Re: [PATCH 4.19 119/131] tracing: Fix event trigger to accept redundant spaces

[Sasha Levin]

On Fri, Jul 03, 2020 at 09:21:03PM +0200, Pavel Machek wrote:
> > > > commit 6784beada631800f2c5afd567e5628c843362cee upstream.
> > > >
> > > > Fix the event trigger to accept redundant spaces in
> > > > the trigger input.
> > > >
> > > > For example, these return -EINVAL
> > > >
> > > > echo " traceon" > events/ftrace/print/trigger
> > > > echo "traceon  if common_pid == 0" > events/ftrace/print/trigger
> > > > echo "disable_event:kmem:kmalloc " > events/ftrace/print/trigger
> > > >
> > > > But these are hard to find what is wrong.
> > > >
> > > > To fix this issue, use skip_spaces() to remove spaces
> > > > in front of actual tokens, and set NULL if there is no
> > > > token.
> > >
> > > For the record, I'm not fan of this one. It is ABI change, not a
> > > bugfix.
> > >
> > > Yes, it makes kernel interface "easier to use". It also changes
> > > interface in the middle of stable series, and if people start relying
> > > on new interface and start putting extra spaces, they'll get nasty
> > > surprise when they move code to the older kernel.
> >
> > If an interface changes anywhere that breaks userspace, it needs to be
> > not done, stable kernels are not an issue here or not.
>
> I'm not saying it is a regression; I'd scream way more if that was the
> case. I'm saying it is nowhere near a fix.
>
> We really don't want userspace doing:
>
> > > > echo " traceon" > events/ftrace/print/trigger
>
> Because it does not work on older kernels. It will work on 4.19.131
> and break on 5.6.19.

Moving to an EOL kernel is very much unsupported.

--
Thanks,
Sasha