dmesg shows /dev/hdg error

v00d00101 · 01-24-2007, 12:33 PM

Hi,

I have come across a problem on FC5 x86_64 with a ghost drive that doesnt physically exist, but is showing up in /dev. I've worked out that this is most likely the cause of some random freezes (hard machine lock) that have occured. The symptoms of which are everything slowly grinds to a halt, then it locks up.

Problems i have found in dmesg and messages log, include some selinux audit problems, which i can ignore for the moment (things that dont audit well Xorg, firefox, etc). But also this /dev/hdg which has bad sectors.

Code:

end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63
end_request: I/O error, dev hdg, sector 63

Research ive done on google point this towards being a RAID problem. But since i dont actually have a raid array setup (neither software or hardware (which was stopped in BIOS)), i cant work out how to cure it.

My hardware for those who might request it is below.

Code:

AMD 64 X2 3800 on an Asrock ALIVESATA2-GLan 
(stock speeds at 27C idle / 36C under load)
1GB OCZ DDR2 Ram (2x 512MB)
80GB WD SATA2 drive at /dev/sda (Linux Drive)
80GB WD IDE drive at /dev/hdc (Primary Boot and Windows XP Drive)
40GB WD IDE drive at /dev/hdb (Shared Data Drive)
NEC DVD Writer at /dev/hda
GF 7300GS (using current kmod-nvidia)

My question is.

How can i remove this hdg entry from the system, so fedora no longer searches for it?

Matir · 01-25-2007, 10:52 AM

When the machine first boots, type "dmesg | grep hdg" and see what that drive is being detected as. It will give you and us a better idea of how to handle this. (Removing a kernel module, for example)

v00d00101 · 01-25-2007, 12:15 PM

Hi Matir, i grabbed this out of dmesg.

Code:

ide3: BM-DMA at 0xe408-0xe40f, BIOS settings: hdg:DMA, hdh:pio
Probing IDE interface ide2...
Probing IDE interface ide3...
hdg: WDC WD800JD-00MSA1, ATA DISK drive
ide3 at 0xe800-0xe807,0xe482 on irq 66

More stuff.

Code:

hdg: max request size: 512KiB
hdg: 156301488 sectors (80026 MB) w/8192KiB Cache, CHS=16383/255/63
hdg: cache flushes supported
 hdg: hdg1 hdg2 hdg3 < hdg5 hdg6 hdg7 hdg8 > hdg4

and

hdg: status error: status=0x7f { DriveReady DeviceFault SeekComplete DataRequest CorrectedError Index Error }
hdg: status error: error=0xff { DriveStatusError BadCRC UncorrectableError SectorIdNotFound TrackZeroNotFound AddrMarkNotFound }, LBAsect=281474976710655, high=16777215, low=16777215, sector=153613286
ide: failed opcode was: unknown
hdg: DMA disabled
hdg: drive not ready for command
ide3: reset: master: error (0xff?); slave: failed
hdg: status error: status=0x7f { DriveReady DeviceFault SeekComplete DataRequest CorrectedError Index Error }
hdg: status error: error=0xff { DriveStatusError BadCRC UncorrectableError SectorIdNotFound TrackZeroNotFound AddrMarkNotFound }, LBAsect=281474976710655, high=16777215, low=16777215, sector=153613286
ide: failed opcode was: unknown
hdg: drive not ready for command
ide3: reset: master: error (0xff?); slave: failed
end_request: I/O error, dev hdg, sector 153613286
Buffer I/O error on device hdg8, logical block 13313792
end_request: I/O error, dev hdg, sector 153613290
Buffer I/O error on device hdg8, logical block 13313793
end_request: I/O error, dev hdg, sector 153613286
Buffer I/O error on device hdg8, logical block 13313792
end_request: I/O error, dev hdg, sector 153613290
Buffer I/O error on device hdg8, logical block 13313793
end_request: I/O error, dev hdg, sector 100358118
Buffer I/O error on device hdg8, logical block 0
end_request: I/O error, dev hdg, sector 100358122
Buffer I/O error on device hdg8, logical block 1
end_request: I/O error, dev hdg, sector 100358118
Buffer I/O error on device hdg8, logical block 0
end_request: I/O error, dev hdg, sector 100358122
Buffer I/O error on device hdg8, logical block 1
end_request: I/O error, dev hdg, sector 100358118

I've made a mistake, this is the SATA Drive that linux is installed on. Its running in SATA-2 mode at the minute. My board supports 2x SATA-1 drives in RAID 0,1, and 2x SATA-2 drives in RAID 0,1. Both IDE 2 and 3 are the SATA controllers, but RAID is currently disabled in the bios for both of them, by me, as i only have 1 SATA drive. The other 3 devices are on ide 1 and 2.

This stil leaves the reason /dev/sda is showing up as /dev/hdg, which appears to be erroring a lot, but /dev/sda seems to be running happily.

A quick(ish) fsck confirms nothing appears to be wrong with /dev/sda.

So the question is what to do about it?

Maybe the next kernel update will fix it.

Matir · 01-25-2007, 12:18 PM

Are you saying you believe both /dev/sda and /dev/hdg point to the same drive? I'd think this to be unlikely, unless multiple SATA chipset drivers are loaded (which the kernel should've prevented).

v00d00101 · 01-25-2007, 12:31 PM

I discovered this post on google, about SATA drives spawning to fast at boot up and causing problems. His board has the Via K8T800 chipset, mine is the K8T890, but he mentions /dev/hdg on Mandrake 10 in relation with RAID.

Since Mandarke and Fedora are similar in some ways, it was maybe going to be more pertinent that some of the other things i came across (ibm unix setup).

The link is below.

Link

v00d00101 · 02-01-2007, 01:25 PM

Finally ive worked out the cause of the problem, the first gig or so of my sata drive is failing. That is what has caused the lockups i think. Still have no idea why it said the sata was mounted at hdg, when it is actually mounted at sda, but with everything else thats been going on, i cant say it surprises me anymore. Maybe a quirk of fedora 5 x86_64.

Weird behaviour has been observed a late, such as, commands not working due to i/o errors, passwords not responding as if changed by someone. But both chkrootkit and rkhunter checks come back negative for signs of any intrusion. Also my last aide update checks back ok, although some of my programs have been updated by yum since then, resulting in hash fails. the passwords were checked from knoppix to see if any had changed since my backup of /etc after i installed and they were intact. I also did some log checking and ran chkrootkit, all with nothing detected.

At some point i will take this hard drive offline and cure the problem, maybe relegate it to scratch space.

Thanks Matir for your help.