On Wed, Nov 16, 2022 at 10:41:39PM -0800, Song Liu wrote:
> On Wed, Nov 16, 2022 at 5:49 PM Luis Chamberlain <mcgrof@xxxxxxxxxx> wrote:
> >
> > On Wed, Nov 16, 2022 at 05:06:18PM -0800, Song Liu wrote:
> > > Add logic to test execmem_[alloc|fill|free] in test_vmalloc.c.
> > > No need to change tools/testing/selftests/vm/test_vmalloc.sh.
> > >
> > > Signed-off-by: Song Liu <song@xxxxxxxxxx>
> > > ---
> > > lib/test_vmalloc.c | 30 ++++++++++++++++++++++++++++++
> > > 1 file changed, 30 insertions(+)
> > >
> > > diff --git a/lib/test_vmalloc.c b/lib/test_vmalloc.c
> > > index cf7780572f5b..6591c4932c3c 100644
> > > --- a/lib/test_vmalloc.c
> > > +++ b/lib/test_vmalloc.c
> > > @@ -50,6 +50,7 @@ __param(int, run_test_mask, INT_MAX,
> > > "\t\tid: 128, name: pcpu_alloc_test\n"
> > > "\t\tid: 256, name: kvfree_rcu_1_arg_vmalloc_test\n"
> > > "\t\tid: 512, name: kvfree_rcu_2_arg_vmalloc_test\n"
> > > + "\t\tid: 1024, name: execmem_alloc_test\n"
> > > /* Add a new test case description here. */
> > > );
> > >
> > > @@ -352,6 +353,34 @@ kvfree_rcu_2_arg_vmalloc_test(void)
> > > return 0;
> > > }
> > >
> > > +static int
> > > +execmem_alloc_test(void)
> > > +{
> > > + void *p, *tmp;
> > > + int i;
> > > +
> > > + for (i = 0; i < test_loop_count; i++) {
> > > + /* allocate variable size, up to 64kB */
> > > + size_t size = (i % 1024 + 1) * 64;
> > > +
> > > + p = execmem_alloc(size, 64);
> > > + if (!p)
> > > + return -1;
> > > +
> > > + tmp = execmem_fill(p, "a", 1);
> > > + if (tmp != p)
> > > + return -1;
> > > +
> > > + tmp = execmem_fill(p + size - 1, "b", 1);
> > > + if (tmp != p + size - 1)
> > > + return -1;
> > > +
> > > + execmem_free(p);
> > > + }
> > > +
> > > + return 0;
> > > +}
> > > +
> >
> > This is a basic test and it is useful.
> >
> > But given all those WARN_ON() and WARN_ON_ONCE() I think the real value
> > test here would be to race 1000 threads doing this at the same time.
>
> test_vmalloc supports parallel tests. We can do something like
>
> tools/testing/selftests/vm/test_vmalloc.sh nr_threads=XXX run_test_mask=1024
Nice, if that is not run by default we won't capture issues which may
arise on selftests on 0day.
Luis
