( that's a capital-aye in the hdparm line )
not even the Magic SysReq key will work.
also, don't
"cd /proc/ide/hde ; cat identify"
... same thing
drive-light comes on, but have to use the power-switch to get the machine
back, ( lost stuff again, fuck )
proc says it's pdc202xx
Promise Ultra series driver Ver 1.20.0.7 2002-05-23
Adapter: Ultra100 on M/B
--
http://www.drawright.com/
- "The New Drawing on the Right Side of the Brain" ( Betty Edwards,
check "Theory", "Gallery", and "Exercises" )
http://www.ldonline.org/ld_indepth/iep/seven_habits.html
- "The 7 Habits of Highly Effective People" ( this site is same
principles as Covey's book )
http://www.eiconsortium.org/research/ei_theory_performance.htm
- "Working With Emotional Intelligence" ( Goleman: this link is
/revised/ theory, "Working. . . " is practical )
http://www.leadershipnow.com/leadershop/1978-5.html
- Corps Business: The 30 /Management Principles/ of the U.S. Marines (
David Freedman )
> have to use the power-switch to get the machine back
If you have another terminal accessible, you could try:
hdparm -w /dev/hda
to reset the interface. I can't guarantee that it wouldn't loose
data, though.
John.
man, the Magic SysReq key didn't work ( at all ):
it were DEAD
The drive-light stayed on for 10+ hours, nothing happening ( that I could
figure out ) the whole time. It /stayed/ dead.
/dev/hde is part of a RAID-5 in my system ( because I no longer trust
anything else ), and this only happens on drives connected onto the
Promise controller.
Oh, yeah, I forgot to include this:
trying to touch/activate/read the S.M.A.R.T. in any drive on the Promise
kills it, too. Can't activate the reliability-system without killing
the kernel? /that's/ ironic, eh?
As for having another terminal connected to my home machine...
1. if the kernel's dead, then how's that gonna work, and
2. why have 2 terminals on one machine when I'm a hermit?
I /do/ thank you for the interface-reset tip, though, I hope I never need
that info : )
On Sun 15 December, 2002 16:39, you wrote:
>> have to use the power-switch to get the machine back
>
>If you have another terminal accessible, you could try:
>
>hdparm -w /dev/hda
>
>to reset the interface. I can't guarantee that it wouldn't loose
>data, though.
>
>John.
--
http://www.drawright.com/
- "The New Drawing on the Right Side of the Brain" ( Betty Edwards,
check "Theory", "Gallery", and "Exercises" )
http://www.ldonline.org/ld_indepth/iep/seven_habits.html
- "The 7 Habits of Highly Effective People" ( this site is same
principles as Covey's book )
http://www.eiconsortium.org/research/ei_theory_performance.htm
- "Working With Emotional Intelligence" ( Goleman: this link is
/revised/ theory, "Working. . . " is practical )
http://www.leadershipnow.com/leadershop/1978-5.html
- Corps Business: The 30 /Management Principles/ of the U.S. Marines (
David Freedman )
> man, the Magic SysReq key didn't work ( at all ):
> it were DEAD
> The drive-light stayed on for 10+ hours, nothing happening ( that I could
> figure out ) the whole time. It /stayed/ dead.
>
> /dev/hde is part of a RAID-5 in my system ( because I no longer trust
> anything else ), and this only happens on drives connected onto the
> Promise controller.
>
> Oh, yeah, I forgot to include this:
> trying to touch/activate/read the S.M.A.R.T. in any drive on the Promise
> kills it, too. Can't activate the reliability-system without killing
> the kernel? /that's/ ironic, eh?
>
>
> As for having another terminal connected to my home machine...
> 1. if the kernel's dead, then how's that gonna work, and
Maybe just the console was not responding.
If I start X with /dev/null as the core pointer, the console locks
completely, but I can still log in on a serial terminal.
I have seen machines which will mostly stop responding when you issue
a sleep command to a disk, E.G.
hdparm -Y /dev/hda
you can't terminate the process with control-C, for example, but if
you are logged in on another virtual terminal, or have another
terminal window open in X, you can reset the interface, and the
machine will respond again.
> 2. why have 2 terminals on one machine when I'm a hermit?
Why not? I read and write a lot of E-Mail on a serial terminal right
next to my main console, and what about debugging SVGALIB applications?
> I /do/ thank you for the interface-reset tip, though, I hope I never need
> that info : )
It can be useful for recovering from a spun-down disk that won't spin
up again :-)
John
The promise chips often respond to starnge situations by locking up the
PCI bus. In particular they assert the wait signal and do not release
it. This locks the system up had the next time the CPU tries to access
the PCI bus. The machine is dead in your case and needs to be reset.
I've sent a PCI bus trace of this happening to Promise and have not yet
heard anything back yet.
Ross
John Bradford wrote:
>>man, the Magic SysReq key didn't work ( at all ):
>>it were DEAD
>>The drive-light stayed on for 10+ hours, nothing happening ( that I could
>>figure out ) the whole time. It /stayed/ dead.
>>
>>/dev/hde is part of a RAID-5 in my system ( because I no longer trust
>>anything else ), and this only happens on drives connected onto the
>>Promise controller.
>>
>>Oh, yeah, I forgot to include this:
>>trying to touch/activate/read the S.M.A.R.T. in any drive on the Promise
>>kills it, too. Can't activate the reliability-system without killing
>>the kernel? /that's/ ironic, eh?
>>
>>
>>As for having another terminal connected to my home machine...
>>1. if the kernel's dead, then how's that gonna work, and
>>
>>
>
>Maybe just the console was not responding.
>
>If I start X with /dev/null as the core pointer, the console locks
>completely, but I can still log in on a serial terminal.
>
>I have seen machines which will mostly stop responding when you issue
>a sleep command to a disk, E.G.
>
>hdparm -Y /dev/hda
>
>you can't terminate the process with control-C, for example, but if
>you are logged in on another virtual terminal, or have another
>terminal window open in X, you can reset the interface, and the
>machine will respond again.
>
>
>
>>2. why have 2 terminals on one machine when I'm a hermit?
>>
>>
>
>Why not? I read and write a lot of E-Mail on a serial terminal right
>next to my main console, and what about debugging SVGALIB applications?
>
>
>
>>I /do/ thank you for the interface-reset tip, though, I hope I never need
>>that info : )
>>
>>
>
>It can be useful for recovering from a spun-down disk that won't spin
>up again :-)
>
>John
>-
>To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>the body of a message to [email protected]
>More majordomo info at http://vger.kernel.org/majordomo-info.html
>Please read the FAQ at http://www.tux.org/lkml/
>
>
There is a bug in the Promise driver that clears an important PIO bit
when switching into DMA mode. When you do an hdparm -I, it issues a
drive command that attempts to transfer data in PIO mode, but since the
PIO mode timing registers are hosed, the machine locks up. It's easy to
reproduce and applies to all drive commands that return data including
SMART commands.
The bit in particular is bit 4 of PCI config register 0x61+4*channel
number (PB bit 4 in Promise terms.) I've got a very unclean fix that I
will attempt to clean up once I can put a few more important issues to bed.
For the time being, you can try to do a work around by putting the drive
into PIO mode with hdparm -X 12 before issuing any drive commands.
Ross
D.A.M. Revok wrote:
>( that's a capital-aye in the hdparm line )
>
>not even the Magic SysReq key will work.
>
>also, don't
>
>"cd /proc/ide/hde ; cat identify"
>
>... same thing
>drive-light comes on, but have to use the power-switch to get the machine
>back, ( lost stuff again, fuck )
>
>
>proc says it's pdc202xx
>
>Promise Ultra series driver Ver 1.20.0.7 2002-05-23
>Adapter: Ultra100 on M/B
>
>
>
On Wed, 18 Dec 2002, Ross Biro wrote:
>
> There is a bug in the Promise driver that clears an important PIO bit
> when switching into DMA mode. When you do an hdparm -I, it issues a
> drive command that attempts to transfer data in PIO mode, but since the
> PIO mode timing registers are hosed, the machine locks up. It's easy to
> reproduce and applies to all drive commands that return data including
> SMART commands.
>
> The bit in particular is bit 4 of PCI config register 0x61+4*channel
> number (PB bit 4 in Promise terms.) I've got a very unclean fix that I
> will attempt to clean up once I can put a few more important issues to bed.
>
> For the time being, you can try to do a work around by putting the drive
> into PIO mode with hdparm -X 12 before issuing any drive commands.
>
> Ross
>
> D.A.M. Revok wrote:
>
> >( that's a capital-aye in the hdparm line )
> >
> >not even the Magic SysReq key will work.
> >
> >also, don't
> >
> >"cd /proc/ide/hde ; cat identify"
> >
> >... same thing
> >drive-light comes on, but have to use the power-switch to get the machine
> >back, ( lost stuff again, fuck )
> >
> >
> >proc says it's pdc202xx
> >
> >Promise Ultra series driver Ver 1.20.0.7 2002-05-23
> >Adapter: Ultra100 on M/B
And this is the drive hack job that Promise did to it in 2.4.19.
This is not my driver version and you need to nail Marcelo for this issue.
Wait, move to 2.4.20 and it may go away. Better yet go back to 2.4.18 and
it should be clean.
Regards,
Andre Hedrick
LAD Storage Consulting Group
Andre Hedrick wrote:
>And this is the drive hack job that Promise did to it in 2.4.19.
>This is not my driver version and you need to nail Marcelo for this issue.
>Wait, move to 2.4.20 and it may go away. Better yet go back to 2.4.18 and
>it should be clean.
>
I'm not sure if the problem code is in the patch from Promise, but I can
say we have applied the promise supplied patch to 2.4.18 and as a whole
it is a nightmare. I don't recomend it if you don't need it.
Ross
Denis Vlasenko wrote:
>OTOH mere mortals are allowed to make full dump of PCI config ;)
>
>
>
Some vendors use index/data registers in the config space, so unless you
know of their existance, a PCI config dump doesn't help.
>Clearly Promise is the one storage vendor whose products are best avoided.
>
>Andre, could you give a recommendation on what add-on IDE controllers are
>not junk hardware and will work nicely with Linux? 'Cos I can't seem to
>remember seeing anything in the shelves other than Promise or CMD64X/68X.
I'm no IDE RAID expert, however I've build quite a few linux servers
for customers.
If IDE raid unix linux is needed here, there's no question that a
linux supported (thanks adam) 3ware card is dropped in, no questions
asked. For my 3 years of working with them under linux THEY JUST
WORK. Native monitoring tools included too..
-Jason
On Sat, 21 Dec 2002, Jason Radford wrote:
> If IDE raid unix linux is needed here, there's no question that a
> linux supported (thanks adam) 3ware card is dropped in, no questions
> asked. For my 3 years of working with them under linux THEY JUST
> WORK. Native monitoring tools included too..
or get an outboard RAID box that uses IDE disks and SCSI for its
connection to the host. plenty of them around if you google. (eg
fibrenetix.co.uk - we have one, and works nicely. reasonably fast
too.)
> -Jason
regards,
--
Paul Jakma Sys Admin Alphyra
[email protected]
Warning: /never/ send email to [email protected] or [email protected]