Hi there: As suggested in the docs, I am forwarding a bug report. I have just converted two partitions to ext3 from ext2 without reformatting on my home box. All is not well, however. Here's the specs AMD K6/2-3D 500 Mhz, 64MB SDRAM, 100 Mhz bus Via Apollo P5MPV3 chipset, 100 Mhz bus Opti Mad 16 931 soundcard Video= AGP S3 Trio, w/8MB RAM, PS/2 mouse. Realtek 8029 pci bus network card. Award BIOS 56K external modem on /dev/ttyS1. PS/2 mouse. /dev/hda=2.5 Gig Maxtor 82560A4 drive(windoze 95) /dev/hdb=6.4 Gig Fujitsu MPC 3064AT Drive Unused usb ports. [root@genius /root]# hdparm /dev/hdb /dev/hdb: multcount = 2 (on) I/O support = 1 (32-bit) unmaskirq = 0 (off) using_dma = 0 (off) keepsettings = 0 (off) nowerr = 0 (off) readonly = 0 (off) readahead = 8 (on) geometry = 784/255/63, sectors = 12594960, start = 0 I have 2 large partitions on /dev/hdb for linux which are now ext3. Before the conversion, and still with a 2.4.3 kernel (not ext3 capable) hdparm -tT reutrns 37MB/sec & 10.5 MB/sec respectively; After converting as per the mini howto (Using defaults), and updating the kernel to 2.4.18, hdparm -tT returns 37MB/sec and 6.5MB/sec. What's worse, in a console, I get this error repeated many times invalidate: busy buffer but in an X terminal, I don't see it at all, although the speeds are similar. Is this dodgy hardware, kernel, or what? Where should I look for a fix? Please copy me direct - I'm not subscribed to this list yet - inbox blues. hdparm -d1 refuses to set dma [root@genius /root]# hdparm -d1 /dev/hdb /dev/hdb: setting using_dma to 1 (on) HDIO_SET_DMA failed: Operation not permitted using_dma = 0 (off) The system is Mandrake 8.0. Regards abd TIA, Declan Moriarty Applied Researches - Ireland's Foremost Electronic Hardware Genius A Slightly Serious(TM) Company Experience is like a comb, that Life gives you - AFTER all your hair has fallen out!
On Apr 04, 2002 18:21 +0100, Declan Moriarty wrote:> I have 2 large partitions on /dev/hdb for linux which are now ext3. Before > the conversion, and still with a 2.4.3 kernel (not ext3 capable) > hdparm -tT reutrns 37MB/sec & 10.5 MB/sec respectively; > After converting as per the mini howto (Using defaults), and updating the > kernel to 2.4.18, hdparm -tT returns 37MB/sec and 6.5MB/sec.That is purely a 2.4.3 -> 2.4.18 kernel issue. hdparm has nothing to do with the filesystem (it is communicating with the IDE hardware directly).> What's worse, in a console, I get this error repeated many times > invalidate: busy bufferThat is because hdparm is probably calling the "BLKFLSBUF" ioctl, which is trying to flush dirty buffers from the system to ensure an accurate test. Ext3 (and also ext2, reiserfs, ...) keeps some buffers on the device in-use while the filesystem is mounted, so it is not possible to flush them at all. In 2.4.3 kernels it silently ignored such busy buffers. In 2.4.18 it prints a warning, but the behaviour is the same.> in an X terminal, I don't see it at all, although the speeds are similar.That's because this is a kernel message and not output from hdparm, so it will only be sent to the console. You can also see it in your syslog or in dmesg output.> Is this dodgy hardware, kernel, or what?Neither.> Please copy me direct - I'm not subscribed to this list yet - inbox blues. > hdparm -d1 refuses to set dma > > [root@genius /root]# hdparm -d1 /dev/hdb > > /dev/hdb: > setting using_dma to 1 (on) > HDIO_SET_DMA failed: Operation not permitted > using_dma = 0 (off)This is the real speed issue. You probably need to compile your kernel with the appropriate PCI IDE chipset support enabled to get DMA. It may be that you need to use an -ac kernel (which has Andre Hedrick's IDE patches in it) to get full DMA support for your chipset. Cheers, Andreas -- Andreas Dilger \ "If a man ate a pound of pasta and a pound of antipasto, \ would they cancel out, leaving him still hungry?" www-mddsp.enel.ucalgary.ca/People/adilger -- Dogbert