Re: [RFC][PATCH] allow sleep inside EH hooks

[Jeff Garzik <jgarzik@pobox.com>]

Christoph Hellwig wrote:
> On Fri, May 27, 2005 at 03:48:40AM -0400, Jeff Garzik wrote:
> 
>>Christoph Hellwig wrote:
>>
>>>On Fri, May 27, 2005 at 12:32:07AM -0400, Jeff Garzik wrote:
>>>
>>>
>>>>SCSI EH processing already serializes things during EH, so this spinlock 
>>>>isn't really needed.
>>>>
>>>>Removing the spinlock outright would break drivers that surround logic 
>>>>with spin_unlock_irq()..spin_lock_irq(), so I introduced ->unlocked_eh 
>>>>option.
>>>
>>>
>>>Linus has vetoed such conditional locking in the past.  However if you do
>>>it don't make it EH specific but introduce a ->concurrent flag that 
>>>disables
>>>taking host_lock for ->queuecommand aswell.
>>
>>Such a 'concurrent' flag violates Linus credo "do what you must, and no 
>>more."  It's also silly and much too invasive.
>>
>>Removing the locking from the EH routines (only), and fixing up all 
>>necessary drivers, is much more appealing.
> 
> 
> No, hav ing the host_lock only held for ->queuecommand which doesn't
> need that locking doesn't make any sense.  An API like the current one

It makes a lot of sense:  LLDs are written with the assumption that 
paths called from ->queuecommand will not be interrupted by their own 
interrupt handler, whereas error handling paths are typically written 
with precisely the -opposite- assumption.

Removing spin_lock_irq() from queuecommand in SCSI EH causes problems, 
and solves nothing.

	Jeff