Date: Sun, 17 Apr 2005 22:20:16 +0200 (CEST) From: Martin Birgmeier <nobody@junk.com> To: FreeBSD-gnats-submit@FreeBSD.org Subject: kern/80042: FreeBSD 5.4 RC2: Promise PDC20265 on A7V mobo + Maxtor 6B200P0/BAH41BM0 gives only 268435455 instead of 398297088 sectors Message-ID: <200504172020.j3HKKGkk001724@gandalf.xyzzy> Resent-Message-ID: <200504172030.j3HKUCHF052713@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
>Number: 80042 >Category: kern >Synopsis: FreeBSD 5.4 RC2: Promise PDC20265 on A7V mobo + Maxtor 6B200P0/BAH41BM0 gives only 268435455 instead of 398297088 sectors >Confidential: no >Severity: serious >Priority: high >Responsible: freebsd-bugs >State: open >Quarter: >Keywords: >Date-Required: >Class: sw-bug >Submitter-Id: current-users >Arrival-Date: Sun Apr 17 20:30:11 GMT 2005 >Closed-Date: >Last-Modified: >Originator: Martin Birgmeier >Release: FreeBSD 5.4 RC2 (built from RELENG_5_4 dated 2005-04-14) >Organization: MBi at home >Environment: FreeBSD 5.4 RC2 (built from RELENG_5_4 dated 2005-04-14) (see also output of scanpci -v and atacontrol cap 3 1 at end of PR) >Description: I cannot use a Maxtor 6B200P0/BAH41BM0 200 GB disk on the Promise controller built into the A7V motherboard (Promise PDC20265) under FreeBSD 5.4 RC2. I can read up to sector 268435454, but reading from sector 268435455 will time out. Note 1: Under RELENG_5_3_0_RELEASE, the system would hang indefinitely. Note 2: Under RELENG_4_11_0_RELEASE (which is what I am using to write this defect report), I can use the disk after applying the following fixes (they are merged from trunk somewhere after RELENG_4 branched, and also include some minor other stuff dealing with additional PCI IDs): ====================================================================== *** sys/dev/ata/ata-all.c.ORIG Fri Jan 14 00:03:12 2005 --- sys/dev/ata/ata-all.c Tue Mar 1 19:05:19 2005 *************** *** 1073,1078 **** --- 1073,1079 ---- ata_prtdev(atadev, "can't translate cmd to 48bit version\n"); return -1; } + atadev->channel->flags |= ATA_48BIT_ACTIVE; } else { ATA_OUTB(atadev->channel->r_io, ATA_FEATURE, feature); *** sys/dev/ata/ata-all.h.ORIG Fri Sep 17 22:28:30 2004 --- sys/dev/ata/ata-all.h Tue Mar 1 19:08:57 2005 *************** *** 212,217 **** --- 212,218 ---- #define ATA_ATAPI_DMA_RO 0x04 #define ATA_QUEUED 0x08 #define ATA_DMA_ACTIVE 0x10 + #define ATA_48BIT_ACTIVE 0x40 struct ata_device device[2]; /* devices on this channel */ #define MASTER 0x00 *** sys/dev/ata/ata-dma.c.ORIG Wed Dec 31 19:05:16 2003 --- sys/dev/ata/ata-dma.c Tue Mar 1 18:52:33 2005 *************** *** 544,549 **** --- 544,550 ---- if (ata_find_dev(parent, 0x31471106, 0) || /* 8233a */ ata_find_dev(parent, 0x31771106, 0) || /* 8235 */ + ata_find_dev(parent, 0x32271106, 0) || /* 8237 */ ata_find_dev(parent, 0x31491106, 0)) { /* 8237 */ udmamode = imin(udmamode, 6); reg_val = via_modes[3]; *************** *** 1081,1088 **** break; case 0x4d69105a: /* Promise TX2 ATA133 controllers */ - case 0x5275105a: /* Promise TX2 ATA133 controllers */ case 0x6269105a: /* Promise TX2 ATA133 controllers */ case 0x7275105a: /* Promise TX2 ATA133 controllers */ ATA_OUTB(atadev->channel->r_bmio, ATA_BMDEVSPEC_0, 0x0b); if (udmamode >= 6 && --- 1082,1090 ---- break; case 0x4d69105a: /* Promise TX2 ATA133 controllers */ case 0x6269105a: /* Promise TX2 ATA133 controllers */ + case 0x1275105a: /* Promise TX2 ATA133 controllers */ + case 0x5275105a: /* Promise TX2 ATA133 controllers */ case 0x7275105a: /* Promise TX2 ATA133 controllers */ ATA_OUTB(atadev->channel->r_bmio, ATA_BMDEVSPEC_0, 0x0b); if (udmamode >= 6 && *************** *** 1410,1415 **** --- 1412,1431 ---- if (ds->flags & ATA_DS_ACTIVE) panic("ata_dmasetup: transfer active on this device!"); + switch(ch->chiptype) { + case 0x0d38105a: /* Promise Fasttrak 66 */ + case 0x4d38105a: /* Promise Ultra/Fasttrak 66 */ + case 0x0d30105a: /* Promise OEM ATA 100 */ + case 0x4d30105a: /* Promise Ultra/Fasttrak 100 */ + if (ch->flags & ATA_48BIT_ACTIVE) { + ATA_OUTB(ch->r_bmio, (ch->unit ? 0x09 : 0x11), + ATA_INB(ch->r_bmio, (ch->unit ? 0x09 : 0x11)) | + (ch->unit ? 0x08 : 0x02)); + ATA_OUTL(ch->r_bmio, (ch->unit ? 0x1c : 0x20), + (dir ? 0x05000000 : 0x06000000) | (count >> 1)); + } + } + cba.dmatab = ds->dmatab; bus_dmamap_sync(ds->cdmatag, ds->cdmamap, BUS_DMASYNC_PREWRITE); if (bus_dmamap_load(ds->ddmatag, ds->ddmamap, data, count, *************** *** 1421,1458 **** BUS_DMASYNC_PREWRITE); ch->flags |= ATA_DMA_ACTIVE; ! ds->flags = ATA_DS_ACTIVE; ! if (dir) ! ds->flags |= ATA_DS_READ; ATA_OUTL(ch->r_bmio, ATA_BMDTP_PORT, ds->mdmatab); ATA_OUTB(ch->r_bmio, ATA_BMCMD_PORT, dir ? ATA_BMCMD_WRITE_READ : 0); ATA_OUTB(ch->r_bmio, ATA_BMSTAT_PORT, ! (ATA_INB(ch->r_bmio, ATA_BMSTAT_PORT) | ! (ATA_BMSTAT_INTERRUPT | ATA_BMSTAT_ERROR))); ATA_OUTB(ch->r_bmio, ATA_BMCMD_PORT, ! ATA_INB(ch->r_bmio, ATA_BMCMD_PORT) | ATA_BMCMD_START_STOP); return 0; } int ata_dmadone(struct ata_device *atadev) { ! struct ata_channel *ch; ! struct ata_dmastate *ds; int error; ! ch = atadev->channel; ! ds = &atadev->dmastate; bus_dmamap_sync(ds->ddmatag, ds->ddmamap, (ds->flags & ATA_DS_READ) != 0 ? BUS_DMASYNC_POSTREAD : BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(ds->ddmatag, ds->ddmamap); - ATA_OUTB(ch->r_bmio, ATA_BMCMD_PORT, - ATA_INB(ch->r_bmio, ATA_BMCMD_PORT) & ~ATA_BMCMD_START_STOP); - error = ATA_INB(ch->r_bmio, ATA_BMSTAT_PORT); - ATA_OUTB(ch->r_bmio, ATA_BMSTAT_PORT, - error | ATA_BMSTAT_INTERRUPT | ATA_BMSTAT_ERROR); ch->flags &= ~ATA_DMA_ACTIVE; ds->flags = 0; return (error & ATA_BMSTAT_MASK); --- 1437,1483 ---- BUS_DMASYNC_PREWRITE); ch->flags |= ATA_DMA_ACTIVE; ! ds->flags = dir ? (ATA_DS_ACTIVE | ATA_DS_READ) : ATA_DS_ACTIVE; ATA_OUTL(ch->r_bmio, ATA_BMDTP_PORT, ds->mdmatab); ATA_OUTB(ch->r_bmio, ATA_BMCMD_PORT, dir ? ATA_BMCMD_WRITE_READ : 0); ATA_OUTB(ch->r_bmio, ATA_BMSTAT_PORT, ! (ATA_INB(ch->r_bmio, ATA_BMSTAT_PORT) | ! (ATA_BMSTAT_INTERRUPT | ATA_BMSTAT_ERROR))); ATA_OUTB(ch->r_bmio, ATA_BMCMD_PORT, ! ATA_INB(ch->r_bmio, ATA_BMCMD_PORT) | ATA_BMCMD_START_STOP); return 0; } int ata_dmadone(struct ata_device *atadev) { ! struct ata_channel *ch = atadev->channel; ! struct ata_dmastate *ds = &atadev->dmastate; int error; ! switch(ch->chiptype) { ! case 0x0d38105a: /* Promise Fasttrak 66 */ ! case 0x4d38105a: /* Promise Ultra/Fasttrak 66 */ ! case 0x0d30105a: /* Promise OEM ATA 100 */ ! case 0x4d30105a: /* Promise Ultra/Fasttrak 100 */ ! if (ch->flags & ATA_48BIT_ACTIVE) { ! ATA_OUTB(ch->r_bmio, (ch->unit ? 0x09 : 0x11), ! ATA_INB(ch->r_bmio, (ch->unit ? 0x09 : 0x11)) & ! ~(ch->unit ? 0x08 : 0x02)); ! ATA_OUTL(ch->r_bmio, (ch->unit ? 0x1c : 0x20), 0); ! } ! } ! ! error = ATA_INB(ch->r_bmio, ATA_BMSTAT_PORT); ! ATA_OUTB(ch->r_bmio, ATA_BMCMD_PORT, ! ATA_INB(ch->r_bmio, ATA_BMCMD_PORT) & ~ATA_BMCMD_START_STOP); ! ATA_OUTB(ch->r_bmio, ATA_BMSTAT_PORT,ATA_BMSTAT_INTERRUPT|ATA_BMSTAT_ERROR); ! bus_dmamap_sync(ds->ddmatag, ds->ddmamap, (ds->flags & ATA_DS_READ) != 0 ? BUS_DMASYNC_POSTREAD : BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(ds->ddmatag, ds->ddmamap); ch->flags &= ~ATA_DMA_ACTIVE; ds->flags = 0; return (error & ATA_BMSTAT_MASK); *** sys/dev/ata/ata-pci.c.ORIG Wed Dec 31 19:05:16 2003 --- sys/dev/ata/ata-pci.c Sun Jul 4 09:16:32 2004 *************** *** 190,195 **** --- 190,197 ---- return "VIA 8233 ATA133 controller"; if (ata_find_dev(dev, 0x31771106, 0)) return "VIA 8235 ATA133 controller"; + if (ata_find_dev(dev, 0x32271106, 0)) + return "VIA 8237 ATA133 controller"; if (ata_find_dev(dev, 0x31491106, 0)) return "VIA 8237 ATA133 controller"; return "VIA Apollo ATA controller"; ====================================================================== Note 3: Since the fixes merged from trunk work on RELENG_4, there seems to be a regression in RELENG_5. Note 4: Whenever I issue the command under "how-to-repeat" below, (under 5.4 RC2, of course), the disk will be trashed. The disk is a dangerously dedicated disk (see environment below). Trashing means that the first few sectors (including the first superblock) are somehow overwritten. The following are the corrupt: - the (pseudo) fdisk information - the disklabel for ad7 (should contain partitions c and e) - the file system, of course I can recover these three things if I run fsck directly in the full disk like this: fsck -b 32 /dev/ad7 Strangely enough, both the fdisk and disklabel seem to be revived using this operation. Note 5: The only thing I am not fully sure is whether the full size which is visible under RELENG_4_11_0_RELEASE using above patches is real... there might be a wraparound which I just have not noticed yet, the disk currently is about 40% full. But on the other hand, up till now everything seems to be fine. I also checked a random sampling of superblocks throughout the disk using fsck -b <from newfs output> -n /dev/ad7 under RELENG_4_11_0_RELEASE, and if there were indeed a wraparound, the file system data (40% disk full) should have already trashed the wrapped-around superblocks, if there were any wraparound at all. >How-To-Repeat: This will lead to a timeout on RELENG_5_4 dated 2005-04-14: dd if=/dev/ad... of=/dev/null bs=512 count=1 iseek=268435455 The same command under RELENG_4_11_0_RELEASE works find. >Fix: >Release-Note: >Audit-Trail: >Unformatted:
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200504172020.j3HKKGkk001724>