NetBSD/sys/arch/vax/uba/uda.c

2338 lines
62 KiB
C
Raw Normal View History

1994-08-03 00:18:46 +04:00
/*
* Copyright (c) 1988 Regents of the University of California.
* All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* Chris Torek.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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.
*
* from: @(#)uda.c 7.32 (Berkeley) 2/13/91
* $Id: uda.c,v 1.1 1994/08/02 20:21:30 ragge Exp $
*/
/*
* UDA50/MSCP device driver
*/
#define POLLSTATS
/*
* TODO
* write bad block forwarding code
*/
#include "uda.h"
#include "ra.h"
#if NUDA > 0
/*
* CONFIGURATION OPTIONS. The next three defines are tunable -- tune away!
*
* COMPAT_42 enables 4.2/4.3 compatibility (label mapping)
*
* NRSPL2 and NCMDL2 control the number of response and command
* packets respectively. They may be any value from 0 to 7, though
* setting them higher than 5 is unlikely to be of any value.
* If you get warnings about your command ring being too small,
* try increasing the values by one.
*
* MAXUNIT controls the maximum unit number (number of drives per
* controller) we are prepared to handle.
*
* DEFAULT_BURST must be at least 1.
*/
#define COMPAT_42
#define NRSPL2 5 /* log2 number of response packets */
#define NCMDL2 5 /* log2 number of command packets */
#define MAXUNIT 8 /* maximum allowed unit number */
#define DEFAULT_BURST 4 /* default DMA burst size */
#include "sys/param.h"
#include "sys/systm.h"
#include "sys/buf.h"
#include "sys/conf.h"
#include "sys/file.h"
#include "sys/ioctl.h"
#include "sys/proc.h"
#include "sys/user.h"
#include "sys/map.h"
#include "sys/device.h"
/* #include "sys/vm.h" */
#include "sys/dkstat.h"
/* #include "sys/cmap.h" */
#include "sys/disklabel.h"
#include "sys/syslog.h"
#include "sys/stat.h"
#include "vax/include/pte.h"
#include "vax/include/sid.h"
#include "vax/include/cpu.h"
#include "vax/uba/ubareg.h"
#include "vax/uba/ubavar.h"
#define NRSP (1 << NRSPL2)
#define NCMD (1 << NCMDL2)
#include "vax/uba/udareg.h"
#include "vax/vax/mscp.h"
#include "vax/vax/mscpvar.h"
#include "vax/include/mtpr.h"
extern int cold;
struct cfdriver udacd;
struct cfdriver racd;
/*
* UDA communications area and MSCP packet pools, per controller.
*/
struct uda {
struct udaca uda_ca; /* communications area */
struct mscp uda_rsp[NRSP]; /* response packets */
struct mscp uda_cmd[NCMD]; /* command packets */
} uda[NUDA];
/*
* Software status, per controller.
*/
struct uda_softc {
struct uda *sc_uda; /* Unibus address of uda struct */
short sc_state; /* UDA50 state; see below */
short sc_flags; /* flags; see below */
int sc_micro; /* microcode revision */
int sc_ivec; /* interrupt vector address */
short sc_ipl; /* interrupt priority, Q-bus */
struct mscp_info sc_mi;/* MSCP info (per mscpvar.h) */
#ifndef POLLSTATS
int sc_wticks; /* watchdog timer ticks */
#else
short sc_wticks;
short sc_ncmd;
#endif
} uda_softc[NUDA];
#ifdef POLLSTATS
struct udastats {
int ncmd;
int cmd[NCMD + 1];
} udastats = { NCMD + 1 };
#endif
/*
* Controller states
*/
#define ST_IDLE 0 /* uninitialised */
#define ST_STEP1 1 /* in `STEP 1' */
#define ST_STEP2 2 /* in `STEP 2' */
#define ST_STEP3 3 /* in `STEP 3' */
#define ST_SETCHAR 4 /* in `Set Controller Characteristics' */
#define ST_RUN 5 /* up and running */
/*
* Flags
*/
#define SC_MAPPED 0x01 /* mapped in Unibus I/O space */
#define SC_INSTART 0x02 /* inside udastart() */
#define SC_GRIPED 0x04 /* griped about cmd ring too small */
#define SC_INSLAVE 0x08 /* inside udaslave() */
#define SC_DOWAKE 0x10 /* wakeup when ctlr init done */
#define SC_STARTPOLL 0x20 /* need to initiate polling */
/*
* Device to unit number and partition and back
*/
#define UNITSHIFT 3
#define UNITMASK 7
#define udaunit(dev) (minor(dev) >> UNITSHIFT)
#define udapart(dev) (minor(dev) & UNITMASK)
#define udaminor(u, p) (((u) << UNITSHIFT) | (p))
/*
* Drive status, per drive
*/
struct ra_info {
daddr_t ra_dsize; /* size in sectors */
/* u_long ra_type; /* drive type */
u_long ra_mediaid; /* media id */
int ra_state; /* open/closed state */
struct ra_geom { /* geometry information */
u_short rg_nsectors; /* sectors/track */
u_short rg_ngroups; /* track groups */
u_short rg_ngpc; /* groups/cylinder */
u_short rg_ntracks; /* ngroups*ngpc */
u_short rg_ncyl; /* ra_dsize/ntracks/nsectors */
#ifdef notyet
u_short rg_rctsize; /* size of rct */
u_short rg_rbns; /* replacement blocks per track */
u_short rg_nrct; /* number of rct copies */
#endif
} ra_geom;
int ra_wlabel; /* label sector is currently writable */
u_long ra_openpart; /* partitions open */
u_long ra_bopenpart; /* block partitions open */
u_long ra_copenpart; /* character partitions open */
} ra_info[NRA];
/*
* Software state, per drive
*/
#define CLOSED 0
#define WANTOPEN 1
#define RDLABEL 2
#define OPEN 3
#define OPENRAW 4
/*
* Definition of the driver for autoconf.
*/
int udaprobe(), udaslave(), udaattach(), udadgo(), udaintr();
struct uba_ctlr *udaminfo[NUDA];
struct uba_device *udadinfo[NRA];
struct disklabel udalabel[NRA];
u_short udastd[] = { 0772150, 0772550, 0777550, 0 };
struct uba_driver udadriver =
{ udaprobe, udaslave, udaattach, udadgo, udastd, "ra", udadinfo, "uda",
udaminfo };
/*
* More driver definitions, for generic MSCP code.
*/
int udadgram(), udactlrdone(), udaunconf(), udaiodone();
int udaonline(), udagotstatus(), udaioerror(), udareplace(), udabb();
struct buf udautab[NRA]; /* per drive transfer queue */
struct mscp_driver udamscpdriver =
{ MAXUNIT, NRA, UNITSHIFT, udautab, udalabel, udadinfo,
udadgram, udactlrdone, udaunconf, udaiodone,
udaonline, udagotstatus, udareplace, udaioerror, udabb,
"uda", "ra" };
/*
* Miscellaneous private variables.
*/
char udasr_bits[] = UDASR_BITS;
struct uba_device *udaip[NUDA][MAXUNIT];
/* inverting pointers: ctlr & unit => Unibus
device pointer */
int udaburst[NUDA] = { 0 }; /* burst size, per UDA50, zero => default;
in data space so patchable via adb */
struct mscp udaslavereply; /* get unit status response packet, set
for udaslave by udaunconf, via udaintr */
static struct uba_ctlr *probeum;/* this is a hack---autoconf should pass ctlr
info to slave routine; instead, we remember
the last ctlr argument to probe */
int udawstart, udawatch(); /* watchdog timer */
/*
* Externals
*/
int hz;
/*
* Poke at a supposed UDA50 to see if it is there.
* This routine duplicates some of the code in udainit() only
* because autoconf has not set up the right information yet.
* We have to do everything `by hand'.
*/
udaprobe(reg, ctlr, um)
caddr_t reg;
int ctlr;
struct uba_ctlr *um;
{
int br, cvec;
struct uda_softc *sc;
struct udadevice *udaddr;
struct mscp_info *mi;
extern int cpu_type;
int timeout, tries;
#ifdef QBA
int s;
#endif
#ifdef VAX750
/*
* The UDA50 wants to share BDPs on 750s, but not on 780s or
* 8600s. (730s have no BDPs anyway.) Toward this end, we
* here set the `keep bdp' flag in the per-driver information
* if this is a 750. (We just need to do it once, but it is
* easiest to do it now, for each UDA50.)
*/
if (MACHID(cpu_type) == VAX_750)
udadriver.ud_keepbdp = 1;
#endif
probeum = um; /* remember for udaslave() */
/*
* Set up the controller-specific generic MSCP driver info.
* Note that this should really be done in the (nonexistent)
* controller attach routine.
*/
sc = &uda_softc[ctlr];
mi = &sc->sc_mi;
mi->mi_md = &udamscpdriver;
mi->mi_ctlr = um->um_ctlr;
mi->mi_tab = &um->um_tab;
mi->mi_ip = udaip[ctlr];
mi->mi_cmd.mri_size = NCMD;
mi->mi_cmd.mri_desc = uda[ctlr].uda_ca.ca_cmddsc;
mi->mi_cmd.mri_ring = uda[ctlr].uda_cmd;
mi->mi_rsp.mri_size = NRSP;
mi->mi_rsp.mri_desc = uda[ctlr].uda_ca.ca_rspdsc;
mi->mi_rsp.mri_ring = uda[ctlr].uda_rsp;
mi->mi_wtab.b_actf = &mi->mi_wtab;
/* Was: mi->mi_wtab.av_forw = mi->mi_wtab.av_back = &mi->mi_wtab; */
/*
* More controller specific variables. Again, this should
* be in the controller attach routine.
*/
if (udaburst[ctlr] == 0)
udaburst[ctlr] = DEFAULT_BURST;
/*
* Get an interrupt vector. Note that even if the controller
* does not respond, we keep the vector. This is not a serious
* problem; but it would be easily fixed if we had a controller
* attach routine. Sigh.
*/
sc->sc_ivec = (uba_hd[numuba].uh_lastiv -= 4);
udaddr = (struct udadevice *) reg;
/*
* Initialise the controller (partially). The UDA50 programmer's
* manual states that if initialisation fails, it should be retried
* at least once, but after a second failure the port should be
* considered `down'; it also mentions that the controller should
* initialise within ten seconds. Or so I hear; I have not seen
* this manual myself.
*/
#if defined(QBA) && !defined(GENERIC)
s = spl6();
#endif
tries = 0;
again:
udaddr->udaip = 0; /* start initialisation */
timeout = todr() + 1000; /* timeout in 10 seconds */
while ((udaddr->udasa & UDA_STEP1) == 0)
if (todr() > timeout)
goto bad;
udaddr->udasa = UDA_ERR | (NCMDL2 << 11) | (NRSPL2 << 8) | UDA_IE |
(sc->sc_ivec >> 2);
while ((udaddr->udasa & UDA_STEP2) == 0)
if (todr() > timeout)
goto bad;
/* should have interrupted by now */
#ifdef QBA
#ifndef GENERIC
sc->sc_ipl = br = qbgetpri();
#else
sc->sc_ipl = br = 0x15;
#endif
#endif
return (sizeof (struct udadevice));
bad:
if (++tries < 2)
goto again;
#if defined(QBA) && !defined(GENERIC)
splx(s);
#endif
return (0);
}
/*
* Find a slave. We allow wildcard slave numbers (something autoconf
* is not really prepared to deal with); and we need to know the
* controller number to talk to the UDA. For the latter, we keep
* track of the last controller probed, since a controller probe
* immediately precedes all slave probes for that controller. For the
* former, we simply put the unit number into ui->ui_slave after we
* have found one.
*
* Note that by the time udaslave is called, the interrupt vector
* for the UDA50 has been set up (so that udaunconf() will be called).
*/
udaslave(ui, reg)
register struct uba_device *ui;
caddr_t reg;
{
register struct uba_ctlr *um = probeum;
register struct mscp *mp;
register struct uda_softc *sc;
int next = 0, timeout, tries, i;
#ifdef lint
i = 0; i = i;
#endif
/*
* Make sure the controller is fully initialised, by waiting
* for it if necessary.
*/
sc = &uda_softc[um->um_ctlr];
if (sc->sc_state == ST_RUN)
goto findunit;
tries = 0;
again:
if (udainit(ui->ui_ctlr))
return (0);
timeout = todr() + 1000; /* 10 seconds */
while (todr() < timeout) {
/* printf("hjagdsfhsafdahsdfasfga %d\n",sc->sc_state); */
if (sc->sc_state == ST_RUN) /* made it */
goto findunit;
}
if (++tries < 2)
goto again;
printf("uda%d: controller hung\n", um->um_ctlr);
return (0);
/*
* The controller is all set; go find the unit. Grab an
* MSCP packet and send out a Get Unit Status command, with
* the `next unit' modifier if we are looking for a generic
* unit. We set the `in slave' flag so that udaunconf()
* knows to copy the response to `udaslavereply'.
*/
findunit:
udaslavereply.mscp_opcode = 0;
sc->sc_flags |= SC_INSLAVE;
if ((mp = mscp_getcp(&sc->sc_mi, MSCP_DONTWAIT)) == NULL)
panic("udaslave"); /* `cannot happen' */
mp->mscp_opcode = M_OP_GETUNITST;
if (ui->ui_slave == '?') {
mp->mscp_unit = next;
mp->mscp_modifier = M_GUM_NEXTUNIT;
} else {
mp->mscp_unit = ui->ui_slave;
mp->mscp_modifier = 0;
}
*mp->mscp_addr |= MSCP_OWN | MSCP_INT;
i = ((struct udadevice *) reg)->udaip; /* initiate polling */
mp = &udaslavereply;
timeout = todr() + 1000;
while (todr() < timeout)
if (mp->mscp_opcode)
goto gotit;
printf("uda%d: no response to Get Unit Status request\n",
um->um_ctlr);
sc->sc_flags &= ~SC_INSLAVE;
return (0);
gotit:
sc->sc_flags &= ~SC_INSLAVE;
/*
* Got a slave response. If the unit is there, use it.
*/
switch (mp->mscp_status & M_ST_MASK) {
case M_ST_SUCCESS: /* worked */
case M_ST_AVAILABLE: /* found another drive */
break; /* use it */
case M_ST_OFFLINE:
/*
* Figure out why it is off line. It may be because
* it is nonexistent, or because it is spun down, or
* for some other reason.
*/
switch (mp->mscp_status & ~M_ST_MASK) {
case M_OFFLINE_UNKNOWN:
/*
* No such drive, and there are none with
* higher unit numbers either, if we are
* using M_GUM_NEXTUNIT.
*/
return (0);
case M_OFFLINE_UNMOUNTED:
/*
* The drive is not spun up. Use it anyway.
*
* N.B.: this seems to be a common occurrance
* after a power failure. The first attempt
* to bring it on line seems to spin it up
* (and thus takes several minutes). Perhaps
* we should note here that the on-line may
* take longer than usual.
*/
break;
default:
/*
* In service, or something else equally unusable.
*/
printf("uda%d: unit %d off line: ", um->um_ctlr,
mp->mscp_unit);
mscp_printevent(mp);
goto try_another;
}
break;
default:
printf("uda%d: unable to get unit status: ", um->um_ctlr);
mscp_printevent(mp);
return (0);
}
/*
* Does this ever happen? What (if anything) does it mean?
*/
if (mp->mscp_unit < next) {
printf("uda%d: unit %d, next %d\n",
um->um_ctlr, mp->mscp_unit, next);
return (0);
}
if (mp->mscp_unit >= MAXUNIT) {
printf("uda%d: cannot handle unit number %d (max is %d)\n",
um->um_ctlr, mp->mscp_unit, MAXUNIT - 1);
return (0);
}
/*
* See if we already handle this drive.
* (Only likely if ui->ui_slave=='?'.)
*/
if (udaip[um->um_ctlr][mp->mscp_unit] != NULL) {
try_another:
if (ui->ui_slave != '?')
return (0);
next = mp->mscp_unit + 1;
goto findunit;
}
/*
* Voila!
*/
uda_rasave(ui->ui_unit, mp, 0);
ui->ui_flags = 0; /* not on line, nor anything else */
ui->ui_slave = mp->mscp_unit;
return (1);
}
/*
* Attach a found slave. Make sure the watchdog timer is running.
* If this disk is being profiled, fill in the `wpms' value (used by
* what?). Set up the inverting pointer, and attempt to bring the
* drive on line and read its label.
*/
udaattach(ui)
register struct uba_device *ui;
{
register int unit = ui->ui_unit;
if (udawstart == 0) {
timeout(udawatch, (caddr_t) 0, hz);
udawstart++;
}
/*
* Floppies cannot be brought on line unless there is
* a disk in the drive. Since an ONLINE while cold
* takes ten seconds to fail, and (when notyet becomes now)
* no sensible person will swap to one, we just
* defer the ONLINE until someone tries to use the drive.
*
* THIS ASSUMES THAT DRIVE TYPES ?X? ARE FLOPPIES
*/
if (MSCP_MID_ECH(1, ra_info[unit].ra_mediaid) == 'X' - '@') {
printf(": floppy");
return;
}
if (ui->ui_dk >= 0)
dk_wpms[ui->ui_dk] = (60 * 31 * 256); /* approx */
udaip[ui->ui_ctlr][ui->ui_slave] = ui;
if (uda_rainit(ui, 0))
printf(": offline");
else if (ra_info[unit].ra_state == OPEN) {
printf(": %s, size = %d sectors",
udalabel[unit].d_typename, ra_info[unit].ra_dsize);
#ifdef notyet
addswap(makedev(UDADEVNUM, udaminor(unit, 0)), &udalabel[unit]);
#endif
}
}
/*
* Initialise a UDA50. Return true iff something goes wrong.
*/
udainit(ctlr)
int ctlr;
{
register struct uda_softc *sc;
register struct udadevice *udaddr;
struct uba_ctlr *um;
int timo, ubinfo;
sc = &uda_softc[ctlr];
um = udaminfo[ctlr];
if ((sc->sc_flags & SC_MAPPED) == 0) {
/*
* Map the communication area and command and
* response packets into Unibus space.
*/
ubinfo = uballoc(um->um_ubanum, (caddr_t) &uda[ctlr],
sizeof (struct uda), UBA_CANTWAIT);
if (ubinfo == 0) {
printf("uda%d: uballoc map failed\n", ctlr);
return (-1);
}
sc->sc_uda = (struct uda *) UBAI_ADDR(ubinfo);
sc->sc_flags |= SC_MAPPED;
}
/*
* While we are thinking about it, reset the next command
* and response indicies.
*/
sc->sc_mi.mi_cmd.mri_next = 0;
sc->sc_mi.mi_rsp.mri_next = 0;
/*
* Start up the hardware initialisation sequence.
*/
#define STEP0MASK (UDA_ERR | UDA_STEP4 | UDA_STEP3 | UDA_STEP2 | \
UDA_STEP1 | UDA_NV)
sc->sc_state = ST_IDLE; /* in case init fails */
udaddr = (struct udadevice *)um->um_addr;
udaddr->udaip = 0;
timo = todr() + 1000;
while ((udaddr->udasa & STEP0MASK) == 0) {
if (todr() > timo) {
printf("uda%d: timeout during init\n", ctlr);
return (-1);
}
}
if ((udaddr->udasa & STEP0MASK) != UDA_STEP1) {
printf("uda%d: init failed, sa=%b\n", ctlr,
udaddr->udasa, udasr_bits);
udasaerror(um, 0);
return (-1);
}
/*
* Success! Record new state, and start step 1 initialisation.
* The rest is done in the interrupt handler.
*/
sc->sc_state = ST_STEP1;
udaddr->udasa = UDA_ERR | (NCMDL2 << 11) | (NRSPL2 << 8) | UDA_IE |
(sc->sc_ivec >> 2);
return (0);
}
/*
* Open a drive.
*/
/*ARGSUSED*/
udaopen(dev, flag, fmt)
dev_t dev;
int flag, fmt;
{
register int unit;
register struct uba_device *ui;
register struct uda_softc *sc;
register struct disklabel *lp;
register struct partition *pp;
register struct ra_info *ra;
int s, i, part, mask, error = 0;
daddr_t start, end;
/*
* Make sure this is a reasonable open request.
*/
unit = udaunit(dev);
if (unit >= NRA || (ui = udadinfo[unit]) == 0 || ui->ui_alive == 0)
return (ENXIO);
/*
* Make sure the controller is running, by (re)initialising it if
* necessary.
*/
sc = &uda_softc[ui->ui_ctlr];
s = splbio();
if (sc->sc_state != ST_RUN) {
if (sc->sc_state == ST_IDLE && udainit(ui->ui_ctlr)) {
splx(s);
return (EIO);
}
/*
* In case it does not come up, make sure we will be
* restarted in 10 seconds. This corresponds to the
* 10 second timeouts in udaprobe() and udaslave().
*/
sc->sc_flags |= SC_DOWAKE;
timeout(wakeup, (caddr_t) sc, 10 * hz);
sleep((caddr_t) sc, PRIBIO);
if (sc->sc_state != ST_RUN) {
splx(s);
printf("uda%d: controller hung\n", ui->ui_ctlr);
return (EIO);
}
untimeout(wakeup, (caddr_t) sc);
}
/*
* Wait for the state to settle
*/
ra = &ra_info[unit];
while (ra->ra_state != OPEN && ra->ra_state != OPENRAW &&
ra->ra_state != CLOSED)
if (error = tsleep((caddr_t)ra, (PZERO + 1) | PCATCH,
devopn, 0)) {
splx(s);
return (error);
}
/*
* If not on line, or we are not sure of the label, reinitialise
* the drive.
*/
if ((ui->ui_flags & UNIT_ONLINE) == 0 ||
(ra->ra_state != OPEN && ra->ra_state != OPENRAW))
error = uda_rainit(ui, flag);
splx(s);
if (error)
return (error);
part = udapart(dev);
lp = &udalabel[unit];
if (part >= lp->d_npartitions)
return (ENXIO);
/*
* Warn if a partition is opened that overlaps another
* already open, unless either is the `raw' partition
* (whole disk).
*/
#define RAWPART 2 /* 'c' partition */ /* XXX */
mask = 1 << part;
if ((ra->ra_openpart & mask) == 0 && part != RAWPART) {
pp = &lp->d_partitions[part];
start = pp->p_offset;
end = pp->p_offset + pp->p_size;
for (pp = lp->d_partitions, i = 0;
i < lp->d_npartitions; pp++, i++) {
if (pp->p_offset + pp->p_size <= start ||
pp->p_offset >= end || i == RAWPART)
continue;
if (ra->ra_openpart & (1 << i))
log(LOG_WARNING,
"ra%d%c: overlaps open partition (%c)\n",
unit, part + 'a', i + 'a');
}
}
switch (fmt) {
case S_IFCHR:
ra->ra_copenpart |= mask;
break;
case S_IFBLK:
ra->ra_bopenpart |= mask;
break;
}
ra->ra_openpart |= mask;
return (0);
}
/* ARGSUSED */
udaclose(dev, flags, fmt)
dev_t dev;
int flags, fmt;
{
register int unit = udaunit(dev);
register struct ra_info *ra = &ra_info[unit];
int s, mask = (1 << udapart(dev));
switch (fmt) {
case S_IFCHR:
ra->ra_copenpart &= ~mask;
break;
case S_IFBLK:
ra->ra_bopenpart &= ~mask;
break;
}
ra->ra_openpart = ra->ra_copenpart | ra->ra_bopenpart;
/*
* Should wait for I/O to complete on this partition even if
* others are open, but wait for work on blkflush().
*/
if (ra->ra_openpart == 0) {
s = splbio();
while (udautab[unit].b_actf)
sleep((caddr_t)&udautab[unit], PZERO - 1);
splx(s);
ra->ra_state = CLOSED;
ra->ra_wlabel = 0;
}
return (0);
}
/*
* Initialise a drive. If it is not already, bring it on line,
* and set a timeout on it in case it fails to respond.
* When on line, read in the pack label.
*/
uda_rainit(ui, flags)
register struct uba_device *ui;
int flags;
{
register struct uda_softc *sc = &uda_softc[ui->ui_ctlr];
register struct disklabel *lp;
register struct mscp *mp;
register int unit = ui->ui_unit;
register struct ra_info *ra;
char *msg, *readdisklabel();
int s, i, udastrategy();
extern int cold;
ra = &ra_info[unit];
if ((ui->ui_flags & UNIT_ONLINE) == 0) {
mp = mscp_getcp(&sc->sc_mi, MSCP_WAIT);
mp->mscp_opcode = M_OP_ONLINE;
mp->mscp_unit = ui->ui_slave;
mp->mscp_cmdref = (long)&ui->ui_flags;
*mp->mscp_addr |= MSCP_OWN | MSCP_INT;
ra->ra_state = WANTOPEN;
if (!cold)
s = splbio();
i = ((struct udadevice *)ui->ui_addr)->udaip;
if (cold) {
i = todr() + 1000;
while ((ui->ui_flags & UNIT_ONLINE) == 0)
if (todr() > i)
break;
} else {
timeout(wakeup, (caddr_t)&ui->ui_flags, 10 * hz);
sleep((caddr_t)&ui->ui_flags, PSWP + 1);
splx(s);
untimeout(wakeup, (caddr_t)&ui->ui_flags);
}
if (ra->ra_state != OPENRAW) {
ra->ra_state = CLOSED;
wakeup((caddr_t)ra);
return (EIO);
}
}
lp = &udalabel[unit];
lp->d_secsize = DEV_BSIZE;
lp->d_secperunit = ra->ra_dsize;
if (flags & O_NDELAY)
return (0);
ra->ra_state = RDLABEL;
/*
* Set up default sizes until we have the label, or longer
* if there is none. Set secpercyl, as readdisklabel wants
* to compute b_cylin (although we do not need it), and set
* nsectors in case diskerr is called.
*/
lp->d_secpercyl = 1;
lp->d_npartitions = 1;
lp->d_secsize = 512;
lp->d_secperunit = ra->ra_dsize;
lp->d_nsectors = ra->ra_geom.rg_nsectors;
lp->d_partitions[0].p_size = lp->d_secperunit;
lp->d_partitions[0].p_offset = 0;
/*
* Read pack label.
*/
if ((msg = readdisklabel(udaminor(unit, 0), udastrategy, lp)) != NULL) {
if (cold)
printf(": %s", msg);
else
log(LOG_ERR, "ra%d: %s", unit, msg);
#ifdef COMPAT_42
if (udamaptype(unit, lp))
ra->ra_state = OPEN;
else
ra->ra_state = OPENRAW;
#else
ra->ra_state = OPENRAW;
uda_makefakelabel(ra, lp);
#endif
} else
ra->ra_state = OPEN;
wakeup((caddr_t)ra);
return (0);
}
/*
* Copy the geometry information for the given ra from a
* GET UNIT STATUS response. If check, see if it changed.
*/
uda_rasave(unit, mp, check)
int unit;
register struct mscp *mp;
int check;
{
register struct ra_info *ra = &ra_info[unit];
if (check && ra->ra_mediaid != mp->mscp_guse.guse_mediaid) {
printf("ra%d: changed types! was %d now %d\n", unit,
ra->ra_mediaid, mp->mscp_guse.guse_mediaid);
ra->ra_state = CLOSED; /* ??? */
}
/* ra->ra_type = mp->mscp_guse.guse_drivetype; */
ra->ra_mediaid = mp->mscp_guse.guse_mediaid;
ra->ra_geom.rg_nsectors = mp->mscp_guse.guse_nspt;
ra->ra_geom.rg_ngroups = mp->mscp_guse.guse_group;
ra->ra_geom.rg_ngpc = mp->mscp_guse.guse_ngpc;
ra->ra_geom.rg_ntracks = ra->ra_geom.rg_ngroups * ra->ra_geom.rg_ngpc;
/* ra_geom.rg_ncyl cannot be computed until we have ra_dsize */
#ifdef notyet
ra->ra_geom.rg_rctsize = mp->mscp_guse.guse_rctsize;
ra->ra_geom.rg_rbns = mp->mscp_guse.guse_nrpt;
ra->ra_geom.rg_nrct = mp->mscp_guse.guse_nrct;
#endif
}
/*
* Queue a transfer request, and if possible, hand it to the controller.
*
* This routine is broken into two so that the internal version
* udastrat1() can be called by the (nonexistent, as yet) bad block
* revectoring routine.
*/
udastrategy(bp)
register struct buf *bp;
{
register int unit;
register struct uba_device *ui;
register struct ra_info *ra;
struct partition *pp;
int p;
daddr_t sz, maxsz;
/*
* Make sure this is a reasonable drive to use.
*/
if ((unit = udaunit(bp->b_dev)) >= NRA ||
(ui = udadinfo[unit]) == NULL || ui->ui_alive == 0 ||
(ra = &ra_info[unit])->ra_state == CLOSED) {
bp->b_error = ENXIO;
goto bad;
}
/*
* If drive is open `raw' or reading label, let it at it.
*/
if (ra->ra_state < OPEN) {
udastrat1(bp);
return;
}
p = udapart(bp->b_dev);
if ((ra->ra_openpart & (1 << p)) == 0) {
bp->b_error = ENODEV;
goto bad;
}
/*
* Determine the size of the transfer, and make sure it is
* within the boundaries of the partition.
*/
pp = &udalabel[unit].d_partitions[p];
maxsz = pp->p_size;
if (pp->p_offset + pp->p_size > ra->ra_dsize)
maxsz = ra->ra_dsize - pp->p_offset;
sz = (bp->b_bcount + DEV_BSIZE - 1) >> DEV_BSHIFT;
if (bp->b_blkno + pp->p_offset <= LABELSECTOR &&
#if LABELSECTOR != 0
bp->b_blkno + pp->p_offset + sz > LABELSECTOR &&
#endif
(bp->b_flags & B_READ) == 0 && ra->ra_wlabel == 0) {
bp->b_error = EROFS;
goto bad;
}
if (bp->b_blkno < 0 || bp->b_blkno + sz > maxsz) {
/* if exactly at end of disk, return an EOF */
if (bp->b_blkno == maxsz) {
bp->b_resid = bp->b_bcount;
biodone(bp);
return;
}
/* or truncate if part of it fits */
sz = maxsz - bp->b_blkno;
if (sz <= 0) {
bp->b_error = EINVAL; /* or hang it up */
goto bad;
}
bp->b_bcount = sz << DEV_BSHIFT;
}
udastrat1(bp);
return;
bad:
bp->b_flags |= B_ERROR;
biodone(bp);
}
/*
* Work routine for udastrategy.
*/
udastrat1(bp)
register struct buf *bp;
{
register int unit = udaunit(bp->b_dev);
register struct uba_ctlr *um;
register struct buf *dp;
struct uba_device *ui;
int s = splbio();
/*
* Append the buffer to the drive queue, and if it is not
* already there, the drive to the controller queue. (However,
* if the drive queue is marked to be requeued, we must be
* awaiting an on line or get unit status command; in this
* case, leave it off the controller queue.)
*/
um = (ui = udadinfo[unit])->ui_mi;
dp = &udautab[unit];
APPEND(bp, dp, b_actf);
if (dp->b_active == 0 && (ui->ui_flags & UNIT_REQUEUE) == 0) {
APPEND(dp, &um->um_tab, b_hash.le_next);
dp->b_active++;
}
/* Was: APPEND(bp, dp, av_forw);
if (dp->b_active == 0 && (ui->ui_flags & UNIT_REQUEUE) == 0) {
APPEND(dp, &um->um_tab, b_forw);
dp->b_active++;
}
*/
/*
* Start activity on the controller. Note that unlike other
* Unibus drivers, we must always do this, not just when the
* controller is not active.
*/
udastart(um);
splx(s);
}
/*
* Start up whatever transfers we can find.
* Note that udastart() must be called at splbio().
*/
udastart(um)
register struct uba_ctlr *um;
{
register struct uda_softc *sc = &uda_softc[um->um_ctlr];
register struct buf *bp, *dp;
register struct mscp *mp;
struct uba_device *ui;
struct udadevice *udaddr;
struct partition *pp;
int i, sz;
#ifdef lint
i = 0; i = i;
#endif
/*
* If it is not running, try (again and again...) to initialise
* it. If it is currently initialising just ignore it for now.
*/
if (sc->sc_state != ST_RUN) {
if (sc->sc_state == ST_IDLE && udainit(um->um_ctlr))
printf("uda%d: still hung\n", um->um_ctlr);
return;
}
/*
* If um_cmd is nonzero, this controller is on the Unibus
* resource wait queue. It will not help to try more requests;
* instead, when the Unibus unblocks and calls udadgo(), we
* will call udastart() again.
*/
if (um->um_cmd)
return;
sc->sc_flags |= SC_INSTART;
udaddr = (struct udadevice *) um->um_addr;
loop:
/*
* Service the drive at the head of the queue. It may not
* need anything, in which case it might be shutting down
* in udaclose().
*/
if ((dp = um->um_tab.b_actf) == NULL)
goto out;
if ((bp = dp->b_actf) == NULL) {
dp->b_active = 0;
um->um_tab.b_actf = dp->b_hash.le_next;
/* Was: um->um_tab.b_actf = dp->b_forw; */
if (ra_info[dp - udautab].ra_openpart == 0)
wakeup((caddr_t)dp); /* finish close protocol */
goto loop;
}
if (udaddr->udasa & UDA_ERR) { /* ctlr fatal error */
udasaerror(um, 1);
goto out;
}
/*
* Get an MSCP packet, then figure out what to do. If
* we cannot get a command packet, the command ring may
* be too small: We should have at least as many command
* packets as credits, for best performance.
*/
if ((mp = mscp_getcp(&sc->sc_mi, MSCP_DONTWAIT)) == NULL) {
if (sc->sc_mi.mi_credits > MSCP_MINCREDITS &&
(sc->sc_flags & SC_GRIPED) == 0) {
log(LOG_NOTICE, "uda%d: command ring too small\n",
um->um_ctlr);
sc->sc_flags |= SC_GRIPED;/* complain only once */
}
goto out;
}
/*
* Bring the drive on line if it is not already. Get its status
* if we do not already have it. Otherwise just start the transfer.
*/
ui = udadinfo[udaunit(bp->b_dev)];
if ((ui->ui_flags & UNIT_ONLINE) == 0) {
mp->mscp_opcode = M_OP_ONLINE;
goto common;
}
if ((ui->ui_flags & UNIT_HAVESTATUS) == 0) {
mp->mscp_opcode = M_OP_GETUNITST;
common:
if (ui->ui_flags & UNIT_REQUEUE) panic("udastart");
/*
* Take the drive off the controller queue. When the
* command finishes, make sure the drive is requeued.
*/
um->um_tab.b_actf = dp->b_hash.le_next;
/* Was: um->um_tab.b_actf = dp->b_forw; */
dp->b_active = 0;
ui->ui_flags |= UNIT_REQUEUE;
mp->mscp_unit = ui->ui_slave;
*mp->mscp_addr |= MSCP_OWN | MSCP_INT;
sc->sc_flags |= SC_STARTPOLL;
#ifdef POLLSTATS
sc->sc_ncmd++;
#endif
goto loop;
}
pp = &udalabel[ui->ui_unit].d_partitions[udapart(bp->b_dev)];
mp->mscp_opcode = (bp->b_flags & B_READ) ? M_OP_READ : M_OP_WRITE;
mp->mscp_unit = ui->ui_slave;
mp->mscp_seq.seq_lbn = bp->b_blkno + pp->p_offset;
sz = (bp->b_bcount + DEV_BSIZE - 1) >> DEV_BSHIFT;
mp->mscp_seq.seq_bytecount = bp->b_blkno + sz > pp->p_size ?
(pp->p_size - bp->b_blkno) >> DEV_BSHIFT : bp->b_bcount;
/* mscp_cmdref is filled in by mscp_go() */
/*
* Drop the packet pointer into the `command' field so udadgo()
* can tell what to start. If ubago returns 1, we can do another
* transfer. If not, um_cmd will still point at mp, so we will
* know that we are waiting for resources.
*/
um->um_cmd = (int)mp;
if (ubago(ui))
goto loop;
/*
* All done, or blocked in ubago(). If we managed to
* issue some commands, start up the beast.
*/
out:
if (sc->sc_flags & SC_STARTPOLL) {
#ifdef POLLSTATS
udastats.cmd[sc->sc_ncmd]++;
sc->sc_ncmd = 0;
#endif
i = ((struct udadevice *)um->um_addr)->udaip;
}
sc->sc_flags &= ~(SC_INSTART | SC_STARTPOLL);
}
/*
* Start a transfer.
*
* If we are not called from within udastart(), we must have been
* blocked, so call udastart to do more requests (if any). If
* this calls us again immediately we will not recurse, because
* that time we will be in udastart(). Clever....
*/
udadgo(um)
register struct uba_ctlr *um;
{
struct uda_softc *sc = &uda_softc[um->um_ctlr];
struct mscp *mp = (struct mscp *)um->um_cmd;
um->um_tab.b_active++; /* another transfer going */
/*
* Fill in the MSCP packet and move the buffer to the
* I/O wait queue. Mark the controller as no longer on
* the resource queue, and remember to initiate polling.
*/
mp->mscp_seq.seq_buffer = UBAI_ADDR(um->um_ubinfo) |
(UBAI_BDP(um->um_ubinfo) << 24);
mscp_go(&sc->sc_mi, mp, um->um_ubinfo);
um->um_cmd = 0;
um->um_ubinfo = 0; /* tyke it awye */
sc->sc_flags |= SC_STARTPOLL;
#ifdef POLLSTATS
sc->sc_ncmd++;
#endif
if ((sc->sc_flags & SC_INSTART) == 0)
udastart(um);
}
udaiodone(mi, bp, info)
register struct mscp_info *mi;
struct buf *bp;
int info;
{
register struct uba_ctlr *um = udaminfo[mi->mi_ctlr];
um->um_ubinfo = info;
ubadone(um);
biodone(bp);
if (um->um_bdp && mi->mi_wtab.b_actf == &mi->mi_wtab)
ubarelse(um->um_ubanum, &um->um_bdp);
/* Was: if (um->um_bdp && mi->mi_wtab.av_forw == &mi->mi_wtab)
ubarelse(um->um_ubanum, &um->um_bdp); */
um->um_tab.b_active--; /* another transfer done */
}
static struct saerr {
int code; /* error code (including UDA_ERR) */
char *desc; /* what it means: Efoo => foo error */
} saerr[] = {
{ 0100001, "Eunibus packet read" },
{ 0100002, "Eunibus packet write" },
{ 0100003, "EUDA ROM and RAM parity" },
{ 0100004, "EUDA RAM parity" },
{ 0100005, "EUDA ROM parity" },
{ 0100006, "Eunibus ring read" },
{ 0100007, "Eunibus ring write" },
{ 0100010, " unibus interrupt master failure" },
{ 0100011, "Ehost access timeout" },
{ 0100012, " host exceeded command limit" },
{ 0100013, " unibus bus master failure" },
{ 0100014, " DM XFC fatal error" },
{ 0100015, " hardware timeout of instruction loop" },
{ 0100016, " invalid virtual circuit id" },
{ 0100017, "Eunibus interrupt write" },
{ 0104000, "Efatal sequence" },
{ 0104040, " D proc ALU" },
{ 0104041, "ED proc control ROM parity" },
{ 0105102, "ED proc w/no BD#2 or RAM parity" },
{ 0105105, "ED proc RAM buffer" },
{ 0105152, "ED proc SDI" },
{ 0105153, "ED proc write mode wrap serdes" },
{ 0105154, "ED proc read mode serdes, RSGEN & ECC" },
{ 0106040, "EU proc ALU" },
{ 0106041, "EU proc control reg" },
{ 0106042, " U proc DFAIL/cntl ROM parity/BD #1 test CNT" },
{ 0106047, " U proc const PROM err w/D proc running SDI test" },
{ 0106055, " unexpected trap" },
{ 0106071, "EU proc const PROM" },
{ 0106072, "EU proc control ROM parity" },
{ 0106200, "Estep 1 data" },
{ 0107103, "EU proc RAM parity" },
{ 0107107, "EU proc RAM buffer" },
{ 0107115, " test count wrong (BD 12)" },
{ 0112300, "Estep 2" },
{ 0122240, "ENPR" },
{ 0122300, "Estep 3" },
{ 0142300, "Estep 4" },
{ 0, " unknown error code" }
};
/*
* If the error bit was set in the controller status register, gripe,
* then (optionally) reset the controller and requeue pending transfers.
*/
udasaerror(um, doreset)
register struct uba_ctlr *um;
int doreset;
{
register int code = ((struct udadevice *)um->um_addr)->udasa;
register struct saerr *e;
if ((code & UDA_ERR) == 0)
return;
for (e = saerr; e->code; e++)
if (e->code == code)
break;
printf("uda%d: controller error, sa=0%o (%s%s)\n",
um->um_ctlr, code, e->desc + 1,
*e->desc == 'E' ? " error" : "");
if (doreset) {
mscp_requeue(&uda_softc[um->um_ctlr].sc_mi);
(void) udainit(um->um_ctlr);
}
}
/*
* Interrupt routine. Depending on the state of the controller,
* continue initialisation, or acknowledge command and response
* interrupts, and process responses.
*/
udaintr(/*ctlr*/)
/* int ctlr; */
{
struct uba_ctlr *um = udaminfo[/*ctlr*/0];
struct uda_softc *sc = &uda_softc[/*ctlr*/0];
struct udadevice *udaddr = (struct udadevice *)um->um_addr;
struct uda *ud;
struct mscp *mp;
int i;
extern int cpu_type;
int ctlr=0; /* XXX Fult f|r att testa interrupter :-/ */
#ifdef QBA
splx(sc->sc_ipl); /* Qbus interrupt protocol is odd */
#endif
sc->sc_wticks = 0; /* reset interrupt watchdog */
/*
* Combinations during steps 1, 2, and 3: STEPnMASK
* corresponds to which bits should be tested;
* STEPnGOOD corresponds to the pattern that should
* appear after the interrupt from STEPn initialisation.
* All steps test the bits in ALLSTEPS.
*/
#define ALLSTEPS (UDA_ERR|UDA_STEP4|UDA_STEP3|UDA_STEP2|UDA_STEP1)
#define STEP1MASK (ALLSTEPS | UDA_IE | UDA_NCNRMASK)
#define STEP1GOOD (UDA_STEP2 | UDA_IE | (NCMDL2 << 3) | NRSPL2)
#define STEP2MASK (ALLSTEPS | UDA_IE | UDA_IVECMASK)
#define STEP2GOOD (UDA_STEP3 | UDA_IE | (sc->sc_ivec >> 2))
#define STEP3MASK ALLSTEPS
#define STEP3GOOD UDA_STEP4
switch (sc->sc_state) {
case ST_IDLE:
/*
* Ignore unsolicited interrupts.
*/
log(LOG_WARNING, "uda%d: stray intr\n", ctlr);
return;
case ST_STEP1:
/*
* Begin step two initialisation.
*/
if ((udaddr->udasa & STEP1MASK) != STEP1GOOD) {
i = 1;
initfailed:
printf("uda%d: init step %d failed, sa=%b\n",
ctlr, i, udaddr->udasa, udasr_bits);
udasaerror(um, 0);
sc->sc_state = ST_IDLE;
if (sc->sc_flags & SC_DOWAKE) {
sc->sc_flags &= ~SC_DOWAKE;
wakeup((caddr_t)sc);
}
return;
}
udaddr->udasa = (int)&sc->sc_uda->uda_ca.ca_rspdsc[0] |
(MACHID(cpu_type) == VAX_780 || MACHID(cpu_type)
== VAX_8600 ? UDA_PI : 0);
sc->sc_state = ST_STEP2;
return;
case ST_STEP2:
/*
* Begin step 3 initialisation.
*/
if ((udaddr->udasa & STEP2MASK) != STEP2GOOD) {
i = 2;
goto initfailed;
}
udaddr->udasa = ((int)&sc->sc_uda->uda_ca.ca_rspdsc[0]) >> 16;
sc->sc_state = ST_STEP3;
return;
case ST_STEP3:
/*
* Set controller characteristics (finish initialisation).
*/
if ((udaddr->udasa & STEP3MASK) != STEP3GOOD) {
i = 3;
goto initfailed;
}
i = udaddr->udasa & 0xff;
if (i != sc->sc_micro) {
sc->sc_micro = i;
printf("uda%d: version %d model %d\n",
ctlr, i & 0xf, i >> 4);
}
/*
* Present the burst size, then remove it. Why this
* should be done this way, I have no idea.
*
* Note that this assumes udaburst[ctlr] > 0.
*/
udaddr->udasa = UDA_GO | (udaburst[ctlr] - 1) << 2;
udaddr->udasa = UDA_GO;
printf("uda%d: DMA burst size set to %d\n",
ctlr, udaburst[ctlr]);
udainitds(ctlr); /* initialise data structures */
/*
* Before we can get a command packet, we need some
* credits. Fake some up to keep mscp_getcp() happy,
* get a packet, and cancel all credits (the right
* number should come back in the response to the
* SCC packet).
*/
sc->sc_mi.mi_credits = MSCP_MINCREDITS + 1;
/* printf("&sc->sc_mi: %x\n",&sc->sc_mi); */
mp = mscp_getcp(&sc->sc_mi, MSCP_DONTWAIT);
if (mp == NULL) /* `cannot happen' */
panic("udaintr");
/* printf("Kom f|rbi mscp_getcp()\n"); */
sc->sc_mi.mi_credits = 0;
mp->mscp_opcode = M_OP_SETCTLRC;
mp->mscp_unit = 0;
mp->mscp_sccc.sccc_ctlrflags = M_CF_ATTN | M_CF_MISC |
M_CF_THIS;
*mp->mscp_addr |= MSCP_OWN | MSCP_INT;
/* printf("udaddr l{ngre ner: %x\n",udaddr); */
i = udaddr->udaip;
sc->sc_state = ST_SETCHAR;
/* printf("ST_SETCHAR satt. %x\n", sc->sc_state); */
return;
case ST_SETCHAR:
case ST_RUN:
/*
* Handle Set Ctlr Characteristics responses and operational
* responses (via mscp_dorsp).
*/
break;
default:
printf("uda%d: driver bug, state %d\n", ctlr, sc->sc_state);
panic("udastate");
}
if (udaddr->udasa & UDA_ERR) { /* ctlr fatal error */
udasaerror(um, 1);
return;
}
ud = &uda[ctlr];
/*
* Handle buffer purge requests.
*/
if (ud->uda_ca.ca_bdp) {
UBAPURGE(um->um_hd->uh_uba, ud->uda_ca.ca_bdp);
ud->uda_ca.ca_bdp = 0;
udaddr->udasa = 0; /* signal purge complete */
}
/*
* Check for response and command ring transitions.
*/
if (ud->uda_ca.ca_rspint) {
ud->uda_ca.ca_rspint = 0;
mscp_dorsp(&sc->sc_mi);
}
if (ud->uda_ca.ca_cmdint) {
ud->uda_ca.ca_cmdint = 0;
MSCP_DOCMD(&sc->sc_mi);
}
udastart(um);
}
/*
* Initialise the various data structures that control the UDA50.
*/
udainitds(ctlr)
int ctlr;
{
register struct uda *ud = &uda[ctlr];
register struct uda *uud = uda_softc[ctlr].sc_uda;
register struct mscp *mp;
register int i;
for (i = 0, mp = ud->uda_rsp; i < NRSP; i++, mp++) {
ud->uda_ca.ca_rspdsc[i] = MSCP_OWN | MSCP_INT |
(long)&uud->uda_rsp[i].mscp_cmdref;
mp->mscp_addr = &ud->uda_ca.ca_rspdsc[i];
mp->mscp_msglen = MSCP_MSGLEN;
}
for (i = 0, mp = ud->uda_cmd; i < NCMD; i++, mp++) {
ud->uda_ca.ca_cmddsc[i] = MSCP_INT |
(long)&uud->uda_cmd[i].mscp_cmdref;
mp->mscp_addr = &ud->uda_ca.ca_cmddsc[i];
mp->mscp_msglen = MSCP_MSGLEN;
}
}
/*
* Handle an error datagram.
*/
udadgram(mi, mp)
struct mscp_info *mi;
struct mscp *mp;
{
mscp_decodeerror(mi->mi_md->md_mname, mi->mi_ctlr, mp);
/*
* SDI status information bytes 10 and 11 are the microprocessor
* error code and front panel code respectively. These vary per
* drive type and are printed purely for field service information.
*/
if (mp->mscp_format == M_FM_SDI)
printf("\tsdi uproc error code 0x%x, front panel code 0x%x\n",
mp->mscp_erd.erd_sdistat[10],
mp->mscp_erd.erd_sdistat[11]);
}
/*
* The Set Controller Characteristics command finished.
* Record the new state of the controller.
*/
udactlrdone(mi, mp)
register struct mscp_info *mi;
struct mscp *mp;
{
register struct uda_softc *sc = &uda_softc[mi->mi_ctlr];
if ((mp->mscp_status & M_ST_MASK) == M_ST_SUCCESS)
sc->sc_state = ST_RUN;
else {
printf("uda%d: SETCTLRC failed: ",
mi->mi_ctlr, mp->mscp_status);
mscp_printevent(mp);
sc->sc_state = ST_IDLE;
}
if (sc->sc_flags & SC_DOWAKE) {
sc->sc_flags &= ~SC_DOWAKE;
wakeup((caddr_t)sc);
}
}
/*
* Received a response from an as-yet unconfigured drive. Configure it
* in, if possible.
*/
udaunconf(mi, mp)
struct mscp_info *mi;
register struct mscp *mp;
{
/*
* If it is a slave response, copy it to udaslavereply for
* udaslave() to look at.
*/
if (mp->mscp_opcode == (M_OP_GETUNITST | M_OP_END) &&
(uda_softc[mi->mi_ctlr].sc_flags & SC_INSLAVE) != 0) {
udaslavereply = *mp;
return (MSCP_DONE);
}
/*
* Otherwise, it had better be an available attention response.
*/
if (mp->mscp_opcode != M_OP_AVAILATTN)
return (MSCP_FAILED);
/* do what autoconf does */
return (MSCP_FAILED); /* not yet, arwhite, not yet */
}
/*
* A drive came on line. Check its type and size. Return DONE if
* we think the drive is truly on line. In any case, awaken anyone
* sleeping on the drive on-line-ness.
*/
udaonline(ui, mp)
register struct uba_device *ui;
struct mscp *mp;
{
register struct ra_info *ra = &ra_info[ui->ui_unit];
wakeup((caddr_t)&ui->ui_flags);
if ((mp->mscp_status & M_ST_MASK) != M_ST_SUCCESS) {
if (!cold)
printf("uda%d: ra%d", ui->ui_ctlr, ui->ui_unit);
printf(": attempt to bring on line failed: ");
mscp_printevent(mp);
ra->ra_state = CLOSED;
return (MSCP_FAILED);
}
ra->ra_state = OPENRAW;
ra->ra_dsize = (daddr_t)mp->mscp_onle.onle_unitsize;
if (!cold)
printf("ra%d: uda%d, unit %d, size = %d sectors\n", ui->ui_unit,
ui->ui_ctlr, mp->mscp_unit, ra->ra_dsize);
/* can now compute ncyl */
ra->ra_geom.rg_ncyl = ra->ra_dsize / ra->ra_geom.rg_ntracks /
ra->ra_geom.rg_nsectors;
return (MSCP_DONE);
}
/*
* We got some (configured) unit's status. Return DONE if it succeeded.
*/
udagotstatus(ui, mp)
register struct uba_device *ui;
register struct mscp *mp;
{
if ((mp->mscp_status & M_ST_MASK) != M_ST_SUCCESS) {
printf("uda%d: attempt to get status for ra%d failed: ",
ui->ui_ctlr, ui->ui_unit);
mscp_printevent(mp);
return (MSCP_FAILED);
}
/* record for (future) bad block forwarding and whatever else */
uda_rasave(ui->ui_unit, mp, 1);
return (MSCP_DONE);
}
/*
* A transfer failed. We get a chance to fix or restart it.
* Need to write the bad block forwaring code first....
*/
/*ARGSUSED*/
udaioerror(ui, mp, bp)
register struct uba_device *ui;
register struct mscp *mp;
struct buf *bp;
{
if (mp->mscp_flags & M_EF_BBLKR) {
/*
* A bad block report. Eventually we will
* restart this transfer, but for now, just
* log it and give up.
*/
log(LOG_ERR, "ra%d: bad block report: %d%s\n",
ui->ui_unit, mp->mscp_seq.seq_lbn,
mp->mscp_flags & M_EF_BBLKU ? " + others" : "");
} else {
/*
* What the heck IS a `serious exception' anyway?
* IT SURE WOULD BE NICE IF DEC SOLD DOCUMENTATION
* FOR THEIR OWN CONTROLLERS.
*/
if (mp->mscp_flags & M_EF_SEREX)
log(LOG_ERR, "ra%d: serious exception reported\n",
ui->ui_unit);
}
return (MSCP_FAILED);
}
/*
* A replace operation finished.
*/
/*ARGSUSED*/
udareplace(ui, mp)
struct uba_device *ui;
struct mscp *mp;
{
panic("udareplace");
}
/*
* A bad block related operation finished.
*/
/*ARGSUSED*/
udabb(ui, mp, bp)
struct uba_device *ui;
struct mscp *mp;
struct buf *bp;
{
panic("udabb");
}
/*
* I/O controls.
*/
udaioctl(dev, cmd, data, flag)
dev_t dev;
int cmd;
caddr_t data;
int flag;
{
register int unit = udaunit(dev);
register struct disklabel *lp;
register struct ra_info *ra = &ra_info[unit];
int error = 0;
lp = &udalabel[unit];
switch (cmd) {
case DIOCGDINFO:
*(struct disklabel *)data = *lp;
break;
case DIOCGPART:
((struct partinfo *)data)->disklab = lp;
((struct partinfo *)data)->part =
&lp->d_partitions[udapart(dev)];
break;
case DIOCSDINFO:
if ((flag & FWRITE) == 0)
error = EBADF;
else
error = setdisklabel(lp, (struct disklabel *)data,
(ra->ra_state == OPENRAW) ? 0 : ra->ra_openpart,0);
break;
case DIOCWLABEL:
if ((flag & FWRITE) == 0)
error = EBADF;
else
ra->ra_wlabel = *(int *)data;
break;
case DIOCWDINFO:
if ((flag & FWRITE) == 0)
error = EBADF;
else if ((error = setdisklabel(lp, (struct disklabel *)data,
(ra->ra_state == OPENRAW) ? 0 : ra->ra_openpart,0)) == 0) {
int wlab;
ra->ra_state = OPEN;
/* simulate opening partition 0 so write succeeds */
ra->ra_openpart |= (1 << 0); /* XXX */
wlab = ra->ra_wlabel;
ra->ra_wlabel = 1;
error = writedisklabel(dev, udastrategy, lp,0);
ra->ra_openpart = ra->ra_copenpart | ra->ra_bopenpart;
ra->ra_wlabel = wlab;
}
break;
#ifdef notyet
case UDAIOCREPLACE:
/*
* Initiate bad block replacement for the given LBN.
* (Should we allow modifiers?)
*/
error = EOPNOTSUPP;
break;
case UDAIOCGMICRO:
/*
* Return the microcode revision for the UDA50 running
* this drive.
*/
*(int *)data = uda_softc[uddinfo[unit]->ui_ctlr].sc_micro;
break;
#endif
default:
error = ENOTTY;
break;
}
return (error);
}
/*
* A Unibus reset has occurred on UBA uban. Reinitialise the controller(s)
* on that Unibus, and requeue outstanding I/O.
*/
udareset(uban)
int uban;
{
register struct uba_ctlr *um;
register struct uda_softc *sc;
register int ctlr;
for (ctlr = 0, sc = uda_softc; ctlr < NUDA; ctlr++, sc++) {
if ((um = udaminfo[ctlr]) == NULL || um->um_ubanum != uban ||
um->um_alive == 0)
continue;
printf(" uda%d", ctlr);
/*
* Our BDP (if any) is gone; our command (if any) is
* flushed; the device is no longer mapped; and the
* UDA50 is not yet initialised.
*/
if (um->um_bdp) {
printf("<%d>", UBAI_BDP(um->um_bdp));
um->um_bdp = 0;
}
um->um_ubinfo = 0;
um->um_cmd = 0;
sc->sc_flags &= ~SC_MAPPED;
sc->sc_state = ST_IDLE;
/* reset queues and requeue pending transfers */
mscp_requeue(&sc->sc_mi);
/*
* If it fails to initialise we will notice later and
* try again (and again...). Do not call udastart()
* here; it will be done after the controller finishes
* initialisation.
*/
if (udainit(ctlr))
printf(" (hung)");
}
}
/*
* Watchdog timer: If the controller is active, and no interrupts
* have occurred for 30 seconds, assume it has gone away.
*/
udawatch()
{
register int i;
register struct uba_ctlr *um;
register struct uda_softc *sc;
timeout(udawatch, (caddr_t) 0, hz); /* every second */
for (i = 0, sc = uda_softc; i < NUDA; i++, sc++) {
if ((um = udaminfo[i]) == 0 || !um->um_alive)
continue;
if (sc->sc_state == ST_IDLE)
continue;
if (sc->sc_state == ST_RUN && !um->um_tab.b_active)
sc->sc_wticks = 0;
else if (++sc->sc_wticks >= 30) {
sc->sc_wticks = 0;
printf("uda%d: lost interrupt\n", i);
ubareset(um->um_ubanum);
}
}
}
/*
* Do a panic dump. We set up the controller for one command packet
* and one response packet, for which we use `struct uda1'.
*/
struct uda1 {
struct uda1ca uda1_ca; /* communications area */
struct mscp uda1_rsp; /* response packet */
struct mscp uda1_cmd; /* command packet */
} uda1;
#define DBSIZE 32 /* dump 16K at a time */
udadump(dev)
dev_t dev;
{
struct udadevice *udaddr;
struct uda1 *ud_ubaddr;
char *start;
int num, blk, unit, maxsz, blkoff, reg;
struct partition *pp;
register struct uba_regs *uba;
register struct uba_device *ui;
register struct uda1 *ud;
register struct pte *io;
register int i;
/*
* Make sure the device is a reasonable place on which to dump.
*/
unit = udaunit(dev);
if (unit >= NRA)
return (ENXIO);
#define phys(cast, addr) ((cast) ((int)addr & 0x7fffffff))
ui = phys(struct uba_device *, udadinfo[unit]);
if (ui == NULL || ui->ui_alive == 0)
return (ENXIO);
/*
* Find and initialise the UBA; get the physical address of the
* device registers, and of communications area and command and
* response packet.
*/
uba = phys(struct uba_hd *, ui->ui_hd)->uh_physuba;
ubainit(uba);
udaddr = (struct udadevice *)ui->ui_physaddr;
ud = phys(struct uda1 *, &uda1);
/*
* Map the ca+packets into Unibus I/O space so the UDA50 can get
* at them. Use the registers at the end of the Unibus map (since
* we will use the registers at the beginning to map the memory
* we are dumping).
*/
num = btoc(sizeof(struct uda1)) + 1;
reg = NUBMREG - num;
io = &uba->uba_map[reg];
for (i = 0; i < num; i++)
*(int *)io++ = UBAMR_MRV | (vax_btop(ud) + i);
ud_ubaddr = (struct uda1 *)(((int)ud & PGOFSET) | (reg << 9));
/*
* Initialise the controller, with one command and one response
* packet.
*/
udaddr->udaip = 0;
if (udadumpwait(udaddr, UDA_STEP1))
return (EFAULT);
udaddr->udasa = UDA_ERR;
if (udadumpwait(udaddr, UDA_STEP2))
return (EFAULT);
udaddr->udasa = (int)&ud_ubaddr->uda1_ca.ca_rspdsc;
if (udadumpwait(udaddr, UDA_STEP3))
return (EFAULT);
udaddr->udasa = ((int)&ud_ubaddr->uda1_ca.ca_rspdsc) >> 16;
if (udadumpwait(udaddr, UDA_STEP4))
return (EFAULT);
uda_softc[ui->ui_ctlr].sc_micro = udaddr->udasa & 0xff;
udaddr->udasa = UDA_GO;
/*
* Set up the command and response descriptor, then set the
* controller characteristics and bring the drive on line.
* Note that all uninitialised locations in uda1_cmd are zero.
*/
ud->uda1_ca.ca_rspdsc = (long)&ud_ubaddr->uda1_rsp.mscp_cmdref;
ud->uda1_ca.ca_cmddsc = (long)&ud_ubaddr->uda1_cmd.mscp_cmdref;
/* ud->uda1_cmd.mscp_sccc.sccc_ctlrflags = 0; */
/* ud->uda1_cmd.mscp_sccc.sccc_version = 0; */
if (udadumpcmd(M_OP_SETCTLRC, ud, ui))
return (EFAULT);
ud->uda1_cmd.mscp_unit = ui->ui_slave;
if (udadumpcmd(M_OP_ONLINE, ud, ui))
return (EFAULT);
pp = phys(struct partition *,
&udalabel[unit].d_partitions[udapart(dev)]);
maxsz = pp->p_size;
blkoff = pp->p_offset;
/*
* Dump all of physical memory, or as much as will fit in the
* space provided.
*/
start = 0;
printf("Dumpar {r inte implementerade {n :) \n");
asm("halt");
/* num = maxfree; */
if (dumplo + num >= maxsz)
num = maxsz - dumplo;
blkoff += dumplo;
/*
* Write out memory, DBSIZE pages at a time.
* N.B.: this code depends on the fact that the sector
* size == the page size.
*/
while (num > 0) {
blk = num > DBSIZE ? DBSIZE : num;
io = uba->uba_map;
/*
* Map in the pages to write, leaving an invalid entry
* at the end to guard against wild Unibus transfers.
* Then do the write.
*/
for (i = 0; i < blk; i++)
*(int *)io++ = UBAMR_MRV | (vax_btop(start) + i);
*(int *)io = 0;
ud->uda1_cmd.mscp_unit = ui->ui_slave;
ud->uda1_cmd.mscp_seq.seq_lbn = vax_btop(start) + blkoff;
ud->uda1_cmd.mscp_seq.seq_bytecount = blk << PGSHIFT;
if (udadumpcmd(M_OP_WRITE, ud, ui))
return (EIO);
start += blk << PGSHIFT;
num -= blk;
}
return (0); /* made it! */
}
/*
* Wait for some of the bits in `bits' to come on. If the error bit
* comes on, or ten seconds pass without response, return true (error).
*/
udadumpwait(udaddr, bits)
register struct udadevice *udaddr;
register int bits;
{
register int timo = todr() + 1000;
while ((udaddr->udasa & bits) == 0) {
if (udaddr->udasa & UDA_ERR) {
printf("udasa=%b\ndump ", udaddr->udasa, udasr_bits);
return (1);
}
if (todr() >= timo) {
printf("timeout\ndump ");
return (1);
}
}
return (0);
}
/*
* Feed a command to the UDA50, wait for its response, and return
* true iff something went wrong.
*/
udadumpcmd(op, ud, ui)
int op;
register struct uda1 *ud;
struct uba_device *ui;
{
register struct udadevice *udaddr;
register int n;
#define mp (&ud->uda1_rsp)
udaddr = (struct udadevice *)ui->ui_physaddr;
ud->uda1_cmd.mscp_opcode = op;
ud->uda1_cmd.mscp_msglen = MSCP_MSGLEN;
ud->uda1_rsp.mscp_msglen = MSCP_MSGLEN;
ud->uda1_ca.ca_rspdsc |= MSCP_OWN | MSCP_INT;
ud->uda1_ca.ca_cmddsc |= MSCP_OWN | MSCP_INT;
if (udaddr->udasa & UDA_ERR) {
printf("udasa=%b\ndump ", udaddr->udasa, udasr_bits);
return (1);
}
n = udaddr->udaip;
n = todr() + 1000;
for (;;) {
if (todr() > n) {
printf("timeout\ndump ");
return (1);
}
if (ud->uda1_ca.ca_cmdint)
ud->uda1_ca.ca_cmdint = 0;
if (ud->uda1_ca.ca_rspint == 0)
continue;
ud->uda1_ca.ca_rspint = 0;
if (mp->mscp_opcode == (op | M_OP_END))
break;
printf("\n");
switch (MSCP_MSGTYPE(mp->mscp_msgtc)) {
case MSCPT_SEQ:
printf("sequential");
break;
case MSCPT_DATAGRAM:
mscp_decodeerror("uda", ui->ui_ctlr, mp);
printf("datagram");
break;
case MSCPT_CREDITS:
printf("credits");
break;
case MSCPT_MAINTENANCE:
printf("maintenance");
break;
default:
printf("unknown (type 0x%x)",
MSCP_MSGTYPE(mp->mscp_msgtc));
break;
}
printf(" ignored\ndump ");
ud->uda1_ca.ca_rspdsc |= MSCP_OWN | MSCP_INT;
}
if ((mp->mscp_status & M_ST_MASK) != M_ST_SUCCESS) {
printf("error: op 0x%x => 0x%x status 0x%x\ndump ", op,
mp->mscp_opcode, mp->mscp_status);
return (1);
}
return (0);
#undef mp
}
/*
* Return the size of a partition, if known, or -1 if not.
*/
udasize(dev)
dev_t dev;
{
register int unit = udaunit(dev);
register struct uba_device *ui;
if (unit >= NRA || (ui = udadinfo[unit]) == NULL ||
ui->ui_alive == 0 || (ui->ui_flags & UNIT_ONLINE) == 0 ||
ra_info[unit].ra_state != OPEN)
return (-1);
return ((int)udalabel[unit].d_partitions[udapart(dev)].p_size);
}
#ifdef COMPAT_42
/*
* Tables mapping unlabelled drives.
*/
struct size {
daddr_t nblocks;
daddr_t blkoff;
} ra60_sizes[8] = {
15884, 0, /* A=sectors 0 thru 15883 */
33440, 15884, /* B=sectors 15884 thru 49323 */
400176, 0, /* C=sectors 0 thru 400175 */
82080, 49324, /* 4.2 G => D=sectors 49324 thru 131403 */
268772, 131404, /* 4.2 H => E=sectors 131404 thru 400175 */
350852, 49324, /* F=sectors 49324 thru 400175 */
157570, 242606, /* UCB G => G=sectors 242606 thru 400175 */
193282, 49324, /* UCB H => H=sectors 49324 thru 242605 */
}, ra70_sizes[8] = {
15884, 0, /* A=blk 0 thru 15883 */
33440, 15972, /* B=blk 15972 thru 49323 */
-1, 0, /* C=blk 0 thru end */
15884, 341220, /* D=blk 341220 thru 357103 */
55936, 357192, /* E=blk 357192 thru 413127 */
-1, 413457, /* F=blk 413457 thru end */
-1, 341220, /* G=blk 341220 thru end */
291346, 49731, /* H=blk 49731 thru 341076 */
}, ra80_sizes[8] = {
15884, 0, /* A=sectors 0 thru 15883 */
33440, 15884, /* B=sectors 15884 thru 49323 */
242606, 0, /* C=sectors 0 thru 242605 */
0, 0, /* D=unused */
193282, 49324, /* UCB H => E=sectors 49324 thru 242605 */
82080, 49324, /* 4.2 G => F=sectors 49324 thru 131403 */
192696, 49910, /* G=sectors 49910 thru 242605 */
111202, 131404, /* 4.2 H => H=sectors 131404 thru 242605 */
}, ra81_sizes[8] ={
/*
* These are the new standard partition sizes for ra81's.
* An RA_COMPAT system is compiled with D, E, and F corresponding
* to the 4.2 partitions for G, H, and F respectively.
*/
#ifndef UCBRA
15884, 0, /* A=sectors 0 thru 15883 */
66880, 16422, /* B=sectors 16422 thru 83301 */
891072, 0, /* C=sectors 0 thru 891071 */
#ifdef RA_COMPAT
82080, 49324, /* 4.2 G => D=sectors 49324 thru 131403 */
759668, 131404, /* 4.2 H => E=sectors 131404 thru 891071 */
478582, 412490, /* 4.2 F => F=sectors 412490 thru 891071 */
#else
15884, 375564, /* D=sectors 375564 thru 391447 */
307200, 391986, /* E=sectors 391986 thru 699185 */
191352, 699720, /* F=sectors 699720 thru 891071 */
#endif RA_COMPAT
515508, 375564, /* G=sectors 375564 thru 891071 */
291346, 83538, /* H=sectors 83538 thru 374883 */
/*
* These partitions correspond to the sizes used by sites at Berkeley,
* and by those sites that have received copies of the Berkeley driver
* with deltas 6.2 or greater (11/15/83).
*/
#else UCBRA
15884, 0, /* A=sectors 0 thru 15883 */
33440, 15884, /* B=sectors 15884 thru 49323 */
891072, 0, /* C=sectors 0 thru 891071 */
15884, 242606, /* D=sectors 242606 thru 258489 */
307200, 258490, /* E=sectors 258490 thru 565689 */
325382, 565690, /* F=sectors 565690 thru 891071 */
648466, 242606, /* G=sectors 242606 thru 891071 */
193282, 49324, /* H=sectors 49324 thru 242605 */
#endif UCBRA
}, ra82_sizes[8] = {
15884, 0, /* A=blk 0 thru 15883 */
66880, 16245, /* B=blk 16245 thru 83124 */
-1, 0, /* C=blk 0 thru end */
15884, 375345, /* D=blk 375345 thru 391228 */
307200, 391590, /* E=blk 391590 thru 698789 */
-1, 699390, /* F=blk 699390 thru end */
-1, 375345, /* G=blk 375345 thru end */
291346, 83790, /* H=blk 83790 thru 375135 */
}, rc25_sizes[8] = {
15884, 0, /* A=blk 0 thru 15883 */
10032, 15884, /* B=blk 15884 thru 49323 */
-1, 0, /* C=blk 0 thru end */
0, 0, /* D=blk 340670 thru 356553 */
0, 0, /* E=blk 356554 thru 412489 */
0, 0, /* F=blk 412490 thru end */
-1, 25916, /* G=blk 49324 thru 131403 */
0, 0, /* H=blk 131404 thru end */
}, rd52_sizes[8] = {
15884, 0, /* A=blk 0 thru 15883 */
9766, 15884, /* B=blk 15884 thru 25649 */
-1, 0, /* C=blk 0 thru end */
0, 0, /* D=unused */
0, 0, /* E=unused */
0, 0, /* F=unused */
-1, 25650, /* G=blk 25650 thru end */
0, 0, /* H=unused */
}, rd53_sizes[8] = {
15884, 0, /* A=blk 0 thru 15883 */
33440, 15884, /* B=blk 15884 thru 49323 */
-1, 0, /* C=blk 0 thru end */
0, 0, /* D=unused */
33440, 0, /* E=blk 0 thru 33439 */
-1, 33440, /* F=blk 33440 thru end */
-1, 49324, /* G=blk 49324 thru end */
-1, 15884, /* H=blk 15884 thru end */
}, rd54_sizes[8] = {
15884, 0, /* A=blk 0 thru 15883 */
33440, 15884, /* B=blk 15884 thru 49323 */
-1, 0, /* C=blk 0 thru end */
130938, 49324, /* D=blk 49324 thru 180261 */
130938, 180262, /* E=blk 180262 thru 311199 (end) */
0, 0, /* F=unused */
261876, 49324, /* G=blk 49324 thru 311199 (end) */
0, 0, /* H=unused */
}, rx50_sizes[8] = {
800, 0, /* A=blk 0 thru 799 */
0, 0,
-1, 0, /* C=blk 0 thru end */
0, 0,
0, 0,
0, 0,
0, 0,
0, 0,
};
/*
* Media ID decoding table.
*/
struct udatypes {
u_long ut_id; /* media drive ID */
char *ut_name; /* drive type name */
struct size *ut_sizes; /* partition tables */
int ut_nsectors, ut_ntracks, ut_ncylinders;
} udatypes[] = {
{ MSCP_MKDRIVE2('R', 'A', 60), "ra60", ra60_sizes, 42, 4, 2382 },
{ MSCP_MKDRIVE2('R', 'A', 70), "ra70", ra70_sizes, 33, 11, 1507 },
{ MSCP_MKDRIVE2('R', 'A', 80), "ra80", ra80_sizes, 31, 14, 559 },
{ MSCP_MKDRIVE2('R', 'A', 81), "ra81", ra81_sizes, 51, 14, 1248 },
{ MSCP_MKDRIVE2('R', 'A', 82), "ra82", ra82_sizes, 57, 15, 1423 },
{ MSCP_MKDRIVE2('R', 'C', 25), "rc25-removable",
rc25_sizes, 42, 4, 302 },
{ MSCP_MKDRIVE3('R', 'C', 'F', 25), "rc25-fixed",
rc25_sizes, 42, 4, 302 },
{ MSCP_MKDRIVE2('R', 'D', 52), "rd52", rd52_sizes, 18, 7, 480 },
{ MSCP_MKDRIVE2('R', 'D', 53), "rd53", rd53_sizes, 18, 8, 963 },
{ MSCP_MKDRIVE2('R', 'D', 32), "rd54-from-rd32",
rd54_sizes, 17, 15, 1220 },
{ MSCP_MKDRIVE2('R', 'D', 54), "rd54", rd54_sizes, 17, 15, 1220 },
{ MSCP_MKDRIVE2('R', 'X', 50), "rx50", rx50_sizes, 10, 1, 80 },
0
};
#define NTYPES (sizeof(udatypes) / sizeof(*udatypes))
udamaptype(unit, lp)
int unit;
register struct disklabel *lp;
{
register struct udatypes *ut;
register struct size *sz;
register struct partition *pp;
register char *p;
register int i;
register struct ra_info *ra = &ra_info[unit];
i = MSCP_MEDIA_DRIVE(ra->ra_mediaid);
for (ut = udatypes; ut->ut_id; ut++)
if (ut->ut_id == i &&
ut->ut_nsectors == ra->ra_geom.rg_nsectors &&
ut->ut_ntracks == ra->ra_geom.rg_ntracks &&
ut->ut_ncylinders == ra->ra_geom.rg_ncyl)
goto found;
/* not one we know; fake up a label for the whole drive */
uda_makefakelabel(ra, lp);
i = ra->ra_mediaid; /* print the port type too */
addlog(": no partition table for %c%c %c%c%c%d, size %d;\n\
using (s,t,c)=(%d,%d,%d)",
MSCP_MID_CHAR(4, i), MSCP_MID_CHAR(3, i),
MSCP_MID_CHAR(2, i), MSCP_MID_CHAR(1, i),
MSCP_MID_CHAR(0, i), MSCP_MID_NUM(i), lp->d_secperunit,
lp->d_nsectors, lp->d_ntracks, lp->d_ncylinders);
if (!cold)
addlog("\n");
return (0);
found:
p = ut->ut_name;
for (i = 0; i < sizeof(lp->d_typename) - 1 && *p; i++)
lp->d_typename[i] = *p++;
lp->d_typename[i] = 0;
sz = ut->ut_sizes;
lp->d_nsectors = ut->ut_nsectors;
lp->d_ntracks = ut->ut_ntracks;
lp->d_ncylinders = ut->ut_ncylinders;
lp->d_npartitions = 8;
lp->d_secpercyl = lp->d_nsectors * lp->d_ntracks;
for (pp = lp->d_partitions; pp < &lp->d_partitions[8]; pp++, sz++) {
pp->p_offset = sz->blkoff;
if ((pp->p_size = sz->nblocks) == (u_long)-1)
pp->p_size = ra->ra_dsize - sz->blkoff;
}
return (1);
}
#endif /* COMPAT_42 */
/*
* Construct a label for a drive from geometry information
* if we have no better information.
*/
uda_makefakelabel(ra, lp)
register struct ra_info *ra;
register struct disklabel *lp;
{
lp->d_nsectors = ra->ra_geom.rg_nsectors;
lp->d_ntracks = ra->ra_geom.rg_ntracks;
lp->d_ncylinders = ra->ra_geom.rg_ncyl;
lp->d_secpercyl = lp->d_nsectors * lp->d_ntracks;
bcopy("ra??", lp->d_typename, sizeof("ra??"));
lp->d_npartitions = 1;
lp->d_partitions[0].p_offset = 0;
lp->d_partitions[0].p_size = lp->d_secperunit;
}
#endif /* NUDA > 0 */