Re: [PATCH 3/4] perf list: Don't write to const memory

[David Laight <david.laight.linux@gmail.com>]

On Wed, 21 Jan 2026 22:10:18 -0300
Arnaldo Carvalho de Melo <acme@kernel.org> wrote:

> On Wed, Jan 21, 2026 at 10:17:13PM +0000, David Laight wrote:
> > On Tue, 20 Jan 2026 19:08:59 -0300
> > Arnaldo Carvalho de Melo <acme@kernel.org> wrote:
> >   
> > > From: Arnaldo Carvalho de Melo <acme@redhat.com>
> > > 
> > > Something now detected on fedora 44, where strchr() returns const if it
> > > is passed a const pointer:
> > > 
> > >   util/print-events.c: In function 'print_sdt_events':
> > >   util/print-events.c:89:29: error: initialization discards 'const' qualifier from pointer target type [-Werror=discarded-qualifiers]
> > >      89 |                 char *bid = strchr(sdt_name->s, '@');
> > >         |                             ^~~~~~
> > > 
> > > Fix it by using strnchr() if strchr finds the separator instead of
> > > temporarily scrubbing it with '\0'.  
> > 
> > I've just looked at the full function to see WTF it is doing.
> > You've fixed the second strchr() not the one the compiler bleated about.  
> 
> It complained about both, no?
> 
> > Line 89 is followed by:
> > 	if (bid)
> > 		*bid++ = 0;
> > but if it is NULL there is a fair chance the code will just explode a bit later on.  
> 
> Why, can you elaborate?

I don't see why it can't end up in:
		if (show_detail) {
			char *path = build_id_cache__origname(bid);
with 'bid == NULL' and the function just dereferences it.

> 
> > I suspect it is an error if the strings aren't "name@bar"  
>  
> > 'show_detail' seems to be set if either the previous or next entries in the
> > list/tree have the same "name" - which seems strange to me.  
> 
> All this is strange, a corner case, it looks as if we may have multiple
> binaries, differentiated by buildids, that would have different SDTs.
> 
> The code seems to be trying to reduce the amount of info (not showing
> the buildid if there are no more at that point multiple versions of a
> DSO with SDTs in place).
> 
> Further analysis of the changesets that put all this in place is needed
> to clarify, but looking just at fixing what the compiler is complaining,
> do you think the patches are bad?

I didn't see a patch to fix the first one.
But can you just change the structure so that 's' is 'char *'?

	David

> 
> - Arnaldo
>  
> > The while thing needs more work :-(
> > 
> > 	David
> >   
> > > 
> > > Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
> > > ---
> > >  tools/perf/util/print-events.c | 13 +++++--------
> > >  1 file changed, 5 insertions(+), 8 deletions(-)
> > > 
> > > diff --git a/tools/perf/util/print-events.c b/tools/perf/util/print-events.c
> > > index 8f3ed83853a9e468..898cf426509790cd 100644
> > > --- a/tools/perf/util/print-events.c
> > > +++ b/tools/perf/util/print-events.c
> > > @@ -97,14 +97,11 @@ void print_sdt_events(const struct print_callbacks *print_cb, void *print_state)
> > >  		} else {
> > >  			next_sdt_name = strlist__next(sdt_name);
> > >  			if (next_sdt_name) {
> > > -				char *bid2 = strchr(next_sdt_name->s, '@');
> > > -
> > > -				if (bid2)
> > > -					*bid2 = '\0';
> > > -				if (strcmp(sdt_name->s, next_sdt_name->s) == 0)
> > > -					show_detail = true;
> > > -				if (bid2)
> > > -					*bid2 = '@';
> > > +				const char *bid2 = strchr(next_sdt_name->s, '@');
> > > +
> > > +				show_detail = bid2 ?
> > > +						strncmp(sdt_name->s, next_sdt_name->s, bid2 - next_sdt_name->s) == 0 :
> > > +						strcmp(sdt_name->s, next_sdt_name->s) == 0;
> > >  			}
> > >  		}
> > >  		last_sdt_name = sdt_name->s;