Re: [PATCH RFC] gcc-plugins: Handle GCC version mismatch for OOT modules

[Greg KH]

On Tue, Jan 26, 2021 at 08:51:55AM -0600, Josh Poimboeuf wrote:
> On Tue, Jan 26, 2021 at 02:51:29PM +0100, Greg KH wrote:
> > On Tue, Jan 26, 2021 at 06:44:44AM -0600, Justin Forbes wrote:
> > > On Tue, Jan 26, 2021 at 2:21 AM Greg KH <greg@xxxxxxxxx> wrote:
> > > >
> > > > On Mon, Jan 25, 2021 at 04:07:57PM -0600, Josh Poimboeuf wrote:
> > > > > On Tue, Jan 26, 2021 at 06:44:35AM +0900, Masahiro Yamada wrote:
> > > > > > > > If people use a different compiler, they must be
> > > > > > > > prepared for any possible problem.
> > > > > > > >
> > > > > > > > Using different compiler flags for in-tree and out-of-tree
> > > > > > > > is even more dangerous.
> > > > > > > >
> > > > > > > > For example, CONFIG_GCC_PLUGIN_RANDSTRUCT is enabled
> > > > > > > > for in-tree build, and then disabled for out-of-tree modules,
> > > > > > > > the struct layout will mismatch, won't it?
> > > > > > >
> > > > > > > If you read the patch you'll notice that it handles that case, when it's
> > > > > > > caused by GCC mismatch.
> > > > > > >
> > > > > > > However, as alluded to in the [1] footnote, it doesn't handle the case
> > > > > > > where the OOT build system doesn't have gcc-plugin-devel installed.
> > > > > > > Then CONFIG_GCC_PLUGIN_RANDSTRUCT gets silently disabled and the build
> > > > > > > succeeds!  That happens even without a GCC mismatch.
> > > > > >
> > > > > >
> > > > > > Ah, sorry.
> > > > > >
> > > > > > I responded too early before reading the patch fully.
> > > > > >
> > > > > > But, I do not like to make RANDSTRUCT a special case.
> > > > > >
> > > > > > I'd rather want to stop building for any plugin.
> > > > >
> > > > > Other than RANDSTRUCT there doesn't seem to be any problem with
> > > > > disabling them (and printing a warning) in the OOT build.  Why not give
> > > > > users that option?  It's harmless, and will make distro's (and their
> > > > > users') lives easier.
> > > > >
> > > > > Either GCC mismatch is ok, or it's not.  Let's not half-enforce it.
> > > >
> > > > As I said earlier, it's not ok, we can not support it at all.
> > > >
> > > 
> > > Support and enforce are 2 completely different things.  To shed a bit
> > > more light on this, the real issue that prompted this was breaking CI
> > > systems.  As we enabled gcc plugins in Fedora, and the toolchain folks
> > > went through 3 different snapshots of gcc 11 in a week. Any CI process
> > > that built an out of tree module failed. I don't think this is nearly
> > > as much of a concern for stable distros, as it is for CI in
> > > development cycles.
> > 
> > It's better to have an obvious break like this than to silently accept
> > things and then have a much harder issue to debug at runtime, right?
> 
> User space mixes compiler versions all the time.  The C ABI is stable.
> 
> What specifically is the harder issue you're referring to?

Have you not noticed include/linux/compiler.h and all of the different
changes/workarounds we do for different versions of gcc/clang/intel
compilers?  We have never guaranteed that a kernel module would work
that was built with a different compiler than the main kernel, and I
doubt we can start now.

thanks,

greg k-h