1999-08-28 01:08:13 +00:00
|
|
|
/* $FreeBSD$ */
|
1998-04-22 18:12:29 +00:00
|
|
|
/*
|
|
|
|
* PCI specific probe and attach routines for Qlogic ISP SCSI adapters.
|
|
|
|
* FreeBSD Version.
|
|
|
|
*
|
2001-03-02 06:28:55 +00:00
|
|
|
* Copyright (c) 1997, 1998, 1999, 2000, 2001 by Matthew Jacob
|
1998-04-22 18:12:29 +00:00
|
|
|
*
|
|
|
|
* 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 immediately at the beginning of the file, without modification,
|
|
|
|
* this list of conditions, and the following disclaimer.
|
2000-09-21 20:16:04 +00:00
|
|
|
* 2. The name of the author may not be used to endorse or promote products
|
|
|
|
* derived from this software without specific prior written permission.
|
1998-04-22 18:12:29 +00:00
|
|
|
*
|
|
|
|
* 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.
|
|
|
|
*/
|
1998-09-15 10:06:23 +00:00
|
|
|
|
2000-02-11 19:45:17 +00:00
|
|
|
#include <sys/param.h>
|
|
|
|
#include <sys/systm.h>
|
|
|
|
#include <sys/kernel.h>
|
|
|
|
#include <sys/module.h>
|
|
|
|
#include <sys/bus.h>
|
1998-04-22 18:12:29 +00:00
|
|
|
|
|
|
|
#include <pci/pcireg.h>
|
|
|
|
#include <pci/pcivar.h>
|
|
|
|
|
1998-09-15 10:06:23 +00:00
|
|
|
#include <machine/bus_memio.h>
|
|
|
|
#include <machine/bus_pio.h>
|
|
|
|
#include <machine/bus.h>
|
2000-02-11 19:45:17 +00:00
|
|
|
#include <machine/resource.h>
|
|
|
|
#include <sys/rman.h>
|
|
|
|
#include <sys/malloc.h>
|
|
|
|
|
|
|
|
#include <dev/isp/isp_freebsd.h>
|
1999-03-17 05:07:18 +00:00
|
|
|
|
2001-08-31 21:39:04 +00:00
|
|
|
static u_int16_t isp_pci_rd_reg(struct ispsoftc *, int);
|
|
|
|
static void isp_pci_wr_reg(struct ispsoftc *, int, u_int16_t);
|
|
|
|
static u_int16_t isp_pci_rd_reg_1080(struct ispsoftc *, int);
|
|
|
|
static void isp_pci_wr_reg_1080(struct ispsoftc *, int, u_int16_t);
|
|
|
|
static int
|
|
|
|
isp_pci_rd_isr(struct ispsoftc *, u_int16_t *, u_int16_t *, u_int16_t *);
|
|
|
|
static int
|
|
|
|
isp_pci_rd_isr_2300(struct ispsoftc *, u_int16_t *, u_int16_t *, u_int16_t *);
|
|
|
|
static int isp_pci_mbxdma(struct ispsoftc *);
|
|
|
|
static int
|
|
|
|
isp_pci_dmasetup(struct ispsoftc *, XS_T *, ispreq_t *, u_int16_t *, u_int16_t);
|
1998-09-15 10:06:23 +00:00
|
|
|
static void
|
2001-08-31 21:39:04 +00:00
|
|
|
isp_pci_dmateardown(struct ispsoftc *, XS_T *, u_int16_t);
|
1998-04-22 18:12:29 +00:00
|
|
|
|
2001-08-31 21:39:04 +00:00
|
|
|
static void isp_pci_reset1(struct ispsoftc *);
|
|
|
|
static void isp_pci_dumpregs(struct ispsoftc *, const char *);
|
1998-04-22 18:12:29 +00:00
|
|
|
|
|
|
|
static struct ispmdvec mdvec = {
|
2001-08-31 21:39:04 +00:00
|
|
|
isp_pci_rd_isr,
|
1998-04-22 18:12:29 +00:00
|
|
|
isp_pci_rd_reg,
|
|
|
|
isp_pci_wr_reg,
|
|
|
|
isp_pci_mbxdma,
|
|
|
|
isp_pci_dmasetup,
|
1998-09-15 10:06:23 +00:00
|
|
|
isp_pci_dmateardown,
|
1998-04-22 18:12:29 +00:00
|
|
|
NULL,
|
|
|
|
isp_pci_reset1,
|
|
|
|
isp_pci_dumpregs,
|
2000-06-18 05:18:55 +00:00
|
|
|
NULL,
|
2000-08-01 05:16:49 +00:00
|
|
|
BIU_BURST_ENABLE|BIU_PCI_CONF1_FIFO_64
|
1998-04-22 18:12:29 +00:00
|
|
|
};
|
|
|
|
|
1999-03-17 05:07:18 +00:00
|
|
|
static struct ispmdvec mdvec_1080 = {
|
2001-08-31 21:39:04 +00:00
|
|
|
isp_pci_rd_isr,
|
1999-03-17 05:07:18 +00:00
|
|
|
isp_pci_rd_reg_1080,
|
|
|
|
isp_pci_wr_reg_1080,
|
|
|
|
isp_pci_mbxdma,
|
|
|
|
isp_pci_dmasetup,
|
|
|
|
isp_pci_dmateardown,
|
|
|
|
NULL,
|
|
|
|
isp_pci_reset1,
|
|
|
|
isp_pci_dumpregs,
|
2000-06-18 05:18:55 +00:00
|
|
|
NULL,
|
2000-08-01 05:16:49 +00:00
|
|
|
BIU_BURST_ENABLE|BIU_PCI_CONF1_FIFO_64
|
1999-03-17 05:07:18 +00:00
|
|
|
};
|
|
|
|
|
2000-02-11 19:45:17 +00:00
|
|
|
static struct ispmdvec mdvec_12160 = {
|
2001-08-31 21:39:04 +00:00
|
|
|
isp_pci_rd_isr,
|
2000-02-11 19:45:17 +00:00
|
|
|
isp_pci_rd_reg_1080,
|
|
|
|
isp_pci_wr_reg_1080,
|
|
|
|
isp_pci_mbxdma,
|
|
|
|
isp_pci_dmasetup,
|
|
|
|
isp_pci_dmateardown,
|
|
|
|
NULL,
|
|
|
|
isp_pci_reset1,
|
|
|
|
isp_pci_dumpregs,
|
2000-06-18 05:18:55 +00:00
|
|
|
NULL,
|
2000-08-01 05:16:49 +00:00
|
|
|
BIU_BURST_ENABLE|BIU_PCI_CONF1_FIFO_64
|
2000-02-11 19:45:17 +00:00
|
|
|
};
|
|
|
|
|
1998-04-22 18:12:29 +00:00
|
|
|
static struct ispmdvec mdvec_2100 = {
|
2001-08-31 21:39:04 +00:00
|
|
|
isp_pci_rd_isr,
|
1998-04-22 18:12:29 +00:00
|
|
|
isp_pci_rd_reg,
|
|
|
|
isp_pci_wr_reg,
|
|
|
|
isp_pci_mbxdma,
|
|
|
|
isp_pci_dmasetup,
|
1998-09-15 10:06:23 +00:00
|
|
|
isp_pci_dmateardown,
|
1998-04-22 18:12:29 +00:00
|
|
|
NULL,
|
|
|
|
isp_pci_reset1,
|
2000-08-01 05:16:49 +00:00
|
|
|
isp_pci_dumpregs
|
1998-04-22 18:12:29 +00:00
|
|
|
};
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
|
|
|
|
static struct ispmdvec mdvec_2200 = {
|
2001-08-31 21:39:04 +00:00
|
|
|
isp_pci_rd_isr,
|
|
|
|
isp_pci_rd_reg,
|
|
|
|
isp_pci_wr_reg,
|
|
|
|
isp_pci_mbxdma,
|
|
|
|
isp_pci_dmasetup,
|
|
|
|
isp_pci_dmateardown,
|
|
|
|
NULL,
|
|
|
|
isp_pci_reset1,
|
|
|
|
isp_pci_dumpregs
|
|
|
|
};
|
|
|
|
|
|
|
|
static struct ispmdvec mdvec_2300 = {
|
|
|
|
isp_pci_rd_isr_2300,
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
isp_pci_rd_reg,
|
|
|
|
isp_pci_wr_reg,
|
|
|
|
isp_pci_mbxdma,
|
|
|
|
isp_pci_dmasetup,
|
|
|
|
isp_pci_dmateardown,
|
|
|
|
NULL,
|
|
|
|
isp_pci_reset1,
|
2000-08-01 05:16:49 +00:00
|
|
|
isp_pci_dumpregs
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
};
|
1999-02-09 01:12:52 +00:00
|
|
|
|
1998-04-22 18:12:29 +00:00
|
|
|
#ifndef PCIM_CMD_INVEN
|
1998-09-15 10:06:23 +00:00
|
|
|
#define PCIM_CMD_INVEN 0x10
|
1998-04-22 18:12:29 +00:00
|
|
|
#endif
|
|
|
|
#ifndef PCIM_CMD_BUSMASTEREN
|
1998-09-15 10:06:23 +00:00
|
|
|
#define PCIM_CMD_BUSMASTEREN 0x0004
|
1998-04-22 18:12:29 +00:00
|
|
|
#endif
|
1999-02-09 01:12:52 +00:00
|
|
|
#ifndef PCIM_CMD_PERRESPEN
|
|
|
|
#define PCIM_CMD_PERRESPEN 0x0040
|
|
|
|
#endif
|
|
|
|
#ifndef PCIM_CMD_SEREN
|
|
|
|
#define PCIM_CMD_SEREN 0x0100
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef PCIR_COMMAND
|
|
|
|
#define PCIR_COMMAND 0x04
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef PCIR_CACHELNSZ
|
|
|
|
#define PCIR_CACHELNSZ 0x0c
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef PCIR_LATTIMER
|
|
|
|
#define PCIR_LATTIMER 0x0d
|
|
|
|
#endif
|
|
|
|
|
1999-04-04 01:14:02 +00:00
|
|
|
#ifndef PCIR_ROMADDR
|
|
|
|
#define PCIR_ROMADDR 0x30
|
|
|
|
#endif
|
|
|
|
|
1998-04-22 18:12:29 +00:00
|
|
|
#ifndef PCI_VENDOR_QLOGIC
|
2000-06-18 05:18:55 +00:00
|
|
|
#define PCI_VENDOR_QLOGIC 0x1077
|
1998-04-22 18:12:29 +00:00
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef PCI_PRODUCT_QLOGIC_ISP1020
|
|
|
|
#define PCI_PRODUCT_QLOGIC_ISP1020 0x1020
|
|
|
|
#endif
|
|
|
|
|
1999-03-17 05:07:18 +00:00
|
|
|
#ifndef PCI_PRODUCT_QLOGIC_ISP1080
|
|
|
|
#define PCI_PRODUCT_QLOGIC_ISP1080 0x1080
|
|
|
|
#endif
|
|
|
|
|
2000-02-11 19:45:17 +00:00
|
|
|
#ifndef PCI_PRODUCT_QLOGIC_ISP12160
|
|
|
|
#define PCI_PRODUCT_QLOGIC_ISP12160 0x1216
|
|
|
|
#endif
|
|
|
|
|
1999-03-17 05:07:18 +00:00
|
|
|
#ifndef PCI_PRODUCT_QLOGIC_ISP1240
|
|
|
|
#define PCI_PRODUCT_QLOGIC_ISP1240 0x1240
|
|
|
|
#endif
|
1998-04-22 18:12:29 +00:00
|
|
|
|
1999-12-16 05:42:02 +00:00
|
|
|
#ifndef PCI_PRODUCT_QLOGIC_ISP1280
|
|
|
|
#define PCI_PRODUCT_QLOGIC_ISP1280 0x1280
|
|
|
|
#endif
|
|
|
|
|
1998-04-22 18:12:29 +00:00
|
|
|
#ifndef PCI_PRODUCT_QLOGIC_ISP2100
|
|
|
|
#define PCI_PRODUCT_QLOGIC_ISP2100 0x2100
|
|
|
|
#endif
|
|
|
|
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
#ifndef PCI_PRODUCT_QLOGIC_ISP2200
|
|
|
|
#define PCI_PRODUCT_QLOGIC_ISP2200 0x2200
|
|
|
|
#endif
|
|
|
|
|
2001-08-31 21:39:04 +00:00
|
|
|
#ifndef PCI_PRODUCT_QLOGIC_ISP2300
|
|
|
|
#define PCI_PRODUCT_QLOGIC_ISP2300 0x2300
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef PCI_PRODUCT_QLOGIC_ISP2312
|
|
|
|
#define PCI_PRODUCT_QLOGIC_ISP2312 0x2312
|
|
|
|
#endif
|
|
|
|
|
2000-06-18 05:18:55 +00:00
|
|
|
#define PCI_QLOGIC_ISP1020 \
|
|
|
|
((PCI_PRODUCT_QLOGIC_ISP1020 << 16) | PCI_VENDOR_QLOGIC)
|
1999-03-17 05:07:18 +00:00
|
|
|
|
|
|
|
#define PCI_QLOGIC_ISP1080 \
|
|
|
|
((PCI_PRODUCT_QLOGIC_ISP1080 << 16) | PCI_VENDOR_QLOGIC)
|
|
|
|
|
2000-02-11 19:45:17 +00:00
|
|
|
#define PCI_QLOGIC_ISP12160 \
|
|
|
|
((PCI_PRODUCT_QLOGIC_ISP12160 << 16) | PCI_VENDOR_QLOGIC)
|
|
|
|
|
1999-03-17 05:07:18 +00:00
|
|
|
#define PCI_QLOGIC_ISP1240 \
|
|
|
|
((PCI_PRODUCT_QLOGIC_ISP1240 << 16) | PCI_VENDOR_QLOGIC)
|
|
|
|
|
1999-12-16 05:42:02 +00:00
|
|
|
#define PCI_QLOGIC_ISP1280 \
|
|
|
|
((PCI_PRODUCT_QLOGIC_ISP1280 << 16) | PCI_VENDOR_QLOGIC)
|
|
|
|
|
1998-04-22 18:12:29 +00:00
|
|
|
#define PCI_QLOGIC_ISP2100 \
|
|
|
|
((PCI_PRODUCT_QLOGIC_ISP2100 << 16) | PCI_VENDOR_QLOGIC)
|
|
|
|
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
#define PCI_QLOGIC_ISP2200 \
|
|
|
|
((PCI_PRODUCT_QLOGIC_ISP2200 << 16) | PCI_VENDOR_QLOGIC)
|
|
|
|
|
2001-08-31 21:39:04 +00:00
|
|
|
#define PCI_QLOGIC_ISP2300 \
|
|
|
|
((PCI_PRODUCT_QLOGIC_ISP2300 << 16) | PCI_VENDOR_QLOGIC)
|
|
|
|
|
|
|
|
#define PCI_QLOGIC_ISP2312 \
|
|
|
|
((PCI_PRODUCT_QLOGIC_ISP2312 << 16) | PCI_VENDOR_QLOGIC)
|
|
|
|
|
2000-09-07 20:27:40 +00:00
|
|
|
/*
|
|
|
|
* Odd case for some AMI raid cards... We need to *not* attach to this.
|
|
|
|
*/
|
|
|
|
#define AMI_RAID_SUBVENDOR_ID 0x101e
|
|
|
|
|
1999-03-17 05:07:18 +00:00
|
|
|
#define IO_MAP_REG 0x10
|
|
|
|
#define MEM_MAP_REG 0x14
|
1998-04-22 18:12:29 +00:00
|
|
|
|
1999-02-09 01:12:52 +00:00
|
|
|
#define PCI_DFLT_LTNCY 0x40
|
|
|
|
#define PCI_DFLT_LNSZ 0x10
|
1998-04-22 18:12:29 +00:00
|
|
|
|
2000-02-11 19:45:17 +00:00
|
|
|
static int isp_pci_probe (device_t);
|
|
|
|
static int isp_pci_attach (device_t);
|
1998-04-22 18:12:29 +00:00
|
|
|
|
2002-04-02 23:36:14 +00:00
|
|
|
|
1998-04-22 18:12:29 +00:00
|
|
|
struct isp_pcisoftc {
|
1998-09-15 10:06:23 +00:00
|
|
|
struct ispsoftc pci_isp;
|
2000-02-11 19:45:17 +00:00
|
|
|
device_t pci_dev;
|
|
|
|
struct resource * pci_reg;
|
1998-09-15 10:06:23 +00:00
|
|
|
bus_space_tag_t pci_st;
|
|
|
|
bus_space_handle_t pci_sh;
|
2000-02-11 19:45:17 +00:00
|
|
|
void * ih;
|
1999-03-17 05:07:18 +00:00
|
|
|
int16_t pci_poff[_NREG_BLKS];
|
2002-04-02 23:36:14 +00:00
|
|
|
bus_dma_tag_t dmat;
|
1999-10-17 19:03:11 +00:00
|
|
|
bus_dmamap_t *dmaps;
|
1998-04-22 18:12:29 +00:00
|
|
|
};
|
2002-07-11 03:25:04 +00:00
|
|
|
extern ispfwfunc *isp_get_firmware_p;
|
1998-04-22 18:12:29 +00:00
|
|
|
|
2000-02-11 19:45:17 +00:00
|
|
|
static device_method_t isp_pci_methods[] = {
|
|
|
|
/* Device interface */
|
|
|
|
DEVMETHOD(device_probe, isp_pci_probe),
|
|
|
|
DEVMETHOD(device_attach, isp_pci_attach),
|
|
|
|
{ 0, 0 }
|
1998-04-22 18:12:29 +00:00
|
|
|
};
|
2001-08-31 21:39:04 +00:00
|
|
|
static void isp_pci_intr(void *);
|
1998-04-22 18:12:29 +00:00
|
|
|
|
2000-02-11 19:45:17 +00:00
|
|
|
static driver_t isp_pci_driver = {
|
|
|
|
"isp", isp_pci_methods, sizeof (struct isp_pcisoftc)
|
|
|
|
};
|
|
|
|
static devclass_t isp_devclass;
|
|
|
|
DRIVER_MODULE(isp, pci, isp_pci_driver, isp_devclass, 0, 0);
|
1998-04-22 18:12:29 +00:00
|
|
|
|
2000-02-11 19:45:17 +00:00
|
|
|
static int
|
|
|
|
isp_pci_probe(device_t dev)
|
1999-03-17 05:07:18 +00:00
|
|
|
{
|
2000-02-11 19:45:17 +00:00
|
|
|
switch ((pci_get_device(dev) << 16) | (pci_get_vendor(dev))) {
|
2000-06-18 05:18:55 +00:00
|
|
|
case PCI_QLOGIC_ISP1020:
|
2000-02-11 19:45:17 +00:00
|
|
|
device_set_desc(dev, "Qlogic ISP 1020/1040 PCI SCSI Adapter");
|
1999-03-17 05:07:18 +00:00
|
|
|
break;
|
|
|
|
case PCI_QLOGIC_ISP1080:
|
2000-02-11 19:45:17 +00:00
|
|
|
device_set_desc(dev, "Qlogic ISP 1080 PCI SCSI Adapter");
|
1999-05-11 04:53:57 +00:00
|
|
|
break;
|
|
|
|
case PCI_QLOGIC_ISP1240:
|
2000-02-11 19:45:17 +00:00
|
|
|
device_set_desc(dev, "Qlogic ISP 1240 PCI SCSI Adapter");
|
1998-04-22 18:12:29 +00:00
|
|
|
break;
|
1999-12-16 05:42:02 +00:00
|
|
|
case PCI_QLOGIC_ISP1280:
|
2000-02-11 19:45:17 +00:00
|
|
|
device_set_desc(dev, "Qlogic ISP 1280 PCI SCSI Adapter");
|
|
|
|
break;
|
|
|
|
case PCI_QLOGIC_ISP12160:
|
2000-09-07 20:27:40 +00:00
|
|
|
if (pci_get_subvendor(dev) == AMI_RAID_SUBVENDOR_ID) {
|
|
|
|
return (ENXIO);
|
|
|
|
}
|
2000-02-11 19:45:17 +00:00
|
|
|
device_set_desc(dev, "Qlogic ISP 12160 PCI SCSI Adapter");
|
1999-12-16 05:42:02 +00:00
|
|
|
break;
|
1998-04-22 18:12:29 +00:00
|
|
|
case PCI_QLOGIC_ISP2100:
|
2000-02-11 19:45:17 +00:00
|
|
|
device_set_desc(dev, "Qlogic ISP 2100 PCI FC-AL Adapter");
|
1998-04-22 18:12:29 +00:00
|
|
|
break;
|
1999-07-05 20:24:46 +00:00
|
|
|
case PCI_QLOGIC_ISP2200:
|
2000-02-11 19:45:17 +00:00
|
|
|
device_set_desc(dev, "Qlogic ISP 2200 PCI FC-AL Adapter");
|
1999-07-05 20:24:46 +00:00
|
|
|
break;
|
2001-08-31 21:39:04 +00:00
|
|
|
case PCI_QLOGIC_ISP2300:
|
|
|
|
device_set_desc(dev, "Qlogic ISP 2300 PCI FC-AL Adapter");
|
|
|
|
break;
|
|
|
|
case PCI_QLOGIC_ISP2312:
|
|
|
|
device_set_desc(dev, "Qlogic ISP 2312 PCI FC-AL Adapter");
|
|
|
|
break;
|
1998-04-22 18:12:29 +00:00
|
|
|
default:
|
2000-02-11 19:45:17 +00:00
|
|
|
return (ENXIO);
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
2002-07-11 03:25:04 +00:00
|
|
|
if (isp_announced == 0 && bootverbose) {
|
2000-08-01 05:16:49 +00:00
|
|
|
printf("Qlogic ISP Driver, FreeBSD Version %d.%d, "
|
1999-10-17 19:03:11 +00:00
|
|
|
"Core Version %d.%d\n",
|
1998-09-15 10:06:23 +00:00
|
|
|
ISP_PLATFORM_VERSION_MAJOR, ISP_PLATFORM_VERSION_MINOR,
|
|
|
|
ISP_CORE_VERSION_MAJOR, ISP_CORE_VERSION_MINOR);
|
2002-07-11 03:25:04 +00:00
|
|
|
isp_announced++;
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
2000-06-18 05:18:55 +00:00
|
|
|
/*
|
|
|
|
* XXXX: Here is where we might load the f/w module
|
|
|
|
* XXXX: (or increase a reference count to it).
|
|
|
|
*/
|
2000-02-11 19:45:17 +00:00
|
|
|
return (0);
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
|
|
|
|
2000-02-11 19:45:17 +00:00
|
|
|
static int
|
|
|
|
isp_pci_attach(device_t dev)
|
1998-04-22 18:12:29 +00:00
|
|
|
{
|
2000-02-11 19:45:17 +00:00
|
|
|
struct resource *regs, *irq;
|
2001-03-01 02:21:36 +00:00
|
|
|
int tval, rtp, rgd, iqd, m1, m2, isp_debug, role;
|
2000-02-11 19:45:17 +00:00
|
|
|
u_int32_t data, cmd, linesz, psize, basetype;
|
1998-04-22 18:12:29 +00:00
|
|
|
struct isp_pcisoftc *pcs;
|
2000-10-25 04:40:49 +00:00
|
|
|
struct ispsoftc *isp = NULL;
|
1999-05-11 04:53:57 +00:00
|
|
|
struct ispmdvec *mdvp;
|
2001-06-15 00:13:18 +00:00
|
|
|
const char *sptr;
|
2000-10-25 04:40:49 +00:00
|
|
|
int locksetup = 0;
|
1998-04-22 18:12:29 +00:00
|
|
|
|
1999-07-06 01:24:20 +00:00
|
|
|
/*
|
|
|
|
* Figure out if we're supposed to skip this one.
|
2001-02-11 03:53:23 +00:00
|
|
|
* If we are, we actually go to ISP_ROLE_NONE.
|
1999-07-06 01:24:20 +00:00
|
|
|
*/
|
2001-03-01 02:21:36 +00:00
|
|
|
|
|
|
|
tval = 0;
|
|
|
|
if (resource_int_value(device_get_name(dev), device_get_unit(dev),
|
|
|
|
"disable", &tval) == 0 && tval) {
|
|
|
|
device_printf(dev, "device is disabled\n");
|
|
|
|
/* but return 0 so the !$)$)*!$*) unit isn't reused */
|
|
|
|
return (0);
|
2001-02-11 03:53:23 +00:00
|
|
|
}
|
2001-03-01 02:21:36 +00:00
|
|
|
|
|
|
|
role = 0;
|
|
|
|
if (resource_int_value(device_get_name(dev), device_get_unit(dev),
|
|
|
|
"role", &role) == 0 &&
|
|
|
|
((role & ~(ISP_ROLE_INITIATOR|ISP_ROLE_TARGET)) == 0)) {
|
|
|
|
device_printf(dev, "setting role to 0x%x\n", role);
|
|
|
|
} else {
|
2001-02-11 03:53:23 +00:00
|
|
|
#ifdef ISP_TARGET_MODE
|
2001-03-01 02:21:36 +00:00
|
|
|
role = ISP_ROLE_INITIATOR|ISP_ROLE_TARGET;
|
2001-02-11 03:53:23 +00:00
|
|
|
#else
|
2001-03-01 02:21:36 +00:00
|
|
|
role = ISP_DEFAULT_ROLES;
|
2001-02-11 03:53:23 +00:00
|
|
|
#endif
|
1999-07-06 01:24:20 +00:00
|
|
|
}
|
|
|
|
|
2000-12-08 21:51:06 +00:00
|
|
|
pcs = malloc(sizeof (struct isp_pcisoftc), M_DEVBUF, M_NOWAIT | M_ZERO);
|
2000-02-11 19:45:17 +00:00
|
|
|
if (pcs == NULL) {
|
|
|
|
device_printf(dev, "cannot allocate softc\n");
|
|
|
|
return (ENOMEM);
|
|
|
|
}
|
|
|
|
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
/*
|
|
|
|
* Figure out which we should try first - memory mapping or i/o mapping?
|
|
|
|
*/
|
2000-06-18 05:18:55 +00:00
|
|
|
#ifdef __alpha__
|
2000-02-11 19:45:17 +00:00
|
|
|
m1 = PCIM_CMD_MEMEN;
|
|
|
|
m2 = PCIM_CMD_PORTEN;
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
#else
|
2000-02-11 19:45:17 +00:00
|
|
|
m1 = PCIM_CMD_PORTEN;
|
|
|
|
m2 = PCIM_CMD_MEMEN;
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
#endif
|
2001-03-01 02:21:36 +00:00
|
|
|
|
|
|
|
tval = 0;
|
|
|
|
if (resource_int_value(device_get_name(dev), device_get_unit(dev),
|
|
|
|
"prefer_iomap", &tval) == 0 && tval != 0) {
|
|
|
|
m1 = PCIM_CMD_PORTEN;
|
|
|
|
m2 = PCIM_CMD_MEMEN;
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
}
|
2001-03-01 02:21:36 +00:00
|
|
|
tval = 0;
|
|
|
|
if (resource_int_value(device_get_name(dev), device_get_unit(dev),
|
|
|
|
"prefer_memmap", &tval) == 0 && tval != 0) {
|
|
|
|
m1 = PCIM_CMD_MEMEN;
|
|
|
|
m2 = PCIM_CMD_PORTEN;
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
}
|
|
|
|
|
1999-04-04 01:14:02 +00:00
|
|
|
linesz = PCI_DFLT_LNSZ;
|
2000-02-11 19:45:17 +00:00
|
|
|
irq = regs = NULL;
|
|
|
|
rgd = rtp = iqd = 0;
|
|
|
|
|
|
|
|
cmd = pci_read_config(dev, PCIR_COMMAND, 1);
|
|
|
|
if (cmd & m1) {
|
|
|
|
rtp = (m1 == PCIM_CMD_MEMEN)? SYS_RES_MEMORY : SYS_RES_IOPORT;
|
|
|
|
rgd = (m1 == PCIM_CMD_MEMEN)? MEM_MAP_REG : IO_MAP_REG;
|
|
|
|
regs = bus_alloc_resource(dev, rtp, &rgd, 0, ~0, 1, RF_ACTIVE);
|
1999-02-09 01:12:52 +00:00
|
|
|
}
|
2000-02-11 19:45:17 +00:00
|
|
|
if (regs == NULL && (cmd & m2)) {
|
|
|
|
rtp = (m2 == PCIM_CMD_MEMEN)? SYS_RES_MEMORY : SYS_RES_IOPORT;
|
|
|
|
rgd = (m2 == PCIM_CMD_MEMEN)? MEM_MAP_REG : IO_MAP_REG;
|
|
|
|
regs = bus_alloc_resource(dev, rtp, &rgd, 0, ~0, 1, RF_ACTIVE);
|
|
|
|
}
|
|
|
|
if (regs == NULL) {
|
|
|
|
device_printf(dev, "unable to map any ports\n");
|
|
|
|
goto bad;
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
if (bootverbose)
|
2000-12-05 07:38:41 +00:00
|
|
|
device_printf(dev, "using %s space register mapping\n",
|
2000-02-11 19:45:17 +00:00
|
|
|
(rgd == IO_MAP_REG)? "I/O" : "Memory");
|
|
|
|
pcs->pci_dev = dev;
|
|
|
|
pcs->pci_reg = regs;
|
|
|
|
pcs->pci_st = rman_get_bustag(regs);
|
|
|
|
pcs->pci_sh = rman_get_bushandle(regs);
|
1998-04-22 18:12:29 +00:00
|
|
|
|
1999-03-17 05:07:18 +00:00
|
|
|
pcs->pci_poff[BIU_BLOCK >> _BLK_REG_SHFT] = BIU_REGS_OFF;
|
|
|
|
pcs->pci_poff[MBOX_BLOCK >> _BLK_REG_SHFT] = PCI_MBOX_REGS_OFF;
|
|
|
|
pcs->pci_poff[SXP_BLOCK >> _BLK_REG_SHFT] = PCI_SXP_REGS_OFF;
|
|
|
|
pcs->pci_poff[RISC_BLOCK >> _BLK_REG_SHFT] = PCI_RISC_REGS_OFF;
|
|
|
|
pcs->pci_poff[DMA_BLOCK >> _BLK_REG_SHFT] = DMA_REGS_OFF;
|
1999-05-11 04:53:57 +00:00
|
|
|
mdvp = &mdvec;
|
|
|
|
basetype = ISP_HA_SCSI_UNKNOWN;
|
|
|
|
psize = sizeof (sdparam);
|
2000-06-18 05:18:55 +00:00
|
|
|
if (pci_get_devid(dev) == PCI_QLOGIC_ISP1020) {
|
1999-05-11 04:53:57 +00:00
|
|
|
mdvp = &mdvec;
|
|
|
|
basetype = ISP_HA_SCSI_UNKNOWN;
|
|
|
|
psize = sizeof (sdparam);
|
1999-03-17 05:07:18 +00:00
|
|
|
}
|
2000-02-11 19:45:17 +00:00
|
|
|
if (pci_get_devid(dev) == PCI_QLOGIC_ISP1080) {
|
1999-05-11 04:53:57 +00:00
|
|
|
mdvp = &mdvec_1080;
|
|
|
|
basetype = ISP_HA_SCSI_1080;
|
|
|
|
psize = sizeof (sdparam);
|
|
|
|
pcs->pci_poff[DMA_BLOCK >> _BLK_REG_SHFT] =
|
|
|
|
ISP1080_DMA_REGS_OFF;
|
|
|
|
}
|
2000-02-11 19:45:17 +00:00
|
|
|
if (pci_get_devid(dev) == PCI_QLOGIC_ISP1240) {
|
1999-05-11 04:53:57 +00:00
|
|
|
mdvp = &mdvec_1080;
|
1999-12-16 05:42:02 +00:00
|
|
|
basetype = ISP_HA_SCSI_1240;
|
|
|
|
psize = 2 * sizeof (sdparam);
|
|
|
|
pcs->pci_poff[DMA_BLOCK >> _BLK_REG_SHFT] =
|
|
|
|
ISP1080_DMA_REGS_OFF;
|
|
|
|
}
|
2000-02-11 19:45:17 +00:00
|
|
|
if (pci_get_devid(dev) == PCI_QLOGIC_ISP1280) {
|
1999-12-16 05:42:02 +00:00
|
|
|
mdvp = &mdvec_1080;
|
|
|
|
basetype = ISP_HA_SCSI_1280;
|
1999-05-11 04:53:57 +00:00
|
|
|
psize = 2 * sizeof (sdparam);
|
1999-03-17 05:07:18 +00:00
|
|
|
pcs->pci_poff[DMA_BLOCK >> _BLK_REG_SHFT] =
|
|
|
|
ISP1080_DMA_REGS_OFF;
|
|
|
|
}
|
2000-02-11 19:45:17 +00:00
|
|
|
if (pci_get_devid(dev) == PCI_QLOGIC_ISP12160) {
|
|
|
|
mdvp = &mdvec_12160;
|
|
|
|
basetype = ISP_HA_SCSI_12160;
|
|
|
|
psize = 2 * sizeof (sdparam);
|
|
|
|
pcs->pci_poff[DMA_BLOCK >> _BLK_REG_SHFT] =
|
|
|
|
ISP1080_DMA_REGS_OFF;
|
|
|
|
}
|
|
|
|
if (pci_get_devid(dev) == PCI_QLOGIC_ISP2100) {
|
1999-05-11 04:53:57 +00:00
|
|
|
mdvp = &mdvec_2100;
|
|
|
|
basetype = ISP_HA_FC_2100;
|
|
|
|
psize = sizeof (fcparam);
|
1999-03-17 05:07:18 +00:00
|
|
|
pcs->pci_poff[MBOX_BLOCK >> _BLK_REG_SHFT] =
|
|
|
|
PCI_MBOX_REGS2100_OFF;
|
2000-02-11 19:45:17 +00:00
|
|
|
if (pci_get_revid(dev) < 3) {
|
1999-04-04 01:14:02 +00:00
|
|
|
/*
|
|
|
|
* XXX: Need to get the actual revision
|
|
|
|
* XXX: number of the 2100 FB. At any rate,
|
|
|
|
* XXX: lower cache line size for early revision
|
|
|
|
* XXX; boards.
|
|
|
|
*/
|
|
|
|
linesz = 1;
|
|
|
|
}
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
2000-02-11 19:45:17 +00:00
|
|
|
if (pci_get_devid(dev) == PCI_QLOGIC_ISP2200) {
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
mdvp = &mdvec_2200;
|
|
|
|
basetype = ISP_HA_FC_2200;
|
|
|
|
psize = sizeof (fcparam);
|
|
|
|
pcs->pci_poff[MBOX_BLOCK >> _BLK_REG_SHFT] =
|
|
|
|
PCI_MBOX_REGS2100_OFF;
|
|
|
|
}
|
2002-02-04 21:04:25 +00:00
|
|
|
if (pci_get_devid(dev) == PCI_QLOGIC_ISP2300) {
|
2001-08-31 21:39:04 +00:00
|
|
|
mdvp = &mdvec_2300;
|
|
|
|
basetype = ISP_HA_FC_2300;
|
|
|
|
psize = sizeof (fcparam);
|
|
|
|
pcs->pci_poff[MBOX_BLOCK >> _BLK_REG_SHFT] =
|
|
|
|
PCI_MBOX_REGS2300_OFF;
|
|
|
|
}
|
2002-02-04 21:04:25 +00:00
|
|
|
if (pci_get_devid(dev) == PCI_QLOGIC_ISP2312) {
|
|
|
|
mdvp = &mdvec_2300;
|
|
|
|
basetype = ISP_HA_FC_2312;
|
|
|
|
psize = sizeof (fcparam);
|
|
|
|
pcs->pci_poff[MBOX_BLOCK >> _BLK_REG_SHFT] =
|
|
|
|
PCI_MBOX_REGS2300_OFF;
|
|
|
|
}
|
1999-05-11 04:53:57 +00:00
|
|
|
isp = &pcs->pci_isp;
|
2000-12-08 21:51:06 +00:00
|
|
|
isp->isp_param = malloc(psize, M_DEVBUF, M_NOWAIT | M_ZERO);
|
1999-05-11 04:53:57 +00:00
|
|
|
if (isp->isp_param == NULL) {
|
2000-02-11 19:45:17 +00:00
|
|
|
device_printf(dev, "cannot allocate parameter data\n");
|
|
|
|
goto bad;
|
1999-05-11 04:53:57 +00:00
|
|
|
}
|
|
|
|
isp->isp_mdvec = mdvp;
|
|
|
|
isp->isp_type = basetype;
|
2000-02-11 19:45:17 +00:00
|
|
|
isp->isp_revision = pci_get_revid(dev);
|
2001-02-11 03:53:23 +00:00
|
|
|
isp->isp_role = role;
|
2001-03-01 02:21:36 +00:00
|
|
|
isp->isp_dev = dev;
|
1998-04-22 18:12:29 +00:00
|
|
|
|
2000-06-18 05:18:55 +00:00
|
|
|
/*
|
|
|
|
* Try and find firmware for this device.
|
|
|
|
*/
|
|
|
|
|
|
|
|
if (isp_get_firmware_p) {
|
|
|
|
int device = (int) pci_get_device(dev);
|
|
|
|
#ifdef ISP_TARGET_MODE
|
|
|
|
(*isp_get_firmware_p)(0, 1, device, &mdvp->dv_ispfw);
|
|
|
|
#else
|
|
|
|
(*isp_get_firmware_p)(0, 0, device, &mdvp->dv_ispfw);
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
1999-02-09 01:12:52 +00:00
|
|
|
/*
|
|
|
|
* Make sure that SERR, PERR, WRITE INVALIDATE and BUSMASTER
|
|
|
|
* are set.
|
|
|
|
*/
|
2000-02-11 19:45:17 +00:00
|
|
|
cmd |= PCIM_CMD_SEREN | PCIM_CMD_PERRESPEN |
|
|
|
|
PCIM_CMD_BUSMASTEREN | PCIM_CMD_INVEN;
|
2002-02-04 21:04:25 +00:00
|
|
|
if (IS_2300(isp)) { /* per QLogic errata */
|
|
|
|
cmd &= ~PCIM_CMD_INVEN;
|
|
|
|
}
|
2002-06-16 04:58:00 +00:00
|
|
|
if (IS_23XX(isp)) {
|
|
|
|
/*
|
|
|
|
* Can't tell if ROM will hang on 'ABOUT FIRMWARE' command.
|
|
|
|
*/
|
|
|
|
isp->isp_touched = 1;
|
|
|
|
|
|
|
|
}
|
2000-02-11 19:45:17 +00:00
|
|
|
pci_write_config(dev, PCIR_COMMAND, cmd, 1);
|
1999-04-04 01:14:02 +00:00
|
|
|
|
1999-02-09 01:12:52 +00:00
|
|
|
/*
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
* Make sure the Cache Line Size register is set sensibly.
|
1999-02-09 01:12:52 +00:00
|
|
|
*/
|
2000-02-11 19:45:17 +00:00
|
|
|
data = pci_read_config(dev, PCIR_CACHELNSZ, 1);
|
1999-04-04 01:14:02 +00:00
|
|
|
if (data != linesz) {
|
1999-02-09 01:12:52 +00:00
|
|
|
data = PCI_DFLT_LNSZ;
|
2000-08-01 05:16:49 +00:00
|
|
|
isp_prt(isp, ISP_LOGCONFIG, "set PCI line size to %d", data);
|
2000-02-11 19:45:17 +00:00
|
|
|
pci_write_config(dev, PCIR_CACHELNSZ, data, 1);
|
1999-02-09 01:12:52 +00:00
|
|
|
}
|
1999-04-04 01:14:02 +00:00
|
|
|
|
1999-02-09 01:12:52 +00:00
|
|
|
/*
|
|
|
|
* Make sure the Latency Timer is sane.
|
|
|
|
*/
|
2000-02-11 19:45:17 +00:00
|
|
|
data = pci_read_config(dev, PCIR_LATTIMER, 1);
|
1999-02-09 01:12:52 +00:00
|
|
|
if (data < PCI_DFLT_LTNCY) {
|
|
|
|
data = PCI_DFLT_LTNCY;
|
2000-08-01 05:16:49 +00:00
|
|
|
isp_prt(isp, ISP_LOGCONFIG, "set PCI latency to %d", data);
|
2000-02-11 19:45:17 +00:00
|
|
|
pci_write_config(dev, PCIR_LATTIMER, data, 1);
|
1999-02-09 01:12:52 +00:00
|
|
|
}
|
1999-04-04 01:14:02 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Make sure we've disabled the ROM.
|
|
|
|
*/
|
2000-02-11 19:45:17 +00:00
|
|
|
data = pci_read_config(dev, PCIR_ROMADDR, 4);
|
1999-04-04 01:14:02 +00:00
|
|
|
data &= ~1;
|
2000-02-11 19:45:17 +00:00
|
|
|
pci_write_config(dev, PCIR_ROMADDR, data, 4);
|
2000-07-18 06:40:22 +00:00
|
|
|
|
2000-02-11 19:45:17 +00:00
|
|
|
iqd = 0;
|
|
|
|
irq = bus_alloc_resource(dev, SYS_RES_IRQ, &iqd, 0, ~0,
|
|
|
|
1, RF_ACTIVE | RF_SHAREABLE);
|
|
|
|
if (irq == NULL) {
|
|
|
|
device_printf(dev, "could not allocate interrupt\n");
|
|
|
|
goto bad;
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
|
|
|
|
2001-03-01 02:21:36 +00:00
|
|
|
tval = 0;
|
|
|
|
if (resource_int_value(device_get_name(dev), device_get_unit(dev),
|
|
|
|
"fwload_disable", &tval) == 0 && tval != 0) {
|
|
|
|
isp->isp_confopts |= ISP_CFG_NORELOAD;
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
}
|
2001-03-01 02:21:36 +00:00
|
|
|
tval = 0;
|
|
|
|
if (resource_int_value(device_get_name(dev), device_get_unit(dev),
|
|
|
|
"ignore_nvram", &tval) == 0 && tval != 0) {
|
|
|
|
isp->isp_confopts |= ISP_CFG_NONVRAM;
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
}
|
2001-03-01 02:21:36 +00:00
|
|
|
tval = 0;
|
|
|
|
if (resource_int_value(device_get_name(dev), device_get_unit(dev),
|
|
|
|
"fullduplex", &tval) == 0 && tval != 0) {
|
|
|
|
isp->isp_confopts |= ISP_CFG_FULL_DUPLEX;
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
}
|
2002-02-18 00:00:34 +00:00
|
|
|
#ifdef ISP_FW_CRASH_DUMP
|
|
|
|
tval = 0;
|
|
|
|
if (resource_int_value(device_get_name(dev), device_get_unit(dev),
|
|
|
|
"fw_dump_enable", &tval) == 0 && tval != 0) {
|
|
|
|
size_t amt = 0;
|
|
|
|
if (IS_2200(isp)) {
|
|
|
|
amt = QLA2200_RISC_IMAGE_DUMP_SIZE;
|
|
|
|
} else if (IS_23XX(isp)) {
|
|
|
|
amt = QLA2300_RISC_IMAGE_DUMP_SIZE;
|
|
|
|
}
|
|
|
|
if (amt) {
|
|
|
|
FCPARAM(isp)->isp_dump_data =
|
|
|
|
malloc(amt, M_DEVBUF, M_WAITOK | M_ZERO);
|
|
|
|
} else {
|
|
|
|
device_printf(dev,
|
|
|
|
"f/w crash dumps not supported for this model\n");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif
|
2001-03-01 02:21:36 +00:00
|
|
|
|
|
|
|
sptr = 0;
|
|
|
|
if (resource_string_value(device_get_name(dev), device_get_unit(dev),
|
2001-07-04 18:51:58 +00:00
|
|
|
"topology", (const char **) &sptr) == 0 && sptr != 0) {
|
2001-03-01 02:21:36 +00:00
|
|
|
if (strcmp(sptr, "lport") == 0) {
|
|
|
|
isp->isp_confopts |= ISP_CFG_LPORT;
|
|
|
|
} else if (strcmp(sptr, "nport") == 0) {
|
2000-02-11 19:45:17 +00:00
|
|
|
isp->isp_confopts |= ISP_CFG_NPORT;
|
2001-03-01 02:21:36 +00:00
|
|
|
} else if (strcmp(sptr, "lport-only") == 0) {
|
|
|
|
isp->isp_confopts |= ISP_CFG_LPORT_ONLY;
|
|
|
|
} else if (strcmp(sptr, "nport-only") == 0) {
|
|
|
|
isp->isp_confopts |= ISP_CFG_NPORT_ONLY;
|
|
|
|
}
|
2000-02-11 19:45:17 +00:00
|
|
|
}
|
2001-03-01 02:21:36 +00:00
|
|
|
|
1999-11-21 02:56:17 +00:00
|
|
|
/*
|
2001-03-01 02:21:36 +00:00
|
|
|
* Because the resource_*_value functions can neither return
|
|
|
|
* 64 bit integer values, nor can they be directly coerced
|
|
|
|
* to interpret the right hand side of the assignment as
|
|
|
|
* you want them to interpret it, we have to force WWN
|
|
|
|
* hint replacement to specify WWN strings with a leading
|
|
|
|
* 'w' (e..g w50000000aaaa0001). Sigh.
|
1999-11-21 02:56:17 +00:00
|
|
|
*/
|
2001-03-01 02:21:36 +00:00
|
|
|
sptr = 0;
|
|
|
|
tval = resource_string_value(device_get_name(dev), device_get_unit(dev),
|
2001-07-04 18:51:58 +00:00
|
|
|
"portwwn", (const char **) &sptr);
|
2001-03-01 02:21:36 +00:00
|
|
|
if (tval == 0 && sptr != 0 && *sptr++ == 'w') {
|
|
|
|
char *eptr = 0;
|
|
|
|
isp->isp_osinfo.default_port_wwn = strtouq(sptr, &eptr, 16);
|
|
|
|
if (eptr < sptr + 16 || isp->isp_osinfo.default_port_wwn == 0) {
|
|
|
|
device_printf(dev, "mangled portwwn hint '%s'\n", sptr);
|
|
|
|
isp->isp_osinfo.default_port_wwn = 0;
|
|
|
|
} else {
|
2002-02-17 06:34:21 +00:00
|
|
|
isp->isp_confopts |= ISP_CFG_OWNWWPN;
|
2001-03-01 02:21:36 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
if (isp->isp_osinfo.default_port_wwn == 0) {
|
|
|
|
isp->isp_osinfo.default_port_wwn = 0x400000007F000009ull;
|
|
|
|
}
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
|
2001-03-01 02:21:36 +00:00
|
|
|
sptr = 0;
|
|
|
|
tval = resource_string_value(device_get_name(dev), device_get_unit(dev),
|
2001-07-04 18:51:58 +00:00
|
|
|
"nodewwn", (const char **) &sptr);
|
2001-03-01 02:21:36 +00:00
|
|
|
if (tval == 0 && sptr != 0 && *sptr++ == 'w') {
|
|
|
|
char *eptr = 0;
|
|
|
|
isp->isp_osinfo.default_node_wwn = strtouq(sptr, &eptr, 16);
|
|
|
|
if (eptr < sptr + 16 || isp->isp_osinfo.default_node_wwn == 0) {
|
|
|
|
device_printf(dev, "mangled nodewwn hint '%s'\n", sptr);
|
|
|
|
isp->isp_osinfo.default_node_wwn = 0;
|
|
|
|
} else {
|
2002-02-17 06:34:21 +00:00
|
|
|
isp->isp_confopts |= ISP_CFG_OWNWWNN;
|
1999-11-21 02:56:17 +00:00
|
|
|
}
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
}
|
2001-03-01 02:21:36 +00:00
|
|
|
if (isp->isp_osinfo.default_node_wwn == 0) {
|
|
|
|
isp->isp_osinfo.default_node_wwn = 0x400000007F000009ull;
|
|
|
|
}
|
|
|
|
|
2002-07-08 17:34:32 +00:00
|
|
|
if (resource_int_value(device_get_name(dev), device_get_unit(dev),
|
|
|
|
"iid", &tval) == 0) {
|
|
|
|
isp->isp_osinfo.default_id = tval;
|
|
|
|
isp->isp_confopts |= ISP_CFG_OWNLOOPID;
|
|
|
|
}
|
|
|
|
if (isp->isp_osinfo.default_id == 0) {
|
|
|
|
if (IS_FC(isp)) {
|
|
|
|
isp->isp_osinfo.default_id = 109;
|
|
|
|
} else {
|
|
|
|
isp->isp_osinfo.default_id = 7;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2000-08-01 05:16:49 +00:00
|
|
|
isp_debug = 0;
|
2001-03-01 02:21:36 +00:00
|
|
|
(void) resource_int_value(device_get_name(dev), device_get_unit(dev),
|
|
|
|
"debug", &isp_debug);
|
2000-12-29 19:10:16 +00:00
|
|
|
|
|
|
|
/* Make sure the lock is set up. */
|
2002-04-04 21:03:38 +00:00
|
|
|
mtx_init(&isp->isp_osinfo.lock, "isp", NULL, MTX_DEF);
|
2000-12-29 19:10:16 +00:00
|
|
|
locksetup++;
|
|
|
|
|
Spring MegaChange #1.
----
Make a device for each ISP- really usable only with devfs and add an ioctl
entry point (this can be used to (re)set debug levels, reset the HBA,
rescan the fabric, issue lips, etc).
----
Add in a kernel thread for Fibre Channel cards. The purpose of this
thread is to be woken up to clean up after Fibre Channel events
block things. Basically, any FC event that casts doubt on the
location or identify of FC devices blocks the queues. When, and
if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED
async event, we activate the kthread which will then, in full thread
context, re-evaluate the local loop and/or the fabric. When it's
satisfied that things are stable, it can then release the blocked
queues and let commands flow again.
The prior mechanism was a lazy evaluation. That is, the next command
to come down the pipe after change events would pay the full price
for re-evaluation. And if this was done off of a softcall, it really
could hang up the system.
These changes brings the FreeBSD port more in line with the Solaris,
Linux and NetBSD ports. It also, more importantly, gets us being
more proactive about topology changes which could then be reflected
upwards to CAM so that the periph driver can be informed sooner
rather than later when things arrive or depart.
---
Add in the (correct) usage of locking macros- we now have lock transition
macros which allow us to transition from holding the CAM lock (Giant)
and grabbing the softc lock and vice versa. Switch over to having this
HBA do real locking. Some folks claim this won't be a win. They're right.
But you have to start somewhere, and this will begin to teach us how
to DTRT for HBAs, etc.
--
Start putting in prototype 2300 support. Add back in LIP
and Loop Reset as async events that each platform will handle.
Add in another int_bogus instrumentation point.
Do some more substantial target mode cleanups.
MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
|
|
|
#ifdef ISP_SMPLOCK
|
|
|
|
#define INTR_FLAGS INTR_TYPE_CAM | INTR_MPSAFE | INTR_ENTROPY
|
2000-12-29 19:10:16 +00:00
|
|
|
#else
|
Spring MegaChange #1.
----
Make a device for each ISP- really usable only with devfs and add an ioctl
entry point (this can be used to (re)set debug levels, reset the HBA,
rescan the fabric, issue lips, etc).
----
Add in a kernel thread for Fibre Channel cards. The purpose of this
thread is to be woken up to clean up after Fibre Channel events
block things. Basically, any FC event that casts doubt on the
location or identify of FC devices blocks the queues. When, and
if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED
async event, we activate the kthread which will then, in full thread
context, re-evaluate the local loop and/or the fabric. When it's
satisfied that things are stable, it can then release the blocked
queues and let commands flow again.
The prior mechanism was a lazy evaluation. That is, the next command
to come down the pipe after change events would pay the full price
for re-evaluation. And if this was done off of a softcall, it really
could hang up the system.
These changes brings the FreeBSD port more in line with the Solaris,
Linux and NetBSD ports. It also, more importantly, gets us being
more proactive about topology changes which could then be reflected
upwards to CAM so that the periph driver can be informed sooner
rather than later when things arrive or depart.
---
Add in the (correct) usage of locking macros- we now have lock transition
macros which allow us to transition from holding the CAM lock (Giant)
and grabbing the softc lock and vice versa. Switch over to having this
HBA do real locking. Some folks claim this won't be a win. They're right.
But you have to start somewhere, and this will begin to teach us how
to DTRT for HBAs, etc.
--
Start putting in prototype 2300 support. Add back in LIP
and Loop Reset as async events that each platform will handle.
Add in another int_bogus instrumentation point.
Do some more substantial target mode cleanups.
MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
|
|
|
#define INTR_FLAGS INTR_TYPE_CAM | INTR_ENTROPY
|
|
|
|
#endif
|
|
|
|
if (bus_setup_intr(dev, irq, INTR_FLAGS, isp_pci_intr, isp, &pcs->ih)) {
|
2000-02-11 19:45:17 +00:00
|
|
|
device_printf(dev, "could not setup interrupt\n");
|
|
|
|
goto bad;
|
|
|
|
}
|
|
|
|
|
2000-08-01 05:16:49 +00:00
|
|
|
/*
|
|
|
|
* Set up logging levels.
|
|
|
|
*/
|
|
|
|
if (isp_debug) {
|
|
|
|
isp->isp_dblev = isp_debug;
|
|
|
|
} else {
|
|
|
|
isp->isp_dblev = ISP_LOGWARN|ISP_LOGERR;
|
|
|
|
}
|
|
|
|
if (bootverbose)
|
2000-12-05 07:38:41 +00:00
|
|
|
isp->isp_dblev |= ISP_LOGCONFIG|ISP_LOGINFO;
|
2000-08-01 05:16:49 +00:00
|
|
|
|
2002-02-04 21:04:25 +00:00
|
|
|
/*
|
|
|
|
* Last minute checks...
|
|
|
|
*/
|
|
|
|
if (IS_2312(isp)) {
|
|
|
|
isp->isp_port = pci_get_function(dev);
|
|
|
|
}
|
|
|
|
|
2000-07-18 06:40:22 +00:00
|
|
|
/*
|
|
|
|
* Make sure we're in reset state.
|
|
|
|
*/
|
2000-10-25 04:40:49 +00:00
|
|
|
ISP_LOCK(isp);
|
1998-04-22 18:12:29 +00:00
|
|
|
isp_reset(isp);
|
|
|
|
if (isp->isp_state != ISP_RESETSTATE) {
|
2000-10-25 04:40:49 +00:00
|
|
|
ISP_UNLOCK(isp);
|
2000-02-11 19:45:17 +00:00
|
|
|
goto bad;
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
|
|
|
isp_init(isp);
|
2001-02-11 03:53:23 +00:00
|
|
|
if (isp->isp_role != ISP_ROLE_NONE && isp->isp_state != ISP_INITSTATE) {
|
|
|
|
isp_uninit(isp);
|
|
|
|
ISP_UNLOCK(isp);
|
|
|
|
goto bad;
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
|
|
|
isp_attach(isp);
|
2001-02-11 03:53:23 +00:00
|
|
|
if (isp->isp_role != ISP_ROLE_NONE && isp->isp_state != ISP_RUNSTATE) {
|
|
|
|
isp_uninit(isp);
|
|
|
|
ISP_UNLOCK(isp);
|
|
|
|
goto bad;
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
2000-06-18 05:18:55 +00:00
|
|
|
/*
|
|
|
|
* XXXX: Here is where we might unload the f/w module
|
|
|
|
* XXXX: (or decrease the reference count to it).
|
|
|
|
*/
|
2000-10-25 04:40:49 +00:00
|
|
|
ISP_UNLOCK(isp);
|
2000-02-11 19:45:17 +00:00
|
|
|
return (0);
|
|
|
|
|
|
|
|
bad:
|
|
|
|
|
|
|
|
if (pcs && pcs->ih) {
|
|
|
|
(void) bus_teardown_intr(dev, irq, pcs->ih);
|
|
|
|
}
|
|
|
|
|
2000-10-25 04:40:49 +00:00
|
|
|
if (locksetup && isp) {
|
|
|
|
mtx_destroy(&isp->isp_osinfo.lock);
|
|
|
|
}
|
|
|
|
|
2000-02-11 19:45:17 +00:00
|
|
|
if (irq) {
|
|
|
|
(void) bus_release_resource(dev, SYS_RES_IRQ, iqd, irq);
|
|
|
|
}
|
2000-10-25 04:40:49 +00:00
|
|
|
|
|
|
|
|
2000-02-11 19:45:17 +00:00
|
|
|
if (regs) {
|
|
|
|
(void) bus_release_resource(dev, rtp, rgd, regs);
|
|
|
|
}
|
2000-10-25 04:40:49 +00:00
|
|
|
|
2000-02-11 19:45:17 +00:00
|
|
|
if (pcs) {
|
|
|
|
if (pcs->pci_isp.isp_param)
|
|
|
|
free(pcs->pci_isp.isp_param, M_DEVBUF);
|
|
|
|
free(pcs, M_DEVBUF);
|
|
|
|
}
|
2000-10-25 04:40:49 +00:00
|
|
|
|
2000-06-18 05:18:55 +00:00
|
|
|
/*
|
|
|
|
* XXXX: Here is where we might unload the f/w module
|
|
|
|
* XXXX: (or decrease the reference count to it).
|
|
|
|
*/
|
2000-02-11 19:45:17 +00:00
|
|
|
return (ENXIO);
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
|
|
|
|
2000-12-29 19:10:16 +00:00
|
|
|
static void
|
|
|
|
isp_pci_intr(void *arg)
|
|
|
|
{
|
|
|
|
struct ispsoftc *isp = arg;
|
2001-08-31 21:39:04 +00:00
|
|
|
u_int16_t isr, sema, mbox;
|
|
|
|
|
2000-12-29 19:10:16 +00:00
|
|
|
ISP_LOCK(isp);
|
2001-08-31 21:39:04 +00:00
|
|
|
isp->isp_intcnt++;
|
|
|
|
if (ISP_READ_ISR(isp, &isr, &sema, &mbox) == 0) {
|
|
|
|
isp->isp_intbogus++;
|
|
|
|
} else {
|
2001-09-04 21:45:57 +00:00
|
|
|
int iok = isp->isp_osinfo.intsok;
|
|
|
|
isp->isp_osinfo.intsok = 0;
|
2001-08-31 21:39:04 +00:00
|
|
|
isp_intr(isp, isr, sema, mbox);
|
2001-09-04 21:45:57 +00:00
|
|
|
isp->isp_osinfo.intsok = iok;
|
2001-08-31 21:39:04 +00:00
|
|
|
}
|
2000-12-29 19:10:16 +00:00
|
|
|
ISP_UNLOCK(isp);
|
|
|
|
}
|
|
|
|
|
2001-08-31 21:39:04 +00:00
|
|
|
|
|
|
|
#define IspVirt2Off(a, x) \
|
|
|
|
(((struct isp_pcisoftc *)a)->pci_poff[((x) & _BLK_REG_MASK) >> \
|
|
|
|
_BLK_REG_SHFT] + ((x) & 0xff))
|
|
|
|
|
|
|
|
#define BXR2(pcs, off) \
|
|
|
|
bus_space_read_2(pcs->pci_st, pcs->pci_sh, off)
|
|
|
|
#define BXW2(pcs, off, v) \
|
|
|
|
bus_space_write_2(pcs->pci_st, pcs->pci_sh, off, v)
|
|
|
|
|
|
|
|
|
|
|
|
static INLINE int
|
|
|
|
isp_pci_rd_debounced(struct ispsoftc *isp, int off, u_int16_t *rp)
|
|
|
|
{
|
|
|
|
struct isp_pcisoftc *pcs = (struct isp_pcisoftc *) isp;
|
|
|
|
u_int16_t val0, val1;
|
|
|
|
int i = 0;
|
|
|
|
|
|
|
|
do {
|
|
|
|
val0 = BXR2(pcs, IspVirt2Off(isp, off));
|
|
|
|
val1 = BXR2(pcs, IspVirt2Off(isp, off));
|
|
|
|
} while (val0 != val1 && ++i < 1000);
|
|
|
|
if (val0 != val1) {
|
|
|
|
return (1);
|
|
|
|
}
|
|
|
|
*rp = val0;
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
isp_pci_rd_isr(struct ispsoftc *isp, u_int16_t *isrp,
|
|
|
|
u_int16_t *semap, u_int16_t *mbp)
|
|
|
|
{
|
|
|
|
struct isp_pcisoftc *pcs = (struct isp_pcisoftc *) isp;
|
|
|
|
u_int16_t isr, sema;
|
|
|
|
|
|
|
|
if (IS_2100(isp)) {
|
|
|
|
if (isp_pci_rd_debounced(isp, BIU_ISR, &isr)) {
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
if (isp_pci_rd_debounced(isp, BIU_SEMA, &sema)) {
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
isr = BXR2(pcs, IspVirt2Off(isp, BIU_ISR));
|
|
|
|
sema = BXR2(pcs, IspVirt2Off(isp, BIU_SEMA));
|
|
|
|
}
|
|
|
|
isp_prt(isp, ISP_LOGDEBUG3, "ISR 0x%x SEMA 0x%x", isr, sema);
|
|
|
|
isr &= INT_PENDING_MASK(isp);
|
|
|
|
sema &= BIU_SEMA_LOCK;
|
|
|
|
if (isr == 0 && sema == 0) {
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
*isrp = isr;
|
|
|
|
if ((*semap = sema) != 0) {
|
|
|
|
if (IS_2100(isp)) {
|
|
|
|
if (isp_pci_rd_debounced(isp, OUTMAILBOX0, mbp)) {
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
*mbp = BXR2(pcs, IspVirt2Off(isp, OUTMAILBOX0));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return (1);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
isp_pci_rd_isr_2300(struct ispsoftc *isp, u_int16_t *isrp,
|
|
|
|
u_int16_t *semap, u_int16_t *mbox0p)
|
|
|
|
{
|
|
|
|
struct isp_pcisoftc *pcs = (struct isp_pcisoftc *) isp;
|
|
|
|
u_int32_t r2hisr;
|
|
|
|
|
2001-10-06 19:34:43 +00:00
|
|
|
if (!(BXR2(pcs, IspVirt2Off(isp, BIU_ISR) & BIU2100_ISR_RISC_INT))) {
|
|
|
|
*isrp = 0;
|
2001-10-06 19:19:24 +00:00
|
|
|
return (0);
|
2001-10-06 19:34:43 +00:00
|
|
|
}
|
2001-08-31 21:39:04 +00:00
|
|
|
r2hisr = bus_space_read_4(pcs->pci_st, pcs->pci_sh,
|
|
|
|
IspVirt2Off(pcs, BIU_R2HSTSLO));
|
|
|
|
isp_prt(isp, ISP_LOGDEBUG3, "RISC2HOST ISR 0x%x", r2hisr);
|
|
|
|
if ((r2hisr & BIU_R2HST_INTR) == 0) {
|
|
|
|
*isrp = 0;
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
switch (r2hisr & BIU_R2HST_ISTAT_MASK) {
|
|
|
|
case ISPR2HST_ROM_MBX_OK:
|
|
|
|
case ISPR2HST_ROM_MBX_FAIL:
|
|
|
|
case ISPR2HST_MBX_OK:
|
|
|
|
case ISPR2HST_MBX_FAIL:
|
|
|
|
case ISPR2HST_ASYNC_EVENT:
|
2002-02-04 21:04:25 +00:00
|
|
|
case ISPR2HST_RIO_16:
|
2001-08-31 21:39:04 +00:00
|
|
|
case ISPR2HST_FPOST:
|
|
|
|
case ISPR2HST_FPOST_CTIO:
|
|
|
|
*isrp = r2hisr & 0xffff;
|
|
|
|
*mbox0p = (r2hisr >> 16);
|
|
|
|
*semap = 1;
|
|
|
|
return (1);
|
|
|
|
case ISPR2HST_RSPQ_UPDATE:
|
|
|
|
*isrp = r2hisr & 0xffff;
|
|
|
|
*mbox0p = 0;
|
|
|
|
*semap = 0;
|
|
|
|
return (1);
|
|
|
|
default:
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
1999-03-17 05:07:18 +00:00
|
|
|
static u_int16_t
|
2001-08-31 21:39:04 +00:00
|
|
|
isp_pci_rd_reg(struct ispsoftc *isp, int regoff)
|
1999-03-17 05:07:18 +00:00
|
|
|
{
|
|
|
|
u_int16_t rv;
|
|
|
|
struct isp_pcisoftc *pcs = (struct isp_pcisoftc *) isp;
|
2001-08-31 21:39:04 +00:00
|
|
|
int oldconf = 0;
|
1998-04-22 18:12:29 +00:00
|
|
|
|
1999-03-17 05:07:18 +00:00
|
|
|
if ((regoff & _BLK_REG_MASK) == SXP_BLOCK) {
|
|
|
|
/*
|
|
|
|
* We will assume that someone has paused the RISC processor.
|
|
|
|
*/
|
2001-08-31 21:39:04 +00:00
|
|
|
oldconf = BXR2(pcs, IspVirt2Off(isp, BIU_CONF1));
|
|
|
|
BXW2(pcs, IspVirt2Off(isp, BIU_CONF1),
|
|
|
|
oldconf | BIU_PCI_CONF1_SXP);
|
1999-03-17 05:07:18 +00:00
|
|
|
}
|
2001-08-31 21:39:04 +00:00
|
|
|
rv = BXR2(pcs, IspVirt2Off(isp, regoff));
|
1999-03-17 05:07:18 +00:00
|
|
|
if ((regoff & _BLK_REG_MASK) == SXP_BLOCK) {
|
2001-08-31 21:39:04 +00:00
|
|
|
BXW2(pcs, IspVirt2Off(isp, BIU_CONF1), oldconf);
|
1999-03-17 05:07:18 +00:00
|
|
|
}
|
|
|
|
return (rv);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2001-08-31 21:39:04 +00:00
|
|
|
isp_pci_wr_reg(struct ispsoftc *isp, int regoff, u_int16_t val)
|
1999-03-17 05:07:18 +00:00
|
|
|
{
|
|
|
|
struct isp_pcisoftc *pcs = (struct isp_pcisoftc *) isp;
|
2001-08-31 21:39:04 +00:00
|
|
|
int oldconf = 0;
|
1999-03-17 05:07:18 +00:00
|
|
|
|
|
|
|
if ((regoff & _BLK_REG_MASK) == SXP_BLOCK) {
|
|
|
|
/*
|
|
|
|
* We will assume that someone has paused the RISC processor.
|
|
|
|
*/
|
2001-08-31 21:39:04 +00:00
|
|
|
oldconf = BXR2(pcs, IspVirt2Off(isp, BIU_CONF1));
|
|
|
|
BXW2(pcs, IspVirt2Off(isp, BIU_CONF1),
|
|
|
|
oldconf | BIU_PCI_CONF1_SXP);
|
1999-03-17 05:07:18 +00:00
|
|
|
}
|
2001-08-31 21:39:04 +00:00
|
|
|
BXW2(pcs, IspVirt2Off(isp, regoff), val);
|
1999-03-17 05:07:18 +00:00
|
|
|
if ((regoff & _BLK_REG_MASK) == SXP_BLOCK) {
|
2001-08-31 21:39:04 +00:00
|
|
|
BXW2(pcs, IspVirt2Off(isp, BIU_CONF1), oldconf);
|
1999-03-17 05:07:18 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
1998-04-22 18:12:29 +00:00
|
|
|
static u_int16_t
|
2001-08-31 21:39:04 +00:00
|
|
|
isp_pci_rd_reg_1080(struct ispsoftc *isp, int regoff)
|
1998-04-22 18:12:29 +00:00
|
|
|
{
|
1999-12-16 05:42:02 +00:00
|
|
|
u_int16_t rv, oc = 0;
|
1998-04-22 18:12:29 +00:00
|
|
|
struct isp_pcisoftc *pcs = (struct isp_pcisoftc *) isp;
|
1999-03-17 05:07:18 +00:00
|
|
|
|
1999-12-16 05:42:02 +00:00
|
|
|
if ((regoff & _BLK_REG_MASK) == SXP_BLOCK ||
|
|
|
|
(regoff & _BLK_REG_MASK) == (SXP_BLOCK|SXP_BANK1_SELECT)) {
|
|
|
|
u_int16_t tc;
|
1998-04-22 18:12:29 +00:00
|
|
|
/*
|
|
|
|
* We will assume that someone has paused the RISC processor.
|
|
|
|
*/
|
2001-08-31 21:39:04 +00:00
|
|
|
oc = BXR2(pcs, IspVirt2Off(isp, BIU_CONF1));
|
1999-12-16 05:42:02 +00:00
|
|
|
tc = oc & ~BIU_PCI1080_CONF1_DMA;
|
|
|
|
if (regoff & SXP_BANK1_SELECT)
|
|
|
|
tc |= BIU_PCI1080_CONF1_SXP1;
|
|
|
|
else
|
|
|
|
tc |= BIU_PCI1080_CONF1_SXP0;
|
2001-08-31 21:39:04 +00:00
|
|
|
BXW2(pcs, IspVirt2Off(isp, BIU_CONF1), tc);
|
1999-03-17 05:07:18 +00:00
|
|
|
} else if ((regoff & _BLK_REG_MASK) == DMA_BLOCK) {
|
2001-08-31 21:39:04 +00:00
|
|
|
oc = BXR2(pcs, IspVirt2Off(isp, BIU_CONF1));
|
|
|
|
BXW2(pcs, IspVirt2Off(isp, BIU_CONF1),
|
|
|
|
oc | BIU_PCI1080_CONF1_DMA);
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
2001-08-31 21:39:04 +00:00
|
|
|
rv = BXR2(pcs, IspVirt2Off(isp, regoff));
|
1999-12-16 05:42:02 +00:00
|
|
|
if (oc) {
|
2001-08-31 21:39:04 +00:00
|
|
|
BXW2(pcs, IspVirt2Off(isp, BIU_CONF1), oc);
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
|
|
|
return (rv);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2001-08-31 21:39:04 +00:00
|
|
|
isp_pci_wr_reg_1080(struct ispsoftc *isp, int regoff, u_int16_t val)
|
1998-04-22 18:12:29 +00:00
|
|
|
{
|
|
|
|
struct isp_pcisoftc *pcs = (struct isp_pcisoftc *) isp;
|
2001-08-31 21:39:04 +00:00
|
|
|
int oc = 0;
|
1999-03-17 05:07:18 +00:00
|
|
|
|
1999-12-16 05:42:02 +00:00
|
|
|
if ((regoff & _BLK_REG_MASK) == SXP_BLOCK ||
|
|
|
|
(regoff & _BLK_REG_MASK) == (SXP_BLOCK|SXP_BANK1_SELECT)) {
|
|
|
|
u_int16_t tc;
|
1998-04-22 18:12:29 +00:00
|
|
|
/*
|
|
|
|
* We will assume that someone has paused the RISC processor.
|
|
|
|
*/
|
2001-08-31 21:39:04 +00:00
|
|
|
oc = BXR2(pcs, IspVirt2Off(isp, BIU_CONF1));
|
1999-12-16 05:42:02 +00:00
|
|
|
tc = oc & ~BIU_PCI1080_CONF1_DMA;
|
|
|
|
if (regoff & SXP_BANK1_SELECT)
|
|
|
|
tc |= BIU_PCI1080_CONF1_SXP1;
|
|
|
|
else
|
|
|
|
tc |= BIU_PCI1080_CONF1_SXP0;
|
2001-08-31 21:39:04 +00:00
|
|
|
BXW2(pcs, IspVirt2Off(isp, BIU_CONF1), tc);
|
1999-03-17 05:07:18 +00:00
|
|
|
} else if ((regoff & _BLK_REG_MASK) == DMA_BLOCK) {
|
2001-08-31 21:39:04 +00:00
|
|
|
oc = BXR2(pcs, IspVirt2Off(isp, BIU_CONF1));
|
|
|
|
BXW2(pcs, IspVirt2Off(isp, BIU_CONF1),
|
|
|
|
oc | BIU_PCI1080_CONF1_DMA);
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
2001-08-31 21:39:04 +00:00
|
|
|
BXW2(pcs, IspVirt2Off(isp, regoff), val);
|
1999-12-16 05:42:02 +00:00
|
|
|
if (oc) {
|
2001-08-31 21:39:04 +00:00
|
|
|
BXW2(pcs, IspVirt2Off(isp, BIU_CONF1), oc);
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
1998-09-15 10:06:23 +00:00
|
|
|
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
struct imush {
|
|
|
|
struct ispsoftc *isp;
|
|
|
|
int error;
|
|
|
|
};
|
|
|
|
|
2002-04-02 23:36:14 +00:00
|
|
|
static void imc(void *, bus_dma_segment_t *, int, int);
|
1998-09-15 10:06:23 +00:00
|
|
|
|
|
|
|
static void
|
2002-04-02 23:36:14 +00:00
|
|
|
imc(void *arg, bus_dma_segment_t *segs, int nseg, int error)
|
1998-09-15 10:06:23 +00:00
|
|
|
{
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
struct imush *imushp = (struct imush *) arg;
|
|
|
|
if (error) {
|
|
|
|
imushp->error = error;
|
|
|
|
} else {
|
2002-04-02 23:36:14 +00:00
|
|
|
struct ispsoftc *isp =imushp->isp;
|
|
|
|
bus_addr_t addr = segs->ds_addr;
|
1998-09-15 10:06:23 +00:00
|
|
|
|
2002-04-02 23:36:14 +00:00
|
|
|
isp->isp_rquest_dma = addr;
|
|
|
|
addr += ISP_QUEUE_SIZE(RQUEST_QUEUE_LEN(isp));
|
|
|
|
isp->isp_result_dma = addr;
|
|
|
|
if (IS_FC(isp)) {
|
|
|
|
addr += ISP_QUEUE_SIZE(RESULT_QUEUE_LEN(isp));
|
|
|
|
FCPARAM(isp)->isp_scdma = addr;
|
|
|
|
}
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
}
|
1998-09-15 10:06:23 +00:00
|
|
|
}
|
|
|
|
|
2002-04-04 23:46:01 +00:00
|
|
|
/*
|
|
|
|
* Should be BUS_SPACE_MAXSIZE, but MAXPHYS is larger than BUS_SPACE_MAXSIZE
|
|
|
|
*/
|
|
|
|
#define ISP_NSEGS ((MAXPHYS / PAGE_SIZE) + 1)
|
2002-04-02 23:36:14 +00:00
|
|
|
|
1998-09-15 10:06:23 +00:00
|
|
|
static int
|
1998-12-28 19:24:23 +00:00
|
|
|
isp_pci_mbxdma(struct ispsoftc *isp)
|
1998-09-15 10:06:23 +00:00
|
|
|
{
|
2002-04-02 23:36:14 +00:00
|
|
|
struct isp_pcisoftc *pcs = (struct isp_pcisoftc *)isp;
|
1998-09-15 10:06:23 +00:00
|
|
|
caddr_t base;
|
|
|
|
u_int32_t len;
|
2002-04-02 23:36:14 +00:00
|
|
|
int i, error, ns;
|
|
|
|
bus_size_t bl;
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
struct imush im;
|
|
|
|
|
1999-10-17 19:03:11 +00:00
|
|
|
/*
|
|
|
|
* Already been here? If so, leave...
|
|
|
|
*/
|
|
|
|
if (isp->isp_rquest) {
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
2002-04-02 23:36:14 +00:00
|
|
|
if (IS_ULTRA2(isp) || IS_FC(isp) || IS_1240(isp)) {
|
|
|
|
bl = BUS_SPACE_UNRESTRICTED;
|
|
|
|
} else {
|
|
|
|
bl = BUS_SPACE_MAXADDR_24BIT;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (bus_dma_tag_create(NULL, 1, 0, BUS_SPACE_MAXADDR,
|
|
|
|
BUS_SPACE_MAXADDR, NULL, NULL, BUS_SPACE_MAXSIZE,
|
|
|
|
ISP_NSEGS, bl, 0, &pcs->dmat)) {
|
|
|
|
isp_prt(isp, ISP_LOGERR, "could not create master dma tag");
|
|
|
|
return(1);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2000-08-01 05:16:49 +00:00
|
|
|
len = sizeof (XS_T **) * isp->isp_maxcmds;
|
2000-12-08 21:51:06 +00:00
|
|
|
isp->isp_xflist = (XS_T **) malloc(len, M_DEVBUF, M_WAITOK | M_ZERO);
|
1999-10-17 19:03:11 +00:00
|
|
|
if (isp->isp_xflist == NULL) {
|
2000-08-01 05:16:49 +00:00
|
|
|
isp_prt(isp, ISP_LOGERR, "cannot alloc xflist array");
|
1999-10-17 19:03:11 +00:00
|
|
|
return (1);
|
|
|
|
}
|
|
|
|
len = sizeof (bus_dmamap_t) * isp->isp_maxcmds;
|
2002-04-02 23:36:14 +00:00
|
|
|
pcs->dmaps = (bus_dmamap_t *) malloc(len, M_DEVBUF, M_WAITOK);
|
|
|
|
if (pcs->dmaps == NULL) {
|
|
|
|
isp_prt(isp, ISP_LOGERR, "can't alloc dma map storage");
|
1999-10-17 19:03:11 +00:00
|
|
|
free(isp->isp_xflist, M_DEVBUF);
|
|
|
|
return (1);
|
|
|
|
}
|
|
|
|
|
1998-09-15 10:06:23 +00:00
|
|
|
/*
|
|
|
|
* Allocate and map the request, result queues, plus FC scratch area.
|
|
|
|
*/
|
2000-08-01 05:16:49 +00:00
|
|
|
len = ISP_QUEUE_SIZE(RQUEST_QUEUE_LEN(isp));
|
|
|
|
len += ISP_QUEUE_SIZE(RESULT_QUEUE_LEN(isp));
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
if (IS_FC(isp)) {
|
1998-09-15 10:06:23 +00:00
|
|
|
len += ISP2100_SCRLEN;
|
|
|
|
}
|
2002-04-02 23:36:14 +00:00
|
|
|
|
|
|
|
ns = (len / PAGE_SIZE) + 1;
|
|
|
|
if (bus_dma_tag_create(pcs->dmat, QENTRY_LEN, 0, BUS_SPACE_MAXADDR,
|
|
|
|
BUS_SPACE_MAXADDR, NULL, NULL, len, ns, bl, 0, &isp->isp_cdmat)) {
|
2000-12-05 07:38:41 +00:00
|
|
|
isp_prt(isp, ISP_LOGERR,
|
|
|
|
"cannot create a dma tag for control spaces");
|
2002-04-02 23:36:14 +00:00
|
|
|
free(pcs->dmaps, M_DEVBUF);
|
1999-10-17 19:03:11 +00:00
|
|
|
free(isp->isp_xflist, M_DEVBUF);
|
1998-09-15 10:06:23 +00:00
|
|
|
return (1);
|
|
|
|
}
|
|
|
|
|
2002-04-02 23:36:14 +00:00
|
|
|
if (bus_dmamem_alloc(isp->isp_cdmat, (void **)&base, BUS_DMA_NOWAIT,
|
|
|
|
&isp->isp_cdmap) != 0) {
|
2000-12-05 07:38:41 +00:00
|
|
|
isp_prt(isp, ISP_LOGERR,
|
2002-04-02 23:36:14 +00:00
|
|
|
"cannot allocate %d bytes of CCB memory", len);
|
|
|
|
bus_dma_tag_destroy(isp->isp_cdmat);
|
1999-10-17 19:03:11 +00:00
|
|
|
free(isp->isp_xflist, M_DEVBUF);
|
2002-04-02 23:36:14 +00:00
|
|
|
free(pcs->dmaps, M_DEVBUF);
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
return (1);
|
|
|
|
}
|
1998-09-15 10:06:23 +00:00
|
|
|
|
1999-10-17 19:03:11 +00:00
|
|
|
for (i = 0; i < isp->isp_maxcmds; i++) {
|
2002-04-02 23:36:14 +00:00
|
|
|
error = bus_dmamap_create(pcs->dmat, 0, &pcs->dmaps[i]);
|
1998-09-15 10:06:23 +00:00
|
|
|
if (error) {
|
2000-12-05 07:38:41 +00:00
|
|
|
isp_prt(isp, ISP_LOGERR,
|
|
|
|
"error %d creating per-cmd DMA maps", error);
|
2002-04-02 23:36:14 +00:00
|
|
|
while (--i >= 0) {
|
|
|
|
bus_dmamap_destroy(pcs->dmat, pcs->dmaps[i]);
|
|
|
|
}
|
|
|
|
goto bad;
|
1998-09-15 10:06:23 +00:00
|
|
|
}
|
|
|
|
}
|
1999-10-17 19:03:11 +00:00
|
|
|
|
2002-04-02 23:36:14 +00:00
|
|
|
im.isp = isp;
|
|
|
|
im.error = 0;
|
|
|
|
bus_dmamap_load(isp->isp_cdmat, isp->isp_cdmap, base, len, imc, &im, 0);
|
|
|
|
if (im.error) {
|
|
|
|
isp_prt(isp, ISP_LOGERR,
|
|
|
|
"error %d loading dma map for control areas", im.error);
|
|
|
|
goto bad;
|
|
|
|
}
|
|
|
|
|
|
|
|
isp->isp_rquest = base;
|
|
|
|
base += ISP_QUEUE_SIZE(RQUEST_QUEUE_LEN(isp));
|
|
|
|
isp->isp_result = base;
|
Remove pre-CAM code. Add in getenv_int calls for variables isp_mem_map,
isp_io_map, isp_no_fwload, isp_fwload, isp_no_nvram, isp_fcduplex
which are all bitmaps of isp instances that should or shouldn't
map memory space, I/O space, not load f/w, load f/w, ignore nvram,
not ignore nvarm, set full duplex mode. Also have an isp_seed value
that we can use to generate a pseudo seed for a synthetic WWN.
Other minor cosmetic cleanup. Add in support for the Qlogic ISP
2200. Very important change where we actually check now to see
whether we were successful in mapping request and response queues
(and fibre channel scratch space).
1999-07-02 23:18:03 +00:00
|
|
|
if (IS_FC(isp)) {
|
2002-04-02 23:36:14 +00:00
|
|
|
base += ISP_QUEUE_SIZE(RESULT_QUEUE_LEN(isp));
|
|
|
|
FCPARAM(isp)->isp_scratch = base;
|
1999-03-25 22:53:56 +00:00
|
|
|
}
|
1998-09-15 10:06:23 +00:00
|
|
|
return (0);
|
2002-04-02 23:36:14 +00:00
|
|
|
|
|
|
|
bad:
|
|
|
|
bus_dmamem_free(isp->isp_cdmat, base, isp->isp_cdmap);
|
|
|
|
bus_dma_tag_destroy(isp->isp_cdmat);
|
|
|
|
free(isp->isp_xflist, M_DEVBUF);
|
|
|
|
free(pcs->dmaps, M_DEVBUF);
|
|
|
|
isp->isp_rquest = NULL;
|
|
|
|
return (1);
|
1998-09-15 10:06:23 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
typedef struct {
|
|
|
|
struct ispsoftc *isp;
|
2000-01-03 23:50:13 +00:00
|
|
|
void *cmd_token;
|
|
|
|
void *rq;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
u_int16_t *nxtip;
|
1999-11-21 02:56:17 +00:00
|
|
|
u_int16_t optr;
|
1998-09-15 10:06:23 +00:00
|
|
|
u_int error;
|
|
|
|
} mush_t;
|
|
|
|
|
1998-09-17 21:11:21 +00:00
|
|
|
#define MUSHERR_NOQENTRIES -2
|
|
|
|
|
2000-01-03 23:50:13 +00:00
|
|
|
#ifdef ISP_TARGET_MODE
|
|
|
|
/*
|
|
|
|
* We need to handle DMA for target mode differently from initiator mode.
|
|
|
|
*
|
|
|
|
* DMA mapping and construction and submission of CTIO Request Entries
|
|
|
|
* and rendevous for completion are very tightly coupled because we start
|
|
|
|
* out by knowing (per platform) how much data we have to move, but we
|
|
|
|
* don't know, up front, how many DMA mapping segments will have to be used
|
|
|
|
* cover that data, so we don't know how many CTIO Request Entries we
|
|
|
|
* will end up using. Further, for performance reasons we may want to
|
|
|
|
* (on the last CTIO for Fibre Channel), send status too (if all went well).
|
|
|
|
*
|
|
|
|
* The standard vector still goes through isp_pci_dmasetup, but the callback
|
|
|
|
* for the DMA mapping routines comes here instead with the whole transfer
|
|
|
|
* mapped and a pointer to a partially filled in already allocated request
|
|
|
|
* queue entry. We finish the job.
|
|
|
|
*/
|
2001-04-04 21:53:59 +00:00
|
|
|
static void tdma_mk(void *, bus_dma_segment_t *, int, int);
|
|
|
|
static void tdma_mkfc(void *, bus_dma_segment_t *, int, int);
|
|
|
|
|
|
|
|
#define STATUS_WITH_DATA 1
|
2000-01-03 23:50:13 +00:00
|
|
|
|
|
|
|
static void
|
2000-07-18 06:40:22 +00:00
|
|
|
tdma_mk(void *arg, bus_dma_segment_t *dm_segs, int nseg, int error)
|
2000-01-03 23:50:13 +00:00
|
|
|
{
|
|
|
|
mush_t *mp;
|
|
|
|
struct ccb_scsiio *csio;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
struct ispsoftc *isp;
|
2002-04-02 23:36:14 +00:00
|
|
|
struct isp_pcisoftc *pcs;
|
2000-01-03 23:50:13 +00:00
|
|
|
bus_dmamap_t *dp;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
ct_entry_t *cto, *qe;
|
2000-07-18 06:40:22 +00:00
|
|
|
u_int8_t scsi_status;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
u_int16_t curi, nxti, handle;
|
2001-09-04 21:45:57 +00:00
|
|
|
u_int32_t sflags;
|
2000-07-18 06:40:22 +00:00
|
|
|
int32_t resid;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
int nth_ctio, nctios, send_status;
|
2000-01-03 23:50:13 +00:00
|
|
|
|
|
|
|
mp = (mush_t *) arg;
|
|
|
|
if (error) {
|
|
|
|
mp->error = error;
|
|
|
|
return;
|
|
|
|
}
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
|
|
|
|
isp = mp->isp;
|
2000-01-03 23:50:13 +00:00
|
|
|
csio = mp->cmd_token;
|
|
|
|
cto = mp->rq;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
curi = isp->isp_reqidx;
|
|
|
|
qe = (ct_entry_t *) ISP_QUEUE_ENTRY(isp->isp_rquest, curi);
|
2000-01-03 23:50:13 +00:00
|
|
|
|
2000-01-15 01:54:36 +00:00
|
|
|
cto->ct_xfrlen = 0;
|
|
|
|
cto->ct_seg_count = 0;
|
2000-07-18 06:40:22 +00:00
|
|
|
cto->ct_header.rqs_entry_count = 1;
|
|
|
|
MEMZERO(cto->ct_dataseg, sizeof(cto->ct_dataseg));
|
|
|
|
|
2000-01-15 01:54:36 +00:00
|
|
|
if (nseg == 0) {
|
2000-07-18 06:40:22 +00:00
|
|
|
cto->ct_header.rqs_seqno = 1;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
isp_prt(isp, ISP_LOGTDEBUG1,
|
2001-04-04 21:53:59 +00:00
|
|
|
"CTIO[%x] lun%d iid%d tag %x flgs %x sts %x ssts %x res %d",
|
2001-03-02 06:28:55 +00:00
|
|
|
cto->ct_fwhandle, csio->ccb_h.target_lun, cto->ct_iid,
|
2001-04-04 21:53:59 +00:00
|
|
|
cto->ct_tag_val, cto->ct_flags, cto->ct_status,
|
|
|
|
cto->ct_scsi_status, cto->ct_resid);
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
ISP_TDQE(isp, "tdma_mk[no data]", curi, cto);
|
|
|
|
isp_put_ctio(isp, cto, qe);
|
2000-01-15 01:54:36 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2000-07-18 06:40:22 +00:00
|
|
|
nctios = nseg / ISP_RQDSEG;
|
|
|
|
if (nseg % ISP_RQDSEG) {
|
|
|
|
nctios++;
|
|
|
|
}
|
|
|
|
|
2000-01-03 23:50:13 +00:00
|
|
|
/*
|
2001-03-02 06:28:55 +00:00
|
|
|
* Save syshandle, and potentially any SCSI status, which we'll
|
|
|
|
* reinsert on the last CTIO we're going to send.
|
2000-01-03 23:50:13 +00:00
|
|
|
*/
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
|
2001-03-02 06:28:55 +00:00
|
|
|
handle = cto->ct_syshandle;
|
|
|
|
cto->ct_syshandle = 0;
|
2000-07-18 06:40:22 +00:00
|
|
|
cto->ct_header.rqs_seqno = 0;
|
|
|
|
send_status = (cto->ct_flags & CT_SENDSTATUS) != 0;
|
2000-01-03 23:50:13 +00:00
|
|
|
|
2000-07-18 06:40:22 +00:00
|
|
|
if (send_status) {
|
|
|
|
sflags = cto->ct_flags & (CT_SENDSTATUS | CT_CCINCR);
|
|
|
|
cto->ct_flags &= ~(CT_SENDSTATUS | CT_CCINCR);
|
|
|
|
/*
|
|
|
|
* Preserve residual.
|
|
|
|
*/
|
|
|
|
resid = cto->ct_resid;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Save actual SCSI status.
|
|
|
|
*/
|
|
|
|
scsi_status = cto->ct_scsi_status;
|
|
|
|
|
2001-04-04 21:53:59 +00:00
|
|
|
#ifndef STATUS_WITH_DATA
|
|
|
|
sflags |= CT_NO_DATA;
|
2000-07-18 06:40:22 +00:00
|
|
|
/*
|
|
|
|
* We can't do a status at the same time as a data CTIO, so
|
|
|
|
* we need to synthesize an extra CTIO at this level.
|
|
|
|
*/
|
2000-01-15 01:54:36 +00:00
|
|
|
nctios++;
|
2001-03-21 00:49:37 +00:00
|
|
|
#endif
|
2000-07-18 06:40:22 +00:00
|
|
|
} else {
|
|
|
|
sflags = scsi_status = resid = 0;
|
2000-01-15 01:54:36 +00:00
|
|
|
}
|
|
|
|
|
2001-09-04 21:45:57 +00:00
|
|
|
cto->ct_resid = 0;
|
2000-07-18 06:40:22 +00:00
|
|
|
cto->ct_scsi_status = 0;
|
|
|
|
|
2002-04-02 23:36:14 +00:00
|
|
|
pcs = (struct isp_pcisoftc *)isp;
|
|
|
|
dp = &pcs->dmaps[isp_handle_index(handle)];
|
2000-01-03 23:50:13 +00:00
|
|
|
if ((csio->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_IN) {
|
2002-04-02 23:36:14 +00:00
|
|
|
bus_dmamap_sync(pcs->dmat, *dp, BUS_DMASYNC_PREREAD);
|
2000-01-03 23:50:13 +00:00
|
|
|
} else {
|
2002-04-02 23:36:14 +00:00
|
|
|
bus_dmamap_sync(pcs->dmat, *dp, BUS_DMASYNC_PREWRITE);
|
2000-01-03 23:50:13 +00:00
|
|
|
}
|
|
|
|
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
nxti = *mp->nxtip;
|
2000-01-03 23:50:13 +00:00
|
|
|
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
for (nth_ctio = 0; nth_ctio < nctios; nth_ctio++) {
|
2000-07-18 06:40:22 +00:00
|
|
|
int seglim;
|
2000-01-03 23:50:13 +00:00
|
|
|
|
|
|
|
seglim = nseg;
|
2000-07-18 06:40:22 +00:00
|
|
|
if (seglim) {
|
|
|
|
int seg;
|
|
|
|
|
|
|
|
if (seglim > ISP_RQDSEG)
|
|
|
|
seglim = ISP_RQDSEG;
|
|
|
|
|
|
|
|
for (seg = 0; seg < seglim; seg++, nseg--) {
|
|
|
|
/*
|
|
|
|
* Unlike normal initiator commands, we don't
|
|
|
|
* do any swizzling here.
|
|
|
|
*/
|
|
|
|
cto->ct_dataseg[seg].ds_count = dm_segs->ds_len;
|
|
|
|
cto->ct_dataseg[seg].ds_base = dm_segs->ds_addr;
|
|
|
|
cto->ct_xfrlen += dm_segs->ds_len;
|
|
|
|
dm_segs++;
|
|
|
|
}
|
|
|
|
cto->ct_seg_count = seg;
|
2000-01-03 23:50:13 +00:00
|
|
|
} else {
|
2000-07-18 06:40:22 +00:00
|
|
|
/*
|
|
|
|
* This case should only happen when we're sending an
|
|
|
|
* extra CTIO with final status.
|
|
|
|
*/
|
|
|
|
if (send_status == 0) {
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
isp_prt(isp, ISP_LOGWARN,
|
2000-12-05 07:38:41 +00:00
|
|
|
"tdma_mk ran out of segments");
|
2000-07-18 06:40:22 +00:00
|
|
|
mp->error = EINVAL;
|
|
|
|
return;
|
|
|
|
}
|
2000-01-03 23:50:13 +00:00
|
|
|
}
|
|
|
|
|
2000-07-18 06:40:22 +00:00
|
|
|
/*
|
|
|
|
* At this point, the fields ct_lun, ct_iid, ct_tagval,
|
|
|
|
* ct_tagtype, and ct_timeout have been carried over
|
|
|
|
* unchanged from what our caller had set.
|
|
|
|
*
|
|
|
|
* The dataseg fields and the seg_count fields we just got
|
|
|
|
* through setting. The data direction we've preserved all
|
|
|
|
* along and only clear it if we're now sending status.
|
|
|
|
*/
|
|
|
|
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
if (nth_ctio == nctios - 1) {
|
2000-01-03 23:50:13 +00:00
|
|
|
/*
|
2000-07-18 06:40:22 +00:00
|
|
|
* We're the last in a sequence of CTIOs, so mark
|
|
|
|
* this CTIO and save the handle to the CCB such that
|
|
|
|
* when this CTIO completes we can free dma resources
|
|
|
|
* and do whatever else we need to do to finish the
|
2001-04-04 21:53:59 +00:00
|
|
|
* rest of the command. We *don't* give this to the
|
|
|
|
* firmware to work on- the caller will do that.
|
2000-01-03 23:50:13 +00:00
|
|
|
*/
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
|
2001-03-02 06:28:55 +00:00
|
|
|
cto->ct_syshandle = handle;
|
2000-07-18 06:40:22 +00:00
|
|
|
cto->ct_header.rqs_seqno = 1;
|
|
|
|
|
|
|
|
if (send_status) {
|
|
|
|
cto->ct_scsi_status = scsi_status;
|
2001-03-21 00:49:37 +00:00
|
|
|
cto->ct_flags |= sflags;
|
2000-07-18 06:40:22 +00:00
|
|
|
cto->ct_resid = resid;
|
|
|
|
}
|
2000-08-01 05:16:49 +00:00
|
|
|
if (send_status) {
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
isp_prt(isp, ISP_LOGTDEBUG1,
|
2001-04-04 21:53:59 +00:00
|
|
|
"CTIO[%x] lun%d iid %d tag %x ct_flags %x "
|
2001-03-02 06:28:55 +00:00
|
|
|
"scsi status %x resid %d",
|
|
|
|
cto->ct_fwhandle, csio->ccb_h.target_lun,
|
2001-04-04 21:53:59 +00:00
|
|
|
cto->ct_iid, cto->ct_tag_val, cto->ct_flags,
|
2000-01-04 03:41:40 +00:00
|
|
|
cto->ct_scsi_status, cto->ct_resid);
|
2000-08-01 05:16:49 +00:00
|
|
|
} else {
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
isp_prt(isp, ISP_LOGTDEBUG1,
|
2001-04-04 21:53:59 +00:00
|
|
|
"CTIO[%x] lun%d iid%d tag %x ct_flags 0x%x",
|
2001-03-02 06:28:55 +00:00
|
|
|
cto->ct_fwhandle, csio->ccb_h.target_lun,
|
2001-04-04 21:53:59 +00:00
|
|
|
cto->ct_iid, cto->ct_tag_val,
|
|
|
|
cto->ct_flags);
|
2000-01-04 03:41:40 +00:00
|
|
|
}
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
isp_put_ctio(isp, cto, qe);
|
|
|
|
ISP_TDQE(isp, "last tdma_mk", curi, cto);
|
|
|
|
if (nctios > 1) {
|
|
|
|
MEMORYBARRIER(isp, SYNC_REQUEST,
|
|
|
|
curi, QENTRY_LEN);
|
|
|
|
}
|
2000-01-03 23:50:13 +00:00
|
|
|
} else {
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
ct_entry_t *oqe = qe;
|
2000-07-18 06:40:22 +00:00
|
|
|
|
|
|
|
/*
|
2001-03-02 06:28:55 +00:00
|
|
|
* Make sure syshandle fields are clean
|
2000-07-18 06:40:22 +00:00
|
|
|
*/
|
2001-03-02 06:28:55 +00:00
|
|
|
cto->ct_syshandle = 0;
|
2000-01-03 23:50:13 +00:00
|
|
|
cto->ct_header.rqs_seqno = 0;
|
2000-07-18 06:40:22 +00:00
|
|
|
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
isp_prt(isp, ISP_LOGTDEBUG1,
|
2001-03-02 06:28:55 +00:00
|
|
|
"CTIO[%x] lun%d for ID%d ct_flags 0x%x",
|
|
|
|
cto->ct_fwhandle, csio->ccb_h.target_lun,
|
|
|
|
cto->ct_iid, cto->ct_flags);
|
2000-07-18 06:40:22 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Get a new CTIO
|
|
|
|
*/
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
qe = (ct_entry_t *)
|
|
|
|
ISP_QUEUE_ENTRY(isp->isp_rquest, nxti);
|
|
|
|
nxti = ISP_NXT_QENTRY(nxti, RQUEST_QUEUE_LEN(isp));
|
|
|
|
if (nxti == mp->optr) {
|
|
|
|
isp_prt(isp, ISP_LOGTDEBUG0,
|
2000-12-05 07:38:41 +00:00
|
|
|
"Queue Overflow in tdma_mk");
|
2000-01-03 23:50:13 +00:00
|
|
|
mp->error = MUSHERR_NOQENTRIES;
|
|
|
|
return;
|
|
|
|
}
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Now that we're done with the old CTIO,
|
|
|
|
* flush it out to the request queue.
|
|
|
|
*/
|
|
|
|
ISP_TDQE(isp, "dma_tgt_fc", curi, cto);
|
|
|
|
isp_put_ctio(isp, cto, oqe);
|
|
|
|
if (nth_ctio != 0) {
|
|
|
|
MEMORYBARRIER(isp, SYNC_REQUEST, curi,
|
|
|
|
QENTRY_LEN);
|
|
|
|
}
|
|
|
|
curi = ISP_NXT_QENTRY(curi, RQUEST_QUEUE_LEN(isp));
|
|
|
|
|
2000-01-03 23:50:13 +00:00
|
|
|
/*
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
* Reset some fields in the CTIO so we can reuse
|
|
|
|
* for the next one we'll flush to the request
|
|
|
|
* queue.
|
2000-01-03 23:50:13 +00:00
|
|
|
*/
|
|
|
|
cto->ct_header.rqs_entry_type = RQSTYPE_CTIO;
|
|
|
|
cto->ct_header.rqs_entry_count = 1;
|
|
|
|
cto->ct_header.rqs_flags = 0;
|
|
|
|
cto->ct_status = 0;
|
|
|
|
cto->ct_scsi_status = 0;
|
|
|
|
cto->ct_xfrlen = 0;
|
|
|
|
cto->ct_resid = 0;
|
|
|
|
cto->ct_seg_count = 0;
|
2000-07-18 06:40:22 +00:00
|
|
|
MEMZERO(cto->ct_dataseg, sizeof(cto->ct_dataseg));
|
2000-01-03 23:50:13 +00:00
|
|
|
}
|
|
|
|
}
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
*mp->nxtip = nxti;
|
2000-01-03 23:50:13 +00:00
|
|
|
}
|
|
|
|
|
2002-06-16 04:58:00 +00:00
|
|
|
/*
|
|
|
|
* We don't have to do multiple CTIOs here. Instead, we can just do
|
|
|
|
* continuation segments as needed. This greatly simplifies the code
|
|
|
|
* improves performance.
|
|
|
|
*/
|
|
|
|
|
2000-01-03 23:50:13 +00:00
|
|
|
static void
|
2000-07-18 06:40:22 +00:00
|
|
|
tdma_mkfc(void *arg, bus_dma_segment_t *dm_segs, int nseg, int error)
|
2000-01-03 23:50:13 +00:00
|
|
|
{
|
|
|
|
mush_t *mp;
|
|
|
|
struct ccb_scsiio *csio;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
struct ispsoftc *isp;
|
|
|
|
ct2_entry_t *cto, *qe;
|
|
|
|
u_int16_t curi, nxti;
|
2002-06-16 04:58:00 +00:00
|
|
|
int segcnt;
|
2000-01-03 23:50:13 +00:00
|
|
|
|
|
|
|
mp = (mush_t *) arg;
|
|
|
|
if (error) {
|
|
|
|
mp->error = error;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
isp = mp->isp;
|
2000-01-15 01:54:36 +00:00
|
|
|
csio = mp->cmd_token;
|
|
|
|
cto = mp->rq;
|
2002-06-16 04:58:00 +00:00
|
|
|
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
curi = isp->isp_reqidx;
|
|
|
|
qe = (ct2_entry_t *) ISP_QUEUE_ENTRY(isp->isp_rquest, curi);
|
2000-01-15 01:54:36 +00:00
|
|
|
|
|
|
|
if (nseg == 0) {
|
|
|
|
if ((cto->ct_flags & CT2_FLAG_MMASK) != CT2_FLAG_MODE1) {
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
isp_prt(isp, ISP_LOGWARN,
|
2000-12-05 07:38:41 +00:00
|
|
|
"dma2_tgt_fc, a status CTIO2 without MODE1 "
|
|
|
|
"set (0x%x)", cto->ct_flags);
|
2000-01-15 01:54:36 +00:00
|
|
|
mp->error = EINVAL;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* We preserve ct_lun, ct_iid, ct_rxid. We set the data
|
|
|
|
* flags to NO DATA and clear relative offset flags.
|
|
|
|
* We preserve the ct_resid and the response area.
|
|
|
|
*/
|
2002-06-16 04:58:00 +00:00
|
|
|
cto->ct_header.rqs_seqno = 1;
|
2000-01-15 01:54:36 +00:00
|
|
|
cto->ct_seg_count = 0;
|
|
|
|
cto->ct_reloff = 0;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
isp_prt(isp, ISP_LOGTDEBUG1,
|
2001-03-02 06:28:55 +00:00
|
|
|
"CTIO2[%x] lun %d->iid%d flgs 0x%x sts 0x%x ssts "
|
2000-08-01 05:16:49 +00:00
|
|
|
"0x%x res %d", cto->ct_rxid, csio->ccb_h.target_lun,
|
|
|
|
cto->ct_iid, cto->ct_flags, cto->ct_status,
|
|
|
|
cto->rsp.m1.ct_scsi_status, cto->ct_resid);
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
isp_put_ctio2(isp, cto, qe);
|
|
|
|
ISP_TDQE(isp, "dma2_tgt_fc[no data]", curi, qe);
|
2000-01-03 23:50:13 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2000-01-15 01:54:36 +00:00
|
|
|
if ((cto->ct_flags & CT2_FLAG_MMASK) != CT2_FLAG_MODE0) {
|
2002-06-16 04:58:00 +00:00
|
|
|
isp_prt(isp, ISP_LOGERR,
|
2000-12-05 07:38:41 +00:00
|
|
|
"dma2_tgt_fc, a data CTIO2 without MODE0 set "
|
|
|
|
"(0x%x)", cto->ct_flags);
|
2000-01-15 01:54:36 +00:00
|
|
|
mp->error = EINVAL;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2002-06-16 04:58:00 +00:00
|
|
|
nxti = *mp->nxtip;
|
2000-01-15 01:54:36 +00:00
|
|
|
|
2000-01-03 23:50:13 +00:00
|
|
|
/*
|
2002-06-16 04:58:00 +00:00
|
|
|
* Set up the CTIO2 data segments.
|
2000-01-03 23:50:13 +00:00
|
|
|
*/
|
2002-06-16 04:58:00 +00:00
|
|
|
for (segcnt = 0; cto->ct_seg_count < ISP_RQDSEG_T2 && segcnt < nseg;
|
|
|
|
cto->ct_seg_count++, segcnt++) {
|
|
|
|
cto->rsp.m0.ct_dataseg[cto->ct_seg_count].ds_base =
|
|
|
|
dm_segs[segcnt].ds_addr;
|
|
|
|
cto->rsp.m0.ct_dataseg[cto->ct_seg_count].ds_count =
|
|
|
|
dm_segs[segcnt].ds_len;
|
|
|
|
cto->rsp.m0.ct_xfrlen += dm_segs[segcnt].ds_len;
|
|
|
|
isp_prt(isp, ISP_LOGTDEBUG1, "isp_send_ctio2: ent0[%d]0x%x:%d",
|
|
|
|
cto->ct_seg_count, dm_segs[segcnt].ds_addr,
|
|
|
|
dm_segs[segcnt].ds_len);
|
|
|
|
}
|
|
|
|
|
|
|
|
while (segcnt < nseg) {
|
|
|
|
u_int16_t curip;
|
|
|
|
int seg;
|
|
|
|
ispcontreq_t local, *crq = &local, *qep;
|
|
|
|
|
|
|
|
qep = (ispcontreq_t *) ISP_QUEUE_ENTRY(isp->isp_rquest, nxti);
|
|
|
|
curip = nxti;
|
|
|
|
nxti = ISP_NXT_QENTRY(curip, RQUEST_QUEUE_LEN(isp));
|
|
|
|
if (nxti == mp->optr) {
|
|
|
|
ISP_UNLOCK(isp);
|
|
|
|
isp_prt(isp, ISP_LOGTDEBUG0,
|
|
|
|
"tdma_mkfc: request queue overflow");
|
|
|
|
mp->error = MUSHERR_NOQENTRIES;
|
|
|
|
return;
|
2000-01-03 23:50:13 +00:00
|
|
|
}
|
2002-06-16 04:58:00 +00:00
|
|
|
cto->ct_header.rqs_entry_count++;
|
|
|
|
MEMZERO((void *)crq, sizeof (*crq));
|
|
|
|
crq->req_header.rqs_entry_count = 1;
|
|
|
|
crq->req_header.rqs_entry_type = RQSTYPE_DATASEG;
|
|
|
|
for (seg = 0; segcnt < nseg && seg < ISP_CDSEG;
|
|
|
|
segcnt++, seg++) {
|
|
|
|
crq->req_dataseg[seg].ds_base = dm_segs[segcnt].ds_addr;
|
|
|
|
crq->req_dataseg[seg].ds_count = dm_segs[segcnt].ds_len;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
isp_prt(isp, ISP_LOGTDEBUG1,
|
2002-06-16 04:58:00 +00:00
|
|
|
"isp_send_ctio2: ent%d[%d]%x:%u",
|
|
|
|
cto->ct_header.rqs_entry_count-1, seg,
|
|
|
|
dm_segs[segcnt].ds_addr, dm_segs[segcnt].ds_len);
|
|
|
|
cto->rsp.m0.ct_xfrlen += dm_segs[segcnt].ds_len;
|
|
|
|
cto->ct_seg_count++;
|
2000-01-03 23:50:13 +00:00
|
|
|
}
|
2002-06-16 04:58:00 +00:00
|
|
|
MEMORYBARRIER(isp, SYNC_REQUEST, curip, QENTRY_LEN);
|
|
|
|
isp_put_cont_req(isp, crq, qep);
|
|
|
|
ISP_TDQE(isp, "cont entry", curi, qep);
|
2000-01-03 23:50:13 +00:00
|
|
|
}
|
2002-06-16 04:58:00 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* No do final twiddling for the CTIO itself.
|
|
|
|
*/
|
|
|
|
cto->ct_header.rqs_seqno = 1;
|
|
|
|
isp_prt(isp, ISP_LOGTDEBUG1,
|
|
|
|
"CTIO2[%x] lun %d->iid%d flgs 0x%x sts 0x%x ssts 0x%x resid %d",
|
|
|
|
cto->ct_rxid, csio->ccb_h.target_lun, (int) cto->ct_iid,
|
|
|
|
cto->ct_flags, cto->ct_status, cto->rsp.m1.ct_scsi_status,
|
|
|
|
cto->ct_resid);
|
|
|
|
isp_put_ctio2(isp, cto, qe);
|
|
|
|
ISP_TDQE(isp, "last dma2_tgt_fc", curi, qe);
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
*mp->nxtip = nxti;
|
2000-01-03 23:50:13 +00:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2001-08-31 21:39:04 +00:00
|
|
|
static void dma2(void *, bus_dma_segment_t *, int, int);
|
2000-01-03 23:50:13 +00:00
|
|
|
|
1998-09-15 10:06:23 +00:00
|
|
|
static void
|
1998-12-28 19:24:23 +00:00
|
|
|
dma2(void *arg, bus_dma_segment_t *dm_segs, int nseg, int error)
|
1998-09-15 10:06:23 +00:00
|
|
|
{
|
|
|
|
mush_t *mp;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
struct ispsoftc *isp;
|
2000-01-03 23:50:13 +00:00
|
|
|
struct ccb_scsiio *csio;
|
2002-04-02 23:36:14 +00:00
|
|
|
struct isp_pcisoftc *pcs;
|
1998-09-15 10:06:23 +00:00
|
|
|
bus_dmamap_t *dp;
|
|
|
|
bus_dma_segment_t *eseg;
|
|
|
|
ispreq_t *rq;
|
2000-01-03 23:50:13 +00:00
|
|
|
int seglim, datalen;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
u_int16_t nxti;
|
1998-09-15 10:06:23 +00:00
|
|
|
|
|
|
|
mp = (mush_t *) arg;
|
|
|
|
if (error) {
|
|
|
|
mp->error = error;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (nseg < 1) {
|
2000-12-05 07:38:41 +00:00
|
|
|
isp_prt(mp->isp, ISP_LOGERR, "bad segment count (%d)", nseg);
|
1998-09-15 10:06:23 +00:00
|
|
|
mp->error = EFAULT;
|
|
|
|
return;
|
|
|
|
}
|
2000-01-03 23:50:13 +00:00
|
|
|
csio = mp->cmd_token;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
isp = mp->isp;
|
1998-09-15 10:06:23 +00:00
|
|
|
rq = mp->rq;
|
2002-04-02 23:36:14 +00:00
|
|
|
pcs = (struct isp_pcisoftc *)mp->isp;
|
|
|
|
dp = &pcs->dmaps[isp_handle_index(rq->req_handle)];
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
nxti = *mp->nxtip;
|
1999-10-17 19:03:11 +00:00
|
|
|
|
2000-01-03 23:50:13 +00:00
|
|
|
if ((csio->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_IN) {
|
2002-04-02 23:36:14 +00:00
|
|
|
bus_dmamap_sync(pcs->dmat, *dp, BUS_DMASYNC_PREREAD);
|
1998-09-15 10:06:23 +00:00
|
|
|
} else {
|
2002-04-02 23:36:14 +00:00
|
|
|
bus_dmamap_sync(pcs->dmat, *dp, BUS_DMASYNC_PREWRITE);
|
1998-09-15 10:06:23 +00:00
|
|
|
}
|
|
|
|
|
2000-01-03 23:50:13 +00:00
|
|
|
datalen = XS_XFRLEN(csio);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We're passed an initial partially filled in entry that
|
|
|
|
* has most fields filled in except for data transfer
|
|
|
|
* related values.
|
|
|
|
*
|
|
|
|
* Our job is to fill in the initial request queue entry and
|
|
|
|
* then to start allocating and filling in continuation entries
|
|
|
|
* until we've covered the entire transfer.
|
|
|
|
*/
|
|
|
|
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
if (IS_FC(isp)) {
|
1998-09-15 10:06:23 +00:00
|
|
|
seglim = ISP_RQDSEG_T2;
|
|
|
|
((ispreqt2_t *)rq)->req_totalcnt = datalen;
|
2000-01-03 23:50:13 +00:00
|
|
|
if ((csio->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_IN) {
|
|
|
|
((ispreqt2_t *)rq)->req_flags |= REQFLAG_DATA_IN;
|
|
|
|
} else {
|
|
|
|
((ispreqt2_t *)rq)->req_flags |= REQFLAG_DATA_OUT;
|
|
|
|
}
|
1998-09-15 10:06:23 +00:00
|
|
|
} else {
|
2000-02-15 00:50:01 +00:00
|
|
|
if (csio->cdb_len > 12) {
|
|
|
|
seglim = 0;
|
|
|
|
} else {
|
|
|
|
seglim = ISP_RQDSEG;
|
|
|
|
}
|
2000-01-03 23:50:13 +00:00
|
|
|
if ((csio->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_IN) {
|
|
|
|
rq->req_flags |= REQFLAG_DATA_IN;
|
|
|
|
} else {
|
|
|
|
rq->req_flags |= REQFLAG_DATA_OUT;
|
|
|
|
}
|
1998-09-15 10:06:23 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
eseg = dm_segs + nseg;
|
|
|
|
|
|
|
|
while (datalen != 0 && rq->req_seg_count < seglim && dm_segs != eseg) {
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
if (IS_FC(isp)) {
|
1998-09-15 10:06:23 +00:00
|
|
|
ispreqt2_t *rq2 = (ispreqt2_t *)rq;
|
|
|
|
rq2->req_dataseg[rq2->req_seg_count].ds_base =
|
|
|
|
dm_segs->ds_addr;
|
|
|
|
rq2->req_dataseg[rq2->req_seg_count].ds_count =
|
|
|
|
dm_segs->ds_len;
|
|
|
|
} else {
|
|
|
|
rq->req_dataseg[rq->req_seg_count].ds_base =
|
|
|
|
dm_segs->ds_addr;
|
|
|
|
rq->req_dataseg[rq->req_seg_count].ds_count =
|
|
|
|
dm_segs->ds_len;
|
|
|
|
}
|
|
|
|
datalen -= dm_segs->ds_len;
|
|
|
|
rq->req_seg_count++;
|
|
|
|
dm_segs++;
|
|
|
|
}
|
|
|
|
|
|
|
|
while (datalen > 0 && dm_segs != eseg) {
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
u_int16_t onxti;
|
|
|
|
ispcontreq_t local, *crq = &local, *cqe;
|
|
|
|
|
|
|
|
cqe = (ispcontreq_t *) ISP_QUEUE_ENTRY(isp->isp_rquest, nxti);
|
|
|
|
onxti = nxti;
|
|
|
|
nxti = ISP_NXT_QENTRY(onxti, RQUEST_QUEUE_LEN(isp));
|
|
|
|
if (nxti == mp->optr) {
|
|
|
|
isp_prt(isp, ISP_LOGDEBUG0, "Request Queue Overflow++");
|
1998-09-17 21:11:21 +00:00
|
|
|
mp->error = MUSHERR_NOQENTRIES;
|
1998-09-15 10:06:23 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
rq->req_header.rqs_entry_count++;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
MEMZERO((void *)crq, sizeof (*crq));
|
1998-09-15 10:06:23 +00:00
|
|
|
crq->req_header.rqs_entry_count = 1;
|
|
|
|
crq->req_header.rqs_entry_type = RQSTYPE_DATASEG;
|
|
|
|
|
|
|
|
seglim = 0;
|
|
|
|
while (datalen > 0 && seglim < ISP_CDSEG && dm_segs != eseg) {
|
|
|
|
crq->req_dataseg[seglim].ds_base =
|
|
|
|
dm_segs->ds_addr;
|
|
|
|
crq->req_dataseg[seglim].ds_count =
|
|
|
|
dm_segs->ds_len;
|
|
|
|
rq->req_seg_count++;
|
|
|
|
dm_segs++;
|
|
|
|
seglim++;
|
|
|
|
datalen -= dm_segs->ds_len;
|
|
|
|
}
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
isp_put_cont_req(isp, crq, cqe);
|
|
|
|
MEMORYBARRIER(isp, SYNC_REQUEST, onxti, QENTRY_LEN);
|
1998-09-15 10:06:23 +00:00
|
|
|
}
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
*mp->nxtip = nxti;
|
1998-09-15 10:06:23 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
2000-01-03 23:50:13 +00:00
|
|
|
isp_pci_dmasetup(struct ispsoftc *isp, struct ccb_scsiio *csio, ispreq_t *rq,
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
u_int16_t *nxtip, u_int16_t optr)
|
1998-09-15 10:06:23 +00:00
|
|
|
{
|
2002-04-02 23:36:14 +00:00
|
|
|
struct isp_pcisoftc *pcs = (struct isp_pcisoftc *)isp;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
ispreq_t *qep;
|
1999-08-16 19:52:29 +00:00
|
|
|
bus_dmamap_t *dp = NULL;
|
1998-09-15 10:06:23 +00:00
|
|
|
mush_t mush, *mp;
|
2001-08-31 21:39:04 +00:00
|
|
|
void (*eptr)(void *, bus_dma_segment_t *, int, int);
|
1998-09-15 10:06:23 +00:00
|
|
|
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
qep = (ispreq_t *) ISP_QUEUE_ENTRY(isp->isp_rquest, isp->isp_reqidx);
|
2000-01-03 23:50:13 +00:00
|
|
|
#ifdef ISP_TARGET_MODE
|
|
|
|
if (csio->ccb_h.func_code == XPT_CONT_TARGET_IO) {
|
|
|
|
if (IS_FC(isp)) {
|
2000-07-18 06:40:22 +00:00
|
|
|
eptr = tdma_mkfc;
|
2000-01-03 23:50:13 +00:00
|
|
|
} else {
|
2000-07-18 06:40:22 +00:00
|
|
|
eptr = tdma_mk;
|
2000-01-03 23:50:13 +00:00
|
|
|
}
|
2000-07-18 06:40:22 +00:00
|
|
|
if ((csio->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_NONE ||
|
|
|
|
(csio->dxfer_len == 0)) {
|
2000-01-15 01:54:36 +00:00
|
|
|
mp = &mush;
|
|
|
|
mp->isp = isp;
|
|
|
|
mp->cmd_token = csio;
|
2001-04-04 21:53:59 +00:00
|
|
|
mp->rq = rq; /* really a ct_entry_t or ct2_entry_t */
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
mp->nxtip = nxtip;
|
2000-01-15 01:54:36 +00:00
|
|
|
mp->optr = optr;
|
|
|
|
mp->error = 0;
|
|
|
|
(*eptr)(mp, NULL, 0, 0);
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
goto mbxsync;
|
2000-01-15 01:54:36 +00:00
|
|
|
}
|
2000-01-03 23:50:13 +00:00
|
|
|
} else
|
|
|
|
#endif
|
|
|
|
eptr = dma2;
|
1998-09-15 10:06:23 +00:00
|
|
|
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
|
2000-07-18 06:40:22 +00:00
|
|
|
if ((csio->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_NONE ||
|
|
|
|
(csio->dxfer_len == 0)) {
|
2000-01-15 01:54:36 +00:00
|
|
|
rq->req_seg_count = 1;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
goto mbxsync;
|
2000-01-15 01:54:36 +00:00
|
|
|
}
|
1998-09-15 10:06:23 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Do a virtual grapevine step to collect info for
|
1998-09-17 21:11:21 +00:00
|
|
|
* the callback dma allocation that we have to use...
|
1998-09-15 10:06:23 +00:00
|
|
|
*/
|
|
|
|
mp = &mush;
|
|
|
|
mp->isp = isp;
|
2000-01-03 23:50:13 +00:00
|
|
|
mp->cmd_token = csio;
|
1998-09-15 10:06:23 +00:00
|
|
|
mp->rq = rq;
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
mp->nxtip = nxtip;
|
1998-09-15 10:06:23 +00:00
|
|
|
mp->optr = optr;
|
|
|
|
mp->error = 0;
|
|
|
|
|
2000-01-03 23:50:13 +00:00
|
|
|
if ((csio->ccb_h.flags & CAM_SCATTER_VALID) == 0) {
|
|
|
|
if ((csio->ccb_h.flags & CAM_DATA_PHYS) == 0) {
|
1998-09-17 21:11:21 +00:00
|
|
|
int error, s;
|
2002-04-02 23:36:14 +00:00
|
|
|
dp = &pcs->dmaps[isp_handle_index(rq->req_handle)];
|
1998-09-15 10:06:23 +00:00
|
|
|
s = splsoftvm();
|
2002-04-02 23:36:14 +00:00
|
|
|
error = bus_dmamap_load(pcs->dmat, *dp,
|
2000-01-03 23:50:13 +00:00
|
|
|
csio->data_ptr, csio->dxfer_len, eptr, mp, 0);
|
1998-09-15 10:06:23 +00:00
|
|
|
if (error == EINPROGRESS) {
|
2002-04-02 23:36:14 +00:00
|
|
|
bus_dmamap_unload(pcs->dmat, *dp);
|
1998-09-15 10:06:23 +00:00
|
|
|
mp->error = EINVAL;
|
2000-12-05 07:38:41 +00:00
|
|
|
isp_prt(isp, ISP_LOGERR,
|
|
|
|
"deferred dma allocation not supported");
|
1998-09-15 10:06:23 +00:00
|
|
|
} else if (error && mp->error == 0) {
|
1999-08-16 19:52:29 +00:00
|
|
|
#ifdef DIAGNOSTIC
|
2001-03-01 02:21:36 +00:00
|
|
|
isp_prt(isp, ISP_LOGERR,
|
|
|
|
"error %d in dma mapping code", error);
|
1999-08-16 19:52:29 +00:00
|
|
|
#endif
|
1998-09-15 10:06:23 +00:00
|
|
|
mp->error = error;
|
|
|
|
}
|
1998-09-17 21:11:21 +00:00
|
|
|
splx(s);
|
1998-09-15 10:06:23 +00:00
|
|
|
} else {
|
|
|
|
/* Pointer to physical buffer */
|
1999-03-17 05:07:18 +00:00
|
|
|
struct bus_dma_segment seg;
|
1998-09-15 10:06:23 +00:00
|
|
|
seg.ds_addr = (bus_addr_t)csio->data_ptr;
|
|
|
|
seg.ds_len = csio->dxfer_len;
|
2000-01-03 23:50:13 +00:00
|
|
|
(*eptr)(mp, &seg, 1, 0);
|
1998-09-15 10:06:23 +00:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
struct bus_dma_segment *segs;
|
|
|
|
|
2000-01-03 23:50:13 +00:00
|
|
|
if ((csio->ccb_h.flags & CAM_DATA_PHYS) != 0) {
|
2000-12-05 07:38:41 +00:00
|
|
|
isp_prt(isp, ISP_LOGERR,
|
|
|
|
"Physical segment pointers unsupported");
|
1998-09-15 10:06:23 +00:00
|
|
|
mp->error = EINVAL;
|
2000-01-03 23:50:13 +00:00
|
|
|
} else if ((csio->ccb_h.flags & CAM_SG_LIST_PHYS) == 0) {
|
2000-12-05 07:38:41 +00:00
|
|
|
isp_prt(isp, ISP_LOGERR,
|
|
|
|
"Virtual segment addresses unsupported");
|
1998-09-15 10:06:23 +00:00
|
|
|
mp->error = EINVAL;
|
|
|
|
} else {
|
|
|
|
/* Just use the segments provided */
|
|
|
|
segs = (struct bus_dma_segment *) csio->data_ptr;
|
2000-01-03 23:50:13 +00:00
|
|
|
(*eptr)(mp, segs, csio->sglist_cnt, 0);
|
1998-09-15 10:06:23 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
if (mp->error) {
|
1998-09-17 21:11:21 +00:00
|
|
|
int retval = CMD_COMPLETE;
|
|
|
|
if (mp->error == MUSHERR_NOQENTRIES) {
|
|
|
|
retval = CMD_EAGAIN;
|
|
|
|
} else if (mp->error == EFBIG) {
|
1999-08-16 19:52:29 +00:00
|
|
|
XS_SETERR(csio, CAM_REQ_TOO_BIG);
|
1998-09-15 10:06:23 +00:00
|
|
|
} else if (mp->error == EINVAL) {
|
1999-08-16 19:52:29 +00:00
|
|
|
XS_SETERR(csio, CAM_REQ_INVALID);
|
1998-09-15 10:06:23 +00:00
|
|
|
} else {
|
1999-08-16 19:52:29 +00:00
|
|
|
XS_SETERR(csio, CAM_UNREC_HBA_ERROR);
|
1998-09-15 10:06:23 +00:00
|
|
|
}
|
1998-09-17 21:11:21 +00:00
|
|
|
return (retval);
|
1998-09-15 10:06:23 +00:00
|
|
|
}
|
Major restructuring for swizzling to the request queue and unswizzling from
the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have
a complete set of inline functions in isp_inline.h. Each platform is
responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32}
macros.
The reason this needs to be done is that we need to have a single set of
functions that will work correctly on multiple architectures for both little
and big endian machines. It also needs to work correctly in the case that
we have the request or response queues in memory that has to be treated
specially (e.g., have ddi_dma_sync called on it for Solaris after we update
it or before we read from it). It also has to handle the SBus cards (for
platforms that have them) which, while on a Big Endian machine, do *not*
require *most* of the request/response queue entry fields to be swizzled
or unswizzled.
One thing that falls out of this is that we no longer build requests in the
request queue itself. Instead, we build the request locally (e.g., on the
stack) and then as part of the swizzling operation, copy it to the request
queue entry we've allocated. I thought long and hard about whether this was
too expensive a change to make as it in a lot of cases requires an extra
copy. On balance, the flexbility is worth it. With any luck, the entry that
we build locally stays in a processor writeback cache (after all, it's only
64 bytes) so that the cost of actually flushing it to the memory area that is
the shared queue with the PCI device is not all that expensive. We may examine
this again and try to get clever in the future to try and avoid copies.
Another change that falls out of this is that MEMORYBARRIER should be taken
a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the
entry being added. But there had been many other places this had been missing.
It's now very important that it be done.
Additional changes:
Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry,
the iptr value that gets returned is the value we intend to eventually plug
into the ISP registers as the entry *one past* the last one we've written-
*not* the current entry we're updating. All along we've been calling sync
functions on the wrong index value. Argh. The 'fix' here is to rename all
'iptr' variables as 'nxti' to remember that this is the 'next' pointer-
not the current pointer.
Devote a single bit to mboxbsy- and set aside bits for output mbox registers
that we need to pick up- we can have at least one command which does not
have any defined output registers (MBOX_EXECUTE_FIRMWARE).
MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
|
|
|
mbxsync:
|
|
|
|
switch (rq->req_header.rqs_entry_type) {
|
|
|
|
case RQSTYPE_REQUEST:
|
|
|
|
isp_put_request(isp, rq, qep);
|
|
|
|
break;
|
|
|
|
case RQSTYPE_CMDONLY:
|
|
|
|
isp_put_extended_request(isp, (ispextreq_t *)rq,
|
|
|
|
(ispextreq_t *)qep);
|
|
|
|
break;
|
|
|
|
case RQSTYPE_T2RQS:
|
|
|
|
isp_put_request_t2(isp, (ispreqt2_t *) rq, (ispreqt2_t *) qep);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return (CMD_QUEUED);
|
1998-09-15 10:06:23 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2001-03-04 18:42:51 +00:00
|
|
|
isp_pci_dmateardown(struct ispsoftc *isp, XS_T *xs, u_int16_t handle)
|
1998-09-15 10:06:23 +00:00
|
|
|
{
|
2002-04-02 23:36:14 +00:00
|
|
|
struct isp_pcisoftc *pcs = (struct isp_pcisoftc *)isp;
|
|
|
|
bus_dmamap_t *dp = &pcs->dmaps[isp_handle_index(handle)];
|
1999-10-17 19:03:11 +00:00
|
|
|
if ((xs->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_IN) {
|
2002-04-02 23:36:14 +00:00
|
|
|
bus_dmamap_sync(pcs->dmat, *dp, BUS_DMASYNC_POSTREAD);
|
1998-09-15 10:06:23 +00:00
|
|
|
} else {
|
2002-04-02 23:36:14 +00:00
|
|
|
bus_dmamap_sync(pcs->dmat, *dp, BUS_DMASYNC_POSTWRITE);
|
1998-09-15 10:06:23 +00:00
|
|
|
}
|
2002-04-02 23:36:14 +00:00
|
|
|
bus_dmamap_unload(pcs->dmat, *dp);
|
1998-09-15 10:06:23 +00:00
|
|
|
}
|
|
|
|
|
1998-04-22 18:12:29 +00:00
|
|
|
|
|
|
|
static void
|
1998-12-28 19:24:23 +00:00
|
|
|
isp_pci_reset1(struct ispsoftc *isp)
|
1998-04-22 18:12:29 +00:00
|
|
|
{
|
|
|
|
/* Make sure the BIOS is disabled */
|
|
|
|
isp_pci_wr_reg(isp, HCCR, PCI_HCCR_CMD_BIOS);
|
2000-07-04 01:01:15 +00:00
|
|
|
/* and enable interrupts */
|
|
|
|
ENABLE_INTS(isp);
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2000-08-01 05:16:49 +00:00
|
|
|
isp_pci_dumpregs(struct ispsoftc *isp, const char *msg)
|
1998-04-22 18:12:29 +00:00
|
|
|
{
|
2002-04-02 23:36:14 +00:00
|
|
|
struct isp_pcisoftc *pcs = (struct isp_pcisoftc *)isp;
|
2000-08-01 05:16:49 +00:00
|
|
|
if (msg)
|
2001-03-01 02:21:36 +00:00
|
|
|
printf("%s: %s\n", device_get_nameunit(isp->isp_dev), msg);
|
|
|
|
else
|
|
|
|
printf("%s:\n", device_get_nameunit(isp->isp_dev));
|
2000-08-01 05:16:49 +00:00
|
|
|
if (IS_SCSI(isp))
|
|
|
|
printf(" biu_conf1=%x", ISP_READ(isp, BIU_CONF1));
|
|
|
|
else
|
|
|
|
printf(" biu_csr=%x", ISP_READ(isp, BIU2100_CSR));
|
|
|
|
printf(" biu_icr=%x biu_isr=%x biu_sema=%x ", ISP_READ(isp, BIU_ICR),
|
|
|
|
ISP_READ(isp, BIU_ISR), ISP_READ(isp, BIU_SEMA));
|
|
|
|
printf("risc_hccr=%x\n", ISP_READ(isp, HCCR));
|
|
|
|
|
|
|
|
|
|
|
|
if (IS_SCSI(isp)) {
|
|
|
|
ISP_WRITE(isp, HCCR, HCCR_CMD_PAUSE);
|
|
|
|
printf(" cdma_conf=%x cdma_sts=%x cdma_fifostat=%x\n",
|
|
|
|
ISP_READ(isp, CDMA_CONF), ISP_READ(isp, CDMA_STATUS),
|
|
|
|
ISP_READ(isp, CDMA_FIFO_STS));
|
|
|
|
printf(" ddma_conf=%x ddma_sts=%x ddma_fifostat=%x\n",
|
|
|
|
ISP_READ(isp, DDMA_CONF), ISP_READ(isp, DDMA_STATUS),
|
|
|
|
ISP_READ(isp, DDMA_FIFO_STS));
|
|
|
|
printf(" sxp_int=%x sxp_gross=%x sxp(scsi_ctrl)=%x\n",
|
|
|
|
ISP_READ(isp, SXP_INTERRUPT),
|
|
|
|
ISP_READ(isp, SXP_GROSS_ERR),
|
|
|
|
ISP_READ(isp, SXP_PINS_CTRL));
|
|
|
|
ISP_WRITE(isp, HCCR, HCCR_CMD_RELEASE);
|
|
|
|
}
|
|
|
|
printf(" mbox regs: %x %x %x %x %x\n",
|
|
|
|
ISP_READ(isp, OUTMAILBOX0), ISP_READ(isp, OUTMAILBOX1),
|
|
|
|
ISP_READ(isp, OUTMAILBOX2), ISP_READ(isp, OUTMAILBOX3),
|
|
|
|
ISP_READ(isp, OUTMAILBOX4));
|
|
|
|
printf(" PCI Status Command/Status=%x\n",
|
2002-04-02 23:36:14 +00:00
|
|
|
pci_read_config(pcs->pci_dev, PCIR_COMMAND, 1));
|
1998-04-22 18:12:29 +00:00
|
|
|
}
|