Bug 27614

We (Red Hat) should really try to resolve this before next release.

I have found further disk corruption, having previously successfully run fsck -f
(single user) without errors. fsck -nf now tells me thing like
Inode 55105 has illegal block(s).
Illegal block #0 (4041469680) in inode 55105.  IGNORED.
and
Error while iterating over blocks in inode 55105: Illegal triply indirect block
found

Also I have spotted errors such as the following occuring in the
/var/log/messages file, which didn't occur when I was running RH6.2 .

Feb 16 09:21:05 itspc116 kernel: hda: status error: status=0x58 { DriveReady
SeekComplete DataRequest }
Feb 16 09:21:06 itspc116 kernel: hda: drive not ready for command

What ide controller, mainboard and disks are you using. What is the exact
version of the
kernel that is running on this system?

This seems to be a kernel problem with resulting disk corruption. I'll reassign
this to the kernel
rpm, but will watch further info about it.

I am not an expert on hardware so I hope this makes sense
Motherboard: ATX (pentium 166) "RM Advanced/ML Pentium Systemboard"
IDE controller: PIIX3 "82371SB PCI ISA/IDE Xcelerator"
Hard Disk: ST32132A
Kernel: 2.4.0-0.99.11

Can you please try newer kernels from
ftp://ftp.redhat.com/pub/rawhide/i386/RedHat/RPMS/
or from ftp://ftp.redhat.com/pub/redhat/beta/wolverine/i386/RedHat/RPMS/ to
check if
newer kernels have this already fixed?

I have upgraded the kernel to that in wolverine (2.4.1-0.1.9). The "drive not
ready" messages are still there (if they are in fact related to the disk
corruption), eg.
Feb 22 11:36:25 itspc116 kernel: hda: status error: status=0x58 { DriveReady
SeekComplete DataRequest }
Feb 22 11:36:25 itspc116 kernel: hda: drive not ready for command
but it may be several days before the disk corruption reappears, if the upgrade
hasn't fixed it.

I have some more evidence that suggests the problem is still there. I upgraded
the XFree packages to wolverine, and afterwards fsck reports some block bitmap
differences, even when the file system is mounted read-only, eg.
Pass 5: Checking group summary information
Block bitmap differences:  -186349 -186350 -186351 -186352 -186353 -186354
-186355 -187286 -187287 -187288 -187289 -187290 -187291 -187292 -195118 -195119
-195120 -195121 -195122 -195123 -195124 -195203 -195204 -195205 -195206 -195207
-195208 -195209

I had some more file corruption yesterday. I logged the fsck session to clean it
(from a second partition) if this information would be useful.

Can you try 2.4.1-0.1.14 from rawhide?

If that doesn't fix it, the next rawhide we put out will have a
"nodma" option that will make it easier to debug this.

I am on 2.4.1-0.1.14 now, I haven't seen any file corruption yet, (though the
system has hung, requiring a reset), but entries in /var/log/messages look
suspicious, for example

Mar  1 14:01:19 itspc116 kernel: hda: status error: status=0x58 { DriveReady
SeekComplete DataRequest }
Mar  1 14:01:19 itspc116 kernel: hda: drive not ready for command
Mar  1 14:01:19 itspc116 kernel: attempt to access beyond end of device
Mar  1 14:01:19 itspc116 kernel: 03:03: rw=0, want=790435384, limit=901152

I have had some more file/directory corruption with the 2.4.2-0.1.19 kernel. I
have a log of the fsck session afterwards (logged to a separate partition) and
messages in /var/log/messages if any of this is useful.

I have had more corruption with 2.4.2-0.1.28 (while I was upgrading to
2.4.2-0.1.49). Again I have more details if you want them.

{ DriveReady SeekComplete DataRequest } is usually an indication that
your cables are outside of the allowed limits. This usually shows up only
when using the higher DMA modes which our kernel now does for a while. 
If you don't want to change cables, you can always boot with "ide=nodma" on the
commandline of the kernel (eg on the lilo prompt)....

Please test this and reopen the bug if this doesn't help.

I encountered this problem, and the ide=nodma boot option avoids it, but I'm
doubtful of a cable problem since this system is a laptop.

Error indications in syslog were:
May 15 10:14:07 cjohnsonPC kernel: hda: status error: status=0x58 { DriveReady
SeekComplete DataRequest }
May 15 10:14:07 cjohnsonPC kernel: hda: drive not ready for command

The HW/SW information in syslog were:
May 15 10:11:52 cjohnsonPC kernel: Uniform Multi-Platform E-IDE driver Revision:
6.31
May 15 10:11:52 cjohnsonPC kernel: ide: Assuming 33MHz system bus speed for PIO
modes; override with idebus=xx
May 15 10:11:52 cjohnsonPC kernel: PIIX4: IDE controller on PCI bus 00 dev 39
May 15 10:11:52 cjohnsonPC kernel:     ide0: BM-DMA at 0x38a0-0x38a7, BIOS
settings: hda:DMA, hdb:DMA
May 15 10:11:52 cjohnsonPC kernel:     ide1: BM-DMA at 0x38a8-0x38af, BIOS
settings: hdc:pio, hdd:pio
May 15 10:11:52 cjohnsonPC kernel: hda: IBM-DARA-212000, ATA DISK drive
May 15 10:11:52 cjohnsonPC kernel: ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
May 15 10:11:52 cjohnsonPC kernel: hda: 23579136 sectors (12073 MB) w/418KiB
Cache, CHS=1559/240/63, UDMA(33)
May 15 10:11:52 cjohnsonPC kernel:  hda: hda1 hda2 hda3 < hda5 hda6 hda7 >

I am running kernel-2.4.2-2 i686 straight out of RH 7.1.
The laptop is a Compaq Armada M700.

If someone at RedHat or Compaq would persue this issue I would gladly provide
any needed info.