linux-kernel - Re: [PATCH] Declare the file

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAPGkw+zhb=zC8fKyK8uUiNxNqCUR9Trg6N7fBjaAcBc9rmkmUg@mail.gmail.com>
Date:   Wed, 1 Sep 2021 20:04:35 +0200
From:   Krish Jain <krishjain02939@...il.com>
To:     Bryan Brattlof <hello@...anbrattlof.com>
Cc:     Greg KH <gregkh@...uxfoundation.org>,
        "Fabio M. De Francesco" <fmdefrancesco@...il.com>,
        linux-staging@...ts.linux.dev, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] Declare the file_operations struct as const

On Wed, Sep 1, 2021 at 7:34 PM Bryan Brattlof <hello@...anbrattlof.com> wrote:
>
> On this day, September  1, 2021, thus sayeth Greg KH:
> > On Wed, Sep 01, 2021 at 05:34:36PM +0200, Krish Jain wrote:
> > >
> > > Can you tell me why this is the case?
> >
> > Again, it depends on your kernel configuration file as to what will, or
> > will not, be built.
> >
> > If you have some things set as modules, they can be built as a module,
> > but the ashmem code can not be built as a module, so you would never
> > build it if you did the above line.
> >
> > Here, look at this sequence, starting with a tree that does nothing if I
> > do a simple 'make' in it, as the whole kernel is already built, and
> > ashmem is enabled in the kernel configuration
> >
> > $ grep ASHMEM .config
> > CONFIG_ASHMEM=y
> > $ make
> > $
> >
> > So, let's change the time stamp on the ashmem.c file and see what gets
> > built if you use the M= option:
> >
> > $ touch drivers/staging/android/ashmem.c
> > $ make M=drivers/staging/android
> >   MODPOST drivers/staging/android/Module.symvers
> > $
> >
> > Nothing gets built as ashmem is NOT a module, and M= only builds any
> > modules in the directory you specified.
> >
> > But, if you tell make to just build the whole subdirectory, no matter
> > what the setting is, it will be built:
> >
> > $ make drivers/staging/android/
> >   CALL    scripts/checksyscalls.sh
> >   CALL    scripts/atomic/check-atomics.sh
> >   DESCEND objtool
> >   CC      drivers/staging/android/ashmem.o
> >   AR      drivers/staging/android/built-in.a
> > $
> >
> > So that's the difference, "M=" builds modules in that directory, but if
> > you tell it to build the subdir, everything in there that needs to be
> > built, will be built.
> >
> > Be careful about your kernel configuration, that is the key for what
> > will, and will not, be built.
> >
>
> Ouch...
>
> I want to *really* apologize to you Krish for introducing so much
> confusion while you, and apparently I, am still learning. And for your
> persistence with seeking the correct answer here Krish.
>
> I did not notice that this could only be build as a built-in object.
> Thank you Greg for pointing out my mistake, and I apologize for dragging
> this out longer than it had to and the frustration this caused.
>
> It seems I will be reading the documentation again, along with Greg's
> book recommendation, "Linux Kernel in a Nutshell" over this merge
> window.
>
> Thank you again Krish and Greg
> ~Bryan
>

Yes, lots of reading to do :) . I had a look at the book and it seems
better than the documentation too, I don't know, maybe the writing
style? Love it, Greg. Lastly just out of curiosity, Bryan,  if this
can only be built as a built-in object then how come "As for your
patch, I built the driver using:

  $ make CCFLAGS=-Werror W=1 M=drivers/staging/android"

got you the errors that I desired? Aren't you building as a module here?


Warm Regards