- 17 Jan, 2013 8 commits
-
-
YOSHIFUJI Hideaki / 吉藤英明 authored
neigh->nud_state and neigh->updated are under protection of neigh->lock. Signed-off-by:
YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Yuval Mintz authored
bnx2x does an internal GRO pass but doesn't provide gso_segs, thus breaking qdisc_pkt_len_init() in case ingress qdisc is used. We store gso_segs in NAPI_GRO_CB(skb)->count, where tcp_gro_complete() expects to find the number of aggregated segments. Signed-off-by:
Yuval Mintz <yuvalmin@broadcom.com> Signed-off-by:
Eric Dumazet <edumazet@google.com> Signed-off-by:
-
Vipul Pandya authored
Signed-off-by:
Jay Hernandez <jay@chelsio.com> Signed-off-by:
Vipul Pandya <vipul@chelsio.com> Signed-off-by:
-
Fabio Baltieri authored
Fix the 64bit optimized version of ipv6_prefix_equal to convert the bitmask to network byte order only after the bit-shift. The bug was introduced in: 38675170 ipv6: 64bit version of ipv6_prefix_equal(). Signed-off-by:
Fabio Baltieri <fabio.baltieri@linaro.org> Signed-off-by:
-
Jesper Dangaard Brouer authored
Increase the amount of memory usage limits for incomplete IP fragments. Arguing for new thresh high/low values: High threshold = 4 MBytes Low threshold = 3 MBytes The fragmentation memory accounting code, tries to account for the real memory usage, by measuring both the size of frag queue struct (inet_frag_queue (ipv4:ipq/ipv6:frag_queue)) and the SKB's truesize. We want to be able to handle/hold-on-to enough fragments, to ensure good performance, without causing incomplete fragments to hurt scalability, by causing the number of inet_frag_queue to grow too much (resulting longer searches for frag queues). For IPv4, how much memory does the largest frag consume. Maximum size fragment is 64K, which is approx 44 fragments with MTU(1500) sized packets. Sizeof(struct ipq) is 200. A 1500 byte packet results in a truesize of 2944 (not 2048 as I first assumed) (44*2944)+200 = 129736 bytes The current default high thresh of 262144 bytes, is obviously problematic, as only two 64K fragments can fit in the queue at the same time. How many 64K fragment can we fit into 4 MBytes: 4*2^20/((44*2944)+200) = 32.34 fragment in queues An attacker could send a separate/distinct fake fragment packets per queue, causing us to allocate one inet_frag_queue per packet, and thus attacking the hash table and its lists. How many frag queue do we need to store, and given a current hash size of 64, what is the average list length. Using one MTU sized fragment per inet_frag_queue, each consuming (2944+200) 3144 bytes. 4*2^20/(2944+200) = 1334 frag queues -> 21 avg list length An attack could send small fragments, the smallest packet I could send resulted in a truesize of 896 bytes (I'm a little surprised by this). 4*2^20/(896+200) = 3827 frag queues -> 59 avg list length When increasing these number, we also need to followup with improvements, that is going to help scalability. Simply increasing the hash size, is not enough as the current implementation does not have a per hash bucket locking. Signed-off-by:
Jesper Dangaard Brouer <brouer@redhat.com> Signed-off-by:
-
Vincent Bernat authored
While a privileged program can open a raw socket, attach some restrictive filter and drop its privileges (or send the socket to an unprivileged program through some Unix socket), the filter can still be removed or modified by the unprivileged program. This commit adds a socket option to lock the filter (SO_LOCK_FILTER) preventing any modification of a socket filter program. This is similar to OpenBSD BIOCLOCK ioctl on bpf sockets, except even root is not allowed change/drop the filter. The state of the lock can be read with getsockopt(). No error is triggered if the state is not changed. -EPERM is returned when a user tries to remove the lock or to change/remove the filter while the lock is active. The check is done directly in sk_attach_filter() and sk_detach_filter() and does not affect only setsockopt() syscall. Signed-off-by:
Vincent Bernat <bernat@luffy.cx> Signed-off-by:
-
Cong Wang authored
__dev_get_by_name() doesn't refcount the network device, so we have to do this by ourselves. Noticed by Eric. Cc: Eric Dumazet <eric.dumazet@gmail.com> Cc: Jiri Pirko <jiri@resnulli.us> Cc: David S. Miller <davem@davemloft.net> Signed-off-by:
Cong Wang <amwang@redhat.com> Signed-off-by:
-
YOSHIFUJI Hideaki authored
Commit 3e4e4c1f ("ipv6: Introduce ip6_flow_hdr() to fill version, tclass and flowlabel.) uses ntohl(), which should be htonl(). Found by Fengguang Wu <fengguang.wu@intel.com>. Signed-off-by:
-
- 16 Jan, 2013 29 commits
-
-
Timo Teräs authored
After commit 6f0333b8 ("r8169: use 50% less ram for RX ring") the rx ring buffers are always copied making dirty_rx useless. Signed-off-by:
Timo Teräs <timo.teras@iki.fi> Acked-by:
Francois Romieu <romieu@fr.zoreil.com> Signed-off-by:
-
Cong Wang authored
v4: hold rtnl lock for the whole netpoll_setup() v3: remove the comment v2: use RCU read lock This patch fixes the following warning: [ 72.013864] RTNL: assertion failed at net/core/dev.c (4955) [ 72.017758] Pid: 668, comm: netpoll-prep-v6 Not tainted 3.8.0-rc1+ #474 [ 72.019582] Call Trace: [ 72.020295] [<ffffffff8176653d>] netdev_master_upper_dev_get+0x35/0x58 [ 72.022545] [<ffffffff81784edd>] netpoll_setup+0x61/0x340 [ 72.024846] [<ffffffff815d837e>] store_enabled+0x82/0xc3 [ 72.027466] [<ffffffff815d7e51>] netconsole_target_attr_store+0x35/0x37 [ 72.029348] [<ffffffff811c3479>] configfs_write_file+0xe2/0x10c [ 72.030959] [<ffffffff8115d239>] vfs_write+0xaf/0xf6 [ 72.032359] [<ffffffff81978a05>] ? sysret_check+0x22/0x5d [ 72.033824] [<ffffffff8115d453>] sys_write+0x5c/0x84 [ 72.035328] [<ffffffff819789d9>] system_call_fastpath+0x16/0x1b In case of other races, hold rtnl lock for the entire netpoll_setup() function. Cc: Eric Dumazet <eric.dumazet@gmail.com> Cc: Jiri Pirko <jiri@resnulli.us> Cc: David S. Miller <davem@davemloft.net> Signed-off-by:
-
Stephen Hemminger authored
The VMXNET3 device provides RSS hash value for received packets, but it is not being used. Signed-off-by:
Stephen Hemminger <stephen@networkplumber.org> Signed-off-by:
-
Stephen Hemminger authored
Rather than generating a different RSS key on each boot, just use a predetermined value that will map same flow to same value on every device for more predictable testing. This is already done on most hardware drivers. Initial key value just some arbitrary bits extracted once from /dev/random. Signed-off-by:
-
Stephen Hemminger authored
This static variable is never set, it initializes to 0 which is VMXNET3_INTR_BUDDYSHARE, and never changes. Signed-off-by:
-
Stephen Hemminger authored
An atomic counter of devices present is maintained but never used. Signed-off-by:
-
Stephen Hemminger authored
Defined but never used. Signed-off-by:
-
Stephen Hemminger authored
Use the standard netdev_xxx() and dev_xxx() wrappers to format log messages. Signed-off-by:
-
Stephen Hemminger authored
Use netdev_dbg() rather than dev_dbg() because the former prints the device name which is more useful than the pci name. Signed-off-by:
-
Stephen Hemminger authored
This messages that occur during boot time from this device when netdev_err is called before calling register_netdevice(). Switch to using dev_XXX macros which correlate message with PCI info which is available. Rather than fixing the features message, just remove it since the information is redundant and available through ethtool. Signed-off-by:
-
Stephen Hemminger authored
The uncommitted[] array was set but never used except in a debug message. Remove it. Signed-off-by:
-
Stephen Hemminger authored
Use netdev_alloc_skb_align, rather than open code using dev_alloc_skb. Change allocation at startup to use GFP_KERNEL. Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net-nextDavid S. Miller authored
Jeff Kirsher says: ==================== This series contains updates to e1000e only. v2- updates patch 09/15 "e1000e: resolve checkpatch PREFER_PR_LEVEL warning" based on feedback from Joe Perches. ==================== Signed-off-by: -
Bruce Allan authored
Cleanup a set of conditional tests. Signed-off-by:
Bruce Allan <bruce.w.allan@intel.com> Tested-by:
Aaron Brown <aaron.f.brown@intel.com> Signed-off-by:
Jeff Kirsher <jeffrey.t.kirsher@intel.com>
-
Bruce Allan authored
The removed code block is duplicated in e1000e_write_itr() so use that instead. Signed-off-by:
-
Bruce Allan authored
Signed-off-by:
-
Bruce Allan authored
Signed-off-by:
-
Bruce Allan authored
Signed-off-by:
-
Bruce Allan authored
rx_long_byte_count can be removed since it is duplicated in rx_bytes Signed-off-by:
-
Bruce Allan authored
WARNING: Prefer netdev_info(netdev, ... then dev_info(dev, ... then pr_info(... to printk(KERN_INFO ... v2 - remove unnecessary "e1000e:" prefix as pointed out by Joe Perches since that produces a redundant "e1000e:" in the log message Cc: Joe Perches <joe@perches.com> Signed-off-by: -
Bruce Allan authored
...discovered during code inspection. Signed-off-by:
-
Bruce Allan authored
When there is heavy traffic and the cable is pulled, the driver must reset the adapter to flush the Tx queue in hardware. This causes the reset path to be scheduled and logs the message "Reset adapter" which could be mis- interpreted as an error by the user. Change how the reset path is invoked for this scenario by using the same method done in an existing work-around for 80003es2lan (i.e. set a flag and if the flag is set in the reset code do not log the "Reset adapter" message since the reset is expected). Re-name the FLAG_RX_RESTART_NOW to FLAG_RESTART_NOW since it is used for resets in both the Rx and Tx specific code. Signed-off-by:
-
Bruce Allan authored
Energy Efficient Ethernet on 82579 and I217 should only be enabled if not disabled by the user, if the link is full duplex and the link partner has similar EEE capabilities (stored in different EMI registers on the two different parts). After enabling EEE, read the IEEE MMD register 3.1 (which is also stored in different EMI registers on the two different parts) to clear the count of received Tx/Rx LPI indications. Also, rename I217_EEE_100_SUPPORTED to I82579_EEE_100_SUPPORTED to indicate the bit is valid starting with I82579 (released before I217). Signed-off-by:
Jeff Pieper <jeffrey.e.pieper@intel.com> Signed-off-by:
-
Bruce Allan authored
When connected to certain switches, the 82577 PHY might drop link unexpectedly. Work around the issue by setting the Mean Square Error higher than the hardware default. Signed-off-by:
-
Bruce Allan authored
The Extended Management Interface (EMI) registers are accessed by first writing the EMI register offset to the EMI_ADDR regiter and then either reading or writing the data to/from the EMI_DATA register. Add helper functions for performing these steps and convert existing EMI register accesses accordingly. Signed-off-by:
-
Eric Dumazet authored
commit 1def9238 (net_sched: more precise pkt_len computation) does a wrong computation of mac + network headers length, as it includes the padding before the frame. Signed-off-by:
-
Bruce Allan authored
On I217, the bit that indicates an invalid EEPROM (NVM) image checksum has changed from previous ICH/PCH LOMs. When validating the EEPROM checksum, check the appropriate bit on different devices. Signed-off-by:
-
Bruce Allan authored
When data blocks are written to the EEPROM, the HW/SW/FW semaphore must be held for the duration. With large data blocks on 80003es2lan, 82571 and 82572, this can take too long and cause the firmware to take ownership of the semaphore and consequently ownership of writes to the EEPROM. Instead, acquire and release the semaphore for each page of the block written. Signed-off-by:
-
Bruce Allan authored
Enables flow control to be set in SerDes autoneg mode. This is what is done for copper, but relies on a different set of register/bit checks since this is all done within the Mac registers. Remove inapplicable comment in defines.h Signed-off-by:
-
- 15 Jan, 2013 3 commits
-
-
Yuval Mintz authored
Signed-off-by:
Ariel Elior <ariele@broadcom.com> Signed-off-by:
Eilon Greenstein <eilong@broadcom.com> Signed-off-by:
-
Yuval Mintz authored
Since submit 621b4d66 the bnx2x driver support FW GRO. However, when using the device with GRO enabled in bridging scenarios throughput is very low, as the bridge expects all incoming packets to be passed with CHECKSUM_PARTIAL - a demand which is satisfied by the SW GRO implementation, but was missed in the bnx2x driver implementation (which returned CHECKSUM_UNNECESSARY). Now, given that the traffic is supported by FW GRO (TCP/IP), the bnx2x driver calculates the pseudo checksum by itself, passing skbs with CHECKSUM_PARTIAL and giving a much better throughput when receiving GRO traffic. Signed-off-by:
-
Yuval Mintz authored
When enabling interrupts, acknowledge the interrupt only after configuring the IGU to the correct interrupt mode (otherwise it would dirty selftests) Signed-off-by:
-
