Re: [PATCH v9 01/15] modpost: fix removing numeric suffixes

[Alexander Lobakin]

From: Borislav Petkov <bp@xxxxxxxxx>
Date: Thu, 23 Dec 2021 17:19:06 +0100

> On Thu, Dec 23, 2021 at 01:21:55AM +0100, Alexander Lobakin wrote:
> > For now, that condition from remove_dot():
> > 
> > if (m && (s[n + m] == '.' || s[n + m] == 0))
> > 
> > which was designed to test if it's a dot or a \0 after the suffix
> > is never satisfied.
> > This is due to that s[n + m] always points to the last digit of a
> > numeric suffix, not on the symbol next to it:
> > 
> > param_set_uint.0, s[n + m] is '0', s[n + m + 1] is '\0'
> > 
> > So it's off by one and was like that since 2014.
> 
> What's the relevance of this? Looking at
> 
>   7d02b490e93c ("Kbuild, lto: Drop .number postfixes in modpost")
> 
> what you're fixing here is something LTO-related. How do you trigger
> this?

It's just a couple lines below. I trigger this using `-z uniq-symbol`
which uses numeric suffixes for globals as well.

> 
> For a Cc:stable patch, I'm missing a lot of context.

It fixes a commit dated 2014, thus Cc:stable. Although the
remove_dot() might've been introduced for neverlanded GCC LTO, but
in fact numeric suffixes are used a lot by the toolchains in regular
builds as well. Just not for globals, that's why it's "well hidden".

> 
> > `-z uniq-symbol` linker flag which we are planning to use to
> 				     ^^
> 
> Who's "we"?

I thought it's a common saying in commit messages, isn't it?

> 
> > simplify livepatching brings numeric suffixes back, fix this.
> > Otherwise:
> > 
> > ERROR: modpost: "param_set_uint.0" [vmlinux] is a static EXPORT_SYMBOL
> > 
> > Fixes: fcd38ed0ff26 ("scripts: modpost: fix compilation warning")
> > Cc: stable@xxxxxxxxxxxxxxx # 3.17+
> > Signed-off-by: Alexander Lobakin <alexandr.lobakin@xxxxxxxxx>
> 
> ...
> 
> -- 
> Regards/Gruss,
>     Boris.
> 
> https://people.kernel.org/tglx/notes-about-netiquette

Thanks,
Al