Re: [PATCH v2] mtd: nand: brcmnand: fix mtd corrected bits stat

From: Miquel Raynal <miquel.raynal@bootlin.com>
To: David Regan <dregan@broadcom.com>
Cc: linux-kernel@vger.kernel.org, linux-mtd@lists.infradead.org,
	bcm-kernel-feedback-list@broadcom.com,
	william.zhang@broadcom.com, anand.gore@broadcom.com,
	florian.fainelli@broadcom.com, kamal.dasu@broadcom.com,
	dan.beygelman@broadcom.com,
	"Álvaro Fernández Rojas" <noltari@gmail.com>,
	rafal@milecki.pl, computersforpeace@gmail.com,
	frieder.schrempf@kontron.de, vigneshr@ti.com, richard@nod.at,
	bbrezillon@kernel.org, kdasu.kdev@gmail.com,
	jaimeliao.tw@gmail.com, kilobyte@angband.pl,
	jonas.gorski@gmail.com, dgcbueu@gmail.com, dregan@mail.com
Subject: Re: [PATCH v2] mtd: nand: brcmnand: fix mtd corrected bits stat
Date: Mon, 09 Jun 2025 11:20:09 +0200	[thread overview]
Message-ID: <8734c9z2ye.fsf@bootlin.com> (raw)
In-Reply-To: <20250606165756.1531164-1-dregan@broadcom.com> (David Regan's message of "Fri, 6 Jun 2025 09:57:03 -0700")

On 06/06/2025 at 09:57:03 -07, David Regan <dregan@broadcom.com> wrote:

> Currently we attempt to get the amount of flipped bits from a hardware
> location which is reset on every subpage. Instead obtain total flipped
> bits stat from hardware accumulator. In addition identify the correct
> maximum subpage corrected bits.
>
> Signed-off-by: David Regan <dregan@broadcom.com>
> Reviewed-by: William Zhang <william.zhang@broadcom.com>
> ---

Hello,

Can you please give the output of nandbiterrs -i /dev/mtdX?

>  v2: Add >= v4 NAND controller support as requested by Jonas.
>      mtd->ecc_stats.corrected accumulates instead of set to total.
>      Remove DMA specific flipped bits count.

The changelog does not mention the fact that you return the maximum
number of corrected bitflips as I requested, and the diff does not show
a straightforward implementation of that. It is very important to get
this right.

If we take the following example of a page with 4 ECC steps, if we get
respectively: 0, 2, 3, 0 bitflips per step, the returned value shall be
3.

To be very certain that this is correct, you can use the nandflipbit
tool from the mtd-utils test suite. You can manually insert bitflips in
various areas of a page and then read the page again with ECC enabled
and see how many bit errors are reported.

Thanks,
Miquèl

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