Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 28 Jun 2018 17:28:20 -0600
From:      Warner Losh <imp@bsdimp.com>
To:        Emmanuel Vadot <manu@bidouilliste.com>
Cc:        Adrian Chadd <adrian.chadd@gmail.com>,  "freebsd-embedded@freebsd.org" <freebsd-embedded@freebsd.org>
Subject:   Re: Atheros AR93xx NAND support
Message-ID:  <CANCZdfrGAuJLy2WM9pp_0uzaBZCTCirOTSo_kn7JzmqpUmM3iA@mail.gmail.com>
In-Reply-To: <20180629011042.6e72464d7fc3294d96dd093f@bidouilliste.com>
References:  <CAF-mK=Xpmi6OStLbosfdwW7KHL4XSPky_7JDZgEosYr3m709aQ@mail.gmail.com> <CAJ-Vmo=JQ-rgPEBOHKMigJB5Kt0QUFVWUKQAh99vvnNqFmoByw@mail.gmail.com> <CANCZdfpC2ExBPytm3dmeno%2B_ZX9nCWRUJb_zmu7TZGbbokc8AQ@mail.gmail.com> <20180629011042.6e72464d7fc3294d96dd093f@bidouilliste.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Jun 28, 2018 at 5:10 PM, Emmanuel Vadot <manu@bidouilliste.com>
wrote:

> On Thu, 28 Jun 2018 14:46:16 -0600
> Warner Losh <imp@bsdimp.com> wrote:
>
> > If you had good access to datasheets, plus errata, plus vendor support,
> you
> > might be able to get a quality implementation in 3-4 months time, which
> > would include time to revamp our NAND system. It doesn't include time to
> > revamp NANDFS, though, which would be another 2-3 months to get rock
> solid.
> >
> > Warner
>
>  I don't know the state of NAND in MIPS world but in the ARM world it's
> clearly fading out. But I've always wondered if instead of NANDFS one
> could do a geom_ftl (Flash Transision Layer) that will do all the nand
> stuff and then we can use UFS on it ?


That's logically appealing, but inefficient.The tables needed get kinda big
and/or you have to do a lot of NAND I/O to get the mappings since you are
hiding things from the UFS layer. NANDFS is a log filesystem, so it knows
how about how to interact with the append store that's NAND. It was one of
the biggest problems with Fusion I/O's driver: it had to keep this
translation table in memory. It made crash recovery difficult. It created
points of contention that limited performance. It could be done, but
experience suggests caution...

Warner



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CANCZdfrGAuJLy2WM9pp_0uzaBZCTCirOTSo_kn7JzmqpUmM3iA>