diff options
author | Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> | 2008-05-29 06:25:23 -0400 |
---|---|---|
committer | David S. Miller <davem@davemloft.net> | 2008-05-29 06:25:23 -0400 |
commit | b79eeeb9e48457579cb742cd02e162fcd673c4a3 (patch) | |
tree | e804011ae95b5f21eb01732e4b0d768691cfd89d /include/linux/tcp.h | |
parent | 75a9cd524c6bb441c61c85bae7020ce5e8b2e807 (diff) |
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>
Diffstat (limited to 'include/linux/tcp.h')
-rw-r--r-- | include/linux/tcp.h | 50 |
1 files changed, 25 insertions, 25 deletions
diff --git a/include/linux/tcp.h b/include/linux/tcp.h index 18e62e3d406f..9881295f3857 100644 --- a/include/linux/tcp.h +++ b/include/linux/tcp.h | |||
@@ -296,10 +296,9 @@ struct tcp_sock { | |||
296 | u32 rcv_ssthresh; /* Current window clamp */ | 296 | u32 rcv_ssthresh; /* Current window clamp */ |
297 | 297 | ||
298 | u32 frto_highmark; /* snd_nxt when RTO occurred */ | 298 | u32 frto_highmark; /* snd_nxt when RTO occurred */ |
299 | u8 reordering; /* Packet reordering metric. */ | 299 | u16 advmss; /* Advertised MSS */ |
300 | u8 frto_counter; /* Number of new acks after RTO */ | 300 | u8 frto_counter; /* Number of new acks after RTO */ |
301 | u8 nonagle; /* Disable Nagle algorithm? */ | 301 | u8 nonagle; /* Disable Nagle algorithm? */ |
302 | u8 keepalive_probes; /* num of allowed keep alive probes */ | ||
303 | 302 | ||
304 | /* RTT measurement */ | 303 | /* RTT measurement */ |
305 | u32 srtt; /* smoothed round trip time << 3 */ | 304 | u32 srtt; /* smoothed round trip time << 3 */ |
@@ -310,6 +309,10 @@ struct tcp_sock { | |||
310 | 309 | ||
311 | u32 packets_out; /* Packets which are "in flight" */ | 310 | u32 packets_out; /* Packets which are "in flight" */ |
312 | u32 retrans_out; /* Retransmitted packets out */ | 311 | u32 retrans_out; /* Retransmitted packets out */ |
312 | |||
313 | u16 urg_data; /* Saved octet of OOB data and control flags */ | ||
314 | u8 urg_mode; /* In urgent mode */ | ||
315 | u8 ecn_flags; /* ECN status bits. */ | ||
313 | /* | 316 | /* |
314 | * Options received (usually on last packet, some only on SYN packets). | 317 | * Options received (usually on last packet, some only on SYN packets). |
315 | */ | 318 | */ |
@@ -325,13 +328,24 @@ struct tcp_sock { | |||
325 | u32 snd_cwnd_used; | 328 | u32 snd_cwnd_used; |
326 | u32 snd_cwnd_stamp; | 329 | u32 snd_cwnd_stamp; |
327 | 330 | ||
328 | struct sk_buff_head out_of_order_queue; /* Out of order segments go here */ | ||
329 | |||
330 | u32 rcv_wnd; /* Current receiver window */ | 331 | u32 rcv_wnd; /* Current receiver window */ |
331 | u32 write_seq; /* Tail(+1) of data held in tcp send buffer */ | 332 | u32 write_seq; /* Tail(+1) of data held in tcp send buffer */ |
332 | u32 pushed_seq; /* Last pushed seq, required to talk to windows */ | 333 | u32 pushed_seq; /* Last pushed seq, required to talk to windows */ |
334 | u32 lost_out; /* Lost packets */ | ||
335 | u32 sacked_out; /* SACK'd packets */ | ||
336 | u32 fackets_out; /* FACK'd packets */ | ||
337 | u32 tso_deferred; | ||
338 | u32 bytes_acked; /* Appropriate Byte Counting - RFC3465 */ | ||
333 | 339 | ||
334 | /* SACKs data */ | 340 | /* from STCP, retrans queue hinting */ |
341 | struct sk_buff* lost_skb_hint; | ||
342 | struct sk_buff *scoreboard_skb_hint; | ||
343 | struct sk_buff *retransmit_skb_hint; | ||
344 | struct sk_buff *forward_skb_hint; | ||
345 | |||
346 | struct sk_buff_head out_of_order_queue; /* Out of order segments go here */ | ||
347 | |||
348 | /* SACKs data, these 2 need to be together (see tcp_build_and_update_options) */ | ||
335 | struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */ | 349 | struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */ |
336 | struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/ | 350 | struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/ |
337 | 351 | ||
@@ -342,23 +356,14 @@ struct tcp_sock { | |||
342 | * sacked_out > 0) | 356 | * sacked_out > 0) |
343 | */ | 357 | */ |
344 | 358 | ||
345 | /* from STCP, retrans queue hinting */ | ||
346 | struct sk_buff* lost_skb_hint; | ||
347 | |||
348 | struct sk_buff *scoreboard_skb_hint; | ||
349 | struct sk_buff *retransmit_skb_hint; | ||
350 | struct sk_buff *forward_skb_hint; | ||
351 | |||
352 | int lost_cnt_hint; | 359 | int lost_cnt_hint; |
353 | int retransmit_cnt_hint; | 360 | int retransmit_cnt_hint; |
354 | 361 | ||
355 | u32 lost_retrans_low; /* Sent seq after any rxmit (lowest) */ | 362 | u32 lost_retrans_low; /* Sent seq after any rxmit (lowest) */ |
356 | 363 | ||
357 | u16 advmss; /* Advertised MSS */ | 364 | u8 reordering; /* Packet reordering metric. */ |
365 | u8 keepalive_probes; /* num of allowed keep alive probes */ | ||
358 | u32 prior_ssthresh; /* ssthresh saved at recovery start */ | 366 | u32 prior_ssthresh; /* ssthresh saved at recovery start */ |
359 | u32 lost_out; /* Lost packets */ | ||
360 | u32 sacked_out; /* SACK'd packets */ | ||
361 | u32 fackets_out; /* FACK'd packets */ | ||
362 | u32 high_seq; /* snd_nxt at onset of congestion */ | 367 | u32 high_seq; /* snd_nxt at onset of congestion */ |
363 | 368 | ||
364 | u32 retrans_stamp; /* Timestamp of the last retransmit, | 369 | u32 retrans_stamp; /* Timestamp of the last retransmit, |
@@ -366,25 +371,18 @@ struct tcp_sock { | |||
366 | * the first SYN. */ | 371 | * the first SYN. */ |
367 | u32 undo_marker; /* tracking retrans started here. */ | 372 | u32 undo_marker; /* tracking retrans started here. */ |
368 | int undo_retrans; /* number of undoable retransmissions. */ | 373 | int undo_retrans; /* number of undoable retransmissions. */ |
374 | u32 total_retrans; /* Total retransmits for entire connection */ | ||
375 | |||
369 | u32 urg_seq; /* Seq of received urgent pointer */ | 376 | u32 urg_seq; /* Seq of received urgent pointer */ |
370 | u16 urg_data; /* Saved octet of OOB data and control flags */ | ||
371 | u8 urg_mode; /* In urgent mode */ | ||
372 | u8 ecn_flags; /* ECN status bits. */ | ||
373 | u32 snd_up; /* Urgent pointer */ | 377 | u32 snd_up; /* Urgent pointer */ |
374 | 378 | ||
375 | u32 total_retrans; /* Total retransmits for entire connection */ | ||
376 | u32 bytes_acked; /* Appropriate Byte Counting - RFC3465 */ | ||
377 | |||
378 | unsigned int keepalive_time; /* time before keep alive takes place */ | 379 | unsigned int keepalive_time; /* time before keep alive takes place */ |
379 | unsigned int keepalive_intvl; /* time interval between keep alive probes */ | 380 | unsigned int keepalive_intvl; /* time interval between keep alive probes */ |
380 | int linger2; | ||
381 | 381 | ||
382 | struct tcp_deferred_accept_info defer_tcp_accept; | 382 | struct tcp_deferred_accept_info defer_tcp_accept; |
383 | 383 | ||
384 | unsigned long last_synq_overflow; | 384 | unsigned long last_synq_overflow; |
385 | 385 | ||
386 | u32 tso_deferred; | ||
387 | |||
388 | /* Receiver side RTT estimation */ | 386 | /* Receiver side RTT estimation */ |
389 | struct { | 387 | struct { |
390 | u32 rtt; | 388 | u32 rtt; |
@@ -412,6 +410,8 @@ struct tcp_sock { | |||
412 | /* TCP MD5 Signagure Option information */ | 410 | /* TCP MD5 Signagure Option information */ |
413 | struct tcp_md5sig_info *md5sig_info; | 411 | struct tcp_md5sig_info *md5sig_info; |
414 | #endif | 412 | #endif |
413 | |||
414 | int linger2; | ||
415 | }; | 415 | }; |
416 | 416 | ||
417 | static inline struct tcp_sock *tcp_sk(const struct sock *sk) | 417 | static inline struct tcp_sock *tcp_sk(const struct sock *sk) |