freebsd-skq/sys/dev/mly/mly.c

3015 lines
94 KiB
C
Raw Normal View History

/*-
* Copyright (c) 2000, 2001 Michael Smith
* Copyright (c) 2000 BSDi
* 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.
*
* $FreeBSD$
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
#include <sys/bus.h>
#include <sys/conf.h>
#include <sys/ctype.h>
#include <sys/ioccom.h>
#include <sys/stat.h>
#include <machine/bus.h>
#include <machine/resource.h>
#include <sys/rman.h>
#include <cam/cam.h>
#include <cam/cam_ccb.h>
#include <cam/cam_periph.h>
#include <cam/cam_sim.h>
#include <cam/cam_xpt_sim.h>
#include <cam/scsi/scsi_all.h>
#include <cam/scsi/scsi_message.h>
#include <dev/pci/pcireg.h>
#include <dev/pci/pcivar.h>
#include <dev/mly/mlyreg.h>
#include <dev/mly/mlyio.h>
#include <dev/mly/mlyvar.h>
#include <dev/mly/mly_tables.h>
static int mly_probe(device_t dev);
static int mly_attach(device_t dev);
static int mly_pci_attach(struct mly_softc *sc);
static int mly_detach(device_t dev);
static int mly_shutdown(device_t dev);
static void mly_intr(void *arg);
static int mly_sg_map(struct mly_softc *sc);
static void mly_sg_map_helper(void *arg, bus_dma_segment_t *segs, int nseg, int error);
static int mly_mmbox_map(struct mly_softc *sc);
static void mly_mmbox_map_helper(void *arg, bus_dma_segment_t *segs, int nseg, int error);
static void mly_free(struct mly_softc *sc);
static int mly_get_controllerinfo(struct mly_softc *sc);
static void mly_scan_devices(struct mly_softc *sc);
static void mly_rescan_btl(struct mly_softc *sc, int bus, int target);
static void mly_complete_rescan(struct mly_command *mc);
static int mly_get_eventstatus(struct mly_softc *sc);
static int mly_enable_mmbox(struct mly_softc *sc);
static int mly_flush(struct mly_softc *sc);
static int mly_ioctl(struct mly_softc *sc, struct mly_command_ioctl *ioctl, void **data,
size_t datasize, u_int8_t *status, void *sense_buffer, size_t *sense_length);
static void mly_check_event(struct mly_softc *sc);
static void mly_fetch_event(struct mly_softc *sc);
static void mly_complete_event(struct mly_command *mc);
static void mly_process_event(struct mly_softc *sc, struct mly_event *me);
static void mly_periodic(void *data);
static int mly_immediate_command(struct mly_command *mc);
static int mly_start(struct mly_command *mc);
static void mly_done(struct mly_softc *sc);
static void mly_complete(struct mly_softc *sc);
static void mly_complete_handler(void *context, int pending);
static int mly_alloc_command(struct mly_softc *sc, struct mly_command **mcp);
static void mly_release_command(struct mly_command *mc);
static void mly_alloc_commands_map(void *arg, bus_dma_segment_t *segs, int nseg, int error);
static int mly_alloc_commands(struct mly_softc *sc);
static void mly_release_commands(struct mly_softc *sc);
static void mly_map_command(struct mly_command *mc);
static void mly_unmap_command(struct mly_command *mc);
static int mly_cam_attach(struct mly_softc *sc);
static void mly_cam_detach(struct mly_softc *sc);
static void mly_cam_rescan_btl(struct mly_softc *sc, int bus, int target);
static void mly_cam_action(struct cam_sim *sim, union ccb *ccb);
static int mly_cam_action_io(struct cam_sim *sim, struct ccb_scsiio *csio);
static void mly_cam_poll(struct cam_sim *sim);
static void mly_cam_complete(struct mly_command *mc);
static struct cam_periph *mly_find_periph(struct mly_softc *sc, int bus, int target);
static int mly_name_device(struct mly_softc *sc, int bus, int target);
static int mly_fwhandshake(struct mly_softc *sc);
static void mly_describe_controller(struct mly_softc *sc);
#ifdef MLY_DEBUG
static void mly_printstate(struct mly_softc *sc);
static void mly_print_command(struct mly_command *mc);
static void mly_print_packet(struct mly_command *mc);
static void mly_panic(struct mly_softc *sc, char *reason);
static void mly_timeout(void *arg);
#endif
void mly_print_controller(int controller);
static d_open_t mly_user_open;
static d_close_t mly_user_close;
static d_ioctl_t mly_user_ioctl;
static int mly_user_command(struct mly_softc *sc, struct mly_user_command *uc);
static int mly_user_health(struct mly_softc *sc, struct mly_user_health *uh);
#define MLY_CMD_TIMEOUT 20
static device_method_t mly_methods[] = {
/* Device interface */
DEVMETHOD(device_probe, mly_probe),
DEVMETHOD(device_attach, mly_attach),
DEVMETHOD(device_detach, mly_detach),
DEVMETHOD(device_shutdown, mly_shutdown),
{ 0, 0 }
};
static driver_t mly_pci_driver = {
"mly",
mly_methods,
sizeof(struct mly_softc)
};
static devclass_t mly_devclass;
DRIVER_MODULE(mly, pci, mly_pci_driver, mly_devclass, 0, 0);
MODULE_DEPEND(mly, pci, 1, 1, 1);
MODULE_DEPEND(mly, cam, 1, 1, 1);
static struct cdevsw mly_cdevsw = {
.d_version = D_VERSION,
.d_open = mly_user_open,
.d_close = mly_user_close,
.d_ioctl = mly_user_ioctl,
.d_name = "mly",
};
/********************************************************************************
********************************************************************************
Device Interface
********************************************************************************
********************************************************************************/
static struct mly_ident
{
u_int16_t vendor;
u_int16_t device;
u_int16_t subvendor;
u_int16_t subdevice;
int hwif;
char *desc;
} mly_identifiers[] = {
{0x1069, 0xba56, 0x1069, 0x0040, MLY_HWIF_STRONGARM, "Mylex eXtremeRAID 2000"},
{0x1069, 0xba56, 0x1069, 0x0030, MLY_HWIF_STRONGARM, "Mylex eXtremeRAID 3000"},
{0x1069, 0x0050, 0x1069, 0x0050, MLY_HWIF_I960RX, "Mylex AcceleRAID 352"},
{0x1069, 0x0050, 0x1069, 0x0052, MLY_HWIF_I960RX, "Mylex AcceleRAID 170"},
{0x1069, 0x0050, 0x1069, 0x0054, MLY_HWIF_I960RX, "Mylex AcceleRAID 160"},
{0, 0, 0, 0, 0, 0}
};
/********************************************************************************
* Compare the provided PCI device with the list we support.
*/
static int
mly_probe(device_t dev)
{
struct mly_ident *m;
debug_called(1);
for (m = mly_identifiers; m->vendor != 0; m++) {
if ((m->vendor == pci_get_vendor(dev)) &&
(m->device == pci_get_device(dev)) &&
((m->subvendor == 0) || ((m->subvendor == pci_get_subvendor(dev)) &&
(m->subdevice == pci_get_subdevice(dev))))) {
device_set_desc(dev, m->desc);
return(BUS_PROBE_DEFAULT); /* allow room to be overridden */
}
}
return(ENXIO);
}
/********************************************************************************
* Initialise the controller and softc
*/
static int
mly_attach(device_t dev)
{
struct mly_softc *sc = device_get_softc(dev);
int error;
debug_called(1);
sc->mly_dev = dev;
mtx_init(&sc->mly_lock, "mly", NULL, MTX_DEF);
callout_init_mtx(&sc->mly_periodic, &sc->mly_lock, 0);
#ifdef MLY_DEBUG
callout_init_mtx(&sc->mly_timeout, &sc->mly_lock, 0);
if (device_get_unit(sc->mly_dev) == 0)
mly_softc0 = sc;
#endif
/*
* Do PCI-specific initialisation.
*/
if ((error = mly_pci_attach(sc)) != 0)
goto out;
/*
* Initialise per-controller queues.
*/
mly_initq_free(sc);
mly_initq_busy(sc);
mly_initq_complete(sc);
/*
* Initialise command-completion task.
*/
TASK_INIT(&sc->mly_task_complete, 0, mly_complete_handler, sc);
/* disable interrupts before we start talking to the controller */
MLY_MASK_INTERRUPTS(sc);
/*
* Wait for the controller to come ready, handshake with the firmware if required.
* This is typically only necessary on platforms where the controller BIOS does not
* run.
*/
if ((error = mly_fwhandshake(sc)))
goto out;
/*
* Allocate initial command buffers.
*/
if ((error = mly_alloc_commands(sc)))
goto out;
/*
* Obtain controller feature information
*/
MLY_LOCK(sc);
error = mly_get_controllerinfo(sc);
MLY_UNLOCK(sc);
if (error)
goto out;
/*
* Reallocate command buffers now we know how many we want.
*/
mly_release_commands(sc);
if ((error = mly_alloc_commands(sc)))
goto out;
/*
* Get the current event counter for health purposes, populate the initial
* health status buffer.
*/
MLY_LOCK(sc);
error = mly_get_eventstatus(sc);
/*
* Enable memory-mailbox mode.
*/
if (error == 0)
error = mly_enable_mmbox(sc);
MLY_UNLOCK(sc);
if (error)
goto out;
/*
* Attach to CAM.
*/
if ((error = mly_cam_attach(sc)))
goto out;
/*
* Print a little information about the controller
*/
mly_describe_controller(sc);
/*
* Mark all attached devices for rescan.
*/
MLY_LOCK(sc);
mly_scan_devices(sc);
/*
* Instigate the first status poll immediately. Rescan completions won't
* happen until interrupts are enabled, which should still be before
* the SCSI subsystem gets to us, courtesy of the "SCSI settling delay".
*/
mly_periodic((void *)sc);
MLY_UNLOCK(sc);
/*
* Create the control device.
*/
sc->mly_dev_t = make_dev(&mly_cdevsw, 0, UID_ROOT, GID_OPERATOR,
S_IRUSR | S_IWUSR, "mly%d", device_get_unit(sc->mly_dev));
sc->mly_dev_t->si_drv1 = sc;
/* enable interrupts now */
MLY_UNMASK_INTERRUPTS(sc);
#ifdef MLY_DEBUG
callout_reset(&sc->mly_timeout, MLY_CMD_TIMEOUT * hz, mly_timeout, sc);
#endif
out:
if (error != 0)
mly_free(sc);
return(error);
}
/********************************************************************************
* Perform PCI-specific initialisation.
*/
static int
mly_pci_attach(struct mly_softc *sc)
{
int i, error;
debug_called(1);
/* assume failure is 'not configured' */
error = ENXIO;
/*
* Verify that the adapter is correctly set up in PCI space.
*/
pci_enable_busmaster(sc->mly_dev);
/*
* Allocate the PCI register window.
*/
sc->mly_regs_rid = PCIR_BAR(0); /* first base address register */
if ((sc->mly_regs_resource = bus_alloc_resource_any(sc->mly_dev,
SYS_RES_MEMORY, &sc->mly_regs_rid, RF_ACTIVE)) == NULL) {
mly_printf(sc, "can't allocate register window\n");
goto fail;
}
/*
* Allocate and connect our interrupt.
*/
sc->mly_irq_rid = 0;
if ((sc->mly_irq = bus_alloc_resource_any(sc->mly_dev, SYS_RES_IRQ,
&sc->mly_irq_rid, RF_SHAREABLE | RF_ACTIVE)) == NULL) {
mly_printf(sc, "can't allocate interrupt\n");
goto fail;
}
if (bus_setup_intr(sc->mly_dev, sc->mly_irq, INTR_TYPE_CAM | INTR_ENTROPY | INTR_MPSAFE, NULL, mly_intr, sc, &sc->mly_intr)) {
mly_printf(sc, "can't set up interrupt\n");
goto fail;
}
/* assume failure is 'out of memory' */
error = ENOMEM;
/*
* Allocate the parent bus DMA tag appropriate for our PCI interface.
*
* Note that all of these controllers are 64-bit capable.
*/
if (bus_dma_tag_create(bus_get_dma_tag(sc->mly_dev),/* PCI parent */
1, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR_32BIT, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
BUS_SPACE_MAXSIZE_32BIT, /* maxsize */
BUS_SPACE_UNRESTRICTED, /* nsegments */
BUS_SPACE_MAXSIZE_32BIT, /* maxsegsize */
BUS_DMA_ALLOCNOW, /* flags */
NULL, /* lockfunc */
NULL, /* lockarg */
&sc->mly_parent_dmat)) {
mly_printf(sc, "can't allocate parent DMA tag\n");
goto fail;
}
/*
* Create DMA tag for mapping buffers into controller-addressable space.
*/
if (bus_dma_tag_create(sc->mly_parent_dmat, /* parent */
1, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
DFLTPHYS, /* maxsize */
MLY_MAX_SGENTRIES, /* nsegments */
BUS_SPACE_MAXSIZE_32BIT, /* maxsegsize */
0, /* flags */
busdma_lock_mutex, /* lockfunc */
&sc->mly_lock, /* lockarg */
&sc->mly_buffer_dmat)) {
mly_printf(sc, "can't allocate buffer DMA tag\n");
goto fail;
}
/*
* Initialise the DMA tag for command packets.
*/
if (bus_dma_tag_create(sc->mly_parent_dmat, /* parent */
1, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
sizeof(union mly_command_packet) * MLY_MAX_COMMANDS, 1, /* maxsize, nsegments */
BUS_SPACE_MAXSIZE_32BIT, /* maxsegsize */
BUS_DMA_ALLOCNOW, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&sc->mly_packet_dmat)) {
mly_printf(sc, "can't allocate command packet DMA tag\n");
goto fail;
}
/*
* Detect the hardware interface version
*/
for (i = 0; mly_identifiers[i].vendor != 0; i++) {
if ((mly_identifiers[i].vendor == pci_get_vendor(sc->mly_dev)) &&
(mly_identifiers[i].device == pci_get_device(sc->mly_dev))) {
sc->mly_hwif = mly_identifiers[i].hwif;
switch(sc->mly_hwif) {
case MLY_HWIF_I960RX:
debug(1, "set hardware up for i960RX");
sc->mly_doorbell_true = 0x00;
sc->mly_command_mailbox = MLY_I960RX_COMMAND_MAILBOX;
sc->mly_status_mailbox = MLY_I960RX_STATUS_MAILBOX;
sc->mly_idbr = MLY_I960RX_IDBR;
sc->mly_odbr = MLY_I960RX_ODBR;
sc->mly_error_status = MLY_I960RX_ERROR_STATUS;
sc->mly_interrupt_status = MLY_I960RX_INTERRUPT_STATUS;
sc->mly_interrupt_mask = MLY_I960RX_INTERRUPT_MASK;
break;
case MLY_HWIF_STRONGARM:
debug(1, "set hardware up for StrongARM");
sc->mly_doorbell_true = 0xff; /* doorbell 'true' is 0 */
sc->mly_command_mailbox = MLY_STRONGARM_COMMAND_MAILBOX;
sc->mly_status_mailbox = MLY_STRONGARM_STATUS_MAILBOX;
sc->mly_idbr = MLY_STRONGARM_IDBR;
sc->mly_odbr = MLY_STRONGARM_ODBR;
sc->mly_error_status = MLY_STRONGARM_ERROR_STATUS;
sc->mly_interrupt_status = MLY_STRONGARM_INTERRUPT_STATUS;
sc->mly_interrupt_mask = MLY_STRONGARM_INTERRUPT_MASK;
break;
}
break;
}
}
/*
* Create the scatter/gather mappings.
*/
if ((error = mly_sg_map(sc)))
goto fail;
/*
* Allocate and map the memory mailbox
*/
if ((error = mly_mmbox_map(sc)))
goto fail;
error = 0;
fail:
return(error);
}
/********************************************************************************
* Shut the controller down and detach all our resources.
*/
static int
mly_detach(device_t dev)
{
int error;
if ((error = mly_shutdown(dev)) != 0)
return(error);
mly_free(device_get_softc(dev));
return(0);
}
/********************************************************************************
* Bring the controller to a state where it can be safely left alone.
*
* Note that it should not be necessary to wait for any outstanding commands,
* as they should be completed prior to calling here.
*
* XXX this applies for I/O, but not status polls; we should beware of
* the case where a status command is running while we detach.
*/
static int
mly_shutdown(device_t dev)
{
struct mly_softc *sc = device_get_softc(dev);
debug_called(1);
MLY_LOCK(sc);
if (sc->mly_state & MLY_STATE_OPEN) {
MLY_UNLOCK(sc);
return(EBUSY);
}
/* kill the periodic event */
callout_stop(&sc->mly_periodic);
#ifdef MLY_DEBUG
callout_stop(&sc->mly_timeout);
#endif
/* flush controller */
mly_printf(sc, "flushing cache...");
printf("%s\n", mly_flush(sc) ? "failed" : "done");
MLY_MASK_INTERRUPTS(sc);
MLY_UNLOCK(sc);
return(0);
}
/*******************************************************************************
* Take an interrupt, or be poked by other code to look for interrupt-worthy
* status.
*/
static void
mly_intr(void *arg)
{
struct mly_softc *sc = (struct mly_softc *)arg;
debug_called(2);
MLY_LOCK(sc);
mly_done(sc);
MLY_UNLOCK(sc);
};
/********************************************************************************
********************************************************************************
Bus-dependant Resource Management
********************************************************************************
********************************************************************************/
/********************************************************************************
* Allocate memory for the scatter/gather tables
*/
static int
mly_sg_map(struct mly_softc *sc)
{
size_t segsize;
debug_called(1);
/*
* Create a single tag describing a region large enough to hold all of
* the s/g lists we will need.
*/
segsize = sizeof(struct mly_sg_entry) * MLY_MAX_COMMANDS *MLY_MAX_SGENTRIES;
if (bus_dma_tag_create(sc->mly_parent_dmat, /* parent */
1, 0, /* alignment,boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
segsize, 1, /* maxsize, nsegments */
BUS_SPACE_MAXSIZE_32BIT, /* maxsegsize */
BUS_DMA_ALLOCNOW, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&sc->mly_sg_dmat)) {
mly_printf(sc, "can't allocate scatter/gather DMA tag\n");
return(ENOMEM);
}
/*
* Allocate enough s/g maps for all commands and permanently map them into
* controller-visible space.
*
* XXX this assumes we can get enough space for all the s/g maps in one
* contiguous slab.
*/
if (bus_dmamem_alloc(sc->mly_sg_dmat, (void **)&sc->mly_sg_table,
BUS_DMA_NOWAIT, &sc->mly_sg_dmamap)) {
mly_printf(sc, "can't allocate s/g table\n");
return(ENOMEM);
}
if (bus_dmamap_load(sc->mly_sg_dmat, sc->mly_sg_dmamap, sc->mly_sg_table,
segsize, mly_sg_map_helper, sc, BUS_DMA_NOWAIT) != 0)
return (ENOMEM);
return(0);
}
/********************************************************************************
* Save the physical address of the base of the s/g table.
*/
static void
mly_sg_map_helper(void *arg, bus_dma_segment_t *segs, int nseg, int error)
{
struct mly_softc *sc = (struct mly_softc *)arg;
debug_called(1);
/* save base of s/g table's address in bus space */
sc->mly_sg_busaddr = segs->ds_addr;
}
/********************************************************************************
* Allocate memory for the memory-mailbox interface
*/
static int
mly_mmbox_map(struct mly_softc *sc)
{
/*
* Create a DMA tag for a single contiguous region large enough for the
* memory mailbox structure.
*/
if (bus_dma_tag_create(sc->mly_parent_dmat, /* parent */
1, 0, /* alignment,boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
sizeof(struct mly_mmbox), 1, /* maxsize, nsegments */
BUS_SPACE_MAXSIZE_32BIT, /* maxsegsize */
BUS_DMA_ALLOCNOW, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&sc->mly_mmbox_dmat)) {
mly_printf(sc, "can't allocate memory mailbox DMA tag\n");
return(ENOMEM);
}
/*
* Allocate the buffer
*/
if (bus_dmamem_alloc(sc->mly_mmbox_dmat, (void **)&sc->mly_mmbox, BUS_DMA_NOWAIT, &sc->mly_mmbox_dmamap)) {
mly_printf(sc, "can't allocate memory mailbox\n");
return(ENOMEM);
}
if (bus_dmamap_load(sc->mly_mmbox_dmat, sc->mly_mmbox_dmamap, sc->mly_mmbox,
sizeof(struct mly_mmbox), mly_mmbox_map_helper, sc,
BUS_DMA_NOWAIT) != 0)
return (ENOMEM);
bzero(sc->mly_mmbox, sizeof(*sc->mly_mmbox));
return(0);
}
/********************************************************************************
* Save the physical address of the memory mailbox
*/
static void
mly_mmbox_map_helper(void *arg, bus_dma_segment_t *segs, int nseg, int error)
{
struct mly_softc *sc = (struct mly_softc *)arg;
debug_called(1);
sc->mly_mmbox_busaddr = segs->ds_addr;
}
/********************************************************************************
* Free all of the resources associated with (sc)
*
* Should not be called if the controller is active.
*/
static void
mly_free(struct mly_softc *sc)
{
debug_called(1);
/* Remove the management device */
destroy_dev(sc->mly_dev_t);
if (sc->mly_intr)
bus_teardown_intr(sc->mly_dev, sc->mly_irq, sc->mly_intr);
callout_drain(&sc->mly_periodic);
#ifdef MLY_DEBUG
callout_drain(&sc->mly_timeout);
#endif
/* detach from CAM */
mly_cam_detach(sc);
/* release command memory */
mly_release_commands(sc);
/* throw away the controllerinfo structure */
if (sc->mly_controllerinfo != NULL)
free(sc->mly_controllerinfo, M_DEVBUF);
/* throw away the controllerparam structure */
if (sc->mly_controllerparam != NULL)
free(sc->mly_controllerparam, M_DEVBUF);
/* destroy data-transfer DMA tag */
if (sc->mly_buffer_dmat)
bus_dma_tag_destroy(sc->mly_buffer_dmat);
/* free and destroy DMA memory and tag for s/g lists */
if (sc->mly_sg_table) {
bus_dmamap_unload(sc->mly_sg_dmat, sc->mly_sg_dmamap);
bus_dmamem_free(sc->mly_sg_dmat, sc->mly_sg_table, sc->mly_sg_dmamap);
}
if (sc->mly_sg_dmat)
bus_dma_tag_destroy(sc->mly_sg_dmat);
/* free and destroy DMA memory and tag for memory mailbox */
if (sc->mly_mmbox) {
bus_dmamap_unload(sc->mly_mmbox_dmat, sc->mly_mmbox_dmamap);
bus_dmamem_free(sc->mly_mmbox_dmat, sc->mly_mmbox, sc->mly_mmbox_dmamap);
}
if (sc->mly_mmbox_dmat)
bus_dma_tag_destroy(sc->mly_mmbox_dmat);
/* disconnect the interrupt handler */
if (sc->mly_irq != NULL)
bus_release_resource(sc->mly_dev, SYS_RES_IRQ, sc->mly_irq_rid, sc->mly_irq);
/* destroy the parent DMA tag */
if (sc->mly_parent_dmat)
bus_dma_tag_destroy(sc->mly_parent_dmat);
/* release the register window mapping */
if (sc->mly_regs_resource != NULL)
bus_release_resource(sc->mly_dev, SYS_RES_MEMORY, sc->mly_regs_rid, sc->mly_regs_resource);
mtx_destroy(&sc->mly_lock);
}
/********************************************************************************
********************************************************************************
Command Wrappers
********************************************************************************
********************************************************************************/
/********************************************************************************
* Fill in the mly_controllerinfo and mly_controllerparam fields in the softc.
*/
static int
mly_get_controllerinfo(struct mly_softc *sc)
{
struct mly_command_ioctl mci;
u_int8_t status;
int error;
debug_called(1);
if (sc->mly_controllerinfo != NULL)
free(sc->mly_controllerinfo, M_DEVBUF);
/* build the getcontrollerinfo ioctl and send it */
bzero(&mci, sizeof(mci));
sc->mly_controllerinfo = NULL;
mci.sub_ioctl = MDACIOCTL_GETCONTROLLERINFO;
if ((error = mly_ioctl(sc, &mci, (void **)&sc->mly_controllerinfo, sizeof(*sc->mly_controllerinfo),
&status, NULL, NULL)))
return(error);
if (status != 0)
return(EIO);
if (sc->mly_controllerparam != NULL)
free(sc->mly_controllerparam, M_DEVBUF);
/* build the getcontrollerparameter ioctl and send it */
bzero(&mci, sizeof(mci));
sc->mly_controllerparam = NULL;
mci.sub_ioctl = MDACIOCTL_GETCONTROLLERPARAMETER;
if ((error = mly_ioctl(sc, &mci, (void **)&sc->mly_controllerparam, sizeof(*sc->mly_controllerparam),
&status, NULL, NULL)))
return(error);
if (status != 0)
return(EIO);
return(0);
}
/********************************************************************************
* Schedule all possible devices for a rescan.
*
*/
static void
mly_scan_devices(struct mly_softc *sc)
{
int bus, target;
debug_called(1);
/*
* Clear any previous BTL information.
*/
bzero(&sc->mly_btl, sizeof(sc->mly_btl));
/*
* Mark all devices as requiring a rescan, and let the next
* periodic scan collect them.
*/
for (bus = 0; bus < sc->mly_cam_channels; bus++)
if (MLY_BUS_IS_VALID(sc, bus))
for (target = 0; target < MLY_MAX_TARGETS; target++)
sc->mly_btl[bus][target].mb_flags = MLY_BTL_RESCAN;
}
/********************************************************************************
* Rescan a device, possibly as a consequence of getting an event which suggests
* that it may have changed.
*
* If we suffer resource starvation, we can abandon the rescan as we'll be
* retried.
*/
static void
mly_rescan_btl(struct mly_softc *sc, int bus, int target)
{
struct mly_command *mc;
struct mly_command_ioctl *mci;
debug_called(1);
/* check that this bus is valid */
if (!MLY_BUS_IS_VALID(sc, bus))
return;
/* get a command */
if (mly_alloc_command(sc, &mc))
return;
/* set up the data buffer */
if ((mc->mc_data = malloc(sizeof(union mly_devinfo), M_DEVBUF, M_NOWAIT | M_ZERO)) == NULL) {
mly_release_command(mc);
return;
}
mc->mc_flags |= MLY_CMD_DATAIN;
mc->mc_complete = mly_complete_rescan;
/*
* Build the ioctl.
*/
mci = (struct mly_command_ioctl *)&mc->mc_packet->ioctl;
mci->opcode = MDACMD_IOCTL;
mci->addr.phys.controller = 0;
mci->timeout.value = 30;
mci->timeout.scale = MLY_TIMEOUT_SECONDS;
if (MLY_BUS_IS_VIRTUAL(sc, bus)) {
mc->mc_length = mci->data_size = sizeof(struct mly_ioctl_getlogdevinfovalid);
mci->sub_ioctl = MDACIOCTL_GETLOGDEVINFOVALID;
mci->addr.log.logdev = MLY_LOGDEV_ID(sc, bus, target);
debug(1, "logical device %d", mci->addr.log.logdev);
} else {
mc->mc_length = mci->data_size = sizeof(struct mly_ioctl_getphysdevinfovalid);
mci->sub_ioctl = MDACIOCTL_GETPHYSDEVINFOVALID;
mci->addr.phys.lun = 0;
mci->addr.phys.target = target;
mci->addr.phys.channel = bus;
debug(1, "physical device %d:%d", mci->addr.phys.channel, mci->addr.phys.target);
}
/*
* Dispatch the command. If we successfully send the command, clear the rescan
* bit.
*/
if (mly_start(mc) != 0) {
mly_release_command(mc);
} else {
sc->mly_btl[bus][target].mb_flags &= ~MLY_BTL_RESCAN; /* success */
}
}
/********************************************************************************
* Handle the completion of a rescan operation
*/
static void
mly_complete_rescan(struct mly_command *mc)
{
struct mly_softc *sc = mc->mc_sc;
struct mly_ioctl_getlogdevinfovalid *ldi;
struct mly_ioctl_getphysdevinfovalid *pdi;
struct mly_command_ioctl *mci;
struct mly_btl btl, *btlp;
int bus, target, rescan;
debug_called(1);
/*
* Recover the bus and target from the command. We need these even in
* the case where we don't have a useful response.
*/
mci = (struct mly_command_ioctl *)&mc->mc_packet->ioctl;
if (mci->sub_ioctl == MDACIOCTL_GETLOGDEVINFOVALID) {
bus = MLY_LOGDEV_BUS(sc, mci->addr.log.logdev);
target = MLY_LOGDEV_TARGET(sc, mci->addr.log.logdev);
} else {
bus = mci->addr.phys.channel;
target = mci->addr.phys.target;
}
/* XXX validate bus/target? */
/* the default result is 'no device' */
bzero(&btl, sizeof(btl));
/* if the rescan completed OK, we have possibly-new BTL data */
if (mc->mc_status == 0) {
if (mc->mc_length == sizeof(*ldi)) {
ldi = (struct mly_ioctl_getlogdevinfovalid *)mc->mc_data;
if ((MLY_LOGDEV_BUS(sc, ldi->logical_device_number) != bus) ||
(MLY_LOGDEV_TARGET(sc, ldi->logical_device_number) != target)) {
mly_printf(sc, "WARNING: BTL rescan for %d:%d returned data for %d:%d instead\n",
bus, target, MLY_LOGDEV_BUS(sc, ldi->logical_device_number),
MLY_LOGDEV_TARGET(sc, ldi->logical_device_number));
/* XXX what can we do about this? */
}
btl.mb_flags = MLY_BTL_LOGICAL;
btl.mb_type = ldi->raid_level;
btl.mb_state = ldi->state;
debug(1, "BTL rescan for %d returns %s, %s", ldi->logical_device_number,
mly_describe_code(mly_table_device_type, ldi->raid_level),
mly_describe_code(mly_table_device_state, ldi->state));
} else if (mc->mc_length == sizeof(*pdi)) {
pdi = (struct mly_ioctl_getphysdevinfovalid *)mc->mc_data;
if ((pdi->channel != bus) || (pdi->target != target)) {
mly_printf(sc, "WARNING: BTL rescan for %d:%d returned data for %d:%d instead\n",
bus, target, pdi->channel, pdi->target);
/* XXX what can we do about this? */
}
btl.mb_flags = MLY_BTL_PHYSICAL;
btl.mb_type = MLY_DEVICE_TYPE_PHYSICAL;
btl.mb_state = pdi->state;
btl.mb_speed = pdi->speed;
btl.mb_width = pdi->width;
if (pdi->state != MLY_DEVICE_STATE_UNCONFIGURED)
sc->mly_btl[bus][target].mb_flags |= MLY_BTL_PROTECTED;
debug(1, "BTL rescan for %d:%d returns %s", bus, target,
mly_describe_code(mly_table_device_state, pdi->state));
} else {
mly_printf(sc, "BTL rescan result invalid\n");
}
}
free(mc->mc_data, M_DEVBUF);
mly_release_command(mc);
/*
* Decide whether we need to rescan the device.
*/
rescan = 0;
/* device type changes (usually between 'nothing' and 'something') */
btlp = &sc->mly_btl[bus][target];
if (btl.mb_flags != btlp->mb_flags) {
debug(1, "flags changed, rescanning");
rescan = 1;
}
/* XXX other reasons? */
/*
* Update BTL information.
*/
*btlp = btl;
/*
* Perform CAM rescan if required.
*/
if (rescan)
mly_cam_rescan_btl(sc, bus, target);
}
/********************************************************************************
* Get the current health status and set the 'next event' counter to suit.
*/
static int
mly_get_eventstatus(struct mly_softc *sc)
{
struct mly_command_ioctl mci;
struct mly_health_status *mh;
u_int8_t status;
int error;
/* build the gethealthstatus ioctl and send it */
bzero(&mci, sizeof(mci));
mh = NULL;
mci.sub_ioctl = MDACIOCTL_GETHEALTHSTATUS;
if ((error = mly_ioctl(sc, &mci, (void **)&mh, sizeof(*mh), &status, NULL, NULL)))
return(error);
if (status != 0)
return(EIO);
/* get the event counter */
sc->mly_event_change = mh->change_counter;
sc->mly_event_waiting = mh->next_event;
sc->mly_event_counter = mh->next_event;
/* save the health status into the memory mailbox */
bcopy(mh, &sc->mly_mmbox->mmm_health.status, sizeof(*mh));
debug(1, "initial change counter %d, event counter %d", mh->change_counter, mh->next_event);
free(mh, M_DEVBUF);
return(0);
}
/********************************************************************************
* Enable the memory mailbox mode.
*/
static int
mly_enable_mmbox(struct mly_softc *sc)
{
struct mly_command_ioctl mci;
u_int8_t *sp, status;
int error;
debug_called(1);
/* build the ioctl and send it */
bzero(&mci, sizeof(mci));
mci.sub_ioctl = MDACIOCTL_SETMEMORYMAILBOX;
/* set buffer addresses */
mci.param.setmemorymailbox.command_mailbox_physaddr =
sc->mly_mmbox_busaddr + offsetof(struct mly_mmbox, mmm_command);
mci.param.setmemorymailbox.status_mailbox_physaddr =
sc->mly_mmbox_busaddr + offsetof(struct mly_mmbox, mmm_status);
mci.param.setmemorymailbox.health_buffer_physaddr =
sc->mly_mmbox_busaddr + offsetof(struct mly_mmbox, mmm_health);
/* set buffer sizes - abuse of data_size field is revolting */
sp = (u_int8_t *)&mci.data_size;
sp[0] = ((sizeof(union mly_command_packet) * MLY_MMBOX_COMMANDS) / 1024);
sp[1] = (sizeof(union mly_status_packet) * MLY_MMBOX_STATUS) / 1024;
mci.param.setmemorymailbox.health_buffer_size = sizeof(union mly_health_region) / 1024;
debug(1, "memory mailbox at %p (0x%llx/%d 0x%llx/%d 0x%llx/%d", sc->mly_mmbox,
mci.param.setmemorymailbox.command_mailbox_physaddr, sp[0],
mci.param.setmemorymailbox.status_mailbox_physaddr, sp[1],
mci.param.setmemorymailbox.health_buffer_physaddr,
mci.param.setmemorymailbox.health_buffer_size);
if ((error = mly_ioctl(sc, &mci, NULL, 0, &status, NULL, NULL)))
return(error);
if (status != 0)
return(EIO);
sc->mly_state |= MLY_STATE_MMBOX_ACTIVE;
debug(1, "memory mailbox active");
return(0);
}
/********************************************************************************
* Flush all pending I/O from the controller.
*/
static int
mly_flush(struct mly_softc *sc)
{
struct mly_command_ioctl mci;
u_int8_t status;
int error;
debug_called(1);
/* build the ioctl */
bzero(&mci, sizeof(mci));
mci.sub_ioctl = MDACIOCTL_FLUSHDEVICEDATA;
mci.param.deviceoperation.operation_device = MLY_OPDEVICE_PHYSICAL_CONTROLLER;
/* pass it off to the controller */
if ((error = mly_ioctl(sc, &mci, NULL, 0, &status, NULL, NULL)))
return(error);
return((status == 0) ? 0 : EIO);
}
/********************************************************************************
* Perform an ioctl command.
*
* If (data) is not NULL, the command requires data transfer. If (*data) is NULL
* the command requires data transfer from the controller, and we will allocate
* a buffer for it. If (*data) is not NULL, the command requires data transfer
* to the controller.
*
* XXX passing in the whole ioctl structure is ugly. Better ideas?
*
* XXX we don't even try to handle the case where datasize > 4k. We should.
*/
static int
mly_ioctl(struct mly_softc *sc, struct mly_command_ioctl *ioctl, void **data, size_t datasize,
u_int8_t *status, void *sense_buffer, size_t *sense_length)
{
struct mly_command *mc;
struct mly_command_ioctl *mci;
int error;
debug_called(1);
MLY_ASSERT_LOCKED(sc);
mc = NULL;
if (mly_alloc_command(sc, &mc)) {
error = ENOMEM;
goto out;
}
/* copy the ioctl structure, but save some important fields and then fixup */
mci = &mc->mc_packet->ioctl;
ioctl->sense_buffer_address = mci->sense_buffer_address;
ioctl->maximum_sense_size = mci->maximum_sense_size;
*mci = *ioctl;
mci->opcode = MDACMD_IOCTL;
mci->timeout.value = 30;
mci->timeout.scale = MLY_TIMEOUT_SECONDS;
/* handle the data buffer */
if (data != NULL) {
if (*data == NULL) {
/* allocate data buffer */
if ((mc->mc_data = malloc(datasize, M_DEVBUF, M_NOWAIT)) == NULL) {
error = ENOMEM;
goto out;
}
mc->mc_flags |= MLY_CMD_DATAIN;
} else {
mc->mc_data = *data;
mc->mc_flags |= MLY_CMD_DATAOUT;
}
mc->mc_length = datasize;
mc->mc_packet->generic.data_size = datasize;
}
/* run the command */
if ((error = mly_immediate_command(mc)))
goto out;
/* clean up and return any data */
*status = mc->mc_status;
if ((mc->mc_sense > 0) && (sense_buffer != NULL)) {
bcopy(mc->mc_packet, sense_buffer, mc->mc_sense);
*sense_length = mc->mc_sense;
goto out;
}
/* should we return a data pointer? */
if ((data != NULL) && (*data == NULL))
*data = mc->mc_data;
/* command completed OK */
error = 0;
out:
if (mc != NULL) {
/* do we need to free a data buffer we allocated? */
if (error && (mc->mc_data != NULL) && (*data == NULL))
free(mc->mc_data, M_DEVBUF);
mly_release_command(mc);
}
return(error);
}
/********************************************************************************
* Check for event(s) outstanding in the controller.
*/
static void
mly_check_event(struct mly_softc *sc)
{
/*
* The controller may have updated the health status information,
* so check for it here. Note that the counters are all in host memory,
* so this check is very cheap. Also note that we depend on checking on
* completion
*/
if (sc->mly_mmbox->mmm_health.status.change_counter != sc->mly_event_change) {
sc->mly_event_change = sc->mly_mmbox->mmm_health.status.change_counter;
debug(1, "event change %d, event status update, %d -> %d", sc->mly_event_change,
sc->mly_event_waiting, sc->mly_mmbox->mmm_health.status.next_event);
sc->mly_event_waiting = sc->mly_mmbox->mmm_health.status.next_event;
/* wake up anyone that might be interested in this */
wakeup(&sc->mly_event_change);
}
if (sc->mly_event_counter != sc->mly_event_waiting)
mly_fetch_event(sc);
}
/********************************************************************************
* Fetch one event from the controller.
*
* If we fail due to resource starvation, we'll be retried the next time a
* command completes.
*/
static void
mly_fetch_event(struct mly_softc *sc)
{
struct mly_command *mc;
struct mly_command_ioctl *mci;
int s;
u_int32_t event;
debug_called(1);
/* get a command */
if (mly_alloc_command(sc, &mc))
return;
/* set up the data buffer */
if ((mc->mc_data = malloc(sizeof(struct mly_event), M_DEVBUF, M_NOWAIT | M_ZERO)) == NULL) {
mly_release_command(mc);
return;
}
mc->mc_length = sizeof(struct mly_event);
mc->mc_flags |= MLY_CMD_DATAIN;
mc->mc_complete = mly_complete_event;
/*
* Get an event number to fetch. It's possible that we've raced with another
* context for the last event, in which case there will be no more events.
*/
s = splcam();
if (sc->mly_event_counter == sc->mly_event_waiting) {
mly_release_command(mc);
splx(s);
return;
}
event = sc->mly_event_counter++;
splx(s);
/*
* Build the ioctl.
*
* At this point we are committed to sending this request, as it
* will be the only one constructed for this particular event number.
*/
mci = (struct mly_command_ioctl *)&mc->mc_packet->ioctl;
mci->opcode = MDACMD_IOCTL;
mci->data_size = sizeof(struct mly_event);
mci->addr.phys.lun = (event >> 16) & 0xff;
mci->addr.phys.target = (event >> 24) & 0xff;
mci->addr.phys.channel = 0;
mci->addr.phys.controller = 0;
mci->timeout.value = 30;
mci->timeout.scale = MLY_TIMEOUT_SECONDS;
mci->sub_ioctl = MDACIOCTL_GETEVENT;
mci->param.getevent.sequence_number_low = event & 0xffff;
debug(1, "fetch event %u", event);
/*
* Submit the command.
*
* Note that failure of mly_start() will result in this event never being
* fetched.
*/
if (mly_start(mc) != 0) {
mly_printf(sc, "couldn't fetch event %u\n", event);
mly_release_command(mc);
}
}
/********************************************************************************
* Handle the completion of an event poll.
*/
static void
mly_complete_event(struct mly_command *mc)
{
struct mly_softc *sc = mc->mc_sc;
struct mly_event *me = (struct mly_event *)mc->mc_data;
debug_called(1);
/*
* If the event was successfully fetched, process it.
*/
if (mc->mc_status == SCSI_STATUS_OK) {
mly_process_event(sc, me);
free(me, M_DEVBUF);
}
mly_release_command(mc);
/*
* Check for another event.
*/
mly_check_event(sc);
}
/********************************************************************************
* Process a controller event.
*/
static void
mly_process_event(struct mly_softc *sc, struct mly_event *me)
{
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
struct scsi_sense_data_fixed *ssd;
char *fp, *tp;
int bus, target, event, class, action;
ssd = (struct scsi_sense_data_fixed *)&me->sense[0];
/*
* Errors can be reported using vendor-unique sense data. In this case, the
* event code will be 0x1c (Request sense data present), the sense key will
* be 0x09 (vendor specific), the MSB of the ASC will be set, and the
* actual event code will be a 16-bit value comprised of the ASCQ (low byte)
* and low seven bits of the ASC (low seven bits of the high byte).
*/
if ((me->code == 0x1c) &&
((ssd->flags & SSD_KEY) == SSD_KEY_Vendor_Specific) &&
(ssd->add_sense_code & 0x80)) {
event = ((int)(ssd->add_sense_code & ~0x80) << 8) + ssd->add_sense_code_qual;
} else {
event = me->code;
}
/* look up event, get codes */
fp = mly_describe_code(mly_table_event, event);
debug(1, "Event %d code 0x%x", me->sequence_number, me->code);
/* quiet event? */
class = fp[0];
if (isupper(class) && bootverbose)
class = tolower(class);
/* get action code, text string */
action = fp[1];
tp = &fp[2];
/*
* Print some information about the event.
*
* This code uses a table derived from the corresponding portion of the Linux
* driver, and thus the parser is very similar.
*/
switch(class) {
case 'p': /* error on physical device */
mly_printf(sc, "physical device %d:%d %s\n", me->channel, me->target, tp);
if (action == 'r')
sc->mly_btl[me->channel][me->target].mb_flags |= MLY_BTL_RESCAN;
break;
case 'l': /* error on logical unit */
case 'm': /* message about logical unit */
bus = MLY_LOGDEV_BUS(sc, me->lun);
target = MLY_LOGDEV_TARGET(sc, me->lun);
mly_name_device(sc, bus, target);
mly_printf(sc, "logical device %d (%s) %s\n", me->lun, sc->mly_btl[bus][target].mb_name, tp);
if (action == 'r')
sc->mly_btl[bus][target].mb_flags |= MLY_BTL_RESCAN;
break;
case 's': /* report of sense data */
if (((ssd->flags & SSD_KEY) == SSD_KEY_NO_SENSE) ||
(((ssd->flags & SSD_KEY) == SSD_KEY_NOT_READY) &&
(ssd->add_sense_code == 0x04) &&
((ssd->add_sense_code_qual == 0x01) || (ssd->add_sense_code_qual == 0x02))))
break; /* ignore NO_SENSE or NOT_READY in one case */
mly_printf(sc, "physical device %d:%d %s\n", me->channel, me->target, tp);
mly_printf(sc, " sense key %d asc %02x ascq %02x\n",
ssd->flags & SSD_KEY, ssd->add_sense_code, ssd->add_sense_code_qual);
mly_printf(sc, " info %4D csi %4D\n", ssd->info, "", ssd->cmd_spec_info, "");
if (action == 'r')
sc->mly_btl[me->channel][me->target].mb_flags |= MLY_BTL_RESCAN;
break;
case 'e':
mly_printf(sc, tp, me->target, me->lun);
2003-02-11 08:19:52 +00:00
printf("\n");
break;
case 'c':
mly_printf(sc, "controller %s\n", tp);
break;
case '?':
mly_printf(sc, "%s - %d\n", tp, me->code);
break;
default: /* probably a 'noisy' event being ignored */
break;
}
}
/********************************************************************************
* Perform periodic activities.
*/
static void
mly_periodic(void *data)
{
struct mly_softc *sc = (struct mly_softc *)data;
int bus, target;
debug_called(2);
MLY_ASSERT_LOCKED(sc);
/*
* Scan devices.
*/
for (bus = 0; bus < sc->mly_cam_channels; bus++) {
if (MLY_BUS_IS_VALID(sc, bus)) {
for (target = 0; target < MLY_MAX_TARGETS; target++) {
/* ignore the controller in this scan */
if (target == sc->mly_controllerparam->initiator_id)
continue;
/* perform device rescan? */
if (sc->mly_btl[bus][target].mb_flags & MLY_BTL_RESCAN)
mly_rescan_btl(sc, bus, target);
}
}
}
/* check for controller events */
mly_check_event(sc);
/* reschedule ourselves */
callout_schedule(&sc->mly_periodic, MLY_PERIODIC_INTERVAL * hz);
}
/********************************************************************************
********************************************************************************
Command Processing
********************************************************************************
********************************************************************************/
/********************************************************************************
* Run a command and wait for it to complete.
*
*/
static int
mly_immediate_command(struct mly_command *mc)
{
struct mly_softc *sc = mc->mc_sc;
int error;
debug_called(1);
MLY_ASSERT_LOCKED(sc);
if ((error = mly_start(mc))) {
return(error);
}
if (sc->mly_state & MLY_STATE_INTERRUPTS_ON) {
/* sleep on the command */
while(!(mc->mc_flags & MLY_CMD_COMPLETE)) {
mtx_sleep(mc, &sc->mly_lock, PRIBIO, "mlywait", 0);
}
} else {
/* spin and collect status while we do */
while(!(mc->mc_flags & MLY_CMD_COMPLETE)) {
mly_done(mc->mc_sc);
}
}
return(0);
}
/********************************************************************************
* Deliver a command to the controller.
*
* XXX it would be good to just queue commands that we can't submit immediately
* and send them later, but we probably want a wrapper for that so that
* we don't hang on a failed submission for an immediate command.
*/
static int
mly_start(struct mly_command *mc)
{
struct mly_softc *sc = mc->mc_sc;
union mly_command_packet *pkt;
debug_called(2);
MLY_ASSERT_LOCKED(sc);
/*
* Set the command up for delivery to the controller.
*/
mly_map_command(mc);
mc->mc_packet->generic.command_id = mc->mc_slot;
#ifdef MLY_DEBUG
mc->mc_timestamp = time_second;
#endif
/*
* Do we have to use the hardware mailbox?
*/
if (!(sc->mly_state & MLY_STATE_MMBOX_ACTIVE)) {
/*
* Check to see if the controller is ready for us.
*/
if (MLY_IDBR_TRUE(sc, MLY_HM_CMDSENT)) {
return(EBUSY);
}
mc->mc_flags |= MLY_CMD_BUSY;
/*
* It's ready, send the command.
*/
MLY_SET_MBOX(sc, sc->mly_command_mailbox, &mc->mc_packetphys);
MLY_SET_REG(sc, sc->mly_idbr, MLY_HM_CMDSENT);
} else { /* use memory-mailbox mode */
pkt = &sc->mly_mmbox->mmm_command[sc->mly_mmbox_command_index];
/* check to see if the next index is free yet */
if (pkt->mmbox.flag != 0) {
return(EBUSY);
}
mc->mc_flags |= MLY_CMD_BUSY;
/* copy in new command */
bcopy(mc->mc_packet->mmbox.data, pkt->mmbox.data, sizeof(pkt->mmbox.data));
/* barrier to ensure completion of previous write before we write the flag */
bus_barrier(sc->mly_regs_resource, 0, 0, BUS_SPACE_BARRIER_WRITE);
/* copy flag last */
pkt->mmbox.flag = mc->mc_packet->mmbox.flag;
/* barrier to ensure completion of previous write before we notify the controller */
bus_barrier(sc->mly_regs_resource, 0, 0, BUS_SPACE_BARRIER_WRITE);
/* signal controller, update index */
MLY_SET_REG(sc, sc->mly_idbr, MLY_AM_CMDSENT);
sc->mly_mmbox_command_index = (sc->mly_mmbox_command_index + 1) % MLY_MMBOX_COMMANDS;
}
mly_enqueue_busy(mc);
return(0);
}
/********************************************************************************
* Pick up command status from the controller, schedule a completion event
*/
static void
mly_done(struct mly_softc *sc)
{
struct mly_command *mc;
union mly_status_packet *sp;
u_int16_t slot;
int worked;
MLY_ASSERT_LOCKED(sc);
worked = 0;
/* pick up hardware-mailbox commands */
if (MLY_ODBR_TRUE(sc, MLY_HM_STSREADY)) {
slot = MLY_GET_REG2(sc, sc->mly_status_mailbox);
if (slot < MLY_SLOT_MAX) {
mc = &sc->mly_command[slot - MLY_SLOT_START];
mc->mc_status = MLY_GET_REG(sc, sc->mly_status_mailbox + 2);
mc->mc_sense = MLY_GET_REG(sc, sc->mly_status_mailbox + 3);
mc->mc_resid = MLY_GET_REG4(sc, sc->mly_status_mailbox + 4);
mly_remove_busy(mc);
mc->mc_flags &= ~MLY_CMD_BUSY;
mly_enqueue_complete(mc);
worked = 1;
} else {
/* slot 0xffff may mean "extremely bogus command" */
mly_printf(sc, "got HM completion for illegal slot %u\n", slot);
}
/* unconditionally acknowledge status */
MLY_SET_REG(sc, sc->mly_odbr, MLY_HM_STSREADY);
MLY_SET_REG(sc, sc->mly_idbr, MLY_HM_STSACK);
}
/* pick up memory-mailbox commands */
if (MLY_ODBR_TRUE(sc, MLY_AM_STSREADY)) {
for (;;) {
sp = &sc->mly_mmbox->mmm_status[sc->mly_mmbox_status_index];
/* check for more status */
if (sp->mmbox.flag == 0)
break;
/* get slot number */
slot = sp->status.command_id;
if (slot < MLY_SLOT_MAX) {
mc = &sc->mly_command[slot - MLY_SLOT_START];
mc->mc_status = sp->status.status;
mc->mc_sense = sp->status.sense_length;
mc->mc_resid = sp->status.residue;
mly_remove_busy(mc);
mc->mc_flags &= ~MLY_CMD_BUSY;
mly_enqueue_complete(mc);
worked = 1;
} else {
/* slot 0xffff may mean "extremely bogus command" */
mly_printf(sc, "got AM completion for illegal slot %u at %d\n",
slot, sc->mly_mmbox_status_index);
}
/* clear and move to next index */
sp->mmbox.flag = 0;
sc->mly_mmbox_status_index = (sc->mly_mmbox_status_index + 1) % MLY_MMBOX_STATUS;
}
/* acknowledge that we have collected status value(s) */
MLY_SET_REG(sc, sc->mly_odbr, MLY_AM_STSREADY);
}
if (worked) {
if (sc->mly_state & MLY_STATE_INTERRUPTS_ON)
taskqueue_enqueue(taskqueue_thread, &sc->mly_task_complete);
else
mly_complete(sc);
}
}
/********************************************************************************
* Process completed commands
*/
static void
mly_complete_handler(void *context, int pending)
{
struct mly_softc *sc = (struct mly_softc *)context;
MLY_LOCK(sc);
mly_complete(sc);
MLY_UNLOCK(sc);
}
static void
mly_complete(struct mly_softc *sc)
{
struct mly_command *mc;
void (* mc_complete)(struct mly_command *mc);
debug_called(2);
/*
* Spin pulling commands off the completed queue and processing them.
*/
while ((mc = mly_dequeue_complete(sc)) != NULL) {
/*
* Free controller resources, mark command complete.
*
* Note that as soon as we mark the command complete, it may be freed
* out from under us, so we need to save the mc_complete field in
* order to later avoid dereferencing mc. (We would not expect to
* have a polling/sleeping consumer with mc_complete != NULL).
*/
mly_unmap_command(mc);
mc_complete = mc->mc_complete;
mc->mc_flags |= MLY_CMD_COMPLETE;
/*
* Call completion handler or wake up sleeping consumer.
*/
if (mc_complete != NULL) {
mc_complete(mc);
} else {
wakeup(mc);
}
}
/*
* XXX if we are deferring commands due to controller-busy status, we should
* retry submitting them here.
*/
}
/********************************************************************************
********************************************************************************
Command Buffer Management
********************************************************************************
********************************************************************************/
/********************************************************************************
* Allocate a command.
*/
static int
mly_alloc_command(struct mly_softc *sc, struct mly_command **mcp)
{
struct mly_command *mc;
debug_called(3);
if ((mc = mly_dequeue_free(sc)) == NULL)
return(ENOMEM);
*mcp = mc;
return(0);
}
/********************************************************************************
* Release a command back to the freelist.
*/
static void
mly_release_command(struct mly_command *mc)
{
debug_called(3);
/*
* Fill in parts of the command that may cause confusion if
* a consumer doesn't when we are later allocated.
*/
mc->mc_data = NULL;
mc->mc_flags = 0;
mc->mc_complete = NULL;
mc->mc_private = NULL;
/*
* By default, we set up to overwrite the command packet with
* sense information.
*/
mc->mc_packet->generic.sense_buffer_address = mc->mc_packetphys;
mc->mc_packet->generic.maximum_sense_size = sizeof(union mly_command_packet);
mly_enqueue_free(mc);
}
/********************************************************************************
* Map helper for command allocation.
*/
static void
mly_alloc_commands_map(void *arg, bus_dma_segment_t *segs, int nseg, int error)
{
struct mly_softc *sc = (struct mly_softc *)arg;
debug_called(1);
sc->mly_packetphys = segs[0].ds_addr;
}
/********************************************************************************
* Allocate and initialise command and packet structures.
*
* If the controller supports fewer than MLY_MAX_COMMANDS commands, limit our
* allocation to that number. If we don't yet know how many commands the
* controller supports, allocate a very small set (suitable for initialisation
* purposes only).
*/
static int
mly_alloc_commands(struct mly_softc *sc)
{
struct mly_command *mc;
int i, ncmd;
if (sc->mly_controllerinfo == NULL) {
ncmd = 4;
} else {
ncmd = min(MLY_MAX_COMMANDS, sc->mly_controllerinfo->maximum_parallel_commands);
}
/*
* Allocate enough space for all the command packets in one chunk and
* map them permanently into controller-visible space.
*/
if (bus_dmamem_alloc(sc->mly_packet_dmat, (void **)&sc->mly_packet,
BUS_DMA_NOWAIT, &sc->mly_packetmap)) {
return(ENOMEM);
}
if (bus_dmamap_load(sc->mly_packet_dmat, sc->mly_packetmap, sc->mly_packet,
ncmd * sizeof(union mly_command_packet),
mly_alloc_commands_map, sc, BUS_DMA_NOWAIT) != 0)
return (ENOMEM);
for (i = 0; i < ncmd; i++) {
mc = &sc->mly_command[i];
bzero(mc, sizeof(*mc));
mc->mc_sc = sc;
mc->mc_slot = MLY_SLOT_START + i;
mc->mc_packet = sc->mly_packet + i;
mc->mc_packetphys = sc->mly_packetphys + (i * sizeof(union mly_command_packet));
if (!bus_dmamap_create(sc->mly_buffer_dmat, 0, &mc->mc_datamap))
mly_release_command(mc);
}
return(0);
}
/********************************************************************************
* Free all the storage held by commands.
*
* Must be called with all commands on the free list.
*/
static void
mly_release_commands(struct mly_softc *sc)
{
struct mly_command *mc;
/* throw away command buffer DMA maps */
while (mly_alloc_command(sc, &mc) == 0)
bus_dmamap_destroy(sc->mly_buffer_dmat, mc->mc_datamap);
/* release the packet storage */
if (sc->mly_packet != NULL) {
bus_dmamap_unload(sc->mly_packet_dmat, sc->mly_packetmap);
bus_dmamem_free(sc->mly_packet_dmat, sc->mly_packet, sc->mly_packetmap);
sc->mly_packet = NULL;
}
}
/********************************************************************************
* Command-mapping helper function - populate this command's s/g table
* with the s/g entries for its data.
*/
static void
mly_map_command_sg(void *arg, bus_dma_segment_t *segs, int nseg, int error)
{
struct mly_command *mc = (struct mly_command *)arg;
struct mly_softc *sc = mc->mc_sc;
struct mly_command_generic *gen = &(mc->mc_packet->generic);
struct mly_sg_entry *sg;
int i, tabofs;
debug_called(2);
/* can we use the transfer structure directly? */
if (nseg <= 2) {
sg = &gen->transfer.direct.sg[0];
gen->command_control.extended_sg_table = 0;
} else {
tabofs = ((mc->mc_slot - MLY_SLOT_START) * MLY_MAX_SGENTRIES);
sg = sc->mly_sg_table + tabofs;
gen->transfer.indirect.entries[0] = nseg;
gen->transfer.indirect.table_physaddr[0] = sc->mly_sg_busaddr + (tabofs * sizeof(struct mly_sg_entry));
gen->command_control.extended_sg_table = 1;
}
/* copy the s/g table */
for (i = 0; i < nseg; i++) {
sg[i].physaddr = segs[i].ds_addr;
sg[i].length = segs[i].ds_len;
}
}
#if 0
/********************************************************************************
* Command-mapping helper function - save the cdb's physical address.
*
* We don't support 'large' SCSI commands at this time, so this is unused.
*/
static void
mly_map_command_cdb(void *arg, bus_dma_segment_t *segs, int nseg, int error)
{
struct mly_command *mc = (struct mly_command *)arg;
debug_called(2);
/* XXX can we safely assume that a CDB will never cross a page boundary? */
if ((segs[0].ds_addr % PAGE_SIZE) >
((segs[0].ds_addr + mc->mc_packet->scsi_large.cdb_length) % PAGE_SIZE))
panic("cdb crosses page boundary");
/* fix up fields in the command packet */
mc->mc_packet->scsi_large.cdb_physaddr = segs[0].ds_addr;
}
#endif
/********************************************************************************
* Map a command into controller-visible space
*/
static void
mly_map_command(struct mly_command *mc)
{
struct mly_softc *sc = mc->mc_sc;
debug_called(2);
/* don't map more than once */
if (mc->mc_flags & MLY_CMD_MAPPED)
return;
/* does the command have a data buffer? */
if (mc->mc_data != NULL) {
if (mc->mc_flags & MLY_CMD_CCB)
bus_dmamap_load_ccb(sc->mly_buffer_dmat, mc->mc_datamap,
mc->mc_data, mly_map_command_sg, mc, 0);
else
bus_dmamap_load(sc->mly_buffer_dmat, mc->mc_datamap,
mc->mc_data, mc->mc_length,
mly_map_command_sg, mc, 0);
if (mc->mc_flags & MLY_CMD_DATAIN)
bus_dmamap_sync(sc->mly_buffer_dmat, mc->mc_datamap, BUS_DMASYNC_PREREAD);
if (mc->mc_flags & MLY_CMD_DATAOUT)
bus_dmamap_sync(sc->mly_buffer_dmat, mc->mc_datamap, BUS_DMASYNC_PREWRITE);
}
mc->mc_flags |= MLY_CMD_MAPPED;
}
/********************************************************************************
* Unmap a command from controller-visible space
*/
static void
mly_unmap_command(struct mly_command *mc)
{
struct mly_softc *sc = mc->mc_sc;
debug_called(2);
if (!(mc->mc_flags & MLY_CMD_MAPPED))
return;
/* does the command have a data buffer? */
if (mc->mc_data != NULL) {
if (mc->mc_flags & MLY_CMD_DATAIN)
bus_dmamap_sync(sc->mly_buffer_dmat, mc->mc_datamap, BUS_DMASYNC_POSTREAD);
if (mc->mc_flags & MLY_CMD_DATAOUT)
bus_dmamap_sync(sc->mly_buffer_dmat, mc->mc_datamap, BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(sc->mly_buffer_dmat, mc->mc_datamap);
}
mc->mc_flags &= ~MLY_CMD_MAPPED;
}
/********************************************************************************
********************************************************************************
CAM interface
********************************************************************************
********************************************************************************/
/********************************************************************************
* Attach the physical and virtual SCSI busses to CAM.
*
* Physical bus numbering starts from 0, virtual bus numbering from one greater
* than the highest physical bus. Physical busses are only registered if
* the kernel environment variable "hw.mly.register_physical_channels" is set.
*
* When we refer to a "bus", we are referring to the bus number registered with
* the SIM, wheras a "channel" is a channel number given to the adapter. In order
* to keep things simple, we map these 1:1, so "bus" and "channel" may be used
* interchangeably.
*/
static int
mly_cam_attach(struct mly_softc *sc)
{
struct cam_devq *devq;
int chn, i;
debug_called(1);
/*
* Allocate a devq for all our channels combined.
*/
if ((devq = cam_simq_alloc(sc->mly_controllerinfo->maximum_parallel_commands)) == NULL) {
mly_printf(sc, "can't allocate CAM SIM queue\n");
return(ENOMEM);
}
/*
* If physical channel registration has been requested, register these first.
* Note that we enable tagged command queueing for physical channels.
*/
if (testenv("hw.mly.register_physical_channels")) {
chn = 0;
for (i = 0; i < sc->mly_controllerinfo->physical_channels_present; i++, chn++) {
if ((sc->mly_cam_sim[chn] = cam_sim_alloc(mly_cam_action, mly_cam_poll, "mly", sc,
device_get_unit(sc->mly_dev),
&sc->mly_lock,
sc->mly_controllerinfo->maximum_parallel_commands,
1, devq)) == NULL) {
return(ENOMEM);
}
MLY_LOCK(sc);
if (xpt_bus_register(sc->mly_cam_sim[chn], sc->mly_dev, chn)) {
MLY_UNLOCK(sc);
mly_printf(sc, "CAM XPT phsyical channel registration failed\n");
return(ENXIO);
}
MLY_UNLOCK(sc);
debug(1, "registered physical channel %d", chn);
}
}
/*
* Register our virtual channels, with bus numbers matching channel numbers.
*/
chn = sc->mly_controllerinfo->physical_channels_present;
for (i = 0; i < sc->mly_controllerinfo->virtual_channels_present; i++, chn++) {
if ((sc->mly_cam_sim[chn] = cam_sim_alloc(mly_cam_action, mly_cam_poll, "mly", sc,
device_get_unit(sc->mly_dev),
&sc->mly_lock,
sc->mly_controllerinfo->maximum_parallel_commands,
0, devq)) == NULL) {
return(ENOMEM);
}
MLY_LOCK(sc);
if (xpt_bus_register(sc->mly_cam_sim[chn], sc->mly_dev, chn)) {
MLY_UNLOCK(sc);
mly_printf(sc, "CAM XPT virtual channel registration failed\n");
return(ENXIO);
}
MLY_UNLOCK(sc);
debug(1, "registered virtual channel %d", chn);
}
/*
* This is the total number of channels that (might have been) registered with
* CAM. Some may not have been; check the mly_cam_sim array to be certain.
*/
sc->mly_cam_channels = sc->mly_controllerinfo->physical_channels_present +
sc->mly_controllerinfo->virtual_channels_present;
return(0);
}
/********************************************************************************
* Detach from CAM
*/
static void
mly_cam_detach(struct mly_softc *sc)
{
int i;
debug_called(1);
MLY_LOCK(sc);
for (i = 0; i < sc->mly_cam_channels; i++) {
if (sc->mly_cam_sim[i] != NULL) {
xpt_bus_deregister(cam_sim_path(sc->mly_cam_sim[i]));
cam_sim_free(sc->mly_cam_sim[i], 0);
}
}
MLY_UNLOCK(sc);
if (sc->mly_cam_devq != NULL)
cam_simq_free(sc->mly_cam_devq);
}
/************************************************************************
* Rescan a device.
*/
static void
mly_cam_rescan_btl(struct mly_softc *sc, int bus, int target)
{
union ccb *ccb;
debug_called(1);
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
if ((ccb = xpt_alloc_ccb()) == NULL) {
mly_printf(sc, "rescan failed (can't allocate CCB)\n");
return;
}
if (xpt_create_path(&ccb->ccb_h.path, NULL,
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
cam_sim_path(sc->mly_cam_sim[bus]), target, 0) != CAM_REQ_CMP) {
mly_printf(sc, "rescan failed (can't create path)\n");
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
xpt_free_ccb(ccb);
return;
}
debug(1, "rescan target %d:%d", bus, target);
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
xpt_rescan(ccb);
}
/********************************************************************************
* Handle an action requested by CAM
*/
static void
mly_cam_action(struct cam_sim *sim, union ccb *ccb)
{
struct mly_softc *sc = cam_sim_softc(sim);
debug_called(2);
MLY_ASSERT_LOCKED(sc);
switch (ccb->ccb_h.func_code) {
/* perform SCSI I/O */
case XPT_SCSI_IO:
if (!mly_cam_action_io(sim, (struct ccb_scsiio *)&ccb->csio))
return;
break;
/* perform geometry calculations */
case XPT_CALC_GEOMETRY:
{
struct ccb_calc_geometry *ccg = &ccb->ccg;
u_int32_t secs_per_cylinder;
debug(2, "XPT_CALC_GEOMETRY %d:%d:%d", cam_sim_bus(sim), ccb->ccb_h.target_id, ccb->ccb_h.target_lun);
if (sc->mly_controllerparam->bios_geometry == MLY_BIOSGEOM_8G) {
ccg->heads = 255;
ccg->secs_per_track = 63;
} else { /* MLY_BIOSGEOM_2G */
ccg->heads = 128;
ccg->secs_per_track = 32;
}
secs_per_cylinder = ccg->heads * ccg->secs_per_track;
ccg->cylinders = ccg->volume_size / secs_per_cylinder;
ccb->ccb_h.status = CAM_REQ_CMP;
break;
}
/* handle path attribute inquiry */
case XPT_PATH_INQ:
{
struct ccb_pathinq *cpi = &ccb->cpi;
debug(2, "XPT_PATH_INQ %d:%d:%d", cam_sim_bus(sim), ccb->ccb_h.target_id, ccb->ccb_h.target_lun);
cpi->version_num = 1;
cpi->hba_inquiry = PI_TAG_ABLE; /* XXX extra flags for physical channels? */
cpi->target_sprt = 0;
cpi->hba_misc = 0;
cpi->max_target = MLY_MAX_TARGETS - 1;
cpi->max_lun = MLY_MAX_LUNS - 1;
cpi->initiator_id = sc->mly_controllerparam->initiator_id;
strncpy(cpi->sim_vid, "FreeBSD", SIM_IDLEN);
strncpy(cpi->hba_vid, "FreeBSD", 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 = 132 * 1024; /* XXX what to set this to? */
cpi->transport = XPORT_SPI;
cpi->transport_version = 2;
cpi->protocol = PROTO_SCSI;
cpi->protocol_version = SCSI_REV_2;
ccb->ccb_h.status = CAM_REQ_CMP;
break;
}
case XPT_GET_TRAN_SETTINGS:
{
struct ccb_trans_settings *cts = &ccb->cts;
int bus, target;
struct ccb_trans_settings_scsi *scsi = &cts->proto_specific.scsi;
struct ccb_trans_settings_spi *spi = &cts->xport_specific.spi;
cts->protocol = PROTO_SCSI;
cts->protocol_version = SCSI_REV_2;
cts->transport = XPORT_SPI;
cts->transport_version = 2;
scsi->flags = 0;
scsi->valid = 0;
spi->flags = 0;
spi->valid = 0;
bus = cam_sim_bus(sim);
target = cts->ccb_h.target_id;
debug(2, "XPT_GET_TRAN_SETTINGS %d:%d", bus, target);
/* logical device? */
if (sc->mly_btl[bus][target].mb_flags & MLY_BTL_LOGICAL) {
/* nothing special for these */
/* physical device? */
} else if (sc->mly_btl[bus][target].mb_flags & MLY_BTL_PHYSICAL) {
/* allow CAM to try tagged transactions */
scsi->flags |= CTS_SCSI_FLAGS_TAG_ENB;
scsi->valid |= CTS_SCSI_VALID_TQ;
/* convert speed (MHz) to usec */
if (sc->mly_btl[bus][target].mb_speed == 0) {
spi->sync_period = 1000000 / 5;
} else {
spi->sync_period = 1000000 / sc->mly_btl[bus][target].mb_speed;
}
/* convert bus width to CAM internal encoding */
switch (sc->mly_btl[bus][target].mb_width) {
case 32:
spi->bus_width = MSG_EXT_WDTR_BUS_32_BIT;
break;
case 16:
spi->bus_width = MSG_EXT_WDTR_BUS_16_BIT;
break;
case 8:
default:
spi->bus_width = MSG_EXT_WDTR_BUS_8_BIT;
break;
}
spi->valid |= CTS_SPI_VALID_SYNC_RATE | CTS_SPI_VALID_BUS_WIDTH;
/* not a device, bail out */
} else {
cts->ccb_h.status = CAM_REQ_CMP_ERR;
break;
}
/* disconnect always OK */
spi->flags |= CTS_SPI_FLAGS_DISC_ENB;
spi->valid |= CTS_SPI_VALID_DISC;
cts->ccb_h.status = CAM_REQ_CMP;
break;
}
default: /* we can't do this */
debug(2, "unspported func_code = 0x%x", ccb->ccb_h.func_code);
ccb->ccb_h.status = CAM_REQ_INVALID;
break;
}
xpt_done(ccb);
}
/********************************************************************************
* Handle an I/O operation requested by CAM
*/
static int
mly_cam_action_io(struct cam_sim *sim, struct ccb_scsiio *csio)
{
struct mly_softc *sc = cam_sim_softc(sim);
struct mly_command *mc;
struct mly_command_scsi_small *ss;
int bus, target;
int error;
bus = cam_sim_bus(sim);
target = csio->ccb_h.target_id;
debug(2, "XPT_SCSI_IO %d:%d:%d", bus, target, csio->ccb_h.target_lun);
/* validate bus number */
if (!MLY_BUS_IS_VALID(sc, bus)) {
debug(0, " invalid bus %d", bus);
csio->ccb_h.status = CAM_REQ_CMP_ERR;
}
/* check for I/O attempt to a protected device */
if (sc->mly_btl[bus][target].mb_flags & MLY_BTL_PROTECTED) {
debug(2, " device protected");
csio->ccb_h.status = CAM_REQ_CMP_ERR;
}
/* check for I/O attempt to nonexistent device */
if (!(sc->mly_btl[bus][target].mb_flags & (MLY_BTL_LOGICAL | MLY_BTL_PHYSICAL))) {
debug(2, " device %d:%d does not exist", bus, target);
csio->ccb_h.status = CAM_REQ_CMP_ERR;
}
/* XXX increase if/when we support large SCSI commands */
if (csio->cdb_len > MLY_CMD_SCSI_SMALL_CDB) {
debug(0, " command too large (%d > %d)", csio->cdb_len, MLY_CMD_SCSI_SMALL_CDB);
csio->ccb_h.status = CAM_REQ_CMP_ERR;
}
/* check that the CDB pointer is not to a physical address */
if ((csio->ccb_h.flags & CAM_CDB_POINTER) && (csio->ccb_h.flags & CAM_CDB_PHYS)) {
debug(0, " CDB pointer is to physical address");
csio->ccb_h.status = CAM_REQ_CMP_ERR;
}
/* abandon aborted ccbs or those that have failed validation */
if ((csio->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_INPROG) {
debug(2, "abandoning CCB due to abort/validation failure");
return(EINVAL);
}
/*
* Get a command, or push the ccb back to CAM and freeze the queue.
*/
if ((error = mly_alloc_command(sc, &mc))) {
xpt_freeze_simq(sim, 1);
csio->ccb_h.status |= CAM_REQUEUE_REQ;
sc->mly_qfrzn_cnt++;
return(error);
}
/* build the command */
mc->mc_data = csio;
mc->mc_length = csio->dxfer_len;
mc->mc_complete = mly_cam_complete;
mc->mc_private = csio;
mc->mc_flags |= MLY_CMD_CCB;
/* XXX This code doesn't set the data direction in mc_flags. */
/* save the bus number in the ccb for later recovery XXX should be a better way */
csio->ccb_h.sim_priv.entries[0].field = bus;
/* build the packet for the controller */
ss = &mc->mc_packet->scsi_small;
ss->opcode = MDACMD_SCSI;
if (csio->ccb_h.flags & CAM_DIS_DISCONNECT)
ss->command_control.disable_disconnect = 1;
if ((csio->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_OUT)
ss->command_control.data_direction = MLY_CCB_WRITE;
ss->data_size = csio->dxfer_len;
ss->addr.phys.lun = csio->ccb_h.target_lun;
ss->addr.phys.target = csio->ccb_h.target_id;
ss->addr.phys.channel = bus;
if (csio->ccb_h.timeout < (60 * 1000)) {
ss->timeout.value = csio->ccb_h.timeout / 1000;
ss->timeout.scale = MLY_TIMEOUT_SECONDS;
} else if (csio->ccb_h.timeout < (60 * 60 * 1000)) {
ss->timeout.value = csio->ccb_h.timeout / (60 * 1000);
ss->timeout.scale = MLY_TIMEOUT_MINUTES;
} else {
ss->timeout.value = csio->ccb_h.timeout / (60 * 60 * 1000); /* overflow? */
ss->timeout.scale = MLY_TIMEOUT_HOURS;
}
ss->maximum_sense_size = csio->sense_len;
ss->cdb_length = csio->cdb_len;
if (csio->ccb_h.flags & CAM_CDB_POINTER) {
bcopy(csio->cdb_io.cdb_ptr, ss->cdb, csio->cdb_len);
} else {
bcopy(csio->cdb_io.cdb_bytes, ss->cdb, csio->cdb_len);
}
/* give the command to the controller */
if ((error = mly_start(mc))) {
xpt_freeze_simq(sim, 1);
csio->ccb_h.status |= CAM_REQUEUE_REQ;
sc->mly_qfrzn_cnt++;
return(error);
}
return(0);
}
/********************************************************************************
* Check for possibly-completed commands.
*/
static void
mly_cam_poll(struct cam_sim *sim)
{
struct mly_softc *sc = cam_sim_softc(sim);
debug_called(2);
mly_done(sc);
}
/********************************************************************************
* Handle completion of a command - pass results back through the CCB
*/
static void
mly_cam_complete(struct mly_command *mc)
{
struct mly_softc *sc = mc->mc_sc;
struct ccb_scsiio *csio = (struct ccb_scsiio *)mc->mc_private;
struct scsi_inquiry_data *inq = (struct scsi_inquiry_data *)csio->data_ptr;
struct mly_btl *btl;
u_int8_t cmd;
int bus, target;
debug_called(2);
csio->scsi_status = mc->mc_status;
switch(mc->mc_status) {
case SCSI_STATUS_OK:
/*
* In order to report logical device type and status, we overwrite
* the result of the INQUIRY command to logical devices.
*/
bus = csio->ccb_h.sim_priv.entries[0].field;
target = csio->ccb_h.target_id;
/* XXX validate bus/target? */
if (sc->mly_btl[bus][target].mb_flags & MLY_BTL_LOGICAL) {
if (csio->ccb_h.flags & CAM_CDB_POINTER) {
cmd = *csio->cdb_io.cdb_ptr;
} else {
cmd = csio->cdb_io.cdb_bytes[0];
}
if (cmd == INQUIRY) {
btl = &sc->mly_btl[bus][target];
padstr(inq->vendor, mly_describe_code(mly_table_device_type, btl->mb_type), 8);
padstr(inq->product, mly_describe_code(mly_table_device_state, btl->mb_state), 16);
padstr(inq->revision, "", 4);
}
}
debug(2, "SCSI_STATUS_OK");
csio->ccb_h.status = CAM_REQ_CMP;
break;
case SCSI_STATUS_CHECK_COND:
debug(1, "SCSI_STATUS_CHECK_COND sense %d resid %d", mc->mc_sense, mc->mc_resid);
csio->ccb_h.status = CAM_SCSI_STATUS_ERROR;
bzero(&csio->sense_data, SSD_FULL_SIZE);
bcopy(mc->mc_packet, &csio->sense_data, mc->mc_sense);
csio->sense_len = mc->mc_sense;
csio->ccb_h.status |= CAM_AUTOSNS_VALID;
csio->resid = mc->mc_resid; /* XXX this is a signed value... */
break;
case SCSI_STATUS_BUSY:
debug(1, "SCSI_STATUS_BUSY");
csio->ccb_h.status = CAM_SCSI_BUSY;
break;
default:
debug(1, "unknown status 0x%x", csio->scsi_status);
csio->ccb_h.status = CAM_REQ_CMP_ERR;
break;
}
if (sc->mly_qfrzn_cnt) {
csio->ccb_h.status |= CAM_RELEASE_SIMQ;
sc->mly_qfrzn_cnt--;
}
xpt_done((union ccb *)csio);
mly_release_command(mc);
}
/********************************************************************************
* Find a peripheral attahed at (bus),(target)
*/
static struct cam_periph *
mly_find_periph(struct mly_softc *sc, int bus, int target)
{
struct cam_periph *periph;
struct cam_path *path;
int status;
status = xpt_create_path(&path, NULL, cam_sim_path(sc->mly_cam_sim[bus]), target, 0);
if (status == CAM_REQ_CMP) {
periph = cam_periph_find(path, NULL);
xpt_free_path(path);
} else {
periph = NULL;
}
return(periph);
}
/********************************************************************************
* Name the device at (bus)(target)
*/
static int
mly_name_device(struct mly_softc *sc, int bus, int target)
{
struct cam_periph *periph;
if ((periph = mly_find_periph(sc, bus, target)) != NULL) {
sprintf(sc->mly_btl[bus][target].mb_name, "%s%d", periph->periph_name, periph->unit_number);
return(0);
}
sc->mly_btl[bus][target].mb_name[0] = 0;
return(ENOENT);
}
/********************************************************************************
********************************************************************************
Hardware Control
********************************************************************************
********************************************************************************/
/********************************************************************************
* Handshake with the firmware while the card is being initialised.
*/
static int
mly_fwhandshake(struct mly_softc *sc)
{
u_int8_t error, param0, param1;
int spinup = 0;
debug_called(1);
/* set HM_STSACK and let the firmware initialise */
MLY_SET_REG(sc, sc->mly_idbr, MLY_HM_STSACK);
DELAY(1000); /* too short? */
/* if HM_STSACK is still true, the controller is initialising */
if (!MLY_IDBR_TRUE(sc, MLY_HM_STSACK))
return(0);
mly_printf(sc, "controller initialisation started\n");
/* spin waiting for initialisation to finish, or for a message to be delivered */
while (MLY_IDBR_TRUE(sc, MLY_HM_STSACK)) {
/* check for a message */
if (MLY_ERROR_VALID(sc)) {
error = MLY_GET_REG(sc, sc->mly_error_status) & ~MLY_MSG_EMPTY;
param0 = MLY_GET_REG(sc, sc->mly_command_mailbox);
param1 = MLY_GET_REG(sc, sc->mly_command_mailbox + 1);
switch(error) {
case MLY_MSG_SPINUP:
if (!spinup) {
mly_printf(sc, "drive spinup in progress\n");
spinup = 1; /* only print this once (should print drive being spun?) */
}
break;
case MLY_MSG_RACE_RECOVERY_FAIL:
mly_printf(sc, "mirror race recovery failed, one or more drives offline\n");
break;
case MLY_MSG_RACE_IN_PROGRESS:
mly_printf(sc, "mirror race recovery in progress\n");
break;
case MLY_MSG_RACE_ON_CRITICAL:
mly_printf(sc, "mirror race recovery on a critical drive\n");
break;
case MLY_MSG_PARITY_ERROR:
mly_printf(sc, "FATAL MEMORY PARITY ERROR\n");
return(ENXIO);
default:
mly_printf(sc, "unknown initialisation code 0x%x\n", error);
}
}
}
return(0);
}
/********************************************************************************
********************************************************************************
Debugging and Diagnostics
********************************************************************************
********************************************************************************/
/********************************************************************************
* Print some information about the controller.
*/
static void
mly_describe_controller(struct mly_softc *sc)
{
struct mly_ioctl_getcontrollerinfo *mi = sc->mly_controllerinfo;
mly_printf(sc, "%16s, %d channel%s, firmware %d.%02d-%d-%02d (%02d%02d%02d%02d), %dMB RAM\n",
mi->controller_name, mi->physical_channels_present, (mi->physical_channels_present) > 1 ? "s" : "",
mi->fw_major, mi->fw_minor, mi->fw_turn, mi->fw_build, /* XXX turn encoding? */
mi->fw_century, mi->fw_year, mi->fw_month, mi->fw_day,
mi->memory_size);
if (bootverbose) {
mly_printf(sc, "%s %s (%x), %dMHz %d-bit %.16s\n",
mly_describe_code(mly_table_oemname, mi->oem_information),
mly_describe_code(mly_table_controllertype, mi->controller_type), mi->controller_type,
mi->interface_speed, mi->interface_width, mi->interface_name);
mly_printf(sc, "%dMB %dMHz %d-bit %s%s%s, cache %dMB\n",
mi->memory_size, mi->memory_speed, mi->memory_width,
mly_describe_code(mly_table_memorytype, mi->memory_type),
mi->memory_parity ? "+parity": "",mi->memory_ecc ? "+ECC": "",
mi->cache_size);
mly_printf(sc, "CPU: %s @ %dMHz\n",
mly_describe_code(mly_table_cputype, mi->cpu[0].type), mi->cpu[0].speed);
if (mi->l2cache_size != 0)
mly_printf(sc, "%dKB L2 cache\n", mi->l2cache_size);
if (mi->exmemory_size != 0)
mly_printf(sc, "%dMB %dMHz %d-bit private %s%s%s\n",
mi->exmemory_size, mi->exmemory_speed, mi->exmemory_width,
mly_describe_code(mly_table_memorytype, mi->exmemory_type),
mi->exmemory_parity ? "+parity": "",mi->exmemory_ecc ? "+ECC": "");
mly_printf(sc, "battery backup %s\n", mi->bbu_present ? "present" : "not installed");
mly_printf(sc, "maximum data transfer %d blocks, maximum sg entries/command %d\n",
mi->maximum_block_count, mi->maximum_sg_entries);
mly_printf(sc, "logical devices present/critical/offline %d/%d/%d\n",
mi->logical_devices_present, mi->logical_devices_critical, mi->logical_devices_offline);
mly_printf(sc, "physical devices present %d\n",
mi->physical_devices_present);
mly_printf(sc, "physical disks present/offline %d/%d\n",
mi->physical_disks_present, mi->physical_disks_offline);
mly_printf(sc, "%d physical channel%s, %d virtual channel%s of %d possible\n",
mi->physical_channels_present, mi->physical_channels_present == 1 ? "" : "s",
mi->virtual_channels_present, mi->virtual_channels_present == 1 ? "" : "s",
mi->virtual_channels_possible);
mly_printf(sc, "%d parallel commands supported\n", mi->maximum_parallel_commands);
mly_printf(sc, "%dMB flash ROM, %d of %d maximum cycles\n",
mi->flash_size, mi->flash_age, mi->flash_maximum_age);
}
}
#ifdef MLY_DEBUG
/********************************************************************************
* Print some controller state
*/
static void
mly_printstate(struct mly_softc *sc)
{
mly_printf(sc, "IDBR %02x ODBR %02x ERROR %02x (%x %x %x)\n",
MLY_GET_REG(sc, sc->mly_idbr),
MLY_GET_REG(sc, sc->mly_odbr),
MLY_GET_REG(sc, sc->mly_error_status),
sc->mly_idbr,
sc->mly_odbr,
sc->mly_error_status);
mly_printf(sc, "IMASK %02x ISTATUS %02x\n",
MLY_GET_REG(sc, sc->mly_interrupt_mask),
MLY_GET_REG(sc, sc->mly_interrupt_status));
mly_printf(sc, "COMMAND %02x %02x %02x %02x %02x %02x %02x %02x\n",
MLY_GET_REG(sc, sc->mly_command_mailbox),
MLY_GET_REG(sc, sc->mly_command_mailbox + 1),
MLY_GET_REG(sc, sc->mly_command_mailbox + 2),
MLY_GET_REG(sc, sc->mly_command_mailbox + 3),
MLY_GET_REG(sc, sc->mly_command_mailbox + 4),
MLY_GET_REG(sc, sc->mly_command_mailbox + 5),
MLY_GET_REG(sc, sc->mly_command_mailbox + 6),
MLY_GET_REG(sc, sc->mly_command_mailbox + 7));
mly_printf(sc, "STATUS %02x %02x %02x %02x %02x %02x %02x %02x\n",
MLY_GET_REG(sc, sc->mly_status_mailbox),
MLY_GET_REG(sc, sc->mly_status_mailbox + 1),
MLY_GET_REG(sc, sc->mly_status_mailbox + 2),
MLY_GET_REG(sc, sc->mly_status_mailbox + 3),
MLY_GET_REG(sc, sc->mly_status_mailbox + 4),
MLY_GET_REG(sc, sc->mly_status_mailbox + 5),
MLY_GET_REG(sc, sc->mly_status_mailbox + 6),
MLY_GET_REG(sc, sc->mly_status_mailbox + 7));
mly_printf(sc, " %04x %08x\n",
MLY_GET_REG2(sc, sc->mly_status_mailbox),
MLY_GET_REG4(sc, sc->mly_status_mailbox + 4));
}
struct mly_softc *mly_softc0 = NULL;
void
mly_printstate0(void)
{
if (mly_softc0 != NULL)
mly_printstate(mly_softc0);
}
/********************************************************************************
* Print a command
*/
static void
mly_print_command(struct mly_command *mc)
{
struct mly_softc *sc = mc->mc_sc;
mly_printf(sc, "COMMAND @ %p\n", mc);
mly_printf(sc, " slot %d\n", mc->mc_slot);
mly_printf(sc, " status 0x%x\n", mc->mc_status);
mly_printf(sc, " sense len %d\n", mc->mc_sense);
mly_printf(sc, " resid %d\n", mc->mc_resid);
mly_printf(sc, " packet %p/0x%llx\n", mc->mc_packet, mc->mc_packetphys);
if (mc->mc_packet != NULL)
mly_print_packet(mc);
mly_printf(sc, " data %p/%d\n", mc->mc_data, mc->mc_length);
mly_printf(sc, " flags %b\n", mc->mc_flags, "\20\1busy\2complete\3slotted\4mapped\5datain\6dataout\n");
mly_printf(sc, " complete %p\n", mc->mc_complete);
mly_printf(sc, " private %p\n", mc->mc_private);
}
/********************************************************************************
* Print a command packet
*/
static void
mly_print_packet(struct mly_command *mc)
{
struct mly_softc *sc = mc->mc_sc;
struct mly_command_generic *ge = (struct mly_command_generic *)mc->mc_packet;
struct mly_command_scsi_small *ss = (struct mly_command_scsi_small *)mc->mc_packet;
struct mly_command_scsi_large *sl = (struct mly_command_scsi_large *)mc->mc_packet;
struct mly_command_ioctl *io = (struct mly_command_ioctl *)mc->mc_packet;
int transfer;
mly_printf(sc, " command_id %d\n", ge->command_id);
mly_printf(sc, " opcode %d\n", ge->opcode);
mly_printf(sc, " command_control fua %d dpo %d est %d dd %s nas %d ddis %d\n",
ge->command_control.force_unit_access,
ge->command_control.disable_page_out,
ge->command_control.extended_sg_table,
(ge->command_control.data_direction == MLY_CCB_WRITE) ? "WRITE" : "READ",
ge->command_control.no_auto_sense,
ge->command_control.disable_disconnect);
mly_printf(sc, " data_size %d\n", ge->data_size);
mly_printf(sc, " sense_buffer_address 0x%llx\n", ge->sense_buffer_address);
mly_printf(sc, " lun %d\n", ge->addr.phys.lun);
mly_printf(sc, " target %d\n", ge->addr.phys.target);
mly_printf(sc, " channel %d\n", ge->addr.phys.channel);
mly_printf(sc, " logical device %d\n", ge->addr.log.logdev);
mly_printf(sc, " controller %d\n", ge->addr.phys.controller);
mly_printf(sc, " timeout %d %s\n",
ge->timeout.value,
(ge->timeout.scale == MLY_TIMEOUT_SECONDS) ? "seconds" :
((ge->timeout.scale == MLY_TIMEOUT_MINUTES) ? "minutes" : "hours"));
mly_printf(sc, " maximum_sense_size %d\n", ge->maximum_sense_size);
switch(ge->opcode) {
case MDACMD_SCSIPT:
case MDACMD_SCSI:
mly_printf(sc, " cdb length %d\n", ss->cdb_length);
mly_printf(sc, " cdb %*D\n", ss->cdb_length, ss->cdb, " ");
transfer = 1;
break;
case MDACMD_SCSILC:
case MDACMD_SCSILCPT:
mly_printf(sc, " cdb length %d\n", sl->cdb_length);
mly_printf(sc, " cdb 0x%llx\n", sl->cdb_physaddr);
transfer = 1;
break;
case MDACMD_IOCTL:
mly_printf(sc, " sub_ioctl 0x%x\n", io->sub_ioctl);
switch(io->sub_ioctl) {
case MDACIOCTL_SETMEMORYMAILBOX:
mly_printf(sc, " health_buffer_size %d\n",
io->param.setmemorymailbox.health_buffer_size);
mly_printf(sc, " health_buffer_phys 0x%llx\n",
io->param.setmemorymailbox.health_buffer_physaddr);
mly_printf(sc, " command_mailbox 0x%llx\n",
io->param.setmemorymailbox.command_mailbox_physaddr);
mly_printf(sc, " status_mailbox 0x%llx\n",
io->param.setmemorymailbox.status_mailbox_physaddr);
transfer = 0;
break;
case MDACIOCTL_SETREALTIMECLOCK:
case MDACIOCTL_GETHEALTHSTATUS:
case MDACIOCTL_GETCONTROLLERINFO:
case MDACIOCTL_GETLOGDEVINFOVALID:
case MDACIOCTL_GETPHYSDEVINFOVALID:
case MDACIOCTL_GETPHYSDEVSTATISTICS:
case MDACIOCTL_GETLOGDEVSTATISTICS:
case MDACIOCTL_GETCONTROLLERSTATISTICS:
case MDACIOCTL_GETBDT_FOR_SYSDRIVE:
case MDACIOCTL_CREATENEWCONF:
case MDACIOCTL_ADDNEWCONF:
case MDACIOCTL_GETDEVCONFINFO:
case MDACIOCTL_GETFREESPACELIST:
case MDACIOCTL_MORE:
case MDACIOCTL_SETPHYSDEVPARAMETER:
case MDACIOCTL_GETPHYSDEVPARAMETER:
case MDACIOCTL_GETLOGDEVPARAMETER:
case MDACIOCTL_SETLOGDEVPARAMETER:
mly_printf(sc, " param %10D\n", io->param.data.param, " ");
transfer = 1;
break;
case MDACIOCTL_GETEVENT:
mly_printf(sc, " event %d\n",
io->param.getevent.sequence_number_low + ((u_int32_t)io->addr.log.logdev << 16));
transfer = 1;
break;
case MDACIOCTL_SETRAIDDEVSTATE:
mly_printf(sc, " state %d\n", io->param.setraiddevstate.state);
transfer = 0;
break;
case MDACIOCTL_XLATEPHYSDEVTORAIDDEV:
mly_printf(sc, " raid_device %d\n", io->param.xlatephysdevtoraiddev.raid_device);
mly_printf(sc, " controller %d\n", io->param.xlatephysdevtoraiddev.controller);
mly_printf(sc, " channel %d\n", io->param.xlatephysdevtoraiddev.channel);
mly_printf(sc, " target %d\n", io->param.xlatephysdevtoraiddev.target);
mly_printf(sc, " lun %d\n", io->param.xlatephysdevtoraiddev.lun);
transfer = 0;
break;
case MDACIOCTL_GETGROUPCONFINFO:
mly_printf(sc, " group %d\n", io->param.getgroupconfinfo.group);
transfer = 1;
break;
case MDACIOCTL_GET_SUBSYSTEM_DATA:
case MDACIOCTL_SET_SUBSYSTEM_DATA:
case MDACIOCTL_STARTDISOCVERY:
case MDACIOCTL_INITPHYSDEVSTART:
case MDACIOCTL_INITPHYSDEVSTOP:
case MDACIOCTL_INITRAIDDEVSTART:
case MDACIOCTL_INITRAIDDEVSTOP:
case MDACIOCTL_REBUILDRAIDDEVSTART:
case MDACIOCTL_REBUILDRAIDDEVSTOP:
case MDACIOCTL_MAKECONSISTENTDATASTART:
case MDACIOCTL_MAKECONSISTENTDATASTOP:
case MDACIOCTL_CONSISTENCYCHECKSTART:
case MDACIOCTL_CONSISTENCYCHECKSTOP:
case MDACIOCTL_RESETDEVICE:
case MDACIOCTL_FLUSHDEVICEDATA:
case MDACIOCTL_PAUSEDEVICE:
case MDACIOCTL_UNPAUSEDEVICE:
case MDACIOCTL_LOCATEDEVICE:
case MDACIOCTL_SETMASTERSLAVEMODE:
case MDACIOCTL_DELETERAIDDEV:
case MDACIOCTL_REPLACEINTERNALDEV:
case MDACIOCTL_CLEARCONF:
case MDACIOCTL_GETCONTROLLERPARAMETER:
case MDACIOCTL_SETCONTRLLERPARAMETER:
case MDACIOCTL_CLEARCONFSUSPMODE:
case MDACIOCTL_STOREIMAGE:
case MDACIOCTL_READIMAGE:
case MDACIOCTL_FLASHIMAGES:
case MDACIOCTL_RENAMERAIDDEV:
default: /* no idea what to print */
transfer = 0;
break;
}
break;
case MDACMD_IOCTLCHECK:
case MDACMD_MEMCOPY:
default:
transfer = 0;
break; /* print nothing */
}
if (transfer) {
if (ge->command_control.extended_sg_table) {
mly_printf(sc, " sg table 0x%llx/%d\n",
ge->transfer.indirect.table_physaddr[0], ge->transfer.indirect.entries[0]);
} else {
mly_printf(sc, " 0000 0x%llx/%lld\n",
ge->transfer.direct.sg[0].physaddr, ge->transfer.direct.sg[0].length);
mly_printf(sc, " 0001 0x%llx/%lld\n",
ge->transfer.direct.sg[1].physaddr, ge->transfer.direct.sg[1].length);
}
}
}
/********************************************************************************
* Panic in a slightly informative fashion
*/
static void
mly_panic(struct mly_softc *sc, char *reason)
{
mly_printstate(sc);
panic(reason);
}
/********************************************************************************
* Print queue statistics, callable from DDB.
*/
void
mly_print_controller(int controller)
{
struct mly_softc *sc;
if ((sc = devclass_get_softc(devclass_find("mly"), controller)) == NULL) {
printf("mly: controller %d invalid\n", controller);
} else {
device_printf(sc->mly_dev, "queue curr max\n");
device_printf(sc->mly_dev, "free %04d/%04d\n",
sc->mly_qstat[MLYQ_FREE].q_length, sc->mly_qstat[MLYQ_FREE].q_max);
device_printf(sc->mly_dev, "busy %04d/%04d\n",
sc->mly_qstat[MLYQ_BUSY].q_length, sc->mly_qstat[MLYQ_BUSY].q_max);
device_printf(sc->mly_dev, "complete %04d/%04d\n",
sc->mly_qstat[MLYQ_COMPLETE].q_length, sc->mly_qstat[MLYQ_COMPLETE].q_max);
}
}
#endif
/********************************************************************************
********************************************************************************
Control device interface
********************************************************************************
********************************************************************************/
/********************************************************************************
* Accept an open operation on the control device.
*/
static int
mly_user_open(struct cdev *dev, int flags, int fmt, struct thread *td)
{
struct mly_softc *sc = dev->si_drv1;
MLY_LOCK(sc);
sc->mly_state |= MLY_STATE_OPEN;
MLY_UNLOCK(sc);
return(0);
}
/********************************************************************************
* Accept the last close on the control device.
*/
static int
mly_user_close(struct cdev *dev, int flags, int fmt, struct thread *td)
{
struct mly_softc *sc = dev->si_drv1;
MLY_LOCK(sc);
sc->mly_state &= ~MLY_STATE_OPEN;
MLY_UNLOCK(sc);
return (0);
}
/********************************************************************************
* Handle controller-specific control operations.
*/
static int
mly_user_ioctl(struct cdev *dev, u_long cmd, caddr_t addr,
int32_t flag, struct thread *td)
{
struct mly_softc *sc = (struct mly_softc *)dev->si_drv1;
struct mly_user_command *uc = (struct mly_user_command *)addr;
struct mly_user_health *uh = (struct mly_user_health *)addr;
switch(cmd) {
case MLYIO_COMMAND:
return(mly_user_command(sc, uc));
case MLYIO_HEALTH:
return(mly_user_health(sc, uh));
default:
return(ENOIOCTL);
}
}
/********************************************************************************
* Execute a command passed in from userspace.
*
* The control structure contains the actual command for the controller, as well
* as the user-space data pointer and data size, and an optional sense buffer
* size/pointer. On completion, the data size is adjusted to the command
* residual, and the sense buffer size to the size of the returned sense data.
*
*/
static int
mly_user_command(struct mly_softc *sc, struct mly_user_command *uc)
{
struct mly_command *mc;
int error;
/* allocate a command */
MLY_LOCK(sc);
if (mly_alloc_command(sc, &mc)) {
MLY_UNLOCK(sc);
error = ENOMEM;
goto out; /* XXX Linux version will wait for a command */
}
MLY_UNLOCK(sc);
/* handle data size/direction */
mc->mc_length = (uc->DataTransferLength >= 0) ? uc->DataTransferLength : -uc->DataTransferLength;
if (mc->mc_length > 0) {
if ((mc->mc_data = malloc(mc->mc_length, M_DEVBUF, M_NOWAIT)) == NULL) {
error = ENOMEM;
goto out;
}
}
if (uc->DataTransferLength > 0) {
mc->mc_flags |= MLY_CMD_DATAIN;
bzero(mc->mc_data, mc->mc_length);
}
if (uc->DataTransferLength < 0) {
mc->mc_flags |= MLY_CMD_DATAOUT;
if ((error = copyin(uc->DataTransferBuffer, mc->mc_data, mc->mc_length)) != 0)
goto out;
}
/* copy the controller command */
bcopy(&uc->CommandMailbox, mc->mc_packet, sizeof(uc->CommandMailbox));
/* clear command completion handler so that we get woken up */
mc->mc_complete = NULL;
/* execute the command */
MLY_LOCK(sc);
if ((error = mly_start(mc)) != 0) {
MLY_UNLOCK(sc);
goto out;
}
while (!(mc->mc_flags & MLY_CMD_COMPLETE))
mtx_sleep(mc, &sc->mly_lock, PRIBIO, "mlyioctl", 0);
MLY_UNLOCK(sc);
/* return the data to userspace */
if (uc->DataTransferLength > 0)
if ((error = copyout(mc->mc_data, uc->DataTransferBuffer, mc->mc_length)) != 0)
goto out;
/* return the sense buffer to userspace */
if ((uc->RequestSenseLength > 0) && (mc->mc_sense > 0)) {
if ((error = copyout(mc->mc_packet, uc->RequestSenseBuffer,
min(uc->RequestSenseLength, mc->mc_sense))) != 0)
goto out;
}
/* return command results to userspace (caller will copy out) */
uc->DataTransferLength = mc->mc_resid;
uc->RequestSenseLength = min(uc->RequestSenseLength, mc->mc_sense);
uc->CommandStatus = mc->mc_status;
error = 0;
out:
if (mc->mc_data != NULL)
free(mc->mc_data, M_DEVBUF);
if (mc != NULL) {
MLY_LOCK(sc);
mly_release_command(mc);
MLY_UNLOCK(sc);
}
return(error);
}
/********************************************************************************
* Return health status to userspace. If the health change index in the user
* structure does not match that currently exported by the controller, we
* return the current status immediately. Otherwise, we block until either
* interrupted or new status is delivered.
*/
static int
mly_user_health(struct mly_softc *sc, struct mly_user_health *uh)
{
struct mly_health_status mh;
int error;
/* fetch the current health status from userspace */
if ((error = copyin(uh->HealthStatusBuffer, &mh, sizeof(mh))) != 0)
return(error);
/* spin waiting for a status update */
MLY_LOCK(sc);
error = EWOULDBLOCK;
while ((error != 0) && (sc->mly_event_change == mh.change_counter))
error = mtx_sleep(&sc->mly_event_change, &sc->mly_lock, PRIBIO | PCATCH,
"mlyhealth", 0);
mh = sc->mly_mmbox->mmm_health.status;
MLY_UNLOCK(sc);
/* copy the controller's health status buffer out */
error = copyout(&mh, uh->HealthStatusBuffer, sizeof(mh));
return(error);
}
#ifdef MLY_DEBUG
static void
mly_timeout(void *arg)
{
struct mly_softc *sc;
struct mly_command *mc;
int deadline;
sc = arg;
MLY_ASSERT_LOCKED(sc);
deadline = time_second - MLY_CMD_TIMEOUT;
TAILQ_FOREACH(mc, &sc->mly_busy, mc_link) {
if ((mc->mc_timestamp < deadline)) {
device_printf(sc->mly_dev,
"COMMAND %p TIMEOUT AFTER %d SECONDS\n", mc,
(int)(time_second - mc->mc_timestamp));
}
}
callout_reset(&sc->mly_timeout, MLY_CMD_TIMEOUT * hz, mly_timeout, sc);
}
#endif