How to distinguish MLC and SLC NANDs?

How to distinguish MLC and SLC NANDs?

[Juergen Beisert]

Hi list,

my NAND controller is able to handle different corrections sum types (a weak 
one for SLC, a more stronger one for MLC). Is there a unique pattern for MLC 
or SLC in the mtd structure available after calling nand_scan_ident() to 
select what correction sum type should be used for the attached NAND?
Or should I always use the strongest type, independent of the NAND type?

Thanks in advance for some hints.

jbe

-- 
Pengutronix e.K.                              | Juergen Beisert             |
Linux Solutions for Science and Industry      | http://www.pengutronix.de/  |

Re: How to distinguish MLC and SLC NANDs?

[Brian Norris]

On Thu, Sep 15, 2011 at 7:33 AM, Juergen Beisert <jbe@pengutronix.de> wrote:
> Is there a unique pattern for MLC
> or SLC in the mtd structure available after calling nand_scan_ident() to
> select what correction sum type should be used for the attached NAND?

Actually, there isn't anything available in struct mtd_info. This
belongs in struct nand_chip, actually, since it's a NAND-specific
feature. In fact, there is some data available there. As you can see
in nand_get_flash_type() in nand_base.c, we fill in the
`chip->cellinfo' field, then later use a mask on it like this:

chip->cellinfo & NAND_CI_CELLTYPE_MSK

According to the data sheets for *most* NAND parts, this is a safe way
of determining SLC vs. MLC (non-zero => MLC; zero => SLC), and we use
it for certain detection code paths. Note that if the chip uses ONFI,
you supply your own ID table, or the chip uses some of the older ID
codes, nand_base.c won't set chip->cellinfo for you; but it will
default to SLC I think, which is the correct option for old parts.

But it's up to you to determine your own reliance on this method. I
don't know of any users who rely on `chip->cellinfo &
NAND_CI_CELLTYPE_MSK' for this type of information.

Brian

Re: How to distinguish MLC and SLC NANDs?

[Juergen Beisert]

Hi Brian,

Brian Norris wrote:
> On Thu, Sep 15, 2011 at 7:33 AM, Juergen Beisert <jbe@pengutronix.de> wrote:
> > Is there a unique pattern for MLC
> > or SLC in the mtd structure available after calling nand_scan_ident() to
> > select what correction sum type should be used for the attached NAND?
>
> Actually, there isn't anything available in struct mtd_info. This
> belongs in struct nand_chip, actually, since it's a NAND-specific
> feature. In fact, there is some data available there. As you can see
> in nand_get_flash_type() in nand_base.c, we fill in the
> `chip->cellinfo' field, then later use a mask on it like this:
>
> chip->cellinfo & NAND_CI_CELLTYPE_MSK
>
> According to the data sheets for *most* NAND parts, this is a safe way
> of determining SLC vs. MLC (non-zero => MLC; zero => SLC), and we use
> it for certain detection code paths.

Hmm, *most* means this information is not reliable? Would it be better to give 
the driver a hint what kind of NAND it should expect via the platform_data?

> Note that if the chip uses ONFI, you supply your own ID table, or the chip
> uses some of the older ID codes, nand_base.c won't set chip->cellinfo for
> you; but it will default to SLC I think, which is the correct option for old
> parts. 

BTW: Is there a known disadvantage or incompatibility if I protect the data of 
an SLC with the more stronger ECC sums intended for MLCs (beside the effect 
the OOB area is more filled than required)?

> But it's up to you to determine your own reliance on this method. I
> don't know of any users who rely on `chip->cellinfo &
> NAND_CI_CELLTYPE_MSK' for this type of information.

I was in the hope of a perfect world and the MTD framework detects the NAND 
type reliable ;-). Now I tend more to use user's info from the platform_data.

Thanks for your answer.

Juergen

-- 
Pengutronix e.K.                              | Juergen Beisert             |
Linux Solutions for Science and Industry      | http://www.pengutronix.de/  |

RE: How to distinguish MLC and SLC NANDs?

[Atlant Schmidt]

Juergen :

> BTW: Is there a known disadvantage or incompatibility if I protect the data of
> an SLC with the more stronger ECC sums intended for MLCs (beside the effect
> the OOB area is more filled than required)?

  Just speed and storage space (which you already realized).

  Better ECCs are otherwise, well, "better"!

  More generally, if you have a relatively small number of
  well-known NAND Flash types, you could probably use the
  Vendor ID and Device Type codes and a table to determine
  which parts are SLC and which are MLC. It's a shame this
  mechanism wasn't built in from the beginning; the same
  question came up just the other day in our shop and we
  ended up collecting all the Device Type codes (we only
  use one vendor and so far, three specific chip types).

                             Atlant


-----Original Message-----
From: linux-mtd-bounces@lists.infradead.org [mailto:linux-mtd-bounces@lists.infradead.org] On Behalf Of Juergen Beisert
Sent: Friday, September 16, 2011 02:13
To: linux-mtd@lists.infradead.org
Cc: Brian Norris
Subject: Re: How to distinguish MLC and SLC NANDs?

Hi Brian,

Brian Norris wrote:
> On Thu, Sep 15, 2011 at 7:33 AM, Juergen Beisert <jbe@pengutronix.de> wrote:
> > Is there a unique pattern for MLC
> > or SLC in the mtd structure available after calling nand_scan_ident() to
> > select what correction sum type should be used for the attached NAND?
>
> Actually, there isn't anything available in struct mtd_info. This
> belongs in struct nand_chip, actually, since it's a NAND-specific
> feature. In fact, there is some data available there. As you can see
> in nand_get_flash_type() in nand_base.c, we fill in the
> `chip->cellinfo' field, then later use a mask on it like this:
>
> chip->cellinfo & NAND_CI_CELLTYPE_MSK
>
> According to the data sheets for *most* NAND parts, this is a safe way
> of determining SLC vs. MLC (non-zero => MLC; zero => SLC), and we use
> it for certain detection code paths.

Hmm, *most* means this information is not reliable? Would it be better to give
the driver a hint what kind of NAND it should expect via the platform_data?

> Note that if the chip uses ONFI, you supply your own ID table, or the chip
> uses some of the older ID codes, nand_base.c won't set chip->cellinfo for
> you; but it will default to SLC I think, which is the correct option for old
> parts.

BTW: Is there a known disadvantage or incompatibility if I protect the data of
an SLC with the more stronger ECC sums intended for MLCs (beside the effect
the OOB area is more filled than required)?

> But it's up to you to determine your own reliance on this method. I
> don't know of any users who rely on `chip->cellinfo &
> NAND_CI_CELLTYPE_MSK' for this type of information.

I was in the hope of a perfect world and the MTD framework detects the NAND
type reliable ;-). Now I tend more to use user's info from the platform_data.

Thanks for your answer.

Juergen

--
Pengutronix e.K.                              | Juergen Beisert             |
Linux Solutions for Science and Industry      | http://www.pengutronix.de/  |

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/


 Click https://www.mailcontrol.com/sr/IXaeqh!Rro7TndxI!oX7Us7Qlo!t9IHtvEZSmYARWeyYDIP5X8F0kaKcsMuDG4ISzSBCnLyI1OpLwdQMGDtFKA==  to report this email as spam.

This e-mail and the information, including any attachments, it contains are intended to be a confidential communication only to the person or entity to whom it is addressed and may contain information that is privileged. If the reader of this message is not the intended recipient, you are hereby notified that any dissemination, distribution or copying of this communication is strictly prohibited. If you have received this communication in error, please immediately notify the sender and destroy the original message.

Thank you.

Please consider the environment before printing this email.

Re: How to distinguish MLC and SLC NANDs?

[Brian Norris]

On Thu, Sep 15, 2011 at 11:12 PM, Juergen Beisert <jbe@pengutronix.de> wrote:
> Brian Norris wrote:
>> chip->cellinfo & NAND_CI_CELLTYPE_MSK
>>
>> According to the data sheets for *most* NAND parts, this is a safe way
>> of determining SLC vs. MLC (non-zero => MLC; zero => SLC), and we use
>> it for certain detection code paths.
>
> Hmm, *most* means this information is not reliable?

*most* means a few things here:

1) this mechanism is only used for a small portion of our detection
code and doesn't have a wide variety of test coverage, I don't think.
If this code were incorrect, most users of this code would at most get
a slightly different bad block scanning pattern; this kind of issue
isn't always real easy to notice. Also, there are some Samsung MLC
that rely on this info for their code path.
2) this mask is defined in many NAND data sheets so that it is safe to
use on any chips that I have seen. Some don't specify everything but
still tend to follow this pattern.
3) NAND does not have a real standard for decoding ID bytes, although
there are a couple of patterns that have become de-facto standards
(with various exceptions, of course; none noticed on this particular
issue)

> Would it be better to give the driver a hint what kind of NAND it should expect via the platform_data?

Your prior knowledge of the NAND chip would probably be more reliable.

> I was in the hope of a perfect world and the MTD framework detects the NAND
> type reliable ;-). Now I tend more to use user's info from the platform_data.

Unfortunately, there is little perfection in NAND standards.

Brian