/* $NetBSD: mt.c,v 1.4 2005/10/15 17:29:12 yamt Exp $ */ /*- * Copyright (c) 1996-2003 The NetBSD Foundation, Inc. * All rights reserved. * * This code is derived from software contributed to The NetBSD Foundation * by Jason R. Thorpe. * * 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 NetBSD * Foundation, Inc. and its contributors. * 4. Neither the name of The NetBSD Foundation 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 NETBSD FOUNDATION, INC. 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 FOUNDATION OR CONTRIBUTORS * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. */ /* * Copyright (c) 1982, 1990, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * the Systems Programming Group of the University of Utah Computer * Science Department. * * 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. 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: Utah $Hdr: rd.c 1.44 92/12/26$ * * @(#)rd.c 8.2 (Berkeley) 5/19/94 */ /* * Copyright (c) 1988 University of Utah. * * This code is derived from software contributed to Berkeley by * the Systems Programming Group of the University of Utah Computer * Science Department. * * 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: Utah $Hdr: rd.c 1.44 92/12/26$ * * @(#)rd.c 8.2 (Berkeley) 5/19/94 */ /* * Magnetic tape driver (HP7974a, HP7978a/b, HP7979a, HP7980a, HP7980xc) * Original version contributed by Mt. Xinu. * Modified for 4.4BSD by Mark Davies and Andrew Vignaux, Department of * Computer Science, Victoria University of Wellington */ #include __KERNEL_RCSID(0, "$NetBSD: mt.c,v 1.4 2005/10/15 17:29:12 yamt Exp $"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef DEBUG int mtdebug = 0; #define MDB_ANY 0xff #define MDB_FOLLOW 0x01 #define DPRINTF(mask, str) if (mtdebug & (mask)) printf str #else #define DPRINTF(mask, str) /* nothing */ #endif struct mt_softc { struct device sc_dev; gpib_chipset_tag_t sc_ic; gpib_handle_t sc_hdl; int sc_slave; /* GPIB slave address (0-6) */ short sc_flags; /* see below */ u_char sc_lastdsj; /* place for DSJ in mtreaddsj() */ u_char sc_lastecmd; /* place for End Command in mtreaddsj() */ short sc_recvtimeo; /* count of gpibsend timeouts to prevent hang */ short sc_statindex; /* index for next sc_stat when MTF_STATTIMEO */ struct mt_stat sc_stat;/* status bytes last read from device */ short sc_density; /* current density of tape (mtio.h format) */ short sc_type; /* tape drive model (hardware IDs) */ tpr_t sc_ttyp; struct bufq_state *sc_tab;/* buf queue */ int sc_active; struct buf sc_bufstore; /* XXX buffer storage */ struct callout sc_start_ch; struct callout sc_intr_ch; }; #define MTUNIT(x) (minor(x) & 0x03) #define B_CMD B_XXX /* command buf instead of data */ #define b_cmd b_blkno /* blkno holds cmd when B_CMD */ int mtmatch(struct device *, struct cfdata *, void *); void mtattach(struct device *, struct device *, void *); CFATTACH_DECL(mt, sizeof(struct mt_softc), mtmatch, mtattach, NULL, NULL); int mtlookup(int, int, int); void mtustart(struct mt_softc *); int mtreaddsj(struct mt_softc *, int); int mtcommand(dev_t, int, int); void mtintr_callout(void *); void mtstart_callout(void *); void mtcallback(void *, int); void mtstart(struct mt_softc *); void mtintr(struct mt_softc *); dev_type_open(mtopen); dev_type_close(mtclose); dev_type_read(mtread); dev_type_write(mtwrite); dev_type_ioctl(mtioctl); dev_type_strategy(mtstrategy); const struct bdevsw mt_bdevsw = { mtopen, mtclose, mtstrategy, mtioctl, nodump, nosize, D_TAPE }; const struct cdevsw mt_cdevsw = { mtopen, mtclose, mtread, mtwrite, mtioctl, nostop, notty, nopoll, nommap, nokqfilter, D_TAPE }; extern struct cfdriver mt_cd; struct mtinfo { u_short hwid; char *desc; } mtinfo[] = { { MT7978ID, "7978" }, { MT7979AID, "7979A" }, { MT7980ID, "7980" }, { MT7974AID, "7974A" }, }; int nmtinfo = sizeof(mtinfo) / sizeof(mtinfo[0]); int mtlookup(id, slave, punit) int id; int slave; int punit; { int i; for (i = 0; i < nmtinfo; i++) if (mtinfo[i].hwid == id) break; if (i == nmtinfo) return (-1); return (0); } int mtmatch(parent, match, aux) struct device *parent; struct cfdata *match; void *aux; { struct cs80bus_attach_args *ca = aux; ca->ca_punit = 0; return (mtlookup(ca->ca_id, ca->ca_slave, ca->ca_punit) == 0); } void mtattach(parent, self, aux) struct device *parent, *self; void *aux; { struct mt_softc *sc = (struct mt_softc *)self; struct cs80bus_attach_args *ca = aux; int type; sc->sc_ic = ca->ca_ic; sc->sc_slave = ca->ca_slave; if ((type = mtlookup(ca->ca_id, ca->ca_slave, ca->ca_punit)) < 0) return; printf(": %s tape\n", mtinfo[type].desc); sc->sc_type = type; sc->sc_flags = MTF_EXISTS; bufq_alloc(&sc->sc_tab, "fcfs", 0); callout_init(&sc->sc_start_ch); callout_init(&sc->sc_intr_ch); if (gpibregister(sc->sc_ic, sc->sc_slave, mtcallback, sc, &sc->sc_hdl)) { printf("%s: can't register callback\n", sc->sc_dev.dv_xname); return; } } /* * Perform a read of "Device Status Jump" register and update the * status if necessary. If status is read, the given "ecmd" is also * performed, unless "ecmd" is zero. Returns DSJ value, -1 on failure * and -2 on "temporary" failure. */ int mtreaddsj(sc, ecmd) struct mt_softc *sc; int ecmd; { int retval; if (sc->sc_flags & MTF_STATTIMEO) goto getstats; retval = gpibrecv(sc->sc_ic, (sc->sc_flags & MTF_DSJTIMEO) ? -1 : sc->sc_slave, MTT_DSJ, &(sc->sc_lastdsj), 1); sc->sc_flags &= ~MTF_DSJTIMEO; if (retval != 1) { DPRINTF(MDB_ANY, ("%s can't gpibrecv DSJ", sc->sc_dev.dv_xname)); if (sc->sc_recvtimeo == 0) sc->sc_recvtimeo = hz; if (--sc->sc_recvtimeo == 0) return (-1); if (retval == 0) sc->sc_flags |= MTF_DSJTIMEO; return (-2); } sc->sc_recvtimeo = 0; sc->sc_statindex = 0; DPRINTF(MDB_ANY, ("%s readdsj: 0x%x", sc->sc_dev.dv_xname, sc->sc_lastdsj)); sc->sc_lastecmd = ecmd; switch (sc->sc_lastdsj) { case 0: if (ecmd & MTE_DSJ_FORCE) break; return (0); case 2: sc->sc_lastecmd = MTE_COMPLETE; case 1: break; default: printf("%s readdsj: DSJ 0x%x\n", sc->sc_dev.dv_xname, sc->sc_lastdsj); return (-1); } getstats: retval = gpibrecv(sc->sc_ic, (sc->sc_flags & MTF_STATCONT) ? -1 : sc->sc_slave, MTT_STAT, ((char *)&(sc->sc_stat)) + sc->sc_statindex, sizeof(sc->sc_stat) - sc->sc_statindex); sc->sc_flags &= ~(MTF_STATTIMEO | MTF_STATCONT); if (retval != sizeof(sc->sc_stat) - sc->sc_statindex) { if (sc->sc_recvtimeo == 0) sc->sc_recvtimeo = hz; if (--sc->sc_recvtimeo != 0) { if (retval >= 0) { sc->sc_statindex += retval; sc->sc_flags |= MTF_STATCONT; } sc->sc_flags |= MTF_STATTIMEO; return (-2); } printf("%s readdsj: can't read status", sc->sc_dev.dv_xname); return (-1); } sc->sc_recvtimeo = 0; sc->sc_statindex = 0; DPRINTF(MDB_ANY, ("%s readdsj: status is %x %x %x %x %x %x", sc->sc_dev.dv_xname, sc->sc_stat1, sc->sc_stat2, sc->sc_stat3, sc->sc_stat4, sc->sc_stat5, sc->sc_stat6)); if (sc->sc_lastecmd) (void) gpibsend(sc->sc_ic, sc->sc_slave, MTL_ECMD, &(sc->sc_lastecmd), 1); return ((int) sc->sc_lastdsj); } int mtopen(dev, flag, mode, p) dev_t dev; int flag, mode; struct proc *p; { struct mt_softc *sc; int req_den; int error; sc = device_lookup(&mt_cd, MTUNIT(dev)); if (sc == NULL || (sc->sc_flags & MTF_EXISTS) == 0) return (ENXIO); if (sc->sc_flags & MTF_OPEN) return (EBUSY); DPRINTF(MDB_ANY, ("%s open: flags 0x%x", sc->sc_dev.dv_xname, sc->sc_flags)); sc->sc_flags |= MTF_OPEN; sc->sc_ttyp = tprintf_open(p); if ((sc->sc_flags & MTF_ALIVE) == 0) { error = mtcommand(dev, MTRESET, 0); if (error != 0 || (sc->sc_flags & MTF_ALIVE) == 0) goto errout; if ((sc->sc_stat1 & (SR1_BOT | SR1_ONLINE)) == SR1_ONLINE) (void) mtcommand(dev, MTREW, 0); } for (;;) { if ((error = mtcommand(dev, MTNOP, 0)) != 0) goto errout; if (!(sc->sc_flags & MTF_REW)) break; if (tsleep((caddr_t) &lbolt, PCATCH | (PZERO + 1), "mt", 0) != 0) { error = EINTR; goto errout; } } if ((flag & FWRITE) && (sc->sc_stat1 & SR1_RO)) { error = EROFS; goto errout; } if (!(sc->sc_stat1 & SR1_ONLINE)) { uprintf("%s: not online\n", sc->sc_dev.dv_xname); error = EIO; goto errout; } /* * Select density: * - find out what density the drive is set to * (i.e. the density of the current tape) * - if we are going to write * - if we're not at the beginning of the tape * - complain if we want to change densities * - otherwise, select the mtcommand to set the density * * If the drive doesn't support it then don't change the recorded * density. * * The original MOREbsd code had these additional conditions * for the mid-tape change * * req_den != T_BADBPI && * sc->sc_density != T_6250BPI * * which suggests that it would be possible to write multiple * densities if req_den == T_BAD_BPI or the current tape * density was 6250. Testing of our 7980 suggests that the * device cannot change densities mid-tape. * * ajv@comp.vuw.ac.nz */ sc->sc_density = (sc->sc_stat2 & SR2_6250) ? T_6250BPI : ( (sc->sc_stat3 & SR3_1600) ? T_1600BPI : ( (sc->sc_stat3 & SR3_800) ? T_800BPI : -1)); req_den = (dev & T_DENSEL); if (flag & FWRITE) { if (!(sc->sc_stat1 & SR1_BOT)) { if (sc->sc_density != req_den) { uprintf("%s: can't change density mid-tape\n", sc->sc_dev.dv_xname); error = EIO; goto errout; } } else { int mtset_density = (req_den == T_800BPI ? MTSET800BPI : ( req_den == T_1600BPI ? MTSET1600BPI : ( req_den == T_6250BPI ? MTSET6250BPI : ( sc->sc_type == MT7980ID ? MTSET6250DC : MTSET6250BPI)))); if (mtcommand(dev, mtset_density, 0) == 0) sc->sc_density = req_den; } } return (0); errout: sc->sc_flags &= ~MTF_OPEN; return (error); } int mtclose(dev, flag, fmt, p) dev_t dev; int flag, fmt; struct proc *p; { struct mt_softc *sc; sc = device_lookup(&mt_cd, MTUNIT(dev)); if (sc == NULL) return (ENXIO); if (sc->sc_flags & MTF_WRT) { (void) mtcommand(dev, MTWEOF, 2); (void) mtcommand(dev, MTBSF, 0); } if ((minor(dev) & T_NOREWIND) == 0) (void) mtcommand(dev, MTREW, 0); sc->sc_flags &= ~MTF_OPEN; tprintf_close(sc->sc_ttyp); return (0); } int mtcommand(dev, cmd, cnt) dev_t dev; int cmd; int cnt; { struct mt_softc *sc; struct buf *bp; int error = 0; sc = device_lookup(&mt_cd, MTUNIT(dev)); bp = &sc->sc_bufstore; if (bp->b_flags & B_BUSY) return (EBUSY); bp->b_cmd = cmd; bp->b_dev = dev; do { bp->b_flags = B_BUSY | B_CMD; mtstrategy(bp); biowait(bp); if (bp->b_flags & B_ERROR) { error = (int) (unsigned) bp->b_error; break; } } while (--cnt > 0); #if 0 bp->b_flags = 0 /*&= ~B_BUSY*/; #else bp->b_flags &= ~B_BUSY; #endif return (error); } /* * Only thing to check here is for legal record lengths (writes only). */ void mtstrategy(bp) struct buf *bp; { struct mt_softc *sc; int s; sc = device_lookup(&mt_cd, MTUNIT(bp->b_dev)); DPRINTF(MDB_ANY, ("%s strategy", sc->sc_dev.dv_xname)); if ((bp->b_flags & (B_CMD | B_READ)) == 0) { #define WRITE_BITS_IGNORED 8 #if 0 if (bp->b_bcount & ((1 << WRITE_BITS_IGNORED) - 1)) { tprintf(sc->sc_ttyp, "%s: write record must be multiple of %d\n", sc->sc_dev.dv_xname, 1 << WRITE_BITS_IGNORED); goto error; } #endif s = 16 * 1024; if (sc->sc_stat2 & SR2_LONGREC) { switch (sc->sc_density) { case T_1600BPI: s = 32 * 1024; break; case T_6250BPI: case T_BADBPI: s = 60 * 1024; break; } } if (bp->b_bcount > s) { tprintf(sc->sc_ttyp, "%s: write record (%ld) too big: limit (%d)\n", sc->sc_dev.dv_xname, bp->b_bcount, s); #if 0 /* XXX see above */ error: #endif bp->b_flags |= B_ERROR; bp->b_error = EIO; biodone(bp); return; } } s = splbio(); BUFQ_PUT(sc->sc_tab, bp); if (sc->sc_active == 0) { sc->sc_active = 1; mtustart(sc); } splx(s); } void mtustart(sc) struct mt_softc *sc; { DPRINTF(MDB_ANY, ("%s ustart", sc->sc_dev.dv_xname)); if (gpibrequest(sc->sc_ic, sc->sc_hdl)) mtstart(sc); } void mtcallback(v, action) void *v; int action; { struct mt_softc *sc = v; DPRINTF(MDB_FOLLOW, ("mtcallback: v=%p, action=%d\n", v, action)); switch (action) { case GPIBCBF_START: mtstart(sc); break; case GPIBCBF_INTR: mtintr(sc); break; #ifdef DEBUG default: printf("mtcallback: unknown action %d\n", action); break; #endif } } void mtintr_callout(arg) void *arg; { struct mt_softc *sc = arg; int s = splbio(); gpibppclear(sc->sc_ic); mtintr(sc); splx(s); } void mtstart_callout(arg) void *arg; { int s = splbio(); mtstart((struct mt_softc *)arg); splx(s); } void mtstart(sc) struct mt_softc *sc; { struct buf *bp; short cmdcount = 1; u_char cmdbuf[2]; DPRINTF(MDB_ANY, ("%s start", sc->sc_dev.dv_xname)); sc->sc_flags &= ~MTF_WRT; bp = BUFQ_PEEK(sc->sc_tab); if ((sc->sc_flags & MTF_ALIVE) == 0 && ((bp->b_flags & B_CMD) == 0 || bp->b_cmd != MTRESET)) goto fatalerror; if (sc->sc_flags & MTF_REW) { if (!gpibpptest(sc->sc_ic, sc->sc_slave)) goto stillrew; switch (mtreaddsj(sc, MTE_DSJ_FORCE|MTE_COMPLETE|MTE_IDLE)) { case 0: case 1: stillrew: if ((sc->sc_stat1 & SR1_BOT) || !(sc->sc_stat1 & SR1_ONLINE)) { sc->sc_flags &= ~MTF_REW; break; } case -2: /* * -2 means "timeout" reading DSJ, which is probably * temporary. This is considered OK when doing a NOP, * but not otherwise. */ if (sc->sc_flags & (MTF_DSJTIMEO | MTF_STATTIMEO)) { callout_reset(&sc->sc_start_ch, hz >> 5, mtstart_callout, sc); return; } case 2: if (bp->b_cmd != MTNOP || !(bp->b_flags & B_CMD)) { bp->b_error = EBUSY; goto errdone; } goto done; default: goto fatalerror; } } if (bp->b_flags & B_CMD) { if (sc->sc_flags & MTF_PASTEOT) { switch(bp->b_cmd) { case MTFSF: case MTWEOF: case MTFSR: bp->b_error = ENOSPC; goto errdone; case MTBSF: case MTOFFL: case MTBSR: case MTREW: sc->sc_flags &= ~(MTF_PASTEOT | MTF_ATEOT); break; } } switch(bp->b_cmd) { case MTFSF: if (sc->sc_flags & MTF_HITEOF) goto done; cmdbuf[0] = MTTC_FSF; break; case MTBSF: if (sc->sc_flags & MTF_HITBOF) goto done; cmdbuf[0] = MTTC_BSF; break; case MTOFFL: sc->sc_flags |= MTF_REW; cmdbuf[0] = MTTC_REWOFF; break; case MTWEOF: cmdbuf[0] = MTTC_WFM; break; case MTBSR: cmdbuf[0] = MTTC_BSR; break; case MTFSR: cmdbuf[0] = MTTC_FSR; break; case MTREW: sc->sc_flags |= MTF_REW; cmdbuf[0] = MTTC_REW; break; case MTNOP: /* * NOP is supposed to set status bits. * Force readdsj to do it. */ switch (mtreaddsj(sc, MTE_DSJ_FORCE | MTE_COMPLETE | MTE_IDLE)) { default: goto done; case -1: /* * If this fails, perform a device clear * to fix any protocol problems and (most * likely) get the status. */ bp->b_cmd = MTRESET; break; case -2: callout_reset(&sc->sc_start_ch, hz >> 5, mtstart_callout, sc); return; } case MTRESET: /* * 1) selected device clear (send with "-2" secondary) * 2) set timeout, then wait for "service request" * 3) interrupt will read DSJ (and END COMPLETE-IDLE) */ if (gpibsend(sc->sc_ic, sc->sc_slave, -2, NULL, 0)){ printf("%s can't reset", sc->sc_dev.dv_xname); goto fatalerror; } callout_reset(&sc->sc_intr_ch, 4*hz, mtintr_callout, sc); gpibawait(sc->sc_ic); return; case MTSET800BPI: cmdbuf[0] = MTTC_800; break; case MTSET1600BPI: cmdbuf[0] = MTTC_1600; break; case MTSET6250BPI: cmdbuf[0] = MTTC_6250; break; case MTSET6250DC: cmdbuf[0] = MTTC_DC6250; break; } } else { if (sc->sc_flags & MTF_PASTEOT) { bp->b_error = ENOSPC; goto errdone; } if (bp->b_flags & B_READ) { sc->sc_flags |= MTF_IO; cmdbuf[0] = MTTC_READ; } else { sc->sc_flags |= MTF_WRT | MTF_IO; cmdbuf[0] = MTTC_WRITE; cmdbuf[1] = (bp->b_bcount +((1 << WRITE_BITS_IGNORED) - 1)) >> WRITE_BITS_IGNORED; cmdcount = 2; } } if (gpibsend(sc->sc_ic, sc->sc_slave, MTL_TCMD, cmdbuf, cmdcount) == cmdcount) { if (sc->sc_flags & MTF_REW) goto done; gpibawait(sc->sc_ic); return; } fatalerror: /* * If anything fails, the drive is probably hosed, so mark it not * "ALIVE" (but it EXISTS and is OPEN or we wouldn't be here, and * if, last we heard, it was REWinding, remember that). */ sc->sc_flags &= MTF_EXISTS | MTF_OPEN | MTF_REW; bp->b_error = EIO; errdone: bp->b_flags |= B_ERROR; done: sc->sc_flags &= ~(MTF_HITEOF | MTF_HITBOF); (void)BUFQ_GET(sc->sc_tab); biodone(bp); gpibrelease(sc->sc_ic, sc->sc_hdl); if ((bp = BUFQ_PEEK(sc->sc_tab)) == NULL) sc->sc_active = 0; else mtustart(sc); } void mtintr(sc) struct mt_softc *sc; { struct buf *bp; int slave, dir, i; u_char cmdbuf[4]; slave = sc->sc_slave; bp = BUFQ_PEEK(sc->sc_tab); if (bp == NULL) { printf("%s intr: bp == NULL", sc->sc_dev.dv_xname); return; } DPRINTF(MDB_ANY, ("%s intr", sc->sc_dev.dv_xname)); /* * Some operation completed. Read status bytes and report errors. * Clear EOF flags here `cause they're set once on specific conditions * below when a command succeeds. * A DSJ of 2 always means keep waiting. If the command was READ * (and we're in data DMA phase) stop data transfer first. */ sc->sc_flags &= ~(MTF_HITEOF | MTF_HITBOF); if ((bp->b_flags & (B_CMD|B_READ)) == B_READ && !(sc->sc_flags & (MTF_IO | MTF_STATTIMEO | MTF_DSJTIMEO))){ cmdbuf[0] = MTE_STOP; (void) gpibsend(sc->sc_ic, slave, MTL_ECMD,cmdbuf,1); } switch (mtreaddsj(sc, 0)) { case 0: break; case 1: /* * If we're in the middle of a READ/WRITE and have yet to * start the data transfer, a DSJ of one should terminate it. */ sc->sc_flags &= ~MTF_IO; break; case 2: (void) gpibawait(sc->sc_ic); return; case -2: /* * -2 means that the drive failed to respond quickly enough * to the request for DSJ. It's probably just "busy" figuring * it out and will know in a little bit... */ callout_reset(&sc->sc_intr_ch, hz >> 5, mtintr_callout, sc); return; default: printf("%s intr: can't get drive stat", sc->sc_dev.dv_xname); goto error; } if (sc->sc_stat1 & (SR1_ERR | SR1_REJECT)) { i = sc->sc_stat4 & SR4_ERCLMASK; printf("%s: %s error, retry %d, SR2/3 %x/%x, code %d", sc->sc_dev.dv_xname, i == SR4_DEVICE ? "device" : (i == SR4_PROTOCOL ? "protocol" : (i == SR4_SELFTEST ? "selftest" : "unknown")), sc->sc_stat4 & SR4_RETRYMASK, sc->sc_stat2, sc->sc_stat3, sc->sc_stat5); if ((bp->b_flags & B_CMD) && bp->b_cmd == MTRESET) callout_stop(&sc->sc_intr_ch); if (sc->sc_stat3 & SR3_POWERUP) sc->sc_flags &= MTF_OPEN | MTF_EXISTS; goto error; } /* * Report and clear any soft errors. */ if (sc->sc_stat1 & SR1_SOFTERR) { printf("%s: soft error, retry %d\n", sc->sc_dev.dv_xname, sc->sc_stat4 & SR4_RETRYMASK); sc->sc_stat1 &= ~SR1_SOFTERR; } /* * We've initiated a read or write, but haven't actually started to * DMA the data yet. At this point, the drive's ready. */ if (sc->sc_flags & MTF_IO) { sc->sc_flags &= ~MTF_IO; dir = (bp->b_flags & B_READ ? GPIB_READ : GPIB_WRITE); gpibxfer(sc->sc_ic, slave, dir == GPIB_READ ? MTT_READ : MTL_WRITE, bp->b_data, bp->b_bcount, dir, dir == GPIB_READ); return; } /* * Check for End Of Tape - we're allowed to hit EOT and then write (or * read) one more record. If we get here and have not already hit EOT, * return ENOSPC to inform the process that it's hit it. If we get * here and HAVE already hit EOT, don't allow any more operations that * move the tape forward. */ if (sc->sc_stat1 & SR1_EOT) { if (sc->sc_flags & MTF_ATEOT) sc->sc_flags |= MTF_PASTEOT; else { bp->b_flags |= B_ERROR; bp->b_error = ENOSPC; sc->sc_flags |= MTF_ATEOT; } } /* * If a motion command was being executed, check for Tape Marks. * If we were doing data, make sure we got the right amount, and * check for hitting tape marks on reads. */ if (bp->b_flags & B_CMD) { if (sc->sc_stat1 & SR1_EOF) { if (bp->b_cmd == MTFSR) sc->sc_flags |= MTF_HITEOF; if (bp->b_cmd == MTBSR) sc->sc_flags |= MTF_HITBOF; } if (bp->b_cmd == MTRESET) { callout_stop(&sc->sc_intr_ch); sc->sc_flags |= MTF_ALIVE; } } else { i = gpibrecv(sc->sc_ic, slave, MTT_BCNT, cmdbuf, 2); if (i != 2) { printf("%s intr: can't get xfer length\n", sc->sc_dev.dv_xname); goto error; } i = (int) *((u_short *) cmdbuf); if (i <= bp->b_bcount) { if (i == 0) sc->sc_flags |= MTF_HITEOF; bp->b_resid = bp->b_bcount - i; DPRINTF(MDB_ANY, ("%s intr: bcount %ld, resid %ld", sc->sc_dev.dv_xname, bp->b_bcount, bp->b_resid)); } else { tprintf(sc->sc_ttyp, "%s: record (%d) larger than wanted (%ld)\n", sc->sc_dev.dv_xname, i, bp->b_bcount); error: sc->sc_flags &= ~MTF_IO; bp->b_error = EIO; bp->b_flags |= B_ERROR; } } /* * The operation is completely done. * Let the drive know with an END command. */ cmdbuf[0] = MTE_COMPLETE | MTE_IDLE; (void) gpibsend(sc->sc_ic, slave, MTL_ECMD, cmdbuf, 1); bp->b_flags &= ~B_CMD; (void)BUFQ_GET(sc->sc_tab); biodone(bp); gpibrelease(sc->sc_ic, sc->sc_hdl); if (BUFQ_PEEK(sc->sc_tab) == NULL) sc->sc_active = 0; else mtustart(sc); } int mtread(dev, uio, flags) dev_t dev; struct uio *uio; int flags; { struct mt_softc *sc; sc = device_lookup(&mt_cd, MTUNIT(dev)); return (physio(mtstrategy, &sc->sc_bufstore, dev, B_READ, minphys, uio)); } int mtwrite(dev, uio, flags) dev_t dev; struct uio *uio; int flags; { struct mt_softc *sc; sc = device_lookup(&mt_cd, MTUNIT(dev)); return (physio(mtstrategy, &sc->sc_bufstore, dev, B_WRITE, minphys, uio)); } int mtioctl(dev, cmd, data, flag, p) dev_t dev; u_long cmd; caddr_t data; int flag; struct proc *p; { struct mtop *op; int cnt; switch (cmd) { case MTIOCTOP: op = (struct mtop *)data; switch(op->mt_op) { case MTWEOF: case MTFSF: case MTBSR: case MTBSF: case MTFSR: cnt = op->mt_count; break; case MTOFFL: case MTREW: case MTNOP: cnt = 0; break; default: return (EINVAL); } return (mtcommand(dev, op->mt_op, cnt)); case MTIOCGET: break; default: return (EINVAL); } return (0); }