[bug report] lib: introduce test_meminit module

[bug report] lib: introduce test_meminit module

[Dan Carpenter]

Hello Alexander Potapenko,

Commit 5015a300a522 ("lib: introduce test_meminit module") from Jul
16, 2019 (linux-next), leads to the following Smatch static checker
warning:

	lib/test_meminit.c:390 test_kmemcache()
	warn: bool mask is always false 'ctor & zero'

lib/test_meminit.c
    378 static int __init test_kmemcache(int *total_failures)
    379 {
    380         int failures = 0, num_tests = 0;
    381         int i, flags, size;
    382         bool ctor, rcu, zero;
    383 
    384         for (i = 0; i < 10; i++) {
    385                 size = 8 << i;
    386                 for (flags = 0; flags < 8; flags++) {
    387                         ctor = flags & 1;
    388                         rcu = flags & 2;
    389                         zero = flags & 4;
--> 390                         if (ctor & zero)
                                    ^^^^^^^^^^^
This is like (BIT(1) & BIT(4)) so it can't be true.  Was && intended?

    391                                 continue;
    392                         num_tests += do_kmem_cache_size(size, ctor, rcu, zero,
    393                                                         &failures);
    394                 }
    395                 num_tests += do_kmem_cache_size_bulk(size, &failures);
    396         }
    397         REPORT_FAILURES_IN_FN();
    398         *total_failures += failures;
    399         return num_tests;
    400 }

This email is a free service from the Smatch-CI project [smatch.sf.net].

regards,
dan carpenter

Re: [bug report] lib: introduce test_meminit module

[Alexander Potapenko]

On Thu, Apr 30, 2026 at 8:15 PM Dan Carpenter <error27@gmail.com> wrote:
>
> Hello Alexander Potapenko,
>
> Commit 5015a300a522 ("lib: introduce test_meminit module") from Jul
> 16, 2019 (linux-next), leads to the following Smatch static checker
> warning:
>
>         lib/test_meminit.c:390 test_kmemcache()
>         warn: bool mask is always false 'ctor & zero'
>
> lib/test_meminit.c
>     378 static int __init test_kmemcache(int *total_failures)
>     379 {
>     380         int failures = 0, num_tests = 0;
>     381         int i, flags, size;
>     382         bool ctor, rcu, zero;
>     383
>     384         for (i = 0; i < 10; i++) {
>     385                 size = 8 << i;
>     386                 for (flags = 0; flags < 8; flags++) {
>     387                         ctor = flags & 1;
>     388                         rcu = flags & 2;
>     389                         zero = flags & 4;
> --> 390                         if (ctor & zero)
>                                     ^^^^^^^^^^^
> This is like (BIT(1) & BIT(4)) so it can't be true.  Was && intended?

Whoa, thanks!
I wonder why Clang never complained about that.

Re: [bug report] lib: introduce test_meminit module

[Alexander Potapenko]

On Mon, May 4, 2026 at 9:36 AM Alexander Potapenko <glider@google.com> wrote:
>
> On Thu, Apr 30, 2026 at 8:15 PM Dan Carpenter <error27@gmail.com> wrote:
> >
> > Hello Alexander Potapenko,
> >
> > Commit 5015a300a522 ("lib: introduce test_meminit module") from Jul
> > 16, 2019 (linux-next), leads to the following Smatch static checker
> > warning:
> >
> >         lib/test_meminit.c:390 test_kmemcache()
> >         warn: bool mask is always false 'ctor & zero'
> >
> > lib/test_meminit.c
> >     378 static int __init test_kmemcache(int *total_failures)
> >     379 {
> >     380         int failures = 0, num_tests = 0;
> >     381         int i, flags, size;
> >     382         bool ctor, rcu, zero;
> >     383
> >     384         for (i = 0; i < 10; i++) {
> >     385                 size = 8 << i;
> >     386                 for (flags = 0; flags < 8; flags++) {
> >     387                         ctor = flags & 1;
> >     388                         rcu = flags & 2;
> >     389                         zero = flags & 4;
> > --> 390                         if (ctor & zero)
> >                                     ^^^^^^^^^^^
> > This is like (BIT(1) & BIT(4)) so it can't be true.  Was && intended?

Well, in fact I think the "(BIT(1) & BIT(4))" part is incorrect.
Both `ctor` and `zero` are _Bool, so their values are converted to 0/1
before `ctor & zero` is calculated.
So while the code is smelly (because it uses `&` on a bool type) and
needs fixing, the warning generated by Smatch is also misleading.