linux/include
Ilpo Järvinen b79eeeb9e4 tcp: Reorganize tcp_sock to fill 64-bit holes & improve locality
I tried to group recovery related fields nearby (non-CA_Open related
variables, to be more accurate) so that one to three cachelines would
not be necessary in CA_Open. These are now contiguously deployed:

  struct sk_buff_head        out_of_order_queue;   /*  1968    80 */
  /* --- cacheline 32 boundary (2048 bytes) --- */
  struct tcp_sack_block      duplicate_sack[1];    /*  2048     8 */
  struct tcp_sack_block      selective_acks[4];    /*  2056    32 */
  struct tcp_sack_block      recv_sack_cache[4];   /*  2088    32 */
  /* --- cacheline 33 boundary (2112 bytes) was 8 bytes ago --- */
  struct sk_buff *           highest_sack;         /*  2120     8 */
  int                        lost_cnt_hint;        /*  2128     4 */
  int                        retransmit_cnt_hint;  /*  2132     4 */
  u32                        lost_retrans_low;     /*  2136     4 */
  u8                         reordering;           /*  2140     1 */
  u8                         keepalive_probes;     /*  2141     1 */

  /* XXX 2 bytes hole, try to pack */

  u32                        prior_ssthresh;       /*  2144     4 */
  u32                        high_seq;             /*  2148     4 */
  u32                        retrans_stamp;        /*  2152     4 */
  u32                        undo_marker;          /*  2156     4 */
  int                        undo_retrans;         /*  2160     4 */
  u32                        total_retrans;        /*  2164     4 */

...and they're then followed by URG slowpath & keepalive related
variables.

Head of the out_of_order_queue always needed for empty checks, if
that's empty (and TCP is in CA_Open), following ~200 bytes (in 64-bit)
shouldn't be necessary for anything. If only OFO queue exists but TCP
is in CA_Open, selective_acks (and possibly duplicate_sack) are
necessary besides the out_of_order_queue but the rest of the block
again shouldn't be (ie., the other direction had losses).

As the cacheline boundaries depend on many factors in the preceeding
stuff, trying to align considering them doesn't make too much sense.

Commented one ordering hazard.

There are number of low utilized u8/16s that could be combined get 2
bytes less in total so that the hole could be made to vanish (includes
at least ecn_flags, urg_data, urg_mode, frto_counter, nonagle).

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Acked-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-05-29 03:25:23 -07:00
..
acpi ACPICA: update Intel copyright 2008-04-23 23:00:13 -04:00
asm-alpha asm-{alpha,h8300,um,v850,xtensa}/param.h: unbreak HZ for userspace 2008-05-14 19:11:14 -07:00
asm-arm [ARM] pxa: Fix RCSR handling 2008-05-08 18:04:02 +01:00
asm-avr32 avr32: types: use <asm-generic/int-*.h> for the avr32 architecture 2008-05-02 16:18:20 -07:00
asm-blackfin Blackfin Serial Driver: abstract away DLAB differences into header 2008-05-07 11:41:26 +08:00
asm-cris cris: types: use <asm-generic/int-*.h> for the cris architecture 2008-05-02 16:18:20 -07:00
asm-frv read_barrier_depends arch fixlets 2008-05-14 10:05:18 -07:00
asm-generic types: add C99-style constructors to <asm-generic/int-*.h> 2008-05-02 16:18:42 -07:00
asm-h8300 asm-{alpha,h8300,um,v850,xtensa}/param.h: unbreak HZ for userspace 2008-05-14 19:11:14 -07:00
asm-ia64 ia64: types: use <asm-generic/int-*.h> for the ia64 architecture 2008-05-02 16:18:21 -07:00
asm-m32r m32r: types: use <asm-generic/int-*.h> for the m32r architecture 2008-05-02 16:18:21 -07:00
asm-m68k m68k: remove old mac_esp cruft 2008-05-05 12:38:50 -07:00
asm-m68knommu m68knommu: rework definition of HZ 2008-05-01 08:08:36 -07:00
asm-mips [MIPS] Get rid of __ilog2 2008-05-12 16:46:55 +01:00
asm-mn10300 MN10300: Make cpu_relax() invoke barrier() 2008-05-08 10:49:39 -07:00
asm-parisc parisc: types: use <asm-generic/int-*.h> for the parisc architecture 2008-05-02 16:18:32 -07:00
asm-powerpc Merge branch 'for-2.6.26' of master.kernel.org:/pub/scm/linux/kernel/git/jwboyer/powerpc-4xx into merge 2008-05-09 20:12:06 +10:00
asm-ppc [POWERPC] ppc: More compile fixes 2008-05-12 22:57:51 +10:00
asm-s390 types: s390: fix #ifdef reversal in <asm-s390/types.h> 2008-05-12 08:49:59 -07:00
asm-sh sh: Fix DMAC base address for SH7709S 2008-05-09 19:04:12 +09:00
asm-sparc sparc: Fix mremap address range validation. 2008-05-12 16:33:33 -07:00
asm-sparc64 sparc64: Use a TS_RESTORE_SIGMASK 2008-05-12 22:45:15 -07:00
asm-um asm-{alpha,h8300,um,v850,xtensa}/param.h: unbreak HZ for userspace 2008-05-14 19:11:14 -07:00
asm-v850 asm-{alpha,h8300,um,v850,xtensa}/param.h: unbreak HZ for userspace 2008-05-14 19:11:14 -07:00
asm-x86 mprotect: prevent alteration of the PAT bits 2008-05-14 19:11:15 -07:00
asm-xtensa asm-{alpha,h8300,um,v850,xtensa}/param.h: unbreak HZ for userspace 2008-05-14 19:11:14 -07:00
crypto [CRYPTO] api: Fix scatterwalk_sg_chain 2008-05-01 18:22:28 +08:00
keys KEYS: Fix the comment to match the file name in rxrpc-type.h. 2008-04-21 22:43:55 +00:00
linux tcp: Reorganize tcp_sock to fill 64-bit holes & improve locality 2008-05-29 03:25:23 -07:00
math-emu
media i2c: Convert some more new-style drivers to use module aliasing 2008-05-11 20:37:06 +02:00
mtd
net Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6 2008-05-25 23:26:10 -07:00
pcmcia
rdma IB: expand ib_umem_get() prototype 2008-04-29 08:06:12 -07:00
rxrpc
scsi Merge git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6 2008-05-02 13:52:35 -07:00
sound [ALSA] ASoC: build fix for snd_soc_info_bool_ext 2008-05-13 14:47:44 +02:00
video fbdev: platforming hecubafb and n411 2008-04-28 08:58:41 -07:00
xen xen: add balloon driver 2008-04-24 23:57:33 +02:00
Kbuild