for SMP\n",
__func__);
ccb->ccb_h.status = CAM_REQ_INVALID;
xpt_done(ccb);
return;
}
/*
* The CAM_SCATTER_VALID flag was originally implemented
* for the XPT_SCSI_IO CCB, which only has one data pointer.
* We have two. So, just take that flag to mean that we
* might have S/G lists, and look at the S/G segment count
* to figure out whether that is the case for each individual
* buffer.
*/
if (ccb->smpio.smp_request_sglist_cnt != 0) {
bus_dma_segment_t *req_sg;
req_sg = (bus_dma_segment_t *)ccb->smpio.smp_request;
request = (uint8_t *)req_sg[0].ds_addr;
} else
request = ccb->smpio.smp_request;
if (ccb->smpio.smp_response_sglist_cnt != 0) {
bus_dma_segment_t *rsp_sg;
rsp_sg = (bus_dma_segment_t *)ccb->smpio.smp_response;
response = (uint8_t *)rsp_sg[0].ds_addr;
} else
response = ccb->smpio.smp_response;
} else {
request = ccb->smpio.smp_request;
response = ccb->smpio.smp_response;
}
cm = mps_alloc_command(sc);
if (cm == NULL) {
mps_printf(sc, "%s: cannot allocate command\n", __func__);
ccb->ccb_h.status = CAM_RESRC_UNAVAIL;
xpt_done(ccb);
return;
}
req = (MPI2_SMP_PASSTHROUGH_REQUEST *)cm->cm_req;
bzero(req, sizeof(*req));
req->Function = MPI2_FUNCTION_SMP_PASSTHROUGH;
/* Allow the chip to use any route to this SAS address. */
req->PhysicalPort = 0xff;
req->RequestDataLength = ccb->smpio.smp_request_len;
req->SGLFlags =
MPI2_SGLFLAGS_SYSTEM_ADDRESS_SPACE | MPI2_SGLFLAGS_SGL_TYPE_MPI;
mps_dprint(sc, MPS_INFO, "%s: sending SMP request to SAS "
"address %#jx\n", __func__, (uintmax_t)sasaddr);
mpi_init_sge(cm, req, &req->SGL);
/*
* Set up a uio to pass into mps_map_command(). This allows us to
* do one map command, and one busdma call in there.
*/
cm->cm_uio.uio_iov = cm->cm_iovec;
cm->cm_uio.uio_iovcnt = 2;
cm->cm_uio.uio_segflg = UIO_SYSSPACE;
/*
* The read/write flag isn't used by busdma, but set it just in
* case. This isn't exactly accurate, either, since we're going in
* both directions.
*/
cm->cm_uio.uio_rw = UIO_WRITE;
cm->cm_iovec[0].iov_base = request;
cm->cm_iovec[0].iov_len = req->RequestDataLength;
cm->cm_iovec[1].iov_base = response;
cm->cm_iovec[1].iov_len = ccb->smpio.smp_response_len;
cm->cm_uio.uio_resid = cm->cm_iovec[0].iov_len +
cm->cm_iovec[1].iov_len;
/*
* Trigger a warning message in mps_data_cb() for the user if we
* wind up exceeding two S/G segments. The chip expects one
* segment for the request and another for the response.
*/
cm->cm_max_segs = 2;
cm->cm_desc.Default.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_DEFAULT_TYPE;
cm->cm_complete = mpssas_smpio_complete;
cm->cm_complete_data = ccb;
/*
* Tell the mapping code that we're using a uio, and that this is
* an SMP passthrough request. There is a little special-case
* logic there (in mps_data_cb()) to handle the bidirectional
* transfer.
*/
cm->cm_flags |= MPS_CM_FLAGS_USE_UIO | MPS_CM_FLAGS_SMP_PASS |
MPS_CM_FLAGS_DATAIN | MPS_CM_FLAGS_DATAOUT;
/* The chip data format is little endian. */
req->SASAddress.High = htole32(sasaddr >> 32);
req->SASAddress.Low = htole32(sasaddr);
/*
* XXX Note that we don't have a timeout/abort mechanism here.
* From the manual, it looks like task management requests only
* work for SCSI IO and SATA passthrough requests. We may need to
* have a mechanism to retry requests in the event of a chip reset
* at least. Hopefully the chip will insure that any errors short
* of that are relayed back to the driver.
*/
error = mps_map_command(sc, cm);
if ((error != 0) && (error != EINPROGRESS)) {
mps_printf(sc, "%s: error %d returned from mps_map_command()\n",
__func__, error);
goto bailout_error;
}
return;
bailout_error:
mps_free_command(sc, cm);
ccb->ccb_h.status = CAM_RESRC_UNAVAIL;
xpt_done(ccb);
return;
}
static void
mpssas_action_smpio(struct mpssas_softc *sassc, union ccb *ccb)
{
struct mps_softc *sc;
struct mpssas_target *targ;
uint64_t sasaddr = 0;
sc = sassc->sc;
/*
* Make sure the target exists.
*/
targ = &sassc->targets[ccb->ccb_h.target_id];
if (targ->handle == 0x0) {
mps_printf(sc, "%s: target %d does not exist!\n", __func__,
ccb->ccb_h.target_id);
ccb->ccb_h.status = CAM_SEL_TIMEOUT;
xpt_done(ccb);
return;
}
/*
* If this device has an embedded SMP target, we'll talk to it
* directly.
* figure out what the expander's address is.
*/
if ((targ->devinfo & MPI2_SAS_DEVICE_INFO_SMP_TARGET) != 0)
sasaddr = targ->sasaddr;
/*
* If we don't have a SAS address for the expander yet, try
* grabbing it from the page 0x83 information cached in the
* transport layer for this target. LSI expanders report the
* expander SAS address as the port-associated SAS address in
* Inquiry VPD page 0x83. Maxim expanders don't report it in page
* 0x83.
*
* XXX KDM disable this for now, but leave it commented out so that
* it is obvious that this is another possible way to get the SAS
* address.
*
* The parent handle method below is a little more reliable, and
* the other benefit is that it works for devices other than SES
* devices. So you can send a SMP request to a da(4) device and it
* will get routed to the expander that device is attached to.
* (Assuming the da(4) device doesn't contain an SMP target...)
*/
#if 0
if (sasaddr == 0)
sasaddr = xpt_path_sas_addr(ccb->ccb_h.path);
#endif
/*
* If we still don't have a SAS address for the expander, look for
* the parent device of this device, which is probably the expander.
*/
if (sasaddr == 0) {
#ifdef OLD_MPS_PROBE
Add Serial Management Protocol (SMP) passthrough support to CAM. This includes support in the kernel, camcontrol(8), libcam and the mps(4) driver for SMP passthrough. The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00 to determine supported pages, and will now fetch page 0x83 in addition to page 0x80 if supported. Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is intended for SMP requests and responses. The ADVINFO is currently used to fetch cached VPD page 0x83 data from the transport layer, but is intended to be extensible to fetch other types of device-specific data. SMP-only devices are not currently represented in the CAM topology, and so the current semantics are that the SIM will route SMP CCBs to either the addressed device, if it contains an SMP target, or its parent, if it contains an SMP target. (This is noted in cam_ccb.h, since it will change later once we have the ability to have SMP-only devices in CAM's topology.) smp_all.c, smp_all.h: New helper routines for SMP. This includes SMP request building routines, response parsing routines, error decoding routines, and structure definitions for a number of SMP commands. libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality is available to userland applications. camcontrol.8, camcontrol.c: Add smp passthrough support to camcontrol. Several new subcommands are now available: 'smpcmd' functions much like 'cmd', except that it allows the user to send generic SMP commands. 'smprg' sends the SMP report general command, and displays the decoded output. It will automatically fetch extended output if it is available. 'smppc' sends the SMP phy control command, with any number of potential options. Among other things, this allows the user to reset a phy on a SAS expander, or disable a phy on an expander. 'smpmaninfo' sends the SMP report manufacturer information and displays the decoded output. 'smpphylist' displays a list of phys on an expander, and the CAM devices attached to those phys, if any. cam.h, cam.c: Add a status value for SMP errors (CAM_SMP_STATUS_ERROR). Add a missing description for CAM_SCSI_IT_NEXUS_LOST. Add support for SMP commands to cam_error_string(). cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP commands are by nature bi-directional, and we may need to support bi-directional SCSI commands later. Add the XPT_SMP_IO CCB. Since SMP commands are bi-directional, there are pointers for both the request and response. Add a fill routine for SMP CCBs. Add the XPT_GDEV_ADVINFO CCB. This is currently used to fetch cached page 0x83 data from the transport later, but is extensible to fetch many other types of data. cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. cam_xpt.c: Add support for executing XPT_SMP_IO CCBs. cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct cam_ed. scsi_all.c: Add scsi_get_sas_addr(), a function that parses VPD page 0x83 data and pulls out a SAS address. scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a prototype for scsi_get_sas_addr(). scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. scsi_xpt.c: In the SCSI probe code, first ask the device for VPD page 0x00. If any VPD pages are supported, that page is required to be implemented. Based on the response, we may probe for the serial number (page 0x80) or device id (page 0x83). Add support for the XPT_GDEV_ADVINFO CCB. sys/conf/files: Add smp_all.c. mps.c: Add support for passing in a uio in mps_map_command(), so we can map a S/G list at once. Add support for SMP passthrough commands in mps_data_cb(). SMP is a special case, because the first buffer in the S/G list is outbound and the second buffer is inbound. Add support for warning the user if the busdma code comes back with more buffers than will work for the command. This will, for example, help the user determine why an SMP command failed if busdma comes back with three buffers. mps_pci.c: Add sys/uio.h. mps_sas.c: Add the SAS address and the parent handle to the list of fields we pull from device page 0 and cache in struct mpssas_target. These are needed for SMP passthrough. Add support for the XPT_SMP_IO CCB. For now, this CCB is routed to the addressed device if it supports SMP, or to its parent if it does not and the parent does. This is necessary because CAM does not currently support SMP-only nodes in the topology. Make SMP passthrough support conditional on __FreeBSD_version >= 900026. This will make it easier to MFC this change to the driver without MFCing the CAM changes as well. mps_user.c: Un-staticize mpi_init_sge() so we can use it for the SMP passthrough code. mpsvar.h: Add a uio and iovecs into struct mps_command for SMP passthrough commands. Add a cm_max_segs field to struct mps_command so that we can warn the user if busdma comes back with too many segments. Clear the cm_reply when a command gets freed. If it is not cleared, reply frames will eventually get freed into the pool multiple times and corrupt the pool. (This fix is from scottl.) Add a prototype for mpi_init_sge(). sys/param.h: Bump __FreeBSD_version to 900026 for the for the inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO CAM CCBs.
2010-11-30 22:39:46 +00:00
struct mpssas_target *parent_target;
#endif
Add Serial Management Protocol (SMP) passthrough support to CAM. This includes support in the kernel, camcontrol(8), libcam and the mps(4) driver for SMP passthrough. The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00 to determine supported pages, and will now fetch page 0x83 in addition to page 0x80 if supported. Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is intended for SMP requests and responses. The ADVINFO is currently used to fetch cached VPD page 0x83 data from the transport layer, but is intended to be extensible to fetch other types of device-specific data. SMP-only devices are not currently represented in the CAM topology, and so the current semantics are that the SIM will route SMP CCBs to either the addressed device, if it contains an SMP target, or its parent, if it contains an SMP target. (This is noted in cam_ccb.h, since it will change later once we have the ability to have SMP-only devices in CAM's topology.) smp_all.c, smp_all.h: New helper routines for SMP. This includes SMP request building routines, response parsing routines, error decoding routines, and structure definitions for a number of SMP commands. libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality is available to userland applications. camcontrol.8, camcontrol.c: Add smp passthrough support to camcontrol. Several new subcommands are now available: 'smpcmd' functions much like 'cmd', except that it allows the user to send generic SMP commands. 'smprg' sends the SMP report general command, and displays the decoded output. It will automatically fetch extended output if it is available. 'smppc' sends the SMP phy control command, with any number of potential options. Among other things, this allows the user to reset a phy on a SAS expander, or disable a phy on an expander. 'smpmaninfo' sends the SMP report manufacturer information and displays the decoded output. 'smpphylist' displays a list of phys on an expander, and the CAM devices attached to those phys, if any. cam.h, cam.c: Add a status value for SMP errors (CAM_SMP_STATUS_ERROR). Add a missing description for CAM_SCSI_IT_NEXUS_LOST. Add support for SMP commands to cam_error_string(). cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP commands are by nature bi-directional, and we may need to support bi-directional SCSI commands later. Add the XPT_SMP_IO CCB. Since SMP commands are bi-directional, there are pointers for both the request and response. Add a fill routine for SMP CCBs. Add the XPT_GDEV_ADVINFO CCB. This is currently used to fetch cached page 0x83 data from the transport later, but is extensible to fetch many other types of data. cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. cam_xpt.c: Add support for executing XPT_SMP_IO CCBs. cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct cam_ed. scsi_all.c: Add scsi_get_sas_addr(), a function that parses VPD page 0x83 data and pulls out a SAS address. scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a prototype for scsi_get_sas_addr(). scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. scsi_xpt.c: In the SCSI probe code, first ask the device for VPD page 0x00. If any VPD pages are supported, that page is required to be implemented. Based on the response, we may probe for the serial number (page 0x80) or device id (page 0x83). Add support for the XPT_GDEV_ADVINFO CCB. sys/conf/files: Add smp_all.c. mps.c: Add support for passing in a uio in mps_map_command(), so we can map a S/G list at once. Add support for SMP passthrough commands in mps_data_cb(). SMP is a special case, because the first buffer in the S/G list is outbound and the second buffer is inbound. Add support for warning the user if the busdma code comes back with more buffers than will work for the command. This will, for example, help the user determine why an SMP command failed if busdma comes back with three buffers. mps_pci.c: Add sys/uio.h. mps_sas.c: Add the SAS address and the parent handle to the list of fields we pull from device page 0 and cache in struct mpssas_target. These are needed for SMP passthrough. Add support for the XPT_SMP_IO CCB. For now, this CCB is routed to the addressed device if it supports SMP, or to its parent if it does not and the parent does. This is necessary because CAM does not currently support SMP-only nodes in the topology. Make SMP passthrough support conditional on __FreeBSD_version >= 900026. This will make it easier to MFC this change to the driver without MFCing the CAM changes as well. mps_user.c: Un-staticize mpi_init_sge() so we can use it for the SMP passthrough code. mpsvar.h: Add a uio and iovecs into struct mps_command for SMP passthrough commands. Add a cm_max_segs field to struct mps_command so that we can warn the user if busdma comes back with too many segments. Clear the cm_reply when a command gets freed. If it is not cleared, reply frames will eventually get freed into the pool multiple times and corrupt the pool. (This fix is from scottl.) Add a prototype for mpi_init_sge(). sys/param.h: Bump __FreeBSD_version to 900026 for the for the inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO CAM CCBs.
2010-11-30 22:39:46 +00:00
if (targ->parent_handle == 0x0) {
mps_printf(sc, "%s: handle %d does not have a valid "
"parent handle!\n", __func__, targ->handle);
ccb->ccb_h.status = CAM_REQ_INVALID;
goto bailout;
}
#ifdef OLD_MPS_PROBE
parent_target = mpssas_find_target_by_handle(sassc, 0,
targ->parent_handle);
Add Serial Management Protocol (SMP) passthrough support to CAM. This includes support in the kernel, camcontrol(8), libcam and the mps(4) driver for SMP passthrough. The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00 to determine supported pages, and will now fetch page 0x83 in addition to page 0x80 if supported. Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is intended for SMP requests and responses. The ADVINFO is currently used to fetch cached VPD page 0x83 data from the transport layer, but is intended to be extensible to fetch other types of device-specific data. SMP-only devices are not currently represented in the CAM topology, and so the current semantics are that the SIM will route SMP CCBs to either the addressed device, if it contains an SMP target, or its parent, if it contains an SMP target. (This is noted in cam_ccb.h, since it will change later once we have the ability to have SMP-only devices in CAM's topology.) smp_all.c, smp_all.h: New helper routines for SMP. This includes SMP request building routines, response parsing routines, error decoding routines, and structure definitions for a number of SMP commands. libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality is available to userland applications. camcontrol.8, camcontrol.c: Add smp passthrough support to camcontrol. Several new subcommands are now available: 'smpcmd' functions much like 'cmd', except that it allows the user to send generic SMP commands. 'smprg' sends the SMP report general command, and displays the decoded output. It will automatically fetch extended output if it is available. 'smppc' sends the SMP phy control command, with any number of potential options. Among other things, this allows the user to reset a phy on a SAS expander, or disable a phy on an expander. 'smpmaninfo' sends the SMP report manufacturer information and displays the decoded output. 'smpphylist' displays a list of phys on an expander, and the CAM devices attached to those phys, if any. cam.h, cam.c: Add a status value for SMP errors (CAM_SMP_STATUS_ERROR). Add a missing description for CAM_SCSI_IT_NEXUS_LOST. Add support for SMP commands to cam_error_string(). cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP commands are by nature bi-directional, and we may need to support bi-directional SCSI commands later. Add the XPT_SMP_IO CCB. Since SMP commands are bi-directional, there are pointers for both the request and response. Add a fill routine for SMP CCBs. Add the XPT_GDEV_ADVINFO CCB. This is currently used to fetch cached page 0x83 data from the transport later, but is extensible to fetch many other types of data. cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. cam_xpt.c: Add support for executing XPT_SMP_IO CCBs. cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct cam_ed. scsi_all.c: Add scsi_get_sas_addr(), a function that parses VPD page 0x83 data and pulls out a SAS address. scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a prototype for scsi_get_sas_addr(). scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. scsi_xpt.c: In the SCSI probe code, first ask the device for VPD page 0x00. If any VPD pages are supported, that page is required to be implemented. Based on the response, we may probe for the serial number (page 0x80) or device id (page 0x83). Add support for the XPT_GDEV_ADVINFO CCB. sys/conf/files: Add smp_all.c. mps.c: Add support for passing in a uio in mps_map_command(), so we can map a S/G list at once. Add support for SMP passthrough commands in mps_data_cb(). SMP is a special case, because the first buffer in the S/G list is outbound and the second buffer is inbound. Add support for warning the user if the busdma code comes back with more buffers than will work for the command. This will, for example, help the user determine why an SMP command failed if busdma comes back with three buffers. mps_pci.c: Add sys/uio.h. mps_sas.c: Add the SAS address and the parent handle to the list of fields we pull from device page 0 and cache in struct mpssas_target. These are needed for SMP passthrough. Add support for the XPT_SMP_IO CCB. For now, this CCB is routed to the addressed device if it supports SMP, or to its parent if it does not and the parent does. This is necessary because CAM does not currently support SMP-only nodes in the topology. Make SMP passthrough support conditional on __FreeBSD_version >= 900026. This will make it easier to MFC this change to the driver without MFCing the CAM changes as well. mps_user.c: Un-staticize mpi_init_sge() so we can use it for the SMP passthrough code. mpsvar.h: Add a uio and iovecs into struct mps_command for SMP passthrough commands. Add a cm_max_segs field to struct mps_command so that we can warn the user if busdma comes back with too many segments. Clear the cm_reply when a command gets freed. If it is not cleared, reply frames will eventually get freed into the pool multiple times and corrupt the pool. (This fix is from scottl.) Add a prototype for mpi_init_sge(). sys/param.h: Bump __FreeBSD_version to 900026 for the for the inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO CAM CCBs.
2010-11-30 22:39:46 +00:00
if (parent_target == NULL) {
mps_printf(sc, "%s: handle %d does not have a valid "
"parent target!\n", __func__, targ->handle);
ccb->ccb_h.status = CAM_REQ_INVALID;
goto bailout;
}
if ((parent_target->devinfo &
MPI2_SAS_DEVICE_INFO_SMP_TARGET) == 0) {
mps_printf(sc, "%s: handle %d parent %d does not "
"have an SMP target!\n", __func__,
targ->handle, parent_target->handle);
ccb->ccb_h.status = CAM_REQ_INVALID;
goto bailout;
}
sasaddr = parent_target->sasaddr;
#else /* OLD_MPS_PROBE */
if ((targ->parent_devinfo &
MPI2_SAS_DEVICE_INFO_SMP_TARGET) == 0) {
mps_printf(sc, "%s: handle %d parent %d does not "
"have an SMP target!\n", __func__,
targ->handle, targ->parent_handle);
ccb->ccb_h.status = CAM_REQ_INVALID;
goto bailout;
}
if (targ->parent_sasaddr == 0x0) {
mps_printf(sc, "%s: handle %d parent handle %d does "
"not have a valid SAS address!\n",
__func__, targ->handle, targ->parent_handle);
ccb->ccb_h.status = CAM_REQ_INVALID;
goto bailout;
}
sasaddr = targ->parent_sasaddr;
#endif /* OLD_MPS_PROBE */
Add Serial Management Protocol (SMP) passthrough support to CAM. This includes support in the kernel, camcontrol(8), libcam and the mps(4) driver for SMP passthrough. The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00 to determine supported pages, and will now fetch page 0x83 in addition to page 0x80 if supported. Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is intended for SMP requests and responses. The ADVINFO is currently used to fetch cached VPD page 0x83 data from the transport layer, but is intended to be extensible to fetch other types of device-specific data. SMP-only devices are not currently represented in the CAM topology, and so the current semantics are that the SIM will route SMP CCBs to either the addressed device, if it contains an SMP target, or its parent, if it contains an SMP target. (This is noted in cam_ccb.h, since it will change later once we have the ability to have SMP-only devices in CAM's topology.) smp_all.c, smp_all.h: New helper routines for SMP. This includes SMP request building routines, response parsing routines, error decoding routines, and structure definitions for a number of SMP commands. libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality is available to userland applications. camcontrol.8, camcontrol.c: Add smp passthrough support to camcontrol. Several new subcommands are now available: 'smpcmd' functions much like 'cmd', except that it allows the user to send generic SMP commands. 'smprg' sends the SMP report general command, and displays the decoded output. It will automatically fetch extended output if it is available. 'smppc' sends the SMP phy control command, with any number of potential options. Among other things, this allows the user to reset a phy on a SAS expander, or disable a phy on an expander. 'smpmaninfo' sends the SMP report manufacturer information and displays the decoded output. 'smpphylist' displays a list of phys on an expander, and the CAM devices attached to those phys, if any. cam.h, cam.c: Add a status value for SMP errors (CAM_SMP_STATUS_ERROR). Add a missing description for CAM_SCSI_IT_NEXUS_LOST. Add support for SMP commands to cam_error_string(). cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP commands are by nature bi-directional, and we may need to support bi-directional SCSI commands later. Add the XPT_SMP_IO CCB. Since SMP commands are bi-directional, there are pointers for both the request and response. Add a fill routine for SMP CCBs. Add the XPT_GDEV_ADVINFO CCB. This is currently used to fetch cached page 0x83 data from the transport later, but is extensible to fetch many other types of data. cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. cam_xpt.c: Add support for executing XPT_SMP_IO CCBs. cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct cam_ed. scsi_all.c: Add scsi_get_sas_addr(), a function that parses VPD page 0x83 data and pulls out a SAS address. scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a prototype for scsi_get_sas_addr(). scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. scsi_xpt.c: In the SCSI probe code, first ask the device for VPD page 0x00. If any VPD pages are supported, that page is required to be implemented. Based on the response, we may probe for the serial number (page 0x80) or device id (page 0x83). Add support for the XPT_GDEV_ADVINFO CCB. sys/conf/files: Add smp_all.c. mps.c: Add support for passing in a uio in mps_map_command(), so we can map a S/G list at once. Add support for SMP passthrough commands in mps_data_cb(). SMP is a special case, because the first buffer in the S/G list is outbound and the second buffer is inbound. Add support for warning the user if the busdma code comes back with more buffers than will work for the command. This will, for example, help the user determine why an SMP command failed if busdma comes back with three buffers. mps_pci.c: Add sys/uio.h. mps_sas.c: Add the SAS address and the parent handle to the list of fields we pull from device page 0 and cache in struct mpssas_target. These are needed for SMP passthrough. Add support for the XPT_SMP_IO CCB. For now, this CCB is routed to the addressed device if it supports SMP, or to its parent if it does not and the parent does. This is necessary because CAM does not currently support SMP-only nodes in the topology. Make SMP passthrough support conditional on __FreeBSD_version >= 900026. This will make it easier to MFC this change to the driver without MFCing the CAM changes as well. mps_user.c: Un-staticize mpi_init_sge() so we can use it for the SMP passthrough code. mpsvar.h: Add a uio and iovecs into struct mps_command for SMP passthrough commands. Add a cm_max_segs field to struct mps_command so that we can warn the user if busdma comes back with too many segments. Clear the cm_reply when a command gets freed. If it is not cleared, reply frames will eventually get freed into the pool multiple times and corrupt the pool. (This fix is from scottl.) Add a prototype for mpi_init_sge(). sys/param.h: Bump __FreeBSD_version to 900026 for the for the inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO CAM CCBs.
2010-11-30 22:39:46 +00:00
}
if (sasaddr == 0) {
mps_printf(sc, "%s: unable to find SAS address for handle %d\n",
__func__, targ->handle);
ccb->ccb_h.status = CAM_REQ_INVALID;
goto bailout;
}
mpssas_send_smpcmd(sassc, ccb, sasaddr);
return;
bailout:
xpt_done(ccb);
}
#endif //__FreeBSD_version >= 900026
Add Serial Management Protocol (SMP) passthrough support to CAM. This includes support in the kernel, camcontrol(8), libcam and the mps(4) driver for SMP passthrough. The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00 to determine supported pages, and will now fetch page 0x83 in addition to page 0x80 if supported. Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is intended for SMP requests and responses. The ADVINFO is currently used to fetch cached VPD page 0x83 data from the transport layer, but is intended to be extensible to fetch other types of device-specific data. SMP-only devices are not currently represented in the CAM topology, and so the current semantics are that the SIM will route SMP CCBs to either the addressed device, if it contains an SMP target, or its parent, if it contains an SMP target. (This is noted in cam_ccb.h, since it will change later once we have the ability to have SMP-only devices in CAM's topology.) smp_all.c, smp_all.h: New helper routines for SMP. This includes SMP request building routines, response parsing routines, error decoding routines, and structure definitions for a number of SMP commands. libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality is available to userland applications. camcontrol.8, camcontrol.c: Add smp passthrough support to camcontrol. Several new subcommands are now available: 'smpcmd' functions much like 'cmd', except that it allows the user to send generic SMP commands. 'smprg' sends the SMP report general command, and displays the decoded output. It will automatically fetch extended output if it is available. 'smppc' sends the SMP phy control command, with any number of potential options. Among other things, this allows the user to reset a phy on a SAS expander, or disable a phy on an expander. 'smpmaninfo' sends the SMP report manufacturer information and displays the decoded output. 'smpphylist' displays a list of phys on an expander, and the CAM devices attached to those phys, if any. cam.h, cam.c: Add a status value for SMP errors (CAM_SMP_STATUS_ERROR). Add a missing description for CAM_SCSI_IT_NEXUS_LOST. Add support for SMP commands to cam_error_string(). cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP commands are by nature bi-directional, and we may need to support bi-directional SCSI commands later. Add the XPT_SMP_IO CCB. Since SMP commands are bi-directional, there are pointers for both the request and response. Add a fill routine for SMP CCBs. Add the XPT_GDEV_ADVINFO CCB. This is currently used to fetch cached page 0x83 data from the transport later, but is extensible to fetch many other types of data. cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. cam_xpt.c: Add support for executing XPT_SMP_IO CCBs. cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct cam_ed. scsi_all.c: Add scsi_get_sas_addr(), a function that parses VPD page 0x83 data and pulls out a SAS address. scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a prototype for scsi_get_sas_addr(). scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. scsi_xpt.c: In the SCSI probe code, first ask the device for VPD page 0x00. If any VPD pages are supported, that page is required to be implemented. Based on the response, we may probe for the serial number (page 0x80) or device id (page 0x83). Add support for the XPT_GDEV_ADVINFO CCB. sys/conf/files: Add smp_all.c. mps.c: Add support for passing in a uio in mps_map_command(), so we can map a S/G list at once. Add support for SMP passthrough commands in mps_data_cb(). SMP is a special case, because the first buffer in the S/G list is outbound and the second buffer is inbound. Add support for warning the user if the busdma code comes back with more buffers than will work for the command. This will, for example, help the user determine why an SMP command failed if busdma comes back with three buffers. mps_pci.c: Add sys/uio.h. mps_sas.c: Add the SAS address and the parent handle to the list of fields we pull from device page 0 and cache in struct mpssas_target. These are needed for SMP passthrough. Add support for the XPT_SMP_IO CCB. For now, this CCB is routed to the addressed device if it supports SMP, or to its parent if it does not and the parent does. This is necessary because CAM does not currently support SMP-only nodes in the topology. Make SMP passthrough support conditional on __FreeBSD_version >= 900026. This will make it easier to MFC this change to the driver without MFCing the CAM changes as well. mps_user.c: Un-staticize mpi_init_sge() so we can use it for the SMP passthrough code. mpsvar.h: Add a uio and iovecs into struct mps_command for SMP passthrough commands. Add a cm_max_segs field to struct mps_command so that we can warn the user if busdma comes back with too many segments. Clear the cm_reply when a command gets freed. If it is not cleared, reply frames will eventually get freed into the pool multiple times and corrupt the pool. (This fix is from scottl.) Add a prototype for mpi_init_sge(). sys/param.h: Bump __FreeBSD_version to 900026 for the for the inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO CAM CCBs.
2010-11-30 22:39:46 +00:00
static void
mpssas_action_resetdev(struct mpssas_softc *sassc, union ccb *ccb)
{
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
struct mps_softc *sc;
struct mps_command *tm;
struct mpssas_target *targ;
mps_dprint(sassc->sc, MPS_TRACE, __func__);
mtx_assert(&sassc->sc->mps_mtx, MA_OWNED);
sc = sassc->sc;
tm = mps_alloc_command(sc);
if (tm == NULL) {
mps_printf(sc, "comand alloc failure in mpssas_action_resetdev\n");
ccb->ccb_h.status = CAM_RESRC_UNAVAIL;
xpt_done(ccb);
return;
}
targ = &sassc->targets[ccb->ccb_h.target_id];
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)tm->cm_req;
req->DevHandle = targ->handle;
req->Function = MPI2_FUNCTION_SCSI_TASK_MGMT;
req->TaskType = MPI2_SCSITASKMGMT_TASKTYPE_TARGET_RESET;
/* SAS Hard Link Reset / SATA Link Reset */
req->MsgFlags = MPI2_SCSITASKMGMT_MSGFLAGS_LINK_RESET;
tm->cm_data = NULL;
tm->cm_desc.HighPriority.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_HIGH_PRIORITY;
tm->cm_complete = mpssas_resetdev_complete;
tm->cm_complete_data = ccb;
mps_map_command(sc, tm);
}
static void
mpssas_resetdev_complete(struct mps_softc *sc, struct mps_command *tm)
{
MPI2_SCSI_TASK_MANAGE_REPLY *resp;
union ccb *ccb;
mps_dprint(sc, MPS_TRACE, __func__);
mtx_assert(&sc->mps_mtx, MA_OWNED);
resp = (MPI2_SCSI_TASK_MANAGE_REPLY *)tm->cm_reply;
ccb = tm->cm_complete_data;
/*
* Currently there should be no way we can hit this case. It only
* happens when we have a failure to allocate chain frames, and
* task management commands don't have S/G lists.
*/
if ((tm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)tm->cm_req;
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
mps_printf(sc, "%s: cm_flags = %#x for reset of handle %#04x! "
"This should not happen!\n", __func__, tm->cm_flags,
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
req->DevHandle);
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
goto bailout;
}
printf("%s: IOCStatus = 0x%x ResponseCode = 0x%x\n", __func__,
resp->IOCStatus, resp->ResponseCode);
if (resp->ResponseCode == MPI2_SCSITASKMGMT_RSP_TM_COMPLETE) {
ccb->ccb_h.status = CAM_REQ_CMP;
mpssas_announce_reset(sc, AC_SENT_BDR, tm->cm_targ->tid,
CAM_LUN_WILDCARD);
}
else
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
Fix several issues with the mps(4) driver. When the driver ran out of DMA chaining buffers, it kept the timeout for the I/O, and I/O would stall. The driver was not freezing the device queue on errors. mps.c: Pull command completion logic into a separate function, and call the callback/wakeup for commands that are never sent due to lack of chain buffers. Add a number of extra diagnostic sysctl variables. Handle pre-hardware errors for configuration I/O. This doesn't panic the system, but it will fail the configuration I/O and there is no retry mechanism. So the device probe will not succeed. This should be a very uncommon situation, however. mps_sas.c: Freeze the SIM queue when we run out of chain buffers, and unfreeze it when more commands complete. Freeze the device queue when errors occur, so that CAM can insure proper command ordering. Report pre-hardware errors for task management commands. In general, that shouldn't be possible because task management commands don't have S/G lists, and that is currently the only error path before we get to the hardware. Handle pre-hardware errors (like out of chain elements) for SMP requests. That shouldn't happen either, since we should have enough space for two S/G elements in the standard request. For commands that end with MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them with CAM_REQUEUE_REQ to retry them unconditionally. These seem to be related to back end, transport related problems that are hopefully transient. We don't want to go through the retry count for something that is not a permanent error. Keep track of the number of outstanding I/Os. mpsvar.h: Track the number of free chain elements. Add variables for the number of outstanding I/Os, and I/O high water mark. Add variables to track the number of free chain buffers and the chain low water mark, as well as the number of chain allocation failures. Add I/O state flags and an attach done flag. MFC after: 3 days
2011-02-18 17:06:06 +00:00
bailout:
mpssas_free_tm(sc, tm);
xpt_done(ccb);
}
static void
mpssas_poll(struct cam_sim *sim)
{
struct mpssas_softc *sassc;
sassc = cam_sim_softc(sim);
if (sassc->sc->mps_debug & MPS_TRACE) {
/* frequent debug messages during a panic just slow
* everything down too much.
*/
mps_printf(sassc->sc, "%s clearing MPS_TRACE\n", __func__);
sassc->sc->mps_debug &= ~MPS_TRACE;
}
mps_intr_locked(sassc->sc);
}
static void
mpssas_rescan_done(struct cam_periph *periph, union ccb *done_ccb)
{
struct mpssas_softc *sassc;
char path_str[64];
if (done_ccb == NULL)
return;
sassc = (struct mpssas_softc *)done_ccb->ccb_h.ppriv_ptr1;
mtx_assert(&sassc->sc->mps_mtx, MA_OWNED);
xpt_path_string(done_ccb->ccb_h.path, path_str, sizeof(path_str));
mps_dprint(sassc->sc, MPS_INFO, "Completing rescan for %s\n", path_str);
xpt_free_path(done_ccb->ccb_h.path);
xpt_free_ccb(done_ccb);
#if __FreeBSD_version < 1000006
/*
* Before completing scan, get EEDP stuff for all of the existing
* targets.
*/
mpssas_check_eedp(sassc);
#endif
}
/* thread to handle bus rescans */
static void
mpssas_scanner_thread(void *arg)
{
struct mpssas_softc *sassc;
struct mps_softc *sc;
union ccb *ccb;
sassc = (struct mpssas_softc *)arg;
sc = sassc->sc;
mps_dprint(sc, MPS_TRACE, "%s\n", __func__);
mps_lock(sc);
for (;;) {
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
/* Sleep for 1 second and check the queue status*/
msleep(&sassc->ccb_scanq, &sc->mps_mtx, PRIBIO,
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
"mps_scanq", 1 * hz);
if (sassc->flags & MPSSAS_SHUTDOWN) {
mps_dprint(sc, MPS_TRACE, "Scanner shutting down\n");
break;
}
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
next_work:
// Get first work.
ccb = (union ccb *)TAILQ_FIRST(&sassc->ccb_scanq);
if (ccb == NULL)
continue;
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
// Got first work.
TAILQ_REMOVE(&sassc->ccb_scanq, &ccb->ccb_h, sim_links.tqe);
xpt_action(ccb);
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
if (sassc->flags & MPSSAS_SHUTDOWN) {
mps_dprint(sc, MPS_TRACE, "Scanner shutting down\n");
break;
}
goto next_work;
}
sassc->flags &= ~MPSSAS_SCANTHREAD;
wakeup(&sassc->flags);
mps_unlock(sc);
mps_dprint(sc, MPS_TRACE, "Scanner exiting\n");
mps_kproc_exit(0);
}
static void
mpssas_rescan(struct mpssas_softc *sassc, union ccb *ccb)
{
char path_str[64];
mps_dprint(sassc->sc, MPS_TRACE, "%s\n", __func__);
mtx_assert(&sassc->sc->mps_mtx, MA_OWNED);
if (ccb == NULL)
return;
xpt_path_string(ccb->ccb_h.path, path_str, sizeof(path_str));
mps_dprint(sassc->sc, MPS_INFO, "Queueing rescan for %s\n", path_str);
/* Prepare request */
ccb->ccb_h.ppriv_ptr1 = sassc;
ccb->ccb_h.cbfcnp = mpssas_rescan_done;
xpt_setup_ccb(&ccb->ccb_h, ccb->ccb_h.path, MPS_PRIORITY_XPT);
TAILQ_INSERT_TAIL(&sassc->ccb_scanq, &ccb->ccb_h, sim_links.tqe);
wakeup(&sassc->ccb_scanq);
}
#if __FreeBSD_version >= 1000006
static void
mpssas_async(void *callback_arg, uint32_t code, struct cam_path *path,
void *arg)
{
struct mps_softc *sc;
sc = (struct mps_softc *)callback_arg;
switch (code) {
case AC_ADVINFO_CHANGED: {
struct mpssas_target *target;
struct mpssas_softc *sassc;
struct scsi_read_capacity_data_long rcap_buf;
struct ccb_dev_advinfo cdai;
struct mpssas_lun *lun;
lun_id_t lunid;
int found_lun;
uintptr_t buftype;
buftype = (uintptr_t)arg;
found_lun = 0;
sassc = sc->sassc;
/*
* We're only interested in read capacity data changes.
*/
if (buftype != CDAI_TYPE_RCAPLONG)
break;
/*
* We're only interested in devices that are attached to
* this controller.
*/
if (xpt_path_path_id(path) != sassc->sim->path_id)
break;
/*
* We should have a handle for this, but check to make sure.
*/
target = &sassc->targets[xpt_path_target_id(path)];
if (target->handle == 0)
break;
lunid = xpt_path_lun_id(path);
SLIST_FOREACH(lun, &target->luns, lun_link) {
if (lun->lun_id == lunid) {
found_lun = 1;
break;
}
}
if (found_lun == 0) {
lun = malloc(sizeof(struct mpssas_lun), M_MPT2,
M_NOWAIT | M_ZERO);
if (lun == NULL) {
mps_dprint(sc, MPS_FAULT, "Unable to alloc "
"LUN for EEDP support.\n");
break;
}
lun->lun_id = lunid;
SLIST_INSERT_HEAD(&target->luns, lun, lun_link);
}
bzero(&rcap_buf, sizeof(rcap_buf));
xpt_setup_ccb(&cdai.ccb_h, path, CAM_PRIORITY_NORMAL);
cdai.ccb_h.func_code = XPT_DEV_ADVINFO;
cdai.ccb_h.flags = CAM_DIR_IN;
cdai.buftype = CDAI_TYPE_RCAPLONG;
cdai.flags = 0;
cdai.bufsiz = sizeof(rcap_buf);
cdai.buf = (uint8_t *)&rcap_buf;
xpt_action((union ccb *)&cdai);
if ((cdai.ccb_h.status & CAM_DEV_QFRZN) != 0)
cam_release_devq(cdai.ccb_h.path,
0, 0, 0, FALSE);
if (((cdai.ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_CMP)
&& (rcap_buf.prot & SRC16_PROT_EN)) {
lun->eedp_formatted = TRUE;
lun->eedp_block_size = scsi_4btoul(rcap_buf.length);
} else {
lun->eedp_formatted = FALSE;
lun->eedp_block_size = 0;
}
break;
}
default:
break;
}
}
#else /* __FreeBSD_version >= 1000006 */
static void
mpssas_check_eedp(struct mpssas_softc *sassc)
{
struct mps_softc *sc = sassc->sc;
struct ccb_scsiio *csio;
struct scsi_read_capacity_16 *scsi_cmd;
struct scsi_read_capacity_eedp *rcap_buf;
union ccb *ccb;
path_id_t pathid = cam_sim_path(sassc->sim);
target_id_t targetid;
lun_id_t lunid;
struct cam_periph *found_periph;
struct mpssas_target *target;
struct mpssas_lun *lun;
uint8_t found_lun;
/*
* Issue a READ CAPACITY 16 command to each LUN of each target. This
* info is used to determine if the LUN is formatted for EEDP support.
*/
for (targetid = 0; targetid < sc->facts->MaxTargets; targetid++) {
target = &sassc->targets[targetid];
if (target->handle == 0x0) {
continue;
}
lunid = 0;
do {
rcap_buf =
malloc(sizeof(struct scsi_read_capacity_eedp),
M_MPT2, M_NOWAIT | M_ZERO);
if (rcap_buf == NULL) {
mps_dprint(sc, MPS_FAULT, "Unable to alloc read "
"capacity buffer for EEDP support.\n");
return;
}
ccb = xpt_alloc_ccb_nowait();
if (ccb == NULL) {
mps_dprint(sc, MPS_FAULT, "Unable to alloc CCB "
"for EEDP support.\n");
free(rcap_buf, M_MPT2);
return;
}
if (xpt_create_path(&ccb->ccb_h.path, xpt_periph,
pathid, targetid, lunid) != CAM_REQ_CMP) {
mps_dprint(sc, MPS_FAULT, "Unable to create "
"path for EEDP support\n");
free(rcap_buf, M_MPT2);
xpt_free_ccb(ccb);
return;
}
/*
* If a periph is returned, the LUN exists. Create an
* entry in the target's LUN list.
*/
if ((found_periph = cam_periph_find(ccb->ccb_h.path,
NULL)) != NULL) {
/*
* If LUN is already in list, don't create a new
* one.
*/
found_lun = FALSE;
SLIST_FOREACH(lun, &target->luns, lun_link) {
if (lun->lun_id == lunid) {
found_lun = TRUE;
break;
}
}
if (!found_lun) {
lun = malloc(sizeof(struct mpssas_lun),
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
M_MPT2, M_NOWAIT | M_ZERO);
if (lun == NULL) {
mps_dprint(sc, MPS_FAULT,
"Unable to alloc LUN for "
"EEDP support.\n");
free(rcap_buf, M_MPT2);
xpt_free_path(ccb->ccb_h.path);
xpt_free_ccb(ccb);
return;
}
lun->lun_id = lunid;
SLIST_INSERT_HEAD(&target->luns, lun,
lun_link);
}
lunid++;
/*
* Issue a READ CAPACITY 16 command for the LUN.
* The mpssas_read_cap_done function will load
* the read cap info into the LUN struct.
*/
csio = &ccb->csio;
csio->ccb_h.func_code = XPT_SCSI_IO;
csio->ccb_h.flags = CAM_DIR_IN;
csio->ccb_h.retry_count = 4;
csio->ccb_h.cbfcnp = mpssas_read_cap_done;
csio->ccb_h.timeout = 60000;
csio->data_ptr = (uint8_t *)rcap_buf;
csio->dxfer_len = sizeof(struct
scsi_read_capacity_eedp);
csio->sense_len = MPS_SENSE_LEN;
csio->cdb_len = sizeof(*scsi_cmd);
csio->tag_action = MSG_SIMPLE_Q_TAG;
scsi_cmd = (struct scsi_read_capacity_16 *)
&csio->cdb_io.cdb_bytes;
bzero(scsi_cmd, sizeof(*scsi_cmd));
scsi_cmd->opcode = 0x9E;
scsi_cmd->service_action = SRC16_SERVICE_ACTION;
((uint8_t *)scsi_cmd)[13] = sizeof(struct
scsi_read_capacity_eedp);
/*
* Set the path, target and lun IDs for the READ
* CAPACITY request.
*/
ccb->ccb_h.path_id =
xpt_path_path_id(ccb->ccb_h.path);
ccb->ccb_h.target_id =
xpt_path_target_id(ccb->ccb_h.path);
ccb->ccb_h.target_lun =
xpt_path_lun_id(ccb->ccb_h.path);
ccb->ccb_h.ppriv_ptr1 = sassc;
xpt_action(ccb);
} else {
free(rcap_buf, M_MPT2);
xpt_free_path(ccb->ccb_h.path);
xpt_free_ccb(ccb);
}
} while (found_periph);
}
}
static void
mpssas_read_cap_done(struct cam_periph *periph, union ccb *done_ccb)
{
struct mpssas_softc *sassc;
struct mpssas_target *target;
struct mpssas_lun *lun;
struct scsi_read_capacity_eedp *rcap_buf;
if (done_ccb == NULL)
return;
Bring in a number of mps(4) driver fixes from LSI: 1. Fixed timeout specification for the msleep in mps_wait_command(). Added 30 second timeout for mps_wait_command() calls in mps_user.c. 2. Make sure we call mps_detach_user() from the kldunload path. 3. Raid Hotplug behavior change. The driver now removes a volume when it goes to a failed state, so we also need to add volume back to the OS when it goes to opitimal/degraded/online from failed/missing. Handle raid volume add and remove from the IR_Volume event. 4. Added some more debugging information. 5. Replace xpt_async(AC_LOST_DEVICE, path, NULL) with mpssas_rescan_target(). This is to work around a panic in CAM that shows up when adding a drive with a rescan and removing another device from the driver thread with an AC_LOST_DEVICE async notification. This problem was encountered in testing with the LSI sas2ircu utility, which was used to create a RAID volume from physical disks. The driver has to create the RAID volume target and remove the physical disk targets, and triggered a panic in the process. The CAM issue needs to be fully diagnosed and fixed, but this works around the issue for now. 6. Fix some memory initialization issues in mps_free_command(). 7. Resolve the "devq freeze forever" issue. This was caused by the internal read capacity command issued in the non-head version of the driver. When the command completed with an error, the driver wasn't unfreezing thd device queue. The version in head uses the CAM infrastructure for getting the read capacity information, and therefore doesn't have the same issue. 8. Bump the version to 13.00.00.00-fbsd. (this is very close to LSI's internal stable driver 13.00.00.00) Submitted by: Kashyap Desai <Kashyap.Desai@lsi.com> MFC after: 3 days
2012-02-09 00:16:12 +00:00
/* Driver need to release devq, it Scsi command is
* generated by driver internally.
* Currently there is a single place where driver
* calls scsi command internally. In future if driver
* calls more scsi command internally, it needs to release
* devq internally, since those command will not go back to
* cam_periph.
*/
if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) ) {
done_ccb->ccb_h.status &= ~CAM_DEV_QFRZN;
xpt_release_devq(done_ccb->ccb_h.path,
/*count*/ 1, /*run_queue*/TRUE);
}
rcap_buf = (struct scsi_read_capacity_eedp *)done_ccb->csio.data_ptr;
/*
* Get the LUN ID for the path and look it up in the LUN list for the
* target.
*/
sassc = (struct mpssas_softc *)done_ccb->ccb_h.ppriv_ptr1;
target = &sassc->targets[done_ccb->ccb_h.target_id];
SLIST_FOREACH(lun, &target->luns, lun_link) {
if (lun->lun_id != done_ccb->ccb_h.target_lun)
continue;
/*
* Got the LUN in the target's LUN list. Fill it in
* with EEDP info. If the READ CAP 16 command had some
* SCSI error (common if command is not supported), mark
* the lun as not supporting EEDP and set the block size
* to 0.
*/
if (((done_ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP)
|| (done_ccb->csio.scsi_status != SCSI_STATUS_OK)) {
lun->eedp_formatted = FALSE;
lun->eedp_block_size = 0;
break;
}
if (rcap_buf->protect & 0x01) {
lun->eedp_formatted = TRUE;
lun->eedp_block_size = scsi_4btoul(rcap_buf->length);
}
break;
}
// Finished with this CCB and path.
free(rcap_buf, M_MPT2);
xpt_free_path(done_ccb->ccb_h.path);
xpt_free_ccb(done_ccb);
}
#endif /* __FreeBSD_version >= 1000006 */
int
mpssas_startup(struct mps_softc *sc)
{
struct mpssas_softc *sassc;
/*
* Send the port enable message and set the wait_for_port_enable flag.
* This flag helps to keep the simq frozen until all discovery events
* are processed.
*/
sassc = sc->sassc;
mpssas_startup_increment(sassc);
sc->wait_for_port_enable = 1;
mpssas_send_portenable(sc);
return (0);
}
static int
mpssas_send_portenable(struct mps_softc *sc)
{
MPI2_PORT_ENABLE_REQUEST *request;
struct mps_command *cm;
mps_dprint(sc, MPS_TRACE, "%s\n", __func__);
if ((cm = mps_alloc_command(sc)) == NULL)
return (EBUSY);
request = (MPI2_PORT_ENABLE_REQUEST *)cm->cm_req;
request->Function = MPI2_FUNCTION_PORT_ENABLE;
request->MsgFlags = 0;
request->VP_ID = 0;
cm->cm_desc.Default.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_DEFAULT_TYPE;
cm->cm_complete = mpssas_portenable_complete;
cm->cm_data = NULL;
cm->cm_sge = NULL;
mps_map_command(sc, cm);
mps_dprint(sc, MPS_TRACE,
"mps_send_portenable finished cm %p req %p complete %p\n",
cm, cm->cm_req, cm->cm_complete);
return (0);
}
static void
mpssas_portenable_complete(struct mps_softc *sc, struct mps_command *cm)
{
MPI2_PORT_ENABLE_REPLY *reply;
struct mpssas_softc *sassc;
struct mpssas_target *target;
int i;
mps_dprint(sc, MPS_TRACE, "%s\n", __func__);
sassc = sc->sassc;
/*
* Currently there should be no way we can hit this case. It only
* happens when we have a failure to allocate chain frames, and
* port enable commands don't have S/G lists.
*/
if ((cm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
mps_printf(sc, "%s: cm_flags = %#x for port enable! "
"This should not happen!\n", __func__, cm->cm_flags);
}
reply = (MPI2_PORT_ENABLE_REPLY *)cm->cm_reply;
if (reply == NULL)
mps_dprint(sc, MPS_FAULT, "Portenable NULL reply\n");
else if ((reply->IOCStatus & MPI2_IOCSTATUS_MASK) !=
MPI2_IOCSTATUS_SUCCESS)
mps_dprint(sc, MPS_FAULT, "Portenable failed\n");
mps_free_command(sc, cm);
if (sc->mps_ich.ich_arg != NULL) {
mps_dprint(sc, MPS_INFO, "disestablish config intrhook\n");
config_intrhook_disestablish(&sc->mps_ich);
sc->mps_ich.ich_arg = NULL;
}
/*
* Get WarpDrive info after discovery is complete but before the scan
* starts. At this point, all devices are ready to be exposed to the
* OS. If devices should be hidden instead, take them out of the
* 'targets' array before the scan. The devinfo for a disk will have
* some info and a volume's will be 0. Use that to remove disks.
*/
mps_wd_config_pages(sc);
if (((sc->mps_flags & MPS_FLAGS_WD_AVAILABLE)
&& (sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS))
|| (sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME))) {
for (i = 0; i < sassc->sc->facts->MaxTargets; i++) {
target = &sassc->targets[i];
if (target->devinfo) {
target->devinfo = 0x0;
target->encl_handle = 0x0;
target->encl_slot = 0x0;
target->handle = 0x0;
target->tid = 0x0;
target->linkrate = 0x0;
target->flags = 0x0;
}
}
}
/*
* Done waiting for port enable to complete. Decrement the refcount.
* If refcount is 0, discovery is complete and a rescan of the bus can
* take place. Since the simq was explicitly frozen before port
* enable, it must be explicitly released here to keep the
* freeze/release count in sync.
*/
sc->wait_for_port_enable = 0;
sc->port_enable_complete = 1;
mpssas_startup_decrement(sassc);
xpt_release_simq(sassc->sim, 1);
}