/* $OpenBSD: aha.c,v 1.73 2010/08/07 03:50:01 krw Exp $ */ /* $NetBSD: aha.c,v 1.11 1996/05/12 23:51:23 mycroft Exp $ */ #undef AHADIAG /* * Copyright (c) 1994, 1996 Charles M. Hannum. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by Charles M. Hannum. * 4. The name of the author may not be used to endorse or promote products * derived from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``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 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. */ /* * Originally written by Julian Elischer (julian@tfs.com) * for TRW Financial Systems for use under the MACH(2.5) operating system. * * TRW Financial Systems, in accordance with their agreement with Carnegie * Mellon University, makes this software available to CMU to distribute * or use in any manner that they see fit as long as this message is kept with * the software. For this reason TFS also grants any other persons or * organisations permission to use or modify this software. * * TFS supplies this software to be publicly redistributed * on the understanding that TFS is not responsible for the correct * functioning of this software in any circumstances. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* XXX fixme: * on i386 at least, xfers to/from user memory * cannot be serviced at interrupt time. */ #ifdef __i386__ #define VOLATILE_XS(xs) \ ((xs)->datalen > 0 && (xs)->bp == NULL && \ ((xs)->flags & SCSI_POLL) == 0) #else #define VOLATILE_XS(xs) 0 #endif /* * Mail box defs etc. * these could be bigger but we need the aha_softc to fit on a single page.. */ #define AHA_MBX_SIZE 16 /* mail box size */ #define AHA_CCB_MAX 16 /* store up to 32 CCBs at one time */ #define CCB_HASH_SIZE 16 /* hash table size for phystokv */ #define CCB_HASH_SHIFT 9 #define CCB_HASH(x) ((((long)(x))>>CCB_HASH_SHIFT) & (CCB_HASH_SIZE - 1)) #define aha_nextmbx(wmb, mbx, mbio) \ if ((wmb) == &(mbx)->mbio[AHA_MBX_SIZE - 1]) \ (wmb) = &(mbx)->mbio[0]; \ else \ (wmb)++; struct aha_mbx { struct aha_mbx_out mbo[AHA_MBX_SIZE]; struct aha_mbx_in mbi[AHA_MBX_SIZE]; struct aha_mbx_out *cmbo; /* Collection Mail Box out */ struct aha_mbx_out *tmbo; /* Target Mail Box out */ struct aha_mbx_in *tmbi; /* Target Mail Box in */ }; struct aha_softc { struct device sc_dev; struct isadev sc_id; void *sc_ih; bus_dma_tag_t sc_dmat; int sc_iobase; int sc_irq, sc_drq; char sc_model[18], sc_firmware[4]; struct aha_mbx *sc_mbx; /* all the mailboxes */ #define wmbx (sc->sc_mbx) struct aha_ccb *sc_ccbhash[CCB_HASH_SIZE]; TAILQ_HEAD(, aha_ccb) sc_free_ccb, sc_waiting_ccb; int sc_numccbs, sc_mbofull; int sc_scsi_dev; /* our scsi id */ struct scsi_link sc_link; }; #ifdef AHADEBUG int aha_debug = 1; #endif /* AHADEBUG */ int aha_cmd(int, struct aha_softc *, int, u_char *, int, u_char *); void aha_finish_ccbs(struct aha_softc *); int ahaintr(void *); void aha_reset_ccb(struct aha_softc *, struct aha_ccb *); void aha_free_ccb(struct aha_softc *, struct aha_ccb *); int aha_init_ccb(struct aha_softc *, struct aha_ccb *, int); struct aha_ccb *aha_get_ccb(struct aha_softc *, int); struct aha_ccb *aha_ccb_phys_kv(struct aha_softc *, u_long); void aha_queue_ccb(struct aha_softc *, struct aha_ccb *); void aha_collect_mbo(struct aha_softc *); void aha_start_ccbs(struct aha_softc *); void aha_done(struct aha_softc *, struct aha_ccb *); int aha_find(struct isa_attach_args *, struct aha_softc *, int); void aha_init(struct aha_softc *); void aha_inquire_setup_information(struct aha_softc *); void ahaminphys(struct buf *, struct scsi_link *); void aha_scsi_cmd(struct scsi_xfer *); int aha_poll(struct aha_softc *, struct scsi_xfer *, int); void aha_timeout(void *arg); struct scsi_adapter aha_switch = { aha_scsi_cmd, ahaminphys, 0, 0, }; int aha_isapnp_probe(struct device *, void *, void *); int ahaprobe(struct device *, void *, void *); void ahaattach(struct device *, struct device *, void *); struct cfattach aha_isapnp_ca = { sizeof(struct aha_softc), aha_isapnp_probe, ahaattach }; struct cfattach aha_isa_ca = { sizeof(struct aha_softc), ahaprobe, ahaattach }; struct cfdriver aha_cd = { NULL, "aha", DV_DULL }; #define AHA_RESET_TIMEOUT 2000 /* time to wait for reset (mSec) */ #define AHA_ABORT_TIMEOUT 2000 /* time to wait for abort (mSec) */ #include "bha.h" /* * aha_cmd(iobase, sc, icnt, ibuf, ocnt, obuf) * * Activate Adapter command * icnt: number of args (outbound bytes including opcode) * ibuf: argument buffer * ocnt: number of expected returned bytes * obuf: result buffer * wait: number of seconds to wait for response * * Performs an adapter command through the ports. Not to be confused with a * scsi command, which is read in via the dma; one of the adapter commands * tells it to read in a scsi command. */ int aha_cmd(iobase, sc, icnt, ibuf, ocnt, obuf) int iobase; struct aha_softc *sc; int icnt, ocnt; u_char *ibuf, *obuf; { const char *name; register int i; int wait; u_char sts; u_char opcode = ibuf[0]; if (sc != NULL) name = sc->sc_dev.dv_xname; else name = "(aha probe)"; /* * Calculate a reasonable timeout for the command. */ switch (opcode) { case AHA_INQUIRE_DEVICES: wait = 15 * 20000; break; default: wait = 1 * 20000; break; } /* * Wait for the adapter to go idle, unless it's one of * the commands which don't need this */ if (opcode != AHA_MBO_INTR_EN) { for (i = 20000; i; i--) { /* 1 sec? */ sts = inb(iobase + AHA_STAT_PORT); if (sts & AHA_STAT_IDLE) break; delay(50); } if (!i) { printf("%s: aha_cmd, host not idle(0x%x)\n", name, sts); return (ENXIO); } } /* * Now that it is idle, if we expect output, preflush the * queue feeding to us. */ if (ocnt) { while ((inb(iobase + AHA_STAT_PORT)) & AHA_STAT_DF) inb(iobase + AHA_DATA_PORT); } /* * Output the command and the number of arguments given * for each byte, first check the port is empty. */ while (icnt--) { for (i = wait; i; i--) { sts = inb(iobase + AHA_STAT_PORT); if (!(sts & AHA_STAT_CDF)) break; delay(50); } if (!i) { if (opcode != AHA_INQUIRE_REVISION) printf("%s: aha_cmd, cmd/data port full\n", name); outb(iobase + AHA_CTRL_PORT, AHA_CTRL_SRST); return (ENXIO); } outb(iobase + AHA_CMD_PORT, *ibuf++); } /* * If we expect input, loop that many times, each time, * looking for the data register to have valid data */ while (ocnt--) { for (i = wait; i; i--) { sts = inb(iobase + AHA_STAT_PORT); if (sts & AHA_STAT_DF) break; delay(50); } if (!i) { if (opcode != AHA_INQUIRE_REVISION) printf("%s: aha_cmd, cmd/data port empty %d\n", name, ocnt); outb(iobase + AHA_CTRL_PORT, AHA_CTRL_SRST); return (ENXIO); } *obuf++ = inb(iobase + AHA_DATA_PORT); } /* * Wait for the board to report a finished instruction. * We may get an extra interrupt for the HACC signal, but this is * unimportant. */ if (opcode != AHA_MBO_INTR_EN) { for (i = 20000; i; i--) { /* 1 sec? */ sts = inb(iobase + AHA_INTR_PORT); /* XXX Need to save this in the interrupt handler? */ if (sts & AHA_INTR_HACC) break; delay(50); } if (!i) { printf("%s: aha_cmd, host not finished(0x%x)\n", name, sts); return (ENXIO); } } outb(iobase + AHA_CTRL_PORT, AHA_CTRL_IRST); return (0); } int aha_isapnp_probe(parent, match, aux) struct device *parent; void *match, *aux; { return (1); } /* * Check if the device can be found at the port given * and if so, set it up ready for further work * as an argument, takes the isa_device structure from * autoconf.c */ int ahaprobe(parent, match, aux) struct device *parent; void *match, *aux; { register struct isa_attach_args *ia = aux; #if NBHA > 0 extern int btports[], nbtports; int i; for (i = 0; i < nbtports; i++) if (btports[i] == ia->ia_iobase) return (0); #endif /* See if there is a unit at this location. */ if (aha_find(ia, NULL, 0) != 0) return (0); ia->ia_msize = 0; ia->ia_iosize = 4; /* IRQ and DRQ set by aha_find(). */ return (1); } /* * Attach all the sub-devices we can find */ void ahaattach(parent, self, aux) struct device *parent, *self; void *aux; { struct isa_attach_args *ia = aux; struct aha_softc *sc = (void *)self; struct scsibus_attach_args saa; int isapnp = !strcmp(parent->dv_cfdata->cf_driver->cd_name, "isapnp"); if (isapnp) { ia->ia_iobase = ia->ipa_io[0].base; isadma_cascade(ia->ia_drq); } if (aha_find(ia, sc, isapnp) != 0) panic("ahaattach: aha_find of %s failed", self->dv_xname); sc->sc_iobase = ia->ia_iobase; sc->sc_dmat = ia->ia_dmat; if (sc->sc_drq != DRQUNK && isapnp == 0) isadma_cascade(sc->sc_drq); aha_inquire_setup_information(sc); aha_init(sc); TAILQ_INIT(&sc->sc_free_ccb); TAILQ_INIT(&sc->sc_waiting_ccb); /* * fill in the prototype scsi_link. */ sc->sc_link.adapter_softc = sc; sc->sc_link.adapter_target = sc->sc_scsi_dev; sc->sc_link.adapter = &aha_switch; sc->sc_link.openings = 2; bzero(&saa, sizeof(saa)); saa.saa_sc_link = &sc->sc_link; sc->sc_ih = isa_intr_establish(ia->ia_ic, sc->sc_irq, IST_EDGE, IPL_BIO, ahaintr, sc, sc->sc_dev.dv_xname); /* * ask the adapter what subunits are present */ config_found(self, &saa, scsiprint); } void aha_finish_ccbs(sc) struct aha_softc *sc; { struct aha_mbx_in *wmbi; struct aha_ccb *ccb; int i; wmbi = wmbx->tmbi; if (wmbi->stat == AHA_MBI_FREE) { for (i = 0; i < AHA_MBX_SIZE; i++) { if (wmbi->stat != AHA_MBI_FREE) { printf("%s: mbi not in round-robin order\n", sc->sc_dev.dv_xname); goto AGAIN; } aha_nextmbx(wmbi, wmbx, mbi); } #ifdef AHADIAGnot printf("%s: mbi interrupt with no full mailboxes\n", sc->sc_dev.dv_xname); #endif return; } AGAIN: do { ccb = aha_ccb_phys_kv(sc, phystol(wmbi->ccb_addr)); if (!ccb) { printf("%s: bad mbi ccb pointer; skipping\n", sc->sc_dev.dv_xname); goto next; } #ifdef AHADEBUG if (aha_debug) { u_char *cp = (u_char *)&ccb->scsi_cmd; printf("op=%x %x %x %x %x %x\n", cp[0], cp[1], cp[2], cp[3], cp[4], cp[5]); printf("stat %x for mbi addr = 0x%08x, ", wmbi->stat, wmbi); printf("ccb addr = 0x%x\n", ccb); } #endif /* AHADEBUG */ switch (wmbi->stat) { case AHA_MBI_OK: case AHA_MBI_ERROR: if ((ccb->flags & CCB_ABORT) != 0) { /* * If we already started an abort, wait for it * to complete before clearing the CCB. We * could instead just clear CCB_SENDING, but * what if the mailbox was already received? * The worst that happens here is that we clear * the CCB a bit later than we need to. BFD. */ goto next; } break; case AHA_MBI_ABORT: case AHA_MBI_UNKNOWN: /* * Even if the CCB wasn't found, we clear it anyway. * See preceding comment. */ break; default: printf("%s: bad mbi status %02x; skipping\n", sc->sc_dev.dv_xname, wmbi->stat); goto next; } if ((ccb->xs->flags & SCSI_POLL) == 0) timeout_del(&ccb->xs->stimeout); bus_dmamap_sync(sc->sc_dmat, ccb->ccb_dmam, 0, ccb->ccb_dmam->dm_mapsize, BUS_DMASYNC_POSTREAD); aha_done(sc, ccb); next: wmbi->stat = AHA_MBI_FREE; aha_nextmbx(wmbi, wmbx, mbi); } while (wmbi->stat != AHA_MBI_FREE); wmbx->tmbi = wmbi; } /* * Catch an interrupt from the adaptor */ int ahaintr(arg) void *arg; { struct aha_softc *sc = arg; int iobase = sc->sc_iobase; u_char sts; #ifdef AHADEBUG if (aha_debug) printf("%s: ahaintr ", sc->sc_dev.dv_xname); #endif /*AHADEBUG */ /* * First acknowlege the interrupt, Then if it's not telling about * a completed operation just return. */ sts = inb(iobase + AHA_INTR_PORT); if ((sts & AHA_INTR_ANYINTR) == 0) return (0); outb(iobase + AHA_CTRL_PORT, AHA_CTRL_IRST); #ifdef AHADIAG /* Make sure we clear CCB_SENDING before finishing a CCB. */ aha_collect_mbo(sc); #endif /* Mail box out empty? */ if (sts & AHA_INTR_MBOA) { struct aha_toggle toggle; toggle.cmd.opcode = AHA_MBO_INTR_EN; toggle.cmd.enable = 0; aha_cmd(iobase, sc, sizeof(toggle.cmd), (u_char *)&toggle.cmd, 0, (u_char *)0); aha_start_ccbs(sc); } /* Mail box in full? */ if (sts & AHA_INTR_MBIF) aha_finish_ccbs(sc); return (1); } void aha_reset_ccb(sc, ccb) struct aha_softc *sc; struct aha_ccb *ccb; { ccb->flags = 0; } /* * A ccb is put onto the free list. */ void aha_free_ccb(sc, ccb) struct aha_softc *sc; struct aha_ccb *ccb; { int s, hashnum; struct aha_ccb **hashccb; s = splbio(); if (ccb->ccb_dmam->dm_segs[0].ds_addr != 0) bus_dmamap_unload(sc->sc_dmat, ccb->ccb_dmam); /* remove from hash table */ hashnum = CCB_HASH(ccb->ccb_dmam->dm_segs[0].ds_addr); hashccb = &sc->sc_ccbhash[hashnum]; while (*hashccb) { if ((*hashccb)->ccb_dmam->dm_segs[0].ds_addr == ccb->ccb_dmam->dm_segs[0].ds_addr) { *hashccb = (*hashccb)->nexthash; break; } hashccb = &(*hashccb)->nexthash; } aha_reset_ccb(sc, ccb); TAILQ_INSERT_HEAD(&sc->sc_free_ccb, ccb, chain); /* * If there were none, wake anybody waiting for one to come free, * starting with queued entries. */ if (TAILQ_NEXT(ccb, chain) == NULL) wakeup(&sc->sc_free_ccb); splx(s); } int aha_init_ccb(sc, ccb, flags) struct aha_softc *sc; struct aha_ccb *ccb; int flags; { int error, wait, state = 0; bzero(ccb, sizeof(struct aha_ccb)); aha_reset_ccb(sc, ccb); wait = (flags & SCSI_NOSLEEP) ? BUS_DMA_NOWAIT : BUS_DMA_WAITOK; /* Create a DMA map for the data area. */ error = bus_dmamap_create(sc->sc_dmat, MAXPHYS, (MAXPHYS / NBPG) + 1, MAXPHYS, 0, wait | BUS_DMA_ALLOCNOW, &ccb->dmam); if (error) goto fail; state++; /* Create a DMA map for the command control block. */ error = bus_dmamap_create(sc->sc_dmat, CCB_PHYS_SIZE, 1, CCB_PHYS_SIZE, 0, wait | BUS_DMA_ALLOCNOW, &ccb->ccb_dmam); if (error) goto fail; return (0); fail: if (state > 0) bus_dmamap_destroy(sc->sc_dmat, ccb->dmam); return (error); } /* * Get a free ccb * * If there are none, see if we can allocate a new one. If so, put it in * the hash table too otherwise either return an error or sleep. */ struct aha_ccb * aha_get_ccb(sc, flags) struct aha_softc *sc; int flags; { struct aha_ccb *ccb; int hashnum, s; s = splbio(); /* * If we can and have to, sleep waiting for one to come free * but only if we can't allocate a new one. */ for (;;) { ccb = TAILQ_FIRST(&sc->sc_free_ccb); if (ccb) { TAILQ_REMOVE(&sc->sc_free_ccb, ccb, chain); break; } if (sc->sc_numccbs < AHA_CCB_MAX) { ccb = malloc(sizeof *ccb, M_DEVBUF, (flags & SCSI_NOSLEEP) ? M_NOWAIT : M_WAITOK); if (ccb == NULL) { printf("%s: can't malloc ccb\n", sc->sc_dev.dv_xname); goto out; } if (aha_init_ccb(sc, ccb, flags) == 0) { sc->sc_numccbs++; break; } free(ccb, M_DEVBUF); ccb = NULL; } if (flags & SCSI_NOSLEEP) goto out; tsleep(&sc->sc_free_ccb, PRIBIO, "ahaccb", 0); } ccb->flags |= CCB_ALLOC; if (bus_dmamap_load(sc->sc_dmat, ccb->ccb_dmam, ccb, CCB_PHYS_SIZE, NULL, BUS_DMA_NOWAIT) != 0) { aha_free_ccb(sc, ccb); ccb = NULL; } else { hashnum = CCB_HASH(ccb->ccb_dmam->dm_segs[0].ds_addr); ccb->nexthash = sc->sc_ccbhash[hashnum]; sc->sc_ccbhash[hashnum] = ccb; } out: splx(s); return (ccb); } /* * Given a physical address, find the ccb that it corresponds to. */ struct aha_ccb * aha_ccb_phys_kv(sc, ccb_phys) struct aha_softc *sc; u_long ccb_phys; { int hashnum = CCB_HASH(ccb_phys); struct aha_ccb *ccb = sc->sc_ccbhash[hashnum]; while (ccb) { if (ccb->ccb_dmam->dm_segs[0].ds_addr == ccb_phys) break; ccb = ccb->nexthash; } return (ccb); } /* * Queue a CCB to be sent to the controller, and send it if possible. */ void aha_queue_ccb(sc, ccb) struct aha_softc *sc; struct aha_ccb *ccb; { TAILQ_INSERT_TAIL(&sc->sc_waiting_ccb, ccb, chain); aha_start_ccbs(sc); } /* * Garbage collect mailboxes that are no longer in use. */ void aha_collect_mbo(sc) struct aha_softc *sc; { struct aha_mbx_out *wmbo; /* Mail Box Out pointer */ #ifdef AHADIAG struct aha_ccb *ccb; #endif wmbo = wmbx->cmbo; while (sc->sc_mbofull > 0) { if (wmbo->cmd != AHA_MBO_FREE) break; #ifdef AHADIAG ccb = aha_ccb_phys_kv(sc, phystol(wmbo->ccb_addr)); if (!ccb) { printf("%s: bad mbo ccb pointer; skipping\n", sc->sc_dev.dv_xname); } else ccb->flags &= ~CCB_SENDING; #endif --sc->sc_mbofull; aha_nextmbx(wmbo, wmbx, mbo); } wmbx->cmbo = wmbo; } /* * Send as many CCBs as we have empty mailboxes for. */ void aha_start_ccbs(sc) struct aha_softc *sc; { int iobase = sc->sc_iobase; struct aha_mbx_out *wmbo; /* Mail Box Out pointer */ struct aha_ccb *ccb; wmbo = wmbx->tmbo; while ((ccb = TAILQ_FIRST(&sc->sc_waiting_ccb)) != NULL) { if (sc->sc_mbofull >= AHA_MBX_SIZE) { aha_collect_mbo(sc); if (sc->sc_mbofull >= AHA_MBX_SIZE) { struct aha_toggle toggle; toggle.cmd.opcode = AHA_MBO_INTR_EN; toggle.cmd.enable = 1; aha_cmd(iobase, sc, sizeof(toggle.cmd), (u_char *)&toggle.cmd, 0, (u_char *)0); break; } } TAILQ_REMOVE(&sc->sc_waiting_ccb, ccb, chain); #ifdef AHADIAG ccb->flags |= CCB_SENDING; #endif /* Link ccb to mbo. */ bus_dmamap_sync(sc->sc_dmat, ccb->ccb_dmam, 0, ccb->ccb_dmam->dm_mapsize, BUS_DMASYNC_PREWRITE); ltophys(ccb->ccb_dmam->dm_segs[0].ds_addr, wmbo->ccb_addr); if (ccb->flags & CCB_ABORT) wmbo->cmd = AHA_MBO_ABORT; else wmbo->cmd = AHA_MBO_START; /* Tell the card to poll immediately. */ outb(iobase + AHA_CMD_PORT, AHA_START_SCSI); if ((ccb->xs->flags & SCSI_POLL) == 0) { timeout_set(&ccb->xs->stimeout, aha_timeout, ccb); timeout_add_msec(&ccb->xs->stimeout, ccb->timeout); } ++sc->sc_mbofull; aha_nextmbx(wmbo, wmbx, mbo); } wmbx->tmbo = wmbo; } /* * We have a ccb which has been processed by the * adaptor, now we look to see how the operation * went. Wake up the owner if waiting */ void aha_done(sc, ccb) struct aha_softc *sc; struct aha_ccb *ccb; { struct scsi_sense_data *s1, *s2; struct scsi_xfer *xs = ccb->xs; SC_DEBUG(xs->sc_link, SDEV_DB2, ("aha_done\n")); /* * Otherwise, put the results of the operation * into the xfer and call whoever started it */ #ifdef AHADIAG if (ccb->flags & CCB_SENDING) panic("%s: exiting ccb still in transit!", sc->sc_dev.dv_xname); #endif if ((ccb->flags & CCB_ALLOC) == 0) panic("%s: exiting ccb not allocated!", sc->sc_dev.dv_xname); if (xs->error == XS_NOERROR) { if (ccb->host_stat != AHA_OK) { switch (ccb->host_stat) { case AHA_SEL_TIMEOUT: /* No response */ xs->error = XS_SELTIMEOUT; break; default: /* Other scsi protocol messes */ printf("%s: host_stat %x\n", sc->sc_dev.dv_xname, ccb->host_stat); xs->error = XS_DRIVER_STUFFUP; break; } } else if (ccb->target_stat != SCSI_OK) { switch (ccb->target_stat) { case SCSI_CHECK: s1 = (struct scsi_sense_data *) (((char *)(&ccb->scsi_cmd)) + ccb->scsi_cmd_length); s2 = &xs->sense; *s2 = *s1; xs->error = XS_SENSE; break; case SCSI_BUSY: xs->error = XS_BUSY; break; default: printf("%s: target_stat %x\n", sc->sc_dev.dv_xname, ccb->target_stat); xs->error = XS_DRIVER_STUFFUP; break; } } else xs->resid = 0; } if (VOLATILE_XS(xs)) { wakeup(ccb); return; } if (ccb->dmam->dm_nsegs > 0) { if (xs->flags & SCSI_DATA_IN) bus_dmamap_sync(sc->sc_dmat, ccb->dmam, 0, ccb->dmam->dm_mapsize, BUS_DMASYNC_POSTREAD); if (xs->flags & SCSI_DATA_OUT) bus_dmamap_sync(sc->sc_dmat, ccb->dmam, 0, ccb->dmam->dm_mapsize, BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(sc->sc_dmat, ccb->dmam); } aha_free_ccb(sc, ccb); scsi_done(xs); } /* * Find the board and find its irq/drq */ int aha_find(ia, sc, isapnp) struct isa_attach_args *ia; struct aha_softc *sc; int isapnp; { int iobase = ia->ia_iobase; int i; u_char sts; struct aha_config config; int irq, drq; /* * reset board, If it doesn't respond, assume * that it's not there.. good for the probe */ outb(iobase + AHA_CTRL_PORT, AHA_CTRL_HRST | AHA_CTRL_SRST); delay(100); for (i = AHA_RESET_TIMEOUT; i; i--) { sts = inb(iobase + AHA_STAT_PORT); if (sts == (AHA_STAT_IDLE | AHA_STAT_INIT)) break; delay(1000); /* calibrated in msec */ } if (!i) { #ifdef AHADEBUG if (aha_debug) printf("aha_find: No answer from adaptec board\n"); #endif /* AHADEBUG */ return (1); } /* * setup dma channel from jumpers and save int * level */ delay(1000); /* for Bustek 545 */ config.cmd.opcode = AHA_INQUIRE_CONFIG; aha_cmd(iobase, sc, sizeof(config.cmd), (u_char *)&config.cmd, sizeof(config.reply), (u_char *)&config.reply); switch (config.reply.chan) { case EISADMA: drq = DRQUNK; /* for EISA/VLB/PCI clones */ break; case CHAN0: drq = 0; break; case CHAN5: drq = 5; break; case CHAN6: drq = 6; break; case CHAN7: drq = 7; break; default: printf("aha_find: illegal drq setting %x\n", config.reply.chan); return (1); } if (isapnp) irq = ia->ia_irq; switch (config.reply.intr) { case INT9: irq = 9; break; case INT10: irq = 10; break; case INT11: irq = 11; break; case INT12: irq = 12; break; case INT14: irq = 14; break; case INT15: irq = 15; break; default: printf("aha_find: illegal irq setting %x\n", config.reply.intr); return (EIO); } if (isapnp) drq = ia->ia_drq; if (sc != NULL) { /* who are we on the scsi bus? */ sc->sc_scsi_dev = config.reply.scsi_dev; sc->sc_iobase = iobase; sc->sc_irq = irq; sc->sc_drq = drq; } else { if (isapnp) return (0); if (ia->ia_irq == IRQUNK) ia->ia_irq = irq; else if (ia->ia_irq != irq) return (1); if (ia->ia_drq == DRQUNK) ia->ia_drq = drq; else if (ia->ia_drq != drq) return (1); } return (0); } /* * Start the board, ready for normal operation */ void aha_init(sc) struct aha_softc *sc; { int iobase = sc->sc_iobase; struct aha_devices devices; struct aha_setup setup; struct aha_mailbox mailbox; struct pglist pglist; struct vm_page *pg; vaddr_t va; vsize_t size; int i; /* * XXX * If we are a 1542C or later, disable the extended BIOS so that the * mailbox interface is unlocked. * No need to check the extended BIOS flags as some of the * extensions that cause us problems are not flagged in that byte. */ if (!strncmp(sc->sc_model, "1542C", 5)) { struct aha_extbios extbios; struct aha_unlock unlock; printf("%s: unlocking mailbox interface\n", sc->sc_dev.dv_xname); extbios.cmd.opcode = AHA_EXT_BIOS; aha_cmd(iobase, sc, sizeof(extbios.cmd), (u_char *)&extbios.cmd, sizeof(extbios.reply), (u_char *)&extbios.reply); #ifdef AHADEBUG printf("%s: flags=%02x, mailboxlock=%02x\n", sc->sc_dev.dv_xname, extbios.reply.flags, extbios.reply.mailboxlock); #endif /* AHADEBUG */ unlock.cmd.opcode = AHA_MBX_ENABLE; unlock.cmd.junk = 0; unlock.cmd.magic = extbios.reply.mailboxlock; aha_cmd(iobase, sc, sizeof(unlock.cmd), (u_char *)&unlock.cmd, 0, (u_char *)0); } #if 0 /* * Change the bus on/off times to not clash with other dma users. */ aha_cmd(sc, 1, 0, 0, 0, AHA_BUS_ON_TIME_SET, 7); aha_cmd(sc, 1, 0, 0, 0, AHA_BUS_OFF_TIME_SET, 4); #endif /* Inquire Installed Devices (to force synchronous negotiation). */ devices.cmd.opcode = AHA_INQUIRE_DEVICES; aha_cmd(iobase, sc, sizeof(devices.cmd), (u_char *)&devices.cmd, sizeof(devices.reply), (u_char *)&devices.reply); /* Obtain setup information from. */ setup.cmd.opcode = AHA_INQUIRE_SETUP; setup.cmd.len = sizeof(setup.reply); aha_cmd(iobase, sc, sizeof(setup.cmd), (u_char *)&setup.cmd, sizeof(setup.reply), (u_char *)&setup.reply); printf("%s: %s, %s\n", sc->sc_dev.dv_xname, setup.reply.sync_neg ? "sync" : "async", setup.reply.parity ? "parity" : "no parity"); for (i = 0; i < 8; i++) { if (!setup.reply.sync[i].valid || (!setup.reply.sync[i].offset && !setup.reply.sync[i].period)) continue; printf("%s targ %d: sync, offset %d, period %dnsec\n", sc->sc_dev.dv_xname, i, setup.reply.sync[i].offset, setup.reply.sync[i].period * 50 + 200); } /* * Set up initial mail box for round-robin operation. */ /* * XXX - this vm juggling is so wrong. use bus_dma instead! */ size = round_page(sizeof(struct aha_mbx)); TAILQ_INIT(&pglist); if (uvm_pglistalloc(size, 0, 0xffffff, PAGE_SIZE, 0, &pglist, 1, UVM_PLA_NOWAIT) || uvm_map(kernel_map, &va, size, NULL, UVM_UNKNOWN_OFFSET, 0, UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_NONE, UVM_ADV_RANDOM, 0))) panic("aha_init: could not allocate mailbox"); wmbx = (struct aha_mbx *)va; for (pg = TAILQ_FIRST(&pglist); pg != NULL; pg = TAILQ_NEXT(pg, pageq)) { pmap_kenter_pa(va, VM_PAGE_TO_PHYS(pg), VM_PROT_READ|VM_PROT_WRITE); va += PAGE_SIZE; } pmap_update(pmap_kernel()); /* * XXXEND */ for (i = 0; i < AHA_MBX_SIZE; i++) { wmbx->mbo[i].cmd = AHA_MBO_FREE; wmbx->mbi[i].stat = AHA_MBI_FREE; } wmbx->cmbo = wmbx->tmbo = &wmbx->mbo[0]; wmbx->tmbi = &wmbx->mbi[0]; sc->sc_mbofull = 0; /* Initialize mail box. */ mailbox.cmd.opcode = AHA_MBX_INIT; mailbox.cmd.nmbx = AHA_MBX_SIZE; ltophys(vtophys((vaddr_t)wmbx), mailbox.cmd.addr); aha_cmd(iobase, sc, sizeof(mailbox.cmd), (u_char *)&mailbox.cmd, 0, (u_char *)0); } void aha_inquire_setup_information(sc) struct aha_softc *sc; { int iobase = sc->sc_iobase; struct aha_revision revision; u_char sts; int i; char *p; strlcpy(sc->sc_model, "unknown", sizeof sc->sc_model); /* * Assume we have a board at this stage, do an adapter inquire * to find out what type of controller it is. If the command * fails, we assume it's either a crusty board or an old 1542 * clone, and skip the board-specific stuff. */ revision.cmd.opcode = AHA_INQUIRE_REVISION; if (aha_cmd(iobase, sc, sizeof(revision.cmd), (u_char *)&revision.cmd, sizeof(revision.reply), (u_char *)&revision.reply)) { /* * aha_cmd() already started the reset. It's not clear we * even need to bother here. */ for (i = AHA_RESET_TIMEOUT; i; i--) { sts = inb(iobase + AHA_STAT_PORT); if (sts == (AHA_STAT_IDLE | AHA_STAT_INIT)) break; delay(1000); } if (!i) { #ifdef AHADEBUG printf("aha_init: soft reset failed\n"); #endif /* AHADEBUG */ return; } #ifdef AHADEBUG printf("aha_init: inquire command failed\n"); #endif /* AHADEBUG */ goto noinquire; } #ifdef AHADEBUG printf("%s: inquire %x, %x, %x, %x\n", sc->sc_dev.dv_xname, revision.reply.boardid, revision.reply.spec_opts, revision.reply.revision_1, revision.reply.revision_2); #endif /* AHADEBUG */ switch (revision.reply.boardid) { case 0x31: strlcpy(sc->sc_model, "1540", sizeof sc->sc_model); break; case 0x41: strlcpy(sc->sc_model, "1540A/1542A/1542B", sizeof sc->sc_model); break; case 0x42: strlcpy(sc->sc_model, "1640", sizeof sc->sc_model); break; case 0x43: case 0x44: /* Is this 1542C or -CF? */ strlcpy(sc->sc_model, "1542C", sizeof sc->sc_model); break; case 0x45: strlcpy(sc->sc_model, "1542CF", sizeof sc->sc_model); break; case 0x46: strlcpy(sc->sc_model, "1542CP", sizeof sc->sc_model); break; } p = sc->sc_firmware; *p++ = revision.reply.revision_1; *p++ = '.'; *p++ = revision.reply.revision_2; *p = '\0'; noinquire: printf(": model AHA-%s, firmware %s\n", sc->sc_model, sc->sc_firmware); } void ahaminphys(struct buf *bp, struct scsi_link *sl) { if (bp->b_bcount > ((AHA_NSEG - 1) << PGSHIFT)) bp->b_bcount = ((AHA_NSEG - 1) << PGSHIFT); minphys(bp); } /* * start a scsi operation given the command and the data address. Also needs * the unit, target and lu. */ void aha_scsi_cmd(xs) struct scsi_xfer *xs; { struct scsi_link *sc_link = xs->sc_link; struct aha_softc *sc = sc_link->adapter_softc; struct aha_ccb *ccb; struct aha_scat_gath *sg; int seg, flags; int s; SC_DEBUG(sc_link, SDEV_DB2, ("aha_scsi_cmd\n")); /* * get a ccb to use. If the transfer * is from a buf (possibly from interrupt time) * then we can't allow it to sleep */ flags = xs->flags; if ((ccb = aha_get_ccb(sc, flags)) == NULL) { xs->error = XS_NO_CCB; scsi_done(xs); return; } ccb->xs = xs; ccb->timeout = xs->timeout; /* * Put all the arguments for the xfer in the ccb */ if (flags & SCSI_RESET) { ccb->opcode = AHA_RESET_CCB; ccb->scsi_cmd_length = 0; } else { /* can't use S/G if zero length */ ccb->opcode = (xs->datalen ? AHA_INIT_SCAT_GATH_CCB : AHA_INITIATOR_CCB); bcopy(xs->cmd, &ccb->scsi_cmd, ccb->scsi_cmd_length = xs->cmdlen); } if (xs->datalen) { sg = ccb->scat_gath; seg = 0; /* * Set up the scatter-gather block. */ if (bus_dmamap_load(sc->sc_dmat, ccb->dmam, xs->data, xs->datalen, NULL, BUS_DMA_NOWAIT) != 0) { aha_free_ccb(sc, ccb); xs->error = XS_NO_CCB; scsi_done(xs); return; } for (seg = 0; seg < ccb->dmam->dm_nsegs; seg++) { ltophys(ccb->dmam->dm_segs[seg].ds_addr, sg[seg].seg_addr); ltophys(ccb->dmam->dm_segs[seg].ds_len, sg[seg].seg_len); } if (flags & SCSI_DATA_OUT) bus_dmamap_sync(sc->sc_dmat, ccb->dmam, 0, ccb->dmam->dm_mapsize, BUS_DMASYNC_PREWRITE); if (flags & SCSI_DATA_IN) bus_dmamap_sync(sc->sc_dmat, ccb->dmam, 0, ccb->dmam->dm_mapsize, BUS_DMASYNC_PREREAD); ltophys((unsigned) ((struct aha_ccb *)(ccb->ccb_dmam->dm_segs[0].ds_addr))-> scat_gath, ccb->data_addr); ltophys(ccb->dmam->dm_nsegs * sizeof(struct aha_scat_gath), ccb->data_length); } else { /* No data xfer, use non S/G values */ ltophys(0, ccb->data_addr); ltophys(0, ccb->data_length); } ccb->data_out = 0; ccb->data_in = 0; ccb->target = sc_link->target; ccb->lun = sc_link->lun; ccb->req_sense_length = sizeof(ccb->scsi_sense); ccb->host_stat = 0x00; ccb->target_stat = 0x00; ccb->link_id = 0; ltophys(0, ccb->link_addr); s = splbio(); aha_queue_ccb(sc, ccb); /* * Usually return SUCCESSFULLY QUEUED */ SC_DEBUG(sc_link, SDEV_DB3, ("cmd_sent\n")); if (VOLATILE_XS(xs)) { while ((ccb->xs->flags & ITSDONE) == 0) { tsleep(ccb, PRIBIO, "ahawait", 0); } if (ccb->dmam->dm_nsegs > 0) { if (flags & SCSI_DATA_OUT) bus_dmamap_sync(sc->sc_dmat, ccb->dmam, 0, ccb->dmam->dm_mapsize, BUS_DMASYNC_POSTWRITE); if (flags & SCSI_DATA_IN) bus_dmamap_sync(sc->sc_dmat, ccb->dmam, 0, ccb->dmam->dm_mapsize, BUS_DMASYNC_POSTREAD); bus_dmamap_unload(sc->sc_dmat, ccb->dmam); } aha_free_ccb(sc, ccb); scsi_done(xs); splx(s); return; } splx(s); if ((flags & SCSI_POLL) == 0) return; /* * If we can't use interrupts, poll on completion */ if (aha_poll(sc, xs, ccb->timeout)) { aha_timeout(ccb); if (aha_poll(sc, xs, ccb->timeout)) aha_timeout(ccb); } } /* * Poll a particular unit, looking for a particular xs */ int aha_poll(sc, xs, count) struct aha_softc *sc; struct scsi_xfer *xs; int count; { int iobase = sc->sc_iobase; int s; /* timeouts are in msec, so we loop in 1000 usec cycles */ while (count) { /* * If we had interrupts enabled, would we * have got an interrupt? */ if (inb(iobase + AHA_INTR_PORT) & AHA_INTR_ANYINTR) { s = splbio(); ahaintr(sc); splx(s); } if (xs->flags & ITSDONE) return (0); delay(1000); /* only happens in boot so ok */ count--; } return (1); } void aha_timeout(arg) void *arg; { struct aha_ccb *ccb = arg; struct scsi_xfer *xs; struct scsi_link *sc_link; struct aha_softc *sc; int s; s = splbio(); xs = ccb->xs; sc_link = xs->sc_link; sc = sc_link->adapter_softc; bus_dmamap_sync(sc->sc_dmat, ccb->ccb_dmam, 0, ccb->ccb_dmam->dm_mapsize, BUS_DMASYNC_POSTREAD); sc_print_addr(sc_link); printf("timed out"); #ifdef AHADIAG /* * If The ccb's mbx is not free, then the board has gone south? */ aha_collect_mbo(sc); if (ccb->flags & CCB_SENDING) panic("%s: not taking commands!", sc->sc_dev.dv_xname); #endif /* * If it has been through before, then * a previous abort has failed, don't * try abort again */ if (ccb->flags & CCB_ABORT) { /* abort timed out */ printf(" AGAIN\n"); /* XXX Must reset! */ } else { /* abort the operation that has timed out */ printf("\n"); ccb->xs->error = XS_TIMEOUT; ccb->timeout = AHA_ABORT_TIMEOUT; ccb->flags |= CCB_ABORT; aha_queue_ccb(sc, ccb); } splx(s); }