Timeout waiting for DMA
Community Support question - to convert into a paid question, click here
Lines in bold below have not yet been seen by the customer - those in blue are from the customer
| Username : Date : Action : Comments [ close all ] |
|
|
|
2hugo : 16/11/01 04:12 PM : Incident created |
|
- Hi,
I am newbie at Linux/Unix and attempted to install Mandrake
Linux 8.1 on a AMD 200 with 64 Mbytes of RAM and two volumes
(4 and 1.5 Giba respectivelly).
On volume 1 I have a:
a) Dos partition of 100 MBytes (for booting)
b) FAT 32 w/ Win95 ( /- 2G)
c) NTFS (a little less than 2G)
On volume 2 is:
a) With Linux (/dev/hdb)
I am currently booting from a floppy.
I have finally got the CD sound card working (tanks to drakxpert 8-))
but now I have the following error:
hdb: timeout waitng DMA
ide_dmaproc: chipset support ide_dma_timeout func only: 14
hdb: status error: status=0x58 { DriverReady SeekComplete DataRequest}
hdb: drive not ready for command
I have done some searching and have not found a definitive answers. Some
people point to chipset errors, cabe errors and disk failure. Others indicate that the DMA is by default on and disk may use PIO (?) which
means you have to switch it off (kernel recompile?, can only off it
and not on). Some still suggest changing the bus speed (hdparm?).
My question is: How do I begin to diagnose the problem or better yet
solve this. Please note I'm a newbie, hence the "?". Any help will be
appreciated.
TIA.
Hugo Ferreira.
|
|
|
|
Linegod_7611 : 16/11/01 05:09 PM : Reply received |
|
- It does sound like a chipset error. Immediately after booting up, as root
type:
# hdparm -d 0 /dev/hdb
This will disable DMA. If you no longer get the error, you can add this
line to the bottom of '/etc/rc.d/rc.local', and it will be issued during
each boot.
----
Note: If this answer resolves your problem, please remember to close this
question.
|
|
|
|
2hugo : 16/11/01 06:11 PM : More info provided |
|
- First and foremost how should I address you? Line? God? ;-)
Second, rest assure I intend to close any (all?) of my requests.
Third, my objective is learn Linux so I am also interested in a
little more information (if you have it of course). More messages means
more points, so you don't loose (provided I close the incident, naturally).
Is this ok w/ you?
Ok, to business. I will try your suggestion. Please note that I
will take some time to respond because I don't have my machine
with me (am at work, will usually have a lag of 1 day).
You say it may be a chipset error, but my machine is a real oldy.
How could I confirm this using the Linux tools?
Another question: does booting from a floppy limit me in any way
to change the OS's parameters. Question is specific to hparam, but
is Lilo the same thing? This is say I do change '/etc/rc.d/rc.local',
when I boot from a floppy, will this work?
TIA.
Hugo.
|
|
|
|
Linegod_7611 : 18/11/01 03:50 AM : Reply received |
|
- Linegod will do :)
'/etc/rc.d/rc.local' is loaded during the 'interactive' part of the bootup,
which is loaded from the harddrive, so booting from floppy or from the
master boot record will not make a difference, it will still be loaded.
I don't know of a way to truly determine if it is a chipset error but you
might do a search against your motherboard chipset on Google, and see if
anybody else has had similar problems.
Another thing to try is running a series of harddrive tests with another
harddrive if you have one available, with DMA enabled (hdparm -d 1 /dev/hdb)
and see if it fails. Just run 'hdparm -tiT /dev/hdb' and it will give an
output of the drive stats (type 'man hdparm' for more info on hdparm
options). If it generates your previous error on one drive but not the
other, it is a drive problem. If it generates the error on both, it is
most likely a chipset error.
|
|
|
|
2hugo : 19/11/01 01:41 PM : More info provided |
|
- Hi Linegod,
I did as you indicated, unfortunatelly something seems amiss.
After deactivating the DMA, the error messages that usually appeared
during KDE launch (only the first time) did not appear. The system
also seemed more responsive, but then again it may have just been me.
So far, so good.
I had checked the "hdparm" command and executed the following:
#root> hdparm -I /dev/hdb
and
#root> hdparm -i /dev/hdb
I also did this for the first volume.
Two things happend:
1. The /dev/hdb shows:
DMA modes: mdma0 mdma1 *mdma2
Seems like the DMA won't change.
2. I get a "segmentation fault" at the end of the above line.
Compared to the dev/hda, two lines are still missing. BTW where
can I see the core dump. Will this give more info. needed to
diagnose the problem?
So, something is not ok. I tried re-installing (pure
hell in this slow machine) and even changed BIOS settings but
it only made it worse.
Questions:
1. Can I have the /dev/hdb as a slave on IDE0
2. Will it make a difference if I set via BIOS the
disk to LBA or Normal?
3. One of your last suggestion is to use "hdparm -tiT /dev/hdb", I
will check waht this does and execute it (after a new install,
did not want to go through that at 2.0 a.m.)
As far as lloking for info. on my chipset, m searches have been
a little more generic, but can I trust Linux's identification of the
chipset's or do I have to open the machine and attempt to identify
the chip?
TIA.
Hugo.
|
|
|
|
Linegod_7611 : 20/11/01 10:14 AM : Reply received |
|
- Check the DMA status with 'hdparm -d /dev/hdb'
1. Your harddrive can be on any controller, in any position.
2. Your best bet would be to set it to LBA
3. Always trust Linux :)
|
|
|
|
2hugo : 21/11/01 11:02 AM : More info provided |
|
- Hi Linegod,
I have installed (once again) ML 8.1 (minimum install), uhhh...
This is more of a report back than anything else. I did as
you suggested and have the following results/conclusions:
1. Because I have installed RedHat 7.1 on this machine and
saw no errors (please note, newbies tend to loose sight of
important details, so this is not a guarantee) I doubt it's
a chipset problem (Intel's 82430VX). A "cat /proc/pci" shows
that the Chipset has been correctly identified.
2. I have searched far and wide and could not find any references
to problems or bugs concerning this chipset. The fact is Triton
chips are even refered in "man hdparm" which suggests a
configuration (hdparm -d1 -X34).
3. I tried disabling the DMA (hdparm -d 0 /dev/hdb) and
checking it's status (hdparm -d /dev/hdb) and this shows that
in fact DMA is off.
4. I then tried testing the drives with "hdparm -tT /dev/hdx"
where x is a or b. Please note that I did not use flags -i or -I
because of the core dump.
First conclusion is that on boot up the DMA is on, errors occur and DMA
remains on. After executing the test the 1st time DMA errors occurr continuously. After doing a Ctrl-C I get the following result (multiple
runs always bring back /- the same result):
/dev/hda/:
Timing buffer-cache reads: 128MB in 4.36 seconds = 29.36 MB/sec
Timing buffer reads: 64MB in 8.19 seconds = 7.81 MB/sec
and
/dev/hda/:
Timing buffer-cache reads: 128MB in 4.38 seconds = 29.22 MB/sec
Timing buffer reads: 64MB in 16.71 seconds = 3.83 MB/sec
Checking the DMA I see it is now off.
5. To set DMA off I guess the "/etc/rc.d/rc.local" is the only solution.
Please note that when testing this cenario, the error still occurs
during boot until the script is executed. I think it is also switched
off when KDE is launched (not sure, can't check no KDE installed).
6. Several problems are still nagging me:
a) The segmentation fault of hdparm. Should I report this
to hdparm's maintainer? Where can I get more info on this?
b) Time-outs when the chipset is supported.
c) The boot shows that: "ide: assuming 33MHz System bus speed
for PIO modes, override with idebus=xx". Can this have any bearing on
the problem?
7. All that seems to remain is the drive conflicts (disk drive,
IRQ's, modules). Please note that I only installed the modem and sound
card (auto-detected). No printer, zip drive or anything else.
Unless you have any more suggestions I will close this incident.
If by any miracle I have any more info, I will look for you here and
inform you.
Thanks for the help.
Hugo.
|
|
|
|
Linegod_7611 : 21/11/01 01:22 PM : Reply received |
|
- All I can say is that for a self-proclaimed 'newbie' you provide more useful
information than some 'experts' I deal with :)
6a. I would update hdparm from a cooker mirror first -
http://www.linux-mandrake.com/en/cookerdevel.php3
6b. After scouting around, the only recommendation I could come up with
would be to try an updated kernel - grab it from the link above.
6c. That is normal. There is a good explanation of it here:
http://list.cobalt.com/pipermail/cobalt-users/2001-May/042555.html
|
|
|
|
2hugo : 21/11/01 03:26 PM : More info provided |
|
- Linegod,
I'll take your first line as a compliment. 8-}
As far as your last recommendations, I will not
venture into such territory as yet (unless this
system does not hold). I need to learn a little more
before doing such 'complicated' stuff (ex: administration in
general and installing rpms specifically).
As for the link, I will look this up.
Again if I get anything interesting, I'll let you know.
In the meantime, I closing this "incident".
Thanks for help.
("Obrigado", in my native language)
Hugo
|
|
|
|
2hugo : 21/11/01 03:30 PM : Incident closed |
|
- I have a feeling that this problem is a direct result of bug specific to
the ML distro. Maybe these "incidents" should be scanned by the ML
engineers. It may actually provide interesting test-beds (if the people are
willig to help) to solve future release problems.
|
|
|
|
Language
Search Mandriva Expert
|
