Re: [RFC PATCH] kbuild: check uniqueness of basename of modules

[Masahiro Yamada]

On Wed, May 15, 2019 at 8:34 PM Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>
> On Wed, May 15, 2019 at 05:57:50PM +0900, Masahiro Yamada wrote:
> > On Wed, May 15, 2019 at 5:14 PM Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> > >
> > > On Wed, May 15, 2019 at 10:08:12AM +0200, Arnd Bergmann wrote:
> > > > On Wed, May 15, 2019 at 9:39 AM Masahiro Yamada
> > > > <yamada.masahiro@xxxxxxxxxxxxx> wrote:
> > > > >
> > > > > In the recent build test of linux-next, Stephen saw a build error
> > > > > caused by a broken .tmp_versions/*.mod file:
> > > > >
> > > > >   https://lkml.org/lkml/2019/5/13/991
> > > > >
> > > > > drivers/net/phy/asix.ko and drivers/net/usb/asix.ko have the same
> > > > > basename, and there is a race in generating .tmp_versions/asix.mod
> > > > >
> > > > > Kbuild has not checked this before, and it occasionally shows up with
> > > > > obscure error message when this kind of race occurs.
> > > > >
> > > > > It is not trivial to catch this potential issue by eyes.
> > > > >
> > > > > Hence, this script.
> > > > >
> > > > > I compile-tested allmodconfig for the latest kernel as of writing,
> > > > > it detected the following:
> > > > >
> > > > > warning: same basename '88pm800.ko' if the following are built as modules:
> > > > >   drivers/regulator/88pm800.ko
> > > > >   drivers/mfd/88pm800.ko
> > > > > warning: same basename 'adv7511.ko' if the following are built as modules:
> > > > >   drivers/gpu/drm/bridge/adv7511/adv7511.ko
> > > > >   drivers/media/i2c/adv7511.ko
> > > > > warning: same basename 'asix.ko' if the following are built as modules:
> > > > >   drivers/net/phy/asix.ko
> > > > >   drivers/net/usb/asix.ko
> > > > > warning: same basename 'coda.ko' if the following are built as modules:
> > > > >   fs/coda/coda.ko
> > > > >   drivers/media/platform/coda/coda.ko
> > > > > warning: same basename 'realtek.ko' if the following are built as modules:
> > > > >   drivers/net/phy/realtek.ko
> > > > >   drivers/net/dsa/realtek.ko
> > > > >
> > > > > Reported-by: Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx>
> > > > > Signed-off-by: Masahiro Yamada <yamada.masahiro@xxxxxxxxxxxxx>
> > > >
> > > > That looks great!
> > > >
> > > > > ---
> > > > >
> > > > >  [Alternative fix ? ]
> > > > >
> > > > > I do not know about the user experience of modprobe etc.
> > > > > when two different modules have the same name.
> > > > > It does not matter if this is correctly handled by modules.order?
> > > > >
> > > > > If this is just a problem of the build system, it is pretty easy to fix.
> > > > > For example, if we prepend the directory path, parallel build will
> > > > > never write to the same file simultanously.
> > > > >
> > > > >   asix.mod -> drivers/net/phy/asix.mod
> > > > >   asix.mod -> drivers/net/usb/asix.mod
> > > >
> > > > non-unique module names cause all kinds of problems, and
> > > > modprobe can certainly not handle them correctly, and there
> > > > are issues with symbols exported from a module when another
> > > > one has the same name.
> > >
> > > /sys/modules/ will fall over when this happens as well.  I thought we
> > > had the "rule" that module names had to be unique, I guess it was only a
> > > matter of time before they started colliding :(
> > >
> > > So warning is good, but forbidding this is better, as things will break.
> > >
> > > Or we need to fix up the places where things will break.
> >
> >
> > If we intentionally break the build,
> > we need to send fix-up patches to subsystems first.
>
> True, but those builds are already broken if anyone tries to use them :)
>
> A warning for now would be nice, that way we can find these and know to
> fix them up over time.


Yes, I think it is a fair point.

Start this with warning, then people will soon notice the problem.

Turning it into error is easy once we fix all instances.


-- 
Best Regards
Masahiro Yamada