Now with the previous patch, __sk_mem_raise_allocated() considers the memory pressure of both global and the socket's memcg on a func- wide level, making the condition of memcg's pressure in question redundant. Signed-off-by: Abel Wu <wuyun.abel@xxxxxxxxxxxxx> --- net/core/sock.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) diff --git a/net/core/sock.c b/net/core/sock.c index 801df091e37a..86735ad2f903 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -3020,9 +3020,15 @@ int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind) if (sk_has_memory_pressure(sk)) { u64 alloc; - if (!sk_under_memory_pressure(sk)) + if (!sk_under_global_memory_pressure(sk)) return 1; + alloc = sk_sockets_allocated_read_positive(sk); + + /* If under global pressure, allow the sockets that are below + * average memory usage to raise, trying to be fair among all + * the sockets under global constrains. + */ if (sk_prot_mem_limits(sk, 2) > alloc * sk_mem_pages(sk->sk_wmem_queued + atomic_read(&sk->sk_rmem_alloc) + -- 2.37.3