On Thursday 27 February 2003 03:19 pm, J. Grant wrote:
> Hi Civilme,
>
> Thanks for the reply.
>
> > well first of all, if you can split ide drives to one per channel, that
> > would be more efficient.  ONly one device per channel on IDE may be
> > active, and two disks on the same channel really slows down disk-to-disk
> > operations
>
> ok, I will give this a go soon.
>
> > Next a DTLA and a DTTA on the same channel is a wide difference in timing
> > specs.  This leaves enough room for echo bounce on a signal gating line
> > to cause mischief.
> >
> > It is not that hard to move a drive if the installed systems have the
> > /etc/fstab modified  (of course with any winsystems we have a little more
> > work) to recognize the new location.
>
> no more winbloat here :)
>
> > The WD should not be paired on any channel with any other hard
> > drive,because its timing is very strange compared to the others and data
> > can be eaten by timing chatter.  It is safely default configured as you
> > can see, peaked at udma2
>
> ok, so is this what the linux kernel does when it boots up? it gives all
> those errors DriveSeek etc, then says device reset DMA disabled i think
> it was.
>
> > I would definitely try limiting the DTLA to udma2 as well --it appears to
> > be set initially for udma3
> >
> > the DTTA can go to udma4 or perhaps more.
>
> in the hpt bios I noticed some strange config. WD80GB can be set upto
> UDMA2 only. DTLA 60GB can be set upto UDMA5 max, DTLA 10GB can be set
> upto UDMA2 max.  So it seems strange that the WD can not go upto a
> faster speed as its ata100 hpt etc.
>
> > drakopt is in /contribs these days, but it does an alternative test and
> > setup for tuning, and it runs in python.  It may be able to set the
> > drives where they run reliably.  It requires some manual asistance yet
> > because I never had time to finish the parser for error messages nor to
> > make the settings within 2% of each other in tested speed "brothers" and
> > to choose the brother with the highest noise immunity.  Still you might
> > find the program useful for performing all the tests of various hdparm
> > settings--none of those are dangerous to existing data.
>
> i downloaded drakopt.  Unfortunatly it does not support my drives:
>
>
>
>
> OPTIONS:
>           (A)--Abort
>           (N)--New (or you changed disks around)
>           (C)--Continue from a previous run
> Your Choice (N/c/a)
> n
>
> Drive WDCWD800BB00CAA1 not found in database
> Attempting to use capabilities line in hdparm -i
> Please report driveid to [EMAIL PROTECTED]
>
> Capabilities line not available this drive model
> No optimizations possible this program
>
>
> Drive IBMDTLA307060 not found in database
> Attempting to use capabilities line in hdparm -i
> Please report driveid to [EMAIL PROTECTED]
>
> Capabilities line not available this drive model
> No optimizations possible this program
>
>
> Drive IBMDTTA351010 not found in database
> Attempting to use capabilities line in hdparm -i
> Please report driveid to [EMAIL PROTECTED]
>
> Capabilities line not available this drive model
> No optimizations possible this program
>
>
>
> I tried it on my laptop and the same problem existed.  I emailed
> [EMAIL PROTECTED]  with my aditions from my laptop and it bounced back.
>   Is there no maintainer now?  Are you still working on it at all?
>
> I added my laptop drive to the config but it still would not run...
>
> > I strongly suspect a timing problem between the two IBM drives and I
> > would suggest splitting them to different channels as a first step.  The
> > slow performance appears to be a fallback setting from probable timing
> > crosstalk.
> >
> > Civileme
>
> Also, there were some interesting logs about my system for you.  Strange
> that:
> /dev/hdh1 /mnt/backup ext3 defaults,user,auto 1 2
> only works when uncomented after boot, it will not mount at all during
> bootup for some reason.
>
>
>
> Feb 23 12:50:34 now1g kernel: ext3: No journal on filesystem on ide3(34,65)
> Feb 23 12:50:34 now1g mount: mount: wrong fs type, bad option, bad
> superblock on /dev/hdh1,
> Feb 23 12:50:34 now1g mount:        or too many mounted file systems
> Feb 23 12:50:34 now1g netfs: Mounting other filesystems:  failed
> Feb 23 12:50:31 now1g kernel: hda: DMA disabled
> Feb 23 12:50:31 now1g kernel: hdc: DMA disabled
>
>
> Feb 23 12:50:31 now1g kernel:     ide2: BM-DMA at 0xe800-0xe807, BIOS
> settings: hde:DMA, hdf:pio
> Feb 23 12:50:31 now1g kernel:     ide3: BM-DMA at 0xe808-0xe80f, BIOS
> settings: hdg:DMA, hdh:DMA
> Feb 23 12:50:31 now1g kernel: hda: LG DVD-ROM DRD-8120B, ATAPI
> CD/DVD-ROM drive
> Feb 23 12:50:31 now1g kernel: hdc: CR-48X5TE, ATAPI CD/DVD-ROM drive
> Feb 23 12:50:31 now1g kernel: hde: WDC WD800BB-00CAA1, ATA DISK drive
> Feb 23 12:50:31 now1g kernel: hdg: IBM-DTLA-307060, ATA DISK drive
> Feb 23 12:50:31 now1g kernel: hdh: IBM-DTTA-351010, ATA DISK drive
> Feb 23 12:50:31 now1g kernel: ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
> Feb 23 12:50:31 now1g kernel: ide1 at 0x170-0x177,0x376 on irq 15
> Feb 23 12:50:31 now1g kernel: ide2 at 0xd800-0xd807,0xdc02 on irq 11
> Feb 23 12:50:31 now1g kernel: ide3 at 0xe000-0xe007,0xe402 on irq 11
> Feb 23 12:50:31 now1g kernel: hde: 156301488 sectors (80026 MB)
> w/2048KiB Cache, CHS=155061/16/63, UDMA(100)
> Feb 23 12:50:31 now1g kernel: hdg: 120103200 sectors (61493 MB)
> w/1916KiB Cache, CHS=119150/16/63, UDMA(44)
> Feb 23 12:50:31 now1g kernel: hdh: 19807200 sectors (10141 MB) w/466KiB
> Cache, CHS=19650/16/63, UDMA(33)
>
>
>
>
> Well, thanks for the help so far, any futher sugestions appreciated.
>
> JG


Leave the WD at udma2.  WDs are not to be trusted above that water mark no 
matter what they claim--their hardware takes a half-hearted swipe at the 57 
byte CRC required for higher speeds and does something stupid and bad, which 
could write errors to the disk that will never iron out short of wiping the 
file written.  If the HPT controller is holding WD at udma2 as well as the 
linux kernel, then I say "bully for HPT".

NO I am not maintaining drakopt.  The kernel took over many of the functions 
of it and began tuning better.  You can look at the source and kill the exit 
function or better yet edit the database file which is plain text.

Civileme 

Want to buy your Pack or Services from MandrakeSoft? 
Go to http://www.mandrakestore.com

Reply via email to