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]

Message-ID: <c82b6aa094f1681272ec1e2a55e47758c435f784.camel@kernel.crashing.org>
Date:   Mon, 15 Jul 2019 20:01:05 +1000
From:   Benjamin Herrenschmidt <benh@...nel.crashing.org>
To:     Maxim Levitsky <mlevitsk@...hat.com>,
        Christoph Hellwig <hch@....de>
Cc:     Keith Busch <kbusch@...nel.org>, Jens Axboe <axboe@...com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        linux-nvme@...ts.infradead.org, Paul Pawlowski <paul@...rm.io>
Subject: Re: [PATCH] nvme: Add support for Apple 2018+ models

On Mon, 2019-07-15 at 12:28 +0300, Maxim Levitsky wrote:
> 
> To be honest, the spec explicitly states that minimum submission queue entry size is 64 
> and minimum completion entry size should be is 16 bytes for NVM command set:
> 
> "Bits 3:0 define the required (i.e., minimum) Submission Queue Entry size when
> using the NVM Command Set. This is the minimum entry size that may be used.
> The value is in bytes and is reported as a power of two (2^n). The required value
> shall be 6, corresponding to 64."

Yes, I saw that :-) Apple seems to ignore this and CC:IOSQES and
effectively hard wire a size of 7 (128 bytes) for the IO queue.

> "Bits 3:0 define the required (i.e., minimum) Completion Queue entry size when using
> the NVM Command Set. This is the minimum entry size that may be used. The value
> is in bytes and is reported as a power of two (2^n). The required value shall be 4,
> corresponding to 16."
> 
> Pages 136/137, NVME 1.3d.
> 
> In theory the spec allows for non NVM IO command set, and for which the sq/cq entry sizes can be of any size,
> as indicated in SQES/CQES and set in CC.IOCQES/CC.IOSQES, but than most of the spec won't apply to it.
> 
> 
> Also FYI, values in CC (IOCQES/IOSQES) are for I/O queues, which kind of implies that admin queue,
> should always use the 64/16 bytes entries, although I haven't found any explicit mention of that.

Right, and it does on the Apple HW as well.

Cheers,
Ben.