Re: [RFC V1 0/8] Add a driver for somagic smi2021

[Ezequiel Garcia <ezequiel.garcia@free-electrons.com>]

Hi Jon,

On Sun, Mar 17, 2013 at 09:01:58PM +0100, Jon Arne Jørgensen wrote:
> On Fri, Mar 15, 2013 at 09:08:58AM -0300, Ezequiel Garcia wrote:
> > On Thu, Mar 14, 2013 at 03:06:56PM +0100, Jon Arne Jørgensen wrote:
> > > This patch-set will add a driver for the Somagic SMI2021 chip.
> > > 
> > > This chip is found inside different usb video-capture devices.
> > > Most of them are branded as EasyCap, but there also seems to be
> > > some other brands selling devices with this chip.
> > > 
> > > This driver is split into two modules, where one is called smi2021-bootloader,
> > > and the other is just called smi2021.
> > > 
> > > The bootloader is responsible for the upload of a firmware that is needed by some
> > > versions of the devices.
> > > 
> > > All Somagic devices that need firmware seems to identify themselves
> > > with the usb product id 0x0007. There is no way for the kernel to know
> > > what firmware to upload to the device without user interaction.
> > > 
> > > If there is only one firmware present on the computer, the kernel
> > > will upload that firmware to any device that identifies as 0x0007.
> > > If there are multiple Somagic firmwares present, the user will have to pass
> > > a module parameter to the smi2021-bootloader module to tell what firmware to use.
> > > 
> > 
> > Nice job!
> >
> Thanks :)
>  
> > I have some minor comments on each patch, but also I don't agree
> > with the patch splitting: what's the point in splitting and sending
> > one patch per file?
> > 
> > It doesn't make it any easier to review, so why don't you just
> > send one patch: "Introduce smi2021 driver"?
> > 
> > The rule is one patch per change, and I believe this whole patchset
> > is just one change: adding a new driver.
> > 
> 
> I think I read another patch to this mailinglist, where someone was told
> to split his patch into one mail per file, but I can't find that thread
> now :)
> 
> I will send the next version as a single mail, and see what happens...
> 

As you will soon realize, the patch preparation is equally important as the
patch content itself. Often, it takes the same time to implement or
fix something, as it takes to prepare the patchset carefully.

When deciding how to prepare your patches you have to keep two main things in mind:

  * The kernel build can never be broken, in any configuration and in any
    point of the kernel history (in other words, by any patch of a patchset).
    This is called 'keep the bisectability' because it's essential
    to make 'git bisect' work properly.

  * Do as much as possible to facilitate reviews from other people.
    This is also important because patches tend to be accepted quicker
    if they recieve attention (reviews, testing, etc.).

In this particular case, I think that the easier way to review is to be
able to see the complete driver in a single patch. Of course, I can be
wrong, so feel free to correct me.

Please note that the reviews I made where almost nitpicks, and the
driver looks good in general. I cannot provide any testing for lack of hardware.

Also, for your next patch, add the output of v4l2-compliance tool,
showing it passes all the tests. This shows your driver is in good shape.
Get v4l2-compliancefrom the git repo, as distribution often provide
an outdated version.

(And another thing, please fix your mail client: the reply-to is pointing
to '20130315120856.GA2989@localhost.arpanet.local'.)

Thanks for your work and best regards,
-- 
Ezequiel García, Free Electrons
Embedded Linux, Kernel and Android Engineering
http://free-electrons.com