On Wed, 2014-07-16 at 15:08 -0400, Martin K. Petersen wrote: > >>>>> "James" == James Bottomley <jbottomley@xxxxxxxxxxxxx> writes: > > James> It's the code we identified in sd.c:read_capacity_16(): > > That's there to support devices that implement thin provisioning but > which predate the LBP VPD page. And thus have no way to tell us their > preferred command variant. > > James> If the inquiry shows LBPME set, we'll do write same regardless of > James> the no_write_same bit. That's why for SPC-2 devices it only > James> shows up on >> 2TB devices because they force RC16. > > But we only do so if they have LBPME set. Generally they don't. > > The problem for hyperv was that LBPME was set. And because it claimed > SBC-2 we did not check the LBP VPD page to see that it prefers UNMAP. > > I believe we did consider unconditionally checking for VPDs when LBPME > was set but I seem to recall that it broke some device that had garbage > in the READ CAPACITY(16) response. I'll see if I can locate the details. > > Otherwise I'm willing to entertain that idea. Well, your judgement: is this situation (support for UNMAP but not for WRITE_SAME) in what is effectively a RAID driver (hv drivers count as RAID) just a silly Microsoft one off? If so, we can go with your force VPD page patch. However, if we get any RAID drivers with strange discard support, we'll likely get the same problem ... I've got to say in the long run, it sounds likely we will given that RAID vendors are only marginally more competent than USB bridge vendors when it comes to following the spec. James -- To unsubscribe from this list: send the line "unsubscribe stable" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html