diff options
| author | Eric Dumazet <edumazet@google.com> | 2015-05-15 12:39:25 -0700 |
|---|---|---|
| committer | Ben Hutchings <ben@decadent.org.uk> | 2017-03-16 02:18:52 +0000 |
| commit | 065d8e5a8975d0a576f0d1cb50108455e79f7c63 (patch) | |
| tree | 362b1e49fc664f4f5f037cd01484a5573f593105 /net | |
| parent | 093582b9b161c768f2038332d1cce5fbd696820e (diff) | |
net: fix sk_mem_reclaim_partial()
commit 1a24e04e4b50939daa3041682b38b82c896ca438 upstream.
sk_mem_reclaim_partial() goal is to ensure each socket has
one SK_MEM_QUANTUM forward allocation. This is needed both for
performance and better handling of memory pressure situations in
follow up patches.
SK_MEM_QUANTUM is currently a page, but might be reduced to 4096 bytes
as some arches have 64KB pages.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
[bwh: Backported to 3.2:
- Keep using atomic_long_sub() directly, not sk_memory_allocated_sub()
- Adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Diffstat (limited to 'net')
| -rw-r--r-- | net/core/sock.c | 9 |
1 files changed, 5 insertions, 4 deletions
diff --git a/net/core/sock.c b/net/core/sock.c index 54a21d437db4..b8f4f353a7d4 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -1750,14 +1750,15 @@ EXPORT_SYMBOL(__sk_mem_schedule); /** * __sk_reclaim - reclaim memory_allocated * @sk: socket + * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple) */ -void __sk_mem_reclaim(struct sock *sk) +void __sk_mem_reclaim(struct sock *sk, int amount) { struct proto *prot = sk->sk_prot; - atomic_long_sub(sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT, - prot->memory_allocated); - sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1; + amount >>= SK_MEM_QUANTUM_SHIFT; + atomic_long_sub(amount, prot->memory_allocated); + sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT; if (prot->memory_pressure && *prot->memory_pressure && (atomic_long_read(prot->memory_allocated) < prot->sysctl_mem[0])) |