Re: [ANNOUNCE] Status of unlocked_qcmds=1 operation for .37

[Boaz Harrosh]

On 10/28/2010 11:10 AM, Andi Kleen wrote:
> On Wed, Oct 27, 2010 at 09:27:38AM -0500, James Bottomley wrote:
>> On Wed, 2010-10-27 at 09:53 +0200, Andi Kleen wrote:
>>>> This sounds like a pretty reasonable compromise that I think is slightly
>>>> less risky for the LLDs with the ghosts and cob-webs hanging off of
>>>> them.
>>>
>>> They won't get tested either next release cycle. Essentially
>>> near nobody uses them.
>>>
>>>>
>>>> What do you think..?
>>>
>>> Standard linux practice is to simply push the locks down. That's a pretty
>>> mechanical operation and shouldn't be too risky
>>>
>>> With some luck you could even do it with coccinelle.
>>
>> Precisely ... if we can do the push down now as a mechanical
>> transformation we can put it in the current merge window as a low risk
>> API change.  This gives us optimal exposure to the rc sequence to sort
>> out any problems that arise (or drivers that got missed) with the lowest
>> risk of such problems actually arising.  Given the corner cases and the
>> late arrival of fixes, the serial number changes are just too risky for
>> the current merge window.  Having an API that changes depending on a
>> flag is also a high risk process because it's prone to further sources
>> of error.
> 
> Here's a coccinelle script I came up with that does the push down.
> It still adds a bogus empty line in front of the irqflags declaration
> which I haven't managed to avoid yet. Other than the it seems
> to DTRT on the SCSI drivers I tried.
> 
> -Andi
> 
> 
> @ rule1 @
> struct scsi_host_template t;
> identifier qc;
> @@
> t.queuecommand = qc;
> 
> @ rule2 @
> identifier rule1.qc;
> identifier cmnd;
> expression E;
> statement S, S2;
> @@
> int qc(struct scsi_cmnd *cmnd, ...) 
> {
> 	... when != S
> +	unsigned long irqflags;
> 
> +	spin_lock_irqsave(&cmnd->device->host->hostlock, irqflags);
> 	S2
> 	...
> +	spin_unlock_irqrestore(&cmnd->device->host->hostlock, irqflags);
> 	return E;
> }
> 

I disagree with your approach this introduces a spin_unlock_irqrestore
call site at every return, in the usually huge queuecommand.

I'd say just do:
- Rename XXX_queuecommand => __XXX_queuecommand_unlocked
- Define new XXX_queuecommand

int qc(struct scsi_cmnd *cmnd, ...) 
{
	unsigned long irqflags;
	int ret;

	spin_lock_irqsave(&cmnd->device->host->hostlock, irqflags);
	ret = __XXX_queuecommand_unlocked(cmnd, ...)
	spin_unlock_irqrestore(&cmnd->device->host->hostlock, irqflags);
	return ret;
}

Then when the driver is manually converted the __queuecommand_unlocked
can be set into the scsi_host_template and the added function can
be dropped.

My $0.017
Boaz

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/