HighPoint 374

DISCUSSION Svetoslav Slavtchev 7 participants 25 messages Oct 19, 2003 – Oct 21, 2003

Share & tools

Svetoslav Slavtchev

author

2003-10-19 19:57 UTC #0 lore raw

i have the same problems with epox 8k9a3+, and may be even strange ones
(like fs coruption when soft raid & / or lvm is used)

and i never had the problems with 8k5a3+,
the drives were actually taken from the 8k5a3+
when it died (me silly tried to update to XP2700)

really strange, isn't it

both boards should be the same, except
8k5a3+ uses kt333
8k9a3+ uses kt400

only mainboard change -> hell of a lot unresolved problems

svetljo
kernels used 2.4.21-2.4.23-pre3 2.6.0-test3-test7bk8

and a nice log when i try to enable TCQ

all Trace: [<c0235ee3>] [<c022e834>] [<c025b0ef>] [<c026e0ed>] [<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>] Badness in as_remove_dispatched_request at drivers/block/as-iosched.c:1022

Call Trace: [<c0235ee3>]  [<c022e834>]  [<c025b0ef>]  [<c026e0ed>] 
[<c025c7e2>]  [<c026e080>]  [<c010df03>]  [<c010e233>]  [<c010c7d8>]
Badness in as_remove_dispatched_request at drivers/block/as-iosched.c:1022
Call Trace: [<c0235ee3>]  [<c022e834>]  [<c025b0ef>]  [<c026e0ed>] 
[<c025c7e2>]  [<c026e080>]  [<c010df03>]  [<c010e233>]  [<c010c7d8>]
Badness in as_remove_dispatched_request at drivers/block/as-iosched.c:1022
Call Trace: [<c0235ee3>]  [<c022e834>]  [<c025b0ef>]  [<c026e0ed>]

[<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>]

Signature · 6 lines

NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

Andre Hedrick

2003-10-19 20:35 UTC #1 lore raw

You do not enable TCQ on highpoint without using the hosted polling timer. Oh and I have not added it, and so hit Bartlomiej up for the additions.

Andre Hedrick
LAD Storage Consulting Group

Quoted 49 lines · Svetoslav Slavtchev

i have the same problems with epox 8k9a3+, and may be even strange ones
(like fs coruption when soft raid & / or lvm is used)
and i never had the problems with 8k5a3+,
the drives were actually taken from the 8k5a3+
when it died (me silly tried to update to XP2700)
really strange, isn't it
both boards should be the same, except
8k5a3+ uses kt333
8k9a3+ uses kt400
only mainboard change -> hell of a lot unresolved problems
svetljo
kernels used 2.4.21-2.4.23-pre3 2.6.0-test3-test7bk8
and a nice log when i try to enable TCQ
all Trace: [<c0235ee3>] [<c022e834>] [<c025b0ef>] [<c026e0ed>] [<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>] Badness in as_remove_dispatched_request at drivers/block/as-iosched.c:1022
Call Trace: [<c0235ee3>]  [<c022e834>]  [<c025b0ef>]  [<c026e0ed>] 
[<c025c7e2>]  [<c026e080>]  [<c010df03>]  [<c010e233>]  [<c010c7d8>]
Badness in as_remove_dispatched_request at drivers/block/as-iosched.c:1022
Call Trace: [<c0235ee3>]  [<c022e834>]  [<c025b0ef>]  [<c026e0ed>] 
[<c025c7e2>]  [<c026e080>]  [<c010df03>]  [<c010e233>]  [<c010c7d8>]
Badness in as_remove_dispatched_request at drivers/block/as-iosched.c:1022
Call Trace: [<c0235ee3>]  [<c022e834>]  [<c025b0ef>]  [<c026e0ed>] 
[<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>]
Signature · 12 lines
NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

-
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/

Svetoslav Slavtchev

author

2003-10-19 20:46 UTC #2 lore raw

Quoted 3 lines

You do not enable TCQ on highpoint without using the hosted polling timer. Oh and I have not added it, and so hit Bartlomiej up for the additions.

good info :-)

but it sometimes seems to enable it

the 4 drives are identical,
and the drive failing is almost always different

and this doesn't explain the md/ lvm corupption,
even when TCQ is not enabled both in 2.4 and 2.5

/dev/hda:

ATA device, with non-removable media powers-up in standby; SET FEATURES subcmd spins-up.

        Model Number:       IC35L080AVVA07-0
        Serial Number:      VNC402A4CVHNEA
        Firmware Revision:  VA4OA52A

Standards:
Used: ATA/ATAPI-5 T13 1321D revision 1
Supported: 5 4 3 2 & some of 6
Configuration:

        Logical         max     current
        cylinders       16383   16383
        heads           16      16
        sectors/track   63      63
        --
        CHS current addressable sectors:   16514064
        LBA    user addressable sectors:  160836480
        device size with M = 1024*1024:       78533 MBytes
        device size with M = 1000*1000:       82348 MBytes (82 GB)

Capabilities:
LBA, IORDY(can be disabled)
bytes avail on r/w long: 52 Queue depth: 32
Standby timer values: spec'd by Standard, no device specific minimum R/W multiple sector transfer: Max = 16 Current = 0 Advanced power management level: unknown setting (0x0000) Recommended acoustic management value: 128, current value: 254
DMA: mdma0 mdma1 mdma2 udma0 udma1 udma2 udma3 udma4 *udma5
Cycle time: min=120ns recommended=120ns
PIO: pio0 pio1 pio2 pio3 pio4
Cycle time: no flow control=240ns IORDY flow control=120ns
Commands/features:
Enabled Supported:
* NOP cmd
* READ BUFFER cmd
* WRITE BUFFER cmd
* Host Protected Area feature set
Release interrupt
* Look-ahead
* Write cache
* Power Management feature set
Security Mode feature set
* SMART feature set
* Mandatory FLUSH CACHE command
* Device Configuration Overlay feature set
.................................................................
* READ/WRITE DMA QUEUED
...................................................................

Quoted 58 lines

Quoted 17 lines
i have the same problems with epox 8k9a3+, and may be even strange ones
(like fs coruption when soft raid & / or lvm is used)
and i never had the problems with 8k5a3+,
the drives were actually taken from the 8k5a3+
when it died (me silly tried to update to XP2700)
really strange, isn't it
both boards should be the same, except
8k5a3+ uses kt333
8k9a3+ uses kt400
only mainboard change -> hell of a lot unresolved problems
Quoted 8 lines
svetljo
kernels used 2.4.21-2.4.23-pre3 2.6.0-test3-test7bk8
and a nice log when i try to enable TCQ
all Trace: [<c0235ee3>] [<c022e834>] [<c025b0ef>] [<c026e0ed>] [<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>] Badness in as_remove_dispatched_request at
drivers/block/as-iosched.c:102
2
Quoted 3 lines
Call Trace: [<c0235ee3>]  [<c022e834>]  [<c025b0ef>]  [<c026e0ed>] 
[<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>] Badness in as_remove_dispatched_request at
drivers/block/as-iosched.c:102
2
Quoted 3 lines
Call Trace: [<c0235ee3>]  [<c022e834>]  [<c025b0ef>]  [<c026e0ed>] 
[<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>] Badness in as_remove_dispatched_request at
drivers/block/as-iosched.c:102
2
Quoted 14 lines
Call Trace: [<c0235ee3>]  [<c022e834>]  [<c025b0ef>]  [<c026e0ed>] 
[<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>]
Signature · 9 lines
NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel"
i
n
Quoted 4 lines
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/

Signature · 6 lines

NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

Bartlomiej Zolnierkiewicz

2003-10-19 20:48 UTC #3 lore raw

In reply to Andre Hedrick

Andre, thanks for helpful hint.
Svetoslav, the right person to whine about TCQ stuff is Jens Axboe 8-).

--bartlomiej

Quoted 41 lines · Andre Hedrick

You do not enable TCQ on highpoint without using the hosted polling timer. Oh and I have not added it, and so hit Bartlomiej up for the additions.
Andre Hedrick
LAD Storage Consulting Group
Quoted 34 lines · Svetoslav Slavtchev
i have the same problems with epox 8k9a3+, and may be even strange ones
(like fs coruption when soft raid & / or lvm is used)
and i never had the problems with 8k5a3+,
the drives were actually taken from the 8k5a3+
when it died (me silly tried to update to XP2700)
really strange, isn't it
both boards should be the same, except
8k5a3+ uses kt333
8k9a3+ uses kt400
only mainboard change -> hell of a lot unresolved problems
svetljo
kernels used 2.4.21-2.4.23-pre3 2.6.0-test3-test7bk8
and a nice log when i try to enable TCQ
all Trace: [<c0235ee3>] [<c022e834>] [<c025b0ef>] [<c026e0ed>] [<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>] Badness in as_remove_dispatched_request at drivers/block/as-iosched.c:1022 Call Trace: [<c0235ee3>] [<c022e834>] [<c025b0ef>] [<c026e0ed>] [<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>] Badness in as_remove_dispatched_request at drivers/block/as-iosched.c:1022 Call Trace: [<c0235ee3>] [<c022e834>] [<c025b0ef>] [<c026e0ed>] [<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>] Badness in as_remove_dispatched_request at drivers/block/as-iosched.c:1022 Call Trace: [<c0235ee3>] [<c022e834>] [<c025b0ef>] [<c026e0ed>] [<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>]

Svetoslav Slavtchev

author

2003-10-19 20:54 UTC #4 lore raw

Quoted 3 lines

Andre, thanks for helpful hint.
Svetoslav, the right person to whine about TCQ stuff is Jens Axboe 8-).

j don't really care about TCQ that much anymore
(i bought the drives because they were marked as supported)

but the fs corruption when md/lvm is used,
and the dma timeouts ;(

svetljo

Quoted 47 lines · Andre Hedrick

Quoted 1 line
You do not enable TCQ on highpoint without using the hosted polling
timer.
Quoted 43 lines
Oh and I have not added it, and so hit Bartlomiej up for the additions.
Andre Hedrick
LAD Storage Consulting Group
Quoted 37 lines · Svetoslav Slavtchev
i have the same problems with epox 8k9a3+, and may be even strange ones
(like fs coruption when soft raid & / or lvm is used)
and i never had the problems with 8k5a3+,
the drives were actually taken from the 8k5a3+
when it died (me silly tried to update to XP2700)
really strange, isn't it
both boards should be the same, except
8k5a3+ uses kt333
8k9a3+ uses kt400
only mainboard change -> hell of a lot unresolved problems
svetljo
kernels used 2.4.21-2.4.23-pre3 2.6.0-test3-test7bk8
and a nice log when i try to enable TCQ
all Trace: [<c0235ee3>] [<c022e834>] [<c025b0ef>] [<c026e0ed>] [<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>] Badness in as_remove_dispatched_request at drivers/block/as-iosched.c:1022 Call Trace: [<c0235ee3>] [<c022e834>]
[<c025b0ef>] [<c026e0ed>] [<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>] Badness in as_remove_dispatched_request at drivers/block/as-iosched.c:1022 Call Trace: [<c0235ee3>] [<c022e834>]
[<c025b0ef>] [<c026e0ed>] [<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>] Badness in as_remove_dispatched_request at drivers/block/as-iosched.c:1022 Call Trace: [<c0235ee3>] [<c022e834>]
[<c025b0ef>] [<c026e0ed>] [<c025c7e2>] [<c026e080>] [<c010df03>] [<c010e233>] [<c010c7d8>]

Signature · 6 lines

NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

Jens Axboe

2003-10-20 06:49 UTC #5 lore raw

In reply to Bartlomiej Zolnierkiewicz

Quoted 3 lines · Bartlomiej Zolnierkiewicz

Andre, thanks for helpful hint.
Svetoslav, the right person to whine about TCQ stuff is Jens Axboe 8-).

Well that's correct, but this looks more like an AS iosched bug :)

Quoted 2 lines

Quoted 2 lines
You do not enable TCQ on highpoint without using the hosted polling timer. Oh and I have not added it, and so hit Bartlomiej up for the additions.

For what? TCQ tests fine on a HPT370 here.

Signature · 1 line

Jens Axboe

Tomi Orava

2003-10-20 07:43 UTC #6 lore raw

Quoted 7 lines · Bartlomiej Zolnierkiewicz

Quoted 3 lines
Andre, thanks for helpful hint.
Svetoslav, the right person to whine about TCQ stuff is Jens Axboe 8-).
Well that's correct, but this looks more like an AS iosched bug :)

Pardon my ignorance, but what is this AS iosched ?

Quoted 6 lines

Quoted 4 lines
Quoted 1 line
You do not enable TCQ on highpoint without using the hosted polling
timer.
Quoted 1 line
Oh and I have not added it, and so hit Bartlomiej up for the
additions.
For what? TCQ tests fine on a HPT370 here.

I'm not entirely sure if that proves something or not, as the whole (interrupt & hang) problem seems to exist only with HPT374 and not with any other version of the chip anyway. I do personally have one mb with a 372-chip and it has always worked just fine with this same driver (perhaps its harder to detect the problem with only two disks?).

Regards,
Tomi Orava

Svetoslav Slavtchev

author

2003-10-20 08:11 UTC #7 lore raw

In reply to Jens Axboe

Quoted 5 lines · Bartlomiej Zolnierkiewicz

Quoted 3 lines
Andre, thanks for helpful hint.
Svetoslav, the right person to whine about TCQ stuff is Jens Axboe 8-).

Quoted 7 lines

Well that's correct, but this looks more like an AS iosched bug :)
Quoted 1 line
Quoted 1 line
You do not enable TCQ on highpoint without using the hosted polling
timer.
Quoted 1 line
Quoted 1 line
Oh and I have not added it, and so hit Bartlomiej up for the
additions.
For what? TCQ tests fine on a HPT370 here.

cmdline : acpi=off pci=noacpi elevator=deadline

hdparm -d1 -X69 -Q32 /dev/hd[a,e,g]
---------------------------------------------------------------------
blk: queue f7dd2e00, I/O limit 4095Mb (mask 0xffffffff)
hda: tagged command queueing enabled, command queue depth 32
blk: queue f7dd2400, I/O limit 4095Mb (mask 0xffffffff)
ide_dmaq_intr: stat=40, not expected
ide_dmaq_intr: stat=40, not expected
hde: tagged command queueing enabled, command queue depth 32
ide_dmaq_intr: stat=40, not expected
hdg: set_drive_speed_status: status=0x58 { DriveReady SeekComplete DataRequest }
blk: queue f7dcce00, I/O limit 4095Mb (mask 0xffffffff)
hdg: dma_timer_expiry: dma status == 0x20
hdg: DMA timeout retry
hdg: timeout waiting for DMA
hdg: status error: status=0x58 { DriveReady SeekComplete DataRequest }

hdg: drive not ready for command
hdg: tagged command queueing enabled, command queue depth 32
ide_dmaq_intr: hdg: error status ff
hdg: ide_dmaq_complete: status=0xff { Busy }
hdg: invalidating tag queue (1 commands)
hdg: ide_intr: huh? expected NULL handler on exit
hdg: drive_cmd: status=0x51 { DriveReady SeekComplete Error }
hdg: drive_cmd: error=0x04 { DriveStatusError }
hdg: set_drive_speed_status: status=0xff { Busy }
ide_tcq_intr_timeout: timeout waiting for completion interrupt
hdg: invalidating tag queue (0 commands)
hdg: status error: status=0x58 { DriveReady SeekComplete DataRequest }

hdg: drive not ready for command
hdg: status error: status=0x58 { DriveReady SeekComplete DataRequest }

hdg: drive not ready for command
hdg: CHECK for good STATUS
ide_dmaq_intr: hde: error status ff
hde: ide_dmaq_complete: status=0xff { Busy }
hde: invalidating tag queue (1 commands)
hde: ide_intr: huh? expected NULL handler on exit
hde: drive_cmd: status=0x51 { DriveReady SeekComplete Error }
hde: drive_cmd: error=0x04 { DriveStatusError
}
--------------------------------------------------------------------------------------

hdparm -m16 -d1 -X69 -Q32
/dev/hd[a,e,g]
--------------------------------------------------------------------------------------
blk: queue f7dd2e00, I/O limit 4095Mb (mask 0xffffffff)
blk: queue f7dd2400, I/O limit 4095Mb (mask 0xffffffff)
hde: tagged command queueing enabled, command queue depth 32
blk: queue f7dcce00, I/O limit 4095Mb (mask 0xffffffff)
hdg: tagged command queueing enabled, command queue depth 32
ide_dmaq_intr: hdg: error status ff
hdg: ide_dmaq_complete: status=0xff { Busy }
hdg: invalidating tag queue (1 commands)
hdg: ide_intr: huh? expected NULL handler on exit
hdg: drive_cmd: status=0x51 { DriveReady SeekComplete Error }
hdg: drive_cmd: error=0x04 { DriveStatusError
}
--------------------------------------------------------------------------------------

hdparm -m0 -d1 -X69 -Q32
/dev/hd[a,e,g]
--------------------------------------------------------------------------------------
blk: queue f7dd2e00, I/O limit 4095Mb (mask 0xffffffff)
blk: queue f7dd2400, I/O limit 4095Mb (mask 0xffffffff)
blk: queue f7dcce00, I/O limit 4095Mb (mask 0xffffffff)
hdg: tagged command queueing enabled, command queue depth 32
ide_dmaq_intr: hdg: error status ff
hdg: ide_dmaq_complete: status=0xff { Busy }
hdg: invalidating tag queue (1 commands)
hdg: ide_intr: huh? expected NULL handler on exit
hdg: drive_cmd: status=0x51 { DriveReady SeekComplete Error }
hdg: drive_cmd: error=0x04 { DriveStatusError
}
-------------------------------------------------------------------------------------

upto now no transfer to the drives, except to the root partition(system activity) which is on hdg

dd if=/dev/zero of=/mnt/tmp/1/10Gb.zeros bs=512k count=20000 [/dev/hde]
------------------------------------------------------------------------------
ide_dmaq_intr: hde: error status ff
hde: ide_dmaq_complete: status=0xff { Busy }
hde: invalidating tag queue (1 commands)
hde: ide_intr: huh? expected NULL handler on exit
hde: drive_cmd: status=0x51 { DriveReady SeekComplete Error }
hde: drive_cmd: error=0x04 { DriveStatusError }

and a bit later

ide_dmaq_intr: hda: error status ff
hda: ide_dmaq_complete: status=0xff { Busy }
hda: invalidating tag queue (1 commands)
hda: ide_intr: huh? expected NULL handler on exit
hda: drive_cmd: status=0x51 { DriveReady SeekComplete Error }
hda: drive_cmd: error=0x04 { DriveStatusError }

only /boot ~ 90Mb was mounted on hda and i don't think the system accessed it

on second thoght i also have a swap partition on it, but i have currently 1.9Mb swap used

svetljo

Signature · 6 lines

NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

Nick Piggin

2003-10-20 08:22 UTC #8 lore raw

Quoted 11 lines · Svetoslav Slavtchev

Quoted 11 lines · Bartlomiej Zolnierkiewicz
Quoted 4 lines
Andre, thanks for helpful hint.
Svetoslav, the right person to whine about TCQ stuff is Jens Axboe 8-).
Well that's correct, but this looks more like an AS iosched bug :)
Quoted 2 lines
Quoted 2 lines
You do not enable TCQ on highpoint without using the hosted polling
timer.

Quoted 9 lines

Quoted 6 lines
Quoted 2 lines
Quoted 2 lines
Oh and I have not added it, and so hit Bartlomiej up for the
additions.
For what? TCQ tests fine on a HPT370 here.
cmdline : acpi=off pci=noacpi elevator=deadline

Thanks, that would be good if you would test IDE TCQ problems with elevator=deadline. If it is working fine there, but you still get problems when using the default (AS) elevator then report them to me please.

Svetoslav Slavtchev

author

2003-10-20 08:30 UTC #9 lore raw

Quoted 22 lines

Quoted 18 lines · Svetoslav Slavtchev
Quoted 17 lines · Bartlomiej Zolnierkiewicz
Quoted 4 lines
Andre, thanks for helpful hint.
Svetoslav, the right person to whine about TCQ stuff is Jens Axboe 8-).
Well that's correct, but this looks more like an AS iosched bug :)
Quoted 2 lines
Quoted 2 lines
You do not enable TCQ on highpoint without using the hosted polling
timer.
Quoted 2 lines
Quoted 2 lines
Oh and I have not added it, and so hit Bartlomiej up for the
additions.
For what? TCQ tests fine on a HPT370 here.

Quoted 8 lines

Quoted 2 lines
cmdline : acpi=off pci=noacpi elevator=deadline
Thanks, that would be good if you would test IDE TCQ problems with elevator=deadline. If it is working fine there, but you still get problems when using the default (AS) elevator then report them to me please.

note taken :-)

svetljo

Signature · 6 lines

NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

Tomi Orava

2003-10-19 21:52 UTC #10 lore raw

In reply to Svetoslav Slavtchev

Quoted 4 lines

i have the same problems with epox 8k9a3+, and may be even strange ones
(like fs coruption when soft raid & / or lvm is used)

I've seen the filesystem corruption with ext3 & xfs and RAID1 (md) as well. However, I don't seem to be able to get that far nowadays, as the machine is being used as NFS-server and thefore there is always "too much" disk-transfers going on and an IDE-system hang happens quite soon after boot. (it seems that my raid-disks get out of sync every time I switch from proprietary driver --> kernel driver and so it might be the raid resync that hangs the system).

Quoted 9 lines

and i never had the problems with 8k5a3+, the drives were actually taken from the 8k5a3+ when it died (me silly tried to update to XP2700)
really strange, isn't it
both boards should be the same, except 8k5a3+ uses kt333
8k9a3+ uses kt400

Yep, but it cannot be strictly via-chipset issue as I have verified that the same problem exists with Epox 4PCA3+ motherboard, which is P4 & Intel 875P based.

Svetoslav Slavtchev

author

2003-10-19 22:00 UTC #11 lore raw

Quoted 5 lines

Quoted 4 lines
i have the same problems with epox 8k9a3+, and may be even strange ones
(like fs coruption when soft raid & / or lvm is used)

Quoted 9 lines

I've seen the filesystem corruption with ext3 & xfs and RAID1 (md) as well. However, I don't seem to be able to get that far nowadays, as the machine is being used as NFS-server and thefore there is always "too much" disk-transfers going on and an IDE-system hang happens quite soon after boot. (it seems that my raid-disks get out of sync every time I switch from proprietary driver --> kernel driver and so it might be the raid resync that hangs the system).

i tested xfs/reiserfs & may be ext2 & jfs both on plain(and with lvm over md) soft raid 0/1/5 never tried the closed source drivers

Quoted 6 lines

Quoted 6 lines
and i never had the problems with 8k5a3+, the drives were actually taken from the 8k5a3+ when it died (me silly tried to update to XP2700)
really strange, isn't it

Quoted 8 lines

Quoted 3 lines
both boards should be the same, except 8k5a3+ uses kt333
8k9a3+ uses kt400
Yep, but it cannot be strictly via-chipset issue as I have verified that the same problem exists with Epox 4PCA3+ motherboard, which is P4 & Intel 875P based.

may be a certain HPT BIOS version ?

i'll reboot to check what's here

svetljo

Signature · 6 lines

NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

Svetoslav Slavtchev

author

2003-10-19 22:11 UTC #12 lore raw

In reply to Tomi Orava

Quoted 21 lines

Quoted 4 lines
i have the same problems with epox 8k9a3+, and may be even strange ones
(like fs coruption when soft raid & / or lvm is used)
I've seen the filesystem corruption with ext3 & xfs and RAID1 (md) as well. However, I don't seem to be able to get that far nowadays, as the machine is being used as NFS-server and thefore there is always "too much" disk-transfers going on and an IDE-system hang happens quite soon after boot. (it seems that my raid-disks get out of sync every time I switch from proprietary driver --> kernel driver and so it might be the raid resync that hangs the system).
Quoted 6 lines
and i never had the problems with 8k5a3+, the drives were actually taken from the 8k5a3+ when it died (me silly tried to update to XP2700)
really strange, isn't it

Quoted 8 lines

Quoted 3 lines
both boards should be the same, except 8k5a3+ uses kt333
8k9a3+ uses kt400
Yep, but it cannot be strictly via-chipset issue as I have verified that the same problem exists with Epox 4PCA3+ motherboard, which is P4 & Intel 875P based.

may be epox got broken HPT's ?
anyone with mainboard from other vendor seeing this problems ?

my HPT BIOS is v1.24
mainboard BIOS last updated in september i think

svetljo

Signature · 6 lines

NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

Tomi Orava

2003-10-19 22:24 UTC #13 lore raw

Quoted 8 lines

Quoted 5 lines
Yep, but it cannot be strictly via-chipset issue as I have verified that the same problem exists with Epox 4PCA3+ motherboard, which is P4 & Intel 875P based.
may be epox got broken HPT's ?
anyone with mainboard from other vendor seeing this problems ?

This whole thread was opened by a person with HPT rocketraid 404 PCI-card ...

Quoted 3 lines

my HPT BIOS is v1.24
mainboard BIOS last updated in september i think

I think I have exactly the same HPT-bios version. The mb's bios is anyway dated 2003-03-07 so I think we have the same version. Unfortunately, I'm unable to check the bios version of the 875P-base mb for now (but I think it was the same 1.24 as well).

Are you capable of trying if the hdparm -m0 trick works for you ?

Regards,
Tomi Orava

Signature · 1 line

[email protected]

Svetoslav Slavtchev

author

2003-10-20 07:53 UTC #14 lore raw

Quoted 16 lines

Quoted 8 lines
Quoted 5 lines
Yep, but it cannot be strictly via-chipset issue as I have verified that the same problem exists with Epox 4PCA3+ motherboard, which is P4 & Intel 875P based.
may be epox got broken HPT's ?
anyone with mainboard from other vendor seeing this problems ?
This whole thread was opened by a person with HPT rocketraid 404 PCI-card ...
Quoted 3 lines
my HPT BIOS is v1.24
mainboard BIOS last updated in september i think

Quoted 8 lines

I think I have exactly the same HPT-bios version. The mb's bios is anyway dated 2003-03-07 so I think we have the same version. Unfortunately, I'm unable to check the bios version of the 875P-base mb for now (but I think it was the same 1.24 as well).
Are you capable of trying if the hdparm -m0 trick works for you ?

you mean the fs corruption on soft raid or the interupts problem ? i dumped the raid setup as i couldn't find a way to debug it and my drives are pretty full again, but i'll try to free some space

for the interupts
with test7-bk8 , acpi=off pci=noacpi &
hdparm -m0 -d1 -X69 /dev/hd[a,e,g]
hdparm -m16 -d1 -X69 /dev/hd[a,e,g]
i don't see timeouts
if i omit -m i do see them sometimes

(i see them both with -m and without -m if i try to use TCQ )

for testing copied several 6-10Gb directories &
dd if=/dev/zero of=/mnt/tmp/1/10Gb.zeros bs=512k count=21000 or
dd if=/dev/zero of=/mnt/tmp/1/10Gb.zeros bs=1024k count=10000

svetljo

Signature · 6 lines

NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

Tomi Orava

2003-10-20 08:03 UTC #15 lore raw

Quoted 8 lines

Quoted 4 lines
Are you capable of trying if the hdparm -m0 trick
works for you ?
you mean the fs corruption on soft raid or the interupts problem ? i dumped the raid setup as i couldn't find a way to debug it and my drives are pretty full again, but i'll try to free some space

I meant the interrupt problem, because if that problem exists there doesn't seem to be a reasonable way to figure out what is the reason for the filesystem corruption.

Quoted 6 lines

for the interupts
with test7-bk8 , acpi=off pci=noacpi &
hdparm -m0 -d1 -X69 /dev/hd[a,e,g]
hdparm -m16 -d1 -X69 /dev/hd[a,e,g]
i don't see timeouts
if i omit -m i do see them sometimes

Ok, so if I understood you correctly, the interrupt problem persists _only_ if you leave the multiple sector setting on its default setting ? If you explicitly disable it, or set it to maximum it works fine ? Does it work with any value ?

Quoted 1 line

(i see them both with -m and without -m if i try to use TCQ )

I'd really like to see the stable series working properly before jumping into the 2.6-series test kernels ... Although, I'm still wondering how come TCQ is known to work with 370-chip as Jens said, but not with 374-chip ...

Regards,
Tomi Orava

Signature · 1 line

[email protected]

Svetoslav Slavtchev

author

2003-10-20 09:03 UTC #16 lore raw

Quoted 20 lines

Quoted 8 lines
Quoted 4 lines
Are you capable of trying if the hdparm -m0 trick
works for you ?
you mean the fs corruption on soft raid or the interupts problem ? i dumped the raid setup as i couldn't find a way to debug it and my drives are pretty full again, but i'll try to free some space
I meant the interrupt problem, because if that problem exists there doesn't seem to be a reasonable way to figure out what is the reason for the filesystem corruption.
Quoted 6 lines
for the interupts
with test7-bk8 , acpi=off pci=noacpi &
hdparm -m0 -d1 -X69 /dev/hd[a,e,g]
hdparm -m16 -d1 -X69 /dev/hd[a,e,g]
i don't see timeouts
if i omit -m i do see them sometimes

Quoted 4 lines

Ok, so if I understood you correctly, the interrupt problem persists _only_ if you leave the multiple sector setting on its default setting ? If you explicitly disable it, or set it to maximum it works fine ? Does it work with any value ?

-m4 & -m8 seems to work OK
each time 3 runs of the dd test
1 to hda
2 to hde

TCQ not activated

Signature · 6 lines

NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

Carlos Fernandez Sanz

2003-10-20 18:01 UTC #17 lore raw

BTW, isn't m0 supposed to reduce perfomance?

[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out

real    0m33.763s
user    0m0.200s
sys     0m10.920s

[root@fulanito wd1200jb]# hdparm -m4 /dev/hdk

/dev/hdk:
setting multcount to 4
multcount = 4 (on)
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out

real    0m30.321s
user    0m0.150s
sys     0m10.630s

[root@fulanito wd1200jb]# hdparm -m0 /dev/hdk

/dev/hdk:
setting multcount to 0
multcount = 0 (off)
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out

real    0m30.749s
user    0m0.130s
sys     0m10.900s

[root@fulanito wd1200jb]#

----- Original Message -----
From: "Svetoslav Slavtchev" <[email protected]>
To: "Tomi Orava" <[email protected]>
Cc: <[email protected]>
Sent: Monday, October 20, 2003 11:03
Subject: Re: HighPoint 374

Quoted 47 lines

Quoted 20 lines
Quoted 8 lines
Quoted 4 lines
Are you capable of trying if the hdparm -m0 trick
works for you ?
you mean the fs corruption on soft raid or the interupts problem ? i dumped the raid setup as i couldn't find a way to debug it and my drives are pretty full again, but i'll try to free some space
I meant the interrupt problem, because if that problem exists there doesn't seem to be a reasonable way to figure out what is the reason for the filesystem corruption.
Quoted 6 lines
for the interupts
with test7-bk8 , acpi=off pci=noacpi &
hdparm -m0 -d1 -X69 /dev/hd[a,e,g]
hdparm -m16 -d1 -X69 /dev/hd[a,e,g]
i don't see timeouts
if i omit -m i do see them sometimes
Quoted 4 lines
Ok, so if I understood you correctly, the interrupt problem persists _only_ if you leave the multiple sector setting on its default setting ? If you explicitly disable it, or set it to maximum it works fine ? Does it work with any value ?
-m4 & -m8 seems to work OK
each time 3 runs of the dd test
1 to hda
2 to hde
TCQ not activated
Signature · 12 lines
NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

-
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/

Svetoslav Slavtchev

author

2003-10-20 19:15 UTC #18 lore raw

Quoted 1 line

BTW, isn't m0 supposed to reduce perfomance?

have to retest probably

how much RAM do you have,
isn't your test file to small

svetljo

Quoted 93 lines

[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m33.763s
user    0m0.200s
sys     0m10.920s
[root@fulanito wd1200jb]# hdparm -m4 /dev/hdk
/dev/hdk:
setting multcount to 4
multcount = 4 (on)
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m30.321s
user    0m0.150s
sys     0m10.630s
[root@fulanito wd1200jb]# hdparm -m0 /dev/hdk
/dev/hdk:
setting multcount to 0
multcount = 0 (off)
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m30.749s
user    0m0.130s
sys     0m10.900s
[root@fulanito wd1200jb]#
----- Original Message -----
From: "Svetoslav Slavtchev" <[email protected]>
To: "Tomi Orava" <[email protected]>
Cc: <[email protected]>
Sent: Monday, October 20, 2003 11:03
Subject: Re: HighPoint 374
Quoted 25 lines
Quoted 20 lines
Quoted 8 lines
Quoted 4 lines
Are you capable of trying if the hdparm -m0 trick
works for you ?
you mean the fs corruption on soft raid or the interupts problem ? i dumped the raid setup as i couldn't find a way to debug it and my drives are pretty full again, but i'll try to free some space
I meant the interrupt problem, because if that problem exists there doesn't seem to be a reasonable way to figure out what is the reason for the filesystem corruption.
Quoted 6 lines
for the interupts
with test7-bk8 , acpi=off pci=noacpi &
hdparm -m0 -d1 -X69 /dev/hd[a,e,g]
hdparm -m16 -d1 -X69 /dev/hd[a,e,g]
i don't see timeouts
if i omit -m i do see them sometimes
Quoted 4 lines
Ok, so if I understood you correctly, the interrupt problem persists _only_ if you leave the multiple sector setting on its default setting ? If you explicitly disable it, or set it to maximum it works fine ? Does it work with any value ?
Quoted 18 lines
-m4 & -m8 seems to work OK
each time 3 runs of the dd test
1 to hda
2 to hde
TCQ not activated
Signature · 9 lines
NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel"
in
Quoted 4 lines
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/

Signature · 6 lines

NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

Carlos Fernandez Sanz

2003-10-20 20:21 UTC #19 lore raw

512 Mb, file was 1.3 Gb.

----- Original Message -----
From: "Svetoslav Slavtchev" <[email protected]>
To: "Carlos Fernandez Sanz" <[email protected]>
Cc: <[email protected]>; <[email protected]>
Sent: Monday, October 20, 2003 21:14
Subject: Re: HighPoint 374

Quoted 62 lines

Quoted 1 line
BTW, isn't m0 supposed to reduce perfomance?
have to retest probably
how much RAM do you have,
isn't your test file to small
svetljo
Quoted 52 lines
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m33.763s
user    0m0.200s
sys     0m10.920s
[root@fulanito wd1200jb]# hdparm -m4 /dev/hdk
/dev/hdk:
setting multcount to 4
multcount = 4 (on)
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m30.321s
user    0m0.150s
sys     0m10.630s
[root@fulanito wd1200jb]# hdparm -m0 /dev/hdk
/dev/hdk:
setting multcount to 0
multcount = 0 (off)
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m30.749s
user    0m0.130s
sys     0m10.900s
[root@fulanito wd1200jb]#
----- Original Message -----
From: "Svetoslav Slavtchev" <[email protected]>
To: "Tomi Orava" <[email protected]>
Cc: <[email protected]>
Sent: Monday, October 20, 2003 11:03
Subject: Re: HighPoint 374
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 4 lines
Are you capable of trying if the hdparm -m0 trick
works for you ?
you mean the fs corruption on soft raid or the interupts problem ? i dumped the raid setup as i couldn't find a way to debug it and my drives are pretty full again, but i'll try to free some

space

Quoted 29 lines

Quoted 29 lines
Quoted 29 lines
Quoted 12 lines
I meant the interrupt problem, because if that problem exists there doesn't seem to be a reasonable way to figure out what is the reason for the filesystem corruption.
Quoted 6 lines
for the interupts
with test7-bk8 , acpi=off pci=noacpi &
hdparm -m0 -d1 -X69 /dev/hd[a,e,g]
hdparm -m16 -d1 -X69 /dev/hd[a,e,g]
i don't see timeouts
if i omit -m i do see them sometimes
Quoted 4 lines
Ok, so if I understood you correctly, the interrupt problem persists _only_ if you leave the multiple sector setting on its default setting ? If you explicitly disable it, or set it to maximum it works fine ? Does it work with any value ?
-m4 & -m8 seems to work OK
each time 3 runs of the dd test
1 to hda
2 to hde
TCQ not activated
Signature · 2 lines
NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Quoted 6 lines

Quoted 6 lines
Quoted 6 lines
Jetzt kostenlos anmelden unter http://www.gmx.net
+++ GMX - die erste Adresse für Mail, Message, More! +++
-
To unsubscribe from this list: send the line "unsubscribe

linux-kernel"

Quoted 21 lines

Quoted 6 lines

in
Quoted 4 lines
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/

Signature · 12 lines

NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

-
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/

Svetoslav Slavtchev

author

2003-10-20 21:07 UTC #20 lore raw

In reply to Carlos Fernandez Sanz

Quoted 1 line

BTW, isn't m0 supposed to reduce perfomance?

hm here it also doesn't seem to change much,
may be it's ignored if UDMA is used

mem is 1024Mb

[root@svetljo kernel]# hdparm -m0 /dev/hde

/dev/hde:
setting multcount to 0
multcount = 0 (off)

[root@svetljo 1]# time dd if=/dev/zero of=10Gb.zeros count=20000 bs=512k 20000+0 records in
20000+0 records out
0.09user 29.38system 5:44.49elapsed 8%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (149major+40minor)pagefaults 0swaps [root@svetljo 1]# hdparm -m8 /dev/hde

/dev/hde:
setting multcount to 8
multcount = 8 (on)
[root@svetljo 1]# time dd if=/dev/zero of=10Gb.zeros count=20000 bs=512k 20000+0 records in
20000+0 records out
0.10user 28.99system 5:50.44elapsed 8%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (150major+41minor)pagefaults 0swaps

root@svetljo 1]# hdparm -m16 /dev/hde

/dev/hde:
setting multcount to 16
multcount = 16 (on)
[root@svetljo 1]# time dd if=/dev/zero of=10Gb.zeros count=20000 bs=512k 20000+0 records in
20000+0 records out
0.09user 28.83system 5:53.57elapsed 8%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (165major+40minor)pagefaults 0swaps [root@svetljo 1]#

Quoted 35 lines

[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m33.763s
user    0m0.200s
sys     0m10.920s
[root@fulanito wd1200jb]# hdparm -m4 /dev/hdk
/dev/hdk:
setting multcount to 4
multcount = 4 (on)
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m30.321s
user    0m0.150s
sys     0m10.630s
[root@fulanito wd1200jb]# hdparm -m0 /dev/hdk
/dev/hdk:
setting multcount to 0
multcount = 0 (off)
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m30.749s
user    0m0.130s
sys     0m10.900s
[root@fulanito wd1200jb]#

Signature · 6 lines

NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

Bartlomiej Zolnierkiewicz

2003-10-20 21:20 UTC #21 lore raw

multicount affects PIO transfers performance only...

Quoted 76 lines · Svetoslav Slavtchev

Quoted 1 line
BTW, isn't m0 supposed to reduce perfomance?
hm here it also doesn't seem to change much,
may be it's ignored if UDMA is used
mem is 1024Mb
[root@svetljo kernel]# hdparm -m0 /dev/hde
/dev/hde:
setting multcount to 0
multcount = 0 (off)
[root@svetljo 1]# time dd if=/dev/zero of=10Gb.zeros count=20000 bs=512k 20000+0 records in
20000+0 records out
0.09user 29.38system 5:44.49elapsed 8%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (149major+40minor)pagefaults 0swaps [root@svetljo 1]# hdparm -m8 /dev/hde
/dev/hde:
setting multcount to 8
multcount = 8 (on)
[root@svetljo 1]# time dd if=/dev/zero of=10Gb.zeros count=20000 bs=512k 20000+0 records in
20000+0 records out
0.10user 28.99system 5:50.44elapsed 8%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (150major+41minor)pagefaults 0swaps
root@svetljo 1]# hdparm -m16 /dev/hde
/dev/hde:
setting multcount to 16
multcount = 16 (on)
[root@svetljo 1]# time dd if=/dev/zero of=10Gb.zeros count=20000 bs=512k 20000+0 records in
20000+0 records out
0.09user 28.83system 5:53.57elapsed 8%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (165major+40minor)pagefaults 0swaps [root@svetljo 1]#
Quoted 35 lines
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m33.763s
user    0m0.200s
sys     0m10.920s
[root@fulanito wd1200jb]# hdparm -m4 /dev/hdk
/dev/hdk:
setting multcount to 4
multcount = 4 (on)
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m30.321s
user    0m0.150s
sys     0m10.630s
[root@fulanito wd1200jb]# hdparm -m0 /dev/hdk
/dev/hdk:
setting multcount to 0
multcount = 0 (off)
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m30.749s
user    0m0.130s
sys     0m10.900s
[root@fulanito wd1200jb]#

Svetoslav Slavtchev

author

2003-10-20 21:25 UTC #22 lore raw

Quoted 2 lines

multicount affects PIO transfers performance only...

how could it then " resolve" DMA timeouts ?

Quoted 81 lines · Svetoslav Slavtchev

Quoted 9 lines
Quoted 1 line
BTW, isn't m0 supposed to reduce perfomance?
hm here it also doesn't seem to change much,
may be it's ignored if UDMA is used
mem is 1024Mb
[root@svetljo kernel]# hdparm -m0 /dev/hde
Quoted 8 lines
/dev/hde:
setting multcount to 0
multcount = 0 (off)
[root@svetljo 1]# time dd if=/dev/zero of=10Gb.zeros count=20000 bs=512k 20000+0 records in
20000+0 records out
0.09user 29.38system 5:44.49elapsed 8%CPU (0avgtext+0avgdata
0maxresident)k
Quoted 10 lines
0inputs+0outputs (149major+40minor)pagefaults 0swaps [root@svetljo 1]# hdparm -m8 /dev/hde
/dev/hde:
setting multcount to 8
multcount = 8 (on)
[root@svetljo 1]# time dd if=/dev/zero of=10Gb.zeros count=20000 bs=512k 20000+0 records in
20000+0 records out
0.10user 28.99system 5:50.44elapsed 8%CPU (0avgtext+0avgdata
0maxresident)k
Quoted 11 lines
0inputs+0outputs (150major+41minor)pagefaults 0swaps
root@svetljo 1]# hdparm -m16 /dev/hde
/dev/hde:
setting multcount to 16
multcount = 16 (on)
[root@svetljo 1]# time dd if=/dev/zero of=10Gb.zeros count=20000 bs=512k 20000+0 records in
20000+0 records out
0.09user 28.83system 5:53.57elapsed 8%CPU (0avgtext+0avgdata
0maxresident)k
Quoted 38 lines
0inputs+0outputs (165major+40minor)pagefaults 0swaps [root@svetljo 1]#
Quoted 35 lines
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m33.763s
user    0m0.200s
sys     0m10.920s
[root@fulanito wd1200jb]# hdparm -m4 /dev/hdk
/dev/hdk:
setting multcount to 4
multcount = 4 (on)
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m30.321s
user    0m0.150s
sys     0m10.630s
[root@fulanito wd1200jb]# hdparm -m0 /dev/hdk
/dev/hdk:
setting multcount to 0
multcount = 0 (off)
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m30.749s
user    0m0.130s
sys     0m10.900s
[root@fulanito wd1200jb]#

Signature · 6 lines

NEU FÜR ALLE - GMX MediaCenter - für Fotos, Musik, Dateien...
Fotoalbum, File Sharing, MMS, Multimedia-Gruß, GMX FotoService

Jetzt kostenlos anmelden unter http://www.gmx.net

+++ GMX - die erste Adresse für Mail, Message, More! +++

Tomi Orava

2003-10-21 18:43 UTC #23 lore raw

In reply to Bartlomiej Zolnierkiewicz

Bartlomiej,

Quoted 1 line

multicount affects PIO transfers performance only...

This makes sense as I was finally able to run some tests and hdparm -m doesn't seem to affect anything.

However, do the following errors tell you anything new ? (They were seen while copying almost 40GB of files from one RAID1-mirror (hpt374) to second RAID1-mirror(

Quoted 88 lines · Svetoslav Slavtchev

Quoted 33 lines
Quoted 1 line
BTW, isn't m0 supposed to reduce perfomance?
hm here it also doesn't seem to change much,
may be it's ignored if UDMA is used
mem is 1024Mb
[root@svetljo kernel]# hdparm -m0 /dev/hde
/dev/hde:
setting multcount to 0
multcount = 0 (off)
[root@svetljo 1]# time dd if=/dev/zero of=10Gb.zeros count=20000 bs=512k 20000+0 records in
20000+0 records out
0.09user 29.38system 5:44.49elapsed 8%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (149major+40minor)pagefaults 0swaps [root@svetljo 1]# hdparm -m8 /dev/hde
/dev/hde:
setting multcount to 8
multcount = 8 (on)
[root@svetljo 1]# time dd if=/dev/zero of=10Gb.zeros count=20000 bs=512k 20000+0 records in
20000+0 records out
0.10user 28.99system 5:50.44elapsed 8%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (150major+41minor)pagefaults 0swaps
root@svetljo 1]# hdparm -m16 /dev/hde
Quoted 46 lines
/dev/hde:
setting multcount to 16
multcount = 16 (on)
[root@svetljo 1]# time dd if=/dev/zero of=10Gb.zeros count=20000 bs=512k 20000+0 records in
20000+0 records out
0.09user 28.83system 5:53.57elapsed 8%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (165major+40minor)pagefaults 0swaps [root@svetljo 1]#
Quoted 35 lines
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m33.763s
user    0m0.200s
sys     0m10.920s
[root@fulanito wd1200jb]# hdparm -m4 /dev/hdk
/dev/hdk:
setting multcount to 4
multcount = 4 (on)
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m30.321s
user    0m0.150s
sys     0m10.630s
[root@fulanito wd1200jb]# hdparm -m0 /dev/hdk
/dev/hdk:
setting multcount to 0
multcount = 0 (off)
[root@fulanito wd1200jb]# time dd if=/dev/zero of=filling count=20000 bs=65536
20000+0 records in
20000+0 records out
real    0m30.749s
user    0m0.130s
sys     0m10.900s
[root@fulanito wd1200jb]#
-
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/

Signature · 1 line

[email protected]

Tomi Orava

2003-10-21 18:50 UTC #24 lore raw

In reply to Bartlomiej Zolnierkiewicz

Ups, hit enter while still writing the message, sorry about that.

Anyway, I just wanted to know if the following error messages tell you anything new.

The test was done by copying almost 40GB data simultaneously from one RAID1-mirror (disks 1&2 on hpt374) to disks 3&4 on hpt374 as well _and_ to disks 1&2 on an external Sil680 PCI-card. Kernel version was 2.4.23-pre7.

In my tests the hdparm -m doesn't solve anything. The hardware was Epox 8K9A3+ with integrated hpt374-chip. Processor is AMD 1.4TB, no overclocking of any kind.

Regards,
Tomi Orava
---------------------------------------------------------------

Oct 21 20:15:05 alderan kernel: hdk: drive_cmd: status=0x51 { DriveReady
SeekComplete Error }
Oct 21 20:15:05 alderan kernel: hdk: drive_cmd: error=0x04 {
DriveStatusError }
Oct 21 20:15:05 alderan kernel: hdk: drive_cmd: status=0x51 { DriveReady
SeekComplete Error }
Oct 21 20:15:05 alderan kernel: hdk: drive_cmd: error=0x04 {
DriveStatusError }
Oct 21 20:15:17 alderan kernel: hdk: drive_cmd: status=0x51 { DriveReady
SeekComplete Error }
Oct 21 20:15:17 alderan kernel: hdk: drive_cmd: error=0x04 {
DriveStatusError }
Oct 21 20:22:36 alderan kernel: hdg: set_drive_speed_status: status=0xff {
Busy }
Oct 21 20:30:57 alderan kernel: hdi: dma_timer_expiry: dma status == 0x20
Oct 21 20:30:57 alderan kernel: hdi: timeout waiting for DMA
Oct 21 20:30:57 alderan kernel: hdi: timeout waiting for DMA
Oct 21 20:30:57 alderan kernel: hdi: (__ide_dma_test_irq) called while not
waiting
Oct 21 20:30:57 alderan kernel: hdi: status error: status=0x58 {
DriveReady SeekComplete DataRequest }
Oct 21 20:30:57 alderan kernel:
Oct 21 20:30:57 alderan kernel: hdi: drive not ready for command
Oct 21 20:30:57 alderan kernel: hdi: status timeout: status=0xd0 { Busy }
Oct 21 20:30:57 alderan kernel:
Oct 21 20:30:57 alderan kernel: hdi: drive not ready for command
Oct 21 20:30:57 alderan kernel: ide4: reset: success
Oct 21 20:30:57 alderan kernel: blk: queue c0498a50, I/O limit 4095Mb
(mask 0xffffffff)
Oct 21 20:30:57 alderan kernel: hdi: status error: status=0x58 {
DriveReady SeekComplete DataRequest }
Oct 21 20:31:01 alderan kernel:
Oct 21 20:31:10 alderan kernel: hdi: drive not ready for command
Oct 21 20:31:17 alderan kernel: hdi: dma_timer_expiry: dma status == 0x20
Oct 21 20:31:20 alderan kernel: hdi: timeout waiting for DMA
Oct 21 20:31:20 alderan kernel: hdi: timeout waiting for DMA
Oct 21 20:31:20 alderan kernel: hdi: (__ide_dma_test_irq) called while not
waiting
Oct 21 20:31:20 alderan kernel: hdi: status error: status=0x58 {
DriveReady SeekComplete DataRequest }
Oct 21 20:31:20 alderan kernel:
Oct 21 20:31:20 alderan kernel: hdi: drive not ready for command
Oct 21 20:31:20 alderan kernel: hdi: status timeout: status=0xd0 { Busy }
Oct 21 20:31:20 alderan kernel:
Oct 21 20:31:20 alderan kernel: hdi: drive not ready for command
Oct 21 20:31:20 alderan kernel: ide4: reset: success
Oct 21 20:32:36 alderan kernel: hdi: lost interrupt

Signature · 1 line

[email protected]