- 11 Jul, 2007 40 commits
-
-
Andrew Morton authored
- save 4 bytes - it's read-mostly. Signed-off-by:
Andrew Morton <akpm@linux-foundation.org> Acked-by:
Vasily Averin <vvs@sw.ru> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Pavel Emelianov authored
This includes /proc/net/protocols, /proc/net/rxrpc_calls and /proc/net/rxrpc_connections files. All three need seq_list_start_head to show some header. Signed-off-by:
Pavel Emelianov <xemul@openvz.org> Signed-off-by:
-
Pavel Emelianov authored
The .show callback receives the list_head pointer now, not the struct br2684_dev one. Signed-off-by:
-
Stephen Hemminger authored
Document the expectations about device MTU handling. The documentation about oversize packet handling is probably too loose. IMHO devices should drop oversize packets for robustness, but many devices allow it now. For example, if you set mtu to 1200 bytes, most ether devices will allow a 1500 byte frame in. Signed-off-by:
Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by:
-
Stephen Hemminger authored
Update the documentation about locking assumptions. Signed-off-by:
-
Ilpo Järvinen authored
Got bored to always recompile it for no reason. Signed-off-by:
Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Signed-off-by:
-
Michael Chan authored
Signed-off-by:
Michael Chan <mchan@broadcom.com> Signed-off-by:
-
Michael Chan authored
Add management firmware version for ethtool -i. Signed-off-by:
-
Michael Chan authored
In addition to the periodic heartbeat, we're adding a heartbeat request interrupt when the heartbeat is late. This is needed during netpoll where the timer is not available. -rt kernels will also benefit since the timer is not as accurate. [ We discussed this patch last time and we decided that the -rt kernel problem alone did not justify this patch. I think the netpoll problem makes this patch necessary. ] Signed-off-by:
-
Michael Chan authored
Spurious interrupts are often encountered especially on systems using the 8259 PIC mode. This is because the I/O write to deassert the interrupt is posted and won't get to the chip immediately. As a result, the IRQ may remain asserted after the IRQ handler exits, causing spurious interrupts. Add read back to flush the I/O write to deassert the IRQ immediately. We also store the last_status_idx immediately in the IRQ handler to help detect whether the interrupt is ours or not when the IRQ is entered again before ->poll gets called. Signed-off-by:
-
Michael Chan authored
Modify the link up dmesg to report remote copper or Serdes link. Signed-off-by:
-
Michael Chan authored
Modify the driver's ethtool_ops->get_settings and set_settings functions to support remote PHY. Users control the remote copper PHY settings by specifying link settings for the tp (twisted pair) port. The nway_reset function is also modified to support remote PHY. mii-tool operations are not supported on remote PHY and we will return -EOPNOTSUPP. Signed-off-by:
-
Michael Chan authored
In blade servers, the Serdes PHY in 5708S can control the remote copper PHY through autonegotiation on the backplane. This patch adds the logic to interface with the firmware to control the remote PHY autonegotiation and to handle remote PHY link events. When remote PHY is present, the 5708S Serdes device practically becomes a copper device with full control over the 1000Base-T link settings. Signed-off-by:
-
Michael Chan authored
Add new fields in struct bnx2 and other bit definitions in shared memory to support remote PHY. Signed-off-by:
-
Michael Chan authored
Put existing code to setup the default link settings in this new function. This makes it easier to support the remote PHY feature in the next few patches. Also change ETHTOOL_ALL_FIBRE_SPEED to include 2500Mbps if supported. Signed-off-by:
-
Balazs Scheidler authored
Signed-off-by:
Balazs Scheidler <bazsi@balabit.hu> Signed-off-by:
Patrick McHardy <kaber@trash.net> Signed-off-by:
-
Yasuyuki Kozakai authored
Queue handlers are registered/unregistered in only process context. Signed-off-by:
Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp> Signed-off-by:
-
Yasuyuki Kozakai authored
The queue handlers registered by ip[6]_queue.ko at initialization should not be unregistered according to requests from userland program using nfnetlink_queue. If we allow that, there is no way to register the handlers of built-in ip[6]_queue again. Signed-off-by:
-
Patrick McHardy authored
Convert DEBUGP to pr_debug and fix lots of non-compiling debug statements. Signed-off-by:
-
Patrick McHardy authored
The ->helper pointer is protected by RCU, no need to take nf_conntrack_lock. Also remove excessive debugging. Signed-off-by:
-
Patrick McHardy authored
Don't spam the ringbuffer with decoding errors. The only printks remaining are for dropped packets when we're certain they are H.323. Signed-off-by:
-
Patrick McHardy authored
Adjust structure size and don't expect pointers passed in from userspace to be valid. Also replace an enum in an ABI structure by a fixed size type. Signed-off-by:
-
Patrick McHardy authored
Signed-off-by:
-
Patrick McHardy authored
When the maximum number of conntrack entries is reached and a new one needs to be allocated, conntrack tries to drop an unassured connection from the same hash bucket the new conntrack would hash to. Since with a properly sized hash the average number of entries per bucket is 1, the chances of actually finding one are not very good. This patch makes it walk the hash until a minimum number of 8 entries are checked. Based on patch by Vasily Averin <vvs@sw.ru>. Signed-off-by:
-
Patrick McHardy authored
Most are __read_mostly already, this changes the remaining ones. Signed-off-by:
-
Patrick McHardy authored
Eliminate the last global list searched for every new connection. Signed-off-by:
-
Patrick McHardy authored
As a last step of preventing DoS by creating lots of expectations, this patch introduces a global maximum and a sysctl to control it. The default is initialized to 4 * the expectation hash table size, which results in 1/64 of the default maxmimum of conntracks. Signed-off-by:
-
Patrick McHardy authored
This patch brings back the per-conntrack expectation list that was removed around 2.6.10 to avoid walking all expectations on expectation eviction and conntrack destruction. As these were the last users of the global expectation list, this patch also kills that. Signed-off-by:
-
Patrick McHardy authored
Convert from the global expectation list to the hash table. Signed-off-by:
-
Patrick McHardy authored
Convert from the global expectation list to the hash table. Signed-off-by:
-
Patrick McHardy authored
Currently all expectations are kept on a global list that - needs to be searched for every new conncetion - needs to be walked for evicting expectations when a master connection has reached its limit - needs to be walked on connection destruction for connections that have open expectations This is obviously not good, especially when considering helpers like H.323 that register *lots* of expectations and can set up permanent expectations, but it also allows for an easy DoS against firewalls using connection tracking helpers. Use a hashtable for expectations to avoid incurring the search overhead for every new connection. The default hash size is 1/256 of the conntrack hash table size, this can be overriden using a module parameter. This patch only introduces the hash table for expectation lookups and keeps other users to reduce the noise, the following patches will get rid of it completely. Signed-off-by:
-
Patrick McHardy authored
Signed-off-by:
-
Patrick McHardy authored
Resync expectation table dumping code with conntrack dumping: don't rely on the unique ID anymore since that requires to walk the list backwards, which doesn't work with the upcoming conversion to hlists. Signed-off-by:
-
Patrick McHardy authored
Don't walk the list when unexpecting an expectation, we already have a reference and the timer check is enough to guarantee that it still is on the list. This comment suggests that it was copied there by mistake from expectation eviction: /* choose the oldest expectation to evict */ Signed-off-by:
-
Patrick McHardy authored
Since conntrack currently allows to use masks for every bit of both helper and expectation tuples, we can't hash them and have to keep them on two global lists that are searched for every new connection. This patch removes the never used ability to use masks for the destination part of the expectation tuple and completely removes masks from helpers since the only reasonable choice is a full match on l3num, protonum and src.u.all. Signed-off-by:
-
Patrick McHardy authored
Signed-off-by:
-
Patrick McHardy authored
Currently there is a wild mix of nf_conntrack_expect_, nf_ct_exp_, expect_, exp_, ... Consistently use nf_ct_ as prefix for exported functions. Signed-off-by:
-
Patrick McHardy authored
Signed-off-by:
-
Patrick McHardy authored
Signed-off-by:
-
Patrick McHardy authored
All callers pass NULL, this also doesn't seem very useful for modules. Signed-off-by:
-
