2.4.22 regression: ide-scsi no longer likes broken hardware

2.4.22 regression: ide-scsi no longer likes broken hardware

[Juliusz Chroboczek]

IDE chipset is Intel Corp. 82801DB ICH4 IDE (rev 01).

Both the CD-ROM on hdc and the CD-RW on hdd are broken, but work
flawlessly under 2.4.20.  According to cdrecord, they are:

  hdc: 'E-IDE   ' 'CD-ROM 56X/AKH  ' 'A8E ' Removable CD-ROM
  hdd: 'AOPEN   ' 'CD-RW CRW4048   ' '1.05' Removable CD-ROM

Both work under 2.4.22 as IDE devices.

Putting ide-scsi over hdc under 2.4.22 leads to a silent hang during
boot (just after the friendly ``ide-cd: passing drive...'' message).

With ide-scsi over hdd, on the other hand, the kernel gets into a loop
reporting that ``the scsi wants to send us more data than expected''.
I am sorry I didn't write the log down, I'll be glad to do that if
anyone thinks it can help.

In either case, a cold boot is necessary.

As an additional data point, cdrecord has the following complaints
about the drives:

  hdc: cdrecord.mmap: Warning: controller returns wrong size for CD capabilities page.  

  hdd: cdrecord.mmap: WARNING: Drive returns wrong startsec (0) using -150

I do realise that this is probably broken hardware.  However, 2.4.20
shows that it is possible to make it work, and I'm wondering what
happened that made 2.4.22 picky.

                                        Juliusz Chroboczek