linux_dsm_epyc7002/include
Eric Dumazet 81c3d5470e [INET]: speedup inet (tcp/dccp) lookups
Arnaldo and I agreed it could be applied now, because I have other
pending patches depending on this one (Thank you Arnaldo)

(The other important patch moves skc_refcnt in a separate cache line,
so that the SMP/NUMA performance doesnt suffer from cache line ping pongs)

1) First some performance data :
--------------------------------

tcp_v4_rcv() wastes a *lot* of time in __inet_lookup_established()

The most time critical code is :

sk_for_each(sk, node, &head->chain) {
     if (INET_MATCH(sk, acookie, saddr, daddr, ports, dif))
         goto hit; /* You sunk my battleship! */
}

The sk_for_each() does use prefetch() hints but only the begining of
"struct sock" is prefetched.

As INET_MATCH first comparison uses inet_sk(__sk)->daddr, wich is far
away from the begining of "struct sock", it has to bring into CPU
cache cold cache line. Each iteration has to use at least 2 cache
lines.

This can be problematic if some chains are very long.

2) The goal
-----------

The idea I had is to change things so that INET_MATCH() may return
FALSE in 99% of cases only using the data already in the CPU cache,
using one cache line per iteration.

3) Description of the patch
---------------------------

Adds a new 'unsigned int skc_hash' field in 'struct sock_common',
filling a 32 bits hole on 64 bits platform.

struct sock_common {
	unsigned short		skc_family;
	volatile unsigned char	skc_state;
	unsigned char		skc_reuse;
	int			skc_bound_dev_if;
	struct hlist_node	skc_node;
	struct hlist_node	skc_bind_node;
	atomic_t		skc_refcnt;
+	unsigned int		skc_hash;
	struct proto		*skc_prot;
};

Store in this 32 bits field the full hash, not masked by (ehash_size -
1) Using this full hash as the first comparison done in INET_MATCH
permits us immediatly skip the element without touching a second cache
line in case of a miss.

Suppress the sk_hashent/tw_hashent fields since skc_hash (aliased to
sk_hash and tw_hash) already contains the slot number if we mask with
(ehash_size - 1)

File include/net/inet_hashtables.h

64 bits platforms :
#define INET_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)\
     (((__sk)->sk_hash == (__hash))
     ((*((__u64 *)&(inet_sk(__sk)->daddr)))== (__cookie))   &&  \
     ((*((__u32 *)&(inet_sk(__sk)->dport))) == (__ports))   &&  \
     (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))

32bits platforms:
#define TCP_IPV4_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)\
     (((__sk)->sk_hash == (__hash))                 &&  \
     (inet_sk(__sk)->daddr          == (__saddr))   &&  \
     (inet_sk(__sk)->rcv_saddr      == (__daddr))   &&  \
     (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))


- Adds a prefetch(head->chain.first) in 
__inet_lookup_established()/__tcp_v4_check_established() and 
__inet6_lookup_established()/__tcp_v6_check_established() and 
__dccp_v4_check_established() to bring into cache the first element of the 
list, before the {read|write}_lock(&head->lock);

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-03 14:13:38 -07:00
..
acpi [ACPI] ACPICA 20050902 2005-09-03 00:15:11 -04:00
asm-alpha [PATCH] slab: alpha inlining fix 2005-09-22 22:17:34 -07:00
asm-arm Merge master.kernel.org:/home/rmk/linux-2.6-arm 2005-09-30 08:39:56 -07:00
asm-arm26 [PATCH] Remove unused var from asm/futex.h 2005-09-21 16:16:29 -07:00
asm-cris [PATCH] Remove unused var from asm/futex.h 2005-09-21 16:16:29 -07:00
asm-frv [PATCH] Remove unused var from asm/futex.h 2005-09-21 16:16:29 -07:00
asm-generic [PATCH] mm: move_pte to remap ZERO_PAGE 2005-09-28 07:46:40 -07:00
asm-h8300 [PATCH] Remove unused var from asm/futex.h 2005-09-21 16:16:29 -07:00
asm-i386 [PATCH] x86: hw_irq.h warning fix 2005-09-30 12:41:17 -07:00
asm-ia64 [PATCH] ia64 basic __user annotations 2005-09-29 08:46:27 -07:00
asm-m32r [PATCH] m32r: more basic __user annotations 2005-09-26 18:29:50 -07:00
asm-m68k [PATCH] Remove unused var from asm/futex.h 2005-09-21 16:16:29 -07:00
asm-m68knommu [PATCH] Remove unused var from asm/futex.h 2005-09-21 16:16:29 -07:00
asm-mips [PATCH] mm: move_pte to remap ZERO_PAGE 2005-09-28 07:46:40 -07:00
asm-parisc [PATCH] Remove unused var from asm/futex.h 2005-09-21 16:16:29 -07:00
asm-powerpc [PATCH] ppc64: Add definitions for new PTRACE calls 2005-09-12 17:19:12 +10:00
asm-ppc [PATCH] missing qualifiers in readb() et.al. on ppc 2005-09-30 08:42:24 -07:00
asm-ppc64 [PATCH] ppc64 get_user annotations 2005-09-29 08:46:27 -07:00
asm-s390 [PATCH] s390 signal annotations 2005-09-29 08:46:27 -07:00
asm-sh [PATCH] Remove unused var from asm/futex.h 2005-09-21 16:16:29 -07:00
asm-sh64 [PATCH] Remove unused var from asm/futex.h 2005-09-21 16:16:29 -07:00
asm-sparc [SPARC]: Declare paging_init() in asm/pgtable.h 2005-09-28 21:31:25 -07:00
asm-sparc64 [SPARC64]: Kill arch/sparc64/prom/memory.c 2005-09-29 18:05:28 -07:00
asm-um [PATCH] uml get_user() NULL noise removal 2005-09-30 08:42:24 -07:00
asm-v850 [PATCH] Remove unused var from asm/futex.h 2005-09-21 16:16:29 -07:00
asm-x86_64 [PATCH] x86_64: Add missing () around arguments of pte_index macro 2005-09-30 08:54:38 -07:00
asm-xtensa [PATCH] xtensa: remove io_remap_page_range and minor clean-ups 2005-09-22 22:17:37 -07:00
linux [INET]: speedup inet (tcp/dccp) lookups 2005-10-03 14:13:38 -07:00
math-emu
media [PATCH] v4l: tveeprom improved to support newer Hauppage cards 2005-09-09 13:57:54 -07:00
mtd
net [INET]: speedup inet (tcp/dccp) lookups 2005-10-03 14:13:38 -07:00
pcmcia [PATCH] yenta: auto-tune EnE bridges for CardBus cards 2005-09-26 13:09:20 +02:00
rdma [IB] uverbs: Close some exploitable races 2005-09-26 13:01:03 -07:00
rxrpc
scsi [SCSI] allow REPORT LUN scanning even for LUN 0 PQ of 3 2005-09-25 12:01:48 -05:00
sound [ALSA] version 1.0.10rc1 2005-09-12 11:14:05 +02:00
video [PATCH] drivers/video: Replace custom macro with isdigit() 2005-09-13 08:22:33 -07:00