- 28 Jan, 2008 40 commits
-
-
Stephen Hemminger authored
This converts dumping (and flushing) of large route tables form O(N^2) to O(N). If the route dump took multiple pages then the dump routine gets called again. The old code kept track of location by counter, the new code instead uses the last key. This is a really big win ( 0.3 sec vs 12 sec) for big route tables. One side effect is that if the table changes during the dump, then the last key will not be found, and we will return -EBUSY. Signed-off-by:
Stephen Hemminger <shemminger@vyatta.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Stephen Hemminger authored
Get rid of extra search that made route deletion O(n). Signed-off-by:
-
Stephen Hemminger authored
It is easier with TRIE to dump the data traversal rather than interating over every possible prefix. This saves some time and makes the dump come out in sorted order. Signed-off-by:
-
Stephen Hemminger authored
Remove the complex loop structure of nextleaf() and replace it with a simpler tree walker. This improves the performance and is much cleaner. Signed-off-by:
-
Stephen Hemminger authored
Match fib_hash, and set NLM_F_MULTI to handle multiple part messages. Signed-off-by:
-
Stephen Hemminger authored
The code to dump can use the existing hash chain rather than doing repeated lookup. Signed-off-by:
Stephen Hemminger <stephen.hemminger@vyatta.com> Signed-off-by:
-
Stephen Hemminger authored
Compute the number of prefixes when needed, rather than doing bookeeping. Signed-off-by:
-
Stephen Hemminger authored
Style cleanups: * make check_leaf return -1 or plen, rather than by reference * Get rid of #ifdef that is always set * split out embedded function calls in if statements. * checkpatch warnings Signed-off-by: -
Stephen Hemminger authored
This improves locality for operations that touch all the leaves. Save space since these entries don't need to be hardware cache aligned. Signed-off-by:
-
Jan Engelhardt authored
Signed-off-by:
Jan Engelhardt <jengelh@computergmbh.de> Signed-off-by:
-
Ross Burton authored
By default, LMP sets up a 3 seconds timer for discovery. We don't need it until discovery is set to 1. Signed-off-by:
Ross Burton <ross@openedhand.com> Signed-off-by:
Samuel Ortiz <samuel@sortiz.org> Signed-off-by:
-
Bruno Randolf authored
always extend the rx timestamp with the local TSF, since this information is also needed for proper IBSS merging. this is done in the tasklet for now, maybe has to be moved to the interrupt handler like in madwifi. drivers/net/wireless/ath5k/base.c: Changes-licensed-under: 3-Clause-BSD Signed-off-by:
Bruno Randolf <bruno@thinktube.com> Signed-off-by:
John W. Linville <linville@tuxdriver.com>
-
Bruno Randolf authored
in "11.1.2.2 Beacon generation in an IBSS" the IEEE802.11 standard says, each STA should... "b) Calculate a random delay uniformly distributed in the range between zero and twice aCWmin × aSlotTime,". configure cwmin and cwmax of the beacon queue in IBSS mode according to this. unfortunately beacon backoff does not work reliably yet, so i suspect we have a problem somewhere else, since the same settings (and similar beacon timer configuration) work for madwifi. drivers/net/wireless/ath5k/base.c: Changes-licensed-under: 3-Clause-BSD Signed-off-by:
-
Bruno Randolf authored
use SWBA (software beacon alert) interrupts to keep track of the next beacon time und check if a HW merge (automatic TSF update) has happened on every received beacon with the same BSSID. this is necessary because the atheros hardware will silently update the local TSF in IBSS mode, but not its beacon timers. if the TSF is ahead of the beacon timers no beacons are sent until the timers wrap around (typically after about 1 minute). this solution is not very nice, since we have to look into every beacon, but there is apparently no other way to detect HW merges. drivers/net/wireless/ath5k/base.c: Changes-licensed-under: 3-Clause-BSD drivers/net/wireless/ath5k/base.h: Changes-licensed-under: 3-Clause-BSD Signed-off-by:
-
Bruno Randolf authored
update ath5k_beacon_update_timers() for better beacon timer calculation in a variety of situations. most important is the possibility to call it with the timestamp of a received beacon, when we detected that a HW merge has happened and we need to reconfigure the beacon timers based on that. we call this from the mac80211 callback reset_tsf now instead of beacon_update, and there will be more use of it in the next patch. drivers/net/wireless/ath5k/base.c: Changes-licensed-under: 3-Clause-BSD Signed-off-by:
-
Bruno Randolf authored
the beacon interval is passed by mac80211 in TU already, so we can directly use it without conversion. also update the comments about TU (1 TU is defined by 802.11 as 1024usec). drivers/net/wireless/ath5k/ath5k.h: Changes-licensed-under: ISC drivers/net/wireless/ath5k/base.c: Changes-licensed-under: 3-Clause-BSD drivers/net/wireless/ath5k/base.h: Changes-licensed-under: 3-Clause-BSD Signed-off-by:
-
Bruno Randolf authored
reviewed beacon timer initialization with register traces from madwifi: what we are doing is correct :). one minor fix: use 3 instead of 0x00000003 - it's more readable. drivers/net/wireless/ath5k/hw.c: Changes-licensed-under: ISC Signed-off-by:
-
Michael Buesch authored
This zeros out all microcode related memory before loading the microcode. This also fixes initialization of the MAC control register. The _only_ place where we overwrite the contents of the MAC control register is at the beginning of b43_chip_init(). All other places must do read() -> mask/set -> write() to not overwrite existing bits. This also adds a longer delay for waiting for the microcode to initialize itself. It seems that the current timeout is sufficient on all available devices, but there's no real reason why we shouldn't wait for up to one second. Slow embedded devices might exist. Better safe than sorry. Signed-off-by:
Michael Buesch <mb@bu3sch.de> Signed-off-by:
-
Masakazu Mokuno authored
As event_type_pk_size[] is not used, Remove it. Signed-off-by:
Masakazu Mokuno <mokuno@sm.sony.co.jp> Signed-off-by:
-
John W. Linville authored
Otherwise it may be impossible to connected to an open network after a resume. This is a modified version of an original patch by Alex Eskin <alexeskin@yahoo.com>: https://bugzilla.redhat.com/show_bug.cgi?id=425950#c8Signed-off-by:
-
John W. Linville authored
...instead of using AR5K_KEYCACHE_SIZE, which would seem to be a typo/thinko... Signed-off-by: -
Michael Buesch authored
We must also store the ID string (filename) for the cached firmware blobs and verify that we really have the right firmware cached before using it. If we don't have the right fw cached, we must free it and request the correct blobs. This fixes bandswitch on A/B/G multi-PHY devices. Signed-off-by:
-
Ron Rindjunsky authored
This patch adds static declarations to functions in the Rx flow in order to eliminate sparse errors Signed-off-by:
Ron Rindjunsky <ron.rindjunsky@intel.com> Signed-off-by:
Tomas Winkler <tomas.winkler@intel.com> Signed-off-by:
-
Ron Rindjunsky authored
This patch changes ieee80211_bar control and start_seq_num to match the proper bitwise attribute expected from ieee 802.11 frame Signed-off-by:
-
Michael Buesch authored
This also adds lots of TODOs. Oh well. Lots of work. :) Signed-off-by:
-
Stefano Brivio authored
Fix last_sample initialization. kzalloc'ing the per-STA data wasn't enough, as jiffies can assume negative values as well. This fixes a bug which prevented correct PID operation for a while after booting. Thanks to Michael Buesch for reporting this. Reported-and-tested-by:
Stefano Brivio <stefano.brivio@polimi.it> Signed-off-by:
-
Johannes Berg authored
Commit c7a51bda ("mac80211: restructure __ieee80211_rx") extracted __ieee80211_rx_handle_packet out of __ieee80211_rx and hence changed the locking rules for __ieee80211_rx_handle_packet(), it is now invoked under RCU lock. There is, however, one instance left where it contains an rcu_read_unlock() in an error path, which is a bug. Signed-off-by:Johannes Berg <johannes@sipsolutions.net> Signed-off-by:
-
Michael Buesch authored
This fixes a sparse warning about weird locking. The spinlock is not needed, so simply remove it. This also adds some sanity checks to the PHY and radio locking to protect against recursive locking. Signed-off-by:
-
Guy Cohen authored
This patch assigns correct TID to frames transmitted between two stations in the same BSS in AP mode. The problem is that skb->protocol is not set to ETH_P_IP and it is wrong to use that field at this stage. The fix compares the LLC/Protocol headers explicitly to check if the encapsulated frame is IP frame Signed-off-by:
Guy Cohen <guy.cohen@intel.com> Signed-off-by:
-
Eric Dumazet authored
On x86_64, sizeof(struct rtable) is 0x148, which is rounded up to 0x180 bytes by SLAB allocator. We can reduce this to exactly 0x140 bytes, without alignment overhead, and store 12 struct rtable per PAGE instead of 10. rate_tokens is currently defined as an "unsigned long", while its content should not exceed 6*HZ. It can safely be converted to an unsigned int. Moving tclassid right after rate_tokens to fill the 4 bytes hole permits to save 8 bytes on 'struct dst_entry', which finally permits to save 8 bytes on 'struct rtable' Signed-off-by:
Eric Dumazet <dada1@cosmosbay.com> Signed-off-by:
-
Pavel Emelyanov authored
On namespace start we mainly prepare the ctl variables. When the namespace is stopped we have to kill all the fragments that point to this namespace. The inet_frags_exit_net() handles it. Signed-off-by:
Pavel Emelyanov <xemul@openvz.org> Signed-off-by:
-
Pavel Emelyanov authored
The inet_frags.lru_list is used for evicting only, so we have to make it per-namespace, to evict only those fragments, who's namespace exceeded its high threshold, but not the whole hash. Besides, this helps to avoid long loops in evictor. The spinlock is not per-namespace because it protects the hash table as well, which is global. Signed-off-by:
-
Pavel Emelyanov authored
Since we have one hashtable to lookup the fragment, having different secret_interval-s for hash rebuild doesn't make sense, so move this one to inet_frags. The inet_frags_ctl becomes empty after this, so remove it. The appropriate ctl table is kept read-only in namespaces. Signed-off-by:
-
Pavel Emelyanov authored
This is the same as with the timeout variable. Currently, after exceeding the high threshold _all_ the fragments are evicted, but it will be fixed in later patch. Signed-off-by:
-
Pavel Emelyanov authored
Move it to the netns_frags, adjust the usage and make the appropriate ctl table writable. Now fragment, that live in different namespaces can live for different times. Signed-off-by:
-
Pavel Emelyanov authored
Each namespace has to have own tables to tune their different parameters, so duplicate the tables and register them. All the tables in sub-namespaces are temporarily made read-only. Signed-off-by:
-
Pavel Emelyanov authored
This is also simple, but introduces more changes, since then mem counter is altered in more places. Signed-off-by:
-
Pavel Emelyanov authored
This is simple - just move the variable from struct inet_frags to struct netns_frags and adjust the usage appropriately. Signed-off-by:
-
Pavel Emelyanov authored
Since fragment management code is consolidated, we cannot have the pointer from inet_frag_queue to struct net, since we must know what king of fragment this is. So, I introduce the netns_frags structure. This one is currently empty, but will be eventually filled with per-namespace attributes. Each inet_frag_queue is tagged with this one. The conntrack_reasm is not "netns-izated", so it has one static netns_frags instance to keep working in init namespace. Signed-off-by:
-
Pavel Emelyanov authored
This is a preparation for sysctl netns-ization. Move the ctl tables to the files, where the tuning variables reside. Plus make the helpers to register the tables. This will simplify the later patches and will keep similar things closer to each other. ipv4, ipv6 and conntrack_reasm are patched differently, but the result is all the tables are in appropriate files. Signed-off-by:
-
