On Wed, Nov 29, 2023 at 4:21 PM Nhat Pham nphamcs@gmail.com wrote:
On Wed, Nov 29, 2023 at 7:17 AM Johannes Weiner hannes@cmpxchg.org wrote:
On Mon, Nov 27, 2023 at 03:45:57PM -0800, Nhat Pham wrote:
static void shrink_worker(struct work_struct *w) { struct zswap_pool *pool = container_of(w, typeof(*pool), shrink_work);
struct mem_cgroup *memcg; int ret, failures = 0;/* global reclaim will select cgroup in a round-robin fashion. */ do {
ret = zswap_reclaim_entry(pool);if (ret) {zswap_reject_reclaim_fail++;if (ret != -EAGAIN)break;
spin_lock(&zswap_pools_lock);memcg = pool->next_shrink =mem_cgroup_iter_online(NULL, pool->next_shrink, NULL, true);/* full round trip */if (!memcg) {spin_unlock(&zswap_pools_lock); if (++failures == MAX_RECLAIM_RETRIES) break;goto resched; }/** Acquire an extra reference to the iterated memcg in case the* original reference is dropped by the zswap offlining callback.*/css_get(&memcg->css);struct mem_cgroup isn't defined when !CONFIG_MEMCG. This needs a mem_cgroup_get() wrapper and a dummy function for no-memcg builds.
I got this exact same issue a couple of versions ago, but it was hidden behind another helper function which can be implemented as a no-op in the case of !CONFIG_MEMCG, so I forgot about it until now. It always strikes me a bit weird that we have mem_cgroup_put() but not an equivalent get - let me correct that.
Actually, I'll instead implement mem_cgroup_tryget_online(), as we have to check for the cgroup's onlineness as well anyway! If it's online, then keep the extra reference - all good. If it's not, then drop the original reference before releasing the lock.
With that fixed, though, everything else looks good to me:
Acked-by: Johannes Weiner hannes@cmpxchg.org
Thanks for the review, Johannes!