Re: [PATCH 2/6] usb:gadget:mass-storage: refactoring the SCSI command handling

From: Alan Stern
Date: Sun Jun 27 2021 - 21:06:47 EST


On Sun, Jun 27, 2021 at 08:14:48PM +0300, i.kononenko wrote:
>
>
> On 27.06.2021 17:23, Alan Stern wrote:
> > On Sun, Jun 27, 2021 at 12:18:15AM +0300, Igor Kononenko wrote:
> >> Implements a universal way to define SCSI commands and configure
> >> precheck handlers.
> >
> > What is the reason for doing this?
>
> I have started implementing a way to specify a backend-file of
> mass-storage images greater than 2.1Gb for cdrom-like mediums.
> I notice the implementation of each scsi-command handler uses too
> many magic-constant, hardcoded indexes and shifts. I decided to
> define structures that contained appropriate SCSI-defined fields
> and constant-values to clarify the code.
>
> Additionally, I noticed, many kernel subsystems use the 'separate
> data and logic' approach, making a code more explicit and readable.
> This looks reasonable to me, and a code looks more clearly, at
> least - we don't need to examine each magic constant and its purpose.
>
> >
> > At first glance, it appears you have added a great deal of complexity
> > to the driver. The patch replaces a large amount of easily understood
> > (albeit rather repetitious) code with an approximately equal amount
> > of rather complicated code. This does not seem like an improvement.
>
> The SCSI-commands table is defined as unifying a way to specify the
> SCSI-command handler, with corresponding required data instead pass
> it to each repeatedly switch-case block, which makes code more readable
> to me. If there isn't, I can keep the definition of SCSI-handlers as is,
> but the SCSI-data structures with their constant-values are still
> required, in my opinion.
>
> >
> > Furthermore, the code you removed is flexible; it easily allows for
> > small variations as neede by some command handlers. But the code you
> > added is all table-driven, which does not easily permit arbitrary
> > variations.
> >
>
> I don't think that the SCSI-command handlers table is an obstacle to
> define variation into a specific handler because the current patch has
> helper macros, which can specify a behavior for each requirement of
> handler.
>
> Anyway, the definition of the scsi-command handlers table may be discarded,
> because this work done to helping developers who will work the
> 'usb:gadget:mass-storage' subsystem in the future.

Can you submit a patch that adds only the data structures without the
commands table?

Alan Stern