Re: [PATCH v3 0/8] memblock tests: update and extend memblock simulator

[Huang, Shaoqin]

All patches looks good to me.
Reviewed-by: Shaoqin Huang <shaoqin.huang@xxxxxxxxx>

Thanks your effort for doing this.

On 8/27/2022 1:42 PM, Rebecca Mckeever wrote:
> These patches update existing tests in memblock simulator, add
> additional tests for memblock functions that are already being tested,
> and add test coverage for additional memblock functions.
>
> Updated tests for:
> - memblock_alloc()
> - memblock_alloc_try_nid()
> - memblock_alloc_from()
>
> The updates to memblock_alloc() tests include the addition of an assert
> that checks whether the entire chunk of allocated memory is cleared. For
> memblock_alloc_try_nid() and memblock_alloc_from(), the assert that checks
> whether the allocated memory is cleared now checks the entire chunk of
> allocated memory instead of just the first byte. To make this more robust,
> setup_memblock() and dummy_physical_memory_init() fill the entire MEM_SIZE
> simulated physical memory with nonzero values by calling fill_memblock().
> setup_memblock() is called at the beginning of most tests for
> memblock_alloc() functions.
>
> Additional tests for:
> - memblock_add()
> - memblock_reserve()
> - memblock_remove()
> - memblock_free()
> - memblock_alloc()
>
> Introducing test coverage for:
> - memblock_alloc_raw()
> - memblock_alloc_try_nid_raw()
> - memblock_set_bottom_up()
> - memblock_bottom_up()
> - memblock_trim_memory()
>
> The tests for the memblock_alloc_*raw() functions test both top-down and
> bottom-up allocation directions. To add coverage for memblock_alloc_raw(),
> the alloc_api was updated so that it runs through all the existing tests
> twice: once for memblock_alloc() and once for memblock_alloc_raw(). When
> the tests run memblock_alloc_raw(), they test that the entire memory
> region is nonzero instead of testing that it is zero.
>
> Similarly, the alloc_nid_api was updated to run through its tests twice:
> once for memblock_alloc_try_nid() and once for
> memblock_alloc_try_nid_raw(). When the tests run
> memblock_alloc_try_nid_raw(), they test that the entire memory region is
> nonzero instead of testing that it is zero.
>
> The patch set also adds labels to verbose output for generic
> memblock_alloc*() tests that indicate which allocation direction is set.
> The function names of those tests do not include this information.
>
> ---
> Changelog
>
> v2 -> v3
> Based on feedback from Mike Rapoport:
> PATCH 1:
> - tests/common.h:
>      + Move (char *) cast inside ASSERT_MEM_EQ().
> PATCH 5:
> - tests/common.h:
>      + Move (char *) cast inside ASSERT_MEM_NE().
>
> Based on feedback from David Hildenbrand:
> PATCH 5:
> - tests/common.h:
>      + Add comments in enum test_flags and rename its values.
>      + Rename verify_mem_content() to assert_mem_content().
> PATCH 5, PATCH 6:
> - Update commit message to explain that the memory region is initialized
>    to nonzero and should remain unchanged when running
>    memblock_alloc_*raw().
> - tests/alloc_api.c, tests/alloc_nid_api.c:
>      + Remove func_testing array.
>      + get_func_testing():
>          * Rename to get_memblock_alloc_*name().
>          * Remove else.
>      + Remove else in run_memblock_alloc*().
> PATCH 7:
> - Commit message:
>      + Change "region unalign" to "unaligned region".
>      + Change "region unaligned" to "unaligned region that is unaligned"
> - tests/basic_api.c:
>      + Change phys_addr_t alignment to const phys_addr_t alignment.
>
> Based on feedback from Shaoqin Huang:
> PATCH 8:
> - tests/basic_api.c:
>      + Change "trimmed at the base" to "trimmed at the end" in
>        memblock_trim_memory_unaligned_end_check().
>
> v1 -> v2
> Updates based on feedback from Shaoqin Huang:
> PATCH 1:
> - tests/alloc_api.c:
>      + Remove fill_memblock() from alloc_no_memory_generic_check().
> - tests/common.c, tests/common.h:
>      + Change fill_memblock() to file static.
> PATCH 3:
> - Shaoqin Huang and I discussed using run_top_down() and run_bottom_up()
>    even for functions with `top_down` and `bottom_up` in the name to
>    maintain a consistent output style. However, this would make the output
>    more redundant, so no changes were made.
> PATCH 4:
> - tests/basic_api.c:
>      + Rename instances of r1_size and r2_size to
>        new_r1_size and new_r2_size.
> PATCH 5, PATCH 6:
> - tests/alloc_api.c, tests/alloc_nid_api.c, tests/common.h:
>      + Change verify_mem_content() to a common function defined in
>        common.h.
> PATCH 8:
> - tests/basic_api.c:
>      + Rename instances of r2_base and r2_size to
>        new_r2_base and new_r2_size.
> ---
>
> Rebecca Mckeever (8):
>    memblock tests: update tests to check if memblock_alloc zeroed memory
>    memblock tests: update zeroed memory check for memblock_alloc_* tests
>    memblock tests: add labels to verbose output for generic alloc tests
>    memblock tests: add additional tests for basic api and memblock_alloc
>    memblock tests: update alloc_api to test memblock_alloc_raw
>    memblock tests: update alloc_nid_api to test
>      memblock_alloc_try_nid_raw
>    memblock tests: add tests for memblock_*bottom_up functions
>    memblock tests: add tests for memblock_trim_memory
>
>   tools/testing/memblock/tests/alloc_api.c      | 168 +++-
>   .../memblock/tests/alloc_helpers_api.c        |  20 +-
>   tools/testing/memblock/tests/alloc_nid_api.c  | 252 +++---
>   tools/testing/memblock/tests/basic_api.c      | 767 ++++++++++++++++++
>   tools/testing/memblock/tests/common.c         |   7 +
>   tools/testing/memblock/tests/common.h         |  55 ++
>   6 files changed, 1082 insertions(+), 187 deletions(-)