Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [High] Missing Memcg Accounting in SVM Caches --
commit c7235f541841c86e76bfeba1fd000ec628328f01 Author: Honglei Huang <[email protected]> drm/amdgpu: add SVM core header and VM integration This commit adds the core header file amdgpu_svm.h defining the data structures, public API, and caching macros for SVM integration. It also adds an experimental SVM pointer to struct amdgpu_vm. > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_svm.h > b/drivers/gpu/drm/amd/amdgpu/amdgpu_svm.h > new file mode 100644 > index 0000000000000..9884a1f834d67 > --- /dev/null > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_svm.h [ ... ] > +#define AMDGPU_SVM_ERR(fmt, ...) \ > + pr_err("%s: " fmt, __func__, ##__VA_ARGS__) > + > +#define AMDGPU_SVM_KMEM_CACHE_CREATE(name, type) \ > + kmem_cache_create((name), sizeof(type), 0, 0, NULL) [Severity: High] Does this hardcoded flag value prevent proper memory cgroup accounting? This macro hardcodes the flags parameter of kmem_cache_create() to 0, which omits SLAB_ACCOUNT. If user-controlled structures like the SVM memory range tracking cache are instantiated using this macro, omitting memory cgroup accounting allows a containerized process to allocate unbounded amounts of kernel memory without being charged to its cgroup. Could this bypass container resource limits and potentially exhaust system kernel memory? > + > +#define AMDGPU_SVM_KMEM_CACHE_DESTROY(cache) \ > + do { \ [ ... ] -- Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=2
