| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
| |
gcc does a poor job at generating code for the memcpy of the frags
array in skb_gro_receive, which is the primary purpose of that
function when merging frags. In particular, it can't utilise the
alignment information of the source and destination. This patch
open-codes the copy so we process words instead of bytes.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
| |
This patch fixes a build warning due to an unused label.
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
|
|
|
| |
Just use the constant 20 to keep things working.
If someone is so motivated, this can be converted over to
dynamic strings. I tried and it's a lot of work.
But for now this is good enough.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
| |
It's going away. Just leave the constant "20" here so that
behavior doesn't change.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
| |
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
| |
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
| |
Use MII_BUS_ID_SIZE instead.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
|
|
| |
BUS_ID_SIZE is really no more, and device names are dynamically
allocated and thus can be any necessary size.
So remove the BUG check here making sure BUS_ID_SIZE is at least
as large as IFNAMSIZ.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
|
|
|
|
| |
The r8169 driver supports 3 different families of network chips
(RTL8169, RTL8168 and RTL8101). When an unknown version is found, the
driver currently always defaults to the RTL8169 variant. This has very
little chance to ever work for chips of the other families. So better
define a per-family default.
Signed-off-by: Jean Delvare <jdelvare@suse.de>
Acked-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Inspired by the patch for 8139too (bda6a15a).
Currently we can't set mac address on a running ucc_geth device.
But this is needed when you use this device as a bonding slave in
bonding device in balance-alb mode. So add this feature for ucc_geth
device.
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
| |
This patch is for supporting C10NEM. C10NEM is a switch module, which
has back-to-back XAUI link connected to blades.
Signed-off-by: Tanli Chang <tanli.chang@sun.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If we loaded the driver with out a SFP module plugged in it would
leave it in a state that make it later unable to link when a module
was plugged in. This patch corrects that by:
ixgbe_probe() - moving the check for IXGBE_ERR_SFP_NOT_PRESENT from
after get_invariants() to after reset_hw() as now reset_hw() is
where this condition will be indentified.
ixgbe_reset_hw_82598() - Enable this function to now return
IXGBE_ERR_SFP_NOT_PRESENT.
ixgbe_identify_sfp_module_generic() - This where the lack of SFP
module is detected. Modifications are added to allow a different
return value for modules that just haven't been plugged in yet.
Other functions were updated to allow correct logging.
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Device ID 0x10d8 is the default silicon device ID for 82599. However, the
device will not be functional without an EEPROM, so we want to prevent the
driver from loading on the device. Otherwise, the driver will load, but no
PHY setup or PCIe setup will occur, causing the device to be unusable. To
prevent users from encountering this, just remove the device ID.
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
|
|
|
| |
The phy port status register has the MDI-X status bit on bit 11, not bit 3
as is currently setup in the define. This patch corrects that so the
correct bit is checked on igp PHY types.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Acked-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Based on previous patch from Jesper Dangaard Brouer.
The RNBC (Receive No Buffers Count) register for the 82576, indicate
that frames were received when there were no available buffers in host
memory to store those frames (receive descriptor head and tail
pointers were equal). The packet is still received by the NIC if
there is space in the FIFO on the NIC.
As the RNBC value is not a packet drop, the driver stores this value
in net_stats.rx_fifo_errors to indicate that there were no system
buffers available for the incoming packet. Actual dropped packets
are counted in the MPC value.
Saving the stats in dev->net_stats makes it visible via
/proc/net/dev as "fifo", and thus viewable to ifconfig
as "overruns" and 'netstat -i' as "RX-OVR".
The Receive No Buffers Count (RNBC) can already be queried by
ethtool -S as "rx_no_buffer_count".
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
CC: Jesper Dangaard Brouer <hawk@comx.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Based on the previous patches from Jesper Dangaard Brouer <hawk@comx.dk>
Implement reading the per queue drop stats register
RQDPC (Receive Queue Drop Packet Count). It counts the number of
packets dropped by a queue due to lack of descriptors available.
Notice RQDPC (Receive Queue Drop Packet Count) stats only gets
incremented, if the DROP_EN bit it set (in the SRRCTL register
for that queue). If DROP_EN bit is NOT set, then the some what
equivalent count is stored in RNBC (not per queue basis).
The RQDPC register is only 12 bit, thus the precision might
suffer due to overrun in-netween the watchdog polling interval.
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
CC: Jesper Dangaard Brouer <hawk@comx.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
| |
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
| |
Also convert DPRINTK to pr_debug and dev_printk to dev_<foo>.
Signed-off-by: Alexander Beregalov <a.beregalov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
| |
Signed-off-by: Alexander Beregalov <a.beregalov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
| |
Signed-off-by: Alexander Beregalov <a.beregalov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
| |
Get rid of private struct net_device_stats in "struct atl1e_adapter",
and use one provided in struct net_device instead.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We would like to get rid of netdev->trans_start = jiffies; that about all net
drivers have to use in their start_xmit() function, and use txq->trans_start
instead.
This can be done generically in core network, as suggested by David.
Some devices, (particularly loopback) dont need trans_start update, because
they dont have transmit watchdog. We could add a new device flag, or rely
on fact that txq->tran_start can be updated is txq->xmit_lock_owner is
different than -1. Use a helper function to hide our choice.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
| |
Signed-off-by: Yoshihiro Shimoda <shimoda.yoshihiro@renesas.com>
Signed-off-by: Kuninori Morimoto <morimoto.kuninori@renesas.com>
Signed-off-by: Nobuhiro Iwamatsu <iwamatsu.nobuhiro@renesas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
| |
This improves readability by collecting CPU dependency code.
Signed-off-by: Yoshihiro Shimoda <shimoda.yoshihiro@renesas.com>
Signed-off-by: Nobuhiro Iwamatsu <iwamatsu.nobuhiro@renesas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
| |
Fix the problem that this driver not update rxdesc in sh_eth_rx.
Signed-off-by: Yoshihiro Shimoda <shimoda.yoshihiro@renesas.com>
Signed-off-by: Nobuhiro Iwamatsu <iwamatsu.nobuhiro@renesas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
| |
Fix the probrem that TX/RX descirptor not set physical memory.
Signed-off-by: Yoshihiro Shimoda <shimoda.yoshihiro@renesas.com>
Signed-off-by: Nobuhiro Iwamatsu <iwamatsu.nobuhiro@renesas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
| |
Fix the problem that may not work receive process by cache coherency issue.
Signed-off-by: Yoshihiro Shimoda <shimoda.yoshihiro@renesas.com>
Signed-off-by: Nobuhiro Iwamatsu <iwamatsu.nobuhiro@renesas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
|
| |
Right-shifts of signed integers are implementation-defined so unportable.
With feedback from: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
| |
Signed-off-by: Gabriel Paubert <paubert@iram.es>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
|
|
| |
The patch makes iwmc3200wifi select CFG80211 instead of LIB80211.
This fixed module link error reported by Randy Dunlap
<randy.dunlap@oracle.com> when compiling iwmc3200wifi without
cfg80211 selected. WIRELESS_EXT is also selected by iwmc3200wifi.
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
| |
All drivers are already converted to new net_device_ops API
and nobody uses old API anymore.
Signed-off-by: Alexander Beregalov <a.beregalov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|\
| |
| |
| |
| |
| |
| |
| |
| | |
master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Conflicts:
drivers/net/wireless/ath/ath5k/phy.c
drivers/net/wireless/iwlwifi/iwl-agn.c
drivers/net/wireless/iwlwifi/iwl3945-base.c
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Since commit 0fd56bb5be6455d0d42241e65aed057244665e5e ("gianfar:
Add support for skb recycling"), gianfar puts skbuffs that are in
the rx ring back onto the recycle list as-is in case there was a
receive error, but this breaks the following invariant: that all
skbuffs on the recycle list have skb->data = skb->head + NET_SKB_PAD.
The RXBUF_ALIGNMENT realignment done in gfar_new_skb() will be done
twice on skbuffs recycled in this way, causing there not to be enough
room in the skb anymore to receive a full packet, eventually leading
to an skb_over_panic from gfar_clean_rx_ring() -> skb_put().
Resetting the skb->data pointer to skb->head + NET_SKB_PAD before
putting the skb back onto the recycle list restores the mentioned
invariant, and should fix this issue.
Reported-by: Michael Guntsche <mike@it-loops.com>
Tested-by: Michael Guntsche <mike@it-loops.com>
Signed-off-by: Lennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
rxrpc_alloc_connection() doesn't return an error code on failure, it just
returns NULL. IS_ERR(NULL) is false.
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
It seems we can fix this by disabling preemption while we re-balance the
trie. This is with the CONFIG_CLASSIC_RCU. It's been stress-tested at high
loads continuesly taking a full BGP table up/down via iproute -batch.
Note. fib_trie is not updated for CONFIG_PREEMPT_RCU
Reported-by: Andrei Popa
Signed-off-by: Robert Olsson <robert.olsson@its.uu.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
typo -- pkt_dev->nflows is for stats only, the number of concurrent
flows is stored in cflows.
Reported-By: Vladimir Ivashchenko <hazard@francoudi.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
| |
| |
| |
| | |
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The use of unspecified protocol in IPv6 initial route prevents quagga to
install IPv6 default route:
# show ipv6 route
S ::/0 [1/0] via fe80::1, eth1_0
K>* ::/0 is directly connected, lo, rej
C>* ::1/128 is directly connected, lo
C>* fe80::/64 is directly connected, eth1_0
# ip -6 route
fe80::/64 dev eth1_0 proto kernel metric 256 mtu 1500 advmss 1440
hoplimit -1
ff00::/8 dev eth1_0 metric 256 mtu 1500 advmss 1440 hoplimit -1
unreachable default dev lo proto none metric -1 error -101 hoplimit 255
The attached patch ensures RTPROT_KERNEL to the default initial route
and fixes the problem for quagga.
This is similar to "ipv6: protocol for address routes"
f410a1fba7afa79d2992620e874a343fdba28332.
# show ipv6 route
S>* ::/0 [1/0] via fe80::1, eth1_0
C>* ::1/128 is directly connected, lo
C>* fe80::/64 is directly connected, eth1_0
# ip -6 route
fe80::/64 dev eth1_0 proto kernel metric 256 mtu 1500 advmss 1440
hoplimit -1
fe80::/64 dev eth1_0 proto kernel metric 256 mtu 1500 advmss 1440
hoplimit -1
ff00::/8 dev eth1_0 metric 256 mtu 1500 advmss 1440 hoplimit -1
default via fe80::1 dev eth1_0 proto zebra metric 1024 mtu 1500
advmss 1440 hoplimit -1
unreachable default dev lo proto kernel metric -1 error -101 hoplimit 255
Signed-off-by: Jean-Mickael Guerin <jean-mickael.guerin@6wind.com>
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |\
| | |
| | |
| | | |
git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-2.6
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Do not go beyond ARRAY_SIZE of intf->crypto_stats
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
enable iwl driver to support 5000 ucode having version 2 of API
Signed-off-by: Jay Sternberg <jay.e.sternberg@linux.intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Its possible for cfg80211 to have scheduled the work and for
the global workqueue to not have kicked in prior to a cfg80211
driver's regulatory hint or wiphy_apply_custom_regulatory().
Although this is very unlikely its possible and should fix
this race. When this race would happen you are expected to have
hit a null pointer dereference panic.
Cc: stable@kernel.org
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Tested-by: Alan Jenkins <alan-jenkins@tuffmail.co.uk>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
"airo: airo_get_encode{,ext} potential buffer overflow" was actually a
no-op, due to an unrecognized type overflow in an assignment. Oddly,
gcc only seems to tell me about it when using -Wextra...grrr...
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
When the EEPROM contains weird values for the power levels we have to
fix the interpolation process.
Signed-off-by: Fabio Rossi <rossi.f@inwind.it>
Acked-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Calling cancel_delayed_work() from inside
spin_lock_irqsave, introduces a potential deadlock.
As explained by Johannes Berg <johannes@sipsolutions.net>
A - lock
T - timer
phase CPU 1 CPU 2
---------------------------------------------
some place that calls
cancel_timer_sync()
(which is the | code)
lock-irq(A)
| "lock-irq"(T)
| "unlock"(T)
| wait(T)
unlock(A)
timer softirq
"lock"(T)
run(T)
"unlock"(T)
irq handler
lock(A)
unlock(A)
Now all that again, interleaved, leading to deadlock:
lock-irq(A)
"lock"(T)
run(T)
IRQ during or maybe
before run(T) --> lock(A)
"lock-irq"(T)
wait(T)
We fix this by moving the call to cancel_delayed_work() into workqueue.
There are cases where the work may not actually be queued or running
at the time we are trying to cancel it, but cancel_delayed_work() is
able to deal with this.
Also cleanup iwl_set_mode related to this call. This function
(iwl_set_mode) is only called when bringing interface up and there will
thus not be any scanning done. No need to try to cancel scanning.
Fixes http://bugzilla.kernel.org/show_bug.cgi?id=13224, which was also
reported at http://marc.info/?l=linux-wireless&m=124081921903223&w=2 .
Tested-by: Miles Lane <miles.lane@gmail.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Acked-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Commit e8f055f0c3b ("ath5k: Update reset code") subtly changed the
code that computes floating point values for the PHY3_TIMING register
such that the exponent is off by a decimal point, which can cause
problems with OFDM channel operation.
get_bitmask_order() actually returns the highest bit set plus one,
whereas the previous code wanted the highest bit set. Instead, use
ilog2 which is what this code is really calculating. Also check
coef_scaled to handle the (invalid) case where we need log2(0).
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Another design flaw in wireless extensions (is anybody
surprised?) in the way it handles the iw_encode_ext
structure: The structure is part of the 'extra' memory
but contains the key length explicitly, instead of it
just being the length of the extra buffer - size of
the struct and using the explicit key length only for
the get operation (which only writes it).
Therefore, we have this layout:
extra: +-------------------------+
| struct iw_encode_ext { |
| ... |
| u16 key_len; |
| u8 key[0]; |
| }; |
+-------------------------+
| key material |
+-------------------------+
Now, all drivers I checked use ext->key_len without
checking that both key_len and the struct fit into the
extra buffer that has been copied from userspace. This
leads to a buffer overrun while reading that buffer,
depending on the driver it may be possible to specify
arbitrary key_len or it may need to be a proper length
for the key algorithm specified.
Thankfully, this is only exploitable by root, but root
can actually cause a segfault or use kernel memory as
a key (which you can even get back with siocgiwencode
or siocgiwencodeext from the key buffer).
Fix this by verifying that key_len fits into the buffer
along with struct iw_encode_ext.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
AR5K_PHY_PLL_40MHZ_5413 should not be ORed with AR5K_PHY_MODE_RAD_RF5112
for 5 GHz channels.
The incorrect PLL value breaks scanning in the countries where 5 GHz
channels are allowed.
Signed-off-by: Pavel Roskin <proski@gnu.org>
Acked-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Alexander V. Lukyanov found a regression in 2.6.29 and made a complete
analysis found in http://bugzilla.kernel.org/show_bug.cgi?id=13339
Quoted here because its a perfect one :
begin_of_quotation
2.6.29 patch has introduced flexible route cache rebuilding. Unfortunately the
patch has at least one critical flaw, and another problem.
rt_intern_hash calculates rthi pointer, which is later used for new entry
insertion. The same loop calculates cand pointer which is used to clean the
list. If the pointers are the same, rtable leak occurs, as first the cand is
removed then the new entry is appended to it.
This leak leads to unregister_netdevice problem (usage count > 0).
Another problem of the patch is that it tries to insert the entries in certain
order, to facilitate counting of entries distinct by all but QoS parameters.
Unfortunately, referencing an existing rtable entry moves it to list beginning,
to speed up further lookups, so the carefully built order is destroyed.
For the first problem the simplest patch it to set rthi=0 when rthi==cand, but
it will also destroy the ordering.
end_of_quotation
Problematic commit is 1080d709fb9d8cd4392f93476ee46a9d6ea05a5b
(net: implement emergency route cache rebulds when gc_elasticity is exceeded)
Trying to keep dst_entries ordered is too complex and breaks the fact that
order should depend on the frequency of use for garbage collection.
A possible fix is to make rt_intern_hash() simpler, and only makes
rt_check_expire() a litle bit smarter, being able to cope with an arbitrary
entries order. The added loop is running on cache hot data, while cpu
is prefetching next object, so should be unnoticied.
Reported-and-analyzed-by: Alexander V. Lukyanov <lav@yar.ru>
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |/
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
rt_check_expire() computes average and standard deviation of chain lengths,
but not correclty reset length to 0 at beginning of each chain.
This probably gives overflows for sum2 (and sum) on loaded machines instead
of meaningful results.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|