Re: how to log reiser and raid0 crash? 2.6.0-t4

Re: how to log reiser and raid0 crash? 2.6.0-t4

[Resident Boxholder]

http://www.promise.com/product/product_detail_eng.asp?productId=87&familyId=3
promise bios flash page--
http://www.promise.com/support/download/download2_eng.asp?productId=87&category=bios&os=100
wget this bios flash if today--(will try later tonight)
http://www.promise.com/support/file/bios/ultra133tx2b220015.zip

Mudama, Eric wrote:

>LBA48 shouldn't affect 60GB drives.
>
>To my understanding, it is only an issue with drives >137GB (128GiB) that
>are moved to some controllers (promise?) after being used on other systems.
>
>--eric
>
True, but I think Alan Cox mentioned something about new code setting 
pio4 and udma6
at the same time so I was hoping new code might relate to system hangs 
on mkfs, fsck,
cp to raid on commodity-promise-controlled drives. I might try the lba48 
patch on a
null-modem serial setup in case there's other code in there. I could use 
a promise
sx-6000 according to Oliver Pitzeier oliver@linux-kernel.at (see below) 
but that's
an expensive onboard raid card, not commodity booty using linux software 
raid.
Lost time moots the expense issue but commodity servers need a cheap 
reliable controller
card, a tulip equivalent in the controller card category, to make a 
cheap net-speed
file-server.
-Bob D

>2.6.0-t4 amd 3000+ 1G four maxtor 60G drives on two 
>> controllers(mb's and promise)
>
As I always tell my customers: Promise means trouble; At least within Linux.

The only Promise controller that works fine for me is a SX-6000 with 6 drives (raid-5, 1 spare).

Best regards,
 Oliver








>
>-----Original Message-----
>From: Resident Boxholder [mailto:resid@boxho.com]
>Sent: Thursday, August 28, 2003 1:29 AM
>To: Oleg Drokin
>Subject: Re: how to log reiser and raid0 crash? 2.6.0-t4
>
>
>2.6.0-t4 amd 3000+ 1G four maxtor 60G drives on two controllers(mb's and 
>promise)
>
>problem isolated to promise card(have to verify the other is error-free 
>longer though)
>
>Oleg Drokin wrote:
>
>  
>
>>Hello!
>>
>>On Tue, Aug 26, 2003 at 03:57:39PM -0400, Resident Boxholder wrote:
>>
>> 
>>
>>    
>>
>>>I cause a lock up by doing a cp -aR /usr/src /mnt/usr...
>>>   
>>>
>>>      
>>>
>>Is there any chance of using sirial console to see if you can capture
>>    
>>
>something on that?
>  
>
>>Bye,
>>   Oleg
>>
>>    
>>
>I have a second linux pc and I can google whether to use null or regular 
>serial cable,
>howto put a console on serial, then log to that, I guess that's all. As 
>a bonus I won't
>need a kvm switch anymore.
>
>Here's some repeatable "high-level(monkey-level)" info.
>
>I have four drives, two on mboard controller and two on a 133 Promise card.
>If I make a four-drive four-partition raid zero md device, I get enough 
>lockups
>to draw my attention, so then I make two-drive md devices to see which
>controller works, if any. Without really flogging it I got ext2 and 
>reiserfs on
>the mboard's two drives to work for mkfs, fsck, and copy /usr/src/ /tmp
>so maybe this is a Promise problem. The mboard controller handles udma6
>no problem for that two-drive md.
>
>With only two drives on the promise card forming a raid zero md device,
>crashes happen on mkfs and fscks and even though no md's are mounted
>on boot, the boot won't happen sometimes after a crash.
>
>I have no errors logging anymore. The only one I got was by switching to
>vc/5 and seeing reiserfs info but I don't see that anymore since fixing one
>thing, it was because of mdadm and debian config conflict, forget that.
>The irq error storm was a false lead as well, no more such errors, no
>errors at all logged, since removing cd's and second promise card and
>turning usb off and turning apic off in bios and letting linux turn apic on.
>ACPI is smooth, no errors, just sudden death.
>
>Things I've tried are no fastrak in promise kernel config, hdparm no dma,
>hdparm udma4 instead of udma6. I'll try pio4 pio3 on the promise md
>and really flog the mboard-controlled md to verify that there is no
>problem except on the promise card. I'm reading what Alan Cox is
>saying about "LBA48 pio and udma" so I will try turning dma off
>and going down in pio modes.
>
>Swap is working with four drives, so two on the promise card, but with 1G of
>ram swap on the promise card may never be used.
>
>-Bob D
>
>-
>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/
>
>
>  
>

reiser4 snapshot for August 26th.

[Oleg Drokin]

Hello!

   I have just released another reiser4 snapshot that I hope all interested
   parties will try. It is released against 2.6.0-test4.
   You can find it at http://namesys.com/snapshots/2003.08.26
   I include release notes below.

Reiser4 snapshot for 2003.08.26

WARNING!!! This code is experimental! WE ARE NOT KIDDING! DO NOT PUT
ANY VALUABLE DATA ON REISER4 YET!

Fixed some bugs. And finally reiser4 should compile on 64bit boxes (hm. somebody try it,
as I am unable to build any 2.6 kernel for ia64). Also reiser4 should now build without
debug enabled.
Important SMP bug was fixed (only was in effect for SMP kernels on boxes with less than 3 CPUs).
There are still some OOM problems sometimes that we are working on, but generally
I hope problems reported by various people about compile failures should be fixed now.
Readonly mounts (and hence - reiser4 as rootfs) are not supported too.

reiser4progs update includes some 64 bit fixes too along with other stuff.
fsck still does not work, so don't even try to run it.

Snapshot is released as a patch against linux-2.6.0-test4
REISER4_NOOPT config option (Disable compiler optimizations for reiser4 code.)
is known to not compile on x86.
Enable CONFIG_REISER4_LARGE_KEY, as reiser4progs are creating filesystems with
large keys by default.
The reiser4-kernelonly.diff file represents only required kernel modifications if you plan to use our reiser4 bk repository.
    It does not contains any reiser4 code
The reiser4.diff is full patch with everything included.

Changelog (extracted by "bk changes"):

ChangeSet@1.1218.1.2, 2003-08-26 12:53:59+04:00, god@laputa.namesys.com
  remove assertion (contradicts recent changes in prepare_twig_cut())

ChangeSet@1.1218.1.1, 2003-08-26 12:53:28+04:00, god@laputa.namesys.com
  remove superfluous check

ChangeSet@1.1217.1.3, 2003-08-26 11:19:03+04:00, zam@crimson.namesys.com
  search_one_bitmap_backward() endless loop fix.

ChangeSet@1.1216.1.1, 2003-08-25 22:01:52+04:00, zam@crimson.namesys.com
  backward search bitmap code fixes

ChangeSet@1.1217.1.1, 2003-08-25 21:26:46+04:00, edward@theta.namesys.com
  ctail.c:
    Fixed bug
  file.c:
    shared reserve_cut_iteration
  cryptcompress.c:
    Added cut_items_cryptcompress(), 
    fixed bugs 

ChangeSet@1.1215.1.6, 2003-08-25 21:06:06+04:00, god@laputa.namesys.com
  sync with recent change in prepare_twig_cut()

ChangeSet@1.1215.1.5, 2003-08-25 21:00:04+04:00, god@laputa.namesys.com
  when removing eottl, clear connected bits on formatted neighbors

ChangeSet@1.1215.1.4, 2003-08-25 20:59:26+04:00, god@laputa.namesys.com
  restore old check.

ChangeSet@1.1215.1.3, 2003-08-25 20:59:13+04:00, god@laputa.namesys.com
  fix race in check_jload()

ChangeSet@1.1215.1.2, 2003-08-25 20:58:52+04:00, god@laputa.namesys.com
  add separate inode to keep bitmaps.
  cleanup

ChangeSet@1.1215.1.1, 2003-08-25 20:45:18+04:00, god@laputa.namesys.com
  fix fence error in wake_up_requestor()

ChangeSet@1.1216, 2003-08-25 18:05:17+04:00, green@angband.namesys.com
  bitmap.c:
    Remove erroneous ON_DEBUG()

ChangeSet@1.1215, 2003-08-25 13:58:15+04:00, god@laputa.namesys.com
  use vmalloc instead of kmalloc

ChangeSet@1.1208.1.1, 2003-08-25 08:49:07+04:00, zam@crimson.namesys.com
  bitmap code reorganization

ChangeSet@1.1213, 2003-08-24 15:56:24+04:00, green@angband.namesys.com
  inode.h:
    64bit compile fix

ChangeSet@1.1212, 2003-08-22 21:05:40+04:00, god@laputa.namesys.com
  fix compilation errors

ChangeSet@1.1211, 2003-08-22 21:00:56+04:00, god@laputa.namesys.com
  restore compilability.

ChangeSet@1.1210, 2003-08-22 20:56:38+04:00, god@laputa.namesys.com
  fix problems with BK's handling of symlinks

ChangeSet@1.1209, 2003-08-22 20:55:24+04:00, god@laputa.namesys.com
  Merge laputa.namesys.com:/home/god/projects/2.5/fs/reiser4
  into laputa.namesys.com:/home/god/projects/i386/fs/reiser4

ChangeSet@1.1185.1.17, 2003-08-22 20:45:52+04:00, god@laputa.namesys.com
  cleanup

ChangeSet@1.1185.1.16, 2003-08-22 20:45:32+04:00, god@laputa.namesys.com
  add scripts to perform mongo automatically.

ChangeSet@1.1185.1.15, 2003-08-22 20:44:24+04:00, god@laputa.namesys.com
  fix race in sibling list code.

ChangeSet@1.1185.1.14, 2003-08-22 20:43:54+04:00, god@laputa.namesys.com
  fix debugging checks.

ChangeSet@1.1185.1.13, 2003-08-22 20:43:15+04:00, god@laputa.namesys.com
  don't do read-ahead if low on memory.

ChangeSet@1.1185.1.12, 2003-08-22 20:42:55+04:00, god@laputa.namesys.com
  add assertion.

ChangeSet@1.1185.1.11, 2003-08-22 20:42:39+04:00, god@laputa.namesys.com
  fix assertion: atom lock should be held to avoid races with fusion.

ChangeSet@1.1185.1.10, 2003-08-22 20:42:13+04:00, god@laputa.namesys.com
  use jnode attached to page to avoid hash table lookup when possible

ChangeSet@1.1185.1.9, 2003-08-22 20:41:38+04:00, god@laputa.namesys.com
  add sysfs file with a list of all active atoms.

ChangeSet@1.1185.1.8, 2003-08-22 20:40:59+04:00, god@laputa.namesys.com
  don't set JNODE_DKSET in add_empty_leaf() to avoid races

ChangeSet@1.1185.1.7, 2003-08-22 20:40:23+04:00, god@laputa.namesys.com
  move check.

ChangeSet@1.1185.1.6, 2003-08-22 20:40:08+04:00, god@laputa.namesys.com
  fix race in eflush_del().
  fix hash function.
  make hash table dynamically sized.

ChangeSet@1.1185.1.5, 2003-08-22 20:39:29+04:00, god@laputa.namesys.com
  mark get_current_context() as const.

ChangeSet@1.1185.1.4, 2003-08-22 20:39:06+04:00, god@laputa.namesys.com
  remove noisy compiler options

Re: reiser4 snapshot for August 26th.

[Steven Cole]

On Tue, 2003-08-26 at 04:22, Oleg Drokin wrote:
> Hello!
> 
>    I have just released another reiser4 snapshot that I hope all interested
>    parties will try. It is released against 2.6.0-test4.
>    You can find it at http://namesys.com/snapshots/2003.08.26
>    I include release notes below.
> 
> Reiser4 snapshot for 2003.08.26
> 

I got this error while attempting to compile with reiser4.
Snippet from .config follows.

Steven

  CC      fs/reiser4/plugin/file/tail_conversion.o
  CC      fs/reiser4/sys_reiser4.o
fs/reiser4/sys_reiser4.c:54:32: parser/parser.code.c: No such file or directory
fs/reiser4/sys_reiser4.c: In function `sys_reiser4':
fs/reiser4/sys_reiser4.c:75: warning: implicit declaration of function `reiser4_pars_init'
fs/reiser4/sys_reiser4.c:75: warning: assignment makes pointer from integer without a cast
fs/reiser4/sys_reiser4.c:80: error: dereferencing pointer to incomplete type
fs/reiser4/sys_reiser4.c:82: warning: implicit declaration of function `yyparse'
fs/reiser4/sys_reiser4.c:83: warning: implicit declaration of function `reiser4_pars_free'
fs/reiser4/sys_reiser4.c:66: warning: unused variable `Gencode'
fs/reiser4/sys_reiser4.c: At top level:
fs/reiser4/parser/parser.h:333: warning: `Fistmsg' defined but not used
fs/reiser4/parser/parser.h:342: warning: `typesOfCommand' defined but not used
fs/reiser4/parser/parser.h:354: warning: `Code' defined but not used
make[2]: *** [fs/reiser4/sys_reiser4.o] Error 1
make[1]: *** [fs/reiser4] Error 2
make: *** [fs] Error 2
[steven@spc1 linux-2.6.0-test4-r4]$ grep REISER4 .config
CONFIG_REISER4_FS=y
CONFIG_REISER4_FS_SYSCALL=y
CONFIG_REISER4_LARGE_KEY=y
# CONFIG_REISER4_CHECK is not set
CONFIG_REISER4_USE_EFLUSH=y
# CONFIG_REISER4_BADBLOCKS is not set

Re: reiser4 snapshot for August 26th.

[Alex Zarochentsev]

On Tue, Aug 26, 2003 at 12:20:38PM -0600, Steven Cole wrote:
> On Tue, 2003-08-26 at 04:22, Oleg Drokin wrote:
> > Hello!
> > 
> >    I have just released another reiser4 snapshot that I hope all interested
> >    parties will try. It is released against 2.6.0-test4.
> >    You can find it at http://namesys.com/snapshots/2003.08.26
> >    I include release notes below.
> > 
> > Reiser4 snapshot for 2003.08.26
> > 
> 
> I got this error while attempting to compile with reiser4.
> Snippet from .config follows.

Disable "reiser4 system call" (CONFIG_REISER4_FS_SYSCALL) support, it is 
not ready.

> 
> Steven
> 
>   CC      fs/reiser4/plugin/file/tail_conversion.o
>   CC      fs/reiser4/sys_reiser4.o
> fs/reiser4/sys_reiser4.c:54:32: parser/parser.code.c: No such file or directory
> fs/reiser4/sys_reiser4.c: In function `sys_reiser4':
> fs/reiser4/sys_reiser4.c:75: warning: implicit declaration of function `reiser4_pars_init'
> fs/reiser4/sys_reiser4.c:75: warning: assignment makes pointer from integer without a cast
> fs/reiser4/sys_reiser4.c:80: error: dereferencing pointer to incomplete type
> fs/reiser4/sys_reiser4.c:82: warning: implicit declaration of function `yyparse'
> fs/reiser4/sys_reiser4.c:83: warning: implicit declaration of function `reiser4_pars_free'
> fs/reiser4/sys_reiser4.c:66: warning: unused variable `Gencode'
> fs/reiser4/sys_reiser4.c: At top level:
> fs/reiser4/parser/parser.h:333: warning: `Fistmsg' defined but not used
> fs/reiser4/parser/parser.h:342: warning: `typesOfCommand' defined but not used
> fs/reiser4/parser/parser.h:354: warning: `Code' defined but not used
> make[2]: *** [fs/reiser4/sys_reiser4.o] Error 1
> make[1]: *** [fs/reiser4] Error 2
> make: *** [fs] Error 2
> [steven@spc1 linux-2.6.0-test4-r4]$ grep REISER4 .config
> CONFIG_REISER4_FS=y
> CONFIG_REISER4_FS_SYSCALL=y
> CONFIG_REISER4_LARGE_KEY=y
> # CONFIG_REISER4_CHECK is not set
> CONFIG_REISER4_USE_EFLUSH=y
> # CONFIG_REISER4_BADBLOCKS is not set
> 
> 

-- 
Alex.

Re: reiser4 snapshot for August 26th.

[Felipe Alfaro Solana]

On Tue, 2003-08-26 at 20:26, Alex Zarochentsev wrote:

> Disable "reiser4 system call" (CONFIG_REISER4_FS_SYSCALL) support, it is 
> not ready.

[...]
  CC      lib/string.o
lib/string.c:435: warning: conflicting types for built-in function
`bcopy'
  CC      lib/vsprintf.o
  AR      lib/lib.a
  LD      arch/i386/lib/built-in.o
  AS      arch/i386/lib/checksum.o
  CC      arch/i386/lib/dec_and_lock.o
  CC      arch/i386/lib/delay.o
  AS      arch/i386/lib/getuser.o
  CC      arch/i386/lib/memcpy.o
  CC      arch/i386/lib/strstr.o
  CC      arch/i386/lib/usercopy.o
  AR      arch/i386/lib/lib.a
  GEN     .version
  CHK     include/linux/compile.h
  UPD     include/linux/compile.h
  CC      init/version.o
  LD      init/built-in.o
  LD      .tmp_vmlinux1
arch/i386/kernel/built-in.o(.data+0x7c4): In function `sys_call_table':
: undefined reference to `sys_reiser4'
make[2]: *** [.tmp_vmlinux1] Error 1
make[1]: *** [vmlinux] Error 2

[...]
CONFIG_REISER4_FS=m
# CONFIG_REISER4_FS_SYSCALL is not set
# CONFIG_REISER4_LARGE_KEY is not set
# CONFIG_REISER4_CHECK is not set
# CONFIG_REISER4_USE_EFLUSH is not set
# CONFIG_REISER4_BADBLOCKS is not set
# CONFIG_REISERFS_FS is not set
# CONFIG_JFS_FS is not set
[...]

how to log reiser and raid0 crash? 2.6.0-t4

[Resident Boxholder]

2.6.0-test4 amd xp 3000+ msi mbo nforce2  four Maxtor 60G 8mbcache raid0

I cause a lock up by doing a cp -aR /usr/src /mnt/usr which moves data 
larger
than total hard buffer cache, to raid0 reiserfs or ext2 ( NOT reiser4!) 
Copy ops
smaller than buffer cache(8mb x 4 = 32mb) don't fail. Nothing fails on a 
single
drive, compiles or copies, just copy to a mounted raid0 device. What 
should I
try, test, or dump?

No irq error storm. No cd drives installed. Smaller copy ops work. Turning
swap off first slows things down enough to work, but swap itself is probably
OK. I have bios turn apic off, then linux turns it on, which is good 
until the
turn apic off before turn apic on patch gets into test5 or whatever.

hdparm sets all four drives the same, udma6 but have tried down to udma4
and pio4 and dma turned off, unmask on or off.

If two drives are on the mboard controller and two on a promise pci card,
does /proc/ide/amd74xx refer to the two drives on the motherboard only?
It seems to mention four drives but the two slow ones might just be a
ref to unoccupied slave drive positions.

I'm wondering what to send in. Maybe I could send a log from successful
copy with swap off, showing reiser logging, and config, in case a stress
condition or misconfig shows up even when catastrophic failure doesn't
occur. With swap on the fail is sudden and no error logging is coming
through.

I could incrementally copy /usr/src to one raid, then do a copy from that
raid to another raid. All that would do is test copying from one balanced
set of four drives/partitions to another balanced set, versus copying from
one drive's /usr/src to that and three other drives' raid set, which is
unbalanced, dragging on one drive.

-Bob

Re: how to log reiser and raid0 crash? 2.6.0-t4

[Resident Boxholder]

I had syslog send just *.info to vc/5 and though the *.info messages 
never make
it into syslog file, I started cp -aR /usr/src/kernel-source-2.6.0 /tmp 
with md/2 mounted
to /tmp and saw this *.info on vc/5 and then it locked up.  How about 
those zeros, skip,
assign id anyway?

journal_1037: journal_read_transaction, offset 3623, len 0 mount_id 0
journal_1039: journal_read_trans skipping because 3623 is too old
journal_1299: Setting newest_mount_id to 12

Anything?

-Bob D

Resident Boxholder wrote:

> 2.6.0-test4 amd xp 3000+ msi mbo nforce2  four Maxtor 60G 8mbcache raid0
>
> I cause a lock up by doing a cp -aR /usr/src /mnt/usr which moves data 
> larger
> than total hard buffer cache, to raid0 reiserfs or ext2 ( NOT 
> reiser4!) Copy ops
> smaller than buffer cache(8mb x 4 = 32mb) don't fail. Nothing fails on 
> a single
> drive, compiles or copies, just copy to a mounted raid0 device. What 
> should I
> try, test, or dump?
>
> No irq error storm. No cd drives installed. Smaller copy ops work. 
> Turning
> swap off first slows things down enough to work, but swap itself is 
> probably
> OK. I have bios turn apic off, then linux turns it on, which is good 
> until the
> turn apic off before turn apic on patch gets into test5 or whatever.
>
> hdparm sets all four drives the same, udma6 but have tried down to udma4
> and pio4 and dma turned off, unmask on or off.
>
> If two drives are on the mboard controller and two on a promise pci card,
> does /proc/ide/amd74xx refer to the two drives on the motherboard only?
> It seems to mention four drives but the two slow ones might just be a
> ref to unoccupied slave drive positions.
>
> I'm wondering what to send in. Maybe I could send a log from successful
> copy with swap off, showing reiser logging, and config, in case a stress
> condition or misconfig shows up even when catastrophic failure doesn't
> occur. With swap on the fail is sudden and no error logging is coming
> through.
>
> I could incrementally copy /usr/src to one raid, then do a copy from that
> raid to another raid. All that would do is test copying from one balanced
> set of four drives/partitions to another balanced set, versus copying 
> from
> one drive's /usr/src to that and three other drives' raid set, which is
> unbalanced, dragging on one drive.
>
> -Bob
>
> -
> 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: how to log reiser and raid0 crash? 2.6.0-t4

[Oleg Drokin]

Hello!

On Tue, Aug 26, 2003 at 03:57:39PM -0400, Resident Boxholder wrote:

> I cause a lock up by doing a cp -aR /usr/src /mnt/usr which moves data 
> larger
> than total hard buffer cache, to raid0 reiserfs or ext2 ( NOT reiser4!) 
> I'm wondering what to send in. Maybe I could send a log from successful
> copy with swap off, showing reiser logging, and config, in case a stress
> condition or misconfig shows up even when catastrophic failure doesn't
> occur. With swap on the fail is sudden and no error logging is coming
> through.

Is there any chance of using sirial console to see if you can capture something on that?

Bye,
    Oleg

Re: how to log reiser and raid0 crash? 2.6.0-t4

[Resident Boxholder]

2.6.0-t4 amd 3000+ 1G four maxtor 60G drives on two controllers(mb's and 
promise)

problem isolated to promise card(have to verify the other is error-free 
longer though)

Oleg Drokin wrote:

>Hello!
>
>On Tue, Aug 26, 2003 at 03:57:39PM -0400, Resident Boxholder wrote:
>
>  
>
>>I cause a lock up by doing a cp -aR /usr/src /mnt/usr...
>>    
>>
>
>Is there any chance of using sirial console to see if you can capture something on that?
>
>Bye,
>    Oleg
>
I have a second linux pc and I can google whether to use null or regular 
serial cable,
howto put a console on serial, then log to that, I guess that's all. As 
a bonus I won't
need a kvm switch anymore.

Here's some repeatable "high-level(monkey-level)" info.

I have four drives, two on mboard controller and two on a 133 Promise card.
If I make a four-drive four-partition raid zero md device, I get enough 
lockups
to draw my attention, so then I make two-drive md devices to see which
controller works, if any. Without really flogging it I got ext2 and 
reiserfs on
the mboard's two drives to work for mkfs, fsck, and copy /usr/src/ /tmp
so maybe this is a Promise problem. The mboard controller handles udma6
no problem for that two-drive md.

With only two drives on the promise card forming a raid zero md device,
crashes happen on mkfs and fscks and even though no md's are mounted
on boot, the boot won't happen sometimes after a crash.

I have no errors logging anymore. The only one I got was by switching to
vc/5 and seeing reiserfs info but I don't see that anymore since fixing one
thing, it was because of mdadm and debian config conflict, forget that.
The irq error storm was a false lead as well, no more such errors, no
errors at all logged, since removing cd's and second promise card and
turning usb off and turning apic off in bios and letting linux turn apic on.
ACPI is smooth, no errors, just sudden death.

Things I've tried are no fastrak in promise kernel config, hdparm no dma,
hdparm udma4 instead of udma6. I'll try pio4 pio3 on the promise md
and really flog the mboard-controlled md to verify that there is no
problem except on the promise card. I'm reading what Alan Cox is
saying about "LBA48 pio and udma" so I will try turning dma off
and going down in pio modes.

Swap is working with four drives, so two on the promise card, but with 1G of
ram swap on the promise card may never be used.

-Bob D