Re: scsipi mid-layer, tagged queueing, and ciss(4)

[Thor Lancelot Simon <tls%panix.com@localhost>]

On Mon, Aug 30, 2010 at 11:23:59AM +0200, Manuel Bouyer wrote:
> 
> A quick look shows that this could be the SCSI_SYNCHRONIZE_CACHE_10
> (BTW, this means that right now, no cache sync is done for ciss I guess),
> which can be worked around with the PQUIRK_NOSYNCCACHE quirk.
> sd_getcache()/sd_setcache() could also issue version 2-only commands, but
> AFAIK these are only called for user IOCTLs.

It seems like -- particularly for something like CISS which has quite a
bit of onboard memory holding data the host thinks have at least reached
the disks' caches -- we should indeed send the version 2 command if we
cannot send the other one, no?

Thor