Re: [PATCH v7] coccinelle: api: add kfree_mismatch script

[Denis Efremov <efremov@linux.com>]

Hi,

On 8/3/20 9:34 PM, Denis Efremov wrote:
> Check that alloc and free types of functions match each other.

Julia, I've just send the patches to fix all the warnings emitted by the script.

[1] https://lore.kernel.org/patchwork/patch/1309731/
[2] https://lore.kernel.org/patchwork/patch/1309273/
[3] https://lore.kernel.org/patchwork/patch/1309275/

Other inconsistencies and bugs detected by this script:

1e814d630fd1 drm/amd/display: Use kfree() to free rgb_user in calculate_user_regamma_ramp()
842540075974 drm/amd/display: Use kvfree() to free coeff in build_regamma()
f5e383ac8b58 iommu/pamu: Use kzfree() in fsl_pamu_probe()
360000b26e37 net/mlx5: Use kfree(ft->g) in arfs_create_groups()
114427b8927a drm/panfrost: Use kvfree() to free bo->sgts
742532d11d83 f2fs: use kfree() instead of kvfree() to free superblock data
47a357de2b6b net/mlx5: DR, Fix freeing in dr_create_rc_qp()
a8c73c1a614f io_uring: use kvfree() in io_sqe_buffer_register()
7f89cc07d22a cxgb4: Use kfree() instead kvfree() where appropriate
bb2359f4dbe9 bpf: Change kvfree to kfree in generic_map_lookup_batch()


> Changes in v2:
>  - Lines are limited to 80 characters where possible
>  - Confidence changed from High to Medium because of 
>    fs/btrfs/send.c:1119 false-positive
>  - __vmalloc_area_node() explicitly excluded from analysis
>    instead of !(file in "mm/vmalloc.c") condition
> Changes in v3:
>  - prints style in org && report modes changed for python2
> Changes in v4:
>  - missing msg argument to print_todo fixed
> Changes in v5:
>  - fix position p in kfree rule
>  - move @kok and @v positions in choice rule after the arguments
>  - remove kvmalloc suggestions
> Changes in v6:
>  - more asterisks added in context mode
>  - second @kok added to the choice rule
> Changes in v7:
>  - file renamed to kfree_mismatch.cocci
>  - python function relevant() removed
>  - additional rule for filtering free positions added
>  - btrfs false-positive fixed
>  - confidence level changed to high
>  - kvfree_switch rule added
>  - names for position variables changed to @a (alloc) and @f (free)

Is there something I can improve in this cocci script to be accepted?

Thanks,
Denis