linux-kernel - Re: [PATCH] nvme: Add support for Apple 2018+ models

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 15 Jul 2019 19:03:31 +1000
From:   Benjamin Herrenschmidt <benh@...nel.crashing.org>
To:     Christoph Hellwig <hch@....de>
Cc:     linux-nvme@...ts.infradead.org,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        Keith Busch <kbusch@...nel.org>, Jens Axboe <axboe@...com>,
        Paul Pawlowski <paul@...rm.io>
Subject: Re: [PATCH] nvme: Add support for Apple 2018+ models

On Mon, 2019-07-15 at 18:43 +1000, Benjamin Herrenschmidt wrote:
> On Mon, 2019-07-15 at 10:10 +0200, Christoph Hellwig wrote:
> > > +	/*
> > > +	 * Apple 2018 and latter variant has a few issues
> > > +	 */
> > > +	NVME_QUIRK_APPLE_2018			= (1 << 10),
> > 
> > We try to have quirks for the actual issue, so this should be one quirk
> > for the irq vectors issues, and another for the sq entry size.  Note that
> > NVMe actually has the concept of an I/O queue entry size (IOSQES in the
> > Cc register based on values reported in the SQES field in Identify
> > Controller.  Do these controllers report anything interesting there?
> 
> Ah good to know, I'll dig.

Interesting... so SQES is 0x76, indicating that it supports the larger
entry size but not that it mandates it.

However, we configure CC:IOSQES with 6 and the HW fails unless we have
the 128 bytes entry size.

So the HW is bogus, but we can probably sort that by doing a better job
at fixing up SQES in the identify on the Apple HW, and then actually
using it for the SQ.

I checked and CC is 0x00460001 so it takes our write of "6" fine. I
think they just ignore the value.

How do you want to proceed here ? Should I go all the way at attempting
to honor sqes "mandatory" size field (and quirk *that*) or just I go
the simpler way and stick to shift 6 unless Apple ?

If I go the complicated path, should I do the same with cq size
(knowing that no known HW has a non-4 mandatory size there and we don't
know of a HW bug... yet).

Cheers,
Ben.