RE: [Open-FCoE] [RFC PATCH] scsi, fcoe, libfc: drop scsi host_lock use from fc_queuecommand

[Vasu Dev <vasu.dev@linux.intel.com>]

On Thu, 2010-09-02 at 12:48 -0700, Nicholas A. Bellinger wrote:
> > Yeah but direct use makes things more obvious at first look. However
> > neatness is worthy using wrapper(macros) if there are several such
> > places. Anycase this is very minor code style thing here and I'm
> fine
> > with wrapper if you want.
> 
> Sure, I am thinking about these simple host_lock wrappers as more of a
> transitional look for LLDs more than anything..
> 
> Btw, I would be happy to include your forthcoming v2 patch into a
> lio-core-2.6.git branch, and give it some testing in the next week.
> 

Awesome, Thanks for your all help Nab,

	I tried to have wrapper instead of checks to drop host_lock before
fc_queuecommand using wrapper something like this :-

+static inline void scsi_qcmd_host_unlock(struct Scsi_Host *shost,
unsigned long irq_flags)
+{
+       if (shost->host_lock_pending) {
+               shost->host_lock_pending = 0;
+               spin_unlock_irqrestore(shost->host_lock, irq_flags);
+       } else if (shost->unlocked_qcmds)
+               spin_unlock_irqrestore(shost->host_lock, irq_flags);
+       else
+               shost->host_lock_pending = 1;
+}
+

This didn't work well beside required lot more checks to track host lock
so that this wrapper can be called w/o checks as:-

	scsi_qcmd_host_unlock(host, flags);
	rtn = host->hostt->queuecommand(cmd, scsi_done);
	scsi_qcmd_host_unlock(host, flags);

I think it is better of with simple checks for now as I posted in my
patch, may be a wrapper can be added in case more places neeeds such
checks as we talked before.

	Thanks
	Vasu