IDE layer adoptations that *fake* a failing drive?

IDE layer adoptations that *fake* a failing drive?

[Leon Woestenberg]

Hello,

I wondered if anyone here is aware of code that sits between the IDE
software layer and the storage device, that emulates a failing disk in 
one way or another.

I.e. for example, write multiple re-tries with time-outs, checksum
errors, failing to complete read/writes to medium etc.

 From my networking stuff, I know there are network simulators that you
can provoke to simulate all kind of network traffic abnormalities, such
as congestion, dupes, lost packets, bursting.

I have access to only a limited number of really bad disks, and want to
do stress tests against fully automatically managed software RAID arrays
under Linux (except for the part where you have to physically swap 
defect disks :-)

Ciao,

Leon.

Re: IDE layer adoptations that *fake* a failing drive?

[Pat LaVarre]

> code that sits between the IDE software layer and
> the storage device, that emulates a failing disk
> in one way or another.

I guess you want intermittent and unreliable failure ... else you could
swap in an ATAPI drive, since ATAPI protocol by definition looks like a
dead ATA drive.

Like you, I want to eject errors and timeouts, please post back here
what you find.  Meanwhile, three thoughts:

1) Read errors

Via ATA pass thru you can force a read error, if your drive supports the
write/ read long legacy of ATA 2 for that purpose.

2) No-data timeouts

Via 48-bit lba ATA you can force a slow timeout by asking to verify
xFFFF lba's = 32 MiB = (1 << (16+9)) bytes, at 512 = x200 = (1 << 9)
bytes/lba, if you drive actually does verify blocks when asked.

3) Write errors, data timeouts, etc.

Tough, without customarily proprietary knowledge of the drive.

Pat LaVarre