Oddities with HighPoint HPT374, 2.4.19-pre10-ac2

Oddities with HighPoint HPT374, 2.4.19-pre10-ac2

[Mike Insch]

Hi,

I have a system with an ABIT AT7 Motherboard (Duron 1200, 256MB PC2100 RAM), 
and have 8 80GB Maxtor IDE HDD's connected to the HPT374 controller on the 
board, configured as a single RAID-0 Array in hardware.

When copying data from another disk connected to the VIA IDE Controller, the 
kernel oopses with an 'Unable to handle NULL pointer dereference at virtual 
address 00000004' after about 5MB has copied.  The process implicated by the 
oops is the kjournald process for the only partition on the RAID array.

Sorry I have no output from the oops - it locks totally, and the oops is never 
logged :(

Copying large amounts of data from a Samba Share on the network (over 15GB!) 
has produced no similar problems - so I'm guessing that kjournald, or the 
HPT374 drivers don't like the 640GB array when data is being committed too 
fast. (All the drives involved are UDMA 133 Drives, both the one on the VIA 
Controller, and all 8 on the HPT Controller).

I know it's hard to say without me giving more info, but does anyone have any 
idea what could possibly cause this?  Have there been updates in newer 
releases to either the HPT374 Code, or to the kjournald code that could solve 
this?  Is it worth me getting and compiling 2.4.19-rc3-ac3 to see if the 
problem is solved there?  Any and all info. which may help tracking down this 
gremlin would be greatly appreciated....

(I can post detailed info. about the hardware (lspci, dmesg etc), and kernel 
configurations if that would be of any use?)

Re: Oddities with HighPoint HPT374, 2.4.19-pre10-ac2

[Andre Hedrick]

The oddities are the fact I have not finished creating the channel and
drive set inner-lock sequence code :-/  If you run it in master mode only
one drive per channel, it is perfectly safe.  The problem happens the
instant you pair drives on the channels.  I have not figured out the
correct interrupt sequence ordering and blocking recipe.

The issue stands to deal with double hwgroups locking and having more than
two channels to the effective HBA.

Instead of 

HPT374
  ide4
  ide5
HPT374
  ide6
  ide7

It has to evlolve to

HPT374
  ide4p
  ide4s
  ide5p
  ide5s

This means extened minor on the second half of each major must be created.
There is not enough sane bandwith in the driver, nor would anybody take
such an exotic solution.

Sorry,

Andre Hedrick
LAD Storage Consulting Group

On Thu, 25 Jul 2002, Mike Insch wrote:

> Hi,
> 
> I have a system with an ABIT AT7 Motherboard (Duron 1200, 256MB PC2100 RAM), 
> and have 8 80GB Maxtor IDE HDD's connected to the HPT374 controller on the 
> board, configured as a single RAID-0 Array in hardware.
> 
> When copying data from another disk connected to the VIA IDE Controller, the 
> kernel oopses with an 'Unable to handle NULL pointer dereference at virtual 
> address 00000004' after about 5MB has copied.  The process implicated by the 
> oops is the kjournald process for the only partition on the RAID array.
> 
> Sorry I have no output from the oops - it locks totally, and the oops is never 
> logged :(
> 
> Copying large amounts of data from a Samba Share on the network (over 15GB!) 
> has produced no similar problems - so I'm guessing that kjournald, or the 
> HPT374 drivers don't like the 640GB array when data is being committed too 
> fast. (All the drives involved are UDMA 133 Drives, both the one on the VIA 
> Controller, and all 8 on the HPT Controller).
> 
> I know it's hard to say without me giving more info, but does anyone have any 
> idea what could possibly cause this?  Have there been updates in newer 
> releases to either the HPT374 Code, or to the kjournald code that could solve 
> this?  Is it worth me getting and compiling 2.4.19-rc3-ac3 to see if the 
> problem is solved there?  Any and all info. which may help tracking down this 
> gremlin would be greatly appreciated....
> 
> (I can post detailed info. about the hardware (lspci, dmesg etc), and kernel 
> configurations if that would be of any use?)
> 
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

Re: Oddities with HighPoint HPT374, 2.4.19-pre10-ac2

[Mike Insch]

OK - Thanks for the Info.

For the moment, the machine in question will be simply running as a SAMBA 
Server, and then only for one or two clients at a time (it's storing disk 
images from PowerQuest's Drive Image Pro Software), and it (so far) seems to 
be stable enough when data is coming in from the network.

If there is any detailed info. I can give you from the system, I'll be happy 
to pop them over to you - similarly, if I can help test any revised versions 
in the future, I'd be happy to do so.

Thanks again... :)

On Thursday 25 July 2002 13:25, Andre Hedrick wrote:
> The oddities are the fact I have not finished creating the channel and
> drive set inner-lock sequence code :-/  If you run it in master mode only
> one drive per channel, it is perfectly safe.  The problem happens the
> instant you pair drives on the channels.  I have not figured out the
> correct interrupt sequence ordering and blocking recipe.
>
> The issue stands to deal with double hwgroups locking and having more than
> two channels to the effective HBA.
>
> Instead of
>
> HPT374
>   ide4
>   ide5
> HPT374
>   ide6
>   ide7
>
> It has to evlolve to
>
> HPT374
>   ide4p
>   ide4s
>   ide5p
>   ide5s
>
> This means extened minor on the second half of each major must be created.
> There is not enough sane bandwith in the driver, nor would anybody take
> such an exotic solution.
>
> Sorry,
>
> Andre Hedrick
> LAD Storage Consulting Group
>