/* $NetBSD: xd.c,v 1.9 1998/02/04 05:15:16 thorpej Exp $ */ /* * * Copyright (c) 1995 Charles D. Cranor * 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 D. Cranor. * 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. */ /* * * x d . c x y l o g i c s 7 5 3 / 7 0 5 3 v m e / s m d d r i v e r * * author: Chuck Cranor * started: 27-Feb-95 * references: [1] Xylogics Model 753 User's Manual * part number: 166-753-001, Revision B, May 21, 1988. * "Your Partner For Performance" * [2] other NetBSD disk device drivers * * Special thanks go to Scott E. Campbell of Xylogics, Inc. for taking * the time to answer some of my questions about the 753/7053. * * note: the 753 and the 7053 are programmed the same way, but are * different sizes. the 753 is a 6U VME card, while the 7053 is a 9U * VME card (found in many VME based suns). */ #undef XDC_DEBUG /* full debug */ #define XDC_DIAG /* extra sanity checks */ #if defined(DIAGNOSTIC) && !defined(XDC_DIAG) #define XDC_DIAG /* link in with master DIAG option */ #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "locators.h" /* * macros */ /* * XDC_TWAIT: add iorq "N" to tail of SC's wait queue */ #define XDC_TWAIT(SC, N) { \ (SC)->waitq[(SC)->waitend] = (N); \ (SC)->waitend = ((SC)->waitend + 1) % XDC_MAXIOPB; \ (SC)->nwait++; \ } /* * XDC_HWAIT: add iorq "N" to head of SC's wait queue */ #define XDC_HWAIT(SC, N) { \ (SC)->waithead = ((SC)->waithead == 0) ? \ (XDC_MAXIOPB - 1) : ((SC)->waithead - 1); \ (SC)->waitq[(SC)->waithead] = (N); \ (SC)->nwait++; \ } /* * XDC_GET_WAITER: gets the first request waiting on the waitq * and removes it (so it can be submitted) */ #define XDC_GET_WAITER(XDCSC, RQ) { \ (RQ) = (XDCSC)->waitq[(XDCSC)->waithead]; \ (XDCSC)->waithead = ((XDCSC)->waithead + 1) % XDC_MAXIOPB; \ xdcsc->nwait--; \ } /* * XDC_FREE: add iorq "N" to SC's free list */ #define XDC_FREE(SC, N) { \ (SC)->freereq[(SC)->nfree++] = (N); \ (SC)->reqs[N].mode = 0; \ if ((SC)->nfree == 1) wakeup(&(SC)->nfree); \ } /* * XDC_RQALLOC: allocate an iorq off the free list (assume nfree > 0). */ #define XDC_RQALLOC(XDCSC) (XDCSC)->freereq[--((XDCSC)->nfree)] /* * XDC_GO: start iopb ADDR (DVMA addr in a u_long) on XDC */ #define XDC_GO(XDC, ADDR) { \ (XDC)->xdc_iopbaddr0 = ((ADDR) & 0xff); \ (ADDR) = ((ADDR) >> 8); \ (XDC)->xdc_iopbaddr1 = ((ADDR) & 0xff); \ (ADDR) = ((ADDR) >> 8); \ (XDC)->xdc_iopbaddr2 = ((ADDR) & 0xff); \ (ADDR) = ((ADDR) >> 8); \ (XDC)->xdc_iopbaddr3 = (ADDR); \ (XDC)->xdc_iopbamod = XDC_ADDRMOD; \ (XDC)->xdc_csr = XDC_ADDIOPB; /* go! */ \ } /* * XDC_WAIT: wait for XDC's csr "BITS" to come on in "TIME". * LCV is a counter. If it goes to zero then we timed out. */ #define XDC_WAIT(XDC, LCV, TIME, BITS) { \ (LCV) = (TIME); \ while ((LCV) > 0) { \ if ((XDC)->xdc_csr & (BITS)) break; \ (LCV) = (LCV) - 1; \ DELAY(1); \ } \ } /* * XDC_DONE: don't need IORQ, get error code and free (done after xdc_cmd) */ #define XDC_DONE(SC,RQ,ER) { \ if ((RQ) == XD_ERR_FAIL) { \ (ER) = (RQ); \ } else { \ if ((SC)->ndone-- == XDC_SUBWAITLIM) \ wakeup(&(SC)->ndone); \ (ER) = (SC)->reqs[RQ].errno; \ XDC_FREE((SC), (RQ)); \ } \ } /* * XDC_ADVANCE: advance iorq's pointers by a number of sectors */ #define XDC_ADVANCE(IORQ, N) { \ if (N) { \ (IORQ)->sectcnt -= (N); \ (IORQ)->blockno += (N); \ (IORQ)->dbuf += ((N)*XDFM_BPS); \ } \ } /* * note - addresses you can sleep on: * [1] & of xd_softc's "state" (waiting for a chance to attach a drive) * [2] & of xdc_softc's "nfree" (waiting for a free iorq/iopb) * [3] & of xdc_softc's "ndone" (waiting for number of done iorq/iopb's * to drop below XDC_SUBWAITLIM) * [4] & an iorq (waiting for an XD_SUB_WAIT iorq to finish) */ /* * function prototypes * "xdc_*" functions are internal, all others are external interfaces */ extern int pil_to_vme[]; /* from obio.c */ /* internals */ int xdc_cmd __P((struct xdc_softc *, int, int, int, int, int, char *, int)); char *xdc_e2str __P((int)); int xdc_error __P((struct xdc_softc *, struct xd_iorq *, struct xd_iopb *, int, int)); int xdc_ioctlcmd __P((struct xd_softc *, dev_t dev, struct xd_iocmd *)); void xdc_perror __P((struct xd_iorq *, struct xd_iopb *, int)); int xdc_piodriver __P((struct xdc_softc *, int, int)); int xdc_remove_iorq __P((struct xdc_softc *)); int xdc_reset __P((struct xdc_softc *, int, int, int, struct xd_softc *)); inline void xdc_rqinit __P((struct xd_iorq *, struct xdc_softc *, struct xd_softc *, int, u_long, int, caddr_t, struct buf *)); void xdc_rqtopb __P((struct xd_iorq *, struct xd_iopb *, int, int)); void xdc_start __P((struct xdc_softc *, int)); int xdc_startbuf __P((struct xdc_softc *, struct xd_softc *, struct buf *)); int xdc_submit_iorq __P((struct xdc_softc *, int, int)); void xdc_tick __P((void *)); void xdc_xdreset __P((struct xdc_softc *, struct xd_softc *)); /* machine interrupt hook */ int xdcintr __P((void *)); /* autoconf */ int xdcmatch __P((struct device *, struct cfdata *, void *)); void xdcattach __P((struct device *, struct device *, void *)); int xdmatch __P((struct device *, struct cfdata *, void *)); void xdattach __P((struct device *, struct device *, void *)); static int xdc_probe __P((void *, void *)); static void xddummystrat __P((struct buf *)); int xdgetdisklabel __P((struct xd_softc *, void *)); /* XXX - think about this more.. xd_machdep? */ void md_setup __P((void)); int XDC_DELAY; #ifdef sparc #include #include void md_setup() { if (CPU_ISSUN4 && cpuinfo.cpu_type == CPUTYP_4_300) XDC_DELAY = XDC_DELAY_4_300; else XDC_DELAY = XDC_DELAY_SPARC; } #endif #ifdef sun3 void md_setup() { XDC_DELAY = XDC_DELAY_SUN3; } #endif /* * cfattach's: device driver interface to autoconfig */ struct cfattach xdc_ca = { sizeof(struct xdc_softc), xdcmatch, xdcattach }; struct cfattach xd_ca = { sizeof(struct xd_softc), xdmatch, xdattach }; extern struct cfdriver xd_cd; struct xdc_attach_args { /* this is the "aux" args to xdattach */ int driveno; /* unit number */ int fullmode; /* submit mode */ int booting; /* are we booting or not? */ }; /* * dkdriver */ struct dkdriver xddkdriver = {xdstrategy}; /* * start: disk label fix code (XXX) */ static void *xd_labeldata; static void xddummystrat(bp) struct buf *bp; { if (bp->b_bcount != XDFM_BPS) panic("xddummystrat"); bcopy(xd_labeldata, bp->b_un.b_addr, XDFM_BPS); bp->b_flags |= B_DONE; bp->b_flags &= ~B_BUSY; } int xdgetdisklabel(xd, b) struct xd_softc *xd; void *b; { char *err; struct sun_disklabel *sdl; /* We already have the label data in `b'; setup for dummy strategy */ xd_labeldata = b; /* Required parameter for readdisklabel() */ xd->sc_dk.dk_label->d_secsize = XDFM_BPS; err = readdisklabel(MAKEDISKDEV(0, xd->sc_dev.dv_unit, RAW_PART), xddummystrat, xd->sc_dk.dk_label, xd->sc_dk.dk_cpulabel); if (err) { printf("%s: %s\n", xd->sc_dev.dv_xname, err); return(XD_ERR_FAIL); } /* Ok, we have the label; fill in `pcyl' if there's SunOS magic */ sdl = (struct sun_disklabel *)xd->sc_dk.dk_cpulabel->cd_block; if (sdl->sl_magic == SUN_DKMAGIC) xd->pcyl = sdl->sl_pcylinders; else { printf("%s: WARNING: no `pcyl' in disk label.\n", xd->sc_dev.dv_xname); xd->pcyl = xd->sc_dk.dk_label->d_ncylinders + xd->sc_dk.dk_label->d_acylinders; printf("%s: WARNING: guessing pcyl=%d (ncyl+acyl)\n", xd->sc_dev.dv_xname, xd->pcyl); } xd->ncyl = xd->sc_dk.dk_label->d_ncylinders; xd->acyl = xd->sc_dk.dk_label->d_acylinders; xd->nhead = xd->sc_dk.dk_label->d_ntracks; xd->nsect = xd->sc_dk.dk_label->d_nsectors; xd->sectpercyl = xd->nhead * xd->nsect; xd->sc_dk.dk_label->d_secsize = XDFM_BPS; /* not handled by * sun->bsd */ return(XD_ERR_AOK); } /* * end: disk label fix code (XXX) */ /* * a u t o c o n f i g f u n c t i o n s */ /* * xdcmatch: determine if xdc is present or not. we do a * soft reset to detect the xdc. */ int xdc_probe(vaddr, arg) void *vaddr; void *arg; { struct xdc *xdc = vaddr; int del = 0; xdc->xdc_csr = XDC_RESET; XDC_WAIT(xdc, del, XDC_RESETUSEC, XDC_RESET); return (del > 0); } int xdcmatch(parent, cf, aux) struct device *parent; struct cfdata *cf; void *aux; { struct vme_attach_args *va = aux; vme_chipset_tag_t ct = va->vma_chipset_tag; bus_space_tag_t bt = va->vma_bustag; vme_mod_t mod; vme_addr_t vaddr; mod = VMEMOD_A16 | VMEMOD_S | VMEMOD_D | VMEMOD_D32; vaddr = va->vma_reg[0] + offsetof(struct xdc, xdc_csr); return (vme_bus_probe(ct, bt, va->vma_reg[0], 1, mod, xdc_probe, 0)); } /* * xdcattach: attach controller */ void xdcattach(parent, self, aux) struct device *parent, *self; void *aux; { struct vme_attach_args *va = aux; vme_chipset_tag_t ct = va->vma_chipset_tag; bus_space_tag_t bt = va->vma_bustag; bus_space_handle_t bh; vme_intr_handle_t ih; vme_mod_t mod; struct xdc_softc *xdc = (void *) self; struct xdc_attach_args xa; int lcv, rqno, error; struct xd_iopb_ctrl *ctl; bus_dma_segment_t seg; int rseg; md_setup(); /* get addressing and intr level stuff from autoconfig and load it * into our xdc_softc. */ xdc->dmatag = va->vma_dmatag; mod = VMEMOD_A16 | VMEMOD_S | VMEMOD_D | VMEMOD_D32; if (vme_bus_map(ct, va->vma_reg[0], sizeof(struct xdc), mod, bt, &bh) != 0) panic("xdc: vme_map"); xdc->xdc = (struct xdc *) bh; xdc->ipl = va->vma_pri; xdc->vector = va->vma_vec; printf(" pri %d", va->vma_pri); for (lcv = 0; lcv < XDC_MAXDEV; lcv++) xdc->sc_drives[lcv] = (struct xd_softc *) 0; /* allocate and zero buffers * * note: we simplify the code by allocating the max number of iopbs and * iorq's up front. thus, we avoid linked lists and the costs * associated with them in exchange for wasting a little memory. */ error = bus_dmamem_alloc(xdc->dmatag, XDC_MAXIOPB * sizeof(struct xd_iopb), NBPG, 0, &seg, 1, &rseg, BUS_DMA_NOWAIT); if (error) { printf("%s: DMA buffer alloc error %d\n", xdc->sc_dev.dv_xname, error); return; } xdc->dvmaiopb = (struct xd_iopb *)seg.ds_addr; error = bus_dmamem_map(xdc->dmatag, &seg, rseg, XDC_MAXIOPB * sizeof(struct xd_iopb), (caddr_t *)&xdc->iopbase, BUS_DMA_NOWAIT|BUS_DMA_COHERENT); if (error) { bus_dmamem_free(xdc->dmatag, &seg, rseg); printf("%s: DMA buffer map error %d\n", xdc->sc_dev.dv_xname, error); return; } bzero(xdc->iopbase, XDC_MAXIOPB * sizeof(struct xd_iopb)); xdc->reqs = (struct xd_iorq *) malloc(XDC_MAXIOPB * sizeof(struct xd_iorq), M_DEVBUF, M_NOWAIT); if (xdc->reqs == NULL) panic("xdc malloc"); bzero(xdc->reqs, XDC_MAXIOPB * sizeof(struct xd_iorq)); /* init free list, iorq to iopb pointers, and non-zero fields in the * iopb which never change. */ for (lcv = 0; lcv < XDC_MAXIOPB; lcv++) { xdc->reqs[lcv].iopb = &xdc->iopbase[lcv]; xdc->freereq[lcv] = lcv; xdc->iopbase[lcv].fixd = 1; /* always the same */ xdc->iopbase[lcv].naddrmod = XDC_ADDRMOD; /* always the same */ xdc->iopbase[lcv].intr_vec = xdc->vector; /* always the same */ error = bus_dmamap_create( xdc->dmatag, MAXPHYS, /* size */ 1, /* nsegments */ MAXPHYS, /* maxsegsz */ 0, /* boundary */ BUS_DMA_NOWAIT, &xdc->reqs[lcv].dmamap); if (error) { printf("%s: DMA buffer map create error %d\n", xdc->sc_dev.dv_xname, error); return; } } xdc->nfree = XDC_MAXIOPB; xdc->nrun = 0; xdc->waithead = xdc->waitend = xdc->nwait = 0; xdc->ndone = 0; /* init queue of waiting bufs */ xdc->sc_wq.b_active = 0; xdc->sc_wq.b_actf = 0; xdc->sc_wq.b_actb = &xdc->sc_wq.b_actf; /* * section 7 of the manual tells us how to init the controller: * - read controller parameters (6/0) * - write controller parameters (5/0) */ /* read controller parameters and insure we have a 753/7053 */ rqno = xdc_cmd(xdc, XDCMD_RDP, XDFUN_CTL, 0, 0, 0, 0, XD_SUB_POLL); if (rqno == XD_ERR_FAIL) { printf(": couldn't read controller params\n"); return; /* shouldn't ever happen */ } ctl = (struct xd_iopb_ctrl *) & xdc->iopbase[rqno]; if (ctl->ctype != XDCT_753) { if (xdc->reqs[rqno].errno) printf(": %s: ", xdc_e2str(xdc->reqs[rqno].errno)); printf(": doesn't identify as a 753/7053\n"); XDC_DONE(xdc, rqno, error); return; } printf(": Xylogics 753/7053, PROM=0x%x.%02x.%02x\n", ctl->eprom_partno, ctl->eprom_lvl, ctl->eprom_rev); XDC_DONE(xdc, rqno, error); /* now write controller parameters (xdc_cmd sets all params for us) */ rqno = xdc_cmd(xdc, XDCMD_WRP, XDFUN_CTL, 0, 0, 0, 0, XD_SUB_POLL); XDC_DONE(xdc, rqno, error); if (error) { printf("%s: controller config error: %s\n", xdc->sc_dev.dv_xname, xdc_e2str(error)); return; } /* link in interrupt with higher level software */ vme_intr_map(ct, va->vma_vec, va->vma_pri, &ih); vme_intr_establish(ct, ih, xdcintr, xdc); evcnt_attach(&xdc->sc_dev, "intr", &xdc->sc_intrcnt); vme_bus_establish(ct, &xdc->sc_dev); /* now we must look for disks using autoconfig */ xa.fullmode = XD_SUB_POLL; xa.booting = 1; for (xa.driveno = 0; xa.driveno < XDC_MAXDEV; xa.driveno++) (void) config_found(self, (void *) &xa, NULL); /* start the watchdog clock */ timeout(xdc_tick, xdc, XDC_TICKCNT); } /* * xdmatch: probe for disk. * * note: we almost always say disk is present. this allows us to * spin up and configure a disk after the system is booted (we can * call xdattach!). */ int xdmatch(parent, cf, aux) struct device *parent; struct cfdata *cf; void *aux; { struct xdc_attach_args *xa = aux; /* looking for autoconf wildcard or exact match */ if (cf->cf_loc[XDCCF_DRIVE] != XDCCF_DRIVE_DEFAULT && cf->cf_loc[XDCCF_DRIVE] != xa->driveno) return 0; return 1; } /* * xdattach: attach a disk. this can be called from autoconf and also * from xdopen/xdstrategy. */ void xdattach(parent, self, aux) struct device *parent, *self; void *aux; { struct xd_softc *xd = (void *) self; struct xdc_softc *xdc = (void *) parent; struct xdc_attach_args *xa = aux; int rqno, spt = 0, mb, blk, lcv, fmode, s = 0, newstate; struct xd_iopb_drive *driopb; struct dkbad *dkb; int rseg, error; bus_dma_segment_t seg; caddr_t dmaddr; caddr_t buf; /* * Always re-initialize the disk structure. We want statistics * to start with a clean slate. */ bzero(&xd->sc_dk, sizeof(xd->sc_dk)); xd->sc_dk.dk_driver = &xddkdriver; xd->sc_dk.dk_name = xd->sc_dev.dv_xname; /* if booting, init the xd_softc */ if (xa->booting) { xd->state = XD_DRIVE_UNKNOWN; /* to start */ xd->flags = 0; xd->parent = xdc; } xd->xd_drive = xa->driveno; fmode = xa->fullmode; xdc->sc_drives[xa->driveno] = xd; /* if not booting, make sure we are the only process in the attach for * this drive. if locked out, sleep on it. */ if (!xa->booting) { s = splbio(); while (xd->state == XD_DRIVE_ATTACHING) { if (tsleep(&xd->state, PRIBIO, "xdattach", 0)) { splx(s); return; } } printf("%s at %s", xd->sc_dev.dv_xname, xd->parent->sc_dev.dv_xname); } /* we now have control */ xd->state = XD_DRIVE_ATTACHING; newstate = XD_DRIVE_UNKNOWN; buf = NULL; error = bus_dmamem_alloc(xdc->dmatag, XDFM_BPS, NBPG, 0, &seg, 1, &rseg, BUS_DMA_NOWAIT); if (error) { printf("%s: DMA buffer alloc error %d\n", xd->sc_dev.dv_xname, error); goto done; } dmaddr = (caddr_t)seg.ds_addr; error = bus_dmamem_map(xdc->dmatag, &seg, rseg, XDFM_BPS, &buf, BUS_DMA_NOWAIT|BUS_DMA_COHERENT); if (error) { printf("%s: DMA buffer alloc error %d\n", xd->sc_dev.dv_xname, error); bus_dmamem_free(xdc->dmatag, &seg, rseg); goto done; } /* first try and reset the drive */ rqno = xdc_cmd(xdc, XDCMD_RST, 0, xd->xd_drive, 0, 0, 0, fmode); XDC_DONE(xdc, rqno, error); if (error == XD_ERR_NRDY) { printf(" drive %d: off-line\n", xa->driveno); goto done; } if (error) { printf(": ERROR 0x%02x (%s)\n", error, xdc_e2str(error)); goto done; } printf(" drive %d: ready\n", xa->driveno); /* now set format parameters */ rqno = xdc_cmd(xdc, XDCMD_WRP, XDFUN_FMT, xd->xd_drive, 0, 0, 0, fmode); XDC_DONE(xdc, rqno, error); if (error) { printf("%s: write format parameters failed: %s\n", xd->sc_dev.dv_xname, xdc_e2str(error)); goto done; } /* get drive parameters */ rqno = xdc_cmd(xdc, XDCMD_RDP, XDFUN_DRV, xd->xd_drive, 0, 0, 0, fmode); if (rqno != XD_ERR_FAIL) { driopb = (struct xd_iopb_drive *) & xdc->iopbase[rqno]; spt = driopb->sectpertrk; } XDC_DONE(xdc, rqno, error); if (error) { printf("%s: read drive parameters failed: %s\n", xd->sc_dev.dv_xname, xdc_e2str(error)); goto done; } /* * now set drive parameters (to semi-bogus values) so we can read the * disk label. */ xd->pcyl = xd->ncyl = 1; xd->acyl = 0; xd->nhead = 1; xd->nsect = 1; xd->sectpercyl = 1; for (lcv = 0; lcv < 126; lcv++) /* init empty bad144 table */ xd->dkb.bt_bad[lcv].bt_cyl = xd->dkb.bt_bad[lcv].bt_trksec = 0xffff; rqno = xdc_cmd(xdc, XDCMD_WRP, XDFUN_DRV, xd->xd_drive, 0, 0, 0, fmode); XDC_DONE(xdc, rqno, error); if (error) { printf("%s: write drive parameters failed: %s\n", xd->sc_dev.dv_xname, xdc_e2str(error)); goto done; } /* read disk label */ rqno = xdc_cmd(xdc, XDCMD_RD, 0, xd->xd_drive, 0, 1, dmaddr, fmode); XDC_DONE(xdc, rqno, error); if (error) { printf("%s: reading disk label failed: %s\n", xd->sc_dev.dv_xname, xdc_e2str(error)); goto done; } newstate = XD_DRIVE_NOLABEL; xd->hw_spt = spt; /* Attach the disk: must be before getdisklabel to malloc label */ disk_attach(&xd->sc_dk); if (xdgetdisklabel(xd, buf) != XD_ERR_AOK) goto done; /* inform the user of what is up */ printf("%s: <%s>, pcyl %d, hw_spt %d\n", xd->sc_dev.dv_xname, buf, xd->pcyl, spt); mb = xd->ncyl * (xd->nhead * xd->nsect) / (1048576 / XDFM_BPS); printf("%s: %dMB, %d cyl, %d head, %d sec, %d bytes/sec\n", xd->sc_dev.dv_xname, mb, xd->ncyl, xd->nhead, xd->nsect, XDFM_BPS); /* now set the real drive parameters! */ rqno = xdc_cmd(xdc, XDCMD_WRP, XDFUN_DRV, xd->xd_drive, 0, 0, 0, fmode); XDC_DONE(xdc, rqno, error); if (error) { printf("%s: write real drive parameters failed: %s\n", xd->sc_dev.dv_xname, xdc_e2str(error)); goto done; } newstate = XD_DRIVE_ONLINE; /* * read bad144 table. this table resides on the first sector of the * last track of the disk (i.e. second cyl of "acyl" area). */ blk = (xd->ncyl + xd->acyl - 1) * (xd->nhead * xd->nsect) + /* last cyl */ (xd->nhead - 1) * xd->nsect; /* last head */ rqno = xdc_cmd(xdc, XDCMD_RD, 0, xd->xd_drive, blk, 1, dmaddr, fmode); XDC_DONE(xdc, rqno, error); if (error) { printf("%s: reading bad144 failed: %s\n", xd->sc_dev.dv_xname, xdc_e2str(error)); goto done; } /* check dkbad for sanity */ dkb = (struct dkbad *) buf; for (lcv = 0; lcv < 126; lcv++) { if ((dkb->bt_bad[lcv].bt_cyl == 0xffff || dkb->bt_bad[lcv].bt_cyl == 0) && dkb->bt_bad[lcv].bt_trksec == 0xffff) continue; /* blank */ if (dkb->bt_bad[lcv].bt_cyl >= xd->ncyl) break; if ((dkb->bt_bad[lcv].bt_trksec >> 8) >= xd->nhead) break; if ((dkb->bt_bad[lcv].bt_trksec & 0xff) >= xd->nsect) break; } if (lcv != 126) { printf("%s: warning: invalid bad144 sector!\n", xd->sc_dev.dv_xname); } else { bcopy(buf, &xd->dkb, XDFM_BPS); } dk_establish(&xd->sc_dk, &xd->sc_dev); /* XXX */ done: if (buf != NULL) { bus_dmamem_unmap(xdc->dmatag, buf, XDFM_BPS); bus_dmamem_free(xdc->dmatag, &seg, rseg); } xd->state = newstate; if (!xa->booting) { wakeup(&xd->state); splx(s); } } /* * end of autoconfig functions */ /* * { b , c } d e v s w f u n c t i o n s */ /* * xdclose: close device */ int xdclose(dev, flag, fmt, p) dev_t dev; int flag, fmt; struct proc *p; { struct xd_softc *xd = xd_cd.cd_devs[DISKUNIT(dev)]; int part = DISKPART(dev); /* clear mask bits */ switch (fmt) { case S_IFCHR: xd->sc_dk.dk_copenmask &= ~(1 << part); break; case S_IFBLK: xd->sc_dk.dk_bopenmask &= ~(1 << part); break; } xd->sc_dk.dk_openmask = xd->sc_dk.dk_copenmask | xd->sc_dk.dk_bopenmask; return 0; } /* * xddump: crash dump system */ int xddump(dev, blkno, va, size) dev_t dev; daddr_t blkno; caddr_t va; size_t size; { int unit, part; struct xd_softc *xd; unit = DISKUNIT(dev); if (unit >= xd_cd.cd_ndevs) return ENXIO; part = DISKPART(dev); xd = xd_cd.cd_devs[unit]; printf("%s%c: crash dump not supported (yet)\n", xd->sc_dev.dv_xname, 'a' + part); return ENXIO; /* outline: globals: "dumplo" == sector number of partition to start * dump at (convert to physical sector with partition table) * "dumpsize" == size of dump in clicks "physmem" == size of physical * memory (clicks, ctob() to get bytes) (normal case: dumpsize == * physmem) * * dump a copy of physical memory to the dump device starting at sector * "dumplo" in the swap partition (make sure > 0). map in pages as * we go. use polled I/O. * * XXX how to handle NON_CONTIG? */ } /* * xdioctl: ioctls on XD drives. based on ioctl's of other netbsd disks. */ int xdioctl(dev, command, addr, flag, p) dev_t dev; u_long command; caddr_t addr; int flag; struct proc *p; { struct xd_softc *xd; struct xd_iocmd *xio; int error, s, unit; unit = DISKUNIT(dev); if (unit >= xd_cd.cd_ndevs || (xd = xd_cd.cd_devs[unit]) == NULL) return (ENXIO); /* switch on ioctl type */ switch (command) { case DIOCSBAD: /* set bad144 info */ if ((flag & FWRITE) == 0) return EBADF; s = splbio(); bcopy(addr, &xd->dkb, sizeof(xd->dkb)); splx(s); return 0; case DIOCGDINFO: /* get disk label */ bcopy(xd->sc_dk.dk_label, addr, sizeof(struct disklabel)); return 0; case DIOCGPART: /* get partition info */ ((struct partinfo *) addr)->disklab = xd->sc_dk.dk_label; ((struct partinfo *) addr)->part = &xd->sc_dk.dk_label->d_partitions[DISKPART(dev)]; return 0; case DIOCSDINFO: /* set disk label */ if ((flag & FWRITE) == 0) return EBADF; error = setdisklabel(xd->sc_dk.dk_label, (struct disklabel *) addr, /* xd->sc_dk.dk_openmask : */ 0, xd->sc_dk.dk_cpulabel); if (error == 0) { if (xd->state == XD_DRIVE_NOLABEL) xd->state = XD_DRIVE_ONLINE; } return error; case DIOCWLABEL: /* change write status of disk label */ if ((flag & FWRITE) == 0) return EBADF; if (*(int *) addr) xd->flags |= XD_WLABEL; else xd->flags &= ~XD_WLABEL; return 0; case DIOCWDINFO: /* write disk label */ if ((flag & FWRITE) == 0) return EBADF; error = setdisklabel(xd->sc_dk.dk_label, (struct disklabel *) addr, /* xd->sc_dk.dk_openmask : */ 0, xd->sc_dk.dk_cpulabel); if (error == 0) { if (xd->state == XD_DRIVE_NOLABEL) xd->state = XD_DRIVE_ONLINE; /* Simulate opening partition 0 so write succeeds. */ xd->sc_dk.dk_openmask |= (1 << 0); error = writedisklabel(MAKEDISKDEV(major(dev), DISKUNIT(dev), RAW_PART), xdstrategy, xd->sc_dk.dk_label, xd->sc_dk.dk_cpulabel); xd->sc_dk.dk_openmask = xd->sc_dk.dk_copenmask | xd->sc_dk.dk_bopenmask; } return error; case DIOSXDCMD: xio = (struct xd_iocmd *) addr; if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) return (error); return (xdc_ioctlcmd(xd, dev, xio)); default: return ENOTTY; } } /* * xdopen: open drive */ int xdopen(dev, flag, fmt, p) dev_t dev; int flag, fmt; struct proc *p; { int unit, part; struct xd_softc *xd; struct xdc_attach_args xa; /* first, could it be a valid target? */ unit = DISKUNIT(dev); if (unit >= xd_cd.cd_ndevs || (xd = xd_cd.cd_devs[unit]) == NULL) return (ENXIO); part = DISKPART(dev); /* do we need to attach the drive? */ if (xd->state == XD_DRIVE_UNKNOWN) { xa.driveno = xd->xd_drive; xa.fullmode = XD_SUB_WAIT; xa.booting = 0; xdattach((struct device *) xd->parent, (struct device *) xd, &xa); if (xd->state == XD_DRIVE_UNKNOWN) { return (EIO); } } /* check for partition */ if (part != RAW_PART && (part >= xd->sc_dk.dk_label->d_npartitions || xd->sc_dk.dk_label->d_partitions[part].p_fstype == FS_UNUSED)) { return (ENXIO); } /* set open masks */ switch (fmt) { case S_IFCHR: xd->sc_dk.dk_copenmask |= (1 << part); break; case S_IFBLK: xd->sc_dk.dk_bopenmask |= (1 << part); break; } xd->sc_dk.dk_openmask = xd->sc_dk.dk_copenmask | xd->sc_dk.dk_bopenmask; return 0; } int xdread(dev, uio, flags) dev_t dev; struct uio *uio; int flags; { return (physio(xdstrategy, NULL, dev, B_READ, minphys, uio)); } int xdwrite(dev, uio, flags) dev_t dev; struct uio *uio; int flags; { return (physio(xdstrategy, NULL, dev, B_WRITE, minphys, uio)); } /* * xdsize: return size of a partition for a dump */ int xdsize(dev) dev_t dev; { struct xd_softc *xdsc; int unit, part, size, omask; /* valid unit? */ unit = DISKUNIT(dev); if (unit >= xd_cd.cd_ndevs || (xdsc = xd_cd.cd_devs[unit]) == NULL) return (-1); part = DISKPART(dev); omask = xdsc->sc_dk.dk_openmask & (1 << part); if (omask == 0 && xdopen(dev, 0, S_IFBLK, NULL) != 0) return (-1); /* do it */ if (xdsc->sc_dk.dk_label->d_partitions[part].p_fstype != FS_SWAP) size = -1; /* only give valid size for swap partitions */ else size = xdsc->sc_dk.dk_label->d_partitions[part].p_size * (xdsc->sc_dk.dk_label->d_secsize / DEV_BSIZE); if (omask == 0 && xdclose(dev, 0, S_IFBLK, NULL) != 0) return (-1); return (size); } /* * xdstrategy: buffering system interface to xd. */ void xdstrategy(bp) struct buf *bp; { struct xd_softc *xd; struct xdc_softc *parent; struct buf *wq; int s, unit; struct xdc_attach_args xa; unit = DISKUNIT(bp->b_dev); /* check for live device */ if (unit >= xd_cd.cd_ndevs || (xd = xd_cd.cd_devs[unit]) == 0 || bp->b_blkno < 0 || (bp->b_bcount % xd->sc_dk.dk_label->d_secsize) != 0) { bp->b_error = EINVAL; goto bad; } /* do we need to attach the drive? */ if (xd->state == XD_DRIVE_UNKNOWN) { xa.driveno = xd->xd_drive; xa.fullmode = XD_SUB_WAIT; xa.booting = 0; xdattach((struct device *)xd->parent, (struct device *)xd, &xa); if (xd->state == XD_DRIVE_UNKNOWN) { bp->b_error = EIO; goto bad; } } if (xd->state != XD_DRIVE_ONLINE && DISKPART(bp->b_dev) != RAW_PART) { /* no I/O to unlabeled disks, unless raw partition */ bp->b_error = EIO; goto bad; } /* short circuit zero length request */ if (bp->b_bcount == 0) goto done; /* check bounds with label (disksubr.c). Determine the size of the * transfer, and make sure it is within the boundaries of the * partition. Adjust transfer if needed, and signal errors or early * completion. */ if (bounds_check_with_label(bp, xd->sc_dk.dk_label, (xd->flags & XD_WLABEL) != 0) <= 0) goto done; /* * now we know we have a valid buf structure that we need to do I/O * on. * * note that we don't disksort because the controller has a sorting * algorithm built into the hardware. */ s = splbio(); /* protect the queues */ /* first, give jobs in front of us a chance */ parent = xd->parent; while (parent->nfree > 0 && parent->sc_wq.b_actf) if (xdc_startbuf(parent, NULL, NULL) != XD_ERR_AOK) break; /* if there are no free iorq's, then we just queue and return. the * buffs will get picked up later by xdcintr(). */ if (parent->nfree == 0) { wq = &xd->parent->sc_wq; bp->b_actf = 0; bp->b_actb = wq->b_actb; *wq->b_actb = bp; wq->b_actb = &bp->b_actf; splx(s); return; } /* now we have free iopb's and we are at splbio... start 'em up */ if (xdc_startbuf(parent, xd, bp) != XD_ERR_AOK) { return; } /* done! */ splx(s); return; bad: /* tells upper layers we have an error */ bp->b_flags |= B_ERROR; done: /* tells upper layers we are done with this * buf */ bp->b_resid = bp->b_bcount; biodone(bp); } /* * end of {b,c}devsw functions */ /* * i n t e r r u p t f u n c t i o n * * xdcintr: hardware interrupt. */ int xdcintr(v) void *v; { struct xdc_softc *xdcsc = v; /* kick the event counter */ xdcsc->sc_intrcnt.ev_count++; /* remove as many done IOPBs as possible */ xdc_remove_iorq(xdcsc); /* start any iorq's already waiting */ xdc_start(xdcsc, XDC_MAXIOPB); /* fill up any remaining iorq's with queue'd buffers */ while (xdcsc->nfree > 0 && xdcsc->sc_wq.b_actf) if (xdc_startbuf(xdcsc, NULL, NULL) != XD_ERR_AOK) break; return (1); } /* * end of interrupt function */ /* * i n t e r n a l f u n c t i o n s */ /* * xdc_rqinit: fill out the fields of an I/O request */ inline void xdc_rqinit(rq, xdc, xd, md, blk, cnt, db, bp) struct xd_iorq *rq; struct xdc_softc *xdc; struct xd_softc *xd; int md; u_long blk; int cnt; caddr_t db; struct buf *bp; { rq->xdc = xdc; rq->xd = xd; rq->ttl = XDC_MAXTTL + 10; rq->mode = md; rq->tries = rq->errno = rq->lasterror = 0; rq->blockno = blk; rq->sectcnt = cnt; rq->dbuf = db; rq->buf = bp; } /* * xdc_rqtopb: load up an IOPB based on an iorq */ void xdc_rqtopb(iorq, iopb, cmd, subfun) struct xd_iorq *iorq; struct xd_iopb *iopb; int cmd, subfun; { u_long block, dp; /* standard stuff */ iopb->errs = iopb->done = 0; iopb->comm = cmd; iopb->errno = iopb->status = 0; iopb->subfun = subfun; if (iorq->xd) iopb->unit = iorq->xd->xd_drive; else iopb->unit = 0; /* check for alternate IOPB format */ if (cmd == XDCMD_WRP) { switch (subfun) { case XDFUN_CTL:{ struct xd_iopb_ctrl *ctrl = (struct xd_iopb_ctrl *) iopb; iopb->lll = 0; iopb->intl = (XD_STATE(iorq->mode) == XD_SUB_POLL) ? 0 : iorq->xdc->ipl; ctrl->param_a = XDPA_TMOD | XDPA_DACF; ctrl->param_b = XDPB_ROR | XDPB_TDT_3_2USEC; ctrl->param_c = XDPC_OVS | XDPC_COP | XDPC_ASR | XDPC_RBC | XDPC_ECC2; ctrl->throttle = XDC_THROTTLE; ctrl->delay = XDC_DELAY; break; } case XDFUN_DRV:{ struct xd_iopb_drive *drv = (struct xd_iopb_drive *)iopb; /* we assume that the disk label has the right * info */ if (XD_STATE(iorq->mode) == XD_SUB_POLL) drv->dparam_ipl = (XDC_DPARAM << 3); else drv->dparam_ipl = (XDC_DPARAM << 3) | iorq->xdc->ipl; drv->maxsect = iorq->xd->nsect - 1; drv->maxsector = drv->maxsect; /* note: maxsector != maxsect only if you are * doing cyl sparing */ drv->headoff = 0; drv->maxcyl = iorq->xd->pcyl - 1; drv->maxhead = iorq->xd->nhead - 1; break; } case XDFUN_FMT:{ struct xd_iopb_format *form = (struct xd_iopb_format *) iopb; if (XD_STATE(iorq->mode) == XD_SUB_POLL) form->interleave_ipl = (XDC_INTERLEAVE << 3); else form->interleave_ipl = (XDC_INTERLEAVE << 3) | iorq->xdc->ipl; form->field1 = XDFM_FIELD1; form->field2 = XDFM_FIELD2; form->field3 = XDFM_FIELD3; form->field4 = XDFM_FIELD4; form->bytespersec = XDFM_BPS; form->field6 = XDFM_FIELD6; form->field7 = XDFM_FIELD7; break; } } } else { /* normal IOPB case (harmless to RDP command) */ iopb->lll = 0; iopb->intl = (XD_STATE(iorq->mode) == XD_SUB_POLL) ? 0 : iorq->xdc->ipl; iopb->sectcnt = iorq->sectcnt; block = iorq->blockno; if (iorq->xd == NULL || block == 0) { iopb->sectno = iopb->headno = iopb->cylno = 0; } else { iopb->sectno = block % iorq->xd->nsect; block = block / iorq->xd->nsect; iopb->headno = block % iorq->xd->nhead; block = block / iorq->xd->nhead; iopb->cylno = block; } dp = (u_long) iorq->dbuf; dp = iopb->daddr = (iorq->dbuf == NULL) ? 0 : dp; iopb->addrmod = ((dp + (XDFM_BPS * iorq->sectcnt)) > 0x1000000) ? XDC_ADDRMOD32 : XDC_ADDRMOD; } } /* * xdc_cmd: front end for POLL'd and WAIT'd commands. Returns rqno. * If you've already got an IORQ, you can call submit directly (currently * there is no need to do this). NORM requests are handled seperately. */ int xdc_cmd(xdcsc, cmd, subfn, unit, block, scnt, dptr, fullmode) struct xdc_softc *xdcsc; int cmd, subfn, unit, block, scnt; char *dptr; int fullmode; { int rqno, submode = XD_STATE(fullmode), retry; struct xd_iorq *iorq; struct xd_iopb *iopb; /* get iorq/iopb */ switch (submode) { case XD_SUB_POLL: while (xdcsc->nfree == 0) { if (xdc_piodriver(xdcsc, 0, 1) != XD_ERR_AOK) return (XD_ERR_FAIL); } break; case XD_SUB_WAIT: retry = 1; while (retry) { while (xdcsc->nfree == 0) { if (tsleep(&xdcsc->nfree, PRIBIO, "xdnfree", 0)) return (XD_ERR_FAIL); } while (xdcsc->ndone > XDC_SUBWAITLIM) { if (tsleep(&xdcsc->ndone, PRIBIO, "xdsubwait", 0)) return (XD_ERR_FAIL); } if (xdcsc->nfree) retry = 0; /* got it */ } break; default: return (XD_ERR_FAIL); /* illegal */ } if (xdcsc->nfree == 0) panic("xdcmd nfree"); rqno = XDC_RQALLOC(xdcsc); iorq = &xdcsc->reqs[rqno]; iopb = iorq->iopb; /* init iorq/iopb */ xdc_rqinit(iorq, xdcsc, (unit == XDC_NOUNIT) ? NULL : xdcsc->sc_drives[unit], fullmode, block, scnt, dptr, NULL); /* load IOPB from iorq */ xdc_rqtopb(iorq, iopb, cmd, subfn); /* submit it for processing */ xdc_submit_iorq(xdcsc, rqno, fullmode); /* error code will be in iorq */ return (rqno); } /* * xdc_startbuf * start a buffer running, assumes nfree > 0 */ int xdc_startbuf(xdcsc, xdsc, bp) struct xdc_softc *xdcsc; struct xd_softc *xdsc; struct buf *bp; { int rqno, partno; struct xd_iorq *iorq; struct xd_iopb *iopb; struct buf *wq; u_long block; /* caddr_t dbuf;*/ int error; if (!xdcsc->nfree) panic("xdc_startbuf free"); rqno = XDC_RQALLOC(xdcsc); iorq = &xdcsc->reqs[rqno]; iopb = iorq->iopb; /* get buf */ if (bp == NULL) { bp = xdcsc->sc_wq.b_actf; if (!bp) panic("xdc_startbuf bp"); wq = bp->b_actf; if (wq) wq->b_actb = bp->b_actb; else xdcsc->sc_wq.b_actb = bp->b_actb; *bp->b_actb = wq; xdsc = xdcsc->sc_drives[DISKUNIT(bp->b_dev)]; } partno = DISKPART(bp->b_dev); #ifdef XDC_DEBUG printf("xdc_startbuf: %s%c: %s block %d\n", xdsc->sc_dev.dv_xname, 'a' + partno, (bp->b_flags & B_READ) ? "read" : "write", bp->b_blkno); printf("xdc_startbuf: b_bcount %d, b_data 0x%x\n", bp->b_bcount, bp->b_data); #endif /* * load request. we have to calculate the correct block number based * on partition info. * * note that iorq points to the buffer as mapped into DVMA space, * where as the bp->b_data points to its non-DVMA mapping. */ block = bp->b_blkno + ((partno == RAW_PART) ? 0 : xdsc->sc_dk.dk_label->d_partitions[partno].p_offset); error = bus_dmamap_load(xdcsc->dmatag, iorq->dmamap, bp->b_data, bp->b_bcount, 0, BUS_DMA_NOWAIT); if (error != 0) { printf("%s: warning: cannot load DMA map\n", xdcsc->sc_dev.dv_xname); XDC_FREE(xdcsc, rqno); wq = &xdcsc->sc_wq; /* put at end of queue */ bp->b_actf = 0; bp->b_actb = wq->b_actb; *wq->b_actb = bp; wq->b_actb = &bp->b_actf; return (XD_ERR_FAIL); /* XXX: need some sort of * call-back scheme here? */ } bus_dmamap_sync(xdcsc->dmatag, iorq->dmamap, 0, iorq->dmamap->dm_mapsize, (bp->b_flags & B_READ) ? BUS_DMASYNC_PREREAD : BUS_DMASYNC_PREWRITE); /* init iorq and load iopb from it */ xdc_rqinit(iorq, xdcsc, xdsc, XD_SUB_NORM | XD_MODE_VERBO, block, bp->b_bcount / XDFM_BPS, (caddr_t)iorq->dmamap->dm_segs[0].ds_addr, bp); xdc_rqtopb(iorq, iopb, (bp->b_flags & B_READ) ? XDCMD_RD : XDCMD_WR, 0); /* Instrumentation. */ disk_busy(&xdsc->sc_dk); /* now submit [note that xdc_submit_iorq can never fail on NORM reqs] */ xdc_submit_iorq(xdcsc, rqno, XD_SUB_NORM); return (XD_ERR_AOK); } /* * xdc_submit_iorq: submit an iorq for processing. returns XD_ERR_AOK * if ok. if it fail returns an error code. type is XD_SUB_*. * * note: caller frees iorq in all cases except NORM * * return value: * NORM: XD_AOK (req pending), XD_FAIL (couldn't submit request) * WAIT: XD_AOK (success), (failed) * POLL: * NOQ : * * there are three sources for i/o requests: * [1] xdstrategy: normal block I/O, using "struct buf" system. * [2] autoconfig/crash dump: these are polled I/O requests, no interrupts. * [3] open/ioctl: these are I/O requests done in the context of a process, * and the process should block until they are done. * * software state is stored in the iorq structure. each iorq has an * iopb structure. the hardware understands the iopb structure. * every command must go through an iopb. a 7053 can only handle * XDC_MAXIOPB (31) active iopbs at one time. iopbs are allocated in * DVMA space at boot up time. what happens if we run out of iopb's? * for i/o type [1], the buffers are queued at the "buff" layer and * picked up later by the interrupt routine. for case [2] the * programmed i/o driver is called with a special flag that says * return when one iopb is free. for case [3] the process can sleep * on the iorq free list until some iopbs are avaliable. */ int xdc_submit_iorq(xdcsc, iorqno, type) struct xdc_softc *xdcsc; int iorqno; int type; { u_long iopbaddr; struct xd_iorq *iorq = &xdcsc->reqs[iorqno]; #ifdef XDC_DEBUG printf("xdc_submit_iorq(%s, no=%d, type=%d)\n", xdcsc->sc_dev.dv_xname, iorqno, type); #endif /* first check and see if controller is busy */ if (xdcsc->xdc->xdc_csr & XDC_ADDING) { #ifdef XDC_DEBUG printf("xdc_submit_iorq: XDC not ready (ADDING)\n"); #endif if (type == XD_SUB_NOQ) return (XD_ERR_FAIL); /* failed */ XDC_TWAIT(xdcsc, iorqno); /* put at end of waitq */ switch (type) { case XD_SUB_NORM: return XD_ERR_AOK; /* success */ case XD_SUB_WAIT: while (iorq->iopb->done == 0) { sleep(iorq, PRIBIO); } return (iorq->errno); case XD_SUB_POLL: return (xdc_piodriver(xdcsc, iorqno, 0)); default: panic("xdc_submit_iorq adding"); } } #ifdef XDC_DEBUG { u_char *rio = (u_char *) iorq->iopb; int sz = sizeof(struct xd_iopb), lcv; printf("%s: aio #%d [", xdcsc->sc_dev.dv_xname, iorq - xdcsc->reqs); for (lcv = 0; lcv < sz; lcv++) printf(" %02x", rio[lcv]); printf("]\n"); } #endif /* XDC_DEBUG */ /* controller not busy, start command */ iopbaddr = (u_long) iorq->iopb; XDC_GO(xdcsc->xdc, iopbaddr); /* go! */ xdcsc->nrun++; /* command now running, wrap it up */ switch (type) { case XD_SUB_NORM: case XD_SUB_NOQ: return (XD_ERR_AOK); /* success */ case XD_SUB_WAIT: while (iorq->iopb->done == 0) { sleep(iorq, PRIBIO); } return (iorq->errno); case XD_SUB_POLL: return (xdc_piodriver(xdcsc, iorqno, 0)); default: panic("xdc_submit_iorq wrap up"); } panic("xdc_submit_iorq"); return 0; /* not reached */ } /* * xdc_piodriver * * programmed i/o driver. this function takes over the computer * and drains off all i/o requests. it returns the status of the iorq * the caller is interesting in. if freeone is true, then it returns * when there is a free iorq. */ int xdc_piodriver(xdcsc, iorqno, freeone) struct xdc_softc *xdcsc; char iorqno; int freeone; { int nreset = 0; int retval = 0; u_long count; struct xdc *xdc = xdcsc->xdc; #ifdef XDC_DEBUG printf("xdc_piodriver(%s, %d, freeone=%d)\n", xdcsc->sc_dev.dv_xname, iorqno, freeone); #endif while (xdcsc->nwait || xdcsc->nrun) { #ifdef XDC_DEBUG printf("xdc_piodriver: wait=%d, run=%d\n", xdcsc->nwait, xdcsc->nrun); #endif XDC_WAIT(xdc, count, XDC_MAXTIME, (XDC_REMIOPB | XDC_F_ERROR)); #ifdef XDC_DEBUG printf("xdc_piodriver: done wait with count = %d\n", count); #endif /* we expect some progress soon */ if (count == 0 && nreset >= 2) { xdc_reset(xdcsc, 0, XD_RSET_ALL, XD_ERR_FAIL, 0); #ifdef XDC_DEBUG printf("xdc_piodriver: timeout\n"); #endif return (XD_ERR_FAIL); } if (count == 0) { if (xdc_reset(xdcsc, 0, (nreset++ == 0) ? XD_RSET_NONE : iorqno, XD_ERR_FAIL, 0) == XD_ERR_FAIL) return (XD_ERR_FAIL); /* flushes all but POLL * requests, resets */ continue; } xdc_remove_iorq(xdcsc); /* could resubmit request */ if (freeone) { if (xdcsc->nrun < XDC_MAXIOPB) { #ifdef XDC_DEBUG printf("xdc_piodriver: done: one free\n"); #endif return (XD_ERR_AOK); } continue; /* don't xdc_start */ } xdc_start(xdcsc, XDC_MAXIOPB); } /* get return value */ retval = xdcsc->reqs[iorqno].errno; #ifdef XDC_DEBUG printf("xdc_piodriver: done, retval = 0x%x (%s)\n", xdcsc->reqs[iorqno].errno, xdc_e2str(xdcsc->reqs[iorqno].errno)); #endif /* now that we've drained everything, start up any bufs that have * queued */ while (xdcsc->nfree > 0 && xdcsc->sc_wq.b_actf) if (xdc_startbuf(xdcsc, NULL, NULL) != XD_ERR_AOK) break; return (retval); } /* * xdc_reset: reset one drive. NOTE: assumes xdc was just reset. * we steal iopb[0] for this, but we put it back when we are done. */ void xdc_xdreset(xdcsc, xdsc) struct xdc_softc *xdcsc; struct xd_softc *xdsc; { struct xd_iopb tmpiopb; u_long addr; int del; bcopy(xdcsc->iopbase, &tmpiopb, sizeof(tmpiopb)); bzero(xdcsc->iopbase, sizeof(tmpiopb)); xdcsc->iopbase->comm = XDCMD_RST; xdcsc->iopbase->unit = xdsc->xd_drive; addr = (u_long) xdcsc->dvmaiopb; XDC_GO(xdcsc->xdc, addr); /* go! */ XDC_WAIT(xdcsc->xdc, del, XDC_RESETUSEC, XDC_REMIOPB); if (del <= 0 || xdcsc->iopbase->errs) { printf("%s: off-line: %s\n", xdcsc->sc_dev.dv_xname, xdc_e2str(xdcsc->iopbase->errno)); xdcsc->xdc->xdc_csr = XDC_RESET; XDC_WAIT(xdcsc->xdc, del, XDC_RESETUSEC, XDC_RESET); if (del <= 0) panic("xdc_reset"); } else { xdcsc->xdc->xdc_csr = XDC_CLRRIO; /* clear RIO */ } bcopy(&tmpiopb, xdcsc->iopbase, sizeof(tmpiopb)); } /* * xdc_reset: reset everything: requests are marked as errors except * a polled request (which is resubmitted) */ int xdc_reset(xdcsc, quiet, blastmode, error, xdsc) struct xdc_softc *xdcsc; int quiet, blastmode, error; struct xd_softc *xdsc; { int del = 0, lcv, retval = XD_ERR_AOK; int oldfree = xdcsc->nfree; /* soft reset hardware */ if (!quiet) printf("%s: soft reset\n", xdcsc->sc_dev.dv_xname); xdcsc->xdc->xdc_csr = XDC_RESET; XDC_WAIT(xdcsc->xdc, del, XDC_RESETUSEC, XDC_RESET); if (del <= 0) { blastmode = XD_RSET_ALL; /* dead, flush all requests */ retval = XD_ERR_FAIL; } if (xdsc) xdc_xdreset(xdcsc, xdsc); /* fix queues based on "blast-mode" */ for (lcv = 0; lcv < XDC_MAXIOPB; lcv++) { register struct xd_iorq *iorq = &xdcsc->reqs[lcv]; if (XD_STATE(iorq->mode) != XD_SUB_POLL && XD_STATE(iorq->mode) != XD_SUB_WAIT && XD_STATE(iorq->mode) != XD_SUB_NORM) /* is it active? */ continue; xdcsc->nrun--; /* it isn't running any more */ if (blastmode == XD_RSET_ALL || blastmode != lcv) { /* failed */ iorq->errno = error; xdcsc->iopbase[lcv].done = xdcsc->iopbase[lcv].errs = 1; switch (XD_STATE(xdcsc->reqs[lcv].mode)) { case XD_SUB_NORM: iorq->buf->b_error = EIO; iorq->buf->b_flags |= B_ERROR; iorq->buf->b_resid = iorq->sectcnt * XDFM_BPS; bus_dmamap_sync(xdcsc->dmatag, iorq->dmamap, 0, iorq->dmamap->dm_mapsize, (iorq->buf->b_flags & B_READ) ? BUS_DMASYNC_POSTREAD : BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(xdcsc->dmatag, iorq->dmamap); disk_unbusy(&xdcsc->reqs[lcv].xd->sc_dk, (xdcsc->reqs[lcv].buf->b_bcount - xdcsc->reqs[lcv].buf->b_resid)); biodone(iorq->buf); XDC_FREE(xdcsc, lcv); /* add to free list */ break; case XD_SUB_WAIT: wakeup(iorq); case XD_SUB_POLL: xdcsc->ndone++; iorq->mode = XD_NEWSTATE(iorq->mode, XD_SUB_DONE); break; } } else { /* resubmit, put at front of wait queue */ XDC_HWAIT(xdcsc, lcv); } } /* * now, if stuff is waiting, start it. * since we just reset it should go */ xdc_start(xdcsc, XDC_MAXIOPB); /* ok, we did it */ if (oldfree == 0 && xdcsc->nfree) wakeup(&xdcsc->nfree); #ifdef XDC_DIAG del = xdcsc->nwait + xdcsc->nrun + xdcsc->nfree + xdcsc->ndone; if (del != XDC_MAXIOPB) printf("%s: diag: xdc_reset miscount (%d should be %d)!\n", xdcsc->sc_dev.dv_xname, del, XDC_MAXIOPB); else if (xdcsc->ndone > XDC_MAXIOPB - XDC_SUBWAITLIM) printf("%s: diag: lots of done jobs (%d)\n", xdcsc->sc_dev.dv_xname, xdcsc->ndone); #endif printf("RESET DONE\n"); return (retval); } /* * xdc_start: start all waiting buffers */ void xdc_start(xdcsc, maxio) struct xdc_softc *xdcsc; int maxio; { int rqno; while (maxio && xdcsc->nwait && (xdcsc->xdc->xdc_csr & XDC_ADDING) == 0) { XDC_GET_WAITER(xdcsc, rqno); /* note: rqno is an "out" * param */ if (xdc_submit_iorq(xdcsc, rqno, XD_SUB_NOQ) != XD_ERR_AOK) panic("xdc_start"); /* should never happen */ maxio--; } } /* * xdc_remove_iorq: remove "done" IOPB's. */ int xdc_remove_iorq(xdcsc) struct xdc_softc *xdcsc; { int errno, rqno, comm, errs; struct xdc *xdc = xdcsc->xdc; struct xd_iopb *iopb; struct xd_iorq *iorq; struct buf *bp; if (xdc->xdc_csr & XDC_F_ERROR) { /* * FATAL ERROR: should never happen under normal use. This * error is so bad, you can't even tell which IOPB is bad, so * we dump them all. */ errno = xdc->xdc_f_err; printf("%s: fatal error 0x%02x: %s\n", xdcsc->sc_dev.dv_xname, errno, xdc_e2str(errno)); if (xdc_reset(xdcsc, 0, XD_RSET_ALL, errno, 0) != XD_ERR_AOK) { printf("%s: soft reset failed!\n", xdcsc->sc_dev.dv_xname); panic("xdc_remove_iorq: controller DEAD"); } return (XD_ERR_AOK); } /* * get iopb that is done * * hmm... I used to read the address of the done IOPB off the VME * registers and calculate the rqno directly from that. that worked * until I started putting a load on the controller. when loaded, i * would get interrupts but neither the REMIOPB or F_ERROR bits would * be set, even after DELAY'ing a while! later on the timeout * routine would detect IOPBs that were marked "running" but their * "done" bit was set. rather than dealing directly with this * problem, it is just easier to look at all running IOPB's for the * done bit. */ if (xdc->xdc_csr & XDC_REMIOPB) { xdc->xdc_csr = XDC_CLRRIO; } for (rqno = 0; rqno < XDC_MAXIOPB; rqno++) { iorq = &xdcsc->reqs[rqno]; if (iorq->mode == 0 || XD_STATE(iorq->mode) == XD_SUB_DONE) continue; /* free, or done */ iopb = &xdcsc->iopbase[rqno]; if (iopb->done == 0) continue; /* not done yet */ #ifdef XDC_DEBUG { u_char *rio = (u_char *) iopb; int sz = sizeof(struct xd_iopb), lcv; printf("%s: rio #%d [", xdcsc->sc_dev.dv_xname, rqno); for (lcv = 0; lcv < sz; lcv++) printf(" %02x", rio[lcv]); printf("]\n"); } #endif /* XDC_DEBUG */ xdcsc->nrun--; comm = iopb->comm; errs = iopb->errs; if (errs) iorq->errno = iopb->errno; else iorq->errno = 0; /* handle non-fatal errors */ if (errs && xdc_error(xdcsc, iorq, iopb, rqno, comm) == XD_ERR_AOK) continue; /* AOK: we resubmitted it */ /* this iorq is now done (hasn't been restarted or anything) */ if ((iorq->mode & XD_MODE_VERBO) && iorq->lasterror) xdc_perror(iorq, iopb, 0); /* now, if read/write check to make sure we got all the data * we needed. (this may not be the case if we got an error in * the middle of a multisector request). */ if ((iorq->mode & XD_MODE_B144) != 0 && errs == 0 && (comm == XDCMD_RD || comm == XDCMD_WR)) { /* we just successfully processed a bad144 sector * note: if we are in bad 144 mode, the pointers have * been advanced already (see above) and are pointing * at the bad144 sector. to exit bad144 mode, we * must advance the pointers 1 sector and issue a new * request if there are still sectors left to process * */ XDC_ADVANCE(iorq, 1); /* advance 1 sector */ /* exit b144 mode */ iorq->mode = iorq->mode & (~XD_MODE_B144); if (iorq->sectcnt) { /* more to go! */ iorq->lasterror = iorq->errno = iopb->errno = 0; iopb->errs = iopb->done = 0; iorq->tries = 0; iopb->sectcnt = iorq->sectcnt; iopb->cylno = iorq->blockno / iorq->xd->sectpercyl; iopb->headno = (iorq->blockno / iorq->xd->nhead) % iorq->xd->nhead; iopb->sectno = iorq->blockno % XDFM_BPS; iopb->daddr = (u_long) iorq->dbuf; XDC_HWAIT(xdcsc, rqno); xdc_start(xdcsc, 1); /* resubmit */ continue; } } /* final cleanup, totally done with this request */ switch (XD_STATE(iorq->mode)) { case XD_SUB_NORM: bp = iorq->buf; if (errs) { bp->b_error = EIO; bp->b_flags |= B_ERROR; bp->b_resid = iorq->sectcnt * XDFM_BPS; } else { bp->b_resid = 0; /* done */ } bus_dmamap_sync(xdcsc->dmatag, iorq->dmamap, 0, iorq->dmamap->dm_mapsize, (bp->b_flags & B_READ) ? BUS_DMASYNC_POSTREAD : BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(xdcsc->dmatag, iorq->dmamap); disk_unbusy(&iorq->xd->sc_dk, (bp->b_bcount - bp->b_resid)); XDC_FREE(xdcsc, rqno); biodone(bp); break; case XD_SUB_WAIT: iorq->mode = XD_NEWSTATE(iorq->mode, XD_SUB_DONE); xdcsc->ndone++; wakeup(iorq); break; case XD_SUB_POLL: iorq->mode = XD_NEWSTATE(iorq->mode, XD_SUB_DONE); xdcsc->ndone++; break; } } return (XD_ERR_AOK); } /* * xdc_perror: print error. * - if still_trying is true: we got an error, retried and got a * different error. in that case lasterror is the old error, * and errno is the new one. * - if still_trying is not true, then if we ever had an error it * is in lasterror. also, if iorq->errno == 0, then we recovered * from that error (otherwise iorq->errno == iorq->lasterror). */ void xdc_perror(iorq, iopb, still_trying) struct xd_iorq *iorq; struct xd_iopb *iopb; int still_trying; { int error = iorq->lasterror; printf("%s", (iorq->xd) ? iorq->xd->sc_dev.dv_xname : iorq->xdc->sc_dev.dv_xname); if (iorq->buf) printf("%c: ", 'a' + DISKPART(iorq->buf->b_dev)); if (iopb->comm == XDCMD_RD || iopb->comm == XDCMD_WR) printf("%s %d/%d/%d: ", (iopb->comm == XDCMD_RD) ? "read" : "write", iopb->cylno, iopb->headno, iopb->sectno); printf("%s", xdc_e2str(error)); if (still_trying) printf(" [still trying, new error=%s]", xdc_e2str(iorq->errno)); else if (iorq->errno == 0) printf(" [recovered in %d tries]", iorq->tries); printf("\n"); } /* * xdc_error: non-fatal error encountered... recover. * return AOK if resubmitted, return FAIL if this iopb is done */ int xdc_error(xdcsc, iorq, iopb, rqno, comm) struct xdc_softc *xdcsc; struct xd_iorq *iorq; struct xd_iopb *iopb; int rqno, comm; { int errno = iorq->errno; int erract = errno & XD_ERA_MASK; int oldmode, advance, i; if (erract == XD_ERA_RSET) { /* some errors require a reset */ oldmode = iorq->mode; iorq->mode = XD_SUB_DONE | (~XD_SUB_MASK & oldmode); xdcsc->ndone++; /* make xdc_start ignore us */ xdc_reset(xdcsc, 1, XD_RSET_NONE, errno, iorq->xd); iorq->mode = oldmode; xdcsc->ndone--; } /* check for read/write to a sector in bad144 table if bad: redirect * request to bad144 area */ if ((comm == XDCMD_RD || comm == XDCMD_WR) && (iorq->mode & XD_MODE_B144) == 0) { advance = iorq->sectcnt - iopb->sectcnt; XDC_ADVANCE(iorq, advance); if ((i = isbad(&iorq->xd->dkb, iorq->blockno / iorq->xd->sectpercyl, (iorq->blockno / iorq->xd->nsect) % iorq->xd->nhead, iorq->blockno % iorq->xd->nsect)) != -1) { iorq->mode |= XD_MODE_B144; /* enter bad144 mode & * redirect */ iopb->errno = iopb->done = iopb->errs = 0; iopb->sectcnt = 1; iopb->cylno = (iorq->xd->ncyl + iorq->xd->acyl) - 2; /* second to last acyl */ i = iorq->xd->sectpercyl - 1 - i; /* follow bad144 * standard */ iopb->headno = i / iorq->xd->nhead; iopb->sectno = i % iorq->xd->nhead; XDC_HWAIT(xdcsc, rqno); xdc_start(xdcsc, 1); /* resubmit */ return (XD_ERR_AOK); /* recovered! */ } } /* * it isn't a bad144 sector, must be real error! see if we can retry * it? */ if ((iorq->mode & XD_MODE_VERBO) && iorq->lasterror) xdc_perror(iorq, iopb, 1); /* inform of error state * change */ iorq->lasterror = errno; if ((erract == XD_ERA_RSET || erract == XD_ERA_HARD) && iorq->tries < XDC_MAXTRIES) { /* retry? */ iorq->tries++; iorq->errno = iopb->errno = iopb->done = iopb->errs = 0; XDC_HWAIT(xdcsc, rqno); xdc_start(xdcsc, 1); /* restart */ return (XD_ERR_AOK); /* recovered! */ } /* failed to recover from this error */ return (XD_ERR_FAIL); } /* * xdc_tick: make sure xd is still alive and ticking (err, kicking). */ void xdc_tick(arg) void *arg; { struct xdc_softc *xdcsc = arg; int lcv, s, reset = 0; #ifdef XDC_DIAG int wait, run, free, done, whd = 0; u_char fqc[XDC_MAXIOPB], wqc[XDC_MAXIOPB], mark[XDC_MAXIOPB]; s = splbio(); wait = xdcsc->nwait; run = xdcsc->nrun; free = xdcsc->nfree; done = xdcsc->ndone; bcopy(xdcsc->waitq, wqc, sizeof(wqc)); bcopy(xdcsc->freereq, fqc, sizeof(fqc)); splx(s); if (wait + run + free + done != XDC_MAXIOPB) { printf("%s: diag: IOPB miscount (got w/f/r/d %d/%d/%d/%d, wanted %d)\n", xdcsc->sc_dev.dv_xname, wait, free, run, done, XDC_MAXIOPB); bzero(mark, sizeof(mark)); printf("FREE: "); for (lcv = free; lcv > 0; lcv--) { printf("%d ", fqc[lcv - 1]); mark[fqc[lcv - 1]] = 1; } printf("\nWAIT: "); lcv = wait; while (lcv > 0) { printf("%d ", wqc[whd]); mark[wqc[whd]] = 1; whd = (whd + 1) % XDC_MAXIOPB; lcv--; } printf("\n"); for (lcv = 0; lcv < XDC_MAXIOPB; lcv++) { if (mark[lcv] == 0) printf("MARK: running %d: mode %d done %d errs %d errno 0x%x ttl %d buf %p\n", lcv, xdcsc->reqs[lcv].mode, xdcsc->iopbase[lcv].done, xdcsc->iopbase[lcv].errs, xdcsc->iopbase[lcv].errno, xdcsc->reqs[lcv].ttl, xdcsc->reqs[lcv].buf); } } else if (done > XDC_MAXIOPB - XDC_SUBWAITLIM) printf("%s: diag: lots of done jobs (%d)\n", xdcsc->sc_dev.dv_xname, done); #endif #ifdef XDC_DEBUG printf("%s: tick: csr 0x%x, w/f/r/d %d/%d/%d/%d\n", xdcsc->sc_dev.dv_xname, xdcsc->xdc->xdc_csr, xdcsc->nwait, xdcsc->nfree, xdcsc->nrun, xdcsc->ndone); for (lcv = 0; lcv < XDC_MAXIOPB; lcv++) { if (xdcsc->reqs[lcv].mode) printf("running %d: mode %d done %d errs %d errno 0x%x\n", lcv, xdcsc->reqs[lcv].mode, xdcsc->iopbase[lcv].done, xdcsc->iopbase[lcv].errs, xdcsc->iopbase[lcv].errno); } #endif /* reduce ttl for each request if one goes to zero, reset xdc */ s = splbio(); for (lcv = 0; lcv < XDC_MAXIOPB; lcv++) { if (xdcsc->reqs[lcv].mode == 0 || XD_STATE(xdcsc->reqs[lcv].mode) == XD_SUB_DONE) continue; xdcsc->reqs[lcv].ttl--; if (xdcsc->reqs[lcv].ttl == 0) reset = 1; } if (reset) { printf("%s: watchdog timeout\n", xdcsc->sc_dev.dv_xname); xdc_reset(xdcsc, 0, XD_RSET_NONE, XD_ERR_FAIL, NULL); } splx(s); /* until next time */ timeout(xdc_tick, xdcsc, XDC_TICKCNT); } /* * xdc_ioctlcmd: this function provides a user level interface to the * controller via ioctl. this allows "format" programs to be written * in user code, and is also useful for some debugging. we return * an error code. called at user priority. */ int xdc_ioctlcmd(xd, dev, xio) struct xd_softc *xd; dev_t dev; struct xd_iocmd *xio; { int s, rqno, dummy; caddr_t dvmabuf = NULL, buf = NULL; struct xdc_softc *xdcsc; int rseg, error; bus_dma_segment_t seg; /* check sanity of requested command */ switch (xio->cmd) { case XDCMD_NOP: /* no op: everything should be zero */ if (xio->subfn || xio->dptr || xio->dlen || xio->block || xio->sectcnt) return (EINVAL); break; case XDCMD_RD: /* read / write sectors (up to XD_IOCMD_MAXS) */ case XDCMD_WR: if (xio->subfn || xio->sectcnt > XD_IOCMD_MAXS || xio->sectcnt * XDFM_BPS != xio->dlen || xio->dptr == NULL) return (EINVAL); break; case XDCMD_SK: /* seek: doesn't seem useful to export this */ return (EINVAL); case XDCMD_WRP: /* write parameters */ return (EINVAL);/* not useful, except maybe drive * parameters... but drive parameters should * go via disklabel changes */ case XDCMD_RDP: /* read parameters */ if (xio->subfn != XDFUN_DRV || xio->dlen || xio->block || xio->dptr) return (EINVAL); /* allow read drive params to * get hw_spt */ xio->sectcnt = xd->hw_spt; /* we already know the answer */ return (0); break; case XDCMD_XRD: /* extended read/write */ case XDCMD_XWR: switch (xio->subfn) { case XDFUN_THD:/* track headers */ if (xio->sectcnt != xd->hw_spt || (xio->block % xd->nsect) != 0 || xio->dlen != XD_IOCMD_HSZ * xd->hw_spt || xio->dptr == NULL) return (EINVAL); xio->sectcnt = 0; break; case XDFUN_FMT:/* NOTE: also XDFUN_VFY */ if (xio->cmd == XDCMD_XRD) return (EINVAL); /* no XDFUN_VFY */ if (xio->sectcnt || xio->dlen || (xio->block % xd->nsect) != 0 || xio->dptr) return (EINVAL); break; case XDFUN_HDR:/* header, header verify, data, data ECC */ return (EINVAL); /* not yet */ case XDFUN_DM: /* defect map */ case XDFUN_DMX:/* defect map (alternate location) */ if (xio->sectcnt || xio->dlen != XD_IOCMD_DMSZ || (xio->block % xd->nsect) != 0 || xio->dptr == NULL) return (EINVAL); break; default: return (EINVAL); } break; case XDCMD_TST: /* diagnostics */ return (EINVAL); default: return (EINVAL);/* ??? */ } xdcsc = xd->parent; /* create DVMA buffer for request if needed */ if (xio->dlen) { error = bus_dmamem_alloc(xdcsc->dmatag, xio->dlen, NBPG, 0, &seg, 1, &rseg, BUS_DMA_WAITOK); if (error) return (error); dvmabuf = (caddr_t)seg.ds_addr; error = bus_dmamem_map(xdcsc->dmatag, &seg, rseg, xio->dlen, &buf, BUS_DMA_WAITOK|BUS_DMA_COHERENT); if (error) { bus_dmamem_free(xdcsc->dmatag, &seg, rseg); return (error); } if (xio->cmd == XDCMD_WR || xio->cmd == XDCMD_XWR) { if ((error = copyin(xio->dptr, buf, xio->dlen)) != 0) { bus_dmamem_unmap(xdcsc->dmatag, buf, xio->dlen); bus_dmamem_free(xdcsc->dmatag, &seg, rseg); return (error); } } } /* do it! */ error = 0; s = splbio(); rqno = xdc_cmd(xdcsc, xio->cmd, xio->subfn, xd->xd_drive, xio->block, xio->sectcnt, dvmabuf, XD_SUB_WAIT); if (rqno == XD_ERR_FAIL) { error = EIO; goto done; } xio->errno = xdcsc->reqs[rqno].errno; xio->tries = xdcsc->reqs[rqno].tries; XDC_DONE(xdcsc, rqno, dummy); if (xio->cmd == XDCMD_RD || xio->cmd == XDCMD_XRD) error = copyout(buf, xio->dptr, xio->dlen); done: splx(s); if (dvmabuf) { bus_dmamem_unmap(xdcsc->dmatag, buf, xio->dlen); bus_dmamem_free(xdcsc->dmatag, &seg, rseg); } return (error); } /* * xdc_e2str: convert error code number into an error string */ char * xdc_e2str(no) int no; { switch (no) { case XD_ERR_FAIL: return ("Software fatal error"); case XD_ERR_AOK: return ("Successful completion"); case XD_ERR_ICYL: return ("Illegal cylinder address"); case XD_ERR_IHD: return ("Illegal head address"); case XD_ERR_ISEC: return ("Illgal sector address"); case XD_ERR_CZER: return ("Count zero"); case XD_ERR_UIMP: return ("Unimplemented command"); case XD_ERR_IF1: return ("Illegal field length 1"); case XD_ERR_IF2: return ("Illegal field length 2"); case XD_ERR_IF3: return ("Illegal field length 3"); case XD_ERR_IF4: return ("Illegal field length 4"); case XD_ERR_IF5: return ("Illegal field length 5"); case XD_ERR_IF6: return ("Illegal field length 6"); case XD_ERR_IF7: return ("Illegal field length 7"); case XD_ERR_ISG: return ("Illegal scatter/gather length"); case XD_ERR_ISPT: return ("Not enough sectors per track"); case XD_ERR_ALGN: return ("Next IOPB address alignment error"); case XD_ERR_SGAL: return ("Scatter/gather address alignment error"); case XD_ERR_SGEC: return ("Scatter/gather with auto-ECC"); case XD_ERR_SECC: return ("Soft ECC corrected"); case XD_ERR_SIGN: return ("ECC ignored"); case XD_ERR_ASEK: return ("Auto-seek retry recovered"); case XD_ERR_RTRY: return ("Soft retry recovered"); case XD_ERR_HECC: return ("Hard data ECC"); case XD_ERR_NHDR: return ("Header not found"); case XD_ERR_NRDY: return ("Drive not ready"); case XD_ERR_TOUT: return ("Operation timeout"); case XD_ERR_VTIM: return ("VMEDMA timeout"); case XD_ERR_DSEQ: return ("Disk sequencer error"); case XD_ERR_HDEC: return ("Header ECC error"); case XD_ERR_RVFY: return ("Read verify"); case XD_ERR_VFER: return ("Fatail VMEDMA error"); case XD_ERR_VBUS: return ("VMEbus error"); case XD_ERR_DFLT: return ("Drive faulted"); case XD_ERR_HECY: return ("Header error/cyliner"); case XD_ERR_HEHD: return ("Header error/head"); case XD_ERR_NOCY: return ("Drive not on-cylinder"); case XD_ERR_SEEK: return ("Seek error"); case XD_ERR_ILSS: return ("Illegal sector size"); case XD_ERR_SEC: return ("Soft ECC"); case XD_ERR_WPER: return ("Write-protect error"); case XD_ERR_IRAM: return ("IRAM self test failure"); case XD_ERR_MT3: return ("Maintenance test 3 failure (DSKCEL RAM)"); case XD_ERR_MT4: return ("Maintenance test 4 failure (header shift reg)"); case XD_ERR_MT5: return ("Maintenance test 5 failure (VMEDMA regs)"); case XD_ERR_MT6: return ("Maintenance test 6 failure (REGCEL chip)"); case XD_ERR_MT7: return ("Maintenance test 7 failure (buffer parity)"); case XD_ERR_MT8: return ("Maintenance test 8 failure (disk FIFO)"); case XD_ERR_IOCK: return ("IOPB checksum miscompare"); case XD_ERR_IODM: return ("IOPB DMA fatal"); case XD_ERR_IOAL: return ("IOPB address alignment error"); case XD_ERR_FIRM: return ("Firmware error"); case XD_ERR_MMOD: return ("Illegal maintenance mode test number"); case XD_ERR_ACFL: return ("ACFAIL asserted"); default: return ("Unknown error"); } }