Re: [RFC 3/7] lib: add rocksoft model crc64

From: Keith Busch <kbusch@kernel.org>
To: Eric Biggers <ebiggers@kernel.org>
Cc: linux-nvme@lists.infradead.org, linux-kernel@vger.kernel.org,
	linux-block@vger.kernel.org, axboe@kernel.dk, hch@lst.de,
	martin.petersen@oracle.com, colyli@suse.de, arnd@arndb.de
Subject: Re: [RFC 3/7] lib: add rocksoft model crc64
Date: Tue, 25 Jan 2022 07:39:50 -0800	[thread overview]
Message-ID: <20220125153950.GB1683949@dhcp-10-100-145-180.wdc.com> (raw)
In-Reply-To: <Ye+GTkrZ5YMe5qGt@sol.localdomain>

On Mon, Jan 24, 2022 at 09:10:38PM -0800, Eric Biggers wrote:
> On Mon, Jan 24, 2022 at 08:01:03AM -0800, Keith Busch wrote:
> > Since this model reflects inputs and outputs, a helper table and
> > function are added to reverse bits of 8 and 64 bit values.
> 
> That's a silly way to do a bit-reflected CRC.  The proper way to do it is to
> reflect the bytes too, so that the bits and bytes are ordered consistently, so
> explicitly reflecting the bits isn't needed.  Most CRC-32's are bit-reflected,
> and they are usually implemented this way.  E.g., see crc32_le() in lib/crc32.c.
> 
> Here's a Python script that shows that the Rocksoft CRC-64 can be computed
> without explicitly reversing the bits.  It passes the tests from your patch 4:

Thanks for the tip! I'll use that for the next version.