IDE DMA Problems...system hangs
Hey guys,
I am attaching my previous email for additional info. Now I am using kernel 2.4.1-ac12 and these problems have not gone away.
Anybody else having these problems with a ide raid 5?
The Raid 5 performance should also be questioned..here are some number returned by hdparam
/dev/hda - IBM DLTA 20GB (ext2)/dev/md0 - 8 IBM DLTA 45GB (Reiserfs)
[root@bertha hdparm-3.9]# ./hdparm -t /dev/hda /dev/hda:
Timing buffered disk reads: 64 MB in 2.36 seconds = 27.12 MB/sec
[root@bertha hdparm-3.9]# ./hdparm -t /dev/md0 /dev/md0:
Timing buffered disk reads: 64 MB in 22.16 seconds = 2.89 MB/sec
Is this to be expected? This much performance loss? Anybody else using IDE raid, I would really appreciate your input on this setup.
Here is the log from syslog
Feb 13 05:23:27 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Feb 13 05:23:27 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
Feb 13 05:23:28 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Feb 13 05:23:28 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
Feb 13 05:23:28 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Feb 13 05:23:28 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
Feb 13 05:23:28 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Feb 13 05:23:28 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
Feb 13 05:23:33 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Feb 13 05:23:33 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
...
Feb 13 08:47:48 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Feb 13 08:47:48 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
...
Feb 13 09:54:07 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Feb 13 09:54:07 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
...
Feb 13 12:10:43 bertha kernel: hds: dma_intr: bad DMA status
Feb 13 12:10:43 bertha kernel: hds: dma_intr: status=0x50 { DriveReady SeekComplete }
Feb 13 12:12:22 bertha kernel: hdg: timeout waiting for DMA
Feb 13 12:12:22 bertha kernel: ide_dmaproc: chipset supported ide_dma_timeout func only: 14
Feb 13 12:12:22 bertha kernel: hdg: irq timeout: status=0x50 { DriveReady SeekComplete }
Feb 13 12:12:42 bertha kernel: hdg: timeout waiting for DMA
Feb 13 12:12:42 bertha kernel: ide_dmaproc: chipset supported ide_dma_timeout func only: 14
Feb 13 12:12:42 bertha kernel: hdg: irq timeout: status=0x50 { DriveReady SeekComplete }
Feb 13 12:13:02 bertha kernel: hdg: timeout waiting for DMA
Feb 13 12:13:02 bertha kernel: ide_dmaproc: chipset supported ide_dma_timeout func only: 14
Feb 13 12:13:02 bertha kernel: hdg: irq timeout: status=0x50 { DriveReady SeekComplete }
Feb 13 12:13:12 bertha kernel: hdg: timeout waiting for DMA
Feb 13 12:13:12 bertha kernel: ide_dmaproc: chipset supported ide_dma_timeout func only: 14
Feb 13 12:13:12 bertha kernel: hdg: irq timeout: status=0x50 { DriveReady SeekComplete }
Feb 13 12:13:12 bertha kernel: hdg: DMA disabled
Feb 13 12:13:12 bertha kernel: ide3: reset: success <------- * SYSTEM HUNG AT THIS POINT *
Feb 13 23:31:13 bertha syslogd 1.3-3: restart.--------------------------------------------------
To: [email protected]
Subject: DMA blues... (Raid5 + Promise ATA/100)
I have a software raid setup using the latest kernel, but the system keeps crashing.
Each drive is connected to the respective ide port via ATA100 capable cables. Each drive is a master..no slaves. The configuration is that /dev/hdc is a hot spare and /dev/hd[e,g,i,k,m,o,q,s] are all setup as raid 5. These are all 75GB drives and are recognized as such.
I have searched the linux-kernel archives and saw many posts addressing the problems that I was experiencing, namely the freezing caused by the code in the body of delay_50ms() in drivers/ide/ide.c. This was fixed in the current patch and as discussed earlier on the linux-kernel mailing list by using mdelay(50). This fixed the problems to some extent, the system seemed very reliable and I did not get a single "DriveStatusError BadCRC" or a "DriveReady SeekComplete Index Error" for a while. But after I had copied a large amount of data to the raid, about 17GB, the system crashed completely and could only be recovered by a cold reboot. Before using the latest patches, the system would usually crash after about 4-6GB of data had been moved. Here are the log entries...
Feb 12 06:41:12 bertha kernel: hdo: dma_intr: status=0x53 { DriveReady SeekComplete Index Error }
Feb 12 06:41:12 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
Feb 12 06:45:42 bertha kernel: hdo: timeout waiting for DMA
Feb 12 06:45:42 bertha kernel: hdo: ide_dma_timeout: Lets do it again!stat = 0x50, dma_stat = 0x20
Feb 12 06:45:42 bertha kernel: hdo: irq timeout: status=0x50 { DriveReady SeekComplete }
Feb 12 06:45:42 bertha kernel: hdo: ide_set_handler: handler not null; old=c01d0710, new=c01dac70
Feb 12 06:45:42 bertha kernel: bug: kernel timer added twice at c01d0585.
Feb 12 09:13:15 bertha syslogd 1.3-3: restart.Let me know If I should post any additional information that might help in troubleshooting. Is this a possible issue with the kernel code or is this a problem with hardware? Any help is appreciated...
- Jasmeet Sidhu
Some other info that might help:
[root@bertha /root]# uname -a
Linux bertha 2.4.1-ac9 #1 Mon Feb 12 02:43:08 PST 2001 i686 unknown
Patches Applied to Kernel 2.4.1:
1) ide.2.4.1-p8.all.01172001.patch
2) patch-2.4.1-ac9
Asus A7V VIA KT133 and Onboard Promise ATA100 Controller (PDC20267) 1GHz AMD Thunderbird Athalon Processor
Four Promise PCI ATA100 Controllers (PDC20267)
Netgear GA620 Gigabit Ethernet Card
Boot Drive (Root + Swap)
hda: IBM-DTLA-307020, ATA DISK drive
Raid 5 Drives:
hdc: IBM-DTLA-307075, ATA DISK drive *SPARE*
hde: IBM-DTLA-307075, ATA DISK drive
hdg: IBM-DTLA-307075, ATA DISK drive
hdi: IBM-DTLA-307075, ATA DISK drive
hdk: IBM-DTLA-307075, ATA DISK drive
hdm: IBM-DTLA-307075, ATA DISK drive
hdo: IBM-DTLA-307075, ATA DISK drive
hdq: IBM-DTLA-307075, ATA DISK drive
hds: IBM-DTLA-307075, ATA DISK drive
Quoted 3 lines
Anybody else having these problems with a ide raid 5?
The Raid 5 performance should also be questioned..here are some number returned by hdparam
You will get horribly bad performance off raid5 if you have stripes on both hda/hdb or hdc/hdd etc.
Quoted 4 lines
Feb 13 05:23:27 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error } Feb 13 05:23:27 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusErrorBadCRC }
You have inadequate cabling. CRC errors are indications of that. Make sure you are using sufficiently short cables for ATA33 and proper 80pin ATA66 cables.
Quoted 3 lines
Feb 13 12:12:42 bertha kernel: hdg: irq timeout: status=0x50 { DriveReady
SeekComplete }
Feb 13 12:13:02 bertha kernel: hdg: timeout waiting for DMAThis could be cabling too, cant be sure
Quoted 1 line
Feb 13 12:13:12 bertha kernel: hdg: DMA disabled
It gave up using DMA
Quoted 2 lines
Feb 13 12:13:12 bertha kernel: ide3: reset: success <------- * SYSTEM HUNGAT THIS POINT *
Ok thats a reasonable behaviour, except it shouldnt have then hung.
Quoted 6 lines · Alan Cox
Quoted 3 lines
Anybody else having these problems with a ide raid 5?
The Raid 5 performance should also be questioned..here are some number returned by hdparamYou will get horribly bad performance off raid5 if you have stripes on both hda/hdb or hdc/hdd etc.
If I am reading this correctly, then by striping on both hda/hdb and /hdc/hdd you mean that I have two drives per ide channel. In other words, you think I have a Master and a Slave type of a setup? This is incorrect. Each drive on the system is a master. I have 5 promise cards in the system (4 PCI and 1 onboard AUS a7v mobo). This gives me the ability to have 10 master drives. Since I am only striping on one drive per ide chanel, the penalty should not be much in terms of performance. Maybe its just that the hdparam utility is not a good tool for benchamarking a raid set?
Quoted 8 lines
Quoted 3 lines
Feb 13 05:23:27 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error } Feb 13 05:23:27 bertha kernel: hdo: dma_intr: error=0x84 {DriveStatusError
Quoted 1 line
BadCRC }
You have inadequate cabling. CRC errors are indications of that. Make sure you are using sufficiently short cables for ATA33 and proper 80pin ATA66 cables.
All the cables are ATA/100 capable. But I cannot think of another reason as to why I might be getting CRC errors. I will invest in better cables and see if it changes anything.
Quoted 9 lines
Quoted 3 lines
Feb 13 12:12:42 bertha kernel: hdg: irq timeout: status=0x50 { DriveReady SeekComplete } Feb 13 12:13:02 bertha kernel: hdg: timeout waiting for DMAThis could be cabling too, cant be sure
Quoted 1 line
Feb 13 12:13:12 bertha kernel: hdg: DMA disabledIt gave up using DMA
agreed.
Quoted 5 lines
Quoted 1 line
Feb 13 12:13:12 bertha kernel: ide3: reset: success <------- * SYSTEMHUNG
Quoted 1 line
AT THIS POINT *
Ok thats a reasonable behaviour, except it shouldnt have then hung.
This is also my main point of frustration. The system should be able to disable DMA if its giving it a lot of problems, but it should not hang. I have been experiencing this for quite a while with the newer kernels. Should I try the latest ac13 patch? I glanced of the changes and didnt seem like anything had changed regarding the ide subsystem.
Is there anyway I can force the kernel to output more messages...maybe that could help narrow down the problem?
J.Sidhu
Quoted 7 lines
Quoted 2 lines
You will get horribly bad performance off raid5 if you have stripes on both hda/hdb or hdc/hdd etc.
If I am reading this correctly, then by striping on both hda/hdb and /hdc/hdd you mean that I have two drives per ide channel. In other words, you think I have a Master and a Slave type of a setup? This is incorrect. Each drive on the system is a master. I have 5 promise cards
Ok then your performance should be fine (at least reasonably so, the lack of tagged queueing does hurt)
Quoted 3 lines
ide chanel, the penalty should not be much in terms of performance. Maybe its just that the hdparam utility is not a good tool for benchamarking a raid set?
Its not a good raid benchmark tool but its a good indication of general problems. Bonnie is a good tool for accurate assessment.
Quoted 4 lines
disable DMA if its giving it a lot of problems, but it should not hang. I have been experiencing this for quite a while with the newer kernels. Should I try the latest ac13 patch? I glanced of the changes and didnt seem like anything had changed regarding the ide subsystem.
I've not changed anything related to DMA handling specifically. The current -ac does have a fix for a couple of cases where an IDE reset on the promise could hang the box dead. That may be the problem.
Quoted 2 lines
Is there anyway I can force the kernel to output more messages...maybe that could help narrow down the problem?
Ask [email protected]. He may know the status of the promise support
Quoted 4 lines
Quoted 4 lines
I've not changed anything related to DMA handling specifically. The current -ac does have a fix for a couple of cases where an IDE reset on the promise could hang the box dead. That may be the problem.
I tried the new patches (2.4.1-ac13) and it seemed very stable. After moving about 50GB of data to the raid5, the system crashed. here is the syslog... (the system had been up for about 20 hours)
Feb 14 03:48:53 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Feb 14 03:48:53 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
<snip - about 40 lines exact same hdo: error>
Feb 14 19:35:52 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
Feb 14 19:35:52 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Feb 14 19:35:52 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
Feb 14 20:13:06 bertha kernel: hdi: dma_intr: bad DMA status
Feb 14 20:13:06 bertha kernel: hdi: dma_intr: status=0x50 { DriveReady SeekComplete }Feb 15 01:26:34 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Feb 15 01:26:34 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
Feb 15 01:26:34 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Feb 15 01:26:34 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
Feb 15 01:26:38 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Feb 15 01:26:38 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
Feb 15 01:45:06 bertha kernel: hdo: dma_intr: status=0x53 { DriveReady SeekComplete Index Error }
Feb 15 01:45:06 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
Feb 15 01:45:06 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Feb 15 01:45:06 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
Feb 15 01:45:06 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Feb 15 01:45:06 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC }
Feb 15 01:54:01 bertha kernel: hdg: timeout waiting for DMA<SYSTEM FROZEN>
Jasmeet
Quoted 34 lines · Alan Cox
Quoted 1 line
Quoted 1 line
You will get horribly bad performance off raid5 if you have stripes on
both
Quoted 6 lines
Quoted 1 line
hda/hdb or hdc/hdd etc.
If I am reading this correctly, then by striping on both hda/hdb and /hdc/hdd you mean that I have two drives per ide channel. In other words, you think I have a Master and a Slave type of a setup? This is incorrect. Each drive on the system is a master. I have 5 promise cards
Ok then your performance should be fine (at least reasonably so, the lack of tagged queueing does hurt)
Quoted 3 lines
ide chanel, the penalty should not be much in terms of performance. Maybe its just that the hdparam utility is not a good tool for benchamarking a raid set?
Its not a good raid benchmark tool but its a good indication of general problems. Bonnie is a good tool for accurate assessment.
Quoted 3 lines
disable DMA if its giving it a lot of problems, but it should not hang. I have been experiencing this for quite a while with the newer kernels. Should I try the latest ac13 patch? I glanced of the changes
and
Quoted 1 line
didnt seem like anything had changed regarding the ide subsystem.
I've not changed anything related to DMA handling specifically. The current -ac does have a fix for a couple of cases where an IDE reset on the promise could hang the box dead. That may be the problem.
Quoted 1 line
Is there anyway I can force the kernel to output more messages...maybe
that
Quoted 1 line
could help narrow down the problem?
Ask [email protected]. He may know the status of the promise support
You have junk for cables or they are noe sheilded correctly from crosstalk. But I do not think this is the case.
Go check your power-supply for stality and load.
Then do a ripple noise test to make sure that underload, it does not cause the clock on the drives to fail.
Quoted 89 lines · Jasmeet Sidhu
Quoted 4 lines
Quoted 4 lines
I've not changed anything related to DMA handling specifically. The current -ac does have a fix for a couple of cases where an IDE reset on the promise could hang the box dead. That may be the problem.
I tried the new patches (2.4.1-ac13) and it seemed very stable. After moving about 50GB of data to the raid5, the system crashed. here is the syslog... (the system had been up for about 20 hours)
Feb 14 03:48:53 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error } Feb 14 03:48:53 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC } <snip - about 40 lines exact same hdo: error> Feb 14 19:35:52 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC } Feb 14 19:35:52 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error } Feb 14 19:35:52 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC } Feb 14 20:13:06 bertha kernel: hdi: dma_intr: bad DMA status Feb 14 20:13:06 bertha kernel: hdi: dma_intr: status=0x50 { DriveReadySeekComplete }
Feb 15 01:26:34 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error } Feb 15 01:26:34 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC } Feb 15 01:26:34 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error } Feb 15 01:26:34 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC } Feb 15 01:26:38 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error } Feb 15 01:26:38 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC } Feb 15 01:45:06 bertha kernel: hdo: dma_intr: status=0x53 { DriveReady SeekComplete Index Error } Feb 15 01:45:06 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC } Feb 15 01:45:06 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error } Feb 15 01:45:06 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC } Feb 15 01:45:06 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error } Feb 15 01:45:06 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusError BadCRC } Feb 15 01:54:01 bertha kernel: hdg: timeout waiting for DMA<SYSTEM FROZEN>
Jasmeet
Quoted 34 lines · Alan Cox
Quoted 1 line
Quoted 1 line
You will get horribly bad performance off raid5 if you have stripes on
both
Quoted 6 lines
Quoted 1 line
hda/hdb or hdc/hdd etc.
If I am reading this correctly, then by striping on both hda/hdb and /hdc/hdd you mean that I have two drives per ide channel. In other words, you think I have a Master and a Slave type of a setup? This is incorrect. Each drive on the system is a master. I have 5 promise cards
Ok then your performance should be fine (at least reasonably so, the lack of tagged queueing does hurt)
Quoted 3 lines
ide chanel, the penalty should not be much in terms of performance. Maybe its just that the hdparam utility is not a good tool for benchamarking a raid set?
Its not a good raid benchmark tool but its a good indication of general problems.
Bonnie is a good tool for accurate assessment.Quoted 3 lines
disable DMA if its giving it a lot of problems, but it should not hang. I have been experiencing this for quite a while with the newer kernels. Should I try the latest ac13 patch? I glanced of the changes
and
Quoted 1 line
didnt seem like anything had changed regarding the ide subsystem.
I've not changed anything related to DMA handling specifically. The current -ac does have a fix for a couple of cases where an IDE reset on the promise could hang the box dead. That may be the problem.
Quoted 1 line
Is there anyway I can force the kernel to output more messages...maybe
that
Quoted 1 line
could help narrow down the problem?
Quoted 8 lines
Quoted 1 line
Ask [email protected]. He may know the status of the promise support
-
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/
Andre Hedrick
Linux ATA Development
ASL Kernel Development
-----------------------------------------------------------------------------
ASL, Inc. Toll free: 1-877-ASL-3535 1757 Houret Court Fax: 1-408-941-2071 Milpitas, CA 95035 Web: www.aslab.com
Quoted 3 lines
I tried the new patches (2.4.1-ac13) and it seemed very stable. After moving about 50GB of data to the raid5, the system crashed. here is the syslog... (the system had been up for about 20 hours)
Ok so better but not perfect
Quoted 2 lines
Feb 15 01:54:01 bertha kernel: hdg: timeout waiting for DMA<SYSTEM FROZEN>
hdg is a promise card ?
Alan
Quoted 7 lines · Alan Cox
Quoted 4 lines
Feb 13 05:23:27 bertha kernel: hdo: dma_intr: status=0x51 { DriveReady SeekComplete Error } Feb 13 05:23:27 bertha kernel: hdo: dma_intr: error=0x84 { DriveStatusErrorBadCRC }
You have inadequate cabling. CRC errors are indications of that. Make sure you are using sufficiently short cables for ATA33 and proper 80pin ATA66 cables.
I've had cases (on VIA chipsets) where, even or ATA33, a 40-pin cable caused CRC errors for ATA33 and an 80-pin cable fixed things. (The same 40-pin cable does ATA33 without problems on an AMD 750 or an Intel BX, though.)
IIRC, Andre Hedrick has said in the past that a marginal PSU or motherboard can also cause CRC errors.
-Barry K. Nathan <[email protected]>
Quoted 18 lines · Jasmeet Sidhu
Hey guys,
I am attaching my previous email for additional info. Now I am using kernel 2.4.1-ac12 and these problems have not gone away.
Anybody else having these problems with a ide raid 5?
The Raid 5 performance should also be questioned..here are some number returned by hdparam
/dev/hda- IBM DLTA 20GB (ext2)/dev/md0- 8 IBM DLTA 45GB (Reiserfs)[root@bertha hdparm-3.9]# ./hdparm -t
/dev/hda/dev/hda:
Timing buffered disk reads: 64 MB in 2.36 seconds = 27.12 MB/sec
Quoted 6 lines
[root@bertha hdparm-3.9]# ./hdparm -t
/dev/md0/dev/md0:
Timing buffered disk reads: 64 MB in 22.16 seconds = 2.89 MB/secIs this to be expected? This much performance loss? Anybody else using IDE raid, I would really appreciate your input on this setup.
md2 = RAID0 ext2
hda = hdb = IBM DTTA-351010 (10GB, 5400RPM, UDMA33)
# hdparm -tT /dev/hda /dev/md2 /dev/hda: Timing buffered disk reads: 64 MB in 5.27 seconds = 12.14 MB/sec Timing buffer-cache reads: 128 MB in 0.82 seconds =156.10 MB/sec
/dev/md2:
Timing buffered disk reads: 64 MB in 3.34 seconds = 19.16 MB/sec Timing buffer-cache reads: 128 MB in 0.80 seconds =160.00 MB/sec
On AMD K7 w/ 7409 (Viper) chipset, DMA66 mode w/ 80-pin cable. kernel = 2.4.1-ac8, no errors in kernel log.
So I get a 58% increase. You should almost max out the bus.
You probably have a bad cable. Try hdparam on each disk and see if any of them have errors/ cause the lockup.
-Thomas