Re: Any hope for ide-scsi (error handling)?

[Willem Riede]

On 2003.03.15 20:23 Alan Cox wrote:
> On Sat, 2003-03-15 at 21:01, Willem Riede wrote:
> > It may not be elegant to schedule(1) with the lock taken, but it
> > does work.
> 
> You can't sleep holding a lock. You also can't null the hwgroup that
> way and you have to deal with some other locking concerns. Have a look
> how HDIO_*_RESET is handled in the very latest 2.4/2.5-ac code and
> you should be able to follow from that. Note your code paths will be
> much like the ioctl as the existing reset code paths are for paths
> where we abort from a known safe state (drive initiated or locked).
> 
> With the stuff there now you should be able to abort the command
> fairly cleanly and then reset the interface.
> 
I hear you, and I will take a hard look at that code. But please
realize, that if we get to this code segment, the drive has _not_ 
responded to the regular command and is in an _unknown_ state.

Thanks, Willem Riede.