2.4.9 to 2.4.14 bug & workaround

2.4.9 to 2.4.14 bug & workaround

[joeja]

I have an internal iomega 'type' (not iomega) IDE zip drive.  It mounts as /dev/hdd instead of /dev/hdd4.  Mounting as /dev/hdd seems okay.

Mounting as /dev/hdd4 will hang my kernel( 2.4.9-2.4.14).  I have read that on some MB you can change the bios to none for the ide device and this works on certain mb.  I tried this and it did not change anything.

MB is ABIT KT7A.  

Bug, some disks will mount as /dev/hdd4 if you try.  This seems to have erratic behavior as sometimes you will have no problems, othertimes it will tell you you have mounted Read only even though mount show rw.  If you copy large files to the drive (38 to 50 meg backups) you may have it hang in the middle of cp.

you can eject the disk from the drive, you cannot kill cp, you cannot properly shutdown system as there are messages about hdd umount failing and this just hangs the system and becomes unusable.

Workaround seems to be mount as /dev/hdd.  This seems wrong though.  

Reproducability is often.

Joe

Re: 2.4.9 to 2.4.14 bug & workaround

[Matthias Andree]

On Mon, 12 Nov 2001, joeja@mindspring.com wrote:

(reformatted quote to heed line length provisions)
> I have an internal iomega 'type' (not iomega) IDE zip drive.  It
> mounts as /dev/hdd instead of /dev/hdd4.  Mounting as /dev/hdd seems
> okay.
> 
> Mounting as /dev/hdd4 will hang my kernel( 2.4.9-2.4.14).  I have read
> that on some MB you can change the bios to none for the ide device and
> this works on certain mb.  I tried this and it did not change
> anything.

Well, all this depends on how the media has been formatted. From what I
heard (I have no exchangable drives since I sold my SyQuest crap), some
ZIP (presumably) media are partitioned and have their fourth partition
formatted, giving your data as /dev/hdd4 or /dev/sdb4 or something other
with 4 to the end. Then, the entire media might be formatted "raw" with
Linux, so you'd have to mount /dev/hdd instead.

The BIOS is only involved for booting and come APM or ACPI (power
saving) stuff, maybe also for VESA frame buffer stuff, but other than
that, its settings are partially read and the BIOS is no longer taken
into account.

Of course, a system hang is Not Nice[tm] and should be fixed by the
respective maintainers IF IT IS an actual bug. Please check the cabling
or use hdparm to use more conservative transfer modes before assuming
it's a bug with Linux -- a "cp" hang certainly warrants investigation.

OTOH, mount should not hang if you mount the wrong part of the disk, but
complain instead, however, I'm not sure if a bogus partition entry can
cause these hangs.

Re: 2.4.9 to 2.4.14 bug & workaround

[Peter Wächtler]

Matthias Andree wrote:
> 
> On Mon, 12 Nov 2001, joeja@mindspring.com wrote:
> 
> (reformatted quote to heed line length provisions)
> > I have an internal iomega 'type' (not iomega) IDE zip drive.  It
> > mounts as /dev/hdd instead of /dev/hdd4.  Mounting as /dev/hdd seems
> > okay.
> >
> > Mounting as /dev/hdd4 will hang my kernel( 2.4.9-2.4.14).  I have read
> > that on some MB you can change the bios to none for the ide device and
> > this works on certain mb.  I tried this and it did not change
> > anything.
> 
> Well, all this depends on how the media has been formatted. From what I
> heard (I have no exchangable drives since I sold my SyQuest crap), some
> ZIP (presumably) media are partitioned and have their fourth partition
> formatted, giving your data as /dev/hdd4 or /dev/sdb4 or something other
> with 4 to the end. Then, the entire media might be formatted "raw" with
> Linux, so you'd have to mount /dev/hdd instead.
> 

Yes, in theory ;-)
I have an internal IDE 100MB Zip @home.
In the office I have a 250MB USB ZIP.

The media is partitioned, with the 4th primary partition formatted
as VFAT.

On the IDE I mount /dev/hdb, on the USB thing I mount sd[ab]4
depending if the flash reader is there or not.
Hmh?

# fdisk -l /dev/sdb
 
Disk /dev/sdb: 4 heads, 32 sectors, 1536 cylinders
Units = cylinders of 128 * 512 bytes
 
   Device Boot    Start       End    Blocks   Id  System
/dev/sdb4             1      1536     98288    6  FAT16



Of course I use different kernels, @home: 2.4.9ac18 and 2.4.1[34]-xfs;
@office for now 2.4.9ac3+bcl.

Until now I thought it had something to do with the different gendisk,
LDM or so.

Re: 2.4.9 to 2.4.14 bug & workaround

[Matthias Andree]

Peter Wächtler schrieb am Dienstag, den 13. November 2001:

> On the IDE I mount /dev/hdb, on the USB thing I mount sd[ab]4
> depending if the flash reader is there or not.
> Hmh?

Do these behave differently? In particular, do the IDE Zip drives hide
the partition structure...

>    Device Boot    Start       End    Blocks   Id  System
> /dev/sdb4             1      1536     98288    6  FAT16

...which is evidently there?

> Until now I thought it had something to do with the different gendisk,
> LDM or so.

Well, you may also see firmware and/or design flaws in the drive
(personally, I have never trusted iomega, because on the CeBIT fair in
Hannover, I once asked them "why should I prefer iomega ZIP or JAZ over
SyQuest" and they had no answer except "we're just better". I later
heard complaints about the SCSI ID only to be chosen from 5 or 6, 25-pin
SCSI connectors and stuff, then there was the click-of-death sabotage
and now there is your "partition entry or not" problem.)

http://www.win.tue.nl/~aeb/linux/zip/zip-1.html has some info which does
not look too promising when you're after consistent behaviour across the
various drive types (interface-wise, that is).

Judging from what's on that page, the IDE driver seems to know it's just
a "floppy" without partitions, but the USB driver sees the (fake)
partitions.

-- 
Matthias Andree

"They that can give up essential liberty to obtain a little temporary
safety deserve neither liberty nor safety."         Benjamin Franklin

Re: 2.4.9 to 2.4.14 bug & workaround

[Peter Wächtler]

Matthias Andree schrieb:
> 
> Peter Wächtler schrieb am Dienstag, den 13. November 2001:
> 
> > On the IDE I mount /dev/hdb, on the USB thing I mount sd[ab]4
> > depending if the flash reader is there or not.
> > Hmh?
> 
> Do these behave differently? In particular, do the IDE Zip drives hide
> the partition structure...
> 
> >    Device Boot    Start       End    Blocks   Id  System
> > /dev/sdb4             1      1536     98288    6  FAT16
> 
> ...which is evidently there?
> 
> > Until now I thought it had something to do with the different gendisk,
> > LDM or so.
> 
> Well, you may also see firmware and/or design flaws in the drive
> (personally, I have never trusted iomega, because on the CeBIT fair in
> Hannover, I once asked them "why should I prefer iomega ZIP or JAZ over
> SyQuest" and they had no answer except "we're just better". I later
> heard complaints about the SCSI ID only to be chosen from 5 or 6, 25-pin
> SCSI connectors and stuff, then there was the click-of-death sabotage
> and now there is your "partition entry or not" problem.)
> 
> http://www.win.tue.nl/~aeb/linux/zip/zip-1.html has some info which does
> not look too promising when you're after consistent behaviour across the
> various drive types (interface-wise, that is).
> 
> Judging from what's on that page, the IDE driver seems to know it's just
> a "floppy" without partitions, but the USB driver sees the (fake)
> partitions.
> 

Wow, thanks for this link. It explains the details very well.

Nov 12 21:50:51 picklock kernel: hdb: 98288kB, 196576 blocks, 512 sector size
Nov 12 21:50:51 picklock kernel: VFS: Disk change detected on device ide0(3,68)
Nov 12 21:50:54 picklock kernel:  hdb: hdb1 hdb2 hdb3 hdb4
Nov 12 21:50:54 picklock kernel: ide-floppy: hdb: I/O error, pc = 28, key =  5,
Nov 12 21:50:54 picklock kernel: end_request: I/O error, dev 03:44 (hdb), sector

<4>hdb: IOMEGA ZIP 100 ATAPI Floppy, ATAPI FLOPPY drive

This was an attempt of (mdir z: with z: mapping to /dev/hdb4)



BTW, on the same day I bought this thing, "Linux" did demolish my only media.
But luckily I also bought a SB Live!, told the dealer, that I am running an
Athlon with 686_A_ (not B) southbridge and the soundcard does not work also.

I got a different soundcard and a new ZIP media for no extra costs ;-)