Re: [PATCH v7 1/4] tracing: add __print_sym() to replace __print_symbolic()

[Johannes Berg <johannes@sipsolutions.net>]

On Wed, 2026-09-23 at 11:10 +0000, sashiko-bot@kernel.org wrote:
> > +#undef __print_sym
> > +#define __print_sym(value, symbol_id)	__print_sym(value:symbol_id)
> 
> [Severity: Medium]
> Does this delimiter choice in the __print_sym() macro in
> include/trace/stages/stage7_class_define.h cause issues for arbitrary C
> expressions?
> 
> This macro transforms the arguments into a string format using a delimiter (:)
> that is unsafe for arbitrary C expressions if parsed left-to-right. If a
> kernel developer uses an expression containing a colon (such as a ternary
> operator like a ? b : c) as the value argument, it seems the string
> representation will contain multiple colons.

[snip]


Technically, yes, but I think it's implausible to have a ternary for
printing a simple enum value *inside* the TP_printk(). You'd usually
want to store the result of the ternary via TP_fast_assign(), I'd think.

> Could a backward search for ':' from the closing parenthesis safely isolate
> the symbol ID?

Technically? But it'd be relatively more complicated, the code would
have to pre-parse to the end and we have seq_put() here etc.

I'm not convinced it's worth doing.

johannes