在 2020/6/2 下午12:05, Hugh Dickins 写道: > Cc'ing Alex Shi, because I noticed this when trying his v11 per-memcg > lru_lock series (which appears to be a big improvement over earlier > versions, thanks in particular to Johannes's memcg swap simplifications); > and Alex's 12/16 makes a change on top of Konstantin's latter check, > which will now just be reverted. I'm not yet confident in Alex's > isolate_migratepages_block(), in part because this muddle. Hi Hugh, Yes, this could make a very tricky change on compaction behavior. I will update the patchset after next update. Thanks a lot for notice this! And looking forward to more comments on per memcg lru_lock patchset! :) Thanks Alex
