Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider: - [Critical] The patch adds error cleanup to `ttm_pool_mgr_init()` (freeing the shrinker and destroying pools), but the caller (`ttm_global_init()`) ignores its return value. This leads to double-frees and NULL pointer dereferences during teardown. - [High] `ttm_pool_init()` ignores the newly added return value of `ttm_pool_type_init()`, resulting in guaranteed NULL pointer dereferences during teardown if initialization fails. - [High] Lockless TOCTOU data race on `shrinker_list` emptiness check in `ttm_pool_shrink()` leads to invalid pointer dereference and memory corruption. --
--- Patch [1]: [PATCH v4] Memory leak error in qxl unbind --- commit ed37ece8c0ad50e1a16deff6653abfe2c17eb1a0 Author: Óscar Megía López <[email protected]> Subject: Memory leak error in qxl unbind This commit addresses an OOM leak and refcount corruption in qxl unbind. It adds proper cleanup and error propagation for list_lru_init() failures during TTM pool initialization to ensure pool types are not left orphaned. Link: https://lore.kernel.org/lkml/[email protected]/ Link: https://lore.kernel.org/dri-devel/[email protected]/ > diff --git a/drivers/gpu/drm/ttm/ttm_pool.c b/drivers/gpu/drm/ttm/ttm_pool.c > index 1bf37023fed66..6821fadf1a8d0 100644 > --- a/drivers/gpu/drm/ttm/ttm_pool.c > +++ b/drivers/gpu/drm/ttm/ttm_pool.c > @@ -374,17 +374,23 @@ static struct page *ttm_pool_type_take(struct > ttm_pool_type *pt, int nid) > } > > /* Initialize and add a pool type to the global shrinker list */ > -static void ttm_pool_type_init(struct ttm_pool_type *pt, struct ttm_pool > *pool, > +static int ttm_pool_type_init(struct ttm_pool_type *pt, struct ttm_pool > *pool, > enum ttm_caching caching, unsigned int order) [Severity: High] Since this function now returns an error code instead of void, do all callers check the new return value? Looking at ttm_pool_init() in drivers/gpu/drm/ttm/ttm_pool.c: ttm_pool_type_init(pt, pool, i, j); If ttm_pool_type_init() returns an error, pt->shrinker_list remains zero-initialized. Later, when ttm_pool_fini() calls ttm_pool_type_fini(), it unconditionally runs: list_del(&pt->shrinker_list); Will this cause a NULL pointer dereference on the zero-initialized list head during teardown? [ ... ] > @@ -458,6 +464,9 @@ static unsigned int ttm_pool_shrink(int nid, unsigned > long num_to_free) > struct ttm_pool_type *pt; > unsigned int num_pages; > > + if ((shrinker_list.prev == &shrinker_list) && (shrinker_list.next == > &shrinker_list)) > + return 0; > + > down_read(&pool_shrink_rwsem); > spin_lock(&shrinker_lock); > pt = list_first_entry(&shrinker_list, typeof(*pt), shrinker_list); [Severity: High] Is it safe to check if shrinker_list is empty without holding shrinker_lock? If the list becomes empty between this check and acquiring shrinker_lock (for example, if a concurrent thread calls ttm_pool_type_fini() during device teardown), list_first_entry() will execute on an empty list. This would return a pointer to the list head itself cast to a struct ttm_pool_type pointer, which could lead to memory corruption when list_move_tail() and list_lru_walk_node() are invoked on this invalid pointer. Could this check be safely moved inside the critical section? [ ... ] > @@ -1506,15 +1576,53 @@ int ttm_pool_mgr_init(unsigned long num_pages) > spin_lock_init(&shrinker_lock); > INIT_LIST_HEAD(&shrinker_list); > > + mm_shrinker = shrinker_alloc(SHRINKER_NUMA_AWARE, "drm-ttm_pool"); > + if (!mm_shrinker) > + return -ENOMEM; > + > for (i = 0; i < NR_PAGE_ORDERS; ++i) { > - ttm_pool_type_init(&global_write_combined[i], NULL, > + ret = ttm_pool_type_init(&global_write_combined[i], NULL, > ttm_write_combined, i); > - ttm_pool_type_init(&global_uncached[i], NULL, ttm_uncached, i); > + if (ret) { > + ttm_pool_type_fini_and_list_lru_destroy(i); > + shrinker_free(mm_shrinker); > + return ret; > + } [Severity: Critical] If ttm_pool_mgr_init() fails here, does the caller properly handle the returned error? If the caller (ttm_global_init()) ignores this error and continues execution, mm_shrinker is left pointing to freed memory while execution proceeds. Then during module unload, ttm_pool_mgr_fini() unconditionally calls: shrinker_free(mm_shrinker); ttm_pool_type_fini_and_list_lru_destroy(NR_PAGE_ORDERS); Could this lead to a double-free of mm_shrinker and potential NULL pointer dereferences on the partially destroyed list structures? -- Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
