On 11/02/2010 05:21 AM, Andi Kleen wrote:
IOW, it's doing the exact opposite of what the previous code did
(release the scsi host lock, before acquiring the ATA port/host
spinlock), not at all an equivalent transformation.
The following sequence would seem to better preserve the existing
lock profile, correct?
Possibly, but it's not a mechanic change.
Oh come on. Anybody can run a script. It's not a mechanical change if
you failed to create an equivalent transformation, fail to maintain
existing lock order, _inverting_ the existing locking.
Have you done any analysis on the correctness of this new locking?
The goal here is not really what comes out of this patch,
but dropping the host lock completely. This is just the first step.
That doesn't excuse lack of analysis or correctness.
Boaz' approach is OBVIOUSLY mechanical, correct and bisectable. Yours
is not.
Jeff
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html