Re: [PATCH v2 12/45] test: Support testing malloc() failures

[Sean Anderson <seanga2@gmail.com>]

On 9/28/22 17:06, Simon Glass wrote:
> Hi Sean,
> 
> On Wed, 28 Sept 2022 at 11:18, Sean Anderson <seanga2@gmail.com> wrote:
>>
>> On 9/6/22 22:27, Simon Glass wrote:
>>> It is helpful to test that out-of-memory checks work correctly in code
>>> that calls malloc().
>>>
>>> Add a simple way to force failure after a given number of malloc() calls.
>>>
>>> Fix a header guard to avoid a build error on sandbox_vpl.
>>>
>>> Signed-off-by: Simon Glass <sjg@chromium.org>
>>> ---
>>>
>>> (no changes since v1)
>>>
>>>    arch/sandbox/include/asm/malloc.h |  1 +
>>>    common/dlmalloc.c                 | 19 +++++++++++++++++++
>>>    include/malloc.h                  | 12 ++++++++++++
>>>    test/test-main.c                  |  1 +
>>>    4 files changed, 33 insertions(+)
>>>
>>> diff --git a/arch/sandbox/include/asm/malloc.h b/arch/sandbox/include/asm/malloc.h
>>> index a1467b5eadd..8aaaa9cb87b 100644
>>> --- a/arch/sandbox/include/asm/malloc.h
>>> +++ b/arch/sandbox/include/asm/malloc.h
>>> @@ -6,6 +6,7 @@
>>>     */
>>>
>>>    #ifndef __ASM_MALLOC_H
>>> +#define __ASM_MALLOC_H
>>>
>>>    void *malloc(size_t size);
>>>    void free(void *ptr);
>>> diff --git a/common/dlmalloc.c b/common/dlmalloc.c
>>> index f48cd2a333d..41c7230424c 100644
>>> --- a/common/dlmalloc.c
>>> +++ b/common/dlmalloc.c
>>> @@ -596,6 +596,9 @@ ulong mem_malloc_start = 0;
>>>    ulong mem_malloc_end = 0;
>>>    ulong mem_malloc_brk = 0;
>>>
>>> +static bool malloc_testing;  /* enable test mode */
>>> +static int malloc_max_allocs;        /* return NULL after this many calls to malloc() */
>>> +
>>>    void *sbrk(ptrdiff_t increment)
>>>    {
>>>        ulong old = mem_malloc_brk;
>>> @@ -1307,6 +1310,11 @@ Void_t* mALLOc(bytes) size_t bytes;
>>>                return malloc_simple(bytes);
>>>    #endif
>>>
>>> +  if (CONFIG_IS_ENABLED(UNIT_TEST) && malloc_testing) {
>>> +    if (--malloc_max_allocs < 0)
>>> +      return NULL;
>>> +  }
>>> +
>>>      /* check if mem_malloc_init() was run */
>>>      if ((mem_malloc_start == 0) && (mem_malloc_end == 0)) {
>>>        /* not initialized yet */
>>> @@ -2470,6 +2478,17 @@ int initf_malloc(void)
>>>        return 0;
>>>    }
>>>
>>> +void malloc_enable_testing(int max_allocs)
>>> +{
>>> +     malloc_testing = true;
>>> +     malloc_max_allocs = max_allocs;
>>> +}
>>> +
>>> +void malloc_disable_testing(void)
>>> +{
>>> +     malloc_testing = false;
>>> +}
>>> +
>>>    /*
>>>
>>>    History:
>>> diff --git a/include/malloc.h b/include/malloc.h
>>> index e8c8b254c0d..161ccbd1298 100644
>>> --- a/include/malloc.h
>>> +++ b/include/malloc.h
>>> @@ -883,6 +883,18 @@ extern Void_t*     sbrk();
>>>
>>>    void malloc_simple_info(void);
>>>
>>> +/**
>>> + * malloc_enable_testing() - Put malloc() into test mode
>>> + *
>>> + * This only works if UNIT_TESTING is enabled
>>> + *
>>> + * @max_allocs: return -ENOMEM after max_allocs calls to malloc()
>>> + */
>>> +void malloc_enable_testing(int max_allocs);
>>> +
>>> +/** malloc_disable_testing() - Put malloc() into normal mode */
>>> +void malloc_disable_testing(void);
>>> +
>>>    #if CONFIG_IS_ENABLED(SYS_MALLOC_SIMPLE)
>>>    #define malloc malloc_simple
>>>    #define realloc realloc_simple
>>> diff --git a/test/test-main.c b/test/test-main.c
>>> index c65cbd7c351..5b6b5ba5bbe 100644
>>> --- a/test/test-main.c
>>> +++ b/test/test-main.c
>>> @@ -46,6 +46,7 @@ static int dm_test_pre_run(struct unit_test_state *uts)
>>>        uts->force_fail_alloc = false;
>>>        uts->skip_post_probe = false;
>>>        gd->dm_root = NULL;
>>> +     malloc_disable_testing();
>>>        if (CONFIG_IS_ENABLED(UT_DM) && !CONFIG_IS_ENABLED(OF_PLATDATA))
>>>                memset(dm_testdrv_op_count, '\0', sizeof(dm_testdrv_op_count));
>>>        arch_reset_for_test();
>>
>> Reviewed-by: Sean Anderson <seanga2@gmail.com>
>>
>> but do we need to instrument rEALLOc too?
> 
> We could, but I'd prefer to have a test which needs it first. Actually
> realloc() calls malloc() in some cases, so we are already messing with
> things here.

Maybe it would be better to do this at the "top-level" instead. E.g.
redefine the malloc() et al macro. That way we remove any double-counting.

--Sean