Re: [PATCH 04/11] regmap: kunit: Run sparse cache tests at non-zero register addresses

[Richard Fitzgerald]

On 16/05/2024 20:53, Guenter Roeck wrote:
> Hi,
>
> On Mon, Apr 08, 2024 at 03:45:53PM +0100, Richard Fitzgerald wrote:
> > Run the cache_drop() and cache_present() tests at blocks of addresses
> > that don't start at zero.
> >
> > This adds a from_reg parameter to struct regmap_test_param. This is
> > used to set the base address of the register defaults created by
> > gen_regmap().
> >
> > Extra entries are added to sparse_cache_types_list[] to test at non-zero
> > from_reg values. The cache_drop() and cache_present() tests are updated
> > to test at the given offset.
> >
> > The aim here is to add test cases to cache_drop() for the bug fixed by
> > commit 00bb549d7d63 ("regmap: maple: Fix cache corruption in
> > regcache_maple_drop()")
> >
> > But the same parameter table is used by the cache_present() test so
> > let's also update that to use from_reg.
> >
> > Signed-off-by: Richard Fitzgerald <rf@xxxxxxxxxxxxxxxxxxxxx>
>
> With this patch in mainline, I get lots of errors such as
>
> [   23.494308] =============================================================================
> [   23.496391] BUG kmalloc-64 (Tainted: G                 N): kmalloc Redzone overwritten

Just a random thought, I wonder whether in gen_regmap() this:

  if (config->num_reg_defaults)
	config->max_register += (config->num_reg_defaults - 1) *
				config->reg_stride;
  else
	config->max_register += (BLOCK_TEST_SIZE * config->reg_stride);

should be:

config->max_register += max(config->num_reg_defaults - 1,
			    BLOCK_TEST_SIZE) * config->reg_stride;

I've only had a quick scan through the code without seeing any other
obvious problem.