[9fans] fossil crash after installation

[9fans] fossil crash after installation

[Antonin Vecera]

[-- Attachment #1: Type: text/plain, Size: 971 bytes --]

Hello all,

I have installed Plan 9 from CD on some older PC. But after a while (from a
few minutes to 10 minutes) the system stalled and on the screen I could see:

     fossil: diskWriteRaw failed: /dev/sdD0/fossil: 0x00004509: date=Thu
Jan  1 21:29:00 EST 1970 part=data addr=17673: i/o error

     fossil(#S/sdD0/fossil)...version...boot: mount /: corrupted meta data
     panic: boot process died: unknown

I had to format the harddisk and made new installation.
Several times but no success. A few minutes after installation when I
started configuration of system files the OS panics...

PC is small booksize PC - Lightbook - from www.lex.com.tw
CPU VIA Eden 500MHz
chipset VIA PLE133 (VT8601A + VT82C686B)
LAN 3x 10/100Base-T, RJ-45, Realtek 8100B

I disabled in BIOS all USB, COM, LPT, audio and left only disk controller +
1 LAN
I have no idea what else.

Does anybody know?

(I tried to install MS-Windows and it was OK)


Antonin

[-- Attachment #2: Type: text/html, Size: 1138 bytes --]

Re: [9fans] fossil crash after installation

[Steve Simon]

> A few minutes after installation when I
> started configuration of system files the OS panics...

It sounds to me like plan9 doesn't know about the quirks of
your IDE controller, even though it obviously thinks it does.

You could try turning off disk i/o DMA just in case that helps,
DMA seems to be the quirkest area in the IDE interface.

	echo 'dma off' > /dev/sd00/ctl

Unfortunately this will have a significant performance hit but
if it works reliably then you could start looking at the datasheets
for your chipset to try to bugfix it.

-Steve

Re: [9fans] fossil crash after installation

[Charles Forsyth]

> DMA seems to be the quirkest area in the IDE interface.

you must remember that dma is remarkable technology
newly devised for the PC platform, and it has been understandably
hard to get it just right.

Re: [9fans] fossil crash after installation

[W B Hacker]

Charles Forsyth wrote:
>> DMA seems to be the quirkest area in the IDE interface.
>
> you must remember that dma is remarkable technology
> newly devised for the PC platform, and it has been understandably
> hard to get it just right.
>
>

Surely you jest?

Bill Godbout's 'Disk One' FDD controller had it on S-100, and it was nowhere
near 'new' even then. Nor dual-ported memory. Both available on the GA-SPC12 IIRC.

BIOS and driver code is what seems to be forever 'hard to get right'.

Smothers Brothers rules seem to apply:  "If you borrow it, break it!"

Bill

Re: [9fans] fossil crash after installation

[Charles Forsyth]

> Surely you jest?

evidently.

Re: [9fans] fossil crash after installation

[Charles Forsyth]

> you must remember that dma is remarkable technology
> newly devised for the PC platform, and it has been understandably
> hard to get it just right.

obviously i refer to the hardware interfaces.
if it weren't new and difficult, IDE/ATA/SATA (... get this right yet!) drive
interfaces would all be the same and all work, wouldn't they?
there can't be any other possible explanation.
it's not like VGA where all the interfaces are simple and the same  ...
just a moment ... oh, no!  that's all different and differently buggy as well.
i wonder how to sack an entire industry.

(have you seen USB-to-Go?)

Re: [9fans] fossil crash after installation

[W B Hacker]

Charles Forsyth wrote:
>> you must remember that dma is remarkable technology
>> newly devised for the PC platform, and it has been understandably
>> hard to get it just right.
>
> obviously i refer to the hardware interfaces.
> if it weren't new and difficult, IDE/ATA/SATA (... get this right yet!) drive
> interfaces would all be the same and all work, wouldn't they?
> there can't be any other possible explanation.
> it's not like VGA where all the interfaces are simple and the same  ...
> just a moment ... oh, no!  that's all different and differently buggy as well.
> i wonder how to sack an entire industry.
>

T'was once both simpler and more obvious wherein the limitations were to be found.

Software - or more specifically, 'firmware' - not hardware.

At least once designers got their collective arms well and truly around clock
distribution, gate-delay, and timing issues in general - most of which were in
the 'long ago solved' category once legacy ISA-bus stuff - IDE in particular -
moved into a ~bridge' chipset.

Unfortunately, today's 'solution' to marginal silicon or PCB implementations
seems to be to program around it with bespoke BIOS / driver kludges that may
satisfy one OS but not another.

> (have you seen USB-to-Go?)
>
>

Not sure I want to...

;-)

Bill

Re: [9fans] fossil crash after installation

[Antonin Vecera]

[-- Attachment #1: Type: text/plain, Size: 383 bytes --]

On 6/4/07, Steve Simon <steve@quintile.net> wrote:
>
>
> You could try turning off disk i/o DMA just in case that helps,
> DMA seems to be the quirkest area in the IDE interface.
>
>         echo 'dma off' > /dev/sd00/ctl
>


Do you think it is enough to disable DMA in BIOS?
Or is it necessary to disable it in Plan 9?
Should I care what is set in PIO mode?

Antonin

[-- Attachment #2: Type: text/html, Size: 793 bytes --]

Re: [9fans] fossil crash after installation

[Antonin Vecera]

[-- Attachment #1: Type: text/plain, Size: 250 bytes --]

On 6/4/07, Steve Simon <steve@quintile.net> wrote:
>
>
>         echo 'dma off' > /dev/sd00/ctl
>
>
Where is the best place to disable DMA ?
This command can be used in termrc (or cpurc) after kernel is loaded. But
isn't it late?

Antonin

[-- Attachment #2: Type: text/html, Size: 582 bytes --]

Re: [9fans] fossil crash after installation

[erik quanstrom]

> Where is the best place to disable DMA ?
> This command can be used in termrc (or cpurc) after kernel is loaded. But
> isn't it late?
>
> Antonin

dma is left as bios left it. c.f. /sys/src/9/pc/sdata.c:/^atadmamode
so disabling dma in bios should do the trick.  you can verify this
with

	; /cat /dev/sdC0/ctl | grep config
	config 427A capabilities 2F00 dma 00550040 dmactl 00000000 rwm 16 rwmctl 0 lba48always off

the number following dmactl will be non-zero if dma is on.

- erik

Re: [9fans] fossil crash after installation

[Russ Cox]

> dma is left as bios left it. c.f. /sys/src/9/pc/sdata.c:/^atadmamode
> so disabling dma in bios should do the trick.  you can verify this
> with
>
> 	; /cat /dev/sdC0/ctl | grep config
> 	config 427A capabilities 2F00 dma 00550040 dmactl 00000000 rwm 16 rwmctl 0 lba48always off
>
> the number following dmactl will be non-zero if dma is on.

not true; dma is left off unless explicitly enabled with

	echo dma on >/dev/sdC0/ctl

that has on occasion been put in termrc but i don't
see it there now.

dma does get enabled in the termrc used for
installation, but i don't believe it stays enabled
in the base system unless you modify your
termrc yourself.

what is true is that the specific dma mode used
when dma is enabled is whatever got set up by
the bios.  but whether or not to use dma at all
is a separate question and not one that plan 9
defers to the bios for.

russ