| View previous topic :: View next topic |
| Author |
Message |
Thulle
n00b
Joined: 11 Nov 2002
Posts: 70
|
 Posted: Thu Nov 27, 2003 3:36 pm Post subject: two promisecards freezes computer |
 |
|
Got some heavy trouble setting up my raid, i got two Promisecards, one Ultra100 tx2, and one Ultra133 tx2, chipsets are 20268 and 20269.. i got 5 wd 250gig drives, each hooked up as master on each chan + the fifth on secondary channel on mobo. The problem is when i try to use all disks, computer freezes and even numlock aint responding. It works to setup a raid5 over hdc (mobo), hde & hdg OR hdc, hdi & hdk and it starts to synch, but if i include disks from both controllers computer locks as soon as it starts to sync (instantly after mkraid). If i do a raid0 over all five discs, it locks when i do mkfs. If i make a partition on hde & hdi and run bonnie++ or something on both at same time, computer freeze.
Motherboard is ABIT KR7A-raid (kt266a) - with buggy acpi-table, so acpi is disabled in kernel & the onboard hpt372-controller is disabled in bios. Running kernel 2.4.20-wolk4.9s, but its the same in 2.4.20-gentoo-r8 & 2.6.0-test10.
The u133-card has been in use for over a year, has worked perfectly, the new u100-card seems to work fine alone too. Also tried different pci-slots to check if they we're buggy, no difference. BIOS on cards are the latest ones (.15) - flashed both yesterday when stuff refused to work, no difference. I have searched the forums and googled alot, founds nothing more than some statement that multiple promise-cards ain't working too good, but i have friends saying the opposite too. Ideas anyone?
edit; im gonna try on an asus kt333 now.. will post result later
edit2; resulted in same thing, just after creation of raid5 it locks (when sync starts) - i managed to cat /proc/mdstat before it froze, and it looked normally, sync had just started and was at block no 40960 |
|
| Back to top |
|
 |
didi666
n00b
Joined: 09 Sep 2003
Posts: 32
Location: Stuttgart, Germany
|
| Posted: Thu Nov 27, 2003 8:23 pm Post subject: |
|
|
Hm, I would think it is some Mb chipset/IRQ issue when it is occurring only while using cards in parallel.
For me, 3x Promise 20268 have been working stable since months, although the Promise bios only shows 2 cards at startup 
Post your output of "lspci" and "cat /proc/interrupts" to see if there is multiple cards on one irq line. My setup looks like this:
| Code: |
# lspci
...
00:09.0 Unknown mass storage controller: Promise Technology, Inc. 20268 (rev 02)
00:0a.0 Unknown mass storage controller: Promise Technology, Inc. 20268 (rev 02)
00:0b.0 Unknown mass storage controller: Promise Technology, Inc. 20268 (rev 02)
...
|
| Code: |
# cat /proc/interrupts
CPU0
0: 1646071343 XT-PIC timer
1: 2 XT-PIC keyboard
2: 0 XT-PIC cascade
3: 5600215 XT-PIC serial
5: 984544719 XT-PIC eth0
8: 2 XT-PIC rtc
9: 0 XT-PIC acpi
10: 51491361 XT-PIC ide2, ide3
11: 173699619 XT-PIC sym53c8xx, eth1
12: 51932547 XT-PIC ide4, ide5
14: 12497553 XT-PIC ide0
15: 53038855 XT-PIC ide6, ide7, usb-uhci
NMI: 0
LOC: 1646120019
ERR: 2017933
MIS: 0
|
|
|
| Back to top |
|
|
Thulle
n00b
Joined: 11 Nov 2002
Posts: 70
|
| Posted: Fri Nov 28, 2003 2:05 am Post subject: |
|
|
yeah, thats what i thought too, but then i got corrected on irc - pci dont have irqconflicts, comeback when you use isa.. a.s.o.
lspci;
| Code: |
00:08.0 Unknown mass storage controller: Promise Technology, Inc. 20269 (rev 02)
00:0f.0 Unknown mass storage controller: Promise Technology, Inc. 20268 (rev 02)
|
interrupts only listed ide0,2&3 - so i rebooted & plugged a drive onto the secondary channel on other card (ide5 that becomes), now it lists ide0,2,3&5 - but not ide1? rebooting.. checked bios, found pci bus master, enabled it..
interrupts;
| Code: |
10: ide5
12: ide2,ide3
14: ide0
15: ide1
|
im pretty sure ide2,3,4,5 were all on irq10 before.. gonna put some load on the disks to see if it holds.. current setup; one disk as hdc on ide1 (ofcorz), 4disks on pdc20269, one 17gig seagate on pdc20268 as hdk.. will be back with result
edit; it seems to work.. interrupts-counter increases on all controllers in use, and so is error-counter, but from the post above - that seems perfectyl normal. I'm gonna get some sleep now, first thing tomorrow is to revert to old setup w/ one disk on each chan.. did pci bus master do it? The bios of both cards actually say PCI busmaster enabled, maybe its something required. Or did the cards automagically decided to work? Or is it just a random moment of happiness, maybe the old seagate ain't putting enough strain on the system... no-one knows, but we'll all se tomorrow... when i...wake...up. *credits start rolling here* *slight music fade-in with computer-fans & disks fading out*
okey, i need my sleep now. badly.
edit2; just checked the other mobo i tried on (kt333) - i found no busmaster thingie in bios, but i dont remeber if i saw it while promise-bios booted, have to test that tomorrow too.. found another nice feature on it though - the ability to assign an irq to a certain pci-slot.. would make it kinda hard for conflicts, wouldnt it? |
|
| Back to top |
|
|
didi666
n00b
Joined: 09 Sep 2003
Posts: 32
Location: Stuttgart, Germany
|
| Posted: Fri Nov 28, 2003 10:09 am Post subject: |
|
|
| Quote: | | yeah, thats what i thought too, but then i got corrected on irc - pci dont have irqconflicts, comeback when you use isa.. a.s.o. |
Yes, yes, in theory I am more inclined to believe what I experience. And since X86 interrupt controllers never improved (apart from APICs for SMP) you will always get some sort of problem when many devices share one IRQ, be it shabby hardware or device driver Just look on google how many people disable Win2000-ACPI because it used to place all devices on one IRQ.
| Quote: | | did pci bus master do it? The bios of both cards actually say PCI busmaster enabled, maybe its something required. |
Definitely. All modern SCSI or IDE controllers I know need PCI Busmastering. Because without it, DMA modes would not be effective.
I think you solved your problem  |
|
| Back to top |
|
|
Thulle
n00b
Joined: 11 Nov 2002
Posts: 70
|
| Posted: Fri Nov 28, 2003 4:48 pm Post subject: |
|
|
| didi666 wrote: | | Quote: | | yeah, thats what i thought too, but then i got corrected on irc - pci dont have irqconflicts, comeback when you use isa.. a.s.o. |
Yes, yes, in theory I am more inclined to believe what I experience. And since X86 interrupt controllers never improved (apart from APICs for SMP) you will always get some sort of problem when many devices share one IRQ, be it shabby hardware or device driver Just look on google how many people disable Win2000-ACPI because it used to place all devices on one IRQ.
hehe, yeah, thats well known.. w2k stacked every unit at irq10
| Quote: | | did pci bus master do it? The bios of both cards actually say PCI busmaster enabled, maybe its something required. |
Definitely. All modern SCSI or IDE controllers I know need PCI Busmastering. Because without it, DMA modes would not be effective.
I think you solved your problem |
Well, it remains to be seen.. spread the disks over the cards again.. rejumpered to all-masters, now both cards said, no disks found - not installing bios - and then the kernel wasnt able to access disks, just got some crc-errors, then it said that disk was busy..
booting, disabled usb in bios too - less stuff demanding an irq.
nah, bios not getting installed.
edit; just checked jumperingmanual on westerndigital.com - there seems to be different settings for master w/ two drives & single master.. trying
damn, how did i forget i how i had drives jumpered? it found all drives now, except one, gonna check power/ide-cable.. if it finds disk now, raid5 should start syncing instantly during boot.. if it doesnt work, the computer will freeze. |
|
| Back to top |
|
|
Thulle
n00b
Joined: 11 Nov 2002
Posts: 70
|
| Posted: Fri Nov 28, 2003 5:21 pm Post subject: |
|
|
YAY, its building!
lets just see if it can skip freezing too
i enabled usb too, current interruptusage is;
| Code: |
5: USB
10: ide2,ide3
12: ide4,ide5
14: ide0
15: ide1
|
edit; though i see an insanely high errorratio on interrupts; ERR: is as high as irq10 & irq12 - but that might be that both irq10 & 12 are fully loaded and thus both getting errors on half of their requests.
atm plan is;
1) see if it survives a sync.
2) see if it survives the addition of an networkcard |
|
| Back to top |
|
|
Thulle
n00b
Joined: 11 Nov 2002
Posts: 70
|
| Posted: Sat Nov 29, 2003 3:39 pm Post subject: |
|
|
| now it syncs totally fine, but i cant mkfs, then it hangs.. |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
Kernel & Hardware
