From mboxrd@z Thu Jan 1 00:00:00 1970 From: OGAWA Hirofumi Subject: Re: Fw: [Bugme-new] [Bug 1340] New: small FAT32 partitions cannot be used properly Date: Mon, 13 Oct 2003 03:13:39 +0900 Sender: linux-ide-owner@vger.kernel.org Message-ID: <87zng6iass.fsf@devron.myhome.or.jp> References: <20031010081436.7652c427.akpm@osdl.org> <87llrrty3z.fsf@devron.myhome.or.jp> <3F89400B.5020607@king.net.pl> <87brsm8tg6.fsf@devron.myhome.or.jp> <3F898EEA.6090907@king.net.pl> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from mail.parknet.co.jp ([210.171.160.6]:21510 "EHLO mail.parknet.co.jp") by vger.kernel.org with ESMTP id S263500AbTJLSOJ (ORCPT ); Sun, 12 Oct 2003 14:14:09 -0400 In-Reply-To: <3F898EEA.6090907@king.net.pl> List-Id: linux-ide@vger.kernel.org To: Paul Osmialowski Cc: Andrew Morton , B.Zolnierkiewicz@elka.pw.edu.pl, linux-ide@vger.kernel.org Hi, Paul, thanks a lot for test. Andrew, this looks like IDE bug with PIO-mode. So could you please handle the bugzilla? I don't know about bugzilla. Bartlomiej, are you interesting the following problem? Any idea? Paul Osmialowski writes: > Dear Ogawa! > I did more test and I discovered: > 1. DriveReady SeekComplete Error was always present and DMA was disabled. > 2. Whenever I connected CF disk alone (as master or slave) I WAS ABLE > TO MOUNT THIS PARTITION IN READ/WRITE MODE!!! That's the solution for > now. > 3. Whenever I connected CF disk as slave with ATAPI disk as master I > couldn't mount this partition properly. > 4. I was always able to read boot sector using fdisk. > > The problem is with operating CF TrueIDE disk as a slave. I'm not sure > but I think I didn't try to mount this partition when I was using CF > disk alone (no master or slave on the same cable), so I was quite > surprised when I saw no dmesg error messages after such mount. > The problem should be fixed anyway, since it's not present on M$ > Windows. It may be caused by some piece of code written for Compact > Flash volumes, since they're mentioned in dmesg as CFA DISK drive > while other HDD's are mentioned as ATA DISK drive. > Since I'm new to bugzilla I'd like to ask you to redirect my problem > report to proper section of kernel developers. Two issues are > important: > 1. DriveReady SeekComplete Error message in dmesg and DMA disabled (I > think CFs are PIO-only anyway). > 2. Partitions cannot be mounted properly when CF is used as a slave > with other drive on the same cable (i.e. hdb with hda present). > Greetings, > Paul > > Uz.ytkownik OGAWA Hirofumi napisa?: > > Paul Osmialowski writes: > > > >>hdb: max request size: 128KiB > >>hdb: 125952 sectors (64 MB) w/1KiB Cache, CHS=246/16/32, BUG > >> hdb:hdb: dma_intr: status=0x51 { DriveReady SeekComplete Error } > >>hdb: dma_intr: error=0x00 { } > >>hdb: dma_intr: status=0x51 { DriveReady SeekComplete Error } > >>hdb: dma_intr: error=0x00 { } > >>hdb: dma_intr: status=0x51 { DriveReady SeekComplete Error } > >>hdb: dma_intr: error=0x00 { } > >>hdb: dma_intr: status=0x51 { DriveReady SeekComplete Error } > >>hdb: dma_intr: error=0x00 { } > >>hda: DMA disabled > >>hdb: DMA disabled > >>ide0: reset: success > > Um. Looks like hdb has hardware or driver(PIO-mode) problem. > > Could you test with /dev/hdb1? If it has corruption problem, > > completely not filesystem layer bug. So it should goes to IDE > > maintainer. > > Thanks. > From: Paul Osmialowski > Subject: Re: Fw: [Bugme-new] [Bug 1340] New: small FAT32 partitions cannot > be used properly > To: OGAWA Hirofumi > Date: Sun, 12 Oct 2003 13:50:35 +0200 > User-Agent: Mozilla/5.0 (X11; U; Linux i686; PL; rv:1.4) Gecko/20030704 > Content-Type: text/plain; charset=ISO-8859-1; format=flowed > > Hello Ogawa! > I've checked twice everything and I've noticed, that problem is > somewhere else. I should have done that before I've wrote anything to > bugzilla, it's my fault, sorry. > First. Hardware background of the problem. To do the test I used my PC: > - popular Microstar motherboard 845PE Max2 with Pentium IV 1.7GHz > - 64MB CompactFlash connected as TrueIDE through CF-25 and IDE25-35 > connectors made by Port80h (http://www.port80h.com.pl/cf25.html, > http://www.port80h.com.pl/ide35_hdd25.html), once as a slave > (/dev/hdb) with 60GB Seagate disk as a master, once alone as a master > (/dev/hdc). Also I was switching on/off BIOS Setup option "Use 32-bit > transfer mode" - it didn't make any differences. (CF-25 internally is > something like that: http://213.25.10.65/~newchief/2.6/IDE-CF.gif with > exception that it can work only in +5V mode, which is acceptable for > most of compactflash cards). > Second. Why I assumed it's FAT module problem: > - fdisk on both Linux and Windows can see and create partitions properly > - mkdosfs can create FAT32 filesystem without any warnings (Windows > create only FAT16, since it has no -F parameter and consequently > cannot create FAT32 on partition smaller than 512MB) > - After creating FAT32 filesystem on Linux using mkdosfs -F32 I was > able to operate it without any problem on Windows98 so I assumed Linux > can't operate non-standard FAT partition properly. I didn't check > kernel code, I looked at filesystem internals and I've found only one > non-standard thing: one sector per cluster (while smallest standard > cluster is 8 sectors). > At that point I wrote to bugzilla. > Finally. After reading your mail I've checked twice dmesg (you can > see it here: http://213.25.10.65/~newchief/2.6/dmesg.txt) (how could I > forget to do that before?) and I've found: > > hdb: max request size: 128KiB > hdb: 125952 sectors (64 MB) w/1KiB Cache, CHS=246/16/32, BUG > hdb:hdb: dma_intr: status=0x51 { DriveReady SeekComplete Error } > hdb: dma_intr: error=0x00 { } > hdb: dma_intr: status=0x51 { DriveReady SeekComplete Error } > hdb: dma_intr: error=0x00 { } > hdb: dma_intr: status=0x51 { DriveReady SeekComplete Error } > hdb: dma_intr: error=0x00 { } > hdb: dma_intr: status=0x51 { DriveReady SeekComplete Error } > hdb: dma_intr: error=0x00 { } > hda: DMA disabled > hdb: DMA disabled > ide0: reset: success > > Gosh! It switched off DMA for both of my HDD! (hda, hdb). I suspect > Microstars 845PE Max2 MB has problems with DMA, sometimes music on > xmms (SB Live!) stops for a short while when disk I/O works heavy > (RedHat9, 2.4.20 kernel with lots of RedHat patches that makes it > closer to 2.6 line). > You can also read hdparm -i /dev/hdb output > (http://213.25.10.65/~newchief/2.6/hdparm-i.txt) and hdparm /dev/hdb > output (http://213.25.10.65/~newchief/2.6/hdparm.txt), both made while > using 2.6.0-test7-bk3. > Maybe TrueIDE protocol of comact flash isn't so true? Anyway, how can > it work properly on Windows9x then? And why boot sector is accessed > properly? Why BIOS Setup discovers disk right way (and says it's > LBA-compatible)? > I was also able to operate properly this 64MB card with non-PC device > that is able to operate FAT32 and it worked fine (I've put it in place > of 2.5'' 20GB HDD that was used by default, again I used CF-25 to do > that). -- OGAWA Hirofumi