Re: [PATCH] drivers, char: add U-Boot bootcount driver

From: Wolfram Sang <w.sang@pengutronix.de>
To: Heiko Schocher <hs@denx.de>
Cc: Wolfgang Denk <wd@denx.de>, Vitaly Bordug <vbordug@ru.mvista.com>,
	devicetree-discuss@lists.ozlabs.org,
	linux-kernel@vger.kernel.org, linux-watchdog@vger.kernel.org
Subject: Re: [PATCH] drivers, char: add U-Boot bootcount driver
Date: Sun, 4 Dec 2011 12:47:41 +0100	[thread overview]
Message-ID: <20111204114741.GA5788@pengutronix.de> (raw)
In-Reply-To: <1322991921-21096-1-git-send-email-hs@denx.de>

[-- Attachment #1: Type: text/plain, Size: 2308 bytes --]

On Sun, Dec 04, 2011 at 10:45:21AM +0100, Heiko Schocher wrote:
> This driver implements the Linux kernel half of the boot count feature -
> the boot counter can only be reset after it is clear that the
> application has been started and is running correctly, which usually
> can only be determined by the application code itself. Thus the reset
> of the boot counter must be done by application code, which thus needs
> an appropriate driver.

An appropriate mechanism, not necessarily a driver, see below.

> Required feature by the Carrier Grade Linux Requirements Definition;
> see for example document "Carrier Grade Linux Requirements Definition
> Overview V3.0" at
> 
> http://www.linuxfoundation.org/collaborate/workgroups/cgl/requirements#SMM.6.0_Boot_Cycle_Detection
> 
>       Description: OSDL CGL specifies that carrier grade Linux
>       shall provide support for detecting a repeating reboot cycle
> 	due to recurring failures. This detection should happen in
> 	user space before system services are started.

So, technically, a flag would be enough, not necessarily a counter? Although a
counter probably has more advantages...

> This driver provides read/write access to the U-Boot bootcounter
> through PROC FS and/or sysFS file.

Why ProcFS? Why ProcFS and/or SysFS? Which has priority? Why not /dev?

> The bootcountregister gets configured via DTS.
> for example on the enbw_cmc board:
> 
> bootcount@0x23060 {
>                   compatible = "uboot,bootcount";

No. I assume you are not the vendor of what is at 0x23060, the actual device.
Only the device must be encoded in the compatible-entry which then implies the
bootcount functionality. Also, keep in mind that your solution should be
generic for bootloaders.

>                   reg = <0x23060 0x20>;

I assume that non-volatile memory would qualify as a boot-counter, so those
could be tied to I2C busses etc? reg would not fit then.

I do wonder if it makes more sense to add such functionality to the
watchdog-core to save the additional device (CCed). Needs a second thought,
though...

Regards,

   Wolfram

-- 
Pengutronix e.K.                           | Wolfram Sang                |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]